diff --git a/llvm/lib/CodeGen/TwoAddressInstructionPass.cpp b/llvm/lib/CodeGen/TwoAddressInstructionPass.cpp
index ebacbc420f858..5b7b3cdc78e51 100644
--- a/llvm/lib/CodeGen/TwoAddressInstructionPass.cpp
+++ b/llvm/lib/CodeGen/TwoAddressInstructionPass.cpp
@@ -921,7 +921,7 @@ bool TwoAddressInstructionPass::rescheduleMIBelowKill(
     // Debug or pseudo instructions cannot be counted against the limit.
     if (OtherMI.isDebugOrPseudoInstr())
       continue;
-    if (NumVisited > 10)  // FIXME: Arbitrary limit to reduce compile time cost.
+    if (NumVisited > 64)
       return false;
     ++NumVisited;
     if (OtherMI.hasUnmodeledSideEffects() || OtherMI.isCall() ||
@@ -1094,7 +1094,7 @@ bool TwoAddressInstructionPass::rescheduleKillAboveMI(
     // Debug or pseudo instructions cannot be counted against the limit.
     if (OtherMI.isDebugOrPseudoInstr())
       continue;
-    if (NumVisited > 10)  // FIXME: Arbitrary limit to reduce compile time cost.
+    if (NumVisited > 64)
       return false;
     ++NumVisited;
     if (OtherMI.hasUnmodeledSideEffects() || OtherMI.isCall() ||
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-extends.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-extends.ll
index c7a89612d278f..8f566b994198e 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-extends.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-extends.ll
@@ -236,22 +236,20 @@ define void @sext_v16i8_v16i64(<16 x i8> %a, ptr %out) {
 ; CHECK-NEXT:    sunpklo z4.d, z2.s
 ; CHECK-NEXT:    ext z2.b, z2.b, z2.b, #8
 ; CHECK-NEXT:    sunpklo z0.s, z0.h
-; CHECK-NEXT:    mov z7.d, z1.d
-; CHECK-NEXT:    sunpklo z2.d, z2.s
+; CHECK-NEXT:    sunpklo z7.d, z1.s
+; CHECK-NEXT:    ext z1.b, z1.b, z1.b, #8
 ; CHECK-NEXT:    sunpklo z5.d, z3.s
 ; CHECK-NEXT:    ext z3.b, z3.b, z3.b, #8
-; CHECK-NEXT:    ext z7.b, z7.b, z1.b, #8
+; CHECK-NEXT:    sunpklo z2.d, z2.s
 ; CHECK-NEXT:    sunpklo z1.d, z1.s
-; CHECK-NEXT:    mov z6.d, z0.d
+; CHECK-NEXT:    sunpklo z6.d, z0.s
+; CHECK-NEXT:    ext z0.b, z0.b, z0.b, #8
 ; CHECK-NEXT:    sunpklo z3.d, z3.s
 ; CHECK-NEXT:    stp q4, q2, [x0]
-; CHECK-NEXT:    sunpklo z4.d, z7.s
-; CHECK-NEXT:    ext z6.b, z6.b, z0.b, #8
 ; CHECK-NEXT:    sunpklo z0.d, z0.s
+; CHECK-NEXT:    stp q7, q1, [x0, #32]
 ; CHECK-NEXT:    stp q5, q3, [x0, #64]
-; CHECK-NEXT:    sunpklo z2.d, z6.s
-; CHECK-NEXT:    stp q1, q4, [x0, #32]
-; CHECK-NEXT:    stp q0, q2, [x0, #96]
+; CHECK-NEXT:    stp q6, q0, [x0, #96]
 ; CHECK-NEXT:    ret
   %b = sext <16 x i8> %a to <16 x i64>
   store <16 x i64> %b, ptr %out
@@ -263,63 +261,57 @@ define void @sext_v32i8_v32i64(ptr %in, ptr %out) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ldp q1, q0, [x0]
 ; CHECK-NEXT:    add z0.b, z0.b, z0.b
-; CHECK-NEXT:    add z1.b, z1.b, z1.b
-; CHECK-NEXT:    mov z2.d, z0.d
-; CHECK-NEXT:    sunpklo z0.h, z0.b
-; CHECK-NEXT:    mov z3.d, z1.d
-; CHECK-NEXT:    sunpklo z1.h, z1.b
+; CHECK-NEXT:    add z2.b, z1.b, z1.b
+; CHECK-NEXT:    sunpklo z3.h, z0.b
+; CHECK-NEXT:    ext z0.b, z0.b, z0.b, #8
+; CHECK-NEXT:    sunpklo z1.h, z2.b
 ; CHECK-NEXT:    ext z2.b, z2.b, z2.b, #8
+; CHECK-NEXT:    sunpklo z0.h, z0.b
+; CHECK-NEXT:    sunpklo z4.s, z3.h
 ; CHECK-NEXT:    ext z3.b, z3.b, z3.b, #8
-; CHECK-NEXT:    sunpklo z4.s, z0.h
-; CHECK-NEXT:    ext z0.b, z0.b, z0.b, #8
 ; CHECK-NEXT:    sunpklo z5.s, z1.h
-; CHECK-NEXT:    ext z1.b, z1.b, z1.b, #8
 ; CHECK-NEXT:    sunpklo z2.h, z2.b
-; CHECK-NEXT:    sunpklo z3.h, z3.b
-; CHECK-NEXT:    sunpklo z0.s, z0.h
-; CHECK-NEXT:    sunpklo z16.d, z4.s
+; CHECK-NEXT:    ext z1.b, z1.b, z1.b, #8
+; CHECK-NEXT:    sunpklo z6.s, z0.h
+; CHECK-NEXT:    sunpklo z3.s, z3.h
+; CHECK-NEXT:    ext z0.b, z0.b, z0.b, #8
+; CHECK-NEXT:    sunpklo z7.d, z4.s
 ; CHECK-NEXT:    ext z4.b, z4.b, z4.b, #8
-; CHECK-NEXT:    sunpklo z1.s, z1.h
-; CHECK-NEXT:    sunpklo z17.d, z5.s
+; CHECK-NEXT:    sunpklo z16.d, z5.s
 ; CHECK-NEXT:    ext z5.b, z5.b, z5.b, #8
-; CHECK-NEXT:    sunpklo z6.s, z2.h
-; CHECK-NEXT:    sunpklo z7.s, z3.h
+; CHECK-NEXT:    sunpklo z17.s, z2.h
 ; CHECK-NEXT:    ext z2.b, z2.b, z2.b, #8
-; CHECK-NEXT:    sunpklo z4.d, z4.s
+; CHECK-NEXT:    sunpklo z1.s, z1.h
+; CHECK-NEXT:    sunpklo z0.s, z0.h
+; CHECK-NEXT:    sunpklo z18.d, z6.s
+; CHECK-NEXT:    ext z6.b, z6.b, z6.b, #8
+; CHECK-NEXT:    sunpklo z19.d, z3.s
 ; CHECK-NEXT:    ext z3.b, z3.b, z3.b, #8
-; CHECK-NEXT:    sunpklo z19.d, z0.s
+; CHECK-NEXT:    sunpklo z4.d, z4.s
 ; CHECK-NEXT:    sunpklo z5.d, z5.s
-; CHECK-NEXT:    ext z0.b, z0.b, z0.b, #8
 ; CHECK-NEXT:    sunpklo z2.s, z2.h
-; CHECK-NEXT:    sunpklo z18.d, z6.s
-; CHECK-NEXT:    ext z6.b, z6.b, z6.b, #8
-; CHECK-NEXT:    sunpklo z3.s, z3.h
-; CHECK-NEXT:    stp q16, q4, [x1, #128]
-; CHECK-NEXT:    mov z16.d, z7.d
-; CHECK-NEXT:    sunpklo z0.d, z0.s
-; CHECK-NEXT:    stp q17, q5, [x1]
-; CHECK-NEXT:    sunpklo z5.d, z7.s
-; CHECK-NEXT:    sunpklo z4.d, z6.s
-; CHECK-NEXT:    mov z6.d, z1.d
-; CHECK-NEXT:    ext z16.b, z16.b, z7.b, #8
-; CHECK-NEXT:    mov z7.d, z2.d
-; CHECK-NEXT:    stp q19, q0, [x1, #160]
-; CHECK-NEXT:    sunpklo z0.d, z2.s
-; CHECK-NEXT:    ext z6.b, z6.b, z1.b, #8
-; CHECK-NEXT:    sunpklo z1.d, z1.s
-; CHECK-NEXT:    stp q18, q4, [x1, #192]
-; CHECK-NEXT:    mov z4.d, z3.d
-; CHECK-NEXT:    ext z7.b, z7.b, z2.b, #8
-; CHECK-NEXT:    sunpklo z16.d, z16.s
 ; CHECK-NEXT:    sunpklo z6.d, z6.s
-; CHECK-NEXT:    ext z4.b, z4.b, z3.b, #8
-; CHECK-NEXT:    sunpklo z2.d, z7.s
 ; CHECK-NEXT:    sunpklo z3.d, z3.s
-; CHECK-NEXT:    stp q5, q16, [x1, #64]
-; CHECK-NEXT:    stp q1, q6, [x1, #32]
-; CHECK-NEXT:    sunpklo z1.d, z4.s
-; CHECK-NEXT:    stp q0, q2, [x1, #224]
-; CHECK-NEXT:    stp q3, q1, [x1, #96]
+; CHECK-NEXT:    stp q16, q5, [x1]
+; CHECK-NEXT:    sunpklo z5.d, z1.s
+; CHECK-NEXT:    ext z1.b, z1.b, z1.b, #8
+; CHECK-NEXT:    stp q7, q4, [x1, #128]
+; CHECK-NEXT:    sunpklo z4.d, z17.s
+; CHECK-NEXT:    ext z17.b, z17.b, z17.b, #8
+; CHECK-NEXT:    stp q18, q6, [x1, #192]
+; CHECK-NEXT:    sunpklo z6.d, z0.s
+; CHECK-NEXT:    ext z0.b, z0.b, z0.b, #8
+; CHECK-NEXT:    stp q19, q3, [x1, #160]
+; CHECK-NEXT:    sunpklo z3.d, z2.s
+; CHECK-NEXT:    ext z2.b, z2.b, z2.b, #8
+; CHECK-NEXT:    sunpklo z7.d, z17.s
+; CHECK-NEXT:    sunpklo z1.d, z1.s
+; CHECK-NEXT:    sunpklo z0.d, z0.s
+; CHECK-NEXT:    sunpklo z2.d, z2.s
+; CHECK-NEXT:    stp q5, q1, [x1, #32]
+; CHECK-NEXT:    stp q4, q7, [x1, #64]
+; CHECK-NEXT:    stp q3, q2, [x1, #96]
+; CHECK-NEXT:    stp q6, q0, [x1, #224]
 ; CHECK-NEXT:    ret
   %a = load <32 x i8>, ptr %in
   %b = add <32 x i8> %a, %a
@@ -661,22 +653,20 @@ define void @zext_v16i8_v16i64(<16 x i8> %a, ptr %out) {
 ; CHECK-NEXT:    uunpklo z4.d, z2.s
 ; CHECK-NEXT:    ext z2.b, z2.b, z2.b, #8
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
-; CHECK-NEXT:    mov z7.d, z1.d
-; CHECK-NEXT:    uunpklo z2.d, z2.s
+; CHECK-NEXT:    uunpklo z7.d, z1.s
+; CHECK-NEXT:    ext z1.b, z1.b, z1.b, #8
 ; CHECK-NEXT:    uunpklo z5.d, z3.s
 ; CHECK-NEXT:    ext z3.b, z3.b, z3.b, #8
-; CHECK-NEXT:    ext z7.b, z7.b, z1.b, #8
+; CHECK-NEXT:    uunpklo z2.d, z2.s
 ; CHECK-NEXT:    uunpklo z1.d, z1.s
-; CHECK-NEXT:    mov z6.d, z0.d
+; CHECK-NEXT:    uunpklo z6.d, z0.s
+; CHECK-NEXT:    ext z0.b, z0.b, z0.b, #8
 ; CHECK-NEXT:    uunpklo z3.d, z3.s
 ; CHECK-NEXT:    stp q4, q2, [x0]
-; CHECK-NEXT:    uunpklo z4.d, z7.s
-; CHECK-NEXT:    ext z6.b, z6.b, z0.b, #8
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
+; CHECK-NEXT:    stp q7, q1, [x0, #32]
 ; CHECK-NEXT:    stp q5, q3, [x0, #64]
-; CHECK-NEXT:    uunpklo z2.d, z6.s
-; CHECK-NEXT:    stp q1, q4, [x0, #32]
-; CHECK-NEXT:    stp q0, q2, [x0, #96]
+; CHECK-NEXT:    stp q6, q0, [x0, #96]
 ; CHECK-NEXT:    ret
   %b = zext <16 x i8> %a to <16 x i64>
   store <16 x i64> %b, ptr %out
@@ -688,63 +678,57 @@ define void @zext_v32i8_v32i64(ptr %in, ptr %out) {
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ldp q1, q0, [x0]
 ; CHECK-NEXT:    add z0.b, z0.b, z0.b
-; CHECK-NEXT:    add z1.b, z1.b, z1.b
-; CHECK-NEXT:    mov z2.d, z0.d
-; CHECK-NEXT:    uunpklo z0.h, z0.b
-; CHECK-NEXT:    mov z3.d, z1.d
-; CHECK-NEXT:    uunpklo z1.h, z1.b
+; CHECK-NEXT:    add z2.b, z1.b, z1.b
+; CHECK-NEXT:    uunpklo z3.h, z0.b
+; CHECK-NEXT:    ext z0.b, z0.b, z0.b, #8
+; CHECK-NEXT:    uunpklo z1.h, z2.b
 ; CHECK-NEXT:    ext z2.b, z2.b, z2.b, #8
+; CHECK-NEXT:    uunpklo z0.h, z0.b
+; CHECK-NEXT:    uunpklo z4.s, z3.h
 ; CHECK-NEXT:    ext z3.b, z3.b, z3.b, #8
-; CHECK-NEXT:    uunpklo z4.s, z0.h
-; CHECK-NEXT:    ext z0.b, z0.b, z0.b, #8
 ; CHECK-NEXT:    uunpklo z5.s, z1.h
-; CHECK-NEXT:    ext z1.b, z1.b, z1.b, #8
 ; CHECK-NEXT:    uunpklo z2.h, z2.b
-; CHECK-NEXT:    uunpklo z3.h, z3.b
-; CHECK-NEXT:    uunpklo z0.s, z0.h
-; CHECK-NEXT:    uunpklo z16.d, z4.s
+; CHECK-NEXT:    ext z1.b, z1.b, z1.b, #8
+; CHECK-NEXT:    uunpklo z6.s, z0.h
+; CHECK-NEXT:    uunpklo z3.s, z3.h
+; CHECK-NEXT:    ext z0.b, z0.b, z0.b, #8
+; CHECK-NEXT:    uunpklo z7.d, z4.s
 ; CHECK-NEXT:    ext z4.b, z4.b, z4.b, #8
-; CHECK-NEXT:    uunpklo z1.s, z1.h
-; CHECK-NEXT:    uunpklo z17.d, z5.s
+; CHECK-NEXT:    uunpklo z16.d, z5.s
 ; CHECK-NEXT:    ext z5.b, z5.b, z5.b, #8
-; CHECK-NEXT:    uunpklo z6.s, z2.h
-; CHECK-NEXT:    uunpklo z7.s, z3.h
+; CHECK-NEXT:    uunpklo z17.s, z2.h
 ; CHECK-NEXT:    ext z2.b, z2.b, z2.b, #8
-; CHECK-NEXT:    uunpklo z4.d, z4.s
+; CHECK-NEXT:    uunpklo z1.s, z1.h
+; CHECK-NEXT:    uunpklo z0.s, z0.h
+; CHECK-NEXT:    uunpklo z18.d, z6.s
+; CHECK-NEXT:    ext z6.b, z6.b, z6.b, #8
+; CHECK-NEXT:    uunpklo z19.d, z3.s
 ; CHECK-NEXT:    ext z3.b, z3.b, z3.b, #8
-; CHECK-NEXT:    uunpklo z19.d, z0.s
+; CHECK-NEXT:    uunpklo z4.d, z4.s
 ; CHECK-NEXT:    uunpklo z5.d, z5.s
-; CHECK-NEXT:    ext z0.b, z0.b, z0.b, #8
 ; CHECK-NEXT:    uunpklo z2.s, z2.h
-; CHECK-NEXT:    uunpklo z18.d, z6.s
-; CHECK-NEXT:    ext z6.b, z6.b, z6.b, #8
-; CHECK-NEXT:    uunpklo z3.s, z3.h
-; CHECK-NEXT:    stp q16, q4, [x1, #128]
-; CHECK-NEXT:    mov z16.d, z7.d
-; CHECK-NEXT:    uunpklo z0.d, z0.s
-; CHECK-NEXT:    stp q17, q5, [x1]
-; CHECK-NEXT:    uunpklo z5.d, z7.s
-; CHECK-NEXT:    uunpklo z4.d, z6.s
-; CHECK-NEXT:    mov z6.d, z1.d
-; CHECK-NEXT:    ext z16.b, z16.b, z7.b, #8
-; CHECK-NEXT:    mov z7.d, z2.d
-; CHECK-NEXT:    stp q19, q0, [x1, #160]
-; CHECK-NEXT:    uunpklo z0.d, z2.s
-; CHECK-NEXT:    ext z6.b, z6.b, z1.b, #8
-; CHECK-NEXT:    uunpklo z1.d, z1.s
-; CHECK-NEXT:    stp q18, q4, [x1, #192]
-; CHECK-NEXT:    mov z4.d, z3.d
-; CHECK-NEXT:    ext z7.b, z7.b, z2.b, #8
-; CHECK-NEXT:    uunpklo z16.d, z16.s
 ; CHECK-NEXT:    uunpklo z6.d, z6.s
-; CHECK-NEXT:    ext z4.b, z4.b, z3.b, #8
-; CHECK-NEXT:    uunpklo z2.d, z7.s
 ; CHECK-NEXT:    uunpklo z3.d, z3.s
-; CHECK-NEXT:    stp q5, q16, [x1, #64]
-; CHECK-NEXT:    stp q1, q6, [x1, #32]
-; CHECK-NEXT:    uunpklo z1.d, z4.s
-; CHECK-NEXT:    stp q0, q2, [x1, #224]
-; CHECK-NEXT:    stp q3, q1, [x1, #96]
+; CHECK-NEXT:    stp q16, q5, [x1]
+; CHECK-NEXT:    uunpklo z5.d, z1.s
+; CHECK-NEXT:    ext z1.b, z1.b, z1.b, #8
+; CHECK-NEXT:    stp q7, q4, [x1, #128]
+; CHECK-NEXT:    uunpklo z4.d, z17.s
+; CHECK-NEXT:    ext z17.b, z17.b, z17.b, #8
+; CHECK-NEXT:    stp q18, q6, [x1, #192]
+; CHECK-NEXT:    uunpklo z6.d, z0.s
+; CHECK-NEXT:    ext z0.b, z0.b, z0.b, #8
+; CHECK-NEXT:    stp q19, q3, [x1, #160]
+; CHECK-NEXT:    uunpklo z3.d, z2.s
+; CHECK-NEXT:    ext z2.b, z2.b, z2.b, #8
+; CHECK-NEXT:    uunpklo z7.d, z17.s
+; CHECK-NEXT:    uunpklo z1.d, z1.s
+; CHECK-NEXT:    uunpklo z0.d, z0.s
+; CHECK-NEXT:    uunpklo z2.d, z2.s
+; CHECK-NEXT:    stp q5, q1, [x1, #32]
+; CHECK-NEXT:    stp q4, q7, [x1, #64]
+; CHECK-NEXT:    stp q3, q2, [x1, #96]
+; CHECK-NEXT:    stp q6, q0, [x1, #224]
 ; CHECK-NEXT:    ret
   %a = load <32 x i8>, ptr %in
   %b = add <32 x i8> %a, %a
diff --git a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-to-fp.ll b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-to-fp.ll
index c110e89326cc0..5d4538acf707f 100644
--- a/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-to-fp.ll
+++ b/llvm/test/CodeGen/AArch64/sve-streaming-mode-fixed-length-int-to-fp.ll
@@ -207,36 +207,31 @@ define void @ucvtf_v16i16_v16f64(ptr %a, ptr %b) {
 ; CHECK-NEXT:    ext z1.b, z1.b, z1.b, #8
 ; CHECK-NEXT:    uunpklo z0.s, z0.h
 ; CHECK-NEXT:    uunpklo z1.s, z1.h
-; CHECK-NEXT:    mov z4.d, z2.d
-; CHECK-NEXT:    mov z7.d, z3.d
-; CHECK-NEXT:    mov z5.d, z0.d
-; CHECK-NEXT:    ext z4.b, z4.b, z2.b, #8
+; CHECK-NEXT:    uunpklo z4.d, z2.s
+; CHECK-NEXT:    ext z2.b, z2.b, z2.b, #8
+; CHECK-NEXT:    uunpklo z7.d, z3.s
+; CHECK-NEXT:    ext z3.b, z3.b, z3.b, #8
+; CHECK-NEXT:    uunpklo z5.d, z0.s
+; CHECK-NEXT:    ext z0.b, z0.b, z0.b, #8
 ; CHECK-NEXT:    uunpklo z2.d, z2.s
-; CHECK-NEXT:    mov z6.d, z1.d
-; CHECK-NEXT:    ext z7.b, z7.b, z3.b, #8
+; CHECK-NEXT:    uunpklo z6.d, z1.s
+; CHECK-NEXT:    ext z1.b, z1.b, z1.b, #8
 ; CHECK-NEXT:    uunpklo z3.d, z3.s
-; CHECK-NEXT:    ext z5.b, z5.b, z0.b, #8
-; CHECK-NEXT:    uunpklo z4.d, z4.s
+; CHECK-NEXT:    ucvtf z4.d, p0/m, z4.d
 ; CHECK-NEXT:    uunpklo z0.d, z0.s
-; CHECK-NEXT:    ext z6.b, z6.b, z1.b, #8
 ; CHECK-NEXT:    uunpklo z1.d, z1.s
 ; CHECK-NEXT:    ucvtf z2.d, p0/m, z2.d
+; CHECK-NEXT:    ucvtf z5.d, p0/m, z5.d
+; CHECK-NEXT:    ucvtf z6.d, p0/m, z6.d
 ; CHECK-NEXT:    ucvtf z3.d, p0/m, z3.d
-; CHECK-NEXT:    uunpklo z7.d, z7.s
-; CHECK-NEXT:    uunpklo z5.d, z5.s
-; CHECK-NEXT:    ucvtf z4.d, p0/m, z4.d
 ; CHECK-NEXT:    ucvtf z0.d, p0/m, z0.d
-; CHECK-NEXT:    uunpklo z6.d, z6.s
 ; CHECK-NEXT:    ucvtf z1.d, p0/m, z1.d
-; CHECK-NEXT:    ucvtf z5.d, p0/m, z5.d
-; CHECK-NEXT:    stp q2, q4, [x1, #64]
-; CHECK-NEXT:    movprfx z2, z6
-; CHECK-NEXT:    ucvtf z2.d, p0/m, z6.d
-; CHECK-NEXT:    stp q1, q2, [x1, #32]
-; CHECK-NEXT:    stp q0, q5, [x1, #96]
-; CHECK-NEXT:    movprfx z0, z7
-; CHECK-NEXT:    ucvtf z0.d, p0/m, z7.d
-; CHECK-NEXT:    stp q3, q0, [x1]
+; CHECK-NEXT:    stp q4, q2, [x1, #64]
+; CHECK-NEXT:    movprfx z2, z7
+; CHECK-NEXT:    ucvtf z2.d, p0/m, z7.d
+; CHECK-NEXT:    stp q2, q3, [x1]
+; CHECK-NEXT:    stp q5, q0, [x1, #96]
+; CHECK-NEXT:    stp q6, q1, [x1, #32]
 ; CHECK-NEXT:    ret
   %op1 = load <16 x i16>, ptr %a
   %res = uitofp <16 x i16> %op1 to <16 x double>
@@ -780,36 +775,31 @@ define void @scvtf_v16i16_v16f64(ptr %a, ptr %b) {
 ; CHECK-NEXT:    ext z1.b, z1.b, z1.b, #8
 ; CHECK-NEXT:    sunpklo z0.s, z0.h
 ; CHECK-NEXT:    sunpklo z1.s, z1.h
-; CHECK-NEXT:    mov z4.d, z2.d
-; CHECK-NEXT:    mov z7.d, z3.d
-; CHECK-NEXT:    mov z5.d, z0.d
-; CHECK-NEXT:    ext z4.b, z4.b, z2.b, #8
+; CHECK-NEXT:    sunpklo z4.d, z2.s
+; CHECK-NEXT:    ext z2.b, z2.b, z2.b, #8
+; CHECK-NEXT:    sunpklo z7.d, z3.s
+; CHECK-NEXT:    ext z3.b, z3.b, z3.b, #8
+; CHECK-NEXT:    sunpklo z5.d, z0.s
+; CHECK-NEXT:    ext z0.b, z0.b, z0.b, #8
 ; CHECK-NEXT:    sunpklo z2.d, z2.s
-; CHECK-NEXT:    mov z6.d, z1.d
-; CHECK-NEXT:    ext z7.b, z7.b, z3.b, #8
+; CHECK-NEXT:    sunpklo z6.d, z1.s
+; CHECK-NEXT:    ext z1.b, z1.b, z1.b, #8
 ; CHECK-NEXT:    sunpklo z3.d, z3.s
-; CHECK-NEXT:    ext z5.b, z5.b, z0.b, #8
-; CHECK-NEXT:    sunpklo z4.d, z4.s
+; CHECK-NEXT:    scvtf z4.d, p0/m, z4.d
 ; CHECK-NEXT:    sunpklo z0.d, z0.s
-; CHECK-NEXT:    ext z6.b, z6.b, z1.b, #8
 ; CHECK-NEXT:    sunpklo z1.d, z1.s
 ; CHECK-NEXT:    scvtf z2.d, p0/m, z2.d
+; CHECK-NEXT:    scvtf z5.d, p0/m, z5.d
+; CHECK-NEXT:    scvtf z6.d, p0/m, z6.d
 ; CHECK-NEXT:    scvtf z3.d, p0/m, z3.d
-; CHECK-NEXT:    sunpklo z7.d, z7.s
-; CHECK-NEXT:    sunpklo z5.d, z5.s
-; CHECK-NEXT:    scvtf z4.d, p0/m, z4.d
 ; CHECK-NEXT:    scvtf z0.d, p0/m, z0.d
-; CHECK-NEXT:    sunpklo z6.d, z6.s
 ; CHECK-NEXT:    scvtf z1.d, p0/m, z1.d
-; CHECK-NEXT:    scvtf z5.d, p0/m, z5.d
-; CHECK-NEXT:    stp q2, q4, [x1, #64]
-; CHECK-NEXT:    movprfx z2, z6
-; CHECK-NEXT:    scvtf z2.d, p0/m, z6.d
-; CHECK-NEXT:    stp q1, q2, [x1, #32]
-; CHECK-NEXT:    stp q0, q5, [x1, #96]
-; CHECK-NEXT:    movprfx z0, z7
-; CHECK-NEXT:    scvtf z0.d, p0/m, z7.d
-; CHECK-NEXT:    stp q3, q0, [x1]
+; CHECK-NEXT:    stp q4, q2, [x1, #64]
+; CHECK-NEXT:    movprfx z2, z7
+; CHECK-NEXT:    scvtf z2.d, p0/m, z7.d
+; CHECK-NEXT:    stp q2, q3, [x1]
+; CHECK-NEXT:    stp q5, q0, [x1, #96]
+; CHECK-NEXT:    stp q6, q1, [x1, #32]
 ; CHECK-NEXT:    ret
   %op1 = load <16 x i16>, ptr %a
   %res = sitofp <16 x i16> %op1 to <16 x double>
diff --git a/llvm/test/CodeGen/ARM/copy-by-struct-i32.ll b/llvm/test/CodeGen/ARM/copy-by-struct-i32.ll
index 34aab4c04b109..8f134e0ac7f18 100644
--- a/llvm/test/CodeGen/ARM/copy-by-struct-i32.ll
+++ b/llvm/test/CodeGen/ARM/copy-by-struct-i32.ll
@@ -22,23 +22,23 @@ define arm_aapcscc void @s(ptr %q, ptr %p) {
 ; ASSEMBLY-NEXT:    ldr r2, [r1, #8]
 ; ASSEMBLY-NEXT:    ldr r3, [r1, #12]
 ; ASSEMBLY-NEXT:    strd r4, r5, [sp, #128]
-; ASSEMBLY-NEXT:    add r5, r1, #16
-; ASSEMBLY-NEXT:    mov r4, sp
-; ASSEMBLY-NEXT:    vld1.32 {d16}, [r5]!
-; ASSEMBLY-NEXT:    vst1.32 {d16}, [r4]!
-; ASSEMBLY-NEXT:    vld1.32 {d16}, [r5]!
-; ASSEMBLY-NEXT:    vst1.32 {d16}, [r4]!
-; ASSEMBLY-NEXT:    vld1.32 {d16}, [r5]!
-; ASSEMBLY-NEXT:    vst1.32 {d16}, [r4]!
-; ASSEMBLY-NEXT:    vld1.32 {d16}, [r5]!
-; ASSEMBLY-NEXT:    vst1.32 {d16}, [r4]!
-; ASSEMBLY-NEXT:    vld1.32 {d16}, [r5]!
-; ASSEMBLY-NEXT:    vst1.32 {d16}, [r4]!
-; ASSEMBLY-NEXT:    vld1.32 {d16}, [r5]!
-; ASSEMBLY-NEXT:    vst1.32 {d16}, [r4]!
-; ASSEMBLY-NEXT:    vld1.32 {d16}, [r5]!
-; ASSEMBLY-NEXT:    vst1.32 {d16}, [r4]!
+; ASSEMBLY-NEXT:    add r4, r1, #16
+; ASSEMBLY-NEXT:    mov r5, sp
+; ASSEMBLY-NEXT:    vld1.32 {d16}, [r4]!
+; ASSEMBLY-NEXT:    vst1.32 {d16}, [r5]!
+; ASSEMBLY-NEXT:    vld1.32 {d16}, [r4]!
+; ASSEMBLY-NEXT:    vst1.32 {d16}, [r5]!
+; ASSEMBLY-NEXT:    vld1.32 {d16}, [r4]!
+; ASSEMBLY-NEXT:    vst1.32 {d16}, [r5]!
+; ASSEMBLY-NEXT:    vld1.32 {d16}, [r4]!
+; ASSEMBLY-NEXT:    vst1.32 {d16}, [r5]!
+; ASSEMBLY-NEXT:    vld1.32 {d16}, [r4]!
+; ASSEMBLY-NEXT:    vst1.32 {d16}, [r5]!
+; ASSEMBLY-NEXT:    vld1.32 {d16}, [r4]!
+; ASSEMBLY-NEXT:    vst1.32 {d16}, [r5]!
+; ASSEMBLY-NEXT:    vld1.32 {d16}, [r4]!
 ; ASSEMBLY-NEXT:    movw r4, #72
+; ASSEMBLY-NEXT:    vst1.32 {d16}, [r5]!
 ; ASSEMBLY-NEXT:  .LBB0_1: @ %entry
 ; ASSEMBLY-NEXT:    @ =>This Inner Loop Header: Depth=1
 ; ASSEMBLY-NEXT:    vld1.32 {d16}, [r1]!
@@ -58,3 +58,5 @@ entry:
 }
 
 declare arm_aapcscc void @r(...)
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; BEFORE-EXPAND: {{.*}}
diff --git a/llvm/test/CodeGen/ARM/vselect_imax.ll b/llvm/test/CodeGen/ARM/vselect_imax.ll
index 37f511fcc68cc..89072683fb01a 100644
--- a/llvm/test/CodeGen/ARM/vselect_imax.ll
+++ b/llvm/test/CodeGen/ARM/vselect_imax.ll
@@ -242,198 +242,195 @@ define void @func_blend20(ptr %loadaddr, ptr %loadaddr2,
                            ptr %blend, ptr %storeaddr) {
 ; CHECK-LABEL: func_blend20:
 ; CHECK:       @ %bb.0:
-; CHECK-NEXT:    .save {r4, r5, r6, r7, r8, r9, r10, lr}
-; CHECK-NEXT:    push {r4, r5, r6, r7, r8, r9, r10, lr}
+; CHECK-NEXT:    .save {r4, r5, r6, r7, r8, r9, r11, lr}
+; CHECK-NEXT:    push {r4, r5, r6, r7, r8, r9, r11, lr}
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11}
-; CHECK-NEXT:    mov r8, r1
-; CHECK-NEXT:    mov lr, r0
-; CHECK-NEXT:    vld1.64 {d16, d17}, [r8:128]!
-; CHECK-NEXT:    add r9, r0, #64
-; CHECK-NEXT:    add r10, r1, #64
+; CHECK-NEXT:    add lr, r1, #64
+; CHECK-NEXT:    add r8, r0, #64
+; CHECK-NEXT:    vld1.64 {d16, d17}, [r1:128]!
+; CHECK-NEXT:    mov r9, #0
 ; CHECK-NEXT:    mov r12, #0
-; CHECK-NEXT:    vld1.64 {d22, d23}, [lr:128]!
-; CHECK-NEXT:    vld1.64 {d18, d19}, [r8:128]!
-; CHECK-NEXT:    vld1.64 {d20, d21}, [lr:128]!
-; CHECK-NEXT:    vmov r6, r4, d19
-; CHECK-NEXT:    vmov r5, r7, d21
-; CHECK-NEXT:    vld1.64 {d4, d5}, [r9:128]!
-; CHECK-NEXT:    vld1.64 {d6, d7}, [r10:128]!
-; CHECK-NEXT:    vld1.64 {d0, d1}, [r10:128]!
-; CHECK-NEXT:    vld1.64 {d2, d3}, [r9:128]!
-; CHECK-NEXT:    subs r6, r5, r6
-; CHECK-NEXT:    sbcs r4, r7, r4
-; CHECK-NEXT:    vmov r5, r6, d18
-; CHECK-NEXT:    vmov r7, r2, d20
-; CHECK-NEXT:    mov r4, #0
-; CHECK-NEXT:    movlt r4, #1
-; CHECK-NEXT:    cmp r4, #0
-; CHECK-NEXT:    mvnne r4, #0
-; CHECK-NEXT:    vdup.32 d31, r4
-; CHECK-NEXT:    subs r5, r7, r5
-; CHECK-NEXT:    sbcs r2, r2, r6
-; CHECK-NEXT:    vmov r4, r5, d3
+; CHECK-NEXT:    vld1.64 {d22, d23}, [r0:128]!
+; CHECK-NEXT:    vld1.64 {d18, d19}, [r1:128]!
+; CHECK-NEXT:    vld1.64 {d20, d21}, [r0:128]!
+; CHECK-NEXT:    vmov r5, r4, d19
+; CHECK-NEXT:    vmov r2, r7, d21
+; CHECK-NEXT:    vld1.64 {d4, d5}, [r8:128]!
+; CHECK-NEXT:    vld1.64 {d6, d7}, [lr:128]!
+; CHECK-NEXT:    vld1.64 {d0, d1}, [lr:128]!
+; CHECK-NEXT:    vld1.64 {d2, d3}, [r8:128]!
+; CHECK-NEXT:    subs r2, r2, r5
+; CHECK-NEXT:    sbcs r2, r7, r4
+; CHECK-NEXT:    vmov r4, r5, d18
+; CHECK-NEXT:    vmov r7, r6, d20
 ; CHECK-NEXT:    mov r2, #0
 ; CHECK-NEXT:    movlt r2, #1
 ; CHECK-NEXT:    cmp r2, #0
 ; CHECK-NEXT:    mvnne r2, #0
+; CHECK-NEXT:    vdup.32 d31, r2
+; CHECK-NEXT:    mov r2, #0
+; CHECK-NEXT:    subs r4, r7, r4
+; CHECK-NEXT:    sbcs r4, r6, r5
+; CHECK-NEXT:    vmov r5, r6, d3
+; CHECK-NEXT:    movlt r2, #1
+; CHECK-NEXT:    cmp r2, #0
+; CHECK-NEXT:    mvnne r2, #0
 ; CHECK-NEXT:    vdup.32 d30, r2
-; CHECK-NEXT:    vmov r0, r2, d1
-; CHECK-NEXT:    subs r0, r4, r0
-; CHECK-NEXT:    sbcs r0, r5, r2
-; CHECK-NEXT:    vmov r4, r5, d2
-; CHECK-NEXT:    mov r0, #0
-; CHECK-NEXT:    movlt r0, #1
-; CHECK-NEXT:    cmp r0, #0
-; CHECK-NEXT:    mvnne r0, #0
-; CHECK-NEXT:    vdup.32 d9, r0
-; CHECK-NEXT:    vmov r0, r2, d0
-; CHECK-NEXT:    subs r0, r4, r0
-; CHECK-NEXT:    sbcs r0, r5, r2
-; CHECK-NEXT:    vmov r4, r5, d5
-; CHECK-NEXT:    mov r0, #0
-; CHECK-NEXT:    movlt r0, #1
-; CHECK-NEXT:    cmp r0, #0
-; CHECK-NEXT:    mvnne r0, #0
-; CHECK-NEXT:    vdup.32 d8, r0
-; CHECK-NEXT:    vmov r0, r2, d7
-; CHECK-NEXT:    subs r0, r4, r0
-; CHECK-NEXT:    sbcs r0, r5, r2
-; CHECK-NEXT:    vmov r4, r5, d4
-; CHECK-NEXT:    mov r0, #0
-; CHECK-NEXT:    movlt r0, #1
-; CHECK-NEXT:    cmp r0, #0
-; CHECK-NEXT:    mvnne r0, #0
-; CHECK-NEXT:    vdup.32 d11, r0
-; CHECK-NEXT:    vmov r0, r2, d6
-; CHECK-NEXT:    subs r0, r4, r0
-; CHECK-NEXT:    sbcs r0, r5, r2
-; CHECK-NEXT:    vmov r4, r5, d23
-; CHECK-NEXT:    mov r0, #0
-; CHECK-NEXT:    movlt r0, #1
-; CHECK-NEXT:    cmp r0, #0
-; CHECK-NEXT:    mvnne r0, #0
-; CHECK-NEXT:    vdup.32 d10, r0
-; CHECK-NEXT:    vmov r0, r2, d17
-; CHECK-NEXT:    subs r0, r4, r0
-; CHECK-NEXT:    sbcs r0, r5, r2
-; CHECK-NEXT:    vmov r4, r5, d22
-; CHECK-NEXT:    mov r0, #0
-; CHECK-NEXT:    movlt r0, #1
-; CHECK-NEXT:    cmp r0, #0
-; CHECK-NEXT:    mvnne r0, #0
-; CHECK-NEXT:    vdup.32 d25, r0
-; CHECK-NEXT:    vmov r0, r2, d16
-; CHECK-NEXT:    subs r0, r4, r0
-; CHECK-NEXT:    sbcs r0, r5, r2
-; CHECK-NEXT:    mov r0, #0
-; CHECK-NEXT:    movlt r0, #1
-; CHECK-NEXT:    cmp r0, #0
-; CHECK-NEXT:    mvnne r0, #0
-; CHECK-NEXT:    vdup.32 d24, r0
+; CHECK-NEXT:    vmov r2, r4, d1
+; CHECK-NEXT:    subs r2, r5, r2
+; CHECK-NEXT:    sbcs r2, r6, r4
+; CHECK-NEXT:    vmov r5, r6, d2
+; CHECK-NEXT:    mov r2, #0
+; CHECK-NEXT:    movlt r2, #1
+; CHECK-NEXT:    cmp r2, #0
+; CHECK-NEXT:    mvnne r2, #0
+; CHECK-NEXT:    vdup.32 d9, r2
+; CHECK-NEXT:    vmov r2, r4, d0
+; CHECK-NEXT:    subs r2, r5, r2
+; CHECK-NEXT:    sbcs r2, r6, r4
+; CHECK-NEXT:    vmov r5, r6, d5
+; CHECK-NEXT:    mov r2, #0
+; CHECK-NEXT:    movlt r2, #1
+; CHECK-NEXT:    cmp r2, #0
+; CHECK-NEXT:    mvnne r2, #0
+; CHECK-NEXT:    vdup.32 d8, r2
+; CHECK-NEXT:    vmov r2, r4, d7
+; CHECK-NEXT:    subs r2, r5, r2
+; CHECK-NEXT:    sbcs r2, r6, r4
+; CHECK-NEXT:    vmov r5, r6, d4
+; CHECK-NEXT:    mov r2, #0
+; CHECK-NEXT:    movlt r2, #1
+; CHECK-NEXT:    cmp r2, #0
+; CHECK-NEXT:    mvnne r2, #0
+; CHECK-NEXT:    vdup.32 d11, r2
+; CHECK-NEXT:    vmov r2, r4, d6
+; CHECK-NEXT:    subs r2, r5, r2
+; CHECK-NEXT:    sbcs r2, r6, r4
+; CHECK-NEXT:    vmov r5, r6, d23
+; CHECK-NEXT:    mov r2, #0
+; CHECK-NEXT:    movlt r2, #1
+; CHECK-NEXT:    cmp r2, #0
+; CHECK-NEXT:    mvnne r2, #0
+; CHECK-NEXT:    vdup.32 d10, r2
+; CHECK-NEXT:    vmov r2, r4, d17
+; CHECK-NEXT:    subs r2, r5, r2
+; CHECK-NEXT:    sbcs r2, r6, r4
+; CHECK-NEXT:    vmov r5, r6, d22
+; CHECK-NEXT:    mov r2, #0
+; CHECK-NEXT:    movlt r2, #1
+; CHECK-NEXT:    cmp r2, #0
+; CHECK-NEXT:    mvnne r2, #0
+; CHECK-NEXT:    vdup.32 d25, r2
+; CHECK-NEXT:    vmov r2, r4, d16
+; CHECK-NEXT:    subs r2, r5, r2
+; CHECK-NEXT:    sbcs r2, r6, r4
+; CHECK-NEXT:    mov r2, #0
+; CHECK-NEXT:    movlt r2, #1
+; CHECK-NEXT:    cmp r2, #0
+; CHECK-NEXT:    mvnne r2, #0
+; CHECK-NEXT:    vdup.32 d24, r2
 ; CHECK-NEXT:    vorr q13, q12, q12
 ; CHECK-NEXT:    vbsl q13, q11, q8
-; CHECK-NEXT:    vld1.64 {d24, d25}, [r9:128]!
+; CHECK-NEXT:    vld1.64 {d24, d25}, [r8:128]!
 ; CHECK-NEXT:    vorr q8, q5, q5
-; CHECK-NEXT:    vld1.64 {d28, d29}, [r10:128]!
+; CHECK-NEXT:    vld1.64 {d28, d29}, [lr:128]!
 ; CHECK-NEXT:    vbsl q8, q2, q3
-; CHECK-NEXT:    vld1.64 {d6, d7}, [r8:128]!
-; CHECK-NEXT:    vld1.64 {d22, d23}, [r8:128]
-; CHECK-NEXT:    vld1.64 {d4, d5}, [lr:128]!
+; CHECK-NEXT:    vld1.64 {d6, d7}, [r1:128]!
+; CHECK-NEXT:    vld1.64 {d4, d5}, [r0:128]!
+; CHECK-NEXT:    vld1.64 {d22, d23}, [r1:128]
 ; CHECK-NEXT:    vbif q10, q9, q15
 ; CHECK-NEXT:    vorr q9, q4, q4
-; CHECK-NEXT:    vmov r0, r2, d22
 ; CHECK-NEXT:    vbsl q9, q1, q0
-; CHECK-NEXT:    vld1.64 {d30, d31}, [lr:128]
-; CHECK-NEXT:    mov lr, #0
-; CHECK-NEXT:    vmov r7, r5, d30
-; CHECK-NEXT:    vld1.64 {d0, d1}, [r9:128]
-; CHECK-NEXT:    vld1.64 {d2, d3}, [r10:128]
-; CHECK-NEXT:    subs r0, r7, r0
-; CHECK-NEXT:    sbcs r0, r5, r2
-; CHECK-NEXT:    vmov r5, r4, d24
-; CHECK-NEXT:    vmov r0, r7, d28
-; CHECK-NEXT:    movlt lr, #1
-; CHECK-NEXT:    cmp lr, #0
-; CHECK-NEXT:    mvnne lr, #0
-; CHECK-NEXT:    subs r0, r5, r0
-; CHECK-NEXT:    sbcs r0, r4, r7
-; CHECK-NEXT:    vmov r7, r5, d29
-; CHECK-NEXT:    vmov r4, r6, d25
-; CHECK-NEXT:    mov r0, #0
-; CHECK-NEXT:    movlt r0, #1
-; CHECK-NEXT:    cmp r0, #0
-; CHECK-NEXT:    mvnne r0, #0
-; CHECK-NEXT:    subs r7, r4, r7
+; CHECK-NEXT:    vld1.64 {d30, d31}, [r0:128]
+; CHECK-NEXT:    vmov r0, r1, d22
+; CHECK-NEXT:    vmov r2, r4, d30
+; CHECK-NEXT:    vld1.64 {d0, d1}, [r8:128]
+; CHECK-NEXT:    vld1.64 {d2, d3}, [lr:128]
+; CHECK-NEXT:    subs r0, r2, r0
+; CHECK-NEXT:    sbcs r0, r4, r1
+; CHECK-NEXT:    vmov r1, r2, d28
+; CHECK-NEXT:    vmov r4, r5, d24
+; CHECK-NEXT:    movlt r9, #1
+; CHECK-NEXT:    cmp r9, #0
+; CHECK-NEXT:    mvnne r9, #0
+; CHECK-NEXT:    subs r1, r4, r1
+; CHECK-NEXT:    sbcs r1, r5, r2
+; CHECK-NEXT:    vmov r2, r4, d29
+; CHECK-NEXT:    vmov r5, r6, d25
+; CHECK-NEXT:    mov r1, #0
+; CHECK-NEXT:    movlt r1, #1
+; CHECK-NEXT:    cmp r1, #0
+; CHECK-NEXT:    mvnne r1, #0
+; CHECK-NEXT:    subs r2, r5, r2
+; CHECK-NEXT:    sbcs r2, r6, r4
+; CHECK-NEXT:    vmov r6, r7, d31
+; CHECK-NEXT:    vmov r2, r5, d23
 ; CHECK-NEXT:    mov r4, #0
-; CHECK-NEXT:    sbcs r7, r6, r5
-; CHECK-NEXT:    vmov r5, r1, d31
-; CHECK-NEXT:    vmov r7, r6, d23
 ; CHECK-NEXT:    movlt r4, #1
 ; CHECK-NEXT:    cmp r4, #0
 ; CHECK-NEXT:    mvnne r4, #0
-; CHECK-NEXT:    subs r7, r5, r7
+; CHECK-NEXT:    subs r2, r6, r2
+; CHECK-NEXT:    sbcs r2, r7, r5
+; CHECK-NEXT:    vmov r7, r0, d5
+; CHECK-NEXT:    vmov r2, r6, d7
 ; CHECK-NEXT:    mov r5, #0
-; CHECK-NEXT:    sbcs r1, r1, r6
-; CHECK-NEXT:    vmov r6, r2, d5
-; CHECK-NEXT:    vmov r1, r7, d7
 ; CHECK-NEXT:    movlt r5, #1
 ; CHECK-NEXT:    cmp r5, #0
 ; CHECK-NEXT:    mvnne r5, #0
-; CHECK-NEXT:    subs r1, r6, r1
-; CHECK-NEXT:    sbcs r1, r2, r7
+; CHECK-NEXT:    subs r2, r7, r2
+; CHECK-NEXT:    sbcs r0, r0, r6
 ; CHECK-NEXT:    vmov r6, r7, d4
-; CHECK-NEXT:    mov r1, #0
-; CHECK-NEXT:    movlt r1, #1
-; CHECK-NEXT:    cmp r1, #0
-; CHECK-NEXT:    mvnne r1, #0
-; CHECK-NEXT:    vdup.32 d9, r1
-; CHECK-NEXT:    vmov r1, r2, d6
-; CHECK-NEXT:    subs r1, r6, r1
-; CHECK-NEXT:    sbcs r1, r7, r2
-; CHECK-NEXT:    vmov r6, r7, d0
-; CHECK-NEXT:    mov r1, #0
-; CHECK-NEXT:    movlt r1, #1
-; CHECK-NEXT:    cmp r1, #0
-; CHECK-NEXT:    mvnne r1, #0
-; CHECK-NEXT:    vdup.32 d8, r1
-; CHECK-NEXT:    vmov r1, r2, d2
+; CHECK-NEXT:    mov r0, #0
+; CHECK-NEXT:    movlt r0, #1
+; CHECK-NEXT:    cmp r0, #0
+; CHECK-NEXT:    mvnne r0, #0
+; CHECK-NEXT:    vdup.32 d9, r0
+; CHECK-NEXT:    vmov r0, r2, d6
+; CHECK-NEXT:    subs r0, r6, r0
+; CHECK-NEXT:    sbcs r0, r7, r2
+; CHECK-NEXT:    vmov r7, r6, d0
+; CHECK-NEXT:    mov r0, #0
+; CHECK-NEXT:    movlt r0, #1
+; CHECK-NEXT:    cmp r0, #0
+; CHECK-NEXT:    mvnne r0, #0
+; CHECK-NEXT:    vdup.32 d8, r0
+; CHECK-NEXT:    vmov r0, r2, d2
 ; CHECK-NEXT:    vbif q2, q3, q4
 ; CHECK-NEXT:    vdup.32 d7, r5
+; CHECK-NEXT:    mov r5, #0
 ; CHECK-NEXT:    vdup.32 d9, r4
-; CHECK-NEXT:    vmov r4, r5, d1
-; CHECK-NEXT:    vdup.32 d8, r0
-; CHECK-NEXT:    mov r0, r3
-; CHECK-NEXT:    vst1.64 {d26, d27}, [r0:128]!
+; CHECK-NEXT:    vdup.32 d8, r1
+; CHECK-NEXT:    add r1, r3, #64
+; CHECK-NEXT:    vst1.64 {d26, d27}, [r3:128]!
 ; CHECK-NEXT:    vbif q12, q14, q4
-; CHECK-NEXT:    vdup.32 d6, lr
+; CHECK-NEXT:    vdup.32 d6, r9
 ; CHECK-NEXT:    vbit q11, q15, q3
-; CHECK-NEXT:    vst1.64 {d20, d21}, [r0:128]!
-; CHECK-NEXT:    subs r1, r6, r1
-; CHECK-NEXT:    mov r6, #0
-; CHECK-NEXT:    sbcs r1, r7, r2
-; CHECK-NEXT:    vmov r1, r2, d3
-; CHECK-NEXT:    movlt r6, #1
-; CHECK-NEXT:    subs r1, r4, r1
-; CHECK-NEXT:    sbcs r1, r5, r2
+; CHECK-NEXT:    vst1.64 {d20, d21}, [r3:128]!
+; CHECK-NEXT:    subs r0, r7, r0
+; CHECK-NEXT:    sbcs r0, r6, r2
+; CHECK-NEXT:    vmov r7, r6, d1
+; CHECK-NEXT:    vmov r0, r2, d3
+; CHECK-NEXT:    movlt r5, #1
+; CHECK-NEXT:    subs r0, r7, r0
+; CHECK-NEXT:    sbcs r0, r6, r2
 ; CHECK-NEXT:    movlt r12, #1
 ; CHECK-NEXT:    cmp r12, #0
 ; CHECK-NEXT:    mvnne r12, #0
-; CHECK-NEXT:    cmp r6, #0
+; CHECK-NEXT:    cmp r5, #0
 ; CHECK-NEXT:    vdup.32 d27, r12
-; CHECK-NEXT:    mvnne r6, #0
-; CHECK-NEXT:    vdup.32 d26, r6
+; CHECK-NEXT:    mvnne r5, #0
+; CHECK-NEXT:    vdup.32 d26, r5
 ; CHECK-NEXT:    vorr q10, q13, q13
 ; CHECK-NEXT:    vbsl q10, q0, q1
-; CHECK-NEXT:    vst1.64 {d4, d5}, [r0:128]!
-; CHECK-NEXT:    vst1.64 {d22, d23}, [r0:128]
-; CHECK-NEXT:    add r0, r3, #64
-; CHECK-NEXT:    vst1.64 {d16, d17}, [r0:128]!
-; CHECK-NEXT:    vst1.64 {d18, d19}, [r0:128]!
-; CHECK-NEXT:    vst1.64 {d24, d25}, [r0:128]!
-; CHECK-NEXT:    vst1.64 {d20, d21}, [r0:128]
+; CHECK-NEXT:    vst1.64 {d16, d17}, [r1:128]!
+; CHECK-NEXT:    vst1.64 {d18, d19}, [r1:128]!
+; CHECK-NEXT:    vst1.64 {d4, d5}, [r3:128]!
+; CHECK-NEXT:    vst1.64 {d24, d25}, [r1:128]!
+; CHECK-NEXT:    vst1.64 {d22, d23}, [r3:128]
+; CHECK-NEXT:    vst1.64 {d20, d21}, [r1:128]
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11}
-; CHECK-NEXT:    pop {r4, r5, r6, r7, r8, r9, r10, lr}
+; CHECK-NEXT:    pop {r4, r5, r6, r7, r8, r9, r11, lr}
 ; CHECK-NEXT:    mov pc, lr
   %v0 = load %T0_20, ptr %loadaddr
   %v1 = load %T0_20, ptr %loadaddr2
diff --git a/llvm/test/CodeGen/Thumb2/mve-shuffle.ll b/llvm/test/CodeGen/Thumb2/mve-shuffle.ll
index f4643f8c6c4a1..a1f64f01887b6 100644
--- a/llvm/test/CodeGen/Thumb2/mve-shuffle.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-shuffle.ll
@@ -346,42 +346,42 @@ entry:
 define arm_aapcs_vfpcc <8 x i16> @shuffle3step_i16(<32 x i16> %src) {
 ; CHECK-LABEL: shuffle3step_i16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9}
-; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    .vsave {d8, d9, d10}
+; CHECK-NEXT:    vpush {d8, d9, d10}
 ; CHECK-NEXT:    vmovx.f16 s12, s0
-; CHECK-NEXT:    vmov.f32 s16, s1
+; CHECK-NEXT:    vmovx.f16 s13, s3
 ; CHECK-NEXT:    vins.f16 s12, s2
 ; CHECK-NEXT:    vmovx.f16 s2, s2
-; CHECK-NEXT:    vins.f16 s16, s2
+; CHECK-NEXT:    vmovx.f16 s20, s1
+; CHECK-NEXT:    vins.f16 s1, s2
 ; CHECK-NEXT:    vmovx.f16 s2, s5
-; CHECK-NEXT:    vmov.f32 s17, s4
-; CHECK-NEXT:    vmovx.f16 s13, s3
-; CHECK-NEXT:    vins.f16 s17, s2
+; CHECK-NEXT:    vins.f16 s13, s5
+; CHECK-NEXT:    vmovx.f16 s5, s4
+; CHECK-NEXT:    vins.f16 s4, s2
 ; CHECK-NEXT:    vmov.f32 s18, s7
 ; CHECK-NEXT:    vmovx.f16 s2, s8
-; CHECK-NEXT:    vmov.f32 s19, s10
 ; CHECK-NEXT:    vins.f16 s18, s2
 ; CHECK-NEXT:    vmovx.f16 s2, s11
-; CHECK-NEXT:    vins.f16 s19, s2
-; CHECK-NEXT:    vmovx.f16 s2, s1
-; CHECK-NEXT:    vins.f16 s0, s2
-; CHECK-NEXT:    vmovx.f16 s2, s4
-; CHECK-NEXT:    vins.f16 s3, s2
-; CHECK-NEXT:    vmovx.f16 s2, s7
+; CHECK-NEXT:    vmov.f32 s17, s4
 ; CHECK-NEXT:    vmovx.f16 s4, s10
+; CHECK-NEXT:    vins.f16 s10, s2
+; CHECK-NEXT:    vmovx.f16 s2, s7
 ; CHECK-NEXT:    vmovx.f16 s14, s6
 ; CHECK-NEXT:    vmovx.f16 s15, s9
 ; CHECK-NEXT:    vins.f16 s6, s2
+; CHECK-NEXT:    vins.f16 s3, s5
 ; CHECK-NEXT:    vins.f16 s9, s4
+; CHECK-NEXT:    vmov.f32 s16, s1
 ; CHECK-NEXT:    vmov.f32 s1, s3
 ; CHECK-NEXT:    vins.f16 s14, s8
 ; CHECK-NEXT:    vins.f16 s15, s11
-; CHECK-NEXT:    vins.f16 s13, s5
+; CHECK-NEXT:    vins.f16 s0, s20
 ; CHECK-NEXT:    vmov.f32 s2, s6
 ; CHECK-NEXT:    vmov.f32 s3, s9
+; CHECK-NEXT:    vmov.f32 s19, s10
 ; CHECK-NEXT:    vadd.i16 q0, q0, q3
 ; CHECK-NEXT:    vadd.i16 q0, q0, q4
-; CHECK-NEXT:    vpop {d8, d9}
+; CHECK-NEXT:    vpop {d8, d9, d10}
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = shufflevector <32 x i16> %src, <32 x i16> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
@@ -1263,42 +1263,42 @@ entry:
 define arm_aapcs_vfpcc <8 x half> @shuffle3step_f16(<32 x half> %src) {
 ; CHECKFP-LABEL: shuffle3step_f16:
 ; CHECKFP:       @ %bb.0: @ %entry
-; CHECKFP-NEXT:    .vsave {d8, d9}
-; CHECKFP-NEXT:    vpush {d8, d9}
+; CHECKFP-NEXT:    .vsave {d8, d9, d10}
+; CHECKFP-NEXT:    vpush {d8, d9, d10}
+; CHECKFP-NEXT:    vmovx.f16 s12, s2
+; CHECKFP-NEXT:    vmovx.f16 s15, s1
+; CHECKFP-NEXT:    vins.f16 s1, s12
+; CHECKFP-NEXT:    vmovx.f16 s12, s5
+; CHECKFP-NEXT:    vmovx.f16 s20, s4
+; CHECKFP-NEXT:    vins.f16 s4, s12
 ; CHECKFP-NEXT:    vmov.f32 s13, s4
-; CHECKFP-NEXT:    vmovx.f16 s4, s4
-; CHECKFP-NEXT:    vmovx.f16 s17, s3
-; CHECKFP-NEXT:    vins.f16 s3, s4
+; CHECKFP-NEXT:    vmovx.f16 s4, s8
+; CHECKFP-NEXT:    vmov.f32 s14, s7
+; CHECKFP-NEXT:    vmovx.f16 s16, s0
+; CHECKFP-NEXT:    vins.f16 s14, s4
 ; CHECKFP-NEXT:    vmovx.f16 s4, s7
 ; CHECKFP-NEXT:    vmovx.f16 s18, s6
-; CHECKFP-NEXT:    vmovx.f16 s16, s0
 ; CHECKFP-NEXT:    vins.f16 s6, s4
-; CHECKFP-NEXT:    vmovx.f16 s14, s2
-; CHECKFP-NEXT:    vmov.f32 s12, s1
 ; CHECKFP-NEXT:    vmovx.f16 s4, s10
 ; CHECKFP-NEXT:    vmovx.f16 s19, s9
-; CHECKFP-NEXT:    vins.f16 s12, s14
-; CHECKFP-NEXT:    vmovx.f16 s14, s5
+; CHECKFP-NEXT:    vmovx.f16 s17, s3
 ; CHECKFP-NEXT:    vins.f16 s16, s2
 ; CHECKFP-NEXT:    vmovx.f16 s2, s11
-; CHECKFP-NEXT:    vmovx.f16 s15, s8
-; CHECKFP-NEXT:    vins.f16 s18, s8
-; CHECKFP-NEXT:    vmovx.f16 s8, s1
+; CHECKFP-NEXT:    vins.f16 s3, s20
 ; CHECKFP-NEXT:    vins.f16 s9, s4
-; CHECKFP-NEXT:    vins.f16 s13, s14
-; CHECKFP-NEXT:    vmov.f32 s14, s7
+; CHECKFP-NEXT:    vmov.f32 s12, s1
 ; CHECKFP-NEXT:    vins.f16 s10, s2
+; CHECKFP-NEXT:    vins.f16 s0, s15
 ; CHECKFP-NEXT:    vmov.f32 s1, s3
+; CHECKFP-NEXT:    vins.f16 s18, s8
 ; CHECKFP-NEXT:    vins.f16 s19, s11
 ; CHECKFP-NEXT:    vins.f16 s17, s5
-; CHECKFP-NEXT:    vins.f16 s0, s8
 ; CHECKFP-NEXT:    vmov.f32 s2, s6
 ; CHECKFP-NEXT:    vmov.f32 s3, s9
-; CHECKFP-NEXT:    vins.f16 s14, s15
 ; CHECKFP-NEXT:    vmov.f32 s15, s10
 ; CHECKFP-NEXT:    vadd.f16 q0, q0, q4
 ; CHECKFP-NEXT:    vadd.f16 q0, q0, q3
-; CHECKFP-NEXT:    vpop {d8, d9}
+; CHECKFP-NEXT:    vpop {d8, d9, d10}
 ; CHECKFP-NEXT:    bx lr
 entry:
   %s1 = shufflevector <32 x half> %src, <32 x half> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
diff --git a/llvm/test/CodeGen/Thumb2/mve-vld3.ll b/llvm/test/CodeGen/Thumb2/mve-vld3.ll
index ccdc996d75970..119cf44f6af77 100644
--- a/llvm/test/CodeGen/Thumb2/mve-vld3.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-vld3.ll
@@ -288,46 +288,46 @@ entry:
 define void @vld3_v8i16(ptr %src, ptr %dst) {
 ; CHECK-LABEL: vld3_v8i16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9}
-; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11}
 ; CHECK-NEXT:    vldrw.u32 q0, [r0]
 ; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #32]
-; CHECK-NEXT:    vmovx.f16 s6, s2
-; CHECK-NEXT:    vmov.f32 s4, s1
-; CHECK-NEXT:    vins.f16 s4, s6
-; CHECK-NEXT:    vmovx.f16 s6, s9
+; CHECK-NEXT:    vldrw.u32 q4, [r0, #32]
+; CHECK-NEXT:    vmovx.f16 s4, s2
+; CHECK-NEXT:    vmovx.f16 s12, s1
+; CHECK-NEXT:    vins.f16 s1, s4
+; CHECK-NEXT:    vmovx.f16 s4, s9
+; CHECK-NEXT:    vmovx.f16 s20, s0
+; CHECK-NEXT:    vmovx.f16 s14, s8
+; CHECK-NEXT:    vins.f16 s8, s4
+; CHECK-NEXT:    vins.f16 s20, s2
 ; CHECK-NEXT:    vmov.f32 s5, s8
-; CHECK-NEXT:    vmovx.f16 s7, s12
-; CHECK-NEXT:    vins.f16 s5, s6
+; CHECK-NEXT:    vmovx.f16 s8, s16
 ; CHECK-NEXT:    vmov.f32 s6, s11
-; CHECK-NEXT:    vins.f16 s6, s7
-; CHECK-NEXT:    vmovx.f16 s16, s15
-; CHECK-NEXT:    vmov.f32 s7, s14
-; CHECK-NEXT:    vmovx.f16 s17, s3
-; CHECK-NEXT:    vins.f16 s7, s16
-; CHECK-NEXT:    vmovx.f16 s16, s0
-; CHECK-NEXT:    vins.f16 s16, s2
-; CHECK-NEXT:    vmovx.f16 s2, s1
-; CHECK-NEXT:    vins.f16 s0, s2
-; CHECK-NEXT:    vmovx.f16 s2, s8
-; CHECK-NEXT:    vins.f16 s3, s2
+; CHECK-NEXT:    vmovx.f16 s13, s18
 ; CHECK-NEXT:    vmovx.f16 s2, s11
-; CHECK-NEXT:    vmovx.f16 s8, s14
-; CHECK-NEXT:    vmovx.f16 s18, s10
-; CHECK-NEXT:    vmovx.f16 s19, s13
+; CHECK-NEXT:    vmovx.f16 s22, s10
+; CHECK-NEXT:    vmovx.f16 s23, s17
+; CHECK-NEXT:    vmovx.f16 s21, s3
 ; CHECK-NEXT:    vins.f16 s10, s2
-; CHECK-NEXT:    vins.f16 s13, s8
+; CHECK-NEXT:    vins.f16 s6, s8
+; CHECK-NEXT:    vmovx.f16 s8, s19
+; CHECK-NEXT:    vins.f16 s3, s14
+; CHECK-NEXT:    vins.f16 s17, s13
+; CHECK-NEXT:    vmov.f32 s4, s1
+; CHECK-NEXT:    vins.f16 s18, s8
 ; CHECK-NEXT:    vmov.f32 s1, s3
-; CHECK-NEXT:    vins.f16 s18, s12
-; CHECK-NEXT:    vins.f16 s19, s15
-; CHECK-NEXT:    vins.f16 s17, s9
+; CHECK-NEXT:    vins.f16 s22, s16
+; CHECK-NEXT:    vins.f16 s23, s19
+; CHECK-NEXT:    vins.f16 s21, s9
+; CHECK-NEXT:    vins.f16 s0, s12
 ; CHECK-NEXT:    vmov.f32 s2, s10
-; CHECK-NEXT:    vmov.f32 s3, s13
-; CHECK-NEXT:    vadd.i16 q0, q0, q4
+; CHECK-NEXT:    vmov.f32 s3, s17
+; CHECK-NEXT:    vmov.f32 s7, s18
+; CHECK-NEXT:    vadd.i16 q0, q0, q5
 ; CHECK-NEXT:    vadd.i16 q0, q0, q1
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vpop {d8, d9}
+; CHECK-NEXT:    vpop {d8, d9, d10, d11}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <24 x i16>, ptr %src, align 4
@@ -343,164 +343,164 @@ entry:
 define void @vld3_v16i16(ptr %src, ptr %dst) {
 ; CHECK-LV-LABEL: vld3_v16i16:
 ; CHECK-LV:       @ %bb.0: @ %entry
-; CHECK-LV-NEXT:    .vsave {d8, d9, d10, d11}
-; CHECK-LV-NEXT:    vpush {d8, d9, d10, d11}
+; CHECK-LV-NEXT:    .vsave {d8, d9, d10, d11, d12}
+; CHECK-LV-NEXT:    vpush {d8, d9, d10, d11, d12}
 ; CHECK-LV-NEXT:    vldrw.u32 q0, [r0, #48]
 ; CHECK-LV-NEXT:    vldrw.u32 q2, [r0, #64]
-; CHECK-LV-NEXT:    vldrw.u32 q3, [r0, #80]
-; CHECK-LV-NEXT:    vmovx.f16 s6, s2
-; CHECK-LV-NEXT:    vmov.f32 s4, s1
-; CHECK-LV-NEXT:    vins.f16 s4, s6
-; CHECK-LV-NEXT:    vmovx.f16 s6, s9
+; CHECK-LV-NEXT:    vldrw.u32 q4, [r0, #80]
+; CHECK-LV-NEXT:    vmovx.f16 s4, s2
+; CHECK-LV-NEXT:    vmovx.f16 s12, s1
+; CHECK-LV-NEXT:    vins.f16 s1, s4
+; CHECK-LV-NEXT:    vmovx.f16 s4, s9
+; CHECK-LV-NEXT:    vmovx.f16 s14, s8
+; CHECK-LV-NEXT:    vins.f16 s8, s4
+; CHECK-LV-NEXT:    vmovx.f16 s20, s0
 ; CHECK-LV-NEXT:    vmov.f32 s5, s8
-; CHECK-LV-NEXT:    vmovx.f16 s7, s12
-; CHECK-LV-NEXT:    vins.f16 s5, s6
 ; CHECK-LV-NEXT:    vmov.f32 s6, s11
-; CHECK-LV-NEXT:    vins.f16 s6, s7
-; CHECK-LV-NEXT:    vmovx.f16 s16, s15
-; CHECK-LV-NEXT:    vmov.f32 s7, s14
-; CHECK-LV-NEXT:    vmovx.f16 s17, s3
-; CHECK-LV-NEXT:    vins.f16 s7, s16
-; CHECK-LV-NEXT:    vmovx.f16 s16, s0
-; CHECK-LV-NEXT:    vins.f16 s16, s2
-; CHECK-LV-NEXT:    vmovx.f16 s2, s1
-; CHECK-LV-NEXT:    vins.f16 s0, s2
-; CHECK-LV-NEXT:    vmovx.f16 s2, s8
-; CHECK-LV-NEXT:    vins.f16 s3, s2
+; CHECK-LV-NEXT:    vmovx.f16 s8, s16
+; CHECK-LV-NEXT:    vins.f16 s6, s8
+; CHECK-LV-NEXT:    vins.f16 s20, s2
+; CHECK-LV-NEXT:    vmovx.f16 s8, s19
+; CHECK-LV-NEXT:    vmovx.f16 s13, s18
 ; CHECK-LV-NEXT:    vmovx.f16 s2, s11
-; CHECK-LV-NEXT:    vmovx.f16 s8, s14
-; CHECK-LV-NEXT:    vmovx.f16 s18, s10
-; CHECK-LV-NEXT:    vmovx.f16 s19, s13
+; CHECK-LV-NEXT:    vmovx.f16 s22, s10
+; CHECK-LV-NEXT:    vmovx.f16 s23, s17
+; CHECK-LV-NEXT:    vmovx.f16 s21, s3
 ; CHECK-LV-NEXT:    vins.f16 s10, s2
-; CHECK-LV-NEXT:    vins.f16 s13, s8
+; CHECK-LV-NEXT:    vins.f16 s18, s8
+; CHECK-LV-NEXT:    vins.f16 s3, s14
+; CHECK-LV-NEXT:    vins.f16 s17, s13
+; CHECK-LV-NEXT:    vmov.f32 s4, s1
+; CHECK-LV-NEXT:    vins.f16 s0, s12
 ; CHECK-LV-NEXT:    vmov.f32 s1, s3
-; CHECK-LV-NEXT:    vins.f16 s18, s12
-; CHECK-LV-NEXT:    vins.f16 s19, s15
-; CHECK-LV-NEXT:    vmov.f32 s3, s13
-; CHECK-LV-NEXT:    vldrw.u32 q3, [r0, #16]
-; CHECK-LV-NEXT:    vins.f16 s17, s9
+; CHECK-LV-NEXT:    vins.f16 s22, s16
+; CHECK-LV-NEXT:    vmov.f32 s7, s18
+; CHECK-LV-NEXT:    vins.f16 s23, s19
+; CHECK-LV-NEXT:    vmov.f32 s3, s17
+; CHECK-LV-NEXT:    vldrw.u32 q4, [r0]
+; CHECK-LV-NEXT:    vins.f16 s21, s9
 ; CHECK-LV-NEXT:    vmov.f32 s2, s10
-; CHECK-LV-NEXT:    vadd.i16 q0, q0, q4
+; CHECK-LV-NEXT:    vldrw.u32 q3, [r0, #16]
+; CHECK-LV-NEXT:    vadd.i16 q0, q0, q5
 ; CHECK-LV-NEXT:    vldrw.u32 q2, [r0, #32]
 ; CHECK-LV-NEXT:    vadd.i16 q0, q0, q1
 ; CHECK-LV-NEXT:    vmovx.f16 s6, s14
-; CHECK-LV-NEXT:    vldrw.u32 q4, [r0]
+; CHECK-LV-NEXT:    vmovx.f16 s4, s16
+; CHECK-LV-NEXT:    vmovx.f16 s5, s19
 ; CHECK-LV-NEXT:    vins.f16 s6, s8
+; CHECK-LV-NEXT:    vins.f16 s4, s18
+; CHECK-LV-NEXT:    vins.f16 s5, s13
+; CHECK-LV-NEXT:    vmovx.f16 s18, s18
+; CHECK-LV-NEXT:    vmovx.f16 s13, s13
 ; CHECK-LV-NEXT:    vmov.f32 s22, s15
 ; CHECK-LV-NEXT:    vmovx.f16 s8, s8
+; CHECK-LV-NEXT:    vmovx.f16 s24, s17
+; CHECK-LV-NEXT:    vins.f16 s17, s18
+; CHECK-LV-NEXT:    vmovx.f16 s18, s12
+; CHECK-LV-NEXT:    vins.f16 s12, s13
 ; CHECK-LV-NEXT:    vins.f16 s22, s8
 ; CHECK-LV-NEXT:    vmovx.f16 s8, s11
-; CHECK-LV-NEXT:    vmov.f32 s23, s10
-; CHECK-LV-NEXT:    vmovx.f16 s4, s16
-; CHECK-LV-NEXT:    vins.f16 s23, s8
-; CHECK-LV-NEXT:    vmovx.f16 s8, s17
-; CHECK-LV-NEXT:    vins.f16 s16, s8
-; CHECK-LV-NEXT:    vmovx.f16 s8, s12
-; CHECK-LV-NEXT:    vmovx.f16 s5, s19
-; CHECK-LV-NEXT:    vins.f16 s19, s8
+; CHECK-LV-NEXT:    vmov.f32 s21, s12
+; CHECK-LV-NEXT:    vmovx.f16 s12, s10
+; CHECK-LV-NEXT:    vins.f16 s10, s8
 ; CHECK-LV-NEXT:    vmovx.f16 s8, s15
 ; CHECK-LV-NEXT:    vmovx.f16 s7, s9
+; CHECK-LV-NEXT:    vins.f16 s19, s18
 ; CHECK-LV-NEXT:    vins.f16 s14, s8
-; CHECK-LV-NEXT:    vmovx.f16 s8, s10
-; CHECK-LV-NEXT:    vins.f16 s4, s18
+; CHECK-LV-NEXT:    vins.f16 s9, s12
 ; CHECK-LV-NEXT:    vmov.f32 s20, s17
-; CHECK-LV-NEXT:    vmovx.f16 s18, s18
-; CHECK-LV-NEXT:    vins.f16 s9, s8
-; CHECK-LV-NEXT:    vins.f16 s5, s13
-; CHECK-LV-NEXT:    vins.f16 s20, s18
-; CHECK-LV-NEXT:    vmov.f32 s17, s19
 ; CHECK-LV-NEXT:    vins.f16 s7, s11
-; CHECK-LV-NEXT:    vmovx.f16 s13, s13
-; CHECK-LV-NEXT:    vmov.f32 s21, s12
+; CHECK-LV-NEXT:    vmov.f32 s17, s19
+; CHECK-LV-NEXT:    vins.f16 s16, s24
 ; CHECK-LV-NEXT:    vmov.f32 s18, s14
-; CHECK-LV-NEXT:    vins.f16 s21, s13
-; CHECK-LV-NEXT:    vmov.f32 s19, s9
 ; CHECK-LV-NEXT:    vstrw.32 q0, [r1, #16]
+; CHECK-LV-NEXT:    vmov.f32 s19, s9
+; CHECK-LV-NEXT:    vmov.f32 s23, s10
 ; CHECK-LV-NEXT:    vadd.i16 q1, q4, q1
 ; CHECK-LV-NEXT:    vadd.i16 q1, q1, q5
 ; CHECK-LV-NEXT:    vstrw.32 q1, [r1]
-; CHECK-LV-NEXT:    vpop {d8, d9, d10, d11}
+; CHECK-LV-NEXT:    vpop {d8, d9, d10, d11, d12}
 ; CHECK-LV-NEXT:    bx lr
 ;
 ; CHECK-LIS-LABEL: vld3_v16i16:
 ; CHECK-LIS:       @ %bb.0: @ %entry
-; CHECK-LIS-NEXT:    .vsave {d8, d9, d10, d11}
-; CHECK-LIS-NEXT:    vpush {d8, d9, d10, d11}
+; CHECK-LIS-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
+; CHECK-LIS-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
 ; CHECK-LIS-NEXT:    vldrw.u32 q0, [r0, #48]
 ; CHECK-LIS-NEXT:    vldrw.u32 q2, [r0, #64]
-; CHECK-LIS-NEXT:    vldrw.u32 q3, [r0, #80]
-; CHECK-LIS-NEXT:    vmovx.f16 s6, s2
-; CHECK-LIS-NEXT:    vmov.f32 s4, s1
-; CHECK-LIS-NEXT:    vins.f16 s4, s6
-; CHECK-LIS-NEXT:    vmovx.f16 s6, s9
+; CHECK-LIS-NEXT:    vldrw.u32 q4, [r0, #80]
+; CHECK-LIS-NEXT:    vmovx.f16 s4, s2
+; CHECK-LIS-NEXT:    vmovx.f16 s12, s1
+; CHECK-LIS-NEXT:    vins.f16 s1, s4
+; CHECK-LIS-NEXT:    vmovx.f16 s4, s9
+; CHECK-LIS-NEXT:    vmovx.f16 s14, s8
+; CHECK-LIS-NEXT:    vins.f16 s8, s4
+; CHECK-LIS-NEXT:    vmovx.f16 s20, s0
 ; CHECK-LIS-NEXT:    vmov.f32 s5, s8
-; CHECK-LIS-NEXT:    vmovx.f16 s7, s12
-; CHECK-LIS-NEXT:    vins.f16 s5, s6
 ; CHECK-LIS-NEXT:    vmov.f32 s6, s11
-; CHECK-LIS-NEXT:    vins.f16 s6, s7
-; CHECK-LIS-NEXT:    vmovx.f16 s16, s15
-; CHECK-LIS-NEXT:    vmov.f32 s7, s14
-; CHECK-LIS-NEXT:    vmovx.f16 s17, s3
-; CHECK-LIS-NEXT:    vins.f16 s7, s16
-; CHECK-LIS-NEXT:    vmovx.f16 s16, s0
-; CHECK-LIS-NEXT:    vins.f16 s16, s2
-; CHECK-LIS-NEXT:    vmovx.f16 s2, s1
-; CHECK-LIS-NEXT:    vins.f16 s0, s2
-; CHECK-LIS-NEXT:    vmovx.f16 s2, s8
-; CHECK-LIS-NEXT:    vins.f16 s3, s2
+; CHECK-LIS-NEXT:    vmovx.f16 s8, s16
+; CHECK-LIS-NEXT:    vins.f16 s6, s8
+; CHECK-LIS-NEXT:    vins.f16 s20, s2
+; CHECK-LIS-NEXT:    vmovx.f16 s8, s19
+; CHECK-LIS-NEXT:    vmovx.f16 s13, s18
 ; CHECK-LIS-NEXT:    vmovx.f16 s2, s11
-; CHECK-LIS-NEXT:    vmovx.f16 s8, s14
-; CHECK-LIS-NEXT:    vmovx.f16 s18, s10
-; CHECK-LIS-NEXT:    vmovx.f16 s19, s13
+; CHECK-LIS-NEXT:    vmovx.f16 s22, s10
+; CHECK-LIS-NEXT:    vmovx.f16 s23, s17
+; CHECK-LIS-NEXT:    vmovx.f16 s21, s3
 ; CHECK-LIS-NEXT:    vins.f16 s10, s2
-; CHECK-LIS-NEXT:    vins.f16 s13, s8
+; CHECK-LIS-NEXT:    vins.f16 s18, s8
+; CHECK-LIS-NEXT:    vins.f16 s3, s14
+; CHECK-LIS-NEXT:    vins.f16 s17, s13
+; CHECK-LIS-NEXT:    vmov.f32 s4, s1
+; CHECK-LIS-NEXT:    vins.f16 s0, s12
 ; CHECK-LIS-NEXT:    vmov.f32 s1, s3
-; CHECK-LIS-NEXT:    vins.f16 s18, s12
-; CHECK-LIS-NEXT:    vins.f16 s19, s15
-; CHECK-LIS-NEXT:    vmov.f32 s3, s13
-; CHECK-LIS-NEXT:    vins.f16 s17, s9
+; CHECK-LIS-NEXT:    vins.f16 s22, s16
+; CHECK-LIS-NEXT:    vmov.f32 s7, s18
+; CHECK-LIS-NEXT:    vins.f16 s23, s19
+; CHECK-LIS-NEXT:    vmov.f32 s3, s17
+; CHECK-LIS-NEXT:    vldrw.u32 q4, [r0, #16]
+; CHECK-LIS-NEXT:    vins.f16 s21, s9
 ; CHECK-LIS-NEXT:    vmov.f32 s2, s10
+; CHECK-LIS-NEXT:    vadd.i16 q0, q0, q5
 ; CHECK-LIS-NEXT:    vldrw.u32 q2, [r0, #32]
-; CHECK-LIS-NEXT:    vadd.i16 q0, q0, q4
-; CHECK-LIS-NEXT:    vldrw.u32 q4, [r0, #16]
 ; CHECK-LIS-NEXT:    vadd.i16 q0, q0, q1
 ; CHECK-LIS-NEXT:    vldrw.u32 q3, [r0]
 ; CHECK-LIS-NEXT:    vmovx.f16 s6, s18
 ; CHECK-LIS-NEXT:    vmov.f32 s22, s19
 ; CHECK-LIS-NEXT:    vins.f16 s6, s8
 ; CHECK-LIS-NEXT:    vmovx.f16 s8, s8
+; CHECK-LIS-NEXT:    vmovx.f16 s7, s9
+; CHECK-LIS-NEXT:    vmovx.f16 s4, s12
 ; CHECK-LIS-NEXT:    vins.f16 s22, s8
 ; CHECK-LIS-NEXT:    vmovx.f16 s8, s11
-; CHECK-LIS-NEXT:    vmov.f32 s23, s10
-; CHECK-LIS-NEXT:    vmovx.f16 s4, s12
-; CHECK-LIS-NEXT:    vins.f16 s23, s8
-; CHECK-LIS-NEXT:    vmovx.f16 s8, s13
-; CHECK-LIS-NEXT:    vins.f16 s12, s8
-; CHECK-LIS-NEXT:    vmovx.f16 s8, s16
-; CHECK-LIS-NEXT:    vmovx.f16 s5, s15
-; CHECK-LIS-NEXT:    vins.f16 s15, s8
-; CHECK-LIS-NEXT:    vmovx.f16 s8, s19
+; CHECK-LIS-NEXT:    vins.f16 s7, s11
+; CHECK-LIS-NEXT:    vmovx.f16 s11, s10
+; CHECK-LIS-NEXT:    vins.f16 s10, s8
 ; CHECK-LIS-NEXT:    vins.f16 s4, s14
-; CHECK-LIS-NEXT:    vmov.f32 s20, s13
+; CHECK-LIS-NEXT:    vmovx.f16 s8, s19
 ; CHECK-LIS-NEXT:    vmovx.f16 s14, s14
-; CHECK-LIS-NEXT:    vins.f16 s18, s8
-; CHECK-LIS-NEXT:    vmovx.f16 s8, s10
-; CHECK-LIS-NEXT:    vmovx.f16 s7, s9
-; CHECK-LIS-NEXT:    vins.f16 s20, s14
+; CHECK-LIS-NEXT:    vmovx.f16 s26, s16
+; CHECK-LIS-NEXT:    vmovx.f16 s5, s15
+; CHECK-LIS-NEXT:    vmovx.f16 s24, s13
+; CHECK-LIS-NEXT:    vins.f16 s13, s14
 ; CHECK-LIS-NEXT:    vmovx.f16 s14, s17
-; CHECK-LIS-NEXT:    vmov.f32 s21, s16
-; CHECK-LIS-NEXT:    vins.f16 s9, s8
-; CHECK-LIS-NEXT:    vins.f16 s21, s14
+; CHECK-LIS-NEXT:    vins.f16 s15, s26
+; CHECK-LIS-NEXT:    vins.f16 s18, s8
+; CHECK-LIS-NEXT:    vins.f16 s9, s11
+; CHECK-LIS-NEXT:    vins.f16 s16, s14
+; CHECK-LIS-NEXT:    vmov.f32 s20, s13
 ; CHECK-LIS-NEXT:    vmov.f32 s13, s15
-; CHECK-LIS-NEXT:    vins.f16 s7, s11
 ; CHECK-LIS-NEXT:    vins.f16 s5, s17
+; CHECK-LIS-NEXT:    vins.f16 s12, s24
 ; CHECK-LIS-NEXT:    vmov.f32 s14, s18
 ; CHECK-LIS-NEXT:    vmov.f32 s15, s9
 ; CHECK-LIS-NEXT:    vstrw.32 q0, [r1, #16]
+; CHECK-LIS-NEXT:    vmov.f32 s21, s16
 ; CHECK-LIS-NEXT:    vadd.i16 q1, q3, q1
+; CHECK-LIS-NEXT:    vmov.f32 s23, s10
 ; CHECK-LIS-NEXT:    vadd.i16 q1, q1, q5
 ; CHECK-LIS-NEXT:    vstrw.32 q1, [r1]
-; CHECK-LIS-NEXT:    vpop {d8, d9, d10, d11}
+; CHECK-LIS-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
 ; CHECK-LIS-NEXT:    bx lr
 entry:
   %l1 = load <48 x i16>, ptr %src, align 4
@@ -1192,27 +1192,27 @@ entry:
 define void @vld3_v4f16(ptr %src, ptr %dst) {
 ; CHECK-LABEL: vld3_v4f16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    ldrd r2, r3, [r0, #16]
 ; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vmov.32 q2[0], r2
-; CHECK-NEXT:    vmovx.f16 s12, s4
-; CHECK-NEXT:    vmov.32 q2[1], r3
-; CHECK-NEXT:    vmovx.f16 s13, s7
-; CHECK-NEXT:    vmovx.f16 s0, s9
-; CHECK-NEXT:    vmov.f32 s1, s8
-; CHECK-NEXT:    vins.f16 s1, s0
-; CHECK-NEXT:    vmovx.f16 s0, s5
-; CHECK-NEXT:    vins.f16 s4, s0
-; CHECK-NEXT:    vmovx.f16 s0, s6
-; CHECK-NEXT:    vins.f16 s5, s0
-; CHECK-NEXT:    vmovx.f16 s0, s8
-; CHECK-NEXT:    vins.f16 s7, s0
-; CHECK-NEXT:    vmov.f32 s0, s5
-; CHECK-NEXT:    vins.f16 s12, s6
-; CHECK-NEXT:    vins.f16 s13, s9
+; CHECK-NEXT:    ldrd r2, r3, [r0, #16]
+; CHECK-NEXT:    vmov.32 q0[0], r2
+; CHECK-NEXT:    vmovx.f16 s8, s4
+; CHECK-NEXT:    vmov.32 q0[1], r3
+; CHECK-NEXT:    vins.f16 s8, s6
+; CHECK-NEXT:    vmovx.f16 s6, s6
+; CHECK-NEXT:    vmovx.f16 s10, s5
+; CHECK-NEXT:    vins.f16 s5, s6
+; CHECK-NEXT:    vmovx.f16 s6, s0
+; CHECK-NEXT:    vmovx.f16 s9, s7
+; CHECK-NEXT:    vmovx.f16 s2, s1
+; CHECK-NEXT:    vins.f16 s7, s6
+; CHECK-NEXT:    vins.f16 s0, s2
+; CHECK-NEXT:    vmov.f32 s12, s5
+; CHECK-NEXT:    vins.f16 s4, s10
+; CHECK-NEXT:    vins.f16 s9, s1
 ; CHECK-NEXT:    vmov.f32 s5, s7
-; CHECK-NEXT:    vadd.f16 q1, q1, q3
-; CHECK-NEXT:    vadd.f16 q0, q1, q0
+; CHECK-NEXT:    vadd.f16 q1, q1, q2
+; CHECK-NEXT:    vmov.f32 s13, s0
+; CHECK-NEXT:    vadd.f16 q0, q1, q3
 ; CHECK-NEXT:    vmov r0, r2, d0
 ; CHECK-NEXT:    strd r0, r2, [r1]
 ; CHECK-NEXT:    bx lr
@@ -1230,79 +1230,79 @@ entry:
 define void @vld3_v8f16(ptr %src, ptr %dst) {
 ; CHECK-LV-LABEL: vld3_v8f16:
 ; CHECK-LV:       @ %bb.0: @ %entry
-; CHECK-LV-NEXT:    .vsave {d8, d9}
-; CHECK-LV-NEXT:    vpush {d8, d9}
-; CHECK-LV-NEXT:    vldrw.u32 q2, [r0, #16]
+; CHECK-LV-NEXT:    .vsave {d8, d9, d10}
+; CHECK-LV-NEXT:    vpush {d8, d9, d10}
 ; CHECK-LV-NEXT:    vldrw.u32 q0, [r0]
+; CHECK-LV-NEXT:    vldrw.u32 q2, [r0, #16]
 ; CHECK-LV-NEXT:    vldrw.u32 q3, [r0, #32]
+; CHECK-LV-NEXT:    vmovx.f16 s4, s2
+; CHECK-LV-NEXT:    vmovx.f16 s7, s1
+; CHECK-LV-NEXT:    vins.f16 s1, s4
+; CHECK-LV-NEXT:    vmovx.f16 s4, s9
+; CHECK-LV-NEXT:    vmovx.f16 s20, s8
+; CHECK-LV-NEXT:    vins.f16 s8, s4
 ; CHECK-LV-NEXT:    vmov.f32 s5, s8
-; CHECK-LV-NEXT:    vmovx.f16 s8, s8
-; CHECK-LV-NEXT:    vmovx.f16 s17, s3
-; CHECK-LV-NEXT:    vins.f16 s3, s8
+; CHECK-LV-NEXT:    vmovx.f16 s8, s12
+; CHECK-LV-NEXT:    vmov.f32 s6, s11
+; CHECK-LV-NEXT:    vmovx.f16 s16, s0
+; CHECK-LV-NEXT:    vins.f16 s6, s8
 ; CHECK-LV-NEXT:    vmovx.f16 s8, s11
 ; CHECK-LV-NEXT:    vmovx.f16 s18, s10
-; CHECK-LV-NEXT:    vmovx.f16 s16, s0
 ; CHECK-LV-NEXT:    vins.f16 s10, s8
-; CHECK-LV-NEXT:    vmovx.f16 s6, s2
-; CHECK-LV-NEXT:    vmov.f32 s4, s1
 ; CHECK-LV-NEXT:    vmovx.f16 s8, s14
 ; CHECK-LV-NEXT:    vmovx.f16 s19, s13
-; CHECK-LV-NEXT:    vins.f16 s4, s6
-; CHECK-LV-NEXT:    vmovx.f16 s6, s9
+; CHECK-LV-NEXT:    vmovx.f16 s17, s3
 ; CHECK-LV-NEXT:    vins.f16 s16, s2
 ; CHECK-LV-NEXT:    vmovx.f16 s2, s15
-; CHECK-LV-NEXT:    vmovx.f16 s7, s12
-; CHECK-LV-NEXT:    vins.f16 s18, s12
-; CHECK-LV-NEXT:    vmovx.f16 s12, s1
+; CHECK-LV-NEXT:    vins.f16 s3, s20
 ; CHECK-LV-NEXT:    vins.f16 s13, s8
-; CHECK-LV-NEXT:    vins.f16 s5, s6
-; CHECK-LV-NEXT:    vmov.f32 s6, s11
+; CHECK-LV-NEXT:    vmov.f32 s4, s1
 ; CHECK-LV-NEXT:    vins.f16 s14, s2
+; CHECK-LV-NEXT:    vins.f16 s0, s7
 ; CHECK-LV-NEXT:    vmov.f32 s1, s3
+; CHECK-LV-NEXT:    vins.f16 s18, s12
 ; CHECK-LV-NEXT:    vins.f16 s19, s15
 ; CHECK-LV-NEXT:    vins.f16 s17, s9
-; CHECK-LV-NEXT:    vins.f16 s0, s12
 ; CHECK-LV-NEXT:    vmov.f32 s2, s10
 ; CHECK-LV-NEXT:    vmov.f32 s3, s13
-; CHECK-LV-NEXT:    vins.f16 s6, s7
 ; CHECK-LV-NEXT:    vmov.f32 s7, s14
 ; CHECK-LV-NEXT:    vadd.f16 q0, q0, q4
 ; CHECK-LV-NEXT:    vadd.f16 q0, q0, q1
 ; CHECK-LV-NEXT:    vstrw.32 q0, [r1]
-; CHECK-LV-NEXT:    vpop {d8, d9}
+; CHECK-LV-NEXT:    vpop {d8, d9, d10}
 ; CHECK-LV-NEXT:    bx lr
 ;
 ; CHECK-LIS-LABEL: vld3_v8f16:
 ; CHECK-LIS:       @ %bb.0: @ %entry
-; CHECK-LIS-NEXT:    .vsave {d8, d9}
-; CHECK-LIS-NEXT:    vpush {d8, d9}
+; CHECK-LIS-NEXT:    .vsave {d8, d9, d10}
+; CHECK-LIS-NEXT:    vpush {d8, d9, d10}
 ; CHECK-LIS-NEXT:    vldrw.u32 q0, [r0]
 ; CHECK-LIS-NEXT:    vldrw.u32 q2, [r0, #16]
 ; CHECK-LIS-NEXT:    vldrw.u32 q4, [r0, #32]
-; CHECK-LIS-NEXT:    vmov.f32 s4, s1
-; CHECK-LIS-NEXT:    vmovx.f16 s6, s2
-; CHECK-LIS-NEXT:    vins.f16 s4, s6
+; CHECK-LIS-NEXT:    vmovx.f16 s4, s2
+; CHECK-LIS-NEXT:    vmovx.f16 s7, s1
+; CHECK-LIS-NEXT:    vins.f16 s1, s4
+; CHECK-LIS-NEXT:    vmovx.f16 s4, s9
+; CHECK-LIS-NEXT:    vmovx.f16 s20, s8
+; CHECK-LIS-NEXT:    vins.f16 s8, s4
 ; CHECK-LIS-NEXT:    vmov.f32 s5, s8
-; CHECK-LIS-NEXT:    vmovx.f16 s6, s9
-; CHECK-LIS-NEXT:    vmovx.f16 s8, s8
-; CHECK-LIS-NEXT:    vmovx.f16 s13, s3
-; CHECK-LIS-NEXT:    vins.f16 s5, s6
-; CHECK-LIS-NEXT:    vins.f16 s3, s8
+; CHECK-LIS-NEXT:    vmovx.f16 s8, s16
 ; CHECK-LIS-NEXT:    vmov.f32 s6, s11
-; CHECK-LIS-NEXT:    vmovx.f16 s12, s16
+; CHECK-LIS-NEXT:    vmovx.f16 s12, s0
+; CHECK-LIS-NEXT:    vins.f16 s6, s8
 ; CHECK-LIS-NEXT:    vmovx.f16 s8, s11
 ; CHECK-LIS-NEXT:    vmovx.f16 s14, s10
-; CHECK-LIS-NEXT:    vins.f16 s6, s12
-; CHECK-LIS-NEXT:    vmovx.f16 s12, s0
 ; CHECK-LIS-NEXT:    vins.f16 s10, s8
 ; CHECK-LIS-NEXT:    vmovx.f16 s8, s18
 ; CHECK-LIS-NEXT:    vmovx.f16 s15, s17
+; CHECK-LIS-NEXT:    vmovx.f16 s13, s3
 ; CHECK-LIS-NEXT:    vins.f16 s12, s2
 ; CHECK-LIS-NEXT:    vmovx.f16 s2, s19
-; CHECK-LIS-NEXT:    vmovx.f16 s1, s1
+; CHECK-LIS-NEXT:    vins.f16 s3, s20
 ; CHECK-LIS-NEXT:    vins.f16 s17, s8
+; CHECK-LIS-NEXT:    vmov.f32 s4, s1
 ; CHECK-LIS-NEXT:    vins.f16 s18, s2
-; CHECK-LIS-NEXT:    vins.f16 s0, s1
+; CHECK-LIS-NEXT:    vins.f16 s0, s7
 ; CHECK-LIS-NEXT:    vmov.f32 s1, s3
 ; CHECK-LIS-NEXT:    vins.f16 s14, s16
 ; CHECK-LIS-NEXT:    vins.f16 s15, s19
@@ -1313,7 +1313,7 @@ define void @vld3_v8f16(ptr %src, ptr %dst) {
 ; CHECK-LIS-NEXT:    vadd.f16 q0, q0, q3
 ; CHECK-LIS-NEXT:    vadd.f16 q0, q0, q1
 ; CHECK-LIS-NEXT:    vstrw.32 q0, [r1]
-; CHECK-LIS-NEXT:    vpop {d8, d9}
+; CHECK-LIS-NEXT:    vpop {d8, d9, d10}
 ; CHECK-LIS-NEXT:    bx lr
 entry:
   %l1 = load <24 x half>, ptr %src, align 4
@@ -1327,167 +1327,86 @@ entry:
 }
 
 define void @vld3_v16f16(ptr %src, ptr %dst) {
-; CHECK-LV-LABEL: vld3_v16f16:
-; CHECK-LV:       @ %bb.0: @ %entry
-; CHECK-LV-NEXT:    .vsave {d8, d9}
-; CHECK-LV-NEXT:    vpush {d8, d9}
-; CHECK-LV-NEXT:    vldrw.u32 q0, [r0, #48]
-; CHECK-LV-NEXT:    vldrw.u32 q2, [r0, #64]
-; CHECK-LV-NEXT:    vldrw.u32 q3, [r0, #80]
-; CHECK-LV-NEXT:    vmovx.f16 s6, s2
-; CHECK-LV-NEXT:    vmov.f32 s4, s1
-; CHECK-LV-NEXT:    vins.f16 s4, s6
-; CHECK-LV-NEXT:    vmovx.f16 s6, s9
-; CHECK-LV-NEXT:    vmov.f32 s5, s8
-; CHECK-LV-NEXT:    vmovx.f16 s7, s12
-; CHECK-LV-NEXT:    vins.f16 s5, s6
-; CHECK-LV-NEXT:    vmov.f32 s6, s11
-; CHECK-LV-NEXT:    vins.f16 s6, s7
-; CHECK-LV-NEXT:    vmovx.f16 s16, s15
-; CHECK-LV-NEXT:    vmov.f32 s7, s14
-; CHECK-LV-NEXT:    vmovx.f16 s17, s3
-; CHECK-LV-NEXT:    vins.f16 s7, s16
-; CHECK-LV-NEXT:    vmovx.f16 s16, s0
-; CHECK-LV-NEXT:    vins.f16 s16, s2
-; CHECK-LV-NEXT:    vmovx.f16 s2, s1
-; CHECK-LV-NEXT:    vins.f16 s0, s2
-; CHECK-LV-NEXT:    vmovx.f16 s2, s8
-; CHECK-LV-NEXT:    vins.f16 s3, s2
-; CHECK-LV-NEXT:    vmovx.f16 s2, s11
-; CHECK-LV-NEXT:    vmovx.f16 s18, s10
-; CHECK-LV-NEXT:    vins.f16 s10, s2
-; CHECK-LV-NEXT:    vmovx.f16 s2, s14
-; CHECK-LV-NEXT:    vmovx.f16 s19, s13
-; CHECK-LV-NEXT:    vins.f16 s13, s2
-; CHECK-LV-NEXT:    vmov.f32 s1, s3
-; CHECK-LV-NEXT:    vins.f16 s18, s12
-; CHECK-LV-NEXT:    vins.f16 s19, s15
-; CHECK-LV-NEXT:    vmov.f32 s3, s13
-; CHECK-LV-NEXT:    vins.f16 s17, s9
-; CHECK-LV-NEXT:    vmov.f32 s2, s10
-; CHECK-LV-NEXT:    vldrw.u32 q3, [r0, #16]
-; CHECK-LV-NEXT:    vadd.f16 q0, q0, q4
-; CHECK-LV-NEXT:    vadd.f16 q2, q0, q1
-; CHECK-LV-NEXT:    vldrw.u32 q0, [r0]
-; CHECK-LV-NEXT:    vldrw.u32 q1, [r0, #32]
-; CHECK-LV-NEXT:    vstrw.32 q2, [r1, #16]
-; CHECK-LV-NEXT:    vmovx.f16 s10, s2
-; CHECK-LV-NEXT:    vmov.f32 s8, s1
-; CHECK-LV-NEXT:    vins.f16 s8, s10
-; CHECK-LV-NEXT:    vmovx.f16 s10, s13
-; CHECK-LV-NEXT:    vmov.f32 s9, s12
-; CHECK-LV-NEXT:    vmovx.f16 s11, s4
-; CHECK-LV-NEXT:    vins.f16 s9, s10
-; CHECK-LV-NEXT:    vmov.f32 s10, s15
-; CHECK-LV-NEXT:    vins.f16 s10, s11
-; CHECK-LV-NEXT:    vmovx.f16 s16, s7
-; CHECK-LV-NEXT:    vmov.f32 s11, s6
-; CHECK-LV-NEXT:    vmovx.f16 s17, s3
-; CHECK-LV-NEXT:    vins.f16 s11, s16
-; CHECK-LV-NEXT:    vmovx.f16 s16, s0
-; CHECK-LV-NEXT:    vins.f16 s16, s2
-; CHECK-LV-NEXT:    vmovx.f16 s2, s1
-; CHECK-LV-NEXT:    vins.f16 s0, s2
-; CHECK-LV-NEXT:    vmovx.f16 s2, s12
-; CHECK-LV-NEXT:    vins.f16 s3, s2
-; CHECK-LV-NEXT:    vmovx.f16 s2, s15
-; CHECK-LV-NEXT:    vmovx.f16 s18, s14
-; CHECK-LV-NEXT:    vins.f16 s14, s2
-; CHECK-LV-NEXT:    vmovx.f16 s2, s6
-; CHECK-LV-NEXT:    vmovx.f16 s19, s5
-; CHECK-LV-NEXT:    vins.f16 s5, s2
-; CHECK-LV-NEXT:    vmov.f32 s1, s3
-; CHECK-LV-NEXT:    vins.f16 s18, s4
-; CHECK-LV-NEXT:    vins.f16 s19, s7
-; CHECK-LV-NEXT:    vins.f16 s17, s13
-; CHECK-LV-NEXT:    vmov.f32 s2, s14
-; CHECK-LV-NEXT:    vmov.f32 s3, s5
-; CHECK-LV-NEXT:    vadd.f16 q0, q0, q4
-; CHECK-LV-NEXT:    vadd.f16 q0, q0, q2
-; CHECK-LV-NEXT:    vstrw.32 q0, [r1]
-; CHECK-LV-NEXT:    vpop {d8, d9}
-; CHECK-LV-NEXT:    bx lr
-;
-; CHECK-LIS-LABEL: vld3_v16f16:
-; CHECK-LIS:       @ %bb.0: @ %entry
-; CHECK-LIS-NEXT:    .vsave {d8, d9}
-; CHECK-LIS-NEXT:    vpush {d8, d9}
-; CHECK-LIS-NEXT:    vldrw.u32 q0, [r0, #48]
-; CHECK-LIS-NEXT:    vldrw.u32 q2, [r0, #64]
-; CHECK-LIS-NEXT:    vldrw.u32 q3, [r0, #80]
-; CHECK-LIS-NEXT:    vmovx.f16 s6, s2
-; CHECK-LIS-NEXT:    vmov.f32 s4, s1
-; CHECK-LIS-NEXT:    vins.f16 s4, s6
-; CHECK-LIS-NEXT:    vmovx.f16 s6, s9
-; CHECK-LIS-NEXT:    vmov.f32 s5, s8
-; CHECK-LIS-NEXT:    vmovx.f16 s7, s12
-; CHECK-LIS-NEXT:    vins.f16 s5, s6
-; CHECK-LIS-NEXT:    vmov.f32 s6, s11
-; CHECK-LIS-NEXT:    vins.f16 s6, s7
-; CHECK-LIS-NEXT:    vmovx.f16 s16, s15
-; CHECK-LIS-NEXT:    vmov.f32 s7, s14
-; CHECK-LIS-NEXT:    vmovx.f16 s17, s3
-; CHECK-LIS-NEXT:    vins.f16 s7, s16
-; CHECK-LIS-NEXT:    vmovx.f16 s16, s0
-; CHECK-LIS-NEXT:    vins.f16 s16, s2
-; CHECK-LIS-NEXT:    vmovx.f16 s2, s1
-; CHECK-LIS-NEXT:    vins.f16 s0, s2
-; CHECK-LIS-NEXT:    vmovx.f16 s2, s8
-; CHECK-LIS-NEXT:    vins.f16 s3, s2
-; CHECK-LIS-NEXT:    vmovx.f16 s2, s11
-; CHECK-LIS-NEXT:    vmovx.f16 s18, s10
-; CHECK-LIS-NEXT:    vins.f16 s10, s2
-; CHECK-LIS-NEXT:    vmovx.f16 s2, s14
-; CHECK-LIS-NEXT:    vmovx.f16 s19, s13
-; CHECK-LIS-NEXT:    vins.f16 s13, s2
-; CHECK-LIS-NEXT:    vmov.f32 s1, s3
-; CHECK-LIS-NEXT:    vins.f16 s18, s12
-; CHECK-LIS-NEXT:    vins.f16 s19, s15
-; CHECK-LIS-NEXT:    vmov.f32 s3, s13
-; CHECK-LIS-NEXT:    vins.f16 s17, s9
-; CHECK-LIS-NEXT:    vmov.f32 s2, s10
-; CHECK-LIS-NEXT:    vldrw.u32 q3, [r0, #16]
-; CHECK-LIS-NEXT:    vadd.f16 q0, q0, q4
-; CHECK-LIS-NEXT:    vldrw.u32 q2, [r0, #32]
-; CHECK-LIS-NEXT:    vadd.f16 q1, q0, q1
-; CHECK-LIS-NEXT:    vldrw.u32 q0, [r0]
-; CHECK-LIS-NEXT:    vstrw.32 q1, [r1, #16]
-; CHECK-LIS-NEXT:    vmov.f32 s5, s12
-; CHECK-LIS-NEXT:    vmovx.f16 s6, s2
-; CHECK-LIS-NEXT:    vmov.f32 s4, s1
-; CHECK-LIS-NEXT:    vins.f16 s4, s6
-; CHECK-LIS-NEXT:    vmovx.f16 s6, s13
-; CHECK-LIS-NEXT:    vins.f16 s5, s6
-; CHECK-LIS-NEXT:    vmov.f32 s6, s15
-; CHECK-LIS-NEXT:    vmovx.f16 s7, s8
-; CHECK-LIS-NEXT:    vmovx.f16 s16, s11
-; CHECK-LIS-NEXT:    vins.f16 s6, s7
-; CHECK-LIS-NEXT:    vmov.f32 s7, s10
-; CHECK-LIS-NEXT:    vins.f16 s7, s16
-; CHECK-LIS-NEXT:    vmovx.f16 s16, s0
-; CHECK-LIS-NEXT:    vins.f16 s16, s2
-; CHECK-LIS-NEXT:    vmovx.f16 s2, s1
-; CHECK-LIS-NEXT:    vins.f16 s0, s2
-; CHECK-LIS-NEXT:    vmovx.f16 s2, s12
-; CHECK-LIS-NEXT:    vmovx.f16 s17, s3
-; CHECK-LIS-NEXT:    vins.f16 s3, s2
-; CHECK-LIS-NEXT:    vmovx.f16 s2, s15
-; CHECK-LIS-NEXT:    vmovx.f16 s18, s14
-; CHECK-LIS-NEXT:    vins.f16 s14, s2
-; CHECK-LIS-NEXT:    vmovx.f16 s2, s10
-; CHECK-LIS-NEXT:    vmovx.f16 s19, s9
-; CHECK-LIS-NEXT:    vins.f16 s9, s2
-; CHECK-LIS-NEXT:    vmov.f32 s1, s3
-; CHECK-LIS-NEXT:    vins.f16 s18, s8
-; CHECK-LIS-NEXT:    vins.f16 s19, s11
-; CHECK-LIS-NEXT:    vins.f16 s17, s13
-; CHECK-LIS-NEXT:    vmov.f32 s2, s14
-; CHECK-LIS-NEXT:    vmov.f32 s3, s9
-; CHECK-LIS-NEXT:    vadd.f16 q0, q0, q4
-; CHECK-LIS-NEXT:    vadd.f16 q0, q0, q1
-; CHECK-LIS-NEXT:    vstrw.32 q0, [r1]
-; CHECK-LIS-NEXT:    vpop {d8, d9}
-; CHECK-LIS-NEXT:    bx lr
+; CHECK-LABEL: vld3_v16f16:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .vsave {d8, d9, d10}
+; CHECK-NEXT:    vpush {d8, d9, d10}
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #48]
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #64]
+; CHECK-NEXT:    vldrw.u32 q3, [r0, #80]
+; CHECK-NEXT:    vmovx.f16 s4, s2
+; CHECK-NEXT:    vmovx.f16 s16, s0
+; CHECK-NEXT:    vmovx.f16 s7, s1
+; CHECK-NEXT:    vins.f16 s1, s4
+; CHECK-NEXT:    vmovx.f16 s4, s9
+; CHECK-NEXT:    vmovx.f16 s20, s8
+; CHECK-NEXT:    vins.f16 s8, s4
+; CHECK-NEXT:    vins.f16 s16, s2
+; CHECK-NEXT:    vmovx.f16 s2, s11
+; CHECK-NEXT:    vmovx.f16 s18, s10
+; CHECK-NEXT:    vmov.f32 s5, s8
+; CHECK-NEXT:    vins.f16 s10, s2
+; CHECK-NEXT:    vmov.f32 s6, s11
+; CHECK-NEXT:    vmovx.f16 s8, s12
+; CHECK-NEXT:    vmovx.f16 s2, s14
+; CHECK-NEXT:    vmovx.f16 s19, s13
+; CHECK-NEXT:    vmovx.f16 s17, s3
+; CHECK-NEXT:    vins.f16 s6, s8
+; CHECK-NEXT:    vmovx.f16 s8, s15
+; CHECK-NEXT:    vins.f16 s13, s2
+; CHECK-NEXT:    vins.f16 s3, s20
+; CHECK-NEXT:    vmov.f32 s4, s1
+; CHECK-NEXT:    vmov.f32 s1, s3
+; CHECK-NEXT:    vins.f16 s18, s12
+; CHECK-NEXT:    vins.f16 s19, s15
+; CHECK-NEXT:    vins.f16 s17, s9
+; CHECK-NEXT:    vins.f16 s0, s7
+; CHECK-NEXT:    vins.f16 s14, s8
+; CHECK-NEXT:    vmov.f32 s2, s10
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
+; CHECK-NEXT:    vmov.f32 s3, s13
+; CHECK-NEXT:    vadd.f16 q0, q0, q4
+; CHECK-NEXT:    vmov.f32 s7, s14
+; CHECK-NEXT:    vadd.f16 q3, q0, q1
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
+; CHECK-NEXT:    vstrw.32 q3, [r1, #16]
+; CHECK-NEXT:    vmovx.f16 s12, s2
+; CHECK-NEXT:    vmovx.f16 s16, s0
+; CHECK-NEXT:    vmovx.f16 s15, s1
+; CHECK-NEXT:    vins.f16 s1, s12
+; CHECK-NEXT:    vmovx.f16 s12, s9
+; CHECK-NEXT:    vmovx.f16 s20, s8
+; CHECK-NEXT:    vins.f16 s8, s12
+; CHECK-NEXT:    vins.f16 s16, s2
+; CHECK-NEXT:    vmovx.f16 s2, s11
+; CHECK-NEXT:    vmovx.f16 s18, s10
+; CHECK-NEXT:    vmov.f32 s13, s8
+; CHECK-NEXT:    vins.f16 s10, s2
+; CHECK-NEXT:    vmov.f32 s14, s11
+; CHECK-NEXT:    vmovx.f16 s8, s4
+; CHECK-NEXT:    vmovx.f16 s2, s6
+; CHECK-NEXT:    vmovx.f16 s19, s5
+; CHECK-NEXT:    vmovx.f16 s17, s3
+; CHECK-NEXT:    vins.f16 s14, s8
+; CHECK-NEXT:    vmovx.f16 s8, s7
+; CHECK-NEXT:    vins.f16 s5, s2
+; CHECK-NEXT:    vins.f16 s3, s20
+; CHECK-NEXT:    vmov.f32 s12, s1
+; CHECK-NEXT:    vmov.f32 s1, s3
+; CHECK-NEXT:    vins.f16 s18, s4
+; CHECK-NEXT:    vins.f16 s19, s7
+; CHECK-NEXT:    vins.f16 s17, s9
+; CHECK-NEXT:    vins.f16 s0, s15
+; CHECK-NEXT:    vins.f16 s6, s8
+; CHECK-NEXT:    vmov.f32 s2, s10
+; CHECK-NEXT:    vmov.f32 s3, s5
+; CHECK-NEXT:    vmov.f32 s15, s6
+; CHECK-NEXT:    vadd.f16 q0, q0, q4
+; CHECK-NEXT:    vadd.f16 q0, q0, q3
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    vpop {d8, d9, d10}
+; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <48 x half>, ptr %src, align 4
   %s1 = shufflevector <48 x half> %l1, <48 x half> undef, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
diff --git a/llvm/test/CodeGen/Thumb2/mve-vldst4.ll b/llvm/test/CodeGen/Thumb2/mve-vldst4.ll
index 219541cffb940..48ebeab4a6b43 100644
--- a/llvm/test/CodeGen/Thumb2/mve-vldst4.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-vldst4.ll
@@ -23,58 +23,58 @@ define void @vldst4(ptr nocapture readonly %pIn, ptr nocapture %pOut, i32 %numRo
 ; CHECK-NEXT:    vldrh.u16 q1, [r0, #32]
 ; CHECK-NEXT:    vldrh.u16 q4, [r0, #48]
 ; CHECK-NEXT:    vldrh.u16 q3, [r0], #64
+; CHECK-NEXT:    vmovx.f16 s2, s5
+; CHECK-NEXT:    vmovx.f16 s8, s7
+; CHECK-NEXT:    vldrh.u16 q5, [r0, #-48]
+; CHECK-NEXT:    vmovx.f16 s3, s17
+; CHECK-NEXT:    vins.f16 s2, s8
+; CHECK-NEXT:    vmovx.f16 s8, s19
 ; CHECK-NEXT:    vmovx.f16 s26, s4
 ; CHECK-NEXT:    vins.f16 s4, s6
 ; CHECK-NEXT:    vmovx.f16 s6, s6
-; CHECK-NEXT:    vldrh.u16 q5, [r0, #-48]
+; CHECK-NEXT:    vmovx.f16 s0, s13
+; CHECK-NEXT:    vins.f16 s3, s8
+; CHECK-NEXT:    vmovx.f16 s8, s15
 ; CHECK-NEXT:    vmovx.f16 s27, s16
 ; CHECK-NEXT:    vins.f16 s26, s6
 ; CHECK-NEXT:    vmovx.f16 s6, s18
-; CHECK-NEXT:    vmovx.f16 s8, s7
-; CHECK-NEXT:    vmovx.f16 s10, s5
+; CHECK-NEXT:    vmovx.f16 s1, s21
+; CHECK-NEXT:    vins.f16 s0, s8
+; CHECK-NEXT:    vmovx.f16 s8, s23
 ; CHECK-NEXT:    vmovx.f16 s24, s12
-; CHECK-NEXT:    vins.f16 s10, s8
 ; CHECK-NEXT:    vins.f16 s27, s6
 ; CHECK-NEXT:    vmovx.f16 s6, s14
-; CHECK-NEXT:    vmovx.f16 s8, s19
-; CHECK-NEXT:    vmovx.f16 s11, s17
-; CHECK-NEXT:    vmov.f32 s0, s13
-; CHECK-NEXT:    vins.f16 s11, s8
+; CHECK-NEXT:    vins.f16 s1, s8
 ; CHECK-NEXT:    vmovx.f16 s25, s20
 ; CHECK-NEXT:    vins.f16 s24, s6
 ; CHECK-NEXT:    vmovx.f16 s6, s22
-; CHECK-NEXT:    vmovx.f16 s1, s15
-; CHECK-NEXT:    vmovx.f16 s8, s13
-; CHECK-NEXT:    vins.f16 s20, s22
+; CHECK-NEXT:    vins.f16 s13, s15
 ; CHECK-NEXT:    vins.f16 s16, s18
-; CHECK-NEXT:    vmov.f32 s2, s5
+; CHECK-NEXT:    vins.f16 s20, s22
 ; CHECK-NEXT:    vins.f16 s25, s6
-; CHECK-NEXT:    vmov.f32 s3, s17
-; CHECK-NEXT:    vins.f16 s0, s15
-; CHECK-NEXT:    vmovx.f16 s9, s21
-; CHECK-NEXT:    vins.f16 s8, s1
-; CHECK-NEXT:    vmovx.f16 s1, s23
-; CHECK-NEXT:    vins.f16 s12, s14
+; CHECK-NEXT:    vmul.f16 q2, q0, r2
+; CHECK-NEXT:    vins.f16 s5, s7
+; CHECK-NEXT:    vins.f16 s17, s19
 ; CHECK-NEXT:    vins.f16 s21, s23
+; CHECK-NEXT:    vmov.f32 s0, s13
+; CHECK-NEXT:    vins.f16 s12, s14
 ; CHECK-NEXT:    vmov.f32 s14, s4
 ; CHECK-NEXT:    vmov.f32 s15, s16
-; CHECK-NEXT:    vins.f16 s9, s1
-; CHECK-NEXT:    vmov.f32 s13, s20
 ; CHECK-NEXT:    vmul.f16 q6, q6, r2
+; CHECK-NEXT:    vmov.f32 s13, s20
+; CHECK-NEXT:    vmovx.f16 s6, s24
 ; CHECK-NEXT:    vmul.f16 q3, q3, r2
-; CHECK-NEXT:    vins.f16 s2, s7
-; CHECK-NEXT:    vins.f16 s3, s19
+; CHECK-NEXT:    vmov.f32 s2, s5
+; CHECK-NEXT:    vmov.f32 s3, s17
+; CHECK-NEXT:    vmovx.f16 s4, s12
 ; CHECK-NEXT:    vmov.f32 s1, s21
+; CHECK-NEXT:    vins.f16 s4, s6
 ; CHECK-NEXT:    vmul.f16 q0, q0, r2
-; CHECK-NEXT:    vmovx.f16 s4, s12
-; CHECK-NEXT:    vmovx.f16 s6, s24
-; CHECK-NEXT:    vmul.f16 q2, q2, r2
+; CHECK-NEXT:    vmovx.f16 s6, s9
+; CHECK-NEXT:    vmovx.f16 s5, s1
 ; CHECK-NEXT:    vmovx.f16 s7, s0
 ; CHECK-NEXT:    vins.f16 s0, s8
 ; CHECK-NEXT:    vmovx.f16 s8, s8
-; CHECK-NEXT:    vins.f16 s4, s6
-; CHECK-NEXT:    vmovx.f16 s5, s1
-; CHECK-NEXT:    vmovx.f16 s6, s9
 ; CHECK-NEXT:    vins.f16 s7, s8
 ; CHECK-NEXT:    vins.f16 s5, s6
 ; CHECK-NEXT:    vmovx.f16 s6, s13
diff --git a/llvm/test/CodeGen/Thumb2/mve-vst3.ll b/llvm/test/CodeGen/Thumb2/mve-vst3.ll
index d80dd5a673e20..b72d5ba29a3d1 100644
--- a/llvm/test/CodeGen/Thumb2/mve-vst3.ll
+++ b/llvm/test/CodeGen/Thumb2/mve-vst3.ll
@@ -403,8 +403,8 @@ define void @vst3_v16i16(ptr %src, ptr %dst) {
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    .pad #48
-; CHECK-NEXT:    sub sp, #48
+; CHECK-NEXT:    .pad #32
+; CHECK-NEXT:    sub sp, #32
 ; CHECK-NEXT:    vldrw.u32 q2, [r0]
 ; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
 ; CHECK-NEXT:    vldrw.u32 q7, [r0, #80]
@@ -414,63 +414,62 @@ define void @vst3_v16i16(ptr %src, ptr %dst) {
 ; CHECK-NEXT:    vins.f16 s0, s7
 ; CHECK-NEXT:    vmov.f32 s2, s11
 ; CHECK-NEXT:    vmov.u16 r2, q1[7]
-; CHECK-NEXT:    vmov.f64 d12, d4
-; CHECK-NEXT:    vstrw.32 q1, [sp, #32] @ 16-byte Spill
+; CHECK-NEXT:    vmov.f64 d10, d4
+; CHECK-NEXT:    vmov q6, q1
 ; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
-; CHECK-NEXT:    vmov.f32 s26, s10
+; CHECK-NEXT:    vstrw.32 q6, [sp] @ 16-byte Spill
+; CHECK-NEXT:    vmov.f32 s22, s10
 ; CHECK-NEXT:    vldrw.u32 q2, [r0, #64]
 ; CHECK-NEXT:    vmov.f32 s13, s0
-; CHECK-NEXT:    vstrw.32 q6, [sp] @ 16-byte Spill
 ; CHECK-NEXT:    vmov.16 q3[6], r2
 ; CHECK-NEXT:    vmovx.f16 s0, s10
 ; CHECK-NEXT:    vins.f16 s12, s0
 ; CHECK-NEXT:    vmovx.f16 s0, s2
+; CHECK-NEXT:    vmovx.f16 s23, s11
+; CHECK-NEXT:    vins.f16 s11, s0
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #48]
 ; CHECK-NEXT:    vmov.f32 s14, s11
-; CHECK-NEXT:    vins.f16 s14, s0
-; CHECK-NEXT:    vmov.f32 s20, s7
-; CHECK-NEXT:    vmov q0, q3
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #48]
-; CHECK-NEXT:    vmov.u16 r2, q3[5]
-; CHECK-NEXT:    vins.f16 s20, s15
+; CHECK-NEXT:    vmov.f32 s11, s7
+; CHECK-NEXT:    vins.f16 s15, s23
+; CHECK-NEXT:    vmov.u16 r2, q0[5]
+; CHECK-NEXT:    vins.f16 s11, s3
 ; CHECK-NEXT:    vmov.16 q4[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q3[7]
-; CHECK-NEXT:    vmov.f32 s17, s20
-; CHECK-NEXT:    vmovx.f16 s20, s31
+; CHECK-NEXT:    vstrw.32 q3, [sp, #16] @ 16-byte Spill
+; CHECK-NEXT:    vmov.f32 s17, s11
+; CHECK-NEXT:    vmov.u16 r2, q0[7]
+; CHECK-NEXT:    vmov.f64 d6, d10
 ; CHECK-NEXT:    vmov.16 q4[6], r2
+; CHECK-NEXT:    vmovx.f16 s11, s31
+; CHECK-NEXT:    vins.f16 s20, s24
+; CHECK-NEXT:    vmov.u16 r0, q6[1]
+; CHECK-NEXT:    vins.f16 s19, s11
+; CHECK-NEXT:    vmovx.f16 s11, s8
 ; CHECK-NEXT:    vmov.f32 s18, s7
 ; CHECK-NEXT:    vmovx.f16 s7, s30
 ; CHECK-NEXT:    vins.f16 s16, s7
 ; CHECK-NEXT:    vmovx.f16 s7, s18
 ; CHECK-NEXT:    vins.f16 s31, s7
-; CHECK-NEXT:    vmovx.f16 s7, s11
-; CHECK-NEXT:    vins.f16 s3, s7
-; CHECK-NEXT:    vins.f16 s19, s20
-; CHECK-NEXT:    vstrw.32 q0, [sp, #16] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #32] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s20, s24
-; CHECK-NEXT:    vmovx.f16 s11, s8
-; CHECK-NEXT:    vmov.f32 s7, s25
-; CHECK-NEXT:    vins.f16 s20, s0
-; CHECK-NEXT:    vmov.u16 r0, q0[1]
-; CHECK-NEXT:    vins.f16 s7, s1
+; CHECK-NEXT:    vmov.f32 s7, s13
+; CHECK-NEXT:    vmov.f32 s14, s22
 ; CHECK-NEXT:    vmov.16 q5[4], r0
-; CHECK-NEXT:    vmov.u16 r0, q3[1]
-; CHECK-NEXT:    vmov.f32 s23, s7
-; CHECK-NEXT:    vmovx.f16 s7, s24
+; CHECK-NEXT:    vins.f16 s7, s25
 ; CHECK-NEXT:    vmov.f32 s24, s4
+; CHECK-NEXT:    vmov.f32 s23, s7
+; CHECK-NEXT:    vmovx.f16 s7, s12
 ; CHECK-NEXT:    vins.f16 s8, s7
-; CHECK-NEXT:    vins.f16 s24, s12
+; CHECK-NEXT:    vins.f16 s24, s0
 ; CHECK-NEXT:    vmov.f32 s21, s8
+; CHECK-NEXT:    vmov.u16 r0, q0[1]
 ; CHECK-NEXT:    vmov.f32 s8, s5
 ; CHECK-NEXT:    vmov.16 q6[4], r0
-; CHECK-NEXT:    vins.f16 s8, s13
+; CHECK-NEXT:    vins.f16 s8, s1
 ; CHECK-NEXT:    vmovx.f16 s4, s4
 ; CHECK-NEXT:    vmov.f32 s27, s8
 ; CHECK-NEXT:    vmovx.f16 s8, s28
 ; CHECK-NEXT:    vins.f16 s28, s4
 ; CHECK-NEXT:    vmov.f32 s4, s6
-; CHECK-NEXT:    vmov.u16 r0, q3[3]
-; CHECK-NEXT:    vins.f16 s4, s14
+; CHECK-NEXT:    vmov.u16 r0, q0[3]
+; CHECK-NEXT:    vins.f16 s4, s2
 ; CHECK-NEXT:    vmov.16 q0[2], r0
 ; CHECK-NEXT:    vins.f16 s26, s8
 ; CHECK-NEXT:    vmov.f32 s2, s4
@@ -480,36 +479,33 @@ define void @vst3_v16i16(ptr %src, ptr %dst) {
 ; CHECK-NEXT:    vmovx.f16 s0, s5
 ; CHECK-NEXT:    vins.f16 s30, s4
 ; CHECK-NEXT:    vldrw.u32 q1, [sp] @ 16-byte Reload
+; CHECK-NEXT:    vmov.f32 s8, s14
 ; CHECK-NEXT:    vins.f16 s29, s0
-; CHECK-NEXT:    vmov.f32 s0, s29
-; CHECK-NEXT:    vins.f16 s22, s11
-; CHECK-NEXT:    vmov.f32 s3, s30
-; CHECK-NEXT:    vstrw.32 q5, [r1]
-; CHECK-NEXT:    vmov.f32 s29, s5
-; CHECK-NEXT:    vstrw.32 q0, [r1, #64]
-; CHECK-NEXT:    vmov.f32 s30, s6
-; CHECK-NEXT:    vmov.f32 s8, s6
-; CHECK-NEXT:    vldrw.u32 q1, [sp, #32] @ 16-byte Reload
 ; CHECK-NEXT:    vmov.f32 s18, s31
 ; CHECK-NEXT:    vmov.u16 r0, q1[3]
 ; CHECK-NEXT:    vins.f16 s8, s6
 ; CHECK-NEXT:    vmov.16 q1[2], r0
-; CHECK-NEXT:    vmov.f32 s25, s28
+; CHECK-NEXT:    vins.f16 s22, s11
 ; CHECK-NEXT:    vmov.f32 s6, s8
 ; CHECK-NEXT:    vmovx.f16 s8, s9
-; CHECK-NEXT:    vmovx.f16 s4, s29
+; CHECK-NEXT:    vmovx.f16 s4, s13
 ; CHECK-NEXT:    vins.f16 s5, s8
-; CHECK-NEXT:    vmovx.f16 s8, s30
+; CHECK-NEXT:    vmovx.f16 s8, s14
 ; CHECK-NEXT:    vins.f16 s9, s4
 ; CHECK-NEXT:    vins.f16 s10, s8
 ; CHECK-NEXT:    vmov.f32 s4, s9
 ; CHECK-NEXT:    vmov.f32 s7, s10
-; CHECK-NEXT:    vstrw.32 q6, [r1, #48]
+; CHECK-NEXT:    vstrw.32 q4, [r1, #80]
 ; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
 ; CHECK-NEXT:    vldrw.u32 q1, [sp, #16] @ 16-byte Reload
-; CHECK-NEXT:    vstrw.32 q4, [r1, #80]
+; CHECK-NEXT:    vmov.f32 s25, s28
+; CHECK-NEXT:    vstrw.32 q5, [r1]
+; CHECK-NEXT:    vmov.f32 s0, s29
 ; CHECK-NEXT:    vstrw.32 q1, [r1, #32]
-; CHECK-NEXT:    add sp, #48
+; CHECK-NEXT:    vmov.f32 s3, s30
+; CHECK-NEXT:    vstrw.32 q6, [r1, #48]
+; CHECK-NEXT:    vstrw.32 q0, [r1, #64]
+; CHECK-NEXT:    add sp, #32
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    bx lr
 entry:
diff --git a/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast_from_memory.ll b/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast_from_memory.ll
index c40053dcb3e70..0e4145fd2ed49 100644
--- a/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast_from_memory.ll
+++ b/llvm/test/CodeGen/X86/any_extend_vector_inreg_of_broadcast_from_memory.ll
@@ -5091,93 +5091,94 @@ define void @vec512_i128_widen_to_i256_factor2_broadcast_to_v2i256_factor2(ptr %
 ;
 ; AVX-LABEL: vec512_i128_widen_to_i256_factor2_broadcast_to_v2i256_factor2:
 ; AVX:       # %bb.0:
+; AVX-NEXT:    pushq %r14
 ; AVX-NEXT:    pushq %rbx
 ; AVX-NEXT:    movq 16(%rdi), %rcx
 ; AVX-NEXT:    movq %rcx, %rax
 ; AVX-NEXT:    movq %rcx, %r8
-; AVX-NEXT:    movq %rcx, %r9
 ; AVX-NEXT:    movq %rcx, %r10
-; AVX-NEXT:    movl %ecx, %r11d
+; AVX-NEXT:    movq %rcx, %r11
+; AVX-NEXT:    movl %ecx, %r9d
 ; AVX-NEXT:    movl %ecx, %ebx
 ; AVX-NEXT:    vmovd %ecx, %xmm0
 ; AVX-NEXT:    shrl $8, %ecx
 ; AVX-NEXT:    vpinsrb $1, %ecx, %xmm0, %xmm0
+; AVX-NEXT:    movq 24(%rdi), %r14
 ; AVX-NEXT:    shrl $16, %ebx
 ; AVX-NEXT:    vpinsrb $2, %ebx, %xmm0, %xmm0
-; AVX-NEXT:    shrl $24, %r11d
-; AVX-NEXT:    vpinsrb $3, %r11d, %xmm0, %xmm0
-; AVX-NEXT:    shrq $32, %r10
-; AVX-NEXT:    vpinsrb $4, %r10d, %xmm0, %xmm0
-; AVX-NEXT:    shrq $40, %r9
-; AVX-NEXT:    vpinsrb $5, %r9d, %xmm0, %xmm0
+; AVX-NEXT:    movq %r14, %rcx
+; AVX-NEXT:    shrl $24, %r9d
+; AVX-NEXT:    vpinsrb $3, %r9d, %xmm0, %xmm0
+; AVX-NEXT:    movq %r14, %r9
+; AVX-NEXT:    shrq $32, %r11
+; AVX-NEXT:    vpinsrb $4, %r11d, %xmm0, %xmm0
+; AVX-NEXT:    movq %r14, %r11
+; AVX-NEXT:    shrq $40, %r10
+; AVX-NEXT:    vpinsrb $5, %r10d, %xmm0, %xmm0
+; AVX-NEXT:    movq %r14, %r10
 ; AVX-NEXT:    shrq $48, %r8
 ; AVX-NEXT:    vpinsrb $6, %r8d, %xmm0, %xmm0
-; AVX-NEXT:    movq 24(%rdi), %rcx
+; AVX-NEXT:    movl %r14d, %r8d
 ; AVX-NEXT:    shrq $56, %rax
 ; AVX-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0
-; AVX-NEXT:    movl %ecx, %eax
-; AVX-NEXT:    shrl $8, %eax
-; AVX-NEXT:    vpinsrb $8, %ecx, %xmm0, %xmm0
-; AVX-NEXT:    vpinsrb $9, %eax, %xmm0, %xmm0
-; AVX-NEXT:    movl %ecx, %eax
+; AVX-NEXT:    movl %r14d, %eax
+; AVX-NEXT:    vpinsrb $8, %r14d, %xmm0, %xmm0
+; AVX-NEXT:    shrl $8, %r14d
+; AVX-NEXT:    vpinsrb $9, %r14d, %xmm0, %xmm0
+; AVX-NEXT:    movq (%rdi), %rbx
 ; AVX-NEXT:    shrl $16, %eax
 ; AVX-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0
-; AVX-NEXT:    movl %ecx, %eax
-; AVX-NEXT:    shrl $24, %eax
-; AVX-NEXT:    vpinsrb $11, %eax, %xmm0, %xmm0
-; AVX-NEXT:    movq %rcx, %rax
-; AVX-NEXT:    shrq $32, %rax
-; AVX-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0
-; AVX-NEXT:    movq %rcx, %rax
-; AVX-NEXT:    shrq $40, %rax
-; AVX-NEXT:    vpinsrb $13, %eax, %xmm0, %xmm0
-; AVX-NEXT:    movq %rcx, %rax
-; AVX-NEXT:    shrq $48, %rax
-; AVX-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0
-; AVX-NEXT:    movq (%rdi), %rax
+; AVX-NEXT:    movq %rbx, %rax
+; AVX-NEXT:    shrl $24, %r8d
+; AVX-NEXT:    vpinsrb $11, %r8d, %xmm0, %xmm0
+; AVX-NEXT:    movq %rbx, %r8
+; AVX-NEXT:    shrq $32, %r10
+; AVX-NEXT:    vpinsrb $12, %r10d, %xmm0, %xmm0
+; AVX-NEXT:    movq %rbx, %r10
+; AVX-NEXT:    shrq $40, %r11
+; AVX-NEXT:    vpinsrb $13, %r11d, %xmm0, %xmm0
+; AVX-NEXT:    movq %rbx, %r11
+; AVX-NEXT:    shrq $48, %r9
+; AVX-NEXT:    vpinsrb $14, %r9d, %xmm0, %xmm0
+; AVX-NEXT:    movl %ebx, %r9d
 ; AVX-NEXT:    shrq $56, %rcx
 ; AVX-NEXT:    vpinsrb $15, %ecx, %xmm0, %xmm0
-; AVX-NEXT:    movl %eax, %ecx
-; AVX-NEXT:    shrl $8, %ecx
-; AVX-NEXT:    vmovd %eax, %xmm1
-; AVX-NEXT:    vpinsrb $1, %ecx, %xmm1, %xmm1
-; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    movl %ebx, %ecx
+; AVX-NEXT:    vmovd %ebx, %xmm1
+; AVX-NEXT:    shrl $8, %ebx
+; AVX-NEXT:    vpinsrb $1, %ebx, %xmm1, %xmm1
+; AVX-NEXT:    movq 8(%rdi), %rbx
 ; AVX-NEXT:    shrl $16, %ecx
 ; AVX-NEXT:    vpinsrb $2, %ecx, %xmm1, %xmm1
-; AVX-NEXT:    movl %eax, %ecx
-; AVX-NEXT:    shrl $24, %ecx
-; AVX-NEXT:    vpinsrb $3, %ecx, %xmm1, %xmm1
-; AVX-NEXT:    movq %rax, %rcx
-; AVX-NEXT:    shrq $32, %rcx
-; AVX-NEXT:    vpinsrb $4, %ecx, %xmm1, %xmm1
-; AVX-NEXT:    movq %rax, %rcx
-; AVX-NEXT:    shrq $40, %rcx
-; AVX-NEXT:    vpinsrb $5, %ecx, %xmm1, %xmm1
-; AVX-NEXT:    movq %rax, %rcx
-; AVX-NEXT:    shrq $48, %rcx
-; AVX-NEXT:    vpinsrb $6, %ecx, %xmm1, %xmm1
-; AVX-NEXT:    movq 8(%rdi), %rcx
+; AVX-NEXT:    movq %rbx, %rcx
+; AVX-NEXT:    shrl $24, %r9d
+; AVX-NEXT:    vpinsrb $3, %r9d, %xmm1, %xmm1
+; AVX-NEXT:    movq %rbx, %rdi
+; AVX-NEXT:    shrq $32, %r11
+; AVX-NEXT:    vpinsrb $4, %r11d, %xmm1, %xmm1
+; AVX-NEXT:    movq %rbx, %r9
+; AVX-NEXT:    shrq $40, %r10
+; AVX-NEXT:    vpinsrb $5, %r10d, %xmm1, %xmm1
+; AVX-NEXT:    movq %rbx, %r10
+; AVX-NEXT:    shrq $48, %r8
+; AVX-NEXT:    vpinsrb $6, %r8d, %xmm1, %xmm1
+; AVX-NEXT:    movl %ebx, %r8d
 ; AVX-NEXT:    shrq $56, %rax
 ; AVX-NEXT:    vpinsrb $7, %eax, %xmm1, %xmm1
-; AVX-NEXT:    movl %ecx, %eax
-; AVX-NEXT:    shrl $8, %eax
-; AVX-NEXT:    vpinsrb $8, %ecx, %xmm1, %xmm1
-; AVX-NEXT:    vpinsrb $9, %eax, %xmm1, %xmm1
-; AVX-NEXT:    movl %ecx, %eax
+; AVX-NEXT:    movl %ebx, %eax
+; AVX-NEXT:    vpinsrb $8, %ebx, %xmm1, %xmm1
+; AVX-NEXT:    shrl $8, %ebx
+; AVX-NEXT:    vpinsrb $9, %ebx, %xmm1, %xmm1
 ; AVX-NEXT:    shrl $16, %eax
 ; AVX-NEXT:    vpinsrb $10, %eax, %xmm1, %xmm1
-; AVX-NEXT:    movl %ecx, %eax
-; AVX-NEXT:    shrl $24, %eax
-; AVX-NEXT:    vpinsrb $11, %eax, %xmm1, %xmm1
-; AVX-NEXT:    movq %rcx, %rax
-; AVX-NEXT:    shrq $32, %rax
-; AVX-NEXT:    vpinsrb $12, %eax, %xmm1, %xmm1
-; AVX-NEXT:    movq %rcx, %rax
-; AVX-NEXT:    shrq $40, %rax
-; AVX-NEXT:    vpinsrb $13, %eax, %xmm1, %xmm1
-; AVX-NEXT:    movq %rcx, %rax
-; AVX-NEXT:    shrq $48, %rax
-; AVX-NEXT:    vpinsrb $14, %eax, %xmm1, %xmm1
+; AVX-NEXT:    shrl $24, %r8d
+; AVX-NEXT:    vpinsrb $11, %r8d, %xmm1, %xmm1
+; AVX-NEXT:    shrq $32, %r10
+; AVX-NEXT:    vpinsrb $12, %r10d, %xmm1, %xmm1
+; AVX-NEXT:    shrq $40, %r9
+; AVX-NEXT:    vpinsrb $13, %r9d, %xmm1, %xmm1
+; AVX-NEXT:    shrq $48, %rdi
+; AVX-NEXT:    vpinsrb $14, %edi, %xmm1, %xmm1
 ; AVX-NEXT:    shrq $56, %rcx
 ; AVX-NEXT:    vpinsrb $15, %ecx, %xmm1, %xmm1
 ; AVX-NEXT:    vpaddb 48(%rsi), %xmm0, %xmm2
@@ -5189,97 +5190,99 @@ define void @vec512_i128_widen_to_i256_factor2_broadcast_to_v2i256_factor2(ptr %
 ; AVX-NEXT:    vmovdqa %xmm3, 32(%rdx)
 ; AVX-NEXT:    vmovdqa %xmm2, 48(%rdx)
 ; AVX-NEXT:    popq %rbx
+; AVX-NEXT:    popq %r14
 ; AVX-NEXT:    retq
 ;
 ; AVX2-LABEL: vec512_i128_widen_to_i256_factor2_broadcast_to_v2i256_factor2:
 ; AVX2:       # %bb.0:
+; AVX2-NEXT:    pushq %r14
 ; AVX2-NEXT:    pushq %rbx
 ; AVX2-NEXT:    movq 16(%rdi), %rcx
 ; AVX2-NEXT:    movq %rcx, %rax
 ; AVX2-NEXT:    movq %rcx, %r8
-; AVX2-NEXT:    movq %rcx, %r9
 ; AVX2-NEXT:    movq %rcx, %r10
-; AVX2-NEXT:    movl %ecx, %r11d
+; AVX2-NEXT:    movq %rcx, %r11
+; AVX2-NEXT:    movl %ecx, %r9d
 ; AVX2-NEXT:    movl %ecx, %ebx
 ; AVX2-NEXT:    vmovd %ecx, %xmm0
 ; AVX2-NEXT:    shrl $8, %ecx
 ; AVX2-NEXT:    vpinsrb $1, %ecx, %xmm0, %xmm0
+; AVX2-NEXT:    movq 24(%rdi), %r14
 ; AVX2-NEXT:    shrl $16, %ebx
 ; AVX2-NEXT:    vpinsrb $2, %ebx, %xmm0, %xmm0
-; AVX2-NEXT:    shrl $24, %r11d
-; AVX2-NEXT:    vpinsrb $3, %r11d, %xmm0, %xmm0
-; AVX2-NEXT:    shrq $32, %r10
-; AVX2-NEXT:    vpinsrb $4, %r10d, %xmm0, %xmm0
-; AVX2-NEXT:    shrq $40, %r9
-; AVX2-NEXT:    vpinsrb $5, %r9d, %xmm0, %xmm0
+; AVX2-NEXT:    movq %r14, %rcx
+; AVX2-NEXT:    shrl $24, %r9d
+; AVX2-NEXT:    vpinsrb $3, %r9d, %xmm0, %xmm0
+; AVX2-NEXT:    movq %r14, %r9
+; AVX2-NEXT:    shrq $32, %r11
+; AVX2-NEXT:    vpinsrb $4, %r11d, %xmm0, %xmm0
+; AVX2-NEXT:    movq %r14, %r11
+; AVX2-NEXT:    shrq $40, %r10
+; AVX2-NEXT:    vpinsrb $5, %r10d, %xmm0, %xmm0
+; AVX2-NEXT:    movq %r14, %r10
 ; AVX2-NEXT:    shrq $48, %r8
 ; AVX2-NEXT:    vpinsrb $6, %r8d, %xmm0, %xmm0
-; AVX2-NEXT:    movq 24(%rdi), %rcx
+; AVX2-NEXT:    movl %r14d, %r8d
 ; AVX2-NEXT:    shrq $56, %rax
 ; AVX2-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0
-; AVX2-NEXT:    movl %ecx, %eax
-; AVX2-NEXT:    shrl $8, %eax
-; AVX2-NEXT:    vpinsrb $8, %ecx, %xmm0, %xmm0
-; AVX2-NEXT:    vpinsrb $9, %eax, %xmm0, %xmm0
-; AVX2-NEXT:    movl %ecx, %eax
+; AVX2-NEXT:    movl %r14d, %eax
+; AVX2-NEXT:    vpinsrb $8, %r14d, %xmm0, %xmm0
+; AVX2-NEXT:    shrl $8, %r14d
+; AVX2-NEXT:    vpinsrb $9, %r14d, %xmm0, %xmm0
+; AVX2-NEXT:    movq (%rdi), %rbx
 ; AVX2-NEXT:    shrl $16, %eax
 ; AVX2-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0
-; AVX2-NEXT:    movl %ecx, %eax
-; AVX2-NEXT:    shrl $24, %eax
-; AVX2-NEXT:    vpinsrb $11, %eax, %xmm0, %xmm0
-; AVX2-NEXT:    movq %rcx, %rax
-; AVX2-NEXT:    shrq $32, %rax
-; AVX2-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0
-; AVX2-NEXT:    movq %rcx, %rax
-; AVX2-NEXT:    shrq $40, %rax
-; AVX2-NEXT:    vpinsrb $13, %eax, %xmm0, %xmm0
-; AVX2-NEXT:    movq %rcx, %rax
-; AVX2-NEXT:    shrq $48, %rax
-; AVX2-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0
-; AVX2-NEXT:    movq (%rdi), %rax
+; AVX2-NEXT:    movq %rbx, %rax
+; AVX2-NEXT:    shrl $24, %r8d
+; AVX2-NEXT:    vpinsrb $11, %r8d, %xmm0, %xmm0
+; AVX2-NEXT:    movq %rbx, %r8
+; AVX2-NEXT:    shrq $32, %r10
+; AVX2-NEXT:    vpinsrb $12, %r10d, %xmm0, %xmm0
+; AVX2-NEXT:    movq %rbx, %r10
+; AVX2-NEXT:    shrq $40, %r11
+; AVX2-NEXT:    vpinsrb $13, %r11d, %xmm0, %xmm0
+; AVX2-NEXT:    movq %rbx, %r11
+; AVX2-NEXT:    shrq $48, %r9
+; AVX2-NEXT:    vpinsrb $14, %r9d, %xmm0, %xmm0
+; AVX2-NEXT:    movl %ebx, %r9d
 ; AVX2-NEXT:    shrq $56, %rcx
 ; AVX2-NEXT:    vpinsrb $15, %ecx, %xmm0, %xmm0
-; AVX2-NEXT:    movl %eax, %ecx
-; AVX2-NEXT:    shrl $8, %ecx
-; AVX2-NEXT:    vmovd %eax, %xmm1
-; AVX2-NEXT:    vpinsrb $1, %ecx, %xmm1, %xmm1
-; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    movl %ebx, %ecx
+; AVX2-NEXT:    vmovd %ebx, %xmm1
+; AVX2-NEXT:    shrl $8, %ebx
+; AVX2-NEXT:    vpinsrb $1, %ebx, %xmm1, %xmm1
+; AVX2-NEXT:    movq 8(%rdi), %rbx
 ; AVX2-NEXT:    shrl $16, %ecx
 ; AVX2-NEXT:    vpinsrb $2, %ecx, %xmm1, %xmm1
-; AVX2-NEXT:    movl %eax, %ecx
-; AVX2-NEXT:    shrl $24, %ecx
-; AVX2-NEXT:    vpinsrb $3, %ecx, %xmm1, %xmm1
-; AVX2-NEXT:    movq %rax, %rcx
-; AVX2-NEXT:    shrq $32, %rcx
-; AVX2-NEXT:    vpinsrb $4, %ecx, %xmm1, %xmm1
-; AVX2-NEXT:    movq %rax, %rcx
-; AVX2-NEXT:    shrq $40, %rcx
-; AVX2-NEXT:    vpinsrb $5, %ecx, %xmm1, %xmm1
-; AVX2-NEXT:    movq %rax, %rcx
-; AVX2-NEXT:    shrq $48, %rcx
-; AVX2-NEXT:    vpinsrb $6, %ecx, %xmm1, %xmm1
-; AVX2-NEXT:    movq 8(%rdi), %rcx
+; AVX2-NEXT:    movq %rbx, %rcx
+; AVX2-NEXT:    shrl $24, %r9d
+; AVX2-NEXT:    vpinsrb $3, %r9d, %xmm1, %xmm1
+; AVX2-NEXT:    movq %rbx, %rdi
+; AVX2-NEXT:    shrq $32, %r11
+; AVX2-NEXT:    vpinsrb $4, %r11d, %xmm1, %xmm1
+; AVX2-NEXT:    movq %rbx, %r9
+; AVX2-NEXT:    shrq $40, %r10
+; AVX2-NEXT:    vpinsrb $5, %r10d, %xmm1, %xmm1
+; AVX2-NEXT:    movq %rbx, %r10
+; AVX2-NEXT:    shrq $48, %r8
+; AVX2-NEXT:    vpinsrb $6, %r8d, %xmm1, %xmm1
+; AVX2-NEXT:    movl %ebx, %r8d
 ; AVX2-NEXT:    shrq $56, %rax
 ; AVX2-NEXT:    vpinsrb $7, %eax, %xmm1, %xmm1
-; AVX2-NEXT:    movl %ecx, %eax
-; AVX2-NEXT:    shrl $8, %eax
-; AVX2-NEXT:    vpinsrb $8, %ecx, %xmm1, %xmm1
-; AVX2-NEXT:    vpinsrb $9, %eax, %xmm1, %xmm1
-; AVX2-NEXT:    movl %ecx, %eax
+; AVX2-NEXT:    movl %ebx, %eax
+; AVX2-NEXT:    vpinsrb $8, %ebx, %xmm1, %xmm1
+; AVX2-NEXT:    shrl $8, %ebx
+; AVX2-NEXT:    vpinsrb $9, %ebx, %xmm1, %xmm1
 ; AVX2-NEXT:    shrl $16, %eax
 ; AVX2-NEXT:    vpinsrb $10, %eax, %xmm1, %xmm1
-; AVX2-NEXT:    movl %ecx, %eax
-; AVX2-NEXT:    shrl $24, %eax
-; AVX2-NEXT:    vpinsrb $11, %eax, %xmm1, %xmm1
-; AVX2-NEXT:    movq %rcx, %rax
-; AVX2-NEXT:    shrq $32, %rax
-; AVX2-NEXT:    vpinsrb $12, %eax, %xmm1, %xmm1
-; AVX2-NEXT:    movq %rcx, %rax
-; AVX2-NEXT:    shrq $40, %rax
-; AVX2-NEXT:    vpinsrb $13, %eax, %xmm1, %xmm1
-; AVX2-NEXT:    movq %rcx, %rax
-; AVX2-NEXT:    shrq $48, %rax
-; AVX2-NEXT:    vpinsrb $14, %eax, %xmm1, %xmm1
+; AVX2-NEXT:    shrl $24, %r8d
+; AVX2-NEXT:    vpinsrb $11, %r8d, %xmm1, %xmm1
+; AVX2-NEXT:    shrq $32, %r10
+; AVX2-NEXT:    vpinsrb $12, %r10d, %xmm1, %xmm1
+; AVX2-NEXT:    shrq $40, %r9
+; AVX2-NEXT:    vpinsrb $13, %r9d, %xmm1, %xmm1
+; AVX2-NEXT:    shrq $48, %rdi
+; AVX2-NEXT:    vpinsrb $14, %edi, %xmm1, %xmm1
 ; AVX2-NEXT:    shrq $56, %rcx
 ; AVX2-NEXT:    vpinsrb $15, %ecx, %xmm1, %xmm1
 ; AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0
@@ -5288,6 +5291,7 @@ define void @vec512_i128_widen_to_i256_factor2_broadcast_to_v2i256_factor2(ptr %
 ; AVX2-NEXT:    vmovdqa %ymm0, (%rdx)
 ; AVX2-NEXT:    vmovdqa %ymm1, 32(%rdx)
 ; AVX2-NEXT:    popq %rbx
+; AVX2-NEXT:    popq %r14
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/machine-cp.ll b/llvm/test/CodeGen/X86/machine-cp.ll
index f84960485840d..420856154dbb5 100644
--- a/llvm/test/CodeGen/X86/machine-cp.ll
+++ b/llvm/test/CodeGen/X86/machine-cp.ll
@@ -99,36 +99,35 @@ while.end:                                        ; preds = %while.body, %entry
 define <16 x float> @foo(<16 x float> %x) {
 ; CHECK-LABEL: foo:
 ; CHECK:       ## %bb.0: ## %bb
-; CHECK-NEXT:    xorps %xmm5, %xmm5
-; CHECK-NEXT:    cvttps2dq %xmm3, %xmm8
 ; CHECK-NEXT:    movaps %xmm3, %xmm4
+; CHECK-NEXT:    xorps %xmm5, %xmm5
+; CHECK-NEXT:    cvttps2dq %xmm3, %xmm3
 ; CHECK-NEXT:    cmpltps %xmm5, %xmm4
 ; CHECK-NEXT:    movaps {{.*#+}} xmm7 = [13,14,15,16]
 ; CHECK-NEXT:    movaps %xmm4, %xmm6
 ; CHECK-NEXT:    orps %xmm7, %xmm6
-; CHECK-NEXT:    cvtdq2ps %xmm8, %xmm3
+; CHECK-NEXT:    cvtdq2ps %xmm3, %xmm3
 ; CHECK-NEXT:    andps %xmm7, %xmm3
 ; CHECK-NEXT:    andps %xmm6, %xmm3
 ; CHECK-NEXT:    andnps %xmm4, %xmm6
 ; CHECK-NEXT:    cvttps2dq %xmm2, %xmm4
-; CHECK-NEXT:    movaps %xmm2, %xmm7
-; CHECK-NEXT:    cmpltps %xmm5, %xmm7
-; CHECK-NEXT:    movaps {{.*#+}} xmm8 = [9,10,11,12]
-; CHECK-NEXT:    movaps %xmm7, %xmm9
-; CHECK-NEXT:    orps %xmm8, %xmm9
-; CHECK-NEXT:    cvtdq2ps %xmm4, %xmm2
-; CHECK-NEXT:    andps %xmm8, %xmm2
-; CHECK-NEXT:    andps %xmm9, %xmm2
-; CHECK-NEXT:    andnps %xmm7, %xmm9
-; CHECK-NEXT:    cvttps2dq %xmm1, %xmm4
-; CHECK-NEXT:    cmpltps %xmm5, %xmm1
-; CHECK-NEXT:    movaps {{.*#+}} xmm7 = [5,6,7,8]
-; CHECK-NEXT:    movaps %xmm1, %xmm8
+; CHECK-NEXT:    cmpltps %xmm5, %xmm2
+; CHECK-NEXT:    movaps {{.*#+}} xmm7 = [9,10,11,12]
+; CHECK-NEXT:    movaps %xmm2, %xmm8
 ; CHECK-NEXT:    orps %xmm7, %xmm8
 ; CHECK-NEXT:    cvtdq2ps %xmm4, %xmm4
 ; CHECK-NEXT:    andps %xmm7, %xmm4
 ; CHECK-NEXT:    andps %xmm8, %xmm4
-; CHECK-NEXT:    andnps %xmm1, %xmm8
+; CHECK-NEXT:    andnps %xmm2, %xmm8
+; CHECK-NEXT:    cvttps2dq %xmm1, %xmm2
+; CHECK-NEXT:    cmpltps %xmm5, %xmm1
+; CHECK-NEXT:    movaps {{.*#+}} xmm7 = [5,6,7,8]
+; CHECK-NEXT:    movaps %xmm1, %xmm9
+; CHECK-NEXT:    orps %xmm7, %xmm9
+; CHECK-NEXT:    cvtdq2ps %xmm2, %xmm2
+; CHECK-NEXT:    andps %xmm7, %xmm2
+; CHECK-NEXT:    andps %xmm9, %xmm2
+; CHECK-NEXT:    andnps %xmm1, %xmm9
 ; CHECK-NEXT:    cvttps2dq %xmm0, %xmm1
 ; CHECK-NEXT:    cmpltps %xmm5, %xmm0
 ; CHECK-NEXT:    movaps {{.*#+}} xmm5 = [1,2,3,4]
@@ -141,14 +140,15 @@ define <16 x float> @foo(<16 x float> %x) {
 ; CHECK-NEXT:    movaps {{.*#+}} xmm0 = [1,1,1,1]
 ; CHECK-NEXT:    andps %xmm0, %xmm7
 ; CHECK-NEXT:    orps %xmm7, %xmm1
-; CHECK-NEXT:    andps %xmm0, %xmm8
-; CHECK-NEXT:    orps %xmm8, %xmm4
 ; CHECK-NEXT:    andps %xmm0, %xmm9
 ; CHECK-NEXT:    orps %xmm9, %xmm2
+; CHECK-NEXT:    andps %xmm0, %xmm8
+; CHECK-NEXT:    orps %xmm8, %xmm4
 ; CHECK-NEXT:    andps %xmm0, %xmm6
 ; CHECK-NEXT:    orps %xmm6, %xmm3
 ; CHECK-NEXT:    movaps %xmm1, %xmm0
-; CHECK-NEXT:    movaps %xmm4, %xmm1
+; CHECK-NEXT:    movaps %xmm2, %xmm1
+; CHECK-NEXT:    movaps %xmm4, %xmm2
 ; CHECK-NEXT:    retq
 bb:
   %v3 = icmp slt <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>, zeroinitializer
diff --git a/llvm/test/CodeGen/X86/oddshuffles.ll b/llvm/test/CodeGen/X86/oddshuffles.ll
index 5da18ee6ad7c4..b7a98d0e2a753 100644
--- a/llvm/test/CodeGen/X86/oddshuffles.ll
+++ b/llvm/test/CodeGen/X86/oddshuffles.ll
@@ -931,37 +931,37 @@ define void @interleave_24i16_out(ptr %p, ptr %q1, ptr %q2, ptr %q3) nounwind {
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,0,65535,65535,0,65535,65535]
 ; SSE2-NEXT:    movdqa %xmm4, %xmm5
 ; SSE2-NEXT:    pandn %xmm2, %xmm5
-; SSE2-NEXT:    movdqa %xmm3, %xmm6
-; SSE2-NEXT:    pand %xmm4, %xmm6
-; SSE2-NEXT:    por %xmm5, %xmm6
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm5 = xmm6[2,1,2,3,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,4,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,3,2,3]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,2,3,0,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,5,5,5]
-; SSE2-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,65535,65535,65535,0,0,0]
-; SSE2-NEXT:    pand %xmm6, %xmm5
+; SSE2-NEXT:    movdqa %xmm4, %xmm6
+; SSE2-NEXT:    pandn %xmm3, %xmm6
+; SSE2-NEXT:    pand %xmm4, %xmm3
+; SSE2-NEXT:    por %xmm5, %xmm3
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[2,1,2,3,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,2,3,0,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,5,5,5]
+; SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,65535,65535,65535,0,0,0]
+; SSE2-NEXT:    pand %xmm5, %xmm3
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm7 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,1,0,3]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,4,5,6]
-; SSE2-NEXT:    movdqa %xmm6, %xmm8
+; SSE2-NEXT:    movdqa %xmm5, %xmm8
 ; SSE2-NEXT:    pandn %xmm7, %xmm8
-; SSE2-NEXT:    por %xmm5, %xmm8
+; SSE2-NEXT:    por %xmm3, %xmm8
 ; SSE2-NEXT:    pand %xmm4, %xmm2
-; SSE2-NEXT:    pandn %xmm3, %xmm4
-; SSE2-NEXT:    por %xmm2, %xmm4
-; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[3,1,2,0]
+; SSE2-NEXT:    por %xmm6, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,0]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,0,3]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,0,3,4,5,6,7]
-; SSE2-NEXT:    pand %xmm6, %xmm2
+; SSE2-NEXT:    pand %xmm5, %xmm2
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
-; SSE2-NEXT:    pandn %xmm0, %xmm6
-; SSE2-NEXT:    por %xmm2, %xmm6
+; SSE2-NEXT:    pandn %xmm0, %xmm5
+; SSE2-NEXT:    por %xmm2, %xmm5
 ; SSE2-NEXT:    movups %xmm1, (%rsi)
 ; SSE2-NEXT:    movdqu %xmm8, (%rdx)
-; SSE2-NEXT:    movdqu %xmm6, (%rcx)
+; SSE2-NEXT:    movdqu %xmm5, (%rcx)
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: interleave_24i16_out:
@@ -1080,38 +1080,38 @@ define void @interleave_24i16_out_reverse(ptr %p, ptr %q1, ptr %q2, ptr %q3) nou
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,0,65535,65535,0,65535,65535]
 ; SSE2-NEXT:    movdqa %xmm4, %xmm5
 ; SSE2-NEXT:    pandn %xmm1, %xmm5
-; SSE2-NEXT:    movdqa %xmm3, %xmm6
-; SSE2-NEXT:    pand %xmm4, %xmm6
-; SSE2-NEXT:    por %xmm5, %xmm6
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm5 = xmm6[0,3,2,3,4,5,6,7]
-; SSE2-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,6,7]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,1,1]
-; SSE2-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[2,1,0,3,4,5,6,7]
-; SSE2-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,65535,65535,65535,0,0,0]
-; SSE2-NEXT:    pand %xmm6, %xmm5
+; SSE2-NEXT:    movdqa %xmm4, %xmm6
+; SSE2-NEXT:    pandn %xmm3, %xmm6
+; SSE2-NEXT:    pand %xmm4, %xmm3
+; SSE2-NEXT:    por %xmm5, %xmm3
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
+; SSE2-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,1,1]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[2,1,0,3,4,5,6,7]
+; SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,65535,65535,65535,0,0,0]
+; SSE2-NEXT:    pand %xmm5, %xmm3
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm7 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,1,2,0]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,4,7]
-; SSE2-NEXT:    movdqa %xmm6, %xmm8
+; SSE2-NEXT:    movdqa %xmm5, %xmm8
 ; SSE2-NEXT:    pandn %xmm7, %xmm8
-; SSE2-NEXT:    por %xmm5, %xmm8
+; SSE2-NEXT:    por %xmm3, %xmm8
 ; SSE2-NEXT:    pand %xmm4, %xmm1
-; SSE2-NEXT:    pandn %xmm3, %xmm4
-; SSE2-NEXT:    por %xmm1, %xmm4
-; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[3,1,2,0]
+; SSE2-NEXT:    por %xmm6, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,0,1,2,4,5,6,7]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
-; SSE2-NEXT:    pand %xmm6, %xmm1
+; SSE2-NEXT:    pand %xmm5, %xmm1
 ; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE2-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,5,4]
-; SSE2-NEXT:    pandn %xmm0, %xmm6
-; SSE2-NEXT:    por %xmm1, %xmm6
+; SSE2-NEXT:    pandn %xmm0, %xmm5
+; SSE2-NEXT:    por %xmm1, %xmm5
 ; SSE2-NEXT:    movups %xmm2, (%rsi)
 ; SSE2-NEXT:    movdqu %xmm8, (%rdx)
-; SSE2-NEXT:    movdqu %xmm6, (%rcx)
+; SSE2-NEXT:    movdqu %xmm5, (%rcx)
 ; SSE2-NEXT:    retq
 ;
 ; SSE42-LABEL: interleave_24i16_out_reverse:
diff --git a/llvm/test/CodeGen/X86/pmulh.ll b/llvm/test/CodeGen/X86/pmulh.ll
index c2a009f06b89d..de5d3b0d7bb7d 100644
--- a/llvm/test/CodeGen/X86/pmulh.ll
+++ b/llvm/test/CodeGen/X86/pmulh.ll
@@ -1389,16 +1389,16 @@ define <64 x i32> @zext_mulhuw_v64i16_lshr(<64 x i16> %a, <64 x i16> %b) {
 ;
 ; SSE41-LABEL: zext_mulhuw_v64i16_lshr:
 ; SSE41:       # %bb.0:
-; SSE41-NEXT:    movdqa %xmm0, %xmm8
+; SSE41-NEXT:    movdqa %xmm1, %xmm9
 ; SSE41-NEXT:    movq %rdi, %rax
-; SSE41-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm8
+; SSE41-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm0
 ; SSE41-NEXT:    pxor %xmm11, %xmm11
-; SSE41-NEXT:    movdqa %xmm8, %xmm0
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
+; SSE41-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE41-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm11[4],xmm0[5],xmm11[5],xmm0[6],xmm11[6],xmm0[7],xmm11[7]
-; SSE41-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE41-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm1
-; SSE41-NEXT:    movdqa %xmm1, %xmm9
-; SSE41-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm11[4],xmm9[5],xmm11[5],xmm9[6],xmm11[6],xmm9[7],xmm11[7]
+; SSE41-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm9
+; SSE41-NEXT:    movdqa %xmm9, %xmm8
+; SSE41-NEXT:    punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm11[4],xmm8[5],xmm11[5],xmm8[6],xmm11[6],xmm8[7],xmm11[7]
 ; SSE41-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm2
 ; SSE41-NEXT:    movdqa %xmm2, %xmm10
 ; SSE41-NEXT:    punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm11[4],xmm10[5],xmm11[5],xmm10[6],xmm11[6],xmm10[7],xmm11[7]
@@ -1415,32 +1415,31 @@ define <64 x i32> @zext_mulhuw_v64i16_lshr(<64 x i16> %a, <64 x i16> %b) {
 ; SSE41-NEXT:    movdqa %xmm6, %xmm15
 ; SSE41-NEXT:    punpckhwd {{.*#+}} xmm15 = xmm15[4],xmm11[4],xmm15[5],xmm11[5],xmm15[6],xmm11[6],xmm15[7],xmm11[7]
 ; SSE41-NEXT:    pmulhuw {{[0-9]+}}(%rsp), %xmm7
-; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm7[0],zero,xmm7[1],zero,xmm7[2],zero,xmm7[3],zero
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm1 = xmm7[0],zero,xmm7[1],zero,xmm7[2],zero,xmm7[3],zero
 ; SSE41-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm11[4],xmm7[5],xmm11[5],xmm7[6],xmm11[6],xmm7[7],xmm11[7]
 ; SSE41-NEXT:    movdqa %xmm7, 240(%rdi)
-; SSE41-NEXT:    movdqa %xmm0, 224(%rdi)
+; SSE41-NEXT:    movdqa %xmm1, 224(%rdi)
 ; SSE41-NEXT:    movdqa %xmm15, 208(%rdi)
-; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero
-; SSE41-NEXT:    movdqa %xmm0, 192(%rdi)
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm1 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero
+; SSE41-NEXT:    movdqa %xmm1, 192(%rdi)
 ; SSE41-NEXT:    movdqa %xmm14, 176(%rdi)
-; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero
-; SSE41-NEXT:    movdqa %xmm0, 160(%rdi)
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm1 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero
+; SSE41-NEXT:    movdqa %xmm1, 160(%rdi)
 ; SSE41-NEXT:    movdqa %xmm13, 144(%rdi)
-; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
-; SSE41-NEXT:    movdqa %xmm0, 128(%rdi)
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm1 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
+; SSE41-NEXT:    movdqa %xmm1, 128(%rdi)
 ; SSE41-NEXT:    movdqa %xmm12, 112(%rdi)
-; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
-; SSE41-NEXT:    movdqa %xmm0, 96(%rdi)
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
+; SSE41-NEXT:    movdqa %xmm1, 96(%rdi)
 ; SSE41-NEXT:    movdqa %xmm10, 80(%rdi)
-; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
-; SSE41-NEXT:    movdqa %xmm0, 64(%rdi)
-; SSE41-NEXT:    movdqa %xmm9, 48(%rdi)
-; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
-; SSE41-NEXT:    movdqa %xmm0, 32(%rdi)
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm1 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
+; SSE41-NEXT:    movdqa %xmm1, 64(%rdi)
+; SSE41-NEXT:    movdqa %xmm8, 48(%rdi)
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm1 = xmm9[0],zero,xmm9[1],zero,xmm9[2],zero,xmm9[3],zero
+; SSE41-NEXT:    movdqa %xmm1, 32(%rdi)
+; SSE41-NEXT:    movdqa %xmm0, 16(%rdi)
 ; SSE41-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE41-NEXT:    movaps %xmm0, 16(%rdi)
-; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm8[0],zero,xmm8[1],zero,xmm8[2],zero,xmm8[3],zero
-; SSE41-NEXT:    movdqa %xmm0, (%rdi)
+; SSE41-NEXT:    movaps %xmm0, (%rdi)
 ; SSE41-NEXT:    retq
 ;
 ; AVX2-LABEL: zext_mulhuw_v64i16_lshr:
@@ -1570,16 +1569,16 @@ define <64 x i32> @mulhsw_v64i16_lshr(<64 x i16> %a, <64 x i16> %b) {
 ;
 ; SSE41-LABEL: mulhsw_v64i16_lshr:
 ; SSE41:       # %bb.0:
-; SSE41-NEXT:    movdqa %xmm0, %xmm8
+; SSE41-NEXT:    movdqa %xmm1, %xmm9
 ; SSE41-NEXT:    movq %rdi, %rax
-; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm8
+; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm0
 ; SSE41-NEXT:    pxor %xmm11, %xmm11
-; SSE41-NEXT:    movdqa %xmm8, %xmm0
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
+; SSE41-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE41-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm11[4],xmm0[5],xmm11[5],xmm0[6],xmm11[6],xmm0[7],xmm11[7]
-; SSE41-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm1
-; SSE41-NEXT:    movdqa %xmm1, %xmm9
-; SSE41-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm11[4],xmm9[5],xmm11[5],xmm9[6],xmm11[6],xmm9[7],xmm11[7]
+; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm9
+; SSE41-NEXT:    movdqa %xmm9, %xmm8
+; SSE41-NEXT:    punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm11[4],xmm8[5],xmm11[5],xmm8[6],xmm11[6],xmm8[7],xmm11[7]
 ; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm2
 ; SSE41-NEXT:    movdqa %xmm2, %xmm10
 ; SSE41-NEXT:    punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm11[4],xmm10[5],xmm11[5],xmm10[6],xmm11[6],xmm10[7],xmm11[7]
@@ -1596,32 +1595,31 @@ define <64 x i32> @mulhsw_v64i16_lshr(<64 x i16> %a, <64 x i16> %b) {
 ; SSE41-NEXT:    movdqa %xmm6, %xmm15
 ; SSE41-NEXT:    punpckhwd {{.*#+}} xmm15 = xmm15[4],xmm11[4],xmm15[5],xmm11[5],xmm15[6],xmm11[6],xmm15[7],xmm11[7]
 ; SSE41-NEXT:    pmulhw {{[0-9]+}}(%rsp), %xmm7
-; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm7[0],zero,xmm7[1],zero,xmm7[2],zero,xmm7[3],zero
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm1 = xmm7[0],zero,xmm7[1],zero,xmm7[2],zero,xmm7[3],zero
 ; SSE41-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm11[4],xmm7[5],xmm11[5],xmm7[6],xmm11[6],xmm7[7],xmm11[7]
 ; SSE41-NEXT:    movdqa %xmm7, 240(%rdi)
-; SSE41-NEXT:    movdqa %xmm0, 224(%rdi)
+; SSE41-NEXT:    movdqa %xmm1, 224(%rdi)
 ; SSE41-NEXT:    movdqa %xmm15, 208(%rdi)
-; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero
-; SSE41-NEXT:    movdqa %xmm0, 192(%rdi)
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm1 = xmm6[0],zero,xmm6[1],zero,xmm6[2],zero,xmm6[3],zero
+; SSE41-NEXT:    movdqa %xmm1, 192(%rdi)
 ; SSE41-NEXT:    movdqa %xmm14, 176(%rdi)
-; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero
-; SSE41-NEXT:    movdqa %xmm0, 160(%rdi)
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm1 = xmm5[0],zero,xmm5[1],zero,xmm5[2],zero,xmm5[3],zero
+; SSE41-NEXT:    movdqa %xmm1, 160(%rdi)
 ; SSE41-NEXT:    movdqa %xmm13, 144(%rdi)
-; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
-; SSE41-NEXT:    movdqa %xmm0, 128(%rdi)
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm1 = xmm4[0],zero,xmm4[1],zero,xmm4[2],zero,xmm4[3],zero
+; SSE41-NEXT:    movdqa %xmm1, 128(%rdi)
 ; SSE41-NEXT:    movdqa %xmm12, 112(%rdi)
-; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
-; SSE41-NEXT:    movdqa %xmm0, 96(%rdi)
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero
+; SSE41-NEXT:    movdqa %xmm1, 96(%rdi)
 ; SSE41-NEXT:    movdqa %xmm10, 80(%rdi)
-; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
-; SSE41-NEXT:    movdqa %xmm0, 64(%rdi)
-; SSE41-NEXT:    movdqa %xmm9, 48(%rdi)
-; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
-; SSE41-NEXT:    movdqa %xmm0, 32(%rdi)
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm1 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero
+; SSE41-NEXT:    movdqa %xmm1, 64(%rdi)
+; SSE41-NEXT:    movdqa %xmm8, 48(%rdi)
+; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm1 = xmm9[0],zero,xmm9[1],zero,xmm9[2],zero,xmm9[3],zero
+; SSE41-NEXT:    movdqa %xmm1, 32(%rdi)
+; SSE41-NEXT:    movdqa %xmm0, 16(%rdi)
 ; SSE41-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE41-NEXT:    movaps %xmm0, 16(%rdi)
-; SSE41-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm8[0],zero,xmm8[1],zero,xmm8[2],zero,xmm8[3],zero
-; SSE41-NEXT:    movdqa %xmm0, (%rdi)
+; SSE41-NEXT:    movaps %xmm0, (%rdi)
 ; SSE41-NEXT:    retq
 ;
 ; AVX2-LABEL: mulhsw_v64i16_lshr:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
index 1436922f9dd11..871a7f1c61331 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-3.ll
@@ -439,37 +439,37 @@ define void @load_i16_stride3_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,0,65535,65535,0,65535,65535]
 ; SSE-NEXT:    movdqa %xmm4, %xmm5
 ; SSE-NEXT:    pandn %xmm2, %xmm5
-; SSE-NEXT:    movdqa %xmm3, %xmm6
-; SSE-NEXT:    pand %xmm4, %xmm6
-; SSE-NEXT:    por %xmm5, %xmm6
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm6[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,2,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,65535,65535,65535,0,0,0]
-; SSE-NEXT:    pand %xmm6, %xmm5
+; SSE-NEXT:    movdqa %xmm4, %xmm6
+; SSE-NEXT:    pandn %xmm3, %xmm6
+; SSE-NEXT:    pand %xmm4, %xmm3
+; SSE-NEXT:    por %xmm5, %xmm3
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,2,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,65535,65535,65535,0,0,0]
+; SSE-NEXT:    pand %xmm5, %xmm3
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,4,5,6]
-; SSE-NEXT:    movdqa %xmm6, %xmm8
+; SSE-NEXT:    movdqa %xmm5, %xmm8
 ; SSE-NEXT:    pandn %xmm7, %xmm8
-; SSE-NEXT:    por %xmm5, %xmm8
+; SSE-NEXT:    por %xmm3, %xmm8
 ; SSE-NEXT:    pand %xmm4, %xmm2
-; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    por %xmm2, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[3,1,2,0]
+; SSE-NEXT:    por %xmm6, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm6, %xmm2
+; SSE-NEXT:    pand %xmm5, %xmm2
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
-; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    por %xmm2, %xmm6
+; SSE-NEXT:    pandn %xmm0, %xmm5
+; SSE-NEXT:    por %xmm2, %xmm5
 ; SSE-NEXT:    movaps %xmm1, (%rsi)
 ; SSE-NEXT:    movdqa %xmm8, (%rdx)
-; SSE-NEXT:    movdqa %xmm6, (%rcx)
+; SSE-NEXT:    movdqa %xmm5, (%rcx)
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i16_stride3_vf8:
@@ -709,101 +709,101 @@ define void @load_i16_stride3_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movdqa (%rdi), %xmm7
 ; SSE-NEXT:    movdqa 16(%rdi), %xmm4
 ; SSE-NEXT:    movdqa 32(%rdi), %xmm3
-; SSE-NEXT:    movdqa 48(%rdi), %xmm2
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,0,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm6, %xmm8
-; SSE-NEXT:    pandn %xmm4, %xmm8
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm7, %xmm9
-; SSE-NEXT:    movdqa %xmm5, %xmm10
-; SSE-NEXT:    pandn %xmm7, %xmm10
-; SSE-NEXT:    pand %xmm6, %xmm7
-; SSE-NEXT:    por %xmm8, %xmm7
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,3,2,1,4,5,6,7]
+; SSE-NEXT:    movdqa 48(%rdi), %xmm6
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,0,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm5, %xmm2
+; SSE-NEXT:    pandn %xmm4, %xmm2
+; SSE-NEXT:    movdqa %xmm7, %xmm8
+; SSE-NEXT:    pand %xmm5, %xmm8
+; SSE-NEXT:    por %xmm2, %xmm8
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm4[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm3[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[3,0],xmm8[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm11[2,0]
-; SSE-NEXT:    movdqa %xmm2, %xmm8
-; SSE-NEXT:    pand %xmm6, %xmm8
-; SSE-NEXT:    pandn %xmm1, %xmm6
-; SSE-NEXT:    por %xmm8, %xmm6
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,3,2,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm3[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[3,0],xmm8[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm9[2,0]
+; SSE-NEXT:    movdqa %xmm6, %xmm8
+; SSE-NEXT:    pand %xmm5, %xmm8
+; SSE-NEXT:    pandn %xmm1, %xmm5
+; SSE-NEXT:    por %xmm8, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm1[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm0[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[3,0],xmm8[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm11[2,0]
-; SSE-NEXT:    movdqa %xmm5, %xmm8
-; SSE-NEXT:    pandn %xmm4, %xmm8
-; SSE-NEXT:    pand %xmm5, %xmm9
-; SSE-NEXT:    por %xmm8, %xmm9
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm9[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[1,2,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm8[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,65535,65535,65535,65535,0,0,0]
-; SSE-NEXT:    pand %xmm8, %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm0[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[3,0],xmm8[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm9[2,0]
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,65535,0,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm8, %xmm9
+; SSE-NEXT:    pandn %xmm4, %xmm9
+; SSE-NEXT:    movdqa %xmm8, %xmm10
+; SSE-NEXT:    pandn %xmm7, %xmm10
+; SSE-NEXT:    pand %xmm8, %xmm7
+; SSE-NEXT:    por %xmm9, %xmm7
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[1,2,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm7[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,65535,65535,65535,0,0,0]
+; SSE-NEXT:    pand %xmm7, %xmm11
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm3[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm9[0,1,2,3,4,4,5,6]
-; SSE-NEXT:    movdqa %xmm8, %xmm9
+; SSE-NEXT:    movdqa %xmm7, %xmm9
 ; SSE-NEXT:    pandn %xmm12, %xmm9
 ; SSE-NEXT:    por %xmm11, %xmm9
-; SSE-NEXT:    movdqa %xmm5, %xmm11
-; SSE-NEXT:    pandn %xmm1, %xmm11
-; SSE-NEXT:    movdqa %xmm2, %xmm12
-; SSE-NEXT:    pand %xmm5, %xmm12
-; SSE-NEXT:    por %xmm11, %xmm12
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm12[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[1,2,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm11[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pand %xmm8, %xmm12
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm0[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm11[0,1,2,3,4,4,5,6]
+; SSE-NEXT:    movdqa %xmm8, %xmm12
+; SSE-NEXT:    pandn %xmm1, %xmm12
 ; SSE-NEXT:    movdqa %xmm8, %xmm11
-; SSE-NEXT:    pandn %xmm13, %xmm11
-; SSE-NEXT:    por %xmm12, %xmm11
-; SSE-NEXT:    pand %xmm5, %xmm4
+; SSE-NEXT:    pandn %xmm6, %xmm11
+; SSE-NEXT:    pand %xmm8, %xmm6
+; SSE-NEXT:    por %xmm12, %xmm6
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[1,2,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm6[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    pand %xmm7, %xmm12
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm0[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm6[0,1,2,3,4,4,5,6]
+; SSE-NEXT:    movdqa %xmm7, %xmm6
+; SSE-NEXT:    pandn %xmm13, %xmm6
+; SSE-NEXT:    por %xmm12, %xmm6
+; SSE-NEXT:    pand %xmm8, %xmm4
 ; SSE-NEXT:    por %xmm10, %xmm4
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[3,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,1,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm8, %xmm4
+; SSE-NEXT:    pand %xmm7, %xmm4
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,2]
-; SSE-NEXT:    movdqa %xmm8, %xmm10
+; SSE-NEXT:    movdqa %xmm7, %xmm10
 ; SSE-NEXT:    pandn %xmm3, %xmm10
 ; SSE-NEXT:    por %xmm4, %xmm10
-; SSE-NEXT:    pand %xmm5, %xmm1
-; SSE-NEXT:    pandn %xmm2, %xmm5
-; SSE-NEXT:    por %xmm1, %xmm5
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[3,1,2,0]
+; SSE-NEXT:    pand %xmm8, %xmm1
+; SSE-NEXT:    por %xmm11, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm8, %xmm1
+; SSE-NEXT:    pand %xmm7, %xmm1
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
-; SSE-NEXT:    pandn %xmm0, %xmm8
-; SSE-NEXT:    por %xmm1, %xmm8
-; SSE-NEXT:    movaps %xmm6, 16(%rsi)
-; SSE-NEXT:    movaps %xmm7, (%rsi)
-; SSE-NEXT:    movdqa %xmm11, 16(%rdx)
+; SSE-NEXT:    pandn %xmm0, %xmm7
+; SSE-NEXT:    por %xmm1, %xmm7
+; SSE-NEXT:    movaps %xmm5, 16(%rsi)
+; SSE-NEXT:    movaps %xmm2, (%rsi)
+; SSE-NEXT:    movdqa %xmm6, 16(%rdx)
 ; SSE-NEXT:    movdqa %xmm9, (%rdx)
-; SSE-NEXT:    movdqa %xmm8, 16(%rcx)
+; SSE-NEXT:    movdqa %xmm7, 16(%rcx)
 ; SSE-NEXT:    movdqa %xmm10, (%rcx)
 ; SSE-NEXT:    retq
 ;
@@ -1235,239 +1235,228 @@ define void @load_i16_stride3_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2) nounwind {
 ; SSE-LABEL: load_i16_stride3_vf32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $40, %rsp
-; SSE-NEXT:    movdqa 96(%rdi), %xmm5
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 176(%rdi), %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 144(%rdi), %xmm13
-; SSE-NEXT:    movdqa 160(%rdi), %xmm9
-; SSE-NEXT:    movdqa 80(%rdi), %xmm11
-; SSE-NEXT:    movdqa (%rdi), %xmm15
-; SSE-NEXT:    movdqa 16(%rdi), %xmm10
-; SSE-NEXT:    movdqa 32(%rdi), %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 48(%rdi), %xmm0
-; SSE-NEXT:    movdqa 64(%rdi), %xmm12
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,0,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    pandn %xmm12, %xmm2
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    pand %xmm1, %xmm3
-; SSE-NEXT:    por %xmm2, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm2[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm12[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[0,1,2,1]
-; SSE-NEXT:    movdqa %xmm11, %xmm8
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,0],xmm2[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[2,0]
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    pandn %xmm10, %xmm2
-; SSE-NEXT:    movdqa %xmm15, %xmm3
-; SSE-NEXT:    pand %xmm1, %xmm3
-; SSE-NEXT:    por %xmm2, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm2[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm10[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    movdqa %xmm10, %xmm11
+; SSE-NEXT:    pushq %rax
+; SSE-NEXT:    movdqa 96(%rdi), %xmm11
+; SSE-NEXT:    movdqa 176(%rdi), %xmm10
 ; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,0],xmm2[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[2,0]
-; SSE-NEXT:    movaps %xmm4, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm9, %xmm7
-; SSE-NEXT:    pandn %xmm9, %xmm2
-; SSE-NEXT:    movdqa %xmm13, %xmm3
-; SSE-NEXT:    pand %xmm1, %xmm3
-; SSE-NEXT:    por %xmm2, %xmm3
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm9[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm6[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,0],xmm2[2,0]
-; SSE-NEXT:    movdqa 112(%rdi), %xmm6
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm4[2,0]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm5, %xmm2
-; SSE-NEXT:    pand %xmm1, %xmm2
-; SSE-NEXT:    pandn %xmm6, %xmm1
-; SSE-NEXT:    por %xmm2, %xmm1
+; SSE-NEXT:    movdqa 144(%rdi), %xmm1
+; SSE-NEXT:    movdqa 160(%rdi), %xmm12
+; SSE-NEXT:    movdqa 80(%rdi), %xmm13
+; SSE-NEXT:    movdqa (%rdi), %xmm0
+; SSE-NEXT:    movdqa 16(%rdi), %xmm4
+; SSE-NEXT:    movdqa 32(%rdi), %xmm8
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 48(%rdi), %xmm2
+; SSE-NEXT:    movdqa 64(%rdi), %xmm9
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,0,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm3, %xmm5
+; SSE-NEXT:    pandn %xmm9, %xmm5
+; SSE-NEXT:    movdqa %xmm2, %xmm7
+; SSE-NEXT:    pand %xmm3, %xmm7
+; SSE-NEXT:    por %xmm5, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm7[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm5[0,3,2,1,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm9[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm13[0,1,2,1]
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,0],xmm5[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,0]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm3, %xmm5
+; SSE-NEXT:    pandn %xmm4, %xmm5
+; SSE-NEXT:    movdqa %xmm0, %xmm7
+; SSE-NEXT:    pand %xmm3, %xmm7
+; SSE-NEXT:    por %xmm5, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm7[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm5[0,3,2,1,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm4[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm8[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,0],xmm5[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,0]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm3, %xmm5
+; SSE-NEXT:    pandn %xmm12, %xmm5
+; SSE-NEXT:    movdqa %xmm1, %xmm8
+; SSE-NEXT:    pand %xmm3, %xmm8
+; SSE-NEXT:    por %xmm5, %xmm8
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm12[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm10[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm7[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[3,0],xmm5[2,0]
+; SSE-NEXT:    movdqa 112(%rdi), %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm8[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,3,2,1,4,5,6,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm10[2,0]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm11, %xmm5
+; SSE-NEXT:    pand %xmm3, %xmm5
+; SSE-NEXT:    pandn %xmm7, %xmm3
+; SSE-NEXT:    por %xmm5, %xmm3
 ; SSE-NEXT:    movdqa 128(%rdi), %xmm5
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm6[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    movdqa %xmm6, %xmm9
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,0],xmm3[2,0]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm14, %xmm4
-; SSE-NEXT:    pandn %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm14, %xmm0
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm7[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,0],xmm8[2,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,1,4,5,6,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [65535,65535,0,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm15, %xmm3
+; SSE-NEXT:    pandn %xmm9, %xmm3
+; SSE-NEXT:    movdqa %xmm15, %xmm10
+; SSE-NEXT:    pandn %xmm2, %xmm10
+; SSE-NEXT:    pand %xmm15, %xmm2
+; SSE-NEXT:    por %xmm3, %xmm2
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm13[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,5,6]
+; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [65535,65535,65535,65535,65535,0,0,0]
+; SSE-NEXT:    movdqa %xmm13, %xmm14
+; SSE-NEXT:    pandn %xmm3, %xmm14
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    pand %xmm13, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm14
+; SSE-NEXT:    movdqa %xmm15, %xmm2
+; SSE-NEXT:    pandn %xmm4, %xmm2
+; SSE-NEXT:    movdqa %xmm15, %xmm8
+; SSE-NEXT:    pandn %xmm0, %xmm8
+; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    por %xmm2, %xmm0
+; SSE-NEXT:    pshuflw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,5,6]
+; SSE-NEXT:    movdqa %xmm13, %xmm5
+; SSE-NEXT:    pandn %xmm2, %xmm5
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    pand %xmm13, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm5
+; SSE-NEXT:    movdqa %xmm15, %xmm0
 ; SSE-NEXT:    pandn %xmm12, %xmm0
-; SSE-NEXT:    pand %xmm14, %xmm1
+; SSE-NEXT:    movdqa %xmm15, %xmm6
+; SSE-NEXT:    pandn %xmm1, %xmm6
+; SSE-NEXT:    pand %xmm15, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm8[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,4,4,5,6]
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,0,0,0]
-; SSE-NEXT:    movdqa %xmm0, %xmm10
-; SSE-NEXT:    pandn %xmm2, %xmm10
+; SSE-NEXT:    movdqa %xmm13, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm3
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm10
-; SSE-NEXT:    movdqa %xmm14, %xmm3
-; SSE-NEXT:    pandn %xmm15, %xmm3
+; SSE-NEXT:    pand %xmm13, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm3
 ; SSE-NEXT:    movdqa %xmm15, %xmm1
-; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    pandn %xmm7, %xmm1
+; SSE-NEXT:    movdqa %xmm15, %xmm2
 ; SSE-NEXT:    pandn %xmm11, %xmm2
-; SSE-NEXT:    pand %xmm14, %xmm1
-; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm11[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,5,6]
-; SSE-NEXT:    movdqa %xmm0, %xmm15
-; SSE-NEXT:    pandn %xmm2, %xmm15
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm15
-; SSE-NEXT:    movdqa %xmm14, %xmm6
-; SSE-NEXT:    pandn %xmm13, %xmm6
+; SSE-NEXT:    pand %xmm15, %xmm11
+; SSE-NEXT:    por %xmm1, %xmm11
+; SSE-NEXT:    pshuflw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,4,5,6]
 ; SSE-NEXT:    movdqa %xmm13, %xmm1
-; SSE-NEXT:    movdqa %xmm14, %xmm13
-; SSE-NEXT:    pandn %xmm7, %xmm13
-; SSE-NEXT:    pand %xmm14, %xmm1
-; SSE-NEXT:    por %xmm13, %xmm1
-; SSE-NEXT:    pshuflw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm13 = mem[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm13[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm13[0,1,2,3,4,4,5,6]
-; SSE-NEXT:    movdqa %xmm0, %xmm13
+; SSE-NEXT:    pandn %xmm0, %xmm1
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm11[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    pand %xmm13, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm1
+; SSE-NEXT:    pand %xmm15, %xmm9
+; SSE-NEXT:    por %xmm10, %xmm9
+; SSE-NEXT:    pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
+; SSE-NEXT:    movdqa %xmm13, %xmm10
+; SSE-NEXT:    pandn %xmm0, %xmm10
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[3,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm13, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm10
+; SSE-NEXT:    pand %xmm15, %xmm4
+; SSE-NEXT:    por %xmm8, %xmm4
+; SSE-NEXT:    pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
+; SSE-NEXT:    movdqa %xmm13, %xmm8
+; SSE-NEXT:    pandn %xmm0, %xmm8
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[3,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm13, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm8
+; SSE-NEXT:    pand %xmm15, %xmm12
+; SSE-NEXT:    por %xmm6, %xmm12
+; SSE-NEXT:    pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
+; SSE-NEXT:    movdqa %xmm13, %xmm4
+; SSE-NEXT:    pandn %xmm0, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[3,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm13, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm4
+; SSE-NEXT:    pand %xmm15, %xmm7
+; SSE-NEXT:    por %xmm2, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[3,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm13, %xmm0
+; SSE-NEXT:    pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
 ; SSE-NEXT:    pandn %xmm2, %xmm13
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm13
-; SSE-NEXT:    movdqa %xmm14, %xmm1
-; SSE-NEXT:    pandn %xmm9, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm8, %xmm2
-; SSE-NEXT:    pand %xmm14, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm5, %xmm9
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm5[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,5,6]
-; SSE-NEXT:    movdqa %xmm0, %xmm5
-; SSE-NEXT:    pandn %xmm1, %xmm5
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm5
-; SSE-NEXT:    pand %xmm14, %xmm12
-; SSE-NEXT:    por %xmm4, %xmm12
-; SSE-NEXT:    pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,2]
-; SSE-NEXT:    movdqa %xmm0, %xmm4
-; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pand %xmm14, %xmm1
-; SSE-NEXT:    por %xmm3, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm11[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,2]
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    pand %xmm14, %xmm7
-; SSE-NEXT:    por %xmm6, %xmm7
-; SSE-NEXT:    pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,2]
-; SSE-NEXT:    movdqa %xmm0, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pand %xmm14, %xmm1
-; SSE-NEXT:    pandn %xmm8, %xmm14
-; SSE-NEXT:    por %xmm1, %xmm14
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm9[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,0,2]
-; SSE-NEXT:    pandn %xmm6, %xmm0
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 32(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 48(%rsi)
-; SSE-NEXT:    movaps (%rsp), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, (%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 16(%rsi)
-; SSE-NEXT:    movdqa %xmm5, 32(%rdx)
-; SSE-NEXT:    movdqa %xmm13, 48(%rdx)
-; SSE-NEXT:    movdqa %xmm15, (%rdx)
-; SSE-NEXT:    movdqa %xmm10, 16(%rdx)
-; SSE-NEXT:    movdqa %xmm0, 32(%rcx)
-; SSE-NEXT:    movdqa %xmm2, 48(%rcx)
-; SSE-NEXT:    movdqa %xmm3, (%rcx)
-; SSE-NEXT:    movdqa %xmm4, 16(%rcx)
-; SSE-NEXT:    addq $40, %rsp
+; SSE-NEXT:    por %xmm0, %xmm13
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rsi)
+; SSE-NEXT:    movdqa %xmm1, 32(%rdx)
+; SSE-NEXT:    movdqa %xmm3, 48(%rdx)
+; SSE-NEXT:    movdqa %xmm5, (%rdx)
+; SSE-NEXT:    movdqa %xmm14, 16(%rdx)
+; SSE-NEXT:    movdqa %xmm13, 32(%rcx)
+; SSE-NEXT:    movdqa %xmm4, 48(%rcx)
+; SSE-NEXT:    movdqa %xmm8, (%rcx)
+; SSE-NEXT:    movdqa %xmm10, 16(%rcx)
+; SSE-NEXT:    popq %rax
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i16_stride3_vf32:
@@ -2250,21 +2239,21 @@ define void @load_i16_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 define void @load_i16_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2) nounwind {
 ; SSE-LABEL: load_i16_stride3_vf64:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $440, %rsp # imm = 0x1B8
+; SSE-NEXT:    subq $424, %rsp # imm = 0x1A8
 ; SSE-NEXT:    movdqa 192(%rdi), %xmm14
-; SSE-NEXT:    movdqa 272(%rdi), %xmm6
-; SSE-NEXT:    movdqa 240(%rdi), %xmm5
-; SSE-NEXT:    movdqa 256(%rdi), %xmm7
-; SSE-NEXT:    movdqa 80(%rdi), %xmm10
+; SSE-NEXT:    movdqa 272(%rdi), %xmm5
+; SSE-NEXT:    movdqa 240(%rdi), %xmm13
+; SSE-NEXT:    movdqa 256(%rdi), %xmm6
+; SSE-NEXT:    movdqa 80(%rdi), %xmm9
 ; SSE-NEXT:    movdqa (%rdi), %xmm15
-; SSE-NEXT:    movdqa 16(%rdi), %xmm9
-; SSE-NEXT:    movdqa 32(%rdi), %xmm8
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 16(%rdi), %xmm8
+; SSE-NEXT:    movdqa 32(%rdi), %xmm7
+; SSE-NEXT:    movdqa %xmm7, (%rsp) # 16-byte Spill
 ; SSE-NEXT:    movdqa 48(%rdi), %xmm0
-; SSE-NEXT:    movdqa 64(%rdi), %xmm11
+; SSE-NEXT:    movdqa 64(%rdi), %xmm10
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,0,65535,65535,0,65535,65535,0]
 ; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    pandn %xmm11, %xmm2
+; SSE-NEXT:    pandn %xmm10, %xmm2
 ; SSE-NEXT:    movdqa %xmm0, %xmm3
 ; SSE-NEXT:    pand %xmm1, %xmm3
 ; SSE-NEXT:    por %xmm2, %xmm3
@@ -2272,42 +2261,45 @@ define void @load_i16_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm2[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm11[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm10[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm10[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    movdqa %xmm10, %xmm12
 ; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[0,1,2,1]
+; SSE-NEXT:    movdqa %xmm9, %xmm11
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,5]
 ; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,0],xmm2[2,0]
 ; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[2,0]
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    pandn %xmm7, %xmm2
-; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    pandn %xmm6, %xmm2
+; SSE-NEXT:    movdqa %xmm13, %xmm3
 ; SSE-NEXT:    pand %xmm1, %xmm3
 ; SSE-NEXT:    por %xmm2, %xmm3
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,1,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm2[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm7[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm6[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm6[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm5[0,1,2,1]
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,5]
 ; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,0],xmm2[2,0]
 ; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[2,0]
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    pandn %xmm9, %xmm2
+; SSE-NEXT:    movdqa %xmm8, %xmm10
+; SSE-NEXT:    pandn %xmm8, %xmm2
 ; SSE-NEXT:    movdqa %xmm15, %xmm3
 ; SSE-NEXT:    pand %xmm1, %xmm3
 ; SSE-NEXT:    por %xmm2, %xmm3
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm9[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm8[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm8[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm7[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,6,5]
 ; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,0],xmm2[2,0]
-; SSE-NEXT:    movdqa 208(%rdi), %xmm8
+; SSE-NEXT:    movdqa 208(%rdi), %xmm7
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,1,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
@@ -2315,8 +2307,7 @@ define void @load_i16_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm4[2,0]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    pandn %xmm8, %xmm2
-; SSE-NEXT:    movdqa %xmm8, (%rsp) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm7, %xmm2
 ; SSE-NEXT:    movdqa %xmm14, %xmm3
 ; SSE-NEXT:    pand %xmm1, %xmm3
 ; SSE-NEXT:    por %xmm2, %xmm3
@@ -2324,7 +2315,8 @@ define void @load_i16_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm8[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm7[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,0],xmm4[2,0]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,1,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
@@ -2335,7 +2327,7 @@ define void @load_i16_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movdqa 160(%rdi), %xmm3
 ; SSE-NEXT:    movdqa %xmm1, %xmm2
 ; SSE-NEXT:    pandn %xmm3, %xmm2
-; SSE-NEXT:    movdqa %xmm3, %xmm4
+; SSE-NEXT:    movdqa %xmm3, %xmm8
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 144(%rdi), %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -2345,7 +2337,7 @@ define void @load_i16_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm8[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,0],xmm4[2,0]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,1,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
@@ -2379,8 +2371,8 @@ define void @load_i16_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pandn %xmm3, %xmm2
 ; SSE-NEXT:    movdqa %xmm3, %xmm4
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 96(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 96(%rdi), %xmm8
+; SSE-NEXT:    movdqa %xmm8, %xmm3
 ; SSE-NEXT:    pand %xmm1, %xmm3
 ; SSE-NEXT:    por %xmm2, %xmm3
 ; SSE-NEXT:    movdqa 128(%rdi), %xmm2
@@ -2414,320 +2406,316 @@ define void @load_i16_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,0]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,0,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm9, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    pandn %xmm11, %xmm0
-; SSE-NEXT:    pand %xmm12, %xmm2
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    pandn %xmm12, %xmm0
+; SSE-NEXT:    pand %xmm9, %xmm2
 ; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm10[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm11[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm0[0,1,2,3,4,4,5,6]
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [65535,65535,65535,65535,65535,0,0,0]
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,0,0,0]
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    pandn %xmm3, %xmm1
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pand %xmm13, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    pandn %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm3
-; SSE-NEXT:    pandn %xmm7, %xmm3
-; SSE-NEXT:    pand %xmm12, %xmm5
-; SSE-NEXT:    por %xmm3, %xmm5
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm6[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    pandn %xmm13, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    pandn %xmm6, %xmm3
+; SSE-NEXT:    pand %xmm9, %xmm13
+; SSE-NEXT:    por %xmm3, %xmm13
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm5[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,5,6]
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm5[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    pandn %xmm3, %xmm1
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm13[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pand %xmm13, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm8
-; SSE-NEXT:    pandn %xmm15, %xmm8
-; SSE-NEXT:    movdqa %xmm12, %xmm3
-; SSE-NEXT:    pandn %xmm9, %xmm3
-; SSE-NEXT:    pand %xmm12, %xmm15
+; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm12
+; SSE-NEXT:    pandn %xmm15, %xmm12
+; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    pandn %xmm10, %xmm3
+; SSE-NEXT:    pand %xmm9, %xmm15
 ; SSE-NEXT:    por %xmm3, %xmm15
-; SSE-NEXT:    pshuflw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    pshuflw $236, (%rsp), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm3 = mem[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,5,6]
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    pandn %xmm3, %xmm1
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm15[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pand %xmm13, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    pandn %xmm14, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm3
-; SSE-NEXT:    pandn (%rsp), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    pand %xmm12, %xmm14
+; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    pandn %xmm14, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    pandn %xmm7, %xmm3
+; SSE-NEXT:    pand %xmm9, %xmm14
 ; SSE-NEXT:    por %xmm3, %xmm14
 ; SSE-NEXT:    pshuflw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm3 = mem[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,5,6]
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    pandn %xmm3, %xmm1
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm14[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pand %xmm13, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm5
+; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm5
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    pandn %xmm2, %xmm5
-; SSE-NEXT:    movdqa %xmm12, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm14, %xmm3
-; SSE-NEXT:    pand %xmm12, %xmm2
+; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    pand %xmm9, %xmm2
 ; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm11[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm14[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,5,6]
-; SSE-NEXT:    movdqa %xmm13, %xmm15
+; SSE-NEXT:    movdqa %xmm0, %xmm15
 ; SSE-NEXT:    pandn %xmm3, %xmm15
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pand %xmm13, %xmm2
+; SSE-NEXT:    pand %xmm0, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm15
-; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    movdqa %xmm9, %xmm1
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm3
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    pand %xmm12, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm2
 ; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    pshuflw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm3 = mem[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,5,6]
-; SSE-NEXT:    movdqa %xmm13, %xmm10
-; SSE-NEXT:    pandn %xmm3, %xmm10
+; SSE-NEXT:    movdqa %xmm0, %xmm11
+; SSE-NEXT:    pandn %xmm3, %xmm11
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pand %xmm13, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm10
-; SSE-NEXT:    movdqa %xmm12, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm3, %xmm2
-; SSE-NEXT:    movdqa %xmm12, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm9, %xmm4
-; SSE-NEXT:    pand %xmm12, %xmm3
-; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm7[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm4[0,1,2,3,4,4,5,6]
-; SSE-NEXT:    movdqa %xmm13, %xmm6
+; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm11
+; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm13, %xmm3
+; SSE-NEXT:    movdqa %xmm9, %xmm4
+; SSE-NEXT:    pandn %xmm8, %xmm4
+; SSE-NEXT:    pand %xmm9, %xmm8
+; SSE-NEXT:    por %xmm3, %xmm8
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm10[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,5,6]
+; SSE-NEXT:    movdqa %xmm0, %xmm7
+; SSE-NEXT:    pandn %xmm3, %xmm7
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm8[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,2,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    pand %xmm0, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm7
+; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pand %xmm9, %xmm8
+; SSE-NEXT:    por %xmm3, %xmm8
+; SSE-NEXT:    pshuflw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,4,4,5,6]
+; SSE-NEXT:    movdqa %xmm0, %xmm6
 ; SSE-NEXT:    pandn %xmm1, %xmm6
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm3[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm8[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pand %xmm13, %xmm1
+; SSE-NEXT:    pand %xmm0, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm6
-; SSE-NEXT:    movdqa %xmm12, %xmm1
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pshuflw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,5,6]
-; SSE-NEXT:    movdqa %xmm13, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pand %xmm13, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
-; SSE-NEXT:    movdqa %xmm13, %xmm4
-; SSE-NEXT:    pandn %xmm0, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm13, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    por %xmm8, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
-; SSE-NEXT:    movdqa %xmm13, %xmm8
-; SSE-NEXT:    pandn %xmm0, %xmm8
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm13, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm8
-; SSE-NEXT:    pand %xmm12, %xmm14
-; SSE-NEXT:    por %xmm5, %xmm14
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm11[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
-; SSE-NEXT:    movdqa %xmm13, %xmm5
-; SSE-NEXT:    pandn %xmm0, %xmm5
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm13, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm5
-; SSE-NEXT:    pand %xmm12, %xmm9
-; SSE-NEXT:    por %xmm2, %xmm9
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm7[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
-; SSE-NEXT:    movdqa %xmm13, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm13, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm2
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pand %xmm12, %xmm1
+; SSE-NEXT:    pand %xmm9, %xmm1
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,2]
+; SSE-NEXT:    movdqa %xmm0, %xmm8
+; SSE-NEXT:    pandn %xmm1, %xmm8
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[3,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm8
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pand %xmm9, %xmm2
+; SSE-NEXT:    por %xmm12, %xmm2
+; SSE-NEXT:    pshufhw $236, (%rsp), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm1[0,1,0,2]
+; SSE-NEXT:    movdqa %xmm0, %xmm3
+; SSE-NEXT:    pandn %xmm12, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm2[3,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm12
+; SSE-NEXT:    por %xmm12, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pand %xmm9, %xmm1
+; SSE-NEXT:    por %xmm5, %xmm1
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm14[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm5[0,1,0,2]
+; SSE-NEXT:    movdqa %xmm0, %xmm5
+; SSE-NEXT:    pandn %xmm12, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm1[3,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm12
+; SSE-NEXT:    por %xmm12, %xmm5
+; SSE-NEXT:    pand %xmm9, %xmm13
+; SSE-NEXT:    por %xmm4, %xmm13
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm10[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm2[0,1,0,2]
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm12, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm13[3,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm12
+; SSE-NEXT:    por %xmm12, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pand %xmm9, %xmm1
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
-; SSE-NEXT:    movdqa %xmm13, %xmm14
-; SSE-NEXT:    pandn %xmm0, %xmm14
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm13, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm14
-; SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
-; SSE-NEXT:    movdqa %xmm13, %xmm11
-; SSE-NEXT:    pandn %xmm0, %xmm11
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm13, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm11
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
-; SSE-NEXT:    movdqa %xmm13, %xmm9
-; SSE-NEXT:    pandn %xmm0, %xmm9
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm13, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm9
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; SSE-NEXT:    por %xmm0, %xmm12
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm13, %xmm0
 ; SSE-NEXT:    pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm12 = mem[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[0,1,0,2]
-; SSE-NEXT:    pandn %xmm12, %xmm13
-; SSE-NEXT:    por %xmm0, %xmm13
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 96(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 32(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 112(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 48(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 64(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, (%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 80(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 16(%rsi)
-; SSE-NEXT:    movdqa %xmm3, 96(%rdx)
-; SSE-NEXT:    movdqa %xmm6, 32(%rdx)
-; SSE-NEXT:    movdqa %xmm10, 112(%rdx)
+; SSE-NEXT:    movdqa %xmm0, %xmm14
+; SSE-NEXT:    pandn %xmm12, %xmm14
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm1[3,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm12
+; SSE-NEXT:    por %xmm12, %xmm14
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pand %xmm9, %xmm1
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = mem[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm12[0,1,0,2]
+; SSE-NEXT:    movdqa %xmm0, %xmm12
+; SSE-NEXT:    pandn %xmm13, %xmm12
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm1[3,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm13[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm13[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm13
+; SSE-NEXT:    por %xmm13, %xmm12
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pand %xmm9, %xmm1
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm10[0,1,0,2]
+; SSE-NEXT:    movdqa %xmm0, %xmm10
+; SSE-NEXT:    pandn %xmm13, %xmm10
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm1[3,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm13[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm13[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm13
+; SSE-NEXT:    por %xmm13, %xmm10
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pand %xmm9, %xmm1
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    por %xmm1, %xmm9
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[3,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm9
+; SSE-NEXT:    pshufhw $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm13 = mem[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm13[0,1,0,2]
+; SSE-NEXT:    pandn %xmm13, %xmm0
+; SSE-NEXT:    por %xmm9, %xmm0
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 96(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 112(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm9, 80(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm9, 16(%rsi)
+; SSE-NEXT:    movdqa %xmm6, 96(%rdx)
+; SSE-NEXT:    movdqa %xmm7, 32(%rdx)
+; SSE-NEXT:    movdqa %xmm11, 112(%rdx)
 ; SSE-NEXT:    movdqa %xmm15, 48(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 64(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, (%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 80(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 16(%rdx)
-; SSE-NEXT:    movdqa %xmm13, 96(%rcx)
-; SSE-NEXT:    movdqa %xmm9, 112(%rcx)
-; SSE-NEXT:    movdqa %xmm11, 64(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 64(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 80(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%rdx)
+; SSE-NEXT:    movdqa %xmm0, 96(%rcx)
+; SSE-NEXT:    movdqa %xmm10, 112(%rcx)
+; SSE-NEXT:    movdqa %xmm12, 64(%rcx)
 ; SSE-NEXT:    movdqa %xmm14, 80(%rcx)
 ; SSE-NEXT:    movdqa %xmm2, 32(%rcx)
 ; SSE-NEXT:    movdqa %xmm5, 48(%rcx)
-; SSE-NEXT:    movdqa %xmm8, (%rcx)
-; SSE-NEXT:    movdqa %xmm4, 16(%rcx)
-; SSE-NEXT:    addq $440, %rsp # imm = 0x1B8
+; SSE-NEXT:    movdqa %xmm3, (%rcx)
+; SSE-NEXT:    movdqa %xmm8, 16(%rcx)
+; SSE-NEXT:    addq $424, %rsp # imm = 0x1A8
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i16_stride3_vf64:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
index 8e55cb48cf7a2..131409d7ea656 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-5.ll
@@ -715,34 +715,34 @@ define void @load_i16_stride5_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-LABEL: load_i16_stride5_vf8:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    movdqa 64(%rdi), %xmm6
-; SSE-NEXT:    movdqa (%rdi), %xmm4
-; SSE-NEXT:    movdqa 16(%rdi), %xmm3
+; SSE-NEXT:    movdqa (%rdi), %xmm5
+; SSE-NEXT:    movdqa 16(%rdi), %xmm4
 ; SSE-NEXT:    movdqa 32(%rdi), %xmm0
-; SSE-NEXT:    movdqa 48(%rdi), %xmm5
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,65535,0,65535,65535,65535]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[0,1,0,3]
-; SSE-NEXT:    pand %xmm1, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm4[0,2,2,3]
+; SSE-NEXT:    movdqa 48(%rdi), %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,65535,0,65535,65535,65535]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,1,0,3]
+; SSE-NEXT:    pand %xmm2, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    por %xmm3, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm5[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm2[0],xmm7[1],xmm2[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm1[2,3]
-; SSE-NEXT:    movaps {{.*#+}} xmm1 = [65535,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT:    andps %xmm1, %xmm7
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm3[0],xmm7[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm2[2,3]
+; SSE-NEXT:    movaps {{.*#+}} xmm2 = [65535,65535,65535,65535,65535,65535,65535,0]
+; SSE-NEXT:    andps %xmm2, %xmm7
 ; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm6[0,1,0,1]
-; SSE-NEXT:    movaps %xmm1, %xmm2
-; SSE-NEXT:    pandn %xmm8, %xmm2
-; SSE-NEXT:    por %xmm7, %xmm2
-; SSE-NEXT:    movdqa %xmm3, %xmm7
+; SSE-NEXT:    movaps %xmm2, %xmm3
+; SSE-NEXT:    pandn %xmm8, %xmm3
+; SSE-NEXT:    por %xmm7, %xmm3
+; SSE-NEXT:    movdqa %xmm4, %xmm7
 ; SSE-NEXT:    psrlq $48, %xmm7
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm4[0,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm5[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[1,2,2,3,4,5,6,7]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [0,0,0,65535,65535,65535,65535,65535]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm5[1,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm1[1,3,2,3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm0[0,2,2,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm10[0,1,2,3,7,5,6,7]
@@ -752,65 +752,64 @@ define void @load_i16_stride5_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pand %xmm7, %xmm9
 ; SSE-NEXT:    pandn %xmm8, %xmm7
 ; SSE-NEXT:    por %xmm9, %xmm7
-; SSE-NEXT:    pand %xmm1, %xmm7
+; SSE-NEXT:    pand %xmm2, %xmm7
 ; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm6[0,1,2,0]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm6[0,1,0,3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm6[0,1,1,3]
 ; SSE-NEXT:    psllq $48, %xmm6
-; SSE-NEXT:    pandn %xmm6, %xmm1
-; SSE-NEXT:    por %xmm7, %xmm1
-; SSE-NEXT:    movdqa %xmm5, %xmm7
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,2,2,3]
-; SSE-NEXT:    movdqa %xmm5, %xmm12
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[1,0],xmm0[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[2,0],xmm0[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,0,1,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[0,1,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm3[2],xmm5[3],xmm3[3]
-; SSE-NEXT:    pand %xmm13, %xmm5
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm4[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm3[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm11[0],xmm14[1],xmm11[1]
-; SSE-NEXT:    movdqa %xmm13, %xmm15
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm14[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm13, %xmm11
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm3[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm4[0,2]
-; SSE-NEXT:    movdqa %xmm13, %xmm4
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[2,1,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm13, %xmm3
-; SSE-NEXT:    pandn %xmm12, %xmm13
-; SSE-NEXT:    por %xmm13, %xmm5
-; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pandn %xmm6, %xmm2
+; SSE-NEXT:    por %xmm7, %xmm2
+; SSE-NEXT:    movdqa %xmm1, %xmm11
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[1,0],xmm0[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[2,0],xmm0[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,0,1,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,1,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm4[2],xmm6[3],xmm4[3]
+; SSE-NEXT:    pand %xmm12, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm5[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm4[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm7[0],xmm13[1],xmm7[1]
+; SSE-NEXT:    movdqa %xmm12, %xmm14
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm13[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[1,0,3,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm12, %xmm7
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm4[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm5[0,2]
+; SSE-NEXT:    movdqa %xmm12, %xmm5
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[2,1,3,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm12, %xmm4
+; SSE-NEXT:    pandn %xmm11, %xmm12
+; SSE-NEXT:    por %xmm12, %xmm6
+; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[3,1],xmm12[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm8[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[3,1],xmm11[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm8[2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm0[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pandn %xmm8, %xmm15
-; SSE-NEXT:    por %xmm15, %xmm11
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm0[3,0]
-; SSE-NEXT:    pandn %xmm0, %xmm4
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm7[0,2]
+; SSE-NEXT:    pandn %xmm8, %xmm14
+; SSE-NEXT:    por %xmm14, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm1[0,2,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[3,0]
+; SSE-NEXT:    pandn %xmm0, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0,2]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,4,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm9[0,1,2,3,4,5,5,6]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,1],xmm0[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,1],xmm7[2,0]
-; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm6[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm10[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,1],xmm0[1,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,0]
-; SSE-NEXT:    movdqa %xmm2, (%rsi)
-; SSE-NEXT:    movdqa %xmm1, (%rdx)
-; SSE-NEXT:    movaps %xmm5, (%rcx)
-; SSE-NEXT:    movaps %xmm11, (%r8)
-; SSE-NEXT:    movaps %xmm3, (%r9)
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm9[0,1,2,3,4,5,5,6]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm1[2,0]
+; SSE-NEXT:    por %xmm5, %xmm4
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm8[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm10[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[1,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm1[2,0]
+; SSE-NEXT:    movdqa %xmm3, (%rsi)
+; SSE-NEXT:    movdqa %xmm2, (%rdx)
+; SSE-NEXT:    movaps %xmm6, (%rcx)
+; SSE-NEXT:    movaps %xmm7, (%r8)
+; SSE-NEXT:    movaps %xmm4, (%r9)
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i16_stride5_vf8:
@@ -1288,222 +1287,221 @@ define void @load_i16_stride5_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i16_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4) nounwind {
 ; SSE-LABEL: load_i16_stride5_vf16:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    movdqa 144(%rdi), %xmm14
-; SSE-NEXT:    movdqa 80(%rdi), %xmm8
-; SSE-NEXT:    movdqa 96(%rdi), %xmm7
-; SSE-NEXT:    movdqa 128(%rdi), %xmm15
-; SSE-NEXT:    movdqa 112(%rdi), %xmm12
-; SSE-NEXT:    movdqa 64(%rdi), %xmm10
-; SSE-NEXT:    movdqa (%rdi), %xmm11
-; SSE-NEXT:    movdqa 16(%rdi), %xmm9
-; SSE-NEXT:    movdqa 32(%rdi), %xmm13
-; SSE-NEXT:    movdqa 48(%rdi), %xmm3
+; SSE-NEXT:    movdqa 144(%rdi), %xmm11
+; SSE-NEXT:    movdqa 80(%rdi), %xmm6
+; SSE-NEXT:    movdqa 96(%rdi), %xmm8
+; SSE-NEXT:    movdqa 128(%rdi), %xmm5
+; SSE-NEXT:    movdqa 112(%rdi), %xmm4
+; SSE-NEXT:    movdqa 64(%rdi), %xmm9
+; SSE-NEXT:    movdqa (%rdi), %xmm7
+; SSE-NEXT:    movdqa 16(%rdi), %xmm10
+; SSE-NEXT:    movdqa 32(%rdi), %xmm12
+; SSE-NEXT:    movdqa 48(%rdi), %xmm13
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,0,65535,65535,65535]
 ; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    pandn %xmm13, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,1,0,3]
-; SSE-NEXT:    movdqa %xmm3, %xmm5
+; SSE-NEXT:    pandn %xmm12, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm13[0,1,0,3]
 ; SSE-NEXT:    pand %xmm0, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[3,1,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm3[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,3]
-; SSE-NEXT:    movaps {{.*#+}} xmm6 = [65535,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT:    andps %xmm6, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[0,1,0,1]
-; SSE-NEXT:    movaps %xmm6, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    por %xmm4, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[0,1,0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[0,2,2,3]
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,3]
+; SSE-NEXT:    movaps {{.*#+}} xmm2 = [65535,65535,65535,65535,65535,65535,65535,0]
+; SSE-NEXT:    andps %xmm2, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,1,0,1]
+; SSE-NEXT:    movdqa %xmm9, %xmm14
+; SSE-NEXT:    movaps %xmm2, %xmm9
+; SSE-NEXT:    pandn %xmm1, %xmm9
+; SSE-NEXT:    por %xmm3, %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[0,1,0,3]
 ; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pandn %xmm12, %xmm0
+; SSE-NEXT:    pandn %xmm4, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[3,1,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3]
-; SSE-NEXT:    andps %xmm6, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[0,1,0,1]
-; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm6[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[2,3]
+; SSE-NEXT:    andps %xmm2, %xmm3
+; SSE-NEXT:    movdqa %xmm11, %xmm9
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[0,1,0,1]
+; SSE-NEXT:    movaps %xmm2, %xmm1
 ; SSE-NEXT:    andnps %xmm0, %xmm1
-; SSE-NEXT:    orps %xmm2, %xmm1
+; SSE-NEXT:    orps %xmm3, %xmm1
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm10, %xmm0
 ; SSE-NEXT:    psrlq $48, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[0,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,2,3,4,5,6,7]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [0,0,0,65535,65535,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm0, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm5, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,3,2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm13[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm4[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    movdqa %xmm0, %xmm3
+; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[1,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm12[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm1[0],xmm11[1],xmm1[1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm11[0,1,2,3,7,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
 ; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm10, %xmm5
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm2
-; SSE-NEXT:    psllq $48, %xmm2
-; SSE-NEXT:    movaps %xmm6, %xmm4
-; SSE-NEXT:    andnps %xmm2, %xmm4
-; SSE-NEXT:    pand %xmm6, %xmm1
-; SSE-NEXT:    orps %xmm1, %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    por %xmm3, %xmm1
+; SSE-NEXT:    movdqa %xmm14, %xmm3
+; SSE-NEXT:    movdqa %xmm14, %xmm15
+; SSE-NEXT:    psllq $48, %xmm3
+; SSE-NEXT:    movaps %xmm2, %xmm7
+; SSE-NEXT:    andnps %xmm3, %xmm7
+; SSE-NEXT:    pand %xmm2, %xmm1
+; SSE-NEXT:    orps %xmm1, %xmm7
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm8, %xmm1
 ; SSE-NEXT:    psrlq $48, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[1,3,2,3]
-; SSE-NEXT:    movdqa %xmm15, %xmm10
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm12[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm4[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm6[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,2,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,3,2,3]
+; SSE-NEXT:    movdqa %xmm5, %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm4[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm14[0,1,2,3,7,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
 ; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pandn %xmm2, %xmm0
+; SSE-NEXT:    pandn %xmm3, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pand %xmm6, %xmm0
-; SSE-NEXT:    movdqa %xmm14, %xmm4
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm1
+; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    psllq $48, %xmm1
-; SSE-NEXT:    pandn %xmm1, %xmm6
-; SSE-NEXT:    por %xmm0, %xmm6
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    movdqa %xmm3, %xmm14
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm13[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm13[2,3]
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm12[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm12[2,3]
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0,1,3]
-; SSE-NEXT:    movaps {{.*#+}} xmm3 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    movaps %xmm3, %xmm1
-; SSE-NEXT:    andnps %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm11[0,1,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm2[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm9[2],xmm15[3],xmm9[3]
-; SSE-NEXT:    pand %xmm3, %xmm15
-; SSE-NEXT:    por %xmm1, %xmm15
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[0,1,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm1[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm14 = xmm14[2],xmm10[2],xmm14[3],xmm10[3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT:    pand %xmm1, %xmm14
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm14
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm1[2,0]
-; SSE-NEXT:    movdqa %xmm10, %xmm5
-; SSE-NEXT:    movdqa %xmm10, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm12[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm12[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0,1,3]
-; SSE-NEXT:    movaps %xmm3, %xmm2
-; SSE-NEXT:    andnps %xmm1, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[0,1,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm7[2],xmm0[3],xmm7[3]
-; SSE-NEXT:    pand %xmm3, %xmm0
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,1],xmm1[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,0]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm13[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm9[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm4[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm3, %xmm1
-; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm14, %xmm4
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm13[3,0]
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    pandn %xmm13, %xmm2
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,1],xmm4[0,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm13[0,1,2,3,7,4,6,7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm10[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,5,5,6]
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[3,1],xmm4[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,0]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm12[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm3, %xmm14
-; SSE-NEXT:    pandn %xmm4, %xmm14
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm8[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm7[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm4[0],xmm13[1],xmm4[1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm13[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm4[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm3, %xmm13
-; SSE-NEXT:    por %xmm14, %xmm13
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[2,0],xmm9[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,1],xmm11[0,2]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,0],xmm7[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm8[0,2]
-; SSE-NEXT:    movdqa %xmm5, %xmm11
-; SSE-NEXT:    movdqa %xmm5, %xmm4
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm12[3,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm9[0,1,2,3,4,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[2,1,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm3, %xmm8
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm15[0,1,2,0]
+; SSE-NEXT:    movdqa %xmm15, %xmm5
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,1],xmm0[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm3[2,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[0,1,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm0[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm8[2],xmm15[3],xmm8[3]
+; SSE-NEXT:    pand %xmm1, %xmm15
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm4[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm4[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0,1,3]
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm15
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,1],xmm0[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm3[2,0]
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm12[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm10[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm9[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    por %xmm3, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[0,2,2,3]
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[2,0],xmm12[3,0]
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm12, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,1],xmm13[0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm12[0,1,2,3,7,4,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm5[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,4,5,5,6]
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[3,1],xmm9[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm12[2,0]
+; SSE-NEXT:    movdqa %xmm1, %xmm9
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm4[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pandn %xmm12, %xmm9
+; SSE-NEXT:    movdqa %xmm6, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm6[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm8[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm13[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[1,0,3,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm1, %xmm12
+; SSE-NEXT:    por %xmm9, %xmm12
+; SSE-NEXT:    movdqa %xmm7, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0],xmm10[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,1],xmm6[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm8[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm10[0,1,2,3,4,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[2,1,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm3, %xmm7
-; SSE-NEXT:    pandn %xmm12, %xmm3
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,1],xmm4[0,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm12[0,1,2,3,7,4,6,7]
+; SSE-NEXT:    pand %xmm1, %xmm7
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm8[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[2,1,3,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm1, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm11[0,2,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[2,0],xmm4[3,0]
+; SSE-NEXT:    pandn %xmm4, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm11[0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,4,6,7]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm9[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,5,6]
 ; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,1],xmm4[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,1],xmm5[2,0]
-; SSE-NEXT:    por %xmm2, %xmm8
-; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[0,2,2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm10[0,1,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,1],xmm2[1,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm4[2,0]
-; SSE-NEXT:    por %xmm7, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm11[0,2,2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm9[0,1,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,1],xmm2[1,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, 16(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, (%rsi)
-; SSE-NEXT:    movdqa %xmm6, 16(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, (%rdx)
-; SSE-NEXT:    movaps %xmm0, 16(%rcx)
-; SSE-NEXT:    movaps %xmm15, (%rcx)
-; SSE-NEXT:    movaps %xmm13, 16(%r8)
-; SSE-NEXT:    movaps %xmm1, (%r8)
-; SSE-NEXT:    movaps %xmm3, 16(%r9)
-; SSE-NEXT:    movaps %xmm8, (%r9)
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,1],xmm5[2,0]
+; SSE-NEXT:    por %xmm3, %xmm7
+; SSE-NEXT:    pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[0,1,1,3]
+; SSE-NEXT:    pshuflw $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,1],xmm4[1,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm3[2,0]
+; SSE-NEXT:    por %xmm1, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,1,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm8[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm3[1,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm1[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rsi)
+; SSE-NEXT:    movdqa %xmm2, 16(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rdx)
+; SSE-NEXT:    movaps %xmm15, 16(%rcx)
+; SSE-NEXT:    movaps %xmm14, (%rcx)
+; SSE-NEXT:    movaps %xmm12, 16(%r8)
+; SSE-NEXT:    movaps %xmm0, (%r8)
+; SSE-NEXT:    movaps %xmm6, 16(%r9)
+; SSE-NEXT:    movaps %xmm7, (%r9)
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i16_stride5_vf16:
@@ -2393,495 +2391,482 @@ define void @load_i16_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 define void @load_i16_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4) nounwind {
 ; SSE-LABEL: load_i16_stride5_vf32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $408, %rsp # imm = 0x198
-; SSE-NEXT:    movdqa 64(%rdi), %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rdi), %xmm10
-; SSE-NEXT:    movdqa 16(%rdi), %xmm13
-; SSE-NEXT:    movdqa 32(%rdi), %xmm9
-; SSE-NEXT:    movdqa 48(%rdi), %xmm5
-; SSE-NEXT:    movdqa 224(%rdi), %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 160(%rdi), %xmm11
-; SSE-NEXT:    movdqa 176(%rdi), %xmm12
+; SSE-NEXT:    subq $344, %rsp # imm = 0x158
+; SSE-NEXT:    movdqa 64(%rdi), %xmm7
+; SSE-NEXT:    movdqa (%rdi), %xmm12
+; SSE-NEXT:    movdqa 16(%rdi), %xmm1
+; SSE-NEXT:    movdqa 32(%rdi), %xmm14
+; SSE-NEXT:    movdqa 48(%rdi), %xmm13
+; SSE-NEXT:    movdqa 224(%rdi), %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 160(%rdi), %xmm8
+; SSE-NEXT:    movdqa 176(%rdi), %xmm11
 ; SSE-NEXT:    movdqa 208(%rdi), %xmm3
-; SSE-NEXT:    movdqa 192(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 192(%rdi), %xmm6
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,0,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    pandn %xmm2, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,1,0,3]
-; SSE-NEXT:    movdqa %xmm3, %xmm8
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[3,1,2,3]
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[0,2,2,3]
-; SSE-NEXT:    movdqa %xmm11, (%rsp) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,3]
-; SSE-NEXT:    movaps {{.*#+}} xmm15 = [65535,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT:    andps %xmm15, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[0,1,0,1]
-; SSE-NEXT:    movaps %xmm15, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    pandn %xmm9, %xmm1
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[0,1,0,3]
-; SSE-NEXT:    movdqa %xmm5, %xmm7
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[3,1,2,3]
-; SSE-NEXT:    movdqa %xmm13, %xmm5
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm6, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; SSE-NEXT:    pand %xmm0, %xmm3
+; SSE-NEXT:    por %xmm2, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm11[3,1,2,3]
+; SSE-NEXT:    movdqa %xmm11, %xmm15
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm8[0,2,2,3]
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[2,3]
+; SSE-NEXT:    movaps {{.*#+}} xmm10 = [65535,65535,65535,65535,65535,65535,65535,0]
+; SSE-NEXT:    andps %xmm10, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[0,1,0,1]
+; SSE-NEXT:    movaps %xmm10, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    por %xmm4, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm14, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm13[0,1,0,3]
 ; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm10[0,2,2,3]
-; SSE-NEXT:    movdqa %xmm10, %xmm6
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,1,0,1]
-; SSE-NEXT:    movaps %xmm15, %xmm2
-; SSE-NEXT:    andnps %xmm1, %xmm2
-; SSE-NEXT:    movdqa 272(%rdi), %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    andps %xmm15, %xmm3
-; SSE-NEXT:    orps %xmm3, %xmm2
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    pandn %xmm4, %xmm1
-; SSE-NEXT:    movdqa 288(%rdi), %xmm2
+; SSE-NEXT:    pand %xmm0, %xmm4
+; SSE-NEXT:    por %xmm2, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[3,1,2,3]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm12[0,2,2,3]
+; SSE-NEXT:    movdqa %xmm12, %xmm11
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm4[2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,1,0,1]
+; SSE-NEXT:    movdqa %xmm7, %xmm12
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm10, %xmm3
+; SSE-NEXT:    andnps %xmm2, %xmm3
+; SSE-NEXT:    movdqa 272(%rdi), %xmm7
+; SSE-NEXT:    andps %xmm10, %xmm5
+; SSE-NEXT:    orps %xmm5, %xmm3
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm7, %xmm2
+; SSE-NEXT:    movdqa 288(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,1,0,3]
+; SSE-NEXT:    pand %xmm0, %xmm4
+; SSE-NEXT:    por %xmm2, %xmm4
+; SSE-NEXT:    movdqa 256(%rdi), %xmm9
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm9[3,1,2,3]
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa 240(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm4[2,3]
+; SSE-NEXT:    movdqa 304(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,1]
+; SSE-NEXT:    movaps %xmm10, %xmm3
+; SSE-NEXT:    andnps %xmm2, %xmm3
+; SSE-NEXT:    andps %xmm10, %xmm5
+; SSE-NEXT:    orps %xmm5, %xmm3
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 128(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
 ; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    movdqa 256(%rdi), %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[3,1,2,3]
-; SSE-NEXT:    movdqa %xmm3, %xmm14
+; SSE-NEXT:    movdqa 112(%rdi), %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    movdqa 240(%rdi), %xmm13
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[0,2,2,3]
-; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,3]
-; SSE-NEXT:    movdqa 304(%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
-; SSE-NEXT:    movaps %xmm15, %xmm2
-; SSE-NEXT:    andnps %xmm1, %xmm2
-; SSE-NEXT:    andps %xmm15, %xmm3
-; SSE-NEXT:    orps %xmm3, %xmm2
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 128(%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movdqa 112(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa 96(%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    movdqa 80(%rdi), %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,2,2,3]
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3]
+; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    por %xmm2, %xmm0
+; SSE-NEXT:    movdqa 96(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, (%rsp) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa 80(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm0[2,3]
 ; SSE-NEXT:    movdqa 144(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
-; SSE-NEXT:    movaps %xmm15, %xmm1
-; SSE-NEXT:    andnps %xmm0, %xmm1
-; SSE-NEXT:    andps %xmm15, %xmm2
-; SSE-NEXT:    orps %xmm2, %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    psrlq $48, %xmm12
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1]
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [0,0,0,65535,65535,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm0, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[1,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm10, %xmm2
-; SSE-NEXT:    psllq $48, %xmm2
-; SSE-NEXT:    movaps %xmm15, %xmm3
-; SSE-NEXT:    andnps %xmm2, %xmm3
-; SSE-NEXT:    pand %xmm15, %xmm1
-; SSE-NEXT:    orps %xmm1, %xmm3
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    psrlq $48, %xmm5
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm6[0,3,2,3]
+; SSE-NEXT:    movaps %xmm10, %xmm2
+; SSE-NEXT:    andnps %xmm0, %xmm2
+; SSE-NEXT:    andps %xmm10, %xmm4
+; SSE-NEXT:    orps %xmm4, %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    psrlq $48, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1]
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    pandn %xmm2, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[1,3,2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [0,0,0,65535,65535,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm0, %xmm4
+; SSE-NEXT:    pandn %xmm2, %xmm4
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[1,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm6[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm5[0,1,2,3,7,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,6,4,7]
 ; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm6, %xmm1
-; SSE-NEXT:    psllq $48, %xmm1
-; SSE-NEXT:    movdqa %xmm15, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    pand %xmm15, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm1
-; SSE-NEXT:    psrlq $48, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm13[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    por %xmm4, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm4
+; SSE-NEXT:    psllq $48, %xmm4
+; SSE-NEXT:    movaps %xmm10, %xmm5
+; SSE-NEXT:    andnps %xmm4, %xmm5
+; SSE-NEXT:    pand %xmm10, %xmm2
+; SSE-NEXT:    orps %xmm2, %xmm5
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    psrlq $48, %xmm2
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm11[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,2,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm4, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm13[1,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm14[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm5[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,4,7]
+; SSE-NEXT:    pand %xmm0, %xmm4
+; SSE-NEXT:    por %xmm2, %xmm4
+; SSE-NEXT:    movdqa %xmm12, %xmm2
+; SSE-NEXT:    psllq $48, %xmm2
+; SSE-NEXT:    movdqa %xmm10, %xmm1
 ; SSE-NEXT:    pandn %xmm2, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[1,3,2,3]
-; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm10, %xmm4
+; SSE-NEXT:    por %xmm4, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    psrlq $48, %xmm9
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm12[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,2,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1]
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm4, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm9[1,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm7[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm5[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,4,7]
+; SSE-NEXT:    pand %xmm0, %xmm4
+; SSE-NEXT:    por %xmm2, %xmm4
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    psllq $48, %xmm2
+; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    pandn %xmm2, %xmm1
+; SSE-NEXT:    pand %xmm10, %xmm4
+; SSE-NEXT:    por %xmm4, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%rsp), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm13, %xmm2
+; SSE-NEXT:    psrlq $48, %xmm2
+; SSE-NEXT:    pshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,2,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSE-NEXT:    pshufd $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[1,3,2,3]
+; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = mem[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm5[0,1,2,3,7,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,6,4,7]
 ; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    psllq $48, %xmm1
-; SSE-NEXT:    movdqa %xmm15, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    pand %xmm15, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm13, %xmm1
-; SSE-NEXT:    psrlq $48, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[1,3,2,3]
-; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pand %xmm15, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm5, %xmm1
-; SSE-NEXT:    psllq $48, %xmm1
-; SSE-NEXT:    pandn %xmm1, %xmm15
-; SSE-NEXT:    por %xmm0, %xmm15
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm4, %xmm0
+; SSE-NEXT:    por %xmm2, %xmm0
+; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    psllq $48, %xmm2
+; SSE-NEXT:    pandn %xmm2, %xmm10
+; SSE-NEXT:    por %xmm0, %xmm10
 ; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    movdqa %xmm11, %xmm8
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm11[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm11[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm6[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm6[2,3]
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0,1,3]
-; SSE-NEXT:    movaps {{.*#+}} xmm11 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    movaps %xmm11, %xmm1
-; SSE-NEXT:    andnps %xmm0, %xmm1
-; SSE-NEXT:    movdqa (%rsp), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,1,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
-; SSE-NEXT:    pand %xmm11, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[0,1,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm2[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm15[2],xmm4[3],xmm15[3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT:    pand %xmm15, %xmm4
+; SSE-NEXT:    movdqa %xmm15, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    movaps %xmm1, %xmm15
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm9[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm9[2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,1],xmm0[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[0,1,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; SSE-NEXT:    pand %xmm15, %xmm4
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm14[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm14[2,3]
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0,1,3]
-; SSE-NEXT:    movaps %xmm11, %xmm1
-; SSE-NEXT:    andnps %xmm0, %xmm1
-; SSE-NEXT:    pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[0,1,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
-; SSE-NEXT:    pand %xmm11, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm15, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[1,0],xmm10[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[2,0],xmm10[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,0,1,3]
-; SSE-NEXT:    movaps %xmm11, %xmm1
-; SSE-NEXT:    andnps %xmm14, %xmm1
-; SSE-NEXT:    pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[0,1,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
-; SSE-NEXT:    pand %xmm11, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm14[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm1[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,1],xmm0[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[0,1,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm0[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[2],mem[2],xmm12[3],mem[3]
+; SSE-NEXT:    pand %xmm15, %xmm12
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[1,0],xmm7[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[2,0],xmm7[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,0,1,3]
+; SSE-NEXT:    movdqa %xmm15, %xmm2
+; SSE-NEXT:    pandn %xmm9, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm12
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm9[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,1],xmm0[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,1],xmm2[2,0]
+; SSE-NEXT:    pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[0,1,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm0[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm13[2],xmm11[3],xmm13[3]
+; SSE-NEXT:    pand %xmm15, %xmm11
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm4[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm4[2,3]
+; SSE-NEXT:    movaps %xmm4, %xmm13
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0,1,3]
-; SSE-NEXT:    movaps %xmm11, %xmm1
-; SSE-NEXT:    andnps %xmm0, %xmm1
-; SSE-NEXT:    pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[0,1,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm13[2],xmm2[3],xmm13[3]
-; SSE-NEXT:    pand %xmm11, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm15, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm11
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm8[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm2
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm8[3,0]
-; SSE-NEXT:    movaps %xmm11, %xmm0
-; SSE-NEXT:    andnps %xmm8, %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm3[0,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm8[0,1,2,3,7,4,6,7]
-; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[0,1,0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,1],xmm0[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,1],xmm2[2,0]
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm6[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pandn %xmm2, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[1,1,1,1]
+; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = mem[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm8[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm2[1,0,3,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm15, %xmm8
+; SSE-NEXT:    por %xmm0, %xmm8
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm6[3,0]
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    pandn %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm6[0,1,2,3,7,4,6,7]
+; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,5,6]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,1],xmm0[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm2[2,0]
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm14[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pandn %xmm2, %xmm0
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm4[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm2[0],xmm6[1],xmm2[1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm6[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm2[1,0,3,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm15, %xmm6
+; SSE-NEXT:    por %xmm0, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm14[3,0]
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    pandn %xmm14, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm0[0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm14[0,1,2,3,7,4,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,6]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm9[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm1[2,0]
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm7[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pandn %xmm1, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm0[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm11, %xmm13
-; SSE-NEXT:    por %xmm1, %xmm13
-; SSE-NEXT:    movaps %xmm15, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm9[3,0]
-; SSE-NEXT:    movdqa %xmm11, %xmm12
-; SSE-NEXT:    pandn %xmm9, %xmm12
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,1],xmm0[0,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm9[0,1,2,3,7,4,6,7]
-; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,6]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,1],xmm1[2,0]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm10[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[1,1,1,1]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm0[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm11, %xmm9
-; SSE-NEXT:    por %xmm1, %xmm9
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm10[3,0]
-; SSE-NEXT:    movdqa %xmm11, %xmm15
-; SSE-NEXT:    pandn %xmm10, %xmm15
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,1],xmm0[0,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm10[0,1,2,3,7,4,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[0,1,0,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm1[1,0,3,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm15, %xmm9
+; SSE-NEXT:    por %xmm0, %xmm9
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm7[3,0]
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    pandn %xmm7, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm0[0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm7[0,1,2,3,7,4,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,6]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
 ; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,1],xmm1[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm7[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm0[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm11, %xmm10
-; SSE-NEXT:    por %xmm1, %xmm10
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm13[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pandn %xmm1, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[1,1,1,1]
+; SSE-NEXT:    movdqa (%rsp), %xmm5 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm1[1,0,3,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm15, %xmm7
+; SSE-NEXT:    por %xmm0, %xmm7
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps (%rsp), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[3,0]
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0,2]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,0],xmm14[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm8[0,2]
-; SSE-NEXT:    movaps %xmm14, %xmm2
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm4[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm1[0,2]
 ; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0],xmm3[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm6[0,2]
-; SSE-NEXT:    movaps %xmm1, %xmm14
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm4[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm5[0,2]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm8, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm7[3,0]
+; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm3[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[0,2]
+; SSE-NEXT:    movdqa %xmm5, %xmm2
+; SSE-NEXT:    movdqa %xmm13, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm5[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm3[0,2]
+; SSE-NEXT:    movaps %xmm2, %xmm3
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm0[2,1,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm11, %xmm3
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm2[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm0[2,1,3,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm15, %xmm13
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,4,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm0[2,1,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm11, %xmm2
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm14[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pand %xmm15, %xmm2
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm0[2,1,3,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm15, %xmm1
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm3[0,1,2,3,4,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[2,1,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm11, %xmm4
-; SSE-NEXT:    pandn %xmm7, %xmm11
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm1[0,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm7[0,1,2,3,7,4,6,7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm6[0,1,0,3]
+; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[0,2,2,3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm4[3,0]
+; SSE-NEXT:    pandn %xmm4, %xmm15
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm4[0,1,2,3,7,4,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm4[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,4,5,5,6]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[3,1],xmm1[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,1],xmm14[2,0]
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[0,2,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[3,1],xmm3[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm14[2,0]
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[0,2,2,3]
 ; SSE-NEXT:    pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm14 = mem[0,1,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,0,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[3,1],xmm1[1,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm14[2,0]
-; SSE-NEXT:    por %xmm12, %xmm2
-; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[0,2,2,3]
-; SSE-NEXT:    pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm12 = mem[0,1,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[3,1],xmm1[1,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm12[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[3,1],xmm3[1,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,1],xmm14[2,0]
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[0,1,1,3]
+; SSE-NEXT:    pshuflw $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,1],xmm14[1,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,0]
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[0,1,1,3]
+; SSE-NEXT:    pshuflw $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,1],xmm14[1,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,0]
 ; SSE-NEXT:    por %xmm15, %xmm0
-; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[0,2,2,3]
-; SSE-NEXT:    pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm5 = mem[0,1,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,1],xmm1[1,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm5[2,0]
-; SSE-NEXT:    por %xmm4, %xmm11
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[0,2,2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm6[0,1,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,1],xmm1[1,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,1],xmm4[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 16(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 48(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, (%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 32(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 16(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 48(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, (%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 32(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 16(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 48(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, (%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 32(%rcx)
-; SSE-NEXT:    movaps %xmm10, 16(%r8)
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,1,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,1],xmm5[1,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, 32(%rsi)
+; SSE-NEXT:    movdqa %xmm10, 16(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, 48(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, (%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, 32(%rdx)
+; SSE-NEXT:    movaps %xmm11, 16(%rcx)
+; SSE-NEXT:    movaps %xmm12, 48(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, (%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, 32(%rcx)
+; SSE-NEXT:    movaps %xmm7, 16(%r8)
 ; SSE-NEXT:    movaps %xmm9, 48(%r8)
-; SSE-NEXT:    movaps %xmm13, (%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 32(%r8)
-; SSE-NEXT:    movaps %xmm11, 16(%r9)
-; SSE-NEXT:    movaps %xmm0, 48(%r9)
+; SSE-NEXT:    movaps %xmm6, (%r8)
+; SSE-NEXT:    movaps %xmm8, 32(%r8)
+; SSE-NEXT:    movaps %xmm0, 16(%r9)
+; SSE-NEXT:    movaps %xmm1, 48(%r9)
 ; SSE-NEXT:    movaps %xmm2, (%r9)
-; SSE-NEXT:    movaps %xmm3, 32(%r9)
-; SSE-NEXT:    addq $408, %rsp # imm = 0x198
+; SSE-NEXT:    movaps %xmm13, 32(%r9)
+; SSE-NEXT:    addq $344, %rsp # imm = 0x158
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i16_stride5_vf32:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll
index ea3bf7b9b7203..44f4c95941b00 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-7.ll
@@ -1044,8 +1044,8 @@ define void @load_i16_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movdqa (%rdi), %xmm0
 ; SSE-NEXT:    movdqa 16(%rdi), %xmm9
 ; SSE-NEXT:    movaps 32(%rdi), %xmm2
-; SSE-NEXT:    movaps 48(%rdi), %xmm8
-; SSE-NEXT:    movdqa 80(%rdi), %xmm7
+; SSE-NEXT:    movaps 48(%rdi), %xmm7
+; SSE-NEXT:    movdqa 80(%rdi), %xmm8
 ; SSE-NEXT:    movdqa 64(%rdi), %xmm1
 ; SSE-NEXT:    movdqa 96(%rdi), %xmm6
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm6[0,0,0,0]
@@ -1053,14 +1053,14 @@ define void @load_i16_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movdqa %xmm11, %xmm4
 ; SSE-NEXT:    pandn %xmm3, %xmm4
 ; SSE-NEXT:    movdqa %xmm1, %xmm5
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm8[0],xmm5[1],xmm8[1]
 ; SSE-NEXT:    pand %xmm11, %xmm5
 ; SSE-NEXT:    por %xmm4, %xmm5
 ; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,65535,65535,0,0,0]
 ; SSE-NEXT:    movdqa %xmm3, %xmm10
 ; SSE-NEXT:    pandn %xmm5, %xmm10
 ; SSE-NEXT:    movaps %xmm2, %xmm5
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,2],xmm8[2,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,2],xmm7[2,2]
 ; SSE-NEXT:    movaps {{.*#+}} xmm4 = [65535,65535,65535,0,0,65535,65535,65535]
 ; SSE-NEXT:    movaps %xmm4, %xmm12
 ; SSE-NEXT:    andnps %xmm5, %xmm12
@@ -1077,49 +1077,49 @@ define void @load_i16_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pslldq {{.*#+}} xmm10 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm10[0,1,2,3,4,5]
 ; SSE-NEXT:    movdqa %xmm11, %xmm12
 ; SSE-NEXT:    pandn %xmm10, %xmm12
-; SSE-NEXT:    movdqa %xmm7, %xmm10
+; SSE-NEXT:    movdqa %xmm8, %xmm10
 ; SSE-NEXT:    psrld $16, %xmm10
-; SSE-NEXT:    movdqa %xmm1, %xmm15
-; SSE-NEXT:    psrldq {{.*#+}} xmm15 = xmm15[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm10[0],xmm15[1],xmm10[1]
-; SSE-NEXT:    pand %xmm11, %xmm15
-; SSE-NEXT:    por %xmm12, %xmm15
-; SSE-NEXT:    movdqa %xmm3, %xmm13
-; SSE-NEXT:    pandn %xmm15, %xmm13
+; SSE-NEXT:    movdqa %xmm1, %xmm13
+; SSE-NEXT:    psrldq {{.*#+}} xmm13 = xmm13[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm10[0],xmm13[1],xmm10[1]
+; SSE-NEXT:    pand %xmm11, %xmm13
+; SSE-NEXT:    por %xmm12, %xmm13
+; SSE-NEXT:    movdqa %xmm3, %xmm15
+; SSE-NEXT:    pandn %xmm13, %xmm15
 ; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,0,65535,65535,65535,65535,65535,65535]
 ; SSE-NEXT:    movdqa %xmm10, %xmm12
 ; SSE-NEXT:    pandn %xmm0, %xmm12
-; SSE-NEXT:    movdqa %xmm9, %xmm15
-; SSE-NEXT:    pand %xmm10, %xmm15
-; SSE-NEXT:    por %xmm12, %xmm15
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm15[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm4, %xmm12
-; SSE-NEXT:    movaps %xmm2, %xmm15
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm15 = xmm15[4],xmm8[4],xmm15[5],xmm8[5],xmm15[6],xmm8[6],xmm15[7],xmm8[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm15[2,1,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm15[0,0,0,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pandn %xmm15, %xmm4
-; SSE-NEXT:    movdqa %xmm1, %xmm15
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm7[2],xmm15[3],xmm7[3]
-; SSE-NEXT:    por %xmm12, %xmm4
+; SSE-NEXT:    movdqa %xmm9, %xmm13
+; SSE-NEXT:    pand %xmm10, %xmm13
+; SSE-NEXT:    por %xmm12, %xmm13
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm13[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm12[1,0,3,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm4, %xmm13
+; SSE-NEXT:    movaps %xmm2, %xmm12
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm12 = xmm12[4],xmm7[4],xmm12[5],xmm7[5],xmm12[6],xmm7[6],xmm12[7],xmm7[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[2,1,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pandn %xmm12, %xmm4
 ; SSE-NEXT:    movdqa %xmm0, %xmm12
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm9[0],xmm12[1],xmm9[1],xmm12[2],xmm9[2],xmm12[3],xmm9[3]
-; SSE-NEXT:    pand %xmm3, %xmm4
 ; SSE-NEXT:    por %xmm13, %xmm4
 ; SSE-NEXT:    movdqa %xmm1, %xmm13
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm13 = xmm13[4],xmm7[4],xmm13[5],xmm7[5],xmm13[6],xmm7[6],xmm13[7],xmm7[7]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm13 = xmm13[4],xmm8[4],xmm13[5],xmm8[5],xmm13[6],xmm8[6],xmm13[7],xmm8[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm13[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,4,7,7]
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm13 = xmm13[4],xmm6[4],xmm13[5],xmm6[5],xmm13[6],xmm6[6],xmm13[7],xmm6[7]
+; SSE-NEXT:    pand %xmm3, %xmm4
+; SSE-NEXT:    por %xmm15, %xmm4
+; SSE-NEXT:    movdqa %xmm1, %xmm15
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm8[2],xmm15[3],xmm8[3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm15[0,1,0,1]
 ; SSE-NEXT:    pand %xmm11, %xmm15
 ; SSE-NEXT:    pandn %xmm14, %xmm11
 ; SSE-NEXT:    por %xmm15, %xmm11
 ; SSE-NEXT:    movdqa %xmm3, %xmm14
 ; SSE-NEXT:    pandn %xmm11, %xmm14
-; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm8[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm7[2,2,3,3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm2[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    punpckhqdq {{.*#+}} xmm11 = xmm11[1],xmm15[1]
@@ -1130,7 +1130,7 @@ define void @load_i16_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    orps %xmm14, %xmm11
 ; SSE-NEXT:    movdqa %xmm10, %xmm14
 ; SSE-NEXT:    pandn %xmm2, %xmm14
-; SSE-NEXT:    movdqa %xmm8, %xmm15
+; SSE-NEXT:    movdqa %xmm7, %xmm15
 ; SSE-NEXT:    pand %xmm10, %xmm15
 ; SSE-NEXT:    por %xmm14, %xmm15
 ; SSE-NEXT:    movdqa %xmm0, %xmm14
@@ -1148,17 +1148,17 @@ define void @load_i16_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm13[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,6,4,7]
 ; SSE-NEXT:    pandn %xmm13, %xmm3
-; SSE-NEXT:    movdqa %xmm2, %xmm13
-; SSE-NEXT:    psrlq $16, %xmm13
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm8[0],xmm13[1],xmm8[1],xmm13[2],xmm8[2],xmm13[3],xmm8[3]
 ; SSE-NEXT:    por %xmm15, %xmm3
-; SSE-NEXT:    movdqa %xmm7, %xmm15
-; SSE-NEXT:    pand %xmm10, %xmm15
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm1[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm13[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm8[2],xmm15[3],xmm8[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm8[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pand %xmm10, %xmm8
 ; SSE-NEXT:    pandn %xmm1, %xmm10
-; SSE-NEXT:    por %xmm15, %xmm10
-; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm1[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm7[2],xmm15[3],xmm7[3]
+; SSE-NEXT:    por %xmm8, %xmm10
+; SSE-NEXT:    movdqa %xmm2, %xmm8
+; SSE-NEXT:    psrlq $16, %xmm8
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm7[0],xmm8[1],xmm7[1],xmm8[2],xmm7[2],xmm8[3],xmm7[3]
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm15 = xmm15[4],xmm6[4],xmm15[5],xmm6[5],xmm15[6],xmm6[6],xmm15[7],xmm6[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,5,4,7,7]
@@ -1166,31 +1166,30 @@ define void @load_i16_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    psrlq $48, %xmm9
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm9[0],xmm12[1],xmm9[1],xmm12[2],xmm9[2],xmm12[3],xmm9[3]
 ; SSE-NEXT:    movdqa %xmm2, %xmm9
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[1,1,1,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm7[0],xmm9[1],xmm7[1],xmm9[2],xmm7[2],xmm9[3],xmm7[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[1,1,1,1]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,0,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm13[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm13[0],xmm12[1],xmm13[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm8[0],xmm12[1],xmm8[1]
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rdi
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm15[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm13[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,5,6,4,7]
-; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm12[0],xmm13[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm15[3,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,5,6,4,7]
+; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm12[0],xmm8[1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[0,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[2,3,2,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm9[0],xmm14[1],xmm9[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm10[0,2]
 ; SSE-NEXT:    psrldq {{.*#+}} xmm2 = xmm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm7[0],xmm2[1],xmm7[1],xmm2[2],xmm7[2],xmm2[3],xmm7[3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,1,1]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1],xmm1[2],xmm13[2],xmm1[3],xmm13[3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm6[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
@@ -1199,7 +1198,7 @@ define void @load_i16_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movdqa %xmm4, (%rdx)
 ; SSE-NEXT:    movaps %xmm11, (%rcx)
 ; SSE-NEXT:    movdqa %xmm3, (%r8)
-; SSE-NEXT:    movapd %xmm13, (%r9)
+; SSE-NEXT:    movapd %xmm8, (%r9)
 ; SSE-NEXT:    movaps %xmm14, (%rdi)
 ; SSE-NEXT:    movapd %xmm1, (%rax)
 ; SSE-NEXT:    retq
@@ -2015,369 +2014,371 @@ define void @load_i16_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i16_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6) nounwind {
 ; SSE-LABEL: load_i16_stride7_vf16:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $232, %rsp
-; SSE-NEXT:    movdqa 80(%rdi), %xmm11
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 64(%rdi), %xmm9
+; SSE-NEXT:    subq $216, %rsp
+; SSE-NEXT:    movdqa 80(%rdi), %xmm13
+; SSE-NEXT:    movdqa 64(%rdi), %xmm15
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 112(%rdi), %xmm8
+; SSE-NEXT:    movdqa 128(%rdi), %xmm9
 ; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 112(%rdi), %xmm12
-; SSE-NEXT:    movdqa 128(%rdi), %xmm6
-; SSE-NEXT:    movaps 160(%rdi), %xmm5
-; SSE-NEXT:    movaps %xmm5, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movaps 144(%rdi), %xmm7
-; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 192(%rdi), %xmm13
-; SSE-NEXT:    movdqa 176(%rdi), %xmm15
-; SSE-NEXT:    movdqa 208(%rdi), %xmm14
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1]
-; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,65535,65535,0,0,0]
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    movdqa %xmm3, %xmm10
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    movaps %xmm7, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm5[2,2]
-; SSE-NEXT:    movaps {{.*#+}} xmm8 = [65535,65535,65535,0,0,65535,65535,65535]
-; SSE-NEXT:    movaps %xmm8, %xmm4
-; SSE-NEXT:    andnps %xmm0, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm12[0,1,0,3]
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm5[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
-; SSE-NEXT:    movdqa 96(%rdi), %xmm0
+; SSE-NEXT:    movaps 160(%rdi), %xmm4
+; SSE-NEXT:    movaps 144(%rdi), %xmm5
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 192(%rdi), %xmm6
+; SSE-NEXT:    movdqa 176(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 208(%rdi), %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,0,0,0]
+; SSE-NEXT:    movdqa %xmm0, %xmm7
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm8, %xmm3
-; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    pand %xmm10, %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
+; SSE-NEXT:    movdqa %xmm6, %xmm11
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm1
+; SSE-NEXT:    por %xmm2, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [65535,65535,65535,65535,65535,0,0,0]
+; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    movaps %xmm5, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,2],xmm4[2,2]
+; SSE-NEXT:    movaps %xmm4, %xmm6
+; SSE-NEXT:    movaps {{.*#+}} xmm10 = [65535,65535,65535,0,0,65535,65535,65535]
+; SSE-NEXT:    movaps %xmm10, %xmm3
+; SSE-NEXT:    andnps %xmm1, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm8[0,1,0,3]
+; SSE-NEXT:    movdqa %xmm8, %xmm9
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    movdqa 96(%rdi), %xmm5
+; SSE-NEXT:    pand %xmm10, %xmm4
+; SSE-NEXT:    por %xmm3, %xmm4
+; SSE-NEXT:    pand %xmm14, %xmm4
+; SSE-NEXT:    por %xmm2, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[0,0,0,0]
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1]
+; SSE-NEXT:    movdqa %xmm13, %xmm8
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm1
+; SSE-NEXT:    por %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    movaps 32(%rdi), %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 48(%rdi), %xmm3
+; SSE-NEXT:    movaps %xmm3, (%rsp) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,2],xmm3[2,2]
+; SSE-NEXT:    movaps %xmm10, %xmm3
+; SSE-NEXT:    andnps %xmm1, %xmm3
+; SSE-NEXT:    movdqa (%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm1[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    movdqa 16(%rdi), %xmm15
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    pand %xmm10, %xmm4
+; SSE-NEXT:    por %xmm3, %xmm4
+; SSE-NEXT:    pand %xmm14, %xmm4
+; SSE-NEXT:    por %xmm2, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    pslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,1,2,3,4,5]
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm11, %xmm1
+; SSE-NEXT:    psrld $16, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm7, %xmm3
+; SSE-NEXT:    psrldq {{.*#+}} xmm3 = xmm3[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    pand %xmm0, %xmm3
 ; SSE-NEXT:    por %xmm2, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm11[0],xmm9[1],xmm11[1]
-; SSE-NEXT:    pand %xmm1, %xmm9
-; SSE-NEXT:    por %xmm2, %xmm9
-; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    movdqa %xmm14, %xmm1
+; SSE-NEXT:    pandn %xmm3, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,0,65535,65535,65535,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm11, %xmm2
+; SSE-NEXT:    movdqa %xmm9, %xmm13
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pandn %xmm9, %xmm2
-; SSE-NEXT:    movaps 32(%rdi), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 48(%rdi), %xmm3
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm3[2,2]
-; SSE-NEXT:    movaps %xmm8, %xmm4
-; SSE-NEXT:    andnps %xmm0, %xmm4
-; SSE-NEXT:    movdqa (%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    movdqa 16(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
-; SSE-NEXT:    pand %xmm8, %xmm3
-; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    pand %xmm10, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm12, %xmm4
+; SSE-NEXT:    pand %xmm11, %xmm4
+; SSE-NEXT:    por %xmm2, %xmm4
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    movdqa %xmm10, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,3,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm10, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm3
+; SSE-NEXT:    pand %xmm14, %xmm3
+; SSE-NEXT:    por %xmm1, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pslldq {{.*#+}} xmm14 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm14[0,1,2,3,4,5]
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    pandn %xmm14, %xmm2
-; SSE-NEXT:    psrld $16, %xmm13
-; SSE-NEXT:    movdqa %xmm15, %xmm4
-; SSE-NEXT:    movdqa %xmm15, %xmm11
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm5[0,1,2,3,4,5]
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm5, %xmm2
+; SSE-NEXT:    psrld $16, %xmm8
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm4
 ; SSE-NEXT:    psrldq {{.*#+}} xmm4 = xmm4[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm13[0],xmm4[1],xmm13[1]
-; SSE-NEXT:    pand %xmm1, %xmm4
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1]
+; SSE-NEXT:    pand %xmm0, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    movdqa %xmm10, %xmm14
-; SSE-NEXT:    pandn %xmm4, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm7, %xmm4
-; SSE-NEXT:    pandn %xmm12, %xmm4
-; SSE-NEXT:    pand %xmm7, %xmm6
-; SSE-NEXT:    por %xmm4, %xmm6
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm9, %xmm4
-; SSE-NEXT:    movdqa (%rsp), %xmm13 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm13[4],xmm4[5],xmm13[5],xmm4[6],xmm13[6],xmm4[7],xmm13[7]
+; SSE-NEXT:    movdqa %xmm14, %xmm1
+; SSE-NEXT:    pandn %xmm4, %xmm1
+; SSE-NEXT:    movdqa %xmm11, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm8, %xmm2
+; SSE-NEXT:    movdqa %xmm15, %xmm5
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm15, %xmm4
+; SSE-NEXT:    pand %xmm11, %xmm4
+; SSE-NEXT:    por %xmm2, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,3,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm10, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movdqa (%rsp), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm15[4],xmm4[5],xmm15[5],xmm4[6],xmm15[6],xmm4[7],xmm15[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,1,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    movdqa %xmm8, %xmm2
-; SSE-NEXT:    pandn %xmm4, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm6[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm8, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm2
-; SSE-NEXT:    pand %xmm10, %xmm2
-; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
-; SSE-NEXT:    movdqa %xmm1, %xmm4
-; SSE-NEXT:    pandn %xmm0, %xmm4
+; SSE-NEXT:    pandn %xmm4, %xmm10
+; SSE-NEXT:    por %xmm2, %xmm10
+; SSE-NEXT:    pand %xmm14, %xmm10
+; SSE-NEXT:    por %xmm1, %xmm10
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,1,0,1]
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; SSE-NEXT:    pand %xmm0, %xmm1
+; SSE-NEXT:    por %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm1[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[2,2,3,3]
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm7 = xmm7[1],xmm1[1]
+; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1],xmm1[2],xmm12[2],xmm1[3],xmm12[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    movss {{.*#+}} xmm7 = xmm4[0],xmm7[1,2,3]
+; SSE-NEXT:    andps %xmm14, %xmm7
+; SSE-NEXT:    orps %xmm2, %xmm7
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm3, %xmm2
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    psrld $16, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm15, %xmm5
-; SSE-NEXT:    psrldq {{.*#+}} xmm5 = xmm5[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
-; SSE-NEXT:    pand %xmm1, %xmm5
-; SSE-NEXT:    por %xmm4, %xmm5
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pandn %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm7, %xmm4
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,1]
+; SSE-NEXT:    pand %xmm0, %xmm2
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm10, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm12, %xmm5
-; SSE-NEXT:    pand %xmm7, %xmm5
-; SSE-NEXT:    por %xmm4, %xmm5
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm8, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm5 = xmm5[4],mem[4],xmm5[5],mem[5],xmm5[6],mem[6],xmm5[7],mem[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,1,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,0,0,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pandn %xmm5, %xmm8
-; SSE-NEXT:    por %xmm4, %xmm8
-; SSE-NEXT:    pand %xmm14, %xmm8
-; SSE-NEXT:    por %xmm0, %xmm8
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm10[0,1,0,1]
+; SSE-NEXT:    pandn %xmm4, %xmm0
+; SSE-NEXT:    por %xmm2, %xmm0
+; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm0[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,2,3,3]
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm12 = xmm12[1],xmm0[1]
+; SSE-NEXT:    movdqa %xmm8, %xmm0
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    movss {{.*#+}} xmm12 = xmm4[0],xmm12[1,2,3]
+; SSE-NEXT:    andps %xmm14, %xmm12
+; SSE-NEXT:    orps %xmm2, %xmm12
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm11, %xmm2
+; SSE-NEXT:    pandn %xmm9, %xmm2
+; SSE-NEXT:    pand %xmm11, %xmm6
+; SSE-NEXT:    por %xmm2, %xmm6
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm6[0,1,1,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
+; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[0,1,0,1]
-; SSE-NEXT:    movdqa %xmm1, %xmm4
-; SSE-NEXT:    pandn %xmm0, %xmm4
+; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,7]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm6[2],xmm11[3],xmm6[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[0,1,0,1]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    por %xmm4, %xmm0
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,4,7]
 ; SSE-NEXT:    movdqa %xmm14, %xmm4
-; SSE-NEXT:    pandn %xmm0, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm0[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm13, %xmm2
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm11 = xmm11[1],xmm0[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm5 = xmm5[0],mem[0],xmm5[1],mem[1],xmm5[2],mem[2],xmm5[3],mem[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    movss {{.*#+}} xmm11 = xmm0[0],xmm11[1,2,3]
-; SSE-NEXT:    andps %xmm14, %xmm11
-; SSE-NEXT:    orps %xmm4, %xmm11
-; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,1,0,1]
-; SSE-NEXT:    pandn %xmm4, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    movaps %xmm14, %xmm15
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[2,2,3,3]
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm4 = xmm4[1],xmm1[1]
-; SSE-NEXT:    movdqa %xmm10, %xmm1
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1],xmm1[2],xmm12[2],xmm1[3],xmm12[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm1[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    movss {{.*#+}} xmm4 = xmm11[0],xmm4[1,2,3]
-; SSE-NEXT:    andps %xmm15, %xmm4
-; SSE-NEXT:    orps %xmm0, %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    pandn %xmm9, %xmm0
-; SSE-NEXT:    pand %xmm7, %xmm2
-; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[0,1,1,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,2,2,2]
-; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm5[0],xmm0[1,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm4, %xmm5
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm6[4],xmm5[5],xmm6[5],xmm5[6],xmm6[6],xmm5[7],xmm6[7]
-; SSE-NEXT:    movdqa %xmm6, %xmm12
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,7,7]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm8[4],xmm5[5],xmm8[5],xmm5[6],xmm8[6],xmm5[7],xmm8[7]
-; SSE-NEXT:    movdqa %xmm8, %xmm6
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,6,4,7]
-; SSE-NEXT:    movaps %xmm15, %xmm2
-; SSE-NEXT:    andnps %xmm5, %xmm2
-; SSE-NEXT:    andps %xmm15, %xmm0
-; SSE-NEXT:    orps %xmm0, %xmm2
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    pandn %xmm13, %xmm0
-; SSE-NEXT:    movdqa %xmm13, %xmm9
-; SSE-NEXT:    movdqa %xmm14, %xmm5
-; SSE-NEXT:    movdqa %xmm14, %xmm13
-; SSE-NEXT:    pand %xmm7, %xmm5
-; SSE-NEXT:    por %xmm0, %xmm5
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm5[0,1,1,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
-; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,7]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
-; SSE-NEXT:    andps %xmm15, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,4,7]
-; SSE-NEXT:    andnps %xmm1, %xmm15
-; SSE-NEXT:    orps %xmm0, %xmm15
-; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm1, %xmm4
+; SSE-NEXT:    andps %xmm14, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm11, %xmm1
+; SSE-NEXT:    pandn %xmm7, %xmm1
+; SSE-NEXT:    pand %xmm11, %xmm15
+; SSE-NEXT:    por %xmm1, %xmm15
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm15[0,1,1,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
+; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm0, %xmm14
-; SSE-NEXT:    psrld $16, %xmm14
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm14 = xmm14[4],xmm2[4],xmm14[5],xmm2[5],xmm14[6],xmm2[6],xmm14[7],xmm2[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm0, %xmm5
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm2[4],xmm5[5],xmm2[5],xmm5[6],xmm2[6],xmm5[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    psrlq $48, %xmm0
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    psrlq $16, %xmm0
-; SSE-NEXT:    movdqa (%rsp), %xmm11 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1],xmm0[2],xmm11[2],xmm0[3],xmm11[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm4, %xmm5
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm12[2],xmm0[3],xmm12[3]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE-NEXT:    movdqa %xmm3, %xmm15
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,4,7]
-; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm10, %xmm15
-; SSE-NEXT:    psrld $16, %xmm15
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,7,7]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm10[4],xmm0[5],xmm10[5],xmm0[6],xmm10[6],xmm0[7],xmm10[7]
+; SSE-NEXT:    movdqa %xmm10, %xmm3
+; SSE-NEXT:    andps %xmm14, %xmm1
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,4,7]
+; SSE-NEXT:    pandn %xmm0, %xmm14
+; SSE-NEXT:    por %xmm1, %xmm14
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm1, %xmm12
+; SSE-NEXT:    psrld $16, %xmm12
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm15 = xmm15[4],xmm2[4],xmm15[5],xmm2[5],xmm15[6],xmm2[6],xmm15[7],xmm2[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,3,2,3]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm12 = xmm12[4],xmm2[4],xmm12[5],xmm2[5],xmm12[6],xmm2[6],xmm12[7],xmm2[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm1, %xmm4
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm2[4],xmm10[5],xmm2[5],xmm10[6],xmm2[6],xmm10[7],xmm2[7]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm2[4],xmm4[5],xmm2[5],xmm4[6],xmm2[6],xmm4[7],xmm2[7]
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    psrlq $48, %xmm1
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    movdqa %xmm9, %xmm4
 ; SSE-NEXT:    movdqa %xmm9, %xmm1
 ; SSE-NEXT:    psrlq $16, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1],xmm1[2],xmm13[2],xmm1[3],xmm13[3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,1,0,3]
+; SSE-NEXT:    movdqa %xmm8, %xmm10
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; SSE-NEXT:    movdqa %xmm5, %xmm8
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1],xmm0[2],xmm11[2],xmm0[3],xmm11[3]
-; SSE-NEXT:    movdqa %xmm11, %xmm8
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    psrld $16, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,2,3]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7]
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    psrlq $48, %xmm4
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; SSE-NEXT:    movdqa %xmm7, %xmm4
+; SSE-NEXT:    psrlq $16, %xmm4
+; SSE-NEXT:    movdqa (%rsp), %xmm5 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    movdqa %xmm15, %xmm6
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm15[2],xmm4[3],xmm15[3]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm3[4],xmm4[5],xmm3[5],xmm4[6],xmm3[6],xmm4[7],xmm3[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[3,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,4,7]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1],xmm0[2],xmm13[2],xmm0[3],xmm13[3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    pandn %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm12, %xmm11
-; SSE-NEXT:    pand %xmm7, %xmm11
-; SSE-NEXT:    por %xmm0, %xmm11
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[0,1,0,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    pandn %xmm10, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm8[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pand %xmm11, %xmm8
+; SSE-NEXT:    por %xmm0, %xmm8
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,7]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm0[0,2]
-; SSE-NEXT:    pand %xmm7, %xmm2
-; SSE-NEXT:    movdqa %xmm9, %xmm11
-; SSE-NEXT:    pandn %xmm9, %xmm7
-; SSE-NEXT:    por %xmm2, %xmm7
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1],xmm0[2],xmm13[2],xmm0[3],xmm13[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,1],xmm0[0,2]
+; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm11, %xmm0
+; SSE-NEXT:    pandn %xmm2, %xmm11
+; SSE-NEXT:    por %xmm0, %xmm11
+; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3]
+; SSE-NEXT:    movdqa %xmm5, %xmm6
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[0,1,0,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,7]
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE-NEXT:    movdqa %xmm3, %xmm11
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm0[0,2]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    psrldq {{.*#+}} xmm2 = xmm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1]
-; SSE-NEXT:    pshufd $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm12[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm6[0,1,0,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,2,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm9[2],xmm0[3],xmm9[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm13[1,1,1,1]
-; SSE-NEXT:    psrldq {{.*#+}} xmm4 = xmm4[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm10[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm2[0],xmm9[1],xmm2[1]
-; SSE-NEXT:    pshuflw $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm2[0],xmm11[1],xmm2[1],xmm11[2],xmm2[2],xmm11[3],xmm2[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,1,0,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[2,2,2,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm2[2],xmm11[3],xmm2[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm9[0],xmm11[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1]
+; SSE-NEXT:    psrldq {{.*#+}} xmm9 = xmm9[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1],xmm9[2],xmm0[2],xmm9[3],xmm0[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[1,1,1,1]
+; SSE-NEXT:    pshufd $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm15[0],xmm10[1],xmm15[1],xmm10[2],xmm15[2],xmm10[3],xmm15[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm8[0,1,0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,2,2,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm5[2],xmm0[3],xmm5[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm6[1,1,1,1]
+; SSE-NEXT:    psrldq {{.*#+}} xmm7 = xmm7[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm3[0],xmm7[1],xmm3[1],xmm7[2],xmm3[2],xmm7[3],xmm3[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[1,1,1,1]
+; SSE-NEXT:    pshufd $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = mem[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
+; SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1],xmm2[2],mem[2],xmm2[3],mem[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[0,1,0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm2[2,2,2,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm3[2],xmm11[3],xmm3[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm5[0],xmm11[1]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm2, (%rsi)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
@@ -2390,20 +2391,19 @@ define void @load_i16_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movaps %xmm2, (%rcx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm2, 16(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, (%r8)
+; SSE-NEXT:    movdqa %xmm14, (%r8)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm2, 16(%r8)
-; SSE-NEXT:    movapd %xmm1, (%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 16(%r9)
+; SSE-NEXT:    movapd %xmm4, (%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, 16(%r9)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movaps %xmm15, (%rax)
-; SSE-NEXT:    movaps %xmm14, 16(%rax)
+; SSE-NEXT:    movaps %xmm1, (%rax)
+; SSE-NEXT:    movaps %xmm12, 16(%rax)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; SSE-NEXT:    movapd %xmm11, (%rax)
 ; SSE-NEXT:    movapd %xmm0, 16(%rax)
-; SSE-NEXT:    addq $232, %rsp
+; SSE-NEXT:    addq $216, %rsp
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i16_stride7_vf16:
@@ -4134,118 +4134,114 @@ define void @load_i16_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 define void @load_i16_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6) nounwind {
 ; SSE-LABEL: load_i16_stride7_vf32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $600, %rsp # imm = 0x258
-; SSE-NEXT:    movdqa 304(%rdi), %xmm5
-; SSE-NEXT:    movdqa 288(%rdi), %xmm6
-; SSE-NEXT:    movdqa 112(%rdi), %xmm13
-; SSE-NEXT:    movdqa 128(%rdi), %xmm8
+; SSE-NEXT:    subq $616, %rsp # imm = 0x268
+; SSE-NEXT:    movdqa 304(%rdi), %xmm8
 ; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 160(%rdi), %xmm7
-; SSE-NEXT:    movaps 144(%rdi), %xmm10
-; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 192(%rdi), %xmm9
-; SSE-NEXT:    movdqa 176(%rdi), %xmm12
-; SSE-NEXT:    movdqa 208(%rdi), %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm1, %xmm11
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 288(%rdi), %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 112(%rdi), %xmm14
+; SSE-NEXT:    movdqa 128(%rdi), %xmm13
+; SSE-NEXT:    movaps 160(%rdi), %xmm4
+; SSE-NEXT:    movaps 144(%rdi), %xmm5
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 192(%rdi), %xmm12
+; SSE-NEXT:    movdqa 176(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 208(%rdi), %xmm9
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[0,0,0,0]
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,65535,65535,65535,65535,0]
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
+; SSE-NEXT:    movdqa %xmm3, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
 ; SSE-NEXT:    pand %xmm2, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [65535,65535,65535,65535,65535,0,0,0]
-; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,65535,65535,65535,0,0,0]
+; SSE-NEXT:    movdqa %xmm10, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movaps %xmm10, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm7[2,2]
-; SSE-NEXT:    movaps %xmm7, %xmm10
-; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{.*#+}} xmm14 = [65535,65535,65535,0,0,65535,65535,65535]
-; SSE-NEXT:    movaps %xmm14, %xmm3
-; SSE-NEXT:    andnps %xmm0, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,2,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm13[0,1,0,3]
-; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
+; SSE-NEXT:    movaps %xmm5, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm4[2,2]
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{.*#+}} xmm15 = [65535,65535,65535,0,0,65535,65535,65535]
+; SSE-NEXT:    movaps %xmm15, %xmm4
+; SSE-NEXT:    andnps %xmm0, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm14[0,1,0,3]
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm0[2],xmm5[3],xmm0[3]
 ; SSE-NEXT:    movdqa 320(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm14, %xmm4
-; SSE-NEXT:    por %xmm3, %xmm4
-; SSE-NEXT:    pand %xmm15, %xmm4
-; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm15, %xmm5
+; SSE-NEXT:    por %xmm4, %xmm5
+; SSE-NEXT:    pand %xmm10, %xmm5
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    movdqa %xmm6, %xmm7
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
-; SSE-NEXT:    movdqa %xmm5, %xmm6
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
 ; SSE-NEXT:    pand %xmm2, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    movdqa %xmm10, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movaps 272(%rdi), %xmm3
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 272(%rdi), %xmm4
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps 256(%rdi), %xmm0
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm3[2,2]
-; SSE-NEXT:    movaps %xmm14, %xmm3
-; SSE-NEXT:    andnps %xmm0, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm4[2,2]
+; SSE-NEXT:    movaps %xmm15, %xmm4
+; SSE-NEXT:    andnps %xmm0, %xmm4
 ; SSE-NEXT:    movdqa 224(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    movdqa 240(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT:    pand %xmm14, %xmm4
-; SSE-NEXT:    por %xmm3, %xmm4
-; SSE-NEXT:    pand %xmm15, %xmm4
-; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm0[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    movdqa 240(%rdi), %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; SSE-NEXT:    pand %xmm15, %xmm5
+; SSE-NEXT:    por %xmm4, %xmm5
+; SSE-NEXT:    pand %xmm10, %xmm5
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 432(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa 416(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 416(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 400(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
 ; SSE-NEXT:    pand %xmm2, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    movdqa %xmm10, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movaps 384(%rdi), %xmm3
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 384(%rdi), %xmm4
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps 368(%rdi), %xmm0
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm3[2,2]
-; SSE-NEXT:    movaps %xmm14, %xmm3
-; SSE-NEXT:    andnps %xmm0, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm4[2,2]
+; SSE-NEXT:    movaps %xmm15, %xmm4
+; SSE-NEXT:    andnps %xmm0, %xmm4
 ; SSE-NEXT:    movdqa 336(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    movdqa 352(%rdi), %xmm8
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT:    pand %xmm14, %xmm4
-; SSE-NEXT:    por %xmm3, %xmm4
-; SSE-NEXT:    pand %xmm15, %xmm4
-; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm0[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    movdqa 352(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; SSE-NEXT:    pand %xmm15, %xmm5
+; SSE-NEXT:    por %xmm4, %xmm5
+; SSE-NEXT:    pand %xmm10, %xmm5
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 96(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
@@ -4258,255 +4254,256 @@ define void @load_i16_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
 ; SSE-NEXT:    pand %xmm2, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    movdqa %xmm10, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    movaps 32(%rdi), %xmm0
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 48(%rdi), %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm4[2,2]
-; SSE-NEXT:    movaps %xmm14, %xmm3
-; SSE-NEXT:    andnps %xmm0, %xmm3
+; SSE-NEXT:    movaps 48(%rdi), %xmm5
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm5[2,2]
+; SSE-NEXT:    movaps %xmm15, %xmm4
+; SSE-NEXT:    andnps %xmm0, %xmm4
 ; SSE-NEXT:    movdqa (%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    movdqa 16(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT:    pand %xmm14, %xmm4
-; SSE-NEXT:    por %xmm3, %xmm4
-; SSE-NEXT:    pand %xmm15, %xmm4
-; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; SSE-NEXT:    pand %xmm15, %xmm5
+; SSE-NEXT:    por %xmm4, %xmm5
+; SSE-NEXT:    pand %xmm10, %xmm5
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pslldq {{.*#+}} xmm9 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm9[0,1,2,3,4,5]
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm0
-; SSE-NEXT:    psrld $16, %xmm0
-; SSE-NEXT:    psrldq {{.*#+}} xmm12 = xmm12[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
-; SSE-NEXT:    pand %xmm2, %xmm12
-; SSE-NEXT:    por %xmm1, %xmm12
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pandn %xmm12, %xmm0
+; SSE-NEXT:    pandn %xmm9, %xmm1
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    psrld $16, %xmm12
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm9, %xmm4
+; SSE-NEXT:    psrldq {{.*#+}} xmm4 = xmm4[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm12[0],xmm4[1],xmm12[1]
+; SSE-NEXT:    pand %xmm2, %xmm4
+; SSE-NEXT:    por %xmm1, %xmm4
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    pandn %xmm4, %xmm0
 ; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,0,65535,65535,65535,65535,65535,65535]
 ; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    pandn %xmm13, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm13, %xmm4
-; SSE-NEXT:    pand %xmm11, %xmm4
-; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm12, %xmm1
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm10[4],xmm1[5],xmm10[5],xmm1[6],xmm10[6],xmm1[7],xmm10[7]
+; SSE-NEXT:    pandn %xmm14, %xmm1
+; SSE-NEXT:    movdqa %xmm13, %xmm12
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm13, %xmm5
+; SSE-NEXT:    pand %xmm11, %xmm5
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    movdqa %xmm14, %xmm3
+; SSE-NEXT:    movdqa %xmm15, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm14, %xmm1
+; SSE-NEXT:    pand %xmm15, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    pand %xmm15, %xmm3
+; SSE-NEXT:    pand %xmm10, %xmm3
 ; SSE-NEXT:    por %xmm0, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm6, %xmm0
 ; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    psrld $16, %xmm6
-; SSE-NEXT:    psrldq {{.*#+}} xmm7 = xmm7[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
-; SSE-NEXT:    pand %xmm2, %xmm7
-; SSE-NEXT:    por %xmm1, %xmm7
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pandn %xmm7, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    psrld $16, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm13, %xmm5
+; SSE-NEXT:    psrldq {{.*#+}} xmm5 = xmm5[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
+; SSE-NEXT:    pand %xmm2, %xmm5
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    pandn %xmm5, %xmm0
 ; SSE-NEXT:    movdqa %xmm11, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pand %xmm11, %xmm4
-; SSE-NEXT:    por %xmm1, %xmm4
+; SSE-NEXT:    pand %xmm11, %xmm7
+; SSE-NEXT:    por %xmm1, %xmm7
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    movdqa %xmm14, %xmm3
+; SSE-NEXT:    movdqa %xmm15, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm14, %xmm1
+; SSE-NEXT:    pand %xmm15, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    pand %xmm15, %xmm3
+; SSE-NEXT:    pand %xmm10, %xmm3
 ; SSE-NEXT:    por %xmm0, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    psrld $16, %xmm0
-; SSE-NEXT:    movdqa (%rsp), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm7, %xmm4
-; SSE-NEXT:    psrldq {{.*#+}} xmm4 = xmm4[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
-; SSE-NEXT:    pand %xmm2, %xmm4
-; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pandn %xmm4, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    psrldq {{.*#+}} xmm5 = xmm5[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
+; SSE-NEXT:    pand %xmm2, %xmm5
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    pandn %xmm5, %xmm0
 ; SSE-NEXT:    movdqa %xmm11, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm8, %xmm4
-; SSE-NEXT:    pand %xmm11, %xmm4
-; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    pand %xmm11, %xmm5
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm7[4],xmm1[5],xmm7[5],xmm1[6],xmm7[6],xmm1[7],xmm7[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    movdqa %xmm14, %xmm8
-; SSE-NEXT:    pandn %xmm1, %xmm8
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,3,2,3]
+; SSE-NEXT:    movdqa %xmm15, %xmm3
+; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm14, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm8
-; SSE-NEXT:    pand %xmm15, %xmm8
-; SSE-NEXT:    por %xmm0, %xmm8
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm15, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    pand %xmm10, %xmm3
+; SSE-NEXT:    por %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5]
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    psrld $16, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm10, %xmm4
-; SSE-NEXT:    psrldq {{.*#+}} xmm4 = xmm4[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
-; SSE-NEXT:    pand %xmm2, %xmm4
-; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pandn %xmm4, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    psrldq {{.*#+}} xmm5 = xmm5[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
+; SSE-NEXT:    pand %xmm2, %xmm5
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    pandn %xmm5, %xmm0
 ; SSE-NEXT:    movdqa %xmm11, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pand %xmm11, %xmm4
-; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    pand %xmm11, %xmm5
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm14, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,1,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,0,0,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pandn %xmm4, %xmm14
-; SSE-NEXT:    por %xmm1, %xmm14
-; SSE-NEXT:    pand %xmm15, %xmm14
-; SSE-NEXT:    por %xmm0, %xmm14
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm15, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm14, %xmm5
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = xmm5[4],mem[4],xmm5[5],mem[5],xmm5[6],mem[6],xmm5[7],mem[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,1,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pandn %xmm5, %xmm15
+; SSE-NEXT:    por %xmm1, %xmm15
+; SSE-NEXT:    pand %xmm10, %xmm15
+; SSE-NEXT:    por %xmm0, %xmm15
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[0,1,0,1]
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm9[2],xmm0[3],xmm9[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = xmm9[2],mem[2],xmm9[3],mem[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[0,1,0,1]
 ; SSE-NEXT:    pand %xmm2, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    movdqa %xmm10, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[0,1,0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm0[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,2,3,3]
 ; SSE-NEXT:    punpckhqdq {{.*#+}} xmm8 = xmm8[1],xmm0[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm13[0],xmm6[1],xmm13[1],xmm6[2],xmm13[2],xmm6[3],xmm13[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm6[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    movss {{.*#+}} xmm8 = xmm4[0],xmm8[1,2,3]
-; SSE-NEXT:    andps %xmm15, %xmm8
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm12[0],xmm3[1],xmm12[1],xmm3[2],xmm12[2],xmm3[3],xmm12[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    movss {{.*#+}} xmm8 = xmm5[0],xmm8[1,2,3]
+; SSE-NEXT:    andps %xmm10, %xmm8
 ; SSE-NEXT:    orps %xmm1, %xmm8
 ; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[0,1,0,1]
-; SSE-NEXT:    movdqa %xmm2, %xmm4
-; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm13[2],xmm1[3],xmm13[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[0,1,0,1]
+; SSE-NEXT:    movdqa %xmm2, %xmm5
+; SSE-NEXT:    pandn %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; SSE-NEXT:    pand %xmm2, %xmm1
-; SSE-NEXT:    por %xmm4, %xmm1
-; SSE-NEXT:    movdqa %xmm15, %xmm4
-; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm1[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    por %xmm5, %xmm1
+; SSE-NEXT:    movdqa %xmm10, %xmm5
+; SSE-NEXT:    pandn %xmm1, %xmm5
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[2,2,3,3]
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm14 = xmm14[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm1[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[2,2,3,3]
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm13 = xmm13[1],xmm1[1]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    movss {{.*#+}} xmm14 = xmm5[0],xmm14[1,2,3]
-; SSE-NEXT:    andps %xmm15, %xmm14
-; SSE-NEXT:    orps %xmm4, %xmm14
-; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[0,1,0,1]
-; SSE-NEXT:    movdqa %xmm2, %xmm5
-; SSE-NEXT:    pandn %xmm4, %xmm5
-; SSE-NEXT:    movdqa %xmm7, %xmm4
-; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,0,1]
-; SSE-NEXT:    pand %xmm2, %xmm4
-; SSE-NEXT:    por %xmm5, %xmm4
-; SSE-NEXT:    movdqa %xmm15, %xmm7
-; SSE-NEXT:    pandn %xmm4, %xmm7
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm4[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[2,2,3,3]
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm4[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    movss {{.*#+}} xmm13 = xmm6[0],xmm13[1,2,3]
+; SSE-NEXT:    andps %xmm10, %xmm13
+; SSE-NEXT:    orps %xmm5, %xmm13
+; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $68, (%rsp), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = mem[0,1,0,1]
+; SSE-NEXT:    movdqa %xmm2, %xmm6
+; SSE-NEXT:    pandn %xmm5, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm15, %xmm5
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,0,1]
+; SSE-NEXT:    pand %xmm2, %xmm5
+; SSE-NEXT:    por %xmm6, %xmm5
+; SSE-NEXT:    movdqa %xmm10, %xmm6
+; SSE-NEXT:    pandn %xmm5, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm13[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm5[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm7[2,2,3,3]
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm5[1]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm5 = xmm5[0],mem[0],xmm5[1],mem[1],xmm5[2],mem[2],xmm5[3],mem[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    movss {{.*#+}} xmm3 = xmm4[0],xmm3[1,2,3]
-; SSE-NEXT:    andps %xmm15, %xmm3
-; SSE-NEXT:    orps %xmm7, %xmm3
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm4
-; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,0,1]
-; SSE-NEXT:    pand %xmm2, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm5[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm7[0],xmm0[1,2,3]
+; SSE-NEXT:    andps %xmm10, %xmm0
+; SSE-NEXT:    orps %xmm6, %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm6 = xmm6[2],mem[2],xmm6[3],mem[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,0,1]
+; SSE-NEXT:    pand %xmm2, %xmm6
 ; SSE-NEXT:    pshufd $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm7 = mem[0,1,0,1]
 ; SSE-NEXT:    pandn %xmm7, %xmm2
-; SSE-NEXT:    por %xmm4, %xmm2
-; SSE-NEXT:    movdqa %xmm15, %xmm4
-; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,1,0,3]
+; SSE-NEXT:    por %xmm6, %xmm2
+; SSE-NEXT:    movdqa %xmm10, %xmm6
+; SSE-NEXT:    pandn %xmm2, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm2[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[2,2,3,3]
 ; SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm2[1]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
@@ -4514,62 +4511,62 @@ define void @load_i16_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm2[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm7[0],xmm0[1,2,3]
-; SSE-NEXT:    andps %xmm15, %xmm0
-; SSE-NEXT:    orps %xmm4, %xmm0
+; SSE-NEXT:    andps %xmm10, %xmm0
+; SSE-NEXT:    orps %xmm6, %xmm0
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm4
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm9, %xmm7
-; SSE-NEXT:    pand %xmm11, %xmm7
-; SSE-NEXT:    por %xmm4, %xmm7
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm7[0,1,1,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm6[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    movdqa %xmm11, %xmm6
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    pand %xmm11, %xmm9
+; SSE-NEXT:    por %xmm6, %xmm9
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm9[0,1,1,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm3[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    movss {{.*#+}} xmm4 = xmm0[0],xmm4[1,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
+; SSE-NEXT:    movss {{.*#+}} xmm6 = xmm0[0],xmm6[1,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,7,7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,4,7]
-; SSE-NEXT:    movdqa %xmm15, %xmm9
-; SSE-NEXT:    pandn %xmm0, %xmm9
-; SSE-NEXT:    andps %xmm15, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm9
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm10, %xmm7
+; SSE-NEXT:    pandn %xmm0, %xmm7
+; SSE-NEXT:    andps %xmm10, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm7
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    pandn %xmm8, %xmm0
-; SSE-NEXT:    movdqa %xmm12, %xmm4
-; SSE-NEXT:    pand %xmm11, %xmm4
-; SSE-NEXT:    por %xmm0, %xmm4
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm4[0,1,1,0,4,5,6,7]
+; SSE-NEXT:    pandn %xmm12, %xmm0
+; SSE-NEXT:    movdqa %xmm4, %xmm6
+; SSE-NEXT:    pand %xmm11, %xmm6
+; SSE-NEXT:    por %xmm0, %xmm6
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm6[0,1,1,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
 ; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm13[4],xmm1[5],xmm13[5],xmm1[6],xmm13[6],xmm1[7],xmm13[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,4,7]
-; SSE-NEXT:    movdqa %xmm15, %xmm4
-; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    andps %xmm15, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm10, %xmm7
+; SSE-NEXT:    pandn %xmm1, %xmm7
+; SSE-NEXT:    andps %xmm10, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm7
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    pandn %xmm13, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    pand %xmm11, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,1,1,0,4,5,6,7]
@@ -4577,312 +4574,315 @@ define void @load_i16_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm5[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
 ; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
-; SSE-NEXT:    movdqa (%rsp), %xmm1 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[4],mem[4],xmm15[5],mem[5],xmm15[6],mem[6],xmm15[7],mem[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,7]
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    punpckhwd (%rsp), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,4,7]
-; SSE-NEXT:    movdqa %xmm15, %xmm4
-; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    andps %xmm15, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    movdqa %xmm14, %xmm1
-; SSE-NEXT:    pand %xmm11, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,1,1,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm2[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
-; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
-; SSE-NEXT:    movdqa %xmm10, %xmm2
-; SSE-NEXT:    movdqa %xmm10, %xmm1
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
-; SSE-NEXT:    andps %xmm15, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,4,7]
-; SSE-NEXT:    pandn %xmm1, %xmm15
-; SSE-NEXT:    por %xmm0, %xmm15
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm9, %xmm0
-; SSE-NEXT:    psrld $16, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,3,2,3]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm1[4],xmm9[5],xmm1[5],xmm9[6],xmm1[6],xmm9[7],xmm1[7]
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    psrlq $48, %xmm1
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    psrlq $16, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm6[2],xmm1[3],xmm6[3]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm7[4],xmm1[5],xmm7[5],xmm1[6],xmm7[6],xmm1[7],xmm7[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    psrld $16, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm1, %xmm6
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm7[4],xmm6[5],xmm7[5],xmm6[6],xmm7[6],xmm6[7],xmm7[7]
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    psrlq $48, %xmm1
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm14, %xmm1
-; SSE-NEXT:    psrlq $16, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1],xmm1[2],xmm10[2],xmm1[3],xmm10[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm13[2],xmm1[3],xmm13[3]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,4,7]
+; SSE-NEXT:    movdqa %xmm10, %xmm5
+; SSE-NEXT:    pandn %xmm1, %xmm5
+; SSE-NEXT:    andps %xmm10, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    pandn %xmm14, %xmm0
+; SSE-NEXT:    pand %xmm11, %xmm8
+; SSE-NEXT:    por %xmm0, %xmm8
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm8[0,1,1,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm2[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
+; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm7
-; SSE-NEXT:    psrld $16, %xmm7
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,7]
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT:    andps %xmm10, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,4,7]
+; SSE-NEXT:    pandn %xmm1, %xmm10
+; SSE-NEXT:    por %xmm0, %xmm10
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm8[4],xmm7[5],xmm8[5],xmm7[6],xmm8[6],xmm7[7],xmm8[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm1, %xmm6
-; SSE-NEXT:    movdqa %xmm8, %xmm1
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm8[4],xmm6[5],xmm8[5],xmm6[6],xmm8[6],xmm6[7],xmm8[7]
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm8, %xmm0
+; SSE-NEXT:    psrld $16, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,3,2,3]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm1[4],xmm8[5],xmm1[5],xmm8[6],xmm1[6],xmm8[7],xmm1[7]
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    psrlq $48, %xmm1
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm13, %xmm1
 ; SSE-NEXT:    psrlq $16, %xmm1
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1],xmm1[2],xmm12[2],xmm1[3],xmm12[3]
+; SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd $196, (%rsp), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[0,1,0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm8[2],xmm1[3],xmm8[3]
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movdqa %xmm3, %xmm15
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    psrld $16, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm6
-; SSE-NEXT:    psrld $16, %xmm6
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm9[4],xmm6[5],xmm9[5],xmm6[6],xmm9[6],xmm6[7],xmm9[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm1, %xmm12
-; SSE-NEXT:    movdqa %xmm9, %xmm1
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm12 = xmm12[4],xmm9[4],xmm12[5],xmm9[5],xmm12[6],xmm9[6],xmm12[7],xmm9[7]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,3,2,3]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm12 = xmm12[4],xmm1[4],xmm12[5],xmm1[5],xmm12[6],xmm1[6],xmm12[7],xmm1[7]
 ; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    psrlq $48, %xmm1
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    psrlq $16, %xmm1
-; SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movdqa %xmm2, %xmm12
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,1,0,3]
+; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,4,7]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm5, %xmm7
+; SSE-NEXT:    psrld $16, %xmm7
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pand %xmm11, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,7]
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1],xmm0[2],xmm10[2],xmm0[3],xmm10[3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm1[4],xmm7[5],xmm1[5],xmm7[6],xmm1[6],xmm7[7],xmm1[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm5, %xmm2
+; SSE-NEXT:    movdqa %xmm1, %xmm5
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    psrlq $48, %xmm5
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm14, %xmm5
+; SSE-NEXT:    psrlq $16, %xmm5
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm12[0],xmm5[1],xmm12[1],xmm5[2],xmm12[2],xmm5[3],xmm12[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = mem[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
+; SSE-NEXT:    punpckhwd (%rsp), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = xmm5[4],mem[4],xmm5[5],mem[5],xmm5[6],mem[6],xmm5[7],mem[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[3,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm5[0,1,2,3,5,6,4,7]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pand %xmm11, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,7]
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    psrld $16, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[2,3,2,3]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    psrlq $48, %xmm5
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    psrlq $16, %xmm5
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1],xmm0[2],xmm10[2],xmm0[3],xmm10[3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    pandn (%rsp), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    pand %xmm11, %xmm8
-; SSE-NEXT:    por %xmm0, %xmm8
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm0[0,2]
-; SSE-NEXT:    pand %xmm11, %xmm2
-; SSE-NEXT:    pandn %xmm12, %xmm11
-; SSE-NEXT:    por %xmm2, %xmm11
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = xmm5[4],mem[4],xmm5[5],mem[5],xmm5[6],mem[6],xmm5[7],mem[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[3,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm5[0,1,2,3,5,6,4,7]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm6[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm13, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm13[0],xmm6[1],xmm13[1],xmm6[2],xmm13[2],xmm6[3],xmm13[3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
+; SSE-NEXT:    movdqa %xmm11, %xmm6
+; SSE-NEXT:    pandn %xmm9, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm5[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm11, %xmm5
+; SSE-NEXT:    por %xmm6, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,5,4,7,7]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm15[4],xmm6[5],xmm15[5],xmm6[6],xmm15[6],xmm6[7],xmm15[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm6[0,2]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm5, %xmm6
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm8[0],xmm6[1],xmm8[1],xmm6[2],xmm8[2],xmm6[3],xmm8[3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
+; SSE-NEXT:    movdqa %xmm11, %xmm6
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,7]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
-; SSE-NEXT:    movdqa %xmm3, %xmm11
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[0,2]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
+; SSE-NEXT:    pandn %xmm8, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm9[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm11, %xmm9
+; SSE-NEXT:    por %xmm6, %xmm9
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm9[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,5,4,7,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm15[4],xmm6[5],xmm15[5],xmm6[6],xmm15[6],xmm6[7],xmm15[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm6[0,2]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm12[0],xmm14[1],xmm12[1],xmm14[2],xmm12[2],xmm14[3],xmm12[3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm14[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
+; SSE-NEXT:    movdqa %xmm11, %xmm9
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm0, %xmm9
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm14[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm11, %xmm14
+; SSE-NEXT:    por %xmm9, %xmm14
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm14[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,5,4,7,7]
+; SSE-NEXT:    movdqa (%rsp), %xmm6 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm6[4],xmm9[5],xmm6[5],xmm9[6],xmm6[6],xmm9[7],xmm6[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm9[0,2]
+; SSE-NEXT:    movdqa %xmm4, %xmm9
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm11, %xmm9
+; SSE-NEXT:    movdqa %xmm9, %xmm14
+; SSE-NEXT:    pandn %xmm1, %xmm11
+; SSE-NEXT:    movdqa %xmm1, %xmm9
+; SSE-NEXT:    por %xmm14, %xmm11
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    psrldq {{.*#+}} xmm1 = xmm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
-; SSE-NEXT:    pshufd $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshuflw $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
-; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[0,1,0,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm2[2,2,2,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm0[2],xmm5[3],xmm0[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm1, %xmm14
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm3[0],xmm14[1],xmm3[1],xmm14[2],xmm3[2],xmm14[3],xmm3[3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm14[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm14[0],xmm2[1],xmm14[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm11[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,4,7,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm11[4],xmm4[5],xmm11[5],xmm4[6],xmm11[6],xmm4[7],xmm11[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm4[0,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm13[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
 ; SSE-NEXT:    psrldq {{.*#+}} xmm14 = xmm14[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1],xmm14[2],xmm0[2],xmm14[3],xmm0[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[1,1,1,1]
-; SSE-NEXT:    pshufd $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE-NEXT:    pshuflw $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,1,0,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[2,2,2,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm2[2],xmm4[3],xmm2[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
-; SSE-NEXT:    psrldq {{.*#+}} xmm13 = xmm13[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1],xmm13[2],xmm0[2],xmm13[3],xmm0[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1]
-; SSE-NEXT:    pshufd $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    pshuflw $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    movdqa (%rsp), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,1,0,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,2,2,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm2[0],xmm3[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[1,1,1,1]
-; SSE-NEXT:    psrldq {{.*#+}} xmm12 = xmm12[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm1[0],xmm12[1],xmm1[1],xmm12[2],xmm1[2],xmm12[3],xmm1[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[1,1,1,1]
-; SSE-NEXT:    pshufd $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    pshuflw $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[0,1,0,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm4[0],xmm14[1],xmm4[1],xmm14[2],xmm4[2],xmm14[3],xmm4[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm14[1,1,1,1]
+; SSE-NEXT:    pshufd $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm14[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm4[0],xmm14[1],xmm4[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = xmm4[0],mem[0],xmm4[1],mem[1],xmm4[2],mem[2],xmm4[3],mem[3]
+; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm13 = mem[0,1,0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,2,2,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm13[2],xmm4[3],xmm13[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm14[0],xmm4[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm13 = mem[1,1,1,1]
+; SSE-NEXT:    psrldq {{.*#+}} xmm5 = xmm5[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm13[0],xmm5[1],xmm13[1],xmm5[2],xmm13[2],xmm5[3],xmm13[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm5[1,1,1,1]
+; SSE-NEXT:    pshufd $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm14[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm13[0],xmm14[1],xmm13[1]
+; SSE-NEXT:    movdqa %xmm8, %xmm13
+; SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm13 = xmm13[0],mem[0],xmm13[1],mem[1],xmm13[2],mem[2],xmm13[3],mem[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm15[0,1,0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm13[2,2,2,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm13 = xmm13[2],xmm8[2],xmm13[3],xmm8[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm14[0],xmm13[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm12[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    psrldq {{.*#+}} xmm5 = xmm5[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm8[0],xmm5[1],xmm8[1],xmm5[2],xmm8[2],xmm5[3],xmm8[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm5[1,1,1,1]
+; SSE-NEXT:    pshufd $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm14[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm8[0],xmm14[1],xmm8[1]
+; SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm6[0,1,0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm0[2,2,2,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm8[2],xmm6[3],xmm8[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm14[0],xmm6[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm3[1,1,1,1]
+; SSE-NEXT:    psrldq {{.*#+}} xmm1 = xmm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm1[1,1,1,1]
+; SSE-NEXT:    pshufd $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm14[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm8[0],xmm14[1],xmm8[1]
+; SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = xmm9[0],mem[0],xmm9[1],mem[1],xmm9[2],mem[2],xmm9[3],mem[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm11[0,1,0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[2,2,2,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm14[0],xmm9[1]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm1, (%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 48(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 32(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm5, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm5, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm8, 16(%rsi)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm1, (%rdx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -4899,34 +4899,34 @@ define void @load_i16_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movaps %xmm1, 32(%rcx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm1, 16(%rcx)
-; SSE-NEXT:    movdqa %xmm15, (%r8)
+; SSE-NEXT:    movdqa %xmm10, (%r8)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm1, 48(%r8)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm1, 32(%r8)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm1, 16(%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, (%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 48(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 32(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 16(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r9)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movaps %xmm6, (%rax)
+; SSE-NEXT:    movaps %xmm2, (%rax)
 ; SSE-NEXT:    movaps %xmm7, 48(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 32(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 16(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movapd %xmm0, (%rax)
-; SSE-NEXT:    movapd %xmm3, 48(%rax)
-; SSE-NEXT:    movapd %xmm4, 32(%rax)
-; SSE-NEXT:    movapd %xmm5, 16(%rax)
-; SSE-NEXT:    addq $600, %rsp # imm = 0x258
+; SSE-NEXT:    movapd %xmm9, (%rax)
+; SSE-NEXT:    movapd %xmm6, 48(%rax)
+; SSE-NEXT:    movapd %xmm13, 32(%rax)
+; SSE-NEXT:    movapd %xmm4, 16(%rax)
+; SSE-NEXT:    addq $616, %rsp # imm = 0x268
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i16_stride7_vf32:
@@ -8606,21 +8606,22 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-LABEL: load_i16_stride7_vf64:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    subq $1352, %rsp # imm = 0x548
-; SSE-NEXT:    movdqa 640(%rdi), %xmm9
-; SSE-NEXT:    movdqa 624(%rdi), %xmm12
-; SSE-NEXT:    movdqa %xmm12, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa 112(%rdi), %xmm8
+; SSE-NEXT:    movdqa 640(%rdi), %xmm11
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 624(%rdi), %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 112(%rdi), %xmm5
 ; SSE-NEXT:    movdqa 128(%rdi), %xmm10
-; SSE-NEXT:    movaps 160(%rdi), %xmm6
-; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 160(%rdi), %xmm7
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps 144(%rdi), %xmm13
 ; SSE-NEXT:    movdqa 192(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 176(%rdi), %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 208(%rdi), %xmm11
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 208(%rdi), %xmm12
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[0,0,0,0]
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,65535,65535,65535,65535,0]
 ; SSE-NEXT:    movdqa %xmm3, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
@@ -8633,19 +8634,19 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    movaps %xmm13, %xmm0
 ; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm6[2,2]
-; SSE-NEXT:    movaps {{.*#+}} xmm6 = [65535,65535,65535,0,0,65535,65535,65535]
-; SSE-NEXT:    movaps %xmm6, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm7[2,2]
+; SSE-NEXT:    movaps {{.*#+}} xmm9 = [65535,65535,65535,0,0,65535,65535,65535]
+; SSE-NEXT:    movaps %xmm9, %xmm2
 ; SSE-NEXT:    andnps %xmm0, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,2,3,3]
 ; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm8[0,1,0,3]
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[0,1,0,3]
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
 ; SSE-NEXT:    movdqa 656(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm6, %xmm4
+; SSE-NEXT:    pand %xmm9, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    pand %xmm14, %xmm4
 ; SSE-NEXT:    por %xmm1, %xmm4
@@ -8653,9 +8654,8 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    movdqa %xmm3, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
 ; SSE-NEXT:    pand %xmm3, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    movdqa %xmm14, %xmm1
@@ -8665,17 +8665,17 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movaps 592(%rdi), %xmm0
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm2[2,2]
-; SSE-NEXT:    movaps %xmm6, %xmm2
+; SSE-NEXT:    movaps %xmm9, %xmm2
 ; SSE-NEXT:    andnps %xmm0, %xmm2
-; SSE-NEXT:    movdqa 560(%rdi), %xmm15
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[0,1,0,3]
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 560(%rdi), %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[0,1,0,3]
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    movdqa 576(%rdi), %xmm5
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 576(%rdi), %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT:    pand %xmm6, %xmm4
+; SSE-NEXT:    pand %xmm9, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    pand %xmm14, %xmm4
 ; SSE-NEXT:    por %xmm1, %xmm4
@@ -8699,17 +8699,17 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movaps 48(%rdi), %xmm4
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm4[2,2]
-; SSE-NEXT:    movaps %xmm6, %xmm2
+; SSE-NEXT:    movaps %xmm9, %xmm2
 ; SSE-NEXT:    andnps %xmm0, %xmm2
 ; SSE-NEXT:    movdqa (%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    movdqa 16(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
+; SSE-NEXT:    movdqa 16(%rdi), %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT:    pand %xmm6, %xmm4
+; SSE-NEXT:    pand %xmm9, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    pand %xmm14, %xmm4
 ; SSE-NEXT:    por %xmm1, %xmm4
@@ -8729,21 +8729,21 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movdqa %xmm14, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    movaps 496(%rdi), %xmm2
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, (%rsp) # 16-byte Spill
 ; SSE-NEXT:    movaps 480(%rdi), %xmm0
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm2[2,2]
-; SSE-NEXT:    movaps %xmm6, %xmm2
+; SSE-NEXT:    movaps %xmm9, %xmm2
 ; SSE-NEXT:    andnps %xmm0, %xmm2
 ; SSE-NEXT:    movdqa 448(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    movdqa 464(%rdi), %xmm12
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 464(%rdi), %xmm15
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT:    pand %xmm6, %xmm4
+; SSE-NEXT:    pand %xmm9, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    pand %xmm14, %xmm4
 ; SSE-NEXT:    por %xmm1, %xmm4
@@ -8767,7 +8767,7 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movaps 368(%rdi), %xmm0
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm2[2,2]
-; SSE-NEXT:    movaps %xmm6, %xmm2
+; SSE-NEXT:    movaps %xmm9, %xmm2
 ; SSE-NEXT:    andnps %xmm0, %xmm2
 ; SSE-NEXT:    movdqa 336(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -8777,7 +8777,7 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT:    pand %xmm6, %xmm4
+; SSE-NEXT:    pand %xmm9, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    pand %xmm14, %xmm4
 ; SSE-NEXT:    por %xmm1, %xmm4
@@ -8787,11 +8787,11 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    movdqa %xmm3, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa 864(%rdi), %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 864(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 848(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
 ; SSE-NEXT:    pand %xmm3, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    movdqa %xmm14, %xmm1
@@ -8801,7 +8801,7 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movaps 816(%rdi), %xmm0
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm2[2,2]
-; SSE-NEXT:    movaps %xmm6, %xmm2
+; SSE-NEXT:    movaps %xmm9, %xmm2
 ; SSE-NEXT:    andnps %xmm0, %xmm2
 ; SSE-NEXT:    movdqa 784(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -8811,7 +8811,7 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT:    pand %xmm6, %xmm4
+; SSE-NEXT:    pand %xmm9, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    pand %xmm14, %xmm4
 ; SSE-NEXT:    por %xmm1, %xmm4
@@ -8835,7 +8835,7 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movaps 256(%rdi), %xmm0
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm2[2,2]
-; SSE-NEXT:    movaps %xmm6, %xmm2
+; SSE-NEXT:    movaps %xmm9, %xmm2
 ; SSE-NEXT:    andnps %xmm0, %xmm2
 ; SSE-NEXT:    movdqa 224(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -8845,7 +8845,7 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT:    pand %xmm6, %xmm4
+; SSE-NEXT:    pand %xmm9, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    pand %xmm14, %xmm4
 ; SSE-NEXT:    por %xmm1, %xmm4
@@ -8855,11 +8855,11 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    movdqa %xmm3, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa 752(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 752(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 736(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
 ; SSE-NEXT:    pand %xmm3, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    movdqa %xmm14, %xmm1
@@ -8869,7 +8869,7 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movaps 704(%rdi), %xmm0
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm2[2,2]
-; SSE-NEXT:    movaps %xmm6, %xmm2
+; SSE-NEXT:    movaps %xmm9, %xmm2
 ; SSE-NEXT:    andnps %xmm0, %xmm2
 ; SSE-NEXT:    movdqa 672(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -8879,14 +8879,14 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT:    pand %xmm6, %xmm4
+; SSE-NEXT:    pand %xmm9, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    pand %xmm14, %xmm4
 ; SSE-NEXT:    por %xmm1, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pslldq {{.*#+}} xmm11 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm11[0,1,2,3,4,5]
+; SSE-NEXT:    pslldq {{.*#+}} xmm12 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm12[0,1,2,3,4,5]
 ; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    pandn %xmm11, %xmm1
+; SSE-NEXT:    pandn %xmm12, %xmm1
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    psrld $16, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
@@ -8896,21 +8896,21 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm14, %xmm1
 ; SSE-NEXT:    pandn %xmm2, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    pandn %xmm8, %xmm2
-; SSE-NEXT:    pand %xmm7, %xmm10
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,0,65535,65535,65535,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm8, %xmm2
+; SSE-NEXT:    pandn %xmm5, %xmm2
+; SSE-NEXT:    pand %xmm8, %xmm10
 ; SSE-NEXT:    por %xmm2, %xmm10
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm13 = xmm13[4],mem[4],xmm13[5],mem[5],xmm13[6],mem[6],xmm13[7],mem[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm13[2,1,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm9, %xmm0
 ; SSE-NEXT:    pandn %xmm2, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm10[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm6, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm0
 ; SSE-NEXT:    pand %xmm14, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
@@ -8919,29 +8919,30 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,1,2,3,4,5]
 ; SSE-NEXT:    movdqa %xmm3, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    psrld $16, %xmm9
-; SSE-NEXT:    movdqa (%rsp), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    psrld $16, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
 ; SSE-NEXT:    psrldq {{.*#+}} xmm4 = xmm4[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
 ; SSE-NEXT:    pand %xmm3, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    movdqa %xmm14, %xmm1
 ; SSE-NEXT:    pandn %xmm4, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    pandn %xmm15, %xmm2
-; SSE-NEXT:    pand %xmm7, %xmm5
-; SSE-NEXT:    por %xmm2, %xmm5
+; SSE-NEXT:    movdqa %xmm8, %xmm2
+; SSE-NEXT:    pandn %xmm7, %xmm2
+; SSE-NEXT:    pand %xmm8, %xmm6
+; SSE-NEXT:    por %xmm2, %xmm6
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm9, %xmm0
 ; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[0,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm6[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm6, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm0
 ; SSE-NEXT:    pand %xmm14, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
@@ -8959,22 +8960,21 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    movdqa %xmm14, %xmm1
 ; SSE-NEXT:    pandn %xmm4, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    movdqa %xmm8, %xmm2
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pand %xmm7, %xmm4
-; SSE-NEXT:    por %xmm2, %xmm4
+; SSE-NEXT:    pand %xmm8, %xmm11
+; SSE-NEXT:    por %xmm2, %xmm11
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm9, %xmm0
 ; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm11[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm6, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm0
 ; SSE-NEXT:    pand %xmm14, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
@@ -8992,21 +8992,21 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    movdqa %xmm14, %xmm1
 ; SSE-NEXT:    pandn %xmm4, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    movdqa %xmm8, %xmm2
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    pand %xmm7, %xmm12
-; SSE-NEXT:    por %xmm2, %xmm12
+; SSE-NEXT:    pand %xmm8, %xmm15
+; SSE-NEXT:    por %xmm2, %xmm15
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    punpckhwd (%rsp), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm9, %xmm0
 ; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm12[0,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm15[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm6, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm0
 ; SSE-NEXT:    pand %xmm14, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
@@ -9024,22 +9024,23 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    movdqa %xmm14, %xmm1
 ; SSE-NEXT:    pandn %xmm4, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    movdqa %xmm8, %xmm2
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pand %xmm7, %xmm4
+; SSE-NEXT:    pand %xmm8, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm13, %xmm2
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm9, %xmm0
 ; SSE-NEXT:    pandn %xmm2, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm6, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm0
 ; SSE-NEXT:    pand %xmm14, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
@@ -9057,10 +9058,10 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    movdqa %xmm14, %xmm1
 ; SSE-NEXT:    pandn %xmm4, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    movdqa %xmm8, %xmm2
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pand %xmm7, %xmm4
+; SSE-NEXT:    pand %xmm8, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
@@ -9068,56 +9069,55 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm9, %xmm0
 ; SSE-NEXT:    pandn %xmm2, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm6, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm0
 ; SSE-NEXT:    pand %xmm14, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm7, %xmm1
 ; SSE-NEXT:    pslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,1,2,3,4,5]
 ; SSE-NEXT:    movdqa %xmm3, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm11, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm10, %xmm1
 ; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm9, %xmm4
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm6, %xmm4
 ; SSE-NEXT:    psrldq {{.*#+}} xmm4 = xmm4[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
 ; SSE-NEXT:    pand %xmm3, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    movdqa %xmm14, %xmm1
 ; SSE-NEXT:    pandn %xmm4, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    movdqa %xmm8, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm11, %xmm2
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm12, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm13, %xmm4
-; SSE-NEXT:    pand %xmm7, %xmm4
+; SSE-NEXT:    movdqa %xmm12, %xmm4
+; SSE-NEXT:    pand %xmm8, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm0, %xmm2
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm5, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm8[4],xmm2[5],xmm8[5],xmm2[6],xmm8[6],xmm2[7],xmm8[7]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    movdqa %xmm6, %xmm15
+; SSE-NEXT:    movdqa %xmm9, %xmm15
 ; SSE-NEXT:    pandn %xmm2, %xmm15
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm6, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm15
 ; SSE-NEXT:    pand %xmm14, %xmm15
 ; SSE-NEXT:    por %xmm1, %xmm15
 ; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    pslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,1,2,3,4,5]
 ; SSE-NEXT:    movdqa %xmm3, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
@@ -9130,25 +9130,26 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    movdqa %xmm14, %xmm1
 ; SSE-NEXT:    pandn %xmm4, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    movdqa %xmm8, %xmm2
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pand %xmm7, %xmm4
+; SSE-NEXT:    pand %xmm8, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,3,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm6, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    pand %xmm9, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm15, %xmm4
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,1,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pandn %xmm4, %xmm6
-; SSE-NEXT:    por %xmm2, %xmm6
-; SSE-NEXT:    pand %xmm14, %xmm6
-; SSE-NEXT:    por %xmm1, %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm4, %xmm9
+; SSE-NEXT:    por %xmm2, %xmm9
+; SSE-NEXT:    pand %xmm14, %xmm9
+; SSE-NEXT:    por %xmm1, %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[0,1,0,1]
 ; SSE-NEXT:    movdqa %xmm3, %xmm2
@@ -9189,38 +9190,37 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    por %xmm2, %xmm1
 ; SSE-NEXT:    movdqa %xmm14, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm1[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm1[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm6 = xmm6[1],xmm1[1]
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm9 = xmm9[1],xmm1[1]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    movss {{.*#+}} xmm6 = xmm4[0],xmm6[1,2,3]
-; SSE-NEXT:    andps %xmm14, %xmm6
-; SSE-NEXT:    orps %xmm2, %xmm6
-; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm10[0,1,0,1]
+; SSE-NEXT:    movss {{.*#+}} xmm9 = xmm4[0],xmm9[1,2,3]
+; SSE-NEXT:    andps %xmm14, %xmm9
+; SSE-NEXT:    orps %xmm2, %xmm9
+; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,1,0,1]
 ; SSE-NEXT:    movdqa %xmm3, %xmm4
 ; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm11[2],xmm9[3],xmm11[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm9[0,1,0,1]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm10[2],xmm6[3],xmm10[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm6[0,1,0,1]
 ; SSE-NEXT:    pand %xmm3, %xmm2
 ; SSE-NEXT:    por %xmm4, %xmm2
 ; SSE-NEXT:    movdqa %xmm14, %xmm4
 ; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[0,1,0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm2[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[2,2,3,3]
 ; SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm2[1]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm13[0],xmm12[1],xmm13[1],xmm12[2],xmm13[2],xmm12[3],xmm13[3]
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm12[2,1,2,3]
+; SSE-NEXT:    movdqa %xmm11, %xmm7
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm12[0],xmm7[1],xmm12[1],xmm7[2],xmm12[2],xmm7[3],xmm12[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm2[0],xmm0[1,2,3]
 ; SSE-NEXT:    andps %xmm14, %xmm0
@@ -9230,7 +9230,7 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    # xmm2 = mem[0,1,0,1]
 ; SSE-NEXT:    movdqa %xmm3, %xmm4
 ; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    movdqa (%rsp), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,1]
@@ -9238,16 +9238,17 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    por %xmm4, %xmm2
 ; SSE-NEXT:    movdqa %xmm14, %xmm4
 ; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[0,1,0,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm12[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm2[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = mem[2,2,3,3]
 ; SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm2[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1],xmm2[2],mem[2],xmm2[3],mem[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm2[2,1,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm6[0],xmm0[1,2,3]
 ; SSE-NEXT:    andps %xmm14, %xmm0
@@ -9268,13 +9269,13 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm4 = mem[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd $250, (%rsp), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm4 = mem[2,2,3,3]
 ; SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm4[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = xmm8[0],mem[0],xmm8[1],mem[1],xmm8[2],mem[2],xmm8[3],mem[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm8[2,1,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1],xmm2[2],mem[2],xmm2[3],mem[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm4[0],xmm0[1,2,3]
 ; SSE-NEXT:    andps %xmm14, %xmm0
@@ -9295,8 +9296,8 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm4 = mem[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm12[2,2,3,3]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[2,2,3,3]
 ; SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm4[1]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
 ; SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
@@ -9307,7 +9308,8 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    andps %xmm14, %xmm0
 ; SSE-NEXT:    orps %xmm6, %xmm0
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm15[0,1,0,1]
+; SSE-NEXT:    pshufd $68, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm6 = mem[0,1,0,1]
 ; SSE-NEXT:    movdqa %xmm3, %xmm10
 ; SSE-NEXT:    pandn %xmm6, %xmm10
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
@@ -9318,8 +9320,7 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    por %xmm10, %xmm6
 ; SSE-NEXT:    movdqa %xmm14, %xmm11
 ; SSE-NEXT:    pandn %xmm6, %xmm11
-; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm6 = mem[0,1,0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm15[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm6[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm6 = mem[2,2,3,3]
@@ -9360,10 +9361,10 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    andps %xmm14, %xmm5
 ; SSE-NEXT:    orps %xmm6, %xmm5
 ; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm6
+; SSE-NEXT:    movdqa %xmm8, %xmm6
 ; SSE-NEXT:    pandn %xmm13, %xmm6
 ; SSE-NEXT:    movdqa %xmm15, %xmm11
-; SSE-NEXT:    pand %xmm7, %xmm11
+; SSE-NEXT:    pand %xmm8, %xmm11
 ; SSE-NEXT:    por %xmm6, %xmm11
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm11[0,1,1,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,7,7,7]
@@ -9382,11 +9383,11 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    andps %xmm14, %xmm6
 ; SSE-NEXT:    por %xmm6, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm3
+; SSE-NEXT:    movdqa %xmm8, %xmm3
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm9, %xmm6
-; SSE-NEXT:    pand %xmm7, %xmm6
+; SSE-NEXT:    pand %xmm8, %xmm6
 ; SSE-NEXT:    por %xmm3, %xmm6
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm6[0,1,1,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,7,7,7]
@@ -9410,15 +9411,14 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    andps %xmm14, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm3
+; SSE-NEXT:    movdqa %xmm8, %xmm3
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pand %xmm7, %xmm5
+; SSE-NEXT:    pand %xmm8, %xmm5
 ; SSE-NEXT:    por %xmm3, %xmm5
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm5[0,1,1,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pshufhw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm5 = mem[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm7[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,2,2,2]
 ; SSE-NEXT:    movss {{.*#+}} xmm3 = xmm5[0],xmm3[1,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
@@ -9426,8 +9426,8 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    # xmm5 = xmm5[4],mem[4],xmm5[5],mem[5],xmm5[6],mem[6],xmm5[7],mem[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,7,7]
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm5 = xmm5[4],mem[4],xmm5[5],mem[5],xmm5[6],mem[6],xmm5[7],mem[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm7[4],xmm5[5],xmm7[5],xmm5[6],xmm7[6],xmm5[7],xmm7[7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,6,4,7]
@@ -9436,10 +9436,10 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    andps %xmm14, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm3
+; SSE-NEXT:    movdqa %xmm8, %xmm3
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pand %xmm7, %xmm5
+; SSE-NEXT:    pand %xmm8, %xmm5
 ; SSE-NEXT:    por %xmm3, %xmm5
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm5[0,1,1,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,7,7,7]
@@ -9463,19 +9463,19 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    andps %xmm14, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm13
 ; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    movdqa %xmm8, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pand %xmm7, %xmm3
+; SSE-NEXT:    movdqa (%rsp), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pand %xmm8, %xmm3
 ; SSE-NEXT:    por %xmm1, %xmm3
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm3[0,1,1,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm8[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm2[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,2,2,2]
 ; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm3[0],xmm1[1,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm8[4],xmm3[5],xmm8[5],xmm3[6],xmm8[6],xmm3[7],xmm8[7]
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,7,7]
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
@@ -9483,22 +9483,23 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,6,4,7]
-; SSE-NEXT:    movdqa %xmm14, %xmm13
-; SSE-NEXT:    pandn %xmm3, %xmm13
+; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    pandn %xmm3, %xmm2
 ; SSE-NEXT:    andps %xmm14, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm13
-; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    pandn %xmm12, %xmm1
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pand %xmm7, %xmm3
+; SSE-NEXT:    pand %xmm8, %xmm3
 ; SSE-NEXT:    por %xmm1, %xmm3
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm3[0,1,1,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufhw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,2,2,2]
 ; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3]
-; SSE-NEXT:    movdqa (%rsp), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
@@ -9513,23 +9514,24 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    andps %xmm14, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    movdqa %xmm8, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pand %xmm7, %xmm2
+; SSE-NEXT:    pand %xmm8, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[0,1,1,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm10[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,2,2,2]
 ; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm2
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,7,7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,4,7]
@@ -9538,11 +9540,12 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    andps %xmm14, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm10
 ; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    movdqa %xmm8, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    pand %xmm7, %xmm12
-; SSE-NEXT:    por %xmm1, %xmm12
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm12[0,1,1,0,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pand %xmm8, %xmm2
+; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[0,1,1,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm4[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,2,2,2]
@@ -9641,8 +9644,7 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm7[4],xmm2[5],xmm7[5],xmm2[6],xmm7[6],xmm2[7],xmm7[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
@@ -9691,14 +9693,15 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    psrlq $16, %xmm2
-; SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    punpcklwd (%rsp), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1],xmm2[2],mem[2],xmm2[3],mem[3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
 ; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = mem[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm8[2],xmm2[3],xmm8[3]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,0]
@@ -9723,11 +9726,11 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1],xmm2[2],xmm10[2],xmm2[3],xmm10[3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; SSE-NEXT:    pshufd $196, (%rsp), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = mem[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm15[2],xmm2[3],xmm15[3]
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,0]
@@ -9746,19 +9749,18 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    psrlq $48, %xmm2
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    psrlq $16, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm7[0],xmm2[1],xmm7[1],xmm2[2],xmm7[2],xmm2[3],xmm7[3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[0,1,0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
@@ -9780,15 +9782,15 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm4, %xmm2
 ; SSE-NEXT:    psrlq $16, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm15[0],xmm2[1],xmm15[1],xmm2[2],xmm15[2],xmm2[3],xmm15[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm9[2],xmm2[3],xmm9[3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,0]
@@ -9804,10 +9806,10 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    movdqa %xmm8, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pand %xmm7, %xmm2
+; SSE-NEXT:    pand %xmm8, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,4,7,7]
@@ -9823,10 +9825,10 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    movdqa %xmm8, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pand %xmm7, %xmm2
+; SSE-NEXT:    pand %xmm8, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,4,7,7]
@@ -9842,10 +9844,10 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    movdqa %xmm8, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pand %xmm7, %xmm2
+; SSE-NEXT:    pand %xmm8, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,4,7,7]
@@ -9855,16 +9857,16 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0,2]
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    punpcklwd (%rsp), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    movdqa %xmm8, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pand %xmm7, %xmm2
+; SSE-NEXT:    pand %xmm8, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,4,7,7]
@@ -9878,82 +9880,83 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    pandn (%rsp), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pand %xmm7, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,1,0,3]
+; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    pand %xmm8, %xmm15
+; SSE-NEXT:    por %xmm1, %xmm15
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,4,7,7]
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
 ; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,1],xmm1[0,2]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm9[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm1[0],xmm12[1],xmm1[1]
-; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    movdqa %xmm8, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pand %xmm7, %xmm2
+; SSE-NEXT:    pand %xmm8, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,4,7,7]
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm15[4],xmm1[5],xmm15[5],xmm1[6],xmm15[6],xmm1[7],xmm15[7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
 ; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,1],xmm1[0,2]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm15[0],xmm4[1],xmm15[1],xmm4[2],xmm15[2],xmm4[3],xmm15[3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm4[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm1[0],xmm11[1],xmm1[1]
-; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    movdqa %xmm8, %xmm1
 ; SSE-NEXT:    pandn %xmm5, %xmm1
-; SSE-NEXT:    pand %xmm7, %xmm3
-; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,1,0,3]
+; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    pand %xmm8, %xmm2
+; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,4,7,7]
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
 ; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,1],xmm1[0,2]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    pand %xmm7, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm5, %xmm7
-; SSE-NEXT:    por %xmm1, %xmm7
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pand %xmm8, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm7, %xmm8
+; SSE-NEXT:    por %xmm2, %xmm8
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm4, %xmm1
+; SSE-NEXT:    movdqa %xmm4, %xmm2
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm1[0],xmm8[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[0,1,0,3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm2[0],xmm9[1],xmm2[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,1],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    psrldq {{.*#+}} xmm1 = xmm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
-; SSE-NEXT:    pshufd $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm3[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,1,0,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm5[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    psrldq {{.*#+}} xmm2 = xmm2[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1]
+; SSE-NEXT:    pshufd $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1],xmm7[2],xmm1[2],xmm7[3],xmm1[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[0,1,0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm7[2,2,2,2]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm10 = xmm10[2],xmm0[2],xmm10[3],xmm0[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm1[0],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm2[0],xmm10[1]
 ; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -10014,7 +10017,7 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm0[2],xmm7[3],xmm0[3]
 ; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm1[0],xmm7[1]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd $85, (%rsp), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    psrldq {{.*#+}} xmm1 = xmm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
@@ -10046,7 +10049,7 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; SSE-NEXT:    pshuflw $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    movdqa (%rsp), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
 ; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[0,1,0,3]
@@ -10066,15 +10069,15 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
 ; SSE-NEXT:    pshuflw $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[0,1,0,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm15[0,1,0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,2,2,2]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm2[2],xmm4[3],xmm2[3]
 ; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[1,1,1,1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    psrldq {{.*#+}} xmm1 = xmm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
@@ -10083,8 +10086,7 @@ define void @load_i16_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    # xmm2 = mem[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    pshuflw $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm3[1,1,1,1,4,5,6,7]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; SSE-NEXT:    pshufd $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-8.ll
index 1b637cd203c8f..7d1316f160b48 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i16-stride-8.ll
@@ -1530,197 +1530,190 @@ define void @load_i16_stride8_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i16_stride8_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6, ptr %out.vec7) nounwind {
 ; SSE-LABEL: load_i16_stride8_vf16:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $168, %rsp
-; SSE-NEXT:    movdqa 112(%rdi), %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 96(%rdi), %xmm14
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    subq $184, %rsp
+; SSE-NEXT:    movdqa 112(%rdi), %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 96(%rdi), %xmm13
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 208(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa 192(%rdi), %xmm5
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 192(%rdi), %xmm3
 ; SSE-NEXT:    movdqa 240(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 224(%rdi), %xmm12
-; SSE-NEXT:    movdqa 144(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 128(%rdi), %xmm10
-; SSE-NEXT:    movdqa 176(%rdi), %xmm2
+; SSE-NEXT:    movdqa 224(%rdi), %xmm10
+; SSE-NEXT:    movdqa %xmm10, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movdqa 144(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 160(%rdi), %xmm11
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm2[0],xmm11[1],xmm2[1],xmm11[2],xmm2[2],xmm11[3],xmm2[3]
-; SSE-NEXT:    movdqa %xmm10, %xmm2
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
-; SSE-NEXT:    movdqa %xmm12, %xmm4
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    movdqa 128(%rdi), %xmm6
+; SSE-NEXT:    movdqa 176(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 160(%rdi), %xmm12
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm4[0],xmm12[1],xmm4[1],xmm12[2],xmm4[2],xmm12[3],xmm4[3]
+; SSE-NEXT:    movdqa %xmm6, %xmm4
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm1[0],xmm10[1],xmm1[1],xmm10[2],xmm1[2],xmm10[3],xmm1[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm10[0,0,0,0]
+; SSE-NEXT:    movdqa %xmm3, %xmm1
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,0,0,0]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT:    movdqa %xmm2, %xmm3
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm11[0],xmm3[1],xmm11[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm12[0],xmm2[1],xmm12[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
 ; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm6[0],xmm14[1],xmm6[1],xmm14[2],xmm6[2],xmm14[3],xmm6[3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm5[0],xmm13[1],xmm5[1],xmm13[2],xmm5[2],xmm13[3],xmm5[3]
 ; SSE-NEXT:    movdqa 80(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 64(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm7
+; SSE-NEXT:    movdqa 64(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, %xmm7
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1],xmm7[2],xmm0[2],xmm7[3],xmm0[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm14[0,0,0,0]
-; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm7[0,0,0,0]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm6[2],xmm15[3],xmm6[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm13[0,0,0,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm7[0,0,0,0]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm14 = xmm14[2],xmm5[2],xmm14[3],xmm5[3]
 ; SSE-NEXT:    movdqa 32(%rdi), %xmm9
 ; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 48(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1],xmm9[2],xmm0[2],xmm9[3],xmm0[3]
-; SSE-NEXT:    movdqa (%rdi), %xmm6
+; SSE-NEXT:    movdqa (%rdi), %xmm5
 ; SSE-NEXT:    movdqa 16(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, %xmm8
+; SSE-NEXT:    movdqa %xmm5, %xmm8
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
 ; SSE-NEXT:    movdqa %xmm8, %xmm0
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm15[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm14[2,3]
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm11[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
-; SSE-NEXT:    movdqa %xmm1, %xmm15
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm4[0],xmm15[1],xmm4[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm15[2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm12[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    movdqa %xmm1, %xmm14
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm10[0],xmm14[1],xmm10[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm14[2,3]
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm9[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm9[1,1,1,1]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
-; SSE-NEXT:    movdqa %xmm7, %xmm15
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm14[0],xmm15[1],xmm14[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm15[2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    movdqa %xmm7, %xmm14
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm13[0],xmm14[1],xmm13[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm14[2,3]
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm4[2,2,2,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm10[2,2,2,2]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,2,2,2]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm15[2],xmm0[3],xmm15[3]
-; SSE-NEXT:    movdqa %xmm2, %xmm15
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm11[2],xmm15[3],xmm11[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm15[0],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm4[3,3,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm12[2],xmm4[3],xmm12[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm4[0],xmm0[1]
 ; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm14[2,2,2,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm13[2,2,2,2]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm15[2],xmm0[3],xmm15[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm8[3,3,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm8[3,3,3,3]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm8 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
 ; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm0[2,3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm14[2],xmm7[3],xmm14[3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm10[2],xmm1[3],xmm10[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,1],xmm1[2,3]
+; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm13[2],xmm7[3],xmm13[3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,1],xmm7[2,3]
-; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm13 = xmm13[4],mem[4],xmm13[5],mem[5],xmm13[6],mem[6],xmm13[7],mem[7]
-; SSE-NEXT:    movdqa %xmm10, %xmm15
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm7[2,3]
+; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = xmm11[4],mem[4],xmm11[5],mem[5],xmm11[6],mem[6],xmm11[7],mem[7]
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm6 = xmm6[4],mem[4],xmm6[5],mem[5],xmm6[6],mem[6],xmm6[7],mem[7]
+; SSE-NEXT:    movdqa (%rsp), %xmm15 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm15 = xmm15[4],mem[4],xmm15[5],mem[5],xmm15[6],mem[6],xmm15[7],mem[7]
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm12 = xmm12[4],mem[4],xmm12[5],mem[5],xmm12[6],mem[6],xmm12[7],mem[7]
-; SSE-NEXT:    movdqa %xmm5, %xmm2
-; SSE-NEXT:    punpckhwd (%rsp), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm0
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[0,0,0,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm0[0,0,0,0]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm14 = xmm14[2],xmm1[2],xmm14[3],xmm1[3]
+; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm1[0],xmm14[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm2[0,0,0,0]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm12 = xmm12[2],xmm1[2],xmm12[3],xmm1[3]
-; SSE-NEXT:    movdqa %xmm15, %xmm1
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm1[0],xmm12[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm10 = xmm10[4],mem[4],xmm10[5],mem[5],xmm10[6],mem[6],xmm10[7],mem[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,0,0,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm2[0,0,0,0]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm1[2],xmm7[3],xmm1[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm10, %xmm14
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,0,0,0]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm5 = xmm5[4],mem[4],xmm5[5],mem[5],xmm5[6],mem[6],xmm5[7],mem[7]
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm6 = xmm6[4],mem[4],xmm6[5],mem[5],xmm6[6],mem[6],xmm6[7],mem[7]
-; SSE-NEXT:    movdqa %xmm6, %xmm1
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm15[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm13[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1]
-; SSE-NEXT:    movdqa %xmm2, %xmm7
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm7[2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm5[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm7[2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm11[1,1,1,1]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm6[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1]
-; SSE-NEXT:    movdqa %xmm3, %xmm9
-; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
+; SSE-NEXT:    movdqa %xmm0, %xmm9
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm15[0],xmm9[1],xmm15[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm9[2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm0[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm0, %xmm10
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm2[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm11[2],xmm9[3],xmm11[3]
-; SSE-NEXT:    movdqa %xmm15, %xmm11
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm13[2],xmm11[3],xmm13[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm11[0],xmm9[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm14[2,2,2,2]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm11[2],xmm0[3],xmm11[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm6[3,3,3,3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm5[2],xmm6[3],xmm5[3]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[2,3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm10[2],xmm2[3],xmm10[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm13[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
-; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm5[0],xmm14[1],xmm5[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm3[2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, (%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 16(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, (%rdx)
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm3[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm5[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
+; SSE-NEXT:    movdqa %xmm2, %xmm10
+; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm4[0],xmm10[1],xmm4[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,1],xmm10[2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm15[2,2,2,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm0[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm10 = xmm10[2],xmm12[2],xmm10[3],xmm12[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm6[3,3,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm11[2],xmm6[3],xmm11[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm6[0],xmm10[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm4[2,2,2,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm2[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm12 = xmm12[2],xmm6[2],xmm12[3],xmm6[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[3,3,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm3[2],xmm5[3],xmm3[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm12[2,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm15[2],xmm0[3],xmm15[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm11[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,1],xmm0[2,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm2[2,3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rsi)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, 16(%rdx)
+; SSE-NEXT:    movaps %xmm2, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rdx)
 ; SSE-NEXT:    movaps %xmm8, (%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, 16(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, (%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, 16(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r8)
 ; SSE-NEXT:    movaps %xmm1, (%r9)
-; SSE-NEXT:    movapd %xmm12, 16(%r9)
+; SSE-NEXT:    movapd %xmm14, 16(%r9)
+; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movaps %xmm9, (%rax)
+; SSE-NEXT:    movaps %xmm7, 16(%rax)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movaps %xmm7, (%rax)
-; SSE-NEXT:    movaps %xmm4, 16(%rax)
+; SSE-NEXT:    movaps %xmm5, (%rax)
+; SSE-NEXT:    movapd %xmm10, 16(%rax)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movaps %xmm13, 16(%rax)
 ; SSE-NEXT:    movaps %xmm6, (%rax)
-; SSE-NEXT:    movapd %xmm9, 16(%rax)
-; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movaps %xmm0, 16(%rax)
-; SSE-NEXT:    movaps %xmm14, (%rax)
-; SSE-NEXT:    addq $168, %rsp
+; SSE-NEXT:    addq $184, %rsp
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i16_stride8_vf16:
@@ -3451,463 +3444,457 @@ define void @load_i16_stride8_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 define void @load_i16_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6, ptr %out.vec7) nounwind {
 ; SSE-LABEL: load_i16_stride8_vf32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $696, %rsp # imm = 0x2B8
+; SSE-NEXT:    subq $712, %rsp # imm = 0x2C8
 ; SSE-NEXT:    movdqa 496(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 480(%rdi), %xmm14
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 480(%rdi), %xmm15
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 208(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 192(%rdi), %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 240(%rdi), %xmm5
 ; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 224(%rdi), %xmm15
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 144(%rdi), %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 128(%rdi), %xmm4
+; SSE-NEXT:    movdqa 224(%rdi), %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 176(%rdi), %xmm7
-; SSE-NEXT:    movdqa %xmm7, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movdqa 144(%rdi), %xmm7
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 128(%rdi), %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 176(%rdi), %xmm8
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 160(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1],xmm0[2],xmm7[2],xmm0[3],xmm7[3]
-; SSE-NEXT:    movdqa %xmm0, %xmm7
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm6[0],xmm4[1],xmm6[1],xmm4[2],xmm6[2],xmm4[3],xmm6[3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3]
+; SSE-NEXT:    movdqa %xmm0, %xmm9
+; SSE-NEXT:    movdqa %xmm6, %xmm8
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm7[0],xmm8[1],xmm7[1],xmm8[2],xmm7[2],xmm8[3],xmm7[3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm5[0],xmm15[1],xmm5[1],xmm15[2],xmm5[2],xmm15[3],xmm5[3]
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm3, %xmm15
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1],xmm15[2],xmm1[2],xmm15[3],xmm1[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[0,0,0,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[0,0,0,0]
+; SSE-NEXT:    movdqa %xmm3, %xmm4
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,0,0,0]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
+; SSE-NEXT:    movdqa %xmm8, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
+; SSE-NEXT:    movdqa %xmm9, %xmm7
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm2[0],xmm14[1],xmm2[1],xmm14[2],xmm2[2],xmm14[3],xmm2[3]
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 464(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 448(%rdi), %xmm8
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[0,0,0,0]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[0,0,0,0]
+; SSE-NEXT:    movdqa 448(%rdi), %xmm15
+; SSE-NEXT:    movdqa %xmm15, (%rsp) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1],xmm15[2],xmm0[2],xmm15[3],xmm0[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,0,0,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[0,0,0,0]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movdqa 432(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 416(%rdi), %xmm11
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1],xmm11[2],xmm0[2],xmm11[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 400(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    movdqa 432(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 416(%rdi), %xmm13
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1],xmm13[2],xmm1[2],xmm13[3],xmm1[3]
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 400(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 384(%rdi), %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm1[0],xmm6[1],xmm1[1],xmm6[2],xmm1[2],xmm6[3],xmm1[3]
+; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 368(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 384(%rdi), %xmm10
+; SSE-NEXT:    movdqa 352(%rdi), %xmm10
 ; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1],xmm10[2],xmm0[2],xmm10[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    movdqa %xmm10, %xmm14
 ; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 368(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 352(%rdi), %xmm9
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1],xmm9[2],xmm0[2],xmm9[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 336(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 320(%rdi), %xmm12
 ; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1],xmm12[2],xmm0[2],xmm12[3],xmm0[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[0,0,0,0]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[0,0,0,0]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movdqa 304(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 288(%rdi), %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[0,0,0,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[0,0,0,0]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT:    movdqa 304(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 288(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1],xmm9[2],xmm1[2],xmm9[3],xmm1[3]
+; SSE-NEXT:    movdqa 272(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 256(%rdi), %xmm11
 ; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1],xmm11[2],xmm0[2],xmm11[3],xmm0[3]
-; SSE-NEXT:    movdqa 272(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 256(%rdi), %xmm10
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1],xmm10[2],xmm0[2],xmm10[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 112(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 96(%rdi), %xmm9
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm1[0],xmm11[1],xmm1[1],xmm11[2],xmm1[2],xmm11[3],xmm1[3]
+; SSE-NEXT:    movdqa %xmm11, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1]
+; SSE-NEXT:    movdqa %xmm9, %xmm10
 ; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1],xmm9[2],xmm0[2],xmm9[3],xmm0[3]
-; SSE-NEXT:    movdqa 80(%rdi), %xmm0
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 112(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 64(%rdi), %xmm13
+; SSE-NEXT:    movdqa 96(%rdi), %xmm13
 ; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1],xmm13[2],xmm0[2],xmm13[3],xmm0[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,0,0,0]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[0,0,0,0]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; SSE-NEXT:    movdqa 32(%rdi), %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 80(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 64(%rdi), %xmm14
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1],xmm14[2],xmm0[2],xmm14[3],xmm0[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[0,0,0,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[0,0,0,0]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE-NEXT:    movdqa 32(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 48(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1],xmm7[2],xmm0[2],xmm7[3],xmm0[3]
-; SSE-NEXT:    movdqa (%rdi), %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1],xmm9[2],xmm0[2],xmm9[3],xmm0[3]
+; SSE-NEXT:    movdqa (%rdi), %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 16(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1],xmm6[2],xmm0[2],xmm6[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm6, %xmm1
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[1,1,1,1]
 ; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
+; SSE-NEXT:    movdqa %xmm14, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,2,2,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,2,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[2,2,2,2]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm5[2],xmm0[3],xmm5[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm8 = xmm8[2],xmm7[2],xmm8[3],xmm7[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm8[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,2,2]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[2,2,2,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm3, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,2,2,2]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movapd %xmm3, %xmm0
-; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm6[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm6[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm10, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,2,2,2]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[2,2,2,2]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm11[2],xmm0[3],xmm11[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm10[2],xmm11[3],xmm10[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm11[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[2,2,2,2]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    movdqa %xmm6, %xmm1
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm7[2],xmm1[3],xmm7[3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[3,3,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm9[2],xmm0[3],xmm9[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm15 = xmm15[2],mem[2],xmm15[3],mem[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm15[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = xmm8[2],mem[2],xmm8[3],mem[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm8[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm12 = xmm12[2],xmm14[2],xmm12[3],xmm14[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[3,3,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm4[2],xmm15[3],xmm4[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm15[2,3]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm12 = xmm12[2],xmm6[2],xmm12[3],xmm6[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm12[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm13 = xmm13[2],xmm9[2],xmm13[3],xmm9[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm14 = xmm14[2],xmm13[2],xmm14[3],xmm13[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm14[2,3]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = xmm11[4],mem[4],xmm11[5],mem[5],xmm11[6],mem[6],xmm11[7],mem[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[0,0,0,0]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    movdqa %xmm4, %xmm10
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpckhwd (%rsp), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm15 = xmm15[4],mem[4],xmm15[5],mem[5],xmm15[6],mem[6],xmm15[7],mem[7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[4],mem[4],xmm7[5],mem[5],xmm7[6],mem[6],xmm7[7],mem[7]
+; SSE-NEXT:    movdqa (%rsp), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,0,0,0]
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm1, %xmm12
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,0,0,0]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE-NEXT:    movdqa %xmm3, %xmm7
+; SSE-NEXT:    movdqa %xmm1, %xmm8
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    movdqa %xmm4, %xmm14
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSE-NEXT:    movdqa %xmm3, %xmm13
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = xmm8[4],mem[4],xmm8[5],mem[5],xmm8[6],mem[6],xmm8[7],mem[7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm8, %xmm11
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[0,0,0,0]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm0[2],xmm5[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    movdqa %xmm4, %xmm14
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movdqa %xmm1, %xmm13
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm0[0],xmm5[1]
-; SSE-NEXT:    movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm10 = xmm10[4],mem[4],xmm10[5],mem[5],xmm10[6],mem[6],xmm10[7],mem[7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm9 = xmm9[4],mem[4],xmm9[5],mem[5],xmm9[6],mem[6],xmm9[7],mem[7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[0,0,0,0]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm0[2],xmm5[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm9, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm0[0],xmm5[1]
-; SSE-NEXT:    movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = xmm8[4],mem[4],xmm8[5],mem[5],xmm8[6],mem[6],xmm8[7],mem[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[0,0,0,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,0,0,0]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    movdqa %xmm4, %xmm12
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
-; SSE-NEXT:    movdqa %xmm4, %xmm3
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm1[2,3]
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm2, %xmm5
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm3
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
-; SSE-NEXT:    movdqa %xmm11, %xmm6
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3]
-; SSE-NEXT:    movaps %xmm2, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm14
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[0,0,0,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[0,0,0,0]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = xmm9[4],mem[4],xmm9[5],mem[5],xmm9[6],mem[6],xmm9[7],mem[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[4],mem[4],xmm15[5],mem[5],xmm15[6],mem[6],xmm15[7],mem[7]
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1]
 ; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm10, %xmm13
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm7, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,1],xmm0[2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm4[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,1],xmm0[2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm5[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm0[2],xmm11[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm0[0],xmm11[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,2,2,2]
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm7[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm12 = xmm12[2],xmm0[2],xmm12[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm13[2],xmm0[3],xmm13[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm0[0],xmm12[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,2,2,2]
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm1[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm1, %xmm3
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm13 = xmm13[2],xmm0[2],xmm13[3],xmm0[3]
-; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    movapd %xmm6, %xmm0
-; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm5[2],xmm0[3],xmm5[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,2,2,2]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm4, %xmm14
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm14 = xmm14[2],xmm2[2],xmm14[3],xmm2[3]
-; SSE-NEXT:    movdqa %xmm2, %xmm8
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm1[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm11[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm10[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, %xmm5
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm15[3,3,3,3]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm5[2,3]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm13[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
+; SSE-NEXT:    movdqa %xmm8, %xmm6
+; SSE-NEXT:    movdqa %xmm8, %xmm5
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm6[2,3]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm3[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm13, %xmm7
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    unpcklps {{.*#+}} xmm7 = xmm7[0],xmm2[0],xmm7[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm7[2,3]
+; SSE-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm9[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm15[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm8[0],xmm6[1],xmm8[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm8[2,3]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm1[2,2,2,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm4[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm10[3,3,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm10 = xmm10[2],xmm11[2],xmm10[3],xmm11[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm10[0],xmm9[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm2[2,2,2,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm13[2,2,2,2]
+; SSE-NEXT:    movaps %xmm13, %xmm2
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm10[2],xmm11[3],xmm10[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm12[3,3,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm12 = xmm12[2],xmm3[2],xmm12[3],xmm3[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm12[0],xmm11[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm7[2,2,2,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm5[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm13 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm14[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm14 = xmm14[2],xmm6[2],xmm14[3],xmm6[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm14[0],xmm13[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm0[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm14[2],xmm0[3],xmm14[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm15[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm3[2],xmm15[3],xmm3[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm0[2,3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
 ; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm1[2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm4[2,3]
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
 ; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3]
-; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
-; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm7 = xmm7[2],mem[2],xmm7[3],mem[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm7[2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 32(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 48(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, (%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm8, 16(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 32(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 48(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, (%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 16(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 32(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 48(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, (%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 16(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 32(%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 48(%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, (%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 16(%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 32(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 48(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, (%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 16(%r9)
+; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,1],xmm2[2,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm7[2],xmm5[3],xmm7[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,1],xmm5[2,3]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm1[2,3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r9)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movaps %xmm9, 32(%rax)
-; SSE-NEXT:    movaps (%rsp), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 48(%rax)
-; SSE-NEXT:    movaps %xmm10, (%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 16(%rax)
+; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; SSE-NEXT:    movapd %xmm13, 48(%rax)
-; SSE-NEXT:    movapd %xmm12, 32(%rax)
-; SSE-NEXT:    movapd %xmm11, 16(%rax)
-; SSE-NEXT:    movaps %xmm14, (%rax)
+; SSE-NEXT:    movapd %xmm11, 32(%rax)
+; SSE-NEXT:    movapd %xmm9, 16(%rax)
+; SSE-NEXT:    movaps %xmm15, (%rax)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movaps %xmm1, 48(%rax)
-; SSE-NEXT:    movaps %xmm2, 32(%rax)
-; SSE-NEXT:    movaps %xmm15, 16(%rax)
-; SSE-NEXT:    movaps %xmm0, (%rax)
-; SSE-NEXT:    addq $696, %rsp # imm = 0x2B8
+; SSE-NEXT:    movaps %xmm12, 48(%rax)
+; SSE-NEXT:    movaps %xmm10, 32(%rax)
+; SSE-NEXT:    movaps %xmm8, 16(%rax)
+; SSE-NEXT:    movaps %xmm14, (%rax)
+; SSE-NEXT:    addq $712, %rsp # imm = 0x2C8
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i16_stride8_vf32:
@@ -7775,7 +7762,7 @@ define void @load_i16_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 define void @load_i16_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6, ptr %out.vec7) nounwind {
 ; SSE-LABEL: load_i16_stride8_vf64:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $1800, %rsp # imm = 0x708
+; SSE-NEXT:    subq $1752, %rsp # imm = 0x6D8
 ; SSE-NEXT:    movdqa 752(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 736(%rdi), %xmm3
@@ -7790,27 +7777,26 @@ define void @load_i16_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 144(%rdi), %xmm7
 ; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 128(%rdi), %xmm8
+; SSE-NEXT:    movdqa 128(%rdi), %xmm15
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 176(%rdi), %xmm8
 ; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 176(%rdi), %xmm9
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 160(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3]
-; SSE-NEXT:    movdqa %xmm0, %xmm9
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm7[0],xmm8[1],xmm7[1],xmm8[2],xmm7[2],xmm8[3],xmm7[3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3]
+; SSE-NEXT:    movdqa %xmm0, %xmm8
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm7[0],xmm15[1],xmm7[1],xmm15[2],xmm7[2],xmm15[3],xmm7[3]
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm6, %xmm12
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,0,0,0]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    movdqa %xmm8, %xmm4
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
-; SSE-NEXT:    movdqa %xmm9, %xmm7
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
@@ -7826,18 +7812,18 @@ define void @load_i16_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; SSE-NEXT:    movdqa 688(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 672(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 656(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 640(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT:    movdqa 672(%rdi), %xmm15
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm2[0],xmm15[1],xmm2[1],xmm15[2],xmm2[2],xmm15[3],xmm2[3]
+; SSE-NEXT:    movdqa 656(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT:    movdqa 640(%rdi), %xmm12
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1],xmm12[2],xmm0[2],xmm12[3],xmm0[3]
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 624(%rdi), %xmm0
@@ -7862,13 +7848,14 @@ define void @load_i16_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm0, %xmm3
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 528(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 512(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 528(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 512(%rdi), %xmm11
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1],xmm11[2],xmm0[2],xmm11[3],xmm0[3]
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm11, %xmm0
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7876,16 +7863,17 @@ define void @load_i16_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 480(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 464(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 448(%rdi), %xmm10
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1],xmm10[2],xmm0[2],xmm10[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,0,0,0]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[0,0,0,0]
+; SSE-NEXT:    movdqa 448(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,0,0,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; SSE-NEXT:    movdqa 432(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7894,12 +7882,12 @@ define void @load_i16_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movdqa %xmm0, %xmm3
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 400(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 384(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT:    movdqa 400(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 384(%rdi), %xmm11
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1],xmm11[2],xmm0[2],xmm11[3],xmm0[3]
+; SSE-NEXT:    movdqa %xmm11, %xmm0
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7907,31 +7895,31 @@ define void @load_i16_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 992(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 976(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 960(%rdi), %xmm13
-; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1],xmm13[2],xmm0[2],xmm13[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,0,0,0]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[0,0,0,0]
+; SSE-NEXT:    movdqa 960(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,0,0,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; SSE-NEXT:    movdqa 944(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 928(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 912(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 896(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT:    movdqa 928(%rdi), %xmm7
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm2[0],xmm7[1],xmm2[1],xmm7[2],xmm2[2],xmm7[3],xmm2[3]
+; SSE-NEXT:    movdqa 912(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT:    movdqa 896(%rdi), %xmm8
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1],xmm8[2],xmm0[2],xmm8[3],xmm0[3]
+; SSE-NEXT:    movdqa %xmm8, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 368(%rdi), %xmm0
@@ -7942,12 +7930,12 @@ define void @load_i16_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 336(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 320(%rdi), %xmm10
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1],xmm10[2],xmm0[2],xmm10[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 320(%rdi), %xmm12
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1],xmm12[2],xmm0[2],xmm12[3],xmm0[3]
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,0,0,0]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[0,0,0,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[0,0,0,0]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; SSE-NEXT:    movdqa 304(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7958,11 +7946,10 @@ define void @load_i16_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 272(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 256(%rdi), %xmm14
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1],xmm14[2],xmm0[2],xmm14[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm0
+; SSE-NEXT:    movdqa 256(%rdi), %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1],xmm6[2],xmm0[2],xmm6[3],xmm0[3]
+; SSE-NEXT:    movdqa %xmm6, %xmm0
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7974,331 +7961,328 @@ define void @load_i16_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 848(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 832(%rdi), %xmm9
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1],xmm9[2],xmm0[2],xmm9[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 832(%rdi), %xmm13
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1],xmm13[2],xmm0[2],xmm13[3],xmm0[3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,0,0,0]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,0,0,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[0,0,0,0]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    movdqa %xmm1, %xmm0
 ; SSE-NEXT:    movdqa 816(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 800(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 784(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 768(%rdi), %xmm15
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1],xmm15[2],xmm0[2],xmm15[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 800(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm1, %xmm5
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3]
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 784(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 768(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 112(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 96(%rdi), %xmm13
-; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1],xmm13[2],xmm0[2],xmm13[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 96(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 80(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 64(%rdi), %xmm5
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[0,0,0,0]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[0,0,0,0]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT:    movdqa 32(%rdi), %xmm8
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 48(%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm8, %xmm2
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rdi), %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 64(%rdi), %xmm14
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1],xmm14[2],xmm0[2],xmm14[3],xmm0[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm1[0,0,0,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[0,0,0,0]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm9[2],xmm0[3],xmm9[3]
+; SSE-NEXT:    movdqa 32(%rdi), %xmm12
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 48(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm3[0],xmm12[1],xmm3[1],xmm12[2],xmm3[2],xmm12[3],xmm3[3]
+; SSE-NEXT:    movdqa (%rdi), %xmm10
+; SSE-NEXT:    movdqa %xmm10, (%rsp) # 16-byte Spill
 ; SSE-NEXT:    movdqa 16(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm1[0],xmm6[1],xmm1[1],xmm6[2],xmm1[2],xmm6[3],xmm1[3]
-; SSE-NEXT:    movdqa %xmm6, %xmm1
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm1[0],xmm10[1],xmm1[1],xmm10[2],xmm1[2],xmm10[3],xmm1[3]
+; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1]
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm7, %xmm14
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[1,1,1,1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    movdqa %xmm12, %xmm4
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm5, %xmm0
+; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[1,1,1,1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, %xmm0
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
 ; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
 ; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
 ; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm5, %xmm0
-; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movdqa %xmm14, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,2,2,2]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[2,2,2,2]
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,2,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,2,2,2]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm14[2],xmm0[3],xmm14[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm9[2],xmm0[3],xmm9[3]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2]
-; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm12 = xmm12[2],xmm13[2],xmm12[3],xmm13[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm12[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,2,2,2]
-; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm10 = xmm10[2],xmm11[2],xmm10[3],xmm11[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm10[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,2,2,2]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[2,2,2,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,2,2,2]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm12[2],xmm0[3],xmm12[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,2,2]
 ; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[2,2,2,2]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm9[0],xmm1[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,2,2,2]
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,2,2]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,2,2,2]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movapd %xmm4, %xmm0
-; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm12[2],xmm11[3],xmm12[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm11[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,2,2,2]
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = xmm8[2],mem[2],xmm8[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm8[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,2,2,2]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[2,2,2,2]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movapd %xmm7, %xmm0
-; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm8[2],xmm0[3],xmm8[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm6[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm2[2],xmm6[3],xmm2[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm6[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[2,2,2,2]
-; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm5, %xmm1
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm6[2],xmm1[3],xmm6[3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm13, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm6[2],xmm4[3],xmm6[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm4[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm10, %xmm13
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,2,2,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm8 = xmm8[2],xmm10[2],xmm8[3],xmm10[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm1[2,3]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = xmm8[2],mem[2],xmm8[3],mem[3]
 ; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, %xmm2
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm13[2],xmm2[3],xmm13[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm8[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm14 = xmm14[2],xmm10[2],xmm14[3],xmm10[3]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm11[2],xmm3[3],xmm11[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm3[2,3]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[2],mem[2],xmm15[3],mem[3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm14[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,1],xmm15[2,3]
+; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
 ; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,1],xmm1[2,3]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
 ; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm11[2],xmm2[3],xmm11[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm1[2,3]
+; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm9[2],xmm7[3],xmm9[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm7[2,3]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
 ; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm9[2],xmm2[3],xmm9[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm6 = xmm6[4],mem[4],xmm6[5],mem[5],xmm6[6],mem[6],xmm6[7],mem[7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm11 = xmm11[4],mem[4],xmm11[5],mem[5],xmm11[6],mem[6],xmm11[7],mem[7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm1[2,3]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm14 = xmm14[2],xmm13[2],xmm14[3],xmm13[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm14[2,3]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = xmm9[4],mem[4],xmm9[5],mem[5],xmm9[6],mem[6],xmm9[7],mem[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
-; SSE-NEXT:    movdqa %xmm6, %xmm13
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,0,0,0]
+; SSE-NEXT:    movdqa %xmm2, %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[0,0,0,0]
+; SSE-NEXT:    movdqa %xmm1, %xmm12
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT:    movdqa %xmm3, %xmm13
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm14 = xmm14[4],mem[4],xmm14[5],mem[5],xmm14[6],mem[6],xmm14[7],mem[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,0,0,0]
+; SSE-NEXT:    movdqa %xmm1, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,0,0,0]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,0,0,0]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
-; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm10 = xmm10[4],mem[4],xmm10[5],mem[5],xmm10[6],mem[6],xmm10[7],mem[7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm6 = xmm6[4],mem[4],xmm6[5],mem[5],xmm6[6],mem[6],xmm6[7],mem[7]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
@@ -8310,86 +8294,89 @@ define void @load_i16_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT:    movdqa %xmm3, %xmm8
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = xmm8[4],mem[4],xmm8[5],mem[5],xmm8[6],mem[6],xmm8[7],mem[7]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = xmm5[4],mem[4],xmm5[5],mem[5],xmm5[6],mem[6],xmm5[7],mem[7]
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm3[0,0,0,0]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm0[2],xmm7[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm0[0],xmm7[1]
-; SSE-NEXT:    movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = xmm4[4],mem[4],xmm4[5],mem[5],xmm4[6],mem[6],xmm4[7],mem[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm6 = xmm6[4],mem[4],xmm6[5],mem[5],xmm6[6],mem[6],xmm6[7],mem[7]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm5 = xmm5[4],mem[4],xmm5[5],mem[5],xmm5[6],mem[6],xmm5[7],mem[7]
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[4],mem[4],xmm1[5],mem[5],xmm1[6],mem[6],xmm1[7],mem[7]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm2[0,0,0,0]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm0[2],xmm9[3],xmm0[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; SSE-NEXT:    movdqa %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm0[0],xmm9[1]
-; SSE-NEXT:    movapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm7 = xmm7[4],mem[4],xmm7[5],mem[5],xmm7[6],mem[6],xmm7[7],mem[7]
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm12 = xmm12[4],mem[4],xmm12[5],mem[5],xmm12[6],mem[6],xmm12[7],mem[7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,0,0,0]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[4],mem[4],xmm15[5],mem[5],xmm15[6],mem[6],xmm15[7],mem[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,0,0,0]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rsp), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
-; SSE-NEXT:    movdqa %xmm2, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = xmm3[4],mem[4],xmm3[5],mem[5],xmm3[6],mem[6],xmm3[7],mem[7]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = xmm5[4],mem[4],xmm5[5],mem[5],xmm5[6],mem[6],xmm5[7],mem[7]
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
@@ -8401,7 +8388,7 @@ define void @load_i16_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm5, %xmm0
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -8416,357 +8403,359 @@ define void @load_i16_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[4],mem[4],xmm2[5],mem[5],xmm2[6],mem[6],xmm2[7],mem[7]
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm7 = xmm7[4],mem[4],xmm7[5],mem[5],xmm7[6],mem[6],xmm7[7],mem[7]
-; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    movdqa %xmm7, %xmm3
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm0, %xmm5
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm1[2,3]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm13, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm14, %xmm9
 ; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    movdqa %xmm7, %xmm14
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm10[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, %xmm0
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[2,3]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm7, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm10, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[2,3]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
 ; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm8, %xmm0
-; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, %xmm0
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
 ; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $85, (%rsp), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $85, (%rsp), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm5, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,2,2,2]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm13[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm0[2],xmm11[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,2,2,2]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[2,2,2,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[2,2,2,2]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm0[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm13[2],xmm0[3],xmm13[3]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,2,2,2]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[2,2,2,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,2,2,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[2,2,2,2]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm9[2],xmm0[3],xmm9[3]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[2,2,2,2]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[2,2,2,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,2,2,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[2,2,2,2]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm8[2],xmm0[3],xmm8[3]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[2,2,2,2]
-; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm11 = mem[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm0[2],xmm11[3],xmm0[3]
-; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    movapd %xmm14, %xmm0
-; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,2,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm0[0],xmm11[1]
-; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[2,2,2,2]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm6[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm10 = xmm10[2],xmm0[2],xmm10[3],xmm0[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,2,2,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm3[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm14 = xmm14[2],xmm0[2],xmm14[3],xmm0[3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm15[2],xmm0[3],xmm15[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm0[0],xmm10[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,2,2,2]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm5[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm0[2],xmm9[3],xmm0[3]
-; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    unpckhps (%rsp), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm0[0],xmm9[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,2,2]
-; SSE-NEXT:    movaps %xmm2, %xmm15
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,2,2,2]
+; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm6 = mem[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm0[2],xmm6[3],xmm0[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm0[3,3,3,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm13, %xmm7
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm12[2],xmm7[3],xmm12[3]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm1[2,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm13[2],xmm0[3],xmm13[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm0[0],xmm6[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm10[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm7[2],xmm0[3],xmm7[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT:    movdqa (%rsp), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,3]
+; SSE-NEXT:    movaps %xmm1, (%rsp) # 16-byte Spill
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = mem[3,3,3,3]
 ; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm1[2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm1[2,3]
+; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm6 = mem[3,3,3,3]
 ; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm1[2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm1[2,3]
+; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm5 = mem[3,3,3,3]
 ; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm1[2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm1[2,3]
+; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
 ; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm1[2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm1[2,3]
+; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm14[3,3,3,3]
 ; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm1[2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm1[2,3]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[3,3,3,3]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm12 = xmm12[2],mem[2],xmm12[3],mem[3]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
-; SSE-NEXT:    pshufd $255, (%rsp), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm12[2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{.*#+}} xmm12 = xmm12[2],xmm15[2],xmm12[3],xmm15[3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[3,3,3,3]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm15 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm12[2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 96(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 32(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 112(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 48(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 64(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, (%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 80(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 16(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 96(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 32(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 112(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 48(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 64(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, (%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 80(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 16(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 96(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 32(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 112(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 48(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 64(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, (%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 80(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 16(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 112(%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 96(%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 80(%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 64(%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 48(%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 32(%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 16(%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, (%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 112(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 96(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 80(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 64(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 48(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 32(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 16(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, (%r9)
-; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm12, 112(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm12, 96(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm12, 80(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm12, 64(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm12, 48(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm12, 32(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 16(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm12, (%rax)
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,1],xmm1[2,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm10 = xmm10[2],xmm12[2],xmm10[3],xmm12[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm10[2,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm8 = xmm8[2],xmm11[2],xmm8[3],xmm11[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm8[2,3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r9)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movapd %xmm9, 112(%rax)
-; SSE-NEXT:    movapd %xmm10, 96(%rax)
-; SSE-NEXT:    movapd %xmm11, 80(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm9, 64(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm9, 48(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm9, 32(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm9, 16(%rax)
-; SSE-NEXT:    movaps %xmm7, (%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rax)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movaps %xmm1, 112(%rax)
-; SSE-NEXT:    movaps %xmm2, 96(%rax)
-; SSE-NEXT:    movaps %xmm3, 80(%rax)
-; SSE-NEXT:    movaps %xmm4, 64(%rax)
-; SSE-NEXT:    movaps %xmm5, 48(%rax)
-; SSE-NEXT:    movaps %xmm6, 32(%rax)
-; SSE-NEXT:    movaps %xmm8, 16(%rax)
+; SSE-NEXT:    movapd %xmm2, 112(%rax)
+; SSE-NEXT:    movapd %xmm6, 96(%rax)
+; SSE-NEXT:    movapd %xmm14, 80(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, (%rax)
-; SSE-NEXT:    addq $1800, %rsp # imm = 0x708
+; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; SSE-NEXT:    movaps %xmm4, 112(%rax)
+; SSE-NEXT:    movaps %xmm9, 96(%rax)
+; SSE-NEXT:    movaps %xmm15, 80(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps %xmm3, (%rax)
+; SSE-NEXT:    addq $1752, %rsp # imm = 0x6D8
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i16_stride8_vf64:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll
index a0ea6ddeca7df..d0ac20725ca06 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-3.ll
@@ -745,99 +745,91 @@ define void @load_i32_stride3_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i32_stride3_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2) nounwind {
 ; SSE-LABEL: load_i32_stride3_vf16:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    movaps 96(%rdi), %xmm6
-; SSE-NEXT:    movaps 128(%rdi), %xmm1
-; SSE-NEXT:    movaps 112(%rdi), %xmm13
-; SSE-NEXT:    movaps 144(%rdi), %xmm11
-; SSE-NEXT:    movaps 176(%rdi), %xmm10
-; SSE-NEXT:    movaps 160(%rdi), %xmm9
-; SSE-NEXT:    movaps (%rdi), %xmm7
-; SSE-NEXT:    movaps 16(%rdi), %xmm8
-; SSE-NEXT:    movaps 32(%rdi), %xmm3
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 48(%rdi), %xmm15
-; SSE-NEXT:    movaps 80(%rdi), %xmm14
-; SSE-NEXT:    movaps 64(%rdi), %xmm2
-; SSE-NEXT:    movaps %xmm2, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm14[1,0]
-; SSE-NEXT:    movaps %xmm15, %xmm5
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm8, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm3[1,0]
-; SSE-NEXT:    movaps %xmm7, %xmm5
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm9, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm10[1,0]
+; SSE-NEXT:    movaps 96(%rdi), %xmm15
+; SSE-NEXT:    movaps 128(%rdi), %xmm11
+; SSE-NEXT:    movaps 112(%rdi), %xmm14
+; SSE-NEXT:    movaps 144(%rdi), %xmm12
+; SSE-NEXT:    movaps 176(%rdi), %xmm2
+; SSE-NEXT:    movaps 160(%rdi), %xmm8
+; SSE-NEXT:    movaps (%rdi), %xmm9
+; SSE-NEXT:    movaps 16(%rdi), %xmm6
+; SSE-NEXT:    movaps 32(%rdi), %xmm10
 ; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm11, %xmm3
-; SSE-NEXT:    movaps %xmm11, %xmm4
-; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm13, %xmm0
+; SSE-NEXT:    movaps 48(%rdi), %xmm5
+; SSE-NEXT:    movaps 80(%rdi), %xmm1
+; SSE-NEXT:    movaps 64(%rdi), %xmm3
+; SSE-NEXT:    movaps %xmm3, %xmm0
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
-; SSE-NEXT:    movaps %xmm1, %xmm12
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm6, %xmm5
-; SSE-NEXT:    movaps %xmm6, %xmm3
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm15, %xmm11
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[1,0],xmm2[0,0]
-; SSE-NEXT:    movaps %xmm2, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm14[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,2],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm4, %xmm6
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[1,0],xmm9[0,0]
-; SSE-NEXT:    movaps %xmm9, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm10[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,2],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm3, %xmm1
-; SSE-NEXT:    movaps %xmm3, %xmm10
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm13[0,0]
-; SSE-NEXT:    movaps %xmm13, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm12[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[1,0],xmm8[0,0]
-; SSE-NEXT:    movaps %xmm8, %xmm12
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[3,1],xmm4[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm12[0,2]
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm2[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm15[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm12[0],xmm2[1],xmm12[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm14[0,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm8[1,1,1,1]
+; SSE-NEXT:    movaps %xmm5, %xmm7
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm10[1,0]
+; SSE-NEXT:    movaps %xmm9, %xmm7
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm8, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm2[1,0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm12, %xmm10
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm14, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm11[1,0]
+; SSE-NEXT:    movaps %xmm11, %xmm13
+; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm15, %xmm7
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,3],xmm0[0,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm5[2,3,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,0],xmm3[0,0]
+; SSE-NEXT:    movaps %xmm3, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm1[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[0,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm12[2,3,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[1,0],xmm8[0,0]
+; SSE-NEXT:    movaps %xmm8, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm2[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,2],xmm0[0,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,3,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[1,0],xmm14[0,0]
+; SSE-NEXT:    movaps %xmm14, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm13[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,2],xmm0[0,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,3,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[1,0],xmm6[0,0]
+; SSE-NEXT:    movaps %xmm6, %xmm13
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[3,1],xmm2[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,2],xmm13[0,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm3[0],xmm11[1],xmm3[1]
+; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = xmm11[0,1],mem[0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm6[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm4[0,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[1,1,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
-; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = xmm8[0,1],mem[0,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm10[2,3,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm8[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
 ; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm4 = xmm4[0,1],mem[0,3]
-; SSE-NEXT:    movaps %xmm5, 32(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 48(%rsi)
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm14[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[0,1],mem[0,3]
+; SSE-NEXT:    movaps %xmm7, 32(%rsi)
+; SSE-NEXT:    movaps %xmm10, 48(%rsi)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm3, (%rsi)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm3, 16(%rsi)
-; SSE-NEXT:    movaps %xmm1, 32(%rdx)
-; SSE-NEXT:    movaps %xmm6, 48(%rdx)
-; SSE-NEXT:    movaps %xmm7, (%rdx)
-; SSE-NEXT:    movaps %xmm11, 16(%rdx)
-; SSE-NEXT:    movaps %xmm4, 32(%rcx)
-; SSE-NEXT:    movaps %xmm8, 48(%rcx)
+; SSE-NEXT:    movaps %xmm15, 32(%rdx)
+; SSE-NEXT:    movaps %xmm12, 48(%rdx)
+; SSE-NEXT:    movaps %xmm9, (%rdx)
+; SSE-NEXT:    movaps %xmm5, 16(%rdx)
+; SSE-NEXT:    movaps %xmm1, 32(%rcx)
+; SSE-NEXT:    movaps %xmm4, 48(%rcx)
 ; SSE-NEXT:    movaps %xmm0, (%rcx)
-; SSE-NEXT:    movaps %xmm2, 16(%rcx)
+; SSE-NEXT:    movaps %xmm11, 16(%rcx)
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i32_stride3_vf16:
@@ -1239,226 +1231,217 @@ define void @load_i32_stride3_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 define void @load_i32_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2) nounwind {
 ; SSE-LABEL: load_i32_stride3_vf32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $392, %rsp # imm = 0x188
-; SSE-NEXT:    movaps 192(%rdi), %xmm4
+; SSE-NEXT:    subq $360, %rsp # imm = 0x168
+; SSE-NEXT:    movaps 192(%rdi), %xmm11
+; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps 224(%rdi), %xmm3
-; SSE-NEXT:    movaps 208(%rdi), %xmm14
-; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 240(%rdi), %xmm7
-; SSE-NEXT:    movaps 272(%rdi), %xmm6
-; SSE-NEXT:    movaps 256(%rdi), %xmm9
-; SSE-NEXT:    movaps (%rdi), %xmm13
+; SSE-NEXT:    movaps 208(%rdi), %xmm13
 ; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 16(%rdi), %xmm8
+; SSE-NEXT:    movaps 240(%rdi), %xmm14
+; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 272(%rdi), %xmm4
+; SSE-NEXT:    movaps 256(%rdi), %xmm6
+; SSE-NEXT:    movaps (%rdi), %xmm10
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 16(%rdi), %xmm7
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 32(%rdi), %xmm8
 ; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 32(%rdi), %xmm11
-; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 48(%rdi), %xmm2
+; SSE-NEXT:    movaps 48(%rdi), %xmm9
 ; SSE-NEXT:    movaps 80(%rdi), %xmm1
 ; SSE-NEXT:    movaps 64(%rdi), %xmm5
 ; SSE-NEXT:    movaps %xmm5, %xmm0
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
 ; SSE-NEXT:    movaps %xmm1, %xmm12
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    movaps %xmm9, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[0,2]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm9, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm6[1,0]
-; SSE-NEXT:    movaps %xmm6, %xmm10
-; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm7, %xmm1
-; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm8, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm11[1,0]
-; SSE-NEXT:    movaps %xmm13, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm14, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm3[1,0]
-; SSE-NEXT:    movaps %xmm3, %xmm13
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm4[1,0]
 ; SSE-NEXT:    movaps %xmm4, %xmm1
-; SSE-NEXT:    movaps %xmm4, %xmm11
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 176(%rdi), %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 160(%rdi), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
-; SSE-NEXT:    movaps 144(%rdi), %xmm3
-; SSE-NEXT:    movaps %xmm3, %xmm1
-; SSE-NEXT:    movaps %xmm3, (%rsp) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 368(%rdi), %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm14, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm7, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm8[1,0]
+; SSE-NEXT:    movaps %xmm10, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm13, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm3[1,0]
+; SSE-NEXT:    movaps %xmm3, %xmm14
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm11, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 176(%rdi), %xmm15
+; SSE-NEXT:    movaps 160(%rdi), %xmm8
+; SSE-NEXT:    movaps %xmm8, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm15[1,0]
+; SSE-NEXT:    movaps %xmm15, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movaps 144(%rdi), %xmm10
+; SSE-NEXT:    movaps %xmm10, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 368(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps 352(%rdi), %xmm0
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
-; SSE-NEXT:    movaps 336(%rdi), %xmm4
-; SSE-NEXT:    movaps %xmm4, %xmm1
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 128(%rdi), %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 112(%rdi), %xmm15
-; SSE-NEXT:    movaps %xmm15, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
-; SSE-NEXT:    movaps 96(%rdi), %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 320(%rdi), %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 304(%rdi), %xmm6
-; SSE-NEXT:    movaps %xmm6, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
-; SSE-NEXT:    movaps 288(%rdi), %xmm8
-; SSE-NEXT:    movaps %xmm8, %xmm1
-; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm2[1,0]
+; SSE-NEXT:    movaps 336(%rdi), %xmm13
+; SSE-NEXT:    movaps %xmm13, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 128(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 112(%rdi), %xmm11
+; SSE-NEXT:    movaps %xmm11, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm2[1,0]
+; SSE-NEXT:    movaps 96(%rdi), %xmm7
+; SSE-NEXT:    movaps %xmm7, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 320(%rdi), %xmm2
+; SSE-NEXT:    movaps 304(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm2[1,0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 288(%rdi), %xmm4
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm5[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[1,0],xmm5[0,0]
 ; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,1],xmm12[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm5[0,2]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm7, %xmm14
-; SSE-NEXT:    movaps %xmm9, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,2],xmm5[0,2]
 ; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[1,0],xmm9[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm10[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,2],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm11, %xmm10
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[1,0],xmm0[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm13[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,2],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm3, %xmm9
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm0
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,0],xmm6[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm1[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[1,0],xmm0[0,0]
-; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,1],mem[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,2],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm4, %xmm7
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm0[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm14[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm10[2,3,2,3]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[1,0],xmm8[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[3,1],xmm15[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,2],xmm8[0,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm13[2,3,2,3]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[1,0],xmm0[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[1,0],xmm0[0,0]
 ; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = xmm0[3,1],mem[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm0[0,2]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, %xmm11
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[1,0],xmm15[0,0]
-; SSE-NEXT:    movaps %xmm15, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,2],xmm0[0,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm7[2,3,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[1,0],xmm11[0,0]
+; SSE-NEXT:    movaps %xmm11, %xmm0
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm12[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,2],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm8, %xmm3
-; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,0],xmm6[0,0]
-; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm6 = xmm6[3,1],mem[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm6[0,2]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm2[2,3,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm0[0,2]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[2,3,2,3]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm0[0,0]
-; SSE-NEXT:    movaps %xmm0, %xmm13
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[3,1],xmm1[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm13[0,2]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm13 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm13[0],xmm8[1],xmm13[1]
-; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = xmm8[0,1],mem[0,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm0[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm13[0],xmm5[1],xmm13[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm1[0,3]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm13 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $238, (%rsp), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm6 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm13[0],xmm6[1],xmm13[1]
-; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm6 = xmm6[0,1],mem[0,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm15[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm13[0],xmm4[1],xmm13[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm12[0,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,0],xmm0[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm2[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm0[0,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[2,3,2,3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[1,0],xmm2[0,0]
+; SSE-NEXT:    movaps %xmm2, %xmm15
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[3,1],xmm0[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,2],xmm15[0,2]
 ; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm15 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm13 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm15[0],xmm13[1],xmm15[1]
-; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm13 = xmm13[0,1],mem[0,3]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm15 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
-; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm15 = xmm15[0,1],mem[0,3]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[0,1],mem[0,3]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
+; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = xmm14[0,1],mem[0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm2[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm15[0],xmm9[1],xmm15[1]
+; SSE-NEXT:    shufps $196, (%rsp), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = xmm9[0,1],mem[0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm11[0],xmm5[1],xmm11[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm12[0,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
 ; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = xmm0[0,1],mem[0,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm11[0],xmm2[1],xmm11[1]
+; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[0,1],mem[0,3]
+; SSE-NEXT:    movaps %xmm2, %xmm12
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm11[0],xmm8[1],xmm11[1]
+; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = xmm8[0,1],mem[0,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm11[0],xmm3[1],xmm11[1]
+; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[0,1],mem[0,3]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm2, 96(%rsi)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm2, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm11, 112(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm11, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm11, 64(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm11, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm11, 80(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm11, 16(%rsi)
+; SSE-NEXT:    movaps %xmm4, 96(%rdx)
+; SSE-NEXT:    movaps %xmm7, 32(%rdx)
+; SSE-NEXT:    movaps %xmm13, 112(%rdx)
+; SSE-NEXT:    movaps %xmm10, 48(%rdx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, 112(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, 48(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, 64(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, (%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, 80(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, 16(%rsi)
-; SSE-NEXT:    movaps %xmm3, 96(%rdx)
-; SSE-NEXT:    movaps %xmm11, 32(%rdx)
-; SSE-NEXT:    movaps %xmm7, 112(%rdx)
-; SSE-NEXT:    movaps %xmm9, 48(%rdx)
-; SSE-NEXT:    movaps %xmm10, 64(%rdx)
+; SSE-NEXT:    movaps %xmm2, 64(%rdx)
+; SSE-NEXT:    movaps %xmm6, (%rdx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, (%rdx)
-; SSE-NEXT:    movaps %xmm14, 80(%rdx)
+; SSE-NEXT:    movaps %xmm2, 80(%rdx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm2, 16(%rdx)
-; SSE-NEXT:    movaps %xmm0, 96(%rcx)
-; SSE-NEXT:    movaps %xmm1, 112(%rcx)
-; SSE-NEXT:    movaps %xmm15, 64(%rcx)
-; SSE-NEXT:    movaps %xmm13, 80(%rcx)
-; SSE-NEXT:    movaps %xmm4, 32(%rcx)
-; SSE-NEXT:    movaps %xmm6, 48(%rcx)
-; SSE-NEXT:    movaps %xmm5, (%rcx)
-; SSE-NEXT:    movaps %xmm8, 16(%rcx)
-; SSE-NEXT:    addq $392, %rsp # imm = 0x188
+; SSE-NEXT:    movaps %xmm3, 96(%rcx)
+; SSE-NEXT:    movaps %xmm8, 112(%rcx)
+; SSE-NEXT:    movaps %xmm12, 64(%rcx)
+; SSE-NEXT:    movaps %xmm0, 80(%rcx)
+; SSE-NEXT:    movaps %xmm5, 32(%rcx)
+; SSE-NEXT:    movaps %xmm9, 48(%rcx)
+; SSE-NEXT:    movaps %xmm1, (%rcx)
+; SSE-NEXT:    movaps %xmm14, 16(%rcx)
+; SSE-NEXT:    addq $360, %rsp # imm = 0x168
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i32_stride3_vf32:
@@ -2230,460 +2213,456 @@ define void @load_i32_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 define void @load_i32_stride3_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2) nounwind {
 ; SSE-LABEL: load_i32_stride3_vf64:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $1112, %rsp # imm = 0x458
+; SSE-NEXT:    subq $1000, %rsp # imm = 0x3E8
 ; SSE-NEXT:    movaps 624(%rdi), %xmm2
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps 656(%rdi), %xmm4
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 640(%rdi), %xmm10
-; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 432(%rdi), %xmm6
-; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 464(%rdi), %xmm5
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 448(%rdi), %xmm11
-; SSE-NEXT:    movaps %xmm11, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movaps 240(%rdi), %xmm7
+; SSE-NEXT:    movaps 640(%rdi), %xmm8
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 432(%rdi), %xmm9
+; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 464(%rdi), %xmm7
 ; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 272(%rdi), %xmm3
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 448(%rdi), %xmm11
+; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 240(%rdi), %xmm12
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 272(%rdi), %xmm6
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps 256(%rdi), %xmm13
 ; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 48(%rdi), %xmm9
+; SSE-NEXT:    movaps 48(%rdi), %xmm3
 ; SSE-NEXT:    movaps 80(%rdi), %xmm1
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 64(%rdi), %xmm12
-; SSE-NEXT:    movaps %xmm12, %xmm0
+; SSE-NEXT:    movaps 64(%rdi), %xmm10
+; SSE-NEXT:    movaps %xmm10, %xmm0
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
-; SSE-NEXT:    movaps %xmm9, %xmm1
-; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm3, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps %xmm13, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm3[1,0]
-; SSE-NEXT:    movaps %xmm7, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm6[1,0]
+; SSE-NEXT:    movaps %xmm12, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps %xmm11, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm5[1,0]
-; SSE-NEXT:    movaps %xmm6, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm10, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm7[1,0]
+; SSE-NEXT:    movaps %xmm9, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm8, %xmm0
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm4[1,0]
-; SSE-NEXT:    movaps %xmm2, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps 16(%rdi), %xmm0
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 32(%rdi), %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
-; SSE-NEXT:    movaps (%rdi), %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 224(%rdi), %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 32(%rdi), %xmm4
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm4[1,0]
+; SSE-NEXT:    movaps (%rdi), %xmm5
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 224(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps 208(%rdi), %xmm0
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
-; SSE-NEXT:    movaps 192(%rdi), %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 416(%rdi), %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm2[1,0]
+; SSE-NEXT:    movaps 192(%rdi), %xmm6
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 416(%rdi), %xmm4
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps 400(%rdi), %xmm0
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
-; SSE-NEXT:    movaps 384(%rdi), %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 608(%rdi), %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm4[1,0]
+; SSE-NEXT:    movaps 384(%rdi), %xmm4
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 608(%rdi), %xmm4
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps 592(%rdi), %xmm0
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
-; SSE-NEXT:    movaps 576(%rdi), %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 176(%rdi), %xmm10
-; SSE-NEXT:    movaps 160(%rdi), %xmm8
-; SSE-NEXT:    movaps %xmm8, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm10[1,0]
-; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 144(%rdi), %xmm2
-; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm4[1,0]
+; SSE-NEXT:    movaps 576(%rdi), %xmm7
+; SSE-NEXT:    movaps %xmm7, (%rsp) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 176(%rdi), %xmm9
+; SSE-NEXT:    movaps 160(%rdi), %xmm6
+; SSE-NEXT:    movaps %xmm6, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm9[1,0]
+; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 144(%rdi), %xmm8
+; SSE-NEXT:    movaps %xmm8, %xmm4
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 368(%rdi), %xmm2
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 368(%rdi), %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 352(%rdi), %xmm15
-; SSE-NEXT:    movaps %xmm15, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
-; SSE-NEXT:    movaps 336(%rdi), %xmm14
-; SSE-NEXT:    movaps %xmm14, %xmm1
-; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 560(%rdi), %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 352(%rdi), %xmm12
+; SSE-NEXT:    movaps %xmm12, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm2[1,0]
+; SSE-NEXT:    movaps 336(%rdi), %xmm13
+; SSE-NEXT:    movaps %xmm13, %xmm4
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 560(%rdi), %xmm4
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps 544(%rdi), %xmm0
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
-; SSE-NEXT:    movaps 528(%rdi), %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 752(%rdi), %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm4[1,0]
+; SSE-NEXT:    movaps 528(%rdi), %xmm7
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 752(%rdi), %xmm7
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps 736(%rdi), %xmm0
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[1,0]
-; SSE-NEXT:    movaps 720(%rdi), %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 128(%rdi), %xmm6
-; SSE-NEXT:    movaps 112(%rdi), %xmm4
-; SSE-NEXT:    movaps %xmm4, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm6[1,0]
-; SSE-NEXT:    movaps 96(%rdi), %xmm3
-; SSE-NEXT:    movaps %xmm3, %xmm7
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,3],xmm1[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm7[1,0]
+; SSE-NEXT:    movaps 720(%rdi), %xmm7
 ; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 320(%rdi), %xmm13
-; SSE-NEXT:    movaps 304(%rdi), %xmm11
-; SSE-NEXT:    movaps %xmm11, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm13[1,0]
-; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 288(%rdi), %xmm5
-; SSE-NEXT:    movaps %xmm5, %xmm7
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 128(%rdi), %xmm5
+; SSE-NEXT:    movaps 112(%rdi), %xmm4
+; SSE-NEXT:    movaps %xmm4, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm5[1,0]
 ; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,3],xmm1[0,2]
+; SSE-NEXT:    movaps 96(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm7
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,3],xmm0[0,2]
 ; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 512(%rdi), %xmm0
+; SSE-NEXT:    movaps 320(%rdi), %xmm15
+; SSE-NEXT:    movaps 304(%rdi), %xmm14
+; SSE-NEXT:    movaps %xmm14, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm15[1,0]
+; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 288(%rdi), %xmm11
+; SSE-NEXT:    movaps %xmm11, %xmm7
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 512(%rdi), %xmm7
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 496(%rdi), %xmm0
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 496(%rdi), %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[1,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm7[1,0]
 ; SSE-NEXT:    movaps 480(%rdi), %xmm7
 ; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,3],xmm1[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,3],xmm0[0,2]
 ; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps 704(%rdi), %xmm7
 ; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 688(%rdi), %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm7[1,0]
+; SSE-NEXT:    movaps 688(%rdi), %xmm0
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm7[1,0]
 ; SSE-NEXT:    movaps 672(%rdi), %xmm7
 ; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,3],xmm1[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,3],xmm0[0,2]
 ; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[1,0],xmm12[0,0]
-; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm12 = xmm12[3,1],mem[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,2],xmm12[0,2]
-; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,0],xmm10[0,0]
+; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = xmm10[3,1],mem[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm10[0,2]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[1,0],xmm6[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[3,1],xmm9[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm6[0,2]
 ; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm8[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[3,1],xmm10[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm8[0,2]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,0],xmm4[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,1],xmm6[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm4[0,2]
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[1,0],xmm1[0,0]
-; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[3,1],mem[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm1[0,2]
-; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm9, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm3[0,0]
-; SSE-NEXT:    movaps %xmm3, %xmm1
-; SSE-NEXT:    movaps %xmm3, %xmm12
-; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[3,1],mem[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[1,0],xmm15[0,0]
-; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm15 = xmm15[3,1],mem[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,2],xmm15[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm4[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,1],xmm5[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm4[0,2]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm0[0,0]
+; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[3,1],mem[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm0[0,0]
+; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[3,1],mem[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm12, %xmm0
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[1,0],xmm12[0,0]
+; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[3,1],mem[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,0],xmm11[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[3,1],xmm13[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm11[0,2]
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[1,0],xmm14[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[3,1],xmm15[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,2],xmm14[0,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm1[2,3,2,3]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps (%rsp), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm1[0,0]
-; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[3,1],mem[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm7, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm4[0,0]
-; SSE-NEXT:    movaps %xmm4, %xmm1
-; SSE-NEXT:    movaps %xmm4, %xmm8
-; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[3,1],mem[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm0[0,0]
+; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[3,1],mem[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm1[2,3,2,3]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm1[0,0]
-; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[3,1],mem[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm0[0,0]
+; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[3,1],mem[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm2[2,3,2,3]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm1[0,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm5[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm0[0,0]
+; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[3,1],mem[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm14[2,3,2,3]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm1[0,0]
-; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[3,1],mem[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[1,0],xmm0[0,0]
+; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[3,1],mem[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,2],xmm0[0,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm1[2,3,2,3]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm14[0,0]
-; SSE-NEXT:    movaps %xmm14, %xmm1
-; SSE-NEXT:    movaps %xmm14, %xmm3
-; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[3,1],mem[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm0[0,0]
+; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[3,1],mem[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%rsp), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[2,3,2,3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm0[0,0]
+; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[3,1],mem[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm1, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm2[2,3,2,3]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm0[0,0]
+; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[3,1],mem[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[2,3,2,3]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm1[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[1,0],xmm1[0,0]
 ; SSE-NEXT:    shufps $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = xmm1[3,1],mem[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,2],xmm1[0,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm2[0,0]
+; SSE-NEXT:    movaps %xmm2, %xmm15
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[1,0],xmm1[0,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm4[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,2],xmm1[0,2]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm2[2,3,2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm0[0,0]
-; SSE-NEXT:    movaps %xmm2, %xmm11
-; SSE-NEXT:    movaps %xmm0, %xmm1
-; SSE-NEXT:    movaps %xmm0, %xmm2
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm0[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,2],xmm1[0,2]
-; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm1[0],xmm10[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,1],xmm0[0,3]
-; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[3,1],xmm1[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm15[0,2]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm2[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm15[0],xmm4[1],xmm15[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm1[0,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
 ; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = xmm0[0,1],mem[0,3]
 ; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm15 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm6[0,3]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm13 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
-; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm13 = xmm13[0,1],mem[0,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm9[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
+; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0,1],mem[0,3]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
+; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0,1],mem[0,3]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
+; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0,1],mem[0,3]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
+; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[0,1],mem[0,3]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm15[0],xmm2[1],xmm15[1]
+; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[0,1],mem[0,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1]
+; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[0,1],mem[0,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm15[0],xmm12[1],xmm15[1]
 ; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm12 = xmm12[0,1],mem[0,3]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm11 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
-; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm11 = xmm11[0,1],mem[0,3]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm10 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
-; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm10 = xmm10[0,1],mem[0,3]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm9 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
-; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm9 = xmm9[0,1],mem[0,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm7[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm15[0],xmm13[1],xmm15[1]
+; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm13 = xmm13[0,1],mem[0,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm15[0],xmm8[1],xmm15[1]
 ; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm8 = xmm8[0,1],mem[0,3]
-; SSE-NEXT:    pshufd $85, (%rsp), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm7 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm15[0],xmm10[1],xmm15[1]
+; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = xmm10[0,1],mem[0,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm15[0],xmm6[1],xmm15[1]
+; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm6 = xmm6[0,1],mem[0,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm15[0],xmm7[1],xmm15[1]
 ; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm7 = xmm7[0,1],mem[0,3]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm6 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm5[0,3]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm5 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
-; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm5 = xmm5[0,1],mem[0,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm15[0],xmm3[1],xmm15[1]
 ; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm3 = xmm3[0,1],mem[0,3]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = xmm2[0,1],mem[0,3]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm4[0,3]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
-; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[0,1],mem[0,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 224(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 160(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 96(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 32(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 240(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 176(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 112(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 48(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 192(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 128(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 64(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, (%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 208(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 144(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 80(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 16(%rsi)
-; SSE-NEXT:    movaps %xmm14, 224(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 240(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 192(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 208(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 160(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 176(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 128(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 144(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 96(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 112(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 64(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 80(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 32(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 48(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, (%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, 16(%rdx)
-; SSE-NEXT:    movaps %xmm0, 240(%rcx)
-; SSE-NEXT:    movaps %xmm1, 224(%rcx)
-; SSE-NEXT:    movaps %xmm2, 208(%rcx)
-; SSE-NEXT:    movaps %xmm3, 192(%rcx)
-; SSE-NEXT:    movaps %xmm5, 176(%rcx)
-; SSE-NEXT:    movaps %xmm6, 160(%rcx)
-; SSE-NEXT:    movaps %xmm7, 144(%rcx)
-; SSE-NEXT:    movaps %xmm8, 128(%rcx)
-; SSE-NEXT:    movaps %xmm9, 112(%rcx)
-; SSE-NEXT:    movaps %xmm10, 96(%rcx)
-; SSE-NEXT:    movaps %xmm11, 80(%rcx)
-; SSE-NEXT:    movaps %xmm12, 64(%rcx)
-; SSE-NEXT:    movaps %xmm13, 48(%rcx)
-; SSE-NEXT:    movaps %xmm15, 32(%rcx)
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm15[0],xmm5[1],xmm15[1]
+; SSE-NEXT:    shufps $196, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = xmm5[0,1],mem[0,3]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 16(%rcx)
+; SSE-NEXT:    movaps %xmm0, 224(%rsi)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, (%rcx)
-; SSE-NEXT:    addq $1112, %rsp # imm = 0x458
+; SSE-NEXT:    movaps %xmm0, 160(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm15, 16(%rsi)
+; SSE-NEXT:    movaps %xmm9, 224(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rdx)
+; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 192(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 208(%rdx)
+; SSE-NEXT:    movaps %xmm14, 160(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rdx)
+; SSE-NEXT:    movaps %xmm11, 96(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rdx)
+; SSE-NEXT:    movaps %xmm5, 240(%rcx)
+; SSE-NEXT:    movaps %xmm3, 224(%rcx)
+; SSE-NEXT:    movaps %xmm7, 208(%rcx)
+; SSE-NEXT:    movaps %xmm6, 192(%rcx)
+; SSE-NEXT:    movaps %xmm10, 176(%rcx)
+; SSE-NEXT:    movaps %xmm8, 160(%rcx)
+; SSE-NEXT:    movaps %xmm13, 144(%rcx)
+; SSE-NEXT:    movaps %xmm12, 128(%rcx)
+; SSE-NEXT:    movaps %xmm1, 112(%rcx)
+; SSE-NEXT:    movaps %xmm2, 96(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rcx)
+; SSE-NEXT:    movaps %xmm4, (%rcx)
+; SSE-NEXT:    addq $1000, %rsp # imm = 0x3E8
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i32_stride3_vf64:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll
index f01aa90e3efc5..7df38621db899 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-5.ll
@@ -1296,212 +1296,204 @@ define void @load_i32_stride5_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i32_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4) nounwind {
 ; SSE-LABEL: load_i32_stride5_vf16:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $312, %rsp # imm = 0x138
-; SSE-NEXT:    movdqa 288(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 272(%rdi), %xmm3
+; SSE-NEXT:    subq $280, %rsp # imm = 0x118
+; SSE-NEXT:    movdqa 288(%rdi), %xmm11
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 272(%rdi), %xmm2
 ; SSE-NEXT:    movdqa 240(%rdi), %xmm14
-; SSE-NEXT:    movdqa 256(%rdi), %xmm8
-; SSE-NEXT:    movdqa (%rdi), %xmm11
-; SSE-NEXT:    movdqa 16(%rdi), %xmm15
-; SSE-NEXT:    movdqa 32(%rdi), %xmm5
-; SSE-NEXT:    movdqa 48(%rdi), %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 256(%rdi), %xmm4
+; SSE-NEXT:    movdqa (%rdi), %xmm5
+; SSE-NEXT:    movdqa 16(%rdi), %xmm7
+; SSE-NEXT:    movdqa 32(%rdi), %xmm8
+; SSE-NEXT:    movdqa 48(%rdi), %xmm12
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 208(%rdi), %xmm6
 ; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 192(%rdi), %xmm7
-; SSE-NEXT:    movdqa 160(%rdi), %xmm10
-; SSE-NEXT:    movdqa 176(%rdi), %xmm13
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    movdqa 192(%rdi), %xmm9
+; SSE-NEXT:    movdqa 160(%rdi), %xmm3
+; SSE-NEXT:    movdqa 176(%rdi), %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm1, %xmm10
+; SSE-NEXT:    movdqa %xmm3, %xmm1
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm7, %xmm9
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm6[2],xmm0[3],xmm6[3]
 ; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
 ; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    movdqa %xmm11, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm5, %xmm1
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm5, %xmm7
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm8, %xmm6
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm12[2],xmm0[3],xmm12[3]
 ; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
 ; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
 ; SSE-NEXT:    movdqa %xmm14, %xmm1
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm3, %xmm5
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 80(%rdi), %xmm4
-; SSE-NEXT:    movdqa 96(%rdi), %xmm11
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 128(%rdi), %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm2, %xmm12
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 112(%rdi), %xmm12
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm11[2],xmm0[3],xmm11[3]
 ; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
 ; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 80(%rdi), %xmm0
+; SSE-NEXT:    movdqa 96(%rdi), %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm2, %xmm8
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    movdqa 128(%rdi), %xmm11
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 112(%rdi), %xmm13
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm11[2],xmm1[3],xmm11[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movdqa 224(%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm9[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    movdqa 224(%rdi), %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[0,0,1,1]
+; SSE-NEXT:    movdqa %xmm11, %xmm15
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm9[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm1[0],xmm11[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm2[0],xmm11[1]
+; SSE-NEXT:    movapd %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm7, (%rsp) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
-; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, %xmm9
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE-NEXT:    movdqa 64(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm6[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm2[0],xmm9[1]
+; SSE-NEXT:    movapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm4, %xmm6
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
-; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm3
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm8, (%rsp) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE-NEXT:    movdqa 304(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
-; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm4, %xmm7
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movdqa 144(%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm12[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm12[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm2[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
-; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[2,2,3,3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm10 = xmm10[2],xmm1[2],xmm10[3],xmm1[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[0,0,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm10[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm0
-; SSE-NEXT:    movdqa %xmm9, %xmm10
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,2,3,3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movdqa 144(%rdi), %xmm4
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,0,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movdqa %xmm4, %xmm12
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm13[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm2[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[3,3,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[0,0,1,1]
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm3[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[2,2,3,3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm5[0,0,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm3[0],xmm12[1],xmm3[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm0[0],xmm12[1]
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[2,2,3,3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm7[0,0,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm0[0],xmm9[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm14, %xmm8
-; SSE-NEXT:    movdqa %xmm14, %xmm15
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[3,3,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm3[2],xmm5[3],xmm3[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm5[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm6[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm14[3,3,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm14 = xmm14[2],xmm3[2],xmm14[3],xmm3[3]
+; SSE-NEXT:    movdqa %xmm14, %xmm3
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm0[0],xmm8[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm4, %xmm6
-; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[3,3,3,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm0[0],xmm6[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm5, %xmm2
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm14[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm6[0],xmm13[1],xmm6[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm3[0],xmm13[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm8[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[3,3,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = mem[0,0,1,1]
+; SSE-NEXT:    movdqa %xmm12, %xmm8
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm12[0],xmm7[1],xmm12[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm0[0],xmm7[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm10, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[2,2,2,2]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm12[0],xmm2[1],xmm12[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm2[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm15
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm11, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm11[2,2,2,2]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[2,2,2,2]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm6[0],xmm4[1],xmm6[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm4[0],xmm14[1]
+; SSE-NEXT:    movdqa %xmm8, %xmm9
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm8[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,2,2,2]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm4[0],xmm8[1],xmm4[1]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm13[0],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm10[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm13[0],xmm15[1],xmm13[1]
-; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm13 = mem[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm13 = xmm13[2],mem[2],xmm13[3],mem[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm15[0],xmm13[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm4[1,1,1,1]
-; SSE-NEXT:    movdqa (%rsp), %xmm14 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm15[0],xmm14[1],xmm15[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm5[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm15 = xmm15[2],mem[2],xmm15[3],mem[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm14[0],xmm15[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm3[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm14[0],xmm11[1],xmm14[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm7[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm14 = xmm14[2],mem[2],xmm14[3],mem[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm11[0],xmm14[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm5[0],xmm8[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm12[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = xmm10[2],mem[2],xmm10[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm5[0],xmm10[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm11[1,1,1,1]
+; SSE-NEXT:    movdqa (%rsp), %xmm5 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm15[2],xmm2[3],xmm15[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm5[0],xmm2[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm6[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm9[2],xmm0[3],xmm9[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm3, 16(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm11, 48(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm11, (%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm11, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 32(%rsi)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm3, 16(%rdx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
@@ -1510,21 +1502,23 @@ define void @load_i32_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movaps %xmm3, (%rdx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm3, 32(%rdx)
-; SSE-NEXT:    movapd %xmm9, 16(%rcx)
-; SSE-NEXT:    movapd %xmm12, 48(%rcx)
+; SSE-NEXT:    movapd %xmm7, 16(%rcx)
+; SSE-NEXT:    movapd %xmm13, 48(%rcx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm3, (%rcx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm3, 32(%rcx)
-; SSE-NEXT:    movapd %xmm1, 16(%r8)
-; SSE-NEXT:    movapd %xmm2, 48(%r8)
-; SSE-NEXT:    movapd %xmm6, (%r8)
-; SSE-NEXT:    movapd %xmm8, 32(%r8)
-; SSE-NEXT:    movapd %xmm14, 16(%r9)
-; SSE-NEXT:    movapd %xmm15, 48(%r9)
-; SSE-NEXT:    movapd %xmm13, (%r9)
-; SSE-NEXT:    movapd %xmm0, 32(%r9)
-; SSE-NEXT:    addq $312, %rsp # imm = 0x138
+; SSE-NEXT:    movapd %xmm8, 16(%r8)
+; SSE-NEXT:    movapd %xmm14, 48(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, (%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, 32(%r8)
+; SSE-NEXT:    movapd %xmm0, 16(%r9)
+; SSE-NEXT:    movapd %xmm1, 48(%r9)
+; SSE-NEXT:    movapd %xmm2, (%r9)
+; SSE-NEXT:    movapd %xmm10, 32(%r9)
+; SSE-NEXT:    addq $280, %rsp # imm = 0x118
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i32_stride5_vf16:
@@ -2522,501 +2516,496 @@ define void @load_i32_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 define void @load_i32_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4) nounwind {
 ; SSE-LABEL: load_i32_stride5_vf32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $904, %rsp # imm = 0x388
-; SSE-NEXT:    movdqa (%rdi), %xmm11
+; SSE-NEXT:    subq $824, %rsp # imm = 0x338
+; SSE-NEXT:    movdqa (%rdi), %xmm14
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 16(%rdi), %xmm5
 ; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 32(%rdi), %xmm9
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 48(%rdi), %xmm8
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 448(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 432(%rdi), %xmm4
-; SSE-NEXT:    movdqa 400(%rdi), %xmm10
-; SSE-NEXT:    movdqa 416(%rdi), %xmm14
-; SSE-NEXT:    movdqa 128(%rdi), %xmm6
+; SSE-NEXT:    movdqa 32(%rdi), %xmm6
 ; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 112(%rdi), %xmm7
-; SSE-NEXT:    movdqa 80(%rdi), %xmm12
-; SSE-NEXT:    movdqa 96(%rdi), %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm1, %xmm15
-; SSE-NEXT:    movdqa %xmm12, %xmm1
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,2,2,2]
+; SSE-NEXT:    movdqa 48(%rdi), %xmm10
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 448(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movdqa 432(%rdi), %xmm3
+; SSE-NEXT:    movdqa 400(%rdi), %xmm11
+; SSE-NEXT:    movdqa 416(%rdi), %xmm4
+; SSE-NEXT:    movdqa 128(%rdi), %xmm7
 ; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm6[2],xmm0[3],xmm6[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm10, %xmm1
-; SSE-NEXT:    movdqa %xmm10, %xmm6
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    movdqa %xmm11, %xmm5
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm8[2],xmm0[3],xmm8[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 320(%rdi), %xmm2
+; SSE-NEXT:    movdqa 112(%rdi), %xmm8
+; SSE-NEXT:    movdqa 80(%rdi), %xmm9
+; SSE-NEXT:    movdqa 96(%rdi), %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm2, %xmm12
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm8, %xmm13
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm7[2],xmm1[3],xmm7[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm4, %xmm8
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm11, %xmm2
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm3, %xmm7
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm10[2],xmm1[3],xmm10[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 320(%rdi), %xmm0
 ; SSE-NEXT:    movdqa 336(%rdi), %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm1, %xmm9
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 368(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 352(%rdi), %xmm11
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 240(%rdi), %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm0, %xmm6
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    movdqa 368(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 352(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 240(%rdi), %xmm3
 ; SSE-NEXT:    movdqa 256(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 288(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm3, %xmm2
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    movdqa 288(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 272(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 560(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 560(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 576(%rdi), %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm1, %xmm8
-; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 608(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 592(%rdi), %xmm13
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 160(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    movdqa 608(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 592(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,2,2,2]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 160(%rdi), %xmm4
 ; SSE-NEXT:    movdqa 176(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 208(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 192(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 480(%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    movdqa 208(%rdi), %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 192(%rdi), %xmm14
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm2[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 480(%rdi), %xmm15
 ; SSE-NEXT:    movdqa 496(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 528(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm15, %xmm5
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; SSE-NEXT:    movdqa 528(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 512(%rdi), %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,2,2,2]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm12[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm15, %xmm10
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    movdqa 144(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
-; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm5[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm12[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm9[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; SSE-NEXT:    movdqa 144(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm13[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm2[0],xmm12[1],xmm2[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm5[0],xmm12[1]
+; SSE-NEXT:    movapd %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm11[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
 ; SSE-NEXT:    movdqa 464(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm4[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm1[0],xmm6[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm0[0],xmm6[1]
-; SSE-NEXT:    movapd %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm5, %xmm12
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm7[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm5[0],xmm8[1]
+; SSE-NEXT:    movapd %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm10[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
 ; SSE-NEXT:    movdqa 64(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
-; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm6[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
 ; SSE-NEXT:    movdqa 384(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm11[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
-; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm13[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; SSE-NEXT:    movdqa 304(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movdqa 304(%rdi), %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,0,1,1]
-; SSE-NEXT:    movdqa %xmm2, %xmm14
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm7 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm0[0],xmm7[1]
-; SSE-NEXT:    movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm8, %xmm11
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm6[2,3,2,3]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = mem[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
 ; SSE-NEXT:    movdqa 624(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm13[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[0,0,1,1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; SSE-NEXT:    movdqa 224(%rdi), %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,0,1,1]
+; SSE-NEXT:    movdqa %xmm7, %xmm12
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm14[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm2[0],xmm7[1],xmm2[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm5[0],xmm7[1]
+; SSE-NEXT:    movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm15[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; SSE-NEXT:    movdqa 544(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm0[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm2[0],xmm7[1],xmm2[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm5[0],xmm7[1]
+; SSE-NEXT:    movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm9[3,3,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm2[2],xmm9[3],xmm2[3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm9[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm10[3,3,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm10 = xmm10[2],xmm2[2],xmm10[3],xmm2[3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm10[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm3[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[3,3,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm4[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm1[3,3,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    pshufd $80, (%rsp), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm0[3,3,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm1[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm6[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = xmm4[0],mem[0],xmm4[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm1[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm14[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm15[3,3,3,3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm6[2],xmm15[3],xmm6[3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[0,0,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm4[0],xmm14[1],xmm4[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm15[0],xmm14[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm2[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = xmm8[0],mem[0],xmm8[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm8[0],xmm2[1]
 ; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movdqa 224(%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm2[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = xmm5[0],mem[0],xmm5[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm5[0],xmm2[1]
 ; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movdqa 544(%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, (%rsp) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm2[2,2,2,2]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm3[0],xmm2[1]
 ; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[2,2,3,3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[0,0,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,2,3,3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm12 = xmm12[2],xmm1[2],xmm12[3],xmm1[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[0,0,1,1]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm12[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[2,2,3,3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm14[0],xmm1[1],xmm14[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[2,2,3,3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[0,0,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[0,0,1,1]
-; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm2[2,2,2,2]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
+; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,2,3,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,0,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[2,2,3,3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,2,2,2]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[0,0,1,1]
-; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[2,2,3,3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[0,0,1,1]
-; SSE-NEXT:    movdqa (%rsp), %xmm7 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm15[0],xmm7[1],xmm15[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm7[0],xmm0[1]
+; SSE-NEXT:    movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,2,3,3]
+; SSE-NEXT:    movdqa (%rsp), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[2,2,2,2]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,2,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
-; SSE-NEXT:    movapd %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm12, %xmm1
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
-; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm6, %xmm13
-; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm8[0],xmm13[1],xmm8[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm13[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, (%rsp) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm9, %xmm12
-; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm0[0],xmm12[1]
-; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm14, %xmm10
-; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm0[0],xmm10[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm11, %xmm9
-; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm0[0],xmm9[1]
-; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm15, %xmm7
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm0[0],xmm7[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = mem[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = xmm8[2],mem[2],xmm8[3],mem[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm2[0],xmm8[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm6 = mem[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[2,2,2,2]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm12[0],xmm9[1],xmm12[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm9[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm13[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,2,2,2]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
+; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm4[0],xmm7[1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm4[0],xmm9[1],xmm4[1]
+; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm9[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm4[0],xmm9[1],xmm4[1]
+; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm9[0],xmm4[1]
+; SSE-NEXT:    movapd %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm4[0],xmm9[1],xmm4[1]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = xmm11[2],mem[2],xmm11[3],mem[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm9[0],xmm11[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm10[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm4[0],xmm9[1],xmm4[1]
 ; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm6 = xmm6[2],mem[2],xmm6[3],mem[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm2[0],xmm6[1]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm5 = mem[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm5 = xmm5[2],mem[2],xmm5[3],mem[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm2[0],xmm5[1]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = xmm4[2],mem[2],xmm4[3],mem[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm2[0],xmm4[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
-; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[2,2,2,2]
+; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm9[0],xmm6[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm15[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm4[0],xmm9[1],xmm4[1]
 ; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm15[0],xmm3[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[2,2,2,2]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm9[0],xmm3[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm8[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm4[0],xmm9[1],xmm4[1]
 ; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
-; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,2,2,2]
-; SSE-NEXT:    punpckhdq (%rsp), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm9[0],xmm2[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm12[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm4[0],xmm9[1],xmm4[1]
+; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = xmm1[2],mem[2],xmm1[3],mem[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm15[0],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
-; SSE-NEXT:    pshufd $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[2,2,2,2]
-; SSE-NEXT:    punpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[2],mem[2],xmm0[3],mem[3]
-; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm15[0],xmm0[1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 96(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 32(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 112(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 48(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 64(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, (%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 80(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 16(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm11, 96(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm11, 32(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm11, 112(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm11, 48(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm11, 64(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm11, (%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 80(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, 16(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm11, 96(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm11, 112(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm11, 64(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm11, 80(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm11, 32(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm11, 48(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm11, (%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm11, 16(%rcx)
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm9[0],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm4[0],xmm9[1],xmm4[1]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm13[2],xmm0[3],xmm13[3]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm9[0],xmm0[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 96(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 112(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 64(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 80(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 96(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 32(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 112(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 48(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 64(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, (%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 80(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 16(%rdx)
+; SSE-NEXT:    movapd %xmm14, 96(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 112(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 64(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 80(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 32(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 48(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, (%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 16(%rcx)
 ; SSE-NEXT:    movapd %xmm7, 112(%r8)
-; SSE-NEXT:    movapd %xmm9, 96(%r8)
-; SSE-NEXT:    movapd %xmm10, 80(%r8)
-; SSE-NEXT:    movapd %xmm12, 64(%r8)
-; SSE-NEXT:    movapd %xmm13, 48(%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm7, 32(%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm7, 16(%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm7, (%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 96(%r8)
+; SSE-NEXT:    movaps (%rsp), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 80(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 64(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 48(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 32(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, 16(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm4, (%r8)
 ; SSE-NEXT:    movapd %xmm0, 112(%r9)
 ; SSE-NEXT:    movapd %xmm1, 96(%r9)
 ; SSE-NEXT:    movapd %xmm2, 80(%r9)
 ; SSE-NEXT:    movapd %xmm3, 64(%r9)
-; SSE-NEXT:    movapd %xmm4, 48(%r9)
-; SSE-NEXT:    movapd %xmm5, 32(%r9)
-; SSE-NEXT:    movapd %xmm6, 16(%r9)
-; SSE-NEXT:    movapd %xmm8, (%r9)
-; SSE-NEXT:    addq $904, %rsp # imm = 0x388
+; SSE-NEXT:    movapd %xmm6, 48(%r9)
+; SSE-NEXT:    movapd %xmm11, 32(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r9)
+; SSE-NEXT:    addq $824, %rsp # imm = 0x338
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i32_stride5_vf32:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll
index aae4d9fa15e24..939a11cb093f6 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-6.ll
@@ -880,122 +880,117 @@ define void @load_i32_stride6_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i32_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5) nounwind {
 ; SSE-LABEL: load_i32_stride6_vf8:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    movdqa 144(%rdi), %xmm4
-; SSE-NEXT:    movdqa 160(%rdi), %xmm2
-; SSE-NEXT:    movdqa 96(%rdi), %xmm6
-; SSE-NEXT:    movdqa 112(%rdi), %xmm3
-; SSE-NEXT:    movdqa 64(%rdi), %xmm5
-; SSE-NEXT:    movdqa (%rdi), %xmm10
-; SSE-NEXT:    movdqa 16(%rdi), %xmm0
+; SSE-NEXT:    movdqa 144(%rdi), %xmm12
+; SSE-NEXT:    movdqa 160(%rdi), %xmm4
+; SSE-NEXT:    movdqa 96(%rdi), %xmm7
+; SSE-NEXT:    movdqa 112(%rdi), %xmm5
+; SSE-NEXT:    movdqa 64(%rdi), %xmm3
+; SSE-NEXT:    movdqa (%rdi), %xmm9
+; SSE-NEXT:    movdqa 16(%rdi), %xmm6
 ; SSE-NEXT:    movdqa 48(%rdi), %xmm8
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm0, %xmm11
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm10[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm10, %xmm7
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,2,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm8[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm7[0],xmm9[1]
-; SSE-NEXT:    movapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[2,3,2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm6[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm6, %xmm9
-; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[2,2,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm4[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm9[0],xmm7[1]
-; SSE-NEXT:    movapd %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm9[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm8[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm1[0],xmm10[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm2[0],xmm10[1]
+; SSE-NEXT:    movapd %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm5, %xmm9
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm8[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm8, %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm7, %xmm10
+; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm1[0],xmm10[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm12[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm1[0],xmm11[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm0[0],xmm11[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[3,3,3,3]
-; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm4, %xmm12
+; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm10[0],xmm11[1]
+; SSE-NEXT:    movapd %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm3, %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm8[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm8, %xmm3
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm12, %xmm1
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
 ; SSE-NEXT:    movdqa 80(%rdi), %xmm14
-; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm15[0],xmm12[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm2[0],xmm12[1]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
-; SSE-NEXT:    movdqa 32(%rdi), %xmm7
-; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm7[0],xmm13[1],xmm7[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm13[0],xmm5[1]
-; SSE-NEXT:    movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm4[2,2,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; SSE-NEXT:    movdqa 32(%rdi), %xmm10
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm10[0],xmm13[1],xmm10[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm13[0],xmm15[1]
+; SSE-NEXT:    movapd %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm1[2,2,3,3]
 ; SSE-NEXT:    movdqa 176(%rdi), %xmm15
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
-; SSE-NEXT:    movdqa 128(%rdi), %xmm5
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
+; SSE-NEXT:    movdqa 128(%rdi), %xmm7
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm7[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[3,3,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm10[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[2,3,2,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm14[0],xmm8[1],xmm14[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm0[0],xmm8[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm15[0],xmm4[1],xmm15[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm0[0],xmm4[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm3, %xmm10
-; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,2,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm9[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm10[0],xmm6[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm5[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,2,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm2[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm1[0],xmm10[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm0[0],xmm10[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[3,3,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm7[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm1[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm15[0],xmm9[1],xmm15[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm0[0],xmm9[1]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 16(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, (%rsi)
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm6[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm6[0],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm15[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm5[0],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm10[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm14[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm3[0],xmm11[1],xmm3[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm2[0],xmm11[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm2[0],xmm6[1],xmm2[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm15[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm6[0],xmm4[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, (%rsi)
 ; SSE-NEXT:    movapd %xmm12, 16(%rdx)
-; SSE-NEXT:    movapd %xmm11, (%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, (%rdx)
 ; SSE-NEXT:    movapd %xmm13, 16(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, (%rcx)
-; SSE-NEXT:    movapd %xmm4, 16(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, (%rcx)
+; SSE-NEXT:    movapd %xmm9, 16(%r8)
 ; SSE-NEXT:    movapd %xmm8, (%r8)
-; SSE-NEXT:    movapd %xmm10, 16(%r9)
-; SSE-NEXT:    movapd %xmm6, (%r9)
+; SSE-NEXT:    movapd %xmm0, 16(%r9)
+; SSE-NEXT:    movapd %xmm1, (%r9)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movapd %xmm2, 16(%rax)
-; SSE-NEXT:    movapd %xmm9, (%rax)
+; SSE-NEXT:    movapd %xmm4, 16(%rax)
+; SSE-NEXT:    movapd %xmm11, (%rax)
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i32_stride6_vf8:
@@ -1692,27 +1687,25 @@ define void @load_i32_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i32_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5) nounwind {
 ; SSE-LABEL: load_i32_stride6_vf16:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $408, %rsp # imm = 0x198
-; SSE-NEXT:    movdqa 240(%rdi), %xmm9
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    subq $376, %rsp # imm = 0x178
+; SSE-NEXT:    movdqa 240(%rdi), %xmm11
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 256(%rdi), %xmm3
-; SSE-NEXT:    movdqa 192(%rdi), %xmm10
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 208(%rdi), %xmm4
-; SSE-NEXT:    movdqa 336(%rdi), %xmm14
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 352(%rdi), %xmm5
-; SSE-NEXT:    movdqa 288(%rdi), %xmm15
+; SSE-NEXT:    movdqa 192(%rdi), %xmm8
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 208(%rdi), %xmm6
+; SSE-NEXT:    movdqa 336(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 352(%rdi), %xmm4
+; SSE-NEXT:    movdqa 288(%rdi), %xmm10
 ; SSE-NEXT:    movdqa 304(%rdi), %xmm7
 ; SSE-NEXT:    movdqa 64(%rdi), %xmm12
-; SSE-NEXT:    movdqa (%rdi), %xmm8
-; SSE-NEXT:    movdqa 16(%rdi), %xmm1
+; SSE-NEXT:    movdqa (%rdi), %xmm5
+; SSE-NEXT:    movdqa 16(%rdi), %xmm15
 ; SSE-NEXT:    movdqa 48(%rdi), %xmm13
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm1, %xmm11
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm8, %xmm1
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm5, %xmm1
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,2,3,3]
 ; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -1722,267 +1715,265 @@ define void @load_i32_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
 ; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    movdqa %xmm10, %xmm1
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[0,0,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm9[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
 ; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm4, %xmm14
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm6, %xmm13
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm8, %xmm1
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; SSE-NEXT:    movdqa %xmm3, %xmm2
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[0,0,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
 ; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 96(%rdi), %xmm4
-; SSE-NEXT:    movdqa 112(%rdi), %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm1, %xmm10
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 96(%rdi), %xmm8
+; SSE-NEXT:    movdqa 112(%rdi), %xmm9
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 144(%rdi), %xmm9
+; SSE-NEXT:    movdqa 144(%rdi), %xmm14
 ; SSE-NEXT:    movdqa 160(%rdi), %xmm3
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
 ; SSE-NEXT:    movdqa %xmm3, %xmm6
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[0,0,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm14[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
 ; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[3,3,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[3,3,3,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm13, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm11, %xmm3
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
 ; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm10, %xmm12
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[3,3,3,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm7, %xmm3
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
 ; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[3,3,3,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm3, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm4, %xmm2
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
 ; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[3,3,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[3,3,3,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm9, %xmm14
-; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    movdqa %xmm14, %xmm9
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
 ; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[2,2,3,3]
 ; SSE-NEXT:    movdqa 80(%rdi), %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,0,1,1]
-; SSE-NEXT:    movdqa %xmm2, %xmm10
+; SSE-NEXT:    movdqa %xmm2, %xmm8
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,3,2,3]
-; SSE-NEXT:    movdqa 32(%rdi), %xmm6
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,3,2,3]
+; SSE-NEXT:    movdqa 32(%rdi), %xmm13
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[2,2,3,3]
 ; SSE-NEXT:    movdqa 368(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm15, %xmm9
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,3,2,3]
-; SSE-NEXT:    movdqa 320(%rdi), %xmm8
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
+; SSE-NEXT:    movdqa %xmm12, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,3,2,3]
+; SSE-NEXT:    movdqa 320(%rdi), %xmm3
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[2,2,3,3]
-; SSE-NEXT:    movdqa 272(%rdi), %xmm15
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[0,0,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[2,2,3,3]
+; SSE-NEXT:    movdqa 272(%rdi), %xmm14
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
-; SSE-NEXT:    movdqa 224(%rdi), %xmm3
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,3,2,3]
+; SSE-NEXT:    movdqa 224(%rdi), %xmm12
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[2,2,3,3]
-; SSE-NEXT:    movdqa 176(%rdi), %xmm11
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[0,0,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm9, %xmm7
+; SSE-NEXT:    movdqa 176(%rdi), %xmm9
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,3,2,3]
-; SSE-NEXT:    movdqa 128(%rdi), %xmm4
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,3,2,3]
+; SSE-NEXT:    movdqa 128(%rdi), %xmm2
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[3,3,3,3]
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[3,3,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1]
+; SSE-NEXT:    movdqa %xmm8, %xmm5
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm3, %xmm11
+; SSE-NEXT:    movdqa %xmm3, (%rsp) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm3, %xmm9
-; SSE-NEXT:    movdqa %xmm3, (%rsp) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[3,3,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm12 = mem[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm15, %xmm3
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm15[0],xmm12[1],xmm15[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm0[0],xmm12[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm4, %xmm7
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm14, %xmm10
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm14[0],xmm1[1],xmm14[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[3,3,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm14[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm11[0],xmm15[1],xmm11[1]
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm6, %xmm1
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm7[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm9[0],xmm8[1],xmm9[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm0[0],xmm8[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm15[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm15[0],xmm7[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm14[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm13[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm14[0],xmm15[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm14[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,2,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm11[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm14[0],xmm2[1]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm10[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm14, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm8[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm1[0],xmm4[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm1, %xmm14
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[2,2,3,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,2,3,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm13[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm1[0],xmm10[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm0[0],xmm10[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[1,1,1,1]
+; SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm14[0],xmm0[1]
 ; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm1[0],xmm6[1],xmm1[1]
 ; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm1[0],xmm8[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm0[0],xmm8[1]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm6[0],xmm14[1]
 ; SSE-NEXT:    pshufd $255, (%rsp), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
 ; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm0[0],xmm9[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm5[0],xmm13[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
 ; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 16(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 32(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 48(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, (%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 16(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 32(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 48(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, (%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 16(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 32(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 48(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, (%rcx)
-; SSE-NEXT:    movapd %xmm15, 16(%r8)
-; SSE-NEXT:    movapd %xmm12, 32(%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 48(%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, (%r8)
-; SSE-NEXT:    movapd %xmm2, 16(%r9)
-; SSE-NEXT:    movapd %xmm3, 32(%r9)
-; SSE-NEXT:    movapd %xmm4, 48(%r9)
-; SSE-NEXT:    movapd %xmm5, (%r9)
+; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm1[0],xmm11[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm4[0],xmm11[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm3[0],xmm9[1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 16(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rcx)
+; SSE-NEXT:    movapd %xmm8, 16(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 32(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, 48(%r8)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%r8)
+; SSE-NEXT:    movapd %xmm0, 16(%r9)
+; SSE-NEXT:    movapd %xmm2, 32(%r9)
+; SSE-NEXT:    movapd %xmm15, 48(%r9)
+; SSE-NEXT:    movapd %xmm7, (%r9)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movapd %xmm13, 16(%rax)
-; SSE-NEXT:    movapd %xmm9, 32(%rax)
-; SSE-NEXT:    movapd %xmm8, 48(%rax)
-; SSE-NEXT:    movapd %xmm10, (%rax)
-; SSE-NEXT:    addq $408, %rsp # imm = 0x198
+; SSE-NEXT:    movapd %xmm9, 16(%rax)
+; SSE-NEXT:    movapd %xmm11, 32(%rax)
+; SSE-NEXT:    movapd %xmm13, 48(%rax)
+; SSE-NEXT:    movapd %xmm14, (%rax)
+; SSE-NEXT:    addq $376, %rsp # imm = 0x178
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i32_stride6_vf16:
@@ -3448,103 +3439,98 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-LABEL: load_i32_stride6_vf32:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    subq $1032, %rsp # imm = 0x408
-; SSE-NEXT:    movdqa 64(%rdi), %xmm5
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rdi), %xmm12
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 16(%rdi), %xmm13
-; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 48(%rdi), %xmm9
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 528(%rdi), %xmm7
+; SSE-NEXT:    movdqa 64(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%rdi), %xmm15
+; SSE-NEXT:    movdqa 16(%rdi), %xmm7
 ; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 48(%rdi), %xmm8
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 528(%rdi), %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 544(%rdi), %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 480(%rdi), %xmm8
-; SSE-NEXT:    movdqa 496(%rdi), %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 480(%rdi), %xmm5
+; SSE-NEXT:    movdqa %xmm5, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movdqa 496(%rdi), %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 144(%rdi), %xmm10
 ; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 160(%rdi), %xmm2
-; SSE-NEXT:    movdqa 96(%rdi), %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 112(%rdi), %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm1, %xmm11
+; SSE-NEXT:    movdqa 160(%rdi), %xmm14
+; SSE-NEXT:    movdqa 96(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    movdqa 112(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm2, %xmm6
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm10[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
 ; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm8, %xmm1
-; SSE-NEXT:    movdqa %xmm8, %xmm4
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm5, %xmm1
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,0,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm9[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
 ; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    movdqa %xmm15, %xmm5
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm9[0,0,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
 ; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 384(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 400(%rdi), %xmm14
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 400(%rdi), %xmm15
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; SSE-NEXT:    movdqa 432(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa 448(%rdi), %xmm9
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 448(%rdi), %xmm10
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
 ; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 288(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 304(%rdi), %xmm15
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    movdqa 288(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 304(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; SSE-NEXT:    movdqa 336(%rdi), %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 352(%rdi), %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm2, %xmm12
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 352(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
 ; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 672(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 688(%rdi), %xmm8
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 688(%rdi), %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 720(%rdi), %xmm3
+; SSE-NEXT:    movdqa 720(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 736(%rdi), %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 736(%rdi), %xmm5
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,0,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
 ; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -3554,439 +3540,438 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 240(%rdi), %xmm2
+; SSE-NEXT:    movdqa 240(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 256(%rdi), %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm2, %xmm13
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 256(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
 ; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 576(%rdi), %xmm7
-; SSE-NEXT:    movdqa 592(%rdi), %xmm13
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 576(%rdi), %xmm8
+; SSE-NEXT:    movdqa 592(%rdi), %xmm12
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 624(%rdi), %xmm10
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 640(%rdi), %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm10[0,0,1,1]
+; SSE-NEXT:    movdqa 624(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 640(%rdi), %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
 ; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm9, %xmm2
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
 ; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
+; SSE-NEXT:    pshufd $85, (%rsp), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
 ; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm11, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm14, %xmm2
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
 ; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
 ; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm5, %xmm2
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
 ; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[3,3,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[3,3,3,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[2,3,2,3]
-; SSE-NEXT:    movdqa (%rsp), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm10, %xmm2
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
 ; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[2,3,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm15, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm4, %xmm2
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
 ; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[3,3,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[3,3,3,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm8, %xmm2
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
-; SSE-NEXT:    movapd %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1]
 ; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm14, %xmm5
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm0[0],xmm5[1]
-; SSE-NEXT:    movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[3,3,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm13, %xmm3
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
+; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[3,3,3,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm7, %xmm3
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
 ; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[2,2,3,3]
-; SSE-NEXT:    movdqa 176(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[2,2,3,3]
+; SSE-NEXT:    movdqa 176(%rdi), %xmm8
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
 ; SSE-NEXT:    movdqa 128(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm5, %xmm12
 ; SSE-NEXT:    movdqa 80(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
-; SSE-NEXT:    movdqa 32(%rdi), %xmm13
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,3,2,3]
+; SSE-NEXT:    movdqa 32(%rdi), %xmm6
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[2,2,3,3]
 ; SSE-NEXT:    movdqa 368(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,3,2,3]
 ; SSE-NEXT:    movdqa 320(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[2,2,3,3]
-; SSE-NEXT:    movdqa 272(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[2,2,3,3]
+; SSE-NEXT:    movdqa 272(%rdi), %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[0,0,1,1]
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm12, %xmm14
-; SSE-NEXT:    movdqa 224(%rdi), %xmm6
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,3,2,3]
+; SSE-NEXT:    movdqa 224(%rdi), %xmm15
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[2,2,3,3]
 ; SSE-NEXT:    movdqa 560(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa (%rsp), %xmm9 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,3,2,3]
 ; SSE-NEXT:    movdqa 512(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rsp), %xmm11 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[2,2,3,3]
-; SSE-NEXT:    movdqa 464(%rdi), %xmm15
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[0,0,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[2,2,3,3]
+; SSE-NEXT:    movdqa 464(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
-; SSE-NEXT:    movdqa 416(%rdi), %xmm4
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,3,2,3]
+; SSE-NEXT:    movdqa 416(%rdi), %xmm10
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[2,2,3,3]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
 ; SSE-NEXT:    movdqa 752(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
-; SSE-NEXT:    movdqa 704(%rdi), %xmm12
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 704(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,2,3,3]
-; SSE-NEXT:    movdqa 656(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[2,2,3,3]
+; SSE-NEXT:    movdqa 656(%rdi), %xmm13
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[0,0,1,1]
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,3,2,3]
-; SSE-NEXT:    movdqa 608(%rdi), %xmm5
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,3,2,3]
+; SSE-NEXT:    movdqa 608(%rdi), %xmm7
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    movdqa %xmm13, %xmm9
-; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm6, %xmm5
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[3,3,3,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1]
+; SSE-NEXT:    movdqa %xmm8, %xmm14
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm14[0],xmm1[1],xmm14[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[1,1,1,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[3,3,3,3]
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, (%rsp) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[3,3,3,3]
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[3,3,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[1,1,1,1]
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1]
+; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm0[0],xmm1[1]
 ; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm1[1,1,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,2,3,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm12[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1]
+; SSE-NEXT:    movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,2,3,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,2,3,3]
-; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm9 = mem[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm1[0],xmm9[1]
-; SSE-NEXT:    movapd %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm7[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm1[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
-; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
-; SSE-NEXT:    movapd %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm1[0],xmm5[1]
+; SSE-NEXT:    movapd %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,3,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm1[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,2,3,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm14[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm9 = xmm1[0],xmm9[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm1[0],xmm15[1]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,3,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm1[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3]
-; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = mem[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm1[0],xmm8[1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,2,3,3]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm11 = xmm1[0],xmm11[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm1[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
 ; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm6 = mem[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm1[0],xmm6[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,2,3,3]
-; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm4 = xmm1[0],xmm4[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[2,2,3,3]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
 ; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm3 = mem[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm0[0],xmm3[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm3 = xmm1[0],xmm3[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[1,1,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,2,3,3]
+; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
 ; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = mem[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
 ; SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm1[0],xmm2[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
-; SSE-NEXT:    movapd %xmm15, %xmm7
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[1,1,1,1]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
-; SSE-NEXT:    movapd %xmm15, %xmm11
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
-; SSE-NEXT:    movapd %xmm14, %xmm12
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
-; SSE-NEXT:    movapd %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
-; SSE-NEXT:    movapd %xmm15, %xmm10
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm15 = xmm0[0],xmm15[1]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm13[0],xmm1[1]
+; SSE-NEXT:    movapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm12 = xmm1[0],xmm12[1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm0[0],xmm13[1]
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
-; SSE-NEXT:    movsd {{.*#+}} xmm14 = xmm0[0],xmm14[1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm13 = xmm14[0],xmm13[1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm10 = xmm8[0],xmm10[1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm8 = xmm7[0],xmm8[1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm9[0],xmm7[1]
+; SSE-NEXT:    movapd %xmm7, %xmm1
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm7 = xmm5[0],xmm7[1]
+; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
+; SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 96(%rsi)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -4053,24 +4038,25 @@ define void @load_i32_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
 ; SSE-NEXT:    movaps %xmm0, (%r8)
 ; SSE-NEXT:    movapd %xmm2, 112(%r9)
 ; SSE-NEXT:    movapd %xmm3, 96(%r9)
-; SSE-NEXT:    movapd %xmm4, 80(%r9)
-; SSE-NEXT:    movapd %xmm6, 64(%r9)
-; SSE-NEXT:    movapd %xmm8, 48(%r9)
-; SSE-NEXT:    movapd %xmm9, 32(%r9)
+; SSE-NEXT:    movapd %xmm6, 80(%r9)
+; SSE-NEXT:    movapd %xmm11, 64(%r9)
+; SSE-NEXT:    movapd %xmm15, 48(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 16(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, (%r9)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movapd %xmm14, 112(%rax)
-; SSE-NEXT:    movapd %xmm13, 96(%rax)
-; SSE-NEXT:    movapd %xmm15, 80(%rax)
-; SSE-NEXT:    movapd %xmm10, 64(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 48(%rax)
-; SSE-NEXT:    movapd %xmm12, 32(%rax)
-; SSE-NEXT:    movapd %xmm11, 16(%rax)
-; SSE-NEXT:    movapd %xmm7, (%rax)
+; SSE-NEXT:    movapd %xmm5, 112(%rax)
+; SSE-NEXT:    movapd %xmm7, 96(%rax)
+; SSE-NEXT:    movapd %xmm1, 80(%rax)
+; SSE-NEXT:    movapd %xmm8, 64(%rax)
+; SSE-NEXT:    movapd %xmm10, 48(%rax)
+; SSE-NEXT:    movapd %xmm13, 32(%rax)
+; SSE-NEXT:    movapd %xmm12, 16(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rax)
 ; SSE-NEXT:    addq $1032, %rsp # imm = 0x408
 ; SSE-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-3.ll
index faecad65c395b..535bb8bb2e0f7 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-3.ll
@@ -642,119 +642,118 @@ define void @load_i8_stride3_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i8_stride3_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2) nounwind {
 ; SSE-LABEL: load_i8_stride3_vf16:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    movdqa (%rdi), %xmm5
-; SSE-NEXT:    movdqa 16(%rdi), %xmm4
+; SSE-NEXT:    movdqa (%rdi), %xmm2
+; SSE-NEXT:    movdqa 16(%rdi), %xmm3
 ; SSE-NEXT:    movdqa 32(%rdi), %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,0,255,255,0,255,255,0,255,255,0,255,255,0,255]
-; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    pandn %xmm4, %xmm1
-; SSE-NEXT:    movdqa %xmm5, %xmm2
-; SSE-NEXT:    pand %xmm3, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    pxor %xmm6, %xmm6
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,0,255,255,0,255,255,0,255,255,0,255,255,0,255]
+; SSE-NEXT:    movdqa %xmm1, %xmm4
+; SSE-NEXT:    pandn %xmm3, %xmm4
 ; SSE-NEXT:    movdqa %xmm2, %xmm7
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm6[8],xmm7[9],xmm6[9],xmm7[10],xmm6[10],xmm7[11],xmm6[11],xmm7[12],xmm6[12],xmm7[13],xmm6[13],xmm7[14],xmm6[14],xmm7[15],xmm6[15]
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,0,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm1, %xmm8
-; SSE-NEXT:    pandn %xmm7, %xmm8
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3],xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
-; SSE-NEXT:    pand %xmm1, %xmm2
-; SSE-NEXT:    por %xmm8, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,5]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm2[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm7[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
-; SSE-NEXT:    packuswb %xmm2, %xmm10
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
-; SSE-NEXT:    pand %xmm2, %xmm10
-; SSE-NEXT:    movdqa %xmm0, %xmm7
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm6[8],xmm7[9],xmm6[9],xmm7[10],xmm6[10],xmm7[11],xmm6[11],xmm7[12],xmm6[12],xmm7[13],xmm6[13],xmm7[14],xmm6[14],xmm7[15],xmm6[15]
-; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm9, %xmm8
-; SSE-NEXT:    pandn %xmm7, %xmm8
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3],xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
-; SSE-NEXT:    movdqa %xmm0, %xmm11
-; SSE-NEXT:    pand %xmm9, %xmm11
-; SSE-NEXT:    por %xmm8, %xmm11
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm11[3,1,2,0]
+; SSE-NEXT:    pand %xmm1, %xmm7
+; SSE-NEXT:    por %xmm4, %xmm7
+; SSE-NEXT:    pxor %xmm4, %xmm4
+; SSE-NEXT:    movdqa %xmm7, %xmm8
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm4[8],xmm8[9],xmm4[9],xmm8[10],xmm4[10],xmm8[11],xmm4[11],xmm8[12],xmm4[12],xmm8[13],xmm4[13],xmm8[14],xmm4[14],xmm8[15],xmm4[15]
+; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,0,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1],xmm7[2],xmm4[2],xmm7[3],xmm4[3],xmm7[4],xmm4[4],xmm7[5],xmm4[5],xmm7[6],xmm4[6],xmm7[7],xmm4[7]
+; SSE-NEXT:    pand %xmm9, %xmm7
+; SSE-NEXT:    movdqa %xmm0, %xmm5
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm4[8],xmm5[9],xmm4[9],xmm5[10],xmm4[10],xmm5[11],xmm4[11],xmm5[12],xmm4[12],xmm5[13],xmm4[13],xmm5[14],xmm4[14],xmm5[15],xmm4[15]
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,0,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm6, %xmm11
+; SSE-NEXT:    pandn %xmm5, %xmm11
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSE-NEXT:    movdqa %xmm0, %xmm12
+; SSE-NEXT:    pand %xmm6, %xmm12
+; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,0,255,255,0,255,255,0,255,255,0,255,255,0,255,255]
+; SSE-NEXT:    movdqa %xmm3, %xmm10
+; SSE-NEXT:    pand %xmm13, %xmm10
+; SSE-NEXT:    pandn %xmm2, %xmm13
+; SSE-NEXT:    por %xmm10, %xmm13
+; SSE-NEXT:    movdqa %xmm13, %xmm14
+; SSE-NEXT:    movdqa %xmm6, %xmm15
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm4[0],xmm13[1],xmm4[1],xmm13[2],xmm4[2],xmm13[3],xmm4[3],xmm13[4],xmm4[4],xmm13[5],xmm4[5],xmm13[6],xmm4[6],xmm13[7],xmm4[7]
+; SSE-NEXT:    pand %xmm6, %xmm13
+; SSE-NEXT:    movdqa %xmm9, %xmm10
+; SSE-NEXT:    pandn %xmm5, %xmm10
+; SSE-NEXT:    pand %xmm6, %xmm5
+; SSE-NEXT:    pandn %xmm0, %xmm6
+; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    pandn %xmm8, %xmm9
+; SSE-NEXT:    por %xmm9, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,7,6,5]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,3,2,1,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm7[0,1,2,3,6,5,4,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm8[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,3,2,3]
+; SSE-NEXT:    packuswb %xmm7, %xmm9
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
+; SSE-NEXT:    por %xmm11, %xmm12
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm12[3,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,3,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    packuswb %xmm8, %xmm11
-; SSE-NEXT:    movdqa %xmm2, %xmm8
+; SSE-NEXT:    movdqa %xmm7, %xmm8
 ; SSE-NEXT:    pandn %xmm11, %xmm8
-; SSE-NEXT:    por %xmm10, %xmm8
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,0,255,255,0,255,255,0,255,255,0,255,255,0,255,255]
-; SSE-NEXT:    movdqa %xmm4, %xmm11
-; SSE-NEXT:    pand %xmm10, %xmm11
-; SSE-NEXT:    pandn %xmm5, %xmm10
-; SSE-NEXT:    por %xmm11, %xmm10
-; SSE-NEXT:    movdqa %xmm10, %xmm11
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm6[8],xmm11[9],xmm6[9],xmm11[10],xmm6[10],xmm11[11],xmm6[11],xmm11[12],xmm6[12],xmm11[13],xmm6[13],xmm11[14],xmm6[14],xmm11[15],xmm6[15]
-; SSE-NEXT:    movdqa %xmm9, %xmm12
-; SSE-NEXT:    pandn %xmm11, %xmm12
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm6[0],xmm10[1],xmm6[1],xmm10[2],xmm6[2],xmm10[3],xmm6[3],xmm10[4],xmm6[4],xmm10[5],xmm6[5],xmm10[6],xmm6[6],xmm10[7],xmm6[7]
-; SSE-NEXT:    pand %xmm9, %xmm10
-; SSE-NEXT:    por %xmm12, %xmm10
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[1,2,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,5,6,7,4]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pand %xmm7, %xmm9
+; SSE-NEXT:    por %xmm9, %xmm8
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm4[8],xmm14[9],xmm4[9],xmm14[10],xmm4[10],xmm14[11],xmm4[11],xmm14[12],xmm4[12],xmm14[13],xmm4[13],xmm14[14],xmm4[14],xmm14[15],xmm4[15]
+; SSE-NEXT:    pandn %xmm14, %xmm15
+; SSE-NEXT:    por %xmm15, %xmm13
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm13[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,6,5,4,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[1,2,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,5,6,7,4]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm14[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[1,2,3,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm11, %xmm10
-; SSE-NEXT:    pand %xmm2, %xmm10
-; SSE-NEXT:    movdqa %xmm7, %xmm11
-; SSE-NEXT:    pand %xmm9, %xmm11
-; SSE-NEXT:    pandn %xmm0, %xmm9
-; SSE-NEXT:    por %xmm11, %xmm9
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,1,2,0]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,5,6,7,4]
-; SSE-NEXT:    packuswb %xmm9, %xmm9
-; SSE-NEXT:    pandn %xmm9, %xmm2
-; SSE-NEXT:    por %xmm10, %xmm2
-; SSE-NEXT:    pand %xmm3, %xmm4
-; SSE-NEXT:    pandn %xmm5, %xmm3
-; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm4
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm6[8],xmm4[9],xmm6[9],xmm4[10],xmm6[10],xmm4[11],xmm6[11],xmm4[12],xmm6[12],xmm4[13],xmm6[13],xmm4[14],xmm6[14],xmm4[15],xmm6[15]
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [0,65535,65535,0,65535,65535,0,65535]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1],xmm3[2],xmm6[2],xmm3[3],xmm6[3],xmm3[4],xmm6[4],xmm3[5],xmm6[5],xmm3[6],xmm6[6],xmm3[7],xmm6[7]
-; SSE-NEXT:    pand %xmm5, %xmm3
-; SSE-NEXT:    pandn %xmm4, %xmm5
-; SSE-NEXT:    por %xmm3, %xmm5
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm5[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,5]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[3,1,2,0]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm4, %xmm3
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,65535,65535,0,0,0]
-; SSE-NEXT:    pand %xmm4, %xmm3
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pandn %xmm7, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,1,3]
+; SSE-NEXT:    packuswb %xmm11, %xmm9
+; SSE-NEXT:    pand %xmm7, %xmm9
+; SSE-NEXT:    por %xmm5, %xmm6
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm6[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,0]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,6,7,4]
+; SSE-NEXT:    packuswb %xmm5, %xmm5
+; SSE-NEXT:    pandn %xmm5, %xmm7
+; SSE-NEXT:    por %xmm9, %xmm7
+; SSE-NEXT:    pand %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm1
+; SSE-NEXT:    por %xmm3, %xmm1
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm4[8],xmm2[9],xmm4[9],xmm2[10],xmm4[10],xmm2[11],xmm4[11],xmm2[12],xmm4[12],xmm2[13],xmm4[13],xmm2[14],xmm4[14],xmm2[15],xmm4[15]
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [0,65535,65535,0,65535,65535,0,65535]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
+; SSE-NEXT:    pand %xmm3, %xmm1
+; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[3,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm2, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,65535,65535,0,0,0]
+; SSE-NEXT:    pand %xmm2, %xmm1
+; SSE-NEXT:    por %xmm10, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pandn %xmm0, %xmm4
-; SSE-NEXT:    por %xmm3, %xmm4
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm8, (%rsi)
-; SSE-NEXT:    movdqa %xmm2, (%rdx)
-; SSE-NEXT:    movdqa %xmm4, (%rcx)
+; SSE-NEXT:    movdqa %xmm7, (%rdx)
+; SSE-NEXT:    movdqa %xmm2, (%rcx)
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i8_stride3_vf16:
@@ -1033,76 +1032,36 @@ define void @load_i8_stride3_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i8_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2) nounwind {
 ; SSE-LABEL: load_i8_stride3_vf32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    movdqa 64(%rdi), %xmm2
+; SSE-NEXT:    subq $56, %rsp
+; SSE-NEXT:    movdqa 64(%rdi), %xmm3
 ; SSE-NEXT:    movdqa (%rdi), %xmm6
-; SSE-NEXT:    movdqa 16(%rdi), %xmm4
-; SSE-NEXT:    movdqa 32(%rdi), %xmm8
-; SSE-NEXT:    movdqa 48(%rdi), %xmm12
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,255,0,255,255,0,255,255,0,255,255,0,255,255,0,255]
-; SSE-NEXT:    movdqa %xmm13, %xmm7
-; SSE-NEXT:    pandn %xmm4, %xmm7
+; SSE-NEXT:    movdqa 16(%rdi), %xmm1
+; SSE-NEXT:    movdqa 32(%rdi), %xmm7
+; SSE-NEXT:    movdqa 48(%rdi), %xmm4
+; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [255,255,0,255,255,0,255,255,0,255,255,0,255,255,0,255]
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    pandn %xmm1, %xmm0
+; SSE-NEXT:    movdqa %xmm1, %xmm13
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,0,255,255,0,255,255,0,255,255,0,255,255,0,255,255]
-; SSE-NEXT:    movdqa %xmm5, %xmm14
-; SSE-NEXT:    pandn %xmm6, %xmm14
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    pandn %xmm6, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm5, %xmm11
+; SSE-NEXT:    pandn %xmm6, %xmm11
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    pandn %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm15, %xmm6
+; SSE-NEXT:    por %xmm0, %xmm6
+; SSE-NEXT:    pxor %xmm1, %xmm1
 ; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    pand %xmm13, %xmm0
-; SSE-NEXT:    por %xmm7, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,0,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm1[0],xmm6[1],xmm1[1],xmm6[2],xmm1[2],xmm6[3],xmm1[3],xmm6[4],xmm1[4],xmm6[5],xmm1[5],xmm6[6],xmm1[6],xmm6[7],xmm1[7]
 ; SSE-NEXT:    pxor %xmm9, %xmm9
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm9[8],xmm1[9],xmm9[9],xmm1[10],xmm9[10],xmm1[11],xmm9[11],xmm1[12],xmm9[12],xmm1[13],xmm9[13],xmm1[14],xmm9[14],xmm1[15],xmm9[15]
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,0,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm6, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
-; SSE-NEXT:    pand %xmm6, %xmm0
-; SSE-NEXT:    por %xmm3, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
-; SSE-NEXT:    pand %xmm7, %xmm0
-; SSE-NEXT:    movdqa %xmm8, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm9[8],xmm3[9],xmm9[9],xmm3[10],xmm9[10],xmm3[11],xmm9[11],xmm3[12],xmm9[12],xmm3[13],xmm9[13],xmm3[14],xmm9[14],xmm3[15],xmm9[15]
-; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm15, %xmm1
-; SSE-NEXT:    pandn %xmm3, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm9[0],xmm8[1],xmm9[1],xmm8[2],xmm9[2],xmm8[3],xmm9[3],xmm8[4],xmm9[4],xmm8[5],xmm9[5],xmm8[6],xmm9[6],xmm8[7],xmm9[7]
-; SSE-NEXT:    movdqa %xmm8, %xmm10
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm15, %xmm10
-; SSE-NEXT:    por %xmm1, %xmm10
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[3,1,2,0]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
-; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm10
-; SSE-NEXT:    pandn %xmm1, %xmm10
-; SSE-NEXT:    por %xmm0, %xmm10
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm1
-; SSE-NEXT:    pand %xmm13, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm9[8],xmm0[9],xmm9[9],xmm0[10],xmm9[10],xmm0[11],xmm9[11],xmm0[12],xmm9[12],xmm0[13],xmm9[13],xmm0[14],xmm9[14],xmm0[15],xmm9[15]
-; SSE-NEXT:    movdqa %xmm6, %xmm10
-; SSE-NEXT:    pandn %xmm0, %xmm10
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1],xmm1[2],xmm9[2],xmm1[3],xmm9[3],xmm1[4],xmm9[4],xmm1[5],xmm9[5],xmm1[6],xmm9[6],xmm1[7],xmm9[7]
-; SSE-NEXT:    pand %xmm6, %xmm1
-; SSE-NEXT:    por %xmm10, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
+; SSE-NEXT:    pand %xmm8, %xmm6
+; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    pandn %xmm0, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[0,2,1,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
@@ -1110,161 +1069,216 @@ define void @load_i8_stride3_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa 80(%rdi), %xmm10
-; SSE-NEXT:    movdqa %xmm10, %xmm12
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
+; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm7, %xmm12
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm9[8],xmm12[9],xmm9[9],xmm12[10],xmm9[10],xmm12[11],xmm9[11],xmm12[12],xmm9[12],xmm12[13],xmm9[13],xmm12[14],xmm9[14],xmm12[15],xmm9[15]
-; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,0,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm10, %xmm0
 ; SSE-NEXT:    pandn %xmm12, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3],xmm10[4],xmm9[4],xmm10[5],xmm9[5],xmm10[6],xmm9[6],xmm10[7],xmm9[7]
-; SSE-NEXT:    movdqa %xmm10, %xmm11
-; SSE-NEXT:    pand %xmm15, %xmm11
-; SSE-NEXT:    por %xmm0, %xmm11
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[3,1,2,0]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1],xmm7[2],xmm9[2],xmm7[3],xmm9[3],xmm7[4],xmm9[4],xmm7[5],xmm9[5],xmm7[6],xmm9[6],xmm7[7],xmm9[7]
+; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    pand %xmm10, %xmm2
+; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[3,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm7, %xmm11
-; SSE-NEXT:    pandn %xmm0, %xmm11
-; SSE-NEXT:    pand %xmm7, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm11
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    pand %xmm5, %xmm0
-; SSE-NEXT:    por %xmm14, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm9[8],xmm1[9],xmm9[9],xmm1[10],xmm9[10],xmm1[11],xmm9[11],xmm1[12],xmm9[12],xmm1[13],xmm9[13],xmm1[14],xmm9[14],xmm1[15],xmm9[15]
-; SSE-NEXT:    movdqa %xmm15, %xmm14
-; SSE-NEXT:    pandn %xmm1, %xmm14
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
-; SSE-NEXT:    pand %xmm15, %xmm0
-; SSE-NEXT:    por %xmm14, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm0[0,1,2,3,5,6,7,4]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm14
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pandn %xmm8, %xmm0
-; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm6, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    movdqa %xmm3, (%rsp) # 16-byte Spill
 ; SSE-NEXT:    pandn %xmm3, %xmm1
-; SSE-NEXT:    pand %xmm15, %xmm3
-; SSE-NEXT:    por %xmm0, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm3[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm7, %xmm3
-; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    pand %xmm7, %xmm14
-; SSE-NEXT:    por %xmm14, %xmm3
-; SSE-NEXT:    pand %xmm5, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm8, %xmm5
-; SSE-NEXT:    por %xmm2, %xmm5
-; SSE-NEXT:    movdqa %xmm5, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm9[8],xmm0[9],xmm9[9],xmm0[10],xmm9[10],xmm0[11],xmm9[11],xmm0[12],xmm9[12],xmm0[13],xmm9[13],xmm0[14],xmm9[14],xmm0[15],xmm9[15]
-; SSE-NEXT:    movdqa %xmm15, %xmm14
-; SSE-NEXT:    pandn %xmm0, %xmm14
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm9[0],xmm5[1],xmm9[1],xmm5[2],xmm9[2],xmm5[3],xmm9[3],xmm5[4],xmm9[4],xmm5[5],xmm9[5],xmm5[6],xmm9[6],xmm5[7],xmm9[7]
-; SSE-NEXT:    pand %xmm15, %xmm5
-; SSE-NEXT:    por %xmm14, %xmm5
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,2,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,6,7,4]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm5
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    pand %xmm15, %xmm0
-; SSE-NEXT:    pandn %xmm10, %xmm15
-; SSE-NEXT:    por %xmm0, %xmm15
-; SSE-NEXT:    pand %xmm7, %xmm5
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm15[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pandn %xmm0, %xmm7
-; SSE-NEXT:    por %xmm5, %xmm7
-; SSE-NEXT:    pand %xmm13, %xmm4
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm9[8],xmm0[9],xmm9[9],xmm0[10],xmm9[10],xmm0[11],xmm9[11],xmm0[12],xmm9[12],xmm0[13],xmm9[13],xmm0[14],xmm9[14],xmm0[15],xmm9[15]
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [0,65535,65535,0,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm5, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3],xmm4[4],xmm9[4],xmm4[5],xmm9[5],xmm4[6],xmm9[6],xmm4[7],xmm9[7]
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm4, %xmm14
+; SSE-NEXT:    pand %xmm15, %xmm14
+; SSE-NEXT:    por %xmm1, %xmm14
+; SSE-NEXT:    movdqa %xmm13, %xmm4
 ; SSE-NEXT:    pand %xmm5, %xmm4
-; SSE-NEXT:    por %xmm2, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[3,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,5]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,0]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm2[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
+; SSE-NEXT:    por %xmm11, %xmm4
+; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    pand %xmm5, %xmm1
+; SSE-NEXT:    pandn %xmm0, %xmm5
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm9[0],xmm14[1],xmm9[1],xmm14[2],xmm9[2],xmm14[3],xmm9[3],xmm14[4],xmm9[4],xmm14[5],xmm9[5],xmm14[6],xmm9[6],xmm14[7],xmm9[7]
+; SSE-NEXT:    movdqa %xmm8, %xmm9
+; SSE-NEXT:    pand %xmm8, %xmm14
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    pandn %xmm7, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm8, %xmm7
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 80(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, %xmm11
+; SSE-NEXT:    pxor %xmm0, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm0[8],xmm11[9],xmm0[9],xmm11[10],xmm0[10],xmm11[11],xmm0[11],xmm11[12],xmm0[12],xmm11[13],xmm0[13],xmm11[14],xmm0[14],xmm11[15],xmm0[15]
+; SSE-NEXT:    movdqa %xmm10, %xmm8
+; SSE-NEXT:    pandn %xmm11, %xmm8
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    pand %xmm10, %xmm1
+; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3],xmm4[4],xmm0[4],xmm4[5],xmm0[5],xmm4[6],xmm0[6],xmm4[7],xmm0[7]
+; SSE-NEXT:    pand %xmm10, %xmm4
+; SSE-NEXT:    movdqa %xmm9, %xmm7
+; SSE-NEXT:    pandn %xmm12, %xmm7
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm10, %xmm12
+; SSE-NEXT:    movdqa %xmm5, %xmm13
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3],xmm5[4],xmm0[4],xmm5[5],xmm0[5],xmm5[6],xmm0[6],xmm5[7],xmm0[7]
+; SSE-NEXT:    pand %xmm10, %xmm5
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    pandn %xmm11, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm10, %xmm11
+; SSE-NEXT:    movdqa %xmm10, %xmm7
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm2, %xmm10
+; SSE-NEXT:    pand %xmm9, %xmm2
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],mem[8],xmm9[9],mem[9],xmm9[10],mem[10],xmm9[11],mem[11],xmm9[12],mem[12],xmm9[13],mem[13],xmm9[14],mem[14],xmm9[15],mem[15]
+; SSE-NEXT:    pandn %xmm9, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm14
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm0[0,1,2,3,6,5,4,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm9[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; SSE-NEXT:    packuswb %xmm0, %xmm14
+; SSE-NEXT:    por %xmm8, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[3,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm4
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,65535,65535,0,0,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm6, %xmm0
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    movdqa %xmm6, %xmm0
 ; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    pand %xmm2, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm0
+; SSE-NEXT:    pand %xmm6, %xmm14
+; SSE-NEXT:    por %xmm14, %xmm0
+; SSE-NEXT:    pxor %xmm8, %xmm8
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm8[8],xmm3[9],xmm8[9],xmm3[10],xmm8[10],xmm3[11],xmm8[11],xmm3[12],xmm8[12],xmm3[13],xmm8[13],xmm3[14],xmm8[14],xmm3[15],xmm8[15]
+; SSE-NEXT:    pandn %xmm3, %xmm7
+; SSE-NEXT:    por %xmm7, %xmm4
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm4[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,2,3,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm3, %xmm1
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm12[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,7,4]
+; SSE-NEXT:    packuswb %xmm3, %xmm4
+; SSE-NEXT:    movdqa %xmm6, %xmm3
+; SSE-NEXT:    pandn %xmm4, %xmm3
+; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm8[8],xmm13[9],xmm8[9],xmm13[10],xmm8[10],xmm13[11],xmm8[11],xmm13[12],xmm8[12],xmm13[13],xmm8[13],xmm13[14],xmm8[14],xmm13[15],xmm8[15]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pand %xmm13, %xmm1
-; SSE-NEXT:    pandn %xmm8, %xmm13
-; SSE-NEXT:    por %xmm1, %xmm13
-; SSE-NEXT:    movdqa %xmm13, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm9[8],xmm1[9],xmm9[9],xmm1[10],xmm9[10],xmm1[11],xmm9[11],xmm1[12],xmm9[12],xmm1[13],xmm9[13],xmm1[14],xmm9[14],xmm1[15],xmm9[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm9[0],xmm13[1],xmm9[1],xmm13[2],xmm9[2],xmm13[3],xmm9[3],xmm13[4],xmm9[4],xmm13[5],xmm9[5],xmm13[6],xmm9[6],xmm13[7],xmm9[7]
-; SSE-NEXT:    pand %xmm5, %xmm13
-; SSE-NEXT:    pandn %xmm1, %xmm5
-; SSE-NEXT:    por %xmm13, %xmm5
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[3,1,2,0]
+; SSE-NEXT:    pandn %xmm13, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm5[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm13[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,2,3,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm4, %xmm1
+; SSE-NEXT:    por %xmm11, %xmm10
+; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm10[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,2,0]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,6,7,4]
+; SSE-NEXT:    packuswb %xmm4, %xmm4
+; SSE-NEXT:    pandn %xmm4, %xmm6
+; SSE-NEXT:    por %xmm1, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pand %xmm15, %xmm7
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm8[8],xmm1[9],xmm8[9],xmm1[10],xmm8[10],xmm1[11],xmm8[11],xmm1[12],xmm8[12],xmm1[13],xmm8[13],xmm1[14],xmm8[14],xmm1[15],xmm8[15]
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [0,65535,65535,0,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm5, %xmm4
+; SSE-NEXT:    pandn %xmm1, %xmm4
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1],xmm7[2],xmm8[2],xmm7[3],xmm8[3],xmm7[4],xmm8[4],xmm7[5],xmm8[5],xmm7[6],xmm8[6],xmm7[7],xmm8[7]
+; SSE-NEXT:    pxor %xmm9, %xmm9
+; SSE-NEXT:    pand %xmm5, %xmm7
+; SSE-NEXT:    por %xmm4, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm7[3,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[3,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,5,4,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm4[0,1,2,3,6,5,4,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm1, %xmm4
-; SSE-NEXT:    pand %xmm6, %xmm10
-; SSE-NEXT:    pandn %xmm12, %xmm6
-; SSE-NEXT:    por %xmm10, %xmm6
-; SSE-NEXT:    pand %xmm2, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,2,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    por %xmm4, %xmm2
-; SSE-NEXT:    movdqa %xmm11, 16(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, (%rsi)
-; SSE-NEXT:    movdqa %xmm7, 16(%rdx)
+; SSE-NEXT:    packuswb %xmm1, %xmm7
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,65535,65535,0,0,0]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,2,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,5,4,7]
+; SSE-NEXT:    packuswb %xmm4, %xmm8
+; SSE-NEXT:    movdqa %xmm1, %xmm4
+; SSE-NEXT:    pandn %xmm8, %xmm4
+; SSE-NEXT:    pand %xmm1, %xmm7
+; SSE-NEXT:    por %xmm7, %xmm4
+; SSE-NEXT:    movdqa (%rsp), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pand %xmm15, %xmm7
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    por %xmm7, %xmm15
+; SSE-NEXT:    movdqa %xmm15, %xmm7
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm9[8],xmm7[9],xmm9[9],xmm7[10],xmm9[10],xmm7[11],xmm9[11],xmm7[12],xmm9[12],xmm7[13],xmm9[13],xmm7[14],xmm9[14],xmm7[15],xmm9[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm9[0],xmm15[1],xmm9[1],xmm15[2],xmm9[2],xmm15[3],xmm9[3],xmm15[4],xmm9[4],xmm15[5],xmm9[5],xmm15[6],xmm9[6],xmm15[7],xmm9[7]
+; SSE-NEXT:    pand %xmm5, %xmm15
+; SSE-NEXT:    pandn %xmm7, %xmm5
+; SSE-NEXT:    por %xmm15, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[3,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,7,6,5]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[3,1,2,0]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,4,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm7, %xmm5
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    pand %xmm1, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,2,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
+; SSE-NEXT:    packuswb %xmm2, %xmm2
+; SSE-NEXT:    pandn %xmm2, %xmm1
+; SSE-NEXT:    por %xmm5, %xmm1
+; SSE-NEXT:    movdqa %xmm0, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rsi)
+; SSE-NEXT:    movdqa %xmm6, 16(%rdx)
 ; SSE-NEXT:    movdqa %xmm3, (%rdx)
-; SSE-NEXT:    movdqa %xmm2, 16(%rcx)
-; SSE-NEXT:    movdqa %xmm0, (%rcx)
+; SSE-NEXT:    movdqa %xmm1, 16(%rcx)
+; SSE-NEXT:    movdqa %xmm4, (%rcx)
+; SSE-NEXT:    addq $56, %rsp
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i8_stride3_vf32:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-4.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-4.ll
index 15f6ef4006fdd..1f8e02f1ae2d6 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-4.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-4.ll
@@ -409,61 +409,61 @@ define void @load_i8_stride4_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i8_stride4_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
 ; SSE-LABEL: load_i8_stride4_vf8:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    movdqa (%rdi), %xmm3
-; SSE-NEXT:    movdqa 16(%rdi), %xmm4
+; SSE-NEXT:    movdqa (%rdi), %xmm1
+; SSE-NEXT:    movdqa 16(%rdi), %xmm2
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,0,255,0,255,0,255,0]
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm4[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm4
-; SSE-NEXT:    pand %xmm3, %xmm0
-; SSE-NEXT:    packuswb %xmm4, %xmm0
+; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    pand %xmm0, %xmm3
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    packuswb %xmm3, %xmm0
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pxor %xmm6, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm6[8],xmm1[9],xmm6[9],xmm1[10],xmm6[10],xmm1[11],xmm6[11],xmm1[12],xmm6[12],xmm1[13],xmm6[13],xmm1[14],xmm6[14],xmm1[15],xmm6[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[0,2,2,3]
+; SSE-NEXT:    pxor %xmm5, %xmm5
+; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm5[8],xmm3[9],xmm5[9],xmm3[10],xmm5[10],xmm3[11],xmm5[11],xmm3[12],xmm5[12],xmm3[13],xmm5[13],xmm3[14],xmm5[14],xmm3[15],xmm5[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3],xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm2[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3],xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm2[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,1,3,4,5,6,7]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
-; SSE-NEXT:    movdqa %xmm3, %xmm4
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm3[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm6[8],xmm3[9],xmm6[9],xmm3[10],xmm6[10],xmm3[11],xmm6[11],xmm3[12],xmm6[12],xmm3[13],xmm6[13],xmm3[14],xmm6[14],xmm3[15],xmm6[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm3[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm6[0],xmm4[1],xmm6[1],xmm4[2],xmm6[2],xmm4[3],xmm6[3],xmm4[4],xmm6[4],xmm4[5],xmm6[5],xmm4[6],xmm6[6],xmm4[7],xmm6[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm4[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm9[0],xmm6[1],xmm9[1]
-; SSE-NEXT:    packuswb %xmm7, %xmm6
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,3,2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    movdqa %xmm1, %xmm4
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm4[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm1[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3],xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm8[0],xmm5[1],xmm8[1]
+; SSE-NEXT:    packuswb %xmm7, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,3,2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,5,6,7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm7, %xmm5
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm7, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm9[0,1,2,3,7,5,6,7]
 ; SSE-NEXT:    pand %xmm7, %xmm8
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm8[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm5, %xmm7
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm7[0,3,2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm6, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm7[0,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT:    packuswb %xmm2, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; SSE-NEXT:    packuswb %xmm2, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
 ; SSE-NEXT:    movq %xmm0, (%rsi)
-; SSE-NEXT:    movq %xmm6, (%rdx)
-; SSE-NEXT:    movq %xmm5, (%rcx)
+; SSE-NEXT:    movq %xmm5, (%rdx)
+; SSE-NEXT:    movq %xmm6, (%rcx)
 ; SSE-NEXT:    movq %xmm1, (%r8)
 ; SSE-NEXT:    retq
 ;
@@ -689,114 +689,110 @@ define void @load_i8_stride4_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i8_stride4_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3) nounwind {
 ; SSE-LABEL: load_i8_stride4_vf16:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    movdqa (%rdi), %xmm1
-; SSE-NEXT:    movdqa 16(%rdi), %xmm3
-; SSE-NEXT:    movdqa 32(%rdi), %xmm9
-; SSE-NEXT:    movdqa 48(%rdi), %xmm10
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,0,255,0,255,0,255,0]
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    movdqa %xmm10, %xmm4
-; SSE-NEXT:    movdqa %xmm10, %xmm7
-; SSE-NEXT:    pand %xmm2, %xmm7
-; SSE-NEXT:    movdqa %xmm9, %xmm5
-; SSE-NEXT:    movdqa %xmm9, %xmm6
-; SSE-NEXT:    movdqa %xmm9, %xmm11
-; SSE-NEXT:    pand %xmm2, %xmm11
-; SSE-NEXT:    packuswb %xmm7, %xmm11
-; SSE-NEXT:    movdqa %xmm3, %xmm7
-; SSE-NEXT:    movdqa %xmm3, %xmm8
-; SSE-NEXT:    movdqa %xmm3, %xmm12
-; SSE-NEXT:    pand %xmm2, %xmm12
-; SSE-NEXT:    pand %xmm1, %xmm2
-; SSE-NEXT:    packuswb %xmm12, %xmm2
-; SSE-NEXT:    packuswb %xmm11, %xmm2
-; SSE-NEXT:    pxor %xmm11, %xmm11
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm11[8],xmm0[9],xmm11[9],xmm0[10],xmm11[10],xmm0[11],xmm11[11],xmm0[12],xmm11[12],xmm0[13],xmm11[13],xmm0[14],xmm11[14],xmm0[15],xmm11[15]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm11[0],xmm4[1],xmm11[1],xmm4[2],xmm11[2],xmm4[3],xmm11[3],xmm4[4],xmm11[4],xmm4[5],xmm11[5],xmm4[6],xmm11[6],xmm4[7],xmm11[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm4[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm13[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm11[8],xmm5[9],xmm11[9],xmm5[10],xmm11[10],xmm5[11],xmm11[11],xmm5[12],xmm11[12],xmm5[13],xmm11[13],xmm5[14],xmm11[14],xmm5[15],xmm11[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm5[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm11[0],xmm6[1],xmm11[1],xmm6[2],xmm11[2],xmm6[3],xmm11[3],xmm6[4],xmm11[4],xmm6[5],xmm11[5],xmm6[6],xmm11[6],xmm6[7],xmm11[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm6[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm14[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm12[0],xmm14[1],xmm12[1]
-; SSE-NEXT:    packuswb %xmm13, %xmm14
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm11[8],xmm7[9],xmm11[9],xmm7[10],xmm11[10],xmm7[11],xmm11[11],xmm7[12],xmm11[12],xmm7[13],xmm11[13],xmm7[14],xmm11[14],xmm7[15],xmm11[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm7[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm11[0],xmm8[1],xmm11[1],xmm8[2],xmm11[2],xmm8[3],xmm11[3],xmm8[4],xmm11[4],xmm8[5],xmm11[5],xmm8[6],xmm11[6],xmm8[7],xmm11[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm8[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm13[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm12[0],xmm15[1],xmm12[1]
-; SSE-NEXT:    movdqa %xmm1, %xmm12
-; SSE-NEXT:    movdqa %xmm1, %xmm13
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm11[8],xmm13[9],xmm11[9],xmm13[10],xmm11[10],xmm13[11],xmm11[11],xmm13[12],xmm11[12],xmm13[13],xmm11[13],xmm13[14],xmm11[14],xmm13[15],xmm11[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm11[0],xmm12[1],xmm11[1],xmm12[2],xmm11[2],xmm12[3],xmm11[3],xmm12[4],xmm11[4],xmm12[5],xmm11[5],xmm12[6],xmm11[6],xmm12[7],xmm11[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm13[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm11[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm12[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
-; SSE-NEXT:    packuswb %xmm15, %xmm11
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm10[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,3],xmm14[0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm10, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm10, %xmm9
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm9
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa (%rdi), %xmm11
+; SSE-NEXT:    movdqa 16(%rdi), %xmm2
+; SSE-NEXT:    movdqa 32(%rdi), %xmm3
+; SSE-NEXT:    movdqa 48(%rdi), %xmm4
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,0,255,0,255,0,255,0]
+; SSE-NEXT:    movdqa %xmm4, %xmm5
+; SSE-NEXT:    pand %xmm0, %xmm5
+; SSE-NEXT:    movdqa %xmm3, %xmm6
+; SSE-NEXT:    pand %xmm0, %xmm6
+; SSE-NEXT:    packuswb %xmm5, %xmm6
+; SSE-NEXT:    movdqa %xmm2, %xmm5
+; SSE-NEXT:    pand %xmm0, %xmm5
+; SSE-NEXT:    pand %xmm11, %xmm0
+; SSE-NEXT:    packuswb %xmm5, %xmm0
+; SSE-NEXT:    packuswb %xmm6, %xmm0
+; SSE-NEXT:    pxor %xmm8, %xmm8
+; SSE-NEXT:    movdqa %xmm4, %xmm5
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm8[8],xmm5[9],xmm8[9],xmm5[10],xmm8[10],xmm5[11],xmm8[11],xmm5[12],xmm8[12],xmm5[13],xmm8[13],xmm5[14],xmm8[14],xmm5[15],xmm8[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm4[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1],xmm4[2],xmm8[2],xmm4[3],xmm8[3],xmm4[4],xmm8[4],xmm4[5],xmm8[5],xmm4[6],xmm8[6],xmm4[7],xmm8[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm4[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
+; SSE-NEXT:    movdqa %xmm3, %xmm6
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm8[8],xmm6[9],xmm8[9],xmm6[10],xmm8[10],xmm6[11],xmm8[11],xmm6[12],xmm8[12],xmm6[13],xmm8[13],xmm6[14],xmm8[14],xmm6[15],xmm8[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm6[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3],xmm3[4],xmm8[4],xmm3[5],xmm8[5],xmm3[6],xmm8[6],xmm3[7],xmm8[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm3[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm12[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm9[0],xmm13[1],xmm9[1]
+; SSE-NEXT:    packuswb %xmm7, %xmm13
+; SSE-NEXT:    movdqa %xmm2, %xmm7
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm8[8],xmm7[9],xmm8[9],xmm7[10],xmm8[10],xmm7[11],xmm8[11],xmm7[12],xmm8[12],xmm7[13],xmm8[13],xmm7[14],xmm8[14],xmm7[15],xmm8[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm7[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm2[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3],xmm2[4],xmm8[4],xmm2[5],xmm8[5],xmm2[6],xmm8[6],xmm2[7],xmm8[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm2[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm14[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm9[0],xmm14[1],xmm9[1]
+; SSE-NEXT:    movdqa %xmm11, %xmm9
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm8[8],xmm9[9],xmm8[9],xmm9[10],xmm8[10],xmm9[11],xmm8[11],xmm9[12],xmm8[12],xmm9[13],xmm8[13],xmm9[14],xmm8[14],xmm9[15],xmm8[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm11[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm8[0],xmm11[1],xmm8[1],xmm11[2],xmm8[2],xmm11[3],xmm8[3],xmm11[4],xmm8[4],xmm11[5],xmm8[5],xmm11[6],xmm8[6],xmm11[7],xmm8[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm9[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm8[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm11[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm1[0],xmm8[1],xmm1[1]
+; SSE-NEXT:    packuswb %xmm14, %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,3],xmm13[0,3]
+; SSE-NEXT:    pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm13, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm13, %xmm10
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm10[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm1, %xmm14
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm12[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm13, %xmm1
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm15[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm13, %xmm10
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm1, %xmm10
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,3],xmm14[0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm10, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm10, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm9[0,3]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm6[3,1,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    packuswb %xmm4, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm11[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
-; SSE-NEXT:    packuswb %xmm3, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm8[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm12[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
-; SSE-NEXT:    packuswb %xmm3, %xmm5
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm4[0,3]
-; SSE-NEXT:    movdqa %xmm2, (%rsi)
-; SSE-NEXT:    movaps %xmm11, (%rdx)
-; SSE-NEXT:    movaps %xmm1, (%rcx)
-; SSE-NEXT:    movaps %xmm5, (%r8)
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT:    packuswb %xmm2, %xmm4
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,3],xmm3[0,3]
+; SSE-NEXT:    movdqa %xmm0, (%rsi)
+; SSE-NEXT:    movaps %xmm8, (%rdx)
+; SSE-NEXT:    movaps %xmm10, (%rcx)
+; SSE-NEXT:    movaps %xmm4, (%r8)
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i8_stride4_vf16:
@@ -1110,248 +1106,242 @@ define void @load_i8_stride4_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-LABEL: load_i8_stride4_vf32:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    subq $136, %rsp
-; SSE-NEXT:    movdqa 64(%rdi), %xmm4
-; SSE-NEXT:    movdqa 80(%rdi), %xmm9
-; SSE-NEXT:    movdqa 96(%rdi), %xmm11
-; SSE-NEXT:    movdqa 112(%rdi), %xmm12
-; SSE-NEXT:    movdqa (%rdi), %xmm13
-; SSE-NEXT:    movdqa 16(%rdi), %xmm14
-; SSE-NEXT:    movdqa 32(%rdi), %xmm2
-; SSE-NEXT:    movdqa 48(%rdi), %xmm10
-; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [255,0,255,0,255,0,255,0]
-; SSE-NEXT:    movdqa %xmm10, %xmm3
-; SSE-NEXT:    movdqa %xmm10, %xmm5
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    movdqa %xmm2, %xmm6
-; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm2, %xmm15
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm8, %xmm1
+; SSE-NEXT:    movdqa 64(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 80(%rdi), %xmm11
+; SSE-NEXT:    movdqa 96(%rdi), %xmm13
+; SSE-NEXT:    movdqa 112(%rdi), %xmm15
+; SSE-NEXT:    movdqa (%rdi), %xmm6
+; SSE-NEXT:    movdqa 16(%rdi), %xmm7
+; SSE-NEXT:    movdqa 32(%rdi), %xmm8
+; SSE-NEXT:    movdqa 48(%rdi), %xmm9
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,0,255,0,255,0,255,0]
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    pand %xmm5, %xmm1
 ; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    movdqa %xmm13, %xmm7
-; SSE-NEXT:    pand %xmm8, %xmm7
-; SSE-NEXT:    packuswb %xmm0, %xmm7
-; SSE-NEXT:    packuswb %xmm1, %xmm7
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    pand %xmm8, %xmm1
+; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    movdqa %xmm6, %xmm2
+; SSE-NEXT:    pand %xmm5, %xmm2
+; SSE-NEXT:    packuswb %xmm0, %xmm2
+; SSE-NEXT:    packuswb %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    pand %xmm5, %xmm1
 ; SSE-NEXT:    packuswb %xmm0, %xmm1
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    pand %xmm3, %xmm5
+; SSE-NEXT:    packuswb %xmm0, %xmm5
+; SSE-NEXT:    packuswb %xmm1, %xmm5
+; SSE-NEXT:    pxor %xmm1, %xmm1
 ; SSE-NEXT:    movdqa %xmm9, %xmm0
-; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    pand %xmm4, %xmm8
-; SSE-NEXT:    packuswb %xmm0, %xmm8
-; SSE-NEXT:    packuswb %xmm1, %xmm8
-; SSE-NEXT:    pxor %xmm2, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[0,2,2,3]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3],xmm5[4],xmm2[4],xmm5[5],xmm2[5],xmm5[6],xmm2[6],xmm5[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm5[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm9[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1],xmm9[2],xmm1[2],xmm9[3],xmm1[3],xmm9[4],xmm1[4],xmm9[5],xmm1[5],xmm9[6],xmm1[6],xmm9[7],xmm1[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm2[8],xmm6[9],xmm2[9],xmm6[10],xmm2[10],xmm6[11],xmm2[11],xmm6[12],xmm2[12],xmm6[13],xmm2[13],xmm6[14],xmm2[14],xmm6[15],xmm2[15]
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[0,2,2,3]
+; SSE-NEXT:    movdqa %xmm8, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm15, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
-; SSE-NEXT:    packuswb %xmm3, %xmm5
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm8[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm1[0],xmm8[1],xmm1[1],xmm8[2],xmm1[2],xmm8[3],xmm1[3],xmm8[4],xmm1[4],xmm8[5],xmm1[5],xmm8[6],xmm1[6],xmm8[7],xmm1[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm8[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    packuswb %xmm3, %xmm4
+; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm14, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm13, %xmm14
-; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
-; SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm7[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1],xmm7[2],xmm1[2],xmm7[3],xmm1[3],xmm7[4],xmm1[4],xmm7[5],xmm1[5],xmm7[6],xmm1[6],xmm7[7],xmm1[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm3[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
+; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm2[0],xmm14[1],xmm2[1],xmm14[2],xmm2[2],xmm14[3],xmm2[3],xmm14[4],xmm2[4],xmm14[5],xmm2[5],xmm14[6],xmm2[6],xmm14[7],xmm2[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm14[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm6[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    packuswb %xmm3, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm5[0,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm15
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm6[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm1[0],xmm6[1],xmm1[1],xmm6[2],xmm1[2],xmm6[3],xmm1[3],xmm6[4],xmm1[4],xmm6[5],xmm1[5],xmm6[6],xmm1[6],xmm6[7],xmm1[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm6[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
+; SSE-NEXT:    packuswb %xmm12, %xmm10
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,3],xmm4[0,3]
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm2[0],xmm15[1],xmm2[1],xmm15[2],xmm2[2],xmm15[3],xmm2[3],xmm15[4],xmm2[4],xmm15[5],xmm2[5],xmm15[6],xmm2[6],xmm15[7],xmm2[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm15[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm15[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1],xmm15[2],xmm1[2],xmm15[3],xmm1[3],xmm15[4],xmm1[4],xmm15[5],xmm1[5],xmm15[6],xmm1[6],xmm15[7],xmm1[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm15[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm4[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
+; SSE-NEXT:    movdqa %xmm13, %xmm14
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm1[8],xmm14[9],xmm1[9],xmm14[10],xmm1[10],xmm14[11],xmm1[11],xmm14[12],xmm1[12],xmm14[13],xmm1[13],xmm14[14],xmm1[14],xmm14[15],xmm1[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm0[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm13[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1],xmm13[2],xmm1[2],xmm13[3],xmm1[3],xmm13[4],xmm1[4],xmm13[5],xmm1[5],xmm13[6],xmm1[6],xmm13[7],xmm1[7]
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm13[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm13[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    packuswb %xmm12, %xmm0
 ; SSE-NEXT:    movdqa %xmm11, %xmm12
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm13
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm2[8],xmm13[9],xmm2[9],xmm13[10],xmm2[10],xmm13[11],xmm2[11],xmm13[12],xmm2[12],xmm13[13],xmm2[13],xmm13[14],xmm2[14],xmm13[15],xmm2[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm2[0],xmm12[1],xmm2[1],xmm12[2],xmm2[2],xmm12[3],xmm2[3],xmm12[4],xmm2[4],xmm12[5],xmm2[5],xmm12[6],xmm2[6],xmm12[7],xmm2[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm12[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm5[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
-; SSE-NEXT:    packuswb %xmm3, %xmm0
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm11
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm2[8],xmm11[9],xmm2[9],xmm11[10],xmm2[10],xmm11[11],xmm2[11],xmm11[12],xmm2[12],xmm11[13],xmm2[13],xmm11[14],xmm2[14],xmm11[15],xmm2[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm2[0],xmm9[1],xmm2[1],xmm9[2],xmm2[2],xmm9[3],xmm2[3],xmm9[4],xmm2[4],xmm9[5],xmm2[5],xmm9[6],xmm2[6],xmm9[7],xmm2[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm9[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm5[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
-; SSE-NEXT:    movdqa %xmm4, %xmm5
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm1[8],xmm12[9],xmm1[9],xmm12[10],xmm1[10],xmm12[11],xmm1[11],xmm12[12],xmm1[12],xmm12[13],xmm1[13],xmm12[14],xmm1[14],xmm12[15],xmm1[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm12[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm11[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm4, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm2[8],xmm6[9],xmm2[9],xmm6[10],xmm2[10],xmm6[11],xmm2[11],xmm6[12],xmm2[12],xmm6[13],xmm2[13],xmm6[14],xmm2[14],xmm6[15],xmm2[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3],xmm5[4],xmm2[4],xmm5[5],xmm2[5],xmm5[6],xmm2[6],xmm5[7],xmm2[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm6[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm5[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm3[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
-; SSE-NEXT:    packuswb %xmm1, %xmm4
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm1[0],xmm11[1],xmm1[1],xmm11[2],xmm1[2],xmm11[3],xmm1[3],xmm11[4],xmm1[4],xmm11[5],xmm1[5],xmm11[6],xmm1[6],xmm11[7],xmm1[7]
+; SSE-NEXT:    movdqa %xmm11, (%rsp) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm11[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm13[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm2[0],xmm13[1],xmm2[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm2, %xmm11
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm1[8],xmm11[9],xmm1[9],xmm11[10],xmm1[10],xmm11[11],xmm1[11],xmm11[12],xmm1[12],xmm11[13],xmm1[13],xmm11[14],xmm1[14],xmm11[15],xmm1[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm2[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm2[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT:    packuswb %xmm13, %xmm4
 ; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,3],xmm0[0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm10[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm3, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm3, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm1
-; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm10 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm3, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm3, %xmm10
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm10
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,3],xmm1[0,3]
-; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm3, %xmm1
+; SSE-NEXT:    pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm2, %xmm1
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm3, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm1, %xmm2
-; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm3, %xmm1
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm3, %xmm0
+; SSE-NEXT:    pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm13 = mem[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm2, %xmm13
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm13[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm13[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm1, %xmm13
+; SSE-NEXT:    pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm2, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm2, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm1, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,3],xmm13[0,3]
+; SSE-NEXT:    pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm2, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm13 = mem[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm2, %xmm13
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm13[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm13[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm1, %xmm13
+; SSE-NEXT:    pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm2, %xmm1
+; SSE-NEXT:    pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm2, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,5,4]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,2,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,3],xmm2[0,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,3],xmm13[0,3]
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[3,1,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm9[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT:    packuswb %xmm2, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm1[0],xmm8[1],xmm1[1]
+; SSE-NEXT:    packuswb %xmm2, %xmm8
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[3,1,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    pshufd $231, (%rsp), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm14[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm14[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
-; SSE-NEXT:    packuswb %xmm2, %xmm14
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,3],xmm3[0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm1[0],xmm6[1],xmm1[1]
+; SSE-NEXT:    packuswb %xmm2, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,3],xmm8[0,3]
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm15[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[3,1,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm12[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT:    packuswb %xmm2, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[3,1,2,3]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1]
+; SSE-NEXT:    packuswb %xmm2, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm9[3,1,2,3]
+; SSE-NEXT:    pshufd $231, (%rsp), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[3,1,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
-; SSE-NEXT:    packuswb %xmm2, %xmm5
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm3[0,3]
-; SSE-NEXT:    movdqa %xmm8, 16(%rsi)
-; SSE-NEXT:    movdqa %xmm7, (%rsi)
-; SSE-NEXT:    movaps %xmm4, 16(%rdx)
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm1[0],xmm8[1],xmm1[1]
+; SSE-NEXT:    packuswb %xmm2, %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,3],xmm7[0,3]
+; SSE-NEXT:    movdqa %xmm5, 16(%rsi)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, (%rdx)
+; SSE-NEXT:    movaps %xmm1, (%rsi)
+; SSE-NEXT:    movaps %xmm4, 16(%rdx)
+; SSE-NEXT:    movaps %xmm10, (%rdx)
 ; SSE-NEXT:    movaps %xmm0, 16(%rcx)
-; SSE-NEXT:    movaps %xmm10, (%rcx)
-; SSE-NEXT:    movaps %xmm5, 16(%r8)
-; SSE-NEXT:    movaps %xmm14, (%r8)
+; SSE-NEXT:    movaps %xmm3, (%rcx)
+; SSE-NEXT:    movaps %xmm8, 16(%r8)
+; SSE-NEXT:    movaps %xmm6, (%r8)
 ; SSE-NEXT:    addq $136, %rsp
 ; SSE-NEXT:    retq
 ;
@@ -1960,515 +1950,512 @@ define void @load_i8_stride4_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    subq $600, %rsp # imm = 0x258
 ; SSE-NEXT:    movdqa 16(%rdi), %xmm8
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 32(%rdi), %xmm15
-; SSE-NEXT:    movdqa 48(%rdi), %xmm14
-; SSE-NEXT:    movdqa 128(%rdi), %xmm4
-; SSE-NEXT:    movdqa 144(%rdi), %xmm7
-; SSE-NEXT:    movdqa 160(%rdi), %xmm10
-; SSE-NEXT:    movdqa 176(%rdi), %xmm12
-; SSE-NEXT:    movdqa 64(%rdi), %xmm6
-; SSE-NEXT:    movdqa 80(%rdi), %xmm13
-; SSE-NEXT:    movdqa 96(%rdi), %xmm2
-; SSE-NEXT:    movdqa 112(%rdi), %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [255,0,255,0,255,0,255,0]
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm1, %xmm3
-; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    movdqa 32(%rdi), %xmm9
+; SSE-NEXT:    movdqa 48(%rdi), %xmm10
+; SSE-NEXT:    movdqa 128(%rdi), %xmm11
+; SSE-NEXT:    movdqa 144(%rdi), %xmm12
+; SSE-NEXT:    movdqa 160(%rdi), %xmm13
+; SSE-NEXT:    movdqa 176(%rdi), %xmm14
+; SSE-NEXT:    movdqa 64(%rdi), %xmm5
+; SSE-NEXT:    movdqa 80(%rdi), %xmm15
+; SSE-NEXT:    movdqa 96(%rdi), %xmm0
+; SSE-NEXT:    movdqa 112(%rdi), %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,0,255,0,255,0,255,0]
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm2, %xmm5
-; SSE-NEXT:    pand %xmm9, %xmm1
-; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    pand %xmm9, %xmm0
-; SSE-NEXT:    movdqa %xmm6, %xmm2
-; SSE-NEXT:    pand %xmm9, %xmm2
-; SSE-NEXT:    packuswb %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm0, %xmm4
+; SSE-NEXT:    pand %xmm6, %xmm2
 ; SSE-NEXT:    packuswb %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    packuswb %xmm1, %xmm0
+; SSE-NEXT:    packuswb %xmm2, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm14, %xmm1
+; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm13, %xmm2
+; SSE-NEXT:    pand %xmm6, %xmm2
+; SSE-NEXT:    packuswb %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    packuswb %xmm1, %xmm0
+; SSE-NEXT:    packuswb %xmm2, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm10, %xmm1
-; SSE-NEXT:    pand %xmm9, %xmm1
-; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    pand %xmm9, %xmm0
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    pand %xmm9, %xmm2
-; SSE-NEXT:    packuswb %xmm0, %xmm2
+; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    pand %xmm6, %xmm2
 ; SSE-NEXT:    packuswb %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    movdqa (%rdi), %xmm7
+; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    packuswb %xmm1, %xmm0
+; SSE-NEXT:    packuswb %xmm2, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 240(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    movdqa 224(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    pand %xmm9, %xmm0
-; SSE-NEXT:    movdqa %xmm15, %xmm1
-; SSE-NEXT:    pand %xmm9, %xmm1
-; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    pand %xmm9, %xmm0
-; SSE-NEXT:    movdqa (%rdi), %xmm11
-; SSE-NEXT:    movdqa %xmm11, %xmm2
-; SSE-NEXT:    pand %xmm9, %xmm2
-; SSE-NEXT:    packuswb %xmm0, %xmm2
+; SSE-NEXT:    pand %xmm6, %xmm2
 ; SSE-NEXT:    packuswb %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 240(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm9, %xmm0
-; SSE-NEXT:    movdqa 224(%rdi), %xmm1
+; SSE-NEXT:    movdqa 208(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm9, %xmm1
-; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa 208(%rdi), %xmm0
+; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    movdqa 192(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm9, %xmm0
-; SSE-NEXT:    movdqa 192(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm2, %xmm9
-; SSE-NEXT:    packuswb %xmm0, %xmm9
-; SSE-NEXT:    packuswb %xmm1, %xmm9
-; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    pand %xmm0, %xmm6
+; SSE-NEXT:    packuswb %xmm1, %xmm6
+; SSE-NEXT:    packuswb %xmm2, %xmm6
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    pxor %xmm8, %xmm8
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
+; SSE-NEXT:    pxor %xmm1, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,2,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm3, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSE-NEXT:    packuswb %xmm3, %xmm4
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
 ; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm5, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
-; SSE-NEXT:    packuswb %xmm3, %xmm5
-; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm5[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    packuswb %xmm3, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,3],xmm4[0,3]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm14, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm14, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
 ; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm6, %xmm1
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm6[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    packuswb %xmm3, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm5[0,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm1
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSE-NEXT:    packuswb %xmm3, %xmm4
+; SSE-NEXT:    movdqa %xmm12, (%rsp) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm10, %xmm1
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm5[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
-; SSE-NEXT:    packuswb %xmm3, %xmm6
-; SSE-NEXT:    movdqa %xmm7, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,2,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm5[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
-; SSE-NEXT:    packuswb %xmm3, %xmm10
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,3],xmm6[0,3]
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm8[0],xmm14[1],xmm8[1],xmm14[2],xmm8[2],xmm14[3],xmm8[3],xmm14[4],xmm8[4],xmm14[5],xmm8[5],xmm14[6],xmm8[6],xmm14[7],xmm8[7]
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm14[0,2,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm5[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm2[0],xmm12[1],xmm2[1]
+; SSE-NEXT:    packuswb %xmm3, %xmm12
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,3],xmm4[0,3]
+; SSE-NEXT:    movdqa %xmm10, %xmm14
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm1[8],xmm14[9],xmm1[9],xmm14[10],xmm1[10],xmm14[11],xmm1[11],xmm14[12],xmm1[12],xmm14[13],xmm1[13],xmm14[14],xmm1[14],xmm14[15],xmm1[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm10[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm1[0],xmm10[1],xmm1[1],xmm10[2],xmm1[2],xmm10[3],xmm1[3],xmm10[4],xmm1[4],xmm10[5],xmm1[5],xmm10[6],xmm1[6],xmm10[7],xmm1[7]
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm10[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; SSE-NEXT:    movdqa %xmm9, %xmm13
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm1[8],xmm13[9],xmm1[9],xmm13[10],xmm1[10],xmm13[11],xmm1[11],xmm13[12],xmm1[12],xmm13[13],xmm1[13],xmm13[14],xmm1[14],xmm13[15],xmm1[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm13[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm9[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm15, %xmm12
-; SSE-NEXT:    movdqa %xmm15, %xmm14
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm8[0],xmm12[1],xmm8[1],xmm12[2],xmm8[2],xmm12[3],xmm8[3],xmm12[4],xmm8[4],xmm12[5],xmm8[5],xmm12[6],xmm8[6],xmm12[7],xmm8[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm12[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm6[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1],xmm9[2],xmm1[2],xmm9[3],xmm1[3],xmm9[4],xmm1[4],xmm9[5],xmm1[5],xmm9[6],xmm1[6],xmm9[7],xmm1[7]
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm9[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm5[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
 ; SSE-NEXT:    packuswb %xmm3, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm15, %xmm13
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm8[8],xmm13[9],xmm8[9],xmm13[10],xmm8[10],xmm13[11],xmm8[11],xmm13[12],xmm8[12],xmm13[13],xmm8[13],xmm13[14],xmm8[14],xmm13[15],xmm8[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[0,2,2,3]
+; SSE-NEXT:    movdqa %xmm8, %xmm10
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm1[8],xmm10[9],xmm1[9],xmm10[10],xmm1[10],xmm10[11],xmm1[11],xmm10[12],xmm1[12],xmm10[13],xmm1[13],xmm10[14],xmm1[14],xmm10[15],xmm1[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm10[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm8[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm1[0],xmm8[1],xmm1[1],xmm8[2],xmm1[2],xmm8[3],xmm1[3],xmm8[4],xmm1[4],xmm8[5],xmm1[5],xmm8[6],xmm1[6],xmm8[7],xmm1[7]
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm8[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm8[0],xmm15[1],xmm8[1],xmm15[2],xmm8[2],xmm15[3],xmm8[3],xmm15[4],xmm8[4],xmm15[5],xmm8[5],xmm15[6],xmm8[6],xmm15[7],xmm8[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm15[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm6[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
-; SSE-NEXT:    movdqa %xmm11, %xmm7
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm8[8],xmm7[9],xmm8[9],xmm7[10],xmm8[10],xmm7[11],xmm8[11],xmm7[12],xmm8[12],xmm7[13],xmm8[13],xmm7[14],xmm8[14],xmm7[15],xmm8[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm11, %xmm6
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm8[0],xmm6[1],xmm8[1],xmm6[2],xmm8[2],xmm6[3],xmm8[3],xmm6[4],xmm8[4],xmm6[5],xmm8[5],xmm6[6],xmm8[6],xmm6[7],xmm8[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm6[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; SSE-NEXT:    movdqa %xmm7, %xmm8
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm1[8],xmm8[9],xmm1[9],xmm8[10],xmm1[10],xmm8[11],xmm1[11],xmm8[12],xmm1[12],xmm8[13],xmm1[13],xmm8[14],xmm1[14],xmm8[15],xmm1[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm7[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1],xmm7[2],xmm1[2],xmm7[3],xmm1[3],xmm7[4],xmm1[4],xmm7[5],xmm1[5],xmm7[6],xmm1[6],xmm7[7],xmm1[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm7[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
-; SSE-NEXT:    packuswb %xmm1, %xmm5
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; SSE-NEXT:    packuswb %xmm3, %xmm5
 ; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm0[0,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm1, %xmm3
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3],xmm3[4],xmm8[4],xmm3[5],xmm8[5],xmm3[6],xmm8[6],xmm3[7],xmm8[7]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm2, %xmm4
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1],xmm4[2],xmm8[2],xmm4[3],xmm8[3],xmm4[4],xmm8[4],xmm4[5],xmm8[5],xmm4[6],xmm8[6],xmm4[7],xmm8[7]
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT:    packuswb %xmm1, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3],xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm0[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3],xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm0[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm3, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm0, %xmm4
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm8[8],xmm4[9],xmm8[9],xmm4[10],xmm8[10],xmm4[11],xmm8[11],xmm4[12],xmm8[12],xmm4[13],xmm8[13],xmm4[14],xmm8[14],xmm4[15],xmm8[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm3[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
+; SSE-NEXT:    packuswb %xmm2, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm3, %xmm4
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3],xmm4[4],xmm1[4],xmm4[5],xmm1[5],xmm4[6],xmm1[6],xmm4[7],xmm1[7]
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3],xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm1[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm3[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm1[8],xmm3[9],xmm1[9],xmm3[10],xmm1[10],xmm3[11],xmm1[11],xmm3[12],xmm1[12],xmm3[13],xmm1[13],xmm3[14],xmm1[14],xmm3[15],xmm1[15]
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm2, %xmm9
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1],xmm9[2],xmm1[2],xmm9[3],xmm1[3],xmm9[4],xmm1[4],xmm9[5],xmm1[5],xmm9[6],xmm1[6],xmm9[7],xmm1[7]
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
-; SSE-NEXT:    packuswb %xmm2, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm3[0,3]
-; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm14[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm0[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm9[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm2[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    packuswb %xmm4, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,3],xmm0[0,3]
+; SSE-NEXT:    pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufhw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm1, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm2[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm3, %xmm4
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm0, %xmm2
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm15[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm11[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm0[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm4, %xmm9
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,3],xmm2[0,3]
 ; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm11[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm2[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm8, %xmm2
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm4[0,3]
-; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm3, %xmm4
-; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm8
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm8[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm3, %xmm11
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,3],xmm4[0,3]
-; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm4 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm1, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm4
+; SSE-NEXT:    pand %xmm1, %xmm4
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm3, %xmm4
-; SSE-NEXT:    pshuflw $231, (%rsp), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm8
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm8[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm3, %xmm14
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,3],xmm4[0,3]
-; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm4[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm2, %xmm11
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm4 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm1, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm4
+; SSE-NEXT:    pand %xmm1, %xmm4
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm3, %xmm4
-; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = mem[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm8
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm0[0,1,2,3,7,6,5,4]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm3, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,3],xmm4[0,3]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm12[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
-; SSE-NEXT:    packuswb %xmm4, %xmm8
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm15[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
-; SSE-NEXT:    packuswb %xmm4, %xmm6
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,3],xmm8[0,3]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm7 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm7[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
-; SSE-NEXT:    packuswb %xmm4, %xmm8
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm7 = mem[3,1,2,3]
+; SSE-NEXT:    packuswb %xmm2, %xmm4
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,3],xmm11[0,3]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm1, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm1, %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm2, %xmm11
+; SSE-NEXT:    pshuflw $231, (%rsp), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm1, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm2[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm15[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm1, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm0, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm11[0,3]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm1, %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm0, %xmm11
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = mem[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pand %xmm1, %xmm15
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,5,4]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,2,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm0, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm11[0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm13 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm13[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
+; SSE-NEXT:    packuswb %xmm11, %xmm13
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm3[0],xmm7[1],xmm3[1]
-; SSE-NEXT:    packuswb %xmm4, %xmm7
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,3],xmm8[0,3]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
+; SSE-NEXT:    packuswb %xmm10, %xmm7
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,3],xmm13[0,3]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm8 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm8[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm3[0],xmm12[1],xmm3[1]
-; SSE-NEXT:    packuswb %xmm4, %xmm12
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
+; SSE-NEXT:    packuswb %xmm8, %xmm10
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm8[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm8 = mem[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
-; SSE-NEXT:    packuswb %xmm4, %xmm8
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,3],xmm12[0,3]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm12 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm3[0],xmm12[1],xmm3[1]
-; SSE-NEXT:    packuswb %xmm4, %xmm12
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    packuswb %xmm11, %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,3],xmm10[0,3]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
+; SSE-NEXT:    packuswb %xmm10, %xmm11
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm10[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm10 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
+; SSE-NEXT:    packuswb %xmm13, %xmm10
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,3],xmm11[0,3]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm13 = mem[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm13[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm3[0],xmm13[1],xmm3[1]
-; SSE-NEXT:    packuswb %xmm4, %xmm13
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,3],xmm12[0,3]
-; SSE-NEXT:    movdqa %xmm9, 48(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, (%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 32(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 16(%rsi)
-; SSE-NEXT:    movaps %xmm1, 48(%rdx)
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
+; SSE-NEXT:    packuswb %xmm11, %xmm13
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = mem[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm14[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
+; SSE-NEXT:    packuswb %xmm11, %xmm14
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,3],xmm13[0,3]
+; SSE-NEXT:    movdqa %xmm6, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rsi)
+; SSE-NEXT:    movaps %xmm3, 48(%rdx)
 ; SSE-NEXT:    movaps %xmm5, (%rdx)
-; SSE-NEXT:    movaps %xmm10, 32(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 16(%rdx)
-; SSE-NEXT:    movaps %xmm0, 48(%rcx)
-; SSE-NEXT:    movaps %xmm14, 32(%rcx)
-; SSE-NEXT:    movaps %xmm11, 16(%rcx)
-; SSE-NEXT:    movaps %xmm2, (%rcx)
-; SSE-NEXT:    movaps %xmm13, 48(%r8)
-; SSE-NEXT:    movaps %xmm8, 32(%r8)
-; SSE-NEXT:    movaps %xmm7, 16(%r8)
-; SSE-NEXT:    movaps %xmm6, (%r8)
+; SSE-NEXT:    movaps %xmm12, 32(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rdx)
+; SSE-NEXT:    movaps %xmm1, 48(%rcx)
+; SSE-NEXT:    movaps %xmm2, 32(%rcx)
+; SSE-NEXT:    movaps %xmm4, 16(%rcx)
+; SSE-NEXT:    movaps %xmm9, (%rcx)
+; SSE-NEXT:    movaps %xmm14, 48(%r8)
+; SSE-NEXT:    movaps %xmm10, 32(%r8)
+; SSE-NEXT:    movaps %xmm8, 16(%r8)
+; SSE-NEXT:    movaps %xmm7, (%r8)
 ; SSE-NEXT:    addq $600, %rsp # imm = 0x258
 ; SSE-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-5.ll
index e05b5ab9ebe02..4b10b1f86d66e 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-5.ll
@@ -1117,277 +1117,278 @@ define void @load_i8_stride5_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i8_stride5_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4) nounwind {
 ; SSE-LABEL: load_i8_stride5_vf16:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    movdqa 64(%rdi), %xmm9
-; SSE-NEXT:    movdqa (%rdi), %xmm1
-; SSE-NEXT:    movdqa 16(%rdi), %xmm6
-; SSE-NEXT:    movdqa 32(%rdi), %xmm10
-; SSE-NEXT:    movdqa 48(%rdi), %xmm2
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255]
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    pandn %xmm10, %xmm0
-; SSE-NEXT:    movdqa %xmm2, %xmm4
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm3, %xmm4
-; SSE-NEXT:    por %xmm0, %xmm4
-; SSE-NEXT:    pxor %xmm8, %xmm8
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3],xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[0,1,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm8[8],xmm4[9],xmm8[9],xmm4[10],xmm8[10],xmm4[11],xmm8[11],xmm4[12],xmm8[12],xmm4[13],xmm8[13],xmm4[14],xmm8[14],xmm4[15],xmm8[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm5[2],xmm4[3],xmm5[3]
-; SSE-NEXT:    packuswb %xmm4, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [255,255,255,255,255,255,255,0,0,0,0,0,0,255,255,255]
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255]
-; SSE-NEXT:    movdqa %xmm4, %xmm5
-; SSE-NEXT:    pandn %xmm6, %xmm5
-; SSE-NEXT:    movdqa %xmm6, %xmm15
-; SSE-NEXT:    movdqa %xmm1, %xmm6
-; SSE-NEXT:    movdqa %xmm1, %xmm13
-; SSE-NEXT:    pand %xmm4, %xmm6
-; SSE-NEXT:    por %xmm5, %xmm6
-; SSE-NEXT:    movdqa %xmm6, %xmm5
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm8[0],xmm5[1],xmm8[1],xmm5[2],xmm8[2],xmm5[3],xmm8[3],xmm5[4],xmm8[4],xmm5[5],xmm8[5],xmm5[6],xmm8[6],xmm5[7],xmm8[7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [0,65535,65535,65535,0,0,65535,65535]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm8[8],xmm6[9],xmm8[9],xmm6[10],xmm8[10],xmm6[11],xmm8[11],xmm6[12],xmm8[12],xmm6[13],xmm8[13],xmm6[14],xmm8[14],xmm6[15],xmm8[15]
-; SSE-NEXT:    pand %xmm7, %xmm6
-; SSE-NEXT:    pandn %xmm5, %xmm7
-; SSE-NEXT:    por %xmm6, %xmm7
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm7[0,2,1,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,6,5,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,2,1,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm5[0,1,2,3,4,6,5,7]
-; SSE-NEXT:    packuswb %xmm7, %xmm7
-; SSE-NEXT:    pand %xmm11, %xmm7
-; SSE-NEXT:    movdqa %xmm11, %xmm5
-; SSE-NEXT:    pandn %xmm0, %xmm5
-; SSE-NEXT:    por %xmm5, %xmm7
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0]
-; SSE-NEXT:    pand %xmm6, %xmm7
-; SSE-NEXT:    movdqa %xmm9, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3],xmm9[4],xmm8[4],xmm9[5],xmm8[5],xmm9[6],xmm8[6],xmm9[7],xmm8[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm8[8],xmm1[9],xmm8[9],xmm1[10],xmm8[10],xmm1[11],xmm8[11],xmm1[12],xmm8[12],xmm1[13],xmm8[13],xmm1[14],xmm8[14],xmm1[15],xmm8[15]
+; SSE-NEXT:    movdqa 64(%rdi), %xmm7
+; SSE-NEXT:    movdqa (%rdi), %xmm2
+; SSE-NEXT:    movdqa 16(%rdi), %xmm5
+; SSE-NEXT:    movdqa 32(%rdi), %xmm1
+; SSE-NEXT:    movdqa 48(%rdi), %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255]
+; SSE-NEXT:    movdqa %xmm8, %xmm0
+; SSE-NEXT:    pandn %xmm1, %xmm0
+; SSE-NEXT:    movdqa %xmm1, %xmm11
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    movdqa %xmm3, %xmm13
+; SSE-NEXT:    pand %xmm8, %xmm1
+; SSE-NEXT:    por %xmm0, %xmm1
+; SSE-NEXT:    pxor %xmm6, %xmm6
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3],xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[0,1,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm6[8],xmm1[9],xmm6[9],xmm1[10],xmm6[10],xmm1[11],xmm6[11],xmm1[12],xmm6[12],xmm1[13],xmm6[13],xmm1[14],xmm6[14],xmm1[15],xmm6[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; SSE-NEXT:    packuswb %xmm1, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,0,2,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,255,255,255,0,0,0,0,0,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm0, %xmm4
+; SSE-NEXT:    pandn %xmm1, %xmm4
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255]
+; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    pandn %xmm5, %xmm1
+; SSE-NEXT:    movdqa %xmm5, %xmm12
+; SSE-NEXT:    movdqa %xmm2, %xmm5
+; SSE-NEXT:    movdqa %xmm2, %xmm10
+; SSE-NEXT:    pand %xmm3, %xmm5
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3],xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [0,65535,65535,65535,0,0,65535,65535]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm6[8],xmm5[9],xmm6[9],xmm5[10],xmm6[10],xmm5[11],xmm6[11],xmm5[12],xmm6[12],xmm5[13],xmm6[13],xmm5[14],xmm6[14],xmm5[15],xmm6[15]
+; SSE-NEXT:    pand %xmm9, %xmm5
+; SSE-NEXT:    pandn %xmm1, %xmm9
+; SSE-NEXT:    por %xmm5, %xmm9
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm9[0,2,1,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,5,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,2,1,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,5,7]
+; SSE-NEXT:    packuswb %xmm1, %xmm1
+; SSE-NEXT:    pand %xmm0, %xmm1
+; SSE-NEXT:    por %xmm4, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0]
+; SSE-NEXT:    pand %xmm15, %xmm1
+; SSE-NEXT:    movdqa %xmm7, %xmm9
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm6[0],xmm9[1],xmm6[1],xmm9[2],xmm6[2],xmm9[3],xmm6[3],xmm9[4],xmm6[4],xmm9[5],xmm6[5],xmm9[6],xmm6[6],xmm9[7],xmm6[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm6[8],xmm7[9],xmm6[9],xmm7[10],xmm6[10],xmm7[11],xmm6[11],xmm7[12],xmm6[12],xmm7[13],xmm6[13],xmm7[14],xmm6[14],xmm7[15],xmm6[15]
+; SSE-NEXT:    movdqa %xmm7, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,0],xmm9[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm9[2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,6,4]
+; SSE-NEXT:    packuswb %xmm5, %xmm5
+; SSE-NEXT:    movdqa %xmm15, %xmm2
+; SSE-NEXT:    pandn %xmm5, %xmm2
+; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm3, %xmm5
+; SSE-NEXT:    pandn %xmm11, %xmm5
+; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    movdqa %xmm13, %xmm4
+; SSE-NEXT:    pand %xmm3, %xmm1
+; SSE-NEXT:    por %xmm5, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, %xmm5
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm9[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm9[2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,4]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm6, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    por %xmm7, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm4, %xmm7
-; SSE-NEXT:    pandn %xmm10, %xmm7
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pand %xmm4, %xmm0
-; SSE-NEXT:    por %xmm7, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm12
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm8[8],xmm12[9],xmm8[9],xmm12[10],xmm8[10],xmm12[11],xmm8[11],xmm12[12],xmm8[12],xmm12[13],xmm8[13],xmm12[14],xmm8[14],xmm12[15],xmm8[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3],xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
-; SSE-NEXT:    movdqa %xmm0, %xmm7
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[1,0],xmm12[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm12[2,3]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm6[8],xmm5[9],xmm6[9],xmm5[10],xmm6[10],xmm5[11],xmm6[11],xmm5[12],xmm6[12],xmm5[13],xmm6[13],xmm5[14],xmm6[14],xmm5[15],xmm6[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3],xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
+; SSE-NEXT:    movdqa %xmm1, %xmm11
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[1,0],xmm5[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[2,0],xmm5[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm11[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,2,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    psllq $48, %xmm1
+; SSE-NEXT:    packuswb %xmm5, %xmm1
 ; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm14, %xmm12
-; SSE-NEXT:    movdqa %xmm13, %xmm1
-; SSE-NEXT:    pandn %xmm13, %xmm12
-; SSE-NEXT:    movdqa %xmm15, %xmm13
-; SSE-NEXT:    movdqa %xmm15, %xmm2
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm14, %xmm13
-; SSE-NEXT:    por %xmm12, %xmm13
-; SSE-NEXT:    movdqa %xmm13, %xmm12
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm8[8],xmm12[9],xmm8[9],xmm12[10],xmm8[10],xmm12[11],xmm8[11],xmm12[12],xmm8[12],xmm12[13],xmm8[13],xmm12[14],xmm8[14],xmm12[15],xmm8[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm8[0],xmm13[1],xmm8[1],xmm13[2],xmm8[2],xmm13[3],xmm8[3],xmm13[4],xmm8[4],xmm13[5],xmm8[5],xmm13[6],xmm8[6],xmm13[7],xmm8[7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [65535,65535,0,0,65535,65535,65535,0]
-; SSE-NEXT:    pand %xmm15, %xmm13
-; SSE-NEXT:    pandn %xmm12, %xmm15
-; SSE-NEXT:    por %xmm13, %xmm15
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm15[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,4,7,6,5]
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[1,2,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,7,4,5,7]
-; SSE-NEXT:    packuswb %xmm12, %xmm12
-; SSE-NEXT:    pand %xmm11, %xmm12
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,2,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[1,2,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    psllq $48, %xmm0
-; SSE-NEXT:    packuswb %xmm7, %xmm0
-; SSE-NEXT:    movdqa %xmm5, %xmm7
-; SSE-NEXT:    pandn %xmm0, %xmm11
+; SSE-NEXT:    movdqa %xmm14, %xmm5
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm10, %xmm5
+; SSE-NEXT:    movdqa %xmm12, %xmm2
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm12, %xmm11
+; SSE-NEXT:    pand %xmm14, %xmm11
+; SSE-NEXT:    por %xmm5, %xmm11
+; SSE-NEXT:    movdqa %xmm11, %xmm5
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm6[8],xmm5[9],xmm6[9],xmm5[10],xmm6[10],xmm5[11],xmm6[11],xmm5[12],xmm6[12],xmm5[13],xmm6[13],xmm5[14],xmm6[14],xmm5[15],xmm6[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm6[0],xmm11[1],xmm6[1],xmm11[2],xmm6[2],xmm11[3],xmm6[3],xmm11[4],xmm6[4],xmm11[5],xmm6[5],xmm11[6],xmm6[6],xmm11[7],xmm6[7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,65535,0,0,65535,65535,65535,0]
+; SSE-NEXT:    pand %xmm12, %xmm11
+; SSE-NEXT:    pandn %xmm5, %xmm12
 ; SSE-NEXT:    por %xmm11, %xmm12
-; SSE-NEXT:    pand %xmm6, %xmm12
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm9[3,0]
-; SSE-NEXT:    movaps %xmm9, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm7[0,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,5,4]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm6, %xmm5
-; SSE-NEXT:    pandn %xmm0, %xmm5
-; SSE-NEXT:    por %xmm12, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm12[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,7,6,5]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,2,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,4,5,7]
+; SSE-NEXT:    packuswb %xmm5, %xmm5
+; SSE-NEXT:    pand %xmm0, %xmm5
+; SSE-NEXT:    pandn %xmm1, %xmm0
+; SSE-NEXT:    por %xmm5, %xmm0
+; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm9[3,0]
+; SSE-NEXT:    movaps %xmm9, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm1[0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm5[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,5,4]
+; SSE-NEXT:    packuswb %xmm1, %xmm1
+; SSE-NEXT:    movdqa %xmm15, %xmm5
+; SSE-NEXT:    pandn %xmm1, %xmm5
+; SSE-NEXT:    por %xmm0, %xmm5
 ; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255]
 ; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm1, %xmm5
-; SSE-NEXT:    pand %xmm12, %xmm2
-; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
+; SSE-NEXT:    pandn %xmm10, %xmm0
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    pand %xmm12, %xmm1
+; SSE-NEXT:    por %xmm0, %xmm1
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm6[8],xmm0[9],xmm6[9],xmm0[10],xmm6[10],xmm0[11],xmm6[11],xmm0[12],xmm6[12],xmm0[13],xmm6[13],xmm0[14],xmm6[14],xmm0[15],xmm6[15]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [65535,65535,65535,0,0,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm13, %xmm15
-; SSE-NEXT:    pandn %xmm0, %xmm15
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3],xmm2[4],xmm8[4],xmm2[5],xmm8[5],xmm2[6],xmm8[6],xmm2[7],xmm8[7]
-; SSE-NEXT:    pand %xmm13, %xmm2
-; SSE-NEXT:    por %xmm15, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm13, %xmm5
+; SSE-NEXT:    pandn %xmm0, %xmm5
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3],xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
+; SSE-NEXT:    pand %xmm13, %xmm1
+; SSE-NEXT:    por %xmm5, %xmm1
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
 ; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [0,0,0,65535,65535,65535,65535,65535]
-; SSE-NEXT:    pandn %xmm0, %xmm15
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm11, %xmm0
-; SSE-NEXT:    movdqa %xmm11, %xmm7
-; SSE-NEXT:    pand %xmm14, %xmm7
-; SSE-NEXT:    pandn %xmm10, %xmm14
-; SSE-NEXT:    pand %xmm12, %xmm11
-; SSE-NEXT:    pandn %xmm10, %xmm12
-; SSE-NEXT:    pand %xmm4, %xmm10
-; SSE-NEXT:    por %xmm0, %xmm10
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm8[0],xmm10[1],xmm8[1],xmm10[2],xmm8[2],xmm10[3],xmm8[3],xmm10[4],xmm8[4],xmm10[5],xmm8[5],xmm10[6],xmm8[6],xmm10[7],xmm8[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[0,1,2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[2,0],xmm0[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm10[0,2]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm15, %xmm11
+; SSE-NEXT:    pandn %xmm0, %xmm11
+; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    pandn %xmm4, %xmm1
+; SSE-NEXT:    movdqa %xmm4, %xmm5
+; SSE-NEXT:    movdqa %xmm4, %xmm10
+; SSE-NEXT:    pand %xmm14, %xmm5
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm0, %xmm14
+; SSE-NEXT:    pand %xmm12, %xmm10
+; SSE-NEXT:    pandn %xmm0, %xmm12
+; SSE-NEXT:    pand %xmm3, %xmm0
+; SSE-NEXT:    por %xmm1, %xmm0
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm6[8],xmm1[9],xmm6[9],xmm1[10],xmm6[10],xmm1[11],xmm6[11],xmm1[12],xmm6[12],xmm1[13],xmm6[13],xmm1[14],xmm6[14],xmm1[15],xmm6[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3],xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,1,2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,6,6,6]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm2[0,1,2,3,4,5,6,5]
 ; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
-; SSE-NEXT:    por %xmm15, %xmm1
+; SSE-NEXT:    pand %xmm15, %xmm1
+; SSE-NEXT:    por %xmm11, %xmm1
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm2[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm10[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm6, %xmm10
-; SSE-NEXT:    pandn %xmm0, %xmm10
-; SSE-NEXT:    pand %xmm6, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm10
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pand %xmm3, %xmm0
-; SSE-NEXT:    pandn %xmm5, %xmm3
-; SSE-NEXT:    por %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3],xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,1],xmm0[2,0]
-; SSE-NEXT:    por %xmm7, %xmm14
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm1[8],xmm14[9],xmm1[9],xmm14[10],xmm1[10],xmm14[11],xmm1[11],xmm14[12],xmm1[12],xmm14[13],xmm1[13],xmm14[14],xmm1[14],xmm14[15],xmm1[15]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[3,1],xmm0[2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm14[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,0,1,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,5,6]
-; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [0,0,0,65535,65535,65535,65535,65535]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm3[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[3,1,2,0]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[3,0,1,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,7,4,6,7]
-; SSE-NEXT:    packuswb %xmm7, %xmm7
-; SSE-NEXT:    pandn %xmm7, %xmm1
-; SSE-NEXT:    movaps %xmm9, %xmm7
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm2[3,0]
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movaps %xmm2, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm7[0,2]
-; SSE-NEXT:    pand %xmm6, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,5]
-; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pandn %xmm1, %xmm6
-; SSE-NEXT:    por %xmm0, %xmm6
-; SSE-NEXT:    por %xmm11, %xmm12
-; SSE-NEXT:    movdqa %xmm12, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm8[8],xmm1[9],xmm8[9],xmm1[10],xmm8[10],xmm1[11],xmm8[11],xmm1[12],xmm8[12],xmm1[13],xmm8[13],xmm1[14],xmm8[14],xmm1[15],xmm8[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm8[0],xmm12[1],xmm8[1],xmm12[2],xmm8[2],xmm12[3],xmm8[3],xmm12[4],xmm8[4],xmm12[5],xmm8[5],xmm12[6],xmm8[6],xmm12[7],xmm8[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[0,1,1,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,3],xmm1[1,2]
-; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    packuswb %xmm0, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0]
+; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    pandn %xmm2, %xmm0
 ; SSE-NEXT:    pand %xmm4, %xmm1
-; SSE-NEXT:    pandn %xmm5, %xmm4
-; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm8[8],xmm4[9],xmm8[9],xmm4[10],xmm8[10],xmm4[11],xmm8[11],xmm4[12],xmm8[12],xmm4[13],xmm8[13],xmm4[14],xmm8[14],xmm4[15],xmm8[15]
-; SSE-NEXT:    pand %xmm13, %xmm4
-; SSE-NEXT:    pandn %xmm1, %xmm13
+; SSE-NEXT:    por %xmm1, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    pand %xmm8, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm11, %xmm8
+; SSE-NEXT:    por %xmm1, %xmm8
+; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    pxor %xmm2, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm2[8],xmm1[9],xmm2[9],xmm1[10],xmm2[10],xmm1[11],xmm2[11],xmm1[12],xmm2[12],xmm1[13],xmm2[13],xmm1[14],xmm2[14],xmm1[15],xmm2[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1],xmm8[2],xmm2[2],xmm8[3],xmm2[3],xmm8[4],xmm2[4],xmm8[5],xmm2[5],xmm8[6],xmm2[6],xmm8[7],xmm2[7]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[3,1],xmm1[2,0]
+; SSE-NEXT:    por %xmm5, %xmm14
+; SSE-NEXT:    movdqa %xmm14, %xmm1
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm2[8],xmm14[9],xmm2[9],xmm14[10],xmm2[10],xmm14[11],xmm2[11],xmm14[12],xmm2[12],xmm14[13],xmm2[13],xmm14[14],xmm2[14],xmm14[15],xmm2[15]
+; SSE-NEXT:    pxor %xmm5, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[3,1],xmm1[2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm14[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,0,1,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,6]
+; SSE-NEXT:    packuswb %xmm2, %xmm1
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm8[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,0]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,0,1,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,4,6,7]
+; SSE-NEXT:    packuswb %xmm2, %xmm2
+; SSE-NEXT:    pand %xmm15, %xmm1
+; SSE-NEXT:    pandn %xmm2, %xmm15
+; SSE-NEXT:    por %xmm1, %xmm15
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,2,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[2,0],xmm7[3,0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm7[3,1,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm9[0,2]
+; SSE-NEXT:    pand %xmm4, %xmm15
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm7[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,7,5]
+; SSE-NEXT:    packuswb %xmm2, %xmm2
+; SSE-NEXT:    pandn %xmm2, %xmm4
+; SSE-NEXT:    por %xmm15, %xmm4
+; SSE-NEXT:    por %xmm10, %xmm12
+; SSE-NEXT:    movdqa %xmm12, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm5[0],xmm12[1],xmm5[1],xmm12[2],xmm5[2],xmm12[3],xmm5[3],xmm12[4],xmm5[4],xmm12[5],xmm5[5],xmm12[6],xmm5[6],xmm12[7],xmm5[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm12[0,1,1,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,3],xmm2[1,2]
+; SSE-NEXT:    pand %xmm3, %xmm6
+; SSE-NEXT:    pandn %xmm11, %xmm3
+; SSE-NEXT:    por %xmm6, %xmm3
+; SSE-NEXT:    movdqa %xmm3, %xmm2
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3],xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm5[8],xmm3[9],xmm5[9],xmm3[10],xmm5[10],xmm3[11],xmm5[11],xmm3[12],xmm5[12],xmm3[13],xmm5[13],xmm3[14],xmm5[14],xmm3[15],xmm5[15]
+; SSE-NEXT:    pand %xmm13, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm13
 ; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,2,3,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm12[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    por %xmm4, %xmm13
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm12[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm7[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    packuswb %xmm2, %xmm6
+; SSE-NEXT:    por %xmm3, %xmm13
 ; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm13[3,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,6,4]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,1,2,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    packuswb %xmm4, %xmm4
-; SSE-NEXT:    pand %xmm3, %xmm4
-; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    por %xmm3, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[3,1,2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm9[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    packuswb %xmm1, %xmm2
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, (%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, (%rdx)
-; SSE-NEXT:    movdqa %xmm10, (%rcx)
-; SSE-NEXT:    movdqa %xmm6, (%r8)
-; SSE-NEXT:    movaps %xmm4, (%r9)
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm13[3,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,6,4]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[3,1,2,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    packuswb %xmm7, %xmm7
+; SSE-NEXT:    pand %xmm3, %xmm7
+; SSE-NEXT:    pandn %xmm6, %xmm3
+; SSE-NEXT:    por %xmm7, %xmm3
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm8[0,1,2,1,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1]
+; SSE-NEXT:    packuswb %xmm2, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm1[2,1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm1, (%rdx)
+; SSE-NEXT:    movdqa %xmm0, (%rcx)
+; SSE-NEXT:    movdqa %xmm4, (%r8)
+; SSE-NEXT:    movaps %xmm3, (%r9)
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i8_stride5_vf16:
@@ -2115,137 +2116,131 @@ define void @load_i8_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    subq $184, %rsp
 ; SSE-NEXT:    movdqa (%rdi), %xmm9
-; SSE-NEXT:    movdqa 16(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 16(%rdi), %xmm11
 ; SSE-NEXT:    movdqa 32(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 48(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 48(%rdi), %xmm5
 ; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255]
 ; SSE-NEXT:    movdqa %xmm4, %xmm0
 ; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm4, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pxor %xmm5, %xmm5
+; SSE-NEXT:    pxor %xmm6, %xmm6
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3],xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,1,1,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm5[8],xmm1[9],xmm5[9],xmm1[10],xmm5[10],xmm1[11],xmm5[11],xmm1[12],xmm5[12],xmm1[13],xmm5[13],xmm1[14],xmm5[14],xmm1[15],xmm5[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm6[8],xmm1[9],xmm6[9],xmm1[10],xmm6[10],xmm1[11],xmm6[11],xmm1[12],xmm6[12],xmm1[13],xmm6[13],xmm1[14],xmm6[14],xmm1[15],xmm6[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
 ; SSE-NEXT:    packuswb %xmm1, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,0,2,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,255,255,255,255,255,255,0,0,0,0,0,0,255,255,255]
-; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,0,0,0,0,0,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm8, %xmm0
 ; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255]
-; SSE-NEXT:    movdqa %xmm15, %xmm1
-; SSE-NEXT:    pandn %xmm3, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255]
-; SSE-NEXT:    pandn %xmm9, %xmm11
-; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255]
-; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255]
+; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    pandn %xmm11, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255]
+; SSE-NEXT:    pandn %xmm9, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm13, %xmm2
 ; SSE-NEXT:    pandn %xmm9, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm4, %xmm2
 ; SSE-NEXT:    pandn %xmm9, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, %xmm2
+; SSE-NEXT:    movdqa %xmm10, %xmm2
 ; SSE-NEXT:    pandn %xmm9, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm15, %xmm9
+; SSE-NEXT:    pand %xmm10, %xmm9
 ; SSE-NEXT:    por %xmm1, %xmm9
 ; SSE-NEXT:    movdqa %xmm9, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3],xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3],xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [0,65535,65535,65535,0,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm1, %xmm6
-; SSE-NEXT:    pandn %xmm2, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm5[8],xmm9[9],xmm5[9],xmm9[10],xmm5[10],xmm9[11],xmm5[11],xmm9[12],xmm5[12],xmm9[13],xmm5[13],xmm9[14],xmm5[14],xmm9[15],xmm5[15]
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm6[8],xmm9[9],xmm6[9],xmm9[10],xmm6[10],xmm9[11],xmm6[11],xmm9[12],xmm6[12],xmm9[13],xmm6[13],xmm9[14],xmm6[14],xmm9[15],xmm6[15]
 ; SSE-NEXT:    pand %xmm1, %xmm9
-; SSE-NEXT:    por %xmm6, %xmm9
+; SSE-NEXT:    por %xmm3, %xmm9
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm9[0,2,1,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,2,1,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,5,7]
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pand %xmm13, %xmm2
+; SSE-NEXT:    pand %xmm8, %xmm2
 ; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa 64(%rdi), %xmm6
-; SSE-NEXT:    movdqa %xmm6, %xmm3
+; SSE-NEXT:    movdqa 64(%rdi), %xmm7
+; SSE-NEXT:    movdqa %xmm7, %xmm15
 ; SSE-NEXT:    pxor %xmm0, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm0[8],xmm6[9],xmm0[9],xmm6[10],xmm0[10],xmm6[11],xmm0[11],xmm6[12],xmm0[12],xmm6[13],xmm0[13],xmm6[14],xmm0[14],xmm6[15],xmm0[15]
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    movdqa %xmm6, %xmm8
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm3[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm3[2,3]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm0[0],xmm15[1],xmm0[1],xmm15[2],xmm0[2],xmm15[3],xmm0[3],xmm15[4],xmm0[4],xmm15[5],xmm0[5],xmm15[6],xmm0[6],xmm15[7],xmm0[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm0[8],xmm7[9],xmm0[9],xmm7[10],xmm0[10],xmm7[11],xmm0[11],xmm7[12],xmm0[12],xmm7[13],xmm0[13],xmm7[14],xmm0[14],xmm7[15],xmm0[15]
+; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm15[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm15[2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,4]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
 ; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0]
-; SSE-NEXT:    movdqa %xmm9, %xmm6
-; SSE-NEXT:    pandn %xmm0, %xmm6
+; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
 ; SSE-NEXT:    pand %xmm9, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 112(%rdi), %xmm10
+; SSE-NEXT:    por %xmm2, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 112(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    pandn %xmm10, %xmm0
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 128(%rdi), %xmm7
-; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm2, %xmm0
+; SSE-NEXT:    movdqa 128(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm4, %xmm2
 ; SSE-NEXT:    por %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm0[0,1,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3],xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[0,1,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm6[8],xmm2[9],xmm6[9],xmm2[10],xmm6[10],xmm2[11],xmm6[11],xmm2[12],xmm6[12],xmm2[13],xmm6[13],xmm2[14],xmm6[14],xmm2[15],xmm6[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm6[2],xmm2[3],xmm6[3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
 ; SSE-NEXT:    packuswb %xmm2, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,3]
-; SSE-NEXT:    movdqa %xmm13, %xmm2
-; SSE-NEXT:    movdqa %xmm13, %xmm3
+; SSE-NEXT:    movdqa %xmm8, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    movdqa 96(%rdi), %xmm4
-; SSE-NEXT:    movdqa %xmm4, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pandn %xmm4, %xmm0
-; SSE-NEXT:    movdqa 80(%rdi), %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm15, %xmm6
-; SSE-NEXT:    por %xmm0, %xmm6
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm5[8],xmm6[9],xmm5[9],xmm6[10],xmm5[10],xmm6[11],xmm5[11],xmm6[12],xmm5[12],xmm6[13],xmm5[13],xmm6[14],xmm5[14],xmm6[15],xmm5[15]
-; SSE-NEXT:    pand %xmm1, %xmm6
+; SSE-NEXT:    movdqa 96(%rdi), %xmm12
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    pandn %xmm12, %xmm0
+; SSE-NEXT:    movdqa 80(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm10, %xmm3
+; SSE-NEXT:    por %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm3, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3],xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm6[8],xmm3[9],xmm6[9],xmm3[10],xmm6[10],xmm3[11],xmm6[11],xmm3[12],xmm6[12],xmm3[13],xmm6[13],xmm3[14],xmm6[14],xmm3[15],xmm6[15]
+; SSE-NEXT:    pand %xmm1, %xmm3
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    por %xmm6, %xmm1
+; SSE-NEXT:    por %xmm3, %xmm1
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,2,1,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,5,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm13, %xmm0
+; SSE-NEXT:    pand %xmm8, %xmm0
 ; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa 144(%rdi), %xmm12
-; SSE-NEXT:    movdqa %xmm12, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3],xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
+; SSE-NEXT:    movdqa 144(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3],xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm5[8],xmm12[9],xmm5[9],xmm12[10],xmm5[10],xmm12[11],xmm5[11],xmm12[12],xmm5[12],xmm12[13],xmm5[13],xmm12[14],xmm5[14],xmm12[15],xmm5[15]
-; SSE-NEXT:    movdqa %xmm12, %xmm1
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm6[8],xmm4[9],xmm6[9],xmm4[10],xmm6[10],xmm4[11],xmm6[11],xmm4[12],xmm6[12],xmm4[13],xmm6[13],xmm4[14],xmm6[14],xmm4[15],xmm6[15]
+; SSE-NEXT:    movdqa %xmm4, %xmm1
+; SSE-NEXT:    movdqa %xmm4, (%rsp) # 16-byte Spill
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm2[0,0]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
@@ -2257,15 +2252,14 @@ define void @load_i8_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pand %xmm9, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    movdqa %xmm10, %xmm0
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pand %xmm15, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm5[8],xmm0[9],xmm5[9],xmm0[10],xmm5[10],xmm0[11],xmm5[11],xmm0[12],xmm5[12],xmm0[13],xmm5[13],xmm0[14],xmm5[14],xmm0[15],xmm5[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3],xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
-; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    pand %xmm10, %xmm5
+; SSE-NEXT:    por %xmm0, %xmm5
+; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm6[8],xmm0[9],xmm6[9],xmm0[10],xmm6[10],xmm0[11],xmm6[11],xmm0[12],xmm6[12],xmm0[13],xmm6[13],xmm0[14],xmm6[14],xmm0[15],xmm6[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3],xmm5[4],xmm6[4],xmm5[5],xmm6[5],xmm5[6],xmm6[6],xmm5[7],xmm6[7]
+; SSE-NEXT:    movdqa %xmm5, %xmm2
 ; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm0[0,0]
 ; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm0[2,3]
 ; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2,1,3]
@@ -2273,37 +2267,37 @@ define void @load_i8_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    psllq $48, %xmm1
-; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm13, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255]
-; SSE-NEXT:    pand %xmm13, %xmm1
-; SSE-NEXT:    por %xmm11, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm5[8],xmm6[9],xmm5[9],xmm6[10],xmm5[10],xmm6[11],xmm5[11],xmm6[12],xmm5[12],xmm6[13],xmm5[13],xmm6[14],xmm5[14],xmm6[15],xmm5[15]
+; SSE-NEXT:    psllq $48, %xmm5
+; SSE-NEXT:    packuswb %xmm0, %xmm5
+; SSE-NEXT:    movdqa %xmm8, %xmm2
+; SSE-NEXT:    pandn %xmm5, %xmm2
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm11, %xmm1
+; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm6[8],xmm3[9],xmm6[9],xmm3[10],xmm6[10],xmm3[11],xmm6[11],xmm3[12],xmm6[12],xmm3[13],xmm6[13],xmm3[14],xmm6[14],xmm3[15],xmm6[15]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,0,0,65535,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm0, %xmm11
-; SSE-NEXT:    pandn %xmm6, %xmm11
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3],xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
+; SSE-NEXT:    movdqa %xmm0, %xmm5
+; SSE-NEXT:    pandn %xmm3, %xmm5
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3],xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
 ; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    por %xmm11, %xmm1
+; SSE-NEXT:    por %xmm5, %xmm1
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,4,5,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm3, %xmm1
-; SSE-NEXT:    movdqa %xmm3, %xmm11
+; SSE-NEXT:    pand %xmm8, %xmm1
 ; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm8, %xmm2
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm6[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm2[0,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm6[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm15[3,0]
+; SSE-NEXT:    movaps %xmm15, %xmm3
+; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,5,4]
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
@@ -2312,70 +2306,71 @@ define void @load_i8_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pand %xmm9, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, %xmm2
-; SSE-NEXT:    pandn %xmm10, %xmm2
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    pand %xmm15, %xmm1
+; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pand %xmm10, %xmm1
 ; SSE-NEXT:    por %xmm2, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3],xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
-; SSE-NEXT:    movdqa %xmm1, %xmm6
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[1,0],xmm2[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0],xmm2[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm6[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm6[8],xmm2[9],xmm6[9],xmm2[10],xmm6[10],xmm2[11],xmm6[11],xmm2[12],xmm6[12],xmm2[13],xmm6[13],xmm2[14],xmm6[14],xmm2[15],xmm6[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3],xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,0],xmm2[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm2[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    psllq $48, %xmm1
 ; SSE-NEXT:    packuswb %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm13, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm7, %xmm2
-; SSE-NEXT:    movdqa (%rsp), %xmm8 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm8, %xmm6
-; SSE-NEXT:    pand %xmm13, %xmm6
-; SSE-NEXT:    por %xmm2, %xmm6
-; SSE-NEXT:    movdqa %xmm6, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3],xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
-; SSE-NEXT:    pand %xmm0, %xmm6
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm8, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm5, %xmm2
+; SSE-NEXT:    movdqa %xmm12, %xmm3
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm8, %xmm3
+; SSE-NEXT:    por %xmm2, %xmm3
+; SSE-NEXT:    movdqa %xmm3, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm6[8],xmm2[9],xmm6[9],xmm2[10],xmm6[10],xmm2[11],xmm6[11],xmm2[12],xmm6[12],xmm2[13],xmm6[13],xmm2[14],xmm6[14],xmm2[15],xmm6[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1],xmm3[2],xmm6[2],xmm3[3],xmm6[3],xmm3[4],xmm6[4],xmm3[5],xmm6[5],xmm3[6],xmm6[6],xmm3[7],xmm6[7]
+; SSE-NEXT:    pand %xmm0, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    por %xmm6, %xmm0
+; SSE-NEXT:    por %xmm3, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,4,5,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm11, %xmm2
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[2,0],xmm2[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm12[0,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm2[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,5,4]
-; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,0,0,0,0,0,0,255,255,255]
+; SSE-NEXT:    pand %xmm2, %xmm0
 ; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pand %xmm9, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm14, %xmm4
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,0,0,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm3, %xmm6
-; SSE-NEXT:    pandn %xmm2, %xmm6
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3],xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
-; SSE-NEXT:    pand %xmm3, %xmm4
-; SSE-NEXT:    por %xmm6, %xmm4
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm4[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm1[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm4[0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,5,4]
+; SSE-NEXT:    packuswb %xmm0, %xmm0
+; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    pandn %xmm0, %xmm1
+; SSE-NEXT:    pand %xmm9, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm13, %xmm11
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm11, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm6[8],xmm1[9],xmm6[9],xmm1[10],xmm6[10],xmm1[11],xmm6[11],xmm1[12],xmm6[12],xmm1[13],xmm6[13],xmm1[14],xmm6[14],xmm1[15],xmm6[15]
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,65535,65535,0,0,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm8, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm6[0],xmm11[1],xmm6[1],xmm11[2],xmm6[2],xmm11[3],xmm6[3],xmm11[4],xmm6[4],xmm11[5],xmm6[5],xmm11[6],xmm6[6],xmm11[7],xmm6[7]
+; SSE-NEXT:    pand %xmm8, %xmm11
+; SSE-NEXT:    por %xmm2, %xmm11
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm11[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
@@ -2383,40 +2378,35 @@ define void @load_i8_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [0,0,0,65535,65535,65535,65535,65535]
 ; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm1, %xmm10
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm13, %xmm6
-; SSE-NEXT:    movdqa %xmm13, %xmm12
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm11, %xmm6
-; SSE-NEXT:    movdqa %xmm14, %xmm4
-; SSE-NEXT:    pandn %xmm11, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm15, %xmm11
-; SSE-NEXT:    movdqa %xmm15, %xmm4
-; SSE-NEXT:    por %xmm0, %xmm11
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm5[8],xmm0[9],xmm5[9],xmm0[10],xmm5[10],xmm0[11],xmm5[11],xmm0[12],xmm5[12],xmm0[13],xmm5[13],xmm0[14],xmm5[14],xmm0[15],xmm5[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm5[0],xmm11[1],xmm5[1],xmm11[2],xmm5[2],xmm11[3],xmm5[3],xmm11[4],xmm5[4],xmm11[5],xmm5[5],xmm11[6],xmm5[6],xmm11[7],xmm5[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm11[0,1,2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[2,0],xmm0[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm11[0,2]
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm4, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm3, %xmm14
+; SSE-NEXT:    movdqa %xmm13, %xmm11
+; SSE-NEXT:    pandn %xmm3, %xmm11
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm10, %xmm3
+; SSE-NEXT:    por %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm3, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm6[8],xmm0[9],xmm6[9],xmm0[10],xmm6[10],xmm0[11],xmm6[11],xmm0[12],xmm6[12],xmm0[13],xmm6[13],xmm0[14],xmm6[14],xmm0[15],xmm6[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1],xmm3[2],xmm6[2],xmm3[3],xmm6[3],xmm3[4],xmm6[4],xmm3[5],xmm6[5],xmm3[6],xmm6[6],xmm3[7],xmm6[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm3[0,1,2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm0[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm3[0,2]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,6,6,6]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm13[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    packuswb %xmm0, %xmm11
-; SSE-NEXT:    pand %xmm10, %xmm11
-; SSE-NEXT:    por %xmm2, %xmm11
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[0,2,2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm11[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    packuswb %xmm0, %xmm3
+; SSE-NEXT:    pand %xmm1, %xmm3
+; SSE-NEXT:    por %xmm2, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[1,1,1,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,2,2,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
@@ -2424,264 +2414,260 @@ define void @load_i8_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
 ; SSE-NEXT:    movdqa %xmm9, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pand %xmm9, %xmm11
-; SSE-NEXT:    por %xmm11, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    pandn %xmm7, %xmm0
-; SSE-NEXT:    movdqa %xmm8, %xmm15
-; SSE-NEXT:    movdqa %xmm8, %xmm2
-; SSE-NEXT:    pand %xmm14, %xmm2
-; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm5[8],xmm0[9],xmm5[9],xmm0[10],xmm5[10],xmm0[11],xmm5[11],xmm0[12],xmm5[12],xmm0[13],xmm5[13],xmm0[14],xmm5[14],xmm0[15],xmm5[15]
-; SSE-NEXT:    movdqa %xmm3, %xmm11
-; SSE-NEXT:    pandn %xmm0, %xmm11
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3],xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
-; SSE-NEXT:    pand %xmm3, %xmm2
-; SSE-NEXT:    por %xmm11, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    pandn %xmm5, %xmm0
+; SSE-NEXT:    pand %xmm13, %xmm12
+; SSE-NEXT:    por %xmm0, %xmm12
+; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm6[8],xmm0[9],xmm6[9],xmm0[10],xmm6[10],xmm0[11],xmm6[11],xmm0[12],xmm6[12],xmm0[13],xmm6[13],xmm0[14],xmm6[14],xmm0[15],xmm6[15]
+; SSE-NEXT:    movdqa %xmm8, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm6[0],xmm12[1],xmm6[1],xmm12[2],xmm6[2],xmm12[3],xmm6[3],xmm12[4],xmm6[4],xmm12[5],xmm6[5],xmm12[6],xmm6[6],xmm12[7],xmm6[7]
+; SSE-NEXT:    pand %xmm8, %xmm12
+; SSE-NEXT:    por %xmm3, %xmm12
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm12[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm10, %xmm13
-; SSE-NEXT:    pandn %xmm0, %xmm13
-; SSE-NEXT:    movdqa %xmm4, %xmm11
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm2, %xmm11
-; SSE-NEXT:    movdqa %xmm1, %xmm5
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm12, %xmm1
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    movdqa %xmm2, %xmm7
-; SSE-NEXT:    pand %xmm12, %xmm7
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm8, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm14, %xmm5
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm14, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pandn %xmm8, %xmm14
-; SSE-NEXT:    pand %xmm4, %xmm8
-; SSE-NEXT:    por %xmm11, %xmm8
-; SSE-NEXT:    movdqa %xmm8, %xmm11
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm1[8],xmm11[9],xmm1[9],xmm11[10],xmm1[10],xmm11[11],xmm1[11],xmm11[12],xmm1[12],xmm11[13],xmm1[13],xmm11[14],xmm1[14],xmm11[15],xmm1[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm1[0],xmm8[1],xmm1[1],xmm8[2],xmm1[2],xmm8[3],xmm1[3],xmm8[4],xmm1[4],xmm8[5],xmm1[5],xmm8[6],xmm1[6],xmm8[7],xmm1[7]
-; SSE-NEXT:    pxor %xmm2, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[0,1,2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,0],xmm11[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,1],xmm8[0,2]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm11[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,6,6,6,6]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    packuswb %xmm8, %xmm1
-; SSE-NEXT:    pand %xmm10, %xmm1
-; SSE-NEXT:    por %xmm13, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm4[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm5[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm8[0],xmm11[1],xmm8[1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm11[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,7,6,5]
-; SSE-NEXT:    packuswb %xmm8, %xmm11
-; SSE-NEXT:    movdqa %xmm9, %xmm12
-; SSE-NEXT:    pandn %xmm11, %xmm12
-; SSE-NEXT:    pand %xmm9, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm12
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255]
-; SSE-NEXT:    pand %xmm13, %xmm1
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm11
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm2[8],xmm11[9],xmm2[9],xmm11[10],xmm2[10],xmm11[11],xmm2[11],xmm11[12],xmm2[12],xmm11[13],xmm2[13],xmm11[14],xmm2[14],xmm11[15],xmm2[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm11[2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,2,0]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,0,1,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,4,6,7]
-; SSE-NEXT:    packuswb %xmm1, %xmm1
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
 ; SSE-NEXT:    movdqa %xmm10, %xmm11
-; SSE-NEXT:    pandn %xmm1, %xmm11
-; SSE-NEXT:    por %xmm6, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm1[2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,0,1,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,6]
-; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm10, %xmm1
-; SSE-NEXT:    por %xmm11, %xmm1
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm10, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm6[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[0,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm6[0,1,2,3,4,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,7,5]
-; SSE-NEXT:    packuswb %xmm0, %xmm6
-; SSE-NEXT:    movdqa %xmm9, %xmm8
-; SSE-NEXT:    pandn %xmm6, %xmm8
-; SSE-NEXT:    pand %xmm9, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm8
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    pand %xmm13, %xmm15
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    por %xmm15, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    pxor %xmm6, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm6[8],xmm1[9],xmm6[9],xmm1[10],xmm6[10],xmm1[11],xmm6[11],xmm1[12],xmm6[12],xmm1[13],xmm6[13],xmm1[14],xmm6[14],xmm1[15],xmm6[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3],xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm1[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    por %xmm7, %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3],xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm6[8],xmm2[9],xmm6[9],xmm2[10],xmm6[10],xmm2[11],xmm6[11],xmm2[12],xmm6[12],xmm2[13],xmm6[13],xmm2[14],xmm6[14],xmm2[15],xmm6[15]
-; SSE-NEXT:    pxor %xmm13, %xmm13
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,1],xmm1[2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,0,1,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,6]
-; SSE-NEXT:    packuswb %xmm2, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [0,0,0,65535,65535,65535,65535,65535]
-; SSE-NEXT:    pand %xmm2, %xmm1
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm0[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[3,1,2,0]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[3,0,1,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,4,6,7]
-; SSE-NEXT:    packuswb %xmm6, %xmm6
-; SSE-NEXT:    pandn %xmm6, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm5, %xmm11
 ; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    movdqa %xmm4, %xmm15
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm5[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm2[0,2]
-; SSE-NEXT:    pand %xmm9, %xmm1
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm5[0,1,2,3,4,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,7,5]
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pandn %xmm2, %xmm9
-; SSE-NEXT:    por %xmm1, %xmm9
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm13[8],xmm1[9],xmm13[9],xmm1[10],xmm13[10],xmm1[11],xmm13[11],xmm1[12],xmm13[12],xmm1[13],xmm13[13],xmm1[14],xmm13[14],xmm1[15],xmm13[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1],xmm0[2],xmm13[2],xmm0[3],xmm13[3],xmm0[4],xmm13[4],xmm0[5],xmm13[5],xmm0[6],xmm13[6],xmm0[7],xmm13[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,1,1,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,3],xmm1[1,2]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2,3,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm0[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    packuswb %xmm1, %xmm2
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm4, %xmm6
-; SSE-NEXT:    pandn %xmm2, %xmm6
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255]
-; SSE-NEXT:    pand %xmm5, %xmm2
+; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255]
+; SSE-NEXT:    pand %xmm4, %xmm0
+; SSE-NEXT:    movdqa %xmm5, %xmm12
+; SSE-NEXT:    pand %xmm4, %xmm12
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm7, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm13, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm13, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm7, %xmm13
+; SSE-NEXT:    pand %xmm10, %xmm7
+; SSE-NEXT:    por %xmm11, %xmm7
+; SSE-NEXT:    movdqa %xmm7, %xmm11
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm6[8],xmm11[9],xmm6[9],xmm11[10],xmm6[10],xmm11[11],xmm6[11],xmm11[12],xmm6[12],xmm11[13],xmm6[13],xmm11[14],xmm6[14],xmm11[15],xmm6[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3],xmm7[4],xmm6[4],xmm7[5],xmm6[5],xmm7[6],xmm6[6],xmm7[7],xmm6[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,1,2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm11[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,1],xmm7[0,2]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm11[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,3,2,1,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,6,6,6]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    packuswb %xmm7, %xmm2
+; SSE-NEXT:    pand %xmm1, %xmm2
+; SSE-NEXT:    por %xmm3, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm10[1,1,1,1]
+; SSE-NEXT:    movdqa (%rsp), %xmm4 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm4[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm3[0],xmm7[1],xmm3[1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm7[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,5]
+; SSE-NEXT:    packuswb %xmm3, %xmm3
+; SSE-NEXT:    movdqa %xmm9, %xmm7
+; SSE-NEXT:    pandn %xmm3, %xmm7
+; SSE-NEXT:    pand %xmm9, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm7
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm15, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255]
+; SSE-NEXT:    pand %xmm6, %xmm2
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1],xmm0[2],xmm13[2],xmm0[3],xmm13[3],xmm0[4],xmm13[4],xmm0[5],xmm13[5],xmm0[6],xmm13[6],xmm0[7],xmm13[7]
-; SSE-NEXT:    movdqa %xmm3, %xmm11
-; SSE-NEXT:    pandn %xmm0, %xmm11
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm13[8],xmm2[9],xmm13[9],xmm2[10],xmm13[10],xmm2[11],xmm13[11],xmm2[12],xmm13[12],xmm2[13],xmm13[13],xmm2[14],xmm13[14],xmm2[15],xmm13[15]
-; SSE-NEXT:    pand %xmm3, %xmm2
-; SSE-NEXT:    por %xmm11, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,4]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    pxor %xmm11, %xmm11
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm11[8],xmm3[9],xmm11[9],xmm3[10],xmm11[10],xmm3[11],xmm11[11],xmm3[12],xmm11[12],xmm3[13],xmm11[13],xmm3[14],xmm11[14],xmm3[15],xmm11[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm11[0],xmm2[1],xmm11[1],xmm2[2],xmm11[2],xmm2[3],xmm11[3],xmm2[4],xmm11[4],xmm2[5],xmm11[5],xmm2[6],xmm11[6],xmm2[7],xmm11[7]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,1],xmm3[2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,0]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,0,1,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,4,6,7]
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pand %xmm4, %xmm2
-; SSE-NEXT:    por %xmm6, %xmm2
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm6 = mem[3,1,2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,1,2,1,4,5,6,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm10 = xmm10[0],xmm6[0],xmm10[1],xmm6[1]
-; SSE-NEXT:    packuswb %xmm1, %xmm10
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm10[2,1]
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm14, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm13[8],xmm1[9],xmm13[9],xmm1[10],xmm13[10],xmm1[11],xmm13[11],xmm1[12],xmm13[12],xmm1[13],xmm13[13],xmm1[14],xmm13[14],xmm1[15],xmm13[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm13[0],xmm14[1],xmm13[1],xmm14[2],xmm13[2],xmm14[3],xmm13[3],xmm14[4],xmm13[4],xmm14[5],xmm13[5],xmm14[6],xmm13[6],xmm14[7],xmm13[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm14[0,1,1,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,3],xmm1[1,2]
-; SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm5, %xmm0
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT:    por %xmm0, %xmm5
-; SSE-NEXT:    movdqa %xmm5, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1],xmm1[2],xmm13[2],xmm1[3],xmm13[3],xmm1[4],xmm13[4],xmm1[5],xmm13[5],xmm1[6],xmm13[6],xmm1[7],xmm13[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm13[8],xmm5[9],xmm13[9],xmm5[10],xmm13[10],xmm5[11],xmm13[11],xmm5[12],xmm13[12],xmm5[13],xmm13[13],xmm5[14],xmm13[14],xmm5[15],xmm13[15]
-; SSE-NEXT:    pand %xmm3, %xmm5
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    por %xmm5, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm3[3,1,0,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    por %xmm14, %xmm0
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm11[0],xmm2[1],xmm11[1],xmm2[2],xmm11[2],xmm2[3],xmm11[3],xmm2[4],xmm11[4],xmm2[5],xmm11[5],xmm2[6],xmm11[6],xmm2[7],xmm11[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm11[8],xmm0[9],xmm11[9],xmm0[10],xmm11[10],xmm0[11],xmm11[11],xmm0[12],xmm11[12],xmm0[13],xmm11[13],xmm0[14],xmm11[14],xmm0[15],xmm11[15]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm2[2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,0,1,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm2[0,1,2,3,4,5,5,6]
+; SSE-NEXT:    packuswb %xmm0, %xmm5
+; SSE-NEXT:    pand %xmm1, %xmm5
+; SSE-NEXT:    por %xmm3, %xmm5
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm14, %xmm0
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm2[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm2[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,7,5]
+; SSE-NEXT:    packuswb %xmm0, %xmm0
+; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm5
+; SSE-NEXT:    por %xmm5, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    por %xmm0, %xmm6
+; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    pxor %xmm3, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm3[8],xmm0[9],xmm3[9],xmm0[10],xmm3[10],xmm0[11],xmm3[11],xmm0[12],xmm3[12],xmm0[13],xmm3[13],xmm0[14],xmm3[14],xmm0[15],xmm3[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3],xmm6[4],xmm3[4],xmm6[5],xmm3[5],xmm6[6],xmm3[6],xmm6[7],xmm3[7]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[3,1],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    por %xmm12, %xmm6
+; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm3[8],xmm6[9],xmm3[9],xmm6[10],xmm3[10],xmm6[11],xmm3[11],xmm6[12],xmm3[12],xmm6[13],xmm3[13],xmm6[14],xmm3[14],xmm6[15],xmm3[15]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[3,1],xmm0[2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm6[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,0,1,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,5,6]
+; SSE-NEXT:    packuswb %xmm3, %xmm0
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm5[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[3,1,2,0]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,0,1,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,4,6,7]
+; SSE-NEXT:    packuswb %xmm3, %xmm3
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    pandn %xmm3, %xmm1
+; SSE-NEXT:    por %xmm0, %xmm1
+; SSE-NEXT:    movdqa %xmm10, %xmm6
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm4[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm0[0,2]
+; SSE-NEXT:    pand %xmm9, %xmm1
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,7,5]
+; SSE-NEXT:    packuswb %xmm0, %xmm0
+; SSE-NEXT:    pandn %xmm0, %xmm9
+; SSE-NEXT:    por %xmm1, %xmm9
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm11[8],xmm0[9],xmm11[9],xmm0[10],xmm11[10],xmm0[11],xmm11[11],xmm0[12],xmm11[12],xmm0[13],xmm11[13],xmm0[14],xmm11[14],xmm0[15],xmm11[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm11[0],xmm3[1],xmm11[1],xmm3[2],xmm11[2],xmm3[3],xmm11[3],xmm3[4],xmm11[4],xmm3[5],xmm11[5],xmm3[6],xmm11[6],xmm3[7],xmm11[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,1,1,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,3],xmm0[1,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2,3,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm0[2,1,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    packuswb %xmm3, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm0, %xmm4
+; SSE-NEXT:    pandn %xmm1, %xmm4
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255]
+; SSE-NEXT:    pand %xmm10, %xmm15
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1],xmm1[2],xmm11[2],xmm1[3],xmm11[3],xmm1[4],xmm11[4],xmm1[5],xmm11[5],xmm1[6],xmm11[6],xmm1[7],xmm11[7]
+; SSE-NEXT:    movdqa %xmm8, %xmm5
+; SSE-NEXT:    pandn %xmm1, %xmm5
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm15 = xmm15[8],xmm11[8],xmm15[9],xmm11[9],xmm15[10],xmm11[10],xmm15[11],xmm11[11],xmm15[12],xmm11[12],xmm15[13],xmm11[13],xmm15[14],xmm11[14],xmm15[15],xmm11[15]
+; SSE-NEXT:    pand %xmm8, %xmm15
+; SSE-NEXT:    por %xmm5, %xmm15
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm15[3,1,0,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,4]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm4, %xmm1
-; SSE-NEXT:    movdqa %xmm4, %xmm7
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,2,3,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm14[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[2,1,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm6[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    packuswb %xmm3, %xmm4
-; SSE-NEXT:    pandn %xmm4, %xmm7
-; SSE-NEXT:    por %xmm7, %xmm1
+; SSE-NEXT:    pand %xmm0, %xmm1
+; SSE-NEXT:    por %xmm4, %xmm1
 ; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm15[0,2,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm14[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,2,1,4,5,6,7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,1,0,3,4,5,6,7]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
 ; SSE-NEXT:    packuswb %xmm3, %xmm5
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 16(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, (%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 16(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, (%rdx)
-; SSE-NEXT:    movdqa %xmm12, 16(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, (%rcx)
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm13, %xmm4
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm11[8],xmm4[9],xmm11[9],xmm4[10],xmm11[10],xmm4[11],xmm11[11],xmm4[12],xmm11[12],xmm4[13],xmm11[13],xmm4[14],xmm11[14],xmm4[15],xmm11[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm11[0],xmm13[1],xmm11[1],xmm13[2],xmm11[2],xmm13[3],xmm11[3],xmm13[4],xmm11[4],xmm13[5],xmm11[5],xmm13[6],xmm11[6],xmm13[7],xmm11[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[0,1,1,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,3],xmm4[1,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm10, %xmm5
+; SSE-NEXT:    pand %xmm10, %xmm4
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    por %xmm4, %xmm5
+; SSE-NEXT:    movdqa %xmm5, %xmm4
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm11[0],xmm4[1],xmm11[1],xmm4[2],xmm11[2],xmm4[3],xmm11[3],xmm4[4],xmm11[4],xmm4[5],xmm11[5],xmm4[6],xmm11[6],xmm4[7],xmm11[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm11[8],xmm5[9],xmm11[9],xmm5[10],xmm11[10],xmm5[11],xmm11[11],xmm5[12],xmm11[12],xmm5[13],xmm11[13],xmm5[14],xmm11[14],xmm5[15],xmm11[15]
+; SSE-NEXT:    pand %xmm8, %xmm5
+; SSE-NEXT:    pandn %xmm4, %xmm8
+; SSE-NEXT:    por %xmm5, %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,2,3,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm13[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[2,1,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    packuswb %xmm4, %xmm3
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm8[3,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,6,4]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[3,1,2,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    packuswb %xmm5, %xmm5
+; SSE-NEXT:    pand %xmm0, %xmm5
+; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    por %xmm5, %xmm0
+; SSE-NEXT:    pshufd $231, (%rsp), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm6[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,2,1,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
+; SSE-NEXT:    packuswb %xmm4, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm5[2,1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, 16(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, (%rdx)
+; SSE-NEXT:    movdqa %xmm7, 16(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, (%rcx)
 ; SSE-NEXT:    movdqa %xmm9, 16(%r8)
-; SSE-NEXT:    movdqa %xmm8, (%r8)
-; SSE-NEXT:    movaps %xmm1, 16(%r9)
-; SSE-NEXT:    movaps %xmm2, (%r9)
+; SSE-NEXT:    movdqa %xmm2, (%r8)
+; SSE-NEXT:    movaps %xmm0, 16(%r9)
+; SSE-NEXT:    movaps %xmm1, (%r9)
 ; SSE-NEXT:    addq $184, %rsp
 ; SSE-NEXT:    retq
 ;
@@ -4067,1131 +4053,1140 @@ define void @load_i8_stride5_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i8_stride5_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4) nounwind {
 ; SSE-LABEL: load_i8_stride5_vf64:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $552, %rsp # imm = 0x228
-; SSE-NEXT:    movdqa 160(%rdi), %xmm9
-; SSE-NEXT:    movdqa 176(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    subq $568, %rsp # imm = 0x238
+; SSE-NEXT:    movdqa 160(%rdi), %xmm10
+; SSE-NEXT:    movdqa 176(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 208(%rdi), %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 192(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255]
-; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255]
+; SSE-NEXT:    movdqa %xmm14, %xmm0
 ; SSE-NEXT:    pandn %xmm1, %xmm0
 ; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    pand %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm2, %xmm14
+; SSE-NEXT:    pand %xmm14, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pxor %xmm12, %xmm12
+; SSE-NEXT:    pxor %xmm9, %xmm9
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1],xmm0[2],xmm12[2],xmm0[3],xmm12[3],xmm0[4],xmm12[4],xmm0[5],xmm12[5],xmm0[6],xmm12[6],xmm0[7],xmm12[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,1,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm12[8],xmm1[9],xmm12[9],xmm1[10],xmm12[10],xmm1[11],xmm12[11],xmm1[12],xmm12[12],xmm1[13],xmm12[13],xmm1[14],xmm12[14],xmm1[15],xmm12[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[0,1,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm9[8],xmm1[9],xmm9[9],xmm1[10],xmm9[10],xmm1[11],xmm9[11],xmm1[12],xmm9[12],xmm1[13],xmm9[13],xmm1[14],xmm9[14],xmm1[15],xmm9[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm3[2],xmm1[3],xmm3[3]
 ; SSE-NEXT:    packuswb %xmm1, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [255,255,255,255,255,255,255,0,0,0,0,0,0,255,255,255]
-; SSE-NEXT:    movdqa %xmm11, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [255,255,255,255,255,255,255,0,0,0,0,0,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm15, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255]
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm2, %xmm3
-; SSE-NEXT:    movdqa %xmm2, %xmm4
-; SSE-NEXT:    pandn %xmm9, %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255]
+; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    pandn %xmm2, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    movdqa %xmm4, %xmm6
+; SSE-NEXT:    pandn %xmm10, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255]
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm3, %xmm4
+; SSE-NEXT:    movdqa %xmm3, %xmm5
+; SSE-NEXT:    pandn %xmm10, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm14, %xmm3
+; SSE-NEXT:    pandn %xmm10, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm7, %xmm2
 ; SSE-NEXT:    movdqa %xmm7, %xmm3
-; SSE-NEXT:    pandn %xmm9, %xmm3
+; SSE-NEXT:    pandn %xmm10, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm2
-; SSE-NEXT:    pandn %xmm9, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm2
-; SSE-NEXT:    pandn %xmm9, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm10, %xmm9
-; SSE-NEXT:    por %xmm0, %xmm9
-; SSE-NEXT:    movdqa %xmm9, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1],xmm0[2],xmm12[2],xmm0[3],xmm12[3],xmm0[4],xmm12[4],xmm0[5],xmm12[5],xmm0[6],xmm12[6],xmm0[7],xmm12[7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [0,65535,65535,65535,0,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm8, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm12[8],xmm9[9],xmm12[9],xmm9[10],xmm12[10],xmm9[11],xmm12[11],xmm9[12],xmm12[12],xmm9[13],xmm12[13],xmm9[14],xmm12[14],xmm9[15],xmm12[15]
-; SSE-NEXT:    pand %xmm8, %xmm9
-; SSE-NEXT:    por %xmm2, %xmm9
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm9[0,2,1,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm7, %xmm10
+; SSE-NEXT:    por %xmm0, %xmm10
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [0,65535,65535,65535,0,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm7, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm9[8],xmm10[9],xmm9[9],xmm10[10],xmm9[10],xmm10[11],xmm9[11],xmm10[12],xmm9[12],xmm10[13],xmm9[13],xmm10[14],xmm9[14],xmm10[15],xmm9[15]
+; SSE-NEXT:    pand %xmm7, %xmm10
+; SSE-NEXT:    por %xmm3, %xmm10
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm10[0,2,1,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,5,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm11, %xmm0
+; SSE-NEXT:    pand %xmm15, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa 224(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm1[8],xmm3[9],xmm1[9],xmm3[10],xmm1[10],xmm3[11],xmm1[11],xmm3[12],xmm1[12],xmm3[13],xmm1[13],xmm3[14],xmm1[14],xmm3[15],xmm1[15]
+; SSE-NEXT:    movdqa 224(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1],xmm3[2],xmm9[2],xmm3[3],xmm9[3],xmm3[4],xmm9[4],xmm3[5],xmm9[5],xmm3[6],xmm9[6],xmm3[7],xmm9[7]
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pxor %xmm9, %xmm9
-; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm2[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[2,3]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm9[8],xmm1[9],xmm9[9],xmm1[10],xmm9[10],xmm1[11],xmm9[11],xmm1[12],xmm9[12],xmm1[13],xmm9[13],xmm1[14],xmm9[14],xmm1[15],xmm9[15]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm3[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm3[2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,4]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0]
-; SSE-NEXT:    movdqa %xmm6, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pand %xmm6, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0]
+; SSE-NEXT:    movdqa %xmm8, %xmm3
+; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    pand %xmm8, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 32(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm14, %xmm0
 ; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa 48(%rdi), %xmm15
-; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    movdqa 48(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, (%rsp) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm14, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,1,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[0,1,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm9[8],xmm1[9],xmm9[9],xmm1[10],xmm9[10],xmm1[11],xmm9[11],xmm1[12],xmm9[12],xmm1[13],xmm9[13],xmm1[14],xmm9[14],xmm1[15],xmm9[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm3[2],xmm1[3],xmm3[3]
 ; SSE-NEXT:    packuswb %xmm1, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,3]
-; SSE-NEXT:    movdqa %xmm11, %xmm1
+; SSE-NEXT:    movdqa %xmm15, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    movdqa 16(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    movdqa (%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm4
-; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm4
-; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm4
-; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm10, %xmm3
-; SSE-NEXT:    por %xmm2, %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm9[0],xmm2[1],xmm9[1],xmm2[2],xmm9[2],xmm2[3],xmm9[3],xmm2[4],xmm9[4],xmm2[5],xmm9[5],xmm2[6],xmm9[6],xmm2[7],xmm9[7]
-; SSE-NEXT:    movdqa %xmm8, %xmm4
-; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm9[8],xmm3[9],xmm9[9],xmm3[10],xmm9[10],xmm3[11],xmm9[11],xmm3[12],xmm9[12],xmm3[13],xmm9[13],xmm3[14],xmm9[14],xmm3[15],xmm9[15]
-; SSE-NEXT:    pand %xmm8, %xmm3
-; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm3[0,2,1,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,5,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,2,1,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,5,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pand %xmm11, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    movdqa 64(%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    movdqa (%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    pandn %xmm4, %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm5, %xmm6
+; SSE-NEXT:    pandn %xmm4, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm14, %xmm5
+; SSE-NEXT:    pandn %xmm4, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm2, %xmm5
+; SSE-NEXT:    pandn %xmm4, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm2, %xmm4
+; SSE-NEXT:    por %xmm3, %xmm4
+; SSE-NEXT:    movdqa %xmm4, %xmm3
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1],xmm3[2],xmm9[2],xmm3[3],xmm9[3],xmm3[4],xmm9[4],xmm3[5],xmm9[5],xmm3[6],xmm9[6],xmm3[7],xmm9[7]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm7, %xmm5
+; SSE-NEXT:    pandn %xmm3, %xmm5
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm9[8],xmm4[9],xmm9[9],xmm4[10],xmm9[10],xmm4[11],xmm9[11],xmm4[12],xmm9[12],xmm4[13],xmm9[13],xmm4[14],xmm9[14],xmm4[15],xmm9[15]
+; SSE-NEXT:    pand %xmm7, %xmm4
+; SSE-NEXT:    por %xmm5, %xmm4
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm4[0,2,1,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,6,5,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,2,1,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,6,5,7]
+; SSE-NEXT:    packuswb %xmm3, %xmm3
+; SSE-NEXT:    pand %xmm15, %xmm3
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    movdqa 64(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, %xmm4
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3],xmm4[4],xmm9[4],xmm4[5],xmm9[5],xmm4[6],xmm9[6],xmm4[7],xmm9[7]
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm9[8],xmm1[9],xmm9[9],xmm1[10],xmm9[10],xmm1[11],xmm9[11],xmm1[12],xmm9[12],xmm1[13],xmm9[13],xmm1[14],xmm9[14],xmm1[15],xmm9[15]
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm3[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm3[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm4[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm4[2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,4]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm6, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    pand %xmm6, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 272(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm1
-; SSE-NEXT:    pandn %xmm2, %xmm1
-; SSE-NEXT:    movdqa 288(%rdi), %xmm13
-; SSE-NEXT:    movdqa %xmm13, %xmm2
-; SSE-NEXT:    pand %xmm14, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1],xmm1[2],xmm9[2],xmm1[3],xmm9[3],xmm1[4],xmm9[4],xmm1[5],xmm9[5],xmm1[6],xmm9[6],xmm1[7],xmm9[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,1,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm9[8],xmm2[9],xmm9[9],xmm2[10],xmm9[10],xmm2[11],xmm9[11],xmm2[12],xmm9[12],xmm2[13],xmm9[13],xmm2[14],xmm9[14],xmm2[15],xmm9[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
-; SSE-NEXT:    packuswb %xmm2, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,2,3]
-; SSE-NEXT:    movdqa %xmm11, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    movdqa 256(%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm4
+; SSE-NEXT:    movdqa %xmm8, %xmm4
 ; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    movdqa 240(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    pandn %xmm3, %xmm1
-; SSE-NEXT:    pandn %xmm3, %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm7
-; SSE-NEXT:    pandn %xmm3, %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm7
-; SSE-NEXT:    pandn %xmm3, %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm10, %xmm3
-; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm4
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3],xmm4[4],xmm9[4],xmm4[5],xmm9[5],xmm4[6],xmm9[6],xmm4[7],xmm9[7]
-; SSE-NEXT:    movdqa %xmm8, %xmm7
-; SSE-NEXT:    pandn %xmm4, %xmm7
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm9[8],xmm3[9],xmm9[9],xmm3[10],xmm9[10],xmm3[11],xmm9[11],xmm3[12],xmm9[12],xmm3[13],xmm9[13],xmm3[14],xmm9[14],xmm3[15],xmm9[15]
 ; SSE-NEXT:    pand %xmm8, %xmm3
-; SSE-NEXT:    por %xmm7, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,2,1,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,6,5,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,2,1,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,6,5,7]
-; SSE-NEXT:    packuswb %xmm3, %xmm3
-; SSE-NEXT:    pand %xmm11, %xmm3
-; SSE-NEXT:    por %xmm2, %xmm3
-; SSE-NEXT:    movdqa 304(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm4
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3],xmm4[4],xmm9[4],xmm4[5],xmm9[5],xmm4[6],xmm9[6],xmm4[7],xmm9[7]
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm9[8],xmm2[9],xmm9[9],xmm2[10],xmm9[10],xmm2[11],xmm9[11],xmm2[12],xmm9[12],xmm2[13],xmm9[13],xmm2[14],xmm9[14],xmm2[15],xmm9[15]
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm4[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm4[2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,4]
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    movdqa %xmm6, %xmm4
-; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    pand %xmm6, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 112(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm2
-; SSE-NEXT:    pandn %xmm3, %xmm2
-; SSE-NEXT:    movdqa 128(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 272(%rdi), %xmm13
+; SSE-NEXT:    movdqa %xmm14, %xmm1
+; SSE-NEXT:    pandn %xmm13, %xmm1
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 288(%rdi), %xmm10
+; SSE-NEXT:    movdqa %xmm10, %xmm3
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm14, %xmm3
-; SSE-NEXT:    por %xmm2, %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm9[0],xmm2[1],xmm9[1],xmm2[2],xmm9[2],xmm2[3],xmm9[3],xmm2[4],xmm9[4],xmm2[5],xmm9[5],xmm2[6],xmm9[6],xmm2[7],xmm9[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[0,1,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1],xmm1[2],xmm9[2],xmm1[3],xmm9[3],xmm1[4],xmm9[4],xmm1[5],xmm9[5],xmm1[6],xmm9[6],xmm1[7],xmm9[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[0,1,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm9[8],xmm3[9],xmm9[9],xmm3[10],xmm9[10],xmm3[11],xmm9[11],xmm3[12],xmm9[12],xmm3[13],xmm9[13],xmm3[14],xmm9[14],xmm3[15],xmm9[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
-; SSE-NEXT:    packuswb %xmm3, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,2,3]
-; SSE-NEXT:    movdqa %xmm11, %xmm3
-; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    movdqa 96(%rdi), %xmm4
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm5[2],xmm3[3],xmm5[3]
+; SSE-NEXT:    packuswb %xmm3, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,2,3]
+; SSE-NEXT:    movdqa %xmm15, %xmm3
+; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    movdqa 256(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm2, %xmm5
+; SSE-NEXT:    pandn %xmm1, %xmm5
+; SSE-NEXT:    movdqa 240(%rdi), %xmm11
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    pandn %xmm11, %xmm1
+; SSE-NEXT:    movdqa %xmm6, %xmm4
+; SSE-NEXT:    pandn %xmm11, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm2
-; SSE-NEXT:    pandn %xmm4, %xmm2
-; SSE-NEXT:    movdqa 80(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm14, %xmm4
+; SSE-NEXT:    pandn %xmm11, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm10, %xmm4
-; SSE-NEXT:    por %xmm2, %xmm4
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm9[0],xmm2[1],xmm9[1],xmm2[2],xmm9[2],xmm2[3],xmm9[3],xmm2[4],xmm9[4],xmm2[5],xmm9[5],xmm2[6],xmm9[6],xmm2[7],xmm9[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm9[8],xmm4[9],xmm9[9],xmm4[10],xmm9[10],xmm4[11],xmm9[11],xmm4[12],xmm9[12],xmm4[13],xmm9[13],xmm4[14],xmm9[14],xmm4[15],xmm9[15]
-; SSE-NEXT:    pand %xmm8, %xmm4
-; SSE-NEXT:    pandn %xmm2, %xmm8
-; SSE-NEXT:    por %xmm4, %xmm8
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm8[0,2,1,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,5,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,2,1,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,5,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pand %xmm11, %xmm2
-; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    movdqa 144(%rdi), %xmm12
-; SSE-NEXT:    movdqa %xmm12, %xmm4
+; SSE-NEXT:    movdqa %xmm2, %xmm4
+; SSE-NEXT:    pandn %xmm11, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm2, %xmm11
+; SSE-NEXT:    por %xmm5, %xmm11
+; SSE-NEXT:    movdqa %xmm11, %xmm5
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm9[0],xmm5[1],xmm9[1],xmm5[2],xmm9[2],xmm5[3],xmm9[3],xmm5[4],xmm9[4],xmm5[5],xmm9[5],xmm5[6],xmm9[6],xmm5[7],xmm9[7]
+; SSE-NEXT:    movdqa %xmm7, %xmm6
+; SSE-NEXT:    pandn %xmm5, %xmm6
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm9[8],xmm11[9],xmm9[9],xmm11[10],xmm9[10],xmm11[11],xmm9[11],xmm11[12],xmm9[12],xmm11[13],xmm9[13],xmm11[14],xmm9[14],xmm11[15],xmm9[15]
+; SSE-NEXT:    pand %xmm7, %xmm11
+; SSE-NEXT:    por %xmm6, %xmm11
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm11[0,2,1,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,6,5,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,2,1,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,6,5,7]
+; SSE-NEXT:    packuswb %xmm5, %xmm5
+; SSE-NEXT:    pand %xmm15, %xmm5
+; SSE-NEXT:    por %xmm3, %xmm5
+; SSE-NEXT:    movdqa 304(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, %xmm4
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3],xmm4[4],xmm9[4],xmm4[5],xmm9[5],xmm4[6],xmm9[6],xmm4[7],xmm9[7]
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm9[8],xmm12[9],xmm9[9],xmm12[10],xmm9[10],xmm12[11],xmm9[11],xmm12[12],xmm9[12],xmm12[13],xmm9[13],xmm12[14],xmm9[14],xmm12[15],xmm9[15]
-; SSE-NEXT:    movdqa %xmm12, %xmm3
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm9[8],xmm3[9],xmm9[9],xmm3[10],xmm9[10],xmm3[11],xmm9[11],xmm3[12],xmm9[12],xmm3[13],xmm9[13],xmm3[14],xmm9[14],xmm3[15],xmm9[15]
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,0],xmm4[0,0]
 ; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm4[2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,4]
 ; SSE-NEXT:    packuswb %xmm3, %xmm3
-; SSE-NEXT:    movdqa %xmm6, %xmm14
-; SSE-NEXT:    movdqa %xmm6, %xmm4
+; SSE-NEXT:    movdqa %xmm8, %xmm4
 ; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    pand %xmm6, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm4
+; SSE-NEXT:    pand %xmm8, %xmm5
+; SSE-NEXT:    por %xmm5, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm2
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pand %xmm10, %xmm3
-; SSE-NEXT:    por %xmm2, %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm9[8],xmm2[9],xmm9[9],xmm2[10],xmm9[10],xmm2[11],xmm9[11],xmm2[12],xmm9[12],xmm2[13],xmm9[13],xmm2[14],xmm9[14],xmm2[15],xmm9[15]
+; SSE-NEXT:    movdqa 112(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm14, %xmm3
+; SSE-NEXT:    pandn %xmm4, %xmm3
+; SSE-NEXT:    movdqa 128(%rdi), %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm14, %xmm5
+; SSE-NEXT:    por %xmm3, %xmm5
+; SSE-NEXT:    movdqa %xmm5, %xmm3
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1],xmm3[2],xmm9[2],xmm3[3],xmm9[3],xmm3[4],xmm9[4],xmm3[5],xmm9[5],xmm3[6],xmm9[6],xmm3[7],xmm9[7]
-; SSE-NEXT:    movdqa %xmm3, %xmm4
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,0],xmm2[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm2[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm4[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    psllq $48, %xmm3
-; SSE-NEXT:    packuswb %xmm2, %xmm3
-; SSE-NEXT:    movdqa %xmm11, %xmm4
-; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm3[0,1,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm9[8],xmm5[9],xmm9[9],xmm5[10],xmm9[10],xmm5[11],xmm9[11],xmm5[12],xmm9[12],xmm5[13],xmm9[13],xmm5[14],xmm9[14],xmm5[15],xmm9[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,4,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; SSE-NEXT:    packuswb %xmm5, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,0,2,3]
+; SSE-NEXT:    movdqa %xmm15, %xmm5
+; SSE-NEXT:    pandn %xmm3, %xmm5
+; SSE-NEXT:    movdqa 96(%rdi), %xmm14
+; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    pandn %xmm14, %xmm3
+; SSE-NEXT:    movdqa 80(%rdi), %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm2, %xmm6
+; SSE-NEXT:    por %xmm3, %xmm6
 ; SSE-NEXT:    movdqa %xmm6, %xmm3
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm3, %xmm7
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm9[8],xmm7[9],xmm9[9],xmm7[10],xmm9[10],xmm7[11],xmm9[11],xmm7[12],xmm9[12],xmm7[13],xmm9[13],xmm7[14],xmm9[14],xmm7[15],xmm9[15]
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,0,0,65535,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm2, %xmm8
-; SSE-NEXT:    pandn %xmm7, %xmm8
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1],xmm3[2],xmm9[2],xmm3[3],xmm9[3],xmm3[4],xmm9[4],xmm3[5],xmm9[5],xmm3[6],xmm9[6],xmm3[7],xmm9[7]
-; SSE-NEXT:    pand %xmm2, %xmm3
-; SSE-NEXT:    por %xmm8, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,5]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm9[8],xmm6[9],xmm9[9],xmm6[10],xmm9[10],xmm6[11],xmm9[11],xmm6[12],xmm9[12],xmm6[13],xmm9[13],xmm6[14],xmm9[14],xmm6[15],xmm9[15]
+; SSE-NEXT:    pand %xmm7, %xmm6
+; SSE-NEXT:    pandn %xmm3, %xmm7
+; SSE-NEXT:    por %xmm6, %xmm7
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm7[0,2,1,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,6,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,2,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,4,5,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,2,1,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,6,5,7]
 ; SSE-NEXT:    packuswb %xmm3, %xmm3
-; SSE-NEXT:    pand %xmm11, %xmm3
-; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm7[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm4[0,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm7[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,6,5,4]
-; SSE-NEXT:    packuswb %xmm4, %xmm4
-; SSE-NEXT:    movdqa %xmm14, %xmm7
-; SSE-NEXT:    pandn %xmm4, %xmm7
-; SSE-NEXT:    pand %xmm14, %xmm3
-; SSE-NEXT:    por %xmm3, %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm3
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm15, %xmm4
-; SSE-NEXT:    pand %xmm10, %xmm4
-; SSE-NEXT:    movdqa %xmm10, %xmm5
+; SSE-NEXT:    pand %xmm15, %xmm3
+; SSE-NEXT:    por %xmm5, %xmm3
+; SSE-NEXT:    movdqa 144(%rdi), %xmm5
+; SSE-NEXT:    movdqa %xmm5, %xmm4
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3],xmm4[4],xmm9[4],xmm4[5],xmm9[5],xmm4[6],xmm9[6],xmm4[7],xmm9[7]
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm9[8],xmm5[9],xmm9[9],xmm5[10],xmm9[10],xmm5[11],xmm9[11],xmm5[12],xmm9[12],xmm5[13],xmm9[13],xmm5[14],xmm9[14],xmm5[15],xmm9[15]
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,0],xmm4[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm4[2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,6,4]
+; SSE-NEXT:    packuswb %xmm5, %xmm5
+; SSE-NEXT:    movdqa %xmm8, %xmm4
+; SSE-NEXT:    pandn %xmm5, %xmm4
+; SSE-NEXT:    pand %xmm8, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm4
-; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    pand %xmm2, %xmm5
+; SSE-NEXT:    movdqa %xmm2, %xmm4
+; SSE-NEXT:    por %xmm3, %xmm5
+; SSE-NEXT:    movdqa %xmm5, %xmm3
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm9[8],xmm3[9],xmm9[9],xmm3[10],xmm9[10],xmm3[11],xmm9[11],xmm3[12],xmm9[12],xmm3[13],xmm9[13],xmm3[14],xmm9[14],xmm3[15],xmm9[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3],xmm4[4],xmm9[4],xmm4[5],xmm9[5],xmm4[6],xmm9[6],xmm4[7],xmm9[7]
-; SSE-NEXT:    movdqa %xmm4, %xmm7
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[1,0],xmm3[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm3[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm7[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm9[0],xmm5[1],xmm9[1],xmm5[2],xmm9[2],xmm5[3],xmm9[3],xmm5[4],xmm9[4],xmm5[5],xmm9[5],xmm5[6],xmm9[6],xmm5[7],xmm9[7]
+; SSE-NEXT:    movdqa %xmm5, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[1,0],xmm3[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0],xmm3[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm6[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    psllq $48, %xmm4
-; SSE-NEXT:    packuswb %xmm3, %xmm4
-; SSE-NEXT:    movdqa %xmm11, %xmm3
-; SSE-NEXT:    pandn %xmm4, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm10, %xmm4
-; SSE-NEXT:    movdqa %xmm0, %xmm8
-; SSE-NEXT:    pand %xmm0, %xmm4
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm9[8],xmm0[9],xmm9[9],xmm0[10],xmm9[10],xmm0[11],xmm9[11],xmm0[12],xmm9[12],xmm0[13],xmm9[13],xmm0[14],xmm9[14],xmm0[15],xmm9[15]
-; SSE-NEXT:    movdqa %xmm2, %xmm7
+; SSE-NEXT:    psllq $48, %xmm5
+; SSE-NEXT:    packuswb %xmm3, %xmm5
+; SSE-NEXT:    movdqa %xmm15, %xmm6
+; SSE-NEXT:    pandn %xmm5, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm2, %xmm5
+; SSE-NEXT:    pand %xmm0, %xmm5
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm5, %xmm7
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm9[8],xmm7[9],xmm9[9],xmm7[10],xmm9[10],xmm7[11],xmm9[11],xmm7[12],xmm9[12],xmm7[13],xmm9[13],xmm7[14],xmm9[14],xmm7[15],xmm9[15]
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,0,0,65535,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm3, %xmm11
+; SSE-NEXT:    pandn %xmm7, %xmm11
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm9[0],xmm5[1],xmm9[1],xmm5[2],xmm9[2],xmm5[3],xmm9[3],xmm5[4],xmm9[4],xmm5[5],xmm9[5],xmm5[6],xmm9[6],xmm5[7],xmm9[7]
+; SSE-NEXT:    pand %xmm3, %xmm5
+; SSE-NEXT:    por %xmm11, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,7,6,5]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,2,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,4,5,7]
+; SSE-NEXT:    packuswb %xmm5, %xmm5
+; SSE-NEXT:    pand %xmm15, %xmm5
+; SSE-NEXT:    por %xmm6, %xmm5
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0],xmm7[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm6[0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm7[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,6,5,4]
+; SSE-NEXT:    packuswb %xmm6, %xmm6
+; SSE-NEXT:    movdqa %xmm8, %xmm7
+; SSE-NEXT:    pandn %xmm6, %xmm7
+; SSE-NEXT:    pand %xmm8, %xmm5
+; SSE-NEXT:    por %xmm5, %xmm7
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm4, %xmm9
+; SSE-NEXT:    movdqa %xmm4, %xmm5
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa (%rsp), %xmm6 # 16-byte Reload
+; SSE-NEXT:    pand %xmm4, %xmm6
+; SSE-NEXT:    por %xmm5, %xmm6
+; SSE-NEXT:    movdqa %xmm6, %xmm5
+; SSE-NEXT:    pxor %xmm4, %xmm4
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm4[8],xmm5[9],xmm4[9],xmm5[10],xmm4[10],xmm5[11],xmm4[11],xmm5[12],xmm4[12],xmm5[13],xmm4[13],xmm5[14],xmm4[14],xmm5[15],xmm4[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1],xmm6[2],xmm4[2],xmm6[3],xmm4[3],xmm6[4],xmm4[4],xmm6[5],xmm4[5],xmm6[6],xmm4[6],xmm6[7],xmm4[7]
+; SSE-NEXT:    pxor %xmm12, %xmm12
+; SSE-NEXT:    movdqa %xmm6, %xmm7
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[1,0],xmm5[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm5[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm7[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,2,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    psllq $48, %xmm6
+; SSE-NEXT:    packuswb %xmm5, %xmm6
+; SSE-NEXT:    movdqa %xmm15, %xmm5
+; SSE-NEXT:    pandn %xmm6, %xmm5
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm4, %xmm6
+; SSE-NEXT:    movdqa %xmm0, %xmm11
+; SSE-NEXT:    pand %xmm0, %xmm6
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm12[8],xmm0[9],xmm12[9],xmm0[10],xmm12[10],xmm0[11],xmm12[11],xmm0[12],xmm12[12],xmm0[13],xmm12[13],xmm0[14],xmm12[14],xmm0[15],xmm12[15]
+; SSE-NEXT:    movdqa %xmm3, %xmm7
 ; SSE-NEXT:    pandn %xmm0, %xmm7
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3],xmm4[4],xmm9[4],xmm4[5],xmm9[5],xmm4[6],xmm9[6],xmm4[7],xmm9[7]
-; SSE-NEXT:    pand %xmm2, %xmm4
-; SSE-NEXT:    por %xmm7, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[0,2,1,3]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1],xmm6[2],xmm12[2],xmm6[3],xmm12[3],xmm6[4],xmm12[4],xmm6[5],xmm12[5],xmm6[6],xmm12[6],xmm6[7],xmm12[7]
+; SSE-NEXT:    pand %xmm3, %xmm6
+; SSE-NEXT:    por %xmm7, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[0,2,1,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,4,5,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    por %xmm3, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm4[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[0,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm4[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,6,5,4]
-; SSE-NEXT:    packuswb %xmm3, %xmm3
-; SSE-NEXT:    movdqa %xmm14, %xmm4
-; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    pand %xmm14, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm5, %xmm0
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm13, %xmm3
-; SSE-NEXT:    pand %xmm5, %xmm3
-; SSE-NEXT:    por %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm9[8],xmm0[9],xmm9[9],xmm0[10],xmm9[10],xmm0[11],xmm9[11],xmm0[12],xmm9[12],xmm0[13],xmm9[13],xmm0[14],xmm9[14],xmm0[15],xmm9[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1],xmm3[2],xmm9[2],xmm3[3],xmm9[3],xmm3[4],xmm9[4],xmm3[5],xmm9[5],xmm3[6],xmm9[6],xmm3[7],xmm9[7]
-; SSE-NEXT:    movdqa %xmm3, %xmm4
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,0],xmm0[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm0[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    por %xmm5, %xmm0
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm6[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm5[0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm6[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,6,5,4]
+; SSE-NEXT:    packuswb %xmm5, %xmm5
+; SSE-NEXT:    movdqa %xmm8, %xmm6
+; SSE-NEXT:    pandn %xmm5, %xmm6
+; SSE-NEXT:    pand %xmm8, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    pandn %xmm13, %xmm0
+; SSE-NEXT:    pand %xmm9, %xmm10
+; SSE-NEXT:    por %xmm0, %xmm10
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm12[8],xmm0[9],xmm12[9],xmm0[10],xmm12[10],xmm0[11],xmm12[11],xmm0[12],xmm12[12],xmm0[13],xmm12[13],xmm0[14],xmm12[14],xmm0[15],xmm12[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm12[0],xmm10[1],xmm12[1],xmm10[2],xmm12[2],xmm10[3],xmm12[3],xmm10[4],xmm12[4],xmm10[5],xmm12[5],xmm10[6],xmm12[6],xmm10[7],xmm12[7]
+; SSE-NEXT:    movdqa %xmm10, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[1,0],xmm0[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0],xmm0[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm6[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    psllq $48, %xmm3
-; SSE-NEXT:    packuswb %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    movdqa (%rsp), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pand %xmm8, %xmm3
-; SSE-NEXT:    movdqa %xmm8, %xmm7
-; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm9[8],xmm1[9],xmm9[9],xmm1[10],xmm9[10],xmm1[11],xmm9[11],xmm1[12],xmm9[12],xmm1[13],xmm9[13],xmm1[14],xmm9[14],xmm1[15],xmm9[15]
-; SSE-NEXT:    movdqa %xmm2, %xmm4
-; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1],xmm3[2],xmm9[2],xmm3[3],xmm9[3],xmm3[4],xmm9[4],xmm3[5],xmm9[5],xmm3[6],xmm9[6],xmm3[7],xmm9[7]
-; SSE-NEXT:    pand %xmm2, %xmm3
-; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,2,1,3]
+; SSE-NEXT:    psllq $48, %xmm10
+; SSE-NEXT:    packuswb %xmm0, %xmm10
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    pandn %xmm10, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm13, %xmm5
+; SSE-NEXT:    pand %xmm11, %xmm5
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm12[8],xmm1[9],xmm12[9],xmm1[10],xmm12[10],xmm1[11],xmm12[11],xmm1[12],xmm12[12],xmm1[13],xmm12[13],xmm1[14],xmm12[14],xmm1[15],xmm12[15]
+; SSE-NEXT:    movdqa %xmm3, %xmm6
+; SSE-NEXT:    pandn %xmm1, %xmm6
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm12[0],xmm5[1],xmm12[1],xmm5[2],xmm12[2],xmm5[3],xmm12[3],xmm5[4],xmm12[4],xmm5[5],xmm12[5],xmm5[6],xmm12[6],xmm5[7],xmm12[7]
+; SSE-NEXT:    pand %xmm3, %xmm5
+; SSE-NEXT:    por %xmm6, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[0,2,1,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,4,5,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm11, %xmm1
+; SSE-NEXT:    pand %xmm15, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm3[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[0,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm3[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm5[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm0[0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm5[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,5,4]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm14, %xmm3
-; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    pand %xmm14, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    movdqa %xmm8, %xmm5
+; SSE-NEXT:    pandn %xmm0, %xmm5
+; SSE-NEXT:    pand %xmm8, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm5, %xmm8
+; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm9, %xmm6
 ; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm9[8],xmm1[9],xmm9[9],xmm1[10],xmm9[10],xmm1[11],xmm9[11],xmm1[12],xmm9[12],xmm1[13],xmm9[13],xmm1[14],xmm9[14],xmm1[15],xmm9[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,0],xmm1[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm1[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm12[8],xmm1[9],xmm12[9],xmm1[10],xmm12[10],xmm1[11],xmm12[11],xmm1[12],xmm12[12],xmm1[13],xmm12[13],xmm1[14],xmm12[14],xmm1[15],xmm12[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1],xmm0[2],xmm12[2],xmm0[3],xmm12[3],xmm0[4],xmm12[4],xmm0[5],xmm12[5],xmm0[6],xmm12[6],xmm0[7],xmm12[7]
+; SSE-NEXT:    movdqa %xmm0, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,0],xmm1[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm1[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm5[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    psllq $48, %xmm0
 ; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm5, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pand %xmm7, %xmm3
-; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm9[8],xmm1[9],xmm9[9],xmm1[10],xmm9[10],xmm1[11],xmm9[11],xmm1[12],xmm9[12],xmm1[13],xmm9[13],xmm1[14],xmm9[14],xmm1[15],xmm9[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1],xmm3[2],xmm9[2],xmm3[3],xmm9[3],xmm3[4],xmm9[4],xmm3[5],xmm9[5],xmm3[6],xmm9[6],xmm3[7],xmm9[7]
-; SSE-NEXT:    pand %xmm2, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,1,3]
+; SSE-NEXT:    movdqa %xmm11, %xmm1
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm14, %xmm5
+; SSE-NEXT:    pand %xmm11, %xmm5
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm12[8],xmm1[9],xmm12[9],xmm1[10],xmm12[10],xmm1[11],xmm12[11],xmm1[12],xmm12[12],xmm1[13],xmm12[13],xmm1[14],xmm12[14],xmm1[15],xmm12[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm12[0],xmm5[1],xmm12[1],xmm5[2],xmm12[2],xmm5[3],xmm12[3],xmm5[4],xmm12[4],xmm5[5],xmm12[5],xmm5[6],xmm12[6],xmm5[7],xmm12[7]
+; SSE-NEXT:    pand %xmm3, %xmm5
+; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    por %xmm5, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,2,1,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,4,5,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm11, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm11
-; SSE-NEXT:    por %xmm11, %xmm1
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[2,0],xmm2[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm12[0,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm2[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pand %xmm15, %xmm1
+; SSE-NEXT:    pandn %xmm0, %xmm15
+; SSE-NEXT:    por %xmm1, %xmm15
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,5,4]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm14, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pand %xmm14, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    pandn %xmm0, %xmm1
+; SSE-NEXT:    pand %xmm8, %xmm15
+; SSE-NEXT:    por %xmm15, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255]
+; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm12[8],xmm1[9],xmm12[9],xmm1[10],xmm12[10],xmm1[11],xmm12[11],xmm1[12],xmm12[12],xmm1[13],xmm12[13],xmm1[14],xmm12[14],xmm1[15],xmm12[15]
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,65535,0,0,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm10, %xmm3
+; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1],xmm0[2],xmm12[2],xmm0[3],xmm12[3],xmm0[4],xmm12[4],xmm0[5],xmm12[5],xmm0[6],xmm12[6],xmm0[7],xmm12[7]
+; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    por %xmm3, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    packuswb %xmm1, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [0,0,0,65535,65535,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm6, %xmm2
 ; SSE-NEXT:    movdqa %xmm6, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [255,255,0,255,255,255,255,0,255,255,255,255,0,255,255,255]
-; SSE-NEXT:    pand %xmm11, %xmm1
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm9[8],xmm2[9],xmm9[9],xmm2[10],xmm9[10],xmm2[11],xmm9[11],xmm2[12],xmm9[12],xmm2[13],xmm9[13],xmm2[14],xmm9[14],xmm2[15],xmm9[15]
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,65535,0,0,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm6, %xmm3
-; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1],xmm1[2],xmm9[2],xmm1[3],xmm9[3],xmm1[4],xmm9[4],xmm1[5],xmm9[5],xmm1[6],xmm9[6],xmm1[7],xmm9[7]
-; SSE-NEXT:    pand %xmm6, %xmm1
-; SSE-NEXT:    por %xmm3, %xmm1
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    pandn %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm2, %xmm6
+; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    por %xmm1, %xmm6
+; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm12[8],xmm1[9],xmm12[9],xmm1[10],xmm12[10],xmm1[11],xmm12[11],xmm1[12],xmm12[12],xmm1[13],xmm12[13],xmm1[14],xmm12[14],xmm1[15],xmm12[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1],xmm6[2],xmm12[2],xmm6[3],xmm12[3],xmm6[4],xmm12[4],xmm6[5],xmm12[5],xmm6[6],xmm12[6],xmm6[7],xmm12[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm6[0,1,2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0],xmm1[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm6[0,2]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,6,6,6]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm7[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    packuswb %xmm1, %xmm6
+; SSE-NEXT:    pand %xmm9, %xmm6
+; SSE-NEXT:    por %xmm3, %xmm6
+; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
+; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm3[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
+; SSE-NEXT:    packuswb %xmm1, %xmm1
+; SSE-NEXT:    movdqa %xmm8, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    pand %xmm8, %xmm6
+; SSE-NEXT:    movdqa %xmm8, %xmm15
+; SSE-NEXT:    por %xmm6, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm5, %xmm4
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm12[8],xmm3[9],xmm12[9],xmm3[10],xmm12[10],xmm3[11],xmm12[11],xmm3[12],xmm12[12],xmm3[13],xmm12[13],xmm3[14],xmm12[14],xmm3[15],xmm12[15]
+; SSE-NEXT:    movdqa %xmm10, %xmm6
+; SSE-NEXT:    pandn %xmm3, %xmm6
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm12[0],xmm4[1],xmm12[1],xmm4[2],xmm12[2],xmm4[3],xmm12[3],xmm4[4],xmm12[4],xmm4[5],xmm12[5],xmm4[6],xmm12[6],xmm4[7],xmm12[7]
+; SSE-NEXT:    pand %xmm10, %xmm4
+; SSE-NEXT:    por %xmm6, %xmm4
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm4[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [0,0,0,65535,65535,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm12, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm8, %xmm1
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm4, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm3
-; SSE-NEXT:    pandn %xmm4, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm8, %xmm4
-; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm9[8],xmm1[9],xmm9[9],xmm1[10],xmm9[10],xmm1[11],xmm9[11],xmm1[12],xmm9[12],xmm1[13],xmm9[13],xmm1[14],xmm9[14],xmm1[15],xmm9[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3],xmm4[4],xmm9[4],xmm4[5],xmm9[5],xmm4[6],xmm9[6],xmm4[7],xmm9[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm4[0,1,2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm1[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm4[0,2]
+; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    movdqa (%rsp), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm11, %xmm8
+; SSE-NEXT:    movdqa %xmm11, %xmm4
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm6, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm5, %xmm4
+; SSE-NEXT:    pandn %xmm6, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm6
+; SSE-NEXT:    movdqa %xmm0, %xmm4
+; SSE-NEXT:    por %xmm1, %xmm6
+; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm12[8],xmm1[9],xmm12[9],xmm1[10],xmm12[10],xmm1[11],xmm12[11],xmm1[12],xmm12[12],xmm1[13],xmm12[13],xmm1[14],xmm12[14],xmm1[15],xmm12[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1],xmm6[2],xmm12[2],xmm6[3],xmm12[3],xmm6[4],xmm12[4],xmm6[5],xmm12[5],xmm6[6],xmm12[6],xmm6[7],xmm12[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm6[0,1,2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0],xmm1[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm6[0,2]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,6,6,6]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm7[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    packuswb %xmm1, %xmm4
-; SSE-NEXT:    pand %xmm12, %xmm4
-; SSE-NEXT:    por %xmm2, %xmm4
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm7[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    packuswb %xmm1, %xmm6
+; SSE-NEXT:    pand %xmm9, %xmm6
+; SSE-NEXT:    por %xmm3, %xmm6
 ; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm3[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm14, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    pand %xmm14, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm11, %xmm10
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm10, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm9[8],xmm2[9],xmm9[9],xmm2[10],xmm9[10],xmm2[11],xmm9[11],xmm2[12],xmm9[12],xmm2[13],xmm9[13],xmm2[14],xmm9[14],xmm2[15],xmm9[15]
-; SSE-NEXT:    movdqa %xmm6, %xmm4
-; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3],xmm10[4],xmm9[4],xmm10[5],xmm9[5],xmm10[6],xmm9[6],xmm10[7],xmm9[7]
-; SSE-NEXT:    pand %xmm6, %xmm10
-; SSE-NEXT:    por %xmm4, %xmm10
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm10[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm15, %xmm11
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    pandn %xmm1, %xmm0
+; SSE-NEXT:    pand %xmm15, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm5, %xmm13
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm13, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm12[8],xmm3[9],xmm12[9],xmm3[10],xmm12[10],xmm3[11],xmm12[11],xmm3[12],xmm12[12],xmm3[13],xmm12[13],xmm3[14],xmm12[14],xmm3[15],xmm12[15]
+; SSE-NEXT:    movdqa %xmm10, %xmm6
+; SSE-NEXT:    pandn %xmm3, %xmm6
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm12[0],xmm13[1],xmm12[1],xmm13[2],xmm12[2],xmm13[3],xmm12[3],xmm13[4],xmm12[4],xmm13[5],xmm12[5],xmm13[6],xmm12[6],xmm13[7],xmm12[7]
+; SSE-NEXT:    pand %xmm10, %xmm13
+; SSE-NEXT:    por %xmm6, %xmm13
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm13[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm12, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm4, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
 ; SSE-NEXT:    pandn %xmm15, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm4, %xmm0
+; SSE-NEXT:    movdqa %xmm8, %xmm0
+; SSE-NEXT:    movdqa %xmm8, %xmm13
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm6, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm3
-; SSE-NEXT:    pandn %xmm4, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm8, %xmm4
-; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm9[8],xmm1[9],xmm9[9],xmm1[10],xmm9[10],xmm1[11],xmm9[11],xmm1[12],xmm9[12],xmm1[13],xmm9[13],xmm1[14],xmm9[14],xmm1[15],xmm9[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3],xmm4[4],xmm9[4],xmm4[5],xmm9[5],xmm4[6],xmm9[6],xmm4[7],xmm9[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm4[0,1,2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm1[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm4[0,2]
+; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    pandn %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm4, %xmm6
+; SSE-NEXT:    movdqa %xmm4, %xmm8
+; SSE-NEXT:    por %xmm1, %xmm6
+; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm12[8],xmm1[9],xmm12[9],xmm1[10],xmm12[10],xmm1[11],xmm12[11],xmm1[12],xmm12[12],xmm1[13],xmm12[13],xmm1[14],xmm12[14],xmm1[15],xmm12[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1],xmm6[2],xmm12[2],xmm6[3],xmm12[3],xmm6[4],xmm12[4],xmm6[5],xmm12[5],xmm6[6],xmm12[6],xmm6[7],xmm12[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm6[0,1,2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0],xmm1[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm6[0,2]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,6,6,6]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm7[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    packuswb %xmm1, %xmm4
-; SSE-NEXT:    pand %xmm12, %xmm4
-; SSE-NEXT:    por %xmm2, %xmm4
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm7[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    packuswb %xmm1, %xmm6
+; SSE-NEXT:    pand %xmm9, %xmm6
+; SSE-NEXT:    por %xmm3, %xmm6
 ; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm3[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm14, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pand %xmm14, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rsp), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pand %xmm11, %xmm1
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm9[8],xmm2[9],xmm9[9],xmm2[10],xmm9[10],xmm2[11],xmm9[11],xmm2[12],xmm9[12],xmm2[13],xmm9[13],xmm2[14],xmm9[14],xmm2[15],xmm9[15]
-; SSE-NEXT:    movdqa %xmm6, %xmm4
-; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1],xmm1[2],xmm9[2],xmm1[3],xmm9[3],xmm1[4],xmm9[4],xmm1[5],xmm9[5],xmm1[6],xmm9[6],xmm1[7],xmm9[7]
-; SSE-NEXT:    pand %xmm6, %xmm1
-; SSE-NEXT:    por %xmm4, %xmm1
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    pandn %xmm1, %xmm0
+; SSE-NEXT:    pand %xmm11, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm5, %xmm7
+; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    pand %xmm5, %xmm14
+; SSE-NEXT:    por %xmm1, %xmm14
+; SSE-NEXT:    movdqa %xmm14, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm12[8],xmm1[9],xmm12[9],xmm1[10],xmm12[10],xmm1[11],xmm12[11],xmm1[12],xmm12[12],xmm1[13],xmm12[13],xmm1[14],xmm12[14],xmm1[15],xmm12[15]
+; SSE-NEXT:    movdqa %xmm10, %xmm6
+; SSE-NEXT:    pandn %xmm1, %xmm6
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm12[0],xmm14[1],xmm12[1],xmm14[2],xmm12[2],xmm14[3],xmm12[3],xmm14[4],xmm12[4],xmm14[5],xmm12[5],xmm14[6],xmm12[6],xmm14[7],xmm12[7]
+; SSE-NEXT:    pand %xmm10, %xmm14
+; SSE-NEXT:    por %xmm6, %xmm14
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm14[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm12, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm8, %xmm4
-; SSE-NEXT:    pandn %xmm13, %xmm4
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm7, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    pandn %xmm7, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm8, %xmm7
-; SSE-NEXT:    movdqa %xmm8, %xmm10
-; SSE-NEXT:    por %xmm4, %xmm7
-; SSE-NEXT:    movdqa %xmm7, %xmm4
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm9[8],xmm4[9],xmm9[9],xmm4[10],xmm9[10],xmm4[11],xmm9[11],xmm4[12],xmm9[12],xmm4[13],xmm9[13],xmm4[14],xmm9[14],xmm4[15],xmm9[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1],xmm7[2],xmm9[2],xmm7[3],xmm9[3],xmm7[4],xmm9[4],xmm7[5],xmm9[5],xmm7[6],xmm9[6],xmm7[7],xmm9[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm7[0,1,2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm4[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm7[0,2]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,6,6,6]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm8[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    packuswb %xmm4, %xmm7
-; SSE-NEXT:    pand %xmm12, %xmm7
-; SSE-NEXT:    por %xmm2, %xmm7
-; SSE-NEXT:    pshufd $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[1,1,1,1]
-; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm4[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,5]
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    movdqa %xmm14, %xmm1
-; SSE-NEXT:    pandn %xmm2, %xmm1
-; SSE-NEXT:    pand %xmm14, %xmm7
-; SSE-NEXT:    por %xmm7, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm8
-; SSE-NEXT:    movdqa %xmm11, %xmm2
-; SSE-NEXT:    pandn %xmm5, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pand %xmm11, %xmm4
-; SSE-NEXT:    por %xmm2, %xmm4
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm9[8],xmm2[9],xmm9[9],xmm2[10],xmm9[10],xmm2[11],xmm9[11],xmm2[12],xmm9[12],xmm2[13],xmm9[13],xmm2[14],xmm9[14],xmm2[15],xmm9[15]
-; SSE-NEXT:    movdqa %xmm6, %xmm7
-; SSE-NEXT:    pandn %xmm2, %xmm7
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3],xmm4[4],xmm9[4],xmm4[5],xmm9[5],xmm4[6],xmm9[6],xmm4[7],xmm9[7]
-; SSE-NEXT:    pand %xmm6, %xmm4
-; SSE-NEXT:    por %xmm7, %xmm4
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm4[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm2[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    packuswb %xmm4, %xmm4
-; SSE-NEXT:    movdqa %xmm12, %xmm3
-; SSE-NEXT:    pandn %xmm4, %xmm3
-; SSE-NEXT:    movdqa %xmm10, %xmm7
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm5, %xmm7
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm1[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    packuswb %xmm3, %xmm3
+; SSE-NEXT:    movdqa %xmm9, %xmm4
+; SSE-NEXT:    pandn %xmm3, %xmm4
+; SSE-NEXT:    movdqa %xmm8, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm3, %xmm6
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm0, %xmm14
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255]
-; SSE-NEXT:    pand %xmm1, %xmm14
-; SSE-NEXT:    movdqa %xmm15, %xmm11
-; SSE-NEXT:    pand %xmm1, %xmm11
-; SSE-NEXT:    movdqa %xmm13, %xmm4
-; SSE-NEXT:    pand %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm5, %xmm2
-; SSE-NEXT:    pand %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm8, %xmm0
+; SSE-NEXT:    movdqa %xmm13, %xmm5
+; SSE-NEXT:    pand %xmm13, %xmm14
+; SSE-NEXT:    movdqa %xmm2, %xmm11
+; SSE-NEXT:    pand %xmm13, %xmm2
+; SSE-NEXT:    movdqa %xmm15, %xmm8
+; SSE-NEXT:    pand %xmm13, %xmm8
+; SSE-NEXT:    movdqa %xmm3, %xmm13
+; SSE-NEXT:    pand %xmm5, %xmm13
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm13, %xmm5
+; SSE-NEXT:    pand %xmm7, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm8, %xmm15
-; SSE-NEXT:    pand %xmm8, %xmm13
-; SSE-NEXT:    pand %xmm8, %xmm5
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pandn %xmm2, %xmm8
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm10, %xmm0
-; SSE-NEXT:    por %xmm7, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm7
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm9[8],xmm7[9],xmm9[9],xmm7[10],xmm9[10],xmm7[11],xmm9[11],xmm7[12],xmm9[12],xmm7[13],xmm9[13],xmm7[14],xmm9[14],xmm7[15],xmm9[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
+; SSE-NEXT:    pand %xmm7, %xmm11
+; SSE-NEXT:    movdqa %xmm11, (%rsp) # 16-byte Spill
+; SSE-NEXT:    pand %xmm7, %xmm15
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm7, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    pandn %xmm13, %xmm7
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; SSE-NEXT:    por %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm0, %xmm6
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm12[8],xmm6[9],xmm12[9],xmm6[10],xmm12[10],xmm6[11],xmm12[11],xmm6[12],xmm12[12],xmm6[13],xmm12[13],xmm6[14],xmm12[14],xmm6[15],xmm12[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1],xmm0[2],xmm12[2],xmm0[3],xmm12[3],xmm0[4],xmm12[4],xmm0[5],xmm12[5],xmm0[6],xmm12[6],xmm0[7],xmm12[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,1,2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm7[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm0[0,2]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm7[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm6[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[0,2]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm6[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,6,6,6]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,6,5]
 ; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm12, %xmm1
-; SSE-NEXT:    por %xmm3, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[1,1,1,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm9, %xmm1
+; SSE-NEXT:    por %xmm4, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[1,1,1,1]
+; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm4[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0]
-; SSE-NEXT:    movdqa %xmm10, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pand %xmm10, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0]
+; SSE-NEXT:    movdqa %xmm12, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    pand %xmm12, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255]
-; SSE-NEXT:    pand %xmm3, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,255,0,255,255,255,255,0,255,255,255,255,0,255,255]
+; SSE-NEXT:    pand %xmm7, %xmm0
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm9[8],xmm1[9],xmm9[9],xmm1[10],xmm9[10],xmm1[11],xmm9[11],xmm1[12],xmm9[12],xmm1[13],xmm9[13],xmm1[14],xmm9[14],xmm1[15],xmm9[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
+; SSE-NEXT:    pxor %xmm3, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm3[8],xmm1[9],xmm3[9],xmm1[10],xmm3[10],xmm1[11],xmm3[11],xmm1[12],xmm3[12],xmm1[13],xmm3[13],xmm1[14],xmm3[14],xmm1[15],xmm3[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm1[2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,0,1,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,4,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    movdqa %xmm9, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm9[8],xmm14[9],xmm9[9],xmm14[10],xmm9[10],xmm14[11],xmm9[11],xmm14[12],xmm9[12],xmm14[13],xmm9[13],xmm14[14],xmm9[14],xmm14[15],xmm9[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm3[8],xmm14[9],xmm3[9],xmm14[10],xmm3[10],xmm14[11],xmm3[11],xmm14[12],xmm3[12],xmm14[13],xmm3[13],xmm14[14],xmm3[14],xmm14[15],xmm3[15]
 ; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[3,1],xmm0[2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm14[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,0,1,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm14[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,0,1,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,5,6]
-; SSE-NEXT:    packuswb %xmm2, %xmm0
-; SSE-NEXT:    pand %xmm12, %xmm0
+; SSE-NEXT:    packuswb %xmm4, %xmm0
+; SSE-NEXT:    pand %xmm9, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm2[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm4[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm1[0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm4[0,1,2,3,4,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,5]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm10, %xmm9
-; SSE-NEXT:    movdqa %xmm10, %xmm14
+; SSE-NEXT:    movdqa %xmm12, %xmm14
 ; SSE-NEXT:    pandn %xmm1, %xmm14
-; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    pand %xmm12, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm14
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm3, %xmm0
+; SSE-NEXT:    pand %xmm7, %xmm0
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    pxor %xmm2, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm2[8],xmm1[9],xmm2[9],xmm1[10],xmm2[10],xmm1[11],xmm2[11],xmm1[12],xmm2[12],xmm1[13],xmm2[13],xmm1[14],xmm2[14],xmm1[15],xmm2[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm3[8],xmm1[9],xmm3[9],xmm1[10],xmm3[10],xmm1[11],xmm3[11],xmm1[12],xmm3[12],xmm1[13],xmm3[13],xmm1[14],xmm3[14],xmm1[15],xmm3[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm1[2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,0,1,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,4,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    movdqa %xmm9, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm2[8],xmm11[9],xmm2[9],xmm11[10],xmm2[10],xmm11[11],xmm2[11],xmm11[12],xmm2[12],xmm11[13],xmm2[13],xmm11[14],xmm2[14],xmm11[15],xmm2[15]
-; SSE-NEXT:    pxor %xmm10, %xmm10
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[3,1],xmm0[2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm11[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,0,1,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm3[8],xmm2[9],xmm3[9],xmm2[10],xmm3[10],xmm2[11],xmm3[11],xmm2[12],xmm3[12],xmm2[13],xmm3[13],xmm2[14],xmm3[14],xmm2[15],xmm3[15]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,1],xmm0[2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm2[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,0,1,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,5,6]
-; SSE-NEXT:    packuswb %xmm2, %xmm0
-; SSE-NEXT:    pand %xmm12, %xmm0
+; SSE-NEXT:    packuswb %xmm4, %xmm0
+; SSE-NEXT:    pand %xmm9, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm2[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm4[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm1[0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm4[0,1,2,3,4,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,5]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm9, %xmm11
-; SSE-NEXT:    pandn %xmm1, %xmm11
-; SSE-NEXT:    pand %xmm9, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm11
-; SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm12, %xmm13
+; SSE-NEXT:    pandn %xmm1, %xmm13
+; SSE-NEXT:    pand %xmm12, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm13
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pand %xmm7, %xmm0
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm10[8],xmm1[9],xmm10[9],xmm1[10],xmm10[10],xmm1[11],xmm10[11],xmm1[12],xmm10[12],xmm1[13],xmm10[13],xmm1[14],xmm10[14],xmm1[15],xmm10[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1],xmm0[2],xmm10[2],xmm0[3],xmm10[3],xmm0[4],xmm10[4],xmm0[5],xmm10[5],xmm0[6],xmm10[6],xmm0[7],xmm10[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm3[8],xmm1[9],xmm3[9],xmm1[10],xmm3[10],xmm1[11],xmm3[11],xmm1[12],xmm3[12],xmm1[13],xmm3[13],xmm1[14],xmm3[14],xmm1[15],xmm3[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm1[2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,0,1,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,4,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    movdqa %xmm9, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1],xmm0[2],xmm10[2],xmm0[3],xmm10[3],xmm0[4],xmm10[4],xmm0[5],xmm10[5],xmm0[6],xmm10[6],xmm0[7],xmm10[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm10[8],xmm4[9],xmm10[9],xmm4[10],xmm10[10],xmm4[11],xmm10[11],xmm4[12],xmm10[12],xmm4[13],xmm10[13],xmm4[14],xmm10[14],xmm4[15],xmm10[15]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,1],xmm0[2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm4[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,0,1,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm8, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm3[8],xmm8[9],xmm3[9],xmm8[10],xmm3[10],xmm8[11],xmm3[11],xmm8[12],xmm3[12],xmm8[13],xmm3[13],xmm8[14],xmm3[14],xmm8[15],xmm3[15]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[3,1],xmm0[2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm8[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,0,1,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,5,6]
-; SSE-NEXT:    packuswb %xmm2, %xmm0
-; SSE-NEXT:    pand %xmm12, %xmm0
+; SSE-NEXT:    packuswb %xmm3, %xmm0
+; SSE-NEXT:    pand %xmm9, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm2[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm3[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm1[0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,4,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,5]
-; SSE-NEXT:    packuswb %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm9, %xmm10
-; SSE-NEXT:    pandn %xmm2, %xmm10
-; SSE-NEXT:    pand %xmm9, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm10
+; SSE-NEXT:    packuswb %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    pandn %xmm3, %xmm1
+; SSE-NEXT:    pand %xmm12, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    pand %xmm3, %xmm0
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm2, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,1],xmm0[2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,0,1,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    movdqa %xmm7, %xmm3
+; SSE-NEXT:    pand %xmm7, %xmm0
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    por %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm3, %xmm0
+; SSE-NEXT:    pxor %xmm2, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,1],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm3, %xmm4
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm2[8],xmm5[9],xmm2[9],xmm5[10],xmm2[10],xmm5[11],xmm2[11],xmm5[12],xmm2[12],xmm5[13],xmm2[13],xmm5[14],xmm2[14],xmm5[15],xmm2[15]
+; SSE-NEXT:    pxor %xmm15, %xmm15
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,1],xmm0[2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm5[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,0,1,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,5,6]
-; SSE-NEXT:    packuswb %xmm2, %xmm0
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm4[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,0]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,0,1,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,4,6,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pandn %xmm2, %xmm12
-; SSE-NEXT:    por %xmm12, %xmm0
-; SSE-NEXT:    movdqa %xmm8, %xmm3
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm8[3,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm5[0,2]
+; SSE-NEXT:    packuswb %xmm3, %xmm0
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm4[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[3,1,2,0]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,0,1,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,4,6,7]
+; SSE-NEXT:    packuswb %xmm3, %xmm3
 ; SSE-NEXT:    pand %xmm9, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,4,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,7,5]
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pandn %xmm2, %xmm9
+; SSE-NEXT:    pandn %xmm3, %xmm9
 ; SSE-NEXT:    por %xmm0, %xmm9
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[2,0],xmm3[3,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm11[0,2]
+; SSE-NEXT:    movdqa %xmm12, %xmm2
+; SSE-NEXT:    pand %xmm12, %xmm9
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm3[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,7,5]
+; SSE-NEXT:    packuswb %xmm0, %xmm0
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    por %xmm9, %xmm2
+; SSE-NEXT:    movdqa %xmm2, %xmm9
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3],xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm15[8],xmm0[9],xmm15[9],xmm0[10],xmm15[10],xmm0[11],xmm15[11],xmm0[12],xmm15[12],xmm0[13],xmm15[13],xmm0[14],xmm15[14],xmm0[15],xmm15[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm15[0],xmm3[1],xmm15[1],xmm3[2],xmm15[2],xmm3[3],xmm15[3],xmm3[4],xmm15[4],xmm3[5],xmm15[5],xmm3[6],xmm15[6],xmm3[7],xmm15[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,1,1,3]
 ; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,3],xmm0[1,2]
 ; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2,3,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm3[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,3,0,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm0[2,1,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm2
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    packuswb %xmm4, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm0, %xmm6
+; SSE-NEXT:    pandn %xmm2, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255]
+; SSE-NEXT:    pand %xmm8, %xmm7
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm15[0],xmm2[1],xmm15[1],xmm2[2],xmm15[2],xmm2[3],xmm15[3],xmm2[4],xmm15[4],xmm2[5],xmm15[5],xmm2[6],xmm15[6],xmm2[7],xmm15[7]
+; SSE-NEXT:    movdqa %xmm10, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,255,255,0,255,255,255,255,0,255,255,255,255,0,255]
-; SSE-NEXT:    pand %xmm12, %xmm8
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm8, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
-; SSE-NEXT:    movdqa %xmm6, %xmm7
-; SSE-NEXT:    pandn %xmm2, %xmm7
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm1[8],xmm8[9],xmm1[9],xmm8[10],xmm1[10],xmm8[11],xmm1[11],xmm8[12],xmm1[12],xmm8[13],xmm1[13],xmm8[14],xmm1[14],xmm8[15],xmm1[15]
-; SSE-NEXT:    pand %xmm6, %xmm8
-; SSE-NEXT:    por %xmm7, %xmm8
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm8[3,1,0,3,4,5,6,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm15[8],xmm7[9],xmm15[9],xmm7[10],xmm15[10],xmm7[11],xmm15[11],xmm7[12],xmm15[12],xmm7[13],xmm15[13],xmm7[14],xmm15[14],xmm7[15],xmm15[15]
+; SSE-NEXT:    pand %xmm10, %xmm7
+; SSE-NEXT:    por %xmm3, %xmm7
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm7[3,1,0,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,4]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pand %xmm4, %xmm2
-; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
-; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm7 = mem[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,2,1,4,5,6,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm3[0],xmm7[1],xmm3[1]
-; SSE-NEXT:    packuswb %xmm0, %xmm7
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm7[2,1]
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm1[0],xmm15[1],xmm1[1],xmm15[2],xmm1[2],xmm15[3],xmm1[3],xmm15[4],xmm1[4],xmm15[5],xmm1[5],xmm15[6],xmm1[6],xmm15[7],xmm1[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm15[0,1,1,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,3],xmm0[1,2]
-; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,2,3,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm15[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm4, %xmm7
-; SSE-NEXT:    pandn %xmm3, %xmm7
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    pand %xmm12, %xmm15
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm15, %xmm3
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3],xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
-; SSE-NEXT:    movdqa %xmm6, %xmm8
-; SSE-NEXT:    pandn %xmm3, %xmm8
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm15 = xmm15[8],xmm1[8],xmm15[9],xmm1[9],xmm15[10],xmm1[10],xmm15[11],xmm1[11],xmm15[12],xmm1[12],xmm15[13],xmm1[13],xmm15[14],xmm1[14],xmm15[15],xmm1[15]
-; SSE-NEXT:    pand %xmm6, %xmm15
-; SSE-NEXT:    por %xmm8, %xmm15
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm15[3,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,4]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm3[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    packuswb %xmm8, %xmm8
-; SSE-NEXT:    pand %xmm4, %xmm8
-; SSE-NEXT:    por %xmm7, %xmm8
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
-; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm7 = mem[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,2,1,4,5,6,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm3[0],xmm7[1],xmm3[1]
-; SSE-NEXT:    packuswb %xmm0, %xmm7
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm7[2,1]
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm1[0],xmm13[1],xmm1[1],xmm13[2],xmm1[2],xmm13[3],xmm1[3],xmm13[4],xmm1[4],xmm13[5],xmm1[5],xmm13[6],xmm1[6],xmm13[7],xmm1[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[0,1,1,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,3],xmm0[1,2]
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,2,3,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm13[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm4, %xmm7
-; SSE-NEXT:    pandn %xmm3, %xmm7
-; SSE-NEXT:    movdqa (%rsp), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pand %xmm12, %xmm13
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm13, %xmm3
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3],xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
-; SSE-NEXT:    movdqa %xmm6, %xmm5
-; SSE-NEXT:    pandn %xmm3, %xmm5
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm1[8],xmm13[9],xmm1[9],xmm13[10],xmm1[10],xmm13[11],xmm1[11],xmm13[12],xmm1[12],xmm13[13],xmm1[13],xmm13[14],xmm1[14],xmm13[15],xmm1[15]
-; SSE-NEXT:    pand %xmm6, %xmm13
-; SSE-NEXT:    por %xmm5, %xmm13
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm13[3,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,4]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm3[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    packuswb %xmm5, %xmm5
-; SSE-NEXT:    pand %xmm4, %xmm5
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm2[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    packuswb %xmm3, %xmm3
+; SSE-NEXT:    pand %xmm0, %xmm3
+; SSE-NEXT:    por %xmm6, %xmm3
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[3,1,2,3]
+; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm6 = mem[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,2,1,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm2[0],xmm6[1],xmm2[1]
+; SSE-NEXT:    packuswb %xmm4, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm6[2,1]
+; SSE-NEXT:    movdqa (%rsp), %xmm5 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm5, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm15[8],xmm2[9],xmm15[9],xmm2[10],xmm15[10],xmm2[11],xmm15[11],xmm2[12],xmm15[12],xmm2[13],xmm15[13],xmm2[14],xmm15[14],xmm2[15],xmm15[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm15[0],xmm5[1],xmm15[1],xmm5[2],xmm15[2],xmm5[3],xmm15[3],xmm5[4],xmm15[4],xmm5[5],xmm15[5],xmm5[6],xmm15[6],xmm5[7],xmm15[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[0,1,1,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm2[1,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2,3,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm5[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    packuswb %xmm2, %xmm4
+; SSE-NEXT:    movdqa %xmm0, %xmm6
+; SSE-NEXT:    pandn %xmm4, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    pand %xmm8, %xmm5
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm5, %xmm4
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm15[0],xmm4[1],xmm15[1],xmm4[2],xmm15[2],xmm4[3],xmm15[3],xmm4[4],xmm15[4],xmm4[5],xmm15[5],xmm4[6],xmm15[6],xmm4[7],xmm15[7]
+; SSE-NEXT:    movdqa %xmm10, %xmm7
+; SSE-NEXT:    pandn %xmm4, %xmm7
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm15[8],xmm5[9],xmm15[9],xmm5[10],xmm15[10],xmm5[11],xmm15[11],xmm5[12],xmm15[12],xmm5[13],xmm15[13],xmm5[14],xmm15[14],xmm5[15],xmm15[15]
+; SSE-NEXT:    pand %xmm10, %xmm5
 ; SSE-NEXT:    por %xmm7, %xmm5
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
-; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm7 = mem[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,2,1,4,5,6,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm3[0],xmm7[1],xmm3[1]
-; SSE-NEXT:    packuswb %xmm0, %xmm7
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm7[2,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1],xmm7[2],xmm1[2],xmm7[3],xmm1[3],xmm7[4],xmm1[4],xmm7[5],xmm1[5],xmm7[6],xmm1[6],xmm7[7],xmm1[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[0,1,1,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,3],xmm0[1,2]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; SSE-NEXT:    por %xmm0, %xmm12
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm1[8],xmm12[9],xmm1[9],xmm12[10],xmm1[10],xmm12[11],xmm1[11],xmm12[12],xmm1[12],xmm12[13],xmm1[13],xmm12[14],xmm1[14],xmm12[15],xmm1[15]
-; SSE-NEXT:    pand %xmm6, %xmm12
-; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    por %xmm12, %xmm6
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm6[3,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,4]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm4, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2,3,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm7[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[2,1,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    packuswb %xmm6, %xmm3
-; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm0
-; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[3,1,2,3]
-; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,2,1,4,5,6,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT:    packuswb %xmm6, %xmm4
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm4[2,1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 16(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 48(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, (%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 32(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 16(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 48(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, (%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, 32(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 16(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 48(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, (%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 32(%rcx)
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm5[3,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,6,4]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,1,2,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm4[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    packuswb %xmm7, %xmm7
+; SSE-NEXT:    pand %xmm0, %xmm7
+; SSE-NEXT:    por %xmm6, %xmm7
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = mem[3,1,2,3]
+; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm6 = mem[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,2,1,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
+; SSE-NEXT:    packuswb %xmm2, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm6[2,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm5, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm15[8],xmm2[9],xmm15[9],xmm2[10],xmm15[10],xmm2[11],xmm15[11],xmm2[12],xmm15[12],xmm2[13],xmm15[13],xmm2[14],xmm15[14],xmm2[15],xmm15[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm15[0],xmm5[1],xmm15[1],xmm5[2],xmm15[2],xmm5[3],xmm15[3],xmm5[4],xmm15[4],xmm5[5],xmm15[5],xmm5[6],xmm15[6],xmm5[7],xmm15[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[0,1,1,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm2[1,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2,3,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm5[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    packuswb %xmm2, %xmm4
+; SSE-NEXT:    movdqa %xmm0, %xmm6
+; SSE-NEXT:    pandn %xmm4, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    pand %xmm8, %xmm5
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm5, %xmm4
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm15[0],xmm4[1],xmm15[1],xmm4[2],xmm15[2],xmm4[3],xmm15[3],xmm4[4],xmm15[4],xmm4[5],xmm15[5],xmm4[6],xmm15[6],xmm4[7],xmm15[7]
+; SSE-NEXT:    movdqa %xmm10, %xmm11
+; SSE-NEXT:    pandn %xmm4, %xmm11
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm15[8],xmm5[9],xmm15[9],xmm5[10],xmm15[10],xmm5[11],xmm15[11],xmm5[12],xmm15[12],xmm5[13],xmm15[13],xmm5[14],xmm15[14],xmm5[15],xmm15[15]
+; SSE-NEXT:    pand %xmm10, %xmm5
+; SSE-NEXT:    por %xmm11, %xmm5
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm5[3,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,6,4]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,1,2,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    packuswb %xmm4, %xmm4
+; SSE-NEXT:    pand %xmm0, %xmm4
+; SSE-NEXT:    por %xmm6, %xmm4
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm6 = mem[3,1,2,3]
+; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = mem[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,1,2,1,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm6[0],xmm11[1],xmm6[1]
+; SSE-NEXT:    packuswb %xmm2, %xmm11
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm11[2,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm5, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm15[8],xmm2[9],xmm15[9],xmm2[10],xmm15[10],xmm2[11],xmm15[11],xmm2[12],xmm15[12],xmm2[13],xmm15[13],xmm2[14],xmm15[14],xmm2[15],xmm15[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm15[0],xmm5[1],xmm15[1],xmm5[2],xmm15[2],xmm5[3],xmm15[3],xmm5[4],xmm15[4],xmm5[5],xmm15[5],xmm5[6],xmm15[6],xmm5[7],xmm15[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,1,1,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,3],xmm2[1,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pand %xmm8, %xmm2
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    por %xmm2, %xmm8
+; SSE-NEXT:    movdqa %xmm8, %xmm2
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm15[0],xmm2[1],xmm15[1],xmm2[2],xmm15[2],xmm2[3],xmm15[3],xmm2[4],xmm15[4],xmm2[5],xmm15[5],xmm2[6],xmm15[6],xmm2[7],xmm15[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm15[8],xmm8[9],xmm15[9],xmm8[10],xmm15[10],xmm8[11],xmm15[11],xmm8[12],xmm15[12],xmm8[13],xmm15[13],xmm8[14],xmm15[14],xmm8[15],xmm15[15]
+; SSE-NEXT:    pand %xmm10, %xmm8
+; SSE-NEXT:    pandn %xmm2, %xmm10
+; SSE-NEXT:    por %xmm8, %xmm10
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2,3,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm5[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    packuswb %xmm2, %xmm6
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm10[3,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,5,6,4]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[3,1,2,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    packuswb %xmm8, %xmm8
+; SSE-NEXT:    pand %xmm0, %xmm8
+; SSE-NEXT:    pandn %xmm6, %xmm0
+; SSE-NEXT:    por %xmm8, %xmm0
+; SSE-NEXT:    pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm6 = mem[3,1,2,3]
+; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = mem[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,1,2,1,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm6[0],xmm8[1],xmm6[1]
+; SSE-NEXT:    packuswb %xmm2, %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm8[2,1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, 48(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, 32(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, 16(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, 48(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, (%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, 32(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, 16(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, 48(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, (%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm2, 32(%rcx)
 ; SSE-NEXT:    movdqa %xmm9, 16(%r8)
-; SSE-NEXT:    movdqa %xmm10, 48(%r8)
-; SSE-NEXT:    movdqa %xmm11, (%r8)
+; SSE-NEXT:    movdqa %xmm1, 48(%r8)
+; SSE-NEXT:    movdqa %xmm13, (%r8)
 ; SSE-NEXT:    movdqa %xmm14, 32(%r8)
 ; SSE-NEXT:    movaps %xmm0, 16(%r9)
-; SSE-NEXT:    movaps %xmm5, 48(%r9)
-; SSE-NEXT:    movaps %xmm8, (%r9)
-; SSE-NEXT:    movaps %xmm2, 32(%r9)
-; SSE-NEXT:    addq $552, %rsp # imm = 0x228
+; SSE-NEXT:    movaps %xmm4, 48(%r9)
+; SSE-NEXT:    movaps %xmm7, (%r9)
+; SSE-NEXT:    movaps %xmm3, 32(%r9)
+; SSE-NEXT:    addq $568, %rsp # imm = 0x238
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i8_stride5_vf64:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll
index c77b232fde969..a148589103d53 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll
@@ -755,145 +755,145 @@ define void @load_i8_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-LABEL: load_i8_stride6_vf8:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movdqa (%rdi), %xmm4
-; SSE-NEXT:    movdqa 16(%rdi), %xmm3
+; SSE-NEXT:    movdqa (%rdi), %xmm8
+; SSE-NEXT:    movdqa 16(%rdi), %xmm2
 ; SSE-NEXT:    movdqa 32(%rdi), %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,0,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    pand %xmm8, %xmm1
-; SSE-NEXT:    pandn %xmm3, %xmm8
-; SSE-NEXT:    por %xmm1, %xmm8
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[0,2,1,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [16711935,16711935,16711935,16711935]
-; SSE-NEXT:    pand %xmm5, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,0,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    pand %xmm7, %xmm1
+; SSE-NEXT:    pandn %xmm2, %xmm7
+; SSE-NEXT:    por %xmm1, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[0,2,1,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [16711935,16711935,16711935,16711935]
+; SSE-NEXT:    pand %xmm4, %xmm1
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm1[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    packuswb %xmm6, %xmm6
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm1[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    packuswb %xmm5, %xmm5
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,65535,65535,65535,65535]
-; SSE-NEXT:    pand %xmm1, %xmm6
-; SSE-NEXT:    movdqa %xmm0, %xmm7
-; SSE-NEXT:    pand %xmm5, %xmm7
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm2[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    pand %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm0, %xmm6
+; SSE-NEXT:    pand %xmm4, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm6[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm3[0,1,2,3,4,5,6,5]
 ; SSE-NEXT:    packuswb %xmm9, %xmm9
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    pandn %xmm9, %xmm2
-; SSE-NEXT:    por %xmm6, %xmm2
-; SSE-NEXT:    pxor %xmm6, %xmm6
-; SSE-NEXT:    movdqa %xmm8, %xmm9
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm6[8],xmm9[9],xmm6[9],xmm9[10],xmm6[10],xmm9[11],xmm6[11],xmm9[12],xmm6[12],xmm9[13],xmm6[13],xmm9[14],xmm6[14],xmm9[15],xmm6[15]
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm9, %xmm3
+; SSE-NEXT:    por %xmm5, %xmm3
+; SSE-NEXT:    pxor %xmm5, %xmm5
+; SSE-NEXT:    movdqa %xmm7, %xmm9
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm5[8],xmm9[9],xmm5[9],xmm9[10],xmm5[10],xmm9[11],xmm5[11],xmm9[12],xmm5[12],xmm9[13],xmm5[13],xmm9[14],xmm5[14],xmm9[15],xmm5[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[2,1,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[1,1,1,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,5,7,6,7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,0,65535,0,0,65535,65535]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3],xmm8[4],xmm6[4],xmm8[5],xmm6[5],xmm8[6],xmm6[6],xmm8[7],xmm6[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[1,3,2,0,4,5,6,7]
-; SSE-NEXT:    pand %xmm10, %xmm8
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm5[0],xmm7[1],xmm5[1],xmm7[2],xmm5[2],xmm7[3],xmm5[3],xmm7[4],xmm5[4],xmm7[5],xmm5[5],xmm7[6],xmm5[6],xmm7[7],xmm5[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[1,3,2,0,4,5,6,7]
+; SSE-NEXT:    pand %xmm10, %xmm7
 ; SSE-NEXT:    pandn %xmm9, %xmm10
-; SSE-NEXT:    por %xmm8, %xmm10
+; SSE-NEXT:    por %xmm7, %xmm10
 ; SSE-NEXT:    packuswb %xmm10, %xmm10
 ; SSE-NEXT:    pand %xmm1, %xmm10
-; SSE-NEXT:    movdqa %xmm0, %xmm8
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm6[8],xmm8[9],xmm6[9],xmm8[10],xmm6[10],xmm8[11],xmm6[11],xmm8[12],xmm6[12],xmm8[13],xmm6[13],xmm8[14],xmm6[14],xmm8[15],xmm6[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3],xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
+; SSE-NEXT:    movdqa %xmm0, %xmm7
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm5[8],xmm7[9],xmm5[9],xmm7[10],xmm5[10],xmm7[11],xmm5[11],xmm7[12],xmm5[12],xmm7[13],xmm5[13],xmm7[14],xmm5[14],xmm7[15],xmm5[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm0[2,2,3,3]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm7[0],xmm9[1],xmm7[1],xmm9[2],xmm7[2],xmm9[3],xmm7[3]
 ; SSE-NEXT:    packuswb %xmm9, %xmm9
 ; SSE-NEXT:    pandn %xmm9, %xmm1
 ; SSE-NEXT:    por %xmm10, %xmm1
 ; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,65535,0,65535,65535,0,65535,65535]
 ; SSE-NEXT:    movdqa %xmm11, %xmm9
-; SSE-NEXT:    pandn %xmm3, %xmm9
-; SSE-NEXT:    movdqa %xmm4, %xmm12
-; SSE-NEXT:    pand %xmm11, %xmm12
-; SSE-NEXT:    por %xmm9, %xmm12
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm12[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pandn %xmm2, %xmm9
+; SSE-NEXT:    movdqa %xmm11, %xmm12
+; SSE-NEXT:    pandn %xmm8, %xmm12
+; SSE-NEXT:    pand %xmm11, %xmm8
+; SSE-NEXT:    por %xmm9, %xmm8
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm8[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pand %xmm5, %xmm9
+; SSE-NEXT:    pand %xmm4, %xmm9
 ; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm9[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    packuswb %xmm13, %xmm13
 ; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [255,255,255,255,255,0,0,0,255,255,255,255,255,255,255,255]
 ; SSE-NEXT:    pand %xmm9, %xmm13
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm7[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm6[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm10[0,1,2,3,4,4,5,6]
 ; SSE-NEXT:    packuswb %xmm14, %xmm14
 ; SSE-NEXT:    movdqa %xmm9, %xmm10
 ; SSE-NEXT:    pandn %xmm14, %xmm10
 ; SSE-NEXT:    por %xmm13, %xmm10
-; SSE-NEXT:    movdqa %xmm12, %xmm13
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm6[0],xmm13[1],xmm6[1],xmm13[2],xmm6[2],xmm13[3],xmm6[3],xmm13[4],xmm6[4],xmm13[5],xmm6[5],xmm13[6],xmm6[6],xmm13[7],xmm6[7]
+; SSE-NEXT:    movdqa %xmm8, %xmm13
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm5[0],xmm13[1],xmm5[1],xmm13[2],xmm5[2],xmm13[3],xmm5[3],xmm13[4],xmm5[4],xmm13[5],xmm5[5],xmm13[6],xmm5[6],xmm13[7],xmm5[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm13[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm13[3,1,2,1,4,5,6,7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [0,65535,65535,0,65535,65535,65535,65535]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm6[8],xmm12[9],xmm6[9],xmm12[10],xmm6[10],xmm12[11],xmm6[11],xmm12[12],xmm6[12],xmm12[13],xmm6[13],xmm12[14],xmm6[14],xmm12[15],xmm6[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[0,1,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pand %xmm14, %xmm12
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm5[8],xmm8[9],xmm5[9],xmm8[10],xmm5[10],xmm8[11],xmm5[11],xmm8[12],xmm5[12],xmm8[13],xmm5[13],xmm8[14],xmm5[14],xmm8[15],xmm5[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[0,1,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pand %xmm14, %xmm8
 ; SSE-NEXT:    pandn %xmm13, %xmm14
-; SSE-NEXT:    por %xmm12, %xmm14
+; SSE-NEXT:    por %xmm8, %xmm14
 ; SSE-NEXT:    packuswb %xmm14, %xmm14
 ; SSE-NEXT:    pand %xmm9, %xmm14
-; SSE-NEXT:    movdqa %xmm8, %xmm12
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[2,0],xmm0[3,0]
+; SSE-NEXT:    movdqa %xmm7, %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,0],xmm0[3,0]
 ; SSE-NEXT:    movaps %xmm0, %xmm13
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,1],xmm12[0,2]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm13[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm12[0,1,0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,1],xmm8[0,2]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm13[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm8[0,1,0,2]
 ; SSE-NEXT:    packuswb %xmm13, %xmm13
-; SSE-NEXT:    movdqa %xmm9, %xmm12
-; SSE-NEXT:    pandn %xmm13, %xmm12
-; SSE-NEXT:    por %xmm14, %xmm12
-; SSE-NEXT:    pand %xmm11, %xmm3
-; SSE-NEXT:    pandn %xmm4, %xmm11
-; SSE-NEXT:    por %xmm3, %xmm11
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[3,1,2,0]
-; SSE-NEXT:    pand %xmm5, %xmm3
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm3[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm4, %xmm4
-; SSE-NEXT:    pand %xmm9, %xmm4
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm7[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[0,1,0,2]
-; SSE-NEXT:    packuswb %xmm5, %xmm5
-; SSE-NEXT:    movdqa %xmm9, %xmm3
-; SSE-NEXT:    pandn %xmm5, %xmm3
-; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    movdqa %xmm11, %xmm4
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm6[8],xmm4[9],xmm6[9],xmm4[10],xmm6[10],xmm4[11],xmm6[11],xmm4[12],xmm6[12],xmm4[13],xmm6[13],xmm4[14],xmm6[14],xmm4[15],xmm6[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[1,1,2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,0,65535,65535,0,65535,65535,65535]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm6[0],xmm11[1],xmm6[1],xmm11[2],xmm6[2],xmm11[3],xmm6[3],xmm11[4],xmm6[4],xmm11[5],xmm6[5],xmm11[6],xmm6[6],xmm11[7],xmm6[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm11[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[3,1,1,2,4,5,6,7]
-; SSE-NEXT:    pand %xmm5, %xmm6
-; SSE-NEXT:    pandn %xmm4, %xmm5
-; SSE-NEXT:    por %xmm6, %xmm5
-; SSE-NEXT:    packuswb %xmm5, %xmm5
-; SSE-NEXT:    pand %xmm9, %xmm5
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm8[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm8[2,3]
+; SSE-NEXT:    movdqa %xmm9, %xmm8
+; SSE-NEXT:    pandn %xmm13, %xmm8
+; SSE-NEXT:    por %xmm14, %xmm8
+; SSE-NEXT:    pand %xmm11, %xmm2
+; SSE-NEXT:    por %xmm12, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm2[3,1,2,0]
+; SSE-NEXT:    pand %xmm4, %xmm11
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm11[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm4[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm11, %xmm11
+; SSE-NEXT:    pand %xmm9, %xmm11
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm6[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm4[0,1,0,2]
+; SSE-NEXT:    packuswb %xmm6, %xmm6
+; SSE-NEXT:    movdqa %xmm9, %xmm4
+; SSE-NEXT:    pandn %xmm6, %xmm4
+; SSE-NEXT:    por %xmm11, %xmm4
+; SSE-NEXT:    movdqa %xmm2, %xmm6
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm5[8],xmm6[9],xmm5[9],xmm6[10],xmm5[10],xmm6[11],xmm5[11],xmm6[12],xmm5[12],xmm6[13],xmm5[13],xmm6[14],xmm5[14],xmm6[15],xmm5[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[1,1,2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,0,65535,65535,0,65535,65535,65535]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3],xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,1,2,4,5,6,7]
+; SSE-NEXT:    pand %xmm11, %xmm2
+; SSE-NEXT:    pandn %xmm6, %xmm11
+; SSE-NEXT:    por %xmm2, %xmm11
+; SSE-NEXT:    packuswb %xmm11, %xmm11
+; SSE-NEXT:    pand %xmm9, %xmm11
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm7[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm7[2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,5,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
 ; SSE-NEXT:    pandn %xmm0, %xmm9
-; SSE-NEXT:    por %xmm5, %xmm9
-; SSE-NEXT:    movq %xmm2, (%rsi)
+; SSE-NEXT:    por %xmm11, %xmm9
+; SSE-NEXT:    movq %xmm3, (%rsi)
 ; SSE-NEXT:    movq %xmm1, (%rdx)
 ; SSE-NEXT:    movq %xmm10, (%rcx)
-; SSE-NEXT:    movq %xmm12, (%r8)
-; SSE-NEXT:    movq %xmm3, (%r9)
+; SSE-NEXT:    movq %xmm8, (%r8)
+; SSE-NEXT:    movq %xmm4, (%r9)
 ; SSE-NEXT:    movq %xmm9, (%rax)
 ; SSE-NEXT:    retq
 ;
@@ -1385,299 +1385,299 @@ define void @load_i8_stride6_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i8_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5) nounwind {
 ; SSE-LABEL: load_i8_stride6_vf16:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    movdqa 64(%rdi), %xmm10
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 64(%rdi), %xmm7
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa (%rdi), %xmm5
-; SSE-NEXT:    movdqa 16(%rdi), %xmm1
-; SSE-NEXT:    movdqa 32(%rdi), %xmm7
-; SSE-NEXT:    movdqa 48(%rdi), %xmm6
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    pandn %xmm7, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,0,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm2, %xmm3
-; SSE-NEXT:    pandn %xmm6, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm4, %xmm3
-; SSE-NEXT:    pandn %xmm6, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm4, %xmm6
-; SSE-NEXT:    por %xmm0, %xmm6
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm0[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,5]
-; SSE-NEXT:    packuswb %xmm3, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm8, %xmm9
-; SSE-NEXT:    pandn %xmm0, %xmm9
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm2, %xmm11
-; SSE-NEXT:    pandn %xmm1, %xmm11
-; SSE-NEXT:    pand %xmm4, %xmm10
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    movdqa 16(%rdi), %xmm0
+; SSE-NEXT:    movdqa 32(%rdi), %xmm15
+; SSE-NEXT:    movdqa 48(%rdi), %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,0,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    pandn %xmm15, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,0,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm8, %xmm2
+; SSE-NEXT:    pandn %xmm3, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm6, %xmm2
+; SSE-NEXT:    pandn %xmm3, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm6, %xmm3
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm1[0,1,2,3,4,5,6,5]
+; SSE-NEXT:    packuswb %xmm2, %xmm4
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm1, %xmm9
+; SSE-NEXT:    pandn %xmm4, %xmm9
+; SSE-NEXT:    movdqa %xmm8, %xmm11
+; SSE-NEXT:    pandn %xmm0, %xmm11
+; SSE-NEXT:    movdqa %xmm7, %xmm10
+; SSE-NEXT:    pand %xmm6, %xmm10
+; SSE-NEXT:    movdqa %xmm6, %xmm4
+; SSE-NEXT:    pandn %xmm0, %xmm4
+; SSE-NEXT:    movdqa %xmm0, %xmm7
 ; SSE-NEXT:    movdqa %xmm5, %xmm14
-; SSE-NEXT:    pand %xmm4, %xmm14
-; SSE-NEXT:    movdqa 80(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm13
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm4, %xmm13
-; SSE-NEXT:    movdqa %xmm7, %xmm15
-; SSE-NEXT:    pand %xmm4, %xmm7
-; SSE-NEXT:    pand %xmm4, %xmm2
+; SSE-NEXT:    pand %xmm6, %xmm14
+; SSE-NEXT:    movdqa 80(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, %xmm13
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm4, %xmm12
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    pandn %xmm5, %xmm4
-; SSE-NEXT:    pand %xmm0, %xmm5
+; SSE-NEXT:    pand %xmm6, %xmm13
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm6, %xmm7
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm6, %xmm7
+; SSE-NEXT:    movdqa %xmm6, %xmm12
+; SSE-NEXT:    pandn %xmm5, %xmm6
+; SSE-NEXT:    pand %xmm8, %xmm5
 ; SSE-NEXT:    por %xmm11, %xmm5
 ; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm5[0,2,1,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm1, %xmm11
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm0, %xmm11
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm11[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    por %xmm9, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm12
-; SSE-NEXT:    por %xmm12, %xmm10
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm10[3,1,2,0]
-; SSE-NEXT:    pand %xmm1, %xmm9
-; SSE-NEXT:    movdqa %xmm1, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,3,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,7,6,5]
-; SSE-NEXT:    packuswb %xmm9, %xmm9
-; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
-; SSE-NEXT:    movdqa %xmm11, %xmm12
-; SSE-NEXT:    pandn %xmm9, %xmm12
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm12
+; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    packuswb %xmm11, %xmm11
+; SSE-NEXT:    pand %xmm1, %xmm11
+; SSE-NEXT:    por %xmm9, %xmm11
+; SSE-NEXT:    pandn %xmm2, %xmm7
+; SSE-NEXT:    por %xmm7, %xmm10
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm10[3,1,2,0]
+; SSE-NEXT:    pand %xmm0, %xmm7
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,3,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,7,6,5]
+; SSE-NEXT:    packuswb %xmm7, %xmm9
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
+; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    pandn %xmm9, %xmm0
+; SSE-NEXT:    pand %xmm7, %xmm11
+; SSE-NEXT:    por %xmm11, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pxor %xmm9, %xmm9
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm9[8],xmm0[9],xmm9[9],xmm0[10],xmm9[10],xmm0[11],xmm9[11],xmm0[12],xmm9[12],xmm0[13],xmm9[13],xmm0[14],xmm9[14],xmm0[15],xmm9[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm9[0],xmm6[1],xmm9[1],xmm6[2],xmm9[2],xmm6[3],xmm9[3],xmm6[4],xmm9[4],xmm6[5],xmm9[5],xmm6[6],xmm9[6],xmm6[7],xmm9[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[2,2,3,3]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    psrld $16, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,5,7,6,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm0[2],xmm6[3],xmm0[3]
-; SSE-NEXT:    packuswb %xmm6, %xmm1
-; SSE-NEXT:    movdqa %xmm5, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm9[8],xmm0[9],xmm9[9],xmm0[10],xmm9[10],xmm0[11],xmm9[11],xmm0[12],xmm9[12],xmm0[13],xmm9[13],xmm0[14],xmm9[14],xmm0[15],xmm9[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,7,6,7]
+; SSE-NEXT:    movdqa %xmm3, %xmm11
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm9[8],xmm11[9],xmm9[9],xmm11[10],xmm9[10],xmm11[11],xmm9[11],xmm11[12],xmm9[12],xmm11[13],xmm9[13],xmm11[14],xmm9[14],xmm11[15],xmm9[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1],xmm3[2],xmm9[2],xmm3[3],xmm9[3],xmm3[4],xmm9[4],xmm3[5],xmm9[5],xmm3[6],xmm9[6],xmm3[7],xmm9[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,2,3,3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1],xmm0[2],xmm11[2],xmm0[3],xmm11[3]
+; SSE-NEXT:    psrld $16, %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,7,6,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm11[2],xmm3[3],xmm11[3]
+; SSE-NEXT:    packuswb %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm9[8],xmm3[9],xmm9[9],xmm3[10],xmm9[10],xmm3[11],xmm9[11],xmm3[12],xmm9[12],xmm3[13],xmm9[13],xmm3[14],xmm9[14],xmm3[15],xmm9[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,7,6,7]
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm9[0],xmm5[1],xmm9[1],xmm5[2],xmm9[2],xmm5[3],xmm9[3],xmm5[4],xmm9[4],xmm5[5],xmm9[5],xmm5[6],xmm9[6],xmm5[7],xmm9[7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,3,2,0,4,5,6,7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,0,65535,0,0,65535,65535]
-; SSE-NEXT:    pand %xmm6, %xmm5
-; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    por %xmm5, %xmm6
-; SSE-NEXT:    packuswb %xmm6, %xmm6
-; SSE-NEXT:    pand %xmm8, %xmm6
-; SSE-NEXT:    pandn %xmm1, %xmm8
-; SSE-NEXT:    por %xmm8, %xmm6
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,65535,0,65535,0,0,65535,65535]
+; SSE-NEXT:    pand %xmm11, %xmm5
+; SSE-NEXT:    pandn %xmm3, %xmm11
+; SSE-NEXT:    por %xmm5, %xmm11
+; SSE-NEXT:    packuswb %xmm11, %xmm11
+; SSE-NEXT:    pand %xmm1, %xmm11
+; SSE-NEXT:    pandn %xmm0, %xmm1
+; SSE-NEXT:    por %xmm11, %xmm1
 ; SSE-NEXT:    movdqa %xmm10, %xmm0
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm9[8],xmm10[9],xmm9[9],xmm10[10],xmm9[10],xmm10[11],xmm9[11],xmm10[12],xmm9[12],xmm10[13],xmm9[13],xmm10[14],xmm9[14],xmm10[15],xmm9[15]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm10[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,7,6,4]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm10[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,7,6,4]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,65535,0,65535,65535,0,65535]
-; SSE-NEXT:    pand %xmm5, %xmm1
+; SSE-NEXT:    pand %xmm5, %xmm3
 ; SSE-NEXT:    pandn %xmm0, %xmm5
-; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    por %xmm3, %xmm5
 ; SSE-NEXT:    packuswb %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm11, %xmm10
+; SSE-NEXT:    movdqa %xmm7, %xmm10
 ; SSE-NEXT:    pandn %xmm0, %xmm10
-; SSE-NEXT:    pand %xmm11, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm10
-; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm15
+; SSE-NEXT:    pand %xmm7, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm10
+; SSE-NEXT:    pand %xmm8, %xmm15
+; SSE-NEXT:    movdqa %xmm8, %xmm11
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pand %xmm3, %xmm0
+; SSE-NEXT:    pand %xmm2, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,3,4,5,6,7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,5,6]
-; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm14[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pand %xmm3, %xmm1
-; SSE-NEXT:    movdqa %xmm3, %xmm8
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm0[0,1,2,3,4,4,5,6]
+; SSE-NEXT:    packuswb %xmm1, %xmm3
+; SSE-NEXT:    por %xmm4, %xmm14
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm14[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    movdqa %xmm2, %xmm8
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm3, %xmm5
-; SSE-NEXT:    pandn %xmm1, %xmm5
-; SSE-NEXT:    pand %xmm3, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm5
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm6, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm13
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm13[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm11, %xmm8
-; SSE-NEXT:    pandn %xmm0, %xmm8
-; SSE-NEXT:    pand %xmm11, %xmm5
-; SSE-NEXT:    por %xmm5, %xmm8
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm0, %xmm4
+; SSE-NEXT:    pandn %xmm1, %xmm4
+; SSE-NEXT:    pand %xmm0, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm4
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    por %xmm12, %xmm13
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm13[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pand %xmm2, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
+; SSE-NEXT:    packuswb %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    pandn %xmm2, %xmm1
+; SSE-NEXT:    pand %xmm7, %xmm4
+; SSE-NEXT:    por %xmm4, %xmm1
+; SSE-NEXT:    movdqa %xmm15, %xmm2
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm9[0],xmm2[1],xmm9[1],xmm2[2],xmm9[2],xmm2[3],xmm9[3],xmm2[4],xmm9[4],xmm2[5],xmm9[5],xmm2[6],xmm9[6],xmm2[7],xmm9[7]
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm15 = xmm15[8],xmm9[8],xmm15[9],xmm9[9],xmm15[10],xmm9[10],xmm15[11],xmm9[11],xmm15[12],xmm9[12],xmm15[13],xmm9[13],xmm15[14],xmm9[14],xmm15[15],xmm9[15]
-; SSE-NEXT:    movdqa %xmm15, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[3,0]
-; SSE-NEXT:    movaps %xmm0, %xmm2
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm15[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm15[2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm2[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,2]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,1,4,5,6,7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm9[8],xmm14[9],xmm9[9],xmm14[10],xmm9[10],xmm14[11],xmm9[11],xmm14[12],xmm9[12],xmm14[13],xmm9[13],xmm14[14],xmm9[14],xmm14[15],xmm9[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[0,3,2,1]
+; SSE-NEXT:    movdqa %xmm15, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm2[3,0]
+; SSE-NEXT:    movaps %xmm2, %xmm4
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,0],xmm15[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm15[2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm4[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,2]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    packuswb %xmm2, %xmm3
+; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm9[0],xmm2[1],xmm9[1],xmm2[2],xmm9[2],xmm2[3],xmm9[3],xmm2[4],xmm9[4],xmm2[5],xmm9[5],xmm2[6],xmm9[6],xmm2[7],xmm9[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,1,4,5,6,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm9[8],xmm14[9],xmm9[9],xmm14[10],xmm9[10],xmm14[11],xmm9[11],xmm14[12],xmm9[12],xmm14[13],xmm9[13],xmm14[14],xmm9[14],xmm14[15],xmm9[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm14[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [0,65535,65535,0,65535,65535,65535,65535]
-; SSE-NEXT:    pand %xmm5, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm5
-; SSE-NEXT:    por %xmm2, %xmm5
-; SSE-NEXT:    pand %xmm3, %xmm1
+; SSE-NEXT:    pand %xmm5, %xmm4
+; SSE-NEXT:    pandn %xmm2, %xmm5
+; SSE-NEXT:    por %xmm4, %xmm5
+; SSE-NEXT:    pand %xmm0, %xmm3
 ; SSE-NEXT:    packuswb %xmm5, %xmm5
-; SSE-NEXT:    pandn %xmm5, %xmm3
-; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm9[8],xmm0[9],xmm9[9],xmm0[10],xmm9[10],xmm0[11],xmm9[11],xmm0[12],xmm9[12],xmm0[13],xmm9[13],xmm0[14],xmm9[14],xmm0[15],xmm9[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,5]
+; SSE-NEXT:    pandn %xmm5, %xmm0
+; SSE-NEXT:    por %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm13, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm9[8],xmm2[9],xmm9[9],xmm2[10],xmm9[10],xmm2[11],xmm9[11],xmm2[12],xmm9[12],xmm2[13],xmm9[13],xmm2[14],xmm9[14],xmm2[15],xmm9[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,5]
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm9[0],xmm13[1],xmm9[1],xmm13[2],xmm9[2],xmm13[3],xmm9[3],xmm13[4],xmm9[4],xmm13[5],xmm9[5],xmm13[6],xmm9[6],xmm13[7],xmm9[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[0,2,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,7,7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    pand %xmm2, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    pand %xmm11, %xmm3
-; SSE-NEXT:    packuswb %xmm2, %xmm0
-; SSE-NEXT:    pandn %xmm0, %xmm11
-; SSE-NEXT:    por %xmm3, %xmm11
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm2, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[0,2,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,7,7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    pand %xmm4, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm4
+; SSE-NEXT:    por %xmm3, %xmm4
+; SSE-NEXT:    pand %xmm7, %xmm0
+; SSE-NEXT:    packuswb %xmm4, %xmm2
+; SSE-NEXT:    pandn %xmm2, %xmm7
+; SSE-NEXT:    por %xmm0, %xmm7
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    pand %xmm8, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
-; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[3,1,2,0]
-; SSE-NEXT:    pand %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm2, %xmm5
+; SSE-NEXT:    packuswb %xmm2, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm12, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[3,1,2,0]
+; SSE-NEXT:    pand %xmm8, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm0[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pand %xmm3, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [65535,0,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm6, %xmm1
-; SSE-NEXT:    pand %xmm13, %xmm1
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm0[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm3, %xmm3
+; SSE-NEXT:    pand %xmm12, %xmm3
+; SSE-NEXT:    por %xmm2, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pand %xmm11, %xmm2
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,0,0,0]
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm13
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[0,2,1,3]
-; SSE-NEXT:    pand %xmm5, %xmm1
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,2,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm0, %xmm6
-; SSE-NEXT:    pandn %xmm1, %xmm6
-; SSE-NEXT:    por %xmm2, %xmm6
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm9[8],xmm1[9],xmm9[9],xmm1[10],xmm9[10],xmm1[11],xmm9[11],xmm1[12],xmm9[12],xmm1[13],xmm9[13],xmm1[14],xmm9[14],xmm1[15],xmm9[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1],xmm7[2],xmm9[2],xmm7[3],xmm9[3],xmm7[4],xmm9[4],xmm7[5],xmm9[5],xmm7[6],xmm9[6],xmm7[7],xmm9[7]
-; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[1,0],xmm1[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm1[2,3]
-; SSE-NEXT:    psrlq $48, %xmm1
-; SSE-NEXT:    psrldq {{.*#+}} xmm2 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm7[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,5,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm9[8],xmm2[9],xmm9[9],xmm2[10],xmm9[10],xmm2[11],xmm9[11],xmm2[12],xmm9[12],xmm2[13],xmm9[13],xmm2[14],xmm9[14],xmm2[15],xmm9[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm2[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,0,65535,65535,0,65535,65535,65535]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3],xmm4[4],xmm9[4],xmm4[5],xmm9[5],xmm4[6],xmm9[6],xmm4[7],xmm9[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,1,1,2,4,5,6,7]
-; SSE-NEXT:    pand %xmm2, %xmm4
-; SSE-NEXT:    pandn %xmm5, %xmm2
-; SSE-NEXT:    por %xmm4, %xmm2
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pand %xmm3, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    pand %xmm0, %xmm3
+; SSE-NEXT:    por %xmm2, %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm11[0,2,1,3]
+; SSE-NEXT:    pand %xmm8, %xmm2
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,2,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
+; SSE-NEXT:    packuswb %xmm2, %xmm4
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm4, %xmm2
 ; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    movdqa %xmm13, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1],xmm1[2],xmm9[2],xmm1[3],xmm9[3],xmm1[4],xmm9[4],xmm1[5],xmm9[5],xmm1[6],xmm9[6],xmm1[7],xmm9[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,1,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm9[8],xmm13[9],xmm9[9],xmm13[10],xmm9[10],xmm13[11],xmm9[11],xmm13[12],xmm9[12],xmm13[13],xmm9[13],xmm13[14],xmm9[14],xmm13[15],xmm9[15]
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,65535,0,65535,0,0]
-; SSE-NEXT:    pand %xmm3, %xmm1
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm13[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,7,4]
-; SSE-NEXT:    pandn %xmm4, %xmm3
-; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    packuswb %xmm3, %xmm1
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm12, (%rsi)
+; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm9[8],xmm3[9],xmm9[9],xmm3[10],xmm9[10],xmm3[11],xmm9[11],xmm3[12],xmm9[12],xmm3[13],xmm9[13],xmm3[14],xmm9[14],xmm3[15],xmm9[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm9[0],xmm5[1],xmm9[1],xmm5[2],xmm9[2],xmm5[3],xmm9[3],xmm5[4],xmm9[4],xmm5[5],xmm9[5],xmm5[6],xmm9[6],xmm5[7],xmm9[7]
+; SSE-NEXT:    movdqa %xmm5, %xmm4
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,0],xmm3[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm3[2,3]
+; SSE-NEXT:    psrlq $48, %xmm3
+; SSE-NEXT:    psrldq {{.*#+}} xmm4 = xmm4[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm5[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,5,7]
+; SSE-NEXT:    packuswb %xmm4, %xmm3
+; SSE-NEXT:    movdqa %xmm6, %xmm4
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm9[8],xmm4[9],xmm9[9],xmm4[10],xmm9[10],xmm4[11],xmm9[11],xmm4[12],xmm9[12],xmm4[13],xmm9[13],xmm4[14],xmm9[14],xmm4[15],xmm9[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[1,1,2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,0,65535,65535,0,65535,65535,65535]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm9[0],xmm6[1],xmm9[1],xmm6[2],xmm9[2],xmm6[3],xmm9[3],xmm6[4],xmm9[4],xmm6[5],xmm9[5],xmm6[6],xmm9[6],xmm6[7],xmm9[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[3,1,1,2,4,5,6,7]
+; SSE-NEXT:    pand %xmm5, %xmm6
+; SSE-NEXT:    pandn %xmm4, %xmm5
+; SSE-NEXT:    por %xmm6, %xmm5
+; SSE-NEXT:    packuswb %xmm5, %xmm5
+; SSE-NEXT:    pand %xmm12, %xmm5
+; SSE-NEXT:    pandn %xmm3, %xmm12
+; SSE-NEXT:    por %xmm5, %xmm12
+; SSE-NEXT:    movdqa %xmm11, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1],xmm3[2],xmm9[2],xmm3[3],xmm9[3],xmm3[4],xmm9[4],xmm3[5],xmm9[5],xmm3[6],xmm9[6],xmm3[7],xmm9[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,1,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm9[8],xmm11[9],xmm9[9],xmm11[10],xmm9[10],xmm11[11],xmm9[11],xmm11[12],xmm9[12],xmm11[13],xmm9[13],xmm11[14],xmm9[14],xmm11[15],xmm9[15]
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,65535,0,65535,0,0]
+; SSE-NEXT:    pand %xmm4, %xmm3
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm11[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,5,7,4]
+; SSE-NEXT:    pandn %xmm5, %xmm4
+; SSE-NEXT:    por %xmm3, %xmm4
+; SSE-NEXT:    pand %xmm0, %xmm12
+; SSE-NEXT:    packuswb %xmm4, %xmm3
+; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    por %xmm12, %xmm0
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, (%rsi)
 ; SSE-NEXT:    movdqa %xmm10, (%rdx)
-; SSE-NEXT:    movdqa %xmm8, (%rcx)
-; SSE-NEXT:    movdqa %xmm11, (%r8)
-; SSE-NEXT:    movdqa %xmm6, (%r9)
+; SSE-NEXT:    movdqa %xmm1, (%rcx)
+; SSE-NEXT:    movdqa %xmm7, (%r8)
+; SSE-NEXT:    movdqa %xmm2, (%r9)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; SSE-NEXT:    movdqa %xmm0, (%rax)
 ; SSE-NEXT:    retq
@@ -2542,229 +2542,229 @@ define void @load_i8_stride6_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i8_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5) nounwind {
 ; SSE-LABEL: load_i8_stride6_vf32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $264, %rsp # imm = 0x108
-; SSE-NEXT:    movdqa 64(%rdi), %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 80(%rdi), %xmm9
-; SSE-NEXT:    movdqa (%rdi), %xmm12
-; SSE-NEXT:    movdqa 16(%rdi), %xmm14
+; SSE-NEXT:    subq $280, %rsp # imm = 0x118
+; SSE-NEXT:    movdqa 64(%rdi), %xmm14
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 80(%rdi), %xmm10
+; SSE-NEXT:    movdqa (%rdi), %xmm6
+; SSE-NEXT:    movdqa 16(%rdi), %xmm13
 ; SSE-NEXT:    movdqa 32(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 48(%rdi), %xmm5
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,0,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm4, %xmm0
 ; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,0,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm11, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,0,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm12, %xmm1
 ; SSE-NEXT:    pandn %xmm5, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    movdqa %xmm4, %xmm1
 ; SSE-NEXT:    pandn %xmm5, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm5, %xmm15
-; SSE-NEXT:    pand %xmm10, %xmm15
+; SSE-NEXT:    pand %xmm4, %xmm15
 ; SSE-NEXT:    por %xmm0, %xmm15
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
 ; SSE-NEXT:    movdqa %xmm15, %xmm0
 ; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm1, %xmm2
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,5]
 ; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,65535,65535,0,0,0,65535,65535]
 ; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    pandn %xmm14, %xmm1
-; SSE-NEXT:    movdqa %xmm12, %xmm8
-; SSE-NEXT:    pand %xmm11, %xmm8
-; SSE-NEXT:    por %xmm1, %xmm8
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[0,2,1,3]
-; SSE-NEXT:    pand %xmm3, %xmm1
-; SSE-NEXT:    movdqa %xmm3, %xmm6
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    packuswb %xmm1, %xmm1
+; SSE-NEXT:    pandn %xmm0, %xmm1
+; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    pandn %xmm13, %xmm0
+; SSE-NEXT:    movdqa %xmm6, %xmm9
+; SSE-NEXT:    pand %xmm12, %xmm9
+; SSE-NEXT:    por %xmm0, %xmm9
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[0,2,1,3]
+; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    packuswb %xmm0, %xmm0
+; SSE-NEXT:    pand %xmm11, %xmm0
+; SSE-NEXT:    por %xmm1, %xmm0
+; SSE-NEXT:    movdqa %xmm4, %xmm1
+; SSE-NEXT:    pandn %xmm10, %xmm1
+; SSE-NEXT:    pand %xmm4, %xmm14
+; SSE-NEXT:    por %xmm1, %xmm14
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[3,1,2,0]
 ; SSE-NEXT:    pand %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm2, %xmm3
 ; SSE-NEXT:    movdqa %xmm2, %xmm5
-; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    por %xmm3, %xmm1
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pandn %xmm9, %xmm0
-; SSE-NEXT:    pand %xmm10, %xmm7
-; SSE-NEXT:    por %xmm0, %xmm7
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[3,1,2,0]
-; SSE-NEXT:    pand %xmm6, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
+; SSE-NEXT:    packuswb %xmm1, %xmm1
 ; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
 ; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pand %xmm3, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    pand %xmm3, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 128(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    movdqa %xmm4, %xmm0
 ; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa 144(%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm11, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    movdqa 144(%rdi), %xmm7
+; SSE-NEXT:    movdqa %xmm12, %xmm2
+; SSE-NEXT:    pandn %xmm7, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    pandn %xmm7, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    pand %xmm10, %xmm2
-; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    pand %xmm4, %xmm7
+; SSE-NEXT:    por %xmm0, %xmm7
+; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    pand %xmm5, %xmm0
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,5]
 ; SSE-NEXT:    packuswb %xmm3, %xmm0
-; SSE-NEXT:    movdqa %xmm5, %xmm6
-; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    movdqa %xmm10, %xmm1
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pandn %xmm12, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 112(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm11, %xmm3
-; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    movdqa 160(%rdi), %xmm5
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm10, %xmm5
-; SSE-NEXT:    movdqa %xmm10, %xmm4
-; SSE-NEXT:    pandn %xmm14, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm10, %xmm12
-; SSE-NEXT:    movdqa %xmm11, %xmm4
-; SSE-NEXT:    pandn %xmm9, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm11
-; SSE-NEXT:    pand %xmm10, %xmm11
-; SSE-NEXT:    movdqa %xmm10, %xmm4
-; SSE-NEXT:    pandn %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 96(%rdi), %xmm13
-; SSE-NEXT:    movdqa %xmm13, %xmm4
-; SSE-NEXT:    pand %xmm10, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 176(%rdi), %xmm4
-; SSE-NEXT:    movdqa %xmm4, %xmm10
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm1, %xmm10
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm9, %xmm10
-; SSE-NEXT:    pand %xmm1, %xmm9
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm1, %xmm14
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm14, %xmm9
-; SSE-NEXT:    pand %xmm1, %xmm14
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm1, %xmm14
+; SSE-NEXT:    movdqa %xmm11, %xmm8
+; SSE-NEXT:    pandn %xmm0, %xmm8
+; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    movdqa %xmm4, %xmm1
+; SSE-NEXT:    pandn %xmm6, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm1, (%rsp) # 16-byte Spill
-; SSE-NEXT:    pandn %xmm13, %xmm1
+; SSE-NEXT:    movdqa 112(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm12, %xmm4
+; SSE-NEXT:    pandn %xmm1, %xmm4
+; SSE-NEXT:    movdqa 160(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm13, %xmm5
+; SSE-NEXT:    pandn %xmm13, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm6
+; SSE-NEXT:    movdqa %xmm6, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm12, %xmm2
+; SSE-NEXT:    pandn %xmm10, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm10, %xmm12
+; SSE-NEXT:    pand %xmm0, %xmm12
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 96(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, %xmm6
+; SSE-NEXT:    pand %xmm0, %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 176(%rdi), %xmm6
+; SSE-NEXT:    movdqa %xmm6, %xmm13
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm13
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm10, %xmm13
+; SSE-NEXT:    pand %xmm0, %xmm10
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm5, %xmm10
+; SSE-NEXT:    pand %xmm0, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm0, %xmm5
+; SSE-NEXT:    pand %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm2, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
-; SSE-NEXT:    por %xmm3, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,2,1,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    packuswb %xmm3, %xmm3
-; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3
-; SSE-NEXT:    por %xmm6, %xmm3
-; SSE-NEXT:    pandn %xmm4, %xmm14
-; SSE-NEXT:    por %xmm14, %xmm5
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[3,1,2,0]
-; SSE-NEXT:    pand %xmm0, %xmm4
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,3,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,5]
+; SSE-NEXT:    por %xmm4, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[0,2,1,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm2, %xmm4
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,1,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    packuswb %xmm4, %xmm4
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    pandn %xmm4, %xmm0
-; SSE-NEXT:    pand %xmm13, %xmm3
-; SSE-NEXT:    por %xmm3, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pxor %xmm4, %xmm4
-; SSE-NEXT:    movdqa %xmm15, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm4[8],xmm3[9],xmm4[9],xmm3[10],xmm4[10],xmm3[11],xmm4[11],xmm3[12],xmm4[12],xmm3[13],xmm4[13],xmm3[14],xmm4[14],xmm3[15],xmm4[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm4[0],xmm15[1],xmm4[1],xmm15[2],xmm4[2],xmm15[3],xmm4[3],xmm15[4],xmm4[4],xmm15[5],xmm4[5],xmm15[6],xmm4[6],xmm15[7],xmm4[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm15[2,2,3,3]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm3[0],xmm14[1],xmm3[1],xmm14[2],xmm3[2],xmm14[3],xmm3[3]
-; SSE-NEXT:    psrld $16, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm15[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,5,7,6,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm3[2],xmm15[3],xmm3[3]
-; SSE-NEXT:    packuswb %xmm15, %xmm14
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm6, %xmm3
-; SSE-NEXT:    pandn %xmm14, %xmm3
-; SSE-NEXT:    movdqa %xmm8, %xmm14
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm4[8],xmm14[9],xmm4[9],xmm14[10],xmm4[10],xmm14[11],xmm4[11],xmm14[12],xmm4[12],xmm14[13],xmm4[13],xmm14[14],xmm4[14],xmm14[15],xmm4[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm14[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm14[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,5,7,6,7]
+; SSE-NEXT:    pand %xmm11, %xmm4
+; SSE-NEXT:    por %xmm8, %xmm4
+; SSE-NEXT:    pandn %xmm6, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[3,1,2,0]
+; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
+; SSE-NEXT:    packuswb %xmm0, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
+; SSE-NEXT:    movdqa %xmm2, %xmm5
+; SSE-NEXT:    pandn %xmm0, %xmm5
+; SSE-NEXT:    pand %xmm2, %xmm4
+; SSE-NEXT:    por %xmm4, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pxor %xmm4, %xmm4
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm4[8],xmm0[9],xmm4[9],xmm0[10],xmm4[10],xmm0[11],xmm4[11],xmm0[12],xmm4[12],xmm0[13],xmm4[13],xmm0[14],xmm4[14],xmm0[15],xmm4[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm4[0],xmm15[1],xmm4[1],xmm15[2],xmm4[2],xmm15[3],xmm4[3],xmm15[4],xmm4[4],xmm15[5],xmm4[5],xmm15[6],xmm4[6],xmm15[7],xmm4[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm15[2,2,3,3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm0[0],xmm6[1],xmm0[1],xmm6[2],xmm0[2],xmm6[3],xmm0[3]
+; SSE-NEXT:    psrld $16, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm15[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,5,7,6,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm0[2],xmm15[3],xmm0[3]
+; SSE-NEXT:    packuswb %xmm15, %xmm6
+; SSE-NEXT:    movdqa %xmm11, %xmm5
+; SSE-NEXT:    pandn %xmm6, %xmm5
+; SSE-NEXT:    movdqa %xmm9, %xmm6
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm4[8],xmm6[9],xmm4[9],xmm6[10],xmm4[10],xmm6[11],xmm4[11],xmm6[12],xmm4[12],xmm6[13],xmm4[13],xmm6[14],xmm4[14],xmm6[15],xmm4[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,5,7,6,7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [65535,65535,0,65535,0,0,65535,65535]
 ; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pandn %xmm14, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm4[0],xmm8[1],xmm4[1],xmm8[2],xmm4[2],xmm8[3],xmm4[3],xmm8[4],xmm4[4],xmm8[5],xmm4[5],xmm8[6],xmm4[6],xmm8[7],xmm4[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm8[1,3,2,0,4,5,6,7]
-; SSE-NEXT:    pand %xmm15, %xmm14
-; SSE-NEXT:    por %xmm0, %xmm14
-; SSE-NEXT:    packuswb %xmm14, %xmm14
-; SSE-NEXT:    pand %xmm6, %xmm14
-; SSE-NEXT:    por %xmm3, %xmm14
-; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    pandn %xmm6, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm4[0],xmm9[1],xmm4[1],xmm9[2],xmm4[2],xmm9[3],xmm4[3],xmm9[4],xmm4[4],xmm9[5],xmm4[5],xmm9[6],xmm4[6],xmm9[7],xmm4[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm9[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[1,3,2,0,4,5,6,7]
+; SSE-NEXT:    pand %xmm15, %xmm6
+; SSE-NEXT:    por %xmm0, %xmm6
+; SSE-NEXT:    packuswb %xmm6, %xmm6
+; SSE-NEXT:    pand %xmm11, %xmm6
+; SSE-NEXT:    por %xmm5, %xmm6
+; SSE-NEXT:    movdqa %xmm14, %xmm0
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,65535,65535,0,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm8, %xmm3
-; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm4[8],xmm7[9],xmm4[9],xmm7[10],xmm4[10],xmm7[11],xmm4[11],xmm7[12],xmm4[12],xmm7[13],xmm4[13],xmm7[14],xmm4[14],xmm7[15],xmm4[15]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm7[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm8, %xmm5
+; SSE-NEXT:    pandn %xmm0, %xmm5
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm4[8],xmm14[9],xmm4[9],xmm14[10],xmm4[10],xmm14[11],xmm4[11],xmm14[12],xmm4[12],xmm14[13],xmm4[13],xmm14[14],xmm4[14],xmm14[15],xmm4[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm14[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,7,6,4]
 ; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    por %xmm3, %xmm0
+; SSE-NEXT:    por %xmm5, %xmm0
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm13, %xmm3
-; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    pand %xmm13, %xmm14
-; SSE-NEXT:    por %xmm14, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    movdqa %xmm2, %xmm5
+; SSE-NEXT:    pandn %xmm0, %xmm5
+; SSE-NEXT:    pand %xmm2, %xmm6
+; SSE-NEXT:    movdqa %xmm2, %xmm9
+; SSE-NEXT:    por %xmm6, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm7, %xmm0
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm4[8],xmm0[9],xmm4[9],xmm0[10],xmm4[10],xmm0[11],xmm4[11],xmm0[12],xmm4[12],xmm0[13],xmm4[13],xmm0[14],xmm4[14],xmm0[15],xmm4[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3],xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[2,2,3,3]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1],xmm7[2],xmm4[2],xmm7[3],xmm4[3],xmm7[4],xmm4[4],xmm7[5],xmm4[5],xmm7[6],xmm4[6],xmm7[7],xmm4[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm7[2,2,3,3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
 ; SSE-NEXT:    psrld $16, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,7,6,7]
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
-; SSE-NEXT:    packuswb %xmm2, %xmm3
+; SSE-NEXT:    packuswb %xmm2, %xmm5
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm4[8],xmm0[9],xmm4[9],xmm0[10],xmm4[10],xmm0[11],xmm4[11],xmm0[12],xmm4[12],xmm0[13],xmm4[13],xmm0[14],xmm4[14],xmm0[15],xmm4[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
@@ -2778,32 +2778,31 @@ define void @load_i8_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pandn %xmm0, %xmm15
 ; SSE-NEXT:    por %xmm1, %xmm15
 ; SSE-NEXT:    packuswb %xmm15, %xmm15
-; SSE-NEXT:    pand %xmm6, %xmm15
-; SSE-NEXT:    pandn %xmm3, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm15
-; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    pand %xmm11, %xmm15
+; SSE-NEXT:    pandn %xmm5, %xmm11
+; SSE-NEXT:    por %xmm15, %xmm11
+; SSE-NEXT:    movdqa %xmm3, %xmm0
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm4[8],xmm5[9],xmm4[9],xmm5[10],xmm4[10],xmm5[11],xmm4[11],xmm5[12],xmm4[12],xmm5[13],xmm4[13],xmm5[14],xmm4[14],xmm5[15],xmm4[15]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm5[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm4[8],xmm3[9],xmm4[9],xmm3[10],xmm4[10],xmm3[11],xmm4[11],xmm3[12],xmm4[12],xmm3[13],xmm4[13],xmm3[14],xmm4[14],xmm3[15],xmm4[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm3[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,7,6,4]
 ; SSE-NEXT:    pand %xmm8, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm8
 ; SSE-NEXT:    por %xmm1, %xmm8
 ; SSE-NEXT:    packuswb %xmm8, %xmm0
-; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    movdqa %xmm9, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm13, %xmm15
-; SSE-NEXT:    movdqa %xmm13, %xmm7
-; SSE-NEXT:    por %xmm15, %xmm1
+; SSE-NEXT:    pand %xmm9, %xmm11
+; SSE-NEXT:    por %xmm11, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,0,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    pand %xmm5, %xmm10
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    pand %xmm5, %xmm13
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm8, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,3,4,5,6,7]
@@ -2811,10 +2810,11 @@ define void @load_i8_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,4,4,5,6]
 ; SSE-NEXT:    packuswb %xmm1, %xmm2
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm12[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa (%rsp), %xmm6 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm6[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    pand %xmm8, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,5,5,5,5]
@@ -2824,26 +2824,25 @@ define void @load_i8_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pandn %xmm1, %xmm3
 ; SSE-NEXT:    pand %xmm0, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm14, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm11
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm11[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    por %xmm1, %xmm12
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm12[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pand %xmm15, %xmm1
+; SSE-NEXT:    pand %xmm8, %xmm1
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm13, %xmm2
+; SSE-NEXT:    movdqa %xmm9, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pand %xmm13, %xmm3
+; SSE-NEXT:    pand %xmm9, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm5, %xmm9
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm9, %xmm1
-; SSE-NEXT:    pand %xmm15, %xmm1
+; SSE-NEXT:    pand %xmm5, %xmm10
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    pand %xmm8, %xmm1
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,3,4,5,6,7]
@@ -2851,11 +2850,11 @@ define void @load_i8_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,5,6]
 ; SSE-NEXT:    packuswb %xmm2, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm13[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm14[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pand %xmm15, %xmm2
+; SSE-NEXT:    pand %xmm8, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
@@ -2864,306 +2863,305 @@ define void @load_i8_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pandn %xmm2, %xmm3
 ; SSE-NEXT:    pand %xmm0, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    movdqa (%rsp), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    por %xmm1, %xmm8
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm8[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    por %xmm1, %xmm11
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm11[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pand %xmm15, %xmm1
+; SSE-NEXT:    pand %xmm8, %xmm1
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,7,4]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    movdqa %xmm9, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pand %xmm7, %xmm3
+; SSE-NEXT:    pand %xmm9, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    movdqa %xmm13, %xmm1
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm4[8],xmm10[9],xmm4[9],xmm10[10],xmm4[10],xmm10[11],xmm4[11],xmm10[12],xmm4[12],xmm10[13],xmm4[13],xmm10[14],xmm4[14],xmm10[15],xmm4[15]
-; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm4[8],xmm13[9],xmm4[9],xmm13[10],xmm4[10],xmm13[11],xmm4[11],xmm13[12],xmm4[12],xmm13[13],xmm4[13],xmm13[14],xmm4[14],xmm13[15],xmm4[15]
+; SSE-NEXT:    movdqa %xmm13, %xmm2
 ; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[3,0]
 ; SSE-NEXT:    movaps %xmm1, %xmm3
 ; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[0,2]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm10[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm10[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,0],xmm13[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm13[2,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,7,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,3,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    movdqa %xmm6, %xmm1
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm1[3,1,2,1,4,5,6,7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [0,65535,65535,0,65535,65535,65535,65535]
 ; SSE-NEXT:    movdqa %xmm1, %xmm5
 ; SSE-NEXT:    pandn %xmm3, %xmm5
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm4[8],xmm12[9],xmm4[9],xmm12[10],xmm4[10],xmm12[11],xmm4[11],xmm12[12],xmm4[12],xmm12[13],xmm4[13],xmm12[14],xmm4[14],xmm12[15],xmm4[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm12[0,3,2,1]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm4[8],xmm6[9],xmm4[9],xmm6[10],xmm4[10],xmm6[11],xmm4[11],xmm6[12],xmm4[12],xmm6[13],xmm4[13],xmm6[14],xmm4[14],xmm6[15],xmm4[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm6[0,3,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pand %xmm1, %xmm3
 ; SSE-NEXT:    por %xmm5, %xmm3
 ; SSE-NEXT:    packuswb %xmm3, %xmm3
-; SSE-NEXT:    movdqa %xmm0, %xmm5
-; SSE-NEXT:    pandn %xmm3, %xmm5
+; SSE-NEXT:    movdqa %xmm0, %xmm6
+; SSE-NEXT:    pandn %xmm3, %xmm6
 ; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm5
-; SSE-NEXT:    movdqa %xmm11, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm6
+; SSE-NEXT:    movdqa %xmm12, %xmm2
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm4[8],xmm2[9],xmm4[9],xmm2[10],xmm4[10],xmm2[11],xmm4[11],xmm2[12],xmm4[12],xmm2[13],xmm4[13],xmm2[14],xmm4[14],xmm2[15],xmm4[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm2[0,1,2,3,7,5,6,5]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm2, %xmm6
-; SSE-NEXT:    pandn %xmm3, %xmm6
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm4[0],xmm11[1],xmm4[1],xmm11[2],xmm4[2],xmm11[3],xmm4[3],xmm11[4],xmm4[4],xmm11[5],xmm4[5],xmm11[6],xmm4[6],xmm11[7],xmm4[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[0,2,0,3]
+; SSE-NEXT:    movdqa %xmm2, %xmm5
+; SSE-NEXT:    pandn %xmm3, %xmm5
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm4[0],xmm12[1],xmm4[1],xmm12[2],xmm4[2],xmm12[3],xmm4[3],xmm12[4],xmm4[4],xmm12[5],xmm4[5],xmm12[6],xmm4[6],xmm12[7],xmm4[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm12[0,2,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,7,7]
 ; SSE-NEXT:    pand %xmm2, %xmm3
-; SSE-NEXT:    por %xmm6, %xmm3
+; SSE-NEXT:    por %xmm5, %xmm3
 ; SSE-NEXT:    packuswb %xmm3, %xmm3
-; SSE-NEXT:    movdqa %xmm7, %xmm6
-; SSE-NEXT:    pandn %xmm3, %xmm6
-; SSE-NEXT:    pand %xmm7, %xmm5
-; SSE-NEXT:    por %xmm5, %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    movdqa %xmm9, %xmm15
+; SSE-NEXT:    pandn %xmm3, %xmm15
+; SSE-NEXT:    pand %xmm9, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm15
+; SSE-NEXT:    movdqa %xmm10, %xmm3
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm4[8],xmm9[9],xmm4[9],xmm9[10],xmm4[10],xmm9[11],xmm4[11],xmm9[12],xmm4[12],xmm9[13],xmm4[13],xmm9[14],xmm4[14],xmm9[15],xmm4[15]
-; SSE-NEXT:    movdqa %xmm9, %xmm5
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm3[3,0]
-; SSE-NEXT:    movaps %xmm3, %xmm6
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm5[0,2]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,0],xmm9[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm9[2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm6[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,0,2]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm4[8],xmm10[9],xmm4[9],xmm10[10],xmm4[10],xmm10[11],xmm4[11],xmm10[12],xmm4[12],xmm10[13],xmm4[13],xmm10[14],xmm4[14],xmm10[15],xmm4[15]
+; SSE-NEXT:    movdqa %xmm10, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0],xmm3[3,0]
+; SSE-NEXT:    movaps %xmm3, %xmm7
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm6[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,0],xmm10[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm10[2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm7[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,0,2]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,3,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm3, %xmm5
-; SSE-NEXT:    movdqa %xmm13, %xmm3
+; SSE-NEXT:    packuswb %xmm3, %xmm6
+; SSE-NEXT:    movdqa %xmm14, %xmm3
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,1,4,5,6,7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm4[8],xmm13[9],xmm4[9],xmm13[10],xmm4[10],xmm13[11],xmm4[11],xmm13[12],xmm4[12],xmm13[13],xmm4[13],xmm13[14],xmm4[14],xmm13[15],xmm4[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm13[0,3,2,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,1,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pand %xmm1, %xmm6
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm4[8],xmm14[9],xmm4[9],xmm14[10],xmm4[10],xmm14[11],xmm4[11],xmm14[12],xmm4[12],xmm14[13],xmm4[13],xmm14[14],xmm4[14],xmm14[15],xmm4[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm14[0,3,2,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pand %xmm1, %xmm7
 ; SSE-NEXT:    pandn %xmm3, %xmm1
-; SSE-NEXT:    por %xmm6, %xmm1
-; SSE-NEXT:    pand %xmm0, %xmm5
+; SSE-NEXT:    por %xmm7, %xmm1
+; SSE-NEXT:    pand %xmm0, %xmm6
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
 ; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    por %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    por %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm11, %xmm1
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm4[8],xmm1[9],xmm4[9],xmm1[10],xmm4[10],xmm1[11],xmm4[11],xmm1[12],xmm4[12],xmm1[13],xmm4[13],xmm1[14],xmm4[14],xmm1[15],xmm4[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,5,6,5]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm4[0],xmm8[1],xmm4[1],xmm8[2],xmm4[2],xmm8[3],xmm4[3],xmm8[4],xmm4[4],xmm8[5],xmm4[5],xmm8[6],xmm4[6],xmm8[7],xmm4[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm8[0,2,0,3]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm4[0],xmm11[1],xmm4[1],xmm11[2],xmm4[2],xmm11[3],xmm4[3],xmm11[4],xmm4[4],xmm11[5],xmm4[5],xmm11[6],xmm4[6],xmm11[7],xmm4[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[0,2,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,7,7]
 ; SSE-NEXT:    pand %xmm2, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm2
 ; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    movdqa %xmm7, %xmm13
-; SSE-NEXT:    pand %xmm7, %xmm0
+; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    pand %xmm9, %xmm0
 ; SSE-NEXT:    packuswb %xmm2, %xmm1
-; SSE-NEXT:    pandn %xmm1, %xmm13
-; SSE-NEXT:    por %xmm0, %xmm13
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    por %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    pand %xmm8, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,2]
-; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[3,1,2,0]
-; SSE-NEXT:    pand %xmm15, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm0[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,1,0,2]
+; SSE-NEXT:    packuswb %xmm1, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    pandn %xmm2, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm10[3,1,2,0]
+; SSE-NEXT:    pand %xmm8, %xmm2
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm2[2,1,0,3,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm3, %xmm3
-; SSE-NEXT:    pand %xmm2, %xmm3
-; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm14, %xmm11
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,0,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    pand %xmm12, %xmm11
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[0,2,1,3]
-; SSE-NEXT:    pand %xmm15, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm5
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,0,0,0]
-; SSE-NEXT:    movdqa %xmm0, %xmm8
-; SSE-NEXT:    pandn %xmm5, %xmm8
 ; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    por %xmm3, %xmm8
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [65535,0,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    pand %xmm13, %xmm12
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[0,2,1,3]
+; SSE-NEXT:    pand %xmm8, %xmm1
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,2,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
+; SSE-NEXT:    packuswb %xmm1, %xmm6
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,65535,65535,0,0,0]
+; SSE-NEXT:    movdqa %xmm1, %xmm5
+; SSE-NEXT:    pandn %xmm6, %xmm5
+; SSE-NEXT:    pand %xmm1, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm5
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm14, %xmm3
-; SSE-NEXT:    pand %xmm15, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm8, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm3[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,2]
-; SSE-NEXT:    packuswb %xmm5, %xmm3
-; SSE-NEXT:    movdqa %xmm2, %xmm5
-; SSE-NEXT:    pandn %xmm3, %xmm5
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[3,1,2,0]
-; SSE-NEXT:    pand %xmm15, %xmm3
+; SSE-NEXT:    packuswb %xmm6, %xmm3
+; SSE-NEXT:    movdqa %xmm0, %xmm6
+; SSE-NEXT:    pandn %xmm3, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[3,1,2,0]
+; SSE-NEXT:    pand %xmm8, %xmm3
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm3[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm6, %xmm6
-; SSE-NEXT:    pand %xmm2, %xmm6
-; SSE-NEXT:    por %xmm5, %xmm6
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm3[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm7, %xmm7
+; SSE-NEXT:    pand %xmm0, %xmm7
+; SSE-NEXT:    por %xmm6, %xmm7
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pand %xmm12, %xmm3
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; SSE-NEXT:    por %xmm3, %xmm12
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm12[0,2,1,3]
-; SSE-NEXT:    pand %xmm15, %xmm3
+; SSE-NEXT:    pand %xmm13, %xmm3
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    por %xmm3, %xmm13
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[0,2,1,3]
+; SSE-NEXT:    pand %xmm8, %xmm3
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,0,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    packuswb %xmm3, %xmm5
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    pandn %xmm5, %xmm3
-; SSE-NEXT:    pand %xmm0, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm3
-; SSE-NEXT:    movdqa %xmm7, %xmm5
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm4[8],xmm5[9],xmm4[9],xmm5[10],xmm4[10],xmm5[11],xmm4[11],xmm5[12],xmm4[12],xmm5[13],xmm4[13],xmm5[14],xmm4[14],xmm5[15],xmm4[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1],xmm7[2],xmm4[2],xmm7[3],xmm4[3],xmm7[4],xmm4[4],xmm7[5],xmm4[5],xmm7[6],xmm4[6],xmm7[7],xmm4[7]
-; SSE-NEXT:    movdqa %xmm7, %xmm6
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[1,0],xmm5[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm5[2,3]
-; SSE-NEXT:    psrlq $48, %xmm5
-; SSE-NEXT:    psrldq {{.*#+}} xmm6 = xmm6[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm7[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,5,7]
-; SSE-NEXT:    packuswb %xmm6, %xmm5
-; SSE-NEXT:    movdqa %xmm2, %xmm6
-; SSE-NEXT:    pandn %xmm5, %xmm6
-; SSE-NEXT:    movdqa %xmm9, %xmm5
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm4[8],xmm5[9],xmm4[9],xmm5[10],xmm4[10],xmm5[11],xmm4[11],xmm5[12],xmm4[12],xmm5[13],xmm4[13],xmm5[14],xmm4[14],xmm5[15],xmm4[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[1,1,2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm5[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,0,65535,65535,0,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm5, %xmm10
-; SSE-NEXT:    pandn %xmm7, %xmm10
+; SSE-NEXT:    packuswb %xmm3, %xmm6
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm6, %xmm3
+; SSE-NEXT:    pand %xmm1, %xmm7
+; SSE-NEXT:    por %xmm7, %xmm3
+; SSE-NEXT:    movdqa %xmm9, %xmm6
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm4[8],xmm6[9],xmm4[9],xmm6[10],xmm4[10],xmm6[11],xmm4[11],xmm6[12],xmm4[12],xmm6[13],xmm4[13],xmm6[14],xmm4[14],xmm6[15],xmm4[15]
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm4[0],xmm9[1],xmm4[1],xmm9[2],xmm4[2],xmm9[3],xmm4[3],xmm9[4],xmm4[4],xmm9[5],xmm4[5],xmm9[6],xmm4[6],xmm9[7],xmm4[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm9[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm7[3,1,1,2,4,5,6,7]
-; SSE-NEXT:    pand %xmm5, %xmm9
-; SSE-NEXT:    por %xmm10, %xmm9
-; SSE-NEXT:    packuswb %xmm9, %xmm9
-; SSE-NEXT:    pand %xmm2, %xmm9
-; SSE-NEXT:    por %xmm6, %xmm9
-; SSE-NEXT:    movdqa %xmm11, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm4[8],xmm11[9],xmm4[9],xmm11[10],xmm4[10],xmm11[11],xmm4[11],xmm11[12],xmm4[12],xmm11[13],xmm4[13],xmm11[14],xmm4[14],xmm11[15],xmm4[15]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm11[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm7[0,1,2,3,5,5,7,4]
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,65535,65535,0,65535,0,0]
-; SSE-NEXT:    movdqa %xmm7, %xmm11
-; SSE-NEXT:    pandn %xmm10, %xmm11
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1],xmm6[2],xmm4[2],xmm6[3],xmm4[3],xmm6[4],xmm4[4],xmm6[5],xmm4[5],xmm6[6],xmm4[6],xmm6[7],xmm4[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,3,1,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm7, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm11
-; SSE-NEXT:    packuswb %xmm11, %xmm10
-; SSE-NEXT:    movdqa %xmm0, %xmm6
-; SSE-NEXT:    pandn %xmm10, %xmm6
-; SSE-NEXT:    pand %xmm0, %xmm9
-; SSE-NEXT:    por %xmm9, %xmm6
-; SSE-NEXT:    movdqa %xmm14, %xmm11
-; SSE-NEXT:    movdqa %xmm14, %xmm9
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm4[8],xmm9[9],xmm4[9],xmm9[10],xmm4[10],xmm9[11],xmm4[11],xmm9[12],xmm4[12],xmm9[13],xmm4[13],xmm9[14],xmm4[14],xmm9[15],xmm4[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm4[0],xmm11[1],xmm4[1],xmm11[2],xmm4[2],xmm11[3],xmm4[3],xmm11[4],xmm4[4],xmm11[5],xmm4[5],xmm11[6],xmm4[6],xmm11[7],xmm4[7]
-; SSE-NEXT:    movdqa %xmm11, %xmm10
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[1,0],xmm9[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[2,0],xmm9[2,3]
-; SSE-NEXT:    psrlq $48, %xmm9
-; SSE-NEXT:    psrldq {{.*#+}} xmm10 = xmm10[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm11[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,4,5,7]
-; SSE-NEXT:    packuswb %xmm10, %xmm9
-; SSE-NEXT:    movdqa %xmm1, %xmm10
+; SSE-NEXT:    movdqa %xmm9, %xmm7
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[1,0],xmm6[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[2,0],xmm6[2,3]
+; SSE-NEXT:    psrlq $48, %xmm6
+; SSE-NEXT:    psrldq {{.*#+}} xmm7 = xmm7[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm9[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,5,7]
+; SSE-NEXT:    packuswb %xmm7, %xmm6
+; SSE-NEXT:    movdqa %xmm0, %xmm7
+; SSE-NEXT:    pandn %xmm6, %xmm7
+; SSE-NEXT:    movdqa %xmm10, %xmm6
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm4[8],xmm6[9],xmm4[9],xmm6[10],xmm4[10],xmm6[11],xmm4[11],xmm6[12],xmm4[12],xmm6[13],xmm4[13],xmm6[14],xmm4[14],xmm6[15],xmm4[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[1,1,2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm6[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,0,65535,65535,0,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm6, %xmm11
+; SSE-NEXT:    pandn %xmm9, %xmm11
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm4[0],xmm10[1],xmm4[1],xmm10[2],xmm4[2],xmm10[3],xmm4[3],xmm10[4],xmm4[4],xmm10[5],xmm4[5],xmm10[6],xmm4[6],xmm10[7],xmm4[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm10[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm9[3,1,1,2,4,5,6,7]
+; SSE-NEXT:    pand %xmm6, %xmm10
+; SSE-NEXT:    por %xmm11, %xmm10
+; SSE-NEXT:    packuswb %xmm10, %xmm10
+; SSE-NEXT:    pand %xmm0, %xmm10
+; SSE-NEXT:    por %xmm7, %xmm10
+; SSE-NEXT:    movdqa %xmm12, %xmm7
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm4[8],xmm12[9],xmm4[9],xmm12[10],xmm4[10],xmm12[11],xmm4[11],xmm12[12],xmm4[12],xmm12[13],xmm4[13],xmm12[14],xmm4[14],xmm12[15],xmm4[15]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm12[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm9[0,1,2,3,5,5,7,4]
+; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,65535,65535,0,65535,0,0]
+; SSE-NEXT:    movdqa %xmm9, %xmm12
+; SSE-NEXT:    pandn %xmm11, %xmm12
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1],xmm7[2],xmm4[2],xmm7[3],xmm4[3],xmm7[4],xmm4[4],xmm7[5],xmm4[5],xmm7[6],xmm4[6],xmm7[7],xmm4[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,3,1,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm9, %xmm7
+; SSE-NEXT:    por %xmm7, %xmm12
+; SSE-NEXT:    packuswb %xmm12, %xmm11
+; SSE-NEXT:    movdqa %xmm1, %xmm7
+; SSE-NEXT:    pandn %xmm11, %xmm7
+; SSE-NEXT:    pand %xmm1, %xmm10
+; SSE-NEXT:    por %xmm10, %xmm7
+; SSE-NEXT:    movdqa %xmm14, %xmm10
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm4[8],xmm10[9],xmm4[9],xmm10[10],xmm4[10],xmm10[11],xmm4[11],xmm10[12],xmm4[12],xmm10[13],xmm4[13],xmm10[14],xmm4[14],xmm10[15],xmm4[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[1,1,2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm1[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[3,1,1,2,4,5,6,7]
-; SSE-NEXT:    pand %xmm5, %xmm11
-; SSE-NEXT:    pandn %xmm10, %xmm5
-; SSE-NEXT:    por %xmm11, %xmm5
-; SSE-NEXT:    packuswb %xmm5, %xmm5
-; SSE-NEXT:    pand %xmm2, %xmm5
-; SSE-NEXT:    pandn %xmm9, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm5
-; SSE-NEXT:    movdqa %xmm12, %xmm2
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm4[0],xmm14[1],xmm4[1],xmm14[2],xmm4[2],xmm14[3],xmm4[3],xmm14[4],xmm4[4],xmm14[5],xmm4[5],xmm14[6],xmm4[6],xmm14[7],xmm4[7]
+; SSE-NEXT:    movdqa %xmm14, %xmm11
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[1,0],xmm10[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[2,0],xmm10[2,3]
+; SSE-NEXT:    psrlq $48, %xmm10
+; SSE-NEXT:    psrldq {{.*#+}} xmm11 = xmm11[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm10[0],xmm11[1],xmm10[1],xmm11[2],xmm10[2],xmm11[3],xmm10[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm14[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,4,4,5,7]
+; SSE-NEXT:    packuswb %xmm11, %xmm10
+; SSE-NEXT:    movdqa %xmm2, %xmm11
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm4[8],xmm11[9],xmm4[9],xmm11[10],xmm4[10],xmm11[11],xmm4[11],xmm11[12],xmm4[12],xmm11[13],xmm4[13],xmm11[14],xmm4[14],xmm11[15],xmm4[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[1,1,2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3],xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm4[8],xmm12[9],xmm4[9],xmm12[10],xmm4[10],xmm12[11],xmm4[11],xmm12[12],xmm4[12],xmm12[13],xmm4[13],xmm12[14],xmm4[14],xmm12[15],xmm4[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,1,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm7, %xmm2
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm12[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,7,4]
-; SSE-NEXT:    pandn %xmm4, %xmm7
-; SSE-NEXT:    por %xmm2, %xmm7
-; SSE-NEXT:    pand %xmm0, %xmm5
-; SSE-NEXT:    packuswb %xmm7, %xmm2
-; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    por %xmm5, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, 16(%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, (%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 16(%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, (%rdx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 16(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, (%rcx)
-; SSE-NEXT:    movdqa %xmm13, 16(%r8)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, (%r8)
+; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm2[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[3,1,1,2,4,5,6,7]
+; SSE-NEXT:    pand %xmm6, %xmm12
+; SSE-NEXT:    pandn %xmm11, %xmm6
+; SSE-NEXT:    por %xmm12, %xmm6
+; SSE-NEXT:    packuswb %xmm6, %xmm6
+; SSE-NEXT:    pand %xmm0, %xmm6
+; SSE-NEXT:    pandn %xmm10, %xmm0
+; SSE-NEXT:    por %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm13, %xmm6
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1],xmm6[2],xmm4[2],xmm6[3],xmm4[3],xmm6[4],xmm4[4],xmm6[5],xmm4[5],xmm6[6],xmm4[6],xmm6[7],xmm4[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm4[8],xmm13[9],xmm4[9],xmm13[10],xmm4[10],xmm13[11],xmm4[11],xmm13[12],xmm4[12],xmm13[13],xmm4[13],xmm13[14],xmm4[14],xmm13[15],xmm4[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm6[0,3,1,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm9, %xmm4
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm13[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,5,5,7,4]
+; SSE-NEXT:    pandn %xmm6, %xmm9
+; SSE-NEXT:    por %xmm4, %xmm9
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    packuswb %xmm9, %xmm4
+; SSE-NEXT:    pandn %xmm4, %xmm1
+; SSE-NEXT:    por %xmm0, %xmm1
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rdx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r8)
+; SSE-NEXT:    movdqa %xmm15, (%r8)
 ; SSE-NEXT:    movdqa %xmm3, 16(%r9)
-; SSE-NEXT:    movdqa %xmm8, (%r9)
+; SSE-NEXT:    movdqa %xmm5, (%r9)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movdqa %xmm0, 16(%rax)
-; SSE-NEXT:    movdqa %xmm6, (%rax)
-; SSE-NEXT:    addq $264, %rsp # imm = 0x108
+; SSE-NEXT:    movdqa %xmm1, 16(%rax)
+; SSE-NEXT:    movdqa %xmm7, (%rax)
+; SSE-NEXT:    addq $280, %rsp # imm = 0x118
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i8_stride6_vf32:
@@ -4624,126 +4622,129 @@ define void @load_i8_stride6_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5) nounwind {
 ; SSE-LABEL: load_i8_stride6_vf64:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $792, %rsp # imm = 0x318
-; SSE-NEXT:    movdqa 64(%rdi), %xmm4
+; SSE-NEXT:    subq $840, %rsp # imm = 0x348
+; SSE-NEXT:    movdqa 64(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 80(%rdi), %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 80(%rdi), %xmm5
+; SSE-NEXT:    movdqa (%rdi), %xmm8
+; SSE-NEXT:    movdqa 16(%rdi), %xmm5
 ; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rdi), %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 16(%rdi), %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 32(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 48(%rdi), %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,0,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm9, %xmm1
 ; SSE-NEXT:    pandn %xmm2, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,0,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm13, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,0,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm2, %xmm6
+; SSE-NEXT:    pandn %xmm0, %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm6
+; SSE-NEXT:    pandn %xmm0, %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm9, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
 ; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm1, %xmm10
+; SSE-NEXT:    movdqa %xmm1, %xmm6
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,5]
 ; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    pandn %xmm6, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    pand %xmm3, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,1,3]
-; SSE-NEXT:    pand %xmm10, %xmm1
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm8, %xmm1
-; SSE-NEXT:    movdqa %xmm8, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm11, %xmm1
+; SSE-NEXT:    pandn %xmm0, %xmm1
+; SSE-NEXT:    movdqa %xmm2, %xmm0
 ; SSE-NEXT:    pandn %xmm5, %xmm0
-; SSE-NEXT:    pand %xmm13, %xmm4
-; SSE-NEXT:    por %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[3,1,2,0]
-; SSE-NEXT:    pand %xmm10, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
+; SSE-NEXT:    movdqa %xmm8, %xmm5
+; SSE-NEXT:    movdqa %xmm8, %xmm7
+; SSE-NEXT:    pand %xmm2, %xmm5
+; SSE-NEXT:    por %xmm0, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[0,2,1,3]
+; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
+; SSE-NEXT:    pand %xmm11, %xmm0
+; SSE-NEXT:    por %xmm1, %xmm0
+; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    pandn %xmm4, %xmm1
+; SSE-NEXT:    pand %xmm9, %xmm3
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[3,1,2,0]
+; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,5]
+; SSE-NEXT:    packuswb %xmm1, %xmm1
 ; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pand %xmm4, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    pand %xmm4, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 320(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    movdqa %xmm9, %xmm0
 ; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa 336(%rdi), %xmm12
-; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    pandn %xmm12, %xmm1
+; SSE-NEXT:    movdqa 336(%rdi), %xmm10
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    pandn %xmm10, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm1
-; SSE-NEXT:    pandn %xmm12, %xmm1
+; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    pandn %xmm10, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm13, %xmm12
-; SSE-NEXT:    por %xmm0, %xmm12
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    pand %xmm9, %xmm10
+; SSE-NEXT:    por %xmm0, %xmm10
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    pand %xmm6, %xmm0
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,5]
 ; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    movdqa %xmm11, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa 304(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm7
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    movdqa 288(%rdi), %xmm6
-; SSE-NEXT:    movdqa %xmm6, %xmm2
-; SSE-NEXT:    pand %xmm3, %xmm2
-; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,2,1,3]
-; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    movdqa 304(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    movdqa 288(%rdi), %xmm12
+; SSE-NEXT:    movdqa %xmm12, %xmm3
+; SSE-NEXT:    pand %xmm2, %xmm3
+; SSE-NEXT:    movdqa %xmm2, %xmm5
+; SSE-NEXT:    por %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[0,2,1,3]
+; SSE-NEXT:    pand %xmm6, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm8, %xmm0
+; SSE-NEXT:    pand %xmm11, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    movdqa 368(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm2
+; SSE-NEXT:    movdqa %xmm9, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    movdqa 352(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm13, %xmm3
+; SSE-NEXT:    movdqa 352(%rdi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pand %xmm9, %xmm3
+; SSE-NEXT:    movdqa %xmm9, %xmm1
 ; SSE-NEXT:    por %xmm2, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[3,1,2,0]
-; SSE-NEXT:    pand %xmm10, %xmm2
+; SSE-NEXT:    pand %xmm6, %xmm2
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,5]
@@ -4754,173 +4755,173 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movdqa %xmm4, %xmm9
 ; SSE-NEXT:    por %xmm0, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 224(%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa 240(%rdi), %xmm11
-; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    pandn %xmm11, %xmm2
+; SSE-NEXT:    movdqa 224(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm2
-; SSE-NEXT:    pandn %xmm11, %xmm2
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    pandn %xmm2, %xmm0
+; SSE-NEXT:    movdqa 240(%rdi), %xmm14
+; SSE-NEXT:    movdqa %xmm5, %xmm2
+; SSE-NEXT:    pandn %xmm14, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm13, %xmm11
-; SSE-NEXT:    por %xmm0, %xmm11
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    pandn %xmm14, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm1, %xmm14
+; SSE-NEXT:    por %xmm0, %xmm14
+; SSE-NEXT:    movdqa %xmm14, %xmm0
+; SSE-NEXT:    pand %xmm6, %xmm0
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,5]
 ; SSE-NEXT:    packuswb %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm8, %xmm2
+; SSE-NEXT:    movdqa %xmm11, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    movdqa 208(%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa 192(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    pand %xmm7, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,1,3]
-; SSE-NEXT:    movdqa %xmm10, %xmm1
-; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    movdqa 208(%rdi), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    pandn %xmm4, %xmm0
+; SSE-NEXT:    movdqa 192(%rdi), %xmm8
+; SSE-NEXT:    movdqa %xmm8, %xmm5
+; SSE-NEXT:    pand %xmm3, %xmm5
+; SSE-NEXT:    movdqa %xmm3, %xmm4
+; SSE-NEXT:    por %xmm0, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[0,2,1,3]
+; SSE-NEXT:    pand %xmm6, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    movdqa %xmm8, %xmm10
+; SSE-NEXT:    pand %xmm11, %xmm0
 ; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa 272(%rdi), %xmm14
-; SSE-NEXT:    movdqa %xmm13, %xmm2
-; SSE-NEXT:    pandn %xmm14, %xmm2
-; SSE-NEXT:    movdqa 256(%rdi), %xmm15
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm13, %xmm15
-; SSE-NEXT:    por %xmm2, %xmm15
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm15[3,1,2,0]
-; SSE-NEXT:    pand %xmm1, %xmm2
+; SSE-NEXT:    movdqa 272(%rdi), %xmm15
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    pandn %xmm15, %xmm2
+; SSE-NEXT:    movdqa 256(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm1, %xmm3
+; SSE-NEXT:    por %xmm2, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[3,1,2,0]
+; SSE-NEXT:    pand %xmm6, %xmm2
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pandn %xmm2, %xmm4
+; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm3
 ; SSE-NEXT:    pand %xmm9, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    por %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 128(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    movdqa %xmm1, %xmm0
 ; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    movdqa 144(%rdi), %xmm9
-; SSE-NEXT:    movdqa %xmm7, %xmm4
-; SSE-NEXT:    pandn %xmm9, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm4
-; SSE-NEXT:    pandn %xmm9, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm13, %xmm9
-; SSE-NEXT:    por %xmm0, %xmm9
-; SSE-NEXT:    movdqa %xmm9, %xmm0
-; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    movdqa 144(%rdi), %xmm10
+; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    pandn %xmm10, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm10, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm1, %xmm10
+; SSE-NEXT:    por %xmm0, %xmm10
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    pand %xmm6, %xmm0
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,3,2,3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,5]
 ; SSE-NEXT:    packuswb %xmm5, %xmm0
-; SSE-NEXT:    pandn %xmm0, %xmm10
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    movdqa %xmm13, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm13, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm1
-; SSE-NEXT:    movdqa %xmm0, %xmm2
-; SSE-NEXT:    pandn %xmm6, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, %xmm5
-; SSE-NEXT:    movdqa %xmm0, %xmm2
-; SSE-NEXT:    pandn %xmm3, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm4
-; SSE-NEXT:    movdqa 112(%rdi), %xmm6
+; SSE-NEXT:    movdqa %xmm11, %xmm13
+; SSE-NEXT:    pandn %xmm0, %xmm13
+; SSE-NEXT:    movdqa %xmm1, %xmm0
 ; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    movdqa %xmm7, %xmm8
-; SSE-NEXT:    pandn %xmm6, %xmm8
-; SSE-NEXT:    movdqa 160(%rdi), %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm7
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm13, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm1
+; SSE-NEXT:    pandn %xmm7, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm3
-; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    pandn %xmm12, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm12, %xmm3
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    pandn %xmm8, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm8, %xmm7
+; SSE-NEXT:    movdqa 112(%rdi), %xmm6
+; SSE-NEXT:    movdqa %xmm4, %xmm5
+; SSE-NEXT:    movdqa %xmm4, %xmm9
+; SSE-NEXT:    pandn %xmm6, %xmm9
+; SSE-NEXT:    movdqa 160(%rdi), %xmm12
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm12
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm4, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm8, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm8
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm0, %xmm8
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm1, %xmm8
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm5, %xmm3
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    pandn %xmm2, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    pand %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm0, %xmm8
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm3, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm3, %xmm8
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm7
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm15, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm15
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm0, %xmm5
+; SSE-NEXT:    pandn %xmm6, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 96(%rdi), %xmm15
+; SSE-NEXT:    movdqa %xmm15, %xmm5
+; SSE-NEXT:    pand %xmm0, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 176(%rdi), %xmm7
+; SSE-NEXT:    movdqa %xmm7, %xmm8
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm8
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm0, %xmm5
 ; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm1, %xmm5
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm0, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm0, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pandn %xmm14, %xmm5
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm14
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm0, %xmm2
-; SSE-NEXT:    pandn %xmm6, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 96(%rdi), %xmm4
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 176(%rdi), %xmm14
-; SSE-NEXT:    movdqa %xmm14, %xmm2
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm13
-; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa (%rsp), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm0, %xmm1
@@ -4928,38 +4929,37 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movdqa %xmm0, %xmm1
 ; SSE-NEXT:    pand %xmm0, %xmm6
 ; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm0, %xmm13
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pandn %xmm4, %xmm1
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm15, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm4, %xmm3
-; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3
-; SSE-NEXT:    por %xmm8, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[0,2,1,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm15, %xmm2
+; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2
+; SSE-NEXT:    por %xmm9, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm2[0,2,1,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm6, %xmm5
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    packuswb %xmm5, %xmm5
-; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    pand %xmm8, %xmm5
-; SSE-NEXT:    por %xmm10, %xmm5
-; SSE-NEXT:    pandn %xmm14, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm7
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[3,1,2,0]
-; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    pand %xmm11, %xmm5
+; SSE-NEXT:    por %xmm13, %xmm5
+; SSE-NEXT:    pandn %xmm7, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm12
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[3,1,2,0]
+; SSE-NEXT:    pand %xmm6, %xmm0
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
-; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
+; SSE-NEXT:    movdqa %xmm7, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm10, %xmm5
+; SSE-NEXT:    pand %xmm7, %xmm5
 ; SSE-NEXT:    por %xmm5, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pxor %xmm5, %xmm5
@@ -4970,194 +4970,194 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[2,2,3,3]
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
 ; SSE-NEXT:    psrld $16, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm1[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,5,7,6,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm14 = xmm14[2],xmm0[2],xmm14[3],xmm0[3]
-; SSE-NEXT:    packuswb %xmm14, %xmm4
-; SSE-NEXT:    movdqa %xmm8, %xmm1
-; SSE-NEXT:    pandn %xmm4, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm2, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm1[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,5,7,6,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm0[2],xmm15[3],xmm0[3]
+; SSE-NEXT:    packuswb %xmm15, %xmm4
+; SSE-NEXT:    movdqa %xmm11, %xmm8
+; SSE-NEXT:    pandn %xmm4, %xmm8
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm1, %xmm4
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,1,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,1,1,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,7,6,7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,0,65535,0,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm0, %xmm14
-; SSE-NEXT:    pandn %xmm4, %xmm14
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3],xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm2[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,3,2,0,4,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm4
-; SSE-NEXT:    por %xmm14, %xmm4
-; SSE-NEXT:    packuswb %xmm4, %xmm4
-; SSE-NEXT:    pand %xmm8, %xmm4
-; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm0, %xmm15
+; SSE-NEXT:    pandn %xmm4, %xmm15
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3],xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    pandn %xmm14, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm5[8],xmm6[9],xmm5[9],xmm6[10],xmm5[10],xmm6[11],xmm5[11],xmm6[12],xmm5[12],xmm6[13],xmm5[13],xmm6[14],xmm5[14],xmm6[15],xmm5[15]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm6[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm14[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,5,7,6,4]
-; SSE-NEXT:    pand %xmm2, %xmm14
-; SSE-NEXT:    por %xmm1, %xmm14
-; SSE-NEXT:    packuswb %xmm14, %xmm1
-; SSE-NEXT:    movdqa %xmm10, %xmm14
-; SSE-NEXT:    pandn %xmm1, %xmm14
-; SSE-NEXT:    pand %xmm10, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm14
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm5[8],xmm1[9],xmm5[9],xmm1[10],xmm5[10],xmm1[11],xmm5[11],xmm1[12],xmm5[12],xmm1[13],xmm5[13],xmm1[14],xmm5[14],xmm1[15],xmm5[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm5[0],xmm12[1],xmm5[1],xmm12[2],xmm5[2],xmm12[3],xmm5[3],xmm12[4],xmm5[4],xmm12[5],xmm5[5],xmm12[6],xmm5[6],xmm12[7],xmm5[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm12[2,2,3,3]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
-; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,5,7,6,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm12 = xmm12[2],xmm1[2],xmm12[3],xmm1[3]
-; SSE-NEXT:    packuswb %xmm12, %xmm4
-; SSE-NEXT:    movdqa %xmm8, %xmm14
-; SSE-NEXT:    movdqa %xmm8, %xmm1
-; SSE-NEXT:    pandn %xmm4, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm6, %xmm4
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,7,6,7]
-; SSE-NEXT:    movdqa %xmm0, %xmm12
-; SSE-NEXT:    pandn %xmm4, %xmm12
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3],xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm6[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm1[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,3,2,0,4,5,6,7]
 ; SSE-NEXT:    pand %xmm0, %xmm4
-; SSE-NEXT:    por %xmm12, %xmm4
+; SSE-NEXT:    por %xmm15, %xmm4
 ; SSE-NEXT:    packuswb %xmm4, %xmm4
-; SSE-NEXT:    pand %xmm8, %xmm4
-; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm6, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3],xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    movdqa %xmm2, %xmm12
-; SSE-NEXT:    pandn %xmm1, %xmm12
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm5[8],xmm6[9],xmm5[9],xmm6[10],xmm5[10],xmm6[11],xmm5[11],xmm6[12],xmm5[12],xmm6[13],xmm5[13],xmm6[14],xmm5[14],xmm6[15],xmm5[15]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm6[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,7,6,4]
-; SSE-NEXT:    pand %xmm2, %xmm1
-; SSE-NEXT:    por %xmm12, %xmm1
-; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm10, %xmm12
-; SSE-NEXT:    pandn %xmm1, %xmm12
-; SSE-NEXT:    pand %xmm10, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm12
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm5[8],xmm1[9],xmm5[9],xmm1[10],xmm5[10],xmm1[11],xmm5[11],xmm1[12],xmm5[12],xmm1[13],xmm5[13],xmm1[14],xmm5[14],xmm1[15],xmm5[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm5[0],xmm11[1],xmm5[1],xmm11[2],xmm5[2],xmm11[3],xmm5[3],xmm11[4],xmm5[4],xmm11[5],xmm5[5],xmm11[6],xmm5[6],xmm11[7],xmm5[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm11[2,2,3,3]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
-; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm11[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,5,7,6,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm8 = xmm8[2],xmm1[2],xmm8[3],xmm1[3]
-; SSE-NEXT:    packuswb %xmm8, %xmm4
-; SSE-NEXT:    movdqa %xmm14, %xmm1
-; SSE-NEXT:    pandn %xmm4, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm6, %xmm4
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,7,6,7]
-; SSE-NEXT:    movdqa %xmm0, %xmm8
-; SSE-NEXT:    pandn %xmm4, %xmm8
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3],xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm6[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[1,3,2,0,4,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm4
+; SSE-NEXT:    pand %xmm11, %xmm4
 ; SSE-NEXT:    por %xmm8, %xmm4
-; SSE-NEXT:    packuswb %xmm4, %xmm4
-; SSE-NEXT:    pand %xmm14, %xmm4
-; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm15, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3],xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    movdqa %xmm2, %xmm8
-; SSE-NEXT:    pandn %xmm1, %xmm8
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm15 = xmm15[8],xmm5[8],xmm15[9],xmm5[9],xmm15[10],xmm5[10],xmm15[11],xmm5[11],xmm15[12],xmm5[12],xmm15[13],xmm5[13],xmm15[14],xmm5[14],xmm15[15],xmm5[15]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm15[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,7,6,4]
-; SSE-NEXT:    pand %xmm2, %xmm1
-; SSE-NEXT:    por %xmm8, %xmm1
-; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm10, %xmm8
-; SSE-NEXT:    pandn %xmm1, %xmm8
-; SSE-NEXT:    pand %xmm10, %xmm4
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm8
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm5[0],xmm8[1],xmm5[1],xmm8[2],xmm5[2],xmm8[3],xmm5[3],xmm8[4],xmm5[4],xmm8[5],xmm5[5],xmm8[6],xmm5[6],xmm8[7],xmm5[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm8[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm1, %xmm8
+; SSE-NEXT:    pandn %xmm15, %xmm8
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm5[8],xmm3[9],xmm5[9],xmm3[10],xmm5[10],xmm3[11],xmm5[11],xmm3[12],xmm5[12],xmm3[13],xmm5[13],xmm3[14],xmm5[14],xmm3[15],xmm5[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm15[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,5,7,6,4]
+; SSE-NEXT:    pand %xmm1, %xmm15
+; SSE-NEXT:    por %xmm8, %xmm15
+; SSE-NEXT:    packuswb %xmm15, %xmm8
+; SSE-NEXT:    movdqa %xmm7, %xmm15
+; SSE-NEXT:    pandn %xmm8, %xmm15
+; SSE-NEXT:    pand %xmm7, %xmm4
+; SSE-NEXT:    por %xmm4, %xmm15
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm4
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3],xmm3[4],xmm5[4],xmm3[5],xmm5[5],xmm3[6],xmm5[6],xmm3[7],xmm5[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm3[2,2,3,3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm4[0],xmm8[1],xmm4[1],xmm8[2],xmm4[2],xmm8[3],xmm4[3]
+; SSE-NEXT:    psrld $16, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm3[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,5,7,6,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm13 = xmm13[2],xmm4[2],xmm13[3],xmm4[3]
+; SSE-NEXT:    packuswb %xmm13, %xmm8
+; SSE-NEXT:    movdqa %xmm11, %xmm4
+; SSE-NEXT:    pandn %xmm8, %xmm4
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm8
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm5[8],xmm8[9],xmm5[9],xmm8[10],xmm5[10],xmm8[11],xmm5[11],xmm8[12],xmm5[12],xmm8[13],xmm5[13],xmm8[14],xmm5[14],xmm8[15],xmm5[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,5,7,6,7]
+; SSE-NEXT:    movdqa %xmm0, %xmm13
+; SSE-NEXT:    pandn %xmm8, %xmm13
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3],xmm3[4],xmm5[4],xmm3[5],xmm5[5],xmm3[6],xmm5[6],xmm3[7],xmm5[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[1,3,2,0,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm8
+; SSE-NEXT:    por %xmm13, %xmm8
+; SSE-NEXT:    packuswb %xmm8, %xmm8
+; SSE-NEXT:    pand %xmm11, %xmm8
 ; SSE-NEXT:    por %xmm4, %xmm8
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm5[8],xmm1[9],xmm5[9],xmm1[10],xmm5[10],xmm1[11],xmm5[11],xmm1[12],xmm5[12],xmm1[13],xmm5[13],xmm1[14],xmm5[14],xmm1[15],xmm5[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm5[0],xmm9[1],xmm5[1],xmm9[2],xmm5[2],xmm9[3],xmm5[3],xmm9[4],xmm5[4],xmm9[5],xmm5[5],xmm9[6],xmm5[6],xmm9[7],xmm5[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm9[2,2,3,3]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
-; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm9[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,5,7,6,7]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm6 = xmm6[2],xmm1[2],xmm6[3],xmm1[3]
-; SSE-NEXT:    packuswb %xmm6, %xmm4
-; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm5[8],xmm1[9],xmm5[9],xmm1[10],xmm5[10],xmm1[11],xmm5[11],xmm1[12],xmm5[12],xmm1[13],xmm5[13],xmm1[14],xmm5[14],xmm1[15],xmm5[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,7,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm4
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3],xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    movdqa %xmm1, %xmm13
+; SSE-NEXT:    pandn %xmm4, %xmm13
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm5[8],xmm3[9],xmm5[9],xmm3[10],xmm5[10],xmm3[11],xmm5[11],xmm3[12],xmm5[12],xmm3[13],xmm5[13],xmm3[14],xmm5[14],xmm3[15],xmm5[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,7,6,4]
+; SSE-NEXT:    pand %xmm1, %xmm4
+; SSE-NEXT:    por %xmm13, %xmm4
+; SSE-NEXT:    packuswb %xmm4, %xmm4
+; SSE-NEXT:    movdqa %xmm7, %xmm13
+; SSE-NEXT:    pandn %xmm4, %xmm13
+; SSE-NEXT:    pand %xmm7, %xmm8
+; SSE-NEXT:    por %xmm8, %xmm13
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm14, %xmm4
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm5[0],xmm14[1],xmm5[1],xmm14[2],xmm5[2],xmm14[3],xmm5[3],xmm14[4],xmm5[4],xmm14[5],xmm5[5],xmm14[6],xmm5[6],xmm14[7],xmm5[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm14[2,2,3,3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm4[0],xmm8[1],xmm4[1],xmm8[2],xmm4[2],xmm8[3],xmm4[3]
+; SSE-NEXT:    psrld $16, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm14[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,5,7,6,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm4[2],xmm9[3],xmm4[3]
+; SSE-NEXT:    packuswb %xmm9, %xmm8
+; SSE-NEXT:    movdqa %xmm11, %xmm4
+; SSE-NEXT:    pandn %xmm8, %xmm4
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm8
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm5[8],xmm8[9],xmm5[9],xmm8[10],xmm5[10],xmm8[11],xmm5[11],xmm8[12],xmm5[12],xmm8[13],xmm5[13],xmm8[14],xmm5[14],xmm8[15],xmm5[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,5,7,6,7]
+; SSE-NEXT:    movdqa %xmm0, %xmm9
+; SSE-NEXT:    pandn %xmm8, %xmm9
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3],xmm3[4],xmm5[4],xmm3[5],xmm5[5],xmm3[6],xmm5[6],xmm3[7],xmm5[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,3,2,0,4,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    por %xmm3, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[1,3,2,0,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm8
+; SSE-NEXT:    por %xmm9, %xmm8
+; SSE-NEXT:    packuswb %xmm8, %xmm8
+; SSE-NEXT:    pand %xmm11, %xmm8
+; SSE-NEXT:    por %xmm4, %xmm8
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm4
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3],xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    movdqa %xmm1, %xmm9
+; SSE-NEXT:    pandn %xmm4, %xmm9
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm5[8],xmm3[9],xmm5[9],xmm3[10],xmm5[10],xmm3[11],xmm5[11],xmm3[12],xmm5[12],xmm3[13],xmm5[13],xmm3[14],xmm5[14],xmm3[15],xmm5[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm3[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,7,6,4]
+; SSE-NEXT:    pand %xmm1, %xmm4
+; SSE-NEXT:    por %xmm9, %xmm4
+; SSE-NEXT:    packuswb %xmm4, %xmm4
+; SSE-NEXT:    movdqa %xmm7, %xmm9
+; SSE-NEXT:    movdqa %xmm7, %xmm3
+; SSE-NEXT:    pandn %xmm4, %xmm3
+; SSE-NEXT:    pand %xmm7, %xmm8
+; SSE-NEXT:    por %xmm8, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm10, %xmm4
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm5[0],xmm10[1],xmm5[1],xmm10[2],xmm5[2],xmm10[3],xmm5[3],xmm10[4],xmm5[4],xmm10[5],xmm5[5],xmm10[6],xmm5[6],xmm10[7],xmm5[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm10[2,2,3,3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm4[0],xmm8[1],xmm4[1],xmm8[2],xmm4[2],xmm8[3],xmm4[3]
+; SSE-NEXT:    psrld $16, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm10[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,7,6,7]
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; SSE-NEXT:    packuswb %xmm3, %xmm8
+; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm5[8],xmm3[9],xmm5[9],xmm3[10],xmm5[10],xmm3[11],xmm5[11],xmm3[12],xmm5[12],xmm3[13],xmm5[13],xmm3[14],xmm5[14],xmm3[15],xmm5[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,7,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3],xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,3,2,0,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    por %xmm2, %xmm0
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm14, %xmm1
-; SSE-NEXT:    pand %xmm14, %xmm0
-; SSE-NEXT:    pandn %xmm4, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm0
+; SSE-NEXT:    pand %xmm11, %xmm0
+; SSE-NEXT:    pandn %xmm8, %xmm11
+; SSE-NEXT:    por %xmm0, %xmm11
+; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm5[8],xmm12[9],xmm5[9],xmm12[10],xmm5[10],xmm12[11],xmm5[11],xmm12[12],xmm5[12],xmm12[13],xmm5[13],xmm12[14],xmm5[14],xmm12[15],xmm5[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm12[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,7,6,4]
+; SSE-NEXT:    pand %xmm1, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm1
+; SSE-NEXT:    por %xmm2, %xmm1
+; SSE-NEXT:    packuswb %xmm1, %xmm0
 ; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3],xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm5[8],xmm7[9],xmm5[9],xmm7[10],xmm5[10],xmm7[11],xmm5[11],xmm7[12],xmm5[12],xmm7[13],xmm5[13],xmm7[14],xmm5[14],xmm7[15],xmm5[15]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm7[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,7,6,4]
-; SSE-NEXT:    pand %xmm2, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    packuswb %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm10, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pand %xmm10, %xmm0
-; SSE-NEXT:    movdqa %xmm10, %xmm11
-; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [65535,0,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    pandn %xmm0, %xmm1
+; SSE-NEXT:    pand %xmm7, %xmm11
+; SSE-NEXT:    por %xmm11, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,0,65535,65535,0,65535,65535,0]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pand %xmm14, %xmm4
+; SSE-NEXT:    pand %xmm7, %xmm4
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    movdqa %xmm6, %xmm13
+; SSE-NEXT:    pand %xmm6, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,3,4,5,6,7]
@@ -5169,41 +5169,38 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm6[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pand %xmm10, %xmm1
-; SSE-NEXT:    movdqa %xmm10, %xmm12
+; SSE-NEXT:    pand %xmm13, %xmm1
+; SSE-NEXT:    movdqa %xmm13, %xmm15
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [0,0,0,0,0,255,255,255,255,255,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm1, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    pand %xmm1, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    por %xmm0, %xmm14
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm14[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    pand %xmm13, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm11, %xmm2
+; SSE-NEXT:    movdqa %xmm9, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pand %xmm11, %xmm3
-; SSE-NEXT:    movdqa %xmm11, %xmm8
+; SSE-NEXT:    pand %xmm9, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pand %xmm14, %xmm13
-; SSE-NEXT:    movdqa %xmm14, %xmm7
+; SSE-NEXT:    pand %xmm7, %xmm13
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    pand %xmm15, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,3,4,5,6,7]
@@ -5211,41 +5208,43 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,5,6]
 ; SSE-NEXT:    packuswb %xmm2, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm14[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pand %xmm10, %xmm2
+; SSE-NEXT:    pand %xmm15, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    movdqa %xmm1, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    pand %xmm1, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm3
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    por %xmm0, %xmm11
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm11[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    por %xmm0, %xmm12
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm12[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    pand %xmm15, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm8, %xmm2
+; SSE-NEXT:    movdqa %xmm9, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pand %xmm8, %xmm3
-; SSE-NEXT:    movdqa %xmm8, %xmm9
+; SSE-NEXT:    pand %xmm9, %xmm3
+; SSE-NEXT:    movdqa %xmm9, %xmm10
 ; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    pand %xmm7, %xmm10
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pand %xmm12, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    pand %xmm7, %xmm11
+; SSE-NEXT:    movdqa %xmm7, %xmm9
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    pand %xmm15, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,3,4,5,6,7]
@@ -5253,19 +5252,18 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,5,6]
 ; SSE-NEXT:    packuswb %xmm2, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm7[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pand %xmm12, %xmm2
+; SSE-NEXT:    pand %xmm15, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    movdqa %xmm1, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    pand %xmm1, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm3
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
@@ -5273,22 +5271,22 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    por %xmm0, %xmm8
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm8[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pand %xmm12, %xmm0
+; SSE-NEXT:    pand %xmm15, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    movdqa %xmm10, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pand %xmm9, %xmm3
-; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    pand %xmm10, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm7, %xmm0
+; SSE-NEXT:    pand %xmm9, %xmm0
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm12, %xmm0
+; SSE-NEXT:    movdqa %xmm15, %xmm3
+; SSE-NEXT:    pand %xmm15, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,3,4,5,6,7]
@@ -5300,38 +5298,36 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm15[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pand %xmm12, %xmm2
-; SSE-NEXT:    movdqa %xmm12, %xmm9
+; SSE-NEXT:    pand %xmm3, %xmm2
+; SSE-NEXT:    movdqa %xmm3, %xmm10
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,2,3,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    movdqa %xmm1, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    pand %xmm1, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm3
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    por %xmm0, %xmm12
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm12[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    por %xmm0, %xmm9
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm9[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    pand %xmm10, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,7,4]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm1, %xmm9
-; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255,255,255,255,0,0,0,0,0]
+; SSE-NEXT:    movdqa %xmm10, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pand %xmm1, %xmm3
+; SSE-NEXT:    pand %xmm10, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm1[8],xmm4[9],xmm1[9],xmm4[10],xmm1[10],xmm4[11],xmm1[11],xmm4[12],xmm1[12],xmm4[13],xmm1[13],xmm4[14],xmm1[14],xmm4[15],xmm1[15]
-; SSE-NEXT:    pxor %xmm7, %xmm7
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm5[8],xmm4[9],xmm5[9],xmm4[10],xmm5[10],xmm4[11],xmm5[11],xmm4[12],xmm5[12],xmm4[13],xmm5[13],xmm4[14],xmm5[14],xmm4[15],xmm5[15]
 ; SSE-NEXT:    movdqa %xmm4, %xmm2
 ; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm0[3,0]
 ; SSE-NEXT:    movaps %xmm0, %xmm3
@@ -5345,46 +5341,44 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,3,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1],xmm0[2],xmm7[2],xmm0[3],xmm7[3],xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,1,4,5,6,7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [0,65535,65535,0,65535,65535,65535,65535]
 ; SSE-NEXT:    movdqa %xmm3, %xmm4
 ; SSE-NEXT:    pandn %xmm0, %xmm4
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm7[8],xmm6[9],xmm7[9],xmm6[10],xmm7[10],xmm6[11],xmm7[11],xmm6[12],xmm7[12],xmm6[13],xmm7[13],xmm6[14],xmm7[14],xmm6[15],xmm7[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm5[8],xmm6[9],xmm5[9],xmm6[10],xmm5[10],xmm6[11],xmm5[11],xmm6[12],xmm5[12],xmm6[13],xmm5[13],xmm6[14],xmm5[14],xmm6[15],xmm5[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[0,3,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pand %xmm3, %xmm0
 ; SSE-NEXT:    por %xmm4, %xmm0
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm5, %xmm6
+; SSE-NEXT:    movdqa %xmm1, %xmm6
 ; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    pand %xmm5, %xmm2
+; SSE-NEXT:    pand %xmm1, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm6
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm7[8],xmm0[9],xmm7[9],xmm0[10],xmm7[10],xmm0[11],xmm7[11],xmm0[12],xmm7[12],xmm0[13],xmm7[13],xmm0[14],xmm7[14],xmm0[15],xmm7[15]
+; SSE-NEXT:    movdqa %xmm14, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm5[8],xmm0[9],xmm5[9],xmm0[10],xmm5[10],xmm0[11],xmm5[11],xmm0[12],xmm5[12],xmm0[13],xmm5[13],xmm0[14],xmm5[14],xmm0[15],xmm5[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,5]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,65535,0,65535,65535,0]
 ; SSE-NEXT:    movdqa %xmm4, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3],xmm1[4],xmm7[4],xmm1[5],xmm7[5],xmm1[6],xmm7[6],xmm1[7],xmm7[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,2,0,3]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm5[0],xmm14[1],xmm5[1],xmm14[2],xmm5[2],xmm14[3],xmm5[3],xmm14[4],xmm5[4],xmm14[5],xmm5[5],xmm14[6],xmm5[6],xmm14[7],xmm5[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[0,2,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,7,7]
 ; SSE-NEXT:    pand %xmm4, %xmm0
 ; SSE-NEXT:    por %xmm2, %xmm0
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    movdqa %xmm10, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pand %xmm9, %xmm6
+; SSE-NEXT:    pand %xmm10, %xmm6
 ; SSE-NEXT:    por %xmm6, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm1[8],xmm13[9],xmm1[9],xmm13[10],xmm1[10],xmm13[11],xmm1[11],xmm13[12],xmm1[12],xmm13[13],xmm1[13],xmm13[14],xmm1[14],xmm13[15],xmm1[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm5[8],xmm13[9],xmm5[9],xmm13[10],xmm5[10],xmm13[11],xmm5[11],xmm13[12],xmm5[12],xmm13[13],xmm5[13],xmm13[14],xmm5[14],xmm13[15],xmm5[15]
 ; SSE-NEXT:    movdqa %xmm13, %xmm2
 ; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm0[3,0]
 ; SSE-NEXT:    movaps %xmm0, %xmm6
@@ -5397,94 +5391,94 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,3,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,1,4,5,6,7]
 ; SSE-NEXT:    movdqa %xmm3, %xmm6
 ; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm1[8],xmm14[9],xmm1[9],xmm14[10],xmm1[10],xmm14[11],xmm1[11],xmm14[12],xmm1[12],xmm14[13],xmm1[13],xmm14[14],xmm1[14],xmm14[15],xmm1[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[0,3,2,1]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm5[8],xmm13[9],xmm5[9],xmm13[10],xmm5[10],xmm13[11],xmm5[11],xmm13[12],xmm5[12],xmm13[13],xmm5[13],xmm13[14],xmm5[14],xmm13[15],xmm5[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[0,3,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pand %xmm3, %xmm0
 ; SSE-NEXT:    por %xmm6, %xmm0
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm5, %xmm6
+; SSE-NEXT:    movdqa %xmm1, %xmm6
 ; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    pand %xmm5, %xmm2
+; SSE-NEXT:    pand %xmm1, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm6
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm5[8],xmm0[9],xmm5[9],xmm0[10],xmm5[10],xmm0[11],xmm5[11],xmm0[12],xmm5[12],xmm0[13],xmm5[13],xmm0[14],xmm5[14],xmm0[15],xmm5[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,5]
 ; SSE-NEXT:    movdqa %xmm4, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm1[0],xmm11[1],xmm1[1],xmm11[2],xmm1[2],xmm11[3],xmm1[3],xmm11[4],xmm1[4],xmm11[5],xmm1[5],xmm11[6],xmm1[6],xmm11[7],xmm1[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[0,2,0,3]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm5[0],xmm12[1],xmm5[1],xmm12[2],xmm5[2],xmm12[3],xmm5[3],xmm12[4],xmm5[4],xmm12[5],xmm5[5],xmm12[6],xmm5[6],xmm12[7],xmm5[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[0,2,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,7,7]
 ; SSE-NEXT:    pand %xmm4, %xmm0
 ; SSE-NEXT:    por %xmm2, %xmm0
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    movdqa %xmm10, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pand %xmm9, %xmm6
+; SSE-NEXT:    pand %xmm10, %xmm6
 ; SSE-NEXT:    por %xmm6, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm1[8],xmm10[9],xmm1[9],xmm10[10],xmm1[10],xmm10[11],xmm1[11],xmm10[12],xmm1[12],xmm10[13],xmm1[13],xmm10[14],xmm1[14],xmm10[15],xmm1[15]
-; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm5[8],xmm11[9],xmm5[9],xmm11[10],xmm5[10],xmm11[11],xmm5[11],xmm11[12],xmm5[12],xmm11[13],xmm5[13],xmm11[14],xmm5[14],xmm11[15],xmm5[15]
+; SSE-NEXT:    movdqa %xmm11, %xmm2
 ; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm0[3,0]
 ; SSE-NEXT:    movaps %xmm0, %xmm6
 ; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm2[0,2]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm10[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm10[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm11[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm11[2,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm6[0,1,2,3,7,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,3,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,1,4,5,6,7]
 ; SSE-NEXT:    movdqa %xmm3, %xmm6
 ; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm1[8],xmm7[9],xmm1[9],xmm7[10],xmm1[10],xmm7[11],xmm1[11],xmm7[12],xmm1[12],xmm7[13],xmm1[13],xmm7[14],xmm1[14],xmm7[15],xmm1[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm5[8],xmm7[9],xmm5[9],xmm7[10],xmm5[10],xmm7[11],xmm5[11],xmm7[12],xmm5[12],xmm7[13],xmm5[13],xmm7[14],xmm5[14],xmm7[15],xmm5[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[0,3,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pand %xmm3, %xmm0
 ; SSE-NEXT:    por %xmm6, %xmm0
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm5, %xmm6
+; SSE-NEXT:    movdqa %xmm1, %xmm6
 ; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    pand %xmm5, %xmm2
+; SSE-NEXT:    pand %xmm1, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm6
 ; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm5[8],xmm0[9],xmm5[9],xmm0[10],xmm5[10],xmm0[11],xmm5[11],xmm0[12],xmm5[12],xmm0[13],xmm5[13],xmm0[14],xmm5[14],xmm0[15],xmm5[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,5]
 ; SSE-NEXT:    movdqa %xmm4, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm1[0],xmm8[1],xmm1[1],xmm8[2],xmm1[2],xmm8[3],xmm1[3],xmm8[4],xmm1[4],xmm8[5],xmm1[5],xmm8[6],xmm1[6],xmm8[7],xmm1[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm5[0],xmm8[1],xmm5[1],xmm8[2],xmm5[2],xmm8[3],xmm5[3],xmm8[4],xmm5[4],xmm8[5],xmm5[5],xmm8[6],xmm5[6],xmm8[7],xmm5[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[0,2,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,7,7]
 ; SSE-NEXT:    pand %xmm4, %xmm0
 ; SSE-NEXT:    por %xmm2, %xmm0
 ; SSE-NEXT:    packuswb %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm10, %xmm0
 ; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    pand %xmm9, %xmm6
+; SSE-NEXT:    pand %xmm10, %xmm6
 ; SSE-NEXT:    por %xmm6, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm0, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3],xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm5[8],xmm0[9],xmm5[9],xmm0[10],xmm5[10],xmm0[11],xmm5[11],xmm0[12],xmm5[12],xmm0[13],xmm5[13],xmm0[14],xmm5[14],xmm0[15],xmm5[15]
 ; SSE-NEXT:    movdqa %xmm0, %xmm6
 ; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0],xmm2[3,0]
 ; SSE-NEXT:    movaps %xmm2, %xmm7
@@ -5498,381 +5492,377 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,3,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm2, %xmm6
 ; SSE-NEXT:    movdqa %xmm15, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3],xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,1,4,5,6,7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm15 = xmm15[8],xmm1[8],xmm15[9],xmm1[9],xmm15[10],xmm1[10],xmm15[11],xmm1[11],xmm15[12],xmm1[12],xmm15[13],xmm1[13],xmm15[14],xmm1[14],xmm15[15],xmm1[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm15 = xmm15[8],xmm5[8],xmm15[9],xmm5[9],xmm15[10],xmm5[10],xmm15[11],xmm5[11],xmm15[12],xmm5[12],xmm15[13],xmm5[13],xmm15[14],xmm5[14],xmm15[15],xmm5[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm15[0,3,2,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pand %xmm3, %xmm7
 ; SSE-NEXT:    pandn %xmm2, %xmm3
 ; SSE-NEXT:    por %xmm7, %xmm3
-; SSE-NEXT:    pand %xmm5, %xmm6
+; SSE-NEXT:    pand %xmm1, %xmm6
 ; SSE-NEXT:    packuswb %xmm3, %xmm3
-; SSE-NEXT:    pandn %xmm3, %xmm5
-; SSE-NEXT:    por %xmm6, %xmm5
-; SSE-NEXT:    movdqa %xmm12, %xmm2
-; SSE-NEXT:    pxor %xmm0, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm0[8],xmm2[9],xmm0[9],xmm2[10],xmm0[10],xmm2[11],xmm0[11],xmm2[12],xmm0[12],xmm2[13],xmm0[13],xmm2[14],xmm0[14],xmm2[15],xmm0[15]
+; SSE-NEXT:    pandn %xmm3, %xmm1
+; SSE-NEXT:    por %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,5]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1],xmm12[2],xmm0[2],xmm12[3],xmm0[3],xmm12[4],xmm0[4],xmm12[5],xmm0[5],xmm12[6],xmm0[6],xmm12[7],xmm0[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm12[0,2,0,3]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm5[0],xmm9[1],xmm5[1],xmm9[2],xmm5[2],xmm9[3],xmm5[3],xmm9[4],xmm5[4],xmm9[5],xmm5[5],xmm9[6],xmm5[6],xmm9[7],xmm5[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[0,2,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,7,7]
 ; SSE-NEXT:    pand %xmm4, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm4
 ; SSE-NEXT:    por %xmm3, %xmm4
-; SSE-NEXT:    pand %xmm9, %xmm5
+; SSE-NEXT:    pand %xmm10, %xmm1
 ; SSE-NEXT:    packuswb %xmm4, %xmm2
-; SSE-NEXT:    pandn %xmm2, %xmm9
-; SSE-NEXT:    por %xmm5, %xmm9
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm2, %xmm10
+; SSE-NEXT:    por %xmm1, %xmm10
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm10, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm12, %xmm1
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,2]
 ; SSE-NEXT:    packuswb %xmm2, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm15, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm4, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[3,1,2,0]
-; SSE-NEXT:    pand %xmm10, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[3,1,2,0]
+; SSE-NEXT:    pand %xmm12, %xmm1
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm1[2,1,0,3,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm6, %xmm6
-; SSE-NEXT:    pand %xmm15, %xmm6
+; SSE-NEXT:    pand %xmm4, %xmm6
 ; SSE-NEXT:    por %xmm2, %xmm6
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,0,65535,65535,0,65535,65535,0]
-; SSE-NEXT:    pand %xmm11, %xmm13
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[0,2,1,3]
-; SSE-NEXT:    pand %xmm10, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [65535,0,65535,65535,0,65535,65535,0]
+; SSE-NEXT:    pand %xmm13, %xmm11
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[0,2,1,3]
+; SSE-NEXT:    pand %xmm12, %xmm1
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm2
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,65535,65535,0,0,0]
-; SSE-NEXT:    movdqa %xmm1, %xmm3
-; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    pand %xmm1, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm14, %xmm2
-; SSE-NEXT:    pand %xmm10, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,65535,65535,0,0,0]
+; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    pandn %xmm2, %xmm1
+; SSE-NEXT:    pand %xmm3, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm12, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm2[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
 ; SSE-NEXT:    packuswb %xmm6, %xmm2
-; SSE-NEXT:    movdqa %xmm15, %xmm6
+; SSE-NEXT:    movdqa %xmm4, %xmm6
 ; SSE-NEXT:    pandn %xmm2, %xmm6
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[3,1,2,0]
-; SSE-NEXT:    pand %xmm10, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[3,1,2,0]
+; SSE-NEXT:    pand %xmm12, %xmm2
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm2[2,1,0,3,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm7, %xmm7
-; SSE-NEXT:    pand %xmm15, %xmm7
+; SSE-NEXT:    pand %xmm4, %xmm7
 ; SSE-NEXT:    por %xmm6, %xmm7
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pand %xmm11, %xmm3
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,1,3]
-; SSE-NEXT:    pand %xmm10, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pand %xmm13, %xmm8
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[0,2,1,3]
+; SSE-NEXT:    pand %xmm12, %xmm2
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
 ; SSE-NEXT:    packuswb %xmm2, %xmm6
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    pandn %xmm6, %xmm2
-; SSE-NEXT:    pand %xmm1, %xmm7
-; SSE-NEXT:    por %xmm7, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    pandn %xmm6, %xmm1
+; SSE-NEXT:    pand %xmm3, %xmm7
+; SSE-NEXT:    por %xmm7, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm10, %xmm6
+; SSE-NEXT:    pand %xmm12, %xmm6
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm6[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,0,2]
 ; SSE-NEXT:    packuswb %xmm7, %xmm6
-; SSE-NEXT:    movdqa %xmm15, %xmm7
+; SSE-NEXT:    movdqa %xmm4, %xmm7
 ; SSE-NEXT:    pandn %xmm6, %xmm7
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm2[3,1,2,0]
-; SSE-NEXT:    pand %xmm10, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[3,1,2,0]
+; SSE-NEXT:    pand %xmm12, %xmm6
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm6[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm8, %xmm8
-; SSE-NEXT:    pand %xmm15, %xmm8
-; SSE-NEXT:    por %xmm7, %xmm8
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pand %xmm11, %xmm2
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm2[0,2,1,3]
-; SSE-NEXT:    pand %xmm10, %xmm6
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm6[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm9, %xmm9
+; SSE-NEXT:    pand %xmm4, %xmm9
+; SSE-NEXT:    por %xmm7, %xmm9
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pand %xmm13, %xmm1
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[0,2,1,3]
+; SSE-NEXT:    pand %xmm12, %xmm6
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,1,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,1,0,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,6,5,4,7]
 ; SSE-NEXT:    packuswb %xmm6, %xmm7
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    pandn %xmm7, %xmm2
-; SSE-NEXT:    pand %xmm1, %xmm8
-; SSE-NEXT:    por %xmm8, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    pandn %xmm7, %xmm1
+; SSE-NEXT:    pand %xmm3, %xmm9
+; SSE-NEXT:    por %xmm9, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa (%rsp), %xmm7 # 16-byte Reload
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm7, (%rsp) # 16-byte Spill
-; SSE-NEXT:    pand %xmm10, %xmm7
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm7[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm12, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm7[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,1,0,2]
-; SSE-NEXT:    packuswb %xmm8, %xmm7
-; SSE-NEXT:    movdqa %xmm15, %xmm8
-; SSE-NEXT:    pandn %xmm7, %xmm8
+; SSE-NEXT:    packuswb %xmm9, %xmm7
+; SSE-NEXT:    movdqa %xmm4, %xmm9
+; SSE-NEXT:    pandn %xmm7, %xmm9
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm6[3,1,2,0]
-; SSE-NEXT:    pand %xmm10, %xmm7
+; SSE-NEXT:    pand %xmm12, %xmm7
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm7[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm9, %xmm9
-; SSE-NEXT:    pand %xmm15, %xmm9
-; SSE-NEXT:    por %xmm8, %xmm9
-; SSE-NEXT:    movdqa %xmm11, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    pand %xmm11, %xmm7
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm7[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm10, %xmm10
+; SSE-NEXT:    pand %xmm4, %xmm10
+; SSE-NEXT:    por %xmm9, %xmm10
+; SSE-NEXT:    movdqa %xmm13, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pand %xmm13, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    por %xmm7, %xmm2
+; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm2[0,2,1,3]
-; SSE-NEXT:    pand %xmm10, %xmm7
+; SSE-NEXT:    pand %xmm12, %xmm7
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,0,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,6,5,4,7]
-; SSE-NEXT:    packuswb %xmm7, %xmm8
-; SSE-NEXT:    movdqa %xmm1, %xmm7
-; SSE-NEXT:    pandn %xmm8, %xmm7
-; SSE-NEXT:    pand %xmm1, %xmm9
-; SSE-NEXT:    por %xmm9, %xmm7
-; SSE-NEXT:    movdqa %xmm0, %xmm8
-; SSE-NEXT:    pxor %xmm5, %xmm5
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm5[8],xmm8[9],xmm5[9],xmm8[10],xmm5[10],xmm8[11],xmm5[11],xmm8[12],xmm5[12],xmm8[13],xmm5[13],xmm8[14],xmm5[14],xmm8[15],xmm5[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
+; SSE-NEXT:    packuswb %xmm7, %xmm9
+; SSE-NEXT:    movdqa %xmm3, %xmm7
+; SSE-NEXT:    pandn %xmm9, %xmm7
+; SSE-NEXT:    pand %xmm3, %xmm10
+; SSE-NEXT:    por %xmm10, %xmm7
 ; SSE-NEXT:    movdqa %xmm0, %xmm9
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm8[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm8[2,3]
-; SSE-NEXT:    psrlq $48, %xmm8
-; SSE-NEXT:    psrldq {{.*#+}} xmm9 = xmm9[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm0[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,5,7]
-; SSE-NEXT:    packuswb %xmm9, %xmm8
-; SSE-NEXT:    movdqa %xmm15, %xmm10
-; SSE-NEXT:    pandn %xmm8, %xmm10
-; SSE-NEXT:    movdqa %xmm12, %xmm8
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm5[8],xmm8[9],xmm5[9],xmm8[10],xmm5[10],xmm8[11],xmm5[11],xmm8[12],xmm5[12],xmm8[13],xmm5[13],xmm8[14],xmm5[14],xmm8[15],xmm5[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[1,1,2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm5[8],xmm9[9],xmm5[9],xmm9[10],xmm5[10],xmm9[11],xmm5[11],xmm9[12],xmm5[12],xmm9[13],xmm5[13],xmm9[14],xmm5[14],xmm9[15],xmm5[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
+; SSE-NEXT:    movdqa %xmm0, %xmm10
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,0],xmm9[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm9[2,3]
+; SSE-NEXT:    psrlq $48, %xmm9
+; SSE-NEXT:    psrldq {{.*#+}} xmm10 = xmm10[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm0[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,4,5,7]
+; SSE-NEXT:    packuswb %xmm10, %xmm9
+; SSE-NEXT:    movdqa %xmm4, %xmm12
+; SSE-NEXT:    pandn %xmm9, %xmm12
+; SSE-NEXT:    movdqa %xmm15, %xmm9
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm5[8],xmm9[9],xmm5[9],xmm9[10],xmm5[10],xmm9[11],xmm5[11],xmm9[12],xmm5[12],xmm9[13],xmm5[13],xmm9[14],xmm5[14],xmm9[15],xmm5[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[1,1,2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,0,65535,65535,0,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm0, %xmm11
-; SSE-NEXT:    pandn %xmm8, %xmm11
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm5[0],xmm12[1],xmm5[1],xmm12[2],xmm5[2],xmm12[3],xmm5[3],xmm12[4],xmm5[4],xmm12[5],xmm5[5],xmm12[6],xmm5[6],xmm12[7],xmm5[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm12[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm8[3,1,1,2,4,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm12
-; SSE-NEXT:    por %xmm11, %xmm12
-; SSE-NEXT:    packuswb %xmm12, %xmm12
-; SSE-NEXT:    pand %xmm15, %xmm12
-; SSE-NEXT:    por %xmm10, %xmm12
-; SSE-NEXT:    movdqa %xmm13, %xmm8
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8],xmm5[8],xmm13[9],xmm5[9],xmm13[10],xmm5[10],xmm13[11],xmm5[11],xmm13[12],xmm5[12],xmm13[13],xmm5[13],xmm13[14],xmm5[14],xmm13[15],xmm5[15]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm13[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,5,5,7,4]
-; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,65535,65535,65535,0,65535,0,0]
+; SSE-NEXT:    movdqa %xmm0, %xmm13
+; SSE-NEXT:    pandn %xmm9, %xmm13
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm5[0],xmm15[1],xmm5[1],xmm15[2],xmm5[2],xmm15[3],xmm5[3],xmm15[4],xmm5[4],xmm15[5],xmm5[5],xmm15[6],xmm5[6],xmm15[7],xmm5[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm15[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm9[3,1,1,2,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm10
+; SSE-NEXT:    por %xmm13, %xmm10
+; SSE-NEXT:    packuswb %xmm10, %xmm10
+; SSE-NEXT:    pand %xmm4, %xmm10
+; SSE-NEXT:    por %xmm12, %xmm10
+; SSE-NEXT:    movdqa %xmm11, %xmm9
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm5[8],xmm11[9],xmm5[9],xmm11[10],xmm5[10],xmm11[11],xmm5[11],xmm11[12],xmm5[12],xmm11[13],xmm5[13],xmm11[14],xmm5[14],xmm11[15],xmm5[15]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm11[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm12[0,1,2,3,5,5,7,4]
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,65535,0,65535,0,0]
+; SSE-NEXT:    movdqa %xmm1, %xmm15
+; SSE-NEXT:    pandn %xmm13, %xmm15
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm5[0],xmm9[1],xmm5[1],xmm9[2],xmm5[2],xmm9[3],xmm5[3],xmm9[4],xmm5[4],xmm9[5],xmm5[5],xmm9[6],xmm5[6],xmm9[7],xmm5[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,3,1,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm1, %xmm9
+; SSE-NEXT:    por %xmm9, %xmm15
+; SSE-NEXT:    packuswb %xmm15, %xmm13
+; SSE-NEXT:    movdqa %xmm3, %xmm9
+; SSE-NEXT:    pandn %xmm13, %xmm9
+; SSE-NEXT:    pand %xmm3, %xmm10
+; SSE-NEXT:    por %xmm10, %xmm9
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm11, %xmm10
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm5[8],xmm10[9],xmm5[9],xmm10[10],xmm5[10],xmm10[11],xmm5[11],xmm10[12],xmm5[12],xmm10[13],xmm5[13],xmm10[14],xmm5[14],xmm10[15],xmm5[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm5[0],xmm11[1],xmm5[1],xmm11[2],xmm5[2],xmm11[3],xmm5[3],xmm11[4],xmm5[4],xmm11[5],xmm5[5],xmm11[6],xmm5[6],xmm11[7],xmm5[7]
 ; SSE-NEXT:    movdqa %xmm11, %xmm13
-; SSE-NEXT:    pandn %xmm10, %xmm13
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm5[0],xmm8[1],xmm5[1],xmm8[2],xmm5[2],xmm8[3],xmm5[3],xmm8[4],xmm5[4],xmm8[5],xmm5[5],xmm8[6],xmm5[6],xmm8[7],xmm5[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,3,1,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm11, %xmm8
-; SSE-NEXT:    por %xmm8, %xmm13
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[1,0],xmm10[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[2,0],xmm10[2,3]
+; SSE-NEXT:    psrlq $48, %xmm10
+; SSE-NEXT:    psrldq {{.*#+}} xmm13 = xmm13[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm10[0],xmm13[1],xmm10[1],xmm13[2],xmm10[2],xmm13[3],xmm10[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm11[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,4,4,5,7]
 ; SSE-NEXT:    packuswb %xmm13, %xmm10
-; SSE-NEXT:    movdqa %xmm1, %xmm8
-; SSE-NEXT:    pandn %xmm10, %xmm8
-; SSE-NEXT:    pand %xmm1, %xmm12
-; SSE-NEXT:    por %xmm12, %xmm8
-; SSE-NEXT:    movdqa %xmm14, %xmm9
+; SSE-NEXT:    movdqa %xmm4, %xmm13
+; SSE-NEXT:    pandn %xmm10, %xmm13
 ; SSE-NEXT:    movdqa %xmm14, %xmm10
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm5[8],xmm10[9],xmm5[9],xmm10[10],xmm5[10],xmm10[11],xmm5[11],xmm10[12],xmm5[12],xmm10[13],xmm5[13],xmm10[14],xmm5[14],xmm10[15],xmm5[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm5[0],xmm9[1],xmm5[1],xmm9[2],xmm5[2],xmm9[3],xmm5[3],xmm9[4],xmm5[4],xmm9[5],xmm5[5],xmm9[6],xmm5[6],xmm9[7],xmm5[7]
-; SSE-NEXT:    movdqa %xmm9, %xmm12
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[1,0],xmm10[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[2,0],xmm10[2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[1,1,2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    movdqa %xmm0, %xmm15
+; SSE-NEXT:    pandn %xmm10, %xmm15
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm5[0],xmm14[1],xmm5[1],xmm14[2],xmm5[2],xmm14[3],xmm5[3],xmm14[4],xmm5[4],xmm14[5],xmm5[5],xmm14[6],xmm5[6],xmm14[7],xmm5[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm14[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[3,1,1,2,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm10
+; SSE-NEXT:    por %xmm15, %xmm10
+; SSE-NEXT:    packuswb %xmm10, %xmm10
+; SSE-NEXT:    pand %xmm4, %xmm10
+; SSE-NEXT:    por %xmm13, %xmm10
+; SSE-NEXT:    movdqa %xmm8, %xmm13
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm5[8],xmm8[9],xmm5[9],xmm8[10],xmm5[10],xmm8[11],xmm5[11],xmm8[12],xmm5[12],xmm8[13],xmm5[13],xmm8[14],xmm5[14],xmm8[15],xmm5[15]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm8[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm15[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,5,5,7,4]
+; SSE-NEXT:    movdqa %xmm1, %xmm12
+; SSE-NEXT:    pandn %xmm15, %xmm12
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm5[0],xmm13[1],xmm5[1],xmm13[2],xmm5[2],xmm13[3],xmm5[3],xmm13[4],xmm5[4],xmm13[5],xmm5[5],xmm13[6],xmm5[6],xmm13[7],xmm5[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm13[0,3,1,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm13[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm1, %xmm13
+; SSE-NEXT:    por %xmm13, %xmm12
+; SSE-NEXT:    packuswb %xmm12, %xmm12
+; SSE-NEXT:    movdqa %xmm3, %xmm13
+; SSE-NEXT:    pandn %xmm12, %xmm13
+; SSE-NEXT:    pand %xmm3, %xmm10
+; SSE-NEXT:    por %xmm10, %xmm13
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm11, %xmm10
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm5[8],xmm10[9],xmm5[9],xmm10[10],xmm5[10],xmm10[11],xmm5[11],xmm10[12],xmm5[12],xmm10[13],xmm5[13],xmm10[14],xmm5[14],xmm10[15],xmm5[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm5[0],xmm11[1],xmm5[1],xmm11[2],xmm5[2],xmm11[3],xmm5[3],xmm11[4],xmm5[4],xmm11[5],xmm5[5],xmm11[6],xmm5[6],xmm11[7],xmm5[7]
+; SSE-NEXT:    movdqa %xmm11, %xmm12
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[1,0],xmm10[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[2,0],xmm10[2,3]
 ; SSE-NEXT:    psrlq $48, %xmm10
 ; SSE-NEXT:    psrldq {{.*#+}} xmm12 = xmm12[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm10[0],xmm12[1],xmm10[1],xmm12[2],xmm10[2],xmm12[3],xmm10[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm9[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm11[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,4,4,5,7]
 ; SSE-NEXT:    packuswb %xmm12, %xmm10
-; SSE-NEXT:    movdqa %xmm15, %xmm12
+; SSE-NEXT:    movdqa %xmm4, %xmm12
 ; SSE-NEXT:    pandn %xmm10, %xmm12
-; SSE-NEXT:    movdqa %xmm4, %xmm10
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm14, %xmm10
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm5[8],xmm10[9],xmm5[9],xmm10[10],xmm5[10],xmm10[11],xmm5[11],xmm10[12],xmm5[12],xmm10[13],xmm5[13],xmm10[14],xmm5[14],xmm10[15],xmm5[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[1,1,2,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    movdqa %xmm0, %xmm14
-; SSE-NEXT:    pandn %xmm10, %xmm14
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3],xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm4[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    movdqa %xmm0, %xmm11
+; SSE-NEXT:    pandn %xmm10, %xmm11
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm5[0],xmm14[1],xmm5[1],xmm14[2],xmm5[2],xmm14[3],xmm5[3],xmm14[4],xmm5[4],xmm14[5],xmm5[5],xmm14[6],xmm5[6],xmm14[7],xmm5[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm14[0,1,2,3,7,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm10[3,1,1,2,4,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm13
-; SSE-NEXT:    por %xmm14, %xmm13
-; SSE-NEXT:    packuswb %xmm13, %xmm13
-; SSE-NEXT:    pand %xmm15, %xmm13
-; SSE-NEXT:    por %xmm12, %xmm13
-; SSE-NEXT:    movdqa %xmm3, %xmm10
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm5[8],xmm3[9],xmm5[9],xmm3[10],xmm5[10],xmm3[11],xmm5[11],xmm3[12],xmm5[12],xmm3[13],xmm5[13],xmm3[14],xmm5[14],xmm3[15],xmm5[15]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm3[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,5,5,7,4]
-; SSE-NEXT:    movdqa %xmm11, %xmm14
-; SSE-NEXT:    pandn %xmm12, %xmm14
+; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm10[3,1,1,2,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm15
+; SSE-NEXT:    por %xmm11, %xmm15
+; SSE-NEXT:    packuswb %xmm15, %xmm15
+; SSE-NEXT:    pand %xmm4, %xmm15
+; SSE-NEXT:    por %xmm12, %xmm15
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm14, %xmm10
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm14 = xmm14[8],xmm5[8],xmm14[9],xmm5[9],xmm14[10],xmm5[10],xmm14[11],xmm5[11],xmm14[12],xmm5[12],xmm14[13],xmm5[13],xmm14[14],xmm5[14],xmm14[15],xmm5[15]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm14[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,5,5,7,4]
+; SSE-NEXT:    movdqa %xmm1, %xmm12
+; SSE-NEXT:    pandn %xmm11, %xmm12
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm5[0],xmm10[1],xmm5[1],xmm10[2],xmm5[2],xmm10[3],xmm5[3],xmm10[4],xmm5[4],xmm10[5],xmm5[5],xmm10[6],xmm5[6],xmm10[7],xmm5[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,3,1,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm11, %xmm10
-; SSE-NEXT:    por %xmm10, %xmm14
-; SSE-NEXT:    packuswb %xmm14, %xmm10
-; SSE-NEXT:    movdqa %xmm1, %xmm12
-; SSE-NEXT:    pandn %xmm10, %xmm12
-; SSE-NEXT:    pand %xmm1, %xmm13
-; SSE-NEXT:    por %xmm13, %xmm12
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm9, %xmm10
-; SSE-NEXT:    pxor %xmm3, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm3[8],xmm10[9],xmm3[9],xmm10[10],xmm3[10],xmm10[11],xmm3[11],xmm10[12],xmm3[12],xmm10[13],xmm3[13],xmm10[14],xmm3[14],xmm10[15],xmm3[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm3[0],xmm9[1],xmm3[1],xmm9[2],xmm3[2],xmm9[3],xmm3[3],xmm9[4],xmm3[4],xmm9[5],xmm3[5],xmm9[6],xmm3[6],xmm9[7],xmm3[7]
-; SSE-NEXT:    pxor %xmm4, %xmm4
-; SSE-NEXT:    movdqa %xmm9, %xmm13
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[1,0],xmm10[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[2,0],xmm10[2,3]
-; SSE-NEXT:    psrlq $48, %xmm10
-; SSE-NEXT:    psrldq {{.*#+}} xmm13 = xmm13[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm10[0],xmm13[1],xmm10[1],xmm13[2],xmm10[2],xmm13[3],xmm10[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm9[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,4,4,5,7]
-; SSE-NEXT:    packuswb %xmm13, %xmm10
-; SSE-NEXT:    movdqa %xmm15, %xmm13
-; SSE-NEXT:    pandn %xmm10, %xmm13
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm3, %xmm10
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm4[8],xmm10[9],xmm4[9],xmm10[10],xmm4[10],xmm10[11],xmm4[11],xmm10[12],xmm4[12],xmm10[13],xmm4[13],xmm10[14],xmm4[14],xmm10[15],xmm4[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[1,1,2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    movdqa %xmm0, %xmm9
-; SSE-NEXT:    pandn %xmm10, %xmm9
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm3[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm10[3,1,1,2,4,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm14
-; SSE-NEXT:    por %xmm9, %xmm14
-; SSE-NEXT:    packuswb %xmm14, %xmm14
-; SSE-NEXT:    pand %xmm15, %xmm14
-; SSE-NEXT:    por %xmm13, %xmm14
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm3, %xmm9
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm4[8],xmm3[9],xmm4[9],xmm3[10],xmm4[10],xmm3[11],xmm4[11],xmm3[12],xmm4[12],xmm3[13],xmm4[13],xmm3[14],xmm4[14],xmm3[15],xmm4[15]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm3[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,5,5,7,4]
-; SSE-NEXT:    movdqa %xmm11, %xmm13
-; SSE-NEXT:    pandn %xmm10, %xmm13
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm4[0],xmm9[1],xmm4[1],xmm9[2],xmm4[2],xmm9[3],xmm4[3],xmm9[4],xmm4[4],xmm9[5],xmm4[5],xmm9[6],xmm4[6],xmm9[7],xmm4[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,3,1,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm11, %xmm9
-; SSE-NEXT:    por %xmm9, %xmm13
-; SSE-NEXT:    packuswb %xmm13, %xmm9
-; SSE-NEXT:    movdqa %xmm1, %xmm13
-; SSE-NEXT:    pandn %xmm9, %xmm13
-; SSE-NEXT:    pand %xmm1, %xmm14
-; SSE-NEXT:    por %xmm14, %xmm13
-; SSE-NEXT:    movdqa (%rsp), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm3, %xmm9
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm4[8],xmm9[9],xmm4[9],xmm9[10],xmm4[10],xmm9[11],xmm4[11],xmm9[12],xmm4[12],xmm9[13],xmm4[13],xmm9[14],xmm4[14],xmm9[15],xmm4[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
+; SSE-NEXT:    pand %xmm1, %xmm10
+; SSE-NEXT:    por %xmm10, %xmm12
+; SSE-NEXT:    packuswb %xmm12, %xmm11
 ; SSE-NEXT:    movdqa %xmm3, %xmm10
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,0],xmm9[0,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm9[2,3]
-; SSE-NEXT:    psrlq $48, %xmm9
-; SSE-NEXT:    psrldq {{.*#+}} xmm10 = xmm10[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm3[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,4,5,7]
-; SSE-NEXT:    packuswb %xmm10, %xmm9
-; SSE-NEXT:    movdqa %xmm6, %xmm10
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm5[8],xmm10[9],xmm5[9],xmm10[10],xmm5[10],xmm10[11],xmm5[11],xmm10[12],xmm5[12],xmm10[13],xmm5[13],xmm10[14],xmm5[14],xmm10[15],xmm5[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[1,1,2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    pandn %xmm11, %xmm10
+; SSE-NEXT:    pand %xmm3, %xmm15
+; SSE-NEXT:    por %xmm15, %xmm10
+; SSE-NEXT:    movdqa (%rsp), %xmm8 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm8, %xmm11
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm5[8],xmm11[9],xmm5[9],xmm11[10],xmm5[10],xmm11[11],xmm5[11],xmm11[12],xmm5[12],xmm11[13],xmm5[13],xmm11[14],xmm5[14],xmm11[15],xmm5[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm5[0],xmm8[1],xmm5[1],xmm8[2],xmm5[2],xmm8[3],xmm5[3],xmm8[4],xmm5[4],xmm8[5],xmm5[5],xmm8[6],xmm5[6],xmm8[7],xmm5[7]
+; SSE-NEXT:    movdqa %xmm8, %xmm12
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[1,0],xmm11[0,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,0],xmm11[2,3]
+; SSE-NEXT:    psrlq $48, %xmm11
+; SSE-NEXT:    psrldq {{.*#+}} xmm12 = xmm12[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm11[0],xmm12[1],xmm11[1],xmm12[2],xmm11[2],xmm12[3],xmm11[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm8[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,4,5,7]
+; SSE-NEXT:    packuswb %xmm12, %xmm11
+; SSE-NEXT:    movdqa %xmm6, %xmm12
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm5[8],xmm12[9],xmm5[9],xmm12[10],xmm5[10],xmm12[11],xmm5[11],xmm12[12],xmm5[12],xmm12[13],xmm5[13],xmm12[14],xmm5[14],xmm12[15],xmm5[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[1,1,2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1],xmm6[2],xmm5[2],xmm6[3],xmm5[3],xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm6[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm14[0,2,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm14[3,1,1,2,4,5,6,7]
-; SSE-NEXT:    pand %xmm0, %xmm14
-; SSE-NEXT:    pandn %xmm10, %xmm0
-; SSE-NEXT:    por %xmm14, %xmm0
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm15, %xmm0
-; SSE-NEXT:    pandn %xmm9, %xmm15
+; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm6[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm15[0,2,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm15[3,1,1,2,4,5,6,7]
+; SSE-NEXT:    pand %xmm0, %xmm15
+; SSE-NEXT:    pandn %xmm12, %xmm0
 ; SSE-NEXT:    por %xmm15, %xmm0
-; SSE-NEXT:    movdqa %xmm2, %xmm4
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3],xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
+; SSE-NEXT:    packuswb %xmm0, %xmm0
+; SSE-NEXT:    pand %xmm4, %xmm0
+; SSE-NEXT:    pandn %xmm11, %xmm4
+; SSE-NEXT:    por %xmm0, %xmm4
+; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,3,1,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    pand %xmm11, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,1,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm1, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm2[0,1,2,3,7,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,5,7,4]
-; SSE-NEXT:    pandn %xmm5, %xmm11
-; SSE-NEXT:    por %xmm4, %xmm11
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    packuswb %xmm11, %xmm4
-; SSE-NEXT:    pandn %xmm4, %xmm1
+; SSE-NEXT:    pandn %xmm5, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
+; SSE-NEXT:    pand %xmm3, %xmm4
+; SSE-NEXT:    packuswb %xmm1, %xmm0
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    por %xmm4, %xmm3
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 16(%rsi)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -5913,11 +5903,11 @@ define void @load_i8_stride6_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, (%r9)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movdqa %xmm1, 16(%rax)
-; SSE-NEXT:    movdqa %xmm13, 32(%rax)
-; SSE-NEXT:    movdqa %xmm12, 48(%rax)
-; SSE-NEXT:    movdqa %xmm8, (%rax)
-; SSE-NEXT:    addq $792, %rsp # imm = 0x318
+; SSE-NEXT:    movdqa %xmm3, 16(%rax)
+; SSE-NEXT:    movdqa %xmm10, 32(%rax)
+; SSE-NEXT:    movdqa %xmm13, 48(%rax)
+; SSE-NEXT:    movdqa %xmm9, (%rax)
+; SSE-NEXT:    addq $840, %rsp # imm = 0x348
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i8_stride6_vf64:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-7.ll
index 0ee10a33c1d0c..56105e4c5dde7 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-7.ll
@@ -942,12 +942,11 @@ define void @load_i8_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pandn %xmm4, %xmm2
 ; SSE-NEXT:    movdqa %xmm4, %xmm11
 ; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    pxor %xmm1, %xmm1
+; SSE-NEXT:    pxor %xmm4, %xmm4
 ; SSE-NEXT:    movdqa %xmm2, %xmm5
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm1[8],xmm5[9],xmm1[9],xmm5[10],xmm1[10],xmm5[11],xmm1[11],xmm5[12],xmm1[12],xmm5[13],xmm1[13],xmm5[14],xmm1[14],xmm5[15],xmm1[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm4[8],xmm5[9],xmm4[9],xmm5[10],xmm4[10],xmm5[11],xmm4[11],xmm5[12],xmm4[12],xmm5[13],xmm4[13],xmm5[14],xmm4[14],xmm5[15],xmm4[15]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,65535,65535,0,65535,0,65535]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
-; SSE-NEXT:    pxor %xmm4, %xmm4
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3],xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7]
 ; SSE-NEXT:    pand %xmm7, %xmm2
 ; SSE-NEXT:    pandn %xmm5, %xmm7
 ; SSE-NEXT:    por %xmm2, %xmm7
@@ -973,9 +972,9 @@ define void @load_i8_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movdqa %xmm11, %xmm1
 ; SSE-NEXT:    pand %xmm5, %xmm10
 ; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,65535,0,65535,65535,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm6, %xmm4
-; SSE-NEXT:    pand %xmm12, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm6, %xmm11
+; SSE-NEXT:    pand %xmm12, %xmm11
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pandn %xmm0, %xmm12
 ; SSE-NEXT:    movaps %xmm0, %xmm14
 ; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[1,0],xmm6[0,0]
@@ -984,8 +983,7 @@ define void @load_i8_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pandn %xmm6, %xmm5
 ; SSE-NEXT:    movdqa %xmm6, %xmm15
-; SSE-NEXT:    pxor %xmm0, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm15 = xmm15[8],xmm0[8],xmm15[9],xmm0[9],xmm15[10],xmm0[10],xmm15[11],xmm0[11],xmm15[12],xmm0[12],xmm15[13],xmm0[13],xmm15[14],xmm0[14],xmm15[15],xmm0[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm15 = xmm15[8],xmm4[8],xmm15[9],xmm4[9],xmm15[10],xmm4[10],xmm15[11],xmm4[11],xmm15[12],xmm4[12],xmm15[13],xmm4[13],xmm15[14],xmm4[14],xmm15[15],xmm4[15]
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm15[0],xmm9[1],xmm15[1],xmm9[2],xmm15[2],xmm9[3],xmm15[3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm9[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,2,1]
@@ -997,21 +995,20 @@ define void @load_i8_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    por %xmm7, %xmm13
 ; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,65535,0,65535,65535,0,65535]
 ; SSE-NEXT:    movdqa %xmm7, %xmm9
-; SSE-NEXT:    movdqa %xmm1, %xmm4
+; SSE-NEXT:    movdqa %xmm1, %xmm6
 ; SSE-NEXT:    pandn %xmm1, %xmm9
 ; SSE-NEXT:    movdqa %xmm3, %xmm2
 ; SSE-NEXT:    pand %xmm7, %xmm3
 ; SSE-NEXT:    por %xmm9, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, %xmm9
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1],xmm9[2],xmm0[2],xmm9[3],xmm0[3],xmm9[4],xmm0[4],xmm9[5],xmm0[5],xmm9[6],xmm0[6],xmm9[7],xmm0[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm4[0],xmm9[1],xmm4[1],xmm9[2],xmm4[2],xmm9[3],xmm4[3],xmm9[4],xmm4[4],xmm9[5],xmm4[5],xmm9[6],xmm4[6],xmm9[7],xmm4[7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,0,65535,65535,65535,65535,0,65535]
 ; SSE-NEXT:    movdqa %xmm0, %xmm1
 ; SSE-NEXT:    pandn %xmm9, %xmm1
-; SSE-NEXT:    pxor %xmm6, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm6[8],xmm3[9],xmm6[9],xmm3[10],xmm6[10],xmm3[11],xmm6[11],xmm3[12],xmm6[12],xmm3[13],xmm6[13],xmm3[14],xmm6[14],xmm3[15],xmm6[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm4[8],xmm3[9],xmm4[9],xmm3[10],xmm4[10],xmm3[11],xmm4[11],xmm3[12],xmm4[12],xmm3[13],xmm4[13],xmm3[14],xmm4[14],xmm3[15],xmm4[15]
 ; SSE-NEXT:    pand %xmm0, %xmm3
 ; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3],xmm8[4],xmm6[4],xmm8[5],xmm6[5],xmm8[6],xmm6[6],xmm8[7],xmm6[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm4[0],xmm8[1],xmm4[1],xmm8[2],xmm4[2],xmm8[3],xmm4[3],xmm8[4],xmm4[4],xmm8[5],xmm4[5],xmm8[6],xmm4[6],xmm8[7],xmm4[7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,65535,65535,65535,65535]
 ; SSE-NEXT:    movdqa %xmm8, %xmm9
 ; SSE-NEXT:    pand %xmm1, %xmm9
@@ -1028,12 +1025,13 @@ define void @load_i8_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
 ; SSE-NEXT:    pand %xmm11, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm9
+; SSE-NEXT:    movdqa %xmm2, %xmm3
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    pandn %xmm2, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm10
 ; SSE-NEXT:    movdqa %xmm10, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm6[8],xmm1[9],xmm6[9],xmm1[10],xmm6[10],xmm1[11],xmm6[11],xmm1[12],xmm6[12],xmm1[13],xmm6[13],xmm1[14],xmm6[14],xmm1[15],xmm6[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm6[0],xmm10[1],xmm6[1],xmm10[2],xmm6[2],xmm10[3],xmm6[3],xmm10[4],xmm6[4],xmm10[5],xmm6[5],xmm10[6],xmm6[6],xmm10[7],xmm6[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm4[8],xmm1[9],xmm4[9],xmm1[10],xmm4[10],xmm1[11],xmm4[11],xmm1[12],xmm4[12],xmm1[13],xmm4[13],xmm1[14],xmm4[14],xmm1[15],xmm4[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm4[0],xmm10[1],xmm4[1],xmm10[2],xmm4[2],xmm10[3],xmm4[3],xmm10[4],xmm4[4],xmm10[5],xmm4[5],xmm10[6],xmm4[6],xmm10[7],xmm4[7]
 ; SSE-NEXT:    pand %xmm0, %xmm10
 ; SSE-NEXT:    pandn %xmm1, %xmm0
 ; SSE-NEXT:    por %xmm10, %xmm0
@@ -1054,14 +1052,15 @@ define void @load_i8_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pandn %xmm1, %xmm8
 ; SSE-NEXT:    por %xmm0, %xmm8
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,0,65535,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm4, %xmm1
+; SSE-NEXT:    movdqa %xmm6, %xmm2
+; SSE-NEXT:    movdqa %xmm6, %xmm1
 ; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm2, %xmm10
+; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm3, %xmm10
 ; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3],xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm6[8],xmm0[9],xmm6[9],xmm0[10],xmm6[10],xmm0[11],xmm6[11],xmm0[12],xmm6[12],xmm0[13],xmm6[13],xmm0[14],xmm6[14],xmm0[15],xmm6[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm4[8],xmm0[9],xmm4[9],xmm0[10],xmm4[10],xmm0[11],xmm4[11],xmm0[12],xmm4[12],xmm0[13],xmm4[13],xmm0[14],xmm4[14],xmm0[15],xmm4[15]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,0,65535,0,65535,65535,65535,65535]
 ; SSE-NEXT:    pand %xmm3, %xmm0
 ; SSE-NEXT:    pandn %xmm1, %xmm3
@@ -1073,7 +1072,7 @@ define void @load_i8_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
 ; SSE-NEXT:    pand %xmm11, %xmm0
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm6[0],xmm12[1],xmm6[1],xmm12[2],xmm6[2],xmm12[3],xmm6[3],xmm12[4],xmm6[4],xmm12[5],xmm6[5],xmm12[6],xmm6[6],xmm12[7],xmm6[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm4[0],xmm12[1],xmm4[1],xmm12[2],xmm4[2],xmm12[3],xmm4[3],xmm12[4],xmm4[4],xmm12[5],xmm4[5],xmm12[6],xmm4[6],xmm12[7],xmm4[7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm12[0,1,2,3,7,5,6,7]
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm12 = xmm12[4],xmm15[4],xmm12[5],xmm15[5],xmm12[6],xmm15[6],xmm12[7],xmm15[7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm12[0,3,2,3,4,5,6,7]
@@ -1084,14 +1083,11 @@ define void @load_i8_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    por %xmm0, %xmm11
 ; SSE-NEXT:    movdqa %xmm11, %xmm6
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    movdqa %xmm2, %xmm3
 ; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm10, %xmm11
 ; SSE-NEXT:    pandn %xmm10, %xmm0
 ; SSE-NEXT:    por %xmm3, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    pxor %xmm4, %xmm4
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm4[8],xmm3[9],xmm4[9],xmm3[10],xmm4[10],xmm3[11],xmm4[11],xmm3[12],xmm4[12],xmm3[13],xmm4[13],xmm3[14],xmm4[14],xmm3[15],xmm4[15]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
@@ -1105,9 +1101,8 @@ define void @load_i8_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pandn %xmm15, %xmm3
 ; SSE-NEXT:    por %xmm1, %xmm3
 ; SSE-NEXT:    packuswb %xmm3, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm0[0,3,2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[1,3,2,3]
-; SSE-NEXT:    movdqa %xmm11, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[0,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[1,3,2,3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,2,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
@@ -1129,17 +1124,19 @@ define void @load_i8_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm1[0,3,2,3]
 ; SSE-NEXT:    movdqa %xmm2, %xmm0
 ; SSE-NEXT:    pand %xmm7, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm7
+; SSE-NEXT:    pandn %xmm10, %xmm7
 ; SSE-NEXT:    por %xmm0, %xmm7
 ; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSE-NEXT:    pxor %xmm1, %xmm1
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,4,6,7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm4[8],xmm7[9],xmm4[9],xmm7[10],xmm4[10],xmm7[11],xmm4[11],xmm7[12],xmm4[12],xmm7[13],xmm4[13],xmm7[14],xmm4[14],xmm7[15],xmm4[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm1[8],xmm7[9],xmm1[9],xmm7[10],xmm1[10],xmm7[11],xmm1[11],xmm7[12],xmm1[12],xmm7[13],xmm1[13],xmm7[14],xmm1[14],xmm7[15],xmm1[15]
+; SSE-NEXT:    pxor %xmm2, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,7,6,7]
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1],xmm5[2],xmm4[2],xmm5[3],xmm4[3],xmm5[4],xmm4[4],xmm5[5],xmm4[5],xmm5[6],xmm4[6],xmm5[7],xmm4[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3],xmm5[4],xmm2[4],xmm5[5],xmm2[5],xmm5[6],xmm2[6],xmm5[7],xmm2[7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,0,65535,65535,65535,65535,65535,65535]
 ; SSE-NEXT:    pand %xmm1, %xmm5
 ; SSE-NEXT:    pandn %xmm15, %xmm1
@@ -1154,7 +1151,7 @@ define void @load_i8_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movq %xmm9, (%rdx)
 ; SSE-NEXT:    movq %xmm8, (%rcx)
 ; SSE-NEXT:    movq %xmm6, (%r8)
-; SSE-NEXT:    movq %xmm10, (%r9)
+; SSE-NEXT:    movq %xmm3, (%r9)
 ; SSE-NEXT:    movq %xmm11, (%rdi)
 ; SSE-NEXT:    movq %xmm0, (%rax)
 ; SSE-NEXT:    retq
@@ -1862,25 +1859,25 @@ define void @load_i8_stride7_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i8_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6) nounwind {
 ; SSE-LABEL: load_i8_stride7_vf16:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $168, %rsp
-; SSE-NEXT:    movdqa 96(%rdi), %xmm15
-; SSE-NEXT:    movdqa 80(%rdi), %xmm4
-; SSE-NEXT:    movdqa 64(%rdi), %xmm7
-; SSE-NEXT:    movdqa (%rdi), %xmm6
-; SSE-NEXT:    movdqa 16(%rdi), %xmm3
-; SSE-NEXT:    movdqa 32(%rdi), %xmm1
-; SSE-NEXT:    movdqa 48(%rdi), %xmm8
+; SSE-NEXT:    subq $200, %rsp
+; SSE-NEXT:    movdqa 96(%rdi), %xmm10
+; SSE-NEXT:    movdqa 80(%rdi), %xmm5
+; SSE-NEXT:    movdqa 64(%rdi), %xmm6
+; SSE-NEXT:    movdqa (%rdi), %xmm4
+; SSE-NEXT:    movdqa 16(%rdi), %xmm8
+; SSE-NEXT:    movdqa 32(%rdi), %xmm7
+; SSE-NEXT:    movdqa 48(%rdi), %xmm13
 ; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,0,65535,65535,65535,0,65535,65535]
 ; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm1, %xmm12
-; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    pandn %xmm7, %xmm0
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm13, %xmm1
 ; SSE-NEXT:    pand %xmm2, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pxor %xmm13, %xmm13
+; SSE-NEXT:    pxor %xmm11, %xmm11
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm13[8],xmm0[9],xmm13[9],xmm0[10],xmm13[10],xmm0[11],xmm13[11],xmm0[12],xmm13[12],xmm0[13],xmm13[13],xmm0[14],xmm13[14],xmm0[15],xmm13[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1],xmm1[2],xmm13[2],xmm1[3],xmm13[3],xmm1[4],xmm13[4],xmm1[5],xmm13[5],xmm1[6],xmm13[6],xmm1[7],xmm13[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm11[8],xmm0[9],xmm11[9],xmm0[10],xmm11[10],xmm0[11],xmm11[11],xmm0[12],xmm11[12],xmm0[13],xmm11[13],xmm0[14],xmm11[14],xmm0[15],xmm11[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1],xmm1[2],xmm11[2],xmm1[3],xmm11[3],xmm1[4],xmm11[4],xmm1[5],xmm11[5],xmm1[6],xmm11[6],xmm1[7],xmm11[7]
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
@@ -1888,57 +1885,54 @@ define void @load_i8_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,0,65535,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    movdqa %xmm3, %xmm9
-; SSE-NEXT:    movdqa %xmm6, %xmm3
-; SSE-NEXT:    movdqa %xmm6, %xmm11
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm10, %xmm3
-; SSE-NEXT:    por %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm13[8],xmm0[9],xmm13[9],xmm0[10],xmm13[10],xmm0[11],xmm13[11],xmm0[12],xmm13[12],xmm0[13],xmm13[13],xmm0[14],xmm13[14],xmm0[15],xmm13[15]
-; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [65535,65535,65535,65535,0,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm14, %xmm6
-; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm13[0],xmm3[1],xmm13[1],xmm3[2],xmm13[2],xmm3[3],xmm13[3],xmm3[4],xmm13[4],xmm3[5],xmm13[5],xmm3[6],xmm13[6],xmm3[7],xmm13[7]
-; SSE-NEXT:    pand %xmm14, %xmm3
-; SSE-NEXT:    por %xmm6, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[0,2,1,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,1,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm2, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,0,65535,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    pandn %xmm8, %xmm2
+; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    movdqa %xmm4, %xmm14
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm9, %xmm3
+; SSE-NEXT:    por %xmm2, %xmm3
+; SSE-NEXT:    movdqa %xmm3, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm11[8],xmm2[9],xmm11[9],xmm2[10],xmm11[10],xmm2[11],xmm11[11],xmm2[12],xmm11[12],xmm2[13],xmm11[13],xmm2[14],xmm11[14],xmm2[15],xmm11[15]
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,65535,65535,65535,0,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm12, %xmm4
+; SSE-NEXT:    pandn %xmm2, %xmm4
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm11[0],xmm3[1],xmm11[1],xmm3[2],xmm11[2],xmm3[3],xmm11[3],xmm3[4],xmm11[4],xmm3[5],xmm11[5],xmm3[6],xmm11[6],xmm3[7],xmm11[7]
+; SSE-NEXT:    pand %xmm12, %xmm3
+; SSE-NEXT:    por %xmm4, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,2,1,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,3,1,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,1,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm2, %xmm2
+; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm0
 ; SSE-NEXT:    por %xmm2, %xmm0
 ; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,0,65535,65535,0,65535]
 ; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    pandn %xmm7, %xmm1
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    movdqa %xmm4, %xmm5
+; SSE-NEXT:    pandn %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm5, %xmm2
 ; SSE-NEXT:    pand %xmm3, %xmm2
-; SSE-NEXT:    movdqa %xmm3, %xmm13
+; SSE-NEXT:    movdqa %xmm3, %xmm11
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    pxor %xmm6, %xmm6
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3],xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
+; SSE-NEXT:    pxor %xmm3, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,4,7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm6[8],xmm2[9],xmm6[9],xmm2[10],xmm6[10],xmm2[11],xmm6[11],xmm2[12],xmm6[12],xmm2[13],xmm6[13],xmm2[14],xmm6[14],xmm2[15],xmm6[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm3[8],xmm2[9],xmm3[9],xmm2[10],xmm3[10],xmm2[11],xmm3[11],xmm2[12],xmm3[12],xmm2[13],xmm3[13],xmm2[14],xmm3[14],xmm2[15],xmm3[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,7,7]
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm15, %xmm2
-; SSE-NEXT:    movdqa %xmm15, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm6[8],xmm3[9],xmm6[9],xmm3[10],xmm6[10],xmm3[11],xmm6[11],xmm3[12],xmm6[12],xmm3[13],xmm6[13],xmm3[14],xmm6[14],xmm3[15],xmm6[15]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3],xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
+; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    movdqa %xmm10, %xmm15
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm15 = xmm15[8],xmm3[8],xmm15[9],xmm3[9],xmm15[10],xmm3[10],xmm15[11],xmm3[11],xmm15[12],xmm3[12],xmm15[13],xmm3[13],xmm15[14],xmm3[14],xmm15[15],xmm3[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3],xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pxor %xmm15, %xmm15
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; SSE-NEXT:    pxor %xmm10, %xmm10
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm15[0],xmm2[1],xmm15[1],xmm2[2],xmm15[2],xmm2[3],xmm15[3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
@@ -1955,13 +1949,13 @@ define void @load_i8_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,0,65535,65535,0,65535,65535]
 ; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pandn %xmm12, %xmm0
-; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    pandn %xmm7, %xmm0
+; SSE-NEXT:    movdqa %xmm13, %xmm1
 ; SSE-NEXT:    pand %xmm2, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm15[8],xmm0[9],xmm15[9],xmm0[10],xmm15[10],xmm0[11],xmm15[11],xmm0[12],xmm15[12],xmm0[13],xmm15[13],xmm0[14],xmm15[14],xmm0[15],xmm15[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1],xmm1[2],xmm15[2],xmm1[3],xmm15[3],xmm1[4],xmm15[4],xmm1[5],xmm15[5],xmm1[6],xmm15[6],xmm1[7],xmm15[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm10[8],xmm0[9],xmm10[9],xmm0[10],xmm10[10],xmm0[11],xmm10[11],xmm0[12],xmm10[12],xmm0[13],xmm10[13],xmm0[14],xmm10[14],xmm0[15],xmm10[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1],xmm1[2],xmm10[2],xmm1[3],xmm10[3],xmm1[4],xmm10[4],xmm1[5],xmm10[5],xmm1[6],xmm10[6],xmm1[7],xmm10[7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,65535,65535,65535,65535]
 ; SSE-NEXT:    pand %xmm2, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm2
@@ -1970,417 +1964,422 @@ define void @load_i8_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,6]
 ; SSE-NEXT:    psrld $16, %xmm0
 ; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,255,0,0,0,0,255,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,0,0,0,0,255,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    movdqa %xmm2, %xmm4
 ; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm13, %xmm1
-; SSE-NEXT:    pandn %xmm9, %xmm1
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm2
-; SSE-NEXT:    pand %xmm13, %xmm2
-; SSE-NEXT:    movdqa %xmm13, %xmm11
+; SSE-NEXT:    movdqa %xmm11, %xmm1
+; SSE-NEXT:    pandn %xmm8, %xmm1
+; SSE-NEXT:    movdqa %xmm8, %xmm7
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    pand %xmm11, %xmm2
+; SSE-NEXT:    movdqa %xmm11, %xmm8
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1],xmm1[2],xmm15[2],xmm1[3],xmm15[3],xmm1[4],xmm15[4],xmm1[5],xmm15[5],xmm1[6],xmm15[6],xmm1[7],xmm15[7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,0,65535,65535,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm6, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1],xmm1[2],xmm10[2],xmm1[3],xmm10[3],xmm1[4],xmm10[4],xmm1[5],xmm10[5],xmm1[6],xmm10[6],xmm1[7],xmm10[7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,0,65535,65535,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm11, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm15[8],xmm2[9],xmm15[9],xmm2[10],xmm15[10],xmm2[11],xmm15[11],xmm2[12],xmm15[12],xmm2[13],xmm15[13],xmm2[14],xmm15[14],xmm2[15],xmm15[15]
-; SSE-NEXT:    pand %xmm6, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm10[8],xmm2[9],xmm10[9],xmm2[10],xmm10[10],xmm2[11],xmm10[11],xmm2[12],xmm10[12],xmm2[13],xmm10[13],xmm2[14],xmm10[14],xmm2[15],xmm10[15]
+; SSE-NEXT:    pand %xmm11, %xmm2
 ; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    packuswb %xmm13, %xmm13
-; SSE-NEXT:    pand %xmm4, %xmm13
-; SSE-NEXT:    por %xmm0, %xmm13
-; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm1[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    packuswb %xmm14, %xmm14
+; SSE-NEXT:    pand %xmm4, %xmm14
+; SSE-NEXT:    por %xmm0, %xmm14
+; SSE-NEXT:    movdqa %xmm9, %xmm0
 ; SSE-NEXT:    pandn %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm5, %xmm6
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    pand %xmm10, %xmm1
+; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    pand %xmm9, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
 ; SSE-NEXT:    pxor %xmm2, %xmm2
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
-; SSE-NEXT:    pand %xmm14, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm14
-; SSE-NEXT:    por %xmm1, %xmm14
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm12, %xmm1
+; SSE-NEXT:    pandn %xmm0, %xmm12
+; SSE-NEXT:    por %xmm1, %xmm12
 ; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    pand %xmm10, %xmm0
-; SSE-NEXT:    pandn %xmm8, %xmm10
-; SSE-NEXT:    por %xmm0, %xmm10
-; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [65535,0,65535,65535,0,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm9, %xmm7
-; SSE-NEXT:    pand %xmm14, %xmm7
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    pandn %xmm13, %xmm9
+; SSE-NEXT:    por %xmm0, %xmm9
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,0,65535,65535,0,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm7, %xmm4
+; SSE-NEXT:    pand %xmm11, %xmm4
+; SSE-NEXT:    movdqa %xmm5, %xmm1
 ; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm5, %xmm15
-; SSE-NEXT:    pand %xmm14, %xmm15
-; SSE-NEXT:    movdqa %xmm11, %xmm3
-; SSE-NEXT:    pandn %xmm8, %xmm3
+; SSE-NEXT:    movdqa %xmm5, %xmm12
+; SSE-NEXT:    pand %xmm11, %xmm12
+; SSE-NEXT:    movdqa %xmm8, %xmm7
+; SSE-NEXT:    movdqa %xmm8, %xmm0
+; SSE-NEXT:    pandn %xmm13, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm13[0,2,2,3]
+; SSE-NEXT:    movdqa %xmm13, %xmm5
+; SSE-NEXT:    pand %xmm11, %xmm13
+; SSE-NEXT:    movdqa %xmm11, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm2, %xmm11
+; SSE-NEXT:    por %xmm13, %xmm11
+; SSE-NEXT:    movdqa %xmm15, %xmm8
+; SSE-NEXT:    pslld $16, %xmm8
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm13, %xmm2
+; SSE-NEXT:    psrldq {{.*#+}} xmm2 = xmm2[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm15[0],xmm2[1],xmm15[1],xmm2[2],xmm15[2],xmm2[3],xmm15[3]
+; SSE-NEXT:    movdqa %xmm13, %xmm3
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm15[4],xmm3[5],xmm15[5],xmm3[6],xmm15[6],xmm3[7],xmm15[7]
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm8[0,2,2,3]
-; SSE-NEXT:    movdqa %xmm8, %xmm4
-; SSE-NEXT:    pand %xmm14, %xmm8
-; SSE-NEXT:    movdqa %xmm14, %xmm9
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pandn %xmm12, %xmm14
-; SSE-NEXT:    por %xmm8, %xmm14
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm0, %xmm5
-; SSE-NEXT:    pslld $16, %xmm5
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm8, %xmm3
-; SSE-NEXT:    psrldq {{.*#+}} xmm3 = xmm3[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm8, %xmm1
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm1[0],xmm10[1],xmm1[1],xmm10[2],xmm1[2],xmm10[3],xmm1[3],xmm10[4],xmm1[4],xmm10[5],xmm1[5],xmm10[6],xmm1[6],xmm10[7],xmm1[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm10[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm2[0,1,2,3,6,4,6,5]
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,65535,65535,65535,65535,65535,0,65535]
-; SSE-NEXT:    pand %xmm12, %xmm10
-; SSE-NEXT:    movdqa %xmm8, %xmm2
-; SSE-NEXT:    pand %xmm12, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, (%rsp) # 16-byte Spill
-; SSE-NEXT:    pandn %xmm0, %xmm12
-; SSE-NEXT:    movdqa %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1],xmm14[2],xmm1[2],xmm14[3],xmm1[3],xmm14[4],xmm1[4],xmm14[5],xmm1[5],xmm14[6],xmm1[6],xmm14[7],xmm1[7]
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pxor %xmm0, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1],xmm9[2],xmm0[2],xmm9[3],xmm0[3],xmm9[4],xmm0[4],xmm9[5],xmm0[5],xmm9[6],xmm0[6],xmm9[7],xmm0[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm9[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm3[0,1,2,3,6,4,6,5]
+; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,65535,65535,65535,65535,0,65535]
+; SSE-NEXT:    pand %xmm9, %xmm10
+; SSE-NEXT:    movdqa %xmm13, %xmm3
+; SSE-NEXT:    pand %xmm9, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm15, %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1],xmm11[2],xmm0[2],xmm11[3],xmm0[3],xmm11[4],xmm0[4],xmm11[5],xmm0[5],xmm11[6],xmm0[6],xmm11[7],xmm0[7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE-NEXT:    pand %xmm0, %xmm14
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm11
+; SSE-NEXT:    pand %xmm0, %xmm15
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pandn %xmm8, %xmm0
+; SSE-NEXT:    pandn %xmm13, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm5[0],xmm8[1],xmm5[1],xmm8[2],xmm5[2],xmm8[3],xmm5[3]
-; SSE-NEXT:    packuswb %xmm8, %xmm5
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm8[0],xmm13[1],xmm8[1],xmm13[2],xmm8[2],xmm13[3],xmm8[3]
+; SSE-NEXT:    packuswb %xmm13, %xmm8
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm0, %xmm8
-; SSE-NEXT:    pandn %xmm5, %xmm8
-; SSE-NEXT:    pshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm5 = mem[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,0,3,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,4,6,7]
-; SSE-NEXT:    packuswb %xmm5, %xmm5
-; SSE-NEXT:    pand %xmm0, %xmm5
-; SSE-NEXT:    por %xmm5, %xmm8
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0]
-; SSE-NEXT:    movdqa %xmm5, %xmm0
-; SSE-NEXT:    pandn %xmm8, %xmm0
-; SSE-NEXT:    pand %xmm5, %xmm13
-; SSE-NEXT:    por %xmm13, %xmm0
+; SSE-NEXT:    movdqa %xmm0, %xmm13
+; SSE-NEXT:    pandn %xmm8, %xmm13
+; SSE-NEXT:    pshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = mem[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[0,0,3,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,5,4,6,7]
+; SSE-NEXT:    packuswb %xmm8, %xmm8
+; SSE-NEXT:    pand %xmm0, %xmm8
+; SSE-NEXT:    movdqa %xmm0, %xmm9
+; SSE-NEXT:    por %xmm8, %xmm13
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0]
+; SSE-NEXT:    movdqa %xmm8, %xmm0
+; SSE-NEXT:    pandn %xmm13, %xmm0
+; SSE-NEXT:    pand %xmm8, %xmm14
+; SSE-NEXT:    por %xmm14, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm2, %xmm8
-; SSE-NEXT:    pandn %xmm6, %xmm8
+; SSE-NEXT:    movdqa %xmm7, %xmm14
+; SSE-NEXT:    movdqa %xmm7, %xmm13
+; SSE-NEXT:    pandn %xmm1, %xmm13
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm2, %xmm0
-; SSE-NEXT:    por %xmm8, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm8
-; SSE-NEXT:    pxor %xmm6, %xmm6
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3],xmm8[4],xmm6[4],xmm8[5],xmm6[5],xmm8[6],xmm6[6],xmm8[7],xmm6[7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [65535,0,65535,65535,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm13, %xmm1
-; SSE-NEXT:    pandn %xmm8, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm6[8],xmm0[9],xmm6[9],xmm0[10],xmm6[10],xmm0[11],xmm6[11],xmm0[12],xmm6[12],xmm0[13],xmm6[13],xmm0[14],xmm6[14],xmm0[15],xmm6[15]
-; SSE-NEXT:    pxor %xmm8, %xmm8
-; SSE-NEXT:    pand %xmm13, %xmm0
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    packuswb %xmm3, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm6, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm6, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm5, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm6, %xmm9
-; SSE-NEXT:    por %xmm9, %xmm7
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm8[8],xmm1[9],xmm8[9],xmm1[10],xmm8[10],xmm1[11],xmm8[11],xmm1[12],xmm8[12],xmm1[13],xmm8[13],xmm1[14],xmm8[14],xmm1[15],xmm8[15]
-; SSE-NEXT:    movdqa %xmm13, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1],xmm7[2],xmm8[2],xmm7[3],xmm8[3],xmm7[4],xmm8[4],xmm7[5],xmm8[5],xmm7[6],xmm8[6],xmm7[7],xmm8[7]
-; SSE-NEXT:    pand %xmm13, %xmm7
-; SSE-NEXT:    por %xmm3, %xmm7
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[1,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm1[0],xmm11[1],xmm1[1]
-; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm8[8],xmm1[9],xmm8[9],xmm1[10],xmm8[10],xmm1[11],xmm8[11],xmm1[12],xmm8[12],xmm1[13],xmm8[13],xmm1[14],xmm8[14],xmm1[15],xmm8[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm8[0],xmm11[1],xmm8[1],xmm11[2],xmm8[2],xmm11[3],xmm8[3],xmm11[4],xmm8[4],xmm11[5],xmm8[5],xmm11[6],xmm8[6],xmm11[7],xmm8[7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,65535,0,65535,65535,65535]
-; SSE-NEXT:    pand %xmm3, %xmm11
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    por %xmm11, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pand %xmm7, %xmm0
+; SSE-NEXT:    por %xmm13, %xmm0
+; SSE-NEXT:    movdqa %xmm0, %xmm13
+; SSE-NEXT:    pxor %xmm7, %xmm7
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm7[0],xmm13[1],xmm7[1],xmm13[2],xmm7[2],xmm13[3],xmm7[3],xmm13[4],xmm7[4],xmm13[5],xmm7[5],xmm13[6],xmm7[6],xmm13[7],xmm7[7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,0,65535,65535,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    pandn %xmm13, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm7[8],xmm0[9],xmm7[9],xmm0[10],xmm7[10],xmm0[11],xmm7[11],xmm0[12],xmm7[12],xmm0[13],xmm7[13],xmm0[14],xmm7[14],xmm0[15],xmm7[15]
+; SSE-NEXT:    pxor %xmm13, %xmm13
+; SSE-NEXT:    pand %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm3, %xmm15
+; SSE-NEXT:    por %xmm1, %xmm0
+; SSE-NEXT:    packuswb %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    packuswb %xmm0, %xmm0
+; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm8, %xmm0
+; SSE-NEXT:    pandn %xmm2, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    movdqa (%rsp), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm7, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm4
+; SSE-NEXT:    movdqa %xmm4, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm13[8],xmm1[9],xmm13[9],xmm1[10],xmm13[10],xmm1[11],xmm13[11],xmm1[12],xmm13[12],xmm1[13],xmm13[13],xmm1[14],xmm13[14],xmm1[15],xmm13[15]
+; SSE-NEXT:    movdqa %xmm3, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm13[0],xmm4[1],xmm13[1],xmm4[2],xmm13[2],xmm4[3],xmm13[3],xmm4[4],xmm13[4],xmm4[5],xmm13[5],xmm4[6],xmm13[6],xmm4[7],xmm13[7]
+; SSE-NEXT:    pand %xmm3, %xmm4
+; SSE-NEXT:    por %xmm2, %xmm4
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm1[0],xmm6[1],xmm1[1]
+; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm13[8],xmm1[9],xmm13[9],xmm1[10],xmm13[10],xmm1[11],xmm13[11],xmm1[12],xmm13[12],xmm1[13],xmm13[13],xmm1[14],xmm13[14],xmm1[15],xmm13[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm13[0],xmm6[1],xmm13[1],xmm6[2],xmm13[2],xmm6[3],xmm13[3],xmm6[4],xmm13[4],xmm6[5],xmm13[5],xmm6[6],xmm13[6],xmm6[7],xmm13[7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,65535,0,65535,65535,65535]
+; SSE-NEXT:    pand %xmm2, %xmm6
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    por %xmm6, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT:    packuswb %xmm1, %xmm3
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,255,255,255,255,0,0,0,0,255,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm13, %xmm8
-; SSE-NEXT:    pandn %xmm3, %xmm8
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[0,2,1,3]
+; SSE-NEXT:    packuswb %xmm1, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,255,255,255,0,0,0,0,255,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm6, %xmm13
+; SSE-NEXT:    pandn %xmm2, %xmm13
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,2,1,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,3,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,0,3,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm13, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm8
-; SSE-NEXT:    pand %xmm5, %xmm8
-; SSE-NEXT:    por %xmm0, %xmm8
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pandn %xmm9, %xmm0
-; SSE-NEXT:    pand %xmm2, %xmm4
-; SSE-NEXT:    por %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm6, %xmm9
+; SSE-NEXT:    por %xmm1, %xmm13
+; SSE-NEXT:    pand %xmm8, %xmm13
+; SSE-NEXT:    por %xmm0, %xmm13
+; SSE-NEXT:    movdqa %xmm14, %xmm0
+; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    pand %xmm14, %xmm5
+; SSE-NEXT:    por %xmm0, %xmm5
+; SSE-NEXT:    movdqa %xmm5, %xmm0
 ; SSE-NEXT:    pxor %xmm1, %xmm1
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3],xmm4[4],xmm1[4],xmm4[5],xmm1[5],xmm4[6],xmm1[6],xmm4[7],xmm1[7]
-; SSE-NEXT:    pxor %xmm2, %xmm2
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm0[4],xmm4[5],xmm0[5],xmm4[6],xmm0[6],xmm4[7],xmm0[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm4[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3],xmm5[4],xmm1[4],xmm5[5],xmm1[5],xmm5[6],xmm1[6],xmm5[7],xmm1[7]
+; SSE-NEXT:    pxor %xmm3, %xmm3
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm0[4],xmm5[5],xmm0[5],xmm5[6],xmm0[6],xmm5[7],xmm0[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm5[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm1[0,1,2,3,4,4,7,6]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,4,4,7,6]
 ; SSE-NEXT:    psrlq $48, %xmm0
-; SSE-NEXT:    packuswb %xmm0, %xmm3
+; SSE-NEXT:    packuswb %xmm0, %xmm2
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,0,65535,65535,65535,0,65535,65535]
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm6, %xmm7
-; SSE-NEXT:    pandn %xmm6, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm9, %xmm4
+; SSE-NEXT:    movdqa %xmm7, %xmm6
+; SSE-NEXT:    pandn %xmm7, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm14, %xmm4
 ; SSE-NEXT:    pand %xmm1, %xmm4
 ; SSE-NEXT:    por %xmm0, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,0,65535,0,65535,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm1, %xmm6
-; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm2[8],xmm4[9],xmm2[9],xmm4[10],xmm2[10],xmm4[11],xmm2[11],xmm4[12],xmm2[12],xmm4[13],xmm2[13],xmm4[14],xmm2[14],xmm4[15],xmm2[15]
+; SSE-NEXT:    movdqa %xmm1, %xmm5
+; SSE-NEXT:    pandn %xmm0, %xmm5
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm3[8],xmm4[9],xmm3[9],xmm4[10],xmm3[10],xmm4[11],xmm3[11],xmm4[12],xmm3[12],xmm4[13],xmm3[13],xmm4[14],xmm3[14],xmm4[15],xmm3[15]
 ; SSE-NEXT:    pand %xmm1, %xmm4
-; SSE-NEXT:    por %xmm6, %xmm4
+; SSE-NEXT:    por %xmm5, %xmm4
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm4[3,2,1,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    packuswb %xmm4, %xmm4
-; SSE-NEXT:    pand %xmm13, %xmm4
-; SSE-NEXT:    pandn %xmm3, %xmm13
-; SSE-NEXT:    por %xmm13, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    packuswb %xmm0, %xmm0
+; SSE-NEXT:    movdqa %xmm9, %xmm4
+; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    pandn %xmm2, %xmm4
+; SSE-NEXT:    por %xmm0, %xmm4
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm6, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm15
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm2[0],xmm15[1],xmm2[1],xmm15[2],xmm2[2],xmm15[3],xmm2[3],xmm15[4],xmm2[4],xmm15[5],xmm2[5],xmm15[6],xmm2[6],xmm15[7],xmm2[7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,0,65535,65535,65535,65535,0,65535]
-; SSE-NEXT:    pand %xmm3, %xmm15
+; SSE-NEXT:    pandn %xmm5, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm12
+; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm3[8],xmm0[9],xmm3[9],xmm0[10],xmm3[10],xmm0[11],xmm3[11],xmm0[12],xmm3[12],xmm0[13],xmm3[13],xmm0[14],xmm3[14],xmm0[15],xmm3[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm3[0],xmm12[1],xmm3[1],xmm12[2],xmm3[2],xmm12[3],xmm3[3],xmm12[4],xmm3[4],xmm12[5],xmm3[5],xmm12[6],xmm3[6],xmm12[7],xmm3[7]
+; SSE-NEXT:    movdqa %xmm15, %xmm3
+; SSE-NEXT:    pand %xmm15, %xmm12
 ; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    por %xmm15, %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm11
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm13[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    por %xmm12, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm9[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm2, %xmm3
-; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm11[0,2,1,0,4,5,6,7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,65535,65535,65535,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm3[0,2,1,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm2, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm5, %xmm15
-; SSE-NEXT:    pandn %xmm3, %xmm15
-; SSE-NEXT:    pand %xmm5, %xmm4
+; SSE-NEXT:    pand %xmm7, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm8, %xmm15
+; SSE-NEXT:    pandn %xmm2, %xmm15
+; SSE-NEXT:    pand %xmm8, %xmm4
 ; SSE-NEXT:    por %xmm4, %xmm15
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    pandn %xmm7, %xmm3
-; SSE-NEXT:    movdqa %xmm9, %xmm4
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm6, %xmm2
+; SSE-NEXT:    movdqa %xmm14, %xmm4
 ; SSE-NEXT:    pand %xmm0, %xmm4
-; SSE-NEXT:    por %xmm3, %xmm4
-; SSE-NEXT:    movdqa %xmm4, %xmm3
-; SSE-NEXT:    pxor %xmm0, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm0[8],xmm3[9],xmm0[9],xmm3[10],xmm0[10],xmm3[11],xmm0[11],xmm3[12],xmm0[12],xmm3[13],xmm0[13],xmm3[14],xmm0[14],xmm3[15],xmm0[15]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3],xmm4[4],xmm0[4],xmm4[5],xmm0[5],xmm4[6],xmm0[6],xmm4[7],xmm0[7]
+; SSE-NEXT:    por %xmm2, %xmm4
+; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    pxor %xmm3, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm3[8],xmm2[9],xmm3[9],xmm2[10],xmm3[10],xmm2[11],xmm3[11],xmm2[12],xmm3[12],xmm2[13],xmm3[13],xmm2[14],xmm3[14],xmm2[15],xmm3[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3],xmm4[4],xmm3[4],xmm4[5],xmm3[5],xmm4[6],xmm3[6],xmm4[7],xmm3[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm0[8],xmm2[9],xmm0[9],xmm2[10],xmm0[10],xmm2[11],xmm0[11],xmm2[12],xmm0[12],xmm2[13],xmm0[13],xmm2[14],xmm0[14],xmm2[15],xmm0[15]
-; SSE-NEXT:    movdqa (%rsp), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm3[8],xmm2[9],xmm3[9],xmm2[10],xmm3[10],xmm2[11],xmm3[11],xmm2[12],xmm3[12],xmm2[13],xmm3[13],xmm2[14],xmm3[14],xmm2[15],xmm3[15]
+; SSE-NEXT:    pxor %xmm12, %xmm12
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    pandn %xmm2, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm10
 ; SSE-NEXT:    packuswb %xmm2, %xmm10
 ; SSE-NEXT:    packuswb %xmm4, %xmm4
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm10[0,1,3,3]
 ; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm4[0],xmm2[1,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm7, %xmm3
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,0,65535,65535,65535,0,65535,65535]
-; SSE-NEXT:    pand %xmm4, %xmm3
-; SSE-NEXT:    pandn %xmm6, %xmm4
-; SSE-NEXT:    movdqa %xmm6, %xmm11
-; SSE-NEXT:    por %xmm3, %xmm4
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm4, %xmm3
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm0[8],xmm4[9],xmm0[9],xmm4[10],xmm0[10],xmm4[11],xmm0[11],xmm4[12],xmm0[12],xmm4[13],xmm0[13],xmm4[14],xmm0[14],xmm4[15],xmm0[15]
-; SSE-NEXT:    pxor %xmm10, %xmm10
-; SSE-NEXT:    pand %xmm1, %xmm4
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,0,65535,65535,65535,0,65535,65535]
+; SSE-NEXT:    pand %xmm7, %xmm3
+; SSE-NEXT:    pandn %xmm5, %xmm7
+; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    por %xmm3, %xmm7
+; SSE-NEXT:    movdqa %xmm7, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm12[0],xmm3[1],xmm12[1],xmm3[2],xmm12[2],xmm3[3],xmm12[3],xmm3[4],xmm12[4],xmm3[5],xmm12[5],xmm3[6],xmm12[6],xmm3[7],xmm12[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm12[8],xmm7[9],xmm12[9],xmm7[10],xmm12[10],xmm7[11],xmm12[11],xmm7[12],xmm12[12],xmm7[13],xmm12[13],xmm7[14],xmm12[14],xmm7[15],xmm12[15]
+; SSE-NEXT:    pand %xmm1, %xmm7
 ; SSE-NEXT:    pandn %xmm3, %xmm1
-; SSE-NEXT:    por %xmm4, %xmm1
+; SSE-NEXT:    por %xmm7, %xmm1
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[0,1,2,1]
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,65535,65535,65535,65535,65535,0]
+; SSE-NEXT:    pand %xmm5, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    packuswb %xmm3, %xmm3
-; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm5, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    andps %xmm5, %xmm2
+; SSE-NEXT:    pandn %xmm3, %xmm5
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    pandn %xmm5, %xmm1
+; SSE-NEXT:    andps %xmm8, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [65535,65535,65535,0,65535,65535,0,65535]
-; SSE-NEXT:    pand %xmm13, %xmm2
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm2, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm10[8],xmm3[9],xmm10[9],xmm3[10],xmm10[10],xmm3[11],xmm10[11],xmm3[12],xmm10[12],xmm3[13],xmm10[13],xmm3[14],xmm10[14],xmm3[15],xmm10[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1],xmm2[2],xmm10[2],xmm2[3],xmm10[3],xmm2[4],xmm10[4],xmm2[5],xmm10[5],xmm2[6],xmm10[6],xmm2[7],xmm10[7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [0,65535,65535,65535,65535,65535,65535,0]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,65535,0,65535,65535,0,65535]
+; SSE-NEXT:    pand %xmm10, %xmm5
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm5, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm12[8],xmm2[9],xmm12[9],xmm2[10],xmm12[10],xmm2[11],xmm12[11],xmm2[12],xmm12[12],xmm2[13],xmm12[13],xmm2[14],xmm12[14],xmm2[15],xmm12[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm12[0],xmm5[1],xmm12[1],xmm5[2],xmm12[2],xmm5[3],xmm12[3],xmm5[4],xmm12[4],xmm5[5],xmm12[5],xmm5[6],xmm12[6],xmm5[7],xmm12[7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [0,65535,65535,65535,65535,65535,65535,0]
+; SSE-NEXT:    pand %xmm3, %xmm5
+; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    por %xmm5, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm3[0,1,2,3,5,4,7,6]
+; SSE-NEXT:    psrldq {{.*#+}} xmm2 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    packuswb %xmm2, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm6[1,3,2,3]
+; SSE-NEXT:    movdqa %xmm6, %xmm9
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm14[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; SSE-NEXT:    movdqa %xmm3, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm12[8],xmm2[9],xmm12[9],xmm2[10],xmm12[10],xmm2[11],xmm12[11],xmm2[12],xmm12[12],xmm2[13],xmm12[13],xmm2[14],xmm12[14],xmm2[15],xmm12[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm12[0],xmm3[1],xmm12[1],xmm3[2],xmm12[2],xmm3[3],xmm12[3],xmm3[4],xmm12[4],xmm3[5],xmm12[5],xmm3[6],xmm12[6],xmm3[7],xmm12[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE-NEXT:    packuswb %xmm3, %xmm3
+; SSE-NEXT:    movss {{.*#+}} xmm7 = xmm3[0],xmm7[1,2,3]
+; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,2,2,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,0,65535,65535,0,65535,65535]
 ; SSE-NEXT:    pand %xmm4, %xmm2
-; SSE-NEXT:    pandn %xmm3, %xmm4
+; SSE-NEXT:    pandn %xmm0, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm2[0,1,2,3,5,4,7,6]
-; SSE-NEXT:    psrldq {{.*#+}} xmm3 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    packuswb %xmm3, %xmm6
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm9[0,2,2,3]
-; SSE-NEXT:    pand %xmm13, %xmm9
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm3, %xmm13
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; SSE-NEXT:    movdqa %xmm2, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm10[8],xmm3[9],xmm10[9],xmm3[10],xmm10[10],xmm3[11],xmm10[11],xmm3[12],xmm10[12],xmm3[13],xmm10[13],xmm3[14],xmm10[14],xmm3[15],xmm10[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1],xmm2[2],xmm10[2],xmm2[3],xmm10[3],xmm2[4],xmm10[4],xmm2[5],xmm10[5],xmm2[6],xmm10[6],xmm2[7],xmm10[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    movss {{.*#+}} xmm6 = xmm2[0],xmm6[1,2,3]
-; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT:    pand %xmm3, %xmm2
-; SSE-NEXT:    pandn %xmm11, %xmm3
-; SSE-NEXT:    por %xmm2, %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1],xmm2[2],xmm10[2],xmm2[3],xmm10[3],xmm2[4],xmm10[4],xmm2[5],xmm10[5],xmm2[6],xmm10[6],xmm2[7],xmm10[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm10[8],xmm3[9],xmm10[9],xmm3[10],xmm10[10],xmm3[11],xmm10[11],xmm3[12],xmm10[12],xmm3[13],xmm10[13],xmm3[14],xmm10[14],xmm3[15],xmm10[15]
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,0,65535,0,65535,65535,65535]
-; SSE-NEXT:    pand %xmm4, %xmm3
-; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    por %xmm3, %xmm4
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm12[0,1,0,3]
+; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm12[0],xmm2[1],xmm12[1],xmm2[2],xmm12[2],xmm2[3],xmm12[3],xmm2[4],xmm12[4],xmm2[5],xmm12[5],xmm2[6],xmm12[6],xmm2[7],xmm12[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm12[8],xmm4[9],xmm12[9],xmm4[10],xmm12[10],xmm4[11],xmm12[11],xmm4[12],xmm12[12],xmm4[13],xmm12[13],xmm4[14],xmm12[14],xmm4[15],xmm12[15]
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,0,65535,0,65535,65535,65535]
+; SSE-NEXT:    pand %xmm5, %xmm4
+; SSE-NEXT:    pandn %xmm2, %xmm5
+; SSE-NEXT:    por %xmm4, %xmm5
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,7,6]
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0]
-; SSE-NEXT:    movdqa %xmm3, %xmm7
-; SSE-NEXT:    pandn %xmm2, %xmm7
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[2,1,0,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0]
+; SSE-NEXT:    movdqa %xmm4, %xmm6
+; SSE-NEXT:    pandn %xmm2, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[2,1,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,0,3,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pand %xmm3, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm7
-; SSE-NEXT:    movdqa %xmm5, %xmm2
-; SSE-NEXT:    pandn %xmm7, %xmm2
-; SSE-NEXT:    andps %xmm5, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm2
-; SSE-NEXT:    movdqa %xmm13, %xmm7
-; SSE-NEXT:    por %xmm9, %xmm7
-; SSE-NEXT:    movdqa %xmm7, %xmm4
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm10[0],xmm4[1],xmm10[1],xmm4[2],xmm10[2],xmm4[3],xmm10[3],xmm4[4],xmm10[4],xmm4[5],xmm10[5],xmm4[6],xmm10[6],xmm4[7],xmm10[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm4[0,1,2,3,6,4,6,7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm10[8],xmm7[9],xmm10[9],xmm7[10],xmm10[10],xmm7[11],xmm10[11],xmm7[12],xmm10[12],xmm7[13],xmm10[13],xmm7[14],xmm10[14],xmm7[15],xmm10[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm7[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,7,6,7]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm4[4],xmm6[5],xmm4[5],xmm6[6],xmm4[6],xmm6[7],xmm4[7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm10[8],xmm4[9],xmm10[9],xmm4[10],xmm10[10],xmm4[11],xmm10[11],xmm4[12],xmm10[12],xmm4[13],xmm10[13],xmm4[14],xmm10[14],xmm4[15],xmm10[15]
+; SSE-NEXT:    pand %xmm4, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm6
+; SSE-NEXT:    movdqa %xmm8, %xmm2
+; SSE-NEXT:    pandn %xmm6, %xmm2
+; SSE-NEXT:    andps %xmm8, %xmm7
+; SSE-NEXT:    por %xmm7, %xmm2
+; SSE-NEXT:    movdqa %xmm10, %xmm7
+; SSE-NEXT:    pand %xmm10, %xmm14
+; SSE-NEXT:    pandn %xmm9, %xmm7
+; SSE-NEXT:    por %xmm14, %xmm7
+; SSE-NEXT:    movdqa %xmm7, %xmm5
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm12[0],xmm5[1],xmm12[1],xmm5[2],xmm12[2],xmm5[3],xmm12[3],xmm5[4],xmm12[4],xmm5[5],xmm12[5],xmm5[6],xmm12[6],xmm5[7],xmm12[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm5[0,1,2,3,6,4,6,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm12[8],xmm7[9],xmm12[9],xmm7[10],xmm12[10],xmm7[11],xmm12[11],xmm7[12],xmm12[12],xmm7[13],xmm12[13],xmm7[14],xmm12[14],xmm7[15],xmm12[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm7[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,7,6,7]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm5[4],xmm6[5],xmm5[5],xmm6[6],xmm5[6],xmm6[7],xmm5[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm12[8],xmm5[9],xmm12[9],xmm5[10],xmm12[10],xmm5[11],xmm12[11],xmm5[12],xmm12[12],xmm5[13],xmm12[13],xmm5[14],xmm12[14],xmm5[15],xmm12[15]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm4, %xmm9
-; SSE-NEXT:    movdqa %xmm4, %xmm7
-; SSE-NEXT:    por %xmm9, %xmm14
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm14[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,5]
+; SSE-NEXT:    pandn %xmm5, %xmm9
+; SSE-NEXT:    movdqa %xmm5, %xmm7
+; SSE-NEXT:    por %xmm9, %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm11[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,7,6,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[3,3,3,3]
-; SSE-NEXT:    packuswb %xmm7, %xmm4
+; SSE-NEXT:    packuswb %xmm7, %xmm5
 ; SSE-NEXT:    packuswb %xmm6, %xmm6
-; SSE-NEXT:    movss {{.*#+}} xmm4 = xmm6[0],xmm4[1,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm11[1,3,2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm0[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
-; SSE-NEXT:    movdqa %xmm7, %xmm6
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm10[0],xmm6[1],xmm10[1],xmm6[2],xmm10[2],xmm6[3],xmm10[3],xmm6[4],xmm10[4],xmm6[5],xmm10[5],xmm6[6],xmm10[6],xmm6[7],xmm10[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm10[8],xmm7[9],xmm10[9],xmm7[10],xmm10[10],xmm7[11],xmm10[11],xmm7[12],xmm10[12],xmm7[13],xmm10[13],xmm7[14],xmm10[14],xmm7[15],xmm10[15]
-; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,65535,0,65535,0,65535,65535]
-; SSE-NEXT:    pand %xmm9, %xmm7
-; SSE-NEXT:    pandn %xmm6, %xmm9
-; SSE-NEXT:    por %xmm7, %xmm9
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm9[2,1,1,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm6, %xmm6
-; SSE-NEXT:    pand %xmm3, %xmm6
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm12[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    packuswb %xmm7, %xmm7
-; SSE-NEXT:    pandn %xmm7, %xmm3
-; SSE-NEXT:    por %xmm3, %xmm6
-; SSE-NEXT:    andps %xmm5, %xmm4
-; SSE-NEXT:    pandn %xmm6, %xmm5
-; SSE-NEXT:    por %xmm4, %xmm5
+; SSE-NEXT:    movss {{.*#+}} xmm5 = xmm6[0],xmm5[1,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm0[1,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1]
+; SSE-NEXT:    movdqa %xmm3, %xmm6
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1],xmm6[2],xmm12[2],xmm6[3],xmm12[3],xmm6[4],xmm12[4],xmm6[5],xmm12[5],xmm6[6],xmm12[6],xmm6[7],xmm12[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm12[8],xmm3[9],xmm12[9],xmm3[10],xmm12[10],xmm3[11],xmm12[11],xmm3[12],xmm12[12],xmm3[13],xmm12[13],xmm3[14],xmm12[14],xmm3[15],xmm12[15]
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,65535,0,65535,0,65535,65535]
+; SSE-NEXT:    pand %xmm7, %xmm3
+; SSE-NEXT:    pandn %xmm6, %xmm7
+; SSE-NEXT:    por %xmm3, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    packuswb %xmm3, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm7[2,1,1,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm6, %xmm6
+; SSE-NEXT:    pand %xmm4, %xmm6
+; SSE-NEXT:    pandn %xmm3, %xmm4
+; SSE-NEXT:    por %xmm6, %xmm4
+; SSE-NEXT:    andps %xmm8, %xmm5
+; SSE-NEXT:    pandn %xmm4, %xmm8
+; SSE-NEXT:    por %xmm5, %xmm8
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm3, (%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, (%rdx)
-; SSE-NEXT:    movdqa %xmm8, (%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm3, (%rdx)
+; SSE-NEXT:    movdqa %xmm13, (%rcx)
 ; SSE-NEXT:    movdqa %xmm15, (%r8)
 ; SSE-NEXT:    movdqa %xmm1, (%r9)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; SSE-NEXT:    movdqa %xmm2, (%rax)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movdqa %xmm5, (%rax)
-; SSE-NEXT:    addq $168, %rsp
+; SSE-NEXT:    movdqa %xmm8, (%rax)
+; SSE-NEXT:    addq $200, %rsp
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i8_stride7_vf16:
@@ -3601,13 +3600,14 @@ define void @load_i8_stride7_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i8_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6) nounwind {
 ; SSE-LABEL: load_i8_stride7_vf32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $648, %rsp # imm = 0x288
-; SSE-NEXT:    movdqa 208(%rdi), %xmm14
+; SSE-NEXT:    subq $632, %rsp # imm = 0x278
+; SSE-NEXT:    movdqa 208(%rdi), %xmm11
 ; SSE-NEXT:    movdqa 192(%rdi), %xmm5
 ; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 176(%rdi), %xmm6
 ; SSE-NEXT:    movdqa 112(%rdi), %xmm4
 ; SSE-NEXT:    movdqa 128(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 160(%rdi), %xmm7
 ; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 144(%rdi), %xmm1
@@ -3619,69 +3619,67 @@ define void @load_i8_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pand %xmm2, %xmm1
 ; SSE-NEXT:    movdqa %xmm2, %xmm9
 ; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pxor %xmm10, %xmm10
+; SSE-NEXT:    pxor %xmm15, %xmm15
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm10[8],xmm0[9],xmm10[9],xmm0[10],xmm10[10],xmm0[11],xmm10[11],xmm0[12],xmm10[12],xmm0[13],xmm10[13],xmm0[14],xmm10[14],xmm0[15],xmm10[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1],xmm1[2],xmm10[2],xmm1[3],xmm10[3],xmm1[4],xmm10[4],xmm1[5],xmm10[5],xmm1[6],xmm10[6],xmm1[7],xmm10[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm15[8],xmm0[9],xmm15[9],xmm0[10],xmm15[10],xmm0[11],xmm15[11],xmm0[12],xmm15[12],xmm0[13],xmm15[13],xmm0[14],xmm15[14],xmm0[15],xmm15[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1],xmm1[2],xmm15[2],xmm1[3],xmm15[3],xmm1[4],xmm15[4],xmm1[5],xmm15[5],xmm1[6],xmm15[6],xmm1[7],xmm15[7]
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,4,4,5,6]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,5,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm2
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255]
+; SSE-NEXT:    packuswb %xmm0, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm8, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
 ; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,0,65535,65535,65535,0,65535]
 ; SSE-NEXT:    movdqa %xmm7, %xmm1
 ; SSE-NEXT:    pandn %xmm3, %xmm1
-; SSE-NEXT:    movdqa %xmm3, %xmm11
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm4, %xmm3
 ; SSE-NEXT:    movdqa %xmm4, %xmm12
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm7, %xmm3
-; SSE-NEXT:    movdqa %xmm7, %xmm8
+; SSE-NEXT:    movdqa %xmm7, %xmm14
 ; SSE-NEXT:    por %xmm1, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm10[8],xmm1[9],xmm10[9],xmm1[10],xmm10[10],xmm1[11],xmm10[11],xmm1[12],xmm10[12],xmm1[13],xmm10[13],xmm1[14],xmm10[14],xmm1[15],xmm10[15]
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,65535,65535,0,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm7, %xmm4
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm15[8],xmm1[9],xmm15[9],xmm1[10],xmm15[10],xmm1[11],xmm15[11],xmm1[12],xmm15[12],xmm1[13],xmm15[13],xmm1[14],xmm15[14],xmm1[15],xmm15[15]
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,0,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm0, %xmm4
 ; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm10[0],xmm3[1],xmm10[1],xmm3[2],xmm10[2],xmm3[3],xmm10[3],xmm3[4],xmm10[4],xmm3[5],xmm10[5],xmm3[6],xmm10[6],xmm3[7],xmm10[7]
-; SSE-NEXT:    pand %xmm7, %xmm3
-; SSE-NEXT:    movdqa %xmm7, %xmm15
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm15[0],xmm3[1],xmm15[1],xmm3[2],xmm15[2],xmm3[3],xmm15[3],xmm3[4],xmm15[4],xmm3[5],xmm15[5],xmm3[6],xmm15[6],xmm3[7],xmm15[7]
+; SSE-NEXT:    pand %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm0, %xmm7
 ; SSE-NEXT:    por %xmm4, %xmm3
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,1,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    por %xmm3, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,65535,0,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    pand %xmm8, %xmm1
+; SSE-NEXT:    por %xmm2, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,0,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm0, %xmm2
 ; SSE-NEXT:    pandn %xmm6, %xmm2
 ; SSE-NEXT:    movdqa %xmm6, %xmm13
 ; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm5, %xmm3
-; SSE-NEXT:    pand %xmm7, %xmm3
+; SSE-NEXT:    pand %xmm0, %xmm3
 ; SSE-NEXT:    por %xmm2, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1],xmm2[2],xmm10[2],xmm2[3],xmm10[3],xmm2[4],xmm10[4],xmm2[5],xmm10[5],xmm2[6],xmm10[6],xmm2[7],xmm10[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm15[0],xmm2[1],xmm15[1],xmm2[2],xmm15[2],xmm2[3],xmm15[3],xmm2[4],xmm15[4],xmm2[5],xmm15[5],xmm2[6],xmm15[6],xmm2[7],xmm15[7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,6,4,7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm10[8],xmm3[9],xmm10[9],xmm3[10],xmm10[10],xmm3[11],xmm10[11],xmm3[12],xmm10[12],xmm3[13],xmm10[13],xmm3[14],xmm10[14],xmm3[15],xmm10[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm15[8],xmm3[9],xmm15[9],xmm3[10],xmm15[10],xmm3[11],xmm15[11],xmm3[12],xmm15[12],xmm3[13],xmm15[13],xmm3[14],xmm15[14],xmm3[15],xmm15[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,7,7]
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
-; SSE-NEXT:    movdqa %xmm14, %xmm3
-; SSE-NEXT:    movdqa %xmm14, %xmm4
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm10[8],xmm4[9],xmm10[9],xmm4[10],xmm10[10],xmm4[11],xmm10[11],xmm4[12],xmm10[12],xmm4[13],xmm10[13],xmm4[14],xmm10[14],xmm4[15],xmm10[15]
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm10[0],xmm3[1],xmm10[1],xmm3[2],xmm10[2],xmm3[3],xmm10[3],xmm3[4],xmm10[4],xmm3[5],xmm10[5],xmm3[6],xmm10[6],xmm3[7],xmm10[7]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; SSE-NEXT:    movdqa %xmm11, %xmm10
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm15[8],xmm10[9],xmm15[9],xmm10[10],xmm15[10],xmm10[11],xmm15[11],xmm10[12],xmm15[12],xmm10[13],xmm15[13],xmm10[14],xmm15[14],xmm10[15],xmm15[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm15[0],xmm11[1],xmm15[1],xmm11[2],xmm15[2],xmm11[3],xmm15[3],xmm11[4],xmm15[4],xmm11[5],xmm15[5],xmm11[6],xmm15[6],xmm11[7],xmm15[7]
+; SSE-NEXT:    movdqa %xmm11, %xmm3
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm10[0],xmm3[1],xmm10[1],xmm3[2],xmm10[2],xmm3[3],xmm10[3]
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    packuswb %xmm3, %xmm3
@@ -3698,17 +3696,17 @@ define void @load_i8_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pand %xmm5, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 32(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 32(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm9, %xmm1
-; SSE-NEXT:    pandn %xmm2, %xmm1
+; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    movdqa 48(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm9, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm10[8],xmm1[9],xmm10[9],xmm1[10],xmm10[10],xmm1[11],xmm10[11],xmm1[12],xmm10[12],xmm1[13],xmm10[13],xmm1[14],xmm10[14],xmm1[15],xmm10[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1],xmm2[2],xmm10[2],xmm2[3],xmm10[3],xmm2[4],xmm10[4],xmm2[5],xmm10[5],xmm2[6],xmm10[6],xmm2[7],xmm10[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm15[8],xmm1[9],xmm15[9],xmm1[10],xmm15[10],xmm1[11],xmm15[11],xmm1[12],xmm15[12],xmm1[13],xmm15[13],xmm1[14],xmm15[14],xmm1[15],xmm15[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm15[0],xmm2[1],xmm15[1],xmm2[2],xmm15[2],xmm2[3],xmm15[3],xmm2[4],xmm15[4],xmm2[5],xmm15[5],xmm2[6],xmm15[6],xmm2[7],xmm15[7]
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
@@ -3717,65 +3715,68 @@ define void @load_i8_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm2
 ; SSE-NEXT:    movdqa 16(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    movdqa %xmm14, %xmm0
+; SSE-NEXT:    movdqa %xmm14, %xmm1
 ; SSE-NEXT:    pandn %xmm3, %xmm1
-; SSE-NEXT:    movdqa %xmm3, %xmm14
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm3, %xmm9
 ; SSE-NEXT:    movdqa (%rdi), %xmm4
 ; SSE-NEXT:    movdqa %xmm4, %xmm3
-; SSE-NEXT:    movdqa %xmm4, %xmm9
+; SSE-NEXT:    movdqa %xmm4, %xmm14
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm8, %xmm3
+; SSE-NEXT:    pand %xmm0, %xmm3
 ; SSE-NEXT:    por %xmm1, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm10[8],xmm1[9],xmm10[9],xmm1[10],xmm10[10],xmm1[11],xmm10[11],xmm1[12],xmm10[12],xmm1[13],xmm10[13],xmm1[14],xmm10[14],xmm1[15],xmm10[15]
-; SSE-NEXT:    movdqa %xmm15, %xmm4
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm15[8],xmm1[9],xmm15[9],xmm1[10],xmm15[10],xmm1[11],xmm15[11],xmm1[12],xmm15[12],xmm1[13],xmm15[13],xmm1[14],xmm15[14],xmm1[15],xmm15[15]
+; SSE-NEXT:    movdqa %xmm7, %xmm4
 ; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm10[0],xmm3[1],xmm10[1],xmm3[2],xmm10[2],xmm3[3],xmm10[3],xmm3[4],xmm10[4],xmm3[5],xmm10[5],xmm3[6],xmm10[6],xmm3[7],xmm10[7]
-; SSE-NEXT:    pand %xmm15, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm15[0],xmm3[1],xmm15[1],xmm3[2],xmm15[2],xmm3[3],xmm15[3],xmm3[4],xmm15[4],xmm3[5],xmm15[5],xmm3[6],xmm15[6],xmm3[7],xmm15[7]
+; SSE-NEXT:    pand %xmm7, %xmm3
 ; SSE-NEXT:    por %xmm4, %xmm3
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,1,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm1
+; SSE-NEXT:    pand %xmm8, %xmm1
+; SSE-NEXT:    pandn %xmm2, %xmm8
+; SSE-NEXT:    por %xmm1, %xmm8
 ; SSE-NEXT:    movdqa 64(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    pandn %xmm2, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,0,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm4, %xmm1
+; SSE-NEXT:    pandn %xmm2, %xmm1
 ; SSE-NEXT:    movdqa %xmm2, %xmm15
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 80(%rdi), %xmm8
-; SSE-NEXT:    movdqa %xmm8, %xmm2
-; SSE-NEXT:    pand %xmm7, %xmm2
-; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1],xmm0[2],xmm10[2],xmm0[3],xmm10[3],xmm0[4],xmm10[4],xmm0[5],xmm10[5],xmm0[6],xmm10[6],xmm0[7],xmm10[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,4,7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm10[8],xmm2[9],xmm10[9],xmm2[10],xmm10[10],xmm2[11],xmm10[11],xmm2[12],xmm10[12],xmm2[13],xmm10[13],xmm2[14],xmm10[14],xmm2[15],xmm10[15]
+; SSE-NEXT:    movdqa 80(%rdi), %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm4, %xmm2
+; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    pxor %xmm3, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,4,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm3[8],xmm2[9],xmm3[9],xmm2[10],xmm3[10],xmm2[11],xmm3[11],xmm2[12],xmm3[12],xmm2[13],xmm3[13],xmm2[14],xmm3[14],xmm2[15],xmm3[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,7,7]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
 ; SSE-NEXT:    movdqa 96(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm10[8],xmm3[9],xmm10[9],xmm3[10],xmm10[10],xmm3[11],xmm10[11],xmm3[12],xmm10[12],xmm3[13],xmm10[13],xmm3[14],xmm10[14],xmm3[15],xmm10[15]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1],xmm2[2],xmm10[2],xmm2[3],xmm10[3],xmm2[4],xmm10[4],xmm2[5],xmm10[5],xmm2[6],xmm10[6],xmm2[7],xmm10[7]
+; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm3[8],xmm0[9],xmm3[9],xmm0[10],xmm3[10],xmm0[11],xmm3[11],xmm0[12],xmm3[12],xmm0[13],xmm3[13],xmm0[14],xmm3[14],xmm0[15],xmm3[15]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3],xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; SSE-NEXT:    pxor %xmm7, %xmm7
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
 ; SSE-NEXT:    movdqa %xmm6, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm6, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm3
-; SSE-NEXT:    pand %xmm5, %xmm1
+; SSE-NEXT:    packuswb %xmm1, %xmm1
+; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    pand %xmm5, %xmm8
 ; SSE-NEXT:    pandn %xmm3, %xmm5
-; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    por %xmm8, %xmm5
 ; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,0,65535,65535,0,65535,65535]
 ; SSE-NEXT:    movdqa %xmm2, %xmm0
@@ -3784,40 +3785,39 @@ define void @load_i8_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pand %xmm2, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm10[8],xmm2[9],xmm10[9],xmm2[10],xmm10[10],xmm2[11],xmm10[11],xmm2[12],xmm10[12],xmm2[13],xmm10[13],xmm2[14],xmm10[14],xmm2[15],xmm10[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm7[8],xmm2[9],xmm7[9],xmm2[10],xmm7[10],xmm2[11],xmm7[11],xmm2[12],xmm7[12],xmm2[13],xmm7[13],xmm2[14],xmm7[14],xmm2[15],xmm7[15]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,0,65535,65535,65535,65535]
 ; SSE-NEXT:    movdqa %xmm0, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1],xmm1[2],xmm10[2],xmm1[3],xmm10[3],xmm1[4],xmm10[4],xmm1[5],xmm10[5],xmm1[6],xmm10[6],xmm1[7],xmm10[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3],xmm1[4],xmm7[4],xmm1[5],xmm7[5],xmm1[6],xmm7[6],xmm1[7],xmm7[7]
 ; SSE-NEXT:    pand %xmm0, %xmm1
 ; SSE-NEXT:    por %xmm3, %xmm1
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,6]
 ; SSE-NEXT:    psrld $16, %xmm2
 ; SSE-NEXT:    packuswb %xmm2, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,255,255,255,255,0,0,0,0,255,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm5, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,255,255,255,0,0,0,0,255,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm6, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    pandn %xmm11, %xmm1
+; SSE-NEXT:    movdqa %xmm4, %xmm1
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm12, %xmm3
-; SSE-NEXT:    pand %xmm7, %xmm3
-; SSE-NEXT:    movdqa %xmm7, %xmm12
+; SSE-NEXT:    pand %xmm4, %xmm3
+; SSE-NEXT:    movdqa %xmm4, %xmm12
 ; SSE-NEXT:    por %xmm1, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1],xmm1[2],xmm10[2],xmm1[3],xmm10[3],xmm1[4],xmm10[4],xmm1[5],xmm10[5],xmm1[6],xmm10[6],xmm1[7],xmm10[7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,0,65535,65535,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm7, %xmm4
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3],xmm1[4],xmm7[4],xmm1[5],xmm7[5],xmm1[6],xmm7[6],xmm1[7],xmm7[7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,0,65535,65535,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm8, %xmm4
 ; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm10[8],xmm3[9],xmm10[9],xmm3[10],xmm10[10],xmm3[11],xmm10[11],xmm3[12],xmm10[12],xmm3[13],xmm10[13],xmm3[14],xmm10[14],xmm3[15],xmm10[15]
-; SSE-NEXT:    pand %xmm7, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm7[8],xmm3[9],xmm7[9],xmm3[10],xmm7[10],xmm3[11],xmm7[11],xmm3[12],xmm7[12],xmm3[13],xmm7[13],xmm3[14],xmm7[14],xmm3[15],xmm7[15]
+; SSE-NEXT:    pand %xmm8, %xmm3
 ; SSE-NEXT:    por %xmm4, %xmm3
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm5, %xmm1
-; SSE-NEXT:    movdqa %xmm5, %xmm7
+; SSE-NEXT:    pand %xmm6, %xmm1
 ; SSE-NEXT:    por %xmm2, %xmm1
 ; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,0,65535,65535,65535,0,65535]
 ; SSE-NEXT:    movdqa %xmm4, %xmm2
@@ -3827,25 +3827,24 @@ define void @load_i8_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movdqa %xmm4, %xmm13
 ; SSE-NEXT:    por %xmm2, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm10[8],xmm2[9],xmm10[9],xmm2[10],xmm10[10],xmm2[11],xmm10[11],xmm2[12],xmm10[12],xmm2[13],xmm10[13],xmm2[14],xmm10[14],xmm2[15],xmm10[15]
-; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,65535,65535,65535,0,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm11, %xmm4
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm7[8],xmm2[9],xmm7[9],xmm2[10],xmm7[10],xmm2[11],xmm7[11],xmm2[12],xmm7[12],xmm2[13],xmm7[13],xmm2[14],xmm7[14],xmm2[15],xmm7[15]
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,65535,65535,0,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm5, %xmm4
 ; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm10[0],xmm3[1],xmm10[1],xmm3[2],xmm10[2],xmm3[3],xmm10[3],xmm3[4],xmm10[4],xmm3[5],xmm10[5],xmm3[6],xmm10[6],xmm3[7],xmm10[7]
-; SSE-NEXT:    pand %xmm11, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm7[0],xmm3[1],xmm7[1],xmm3[2],xmm7[2],xmm3[3],xmm7[3],xmm3[4],xmm7[4],xmm3[5],xmm7[5],xmm3[6],xmm7[6],xmm3[7],xmm7[7]
+; SSE-NEXT:    pand %xmm5, %xmm3
 ; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pslld $16, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
-; SSE-NEXT:    packuswb %xmm4, %xmm2
-; SSE-NEXT:    movdqa %xmm6, %xmm4
+; SSE-NEXT:    pslld $16, %xmm10
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm10[0],xmm11[1],xmm10[1],xmm11[2],xmm10[2],xmm11[3],xmm10[3]
+; SSE-NEXT:    packuswb %xmm11, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,65535,65535,65535,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm7, %xmm4
 ; SSE-NEXT:    pandn %xmm2, %xmm4
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,0,3,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,4,6,7]
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pand %xmm6, %xmm2
+; SSE-NEXT:    pand %xmm7, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0]
 ; SSE-NEXT:    movdqa %xmm3, %xmm2
@@ -3855,8 +3854,7 @@ define void @load_i8_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,0,65535,65535,0,65535,65535]
 ; SSE-NEXT:    movdqa %xmm5, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm6, %xmm1
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm10, %xmm2
 ; SSE-NEXT:    pand %xmm5, %xmm2
@@ -3865,7 +3863,7 @@ define void @load_i8_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pxor %xmm3, %xmm3
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm3[8],xmm1[9],xmm3[9],xmm1[10],xmm3[10],xmm1[11],xmm3[11],xmm1[12],xmm3[12],xmm1[13],xmm3[13],xmm1[14],xmm3[14],xmm1[15],xmm3[15]
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3],xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
-; SSE-NEXT:    pxor %xmm5, %xmm5
+; SSE-NEXT:    pxor %xmm4, %xmm4
 ; SSE-NEXT:    pand %xmm0, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm0
 ; SSE-NEXT:    por %xmm2, %xmm0
@@ -3873,788 +3871,804 @@ define void @load_i8_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,7,6]
 ; SSE-NEXT:    psrld $16, %xmm1
 ; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm7, %xmm4
-; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    movdqa %xmm6, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    pandn %xmm14, %xmm0
-; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    pandn %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm9, %xmm5
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm14, %xmm2
 ; SSE-NEXT:    pand %xmm12, %xmm2
 ; SSE-NEXT:    por %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,0,65535,65535,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm7, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSE-NEXT:    movdqa %xmm8, %xmm3
 ; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
-; SSE-NEXT:    pand %xmm7, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm4[8],xmm2[9],xmm4[9],xmm2[10],xmm4[10],xmm2[11],xmm4[11],xmm2[12],xmm4[12],xmm2[13],xmm4[13],xmm2[14],xmm4[14],xmm2[15],xmm4[15]
+; SSE-NEXT:    pand %xmm8, %xmm2
 ; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm4, %xmm0
+; SSE-NEXT:    pand %xmm6, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    pandn %xmm8, %xmm0
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm2, %xmm0
 ; SSE-NEXT:    movdqa %xmm15, %xmm1
 ; SSE-NEXT:    pand %xmm13, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    pxor %xmm2, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm11, %xmm2
-; SSE-NEXT:    pand %xmm11, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pxor %xmm3, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm3[8],xmm0[9],xmm3[9],xmm0[10],xmm3[10],xmm0[11],xmm3[11],xmm0[12],xmm3[12],xmm0[13],xmm3[13],xmm0[14],xmm3[14],xmm0[15],xmm3[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,65535,0,65535,0,65535]
+; SSE-NEXT:    pand %xmm3, %xmm1
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm7, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm11
+; SSE-NEXT:    pand %xmm13, %xmm11
+; SSE-NEXT:    por %xmm0, %xmm11
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm5, %xmm9
-; SSE-NEXT:    pand %xmm13, %xmm9
-; SSE-NEXT:    por %xmm0, %xmm9
-; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm1, %xmm0
 ; SSE-NEXT:    pand %xmm13, %xmm0
+; SSE-NEXT:    movdqa %xmm10, %xmm4
 ; SSE-NEXT:    pandn %xmm10, %xmm13
 ; SSE-NEXT:    por %xmm0, %xmm13
 ; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,0,65535,65535,0,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pandn %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm12, %xmm7
-; SSE-NEXT:    movdqa %xmm12, %xmm5
-; SSE-NEXT:    pandn %xmm1, %xmm5
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm1[0,2,2,3]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm2, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,0,65535,65535,0,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm12, %xmm6
+; SSE-NEXT:    movdqa %xmm12, %xmm3
+; SSE-NEXT:    pandn %xmm7, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm7[0,2,2,3]
+; SSE-NEXT:    movdqa %xmm7, %xmm15
+; SSE-NEXT:    pand %xmm10, %xmm7
+; SSE-NEXT:    por %xmm0, %xmm7
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pand %xmm10, %xmm8
+; SSE-NEXT:    movdqa %xmm5, %xmm12
+; SSE-NEXT:    pand %xmm10, %xmm12
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pand %xmm2, %xmm13
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    pand %xmm2, %xmm12
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    pand %xmm2, %xmm14
-; SSE-NEXT:    pand %xmm2, %xmm8
-; SSE-NEXT:    movdqa %xmm8, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    pandn %xmm10, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[0,2,2,3]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm2, %xmm10
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pandn %xmm6, %xmm2
-; SSE-NEXT:    por %xmm10, %xmm2
+; SSE-NEXT:    pand %xmm10, %xmm13
+; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm6, %xmm2
+; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    pandn %xmm4, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm5
-; SSE-NEXT:    movdqa %xmm2, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm6, %xmm8
-; SSE-NEXT:    pslld $16, %xmm8
-; SSE-NEXT:    psrldq {{.*#+}} xmm5 = xmm5[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,2,2,3]
+; SSE-NEXT:    movdqa %xmm2, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm1, %xmm10
+; SSE-NEXT:    por %xmm0, %xmm10
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,0,65535,65535,65535,65535,65535,65535]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm15
-; SSE-NEXT:    psrldq {{.*#+}} xmm15 = xmm15[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm6[0],xmm15[1],xmm6[1],xmm15[2],xmm6[2],xmm15[3],xmm6[3]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm0, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
+; SSE-NEXT:    movdqa %xmm0, %xmm4
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm7, %xmm6
+; SSE-NEXT:    pslld $16, %xmm6
+; SSE-NEXT:    psrldq {{.*#+}} xmm3 = xmm3[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm2, %xmm14
+; SSE-NEXT:    psrldq {{.*#+}} xmm14 = xmm14[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm7[0],xmm14[1],xmm7[1],xmm14[2],xmm7[2],xmm14[3],xmm7[3]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm0[4],xmm4[5],xmm0[5],xmm4[6],xmm0[6],xmm4[7],xmm0[7]
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm0, %xmm4
+; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pxor %xmm10, %xmm10
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1],xmm9[2],xmm10[2],xmm9[3],xmm10[3],xmm9[4],xmm10[4],xmm9[5],xmm10[5],xmm9[6],xmm10[6],xmm9[7],xmm10[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm9[0,1,2,3,6,4,6,5]
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,65535,65535,65535,0,65535]
-; SSE-NEXT:    pand %xmm4, %xmm0
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pxor %xmm5, %xmm5
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm5[0],xmm11[1],xmm5[1],xmm11[2],xmm5[2],xmm11[3],xmm5[3],xmm11[4],xmm5[4],xmm11[5],xmm5[5],xmm11[6],xmm5[6],xmm11[7],xmm5[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm11[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm10[0,1,2,3,6,4,6,5]
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,65535,65535,65535,65535,0,65535]
+; SSE-NEXT:    pand %xmm10, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1],xmm0[2],xmm10[2],xmm0[3],xmm10[3],xmm0[4],xmm10[4],xmm0[5],xmm10[5],xmm0[6],xmm10[6],xmm0[7],xmm10[7]
-; SSE-NEXT:    pxor %xmm9, %xmm9
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm0[0,1,2,3,7,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,6,4,6,5]
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    pand %xmm4, %xmm10
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm1, %xmm4
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm0[0,1,2,3,7,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm11[0,1,2,3,6,4,6,5]
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    pand %xmm10, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm10, %xmm5
+; SSE-NEXT:    movdqa %xmm4, %xmm10
+; SSE-NEXT:    pandn %xmm4, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm0, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm2, %xmm4
 ; SSE-NEXT:    pand %xmm0, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, %xmm4
-; SSE-NEXT:    pandn %xmm6, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
-; SSE-NEXT:    pand %xmm7, %xmm0
+; SSE-NEXT:    movdqa %xmm7, %xmm5
+; SSE-NEXT:    pandn %xmm7, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm7, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm9[0],xmm6[1],xmm9[1],xmm6[2],xmm9[2],xmm6[3],xmm9[3],xmm6[4],xmm9[4],xmm6[5],xmm9[5],xmm6[6],xmm9[6],xmm6[7],xmm9[7]
-; SSE-NEXT:    pand %xmm7, %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pxor %xmm0, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3],xmm4[4],xmm0[4],xmm4[5],xmm0[5],xmm4[6],xmm0[6],xmm4[7],xmm0[7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,0,65535,65535,65535,65535,65535,65535]
 ; SSE-NEXT:    pand %xmm7, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm7, %xmm10
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3],xmm4[4],xmm0[4],xmm4[5],xmm0[5],xmm4[6],xmm0[6],xmm4[7],xmm0[7]
+; SSE-NEXT:    pxor %xmm10, %xmm10
+; SSE-NEXT:    pand %xmm7, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm7, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pandn %xmm1, %xmm7
+; SSE-NEXT:    pandn %xmm2, %xmm7
 ; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3]
-; SSE-NEXT:    packuswb %xmm1, %xmm1
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3]
+; SSE-NEXT:    packuswb %xmm2, %xmm2
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm0, %xmm10
-; SSE-NEXT:    pandn %xmm1, %xmm10
-; SSE-NEXT:    pshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,3,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,4,6,7]
-; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm0, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm10
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0]
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    pandn %xmm10, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm0, %xmm8
-; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,0,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm0, %xmm11
+; SSE-NEXT:    pandn %xmm2, %xmm11
+; SSE-NEXT:    pshufd $236, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = mem[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,0,3,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,4,6,7]
+; SSE-NEXT:    packuswb %xmm2, %xmm2
+; SSE-NEXT:    pand %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    pand %xmm0, %xmm10
-; SSE-NEXT:    por %xmm1, %xmm10
-; SSE-NEXT:    movdqa %xmm10, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1],xmm1[2],xmm9[2],xmm1[3],xmm9[3],xmm1[4],xmm9[4],xmm1[5],xmm9[5],xmm1[6],xmm9[6],xmm1[7],xmm9[7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,0,65535,65535,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm9[8],xmm10[9],xmm9[9],xmm10[10],xmm9[10],xmm10[11],xmm9[11],xmm10[12],xmm9[12],xmm10[13],xmm9[13],xmm10[14],xmm9[14],xmm10[15],xmm9[15]
-; SSE-NEXT:    pand %xmm3, %xmm10
-; SSE-NEXT:    por %xmm0, %xmm10
-; SSE-NEXT:    packuswb %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    por %xmm2, %xmm11
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0]
+; SSE-NEXT:    movdqa %xmm7, %xmm4
+; SSE-NEXT:    pandn %xmm11, %xmm4
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pand %xmm7, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,0,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    pand %xmm0, %xmm11
+; SSE-NEXT:    por %xmm2, %xmm11
+; SSE-NEXT:    movdqa %xmm11, %xmm2
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1],xmm2[2],xmm10[2],xmm2[3],xmm10[3],xmm2[4],xmm10[4],xmm2[5],xmm10[5],xmm2[6],xmm10[6],xmm2[7],xmm10[7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,0,65535,65535,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    pandn %xmm2, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm10[8],xmm11[9],xmm10[9],xmm11[10],xmm10[10],xmm11[11],xmm10[11],xmm11[12],xmm10[12],xmm11[13],xmm10[13],xmm11[14],xmm10[14],xmm11[15],xmm10[15]
+; SSE-NEXT:    pand %xmm4, %xmm11
+; SSE-NEXT:    por %xmm0, %xmm11
+; SSE-NEXT:    packuswb %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm1, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[0,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
 ; SSE-NEXT:    pand %xmm2, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm1
+; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    por %xmm0, %xmm8
 ; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    por %xmm1, %xmm13
-; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    pxor %xmm11, %xmm11
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm11[8],xmm0[9],xmm11[9],xmm0[10],xmm11[10],xmm0[11],xmm11[11],xmm0[12],xmm11[12],xmm0[13],xmm11[13],xmm0[14],xmm11[14],xmm0[15],xmm11[15]
+; SSE-NEXT:    movdqa %xmm4, %xmm1
+; SSE-NEXT:    pandn %xmm0, %xmm1
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm11[0],xmm8[1],xmm11[1],xmm8[2],xmm11[2],xmm8[3],xmm11[3],xmm8[4],xmm11[4],xmm8[5],xmm11[5],xmm8[6],xmm11[6],xmm8[7],xmm11[7]
+; SSE-NEXT:    pand %xmm4, %xmm8
+; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    por %xmm1, %xmm8
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[1,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm0[0],xmm9[1],xmm0[1]
+; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm11[8],xmm1[9],xmm11[9],xmm1[10],xmm11[10],xmm1[11],xmm11[11],xmm1[12],xmm11[12],xmm1[13],xmm11[13],xmm1[14],xmm11[14],xmm1[15],xmm11[15]
 ; SSE-NEXT:    pxor %xmm6, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm6[8],xmm1[9],xmm6[9],xmm1[10],xmm6[10],xmm1[11],xmm6[11],xmm1[12],xmm6[12],xmm1[13],xmm6[13],xmm1[14],xmm6[14],xmm1[15],xmm6[15]
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm6[0],xmm13[1],xmm6[1],xmm13[2],xmm6[2],xmm13[3],xmm6[3],xmm13[4],xmm6[4],xmm13[5],xmm6[5],xmm13[6],xmm6[6],xmm13[7],xmm6[7]
-; SSE-NEXT:    pand %xmm3, %xmm13
-; SSE-NEXT:    movdqa %xmm3, %xmm5
-; SSE-NEXT:    por %xmm2, %xmm13
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[1,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm1[0],xmm11[1],xmm1[1]
-; SSE-NEXT:    movdqa %xmm11, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm6[8],xmm2[9],xmm6[9],xmm2[10],xmm6[10],xmm2[11],xmm6[11],xmm2[12],xmm6[12],xmm2[13],xmm6[13],xmm2[14],xmm6[14],xmm2[15],xmm6[15]
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,65535,0,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm1, %xmm10
-; SSE-NEXT:    pandn %xmm2, %xmm10
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm6[0],xmm11[1],xmm6[1],xmm11[2],xmm6[2],xmm11[3],xmm6[3],xmm11[4],xmm6[4],xmm11[5],xmm6[5],xmm11[6],xmm6[6],xmm11[7],xmm6[7]
-; SSE-NEXT:    pand %xmm1, %xmm11
-; SSE-NEXT:    por %xmm10, %xmm11
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,1,1]
-; SSE-NEXT:    packuswb %xmm2, %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,0,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm0, %xmm11
+; SSE-NEXT:    pandn %xmm1, %xmm11
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm6[0],xmm9[1],xmm6[1],xmm9[2],xmm6[2],xmm9[3],xmm6[3],xmm9[4],xmm6[4],xmm9[5],xmm6[5],xmm9[6],xmm6[6],xmm9[7],xmm6[7]
+; SSE-NEXT:    pand %xmm0, %xmm9
+; SSE-NEXT:    por %xmm11, %xmm9
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm9[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
+; SSE-NEXT:    packuswb %xmm1, %xmm4
 ; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,255,255,255,0,0,0,0,255,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm6, %xmm4
-; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm13[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,3,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pand %xmm6, %xmm2
-; SSE-NEXT:    movdqa %xmm6, %xmm13
-; SSE-NEXT:    por %xmm2, %xmm4
-; SSE-NEXT:    pand %xmm8, %xmm4
-; SSE-NEXT:    por %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,65535,0,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm6, %xmm5
+; SSE-NEXT:    pandn %xmm4, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,3,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm1, %xmm1
+; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm6, %xmm9
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    pand %xmm7, %xmm5
+; SSE-NEXT:    por %xmm2, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,65535,65535,0,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm11, %xmm1
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pand %xmm10, %xmm2
-; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pxor %xmm6, %xmm6
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3],xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
-; SSE-NEXT:    movdqa %xmm5, %xmm11
-; SSE-NEXT:    movdqa %xmm5, %xmm3
-; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm6[8],xmm2[9],xmm6[9],xmm2[10],xmm6[10],xmm2[11],xmm6[11],xmm2[12],xmm6[12],xmm2[13],xmm6[13],xmm2[14],xmm6[14],xmm2[15],xmm6[15]
+; SSE-NEXT:    pand %xmm11, %xmm2
+; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    pxor %xmm4, %xmm4
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3],xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
+; SSE-NEXT:    movdqa %xmm3, %xmm5
+; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm4[8],xmm2[9],xmm4[9],xmm2[10],xmm4[10],xmm2[11],xmm4[11],xmm2[12],xmm4[12],xmm2[13],xmm4[13],xmm2[14],xmm4[14],xmm2[15],xmm4[15]
 ; SSE-NEXT:    pand %xmm5, %xmm2
+; SSE-NEXT:    movdqa %xmm5, %xmm6
 ; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    packuswb %xmm15, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm4, %xmm3
-; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,3,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm4, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    movdqa %xmm8, %xmm15
-; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    packuswb %xmm14, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,65535,65535,65535,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,3,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
+; SSE-NEXT:    packuswb %xmm1, %xmm1
+; SSE-NEXT:    pand %xmm5, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    movdqa %xmm7, %xmm5
+; SSE-NEXT:    pandn %xmm3, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm5, %xmm2
+; SSE-NEXT:    pandn %xmm8, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm12
 ; SSE-NEXT:    movdqa %xmm12, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm6[8],xmm2[9],xmm6[9],xmm2[10],xmm6[10],xmm2[11],xmm6[11],xmm2[12],xmm6[12],xmm2[13],xmm6[13],xmm2[14],xmm6[14],xmm2[15],xmm6[15]
-; SSE-NEXT:    movdqa %xmm11, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm4[8],xmm2[9],xmm4[9],xmm2[10],xmm4[10],xmm2[11],xmm4[11],xmm2[12],xmm4[12],xmm2[13],xmm4[13],xmm2[14],xmm4[14],xmm2[15],xmm4[15]
+; SSE-NEXT:    movdqa %xmm6, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm6[0],xmm12[1],xmm6[1],xmm12[2],xmm6[2],xmm12[3],xmm6[3],xmm12[4],xmm6[4],xmm12[5],xmm6[5],xmm12[6],xmm6[6],xmm12[7],xmm6[7]
-; SSE-NEXT:    pand %xmm11, %xmm12
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm4[0],xmm12[1],xmm4[1],xmm12[2],xmm4[2],xmm12[3],xmm4[3],xmm12[4],xmm4[4],xmm12[5],xmm4[5],xmm12[6],xmm4[6],xmm12[7],xmm4[7]
+; SSE-NEXT:    pand %xmm6, %xmm12
 ; SSE-NEXT:    por %xmm3, %xmm12
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[1,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm6[1,3,2,3]
+; SSE-NEXT:    movdqa (%rsp), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
 ; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm6[8],xmm2[9],xmm6[9],xmm2[10],xmm6[10],xmm2[11],xmm6[11],xmm2[12],xmm6[12],xmm2[13],xmm6[13],xmm2[14],xmm6[14],xmm2[15],xmm6[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1],xmm3[2],xmm6[2],xmm3[3],xmm6[3],xmm3[4],xmm6[4],xmm3[5],xmm6[5],xmm3[6],xmm6[6],xmm3[7],xmm6[7]
-; SSE-NEXT:    pand %xmm1, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm4[8],xmm2[9],xmm4[9],xmm2[10],xmm4[10],xmm2[11],xmm4[11],xmm2[12],xmm4[12],xmm2[13],xmm4[13],xmm2[14],xmm4[14],xmm2[15],xmm4[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
+; SSE-NEXT:    pxor %xmm7, %xmm7
+; SSE-NEXT:    pand %xmm0, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm0
+; SSE-NEXT:    por %xmm3, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm0[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,1,1]
+; SSE-NEXT:    packuswb %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm12[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,3,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,0,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm2, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm0
+; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    por %xmm1, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    movdqa %xmm11, %xmm1
+; SSE-NEXT:    pandn %xmm10, %xmm1
+; SSE-NEXT:    pand %xmm11, %xmm15
+; SSE-NEXT:    por %xmm1, %xmm15
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm7[8],xmm1[9],xmm7[9],xmm1[10],xmm7[10],xmm1[11],xmm7[11],xmm1[12],xmm7[12],xmm1[13],xmm7[13],xmm1[14],xmm7[14],xmm1[15],xmm7[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm7[0],xmm15[1],xmm7[1],xmm15[2],xmm7[2],xmm15[3],xmm7[3],xmm15[4],xmm7[4],xmm15[5],xmm7[5],xmm15[6],xmm7[6],xmm15[7],xmm7[7]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm15 = xmm15[4],xmm1[4],xmm15[5],xmm1[5],xmm15[6],xmm1[6],xmm15[7],xmm1[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm15[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,7,6]
+; SSE-NEXT:    psrlq $48, %xmm1
+; SSE-NEXT:    packuswb %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm9, %xmm14
+; SSE-NEXT:    movdqa %xmm9, %xmm1
 ; SSE-NEXT:    pandn %xmm2, %xmm1
-; SSE-NEXT:    por %xmm3, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,1,1]
-; SSE-NEXT:    packuswb %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm13, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,3,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,0,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm13, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    pand %xmm15, %xmm2
-; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pandn %xmm7, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pand %xmm10, %xmm2
-; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm6[8],xmm0[9],xmm6[9],xmm0[10],xmm6[10],xmm0[11],xmm6[11],xmm0[12],xmm6[12],xmm0[13],xmm6[13],xmm0[14],xmm6[14],xmm0[15],xmm6[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3],xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,6]
-; SSE-NEXT:    psrlq $48, %xmm0
-; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,0,65535,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,0,65535,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm11, %xmm2
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm9, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    pand %xmm3, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3],xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
+; SSE-NEXT:    pandn %xmm9, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm10, %xmm3
+; SSE-NEXT:    pand %xmm11, %xmm3
+; SSE-NEXT:    movdqa %xmm11, %xmm12
+; SSE-NEXT:    por %xmm2, %xmm3
+; SSE-NEXT:    movdqa %xmm3, %xmm2
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm7[0],xmm2[1],xmm7[1],xmm2[2],xmm7[2],xmm2[3],xmm7[3],xmm2[4],xmm7[4],xmm2[5],xmm7[5],xmm2[6],xmm7[6],xmm2[7],xmm7[7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,0,65535,0,65535,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    movdqa %xmm4, %xmm5
+; SSE-NEXT:    pandn %xmm2, %xmm5
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm7[8],xmm3[9],xmm7[9],xmm3[10],xmm7[10],xmm3[11],xmm7[11],xmm3[12],xmm7[12],xmm3[13],xmm7[13],xmm3[14],xmm7[14],xmm3[15],xmm7[15]
+; SSE-NEXT:    pand %xmm4, %xmm3
+; SSE-NEXT:    por %xmm5, %xmm3
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm3[3,2,1,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    packuswb %xmm2, %xmm2
+; SSE-NEXT:    pand %xmm14, %xmm2
+; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm11, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm13
+; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm7[8],xmm1[9],xmm7[9],xmm1[10],xmm7[10],xmm1[11],xmm7[11],xmm1[12],xmm7[12],xmm1[13],xmm7[13],xmm1[14],xmm7[14],xmm1[15],xmm7[15]
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,0,65535,65535,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm5, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm6[8],xmm2[9],xmm6[9],xmm2[10],xmm6[10],xmm2[11],xmm6[11],xmm2[12],xmm6[12],xmm2[13],xmm6[13],xmm2[14],xmm6[14],xmm2[15],xmm6[15]
-; SSE-NEXT:    pand %xmm4, %xmm2
-; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[3,2,1,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm7[0],xmm13[1],xmm7[1],xmm13[2],xmm7[2],xmm13[3],xmm7[3],xmm13[4],xmm7[4],xmm13[5],xmm7[5],xmm13[6],xmm7[6],xmm13[7],xmm7[7]
+; SSE-NEXT:    pand %xmm5, %xmm13
+; SSE-NEXT:    por %xmm3, %xmm13
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm14[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm13, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,65535,65535,65535,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm13[0,2,1,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,6,7]
+; SSE-NEXT:    packuswb %xmm1, %xmm1
+; SSE-NEXT:    pand %xmm5, %xmm1
+; SSE-NEXT:    movdqa %xmm5, %xmm15
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0]
+; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    pandn %xmm3, %xmm1
+; SSE-NEXT:    pand %xmm5, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm1, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    pandn %xmm6, %xmm1
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm12, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm14
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm6[8],xmm0[9],xmm6[9],xmm0[10],xmm6[10],xmm0[11],xmm6[11],xmm0[12],xmm6[12],xmm0[13],xmm6[13],xmm0[14],xmm6[14],xmm0[15],xmm6[15]
-; SSE-NEXT:    movdqa %xmm11, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm6[0],xmm14[1],xmm6[1],xmm14[2],xmm6[2],xmm14[3],xmm6[3],xmm14[4],xmm6[4],xmm14[5],xmm6[5],xmm14[6],xmm6[6],xmm14[7],xmm6[7]
-; SSE-NEXT:    pand %xmm11, %xmm14
-; SSE-NEXT:    por %xmm2, %xmm14
-; SSE-NEXT:    pshuflw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm10, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm14[0,2,1,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm10, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    pand %xmm15, %xmm1
+; SSE-NEXT:    pand %xmm2, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,65535,65535,0,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    pandn %xmm8, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pand %xmm11, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm6[8],xmm0[9],xmm6[9],xmm0[10],xmm6[10],xmm0[11],xmm6[11],xmm0[12],xmm6[12],xmm0[13],xmm6[13],xmm0[14],xmm6[14],xmm0[15],xmm6[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3],xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,6]
-; SSE-NEXT:    psrlq $48, %xmm0
-; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,0,65535,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    movdqa %xmm5, %xmm11
-; SSE-NEXT:    pandn %xmm5, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    pand %xmm8, %xmm2
-; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3],xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm7[8],xmm1[9],xmm7[9],xmm1[10],xmm7[10],xmm1[11],xmm7[11],xmm1[12],xmm7[12],xmm1[13],xmm7[13],xmm1[14],xmm7[14],xmm1[15],xmm7[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1],xmm0[2],xmm7[2],xmm0[3],xmm7[3],xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm0[0,3,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,7,6]
+; SSE-NEXT:    psrlq $48, %xmm1
+; SSE-NEXT:    packuswb %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    movdqa %xmm8, %xmm13
+; SSE-NEXT:    pandn %xmm8, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm0, %xmm3
+; SSE-NEXT:    pand %xmm12, %xmm3
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3],xmm1[4],xmm7[4],xmm1[5],xmm7[5],xmm1[6],xmm7[6],xmm1[7],xmm7[7]
 ; SSE-NEXT:    movdqa %xmm4, %xmm5
-; SSE-NEXT:    pandn %xmm0, %xmm5
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm6[8],xmm2[9],xmm6[9],xmm2[10],xmm6[10],xmm2[11],xmm6[11],xmm2[12],xmm6[12],xmm2[13],xmm6[13],xmm2[14],xmm6[14],xmm2[15],xmm6[15]
-; SSE-NEXT:    pand %xmm4, %xmm2
-; SSE-NEXT:    por %xmm5, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[3,2,1,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm13, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm13
-; SSE-NEXT:    por %xmm13, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm1, %xmm5
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm7[8],xmm3[9],xmm7[9],xmm3[10],xmm7[10],xmm3[11],xmm7[11],xmm3[12],xmm7[12],xmm3[13],xmm7[13],xmm3[14],xmm7[14],xmm3[15],xmm7[15]
+; SSE-NEXT:    pand %xmm4, %xmm3
+; SSE-NEXT:    por %xmm5, %xmm3
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm3[3,2,1,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    packuswb %xmm1, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,0,0,0,0,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm3, %xmm1
+; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm3, %xmm5
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm8, %xmm1
-; SSE-NEXT:    movdqa (%rsp), %xmm5 # 16-byte Reload
-; SSE-NEXT:    por %xmm1, %xmm5
-; SSE-NEXT:    movdqa %xmm5, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm6[8],xmm1[9],xmm6[9],xmm1[10],xmm6[10],xmm1[11],xmm6[11],xmm1[12],xmm6[12],xmm1[13],xmm6[13],xmm1[14],xmm6[14],xmm1[15],xmm6[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3],xmm5[4],xmm6[4],xmm5[5],xmm6[5],xmm5[6],xmm6[6],xmm5[7],xmm6[7]
+; SSE-NEXT:    pandn %xmm3, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    por %xmm1, %xmm6
+; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm7[8],xmm1[9],xmm7[9],xmm1[10],xmm7[10],xmm1[11],xmm7[11],xmm1[12],xmm7[12],xmm1[13],xmm7[13],xmm1[14],xmm7[14],xmm1[15],xmm7[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3],xmm6[4],xmm7[4],xmm6[5],xmm7[5],xmm6[6],xmm7[6],xmm6[7],xmm7[7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,0,65535,65535,65535,65535,0,65535]
-; SSE-NEXT:    pand %xmm2, %xmm5
+; SSE-NEXT:    pand %xmm2, %xmm6
 ; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    por %xmm5, %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm5
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm13[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    por %xmm6, %xmm2
+; SSE-NEXT:    movdqa %xmm2, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm12[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    movdqa %xmm15, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm5[0,2,1,0,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm6[0,2,1,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm10, %xmm1
+; SSE-NEXT:    pand %xmm15, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0]
 ; SSE-NEXT:    movdqa %xmm15, %xmm1
 ; SSE-NEXT:    pandn %xmm2, %xmm1
-; SSE-NEXT:    pand %xmm15, %xmm0
-; SSE-NEXT:    movdqa %xmm15, %xmm14
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pandn %xmm9, %xmm0
-; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    pand %xmm15, %xmm2
-; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSE-NEXT:    pand %xmm15, %xmm5
+; SSE-NEXT:    por %xmm5, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,0,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    pandn %xmm9, %xmm1
+; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    pand %xmm5, %xmm2
+; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm7[8],xmm1[9],xmm7[9],xmm1[10],xmm7[10],xmm1[11],xmm7[11],xmm1[12],xmm7[12],xmm1[13],xmm7[13],xmm1[14],xmm7[14],xmm1[15],xmm7[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm7[0],xmm2[1],xmm7[1],xmm2[2],xmm7[2],xmm2[3],xmm7[3],xmm2[4],xmm7[4],xmm2[5],xmm7[5],xmm2[6],xmm7[6],xmm2[7],xmm7[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm7[8],xmm1[9],xmm7[9],xmm1[10],xmm7[10],xmm1[11],xmm7[11],xmm1[12],xmm7[12],xmm1[13],xmm7[13],xmm1[14],xmm7[14],xmm1[15],xmm7[15]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm0, %xmm6
+; SSE-NEXT:    pandn %xmm1, %xmm6
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; SSE-NEXT:    por %xmm6, %xmm5
-; SSE-NEXT:    packuswb %xmm0, %xmm5
+; SSE-NEXT:    packuswb %xmm1, %xmm5
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[0,1,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm2[0],xmm0[1,2,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,0,65535,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm9, %xmm2
-; SSE-NEXT:    pandn %xmm12, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm7, %xmm5
-; SSE-NEXT:    pand %xmm9, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[0,1,3,3]
+; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,0,65535,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm6, %xmm2
+; SSE-NEXT:    pandn %xmm11, %xmm2
+; SSE-NEXT:    movdqa %xmm11, %xmm8
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm9, %xmm5
+; SSE-NEXT:    pand %xmm6, %xmm5
+; SSE-NEXT:    movdqa %xmm6, %xmm11
 ; SSE-NEXT:    por %xmm2, %xmm5
 ; SSE-NEXT:    movdqa %xmm5, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm7[0],xmm2[1],xmm7[1],xmm2[2],xmm7[2],xmm2[3],xmm7[3],xmm2[4],xmm7[4],xmm2[5],xmm7[5],xmm2[6],xmm7[6],xmm2[7],xmm7[7]
 ; SSE-NEXT:    movdqa %xmm4, %xmm6
 ; SSE-NEXT:    pandn %xmm2, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm1[8],xmm5[9],xmm1[9],xmm5[10],xmm1[10],xmm5[11],xmm1[11],xmm5[12],xmm1[12],xmm5[13],xmm1[13],xmm5[14],xmm1[14],xmm5[15],xmm1[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm7[8],xmm5[9],xmm7[9],xmm5[10],xmm7[10],xmm5[11],xmm7[11],xmm5[12],xmm7[12],xmm5[13],xmm7[13],xmm5[14],xmm7[14],xmm5[15],xmm7[15]
 ; SSE-NEXT:    pand %xmm4, %xmm5
 ; SSE-NEXT:    por %xmm6, %xmm5
-; SSE-NEXT:    pshufd $100, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[0,1,2,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    movdqa %xmm10, %xmm6
+; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [65535,65535,65535,65535,65535,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm14, %xmm6
 ; SSE-NEXT:    pandn %xmm2, %xmm6
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[0,1,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pand %xmm10, %xmm2
+; SSE-NEXT:    pand %xmm14, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm6
-; SSE-NEXT:    movdqa %xmm14, %xmm1
-; SSE-NEXT:    pandn %xmm6, %xmm14
-; SSE-NEXT:    andps %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm1, %xmm6
-; SSE-NEXT:    por %xmm0, %xmm14
-; SSE-NEXT:    movdqa %xmm15, %xmm1
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    movdqa %xmm11, %xmm15
-; SSE-NEXT:    pandn %xmm11, %xmm0
-; SSE-NEXT:    pand %xmm1, %xmm3
-; SSE-NEXT:    por %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3],xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[2,1,2,3]
+; SSE-NEXT:    movdqa %xmm15, %xmm2
+; SSE-NEXT:    pandn %xmm6, %xmm2
+; SSE-NEXT:    andps %xmm15, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,0,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    pandn %xmm13, %xmm1
+; SSE-NEXT:    movdqa %xmm0, %xmm5
+; SSE-NEXT:    pand %xmm2, %xmm5
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    pxor %xmm0, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm0[8],xmm1[9],xmm0[9],xmm1[10],xmm0[10],xmm1[11],xmm0[11],xmm1[12],xmm0[12],xmm1[13],xmm0[13],xmm1[14],xmm0[14],xmm1[15],xmm0[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3],xmm5[4],xmm0[4],xmm5[5],xmm0[5],xmm5[6],xmm0[6],xmm5[7],xmm0[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    packuswb %xmm0, %xmm2
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm0[8],xmm1[9],xmm0[9],xmm1[10],xmm0[10],xmm1[11],xmm0[11],xmm1[12],xmm0[12],xmm1[13],xmm0[13],xmm1[14],xmm0[14],xmm1[15],xmm0[15]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm1, %xmm7
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    por %xmm7, %xmm6
+; SSE-NEXT:    packuswb %xmm1, %xmm6
 ; SSE-NEXT:    packuswb %xmm5, %xmm5
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,1,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm5[0],xmm0[1,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm2, %xmm5
-; SSE-NEXT:    movdqa %xmm2, %xmm3
-; SSE-NEXT:    movdqa %xmm9, %xmm2
-; SSE-NEXT:    pand %xmm9, %xmm5
-; SSE-NEXT:    pandn %xmm8, %xmm2
-; SSE-NEXT:    movdqa %xmm8, %xmm9
-; SSE-NEXT:    por %xmm5, %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm5
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3],xmm5[4],xmm1[4],xmm5[5],xmm1[5],xmm5[6],xmm1[6],xmm5[7],xmm1[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
-; SSE-NEXT:    pand %xmm4, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[0,1,3,3]
+; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm5[0],xmm1[1,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm11, %xmm6
+; SSE-NEXT:    pand %xmm11, %xmm5
+; SSE-NEXT:    pandn %xmm3, %xmm6
+; SSE-NEXT:    por %xmm5, %xmm6
+; SSE-NEXT:    movdqa %xmm6, %xmm5
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3],xmm5[4],xmm0[4],xmm5[5],xmm0[5],xmm5[6],xmm0[6],xmm5[7],xmm0[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm0[8],xmm6[9],xmm0[9],xmm6[10],xmm0[10],xmm6[11],xmm0[11],xmm6[12],xmm0[12],xmm6[13],xmm0[13],xmm6[14],xmm0[14],xmm6[15],xmm0[15]
+; SSE-NEXT:    pxor %xmm13, %xmm13
+; SSE-NEXT:    pand %xmm4, %xmm6
 ; SSE-NEXT:    pandn %xmm5, %xmm4
-; SSE-NEXT:    por %xmm2, %xmm4
+; SSE-NEXT:    por %xmm6, %xmm4
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    packuswb %xmm4, %xmm4
-; SSE-NEXT:    pand %xmm10, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm13[0,1,2,1]
+; SSE-NEXT:    movdqa %xmm14, %xmm0
+; SSE-NEXT:    pand %xmm14, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm12[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    packuswb %xmm5, %xmm5
-; SSE-NEXT:    pandn %xmm5, %xmm10
-; SSE-NEXT:    por %xmm4, %xmm10
-; SSE-NEXT:    movdqa %xmm6, %xmm4
-; SSE-NEXT:    pandn %xmm10, %xmm4
-; SSE-NEXT:    andps %xmm6, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,65535,65535,0,65535]
-; SSE-NEXT:    pand %xmm2, %xmm0
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm0, %xmm5
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm1[8],xmm5[9],xmm1[9],xmm5[10],xmm1[10],xmm5[11],xmm1[11],xmm5[12],xmm1[12],xmm5[13],xmm1[13],xmm5[14],xmm1[14],xmm5[15],xmm1[15]
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [0,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm10, %xmm6
-; SSE-NEXT:    pandn %xmm5, %xmm6
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; SSE-NEXT:    pand %xmm10, %xmm0
-; SSE-NEXT:    por %xmm6, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm0[0,1,2,3,5,4,7,6]
+; SSE-NEXT:    pandn %xmm5, %xmm0
+; SSE-NEXT:    por %xmm4, %xmm0
+; SSE-NEXT:    movdqa %xmm15, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    andps %xmm15, %xmm1
+; SSE-NEXT:    movdqa %xmm15, %xmm14
+; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,0,65535,65535,0,65535]
+; SSE-NEXT:    pand %xmm3, %xmm1
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm1, %xmm5
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm13[8],xmm5[9],xmm13[9],xmm5[10],xmm13[10],xmm5[11],xmm13[11],xmm5[12],xmm13[12],xmm5[13],xmm13[13],xmm5[14],xmm13[14],xmm5[15],xmm13[15]
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [0,65535,65535,65535,65535,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm6, %xmm7
+; SSE-NEXT:    pandn %xmm5, %xmm7
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1],xmm1[2],xmm13[2],xmm1[3],xmm13[3],xmm1[4],xmm13[4],xmm1[5],xmm13[5],xmm1[6],xmm13[6],xmm1[7],xmm13[7]
+; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    por %xmm7, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm1[0,1,2,3,5,4,7,6]
 ; SSE-NEXT:    psrldq {{.*#+}} xmm5 = xmm5[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    packuswb %xmm5, %xmm8
-; SSE-NEXT:    movdqa %xmm2, %xmm11
+; SSE-NEXT:    packuswb %xmm5, %xmm7
+; SSE-NEXT:    movdqa %xmm3, %xmm11
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    pandn %xmm0, %xmm11
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm13[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm5, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3],xmm5[4],xmm1[4],xmm5[5],xmm1[5],xmm5[6],xmm1[6],xmm5[7],xmm1[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,3,2,3]
+; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm10[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
+; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm13[8],xmm1[9],xmm13[9],xmm1[10],xmm13[10],xmm1[11],xmm13[11],xmm1[12],xmm13[12],xmm1[13],xmm13[13],xmm1[14],xmm13[14],xmm1[15],xmm13[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm13[0],xmm5[1],xmm13[1],xmm5[2],xmm13[2],xmm5[3],xmm13[3],xmm5[4],xmm13[4],xmm5[5],xmm13[5],xmm5[6],xmm13[6],xmm5[7],xmm13[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1],xmm5[2],xmm1[2],xmm5[3],xmm1[3]
 ; SSE-NEXT:    packuswb %xmm5, %xmm5
-; SSE-NEXT:    movss {{.*#+}} xmm8 = xmm5[0],xmm8[1,2,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    pandn %xmm12, %xmm0
-; SSE-NEXT:    movdqa %xmm7, %xmm5
-; SSE-NEXT:    pand %xmm6, %xmm5
-; SSE-NEXT:    por %xmm0, %xmm5
-; SSE-NEXT:    movdqa %xmm5, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,0,65535,0,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm6, %xmm7
-; SSE-NEXT:    pandn %xmm0, %xmm7
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm1[8],xmm5[9],xmm1[9],xmm5[10],xmm1[10],xmm5[11],xmm1[11],xmm5[12],xmm1[12],xmm5[13],xmm1[13],xmm5[14],xmm1[14],xmm5[15],xmm1[15]
-; SSE-NEXT:    pand %xmm6, %xmm5
-; SSE-NEXT:    por %xmm7, %xmm5
+; SSE-NEXT:    movss {{.*#+}} xmm7 = xmm5[0],xmm7[1,2,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [65535,65535,0,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    pandn %xmm8, %xmm1
+; SSE-NEXT:    movdqa %xmm9, %xmm5
+; SSE-NEXT:    pand %xmm15, %xmm5
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm5, %xmm9
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm13[0],xmm9[1],xmm13[1],xmm9[2],xmm13[2],xmm9[3],xmm13[3],xmm9[4],xmm13[4],xmm9[5],xmm13[5],xmm9[6],xmm13[6],xmm9[7],xmm13[7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,0,65535,0,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm1, %xmm10
+; SSE-NEXT:    pandn %xmm9, %xmm10
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm13[8],xmm5[9],xmm13[9],xmm5[10],xmm13[10],xmm5[11],xmm13[11],xmm5[12],xmm13[12],xmm5[13],xmm13[13],xmm5[14],xmm13[14],xmm5[15],xmm13[15]
+; SSE-NEXT:    pand %xmm1, %xmm5
+; SSE-NEXT:    por %xmm10, %xmm5
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,7,6]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0]
-; SSE-NEXT:    movdqa %xmm7, %xmm12
-; SSE-NEXT:    pandn %xmm0, %xmm12
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,3,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm7, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm12
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0]
-; SSE-NEXT:    movdqa %xmm0, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm0[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,4,7,6]
+; SSE-NEXT:    packuswb %xmm9, %xmm9
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0]
+; SSE-NEXT:    movdqa %xmm10, %xmm12
+; SSE-NEXT:    pandn %xmm9, %xmm12
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,0,3,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    packuswb %xmm5, %xmm5
+; SSE-NEXT:    pand %xmm10, %xmm5
+; SSE-NEXT:    por %xmm5, %xmm12
+; SSE-NEXT:    movdqa %xmm14, %xmm5
 ; SSE-NEXT:    pandn %xmm12, %xmm5
-; SSE-NEXT:    andps %xmm0, %xmm8
-; SSE-NEXT:    por %xmm8, %xmm5
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm2, %xmm0
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm0, %xmm12
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm1[8],xmm12[9],xmm1[9],xmm12[10],xmm1[10],xmm12[11],xmm1[11],xmm12[12],xmm1[12],xmm12[13],xmm1[13],xmm12[14],xmm1[14],xmm12[15],xmm1[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; SSE-NEXT:    pand %xmm10, %xmm0
-; SSE-NEXT:    pandn %xmm12, %xmm10
-; SSE-NEXT:    por %xmm0, %xmm10
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm0[0,1,2,3,5,4,7,6]
-; SSE-NEXT:    psrldq {{.*#+}} xmm12 = xmm12[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    packuswb %xmm12, %xmm8
-; SSE-NEXT:    movdqa %xmm13, %xmm12
-; SSE-NEXT:    pand %xmm2, %xmm12
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[0,2,2,3]
-; SSE-NEXT:    pand %xmm2, %xmm10
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pandn %xmm15, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm15[1,3,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
-; SSE-NEXT:    movdqa %xmm0, %xmm10
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8],xmm1[8],xmm10[9],xmm1[9],xmm10[10],xmm1[10],xmm10[11],xmm1[11],xmm10[12],xmm1[12],xmm10[13],xmm1[13],xmm10[14],xmm1[14],xmm10[15],xmm1[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm10[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1],xmm0[2],xmm10[2],xmm0[3],xmm10[3]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movss {{.*#+}} xmm8 = xmm0[0],xmm8[1,2,3]
-; SSE-NEXT:    movdqa %xmm3, %xmm13
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT:    pand %xmm3, %xmm0
-; SSE-NEXT:    pandn %xmm9, %xmm3
-; SSE-NEXT:    movdqa %xmm9, %xmm15
-; SSE-NEXT:    por %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm1[8],xmm3[9],xmm1[9],xmm3[10],xmm1[10],xmm3[11],xmm1[11],xmm3[12],xmm1[12],xmm3[13],xmm1[13],xmm3[14],xmm1[14],xmm3[15],xmm1[15]
-; SSE-NEXT:    pand %xmm6, %xmm3
-; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    por %xmm3, %xmm6
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,7,6]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
+; SSE-NEXT:    andps %xmm14, %xmm7
+; SSE-NEXT:    por %xmm7, %xmm5
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pand %xmm3, %xmm7
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm7, %xmm9
-; SSE-NEXT:    pandn %xmm0, %xmm9
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,3,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm7, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm9
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm13[8],xmm9[9],xmm13[9],xmm9[10],xmm13[10],xmm9[11],xmm13[11],xmm9[12],xmm13[12],xmm9[13],xmm13[13],xmm9[14],xmm13[14],xmm9[15],xmm13[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm13[0],xmm7[1],xmm13[1],xmm7[2],xmm13[2],xmm7[3],xmm13[3],xmm7[4],xmm13[4],xmm7[5],xmm13[5],xmm7[6],xmm13[6],xmm7[7],xmm13[7]
+; SSE-NEXT:    pand %xmm6, %xmm7
+; SSE-NEXT:    pandn %xmm9, %xmm6
+; SSE-NEXT:    por %xmm7, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm6[0,1,2,3,5,4,7,6]
+; SSE-NEXT:    psrldq {{.*#+}} xmm9 = xmm9[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    packuswb %xmm9, %xmm7
+; SSE-NEXT:    movdqa %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm2, %xmm12
+; SSE-NEXT:    pand %xmm3, %xmm12
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm8[0,2,2,3]
+; SSE-NEXT:    pand %xmm3, %xmm8
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm9, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[1,3,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm9[0],xmm6[1],xmm9[1]
+; SSE-NEXT:    movdqa %xmm6, %xmm9
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm13[8],xmm9[9],xmm13[9],xmm9[10],xmm13[10],xmm9[11],xmm13[11],xmm9[12],xmm13[12],xmm9[13],xmm13[13],xmm9[14],xmm13[14],xmm9[15],xmm13[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm13[0],xmm6[1],xmm13[1],xmm6[2],xmm13[2],xmm6[3],xmm13[3],xmm6[4],xmm13[4],xmm6[5],xmm13[5],xmm6[6],xmm13[6],xmm6[7],xmm13[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm9[0],xmm6[1],xmm9[1],xmm6[2],xmm9[2],xmm6[3],xmm9[3]
+; SSE-NEXT:    packuswb %xmm6, %xmm6
+; SSE-NEXT:    movss {{.*#+}} xmm7 = xmm6[0],xmm7[1,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm4, %xmm6
+; SSE-NEXT:    movdqa %xmm15, %xmm3
+; SSE-NEXT:    pand %xmm15, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    por %xmm6, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, %xmm6
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm13[0],xmm6[1],xmm13[1],xmm6[2],xmm13[2],xmm6[3],xmm13[3],xmm6[4],xmm13[4],xmm6[5],xmm13[5],xmm6[6],xmm13[6],xmm6[7],xmm13[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm13[8],xmm3[9],xmm13[9],xmm3[10],xmm13[10],xmm3[11],xmm13[11],xmm3[12],xmm13[12],xmm3[13],xmm13[13],xmm3[14],xmm13[14],xmm3[15],xmm13[15]
+; SSE-NEXT:    pand %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm6, %xmm1
+; SSE-NEXT:    por %xmm3, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm3[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,7,6]
+; SSE-NEXT:    packuswb %xmm6, %xmm6
+; SSE-NEXT:    movdqa %xmm10, %xmm9
+; SSE-NEXT:    pandn %xmm6, %xmm9
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,3,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    packuswb %xmm1, %xmm1
+; SSE-NEXT:    pand %xmm10, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm9
+; SSE-NEXT:    movdqa %xmm14, %xmm6
 ; SSE-NEXT:    pandn %xmm9, %xmm6
-; SSE-NEXT:    andps %xmm3, %xmm8
-; SSE-NEXT:    por %xmm8, %xmm6
-; SSE-NEXT:    movdqa %xmm12, %xmm1
-; SSE-NEXT:    por %xmm11, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    andps %xmm14, %xmm7
+; SSE-NEXT:    por %xmm7, %xmm6
+; SSE-NEXT:    movdqa %xmm12, %xmm7
+; SSE-NEXT:    por %xmm11, %xmm7
+; SSE-NEXT:    movdqa %xmm7, %xmm1
 ; SSE-NEXT:    pxor %xmm9, %xmm9
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3],xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,4,6,7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm9[8],xmm1[9],xmm9[9],xmm1[10],xmm9[10],xmm1[11],xmm9[11],xmm1[12],xmm9[12],xmm1[13],xmm9[13],xmm1[14],xmm9[14],xmm1[15],xmm9[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm1[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,5,7,6,7]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm9[8],xmm8[9],xmm9[9],xmm8[10],xmm9[10],xmm8[11],xmm9[11],xmm8[12],xmm9[12],xmm8[13],xmm9[13],xmm8[14],xmm9[14],xmm8[15],xmm9[15]
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm8, %xmm10
-; SSE-NEXT:    movdqa %xmm8, %xmm9
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    por %xmm10, %xmm8
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm8[0,1,2,3,4,7,6,5]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm9[3,3,3,3]
-; SSE-NEXT:    packuswb %xmm8, %xmm10
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movss {{.*#+}} xmm10 = xmm0[0],xmm10[1,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm7, %xmm8
-; SSE-NEXT:    pandn %xmm0, %xmm8
-; SSE-NEXT:    pshufd $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[1,3,2,3]
-; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm11 = mem[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,65535,0,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm9, %xmm12
-; SSE-NEXT:    pandn %xmm0, %xmm12
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm1[8],xmm11[9],xmm1[9],xmm11[10],xmm1[10],xmm11[11],xmm1[11],xmm11[12],xmm1[12],xmm11[13],xmm1[13],xmm11[14],xmm1[14],xmm11[15],xmm1[15]
-; SSE-NEXT:    pand %xmm9, %xmm11
-; SSE-NEXT:    por %xmm12, %xmm11
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,1,1,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm7, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1],xmm1[2],xmm9[2],xmm1[3],xmm9[3],xmm1[4],xmm9[4],xmm1[5],xmm9[5],xmm1[6],xmm9[6],xmm1[7],xmm9[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,4,6,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm9[8],xmm7[9],xmm9[9],xmm7[10],xmm9[10],xmm7[11],xmm9[11],xmm7[12],xmm9[12],xmm7[13],xmm9[13],xmm7[14],xmm9[14],xmm7[15],xmm9[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,5,7,6,7]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm7[4],xmm1[5],xmm7[5],xmm1[6],xmm7[6],xmm1[7],xmm7[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm9[8],xmm3[9],xmm9[9],xmm3[10],xmm9[10],xmm3[11],xmm9[11],xmm3[12],xmm9[12],xmm3[13],xmm9[13],xmm3[14],xmm9[14],xmm3[15],xmm9[15]
+; SSE-NEXT:    pxor %xmm15, %xmm15
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm3, %xmm9
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    por %xmm9, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm7[0,1,2,3,4,7,6,5]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm3[3,3,3,3]
+; SSE-NEXT:    packuswb %xmm7, %xmm9
+; SSE-NEXT:    packuswb %xmm1, %xmm1
+; SSE-NEXT:    movss {{.*#+}} xmm9 = xmm1[0],xmm9[1,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    packuswb %xmm1, %xmm1
+; SSE-NEXT:    movdqa %xmm10, %xmm11
+; SSE-NEXT:    pandn %xmm1, %xmm11
+; SSE-NEXT:    pshufd $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[1,3,2,3]
+; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = mem[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm1[0],xmm12[1],xmm1[1]
+; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1],xmm1[2],xmm15[2],xmm1[3],xmm15[3],xmm1[4],xmm15[4],xmm1[5],xmm15[5],xmm1[6],xmm15[6],xmm1[7],xmm15[7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,65535,0,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm7, %xmm13
+; SSE-NEXT:    pandn %xmm1, %xmm13
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm15[8],xmm12[9],xmm15[9],xmm12[10],xmm15[10],xmm12[11],xmm15[11],xmm12[12],xmm15[12],xmm12[13],xmm15[13],xmm12[14],xmm15[14],xmm12[15],xmm15[15]
+; SSE-NEXT:    pand %xmm7, %xmm12
+; SSE-NEXT:    por %xmm13, %xmm12
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[2,1,1,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm1, %xmm12
+; SSE-NEXT:    pand %xmm10, %xmm12
+; SSE-NEXT:    por %xmm11, %xmm12
+; SSE-NEXT:    movdqa %xmm14, %xmm1
+; SSE-NEXT:    pandn %xmm12, %xmm1
+; SSE-NEXT:    andps %xmm14, %xmm9
+; SSE-NEXT:    por %xmm9, %xmm1
 ; SSE-NEXT:    por %xmm8, %xmm0
-; SSE-NEXT:    movaps %xmm3, %xmm1
-; SSE-NEXT:    movdqa %xmm3, %xmm8
-; SSE-NEXT:    pandn %xmm0, %xmm8
-; SSE-NEXT:    andps %xmm3, %xmm10
-; SSE-NEXT:    por %xmm10, %xmm8
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pxor %xmm11, %xmm11
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1],xmm0[2],xmm11[2],xmm0[3],xmm11[3],xmm0[4],xmm11[4],xmm0[5],xmm11[5],xmm0[6],xmm11[6],xmm0[7],xmm11[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,4,6,7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm11[8],xmm2[9],xmm11[9],xmm2[10],xmm11[10],xmm2[11],xmm11[11],xmm2[12],xmm11[12],xmm2[13],xmm11[13],xmm2[14],xmm11[14],xmm2[15],xmm11[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm2[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,5,7,6,7]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm10[4],xmm0[5],xmm10[5],xmm0[6],xmm10[6],xmm0[7],xmm10[7]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm11[8],xmm2[9],xmm11[9],xmm2[10],xmm11[10],xmm2[11],xmm11[11],xmm2[12],xmm11[12],xmm2[13],xmm11[13],xmm2[14],xmm11[14],xmm2[15],xmm11[15]
+; SSE-NEXT:    movdqa %xmm0, %xmm9
 ; SSE-NEXT:    pxor %xmm12, %xmm12
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm2, %xmm10
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    por %xmm10, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm3[0,2,1,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,4,7,6,5]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm2[3,3,3,3]
-; SSE-NEXT:    packuswb %xmm11, %xmm10
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movss {{.*#+}} xmm10 = xmm0[0],xmm10[1,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[1,3,2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm13[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1],xmm0[2],xmm12[2],xmm0[3],xmm12[3],xmm0[4],xmm12[4],xmm0[5],xmm12[5],xmm0[6],xmm12[6],xmm0[7],xmm12[7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm12[8],xmm11[9],xmm12[9],xmm11[10],xmm12[10],xmm11[11],xmm12[11],xmm11[12],xmm12[12],xmm11[13],xmm12[13],xmm11[14],xmm12[14],xmm11[15],xmm12[15]
-; SSE-NEXT:    pand %xmm9, %xmm11
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm12[0],xmm9[1],xmm12[1],xmm9[2],xmm12[2],xmm9[3],xmm12[3],xmm9[4],xmm12[4],xmm9[5],xmm12[5],xmm9[6],xmm12[6],xmm9[7],xmm12[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm9[0,1,2,3,6,4,6,7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm12[8],xmm0[9],xmm12[9],xmm0[10],xmm12[10],xmm0[11],xmm12[11],xmm0[12],xmm12[12],xmm0[13],xmm12[13],xmm0[14],xmm12[14],xmm0[15],xmm12[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm0[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,5,7,6,7]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm11 = xmm11[4],xmm9[4],xmm11[5],xmm9[5],xmm11[6],xmm9[6],xmm11[7],xmm9[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm12[8],xmm0[9],xmm12[9],xmm0[10],xmm12[10],xmm0[11],xmm12[11],xmm0[12],xmm12[12],xmm0[13],xmm12[13],xmm0[14],xmm12[14],xmm0[15],xmm12[15]
+; SSE-NEXT:    pxor %xmm13, %xmm13
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
 ; SSE-NEXT:    pandn %xmm0, %xmm9
-; SSE-NEXT:    por %xmm11, %xmm9
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,1,1,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm7, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm2[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    packuswb %xmm9, %xmm9
-; SSE-NEXT:    pandn %xmm9, %xmm7
-; SSE-NEXT:    por %xmm7, %xmm0
-; SSE-NEXT:    andps %xmm1, %xmm10
-; SSE-NEXT:    andnps %xmm0, %xmm1
-; SSE-NEXT:    orps %xmm10, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    por %xmm9, %xmm8
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm8[0,2,1,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,7,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,1,2,0]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm9[0,1,2,3,4,7,6,5]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm0[3,3,3,3]
+; SSE-NEXT:    packuswb %xmm12, %xmm9
+; SSE-NEXT:    packuswb %xmm11, %xmm11
+; SSE-NEXT:    movss {{.*#+}} xmm9 = xmm11[0],xmm9[1,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm2[1,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm4[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm11[0],xmm12[1],xmm11[1]
+; SSE-NEXT:    movdqa %xmm12, %xmm11
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm13[0],xmm11[1],xmm13[1],xmm11[2],xmm13[2],xmm11[3],xmm13[3],xmm11[4],xmm13[4],xmm11[5],xmm13[5],xmm11[6],xmm13[6],xmm11[7],xmm13[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm12 = xmm12[8],xmm13[8],xmm12[9],xmm13[9],xmm12[10],xmm13[10],xmm12[11],xmm13[11],xmm12[12],xmm13[12],xmm12[13],xmm13[13],xmm12[14],xmm13[14],xmm12[15],xmm13[15]
+; SSE-NEXT:    pand %xmm7, %xmm12
+; SSE-NEXT:    pandn %xmm11, %xmm7
+; SSE-NEXT:    por %xmm12, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm0[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    packuswb %xmm11, %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[2,1,1,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    packuswb %xmm7, %xmm7
+; SSE-NEXT:    pand %xmm10, %xmm7
+; SSE-NEXT:    pandn %xmm11, %xmm10
+; SSE-NEXT:    por %xmm7, %xmm10
+; SSE-NEXT:    movdqa %xmm14, %xmm0
+; SSE-NEXT:    andps %xmm14, %xmm9
+; SSE-NEXT:    pandn %xmm10, %xmm0
+; SSE-NEXT:    por %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm0, %xmm8
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, (%rsi)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 16(%rsi)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm7, 16(%rsi)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, (%rdx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -4663,19 +4677,21 @@ define void @load_i8_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movaps %xmm0, (%rcx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 16(%rcx)
-; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, (%r8)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r8)
+; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 16(%r8)
-; SSE-NEXT:    movdqa %xmm4, (%r9)
-; SSE-NEXT:    movdqa %xmm14, 16(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%r9)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; SSE-NEXT:    movdqa %xmm6, (%rax)
 ; SSE-NEXT:    movdqa %xmm5, 16(%rax)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movaps %xmm1, (%rax)
-; SSE-NEXT:    movdqa %xmm8, 16(%rax)
-; SSE-NEXT:    addq $648, %rsp # imm = 0x288
+; SSE-NEXT:    movdqa %xmm8, (%rax)
+; SSE-NEXT:    movdqa %xmm1, 16(%rax)
+; SSE-NEXT:    addq $632, %rsp # imm = 0x278
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i8_stride7_vf32:
@@ -7198,7 +7214,7 @@ define void @load_i8_stride7_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr %out.vec2, ptr %out.vec3, ptr %out.vec4, ptr %out.vec5, ptr %out.vec6) nounwind {
 ; SSE-LABEL: load_i8_stride7_vf64:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $1528, %rsp # imm = 0x5F8
+; SSE-NEXT:    subq $1512, %rsp # imm = 0x5E8
 ; SSE-NEXT:    movdqa 208(%rdi), %xmm12
 ; SSE-NEXT:    movdqa 192(%rdi), %xmm5
 ; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -7209,13 +7225,15 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movdqa 128(%rdi), %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 160(%rdi), %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 144(%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,0,65535,65535,65535,0,65535,65535]
 ; SSE-NEXT:    movdqa %xmm2, %xmm0
 ; SSE-NEXT:    pandn %xmm1, %xmm0
+; SSE-NEXT:    movdqa %xmm1, %xmm13
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm6, %xmm14
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm2, %xmm1
 ; SSE-NEXT:    movdqa %xmm2, %xmm7
 ; SSE-NEXT:    por %xmm0, %xmm1
@@ -7226,11 +7244,13 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm1[0,1,2,3,4,4,5,6]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,5,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,3,4,5,6,7]
-; SSE-NEXT:    packuswb %xmm0, %xmm2
+; SSE-NEXT:    packuswb %xmm0, %xmm1
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,255,0,0,0,0,0,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
 ; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,65535,0,65535,65535,65535,0,65535]
 ; SSE-NEXT:    movdqa %xmm11, %xmm1
 ; SSE-NEXT:    pandn %xmm3, %xmm1
@@ -7239,11 +7259,11 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    por %xmm1, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, %xmm1
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm6[8],xmm1[9],xmm6[9],xmm1[10],xmm6[10],xmm1[11],xmm6[11],xmm1[12],xmm6[12],xmm1[13],xmm6[13],xmm1[14],xmm6[14],xmm1[15],xmm6[15]
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,65535,65535,0,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm10, %xmm4
+; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,65535,65535,0,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm9, %xmm4
 ; SSE-NEXT:    pandn %xmm1, %xmm4
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1],xmm3[2],xmm6[2],xmm3[3],xmm6[3],xmm3[4],xmm6[4],xmm3[5],xmm6[5],xmm3[6],xmm6[6],xmm3[7],xmm6[7]
-; SSE-NEXT:    pand %xmm10, %xmm3
+; SSE-NEXT:    pand %xmm9, %xmm3
 ; SSE-NEXT:    por %xmm4, %xmm3
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
@@ -7251,14 +7271,12 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
 ; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    por %xmm3, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,65535,0,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,65535,0,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm10, %xmm2
 ; SSE-NEXT:    pandn %xmm8, %xmm2
 ; SSE-NEXT:    movdqa %xmm5, %xmm3
-; SSE-NEXT:    pand %xmm9, %xmm3
+; SSE-NEXT:    pand %xmm10, %xmm3
 ; SSE-NEXT:    por %xmm2, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, %xmm2
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3],xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
@@ -7294,7 +7312,7 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movdqa %xmm7, %xmm1
 ; SSE-NEXT:    pandn %xmm2, %xmm1
 ; SSE-NEXT:    movdqa 272(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm7, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
@@ -7319,10 +7337,10 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm6[8],xmm1[9],xmm6[9],xmm1[10],xmm6[10],xmm1[11],xmm6[11],xmm1[12],xmm6[12],xmm1[13],xmm6[13],xmm1[14],xmm6[14],xmm1[15],xmm6[15]
-; SSE-NEXT:    movdqa %xmm10, %xmm4
+; SSE-NEXT:    movdqa %xmm9, %xmm4
 ; SSE-NEXT:    pandn %xmm1, %xmm4
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3],xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
-; SSE-NEXT:    pand %xmm10, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm2
 ; SSE-NEXT:    por %xmm4, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
@@ -7333,11 +7351,11 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    por %xmm3, %xmm1
 ; SSE-NEXT:    movdqa 288(%rdi), %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    movdqa %xmm10, %xmm2
 ; SSE-NEXT:    pandn %xmm3, %xmm2
 ; SSE-NEXT:    movdqa 304(%rdi), %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm9, %xmm3
+; SSE-NEXT:    pand %xmm10, %xmm3
 ; SSE-NEXT:    por %xmm2, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, %xmm2
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3],xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
@@ -7396,10 +7414,10 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm6[8],xmm1[9],xmm6[9],xmm1[10],xmm6[10],xmm1[11],xmm6[11],xmm1[12],xmm6[12],xmm1[13],xmm6[13],xmm1[14],xmm6[14],xmm1[15],xmm6[15]
-; SSE-NEXT:    movdqa %xmm10, %xmm4
+; SSE-NEXT:    movdqa %xmm9, %xmm4
 ; SSE-NEXT:    pandn %xmm1, %xmm4
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3],xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
-; SSE-NEXT:    pand %xmm10, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm2
 ; SSE-NEXT:    por %xmm4, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
@@ -7410,12 +7428,11 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    por %xmm3, %xmm1
 ; SSE-NEXT:    movdqa 400(%rdi), %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    movdqa %xmm10, %xmm2
 ; SSE-NEXT:    pandn %xmm3, %xmm2
-; SSE-NEXT:    movdqa 416(%rdi), %xmm14
-; SSE-NEXT:    movdqa %xmm14, %xmm3
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm9, %xmm3
+; SSE-NEXT:    movdqa 416(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, (%rsp) # 16-byte Spill
+; SSE-NEXT:    pand %xmm10, %xmm3
 ; SSE-NEXT:    por %xmm2, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, %xmm2
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3],xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
@@ -7471,10 +7488,10 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    por %xmm1, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, %xmm1
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm6[8],xmm1[9],xmm6[9],xmm1[10],xmm6[10],xmm1[11],xmm6[11],xmm1[12],xmm6[12],xmm1[13],xmm6[13],xmm1[14],xmm6[14],xmm1[15],xmm6[15]
-; SSE-NEXT:    movdqa %xmm10, %xmm5
+; SSE-NEXT:    movdqa %xmm9, %xmm5
 ; SSE-NEXT:    pandn %xmm1, %xmm5
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm6[0],xmm4[1],xmm6[1],xmm4[2],xmm6[2],xmm4[3],xmm6[3],xmm4[4],xmm6[4],xmm4[5],xmm6[5],xmm4[6],xmm6[6],xmm4[7],xmm6[7]
-; SSE-NEXT:    pand %xmm10, %xmm4
+; SSE-NEXT:    pand %xmm9, %xmm4
 ; SSE-NEXT:    por %xmm5, %xmm4
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,2,1,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
@@ -7483,23 +7500,23 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
 ; SSE-NEXT:    pand %xmm0, %xmm1
 ; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    movdqa 64(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm0
-; SSE-NEXT:    pandn %xmm2, %xmm0
+; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    pandn %xmm2, %xmm1
 ; SSE-NEXT:    movdqa 80(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm9, %xmm2
-; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    pand %xmm10, %xmm2
+; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    pxor %xmm5, %xmm5
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,4,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3],xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,4,7]
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,7,7]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
 ; SSE-NEXT:    movdqa 96(%rdi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm3
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm5[8],xmm3[9],xmm5[9],xmm3[10],xmm5[10],xmm3[11],xmm5[11],xmm3[12],xmm5[12],xmm3[13],xmm5[13],xmm3[14],xmm5[14],xmm3[15],xmm5[15]
@@ -7513,19 +7530,18 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
 ; SSE-NEXT:    movdqa %xmm12, %xmm4
 ; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm4
-; SSE-NEXT:    pand %xmm8, %xmm1
+; SSE-NEXT:    packuswb %xmm1, %xmm1
+; SSE-NEXT:    pand %xmm12, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm4
+; SSE-NEXT:    pand %xmm8, %xmm0
 ; SSE-NEXT:    pandn %xmm4, %xmm8
-; SSE-NEXT:    por %xmm1, %xmm8
+; SSE-NEXT:    por %xmm0, %xmm8
 ; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pand %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm2, %xmm13
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,0,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    pandn %xmm13, %xmm0
+; SSE-NEXT:    movdqa %xmm14, %xmm1
+; SSE-NEXT:    pand %xmm6, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, %xmm2
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm7[8],xmm2[9],xmm7[9],xmm2[10],xmm7[10],xmm2[11],xmm7[11],xmm2[12],xmm7[12],xmm2[13],xmm7[13],xmm2[14],xmm7[14],xmm2[15],xmm7[15]
@@ -7539,29 +7555,30 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,6]
 ; SSE-NEXT:    psrld $16, %xmm2
 ; SSE-NEXT:    packuswb %xmm2, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,255,0,0,0,0,255,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    movdqa %xmm4, %xmm8
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,0,0,0,0,255,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm3, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    movdqa %xmm10, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pand %xmm9, %xmm4
+; SSE-NEXT:    pand %xmm10, %xmm4
+; SSE-NEXT:    movdqa %xmm10, %xmm8
 ; SSE-NEXT:    por %xmm1, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, %xmm1
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3],xmm1[4],xmm7[4],xmm1[5],xmm7[5],xmm1[6],xmm7[6],xmm1[7],xmm7[7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,0,65535,65535,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm6, %xmm5
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,0,65535,65535,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm10, %xmm5
 ; SSE-NEXT:    pandn %xmm1, %xmm5
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm7[8],xmm4[9],xmm7[9],xmm4[10],xmm7[10],xmm4[11],xmm7[11],xmm4[12],xmm7[12],xmm4[13],xmm7[13],xmm4[14],xmm7[14],xmm4[15],xmm7[15]
-; SSE-NEXT:    pand %xmm6, %xmm4
+; SSE-NEXT:    pand %xmm10, %xmm4
 ; SSE-NEXT:    por %xmm5, %xmm4
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm8, %xmm1
+; SSE-NEXT:    pand %xmm3, %xmm1
 ; SSE-NEXT:    por %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm11, %xmm13
 ; SSE-NEXT:    movdqa %xmm11, %xmm2
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
@@ -7569,10 +7586,10 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, %xmm2
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm7[8],xmm2[9],xmm7[9],xmm2[10],xmm7[10],xmm2[11],xmm7[11],xmm2[12],xmm7[12],xmm2[13],xmm7[13],xmm2[14],xmm7[14],xmm2[15],xmm7[15]
-; SSE-NEXT:    movdqa %xmm10, %xmm5
+; SSE-NEXT:    movdqa %xmm9, %xmm5
 ; SSE-NEXT:    pandn %xmm2, %xmm5
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1],xmm4[2],xmm7[2],xmm4[3],xmm7[3],xmm4[4],xmm7[4],xmm4[5],xmm7[5],xmm4[6],xmm7[6],xmm4[7],xmm7[7]
-; SSE-NEXT:    pand %xmm10, %xmm4
+; SSE-NEXT:    pand %xmm9, %xmm4
 ; SSE-NEXT:    por %xmm5, %xmm4
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    pslld $16, %xmm2
@@ -7587,17 +7604,16 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
 ; SSE-NEXT:    pand %xmm12, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm5
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0]
-; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0]
+; SSE-NEXT:    movdqa %xmm14, %xmm2
 ; SSE-NEXT:    pandn %xmm5, %xmm2
-; SSE-NEXT:    pand %xmm4, %xmm1
-; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    pand %xmm14, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    movdqa %xmm6, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa (%rsp), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pand %xmm13, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pand %xmm6, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm7[8],xmm1[9],xmm7[9],xmm1[10],xmm7[10],xmm1[11],xmm7[11],xmm1[12],xmm7[12],xmm1[13],xmm7[13],xmm1[14],xmm7[14],xmm1[15],xmm7[15]
@@ -7610,43 +7626,45 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,7,6]
 ; SSE-NEXT:    psrld $16, %xmm1
 ; SSE-NEXT:    packuswb %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm8, %xmm4
+; SSE-NEXT:    movdqa %xmm3, %xmm4
 ; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    movdqa %xmm8, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pand %xmm9, %xmm2
+; SSE-NEXT:    pand %xmm8, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3],xmm1[4],xmm7[4],xmm1[5],xmm7[5],xmm1[6],xmm7[6],xmm1[7],xmm7[7]
-; SSE-NEXT:    movdqa %xmm6, %xmm5
+; SSE-NEXT:    movdqa %xmm10, %xmm5
 ; SSE-NEXT:    pandn %xmm1, %xmm5
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm7[8],xmm2[9],xmm7[9],xmm2[10],xmm7[10],xmm2[11],xmm7[11],xmm2[12],xmm7[12],xmm2[13],xmm7[13],xmm2[14],xmm7[14],xmm2[15],xmm7[15]
-; SSE-NEXT:    pand %xmm6, %xmm2
+; SSE-NEXT:    pand %xmm10, %xmm2
 ; SSE-NEXT:    por %xmm5, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm8, %xmm1
+; SSE-NEXT:    pand %xmm3, %xmm1
+; SSE-NEXT:    movdqa %xmm3, %xmm11
 ; SSE-NEXT:    por %xmm4, %xmm1
-; SSE-NEXT:    movdqa %xmm11, %xmm2
+; SSE-NEXT:    movdqa %xmm13, %xmm2
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pand %xmm11, %xmm4
+; SSE-NEXT:    pand %xmm13, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, %xmm2
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm7[8],xmm2[9],xmm7[9],xmm2[10],xmm7[10],xmm2[11],xmm7[11],xmm2[12],xmm7[12],xmm2[13],xmm7[13],xmm2[14],xmm7[14],xmm2[15],xmm7[15]
-; SSE-NEXT:    movdqa %xmm10, %xmm5
+; SSE-NEXT:    movdqa %xmm9, %xmm5
 ; SSE-NEXT:    pandn %xmm2, %xmm5
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1],xmm4[2],xmm7[2],xmm4[3],xmm7[3],xmm4[4],xmm7[4],xmm4[5],xmm7[5],xmm4[6],xmm7[6],xmm4[7],xmm7[7]
-; SSE-NEXT:    pand %xmm10, %xmm4
+; SSE-NEXT:    pand %xmm9, %xmm4
 ; SSE-NEXT:    por %xmm5, %xmm4
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    pslld $16, %xmm2
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3]
 ; SSE-NEXT:    packuswb %xmm5, %xmm2
+; SSE-NEXT:    movdqa %xmm12, %xmm3
 ; SSE-NEXT:    movdqa %xmm12, %xmm5
 ; SSE-NEXT:    pandn %xmm2, %xmm5
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,3,2,3]
@@ -7655,15 +7673,17 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
 ; SSE-NEXT:    pand %xmm12, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm5
-; SSE-NEXT:    movdqa %xmm3, %xmm2
+; SSE-NEXT:    movdqa %xmm14, %xmm12
+; SSE-NEXT:    movdqa %xmm14, %xmm2
 ; SSE-NEXT:    pandn %xmm5, %xmm2
-; SSE-NEXT:    pand %xmm3, %xmm1
+; SSE-NEXT:    pand %xmm14, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    movdqa %xmm6, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pand %xmm13, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    pand %xmm6, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm7[8],xmm1[9],xmm7[9],xmm1[10],xmm7[10],xmm1[11],xmm7[11],xmm1[12],xmm7[12],xmm1[13],xmm7[13],xmm1[14],xmm7[14],xmm1[15],xmm7[15]
@@ -7676,68 +7696,67 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,7,6]
 ; SSE-NEXT:    psrld $16, %xmm1
 ; SSE-NEXT:    packuswb %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm8, %xmm4
+; SSE-NEXT:    movdqa %xmm11, %xmm4
 ; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    movdqa %xmm8, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pand %xmm9, %xmm2
+; SSE-NEXT:    pand %xmm8, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3],xmm1[4],xmm7[4],xmm1[5],xmm7[5],xmm1[6],xmm7[6],xmm1[7],xmm7[7]
-; SSE-NEXT:    movdqa %xmm6, %xmm5
+; SSE-NEXT:    movdqa %xmm10, %xmm5
 ; SSE-NEXT:    pandn %xmm1, %xmm5
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm7[8],xmm2[9],xmm7[9],xmm2[10],xmm7[10],xmm2[11],xmm7[11],xmm2[12],xmm7[12],xmm2[13],xmm7[13],xmm2[14],xmm7[14],xmm2[15],xmm7[15]
-; SSE-NEXT:    pand %xmm6, %xmm2
+; SSE-NEXT:    pand %xmm10, %xmm2
 ; SSE-NEXT:    por %xmm5, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,0,3,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm8, %xmm1
+; SSE-NEXT:    pand %xmm11, %xmm1
 ; SSE-NEXT:    por %xmm4, %xmm1
-; SSE-NEXT:    movdqa %xmm11, %xmm2
-; SSE-NEXT:    pandn %xmm14, %xmm2
+; SSE-NEXT:    movdqa %xmm13, %xmm2
+; SSE-NEXT:    pandn (%rsp), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pand %xmm11, %xmm4
+; SSE-NEXT:    pand %xmm13, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, %xmm2
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm7[8],xmm2[9],xmm7[9],xmm2[10],xmm7[10],xmm2[11],xmm7[11],xmm2[12],xmm7[12],xmm2[13],xmm7[13],xmm2[14],xmm7[14],xmm2[15],xmm7[15]
-; SSE-NEXT:    movdqa %xmm10, %xmm5
+; SSE-NEXT:    movdqa %xmm9, %xmm5
 ; SSE-NEXT:    pandn %xmm2, %xmm5
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1],xmm4[2],xmm7[2],xmm4[3],xmm7[3],xmm4[4],xmm7[4],xmm4[5],xmm7[5],xmm4[6],xmm7[6],xmm4[7],xmm7[7]
-; SSE-NEXT:    pand %xmm10, %xmm4
+; SSE-NEXT:    pand %xmm9, %xmm4
 ; SSE-NEXT:    por %xmm5, %xmm4
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    pslld $16, %xmm2
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3]
 ; SSE-NEXT:    packuswb %xmm5, %xmm2
-; SSE-NEXT:    movdqa %xmm12, %xmm5
+; SSE-NEXT:    movdqa %xmm3, %xmm5
 ; SSE-NEXT:    pandn %xmm2, %xmm5
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,0,3,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,4,6,7]
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pand %xmm12, %xmm2
+; SSE-NEXT:    pand %xmm3, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm5
-; SSE-NEXT:    movdqa %xmm3, %xmm2
+; SSE-NEXT:    movdqa %xmm12, %xmm2
 ; SSE-NEXT:    pandn %xmm5, %xmm2
-; SSE-NEXT:    pand %xmm3, %xmm1
+; SSE-NEXT:    pand %xmm12, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    movdqa %xmm6, %xmm1
 ; SSE-NEXT:    pandn %xmm15, %xmm1
-; SSE-NEXT:    movdqa %xmm15, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm15, %xmm2
-; SSE-NEXT:    pand %xmm13, %xmm2
+; SSE-NEXT:    movdqa %xmm15, %xmm12
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm5, %xmm2
+; SSE-NEXT:    pand %xmm6, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    pxor %xmm4, %xmm4
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm4[8],xmm1[9],xmm4[9],xmm1[10],xmm4[10],xmm1[11],xmm4[11],xmm1[12],xmm4[12],xmm1[13],xmm4[13],xmm1[14],xmm4[14],xmm1[15],xmm4[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3],xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7]
-; SSE-NEXT:    pxor %xmm5, %xmm5
+; SSE-NEXT:    pxor %xmm3, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm3[8],xmm1[9],xmm3[9],xmm1[10],xmm3[10],xmm1[11],xmm3[11],xmm1[12],xmm3[12],xmm1[13],xmm3[13],xmm1[14],xmm3[14],xmm1[15],xmm3[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3],xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
 ; SSE-NEXT:    pand %xmm0, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm0
 ; SSE-NEXT:    por %xmm2, %xmm0
@@ -7745,204 +7764,200 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,7,6]
 ; SSE-NEXT:    psrld $16, %xmm1
 ; SSE-NEXT:    packuswb %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    movdqa %xmm11, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm8, %xmm0
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pand %xmm9, %xmm2
-; SSE-NEXT:    movdqa %xmm9, %xmm12
+; SSE-NEXT:    pand %xmm8, %xmm2
 ; SSE-NEXT:    por %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3],xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
-; SSE-NEXT:    movdqa %xmm6, %xmm4
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
+; SSE-NEXT:    movdqa %xmm10, %xmm4
 ; SSE-NEXT:    pandn %xmm0, %xmm4
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm5[8],xmm2[9],xmm5[9],xmm2[10],xmm5[10],xmm2[11],xmm5[11],xmm2[12],xmm5[12],xmm2[13],xmm5[13],xmm2[14],xmm5[14],xmm2[15],xmm5[15]
-; SSE-NEXT:    pand %xmm6, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm3[8],xmm2[9],xmm3[9],xmm2[10],xmm3[10],xmm2[11],xmm3[11],xmm2[12],xmm3[12],xmm2[13],xmm3[13],xmm2[14],xmm3[14],xmm2[15],xmm3[15]
+; SSE-NEXT:    pand %xmm10, %xmm2
 ; SSE-NEXT:    por %xmm4, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[1,0,3,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm8, %xmm0
+; SSE-NEXT:    pand %xmm11, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm14, %xmm0
+; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm15, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pand %xmm11, %xmm1
+; SSE-NEXT:    pand %xmm13, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
 ; SSE-NEXT:    pxor %xmm2, %xmm2
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15]
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
-; SSE-NEXT:    pand %xmm10, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm10
-; SSE-NEXT:    por %xmm1, %xmm10
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    pand %xmm9, %xmm1
+; SSE-NEXT:    pandn %xmm0, %xmm9
+; SSE-NEXT:    por %xmm1, %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm13, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
 ; SSE-NEXT:    pandn %xmm7, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    pand %xmm11, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    movdqa (%rsp), %xmm8 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm8, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm5, %xmm1
-; SSE-NEXT:    pand %xmm11, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm9, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm4
-; SSE-NEXT:    pand %xmm11, %xmm4
-; SSE-NEXT:    por %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    movdqa %xmm15, %xmm6
-; SSE-NEXT:    pandn %xmm15, %xmm11
-; SSE-NEXT:    por %xmm0, %xmm11
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [65535,0,65535,65535,0,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm12, %xmm2
-; SSE-NEXT:    movdqa %xmm7, %xmm4
-; SSE-NEXT:    pandn %xmm7, %xmm2
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    pand %xmm13, %xmm2
+; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm8, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm2
+; SSE-NEXT:    pand %xmm13, %xmm2
+; SSE-NEXT:    por %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,2,2,3]
+; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    pandn %xmm14, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    pand %xmm13, %xmm2
+; SSE-NEXT:    por %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    pand %xmm13, %xmm0
+; SSE-NEXT:    movdqa %xmm5, %xmm11
+; SSE-NEXT:    pandn %xmm5, %xmm13
+; SSE-NEXT:    por %xmm0, %xmm13
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,0,65535,65535,0,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    pandn %xmm1, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm1, %xmm4
+; SSE-NEXT:    movdqa %xmm7, %xmm5
+; SSE-NEXT:    pandn %xmm7, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm7[0,2,2,3]
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm15, %xmm4
-; SSE-NEXT:    por %xmm0, %xmm4
+; SSE-NEXT:    pand %xmm6, %xmm5
+; SSE-NEXT:    por %xmm0, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm1, %xmm4
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm8, %xmm2
+; SSE-NEXT:    pandn %xmm8, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pandn %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm12, %xmm2
-; SSE-NEXT:    movdqa %xmm8, %xmm4
-; SSE-NEXT:    pandn %xmm8, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[0,2,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[0,2,2,3]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm6, %xmm2
+; SSE-NEXT:    por %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm15, %xmm4
-; SSE-NEXT:    por %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm4, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm12, %xmm2
-; SSE-NEXT:    movdqa %xmm12, %xmm1
-; SSE-NEXT:    movdqa %xmm9, %xmm4
-; SSE-NEXT:    pandn %xmm9, %xmm1
+; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    pandn %xmm10, %xmm0
+; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    pandn %xmm14, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,2,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[0,2,2,3]
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm15, %xmm4
-; SSE-NEXT:    por %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    pand %xmm15, %xmm9
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    pand %xmm15, %xmm12
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm15, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm6, %xmm14
+; SSE-NEXT:    por %xmm0, %xmm14
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pand %xmm6, %xmm7
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pand %xmm6, %xmm10
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    pand %xmm6, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    pand %xmm6, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    pand %xmm6, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pand %xmm15, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    pand %xmm6, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm4
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    pandn %xmm6, %xmm4
+; SSE-NEXT:    movdqa (%rsp), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm6, %xmm15
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm3, %xmm4
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    pandn %xmm11, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[0,2,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[0,2,2,3]
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm12, %xmm6
+; SSE-NEXT:    por %xmm0, %xmm6
 ; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm15, %xmm0
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pandn %xmm3, %xmm15
-; SSE-NEXT:    por %xmm0, %xmm15
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,0,65535,65535,65535,65535,65535,65535]
 ; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm7, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm5, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm10
-; SSE-NEXT:    movdqa %xmm7, %xmm4
+; SSE-NEXT:    movdqa %xmm5, %xmm6
+; SSE-NEXT:    movdqa %xmm5, %xmm8
 ; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm6, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm4, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, %xmm8
+; SSE-NEXT:    movdqa %xmm4, %xmm13
+; SSE-NEXT:    movdqa %xmm4, %xmm11
 ; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm5, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm5, %xmm1
-; SSE-NEXT:    movdqa %xmm5, %xmm11
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm1, %xmm9
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm3, %xmm13
-; SSE-NEXT:    pslld $16, %xmm13
-; SSE-NEXT:    psrldq {{.*#+}} xmm10 = xmm10[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT:    movdqa %xmm3, %xmm12
+; SSE-NEXT:    pslld $16, %xmm12
+; SSE-NEXT:    psrldq {{.*#+}} xmm6 = xmm6[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm14[0],xmm10[1],xmm14[1],xmm10[2],xmm14[2],xmm10[3],xmm14[3]
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    psrldq {{.*#+}} xmm0 = xmm0[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm14[0],xmm6[1],xmm14[1],xmm6[2],xmm14[2],xmm6[3],xmm14[3]
+; SSE-NEXT:    psrldq {{.*#+}} xmm13 = xmm13[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1],xmm0[2],xmm15[2],xmm0[3],xmm15[3]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    psrldq {{.*#+}} xmm1 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm15[0],xmm13[1],xmm15[1],xmm13[2],xmm15[2],xmm13[3],xmm15[3]
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    psrldq {{.*#+}} xmm2 = xmm2[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm14[4],xmm4[5],xmm14[5],xmm4[6],xmm14[6],xmm4[7],xmm14[7]
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm4
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm15[4],xmm8[5],xmm15[5],xmm8[6],xmm15[6],xmm8[7],xmm15[7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm2, %xmm13
+; SSE-NEXT:    psrldq {{.*#+}} xmm13 = xmm13[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm3[0],xmm13[1],xmm3[1],xmm13[2],xmm3[2],xmm13[3],xmm3[3]
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm14[4],xmm8[5],xmm14[5],xmm8[6],xmm14[6],xmm8[7],xmm14[7]
 ; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, %xmm8
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm11 = xmm11[4],xmm0[4],xmm11[5],xmm0[5],xmm11[6],xmm0[6],xmm11[7],xmm0[7]
+; SSE-NEXT:    movdqa %xmm14, %xmm8
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm11 = xmm11[4],xmm15[4],xmm11[5],xmm15[5],xmm11[6],xmm15[6],xmm11[7],xmm15[7]
 ; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm0, %xmm11
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm15, %xmm13
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm0[4],xmm9[5],xmm0[5],xmm9[6],xmm0[6],xmm9[7],xmm0[7]
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm0, %xmm9
+; SSE-NEXT:    movdqa %xmm2, %xmm0
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    movdqa %xmm3, %xmm11
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pxor %xmm0, %xmm0
@@ -7975,35 +7990,33 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm0[0,1,2,3,7,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm15[0,1,2,0]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm15[0,1,2,3,6,4,6,5]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm15[0,1,2,3,6,4,6,5]
+; SSE-NEXT:    pand %xmm14, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    pand %xmm14, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm3
-; SSE-NEXT:    pandn %xmm4, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm14, %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm3
-; SSE-NEXT:    pandn %xmm8, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm14, %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm3
-; SSE-NEXT:    movdqa %xmm11, %xmm6
-; SSE-NEXT:    pandn %xmm11, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm8, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pand %xmm14, %xmm5
 ; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm3
-; SSE-NEXT:    pand %xmm14, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm14, %xmm0
+; SSE-NEXT:    pandn %xmm13, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm14, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm14, %xmm0
+; SSE-NEXT:    pandn %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm14, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    pand %xmm14, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm11, %xmm4
+; SSE-NEXT:    pandn %xmm11, %xmm14
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pxor %xmm0, %xmm0
@@ -8011,37 +8024,37 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,0,65535,65535,65535,65535,65535,65535]
 ; SSE-NEXT:    pand %xmm11, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm11, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rsp), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pand %xmm11, %xmm8
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
 ; SSE-NEXT:    pand %xmm11, %xmm3
-; SSE-NEXT:    movdqa %xmm3, (%rsp) # 16-byte Spill
-; SSE-NEXT:    pand %xmm11, %xmm8
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm11, %xmm13
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3],xmm3[4],xmm0[4],xmm3[5],xmm0[5],xmm3[6],xmm0[6],xmm3[7],xmm0[7]
-; SSE-NEXT:    pxor %xmm8, %xmm8
+; SSE-NEXT:    pxor %xmm13, %xmm13
 ; SSE-NEXT:    pand %xmm11, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm11, %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm11, %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3],xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1],xmm0[2],xmm13[2],xmm0[3],xmm13[3],xmm0[4],xmm13[4],xmm0[5],xmm13[5],xmm0[6],xmm13[6],xmm0[7],xmm13[7]
 ; SSE-NEXT:    pand %xmm11, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm11, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm11, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pandn %xmm2, %xmm11
 ; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm13[0],xmm2[1],xmm13[1],xmm2[2],xmm13[2],xmm2[3],xmm13[3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm12[0],xmm2[1],xmm12[1],xmm2[2],xmm12[2],xmm2[3],xmm12[3]
 ; SSE-NEXT:    packuswb %xmm2, %xmm3
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,65535,65535,0]
 ; SSE-NEXT:    movdqa %xmm0, %xmm15
@@ -8052,14 +8065,14 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,4,6,7]
 ; SSE-NEXT:    packuswb %xmm3, %xmm3
 ; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm0, %xmm4
+; SSE-NEXT:    movdqa %xmm0, %xmm8
 ; SSE-NEXT:    por %xmm3, %xmm15
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0]
 ; SSE-NEXT:    movdqa %xmm0, %xmm3
 ; SSE-NEXT:    pandn %xmm15, %xmm3
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm0, %xmm13
+; SSE-NEXT:    movdqa %xmm0, %xmm5
 ; SSE-NEXT:    por %xmm2, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,0,65535,65535,0,65535]
@@ -8067,201 +8080,204 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
 ; SSE-NEXT:    pand %xmm0, %xmm15
-; SSE-NEXT:    movdqa %xmm0, %xmm5
+; SSE-NEXT:    movdqa %xmm0, %xmm9
 ; SSE-NEXT:    por %xmm3, %xmm15
 ; SSE-NEXT:    movdqa %xmm15, %xmm3
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3],xmm3[4],xmm8[4],xmm3[5],xmm8[5],xmm3[6],xmm8[6],xmm3[7],xmm8[7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [65535,0,65535,65535,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm14, %xmm0
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm13[0],xmm3[1],xmm13[1],xmm3[2],xmm13[2],xmm3[3],xmm13[3],xmm3[4],xmm13[4],xmm3[5],xmm13[5],xmm3[6],xmm13[6],xmm3[7],xmm13[7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,0,65535,65535,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm2, %xmm0
 ; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm15 = xmm15[8],xmm8[8],xmm15[9],xmm8[9],xmm15[10],xmm8[10],xmm15[11],xmm8[11],xmm15[12],xmm8[12],xmm15[13],xmm8[13],xmm15[14],xmm8[14],xmm15[15],xmm8[15]
-; SSE-NEXT:    pand %xmm14, %xmm15
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm15 = xmm15[8],xmm13[8],xmm15[9],xmm13[9],xmm15[10],xmm13[10],xmm15[11],xmm13[11],xmm15[12],xmm13[12],xmm15[13],xmm13[13],xmm15[14],xmm13[14],xmm15[15],xmm13[15]
+; SSE-NEXT:    pand %xmm2, %xmm15
+; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm15
-; SSE-NEXT:    packuswb %xmm10, %xmm0
-; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    packuswb %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm8, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm4, %xmm0
+; SSE-NEXT:    pand %xmm8, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm13, %xmm3
+; SSE-NEXT:    movdqa %xmm5, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    por %xmm0, %xmm9
-; SSE-NEXT:    movdqa %xmm9, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
-; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    por %xmm0, %xmm7
+; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm13[8],xmm0[9],xmm13[9],xmm0[10],xmm13[10],xmm0[11],xmm13[11],xmm0[12],xmm13[12],xmm0[13],xmm13[13],xmm0[14],xmm13[14],xmm0[15],xmm13[15]
+; SSE-NEXT:    movdqa %xmm1, %xmm6
+; SSE-NEXT:    movdqa %xmm1, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3],xmm9[4],xmm8[4],xmm9[5],xmm8[5],xmm9[6],xmm8[6],xmm9[7],xmm8[7]
-; SSE-NEXT:    pand %xmm14, %xmm9
-; SSE-NEXT:    por %xmm2, %xmm9
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,3,2,3]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm13[0],xmm7[1],xmm13[1],xmm7[2],xmm13[2],xmm7[3],xmm13[3],xmm7[4],xmm13[4],xmm7[5],xmm13[5],xmm7[6],xmm13[6],xmm7[7],xmm13[7]
+; SSE-NEXT:    pand %xmm1, %xmm7
+; SSE-NEXT:    por %xmm2, %xmm7
+; SSE-NEXT:    pshufd $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,3,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm8[8],xmm2[9],xmm8[9],xmm2[10],xmm8[10],xmm2[11],xmm8[11],xmm2[12],xmm8[12],xmm2[13],xmm8[13],xmm2[14],xmm8[14],xmm2[15],xmm8[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm13[8],xmm2[9],xmm13[9],xmm2[10],xmm13[10],xmm2[11],xmm13[11],xmm2[12],xmm13[12],xmm2[13],xmm13[13],xmm2[14],xmm13[14],xmm2[15],xmm13[15]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,0,65535,65535,65535]
 ; SSE-NEXT:    movdqa %xmm0, %xmm15
 ; SSE-NEXT:    pandn %xmm2, %xmm15
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1],xmm1[2],xmm13[2],xmm1[3],xmm13[3],xmm1[4],xmm13[4],xmm1[5],xmm13[5],xmm1[6],xmm13[6],xmm1[7],xmm13[7]
 ; SSE-NEXT:    pand %xmm0, %xmm1
 ; SSE-NEXT:    por %xmm15, %xmm1
 ; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm1[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,1,1]
 ; SSE-NEXT:    packuswb %xmm2, %xmm11
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,255,255,255,0,0,0,0,255,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm6, %xmm2
-; SSE-NEXT:    pandn %xmm11, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,2,1,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,0,0,0,0,255,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm2, %xmm4
+; SSE-NEXT:    pandn %xmm11, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[0,2,1,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,3,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,1,0,3,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm6, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    pand %xmm13, %xmm2
-; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm5, %xmm15
-; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    pand %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm2, %xmm12
+; SSE-NEXT:    por %xmm1, %xmm4
+; SSE-NEXT:    pand %xmm5, %xmm4
+; SSE-NEXT:    por %xmm3, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm15
+; SSE-NEXT:    movdqa %xmm9, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pand %xmm5, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
-; SSE-NEXT:    movdqa %xmm14, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1],xmm1[2],xmm13[2],xmm1[3],xmm13[3],xmm1[4],xmm13[4],xmm1[5],xmm13[5],xmm1[6],xmm13[6],xmm1[7],xmm13[7]
+; SSE-NEXT:    movdqa %xmm6, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm8[8],xmm2[9],xmm8[9],xmm2[10],xmm8[10],xmm2[11],xmm8[11],xmm2[12],xmm8[12],xmm2[13],xmm8[13],xmm2[14],xmm8[14],xmm2[15],xmm8[15]
-; SSE-NEXT:    pand %xmm14, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm13[8],xmm2[9],xmm13[9],xmm2[10],xmm13[10],xmm2[11],xmm13[11],xmm2[12],xmm13[12],xmm2[13],xmm13[13],xmm2[14],xmm13[14],xmm2[15],xmm13[15]
+; SSE-NEXT:    pand %xmm6, %xmm2
 ; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    packuswb {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    movdqa %xmm8, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm3
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm4, %xmm1
-; SSE-NEXT:    movdqa %xmm4, %xmm10
+; SSE-NEXT:    pand %xmm8, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    movdqa %xmm5, %xmm9
 ; SSE-NEXT:    pandn %xmm3, %xmm1
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    por %xmm2, %xmm12
-; SSE-NEXT:    movdqa %xmm12, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm8[8],xmm2[9],xmm8[9],xmm2[10],xmm8[10],xmm2[11],xmm8[11],xmm2[12],xmm8[12],xmm2[13],xmm8[13],xmm2[14],xmm8[14],xmm2[15],xmm8[15]
-; SSE-NEXT:    movdqa %xmm14, %xmm3
+; SSE-NEXT:    por %xmm2, %xmm10
+; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm13[8],xmm2[9],xmm13[9],xmm2[10],xmm13[10],xmm2[11],xmm13[11],xmm2[12],xmm13[12],xmm2[13],xmm13[13],xmm2[14],xmm13[14],xmm2[15],xmm13[15]
+; SSE-NEXT:    movdqa %xmm6, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm8[0],xmm12[1],xmm8[1],xmm12[2],xmm8[2],xmm12[3],xmm8[3],xmm12[4],xmm8[4],xmm12[5],xmm8[5],xmm12[6],xmm8[6],xmm12[7],xmm8[7]
-; SSE-NEXT:    pand %xmm14, %xmm12
-; SSE-NEXT:    por %xmm3, %xmm12
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm13[0],xmm10[1],xmm13[1],xmm10[2],xmm13[2],xmm10[3],xmm13[3],xmm10[4],xmm13[4],xmm10[5],xmm13[5],xmm10[6],xmm13[6],xmm10[7],xmm13[7]
+; SSE-NEXT:    pand %xmm6, %xmm10
+; SSE-NEXT:    por %xmm3, %xmm10
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm11[1,3,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
 ; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm8[8],xmm2[9],xmm8[9],xmm2[10],xmm8[10],xmm2[11],xmm8[11],xmm2[12],xmm8[12],xmm2[13],xmm8[13],xmm2[14],xmm8[14],xmm2[15],xmm8[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm13[8],xmm2[9],xmm13[9],xmm2[10],xmm13[10],xmm2[11],xmm13[11],xmm2[12],xmm13[12],xmm2[13],xmm13[13],xmm2[14],xmm13[14],xmm2[15],xmm13[15]
 ; SSE-NEXT:    movdqa %xmm0, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1],xmm4[2],xmm8[2],xmm4[3],xmm8[3],xmm4[4],xmm8[4],xmm4[5],xmm8[5],xmm4[6],xmm8[6],xmm4[7],xmm8[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm13[0],xmm4[1],xmm13[1],xmm4[2],xmm13[2],xmm4[3],xmm13[3],xmm4[4],xmm13[4],xmm4[5],xmm13[5],xmm4[6],xmm13[6],xmm4[7],xmm13[7]
 ; SSE-NEXT:    pand %xmm0, %xmm4
 ; SSE-NEXT:    por %xmm3, %xmm4
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,1,1]
 ; SSE-NEXT:    packuswb %xmm2, %xmm3
-; SSE-NEXT:    movdqa %xmm6, %xmm4
+; SSE-NEXT:    movdqa %xmm12, %xmm4
 ; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm12[0,2,1,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm10[0,2,1,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,3,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,0,3,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pand %xmm6, %xmm2
+; SSE-NEXT:    pand %xmm12, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm4
-; SSE-NEXT:    pand %xmm13, %xmm4
+; SSE-NEXT:    pand %xmm5, %xmm4
 ; SSE-NEXT:    por %xmm1, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm5, %xmm1
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    pandn (%rsp), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pand %xmm5, %xmm2
+; SSE-NEXT:    pand %xmm15, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
-; SSE-NEXT:    movdqa %xmm14, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1],xmm1[2],xmm13[2],xmm1[3],xmm13[3],xmm1[4],xmm13[4],xmm1[5],xmm13[5],xmm1[6],xmm13[6],xmm1[7],xmm13[7]
+; SSE-NEXT:    movdqa %xmm6, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm8[8],xmm2[9],xmm8[9],xmm2[10],xmm8[10],xmm2[11],xmm8[11],xmm2[12],xmm8[12],xmm2[13],xmm8[13],xmm2[14],xmm8[14],xmm2[15],xmm8[15]
-; SSE-NEXT:    pand %xmm14, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm13[8],xmm2[9],xmm13[9],xmm2[10],xmm13[10],xmm2[11],xmm13[11],xmm2[12],xmm13[12],xmm2[13],xmm13[13],xmm2[14],xmm13[14],xmm2[15],xmm13[15]
+; SSE-NEXT:    pand %xmm6, %xmm2
 ; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    packuswb {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm10, %xmm3
+; SSE-NEXT:    movdqa %xmm8, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm3
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,3,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm10, %xmm1
+; SSE-NEXT:    pand %xmm8, %xmm1
+; SSE-NEXT:    movdqa %xmm8, %xmm10
 ; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    movdqa %xmm5, %xmm1
 ; SSE-NEXT:    pandn %xmm3, %xmm1
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm8[8],xmm2[9],xmm8[9],xmm2[10],xmm8[10],xmm2[11],xmm8[11],xmm2[12],xmm8[12],xmm2[13],xmm8[13],xmm2[14],xmm8[14],xmm2[15],xmm8[15]
-; SSE-NEXT:    movdqa %xmm14, %xmm3
-; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1],xmm4[2],xmm8[2],xmm4[3],xmm8[3],xmm4[4],xmm8[4],xmm4[5],xmm8[5],xmm4[6],xmm8[6],xmm4[7],xmm8[7]
-; SSE-NEXT:    pand %xmm14, %xmm4
-; SSE-NEXT:    por %xmm3, %xmm4
-; SSE-NEXT:    movdqa %xmm4, %xmm5
-; SSE-NEXT:    pshufd $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[1,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm8[8],xmm2[9],xmm8[9],xmm2[10],xmm8[10],xmm2[11],xmm8[11],xmm2[12],xmm8[12],xmm2[13],xmm8[13],xmm2[14],xmm8[14],xmm2[15],xmm8[15]
-; SSE-NEXT:    movdqa %xmm0, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm13[8],xmm2[9],xmm13[9],xmm2[10],xmm13[10],xmm2[11],xmm13[11],xmm2[12],xmm13[12],xmm2[13],xmm13[13],xmm2[14],xmm13[14],xmm2[15],xmm13[15]
+; SSE-NEXT:    movdqa %xmm6, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1],xmm4[2],xmm8[2],xmm4[3],xmm8[3],xmm4[4],xmm8[4],xmm4[5],xmm8[5],xmm4[6],xmm8[6],xmm4[7],xmm8[7]
-; SSE-NEXT:    pand %xmm0, %xmm4
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm13[0],xmm4[1],xmm13[1],xmm4[2],xmm13[2],xmm4[3],xmm13[3],xmm4[4],xmm13[4],xmm4[5],xmm13[5],xmm4[6],xmm13[6],xmm4[7],xmm13[7]
+; SSE-NEXT:    pand %xmm6, %xmm4
 ; SSE-NEXT:    por %xmm3, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,1,2,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[1,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; SSE-NEXT:    movdqa %xmm5, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm13[8],xmm2[9],xmm13[9],xmm2[10],xmm13[10],xmm2[11],xmm13[11],xmm2[12],xmm13[12],xmm2[13],xmm13[13],xmm2[14],xmm13[14],xmm2[15],xmm13[15]
+; SSE-NEXT:    movdqa %xmm0, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm13[0],xmm5[1],xmm13[1],xmm5[2],xmm13[2],xmm5[3],xmm13[3],xmm5[4],xmm13[4],xmm5[5],xmm13[5],xmm5[6],xmm13[6],xmm5[7],xmm13[7]
+; SSE-NEXT:    pand %xmm0, %xmm5
+; SSE-NEXT:    por %xmm3, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm5[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,1,1]
 ; SSE-NEXT:    packuswb %xmm2, %xmm3
-; SSE-NEXT:    movdqa %xmm6, %xmm4
-; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[0,2,1,3]
+; SSE-NEXT:    movdqa %xmm12, %xmm5
+; SSE-NEXT:    pandn %xmm3, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,2,1,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,3,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,1,0,3,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pand %xmm6, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm4
-; SSE-NEXT:    pand %xmm13, %xmm4
-; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm12, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm5
+; SSE-NEXT:    pand %xmm9, %xmm5
+; SSE-NEXT:    por %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm15, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    pand %xmm15, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
-; SSE-NEXT:    movdqa %xmm14, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1],xmm1[2],xmm13[2],xmm1[3],xmm13[3],xmm1[4],xmm13[4],xmm1[5],xmm13[5],xmm1[6],xmm13[6],xmm1[7],xmm13[7]
+; SSE-NEXT:    movdqa %xmm6, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm8[8],xmm2[9],xmm8[9],xmm2[10],xmm8[10],xmm2[11],xmm8[11],xmm2[12],xmm8[12],xmm2[13],xmm8[13],xmm2[14],xmm8[14],xmm2[15],xmm8[15]
-; SSE-NEXT:    pand %xmm14, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm13[8],xmm2[9],xmm13[9],xmm2[10],xmm13[10],xmm2[11],xmm13[11],xmm2[12],xmm13[12],xmm2[13],xmm13[13],xmm2[14],xmm13[14],xmm2[15],xmm13[15]
+; SSE-NEXT:    pand %xmm6, %xmm2
 ; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    packuswb {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm10, %xmm3
@@ -8272,26 +8288,27 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
 ; SSE-NEXT:    pand %xmm10, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    movdqa %xmm9, %xmm7
 ; SSE-NEXT:    pandn %xmm3, %xmm1
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    por %xmm2, %xmm4
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm8[8],xmm2[9],xmm8[9],xmm2[10],xmm8[10],xmm2[11],xmm8[11],xmm2[12],xmm8[12],xmm2[13],xmm8[13],xmm2[14],xmm8[14],xmm2[15],xmm8[15]
-; SSE-NEXT:    movdqa %xmm14, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    por %xmm2, %xmm5
+; SSE-NEXT:    movdqa %xmm5, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm13[8],xmm2[9],xmm13[9],xmm2[10],xmm13[10],xmm2[11],xmm13[11],xmm2[12],xmm13[12],xmm2[13],xmm13[13],xmm2[14],xmm13[14],xmm2[15],xmm13[15]
+; SSE-NEXT:    movdqa %xmm6, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1],xmm4[2],xmm8[2],xmm4[3],xmm8[3],xmm4[4],xmm8[4],xmm4[5],xmm8[5],xmm4[6],xmm8[6],xmm4[7],xmm8[7]
-; SSE-NEXT:    pand %xmm14, %xmm4
-; SSE-NEXT:    por %xmm3, %xmm4
-; SSE-NEXT:    pshufd $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[1,3,2,3]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm13[0],xmm5[1],xmm13[1],xmm5[2],xmm13[2],xmm5[3],xmm13[3],xmm5[4],xmm13[4],xmm5[5],xmm13[5],xmm5[6],xmm13[6],xmm5[7],xmm13[7]
+; SSE-NEXT:    pand %xmm6, %xmm5
+; SSE-NEXT:    por %xmm3, %xmm5
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm9[1,3,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
 ; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm8[8],xmm2[9],xmm8[9],xmm2[10],xmm8[10],xmm2[11],xmm8[11],xmm2[12],xmm8[12],xmm2[13],xmm8[13],xmm2[14],xmm8[14],xmm2[15],xmm8[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3],xmm3[4],xmm8[4],xmm3[5],xmm8[5],xmm3[6],xmm8[6],xmm3[7],xmm8[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm13[8],xmm2[9],xmm13[9],xmm2[10],xmm13[10],xmm2[11],xmm13[11],xmm2[12],xmm13[12],xmm2[13],xmm13[13],xmm2[14],xmm13[14],xmm2[15],xmm13[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm13[0],xmm3[1],xmm13[1],xmm3[2],xmm13[2],xmm3[3],xmm13[3],xmm3[4],xmm13[4],xmm3[5],xmm13[5],xmm3[6],xmm13[6],xmm3[7],xmm13[7]
 ; SSE-NEXT:    pand %xmm0, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm0
 ; SSE-NEXT:    por %xmm3, %xmm0
@@ -8299,128 +8316,128 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[1,1,1,1]
 ; SSE-NEXT:    packuswb %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm6, %xmm2
+; SSE-NEXT:    movdqa %xmm12, %xmm4
+; SSE-NEXT:    movdqa %xmm12, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[0,2,1,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[0,2,1,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,3,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    pand %xmm12, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    pand %xmm13, %xmm2
+; SSE-NEXT:    pand %xmm7, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, %xmm9
 ; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pandn %xmm7, %xmm0
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    pand %xmm15, %xmm1
+; SSE-NEXT:    movdqa %xmm15, %xmm12
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm13[8],xmm0[9],xmm13[9],xmm0[10],xmm13[10],xmm0[11],xmm13[11],xmm0[12],xmm13[12],xmm0[13],xmm13[13],xmm0[14],xmm13[14],xmm0[15],xmm13[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1],xmm1[2],xmm13[2],xmm1[3],xmm13[3],xmm1[4],xmm13[4],xmm1[5],xmm13[5],xmm1[6],xmm13[6],xmm1[7],xmm13[7]
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,6]
 ; SSE-NEXT:    psrlq $48, %xmm0
 ; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm4, %xmm0
 ; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,0,65535,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,0,65535,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm3, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pand %xmm12, %xmm2
+; SSE-NEXT:    pand %xmm3, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1],xmm1[2],xmm13[2],xmm1[3],xmm13[3],xmm1[4],xmm13[4],xmm1[5],xmm13[5],xmm1[6],xmm13[6],xmm1[7],xmm13[7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,0,65535,0,65535,65535,65535,65535]
 ; SSE-NEXT:    movdqa %xmm5, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm8[8],xmm2[9],xmm8[9],xmm2[10],xmm8[10],xmm2[11],xmm8[11],xmm2[12],xmm8[12],xmm2[13],xmm8[13],xmm2[14],xmm8[14],xmm2[15],xmm8[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm13[8],xmm2[9],xmm13[9],xmm2[10],xmm13[10],xmm2[11],xmm13[11],xmm2[12],xmm13[12],xmm2[13],xmm13[13],xmm2[14],xmm13[14],xmm2[15],xmm13[15]
 ; SSE-NEXT:    pand %xmm5, %xmm2
 ; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[3,2,1,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    pand %xmm4, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    por %xmm0, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
-; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm13[8],xmm0[9],xmm13[9],xmm0[10],xmm13[10],xmm0[11],xmm13[11],xmm0[12],xmm13[12],xmm0[13],xmm13[13],xmm0[14],xmm13[14],xmm0[15],xmm13[15]
+; SSE-NEXT:    movdqa %xmm6, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3],xmm3[4],xmm8[4],xmm3[5],xmm8[5],xmm3[6],xmm8[6],xmm3[7],xmm8[7]
-; SSE-NEXT:    pand %xmm14, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm13[0],xmm3[1],xmm13[1],xmm3[2],xmm13[2],xmm3[3],xmm13[3],xmm3[4],xmm13[4],xmm3[5],xmm13[5],xmm3[6],xmm13[6],xmm3[7],xmm13[7]
+; SSE-NEXT:    pand %xmm6, %xmm3
 ; SSE-NEXT:    por %xmm2, %xmm3
 ; SSE-NEXT:    pshuflw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[2,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm10, %xmm7
-; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,65535,65535,65535,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm7, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm3[0,2,1,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    pand %xmm7, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm13, %xmm10
-; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [255,255,255,255,255,255,255,255,255,0,0,0,0,0,0,0]
+; SSE-NEXT:    movdqa %xmm15, %xmm0
 ; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    pand %xmm13, %xmm1
+; SSE-NEXT:    pand %xmm15, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    movdqa %xmm12, %xmm0
 ; SSE-NEXT:    pandn %xmm11, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pand %xmm15, %xmm1
+; SSE-NEXT:    pand %xmm12, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm13[8],xmm0[9],xmm13[9],xmm0[10],xmm13[10],xmm0[11],xmm13[11],xmm0[12],xmm13[12],xmm0[13],xmm13[13],xmm0[14],xmm13[14],xmm0[15],xmm13[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1],xmm1[2],xmm13[2],xmm1[3],xmm13[3],xmm1[4],xmm13[4],xmm1[5],xmm13[5],xmm1[6],xmm13[6],xmm1[7],xmm13[7]
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,6]
 ; SSE-NEXT:    psrlq $48, %xmm0
 ; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm4, %xmm0
 ; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm12, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm15, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    pand %xmm12, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,0,65535,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pand %xmm10, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1],xmm1[2],xmm13[2],xmm1[3],xmm13[3],xmm1[4],xmm13[4],xmm1[5],xmm13[5],xmm1[6],xmm13[6],xmm1[7],xmm13[7]
 ; SSE-NEXT:    movdqa %xmm5, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm8[8],xmm2[9],xmm8[9],xmm2[10],xmm8[10],xmm2[11],xmm8[11],xmm2[12],xmm8[12],xmm2[13],xmm8[13],xmm2[14],xmm8[14],xmm2[15],xmm8[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm13[8],xmm2[9],xmm13[9],xmm2[10],xmm13[10],xmm2[11],xmm13[11],xmm2[12],xmm13[12],xmm2[13],xmm13[13],xmm2[14],xmm13[14],xmm2[15],xmm13[15]
 ; SSE-NEXT:    pand %xmm5, %xmm2
 ; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[3,2,1,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    pand %xmm4, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    por %xmm0, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
-; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm13[8],xmm0[9],xmm13[9],xmm0[10],xmm13[10],xmm0[11],xmm13[11],xmm0[12],xmm13[12],xmm0[13],xmm13[13],xmm0[14],xmm13[14],xmm0[15],xmm13[15]
+; SSE-NEXT:    movdqa %xmm6, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3],xmm3[4],xmm8[4],xmm3[5],xmm8[5],xmm3[6],xmm8[6],xmm3[7],xmm8[7]
-; SSE-NEXT:    pand %xmm14, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm13[0],xmm3[1],xmm13[1],xmm3[2],xmm13[2],xmm3[3],xmm13[3],xmm3[4],xmm13[4],xmm3[5],xmm13[5],xmm3[6],xmm13[6],xmm3[7],xmm13[7]
+; SSE-NEXT:    pand %xmm6, %xmm3
 ; SSE-NEXT:    por %xmm2, %xmm3
 ; SSE-NEXT:    pshuflw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[2,1,2,3,4,5,6,7]
@@ -8433,57 +8450,60 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
 ; SSE-NEXT:    pand %xmm7, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    movdqa %xmm15, %xmm0
 ; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    pand %xmm13, %xmm1
+; SSE-NEXT:    pand %xmm15, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm0
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    pandn %xmm8, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pand %xmm9, %xmm1
-; SSE-NEXT:    movdqa %xmm9, %xmm13
+; SSE-NEXT:    pand %xmm12, %xmm1
+; SSE-NEXT:    movdqa %xmm12, %xmm8
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm13[8],xmm0[9],xmm13[9],xmm0[10],xmm13[10],xmm0[11],xmm13[11],xmm0[12],xmm13[12],xmm0[13],xmm13[13],xmm0[14],xmm13[14],xmm0[15],xmm13[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1],xmm1[2],xmm13[2],xmm1[3],xmm13[3],xmm1[4],xmm13[4],xmm1[5],xmm13[5],xmm1[6],xmm13[6],xmm1[7],xmm13[7]
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,6]
 ; SSE-NEXT:    psrlq $48, %xmm0
 ; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm4, %xmm12
+; SSE-NEXT:    movdqa %xmm4, %xmm0
 ; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm12, %xmm1
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm9, %xmm2
-; SSE-NEXT:    pand %xmm12, %xmm2
+; SSE-NEXT:    movdqa %xmm10, %xmm4
+; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm10, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm11, %xmm2
+; SSE-NEXT:    pand %xmm4, %xmm2
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3],xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1],xmm1[2],xmm13[2],xmm1[3],xmm13[3],xmm1[4],xmm13[4],xmm1[5],xmm13[5],xmm1[6],xmm13[6],xmm1[7],xmm13[7]
 ; SSE-NEXT:    movdqa %xmm5, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm8[8],xmm2[9],xmm8[9],xmm2[10],xmm8[10],xmm2[11],xmm8[11],xmm2[12],xmm8[12],xmm2[13],xmm8[13],xmm2[14],xmm8[14],xmm2[15],xmm8[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm13[8],xmm2[9],xmm13[9],xmm2[10],xmm13[10],xmm2[11],xmm13[11],xmm2[12],xmm13[12],xmm2[13],xmm13[13],xmm2[14],xmm13[14],xmm2[15],xmm13[15]
 ; SSE-NEXT:    pand %xmm5, %xmm2
 ; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[3,2,1,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    pand %xmm12, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm11, %xmm0
+; SSE-NEXT:    pandn %xmm12, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    por %xmm0, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
-; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm13[8],xmm0[9],xmm13[9],xmm0[10],xmm13[10],xmm0[11],xmm13[11],xmm0[12],xmm13[12],xmm0[13],xmm13[13],xmm0[14],xmm13[14],xmm0[15],xmm13[15]
+; SSE-NEXT:    movdqa %xmm6, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3],xmm3[4],xmm8[4],xmm3[5],xmm8[5],xmm3[6],xmm8[6],xmm3[7],xmm8[7]
-; SSE-NEXT:    pand %xmm14, %xmm3
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm13[0],xmm3[1],xmm13[1],xmm3[2],xmm13[2],xmm3[3],xmm13[3],xmm3[4],xmm13[4],xmm3[5],xmm13[5],xmm3[6],xmm13[6],xmm3[7],xmm13[7]
+; SSE-NEXT:    pand %xmm6, %xmm3
 ; SSE-NEXT:    por %xmm2, %xmm3
 ; SSE-NEXT:    pshuflw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[2,1,2,3,4,5,6,7]
@@ -8496,75 +8516,130 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
 ; SSE-NEXT:    pand %xmm7, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    movdqa %xmm15, %xmm0
 ; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    pand %xmm10, %xmm1
+; SSE-NEXT:    pand %xmm15, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm8, %xmm0
+; SSE-NEXT:    pandn %xmm9, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pand %xmm13, %xmm2
+; SSE-NEXT:    pand %xmm8, %xmm2
 ; SSE-NEXT:    por %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm8[8],xmm0[9],xmm8[9],xmm0[10],xmm8[10],xmm0[11],xmm8[11],xmm0[12],xmm8[12],xmm0[13],xmm8[13],xmm0[14],xmm8[14],xmm0[15],xmm8[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3],xmm2[4],xmm8[4],xmm2[5],xmm8[5],xmm2[6],xmm8[6],xmm2[7],xmm8[7]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm13[8],xmm0[9],xmm13[9],xmm0[10],xmm13[10],xmm0[11],xmm13[11],xmm0[12],xmm13[12],xmm0[13],xmm13[13],xmm0[14],xmm13[14],xmm0[15],xmm13[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm13[0],xmm2[1],xmm13[1],xmm2[2],xmm13[2],xmm2[3],xmm13[3],xmm2[4],xmm13[4],xmm2[5],xmm13[5],xmm2[6],xmm13[6],xmm2[7],xmm13[7]
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm2[0,3,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,0]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,7,6]
 ; SSE-NEXT:    psrlq $48, %xmm0
 ; SSE-NEXT:    packuswb %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm13, %xmm0
+; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm8, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pand %xmm12, %xmm2
+; SSE-NEXT:    pand %xmm4, %xmm2
 ; SSE-NEXT:    por %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3],xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1],xmm0[2],xmm13[2],xmm0[3],xmm13[3],xmm0[4],xmm13[4],xmm0[5],xmm13[5],xmm0[6],xmm13[6],xmm0[7],xmm13[7]
 ; SSE-NEXT:    movdqa %xmm5, %xmm3
 ; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm8[8],xmm2[9],xmm8[9],xmm2[10],xmm8[10],xmm2[11],xmm8[11],xmm2[12],xmm8[12],xmm2[13],xmm8[13],xmm2[14],xmm8[14],xmm2[15],xmm8[15]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm13[8],xmm2[9],xmm13[9],xmm2[10],xmm13[10],xmm2[11],xmm13[11],xmm2[12],xmm13[12],xmm2[13],xmm13[13],xmm2[14],xmm13[14],xmm2[15],xmm13[15]
 ; SSE-NEXT:    pand %xmm5, %xmm2
 ; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[3,2,1,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    pand %xmm6, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,0,0,0,0,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    por %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm3, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm13[8],xmm0[9],xmm13[9],xmm0[10],xmm13[10],xmm0[11],xmm13[11],xmm0[12],xmm13[12],xmm0[13],xmm13[13],xmm0[14],xmm13[14],xmm0[15],xmm13[15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm13[0],xmm3[1],xmm13[1],xmm3[2],xmm13[2],xmm3[3],xmm13[3],xmm3[4],xmm13[4],xmm3[5],xmm13[5],xmm3[6],xmm13[6],xmm3[7],xmm13[7]
+; SSE-NEXT:    pand %xmm6, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm6
+; SSE-NEXT:    por %xmm3, %xmm6
+; SSE-NEXT:    pshuflw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[2,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; SSE-NEXT:    packuswb %xmm0, %xmm0
+; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    pandn %xmm0, %xmm1
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm6[0,2,1,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,6,6,7]
+; SSE-NEXT:    packuswb %xmm0, %xmm0
+; SSE-NEXT:    pand %xmm7, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm1
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    pandn %xmm1, %xmm0
+; SSE-NEXT:    pand %xmm15, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,0,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    pand %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm2, %xmm6
+; SSE-NEXT:    por %xmm0, %xmm1
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm13[8],xmm0[9],xmm13[9],xmm0[10],xmm13[10],xmm0[11],xmm13[11],xmm0[12],xmm13[12],xmm0[13],xmm13[13],xmm0[14],xmm13[14],xmm0[15],xmm13[15]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1],xmm1[2],xmm13[2],xmm1[3],xmm13[3],xmm1[4],xmm13[4],xmm1[5],xmm13[5],xmm1[6],xmm13[6],xmm1[7],xmm13[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm13[8],xmm0[9],xmm13[9],xmm0[10],xmm13[10],xmm0[11],xmm13[11],xmm0[12],xmm13[12],xmm0[13],xmm13[13],xmm0[14],xmm13[14],xmm0[15],xmm13[15]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm8[8],xmm1[9],xmm8[9],xmm1[10],xmm8[10],xmm1[11],xmm8[11],xmm1[12],xmm8[12],xmm1[13],xmm8[13],xmm1[14],xmm8[14],xmm1[15],xmm8[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3],xmm3[4],xmm8[4],xmm3[5],xmm8[5],xmm3[6],xmm8[6],xmm3[7],xmm8[7]
-; SSE-NEXT:    pand %xmm14, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm14
-; SSE-NEXT:    por %xmm3, %xmm14
-; SSE-NEXT:    pshuflw $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    por %xmm3, %xmm2
+; SSE-NEXT:    packuswb %xmm0, %xmm2
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm14[0,2,1,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,1,3,3]
+; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
+; SSE-NEXT:    movdqa %xmm4, %xmm1
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    pand %xmm4, %xmm2
+; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1],xmm1[2],xmm13[2],xmm1[3],xmm13[3],xmm1[4],xmm13[4],xmm1[5],xmm13[5],xmm1[6],xmm13[6],xmm1[7],xmm13[7]
+; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm13[8],xmm2[9],xmm13[9],xmm2[10],xmm13[10],xmm2[11],xmm13[11],xmm2[12],xmm13[12],xmm2[13],xmm13[13],xmm2[14],xmm13[14],xmm2[15],xmm13[15]
+; SSE-NEXT:    pand %xmm5, %xmm2
+; SSE-NEXT:    por %xmm3, %xmm2
+; SSE-NEXT:    pshufd $100, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    packuswb %xmm1, %xmm1
+; SSE-NEXT:    movdqa %xmm7, %xmm3
+; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
 ; SSE-NEXT:    pand %xmm7, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm10, %xmm1
-; SSE-NEXT:    pandn %xmm2, %xmm1
-; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    pandn %xmm3, %xmm1
+; SSE-NEXT:    andps %xmm15, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm8, %xmm0
+; SSE-NEXT:    movdqa %xmm6, %xmm13
+; SSE-NEXT:    movdqa %xmm6, %xmm0
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pand %xmm8, %xmm1
+; SSE-NEXT:    pand %xmm6, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
 ; SSE-NEXT:    pxor %xmm2, %xmm2
@@ -8585,115 +8660,64 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,1,3,3]
 ; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
-; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,0,65535,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm4, %xmm1
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm14, %xmm2
-; SSE-NEXT:    pand %xmm12, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    pand %xmm4, %xmm3
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm3, %xmm1
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3],xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
-; SSE-NEXT:    movdqa %xmm5, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm6[8],xmm2[9],xmm6[9],xmm2[10],xmm6[10],xmm2[11],xmm6[11],xmm2[12],xmm6[12],xmm2[13],xmm6[13],xmm2[14],xmm6[14],xmm2[15],xmm6[15]
-; SSE-NEXT:    pand %xmm5, %xmm2
-; SSE-NEXT:    por %xmm3, %xmm2
+; SSE-NEXT:    movdqa %xmm5, %xmm4
+; SSE-NEXT:    pandn %xmm1, %xmm4
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm6[8],xmm3[9],xmm6[9],xmm3[10],xmm6[10],xmm3[11],xmm6[11],xmm3[12],xmm6[12],xmm3[13],xmm6[13],xmm3[14],xmm6[14],xmm3[15],xmm6[15]
+; SSE-NEXT:    pand %xmm5, %xmm3
+; SSE-NEXT:    por %xmm4, %xmm3
 ; SSE-NEXT:    pshufd $100, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,1,0,3]
+; SSE-NEXT:    movdqa %xmm7, %xmm4
+; SSE-NEXT:    pandn %xmm1, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,1,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
 ; SSE-NEXT:    pand %xmm7, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm10, %xmm1
-; SSE-NEXT:    pandn %xmm3, %xmm1
-; SSE-NEXT:    andps %xmm10, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    pandn %xmm15, %xmm0
-; SSE-NEXT:    pand %xmm8, %xmm4
-; SSE-NEXT:    por %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3],xmm4[4],xmm1[4],xmm4[5],xmm1[5],xmm4[6],xmm1[6],xmm4[7],xmm1[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm0, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    packuswb %xmm0, %xmm3
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[0,1,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm2[0],xmm0[1,2,3]
-; SSE-NEXT:    movdqa %xmm12, %xmm2
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm15, %xmm3
-; SSE-NEXT:    pand %xmm12, %xmm3
-; SSE-NEXT:    por %xmm2, %xmm3
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
-; SSE-NEXT:    movdqa %xmm5, %xmm4
-; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm1[8],xmm3[9],xmm1[9],xmm3[10],xmm1[10],xmm3[11],xmm1[11],xmm3[12],xmm1[12],xmm3[13],xmm1[13],xmm3[14],xmm1[14],xmm3[15],xmm1[15]
-; SSE-NEXT:    pand %xmm5, %xmm3
-; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    pshufd $100, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[0,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,4,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    movdqa %xmm7, %xmm4
-; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,3,2,1,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,6,7]
-; SSE-NEXT:    packuswb %xmm2, %xmm2
-; SSE-NEXT:    pand %xmm7, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm4
-; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm4
+; SSE-NEXT:    movdqa %xmm15, %xmm1
 ; SSE-NEXT:    pandn %xmm4, %xmm1
-; SSE-NEXT:    andps %xmm10, %xmm0
+; SSE-NEXT:    andps %xmm15, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    pand %xmm8, %xmm9
-; SSE-NEXT:    por %xmm0, %xmm9
-; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    pandn %xmm10, %xmm0
+; SSE-NEXT:    pand %xmm13, %xmm11
+; SSE-NEXT:    por %xmm0, %xmm11
+; SSE-NEXT:    movdqa %xmm11, %xmm0
 ; SSE-NEXT:    pxor %xmm1, %xmm1
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1],xmm9[2],xmm1[2],xmm9[3],xmm1[3],xmm9[4],xmm1[4],xmm9[5],xmm1[5],xmm9[6],xmm1[6],xmm9[7],xmm1[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[2,1,2,3]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm1[0],xmm11[1],xmm1[1],xmm11[2],xmm1[2],xmm11[3],xmm1[3],xmm11[4],xmm1[4],xmm11[5],xmm1[5],xmm11[6],xmm1[6],xmm11[7],xmm1[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,2,2,3,4,5,6,7]
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm0, %xmm6
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm0, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    por %xmm4, %xmm2
-; SSE-NEXT:    packuswb %xmm0, %xmm2
+; SSE-NEXT:    por %xmm6, %xmm4
+; SSE-NEXT:    packuswb %xmm0, %xmm4
 ; SSE-NEXT:    packuswb %xmm3, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,1,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[0,1,3,3]
 ; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm3[0],xmm0[1,2,3]
-; SSE-NEXT:    movdqa %xmm12, %xmm3
-; SSE-NEXT:    pandn %xmm11, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm2, %xmm4
-; SSE-NEXT:    pand %xmm12, %xmm4
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,0,65535,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm10, %xmm3
+; SSE-NEXT:    pandn %xmm12, %xmm3
+; SSE-NEXT:    movdqa (%rsp), %xmm4 # 16-byte Reload
+; SSE-NEXT:    pand %xmm10, %xmm4
 ; SSE-NEXT:    por %xmm3, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, %xmm3
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3],xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
@@ -8714,15 +8738,15 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    packuswb %xmm3, %xmm3
 ; SSE-NEXT:    pand %xmm7, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm6
-; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    movdqa %xmm15, %xmm1
 ; SSE-NEXT:    pandn %xmm6, %xmm1
-; SSE-NEXT:    andps %xmm10, %xmm0
+; SSE-NEXT:    andps %xmm15, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    pandn %xmm13, %xmm0
+; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    pandn %xmm8, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pand %xmm8, %xmm4
+; SSE-NEXT:    pand %xmm13, %xmm4
 ; SSE-NEXT:    por %xmm0, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, %xmm0
 ; SSE-NEXT:    pxor %xmm1, %xmm1
@@ -8743,8 +8767,8 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[0,1,3,3]
 ; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm4[0],xmm0[1,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm12, %xmm3
-; SSE-NEXT:    pand %xmm12, %xmm4
+; SSE-NEXT:    movdqa %xmm10, %xmm3
+; SSE-NEXT:    pand %xmm10, %xmm4
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    por %xmm4, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, %xmm4
@@ -8765,144 +8789,143 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    packuswb %xmm5, %xmm5
 ; SSE-NEXT:    pandn %xmm5, %xmm7
 ; SSE-NEXT:    por %xmm4, %xmm7
-; SSE-NEXT:    movdqa %xmm10, %xmm3
+; SSE-NEXT:    movdqa %xmm15, %xmm3
 ; SSE-NEXT:    pandn %xmm7, %xmm3
-; SSE-NEXT:    andps %xmm10, %xmm0
+; SSE-NEXT:    andps %xmm15, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [65535,65535,65535,0,65535,65535,0,65535]
-; SSE-NEXT:    pand %xmm13, %xmm4
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,65535,0,65535,65535,0,65535]
+; SSE-NEXT:    pand %xmm10, %xmm4
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm4, %xmm6
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm12[8],xmm6[9],xmm12[9],xmm6[10],xmm12[10],xmm6[11],xmm12[11],xmm6[12],xmm12[12],xmm6[13],xmm12[13],xmm6[14],xmm12[14],xmm6[15],xmm12[15]
+; SSE-NEXT:    movdqa %xmm4, %xmm7
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm12[8],xmm7[9],xmm12[9],xmm7[10],xmm12[10],xmm7[11],xmm12[11],xmm7[12],xmm12[12],xmm7[13],xmm12[13],xmm7[14],xmm12[14],xmm7[15],xmm12[15]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [0,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm0, %xmm7
-; SSE-NEXT:    pandn %xmm6, %xmm7
+; SSE-NEXT:    movdqa %xmm0, %xmm6
+; SSE-NEXT:    pandn %xmm7, %xmm6
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm12[0],xmm4[1],xmm12[1],xmm4[2],xmm12[2],xmm4[3],xmm12[3],xmm4[4],xmm12[4],xmm4[5],xmm12[5],xmm4[6],xmm12[6],xmm4[7],xmm12[7]
 ; SSE-NEXT:    pand %xmm0, %xmm4
-; SSE-NEXT:    por %xmm7, %xmm4
+; SSE-NEXT:    por %xmm6, %xmm4
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm4[0,1,2,3,5,4,7,6]
-; SSE-NEXT:    psrldq {{.*#+}} xmm6 = xmm6[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    packuswb %xmm6, %xmm7
-; SSE-NEXT:    movdqa %xmm13, %xmm3
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm4[0,1,2,3,5,4,7,6]
+; SSE-NEXT:    psrldq {{.*#+}} xmm7 = xmm7[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    packuswb %xmm7, %xmm6
+; SSE-NEXT:    movdqa %xmm10, %xmm3
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    pandn %xmm1, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[1,3,2,3]
-; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm6 = mem[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
-; SSE-NEXT:    movdqa %xmm6, %xmm4
+; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = mem[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
+; SSE-NEXT:    movdqa %xmm7, %xmm4
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm12[8],xmm4[9],xmm12[9],xmm4[10],xmm12[10],xmm4[11],xmm12[11],xmm4[12],xmm12[12],xmm4[13],xmm12[13],xmm4[14],xmm12[14],xmm4[15],xmm12[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1],xmm6[2],xmm12[2],xmm6[3],xmm12[3],xmm6[4],xmm12[4],xmm6[5],xmm12[5],xmm6[6],xmm12[6],xmm6[7],xmm12[7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1],xmm6[2],xmm4[2],xmm6[3],xmm4[3]
-; SSE-NEXT:    packuswb %xmm6, %xmm6
-; SSE-NEXT:    movss {{.*#+}} xmm7 = xmm6[0],xmm7[1,2,3]
-; SSE-NEXT:    movdqa %xmm8, %xmm1
-; SSE-NEXT:    movdqa %xmm8, %xmm4
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm12[0],xmm7[1],xmm12[1],xmm7[2],xmm12[2],xmm7[3],xmm12[3],xmm7[4],xmm12[4],xmm7[5],xmm12[5],xmm7[6],xmm12[6],xmm7[7],xmm12[7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1],xmm7[2],xmm4[2],xmm7[3],xmm4[3]
+; SSE-NEXT:    packuswb %xmm7, %xmm7
+; SSE-NEXT:    movss {{.*#+}} xmm6 = xmm7[0],xmm6[1,2,3]
+; SSE-NEXT:    movdqa %xmm13, %xmm11
+; SSE-NEXT:    movdqa %xmm13, %xmm4
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm14, %xmm8
-; SSE-NEXT:    pand %xmm1, %xmm8
-; SSE-NEXT:    movdqa %xmm1, %xmm14
+; SSE-NEXT:    movdqa %xmm9, %xmm8
+; SSE-NEXT:    pand %xmm13, %xmm8
 ; SSE-NEXT:    por %xmm4, %xmm8
 ; SSE-NEXT:    movdqa %xmm8, %xmm4
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm12[0],xmm4[1],xmm12[1],xmm4[2],xmm12[2],xmm4[3],xmm12[3],xmm4[4],xmm12[4],xmm4[5],xmm12[5],xmm4[6],xmm12[6],xmm4[7],xmm12[7]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,0,65535,0,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm1, %xmm6
-; SSE-NEXT:    pandn %xmm4, %xmm6
+; SSE-NEXT:    movdqa %xmm1, %xmm7
+; SSE-NEXT:    pandn %xmm4, %xmm7
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm12[8],xmm8[9],xmm12[9],xmm8[10],xmm12[10],xmm8[11],xmm12[11],xmm8[12],xmm12[12],xmm8[13],xmm12[13],xmm8[14],xmm12[14],xmm8[15],xmm12[15]
 ; SSE-NEXT:    pand %xmm1, %xmm8
-; SSE-NEXT:    por %xmm6, %xmm8
+; SSE-NEXT:    por %xmm7, %xmm8
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,7,6]
 ; SSE-NEXT:    packuswb %xmm4, %xmm4
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0]
-; SSE-NEXT:    movdqa %xmm6, %xmm9
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,255,255,255,255,255,255,255,255,255,255,255,0,0,0]
+; SSE-NEXT:    movdqa %xmm7, %xmm9
 ; SSE-NEXT:    pandn %xmm4, %xmm9
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm8[2,1,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,0,3,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    packuswb %xmm4, %xmm4
-; SSE-NEXT:    pand %xmm6, %xmm4
+; SSE-NEXT:    pand %xmm7, %xmm4
 ; SSE-NEXT:    por %xmm4, %xmm9
-; SSE-NEXT:    movdqa %xmm10, %xmm3
+; SSE-NEXT:    movdqa %xmm15, %xmm3
 ; SSE-NEXT:    pandn %xmm9, %xmm3
-; SSE-NEXT:    andps %xmm10, %xmm7
-; SSE-NEXT:    movdqa %xmm10, %xmm5
-; SSE-NEXT:    por %xmm7, %xmm3
+; SSE-NEXT:    andps %xmm15, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm13, %xmm10
-; SSE-NEXT:    pand %xmm13, %xmm7
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm7, %xmm8
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm10, %xmm4
+; SSE-NEXT:    pand %xmm10, %xmm6
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm6, %xmm8
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm12[8],xmm8[9],xmm12[9],xmm8[10],xmm12[10],xmm8[11],xmm12[11],xmm8[12],xmm12[12],xmm8[13],xmm12[13],xmm8[14],xmm12[14],xmm8[15],xmm12[15]
 ; SSE-NEXT:    movdqa %xmm0, %xmm9
 ; SSE-NEXT:    pandn %xmm8, %xmm9
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm12[0],xmm7[1],xmm12[1],xmm7[2],xmm12[2],xmm7[3],xmm12[3],xmm7[4],xmm12[4],xmm7[5],xmm12[5],xmm7[6],xmm12[6],xmm7[7],xmm12[7]
-; SSE-NEXT:    pand %xmm0, %xmm7
-; SSE-NEXT:    por %xmm9, %xmm7
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm7[0,1,2,3,5,4,7,6]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1],xmm6[2],xmm12[2],xmm6[3],xmm12[3],xmm6[4],xmm12[4],xmm6[5],xmm12[5],xmm6[6],xmm12[6],xmm6[7],xmm12[7]
+; SSE-NEXT:    pand %xmm0, %xmm6
+; SSE-NEXT:    por %xmm9, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm6[0,1,2,3,5,4,7,6]
 ; SSE-NEXT:    psrldq {{.*#+}} xmm8 = xmm8[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; SSE-NEXT:    packuswb %xmm8, %xmm9
-; SSE-NEXT:    movdqa %xmm13, %xmm4
+; SSE-NEXT:    movdqa %xmm10, %xmm6
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm3[1,3,2,3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm4[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
-; SSE-NEXT:    movdqa %xmm8, %xmm7
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm7 = xmm7[8],xmm12[8],xmm7[9],xmm12[9],xmm7[10],xmm12[10],xmm7[11],xmm12[11],xmm7[12],xmm12[12],xmm7[13],xmm12[13],xmm7[14],xmm12[14],xmm7[15],xmm12[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    pandn %xmm3, %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm3[1,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm10[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm6[0],xmm8[1],xmm6[1]
+; SSE-NEXT:    movdqa %xmm8, %xmm6
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm6 = xmm6[8],xmm12[8],xmm6[9],xmm12[9],xmm6[10],xmm12[10],xmm6[11],xmm12[11],xmm6[12],xmm12[12],xmm6[13],xmm12[13],xmm6[14],xmm12[14],xmm6[15],xmm12[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,2,2,3,4,5,6,7]
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm12[0],xmm8[1],xmm12[1],xmm8[2],xmm12[2],xmm8[3],xmm12[3],xmm8[4],xmm12[4],xmm8[5],xmm12[5],xmm8[6],xmm12[6],xmm8[7],xmm12[7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm7[0],xmm8[1],xmm7[1],xmm8[2],xmm7[2],xmm8[3],xmm7[3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm6[0],xmm8[1],xmm6[1],xmm8[2],xmm6[2],xmm8[3],xmm6[3]
 ; SSE-NEXT:    packuswb %xmm8, %xmm8
 ; SSE-NEXT:    movss {{.*#+}} xmm9 = xmm8[0],xmm9[1,2,3]
-; SSE-NEXT:    movdqa %xmm14, %xmm7
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm15, %xmm8
-; SSE-NEXT:    pand %xmm14, %xmm8
-; SSE-NEXT:    por %xmm7, %xmm8
-; SSE-NEXT:    movdqa %xmm8, %xmm7
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm12[0],xmm7[1],xmm12[1],xmm7[2],xmm12[2],xmm7[3],xmm12[3],xmm7[4],xmm12[4],xmm7[5],xmm12[5],xmm7[6],xmm12[6],xmm7[7],xmm12[7]
+; SSE-NEXT:    movdqa %xmm13, %xmm6
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm2, %xmm8
+; SSE-NEXT:    pand %xmm13, %xmm8
+; SSE-NEXT:    por %xmm6, %xmm8
+; SSE-NEXT:    movdqa %xmm8, %xmm6
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm12[0],xmm6[1],xmm12[1],xmm6[2],xmm12[2],xmm6[3],xmm12[3],xmm6[4],xmm12[4],xmm6[5],xmm12[5],xmm6[6],xmm12[6],xmm6[7],xmm12[7]
 ; SSE-NEXT:    movdqa %xmm1, %xmm13
-; SSE-NEXT:    pandn %xmm7, %xmm13
+; SSE-NEXT:    pandn %xmm6, %xmm13
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm12[8],xmm8[9],xmm12[9],xmm8[10],xmm12[10],xmm8[11],xmm12[11],xmm8[12],xmm12[12],xmm8[13],xmm12[13],xmm8[14],xmm12[14],xmm8[15],xmm12[15]
 ; SSE-NEXT:    pand %xmm1, %xmm8
 ; SSE-NEXT:    por %xmm13, %xmm8
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm3[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,4,7,6]
-; SSE-NEXT:    packuswb %xmm7, %xmm7
-; SSE-NEXT:    movdqa %xmm6, %xmm13
-; SSE-NEXT:    pandn %xmm7, %xmm13
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm8[2,1,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,0,3,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    packuswb %xmm7, %xmm7
-; SSE-NEXT:    pand %xmm6, %xmm7
-; SSE-NEXT:    por %xmm7, %xmm13
-; SSE-NEXT:    movdqa %xmm5, %xmm7
-; SSE-NEXT:    pandn %xmm13, %xmm7
-; SSE-NEXT:    andps %xmm5, %xmm9
-; SSE-NEXT:    por %xmm9, %xmm7
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm2[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,7,6]
+; SSE-NEXT:    packuswb %xmm6, %xmm6
+; SSE-NEXT:    movdqa %xmm7, %xmm13
+; SSE-NEXT:    pandn %xmm6, %xmm13
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm8[2,1,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[0,0,3,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    packuswb %xmm6, %xmm6
+; SSE-NEXT:    pand %xmm7, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm13
+; SSE-NEXT:    movdqa %xmm15, %xmm6
+; SSE-NEXT:    pandn %xmm13, %xmm6
+; SSE-NEXT:    andps %xmm15, %xmm9
+; SSE-NEXT:    movdqa %xmm15, %xmm5
+; SSE-NEXT:    por %xmm9, %xmm6
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    pand %xmm10, %xmm8
+; SSE-NEXT:    pand %xmm4, %xmm8
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm8, %xmm9
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm12[8],xmm9[9],xmm12[9],xmm9[10],xmm12[10],xmm9[11],xmm12[11],xmm9[12],xmm12[12],xmm9[13],xmm12[13],xmm9[14],xmm12[14],xmm9[15],xmm12[15]
@@ -8915,10 +8938,10 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm8[0,1,2,3,5,4,7,6]
 ; SSE-NEXT:    psrldq {{.*#+}} xmm9 = xmm9[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; SSE-NEXT:    packuswb %xmm9, %xmm15
-; SSE-NEXT:    movdqa %xmm10, %xmm13
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pandn %xmm3, %xmm13
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm3[1,3,2,3]
+; SSE-NEXT:    movdqa %xmm4, %xmm13
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pandn %xmm2, %xmm13
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm2[1,3,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm3[0,2,2,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
@@ -8932,10 +8955,10 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm8[0],xmm9[1],xmm8[1],xmm9[2],xmm8[2],xmm9[3],xmm8[3]
 ; SSE-NEXT:    packuswb %xmm9, %xmm9
 ; SSE-NEXT:    movss {{.*#+}} xmm15 = xmm9[0],xmm15[1,2,3]
-; SSE-NEXT:    movdqa %xmm14, %xmm8
-; SSE-NEXT:    pandn %xmm11, %xmm8
-; SSE-NEXT:    movdqa %xmm2, %xmm9
-; SSE-NEXT:    pand %xmm14, %xmm9
+; SSE-NEXT:    movdqa %xmm11, %xmm8
+; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa (%rsp), %xmm9 # 16-byte Reload
+; SSE-NEXT:    pand %xmm11, %xmm9
 ; SSE-NEXT:    por %xmm8, %xmm9
 ; SSE-NEXT:    movdqa %xmm9, %xmm8
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm12[0],xmm8[1],xmm12[1],xmm8[2],xmm12[2],xmm8[3],xmm12[3],xmm8[4],xmm12[4],xmm8[5],xmm12[5],xmm8[6],xmm12[6],xmm8[7],xmm12[7]
@@ -8949,20 +8972,20 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm2[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,7,6]
 ; SSE-NEXT:    packuswb %xmm8, %xmm8
-; SSE-NEXT:    movdqa %xmm6, %xmm11
+; SSE-NEXT:    movdqa %xmm7, %xmm11
 ; SSE-NEXT:    pandn %xmm8, %xmm11
 ; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm9[2,1,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[0,0,3,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    packuswb %xmm8, %xmm8
-; SSE-NEXT:    pand %xmm6, %xmm8
+; SSE-NEXT:    pand %xmm7, %xmm8
 ; SSE-NEXT:    por %xmm8, %xmm11
 ; SSE-NEXT:    movdqa %xmm5, %xmm9
 ; SSE-NEXT:    pandn %xmm11, %xmm9
 ; SSE-NEXT:    andps %xmm5, %xmm15
 ; SSE-NEXT:    por %xmm15, %xmm9
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    pand %xmm10, %xmm8
+; SSE-NEXT:    pand %xmm4, %xmm8
 ; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
 ; SSE-NEXT:    movdqa %xmm8, %xmm11
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8],xmm12[8],xmm11[9],xmm12[9],xmm11[10],xmm12[10],xmm11[11],xmm12[11],xmm11[12],xmm12[12],xmm11[13],xmm12[13],xmm11[14],xmm12[14],xmm11[15],xmm12[15]
@@ -8974,15 +8997,16 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,6]
 ; SSE-NEXT:    psrldq {{.*#+}} xmm11 = xmm11[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; SSE-NEXT:    packuswb %xmm11, %xmm0
-; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    movdqa %xmm4, %xmm2
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    pand %xmm10, %xmm15
-; SSE-NEXT:    pand %xmm10, %xmm4
-; SSE-NEXT:    pand %xmm10, %xmm3
+; SSE-NEXT:    pand %xmm4, %xmm15
+; SSE-NEXT:    movdqa %xmm10, %xmm4
+; SSE-NEXT:    pand %xmm2, %xmm4
+; SSE-NEXT:    pand %xmm2, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm3[0,2,2,3]
-; SSE-NEXT:    pand %xmm10, %xmm3
+; SSE-NEXT:    pand %xmm2, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    pandn %xmm3, %xmm2
@@ -8999,8 +9023,8 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    packuswb %xmm8, %xmm8
 ; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm8[0],xmm0[1,2,3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm14, %xmm3
-; SSE-NEXT:    pand %xmm14, %xmm8
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,0,65535,65535,0,65535,65535]
+; SSE-NEXT:    pand %xmm3, %xmm8
 ; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    por %xmm8, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, %xmm8
@@ -9009,18 +9033,17 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pand %xmm1, %xmm3
 ; SSE-NEXT:    pandn %xmm8, %xmm1
 ; SSE-NEXT:    por %xmm3, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm3[0,1,0,3]
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm14[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,7,6]
 ; SSE-NEXT:    packuswb %xmm8, %xmm8
-; SSE-NEXT:    movdqa %xmm6, %xmm14
+; SSE-NEXT:    movdqa %xmm7, %xmm14
 ; SSE-NEXT:    pandn %xmm8, %xmm14
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,0,3,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    pand %xmm7, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm14
 ; SSE-NEXT:    movdqa %xmm5, %xmm11
 ; SSE-NEXT:    pandn %xmm14, %xmm11
@@ -9056,7 +9079,7 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm6, %xmm8
+; SSE-NEXT:    movdqa %xmm7, %xmm8
 ; SSE-NEXT:    pandn %xmm0, %xmm8
 ; SSE-NEXT:    pshufd $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[1,3,2,3]
@@ -9074,26 +9097,27 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,1,1,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,3,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm0, %xmm14
-; SSE-NEXT:    pand %xmm6, %xmm14
+; SSE-NEXT:    pand %xmm7, %xmm14
 ; SSE-NEXT:    por %xmm8, %xmm14
 ; SSE-NEXT:    movdqa %xmm5, %xmm3
 ; SSE-NEXT:    pandn %xmm14, %xmm3
 ; SSE-NEXT:    andps %xmm5, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm4, %xmm1
+; SSE-NEXT:    movdqa %xmm4, %xmm8
+; SSE-NEXT:    por {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    movdqa %xmm8, %xmm1
 ; SSE-NEXT:    pxor %xmm0, %xmm0
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,4,6,7]
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm0[8],xmm4[9],xmm0[9],xmm4[10],xmm0[10],xmm4[11],xmm0[11],xmm4[12],xmm0[12],xmm4[13],xmm0[13],xmm4[14],xmm0[14],xmm4[15],xmm0[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm4[0,1,2,1]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm0[8],xmm8[9],xmm0[9],xmm8[10],xmm0[10],xmm8[11],xmm0[11],xmm8[12],xmm0[12],xmm8[13],xmm0[13],xmm8[14],xmm0[14],xmm8[15],xmm0[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,5,7,6,7]
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm0[8],xmm4[9],xmm0[9],xmm4[10],xmm0[10],xmm4[11],xmm0[11],xmm4[12],xmm0[12],xmm4[13],xmm0[13],xmm4[14],xmm0[14],xmm4[15],xmm0[15]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
 ; SSE-NEXT:    pandn %xmm4, %xmm12
-; SSE-NEXT:    movdqa (%rsp), %xmm8 # 16-byte Reload
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; SSE-NEXT:    por %xmm12, %xmm8
 ; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,2,1,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,7,6,7]
@@ -9108,7 +9132,7 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm6, %xmm12
+; SSE-NEXT:    movdqa %xmm7, %xmm12
 ; SSE-NEXT:    pandn %xmm1, %xmm12
 ; SSE-NEXT:    pshufd $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[1,3,2,3]
@@ -9125,7 +9149,7 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[2,1,1,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    pand %xmm7, %xmm1
 ; SSE-NEXT:    por %xmm12, %xmm1
 ; SSE-NEXT:    movdqa %xmm5, %xmm12
 ; SSE-NEXT:    pandn %xmm1, %xmm12
@@ -9162,11 +9186,11 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    movdqa %xmm6, %xmm13
+; SSE-NEXT:    movdqa %xmm7, %xmm13
 ; SSE-NEXT:    pandn %xmm1, %xmm13
 ; SSE-NEXT:    pshufd $237, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[1,3,2,3]
-; SSE-NEXT:    pshufd $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    pshufd $232, (%rsp), %xmm14 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm14 = mem[0,2,2,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
 ; SSE-NEXT:    movdqa %xmm14, %xmm1
@@ -9179,7 +9203,7 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[2,1,1,1]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    pand %xmm7, %xmm1
 ; SSE-NEXT:    por %xmm13, %xmm1
 ; SSE-NEXT:    movdqa %xmm4, %xmm0
 ; SSE-NEXT:    movdqa %xmm4, %xmm13
@@ -9223,17 +9247,17 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    pand %xmm10, %xmm8
 ; SSE-NEXT:    pandn %xmm1, %xmm10
 ; SSE-NEXT:    por %xmm8, %xmm10
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[2,1,1,1]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,0,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,1,0,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,4,7]
 ; SSE-NEXT:    packuswb %xmm1, %xmm1
-; SSE-NEXT:    pand %xmm6, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm2[0,1,0,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,5,4,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm10[2,1,1,1]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[0,1,0,3,4,5,6,7]
 ; SSE-NEXT:    packuswb %xmm8, %xmm8
-; SSE-NEXT:    pandn %xmm8, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm1
+; SSE-NEXT:    pand %xmm7, %xmm8
+; SSE-NEXT:    pandn %xmm1, %xmm7
+; SSE-NEXT:    por %xmm8, %xmm7
 ; SSE-NEXT:    andps %xmm0, %xmm14
-; SSE-NEXT:    pandn %xmm1, %xmm0
+; SSE-NEXT:    pandn %xmm7, %xmm0
 ; SSE-NEXT:    por %xmm14, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, %xmm1
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -9279,7 +9303,7 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; SSE-NEXT:    movdqa %xmm11, (%rax)
 ; SSE-NEXT:    movdqa %xmm9, 48(%rax)
-; SSE-NEXT:    movdqa %xmm7, 32(%rax)
+; SSE-NEXT:    movdqa %xmm6, 32(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 16(%rax)
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
@@ -9287,7 +9311,7 @@ define void @load_i8_stride7_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
 ; SSE-NEXT:    movdqa %xmm13, 48(%rax)
 ; SSE-NEXT:    movdqa %xmm12, 32(%rax)
 ; SSE-NEXT:    movdqa %xmm3, 16(%rax)
-; SSE-NEXT:    addq $1528, %rsp # imm = 0x5F8
+; SSE-NEXT:    addq $1512, %rsp # imm = 0x5E8
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: load_i8_stride7_vf64:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-5.ll
index c1e7f1e8c6c72..c2c39c9f6dced 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-5.ll
@@ -269,29 +269,29 @@ define void @store_i16_stride5_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-LABEL: store_i16_stride5_vf4:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    movq {{.*#+}} xmm2 = mem[0],zero
-; SSE-NEXT:    movq {{.*#+}} xmm5 = mem[0],zero
+; SSE-NEXT:    movq {{.*#+}} xmm4 = mem[0],zero
 ; SSE-NEXT:    movq {{.*#+}} xmm3 = mem[0],zero
 ; SSE-NEXT:    movq {{.*#+}} xmm1 = mem[0],zero
 ; SSE-NEXT:    movq {{.*#+}} xmm0 = mem[0],zero
-; SSE-NEXT:    movdqa %xmm1, %xmm4
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3]
-; SSE-NEXT:    movdqa %xmm5, %xmm6
+; SSE-NEXT:    movdqa %xmm1, %xmm5
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; SSE-NEXT:    movdqa %xmm4, %xmm6
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm6[3,3,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm6[3,3,3,3]
 ; SSE-NEXT:    psrldq {{.*#+}} xmm6 = xmm6[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero
 ; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,65535,0,0,65535,65535,65535]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[2,1,1,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,6,7]
-; SSE-NEXT:    pand %xmm7, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[2,1,1,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,6,7]
+; SSE-NEXT:    pand %xmm7, %xmm5
 ; SSE-NEXT:    pandn %xmm6, %xmm7
-; SSE-NEXT:    por %xmm4, %xmm7
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,0,65535,65535,65535,65535,0,65535]
-; SSE-NEXT:    pand %xmm4, %xmm7
+; SSE-NEXT:    por %xmm5, %xmm7
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,0,65535,65535,65535,65535,0,65535]
+; SSE-NEXT:    pand %xmm5, %xmm7
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm0[0,1,0,1]
-; SSE-NEXT:    pandn %xmm6, %xmm4
-; SSE-NEXT:    por %xmm7, %xmm4
+; SSE-NEXT:    pandn %xmm6, %xmm5
+; SSE-NEXT:    por %xmm7, %xmm5
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,0,2,1]
@@ -307,13 +307,13 @@ define void @store_i16_stride5_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    por %xmm7, %xmm2
 ; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,0,65535,65535,65535,65535]
 ; SSE-NEXT:    psrlq $48, %xmm1
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
-; SSE-NEXT:    pand %xmm3, %xmm5
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSE-NEXT:    pand %xmm3, %xmm4
 ; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    por %xmm5, %xmm3
+; SSE-NEXT:    por %xmm4, %xmm3
 ; SSE-NEXT:    movq %xmm3, 32(%r9)
 ; SSE-NEXT:    movdqa %xmm2, (%r9)
-; SSE-NEXT:    movdqa %xmm4, 16(%r9)
+; SSE-NEXT:    movdqa %xmm5, 16(%r9)
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: store_i16_stride5_vf4:
@@ -2132,397 +2132,394 @@ define void @store_i16_stride5_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 define void @store_i16_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %in.vecptr3, ptr %in.vecptr4, ptr %out.vec) nounwind {
 ; SSE-LABEL: store_i16_stride5_vf32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $248, %rsp
-; SSE-NEXT:    movdqa (%rdi), %xmm5
-; SSE-NEXT:    movdqa 16(%rdi), %xmm9
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rsi), %xmm6
-; SSE-NEXT:    movdqa 16(%rsi), %xmm12
+; SSE-NEXT:    subq $232, %rsp
+; SSE-NEXT:    movdqa (%rdi), %xmm9
+; SSE-NEXT:    movdqa 16(%rdi), %xmm12
 ; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rdx), %xmm2
-; SSE-NEXT:    movdqa (%rcx), %xmm8
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%rsi), %xmm7
+; SSE-NEXT:    movdqa 16(%rsi), %xmm13
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%rdx), %xmm4
+; SSE-NEXT:    movdqa (%rcx), %xmm10
 ; SSE-NEXT:    movdqa 16(%rcx), %xmm14
 ; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%r8), %xmm0
+; SSE-NEXT:    movdqa (%r8), %xmm15
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,65535,65535,0,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    pandn %xmm9, %xmm1
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm7[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    pand %xmm6, %xmm2
+; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [0,65535,65535,65535,0,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[1,1,2,2]
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,65535,0,65535,65535,65535,65535,0]
+; SSE-NEXT:    pand %xmm11, %xmm2
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm10[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    pandn %xmm5, %xmm11
+; SSE-NEXT:    por %xmm2, %xmm11
+; SSE-NEXT:    pand %xmm1, %xmm11
+; SSE-NEXT:    por %xmm3, %xmm11
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,65535,65535,0,65535,65535,65535,65535]
+; SSE-NEXT:    pand %xmm8, %xmm11
+; SSE-NEXT:    movdqa %xmm8, %xmm0
+; SSE-NEXT:    pandn %xmm15, %xmm0
+; SSE-NEXT:    por %xmm11, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [65535,65535,65535,65535,0,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm15, %xmm1
-; SSE-NEXT:    pandn %xmm5, %xmm1
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm6[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm6, %xmm2
+; SSE-NEXT:    pandn %xmm12, %xmm2
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm13[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm15, %xmm3
-; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [0,65535,65535,65535,0,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm1, %xmm4
-; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm2[1,1,2,2]
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [65535,65535,0,65535,65535,65535,65535,0]
-; SSE-NEXT:    pand %xmm13, %xmm7
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm13, %xmm11
-; SSE-NEXT:    pandn %xmm8, %xmm11
-; SSE-NEXT:    por %xmm7, %xmm11
+; SSE-NEXT:    pand %xmm6, %xmm3
+; SSE-NEXT:    por %xmm2, %xmm3
+; SSE-NEXT:    movdqa %xmm1, %xmm5
+; SSE-NEXT:    pandn %xmm3, %xmm5
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm14[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,0,65535,65535,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm0, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    movdqa 16(%rdx), %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm2[1,1,2,2]
+; SSE-NEXT:    pand %xmm0, %xmm11
+; SSE-NEXT:    movdqa %xmm0, %xmm12
+; SSE-NEXT:    por %xmm11, %xmm3
+; SSE-NEXT:    pand %xmm1, %xmm3
+; SSE-NEXT:    por %xmm5, %xmm3
+; SSE-NEXT:    movdqa 16(%r8), %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm8, %xmm3
+; SSE-NEXT:    movdqa %xmm8, %xmm0
+; SSE-NEXT:    pandn %xmm5, %xmm0
+; SSE-NEXT:    por %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 32(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm6, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    movdqa 32(%rsi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    pand %xmm6, %xmm5
+; SSE-NEXT:    por %xmm3, %xmm5
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm5, %xmm3
+; SSE-NEXT:    movdqa 32(%rcx), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    movdqa %xmm12, %xmm11
+; SSE-NEXT:    pandn %xmm5, %xmm11
+; SSE-NEXT:    movdqa 32(%rdx), %xmm14
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm14[1,1,2,2]
+; SSE-NEXT:    pand %xmm0, %xmm12
+; SSE-NEXT:    movdqa %xmm0, %xmm5
+; SSE-NEXT:    por %xmm12, %xmm11
 ; SSE-NEXT:    pand %xmm1, %xmm11
-; SSE-NEXT:    por %xmm4, %xmm11
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,65535,0,65535,65535,65535,65535]
-; SSE-NEXT:    pand %xmm10, %xmm11
-; SSE-NEXT:    movdqa %xmm10, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm11
+; SSE-NEXT:    pand %xmm8, %xmm11
+; SSE-NEXT:    movdqa 32(%r8), %xmm12
+; SSE-NEXT:    movdqa %xmm8, %xmm0
+; SSE-NEXT:    pandn %xmm12, %xmm0
+; SSE-NEXT:    por %xmm11, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 48(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm6, %xmm3
 ; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    por %xmm11, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, %xmm4
-; SSE-NEXT:    pandn %xmm9, %xmm4
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm12[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm15, %xmm7
-; SSE-NEXT:    por %xmm4, %xmm7
-; SSE-NEXT:    movdqa %xmm1, %xmm8
-; SSE-NEXT:    pandn %xmm7, %xmm8
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm14[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm13, %xmm7
-; SSE-NEXT:    pandn %xmm4, %xmm7
-; SSE-NEXT:    movdqa 16(%rdx), %xmm14
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm14[1,1,2,2]
-; SSE-NEXT:    pand %xmm13, %xmm11
-; SSE-NEXT:    por %xmm11, %xmm7
-; SSE-NEXT:    pand %xmm1, %xmm7
-; SSE-NEXT:    por %xmm8, %xmm7
-; SSE-NEXT:    movdqa 16(%r8), %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm10, %xmm7
-; SSE-NEXT:    movdqa %xmm10, %xmm3
-; SSE-NEXT:    pandn %xmm4, %xmm3
-; SSE-NEXT:    por %xmm7, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 32(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, %xmm7
-; SSE-NEXT:    pandn %xmm3, %xmm7
-; SSE-NEXT:    movdqa 32(%rsi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm3[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm15, %xmm8
-; SSE-NEXT:    por %xmm7, %xmm8
-; SSE-NEXT:    movdqa %xmm1, %xmm7
-; SSE-NEXT:    pandn %xmm8, %xmm7
-; SSE-NEXT:    movdqa 32(%rcx), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm3[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm13, %xmm12
-; SSE-NEXT:    pandn %xmm8, %xmm12
-; SSE-NEXT:    movdqa 32(%rdx), %xmm11
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm11[1,1,2,2]
-; SSE-NEXT:    pand %xmm13, %xmm8
-; SSE-NEXT:    por %xmm8, %xmm12
-; SSE-NEXT:    pand %xmm1, %xmm12
-; SSE-NEXT:    por %xmm7, %xmm12
-; SSE-NEXT:    pand %xmm10, %xmm12
-; SSE-NEXT:    movdqa 32(%r8), %xmm7
-; SSE-NEXT:    movdqa %xmm7, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm3
-; SSE-NEXT:    pandn %xmm7, %xmm3
-; SSE-NEXT:    por %xmm12, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 48(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, %xmm7
-; SSE-NEXT:    pandn %xmm3, %xmm7
-; SSE-NEXT:    movdqa 48(%rsi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm3[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm15, %xmm8
-; SSE-NEXT:    por %xmm7, %xmm8
-; SSE-NEXT:    movdqa 48(%rcx), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm3[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm13, %xmm12
-; SSE-NEXT:    pandn %xmm7, %xmm12
-; SSE-NEXT:    movdqa 48(%rdx), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm3[1,1,2,2]
-; SSE-NEXT:    pand %xmm13, %xmm7
-; SSE-NEXT:    por %xmm7, %xmm12
-; SSE-NEXT:    pand %xmm1, %xmm12
-; SSE-NEXT:    pandn %xmm8, %xmm1
-; SSE-NEXT:    por %xmm12, %xmm1
-; SSE-NEXT:    pand %xmm10, %xmm1
-; SSE-NEXT:    movdqa 48(%r8), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pandn %xmm3, %xmm10
-; SSE-NEXT:    por %xmm1, %xmm10
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    movdqa 48(%rsi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    pand %xmm6, %xmm11
+; SSE-NEXT:    por %xmm3, %xmm11
+; SSE-NEXT:    movdqa 48(%rcx), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    movdqa %xmm5, %xmm15
+; SSE-NEXT:    pandn %xmm3, %xmm15
+; SSE-NEXT:    movdqa 48(%rdx), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[1,1,2,2]
+; SSE-NEXT:    pand %xmm5, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm15
+; SSE-NEXT:    pand %xmm1, %xmm15
+; SSE-NEXT:    pandn %xmm11, %xmm1
+; SSE-NEXT:    por %xmm15, %xmm1
+; SSE-NEXT:    pand %xmm8, %xmm1
+; SSE-NEXT:    movdqa 48(%r8), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm0, %xmm8
+; SSE-NEXT:    por %xmm1, %xmm8
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm4, %xmm1
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1],xmm1[2],xmm10[2],xmm1[3],xmm10[3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,2,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm1[0,0,2,1]
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,0,0,65535,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm3, %xmm8
-; SSE-NEXT:    pandn %xmm7, %xmm8
-; SSE-NEXT:    movdqa %xmm5, %xmm7
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm7[0,1,3,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[0,1,1,1]
-; SSE-NEXT:    pand %xmm3, %xmm12
-; SSE-NEXT:    por %xmm8, %xmm12
-; SSE-NEXT:    pand %xmm15, %xmm12
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm4[0,1,0,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,0,2,1]
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,0,0,65535,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm5, %xmm11
+; SSE-NEXT:    pandn %xmm3, %xmm11
+; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm7[0],xmm3[1],xmm7[1],xmm3[2],xmm7[2],xmm3[3],xmm7[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm3[0,1,3,2,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm15[0,1,1,1]
+; SSE-NEXT:    pand %xmm5, %xmm15
+; SSE-NEXT:    por %xmm11, %xmm15
+; SSE-NEXT:    pand %xmm6, %xmm15
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm8[0,1,0,1]
+; SSE-NEXT:    pandn %xmm11, %xmm6
+; SSE-NEXT:    por %xmm15, %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm10, %xmm15
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm4[0],xmm15[1],xmm4[1],xmm15[2],xmm4[2],xmm15[3],xmm4[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm15[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,5,4,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [65535,0,65535,65535,65535,65535,0,65535]
 ; SSE-NEXT:    movdqa %xmm15, %xmm1
-; SSE-NEXT:    pandn %xmm8, %xmm1
-; SSE-NEXT:    por %xmm12, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,6,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm0, %xmm12
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm2[0],xmm12[1],xmm2[1],xmm12[2],xmm2[2],xmm12[3],xmm2[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,5,4,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm7[0],xmm12[1],xmm7[1]
-; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,0,65535,65535,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm9, %xmm1
-; SSE-NEXT:    pandn %xmm8, %xmm1
-; SSE-NEXT:    pand %xmm9, %xmm12
-; SSE-NEXT:    por %xmm12, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
-; SSE-NEXT:    movdqa %xmm2, %xmm7
-; SSE-NEXT:    pslldq {{.*#+}} xmm7 = zero,zero,xmm7[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,0,0,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm0, %xmm8
-; SSE-NEXT:    pandn %xmm7, %xmm8
-; SSE-NEXT:    movdqa %xmm5, %xmm7
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm6[4],xmm7[5],xmm6[5],xmm7[6],xmm6[6],xmm7[7],xmm6[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,2,3,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm7[0,1,2,2]
-; SSE-NEXT:    pand %xmm0, %xmm12
-; SSE-NEXT:    por %xmm8, %xmm12
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [0,65535,65535,65535,65535,0,65535,65535]
-; SSE-NEXT:    pand %xmm10, %xmm12
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm4[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm10, %xmm1
-; SSE-NEXT:    pandn %xmm8, %xmm1
-; SSE-NEXT:    por %xmm12, %xmm1
+; SSE-NEXT:    pandn %xmm11, %xmm1
+; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    psrlq $48, %xmm6
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm5 = xmm5[1],xmm6[1]
-; SSE-NEXT:    movdqa %xmm0, %xmm6
-; SSE-NEXT:    pandn %xmm5, %xmm6
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,7,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,3,3]
-; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm10[4],xmm4[5],xmm10[5],xmm4[6],xmm10[6],xmm4[7],xmm10[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm4[0,1,2,3,4,5,7,6]
+; SSE-NEXT:    pslldq {{.*#+}} xmm4 = zero,zero,xmm4[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,0,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm1, %xmm10
+; SSE-NEXT:    pandn %xmm4, %xmm10
+; SSE-NEXT:    movdqa %xmm9, %xmm4
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm7[4],xmm4[5],xmm7[5],xmm4[6],xmm7[6],xmm4[7],xmm7[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,2,3,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm4[0,1,2,2]
+; SSE-NEXT:    pand %xmm1, %xmm11
+; SSE-NEXT:    por %xmm10, %xmm11
+; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [0,65535,65535,65535,65535,0,65535,65535]
+; SSE-NEXT:    pand %xmm13, %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm8[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    pandn %xmm10, %xmm0
+; SSE-NEXT:    por %xmm11, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    psrlq $48, %xmm7
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm9 = xmm9[1],xmm7[1]
+; SSE-NEXT:    movdqa %xmm1, %xmm7
+; SSE-NEXT:    pandn %xmm9, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,3,3]
+; SSE-NEXT:    pand %xmm1, %xmm3
+; SSE-NEXT:    por %xmm7, %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,0,65535,65535,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    pandn %xmm10, %xmm0
+; SSE-NEXT:    pand %xmm4, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1],xmm3[2],xmm9[2],xmm3[3],xmm9[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,2,2,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,0,2,1]
+; SSE-NEXT:    movdqa %xmm5, %xmm6
+; SSE-NEXT:    pandn %xmm3, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm3[0,1,3,2,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,1,1,1]
+; SSE-NEXT:    pand %xmm5, %xmm7
+; SSE-NEXT:    por %xmm6, %xmm7
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,0,65535,65535,65535]
+; SSE-NEXT:    pand %xmm0, %xmm7
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm10[0,1,0,1]
+; SSE-NEXT:    movdqa %xmm0, %xmm11
+; SSE-NEXT:    pandn %xmm6, %xmm11
+; SSE-NEXT:    por %xmm7, %xmm11
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm9, %xmm7
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm2[0],xmm7[1],xmm2[1],xmm7[2],xmm2[2],xmm7[3],xmm2[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,5,4,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm3[0],xmm7[1],xmm3[1]
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    pandn %xmm6, %xmm0
+; SSE-NEXT:    pand %xmm15, %xmm7
+; SSE-NEXT:    por %xmm7, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm9[4],xmm2[5],xmm9[5],xmm2[6],xmm9[6],xmm2[7],xmm9[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm2[0,1,2,3,4,5,7,6]
+; SSE-NEXT:    pslldq {{.*#+}} xmm2 = zero,zero,xmm2[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT:    movdqa %xmm1, %xmm6
+; SSE-NEXT:    pandn %xmm2, %xmm6
+; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm8[4],xmm2[5],xmm8[5],xmm2[6],xmm8[6],xmm2[7],xmm8[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,2,3,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,2]
+; SSE-NEXT:    pand %xmm1, %xmm2
 ; SSE-NEXT:    por %xmm6, %xmm2
-; SSE-NEXT:    movdqa %xmm13, %xmm1
-; SSE-NEXT:    pandn %xmm8, %xmm1
 ; SSE-NEXT:    pand %xmm13, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm10[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm13, %xmm7
+; SSE-NEXT:    pandn %xmm6, %xmm7
+; SSE-NEXT:    por %xmm2, %xmm7
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm8, %xmm2
+; SSE-NEXT:    psrlq $48, %xmm2
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    pandn %xmm0, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,3,3]
+; SSE-NEXT:    pand %xmm1, %xmm3
+; SSE-NEXT:    por %xmm2, %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,0,65535,65535,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    pandn %xmm6, %xmm0
+; SSE-NEXT:    pand %xmm10, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm14, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm9[0],xmm2[1],xmm9[1],xmm2[2],xmm9[2],xmm2[3],xmm9[3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,2,1]
-; SSE-NEXT:    movdqa %xmm3, %xmm5
-; SSE-NEXT:    pandn %xmm2, %xmm5
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm3
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm7[0],xmm2[1],xmm7[1],xmm2[2],xmm7[2],xmm2[3],xmm7[3]
+; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    movdqa (%rsp), %xmm8 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm2[0,1,3,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,1,1]
-; SSE-NEXT:    pand %xmm3, %xmm6
-; SSE-NEXT:    por %xmm5, %xmm6
-; SSE-NEXT:    pand %xmm15, %xmm6
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[0,1,0,1]
-; SSE-NEXT:    movdqa %xmm15, %xmm12
-; SSE-NEXT:    pandn %xmm5, %xmm12
-; SSE-NEXT:    por %xmm6, %xmm12
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm8, %xmm6
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm14[0],xmm6[1],xmm14[1],xmm6[2],xmm14[2],xmm6[3],xmm14[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,5,4,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm2[0],xmm6[1],xmm2[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[0,1,1,1]
+; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    por %xmm3, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,65535,0,65535,65535,65535]
+; SSE-NEXT:    pand %xmm4, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm12[0,1,0,1]
+; SSE-NEXT:    movdqa %xmm4, %xmm6
+; SSE-NEXT:    pandn %xmm3, %xmm6
+; SSE-NEXT:    por %xmm0, %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm2[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
 ; SSE-NEXT:    movdqa %xmm9, %xmm2
-; SSE-NEXT:    pandn %xmm5, %xmm2
-; SSE-NEXT:    pand %xmm9, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm14 = xmm14[4],xmm8[4],xmm14[5],xmm8[5],xmm14[6],xmm8[6],xmm14[7],xmm8[7]
-; SSE-NEXT:    movdqa %xmm14, %xmm2
-; SSE-NEXT:    pslldq {{.*#+}} xmm2 = zero,zero,xmm2[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    movdqa %xmm0, %xmm5
-; SSE-NEXT:    pandn %xmm2, %xmm5
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm7[4],xmm2[5],xmm7[5],xmm2[6],xmm7[6],xmm2[7],xmm7[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,2,3,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,2]
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    por %xmm5, %xmm2
-; SSE-NEXT:    pand %xmm10, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm4[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm10, %xmm4
-; SSE-NEXT:    pandn %xmm5, %xmm4
-; SSE-NEXT:    por %xmm2, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    psrlq $48, %xmm2
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; SSE-NEXT:    movdqa %xmm0, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm14[0,1,2,3,4,5,7,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,1,3,3]
-; SSE-NEXT:    pand %xmm0, %xmm4
-; SSE-NEXT:    por %xmm2, %xmm4
-; SSE-NEXT:    movdqa %xmm13, %xmm1
-; SSE-NEXT:    pandn %xmm5, %xmm1
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm14[0],xmm2[1],xmm14[1],xmm2[2],xmm14[2],xmm2[3],xmm14[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,4,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
+; SSE-NEXT:    movdqa %xmm15, %xmm11
+; SSE-NEXT:    pandn %xmm3, %xmm11
+; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm11
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm14 = xmm14[4],xmm9[4],xmm14[5],xmm9[5],xmm14[6],xmm9[6],xmm14[7],xmm9[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm14[0,1,2,3,4,5,7,6]
+; SSE-NEXT:    pslldq {{.*#+}} xmm14 = zero,zero,xmm14[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    pandn %xmm14, %xmm0
+; SSE-NEXT:    movdqa %xmm7, %xmm4
+; SSE-NEXT:    movdqa %xmm8, %xmm6
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm8[4],xmm4[5],xmm8[5],xmm4[6],xmm8[6],xmm4[7],xmm8[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[0,2,3,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,1,2,2]
+; SSE-NEXT:    pand %xmm1, %xmm4
+; SSE-NEXT:    por %xmm0, %xmm4
 ; SSE-NEXT:    pand %xmm13, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,2,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,2,1]
-; SSE-NEXT:    movdqa %xmm3, %xmm4
-; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm12[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm13, %xmm8
+; SSE-NEXT:    pandn %xmm14, %xmm8
+; SSE-NEXT:    por %xmm4, %xmm8
+; SSE-NEXT:    movdqa %xmm6, %xmm4
+; SSE-NEXT:    psrlq $48, %xmm4
+; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm4[1]
+; SSE-NEXT:    movdqa %xmm1, %xmm4
+; SSE-NEXT:    pandn %xmm0, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[2,1,3,3]
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    por %xmm4, %xmm0
+; SSE-NEXT:    movdqa %xmm10, %xmm3
+; SSE-NEXT:    pandn %xmm14, %xmm3
+; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm12, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm2[0,1,3,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,1,1]
-; SSE-NEXT:    pand %xmm3, %xmm5
-; SSE-NEXT:    por %xmm4, %xmm5
-; SSE-NEXT:    pand %xmm15, %xmm5
-; SSE-NEXT:    movdqa (%rsp), %xmm12 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[0,1,0,1]
-; SSE-NEXT:    movdqa %xmm15, %xmm4
-; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    por %xmm5, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm8, %xmm5
-; SSE-NEXT:    movdqa %xmm8, %xmm4
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm11[0],xmm5[1],xmm11[1],xmm5[2],xmm11[2],xmm5[3],xmm11[3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,2,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm10, %xmm4
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm4[0,1,3,2,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm14[0,1,1,1]
+; SSE-NEXT:    pand %xmm5, %xmm14
+; SSE-NEXT:    pandn %xmm0, %xmm5
+; SSE-NEXT:    por %xmm14, %xmm5
+; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [65535,65535,65535,65535,0,65535,65535,65535]
+; SSE-NEXT:    pand %xmm14, %xmm5
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[0,1,0,1]
+; SSE-NEXT:    pandn %xmm2, %xmm14
+; SSE-NEXT:    por %xmm5, %xmm14
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm6, %xmm5
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm12[0],xmm5[1],xmm12[1],xmm5[2],xmm12[2],xmm5[3],xmm12[3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,4,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
-; SSE-NEXT:    movdqa %xmm9, %xmm8
-; SSE-NEXT:    pandn %xmm1, %xmm8
-; SSE-NEXT:    pand %xmm9, %xmm5
-; SSE-NEXT:    por %xmm5, %xmm8
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm11 = xmm11[4],xmm4[4],xmm11[5],xmm4[5],xmm11[6],xmm4[6],xmm11[7],xmm4[7]
-; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    pslldq {{.*#+}} xmm1 = zero,zero,xmm1[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    movdqa %xmm0, %xmm5
-; SSE-NEXT:    pandn %xmm1, %xmm5
-; SSE-NEXT:    movdqa %xmm7, %xmm4
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,2,3,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,2]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    por %xmm5, %xmm1
-; SSE-NEXT:    pand %xmm10, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm12[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm10, %xmm9
-; SSE-NEXT:    pandn %xmm2, %xmm9
-; SSE-NEXT:    por %xmm1, %xmm9
-; SSE-NEXT:    movdqa %xmm6, %xmm1
-; SSE-NEXT:    psrlq $48, %xmm1
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm4 = xmm4[1],xmm1[1]
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    pandn %xmm4, %xmm1
-; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,5,7,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[2,1,3,3]
-; SSE-NEXT:    pand %xmm0, %xmm11
-; SSE-NEXT:    por %xmm1, %xmm11
-; SSE-NEXT:    movdqa %xmm13, %xmm14
-; SSE-NEXT:    pandn %xmm2, %xmm14
-; SSE-NEXT:    pand %xmm13, %xmm11
-; SSE-NEXT:    por %xmm11, %xmm14
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1],xmm2[2],xmm6[2],xmm2[3],xmm6[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,2,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,2,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm7, %xmm11
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm5[0],xmm11[1],xmm5[1],xmm11[2],xmm5[2],xmm11[3],xmm5[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm11[0,1,3,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[0,1,1,1]
-; SSE-NEXT:    pand %xmm3, %xmm12
-; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    por %xmm12, %xmm3
-; SSE-NEXT:    pand %xmm15, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm4[0,1,0,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSE-NEXT:    pand %xmm15, %xmm5
 ; SSE-NEXT:    pandn %xmm2, %xmm15
-; SSE-NEXT:    por %xmm3, %xmm15
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm11[0,1,2,3,4,5,6,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm6, %xmm11
-; SSE-NEXT:    movdqa %xmm1, %xmm12
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm1[0],xmm11[1],xmm1[1],xmm11[2],xmm1[2],xmm11[3],xmm1[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm11[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,5,4,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm3[0],xmm11[1],xmm3[1]
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,0,65535,65535,65535,65535,0,65535]
-; SSE-NEXT:    pand %xmm1, %xmm11
-; SSE-NEXT:    pandn %xmm2, %xmm1
-; SSE-NEXT:    por %xmm11, %xmm1
-; SSE-NEXT:    movdqa %xmm12, %xmm3
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm6[4],xmm3[5],xmm6[5],xmm3[6],xmm6[6],xmm3[7],xmm6[7]
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    movdqa %xmm3, %xmm6
-; SSE-NEXT:    pslldq {{.*#+}} xmm2 = zero,zero,xmm2[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,2,3,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,2]
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    pand %xmm10, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[2,3,2,3]
-; SSE-NEXT:    pandn %xmm3, %xmm10
-; SSE-NEXT:    por %xmm2, %xmm10
-; SSE-NEXT:    psrlq $48, %xmm5
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm7 = xmm7[1],xmm5[1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm6[0,1,2,3,4,5,7,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,3,3]
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    pandn %xmm7, %xmm0
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    pand %xmm13, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm13
-; SSE-NEXT:    por %xmm0, %xmm13
-; SSE-NEXT:    movdqa %xmm13, 304(%r9)
-; SSE-NEXT:    movdqa %xmm10, 288(%r9)
-; SSE-NEXT:    movdqa %xmm1, 256(%r9)
-; SSE-NEXT:    movdqa %xmm15, 240(%r9)
-; SSE-NEXT:    movdqa %xmm14, 224(%r9)
-; SSE-NEXT:    movdqa %xmm9, 208(%r9)
-; SSE-NEXT:    movdqa %xmm8, 176(%r9)
+; SSE-NEXT:    por %xmm5, %xmm15
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm12 = xmm12[4],xmm6[4],xmm12[5],xmm6[5],xmm12[6],xmm6[6],xmm12[7],xmm6[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm12[0,1,2,3,4,5,7,6]
+; SSE-NEXT:    pslldq {{.*#+}} xmm12 = zero,zero,xmm12[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT:    movdqa %xmm1, %xmm4
+; SSE-NEXT:    pandn %xmm12, %xmm4
+; SSE-NEXT:    movdqa %xmm10, %xmm5
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm9[4],xmm5[5],xmm9[5],xmm5[6],xmm9[6],xmm5[7],xmm9[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,2,3,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,2]
+; SSE-NEXT:    pand %xmm1, %xmm5
+; SSE-NEXT:    por %xmm4, %xmm5
+; SSE-NEXT:    pand %xmm13, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm7[2,3,2,3]
+; SSE-NEXT:    pandn %xmm4, %xmm13
+; SSE-NEXT:    por %xmm5, %xmm13
+; SSE-NEXT:    psrlq $48, %xmm9
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm10 = xmm10[1],xmm9[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,1,3,3]
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    pandn %xmm10, %xmm1
+; SSE-NEXT:    por %xmm0, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,0,65535,65535,65535,65535,0]
+; SSE-NEXT:    pand %xmm0, %xmm1
+; SSE-NEXT:    pandn %xmm4, %xmm0
+; SSE-NEXT:    por %xmm1, %xmm0
+; SSE-NEXT:    movdqa %xmm0, 304(%r9)
+; SSE-NEXT:    movdqa %xmm13, 288(%r9)
+; SSE-NEXT:    movdqa %xmm15, 256(%r9)
+; SSE-NEXT:    movdqa %xmm14, 240(%r9)
+; SSE-NEXT:    movdqa %xmm3, 224(%r9)
+; SSE-NEXT:    movdqa %xmm8, 208(%r9)
+; SSE-NEXT:    movdqa %xmm11, 176(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 160(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -2549,7 +2546,7 @@ define void @store_i16_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; SSE-NEXT:    movaps %xmm0, 112(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 32(%r9)
-; SSE-NEXT:    addq $248, %rsp
+; SSE-NEXT:    addq $232, %rsp
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: store_i16_stride5_vf32:
@@ -4210,488 +4207,555 @@ define void @store_i16_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; SSE-LABEL: store_i16_stride5_vf64:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    subq $616, %rsp # imm = 0x268
-; SSE-NEXT:    movdqa (%rdi), %xmm14
-; SSE-NEXT:    movdqa 16(%rdi), %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rsi), %xmm12
+; SSE-NEXT:    movdqa (%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 16(%rdi), %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%rsi), %xmm13
 ; SSE-NEXT:    movdqa 16(%rsi), %xmm11
 ; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa (%rdx), %xmm7
-; SSE-NEXT:    movdqa (%rcx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 16(%rcx), %xmm13
-; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%r8), %xmm15
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,65535,65535,0,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm9, %xmm1
-; SSE-NEXT:    pandn %xmm14, %xmm1
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm12[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm9, %xmm3
-; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    movdqa (%rcx), %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 16(%rcx), %xmm12
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%r8), %xmm14
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,65535,65535,0,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    pandn %xmm0, %xmm1
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm13[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    pand %xmm10, %xmm4
+; SSE-NEXT:    por %xmm1, %xmm4
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [0,65535,65535,65535,0,0,65535,65535]
 ; SSE-NEXT:    movdqa %xmm1, %xmm5
-; SSE-NEXT:    pandn %xmm3, %xmm5
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[1,1,2,2]
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,0,65535,65535,65535,65535,0]
-; SSE-NEXT:    pand %xmm10, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pandn %xmm4, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm7[1,1,2,2]
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,0,65535,65535,65535,65535,0]
+; SSE-NEXT:    pand %xmm2, %xmm4
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm6[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm10, %xmm8
+; SSE-NEXT:    movdqa %xmm2, %xmm8
 ; SSE-NEXT:    pandn %xmm6, %xmm8
-; SSE-NEXT:    por %xmm3, %xmm8
+; SSE-NEXT:    por %xmm4, %xmm8
 ; SSE-NEXT:    pand %xmm1, %xmm8
 ; SSE-NEXT:    por %xmm5, %xmm8
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,65535,65535,65535,65535]
-; SSE-NEXT:    pand %xmm2, %xmm8
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pandn %xmm15, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,65535,0,65535,65535,65535,65535]
+; SSE-NEXT:    pand %xmm9, %xmm8
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    pandn %xmm14, %xmm0
 ; SSE-NEXT:    por %xmm8, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm3
-; SSE-NEXT:    pandn %xmm4, %xmm3
+; SSE-NEXT:    movdqa %xmm10, %xmm4
+; SSE-NEXT:    pandn %xmm3, %xmm4
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm11[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm9, %xmm5
-; SSE-NEXT:    por %xmm3, %xmm5
-; SSE-NEXT:    movdqa %xmm1, %xmm3
-; SSE-NEXT:    pandn %xmm5, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm13[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm10, %xmm6
+; SSE-NEXT:    pand %xmm10, %xmm5
+; SSE-NEXT:    por %xmm4, %xmm5
+; SSE-NEXT:    movdqa %xmm1, %xmm6
 ; SSE-NEXT:    pandn %xmm5, %xmm6
-; SSE-NEXT:    movdqa 16(%rdx), %xmm5
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm5[1,1,2,2]
-; SSE-NEXT:    pand %xmm10, %xmm8
-; SSE-NEXT:    por %xmm8, %xmm6
-; SSE-NEXT:    pand %xmm1, %xmm6
-; SSE-NEXT:    por %xmm3, %xmm6
-; SSE-NEXT:    movdqa 16(%r8), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pand %xmm2, %xmm6
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    por %xmm6, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm12[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    movdqa %xmm2, %xmm5
+; SSE-NEXT:    pandn %xmm4, %xmm5
+; SSE-NEXT:    movdqa 16(%rdx), %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm3[1,1,2,2]
+; SSE-NEXT:    pand %xmm2, %xmm8
+; SSE-NEXT:    por %xmm8, %xmm5
+; SSE-NEXT:    pand %xmm1, %xmm5
+; SSE-NEXT:    por %xmm6, %xmm5
+; SSE-NEXT:    movdqa 16(%r8), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm9, %xmm5
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    pandn %xmm4, %xmm0
+; SSE-NEXT:    por %xmm5, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 32(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm3
-; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm10, %xmm5
+; SSE-NEXT:    pandn %xmm0, %xmm5
 ; SSE-NEXT:    movdqa 32(%rsi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm0[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm9, %xmm6
-; SSE-NEXT:    por %xmm3, %xmm6
-; SSE-NEXT:    movdqa %xmm1, %xmm3
-; SSE-NEXT:    pandn %xmm6, %xmm3
+; SSE-NEXT:    pand %xmm10, %xmm6
+; SSE-NEXT:    por %xmm5, %xmm6
+; SSE-NEXT:    movdqa %xmm1, %xmm8
+; SSE-NEXT:    pandn %xmm6, %xmm8
 ; SSE-NEXT:    movdqa 32(%rcx), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm0[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm10, %xmm8
-; SSE-NEXT:    pandn %xmm6, %xmm8
-; SSE-NEXT:    movdqa 32(%rdx), %xmm6
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm6[1,1,2,2]
-; SSE-NEXT:    pand %xmm10, %xmm11
-; SSE-NEXT:    por %xmm11, %xmm8
-; SSE-NEXT:    pand %xmm1, %xmm8
-; SSE-NEXT:    por %xmm3, %xmm8
-; SSE-NEXT:    pand %xmm2, %xmm8
-; SSE-NEXT:    movdqa 32(%r8), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    por %xmm8, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    movdqa %xmm2, %xmm6
+; SSE-NEXT:    pandn %xmm5, %xmm6
+; SSE-NEXT:    movdqa 32(%rdx), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm0[1,1,2,2]
+; SSE-NEXT:    pand %xmm2, %xmm11
+; SSE-NEXT:    por %xmm11, %xmm6
+; SSE-NEXT:    pand %xmm1, %xmm6
+; SSE-NEXT:    por %xmm8, %xmm6
+; SSE-NEXT:    pand %xmm9, %xmm6
+; SSE-NEXT:    movdqa 32(%r8), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    pandn %xmm4, %xmm0
+; SSE-NEXT:    por %xmm6, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 48(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm3
-; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm10, %xmm6
+; SSE-NEXT:    pandn %xmm0, %xmm6
 ; SSE-NEXT:    movdqa 48(%rsi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm0[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm9, %xmm8
-; SSE-NEXT:    por %xmm3, %xmm8
-; SSE-NEXT:    movdqa %xmm1, %xmm3
-; SSE-NEXT:    pandn %xmm8, %xmm3
+; SSE-NEXT:    pand %xmm10, %xmm8
+; SSE-NEXT:    por %xmm6, %xmm8
+; SSE-NEXT:    movdqa %xmm1, %xmm11
+; SSE-NEXT:    pandn %xmm8, %xmm11
 ; SSE-NEXT:    movdqa 48(%rcx), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm0[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm10, %xmm11
-; SSE-NEXT:    pandn %xmm8, %xmm11
-; SSE-NEXT:    movdqa 48(%rdx), %xmm8
-; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm8[1,1,2,2]
-; SSE-NEXT:    pand %xmm10, %xmm15
-; SSE-NEXT:    por %xmm15, %xmm11
-; SSE-NEXT:    pand %xmm1, %xmm11
-; SSE-NEXT:    por %xmm3, %xmm11
-; SSE-NEXT:    pand %xmm2, %xmm11
-; SSE-NEXT:    movdqa 48(%r8), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    por %xmm11, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    movdqa %xmm2, %xmm8
+; SSE-NEXT:    pandn %xmm6, %xmm8
+; SSE-NEXT:    movdqa 48(%rdx), %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm6[1,1,2,2]
+; SSE-NEXT:    pand %xmm2, %xmm12
+; SSE-NEXT:    por %xmm12, %xmm8
+; SSE-NEXT:    pand %xmm1, %xmm8
+; SSE-NEXT:    por %xmm11, %xmm8
+; SSE-NEXT:    pand %xmm9, %xmm8
+; SSE-NEXT:    movdqa 48(%r8), %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    pandn %xmm4, %xmm0
+; SSE-NEXT:    por %xmm8, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 64(%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm3
-; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm10, %xmm8
+; SSE-NEXT:    pandn %xmm0, %xmm8
 ; SSE-NEXT:    movdqa 64(%rsi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm0[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm9, %xmm11
-; SSE-NEXT:    por %xmm3, %xmm11
-; SSE-NEXT:    movdqa %xmm1, %xmm15
-; SSE-NEXT:    pandn %xmm11, %xmm15
+; SSE-NEXT:    pand %xmm10, %xmm11
+; SSE-NEXT:    por %xmm8, %xmm11
+; SSE-NEXT:    movdqa %xmm1, %xmm12
+; SSE-NEXT:    pandn %xmm11, %xmm12
 ; SSE-NEXT:    movdqa 64(%rcx), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm0[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm10, %xmm11
-; SSE-NEXT:    pandn %xmm3, %xmm11
-; SSE-NEXT:    movdqa 64(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,2,2]
-; SSE-NEXT:    pand %xmm10, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm11
-; SSE-NEXT:    pand %xmm1, %xmm11
-; SSE-NEXT:    por %xmm15, %xmm11
-; SSE-NEXT:    pand %xmm2, %xmm11
-; SSE-NEXT:    movdqa 64(%r8), %xmm3
-; SSE-NEXT:    movdqa %xmm3, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    movdqa %xmm2, %xmm11
+; SSE-NEXT:    pandn %xmm8, %xmm11
+; SSE-NEXT:    movdqa 64(%rdx), %xmm8
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm8[1,1,2,2]
+; SSE-NEXT:    pand %xmm2, %xmm14
+; SSE-NEXT:    por %xmm14, %xmm11
+; SSE-NEXT:    pand %xmm1, %xmm11
+; SSE-NEXT:    por %xmm12, %xmm11
+; SSE-NEXT:    pand %xmm9, %xmm11
+; SSE-NEXT:    movdqa 64(%r8), %xmm4
+; SSE-NEXT:    movdqa %xmm4, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    pandn %xmm4, %xmm0
 ; SSE-NEXT:    por %xmm11, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 80(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    movdqa 80(%rsi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm3[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm9, %xmm11
-; SSE-NEXT:    por %xmm0, %xmm11
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    pandn %xmm11, %xmm0
-; SSE-NEXT:    movdqa 80(%rcx), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm3[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa 80(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm10, %xmm11
+; SSE-NEXT:    pandn %xmm0, %xmm11
+; SSE-NEXT:    movdqa 80(%rsi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    pand %xmm10, %xmm12
+; SSE-NEXT:    por %xmm11, %xmm12
+; SSE-NEXT:    movdqa %xmm1, %xmm14
+; SSE-NEXT:    pandn %xmm12, %xmm14
+; SSE-NEXT:    movdqa 80(%rcx), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm0[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm10, %xmm15
-; SSE-NEXT:    pandn %xmm11, %xmm15
-; SSE-NEXT:    movdqa 80(%rdx), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm3[1,1,2,2]
-; SSE-NEXT:    pand %xmm10, %xmm11
-; SSE-NEXT:    por %xmm11, %xmm15
-; SSE-NEXT:    pand %xmm1, %xmm15
-; SSE-NEXT:    por %xmm0, %xmm15
-; SSE-NEXT:    pand %xmm2, %xmm15
-; SSE-NEXT:    movdqa 80(%r8), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    por %xmm15, %xmm0
+; SSE-NEXT:    movdqa %xmm2, %xmm12
+; SSE-NEXT:    pandn %xmm11, %xmm12
+; SSE-NEXT:    movdqa 80(%rdx), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 96(%rdi), %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm0[1,1,2,2]
+; SSE-NEXT:    pand %xmm2, %xmm15
+; SSE-NEXT:    por %xmm15, %xmm12
+; SSE-NEXT:    pand %xmm1, %xmm12
+; SSE-NEXT:    por %xmm14, %xmm12
+; SSE-NEXT:    pand %xmm9, %xmm12
+; SSE-NEXT:    movdqa 80(%r8), %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm9, %xmm0
 ; SSE-NEXT:    pandn %xmm4, %xmm0
-; SSE-NEXT:    movdqa 96(%rsi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm3[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm9, %xmm11
-; SSE-NEXT:    por %xmm0, %xmm11
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    pandn %xmm11, %xmm0
-; SSE-NEXT:    movdqa 96(%rcx), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm3[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm10, %xmm15
-; SSE-NEXT:    pandn %xmm11, %xmm15
-; SSE-NEXT:    movdqa 96(%rdx), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm3[1,1,2,2]
-; SSE-NEXT:    pand %xmm10, %xmm11
-; SSE-NEXT:    por %xmm11, %xmm15
-; SSE-NEXT:    pand %xmm1, %xmm15
-; SSE-NEXT:    por %xmm0, %xmm15
-; SSE-NEXT:    pand %xmm2, %xmm15
-; SSE-NEXT:    movdqa 96(%r8), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    por %xmm15, %xmm0
+; SSE-NEXT:    por %xmm12, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 96(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm10, %xmm12
+; SSE-NEXT:    pandn %xmm0, %xmm12
+; SSE-NEXT:    movdqa 96(%rsi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    pand %xmm10, %xmm14
+; SSE-NEXT:    por %xmm12, %xmm14
+; SSE-NEXT:    movdqa %xmm1, %xmm12
+; SSE-NEXT:    pandn %xmm14, %xmm12
+; SSE-NEXT:    movdqa 96(%rcx), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 112(%rdi), %xmm4
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    movdqa %xmm2, %xmm15
+; SSE-NEXT:    pandn %xmm14, %xmm15
+; SSE-NEXT:    movdqa 96(%rdx), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm0[1,1,2,2]
+; SSE-NEXT:    pand %xmm2, %xmm14
+; SSE-NEXT:    por %xmm14, %xmm15
+; SSE-NEXT:    pand %xmm1, %xmm15
+; SSE-NEXT:    por %xmm12, %xmm15
+; SSE-NEXT:    pand %xmm9, %xmm15
+; SSE-NEXT:    movdqa 96(%r8), %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm9, %xmm0
 ; SSE-NEXT:    pandn %xmm4, %xmm0
-; SSE-NEXT:    movdqa 112(%rsi), %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm4[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm11 = xmm11[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm9, %xmm11
-; SSE-NEXT:    por %xmm0, %xmm11
+; SSE-NEXT:    por %xmm15, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 112(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm10, %xmm12
+; SSE-NEXT:    pandn %xmm0, %xmm12
+; SSE-NEXT:    movdqa 112(%rsi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    pand %xmm10, %xmm14
+; SSE-NEXT:    por %xmm12, %xmm14
 ; SSE-NEXT:    movdqa 112(%rcx), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm10, %xmm15
-; SSE-NEXT:    pandn %xmm0, %xmm15
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    movdqa %xmm2, %xmm15
+; SSE-NEXT:    pandn %xmm12, %xmm15
 ; SSE-NEXT:    movdqa 112(%rdx), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,2,2]
-; SSE-NEXT:    pand %xmm10, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm15
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm0[1,1,2,2]
+; SSE-NEXT:    pand %xmm2, %xmm12
+; SSE-NEXT:    por %xmm12, %xmm15
 ; SSE-NEXT:    pand %xmm1, %xmm15
-; SSE-NEXT:    pandn %xmm11, %xmm1
+; SSE-NEXT:    pandn %xmm14, %xmm1
 ; SSE-NEXT:    por %xmm15, %xmm1
-; SSE-NEXT:    pand %xmm2, %xmm1
+; SSE-NEXT:    pand %xmm9, %xmm1
 ; SSE-NEXT:    movdqa 112(%r8), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,0,0,65535,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm1, %xmm11
-; SSE-NEXT:    pandn %xmm0, %xmm11
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1],xmm0[2],xmm12[2],xmm0[3],xmm12[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm0[0,1,3,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm15[0,1,1,1]
-; SSE-NEXT:    pand %xmm1, %xmm15
-; SSE-NEXT:    por %xmm11, %xmm15
-; SSE-NEXT:    movdqa %xmm9, %xmm13
-; SSE-NEXT:    pand %xmm9, %xmm15
+; SSE-NEXT:    pandn %xmm0, %xmm9
+; SSE-NEXT:    por %xmm1, %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,2,2,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm1[0,0,2,1]
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,65535,0,0,65535,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm11, %xmm14
+; SSE-NEXT:    pandn %xmm12, %xmm14
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm4[0,1,0,1]
-; SSE-NEXT:    movdqa %xmm9, %xmm1
-; SSE-NEXT:    pandn %xmm11, %xmm1
+; SSE-NEXT:    movdqa %xmm4, %xmm12
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm12 = xmm12[0],xmm13[0],xmm12[1],xmm13[1],xmm12[2],xmm13[2],xmm12[3],xmm13[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm12[0,1,3,2,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm15[0,1,1,1]
+; SSE-NEXT:    pand %xmm11, %xmm15
+; SSE-NEXT:    por %xmm14, %xmm15
+; SSE-NEXT:    pand %xmm10, %xmm15
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[0,1,0,1]
+; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    por %xmm15, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm2, %xmm15
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm7[0],xmm15[1],xmm7[1],xmm15[2],xmm7[2],xmm15[3],xmm7[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm15[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,5,4,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [65535,0,65535,65535,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm15, %xmm1
-; SSE-NEXT:    pandn %xmm11, %xmm1
-; SSE-NEXT:    pand %xmm15, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm2[4],xmm7[5],xmm2[5],xmm7[6],xmm2[6],xmm7[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    pslldq {{.*#+}} xmm1 = zero,zero,xmm1[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,65535,0,0,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm9, %xmm11
-; SSE-NEXT:    pandn %xmm1, %xmm11
-; SSE-NEXT:    movdqa %xmm14, %xmm1
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm12[4],xmm1[5],xmm12[5],xmm1[6],xmm12[6],xmm1[7],xmm12[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,2,3,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,2]
-; SSE-NEXT:    pand %xmm9, %xmm1
-; SSE-NEXT:    por %xmm11, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [0,65535,65535,65535,65535,0,65535,65535]
-; SSE-NEXT:    pand %xmm3, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm3, %xmm11
-; SSE-NEXT:    pandn %xmm0, %xmm11
-; SSE-NEXT:    por %xmm1, %xmm11
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    psrlq $48, %xmm12
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm14 = xmm14[1],xmm12[1]
-; SSE-NEXT:    movdqa %xmm9, %xmm1
-; SSE-NEXT:    pandn %xmm14, %xmm1
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,7,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[2,1,3,3]
-; SSE-NEXT:    pand %xmm9, %xmm7
-; SSE-NEXT:    por %xmm1, %xmm7
-; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm5, %xmm14
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm7[0],xmm14[1],xmm7[1],xmm14[2],xmm7[2],xmm14[3],xmm7[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm14[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,5,4,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm14[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm12[0],xmm15[1],xmm12[1]
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,0,65535,65535,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm12, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm10, %xmm7
-; SSE-NEXT:    por %xmm7, %xmm1
+; SSE-NEXT:    pand %xmm12, %xmm15
+; SSE-NEXT:    por %xmm15, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm5[4],xmm7[5],xmm5[5],xmm7[6],xmm5[6],xmm7[7],xmm5[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm7[0,1,2,3,4,5,7,6]
+; SSE-NEXT:    pslldq {{.*#+}} xmm7 = zero,zero,xmm7[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,0,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm1, %xmm12
+; SSE-NEXT:    pandn %xmm7, %xmm12
+; SSE-NEXT:    movdqa %xmm4, %xmm7
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm13[4],xmm7[5],xmm13[5],xmm7[6],xmm13[6],xmm7[7],xmm13[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,2,3,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,1,2,2]
+; SSE-NEXT:    pand %xmm1, %xmm7
+; SSE-NEXT:    por %xmm12, %xmm7
+; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [0,65535,65535,65535,65535,0,65535,65535]
+; SSE-NEXT:    pand %xmm14, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm14, %xmm9
+; SSE-NEXT:    pandn %xmm0, %xmm9
+; SSE-NEXT:    por %xmm7, %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    psrlq $48, %xmm13
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm4 = xmm4[1],xmm13[1]
+; SSE-NEXT:    movdqa %xmm1, %xmm7
+; SSE-NEXT:    pandn %xmm4, %xmm7
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm15[2,1,3,3]
+; SSE-NEXT:    pand %xmm1, %xmm13
+; SSE-NEXT:    por %xmm7, %xmm13
+; SSE-NEXT:    movdqa %xmm2, %xmm7
+; SSE-NEXT:    pandn %xmm0, %xmm7
+; SSE-NEXT:    pand %xmm2, %xmm13
+; SSE-NEXT:    por %xmm13, %xmm7
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm3, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,0,0,65535,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1],xmm0[2],xmm14[2],xmm0[3],xmm14[3]
+; SSE-NEXT:    movdqa %xmm11, %xmm12
+; SSE-NEXT:    movdqa %xmm11, %xmm7
+; SSE-NEXT:    pandn %xmm0, %xmm7
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm15, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm0[0,1,3,2,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm13[0,1,1,1]
+; SSE-NEXT:    pand %xmm11, %xmm13
+; SSE-NEXT:    por %xmm7, %xmm13
+; SSE-NEXT:    pand %xmm10, %xmm13
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm5[0,1,0,1]
+; SSE-NEXT:    movdqa %xmm10, %xmm9
+; SSE-NEXT:    pandn %xmm7, %xmm9
+; SSE-NEXT:    por %xmm13, %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm2, %xmm13
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm3[0],xmm13[1],xmm3[1],xmm13[2],xmm3[2],xmm13[3],xmm3[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm13[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,5,4,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm13[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm0[0],xmm13[1],xmm0[1]
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,0,65535,65535,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    pandn %xmm7, %xmm0
+; SSE-NEXT:    pand %xmm11, %xmm13
+; SSE-NEXT:    por %xmm13, %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm3[0,1,2,3,4,5,7,6]
+; SSE-NEXT:    pslldq {{.*#+}} xmm3 = zero,zero,xmm3[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT:    movdqa %xmm1, %xmm7
+; SSE-NEXT:    pandn %xmm3, %xmm7
+; SSE-NEXT:    movdqa %xmm15, %xmm3
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,2,3,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,2,2]
+; SSE-NEXT:    pand %xmm1, %xmm3
+; SSE-NEXT:    por %xmm7, %xmm3
+; SSE-NEXT:    pand %xmm14, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm5[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm14, %xmm9
+; SSE-NEXT:    pandn %xmm7, %xmm9
+; SSE-NEXT:    por %xmm3, %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    psrlq $48, %xmm4
+; SSE-NEXT:    movdqa %xmm15, %xmm9
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm9 = xmm9[1],xmm4[1]
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm9, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,3,3]
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    por %xmm3, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,0,65535,65535,65535,65535,0]
+; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    pandn %xmm7, %xmm3
+; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    movdqa %xmm2, %xmm9
+; SSE-NEXT:    por %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1],xmm0[2],xmm15[2],xmm0[3],xmm15[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,2,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
+; SSE-NEXT:    movdqa %xmm12, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm10, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1],xmm0[2],xmm13[2],xmm0[3],xmm13[3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm0[0,1,3,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,1,1,1]
-; SSE-NEXT:    pand %xmm4, %xmm7
-; SSE-NEXT:    por %xmm1, %xmm7
-; SSE-NEXT:    pand %xmm13, %xmm7
+; SSE-NEXT:    pand %xmm12, %xmm7
+; SSE-NEXT:    por %xmm3, %xmm7
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,65535,65535,65535,0,65535,65535,65535]
+; SSE-NEXT:    pand %xmm12, %xmm7
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,1,0,1]
-; SSE-NEXT:    movdqa %xmm13, %xmm11
-; SSE-NEXT:    pandn %xmm1, %xmm11
-; SSE-NEXT:    por %xmm7, %xmm11
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[0,1,0,1]
+; SSE-NEXT:    movdqa %xmm12, %xmm2
+; SSE-NEXT:    pandn %xmm3, %xmm2
+; SSE-NEXT:    por %xmm7, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm2, %xmm7
+; SSE-NEXT:    movdqa %xmm15, %xmm7
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm5[0],xmm7[1],xmm5[1],xmm7[2],xmm5[2],xmm7[3],xmm5[3]
+; SSE-NEXT:    movdqa %xmm5, %xmm2
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,5,4,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,2,2,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    pand %xmm15, %xmm7
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    pand %xmm11, %xmm7
 ; SSE-NEXT:    por %xmm7, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm2[4],xmm5[5],xmm2[5],xmm5[6],xmm2[6],xmm5[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm5, %xmm0
-; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    movdqa %xmm9, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm12, %xmm11
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    movdqa %xmm14, %xmm12
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm14[4],xmm0[5],xmm14[5],xmm0[6],xmm14[6],xmm0[7],xmm14[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,3,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,2]
-; SSE-NEXT:    pand %xmm9, %xmm0
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pand %xmm3, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm3, %xmm7
-; SSE-NEXT:    movdqa %xmm3, %xmm14
-; SSE-NEXT:    pandn %xmm1, %xmm7
-; SSE-NEXT:    por %xmm0, %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    psrlq $48, %xmm0
-; SSE-NEXT:    movdqa %xmm11, %xmm3
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm9, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,7,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,1,3,3]
-; SSE-NEXT:    pand %xmm9, %xmm5
-; SSE-NEXT:    por %xmm0, %xmm5
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    pand %xmm10, %xmm5
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm15[4],xmm2[5],xmm15[5],xmm2[6],xmm15[6],xmm2[7],xmm15[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm2[0,1,2,3,4,5,7,6]
+; SSE-NEXT:    pslldq {{.*#+}} xmm2 = zero,zero,xmm2[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    movdqa %xmm10, %xmm5
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm13[4],xmm5[5],xmm13[5],xmm5[6],xmm13[6],xmm5[7],xmm13[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,2,3,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,2]
+; SSE-NEXT:    pand %xmm1, %xmm5
+; SSE-NEXT:    por %xmm3, %xmm5
+; SSE-NEXT:    pand %xmm14, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm14, %xmm10
+; SSE-NEXT:    pandn %xmm3, %xmm10
+; SSE-NEXT:    por %xmm5, %xmm10
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm13, %xmm5
+; SSE-NEXT:    psrlq $48, %xmm5
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm5[1]
+; SSE-NEXT:    movdqa %xmm1, %xmm5
+; SSE-NEXT:    pandn %xmm2, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,3,3]
+; SSE-NEXT:    pand %xmm1, %xmm0
 ; SSE-NEXT:    por %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    pandn %xmm3, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm9, %xmm13
+; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1],xmm0[2],xmm12[2],xmm0[3],xmm12[3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1],xmm0[2],xmm9[2],xmm0[3],xmm9[3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,0,0,65535,65535,65535,0]
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
+; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm2, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1],xmm0[2],xmm11[2],xmm0[3],xmm11[3]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1],xmm0[2],xmm7[2],xmm0[3],xmm7[3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm0[0,1,3,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,1,1]
 ; SSE-NEXT:    pand %xmm4, %xmm5
-; SSE-NEXT:    por %xmm1, %xmm5
-; SSE-NEXT:    movdqa %xmm13, %xmm2
-; SSE-NEXT:    pand %xmm13, %xmm5
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[0,1,0,1]
-; SSE-NEXT:    movdqa %xmm2, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    por %xmm5, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm4, %xmm15
+; SSE-NEXT:    por %xmm3, %xmm5
+; SSE-NEXT:    pand %xmm12, %xmm5
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm10[0,1,0,1]
+; SSE-NEXT:    movdqa %xmm12, %xmm4
+; SSE-NEXT:    pandn %xmm3, %xmm4
+; SSE-NEXT:    por %xmm5, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm12, %xmm5
+; SSE-NEXT:    movdqa %xmm9, %xmm5
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,4,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,2,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    pand %xmm15, %xmm5
-; SSE-NEXT:    por %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm12[4],xmm6[5],xmm12[5],xmm6[6],xmm12[6],xmm6[7],xmm12[7]
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    movdqa %xmm9, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm3
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm11[4],xmm0[5],xmm11[5],xmm0[6],xmm11[6],xmm0[7],xmm11[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,3,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,2]
-; SSE-NEXT:    pand %xmm9, %xmm0
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pand %xmm14, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm14, %xmm7
-; SSE-NEXT:    pandn %xmm1, %xmm7
-; SSE-NEXT:    por %xmm0, %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    psrlq $48, %xmm0
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm9, %xmm0
 ; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm6[0,1,2,3,4,5,7,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,1,3,3]
-; SSE-NEXT:    pand %xmm9, %xmm5
-; SSE-NEXT:    por %xmm0, %xmm5
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    pand %xmm10, %xmm5
+; SSE-NEXT:    pand %xmm11, %xmm5
 ; SSE-NEXT:    por %xmm5, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm9[4],xmm6[5],xmm9[5],xmm6[6],xmm9[6],xmm6[7],xmm9[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm6[0,1,2,3,4,5,7,6]
+; SSE-NEXT:    pslldq {{.*#+}} xmm6 = zero,zero,xmm6[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm6, %xmm3
+; SSE-NEXT:    movdqa %xmm2, %xmm5
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm7[4],xmm5[5],xmm7[5],xmm5[6],xmm7[6],xmm5[7],xmm7[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,2,3,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,2]
+; SSE-NEXT:    pand %xmm1, %xmm5
+; SSE-NEXT:    por %xmm3, %xmm5
+; SSE-NEXT:    pand %xmm14, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm10[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm14, %xmm4
+; SSE-NEXT:    pandn %xmm3, %xmm4
+; SSE-NEXT:    por %xmm5, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm7, %xmm5
+; SSE-NEXT:    psrlq $48, %xmm5
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm5[1]
+; SSE-NEXT:    movdqa %xmm1, %xmm5
+; SSE-NEXT:    pandn %xmm2, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,3,3]
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    por %xmm5, %xmm0
+; SSE-NEXT:    movdqa %xmm13, %xmm2
+; SSE-NEXT:    pandn %xmm3, %xmm2
+; SSE-NEXT:    pand %xmm13, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm8, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1],xmm0[2],xmm7[2],xmm0[3],xmm7[3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm15, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm2, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm0[0,1,3,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,1,1]
-; SSE-NEXT:    pand %xmm4, %xmm5
-; SSE-NEXT:    por %xmm1, %xmm5
-; SSE-NEXT:    pand %xmm2, %xmm5
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[0,1,0,1]
-; SSE-NEXT:    movdqa %xmm2, %xmm12
-; SSE-NEXT:    pandn %xmm1, %xmm12
-; SSE-NEXT:    por %xmm5, %xmm12
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm15, %xmm5
+; SSE-NEXT:    movdqa %xmm15, %xmm4
+; SSE-NEXT:    por %xmm3, %xmm5
+; SSE-NEXT:    pand %xmm12, %xmm5
+; SSE-NEXT:    movdqa (%rsp), %xmm9 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[0,1,0,1]
+; SSE-NEXT:    movdqa %xmm12, %xmm10
+; SSE-NEXT:    pandn %xmm3, %xmm10
+; SSE-NEXT:    por %xmm5, %xmm10
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; SSE-NEXT:    movdqa %xmm7, %xmm5
@@ -4700,322 +4764,243 @@ define void @store_i16_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,4,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,2,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    pand %xmm15, %xmm5
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    pand %xmm11, %xmm5
 ; SSE-NEXT:    por %xmm5, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm7[4],xmm8[5],xmm7[5],xmm8[6],xmm7[6],xmm8[7],xmm7[7]
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    movdqa %xmm9, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm3, %xmm0
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm8[0,1,2,3,4,5,7,6]
+; SSE-NEXT:    pslldq {{.*#+}} xmm8 = zero,zero,xmm8[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm8, %xmm3
+; SSE-NEXT:    movdqa %xmm2, %xmm5
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm6[4],xmm5[5],xmm6[5],xmm5[6],xmm6[6],xmm5[7],xmm6[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,2,3,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,2]
+; SSE-NEXT:    pand %xmm1, %xmm5
+; SSE-NEXT:    por %xmm3, %xmm5
+; SSE-NEXT:    pand %xmm14, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm14, %xmm7
+; SSE-NEXT:    pandn %xmm3, %xmm7
+; SSE-NEXT:    por %xmm5, %xmm7
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm6, %xmm5
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,3,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,2]
-; SSE-NEXT:    pand %xmm9, %xmm0
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pand %xmm14, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm14, %xmm6
-; SSE-NEXT:    pandn %xmm1, %xmm6
-; SSE-NEXT:    por %xmm0, %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm5, %xmm0
-; SSE-NEXT:    psrlq $48, %xmm0
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm9, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm8[0,1,2,3,4,5,7,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,1,3,3]
-; SSE-NEXT:    pand %xmm9, %xmm5
-; SSE-NEXT:    por %xmm0, %xmm5
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    pand %xmm10, %xmm5
+; SSE-NEXT:    psrlq $48, %xmm5
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm5[1]
+; SSE-NEXT:    movdqa %xmm1, %xmm5
+; SSE-NEXT:    pandn %xmm2, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,3,3]
+; SSE-NEXT:    pand %xmm1, %xmm0
 ; SSE-NEXT:    por %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm13, %xmm2
+; SSE-NEXT:    pandn %xmm3, %xmm2
+; SSE-NEXT:    pand %xmm13, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm9, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1],xmm0[2],xmm7[2],xmm0[3],xmm7[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,2,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
+; SSE-NEXT:    movdqa %xmm15, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm2, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm0[0,1,3,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,1,1]
-; SSE-NEXT:    pand %xmm4, %xmm5
-; SSE-NEXT:    por %xmm1, %xmm5
-; SSE-NEXT:    pand %xmm2, %xmm5
-; SSE-NEXT:    movdqa (%rsp), %xmm8 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[0,1,0,1]
-; SSE-NEXT:    movdqa %xmm2, %xmm12
-; SSE-NEXT:    pandn %xmm1, %xmm12
-; SSE-NEXT:    por %xmm5, %xmm12
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm15, %xmm5
+; SSE-NEXT:    por %xmm3, %xmm5
+; SSE-NEXT:    pand %xmm12, %xmm5
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm8[0,1,0,1]
+; SSE-NEXT:    movdqa %xmm12, %xmm10
+; SSE-NEXT:    pandn %xmm3, %xmm10
+; SSE-NEXT:    por %xmm5, %xmm10
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm6, %xmm5
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm11[0],xmm5[1],xmm11[1],xmm5[2],xmm11[2],xmm5[3],xmm11[3]
+; SSE-NEXT:    movdqa %xmm7, %xmm5
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm9[0],xmm5[1],xmm9[1],xmm5[2],xmm9[2],xmm5[3],xmm9[3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,4,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,2,3]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    pand %xmm15, %xmm5
-; SSE-NEXT:    por %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm1, %xmm6
-; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    movdqa %xmm9, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    movdqa %xmm7, %xmm5
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,3,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,2]
-; SSE-NEXT:    pand %xmm9, %xmm0
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pand %xmm14, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm14, %xmm7
-; SSE-NEXT:    pandn %xmm1, %xmm7
-; SSE-NEXT:    por %xmm0, %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm5, %xmm0
-; SSE-NEXT:    psrlq $48, %xmm0
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm11, %xmm0
 ; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm6[0,1,2,3,4,5,7,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,3,3]
-; SSE-NEXT:    pand %xmm9, %xmm3
-; SSE-NEXT:    por %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    pand %xmm10, %xmm3
-; SSE-NEXT:    por %xmm3, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm7, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm3[0,1,3,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm0[0,1,1,1]
-; SSE-NEXT:    pand %xmm4, %xmm5
-; SSE-NEXT:    por %xmm1, %xmm5
-; SSE-NEXT:    pand %xmm2, %xmm5
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[0,1,0,1]
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm0
+; SSE-NEXT:    pand %xmm11, %xmm5
+; SSE-NEXT:    movdqa %xmm11, %xmm10
 ; SSE-NEXT:    por %xmm5, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm6, %xmm5
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm12[0],xmm5[1],xmm12[1],xmm5[2],xmm12[2],xmm5[3],xmm12[3]
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,4,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
-; SSE-NEXT:    movdqa %xmm15, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    pand %xmm15, %xmm5
-; SSE-NEXT:    movdqa %xmm15, %xmm13
-; SSE-NEXT:    por %xmm5, %xmm3
-; SSE-NEXT:    movdqa %xmm3, (%rsp) # 16-byte Spill
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm0, %xmm6
-; SSE-NEXT:    pslldq {{.*#+}} xmm1 = zero,zero,xmm1[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm9, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm8[4],xmm1[5],xmm8[5],xmm1[6],xmm8[6],xmm1[7],xmm8[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,2,3,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,2]
-; SSE-NEXT:    pand %xmm9, %xmm1
-; SSE-NEXT:    por %xmm3, %xmm1
-; SSE-NEXT:    pand %xmm14, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[2,3,2,3]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm7[4],xmm3[5],xmm7[5],xmm3[6],xmm7[6],xmm3[7],xmm7[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm3[0,1,2,3,4,5,7,6]
+; SSE-NEXT:    pslldq {{.*#+}} xmm3 = zero,zero,xmm3[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT:    movdqa %xmm3, %xmm5
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm5, %xmm3
+; SSE-NEXT:    movdqa %xmm2, %xmm5
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm6[4],xmm5[5],xmm6[5],xmm5[6],xmm6[6],xmm5[7],xmm6[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,2,3,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,2]
+; SSE-NEXT:    pand %xmm1, %xmm5
+; SSE-NEXT:    por %xmm3, %xmm5
+; SSE-NEXT:    pand %xmm14, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm8[2,3,2,3]
 ; SSE-NEXT:    movdqa %xmm14, %xmm7
 ; SSE-NEXT:    pandn %xmm3, %xmm7
-; SSE-NEXT:    por %xmm1, %xmm7
+; SSE-NEXT:    por %xmm5, %xmm7
 ; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm8, %xmm1
-; SSE-NEXT:    psrlq $48, %xmm1
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; SSE-NEXT:    movdqa %xmm9, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm6[0,1,2,3,4,5,7,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm5[2,1,3,3]
-; SSE-NEXT:    pand %xmm9, %xmm8
-; SSE-NEXT:    por %xmm1, %xmm8
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    pand %xmm10, %xmm8
-; SSE-NEXT:    por %xmm8, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,2,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,2,1]
-; SSE-NEXT:    movdqa %xmm4, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm5, %xmm8
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm12[0],xmm8[1],xmm12[1],xmm8[2],xmm12[2],xmm8[3],xmm12[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm8[0,1,3,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm14 = xmm1[0,1,1,1]
-; SSE-NEXT:    pand %xmm4, %xmm14
-; SSE-NEXT:    por %xmm3, %xmm14
-; SSE-NEXT:    pand %xmm2, %xmm14
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[0,1,0,1]
-; SSE-NEXT:    movdqa %xmm2, %xmm15
-; SSE-NEXT:    pandn %xmm3, %xmm15
-; SSE-NEXT:    por %xmm14, %xmm15
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,5,6,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm0, %xmm14
-; SSE-NEXT:    movdqa %xmm0, %xmm6
-; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm11[0],xmm14[1],xmm11[1],xmm14[2],xmm11[2],xmm14[3],xmm11[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm14[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,5,4,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
-; SSE-NEXT:    movdqa %xmm13, %xmm14
-; SSE-NEXT:    pandn %xmm3, %xmm14
+; SSE-NEXT:    movdqa %xmm6, %xmm5
+; SSE-NEXT:    psrlq $48, %xmm5
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm5[1]
+; SSE-NEXT:    movdqa %xmm1, %xmm5
+; SSE-NEXT:    pandn %xmm2, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,3,3]
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    por %xmm5, %xmm0
+; SSE-NEXT:    movdqa %xmm13, %xmm2
+; SSE-NEXT:    pandn %xmm3, %xmm2
 ; SSE-NEXT:    pand %xmm13, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm14
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm1, %xmm6
+; SSE-NEXT:    movdqa %xmm13, %xmm15
+; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1],xmm0[2],xmm5[2],xmm0[3],xmm5[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,1,2,2,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
+; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm2, %xmm7
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm7[0,1,3,2,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm0[0,1,1,1]
+; SSE-NEXT:    pand %xmm4, %xmm13
+; SSE-NEXT:    por %xmm3, %xmm13
+; SSE-NEXT:    movdqa %xmm12, %xmm11
+; SSE-NEXT:    pand %xmm12, %xmm13
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm8[0,1,0,1]
+; SSE-NEXT:    pandn %xmm3, %xmm11
+; SSE-NEXT:    por %xmm13, %xmm11
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm5, %xmm13
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm9[0],xmm13[1],xmm9[1],xmm13[2],xmm9[2],xmm13[3],xmm9[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm13[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,5,4,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
+; SSE-NEXT:    movdqa %xmm10, %xmm13
+; SSE-NEXT:    pandn %xmm3, %xmm13
+; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm13
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm5[4],xmm0[5],xmm5[5],xmm0[6],xmm5[6],xmm0[7],xmm5[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm0[0,1,2,3,4,5,7,6]
 ; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    movdqa %xmm1, %xmm3
 ; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm5, %xmm0
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm12[4],xmm0[5],xmm12[5],xmm0[6],xmm12[6],xmm0[7],xmm12[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,3,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,2,2]
-; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm2, %xmm7
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm6[4],xmm7[5],xmm6[5],xmm7[6],xmm6[6],xmm7[7],xmm6[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm7[0,2,3,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[0,1,2,2]
+; SSE-NEXT:    pand %xmm1, %xmm0
 ; SSE-NEXT:    por %xmm3, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [0,65535,65535,65535,65535,0,65535,65535]
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[2,3,2,3]
-; SSE-NEXT:    pandn %xmm1, %xmm11
-; SSE-NEXT:    por %xmm0, %xmm11
-; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    pand %xmm14, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm8[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm14, %xmm12
+; SSE-NEXT:    pandn %xmm3, %xmm12
+; SSE-NEXT:    por %xmm0, %xmm12
+; SSE-NEXT:    movdqa %xmm6, %xmm0
 ; SSE-NEXT:    psrlq $48, %xmm0
-; SSE-NEXT:    movdqa %xmm5, %xmm3
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm9, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm6[0,1,2,3,4,5,7,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,3,3]
-; SSE-NEXT:    pand %xmm9, %xmm3
-; SSE-NEXT:    por %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm10, %xmm12
-; SSE-NEXT:    pandn %xmm1, %xmm12
-; SSE-NEXT:    pand %xmm10, %xmm3
-; SSE-NEXT:    por %xmm3, %xmm12
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,2,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,2,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm8, %xmm3
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm0[1]
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    pandn %xmm2, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,1,3,3]
+; SSE-NEXT:    pand %xmm1, %xmm5
+; SSE-NEXT:    por %xmm0, %xmm5
+; SSE-NEXT:    movdqa %xmm15, %xmm10
+; SSE-NEXT:    pandn %xmm3, %xmm10
+; SSE-NEXT:    pand %xmm15, %xmm5
+; SSE-NEXT:    por %xmm5, %xmm10
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1],xmm3[2],xmm6[2],xmm3[3],xmm6[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm3[0,1,3,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm13[0,1,1,1]
-; SSE-NEXT:    pand %xmm4, %xmm13
-; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    por %xmm13, %xmm4
-; SSE-NEXT:    pand %xmm2, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[0,1,0,1]
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    por %xmm4, %xmm2
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm7, %xmm4
-; SSE-NEXT:    movdqa %xmm0, %xmm13
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,4,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,0,65535,65535,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm6, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,2,2,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,0,2,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm8, %xmm5
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm9[0],xmm5[1],xmm9[1],xmm5[2],xmm9[2],xmm5[3],xmm9[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm5[0,1,3,2,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm15[0,1,1,1]
+; SSE-NEXT:    pand %xmm4, %xmm15
+; SSE-NEXT:    pandn %xmm3, %xmm4
+; SSE-NEXT:    por %xmm15, %xmm4
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,0,65535,65535,65535]
 ; SSE-NEXT:    pand %xmm0, %xmm4
-; SSE-NEXT:    pandn %xmm1, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[0,1,0,1]
+; SSE-NEXT:    pandn %xmm3, %xmm0
 ; SSE-NEXT:    por %xmm4, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm3
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm7[4],xmm3[5],xmm7[5],xmm3[6],xmm7[6],xmm3[7],xmm7[7]
-; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    movdqa %xmm3, %xmm7
-; SSE-NEXT:    pslldq {{.*#+}} xmm1 = zero,zero,xmm1[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    movdqa %xmm9, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm8, %xmm1
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,2,3,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,2,2]
-; SSE-NEXT:    pand %xmm9, %xmm1
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm5[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
+; SSE-NEXT:    movdqa %xmm2, %xmm15
+; SSE-NEXT:    movdqa %xmm2, %xmm5
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,4,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,2,3]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,0,65535,65535,65535,65535,0,65535]
+; SSE-NEXT:    pand %xmm2, %xmm5
+; SSE-NEXT:    pandn %xmm3, %xmm2
+; SSE-NEXT:    por %xmm5, %xmm2
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm15[4],xmm6[5],xmm15[5],xmm6[6],xmm15[6],xmm6[7],xmm15[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm6[0,1,2,3,4,5,7,6]
+; SSE-NEXT:    pslldq {{.*#+}} xmm6 = zero,zero,xmm6[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT:    movdqa %xmm1, %xmm4
+; SSE-NEXT:    pandn %xmm6, %xmm4
+; SSE-NEXT:    movdqa %xmm8, %xmm5
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm9[4],xmm5[5],xmm9[5],xmm5[6],xmm9[6],xmm5[7],xmm9[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm5[0,2,3,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,1,2,2]
+; SSE-NEXT:    pand %xmm1, %xmm5
+; SSE-NEXT:    por %xmm4, %xmm5
+; SSE-NEXT:    pand %xmm14, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm7[2,3,2,3]
+; SSE-NEXT:    pandn %xmm4, %xmm14
+; SSE-NEXT:    por %xmm5, %xmm14
+; SSE-NEXT:    psrlq $48, %xmm9
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm8 = xmm8[1],xmm9[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,3,3]
+; SSE-NEXT:    pand %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm8, %xmm1
 ; SSE-NEXT:    por %xmm3, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [0,65535,65535,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,0,65535,65535,65535,65535,0]
 ; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm5[2,3,2,3]
-; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    pandn %xmm4, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    psrlq $48, %xmm6
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm8 = xmm8[1],xmm6[1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm7[0,1,2,3,4,5,7,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,3,3]
-; SSE-NEXT:    pand %xmm9, %xmm1
-; SSE-NEXT:    pandn %xmm8, %xmm9
-; SSE-NEXT:    por %xmm1, %xmm9
-; SSE-NEXT:    pand %xmm10, %xmm9
-; SSE-NEXT:    pandn %xmm3, %xmm10
-; SSE-NEXT:    por %xmm9, %xmm10
-; SSE-NEXT:    movdqa %xmm10, 624(%r9)
-; SSE-NEXT:    movdqa %xmm0, 608(%r9)
+; SSE-NEXT:    movdqa %xmm0, 624(%r9)
+; SSE-NEXT:    movdqa %xmm14, 608(%r9)
+; SSE-NEXT:    movdqa %xmm2, 576(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 576(%r9)
-; SSE-NEXT:    movdqa %xmm2, 560(%r9)
-; SSE-NEXT:    movdqa %xmm12, 544(%r9)
-; SSE-NEXT:    movdqa %xmm11, 528(%r9)
-; SSE-NEXT:    movdqa %xmm14, 496(%r9)
-; SSE-NEXT:    movdqa %xmm15, 480(%r9)
+; SSE-NEXT:    movaps %xmm0, 560(%r9)
+; SSE-NEXT:    movdqa %xmm10, 544(%r9)
+; SSE-NEXT:    movdqa %xmm12, 528(%r9)
+; SSE-NEXT:    movdqa %xmm13, 496(%r9)
+; SSE-NEXT:    movdqa %xmm11, 480(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 464(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll
index 79cc8e49f1fdb..a9d45ceb45aae 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll
@@ -360,77 +360,77 @@ define void @store_i16_stride7_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %r10
-; SSE-NEXT:    movq {{.*#+}} xmm3 = mem[0],zero
 ; SSE-NEXT:    movq {{.*#+}} xmm5 = mem[0],zero
+; SSE-NEXT:    movq {{.*#+}} xmm6 = mem[0],zero
 ; SSE-NEXT:    movq {{.*#+}} xmm0 = mem[0],zero
 ; SSE-NEXT:    movq {{.*#+}} xmm4 = mem[0],zero
 ; SSE-NEXT:    movq {{.*#+}} xmm1 = mem[0],zero
-; SSE-NEXT:    movq {{.*#+}} xmm6 = mem[0],zero
+; SSE-NEXT:    movq {{.*#+}} xmm3 = mem[0],zero
 ; SSE-NEXT:    movq {{.*#+}} xmm2 = mem[0],zero
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm6[0]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm1[0,1,0,1]
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,65535,65535,65535,0,65535,65535]
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[0,1,0,1]
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,65535,65535,65535,0,65535,65535]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm2[0,0,2,3]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pand %xmm6, %xmm8
-; SSE-NEXT:    pandn %xmm7, %xmm6
-; SSE-NEXT:    por %xmm8, %xmm6
+; SSE-NEXT:    pand %xmm7, %xmm8
+; SSE-NEXT:    pandn %xmm3, %xmm7
+; SSE-NEXT:    por %xmm8, %xmm7
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,0,0,0,65535,65535]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm4[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm8 = xmm8[0],xmm3[0]
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,0,65535,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm5, %xmm10
-; SSE-NEXT:    movdqa %xmm3, %xmm9
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm3[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1],xmm3[2],xmm5[2],xmm3[3],xmm5[3]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[1,1,2,3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm5[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm7, %xmm12
-; SSE-NEXT:    pandn %xmm8, %xmm7
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    por %xmm12, %xmm7
-; SSE-NEXT:    pand %xmm5, %xmm7
-; SSE-NEXT:    pandn %xmm6, %xmm5
-; SSE-NEXT:    por %xmm7, %xmm5
-; SSE-NEXT:    psrld $16, %xmm10
-; SSE-NEXT:    punpckldq {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1]
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,65535,65535,65535,0,0,65535]
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm8 = xmm8[0],xmm5[0]
+; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,0,65535,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm6, %xmm10
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm0[0],xmm10[1],xmm0[1],xmm10[2],xmm0[2],xmm10[3],xmm0[3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[1,1,2,3]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    pand %xmm9, %xmm10
+; SSE-NEXT:    pandn %xmm8, %xmm9
+; SSE-NEXT:    por %xmm10, %xmm9
+; SSE-NEXT:    pand %xmm3, %xmm9
+; SSE-NEXT:    pandn %xmm7, %xmm3
+; SSE-NEXT:    por %xmm9, %xmm3
+; SSE-NEXT:    movdqa %xmm5, %xmm7
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3]
+; SSE-NEXT:    psrld $16, %xmm6
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm5[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,65535,65535,65535,0,0,65535]
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm0[0,1,2,3,4,5,6,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,1,2,3]
-; SSE-NEXT:    pand %xmm7, %xmm6
-; SSE-NEXT:    pandn %xmm9, %xmm7
-; SSE-NEXT:    por %xmm6, %xmm7
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,0,0,0,65535,65535,65535]
-; SSE-NEXT:    pand %xmm6, %xmm7
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm2[3,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm8[0,1,2,0,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm9 = xmm9[0],xmm10[0]
-; SSE-NEXT:    pandn %xmm9, %xmm6
-; SSE-NEXT:    por %xmm7, %xmm6
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,65535,0,65535,65535,65535,65535]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[0,3,1,3,4,5,6,7]
-; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm8
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm0[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[2,1,2,3]
+; SSE-NEXT:    pand %xmm6, %xmm9
+; SSE-NEXT:    pandn %xmm5, %xmm6
+; SSE-NEXT:    por %xmm9, %xmm6
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,0,0,0,65535,65535,65535]
+; SSE-NEXT:    pand %xmm5, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm2[3,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm9[0,1,2,0,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm1[1,1,1,1]
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm10 = xmm10[0],xmm11[0]
+; SSE-NEXT:    pandn %xmm10, %xmm5
+; SSE-NEXT:    por %xmm6, %xmm5
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,65535,0,65535,65535,65535,65535]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm9[0,3,1,3,4,5,6,7]
+; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm9
 ; SSE-NEXT:    psrlq $48, %xmm4
-; SSE-NEXT:    por %xmm8, %xmm4
-; SSE-NEXT:    pand %xmm7, %xmm4
-; SSE-NEXT:    pandn %xmm1, %xmm7
-; SSE-NEXT:    por %xmm4, %xmm7
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm11[2,1]
-; SSE-NEXT:    movaps {{.*#+}} xmm3 = [65535,65535,65535,65535,0,0,0,65535]
-; SSE-NEXT:    andps %xmm3, %xmm0
+; SSE-NEXT:    por %xmm9, %xmm4
+; SSE-NEXT:    pand %xmm6, %xmm4
+; SSE-NEXT:    pandn %xmm1, %xmm6
+; SSE-NEXT:    por %xmm4, %xmm6
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm7[0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm8[2,1]
+; SSE-NEXT:    movaps {{.*#+}} xmm4 = [65535,65535,65535,65535,0,0,0,65535]
+; SSE-NEXT:    andps %xmm4, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,0,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    andnps %xmm2, %xmm3
-; SSE-NEXT:    orps %xmm0, %xmm3
-; SSE-NEXT:    movaps %xmm3, (%rax)
-; SSE-NEXT:    movq %xmm7, 48(%rax)
-; SSE-NEXT:    movdqa %xmm6, 32(%rax)
-; SSE-NEXT:    movdqa %xmm5, 16(%rax)
+; SSE-NEXT:    andnps %xmm2, %xmm4
+; SSE-NEXT:    orps %xmm0, %xmm4
+; SSE-NEXT:    movaps %xmm4, (%rax)
+; SSE-NEXT:    movq %xmm6, 48(%rax)
+; SSE-NEXT:    movdqa %xmm5, 32(%rax)
+; SSE-NEXT:    movdqa %xmm3, 16(%rax)
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: store_i16_stride7_vf4:
@@ -906,145 +906,143 @@ define void @store_i16_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-LABEL: store_i16_stride7_vf8:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movdqa (%rdi), %xmm3
+; SSE-NEXT:    movdqa (%rdi), %xmm7
 ; SSE-NEXT:    movdqa (%rsi), %xmm8
-; SSE-NEXT:    movdqa (%rdx), %xmm5
+; SSE-NEXT:    movdqa (%rdx), %xmm0
 ; SSE-NEXT:    movdqa (%rcx), %xmm11
-; SSE-NEXT:    movdqa (%r8), %xmm4
+; SSE-NEXT:    movdqa (%r8), %xmm3
 ; SSE-NEXT:    movdqa (%r9), %xmm10
 ; SSE-NEXT:    movdqa (%rax), %xmm2
-; SSE-NEXT:    movdqa %xmm5, %xmm0
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm11[4],xmm0[5],xmm11[5],xmm0[6],xmm11[6],xmm0[7],xmm11[7]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm6
+; SSE-NEXT:    movdqa %xmm0, %xmm5
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm11[4],xmm5[5],xmm11[5],xmm5[6],xmm11[6],xmm5[7],xmm11[7]
+; SSE-NEXT:    movdqa %xmm7, %xmm6
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm8[4],xmm6[5],xmm8[5],xmm6[6],xmm8[6],xmm6[7],xmm8[7]
 ; SSE-NEXT:    movdqa %xmm6, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,2],xmm0[2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm2[2,3,2,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm4, %xmm7
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm7 = xmm7[0],xmm10[0],xmm7[1],xmm10[1],xmm7[2],xmm10[2],xmm7[3],xmm10[3]
-; SSE-NEXT:    movdqa %xmm10, %xmm13
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm13 = xmm13[4],xmm4[4],xmm13[5],xmm4[5],xmm13[6],xmm4[6],xmm13[7],xmm4[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm13 = xmm13[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm13 = xmm13[0,1,2,3,4,5,5,4]
-; SSE-NEXT:    pand %xmm12, %xmm13
-; SSE-NEXT:    pandn %xmm9, %xmm12
-; SSE-NEXT:    por %xmm13, %xmm12
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,2],xmm12[0,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,65535,0,65535,65535,65535,65535]
-; SSE-NEXT:    pandn %xmm2, %xmm9
-; SSE-NEXT:    movdqa %xmm7, %xmm12
-; SSE-NEXT:    movdqa %xmm7, %xmm13
-; SSE-NEXT:    psrldq {{.*#+}} xmm13 = xmm13[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    por %xmm9, %xmm13
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm5[2,2,2,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,2],xmm5[2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[2,3,2,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,0,65535,65535,65535,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm10, %xmm12
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm12 = xmm12[4],xmm3[4],xmm12[5],xmm3[5],xmm12[6],xmm3[6],xmm12[7],xmm3[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm12 = xmm12[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm12[0,1,2,3,4,5,5,4]
+; SSE-NEXT:    pand %xmm9, %xmm12
+; SSE-NEXT:    pandn %xmm4, %xmm9
+; SSE-NEXT:    por %xmm12, %xmm9
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,2],xmm9[0,3]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,0,65535,65535,65535,65535]
+; SSE-NEXT:    pandn %xmm2, %xmm4
+; SSE-NEXT:    movdqa %xmm3, %xmm9
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm10[0],xmm9[1],xmm10[1],xmm9[2],xmm10[2],xmm9[3],xmm10[3]
+; SSE-NEXT:    movdqa %xmm9, %xmm12
+; SSE-NEXT:    psrldq {{.*#+}} xmm12 = xmm12[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    por %xmm4, %xmm12
+; SSE-NEXT:    movaps {{.*#+}} xmm4 = [65535,0,0,0,65535,65535,65535,65535]
+; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm0[2,2,2,2]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [65535,65535,65535,65535,65535,65535,0,65535]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm11[0,1,2,3,4,4,4,4]
 ; SSE-NEXT:    pand %xmm14, %xmm15
-; SSE-NEXT:    pandn %xmm9, %xmm14
-; SSE-NEXT:    movaps {{.*#+}} xmm9 = [65535,0,0,0,65535,65535,65535,65535]
+; SSE-NEXT:    pandn %xmm13, %xmm14
 ; SSE-NEXT:    por %xmm15, %xmm14
-; SSE-NEXT:    movdqa %xmm6, %xmm15
-; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm14[3,3]
+; SSE-NEXT:    movdqa %xmm6, %xmm13
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,1],xmm14[3,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm11[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm15[0,2]
-; SSE-NEXT:    andps %xmm9, %xmm14
-; SSE-NEXT:    andnps %xmm13, %xmm9
-; SSE-NEXT:    orps %xmm14, %xmm9
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm13[0,2]
+; SSE-NEXT:    andps %xmm4, %xmm14
+; SSE-NEXT:    andnps %xmm12, %xmm4
+; SSE-NEXT:    orps %xmm14, %xmm4
 ; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm2[0,1,0,1]
-; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [65535,65535,65535,65535,65535,0,65535,65535]
-; SSE-NEXT:    pslldq {{.*#+}} xmm12 = zero,zero,xmm12[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    pand %xmm14, %xmm12
-; SSE-NEXT:    pandn %xmm13, %xmm14
-; SSE-NEXT:    por %xmm12, %xmm14
-; SSE-NEXT:    movdqa %xmm11, %xmm12
-; SSE-NEXT:    psrld $16, %xmm12
-; SSE-NEXT:    movdqa %xmm5, %xmm13
-; SSE-NEXT:    punpckldq {{.*#+}} xmm13 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,0,0,65535,65535,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm8, %xmm15
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm3[0],xmm15[1],xmm3[1],xmm15[2],xmm3[2],xmm15[3],xmm3[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm15[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm15 = xmm15[0,1,2,3,4,5,5,4]
-; SSE-NEXT:    pand %xmm12, %xmm15
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,65535,65535,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm9, %xmm14
+; SSE-NEXT:    pslldq {{.*#+}} xmm14 = zero,zero,xmm14[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT:    pand %xmm12, %xmm14
 ; SSE-NEXT:    pandn %xmm13, %xmm12
-; SSE-NEXT:    movdqa %xmm5, %xmm13
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm11[0],xmm5[1],xmm11[1],xmm5[2],xmm11[2],xmm5[3],xmm11[3]
-; SSE-NEXT:    por %xmm15, %xmm12
-; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    pand %xmm15, %xmm12
-; SSE-NEXT:    pandn %xmm14, %xmm15
-; SSE-NEXT:    movdqa %xmm4, %xmm14
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm14 = xmm14[4],xmm10[4],xmm14[5],xmm10[5],xmm14[6],xmm10[6],xmm14[7],xmm10[7]
-; SSE-NEXT:    por %xmm12, %xmm15
+; SSE-NEXT:    por %xmm14, %xmm12
+; SSE-NEXT:    movdqa %xmm11, %xmm13
+; SSE-NEXT:    psrld $16, %xmm13
+; SSE-NEXT:    movdqa %xmm0, %xmm15
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm13[0],xmm15[1],xmm13[1]
+; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [65535,0,0,65535,65535,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm8, %xmm14
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm7[0],xmm14[1],xmm7[1],xmm14[2],xmm7[2],xmm14[3],xmm7[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm14 = xmm14[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm14 = xmm14[0,1,2,3,4,5,5,4]
+; SSE-NEXT:    pand %xmm13, %xmm14
+; SSE-NEXT:    pandn %xmm15, %xmm13
+; SSE-NEXT:    movdqa %xmm0, %xmm15
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1],xmm0[2],xmm11[2],xmm0[3],xmm11[3]
+; SSE-NEXT:    por %xmm14, %xmm13
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT:    pand %xmm1, %xmm13
+; SSE-NEXT:    pandn %xmm12, %xmm1
+; SSE-NEXT:    por %xmm13, %xmm1
 ; SSE-NEXT:    psrlq $48, %xmm11
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm13 = xmm13[1],xmm11[1]
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm15 = xmm15[1],xmm11[1]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,65535,65535,0,0,65535,65535,65535]
-; SSE-NEXT:    pandn %xmm13, %xmm11
-; SSE-NEXT:    movdqa %xmm6, %xmm13
-; SSE-NEXT:    psrldq {{.*#+}} xmm13 = xmm13[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    por %xmm11, %xmm13
+; SSE-NEXT:    pandn %xmm15, %xmm11
+; SSE-NEXT:    movdqa %xmm6, %xmm14
+; SSE-NEXT:    psrldq {{.*#+}} xmm14 = xmm14[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    por %xmm11, %xmm14
+; SSE-NEXT:    movdqa %xmm3, %xmm13
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm13 = xmm13[4],xmm10[4],xmm13[5],xmm10[5],xmm13[6],xmm10[6],xmm13[7],xmm10[7]
 ; SSE-NEXT:    psrld $16, %xmm10
-; SSE-NEXT:    movdqa %xmm4, %xmm11
+; SSE-NEXT:    movdqa %xmm3, %xmm11
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm10[2],xmm11[3],xmm10[3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,0,0,65535]
-; SSE-NEXT:    movdqa %xmm0, %xmm12
-; SSE-NEXT:    pandn %xmm11, %xmm12
-; SSE-NEXT:    por %xmm13, %xmm12
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,65535,65535,65535,65535,0,0,65535]
+; SSE-NEXT:    movdqa %xmm12, %xmm15
+; SSE-NEXT:    pandn %xmm11, %xmm15
+; SSE-NEXT:    por %xmm14, %xmm15
 ; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [0,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT:    pand %xmm10, %xmm12
+; SSE-NEXT:    pand %xmm10, %xmm15
 ; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm2[3,3,3,3]
 ; SSE-NEXT:    pandn %xmm11, %xmm10
-; SSE-NEXT:    por %xmm12, %xmm10
-; SSE-NEXT:    movdqa %xmm3, %xmm12
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm3[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm8[0],xmm3[1],xmm8[1],xmm3[2],xmm8[2],xmm3[3],xmm8[3]
+; SSE-NEXT:    por %xmm15, %xmm10
+; SSE-NEXT:    movdqa %xmm7, %xmm14
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm8[0],xmm14[1],xmm8[1],xmm14[2],xmm8[2],xmm14[3],xmm8[3]
 ; SSE-NEXT:    psrld $16, %xmm8
-; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm8[0],xmm12[1],xmm8[1]
-; SSE-NEXT:    movdqa %xmm0, %xmm8
-; SSE-NEXT:    pandn %xmm12, %xmm8
-; SSE-NEXT:    pshufhw {{.*#+}} xmm12 = xmm5[0,1,2,3,4,5,6,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm12[2,1,2,3]
-; SSE-NEXT:    pand %xmm0, %xmm12
-; SSE-NEXT:    por %xmm8, %xmm12
-; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,65535,0,0,0,65535,65535,65535]
-; SSE-NEXT:    pand %xmm8, %xmm12
-; SSE-NEXT:    movdqa %xmm7, %xmm13
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[2,2],xmm2[1,1]
-; SSE-NEXT:    pandn %xmm13, %xmm8
-; SSE-NEXT:    por %xmm12, %xmm8
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    pslldq {{.*#+}} xmm12 = zero,zero,zero,zero,zero,zero,xmm12[0,1,2,3,4,5,6,7,8,9]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm15 = xmm7[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
+; SSE-NEXT:    movdqa %xmm12, %xmm8
+; SSE-NEXT:    pandn %xmm7, %xmm8
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm0[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm7[2,1,2,3]
+; SSE-NEXT:    pand %xmm12, %xmm11
+; SSE-NEXT:    por %xmm8, %xmm11
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,0,0,0,65535,65535,65535]
+; SSE-NEXT:    pand %xmm7, %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm9[0,0,1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[2,2],xmm2[1,1]
+; SSE-NEXT:    pandn %xmm9, %xmm7
+; SSE-NEXT:    por %xmm11, %xmm7
+; SSE-NEXT:    pslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,xmm5[0,1,2,3,4,5,6,7,8,9]
 ; SSE-NEXT:    pslldq {{.*#+}} xmm6 = zero,zero,xmm6[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    pand %xmm0, %xmm6
-; SSE-NEXT:    pandn %xmm12, %xmm0
-; SSE-NEXT:    por %xmm6, %xmm0
-; SSE-NEXT:    movaps %xmm2, %xmm6
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,3],xmm14[0,1]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4,4,5,5,6,6,7,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0],xmm4[2,1]
-; SSE-NEXT:    movaps {{.*#+}} xmm4 = [65535,65535,65535,0,0,0,0,65535]
-; SSE-NEXT:    andps %xmm4, %xmm6
-; SSE-NEXT:    andnps %xmm0, %xmm4
-; SSE-NEXT:    orps %xmm6, %xmm4
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm3[0]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm11[2,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[0,0,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    pand %xmm12, %xmm6
+; SSE-NEXT:    pandn %xmm5, %xmm12
+; SSE-NEXT:    por %xmm6, %xmm12
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm2[0,0,1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,3],xmm13[0,1]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm3[2,1]
+; SSE-NEXT:    movaps {{.*#+}} xmm3 = [65535,65535,65535,0,0,0,0,65535]
+; SSE-NEXT:    andps %xmm3, %xmm2
+; SSE-NEXT:    andnps %xmm12, %xmm3
+; SSE-NEXT:    orps %xmm2, %xmm3
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm14[0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm15[2,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm5[0],xmm8[1],xmm5[1]
 ; SSE-NEXT:    movaps {{.*#+}} xmm2 = [65535,65535,65535,65535,0,0,0,65535]
-; SSE-NEXT:    andps %xmm2, %xmm5
-; SSE-NEXT:    andnps %xmm0, %xmm2
-; SSE-NEXT:    orps %xmm5, %xmm2
+; SSE-NEXT:    andps %xmm2, %xmm0
+; SSE-NEXT:    andnps %xmm8, %xmm2
+; SSE-NEXT:    orps %xmm0, %xmm2
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; SSE-NEXT:    movaps %xmm2, (%rax)
-; SSE-NEXT:    movaps %xmm4, 64(%rax)
-; SSE-NEXT:    movdqa %xmm15, 16(%rax)
-; SSE-NEXT:    movdqa %xmm8, 32(%rax)
-; SSE-NEXT:    movaps %xmm9, 48(%rax)
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0,1,3]
-; SSE-NEXT:    movaps %xmm1, 80(%rax)
+; SSE-NEXT:    movaps %xmm3, 64(%rax)
+; SSE-NEXT:    movdqa %xmm1, 16(%rax)
+; SSE-NEXT:    movdqa %xmm7, 32(%rax)
+; SSE-NEXT:    movaps %xmm4, 48(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0,1,3]
+; SSE-NEXT:    movaps %xmm0, 80(%rax)
 ; SSE-NEXT:    movdqa %xmm10, 96(%rax)
 ; SSE-NEXT:    retq
 ;
@@ -1706,333 +1704,335 @@ define void @store_i16_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 define void @store_i16_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %in.vecptr3, ptr %in.vecptr4, ptr %in.vecptr5, ptr %in.vecptr6, ptr %out.vec) nounwind {
 ; SSE-LABEL: store_i16_stride7_vf16:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $216, %rsp
+; SSE-NEXT:    subq $232, %rsp
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movdqa 16(%rdi), %xmm6
-; SSE-NEXT:    movdqa 16(%rsi), %xmm4
-; SSE-NEXT:    movdqa 16(%rdx), %xmm15
+; SSE-NEXT:    movdqa 16(%rdi), %xmm5
+; SSE-NEXT:    movdqa 16(%rsi), %xmm6
+; SSE-NEXT:    movdqa 16(%rdx), %xmm2
 ; SSE-NEXT:    movdqa 16(%rcx), %xmm1
-; SSE-NEXT:    movdqa 16(%r8), %xmm8
-; SSE-NEXT:    movdqa 16(%r9), %xmm7
-; SSE-NEXT:    movdqa 16(%rax), %xmm3
+; SSE-NEXT:    movdqa 16(%r8), %xmm4
+; SSE-NEXT:    movdqa 16(%r9), %xmm8
+; SSE-NEXT:    movdqa 16(%rax), %xmm10
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm1, %xmm12
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    psrlq $48, %xmm0
-; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm2, %xmm13
 ; SSE-NEXT:    punpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm0[1]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,0,65535,65535,65535]
 ; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    movdqa %xmm6, %xmm10
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7]
+; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    movdqa %xmm5, %xmm9
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm4, %xmm9
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm6, %xmm11
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[2,2,3,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,65535,65535,0,65535,65535]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[2,2,3,3]
 ; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm3, %xmm4
+; SSE-NEXT:    pandn %xmm1, %xmm4
+; SSE-NEXT:    por %xmm0, %xmm4
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1],xmm1[2],xmm8[2],xmm1[3],xmm8[3]
+; SSE-NEXT:    movdqa %xmm1, %xmm6
 ; SSE-NEXT:    movdqa %xmm8, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm6
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3]
-; SSE-NEXT:    movdqa %xmm1, %xmm7
-; SSE-NEXT:    movdqa %xmm6, %xmm1
 ; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,0],xmm2[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,0],xmm4[2,0]
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
-; SSE-NEXT:    movaps {{.*#+}} xmm6 = [0,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT:    andps %xmm6, %xmm0
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[3,3,3,3]
-; SSE-NEXT:    andnps %xmm1, %xmm6
-; SSE-NEXT:    orps %xmm0, %xmm6
-; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[0,1,0,1]
-; SSE-NEXT:    movdqa %xmm4, %xmm1
+; SSE-NEXT:    movaps {{.*#+}} xmm4 = [0,65535,65535,65535,65535,65535,65535,0]
+; SSE-NEXT:    andps %xmm4, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[3,3,3,3]
+; SSE-NEXT:    andnps %xmm1, %xmm4
+; SSE-NEXT:    orps %xmm0, %xmm4
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[0,1,0,1]
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm3, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    movdqa %xmm7, %xmm8
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm6, %xmm14
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    pand %xmm4, %xmm0
+; SSE-NEXT:    pand %xmm3, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm2, %xmm6
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    movdqa %xmm12, %xmm0
 ; SSE-NEXT:    psrld $16, %xmm0
-; SSE-NEXT:    movdqa %xmm15, %xmm3
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,0,0,65535,65535,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm0, %xmm6
-; SSE-NEXT:    pandn %xmm3, %xmm6
-; SSE-NEXT:    movdqa %xmm9, %xmm3
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm10[0],xmm3[1],xmm10[1],xmm3[2],xmm10[2],xmm3[3],xmm10[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,5,4]
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    por %xmm6, %xmm3
-; SSE-NEXT:    pand %xmm7, %xmm3
-; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rax), %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[2,3,2,3]
-; SSE-NEXT:    movdqa %xmm3, %xmm9
-; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm11, %xmm3
-; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    movdqa (%r8), %xmm2
-; SSE-NEXT:    movdqa (%r9), %xmm10
-; SSE-NEXT:    movdqa %xmm10, %xmm1
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm13, %xmm4
+; SSE-NEXT:    movdqa %xmm13, %xmm8
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,0,0,65535,65535,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm2, %xmm5
+; SSE-NEXT:    pandn %xmm4, %xmm5
+; SSE-NEXT:    movdqa %xmm11, %xmm4
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,5,4]
+; SSE-NEXT:    pand %xmm2, %xmm4
+; SSE-NEXT:    por %xmm5, %xmm4
+; SSE-NEXT:    pand %xmm6, %xmm4
+; SSE-NEXT:    por %xmm1, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%rax), %xmm12
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[2,3,2,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,0,65535,65535,65535,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm0, %xmm4
+; SSE-NEXT:    pandn %xmm1, %xmm4
+; SSE-NEXT:    movdqa (%r8), %xmm6
+; SSE-NEXT:    movdqa (%r9), %xmm7
+; SSE-NEXT:    movdqa %xmm7, %xmm1
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm1[0,1,2,3,4,5,5,4]
-; SSE-NEXT:    pand %xmm11, %xmm6
-; SSE-NEXT:    por %xmm3, %xmm6
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm1[0,1,2,3,4,5,5,4]
+; SSE-NEXT:    pand %xmm0, %xmm5
+; SSE-NEXT:    por %xmm4, %xmm5
 ; SSE-NEXT:    movdqa (%rdx), %xmm13
-; SSE-NEXT:    movdqa (%rcx), %xmm5
-; SSE-NEXT:    movdqa %xmm13, %xmm14
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm14 = xmm14[4],xmm5[4],xmm14[5],xmm5[5],xmm14[6],xmm5[6],xmm14[7],xmm5[7]
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rdi), %xmm11
-; SSE-NEXT:    movdqa (%rsi), %xmm1
-; SSE-NEXT:    movdqa %xmm11, %xmm12
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm12 = xmm12[4],xmm1[4],xmm12[5],xmm1[5],xmm12[6],xmm1[6],xmm12[7],xmm1[7]
+; SSE-NEXT:    movdqa (%rcx), %xmm15
+; SSE-NEXT:    movdqa %xmm13, %xmm9
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm15[4],xmm9[5],xmm15[5],xmm9[6],xmm15[6],xmm9[7],xmm15[7]
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%rdi), %xmm1
+; SSE-NEXT:    movdqa (%rsi), %xmm0
+; SSE-NEXT:    movdqa %xmm1, %xmm11
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm3
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,2],xmm14[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,2],xmm6[0,3]
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1],xmm2[2],xmm10[2],xmm2[3],xmm10[3]
-; SSE-NEXT:    movdqa %xmm2, (%rsp) # 16-byte Spill
-; SSE-NEXT:    pslldq {{.*#+}} xmm2 = zero,zero,xmm2[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    pand %xmm4, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm9[0,1,0,1]
-; SSE-NEXT:    movdqa %xmm9, %xmm14
-; SSE-NEXT:    pandn %xmm6, %xmm4
-; SSE-NEXT:    por %xmm2, %xmm4
-; SSE-NEXT:    movdqa %xmm5, %xmm3
-; SSE-NEXT:    movdqa %xmm5, %xmm2
-; SSE-NEXT:    psrld $16, %xmm3
-; SSE-NEXT:    movdqa %xmm13, %xmm6
-; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1],xmm1[2],xmm11[2],xmm1[3],xmm11[3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm1[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,5,4]
-; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    pandn %xmm6, %xmm0
-; SSE-NEXT:    por %xmm3, %xmm0
-; SSE-NEXT:    pand %xmm7, %xmm0
-; SSE-NEXT:    pandn %xmm4, %xmm7
-; SSE-NEXT:    por %xmm0, %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,65535,0,65535,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm7, %xmm0
-; SSE-NEXT:    pandn {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    movdqa %xmm8, %xmm3
-; SSE-NEXT:    psrldq {{.*#+}} xmm3 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    por %xmm0, %xmm3
-; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,0,0,0,65535,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm8, %xmm6
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm11 = xmm11[4],xmm0[4],xmm11[5],xmm0[5],xmm11[6],xmm0[6],xmm11[7],xmm0[7]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm11, %xmm4
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,2],xmm9[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,2],xmm5[0,3]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3]
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pslldq {{.*#+}} xmm6 = zero,zero,xmm6[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT:    pand %xmm3, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm12[0,1,0,1]
+; SSE-NEXT:    pandn %xmm5, %xmm3
+; SSE-NEXT:    por %xmm6, %xmm3
+; SSE-NEXT:    movdqa %xmm15, %xmm4
+; SSE-NEXT:    psrld $16, %xmm4
+; SSE-NEXT:    movdqa %xmm13, %xmm5
+; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm0[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,5,4]
+; SSE-NEXT:    pand %xmm2, %xmm4
+; SSE-NEXT:    pandn %xmm5, %xmm2
+; SSE-NEXT:    por %xmm4, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT:    pand %xmm6, %xmm2
 ; SSE-NEXT:    pandn %xmm3, %xmm6
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm15[2,2,2,2]
+; SSE-NEXT:    por %xmm2, %xmm6
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,65535,0,65535,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    pandn %xmm10, %xmm0
+; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    psrldq {{.*#+}} xmm2 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,0,0,0,65535,65535,65535,65535]
+; SSE-NEXT:    pandn %xmm2, %xmm9
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[2,2,2,2]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,65535,0,65535]
 ; SSE-NEXT:    movdqa %xmm0, %xmm4
-; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm5[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    pandn %xmm2, %xmm4
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    por %xmm4, %xmm2
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm5, %xmm14
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm2[3,3]
+; SSE-NEXT:    movdqa %xmm8, %xmm10
+; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
+; SSE-NEXT:    movdqa %xmm1, %xmm8
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm3[0],xmm10[1],xmm3[1],xmm10[2],xmm3[2],xmm10[3],xmm3[3]
+; SSE-NEXT:    movdqa %xmm10, %xmm7
+; SSE-NEXT:    movdqa %xmm10, (%rsp) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm3[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm14[0,2]
+; SSE-NEXT:    movaps {{.*#+}} xmm10 = [65535,0,0,0,65535,65535,65535,65535]
+; SSE-NEXT:    andps %xmm10, %xmm1
+; SSE-NEXT:    orps %xmm9, %xmm1
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pandn %xmm12, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    psrldq {{.*#+}} xmm2 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    por %xmm6, %xmm2
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm15[0,1,2,3,4,4,4,4]
 ; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,1],xmm3[3,3]
-; SSE-NEXT:    movdqa %xmm15, %xmm10
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm10 = xmm10[4],xmm5[4],xmm10[5],xmm5[5],xmm10[6],xmm5[6],xmm10[7],xmm5[7]
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm5[0],xmm15[1],xmm5[1],xmm15[2],xmm5[2],xmm15[3],xmm5[3]
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm5[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm9[0,2]
-; SSE-NEXT:    andps %xmm8, %xmm1
-; SSE-NEXT:    orps %xmm6, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm13[2,2,2,2]
+; SSE-NEXT:    pandn %xmm4, %xmm0
+; SSE-NEXT:    por %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm11, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[3,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm15[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm3[0,2]
+; SSE-NEXT:    movaps %xmm10, %xmm1
+; SSE-NEXT:    andps %xmm10, %xmm0
+; SSE-NEXT:    andnps %xmm2, %xmm1
+; SSE-NEXT:    orps %xmm0, %xmm1
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm11
-; SSE-NEXT:    pandn %xmm14, %xmm7
-; SSE-NEXT:    movdqa (%rsp), %xmm3 # 16-byte Reload
-; SSE-NEXT:    psrldq {{.*#+}} xmm3 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    por %xmm7, %xmm3
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm2[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm0, %xmm4
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm13[2,2,2,2]
-; SSE-NEXT:    pandn %xmm5, %xmm0
-; SSE-NEXT:    por %xmm4, %xmm0
-; SSE-NEXT:    movdqa %xmm12, %xmm4
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm0[3,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm4[0,2]
-; SSE-NEXT:    andps %xmm8, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm8
-; SSE-NEXT:    por %xmm0, %xmm8
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm2[0],xmm13[1],xmm2[1],xmm13[2],xmm2[2],xmm13[3],xmm2[3]
-; SSE-NEXT:    psrlq $48, %xmm2
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm15[0],xmm13[1],xmm15[1],xmm13[2],xmm15[2],xmm13[3],xmm15[3]
+; SSE-NEXT:    psrlq $48, %xmm15
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm15[1]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,0,65535,65535,65535]
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    movdqa %xmm11, %xmm0
 ; SSE-NEXT:    psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm3, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm2, %xmm3
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; SSE-NEXT:    movdqa %xmm14, %xmm1
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [0,65535,65535,65535,65535,0,0,0]
-; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; SSE-NEXT:    movdqa %xmm12, %xmm2
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [0,65535,65535,65535,65535,0,0,0]
+; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm2[0,1,2,3,6,5,4,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,5,4]
-; SSE-NEXT:    pandn %xmm1, %xmm8
-; SSE-NEXT:    por %xmm0, %xmm8
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,0],xmm10[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,2],xmm0[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm0[4],xmm4[5],xmm0[5],xmm4[6],xmm0[6],xmm4[7],xmm0[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,6,7]
+; SSE-NEXT:    pandn %xmm1, %xmm10
+; SSE-NEXT:    por %xmm0, %xmm10
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,0],xmm8[2,0]
+; SSE-NEXT:    movaps %xmm5, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm1[2,0]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm1[4],xmm6[5],xmm1[5],xmm6[6],xmm1[6],xmm6[7],xmm1[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,6,7]
 ; SSE-NEXT:    movaps {{.*#+}} xmm2 = [0,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT:    andps %xmm2, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    andnps %xmm0, %xmm2
-; SSE-NEXT:    orps %xmm1, %xmm2
+; SSE-NEXT:    andps %xmm2, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
+; SSE-NEXT:    andnps %xmm1, %xmm2
+; SSE-NEXT:    orps %xmm0, %xmm2
 ; SSE-NEXT:    movaps {{.*#+}} xmm1 = [65535,0,65535,65535,65535,65535,65535,65535]
 ; SSE-NEXT:    andps %xmm1, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,3,2,3]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,3,2,3]
 ; SSE-NEXT:    andnps %xmm0, %xmm1
 ; SSE-NEXT:    orps %xmm2, %xmm1
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    movdqa %xmm0, %xmm9
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1],xmm9[2],xmm1[2],xmm9[3],xmm1[3]
+; SSE-NEXT:    psrld $16, %xmm1
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm0[1,1,1,1,4,5,6,7]
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    psrld $16, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,0,0,65535]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
 ; SSE-NEXT:    movdqa %xmm0, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm15[0,1,2,3,4,5,6,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm14, %xmm5
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,2],xmm7[1,1]
-; SSE-NEXT:    movaps {{.*#+}} xmm6 = [65535,65535,0,0,0,65535,65535,65535]
-; SSE-NEXT:    movaps %xmm6, %xmm3
-; SSE-NEXT:    andnps %xmm5, %xmm3
-; SSE-NEXT:    pand %xmm6, %xmm1
-; SSE-NEXT:    orps %xmm1, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm15, %xmm1
-; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm15[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm9[0],xmm15[1],xmm9[1],xmm15[2],xmm9[2],xmm15[3],xmm9[3]
-; SSE-NEXT:    psrld $16, %xmm9
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1]
-; SSE-NEXT:    movdqa %xmm0, %xmm9
-; SSE-NEXT:    pandn %xmm1, %xmm9
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm13[0,1,2,3,4,5,6,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    por %xmm9, %xmm1
-; SSE-NEXT:    movaps (%rsp), %xmm5 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm5, %xmm9
-; SSE-NEXT:    movdqa %xmm11, %xmm2
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[2,2],xmm11[1,1]
-; SSE-NEXT:    pand %xmm6, %xmm1
-; SSE-NEXT:    andnps %xmm9, %xmm6
-; SSE-NEXT:    orps %xmm1, %xmm6
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    pslldq {{.*#+}} xmm9 = zero,zero,zero,zero,zero,zero,xmm9[0,1,2,3,4,5,6,7,8,9]
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    pandn %xmm9, %xmm1
-; SSE-NEXT:    pslldq {{.*#+}} xmm12 = zero,zero,xmm12[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    pand %xmm0, %xmm12
-; SSE-NEXT:    por %xmm1, %xmm12
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,0,0,0,65535]
-; SSE-NEXT:    movdqa %xmm1, %xmm9
-; SSE-NEXT:    pandn %xmm12, %xmm9
-; SSE-NEXT:    shufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm11 = xmm11[1],mem[0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm12 = xmm12[4,4,5,5,6,6,7,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[2,0],xmm12[2,1]
-; SSE-NEXT:    andps %xmm1, %xmm11
-; SSE-NEXT:    orps %xmm9, %xmm11
-; SSE-NEXT:    pslldq {{.*#+}} xmm10 = zero,zero,zero,zero,zero,zero,xmm10[0,1,2,3,4,5,6,7,8,9]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    pslldq {{.*#+}} xmm9 = zero,zero,xmm9[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    pand %xmm0, %xmm9
-; SSE-NEXT:    pandn %xmm10, %xmm0
-; SSE-NEXT:    por %xmm9, %xmm0
-; SSE-NEXT:    movdqa %xmm7, %xmm9
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,3],xmm4[0,1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4,4,5,5,6,6,7,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm4[2,1]
-; SSE-NEXT:    andps %xmm1, %xmm7
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,65535,65535,0,0,65535]
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    pandn %xmm2, %xmm0
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm7[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,2,3]
+; SSE-NEXT:    pand %xmm1, %xmm2
+; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm0[0,0,1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm14[1,1]
+; SSE-NEXT:    movaps {{.*#+}} xmm7 = [65535,65535,0,0,0,65535,65535,65535]
+; SSE-NEXT:    movaps %xmm7, %xmm8
+; SSE-NEXT:    andnps %xmm0, %xmm8
+; SSE-NEXT:    pand %xmm7, %xmm2
+; SSE-NEXT:    orps %xmm2, %xmm8
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm0, %xmm4
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
+; SSE-NEXT:    psrld $16, %xmm2
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm0[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm13[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[2,1,2,3]
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    por %xmm3, %xmm0
+; SSE-NEXT:    pand %xmm7, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,0,1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,2],xmm12[1,1]
+; SSE-NEXT:    andnps %xmm3, %xmm7
+; SSE-NEXT:    orps %xmm0, %xmm7
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pslldq {{.*#+}} xmm3 = zero,zero,zero,zero,zero,zero,xmm3[0,1,2,3,4,5,6,7,8,9]
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    pslldq {{.*#+}} xmm11 = zero,zero,xmm11[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT:    pand %xmm1, %xmm11
+; SSE-NEXT:    por %xmm0, %xmm11
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,65535,0,0,0,0,65535]
+; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    pandn %xmm11, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm12[0,0,1,1]
+; SSE-NEXT:    shufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[1],mem[0]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[2,0],xmm0[2,1]
+; SSE-NEXT:    andps %xmm5, %xmm12
+; SSE-NEXT:    orps %xmm3, %xmm12
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pslldq {{.*#+}} xmm3 = zero,zero,xmm3[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT:    pand %xmm1, %xmm3
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    por %xmm7, %xmm1
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm13 = xmm13[0],xmm15[0]
+; SSE-NEXT:    por %xmm3, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm14[0,0,1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[2,3],xmm6[0,1]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4,4,5,5,6,6,7,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[2,0],xmm6[2,1]
+; SSE-NEXT:    andps %xmm5, %xmm14
+; SSE-NEXT:    pandn %xmm1, %xmm5
+; SSE-NEXT:    por %xmm14, %xmm5
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm13 = xmm13[0],xmm4[0]
 ; SSE-NEXT:    shufps $98, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm13 = xmm13[2,0],mem[2,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[0,0,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,65535,0,0,0,65535]
-; SSE-NEXT:    movdqa %xmm4, %xmm5
-; SSE-NEXT:    pandn %xmm0, %xmm5
-; SSE-NEXT:    andps %xmm4, %xmm13
-; SSE-NEXT:    por %xmm13, %xmm5
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[0],mem[0]
-; SSE-NEXT:    shufps $98, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[2,0],mem[2,1]
-; SSE-NEXT:    movaps %xmm0, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[0,0,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm9[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm7[0],xmm0[1],xmm7[1]
-; SSE-NEXT:    andps %xmm4, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm4
-; SSE-NEXT:    por %xmm2, %xmm4
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm11[0],xmm2[1],xmm11[1]
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,65535,0,0,0,65535]
+; SSE-NEXT:    movdqa %xmm1, %xmm4
+; SSE-NEXT:    pandn %xmm2, %xmm4
+; SSE-NEXT:    andps %xmm1, %xmm13
+; SSE-NEXT:    por %xmm13, %xmm4
+; SSE-NEXT:    movdqa (%rsp), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm9[0]
+; SSE-NEXT:    shufps $98, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[2,0],mem[2,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm3[0],xmm15[1],xmm3[1]
+; SSE-NEXT:    andps %xmm1, %xmm2
+; SSE-NEXT:    pandn %xmm15, %xmm1
+; SSE-NEXT:    por %xmm2, %xmm1
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movdqa %xmm4, 112(%rax)
-; SSE-NEXT:    movdqa %xmm5, (%rax)
-; SSE-NEXT:    movdqa %xmm1, 176(%rax)
-; SSE-NEXT:    movaps %xmm11, 64(%rax)
-; SSE-NEXT:    movaps %xmm6, 32(%rax)
+; SSE-NEXT:    movdqa %xmm1, 112(%rax)
+; SSE-NEXT:    movdqa %xmm4, (%rax)
+; SSE-NEXT:    movdqa %xmm5, 176(%rax)
+; SSE-NEXT:    movaps %xmm12, 64(%rax)
+; SSE-NEXT:    movaps %xmm7, 32(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 48(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -2044,13 +2044,13 @@ define void @store_i16_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; SSE-NEXT:    movaps %xmm0, 80(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 128(%rax)
-; SSE-NEXT:    movaps %xmm3, 144(%rax)
+; SSE-NEXT:    movaps %xmm8, 144(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 192(%rax)
-; SSE-NEXT:    movdqa %xmm8, 96(%rax)
+; SSE-NEXT:    movdqa %xmm10, 96(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 208(%rax)
-; SSE-NEXT:    addq $216, %rsp
+; SSE-NEXT:    addq $232, %rsp
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: store_i16_stride7_vf16:
@@ -3558,39 +3558,40 @@ define void @store_i16_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %in.vecptr3, ptr %in.vecptr4, ptr %in.vecptr5, ptr %in.vecptr6, ptr %out.vec) nounwind {
 ; SSE-LABEL: store_i16_stride7_vf32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $680, %rsp # imm = 0x2A8
+; SSE-NEXT:    subq $696, %rsp # imm = 0x2B8
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movdqa 48(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 48(%rdi), %xmm14
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 48(%rsi), %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 48(%rdx), %xmm1
 ; SSE-NEXT:    movdqa 48(%rcx), %xmm5
-; SSE-NEXT:    movdqa 48(%r8), %xmm9
+; SSE-NEXT:    movdqa 48(%r8), %xmm3
 ; SSE-NEXT:    movdqa 48(%r9), %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 48(%rax), %xmm7
+; SSE-NEXT:    movaps 48(%rax), %xmm13
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm1, %xmm10
+; SSE-NEXT:    movdqa %xmm1, %xmm6
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,65535,65535,65535,65535,0,65535]
-; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,65535,65535,65535,65535,0,65535]
+; SSE-NEXT:    movdqa %xmm10, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm5[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm5, %xmm11
+; SSE-NEXT:    movdqa %xmm5, %xmm8
 ; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    pand %xmm10, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm3, %xmm1
-; SSE-NEXT:    movdqa %xmm3, %xmm5
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm14 = xmm14[4],xmm2[4],xmm14[5],xmm2[5],xmm14[6],xmm2[6],xmm14[7],xmm2[7]
+; SSE-NEXT:    movdqa %xmm14, %xmm1
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[3,0]
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE-NEXT:    pand %xmm3, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm9[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,0,65535,65535,65535,65535,65535,65535]
+; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[1,1,1,1]
+; SSE-NEXT:    movdqa %xmm3, %xmm7
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm5, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
 ; SSE-NEXT:    por %xmm0, %xmm3
 ; SSE-NEXT:    movdqa %xmm4, %xmm0
@@ -3600,29 +3601,29 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[0,2]
 ; SSE-NEXT:    movaps {{.*#+}} xmm1 = [65535,65535,65535,0,65535,65535,65535,65535]
 ; SSE-NEXT:    andps %xmm1, %xmm2
-; SSE-NEXT:    andnps %xmm7, %xmm1
-; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    andnps %xmm13, %xmm1
+; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    orps %xmm2, %xmm1
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm1
+; SSE-NEXT:    movdqa %xmm8, %xmm1
 ; SSE-NEXT:    psrlq $48, %xmm1
-; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    movdqa %xmm6, %xmm2
 ; SSE-NEXT:    punpckhqdq {{.*#+}} xmm2 = xmm2[1],xmm1[1]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,0,65535,65535,65535]
 ; SSE-NEXT:    pandn %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm5, %xmm2
+; SSE-NEXT:    movdqa %xmm14, %xmm2
 ; SSE-NEXT:    psrldq {{.*#+}} xmm2 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[2,2,3,3]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,65535,65535,0,65535,65535]
 ; SSE-NEXT:    movdqa %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm1, %xmm8
+; SSE-NEXT:    movdqa %xmm1, %xmm5
 ; SSE-NEXT:    pandn %xmm3, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,0],xmm4[2,0]
 ; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [0,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[3,3,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[3,3,3,3]
 ; SSE-NEXT:    movdqa %xmm1, %xmm3
 ; SSE-NEXT:    pandn %xmm0, %xmm3
 ; SSE-NEXT:    andps %xmm1, %xmm2
@@ -3630,279 +3631,270 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa (%rax), %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,1,0,1]
-; SSE-NEXT:    movdqa %xmm2, %xmm7
-; SSE-NEXT:    movdqa %xmm8, %xmm2
+; SSE-NEXT:    movdqa %xmm2, %xmm11
+; SSE-NEXT:    movdqa %xmm5, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    movdqa (%r8), %xmm0
-; SSE-NEXT:    movdqa (%r9), %xmm1
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm0, %xmm12
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1],xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; SSE-NEXT:    movdqa %xmm1, %xmm13
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%r8), %xmm9
+; SSE-NEXT:    movdqa (%r9), %xmm6
+; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1],xmm3[2],xmm6[2],xmm3[3],xmm6[3]
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    movdqa %xmm3, %xmm5
+; SSE-NEXT:    movdqa %xmm3, %xmm7
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    pand %xmm8, %xmm0
+; SSE-NEXT:    pand %xmm5, %xmm0
 ; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm1, %xmm10
+; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm13, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
 ; SSE-NEXT:    movdqa (%rcx), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
 ; SSE-NEXT:    psrld $16, %xmm0
 ; SSE-NEXT:    movdqa (%rdx), %xmm4
 ; SSE-NEXT:    movdqa %xmm4, %xmm3
-; SSE-NEXT:    movdqa %xmm4, %xmm14
+; SSE-NEXT:    movdqa %xmm4, %xmm8
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [65535,0,0,65535,65535,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm15, %xmm4
+; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [65535,0,0,65535,65535,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm14, %xmm4
 ; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    movdqa (%rdi), %xmm11
-; SSE-NEXT:    movdqa (%rsi), %xmm9
-; SSE-NEXT:    movdqa %xmm9, %xmm3
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm11[0],xmm3[1],xmm11[1],xmm3[2],xmm11[2],xmm3[3],xmm11[3]
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%rdi), %xmm12
+; SSE-NEXT:    movdqa (%rsi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm12[0],xmm3[1],xmm12[1],xmm3[2],xmm12[2],xmm3[3],xmm12[3]
+; SSE-NEXT:    movdqa %xmm12, %xmm15
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,5,4]
-; SSE-NEXT:    pand %xmm15, %xmm3
+; SSE-NEXT:    pand %xmm14, %xmm3
 ; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    pand %xmm10, %xmm3
+; SSE-NEXT:    pand %xmm13, %xmm3
 ; SSE-NEXT:    por %xmm2, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,65535,65535,65535,65535]
-; SSE-NEXT:    pandn %xmm7, %xmm2
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,65535,65535,0,65535,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm12, %xmm2
+; SSE-NEXT:    pandn %xmm11, %xmm2
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm7, %xmm3
 ; SSE-NEXT:    psrldq {{.*#+}} xmm3 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; SSE-NEXT:    por %xmm2, %xmm3
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,0,0,0,65535,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,0,0,0,65535,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm5, %xmm2
 ; SSE-NEXT:    pandn %xmm3, %xmm2
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm14[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm6, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm8[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm10, %xmm4
 ; SSE-NEXT:    pandn %xmm3, %xmm4
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm1[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm6, %xmm3
+; SSE-NEXT:    pand %xmm10, %xmm3
 ; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm11 = xmm11[4],xmm9[4],xmm11[5],xmm9[5],xmm11[6],xmm9[6],xmm11[7],xmm9[7]
-; SSE-NEXT:    movdqa %xmm11, %xmm4
+; SSE-NEXT:    movdqa %xmm15, %xmm7
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm0[4],xmm7[5],xmm0[5],xmm7[6],xmm0[6],xmm7[7],xmm0[7]
+; SSE-NEXT:    movdqa %xmm7, %xmm4
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[3,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm1[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm4[0,2]
-; SSE-NEXT:    andps %xmm0, %xmm3
+; SSE-NEXT:    andps %xmm5, %xmm3
 ; SSE-NEXT:    orps %xmm2, %xmm3
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm11[2,3,2,3]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,0,65535,65535,65535,65535,65535,65535]
 ; SSE-NEXT:    movdqa %xmm0, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    movdqa %xmm13, %xmm2
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm12[4],xmm2[5],xmm12[5],xmm2[6],xmm12[6],xmm2[7],xmm12[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm9[4],xmm6[5],xmm9[5],xmm6[6],xmm9[6],xmm6[7],xmm9[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm6[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,5,4]
 ; SSE-NEXT:    pand %xmm0, %xmm2
 ; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    movdqa %xmm14, %xmm4
+; SSE-NEXT:    movdqa %xmm8, %xmm4
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm1[4],xmm4[5],xmm1[5],xmm4[6],xmm1[6],xmm4[7],xmm1[7]
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm3
+; SSE-NEXT:    movdqa %xmm7, %xmm3
 ; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,2],xmm4[2,3]
 ; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,2],xmm2[0,3]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 16(%rax), %xmm3
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,1,0,1]
-; SSE-NEXT:    movdqa %xmm3, %xmm7
-; SSE-NEXT:    movdqa %xmm8, %xmm5
-; SSE-NEXT:    movdqa %xmm8, %xmm3
+; SSE-NEXT:    movdqa %xmm3, %xmm11
+; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [65535,65535,65535,65535,65535,0,65535,65535]
+; SSE-NEXT:    movdqa %xmm9, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    movdqa 16(%r8), %xmm10
-; SSE-NEXT:    movdqa 16(%r9), %xmm8
-; SSE-NEXT:    movdqa %xmm10, %xmm4
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1],xmm4[2],xmm8[2],xmm4[3],xmm8[3]
-; SSE-NEXT:    movdqa %xmm8, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movdqa 16(%r8), %xmm8
+; SSE-NEXT:    movdqa 16(%r9), %xmm5
+; SSE-NEXT:    movdqa %xmm8, %xmm4
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; SSE-NEXT:    movdqa %xmm5, (%rsp) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    movdqa %xmm4, %xmm12
+; SSE-NEXT:    movdqa %xmm4, %xmm7
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pslldq {{.*#+}} xmm2 = zero,zero,xmm2[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    pand %xmm5, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm2
 ; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm13, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    movdqa 16(%rcx), %xmm5
-; SSE-NEXT:    movdqa %xmm5, %xmm2
+; SSE-NEXT:    movdqa 16(%rcx), %xmm0
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm0, %xmm6
 ; SSE-NEXT:    psrld $16, %xmm2
-; SSE-NEXT:    movdqa 16(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm0, %xmm9
+; SSE-NEXT:    movdqa 16(%rdx), %xmm1
+; SSE-NEXT:    movdqa %xmm1, %xmm4
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
-; SSE-NEXT:    movdqa %xmm15, %xmm2
+; SSE-NEXT:    movdqa %xmm14, %xmm2
 ; SSE-NEXT:    pandn %xmm4, %xmm2
-; SSE-NEXT:    movdqa 16(%rdi), %xmm0
-; SSE-NEXT:    movdqa 16(%rsi), %xmm13
-; SSE-NEXT:    movdqa %xmm13, %xmm4
-; SSE-NEXT:    movdqa %xmm13, %xmm14
-; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT:    movdqa %xmm0, %xmm13
+; SSE-NEXT:    movdqa 16(%rdi), %xmm15
+; SSE-NEXT:    movdqa 16(%rsi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, %xmm4
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm15[0],xmm4[1],xmm15[1],xmm4[2],xmm15[2],xmm4[3],xmm15[3]
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,5,4]
-; SSE-NEXT:    pand %xmm15, %xmm4
+; SSE-NEXT:    pand %xmm14, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
-; SSE-NEXT:    pand %xmm1, %xmm4
+; SSE-NEXT:    pand %xmm13, %xmm4
 ; SSE-NEXT:    por %xmm3, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,65535,65535,65535,65535]
-; SSE-NEXT:    pandn %xmm7, %xmm2
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm3
+; SSE-NEXT:    pandn %xmm11, %xmm12
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm7, %xmm3
 ; SSE-NEXT:    psrldq {{.*#+}} xmm3 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    por %xmm2, %xmm3
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,0,0,0,65535,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    por %xmm12, %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [65535,0,0,0,65535,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm12, %xmm2
 ; SSE-NEXT:    pandn %xmm3, %xmm2
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm9[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm6, %xmm4
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm10, %xmm4
 ; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm5[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm6, %xmm3
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm6[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    pand %xmm10, %xmm3
 ; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm13 = xmm13[4],xmm14[4],xmm13[5],xmm14[5],xmm13[6],xmm14[6],xmm13[7],xmm14[7]
-; SSE-NEXT:    movdqa %xmm13, %xmm4
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm15 = xmm15[4],xmm0[4],xmm15[5],xmm0[5],xmm15[6],xmm0[6],xmm15[7],xmm0[7]
+; SSE-NEXT:    movdqa %xmm15, %xmm4
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[3,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm5[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm6[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm4[0,2]
-; SSE-NEXT:    andps %xmm0, %xmm3
+; SSE-NEXT:    andps %xmm12, %xmm3
 ; SSE-NEXT:    orps %xmm2, %xmm3
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[2,3,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm11[2,3,2,3]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,0,65535,65535,65535,65535,65535,65535]
 ; SSE-NEXT:    movdqa %xmm0, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm10[4],xmm8[5],xmm10[5],xmm8[6],xmm10[6],xmm8[7],xmm10[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm8[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm8[4],xmm5[5],xmm8[5],xmm5[6],xmm8[6],xmm5[7],xmm8[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm5[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,5,4]
 ; SSE-NEXT:    pand %xmm0, %xmm2
 ; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    movdqa %xmm9, %xmm4
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm3
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,2],xmm4[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,2],xmm2[0,3]
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 32(%rax), %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[0,1,0,1]
-; SSE-NEXT:    movdqa %xmm3, %xmm7
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,65535,65535,65535,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[2,2],xmm1[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[1,2],xmm2[0,3]
+; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 32(%rax), %xmm8
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[0,1,0,1]
+; SSE-NEXT:    movdqa %xmm9, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    movdqa 32(%r8), %xmm10
-; SSE-NEXT:    movdqa 32(%r9), %xmm9
-; SSE-NEXT:    movdqa %xmm10, %xmm4
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3]
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 32(%r8), %xmm7
+; SSE-NEXT:    movdqa 32(%r9), %xmm1
+; SSE-NEXT:    movdqa %xmm7, %xmm4
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    movdqa %xmm1, %xmm15
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    movdqa %xmm4, %xmm14
+; SSE-NEXT:    movdqa %xmm4, %xmm0
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pslldq {{.*#+}} xmm2 = zero,zero,xmm2[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    pand %xmm5, %xmm2
+; SSE-NEXT:    pand %xmm9, %xmm2
 ; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    movdqa %xmm13, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    movdqa 32(%rcx), %xmm12
-; SSE-NEXT:    movdqa %xmm12, %xmm2
+; SSE-NEXT:    movdqa 32(%rcx), %xmm6
+; SSE-NEXT:    movdqa %xmm6, %xmm2
 ; SSE-NEXT:    psrld $16, %xmm2
-; SSE-NEXT:    movdqa 32(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm0, %xmm8
+; SSE-NEXT:    movdqa 32(%rdx), %xmm11
+; SSE-NEXT:    movdqa %xmm11, %xmm4
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
-; SSE-NEXT:    movdqa %xmm15, %xmm2
+; SSE-NEXT:    movdqa %xmm14, %xmm2
 ; SSE-NEXT:    pandn %xmm4, %xmm2
-; SSE-NEXT:    movdqa 32(%rdi), %xmm1
-; SSE-NEXT:    movdqa 32(%rsi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1],xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    movdqa 32(%rdi), %xmm5
+; SSE-NEXT:    movdqa 32(%rsi), %xmm1
+; SSE-NEXT:    movdqa %xmm1, %xmm4
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,5,4]
-; SSE-NEXT:    pand %xmm15, %xmm4
+; SSE-NEXT:    pand %xmm14, %xmm4
 ; SSE-NEXT:    por %xmm2, %xmm4
-; SSE-NEXT:    pand %xmm5, %xmm4
+; SSE-NEXT:    pand %xmm13, %xmm4
 ; SSE-NEXT:    por %xmm3, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [65535,65535,65535,0,65535,65535,65535,65535]
-; SSE-NEXT:    pandn %xmm7, %xmm3
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    pandn %xmm8, %xmm3
+; SSE-NEXT:    movdqa %xmm0, %xmm2
 ; SSE-NEXT:    psrldq {{.*#+}} xmm2 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm12[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm6, %xmm3
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm8[2,2,2,2]
-; SSE-NEXT:    pandn %xmm4, %xmm6
-; SSE-NEXT:    por %xmm3, %xmm6
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
-; SSE-NEXT:    movdqa %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm6[3,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm12[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[0,2]
-; SSE-NEXT:    movaps {{.*#+}} xmm5 = [65535,0,0,0,65535,65535,65535,65535]
-; SSE-NEXT:    andps %xmm5, %xmm4
-; SSE-NEXT:    andnps %xmm2, %xmm5
-; SSE-NEXT:    orps %xmm4, %xmm5
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[2,3,2,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [65535,0,65535,65535,65535,65535,65535,65535]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm6[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    pand %xmm10, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm11[2,2,2,2]
+; SSE-NEXT:    pandn %xmm4, %xmm10
+; SSE-NEXT:    por %xmm3, %xmm10
+; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
 ; SSE-NEXT:    movdqa %xmm0, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm10[3,3]
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm6[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[0,2]
+; SSE-NEXT:    andps %xmm12, %xmm4
+; SSE-NEXT:    pandn %xmm2, %xmm12
+; SSE-NEXT:    por %xmm4, %xmm12
+; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[2,3,2,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [65535,0,65535,65535,65535,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm8, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm10[4],xmm9[5],xmm10[5],xmm9[6],xmm10[6],xmm9[7],xmm10[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm9[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm15, %xmm2
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm7[4],xmm2[5],xmm7[5],xmm2[6],xmm7[6],xmm2[7],xmm7[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,5,4]
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm0, %xmm9
+; SSE-NEXT:    pand %xmm8, %xmm2
 ; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm12[4],xmm0[5],xmm12[5],xmm0[6],xmm12[6],xmm0[7],xmm12[7]
+; SSE-NEXT:    movdqa %xmm11, %xmm3
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm6[4],xmm3[5],xmm6[5],xmm3[6],xmm6[6],xmm3[7],xmm6[7]
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm0, %xmm12
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,2],xmm0[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,2],xmm2[0,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1],xmm2[2],xmm8[2],xmm2[3],xmm8[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2],xmm3[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,2],xmm2[0,3]
+; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm0, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3]
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pslldq {{.*#+}} xmm2 = zero,zero,xmm2[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [65535,65535,65535,65535,65535,0,65535,65535]
-; SSE-NEXT:    pand %xmm10, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[0,1,0,1]
-; SSE-NEXT:    pandn %xmm3, %xmm10
-; SSE-NEXT:    por %xmm2, %xmm10
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm5, %xmm6
+; SSE-NEXT:    pand %xmm9, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm10[0,1,0,1]
+; SSE-NEXT:    pandn %xmm3, %xmm9
+; SSE-NEXT:    por %xmm2, %xmm9
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm6, %xmm5
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4],xmm3[4],xmm6[5],xmm3[5],xmm6[6],xmm3[6],xmm6[7],xmm3[7]
-; SSE-NEXT:    movdqa %xmm5, %xmm2
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3]
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm3[4],xmm5[5],xmm3[5],xmm5[6],xmm3[6],xmm5[7],xmm3[7]
+; SSE-NEXT:    movdqa %xmm6, %xmm2
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    psrld $16, %xmm3
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
@@ -3910,83 +3902,86 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; SSE-NEXT:    # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1],xmm3[2],mem[2],xmm3[3],mem[3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,5,4]
-; SSE-NEXT:    pand %xmm15, %xmm3
-; SSE-NEXT:    pandn %xmm2, %xmm15
-; SSE-NEXT:    por %xmm3, %xmm15
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT:    pand %xmm2, %xmm15
-; SSE-NEXT:    pandn %xmm10, %xmm2
-; SSE-NEXT:    por %xmm15, %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,2,2,2]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,0],xmm6[2,0]
+; SSE-NEXT:    pand %xmm14, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm14
+; SSE-NEXT:    por %xmm3, %xmm14
+; SSE-NEXT:    pand %xmm13, %xmm14
+; SSE-NEXT:    pandn %xmm9, %xmm13
+; SSE-NEXT:    por %xmm14, %xmm13
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,0],xmm5[2,0]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pslldq {{.*#+}} xmm6 = zero,zero,zero,zero,zero,zero,xmm6[0,1,2,3,4,5,6,7,8,9]
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,65535,65535,0,0,65535]
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    pandn %xmm6, %xmm2
+; SSE-NEXT:    pslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,xmm5[0,1,2,3,4,5,6,7,8,9]
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,65535,65535,65535,0,0,65535]
+; SSE-NEXT:    movdqa %xmm6, %xmm2
+; SSE-NEXT:    pandn %xmm5, %xmm2
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    pslldq {{.*#+}} xmm3 = zero,zero,xmm3[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    pand %xmm4, %xmm3
+; SSE-NEXT:    pand %xmm6, %xmm3
 ; SSE-NEXT:    por %xmm2, %xmm3
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [0,65535,65535,65535,65535,65535,65535,0]
 ; SSE-NEXT:    movdqa %xmm1, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
 ; SSE-NEXT:    pand %xmm1, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    psrldq {{.*#+}} xmm8 = xmm8[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    movdqa %xmm9, %xmm3
-; SSE-NEXT:    pandn %xmm8, %xmm3
-; SSE-NEXT:    pand %xmm9, %xmm2
+; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    psrldq {{.*#+}} xmm0 = xmm0[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    movdqa %xmm8, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    pand %xmm8, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm3
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,0,65535,65535,65535,65535,65535]
 ; SSE-NEXT:    pand %xmm1, %xmm3
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[2,2,3,3]
 ; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    por %xmm3, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm15[0],xmm8[1],xmm15[1],xmm8[2],xmm15[2],xmm8[3],xmm15[3]
-; SSE-NEXT:    psrlq $48, %xmm15
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm15[1]
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,65535,0,0,65535,65535,65535]
-; SSE-NEXT:    movdqa %xmm6, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm2[0],xmm9[1],xmm2[1],xmm9[2],xmm2[2],xmm9[3],xmm2[3]
+; SSE-NEXT:    psrlq $48, %xmm2
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,65535,0,0,65535,65535,65535]
+; SSE-NEXT:    movdqa %xmm5, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm3, %xmm1
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    psrld $16, %xmm2
-; SSE-NEXT:    movdqa %xmm1, %xmm3
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm5, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm4, %xmm2
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,5,4]
-; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [0,65535,65535,65535,65535,0,0,0]
-; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [0,65535,65535,65535,65535,0,0,0]
+; SSE-NEXT:    movdqa %xmm10, %xmm1
 ; SSE-NEXT:    pandn %xmm2, %xmm1
-; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    pand %xmm10, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm9, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm14[0],xmm9[1],xmm14[1],xmm9[2],xmm14[2],xmm9[3],xmm14[3]
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    psrlq $48, %xmm14
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm14[1]
-; SSE-NEXT:    movdqa %xmm6, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE-NEXT:    movdqa %xmm1, %xmm13
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    psrlq $48, %xmm2
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; SSE-NEXT:    movdqa %xmm5, %xmm2
+; SSE-NEXT:    movdqa %xmm5, %xmm8
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    movdqa %xmm13, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm15, %xmm0
 ; SSE-NEXT:    psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; SSE-NEXT:    por %xmm2, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -3997,46 +3992,45 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; SSE-NEXT:    psrld $16, %xmm2
 ; SSE-NEXT:    movdqa %xmm1, %xmm3
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm5, %xmm2
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,5,4]
-; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    movdqa %xmm10, %xmm1
 ; SSE-NEXT:    pandn %xmm2, %xmm1
-; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    pand %xmm10, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1],xmm1[2],xmm12[2],xmm1[3],xmm12[3]
-; SSE-NEXT:    movdqa %xmm1, %xmm14
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    psrlq $48, %xmm12
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm12[1]
-; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm2[0],xmm11[1],xmm2[1],xmm11[2],xmm2[2],xmm11[3],xmm2[3]
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    psrlq $48, %xmm2
+; SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; SSE-NEXT:    movdqa %xmm8, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm12, %xmm0
 ; SSE-NEXT:    psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm6, %xmm3
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm7, %xmm3
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7]
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    psrld $16, %xmm2
-; SSE-NEXT:    movdqa %xmm6, %xmm3
+; SSE-NEXT:    movdqa %xmm7, %xmm3
 ; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm7, %xmm2
 ; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7]
-; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    pand %xmm10, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,4,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[3,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,7,5,4]
-; SSE-NEXT:    pandn %xmm2, %xmm15
-; SSE-NEXT:    por %xmm0, %xmm15
+; SSE-NEXT:    pandn %xmm2, %xmm10
+; SSE-NEXT:    por %xmm0, %xmm10
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[1,1,1,1,4,5,6,7]
@@ -4046,217 +4040,216 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    psrld $16, %xmm2
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    movdqa %xmm6, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm8[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm9[0,1,2,3,4,5,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
-; SSE-NEXT:    pand %xmm4, %xmm0
+; SSE-NEXT:    pand %xmm6, %xmm0
 ; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,2],xmm5[1,1]
-; SSE-NEXT:    movaps %xmm5, %xmm3
-; SSE-NEXT:    movaps {{.*#+}} xmm12 = [65535,65535,0,0,0,65535,65535,65535]
-; SSE-NEXT:    movaps %xmm12, %xmm1
-; SSE-NEXT:    andnps %xmm2, %xmm1
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    orps %xmm0, %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,2],xmm4[1,1]
+; SSE-NEXT:    movaps {{.*#+}} xmm2 = [65535,65535,0,0,0,65535,65535,65535]
+; SSE-NEXT:    movaps %xmm2, %xmm14
+; SSE-NEXT:    andnps %xmm3, %xmm14
+; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    orps %xmm0, %xmm14
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm1[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    psrld $16, %xmm2
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm9[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    psrld $16, %xmm3
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT:    movdqa %xmm6, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm13[0,1,2,3,4,5,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
-; SSE-NEXT:    pand %xmm4, %xmm0
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,2],xmm10[1,1]
-; SSE-NEXT:    movaps %xmm12, %xmm1
-; SSE-NEXT:    andnps %xmm2, %xmm1
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    orps %xmm0, %xmm1
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    por %xmm3, %xmm0
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,2],xmm5[1,1]
+; SSE-NEXT:    movaps %xmm2, %xmm8
+; SSE-NEXT:    andnps %xmm3, %xmm8
+; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    orps %xmm0, %xmm8
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm1[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    psrld $16, %xmm2
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm14[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    psrld $16, %xmm3
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; SSE-NEXT:    movdqa %xmm6, %xmm3
+; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm11[0,1,2,3,4,5,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
-; SSE-NEXT:    pand %xmm4, %xmm0
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm9, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,2],xmm7[1,1]
-; SSE-NEXT:    movaps %xmm7, %xmm5
-; SSE-NEXT:    movaps %xmm12, %xmm10
-; SSE-NEXT:    andnps %xmm1, %xmm10
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    orps %xmm0, %xmm10
+; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    por %xmm3, %xmm0
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,2],xmm7[1,1]
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    andnps %xmm4, %xmm5
+; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    orps %xmm0, %xmm5
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm1[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    psrld $16, %xmm1
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pshufhw $164, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    psrld $16, %xmm4
+; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
+; SSE-NEXT:    movdqa %xmm6, %xmm4
+; SSE-NEXT:    pandn %xmm0, %xmm4
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm12[0,1,2,3,4,5,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
-; SSE-NEXT:    pand %xmm4, %xmm0
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,2],xmm14[1,1]
-; SSE-NEXT:    andnps %xmm1, %xmm12
-; SSE-NEXT:    orps %xmm0, %xmm12
+; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    por %xmm4, %xmm0
+; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm11, %xmm4
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,2],xmm13[1,1]
+; SSE-NEXT:    andnps %xmm4, %xmm2
+; SSE-NEXT:    orps %xmm0, %xmm2
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    shufps $42, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = xmm1[2,2],mem[2,0]
 ; SSE-NEXT:    pshufhw $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[0,1,2,3,5,6,6,7]
-; SSE-NEXT:    movaps {{.*#+}} xmm2 = [0,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT:    andps %xmm2, %xmm1
+; SSE-NEXT:    movaps {{.*#+}} xmm3 = [0,65535,65535,65535,65535,65535,65535,0]
+; SSE-NEXT:    andps %xmm3, %xmm1
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    andnps %xmm0, %xmm2
-; SSE-NEXT:    orps %xmm1, %xmm2
+; SSE-NEXT:    andnps %xmm0, %xmm3
+; SSE-NEXT:    orps %xmm1, %xmm3
 ; SSE-NEXT:    movaps {{.*#+}} xmm1 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE-NEXT:    andps %xmm1, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,3,2,3]
+; SSE-NEXT:    andps %xmm1, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,3,2,3]
 ; SSE-NEXT:    andnps %xmm0, %xmm1
-; SSE-NEXT:    orps %xmm2, %xmm1
+; SSE-NEXT:    orps %xmm3, %xmm1
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    pslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,xmm1[0,1,2,3,4,5,6,7,8,9]
-; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    movdqa %xmm6, %xmm0
 ; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    pslldq {{.*#+}} xmm11 = zero,zero,xmm11[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    pand %xmm4, %xmm11
-; SSE-NEXT:    por %xmm0, %xmm11
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,0,0,0,65535]
-; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    pandn %xmm11, %xmm1
-; SSE-NEXT:    movaps %xmm3, %xmm11
-; SSE-NEXT:    shufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm11 = xmm11[1],mem[0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[2,0],xmm0[2,1]
-; SSE-NEXT:    andps %xmm2, %xmm11
-; SSE-NEXT:    orps %xmm1, %xmm11
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9]
-; SSE-NEXT:    movdqa %xmm4, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pslldq {{.*#+}} xmm13 = zero,zero,xmm13[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    pand %xmm4, %xmm13
-; SSE-NEXT:    por %xmm1, %xmm13
-; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    pandn %xmm13, %xmm1
-; SSE-NEXT:    movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movapd %xmm7, %xmm13
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pslldq {{.*#+}} xmm1 = zero,zero,xmm1[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT:    pand %xmm6, %xmm1
+; SSE-NEXT:    por %xmm0, %xmm1
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,0,0,0,65535]
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm13[0,0,1,1]
 ; SSE-NEXT:    shufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm13 = xmm13[1],mem[0]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[2,0],xmm3[2,1]
+; SSE-NEXT:    andps %xmm1, %xmm13
+; SSE-NEXT:    orps %xmm0, %xmm13
+; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pslldq {{.*#+}} xmm3 = zero,zero,zero,zero,zero,zero,xmm3[0,1,2,3,4,5,6,7,8,9]
+; SSE-NEXT:    movdqa %xmm6, %xmm0
+; SSE-NEXT:    pandn %xmm3, %xmm0
+; SSE-NEXT:    pslldq {{.*#+}} xmm15 = zero,zero,xmm15[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT:    pand %xmm6, %xmm15
+; SSE-NEXT:    por %xmm0, %xmm15
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    pandn %xmm15, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm15 = xmm13[0,0,1,1]
+; SSE-NEXT:    shufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm13 = xmm13[1],mem[0]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[2,0],xmm3[2,1]
+; SSE-NEXT:    andps %xmm1, %xmm13
+; SSE-NEXT:    orps %xmm0, %xmm13
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pslldq {{.*#+}} xmm3 = zero,zero,zero,zero,zero,zero,xmm3[0,1,2,3,4,5,6,7,8,9]
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[2,0],xmm0[2,1]
-; SSE-NEXT:    andps %xmm2, %xmm13
-; SSE-NEXT:    orps %xmm1, %xmm13
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pslldq {{.*#+}} xmm1 = zero,zero,xmm1[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT:    pand %xmm4, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm4
-; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    movaps %xmm5, %xmm1
-; SSE-NEXT:    shufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[1],mem[0]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm6 = xmm6[4,4,5,5,6,6,7,7]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm6[2,1]
-; SSE-NEXT:    andps %xmm2, %xmm1
-; SSE-NEXT:    pandn %xmm4, %xmm2
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = xmm8[0],mem[0]
-; SSE-NEXT:    shufps $98, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = xmm8[2,0],mem[2,1]
-; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[0,0,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,65535,65535,0,0,0,65535]
-; SSE-NEXT:    movdqa %xmm6, %xmm1
-; SSE-NEXT:    pandn %xmm4, %xmm1
-; SSE-NEXT:    andps %xmm6, %xmm8
-; SSE-NEXT:    por %xmm8, %xmm1
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = xmm3[0],mem[0]
-; SSE-NEXT:    shufps $98, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = xmm3[2,0],mem[2,1]
+; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    pandn %xmm3, %xmm6
+; SSE-NEXT:    por %xmm0, %xmm6
+; SSE-NEXT:    movaps %xmm7, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm7[0,0,1,1]
+; SSE-NEXT:    shufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[1],mem[0]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm7 = xmm7[4,4,5,5,6,6,7,7]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm7[2,1]
+; SSE-NEXT:    andps %xmm1, %xmm0
+; SSE-NEXT:    pandn %xmm6, %xmm1
+; SSE-NEXT:    por %xmm0, %xmm1
+; SSE-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = xmm9[0],mem[0]
+; SSE-NEXT:    shufps $98, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = xmm9[2,0],mem[2,1]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm6 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [65535,65535,65535,65535,0,0,0,65535]
+; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    pandn %xmm6, %xmm0
+; SSE-NEXT:    andps %xmm7, %xmm9
+; SSE-NEXT:    por %xmm9, %xmm0
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = xmm9[0],mem[0]
+; SSE-NEXT:    shufps $98, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = xmm9[2,0],mem[2,1]
 ; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm4 = mem[0,0,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm7[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1]
-; SSE-NEXT:    movdqa %xmm6, %xmm8
-; SSE-NEXT:    pandn %xmm4, %xmm8
-; SSE-NEXT:    andps %xmm6, %xmm3
-; SSE-NEXT:    por %xmm3, %xmm8
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[0],mem[0]
-; SSE-NEXT:    shufps $98, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[2,0],mem[2,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm9[0,0,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm5[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1]
-; SSE-NEXT:    movdqa %xmm6, %xmm9
-; SSE-NEXT:    pandn %xmm4, %xmm9
-; SSE-NEXT:    andps %xmm6, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm9
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[0],mem[0]
-; SSE-NEXT:    shufps $98, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[2,0],mem[2,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm15[0],xmm4[1],xmm15[1]
+; SSE-NEXT:    movdqa %xmm7, %xmm6
+; SSE-NEXT:    pandn %xmm4, %xmm6
+; SSE-NEXT:    andps %xmm7, %xmm9
+; SSE-NEXT:    por %xmm9, %xmm6
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = xmm9[0],mem[0]
+; SSE-NEXT:    shufps $98, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = xmm9[2,0],mem[2,1]
 ; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm4 = mem[0,0,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm14[0,0,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1]
-; SSE-NEXT:    andps %xmm6, %xmm0
-; SSE-NEXT:    pandn %xmm4, %xmm6
-; SSE-NEXT:    por %xmm0, %xmm6
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT:    movdqa %xmm7, %xmm3
+; SSE-NEXT:    pandn %xmm4, %xmm3
+; SSE-NEXT:    andps %xmm7, %xmm9
+; SSE-NEXT:    por %xmm9, %xmm3
+; SSE-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[0],mem[0]
+; SSE-NEXT:    shufps $98, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[2,0],mem[2,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm11[0,0,1,1]
+; SSE-NEXT:    pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm11 = mem[0,0,1,1]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm11[0],xmm4[1],xmm11[1]
+; SSE-NEXT:    andps %xmm7, %xmm12
+; SSE-NEXT:    pandn %xmm4, %xmm7
+; SSE-NEXT:    por %xmm12, %xmm7
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movdqa %xmm6, 336(%rax)
-; SSE-NEXT:    movdqa %xmm9, 224(%rax)
-; SSE-NEXT:    movdqa %xmm8, 112(%rax)
-; SSE-NEXT:    movdqa %xmm1, (%rax)
-; SSE-NEXT:    movdqa %xmm2, 288(%rax)
+; SSE-NEXT:    movdqa %xmm7, 336(%rax)
+; SSE-NEXT:    movdqa %xmm3, 224(%rax)
+; SSE-NEXT:    movdqa %xmm6, 112(%rax)
+; SSE-NEXT:    movdqa %xmm0, (%rax)
+; SSE-NEXT:    movdqa %xmm1, 288(%rax)
 ; SSE-NEXT:    movaps %xmm13, 176(%rax)
-; SSE-NEXT:    movaps %xmm11, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 416(%rax)
-; SSE-NEXT:    movaps %xmm12, 368(%rax)
+; SSE-NEXT:    movaps %xmm2, 368(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 352(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -4264,7 +4257,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; SSE-NEXT:    movaps %xmm0, 304(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 272(%rax)
-; SSE-NEXT:    movaps %xmm10, 256(%rax)
+; SSE-NEXT:    movaps %xmm5, 256(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 240(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -4272,8 +4265,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; SSE-NEXT:    movaps %xmm0, 192(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 160(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 144(%rax)
+; SSE-NEXT:    movaps %xmm8, 144(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 128(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -4281,11 +4273,10 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; SSE-NEXT:    movaps %xmm0, 80(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 48(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 32(%rax)
+; SSE-NEXT:    movaps %xmm14, 32(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 16(%rax)
-; SSE-NEXT:    movdqa %xmm15, 320(%rax)
+; SSE-NEXT:    movdqa %xmm10, 320(%rax)
 ; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 208(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -4296,7 +4287,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; SSE-NEXT:    movaps %xmm0, 400(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 384(%rax)
-; SSE-NEXT:    addq $680, %rsp # imm = 0x2A8
+; SSE-NEXT:    addq $696, %rsp # imm = 0x2B8
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: store_i16_stride7_vf32:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-3.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-3.ll
index de2e1df4c5566..f82648fa78f0e 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-3.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-3.ll
@@ -1195,191 +1195,183 @@ define void @store_i32_stride3_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    subq $152, %rsp
 ; SSE-NEXT:    movaps (%rdi), %xmm1
-; SSE-NEXT:    movaps 16(%rdi), %xmm2
-; SSE-NEXT:    movaps 32(%rdi), %xmm3
-; SSE-NEXT:    movaps 48(%rdi), %xmm4
-; SSE-NEXT:    movaps (%rsi), %xmm10
-; SSE-NEXT:    movaps 16(%rsi), %xmm13
-; SSE-NEXT:    movaps 32(%rsi), %xmm12
-; SSE-NEXT:    movaps 48(%rsi), %xmm9
-; SSE-NEXT:    movaps (%rdx), %xmm5
-; SSE-NEXT:    movaps 16(%rdx), %xmm6
-; SSE-NEXT:    movaps 32(%rdx), %xmm7
-; SSE-NEXT:    movaps 48(%rdx), %xmm8
+; SSE-NEXT:    movaps 16(%rdi), %xmm14
+; SSE-NEXT:    movaps 32(%rdi), %xmm13
+; SSE-NEXT:    movaps 48(%rdi), %xmm12
+; SSE-NEXT:    movaps (%rsi), %xmm2
+; SSE-NEXT:    movaps 16(%rsi), %xmm3
+; SSE-NEXT:    movaps 32(%rsi), %xmm4
+; SSE-NEXT:    movaps 48(%rsi), %xmm5
+; SSE-NEXT:    movaps (%rdx), %xmm10
+; SSE-NEXT:    movaps %xmm10, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movaps 16(%rdx), %xmm9
+; SSE-NEXT:    movaps 32(%rdx), %xmm8
+; SSE-NEXT:    movaps 48(%rdx), %xmm7
 ; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm5[0,3]
-; SSE-NEXT:    movaps %xmm5, %xmm11
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm1, %xmm5
-; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm10[0],xmm5[1],xmm10[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm10[0,3]
+; SSE-NEXT:    movaps %xmm1, %xmm6
+; SSE-NEXT:    unpcklps {{.*#+}} xmm6 = xmm6[0],xmm2[0],xmm6[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm10[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm10[3,3]
-; SSE-NEXT:    movaps %xmm1, %xmm15
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[1,1],xmm11[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,2],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm2, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm6[0,3]
-; SSE-NEXT:    movaps %xmm2, %xmm1
-; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm2[3,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm2, %xmm0
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm13[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm13[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm10[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm0[0,2]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[1,1],xmm6[1,1]
-; SSE-NEXT:    movaps %xmm6, %xmm14
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,2],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm7[0,3]
-; SSE-NEXT:    movaps %xmm3, %xmm2
-; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm12[0],xmm2[1],xmm12[1]
+; SSE-NEXT:    movaps %xmm14, %xmm0
+; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm9[0,3]
+; SSE-NEXT:    movaps %xmm14, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm12[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,3],xmm12[3,3]
-; SSE-NEXT:    movaps %xmm3, (%rsp) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[1,1],xmm7[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,2],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm4, %xmm0
+; SSE-NEXT:    movaps %xmm14, %xmm0
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm9[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm13, %xmm0
 ; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm8[0,3]
-; SSE-NEXT:    movaps %xmm4, %xmm1
-; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm9[0],xmm1[1],xmm9[1]
+; SSE-NEXT:    movaps %xmm13, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm4[0],xmm1[1],xmm4[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm4, %xmm0
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm9[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,3],xmm9[3,3]
+; SSE-NEXT:    movaps %xmm13, %xmm0
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[3,3],xmm4[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm8[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm0[0,2]
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[1,1],xmm8[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,2],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 64(%rdi), %xmm9
+; SSE-NEXT:    movaps %xmm12, %xmm0
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm7[0,3]
+; SSE-NEXT:    movaps %xmm12, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm12, %xmm0
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm5[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[3,3],xmm5[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 64(%rdi), %xmm15
 ; SSE-NEXT:    movaps 64(%rdx), %xmm1
-; SSE-NEXT:    movaps %xmm9, %xmm0
+; SSE-NEXT:    movaps %xmm15, %xmm0
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[0,3]
 ; SSE-NEXT:    movaps %xmm1, %xmm2
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 64(%rsi), %xmm12
-; SSE-NEXT:    movaps %xmm9, %xmm1
-; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1]
+; SSE-NEXT:    movaps 64(%rsi), %xmm7
+; SSE-NEXT:    movaps %xmm15, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm9, %xmm0
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm12[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[3,3],xmm12[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[1,1],xmm2[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm15, %xmm0
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm7[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[3,3],xmm7[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[1,1],xmm2[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm0[0,2]
 ; SSE-NEXT:    movaps 80(%rdi), %xmm10
 ; SSE-NEXT:    movaps 80(%rdx), %xmm1
 ; SSE-NEXT:    movaps %xmm10, %xmm0
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[0,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 80(%rsi), %xmm8
+; SSE-NEXT:    movaps 80(%rsi), %xmm6
 ; SSE-NEXT:    movaps %xmm10, %xmm11
-; SSE-NEXT:    unpcklps {{.*#+}} xmm11 = xmm11[0],xmm8[0],xmm11[1],xmm8[1]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm11 = xmm11[0],xmm6[0],xmm11[1],xmm6[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,1],xmm0[2,0]
 ; SSE-NEXT:    movaps %xmm10, %xmm0
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm8[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[3,3],xmm8[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[1,1],xmm1[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm0[0,2]
-; SSE-NEXT:    movaps 96(%rdi), %xmm4
-; SSE-NEXT:    movaps 96(%rdx), %xmm13
-; SSE-NEXT:    movaps %xmm4, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm13[0,3]
-; SSE-NEXT:    movaps 96(%rsi), %xmm5
-; SSE-NEXT:    movaps %xmm4, %xmm6
-; SSE-NEXT:    unpcklps {{.*#+}} xmm6 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm4, %xmm0
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm5[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,3],xmm5[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm13[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[0,2]
-; SSE-NEXT:    movaps 112(%rdi), %xmm0
-; SSE-NEXT:    movaps 112(%rdx), %xmm7
-; SSE-NEXT:    movaps %xmm0, %xmm2
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm7[0,3]
-; SSE-NEXT:    movaps 112(%rsi), %xmm1
-; SSE-NEXT:    movaps %xmm0, %xmm3
-; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,0]
-; SSE-NEXT:    movaps %xmm0, %xmm2
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm7[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm2[0,2]
-; SSE-NEXT:    movaps %xmm15, %xmm2
-; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm6[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[3,3],xmm6[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[1,1],xmm1[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps 96(%rdi), %xmm5
+; SSE-NEXT:    movaps 96(%rdx), %xmm1
+; SSE-NEXT:    movaps %xmm5, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[0,3]
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 96(%rsi), %xmm1
+; SSE-NEXT:    movaps %xmm5, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm1[0],xmm8[1],xmm1[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm5, %xmm0
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm2[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps 112(%rdi), %xmm3
+; SSE-NEXT:    movaps 112(%rdx), %xmm9
+; SSE-NEXT:    movaps %xmm3, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm9[0,3]
+; SSE-NEXT:    movaps 112(%rsi), %xmm0
+; SSE-NEXT:    movaps %xmm3, %xmm4
+; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm3, %xmm2
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,3],xmm0[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm9[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    shufps $233, (%rsp), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = xmm2[1,2],mem[2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[1,2],xmm14[2,3]
-; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps (%rsp), %xmm15 # 16-byte Reload
-; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm15 = xmm15[1,2],mem[2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
 ; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm14 = xmm14[1,2],mem[2,3]
-; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm9 = xmm9[1,2],mem[2,3]
+; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm13 = xmm13[1,2],mem[2,3]
+; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[1,2],mem[2,3]
+; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[1,2],mem[2,3]
 ; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm10 = xmm10[1,2],mem[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,2],xmm13[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,2],xmm7[2,3]
-; SSE-NEXT:    movaps %xmm1, 352(%rcx)
-; SSE-NEXT:    movaps %xmm3, 336(%rcx)
-; SSE-NEXT:    movaps %xmm5, 304(%rcx)
-; SSE-NEXT:    movaps %xmm6, 288(%rcx)
-; SSE-NEXT:    movaps %xmm8, 256(%rcx)
+; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = xmm5[1,2],mem[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,2],xmm9[2,3]
+; SSE-NEXT:    movaps %xmm0, 352(%rcx)
+; SSE-NEXT:    movaps %xmm4, 336(%rcx)
+; SSE-NEXT:    movaps %xmm1, 304(%rcx)
+; SSE-NEXT:    movaps %xmm8, 288(%rcx)
+; SSE-NEXT:    movaps %xmm6, 256(%rcx)
 ; SSE-NEXT:    movaps %xmm11, 240(%rcx)
-; SSE-NEXT:    movaps %xmm12, 208(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 192(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 160(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 144(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 112(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 96(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 64(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 48(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 16(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, (%rcx)
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0,1,3]
-; SSE-NEXT:    movaps %xmm0, 368(%rcx)
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0,1,3]
-; SSE-NEXT:    movaps %xmm4, 320(%rcx)
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[2,0,1,3]
-; SSE-NEXT:    movaps %xmm10, 272(%rcx)
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[2,0,1,3]
-; SSE-NEXT:    movaps %xmm9, 224(%rcx)
+; SSE-NEXT:    movaps %xmm7, 208(%rcx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0,1,3]
-; SSE-NEXT:    movaps %xmm0, 176(%rcx)
-; SSE-NEXT:    movaps %xmm15, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm15[1,3]
-; SSE-NEXT:    movaps %xmm0, 128(%rcx)
+; SSE-NEXT:    movaps %xmm0, 192(%rcx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0,1,3]
-; SSE-NEXT:    movaps %xmm0, 80(%rcx)
+; SSE-NEXT:    movaps %xmm0, 160(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 96(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 48(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rcx)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rcx)
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0,1,3]
+; SSE-NEXT:    movaps %xmm3, 368(%rcx)
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0,1,3]
+; SSE-NEXT:    movaps %xmm5, 320(%rcx)
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[2,0,1,3]
+; SSE-NEXT:    movaps %xmm10, 272(%rcx)
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[2,0,1,3]
+; SSE-NEXT:    movaps %xmm15, 224(%rcx)
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[2,0,1,3]
+; SSE-NEXT:    movaps %xmm12, 176(%rcx)
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[2,0,1,3]
+; SSE-NEXT:    movaps %xmm13, 128(%rcx)
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[2,0,1,3]
+; SSE-NEXT:    movaps %xmm14, 80(%rcx)
 ; SSE-NEXT:    movaps %xmm2, %xmm0
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm2[1,3]
 ; SSE-NEXT:    movaps %xmm0, 32(%rcx)
@@ -2131,75 +2123,75 @@ define void @store_i32_stride3_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; SSE-LABEL: store_i32_stride3_vf64:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    subq $664, %rsp # imm = 0x298
-; SSE-NEXT:    movaps (%rdi), %xmm2
-; SSE-NEXT:    movaps 16(%rdi), %xmm4
-; SSE-NEXT:    movaps 32(%rdi), %xmm5
-; SSE-NEXT:    movaps 48(%rdi), %xmm6
-; SSE-NEXT:    movaps (%rsi), %xmm0
+; SSE-NEXT:    movaps (%rdi), %xmm1
+; SSE-NEXT:    movaps 16(%rdi), %xmm2
+; SSE-NEXT:    movaps 32(%rdi), %xmm3
+; SSE-NEXT:    movaps 48(%rdi), %xmm4
+; SSE-NEXT:    movaps (%rsi), %xmm9
 ; SSE-NEXT:    movaps 16(%rsi), %xmm11
 ; SSE-NEXT:    movaps 32(%rsi), %xmm14
-; SSE-NEXT:    movaps 48(%rsi), %xmm3
-; SSE-NEXT:    movaps (%rdx), %xmm7
-; SSE-NEXT:    movaps 16(%rdx), %xmm8
-; SSE-NEXT:    movaps 32(%rdx), %xmm9
-; SSE-NEXT:    movaps 48(%rdx), %xmm10
-; SSE-NEXT:    movaps %xmm2, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm7[0,3]
-; SSE-NEXT:    movaps %xmm7, %xmm12
-; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm2, %xmm7
-; SSE-NEXT:    unpcklps {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm1[2,0]
-; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 48(%rsi), %xmm13
+; SSE-NEXT:    movaps (%rdx), %xmm5
+; SSE-NEXT:    movaps 16(%rdx), %xmm6
+; SSE-NEXT:    movaps 32(%rdx), %xmm7
+; SSE-NEXT:    movaps 48(%rdx), %xmm8
+; SSE-NEXT:    movaps %xmm1, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm5[0,3]
+; SSE-NEXT:    movaps %xmm5, %xmm10
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm1, %xmm5
+; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm9[0],xmm5[1],xmm9[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm1, %xmm0
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm9[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm9[3,3]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[1,1],xmm10[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm0
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm6[0,3]
 ; SSE-NEXT:    movaps %xmm2, %xmm1
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm0[3,3]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm12[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm4, %xmm0
-; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm8[0,3]
-; SSE-NEXT:    movaps %xmm4, %xmm1
 ; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm4, %xmm0
+; SSE-NEXT:    movaps %xmm2, %xmm0
 ; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm11[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,3],xmm11[3,3]
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[1,1],xmm8[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm11[3,3]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[1,1],xmm6[1,1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,2],xmm0[0,2]
 ; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm5, %xmm0
-; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm9[0,3]
-; SSE-NEXT:    movaps %xmm5, %xmm1
+; SSE-NEXT:    movaps %xmm3, %xmm0
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm7[0,3]
+; SSE-NEXT:    movaps %xmm3, %xmm1
 ; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm14[0],xmm1[1],xmm14[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm5, %xmm0
+; SSE-NEXT:    movaps %xmm3, %xmm0
 ; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm14[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,3],xmm14[3,3]
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[1,1],xmm9[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,3],xmm14[3,3]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[1,1],xmm7[1,1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,2],xmm0[0,2]
 ; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm6, %xmm0
-; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm10[0,3]
-; SSE-NEXT:    movaps %xmm6, %xmm1
-; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; SSE-NEXT:    movaps %xmm4, %xmm0
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm8[0,3]
+; SSE-NEXT:    movaps %xmm4, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm6, %xmm0
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[3,3],xmm3[3,3]
-; SSE-NEXT:    movaps %xmm6, (%rsp) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm10[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm4, %xmm0
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm13[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,3],xmm13[3,3]
+; SSE-NEXT:    movaps %xmm4, (%rsp) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[1,1],xmm8[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps 64(%rdi), %xmm2
 ; SSE-NEXT:    movaps 64(%rdx), %xmm1
 ; SSE-NEXT:    movaps %xmm2, %xmm0
@@ -2290,179 +2282,175 @@ define void @store_i32_stride3_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm4[1,1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 144(%rdi), %xmm2
-; SSE-NEXT:    movaps 144(%rdx), %xmm1
-; SSE-NEXT:    movaps %xmm2, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[0,3]
-; SSE-NEXT:    movaps %xmm1, %xmm4
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 144(%rsi), %xmm1
-; SSE-NEXT:    movaps %xmm2, %xmm3
-; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm2, %xmm0
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm1[3,3]
+; SSE-NEXT:    movaps 144(%rdi), %xmm1
+; SSE-NEXT:    movaps 144(%rdx), %xmm4
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm1, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm4[0,3]
+; SSE-NEXT:    movaps 144(%rsi), %xmm3
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm4[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 160(%rdi), %xmm14
-; SSE-NEXT:    movaps 160(%rdx), %xmm1
-; SSE-NEXT:    movaps %xmm14, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[0,3]
-; SSE-NEXT:    movaps %xmm1, %xmm3
+; SSE-NEXT:    movaps %xmm1, %xmm0
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm3[3,3]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 160(%rsi), %xmm1
-; SSE-NEXT:    movaps %xmm14, %xmm2
-; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm4[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 160(%rdi), %xmm15
+; SSE-NEXT:    movaps 160(%rdx), %xmm2
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm14, %xmm0
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[3,3],xmm1[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm3[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm15, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm2[0,3]
+; SSE-NEXT:    movaps 160(%rsi), %xmm13
+; SSE-NEXT:    movaps %xmm15, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm13[0],xmm1[1],xmm13[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm15, %xmm0
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm13[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[3,3],xmm13[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[1,1],xmm2[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,2],xmm0[0,2]
 ; SSE-NEXT:    movaps 176(%rdi), %xmm12
-; SSE-NEXT:    movaps 176(%rdx), %xmm1
-; SSE-NEXT:    movaps %xmm12, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[0,3]
-; SSE-NEXT:    movaps %xmm1, %xmm3
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 176(%rsi), %xmm1
-; SSE-NEXT:    movaps %xmm12, %xmm2
-; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
+; SSE-NEXT:    movaps 176(%rdx), %xmm2
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps %xmm12, %xmm0
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[3,3],xmm1[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm3[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 192(%rdi), %xmm13
-; SSE-NEXT:    movaps 192(%rdx), %xmm1
-; SSE-NEXT:    movaps %xmm13, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[0,3]
-; SSE-NEXT:    movaps %xmm1, %xmm2
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 192(%rsi), %xmm11
-; SSE-NEXT:    movaps %xmm13, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm2[0,3]
+; SSE-NEXT:    movaps 176(%rsi), %xmm11
+; SSE-NEXT:    movaps %xmm12, %xmm1
 ; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm13, %xmm0
+; SSE-NEXT:    movaps %xmm12, %xmm0
 ; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm11[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[3,3],xmm11[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[3,3],xmm11[3,3]
 ; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[1,1],xmm2[1,1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,2],xmm0[0,2]
-; SSE-NEXT:    movaps 208(%rdi), %xmm6
-; SSE-NEXT:    movaps 208(%rdx), %xmm1
-; SSE-NEXT:    movaps %xmm6, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[0,3]
+; SSE-NEXT:    movaps 192(%rdi), %xmm9
+; SSE-NEXT:    movaps 192(%rdx), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm9, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm2[0,3]
+; SSE-NEXT:    movaps 192(%rsi), %xmm8
+; SSE-NEXT:    movaps %xmm9, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm8[0],xmm1[1],xmm8[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 208(%rsi), %xmm8
-; SSE-NEXT:    movaps %xmm6, %xmm10
-; SSE-NEXT:    unpcklps {{.*#+}} xmm10 = xmm10[0],xmm8[0],xmm10[1],xmm8[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm6, %xmm0
+; SSE-NEXT:    movaps %xmm9, %xmm0
 ; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm8[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[3,3],xmm8[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[1,1],xmm1[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[3,3],xmm8[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[1,1],xmm2[1,1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm0[0,2]
-; SSE-NEXT:    movaps 224(%rdi), %xmm5
-; SSE-NEXT:    movaps 224(%rdx), %xmm15
-; SSE-NEXT:    movaps %xmm5, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm15[0,3]
-; SSE-NEXT:    movaps 224(%rsi), %xmm4
-; SSE-NEXT:    movaps %xmm5, %xmm7
-; SSE-NEXT:    unpcklps {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm5, %xmm0
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,3],xmm4[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm15[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm0[0,2]
-; SSE-NEXT:    movaps 240(%rdi), %xmm2
-; SSE-NEXT:    movaps 240(%rdx), %xmm9
-; SSE-NEXT:    movaps %xmm2, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm9[0,3]
-; SSE-NEXT:    movaps 240(%rsi), %xmm0
-; SSE-NEXT:    movaps %xmm2, %xmm3
-; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm1[2,0]
-; SSE-NEXT:    movaps %xmm2, %xmm1
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm0[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm9[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[1,2],mem[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[1,2],mem[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[1,2],mem[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps (%rsp), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[1,2],mem[2,3]
-; SSE-NEXT:    movaps %xmm1, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[1,2],mem[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[1,2],mem[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[1,2],mem[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[1,2],mem[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[1,2],mem[2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[1,2],mem[2,3]
+; SSE-NEXT:    movaps 208(%rdi), %xmm14
+; SSE-NEXT:    movaps 208(%rdx), %xmm1
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm14 = xmm14[1,2],mem[2,3]
+; SSE-NEXT:    movaps %xmm14, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[0,3]
+; SSE-NEXT:    movaps 208(%rsi), %xmm6
+; SSE-NEXT:    movaps %xmm14, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm14, %xmm0
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm6[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[3,3],xmm6[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[1,1],xmm1[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps 224(%rdi), %xmm4
+; SSE-NEXT:    movaps 224(%rdx), %xmm3
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm4, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm3[0,3]
+; SSE-NEXT:    movaps 224(%rsi), %xmm1
+; SSE-NEXT:    movaps %xmm4, %xmm10
+; SSE-NEXT:    unpcklps {{.*#+}} xmm10 = xmm10[0],xmm1[0],xmm10[1],xmm1[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,1],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm4, %xmm0
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm3[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps 240(%rdi), %xmm5
+; SSE-NEXT:    movaps 240(%rdx), %xmm2
+; SSE-NEXT:    movaps %xmm5, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm2[0,3]
+; SSE-NEXT:    movaps 240(%rsi), %xmm0
+; SSE-NEXT:    movaps %xmm5, %xmm7
+; SSE-NEXT:    unpcklps {{.*#+}} xmm7 = xmm7[0],xmm0[0],xmm7[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm3[2,0]
+; SSE-NEXT:    movaps %xmm5, %xmm3
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,3],xmm0[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm2[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm3[0,2]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[1,2],mem[2,3]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[1,2],mem[2,3]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[1,2],mem[2,3]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps (%rsp), %xmm3 # 16-byte Reload
+; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[1,2],mem[2,3]
+; SSE-NEXT:    movaps %xmm3, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[1,2],mem[2,3]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[1,2],mem[2,3]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[1,2],mem[2,3]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[1,2],mem[2,3]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[1,2],mem[2,3]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = xmm3[1,2],mem[2,3]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[1,2],mem[2,3]
 ; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm12 = xmm12[1,2],mem[2,3]
-; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm13 = xmm13[1,2],mem[2,3]
-; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm6 = xmm6[1,2],mem[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,2],xmm15[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,2],xmm9[2,3]
+; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = xmm9[1,2],mem[2,3]
+; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = xmm14[1,2],mem[2,3]
+; SSE-NEXT:    shufps $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm4 = xmm4[1,2],mem[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,2],xmm2[2,3]
 ; SSE-NEXT:    movaps %xmm0, 736(%rcx)
-; SSE-NEXT:    movaps %xmm3, 720(%rcx)
-; SSE-NEXT:    movaps %xmm4, 688(%rcx)
-; SSE-NEXT:    movaps %xmm7, 672(%rcx)
-; SSE-NEXT:    movaps %xmm8, 640(%rcx)
-; SSE-NEXT:    movaps %xmm10, 624(%rcx)
-; SSE-NEXT:    movaps %xmm11, 592(%rcx)
+; SSE-NEXT:    movaps %xmm7, 720(%rcx)
+; SSE-NEXT:    movaps %xmm1, 688(%rcx)
+; SSE-NEXT:    movaps %xmm10, 672(%rcx)
+; SSE-NEXT:    movaps %xmm6, 640(%rcx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 576(%rcx)
+; SSE-NEXT:    movaps %xmm0, 624(%rcx)
+; SSE-NEXT:    movaps %xmm8, 592(%rcx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 544(%rcx)
+; SSE-NEXT:    movaps %xmm0, 576(%rcx)
+; SSE-NEXT:    movaps %xmm11, 544(%rcx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 528(%rcx)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 496(%rcx)
+; SSE-NEXT:    movaps %xmm13, 496(%rcx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 480(%rcx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -2505,18 +2493,18 @@ define void @store_i32_stride3_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
 ; SSE-NEXT:    movaps %xmm0, 16(%rcx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, (%rcx)
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0,1,3]
-; SSE-NEXT:    movaps %xmm2, 752(%rcx)
 ; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0,1,3]
-; SSE-NEXT:    movaps %xmm5, 704(%rcx)
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0,1,3]
-; SSE-NEXT:    movaps %xmm6, 656(%rcx)
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[2,0,1,3]
-; SSE-NEXT:    movaps %xmm13, 608(%rcx)
+; SSE-NEXT:    movaps %xmm5, 752(%rcx)
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0,1,3]
+; SSE-NEXT:    movaps %xmm4, 704(%rcx)
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[2,0,1,3]
+; SSE-NEXT:    movaps %xmm14, 656(%rcx)
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[2,0,1,3]
+; SSE-NEXT:    movaps %xmm9, 608(%rcx)
 ; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[2,0,1,3]
 ; SSE-NEXT:    movaps %xmm12, 560(%rcx)
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[2,0,1,3]
-; SSE-NEXT:    movaps %xmm14, 512(%rcx)
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[2,0,1,3]
+; SSE-NEXT:    movaps %xmm15, 512(%rcx)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0,1,3]
 ; SSE-NEXT:    movaps %xmm0, 464(%rcx)
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-6.ll
index 0cd72be39557a..4d1216ca8fe4c 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-6.ll
@@ -697,76 +697,73 @@ define void @store_i32_stride6_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 define void @store_i32_stride6_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %in.vecptr3, ptr %in.vecptr4, ptr %in.vecptr5, ptr %out.vec) nounwind {
 ; SSE-LABEL: store_i32_stride6_vf8:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    movaps (%rdi), %xmm4
+; SSE-NEXT:    movaps (%rdi), %xmm0
 ; SSE-NEXT:    movaps 16(%rdi), %xmm1
-; SSE-NEXT:    movaps (%rsi), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 16(%rsi), %xmm10
-; SSE-NEXT:    movaps (%rdx), %xmm8
+; SSE-NEXT:    movaps (%rsi), %xmm10
+; SSE-NEXT:    movaps 16(%rsi), %xmm12
+; SSE-NEXT:    movaps (%rdx), %xmm9
 ; SSE-NEXT:    movaps 16(%rdx), %xmm2
-; SSE-NEXT:    movaps (%rcx), %xmm6
-; SSE-NEXT:    movaps 16(%rcx), %xmm9
+; SSE-NEXT:    movaps (%rcx), %xmm8
+; SSE-NEXT:    movaps 16(%rcx), %xmm15
 ; SSE-NEXT:    movaps (%r8), %xmm5
-; SSE-NEXT:    movaps 16(%r8), %xmm11
+; SSE-NEXT:    movaps 16(%r8), %xmm13
 ; SSE-NEXT:    movaps (%r9), %xmm7
-; SSE-NEXT:    movaps 16(%r9), %xmm3
-; SSE-NEXT:    movaps %xmm9, %xmm14
+; SSE-NEXT:    movaps 16(%r9), %xmm4
+; SSE-NEXT:    movaps %xmm15, %xmm14
 ; SSE-NEXT:    unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm2[1]
-; SSE-NEXT:    movaps %xmm1, %xmm12
-; SSE-NEXT:    unpckhps {{.*#+}} xmm12 = xmm12[2],xmm10[2],xmm12[3],xmm10[3]
-; SSE-NEXT:    movaps %xmm11, %xmm13
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm13 = xmm13[1],xmm3[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,2],xmm12[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,1],xmm14[2,0]
-; SSE-NEXT:    movaps %xmm11, %xmm14
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[1,1],xmm3[1,1]
-; SSE-NEXT:    movaps %xmm2, %xmm15
-; SSE-NEXT:    unpcklps {{.*#+}} xmm15 = xmm15[0],xmm9[0],xmm15[1],xmm9[1]
-; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm10[0],xmm1[1],xmm10[1]
-; SSE-NEXT:    movaps %xmm11, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm3[3,3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm11[0]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm1[2,3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm15[0]
-; SSE-NEXT:    movaps %xmm15, %xmm10
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[2,3],xmm14[0,2]
-; SSE-NEXT:    movaps %xmm5, %xmm14
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[3,3],xmm7[3,3]
-; SSE-NEXT:    movaps %xmm8, %xmm11
-; SSE-NEXT:    unpckhps {{.*#+}} xmm11 = xmm11[2],xmm6[2],xmm11[3],xmm6[3]
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[2,3],xmm14[0,2]
-; SSE-NEXT:    movaps %xmm4, %xmm14
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{.*#+}} xmm14 = xmm14[2],xmm15[2],xmm14[3],xmm15[3]
-; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm15[0],xmm4[1],xmm15[1]
-; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm9[2],xmm2[3],xmm9[3]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,3],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm8, %xmm0
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm8[1]
-; SSE-NEXT:    movaps %xmm5, %xmm8
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm7[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm14[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm6[2,0]
-; SSE-NEXT:    movaps %xmm7, %xmm6
-; SSE-NEXT:    movlhps {{.*#+}} xmm6 = xmm6[0],xmm5[0]
+; SSE-NEXT:    movaps %xmm1, %xmm3
+; SSE-NEXT:    unpckhps {{.*#+}} xmm3 = xmm3[2],xmm12[2],xmm3[3],xmm12[3]
+; SSE-NEXT:    movaps %xmm5, %xmm11
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[3,3],xmm7[3,3]
+; SSE-NEXT:    movaps %xmm9, %xmm6
+; SSE-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm8[2],xmm6[3],xmm8[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,3],xmm11[0,2]
+; SSE-NEXT:    movaps %xmm0, %xmm11
+; SSE-NEXT:    unpckhps {{.*#+}} xmm11 = xmm11[2],xmm10[2],xmm11[3],xmm10[3]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
+; SSE-NEXT:    movaps %xmm2, %xmm10
+; SSE-NEXT:    unpcklps {{.*#+}} xmm10 = xmm10[0],xmm15[0],xmm10[1],xmm15[1]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm15[2],xmm2[3],xmm15[3]
+; SSE-NEXT:    movaps %xmm13, %xmm15
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[3,3],xmm4[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,3],xmm15[0,2]
+; SSE-NEXT:    movaps %xmm13, %xmm15
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm15 = xmm15[1],xmm4[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,2],xmm3[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm14[2,0]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1]
+; SSE-NEXT:    movaps %xmm13, %xmm12
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[1,1],xmm4[1,1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm13[0]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm1[2,3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm10[0]
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[2,3],xmm12[0,2]
+; SSE-NEXT:    movaps %xmm9, %xmm12
+; SSE-NEXT:    unpcklps {{.*#+}} xmm12 = xmm12[0],xmm8[0],xmm12[1],xmm8[1]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm9[1]
+; SSE-NEXT:    movaps %xmm5, %xmm9
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm9 = xmm9[1],xmm7[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,2],xmm11[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,1],xmm8[2,0]
+; SSE-NEXT:    movaps %xmm7, %xmm8
+; SSE-NEXT:    movlhps {{.*#+}} xmm8 = xmm8[0],xmm5[0]
 ; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0],xmm4[2,3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm0[0]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,3],xmm5[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,0],xmm0[2,3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm12[0]
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[2,3],xmm5[0,2]
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movaps %xmm0, 32(%rax)
-; SSE-NEXT:    movaps %xmm14, 48(%rax)
+; SSE-NEXT:    movaps %xmm12, 32(%rax)
+; SSE-NEXT:    movaps %xmm11, 48(%rax)
 ; SSE-NEXT:    movaps %xmm1, 96(%rax)
-; SSE-NEXT:    movaps %xmm3, 112(%rax)
-; SSE-NEXT:    movaps %xmm13, 160(%rax)
+; SSE-NEXT:    movaps %xmm4, 112(%rax)
+; SSE-NEXT:    movaps %xmm15, 160(%rax)
 ; SSE-NEXT:    movaps %xmm2, 176(%rax)
-; SSE-NEXT:    movaps %xmm4, (%rax)
-; SSE-NEXT:    movaps %xmm6, 16(%rax)
-; SSE-NEXT:    movaps %xmm8, 64(%rax)
-; SSE-NEXT:    movaps %xmm11, 80(%rax)
+; SSE-NEXT:    movaps %xmm0, (%rax)
+; SSE-NEXT:    movaps %xmm8, 16(%rax)
+; SSE-NEXT:    movaps %xmm9, 64(%rax)
+; SSE-NEXT:    movaps %xmm6, 80(%rax)
 ; SSE-NEXT:    movaps %xmm10, 128(%rax)
-; SSE-NEXT:    movaps %xmm12, 144(%rax)
+; SSE-NEXT:    movaps %xmm3, 144(%rax)
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: store_i32_stride6_vf8:
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
index 0495e240ba968..c72b8518900e5 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
@@ -690,74 +690,74 @@ define void @store_i8_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; SSE-NEXT:    movq {{.*#+}} xmm3 = mem[0],zero
-; SSE-NEXT:    movq {{.*#+}} xmm10 = mem[0],zero
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movq {{.*#+}} xmm9 = mem[0],zero
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movq {{.*#+}} xmm2 = mem[0],zero
-; SSE-NEXT:    movq {{.*#+}} xmm4 = mem[0],zero
+; SSE-NEXT:    movq {{.*#+}} xmm5 = mem[0],zero
 ; SSE-NEXT:    movq {{.*#+}} xmm14 = mem[0],zero
 ; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movq {{.*#+}} xmm0 = mem[0],zero
-; SSE-NEXT:    movq {{.*#+}} xmm5 = mem[0],zero
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm0[0]
+; SSE-NEXT:    movq {{.*#+}} xmm4 = mem[0],zero
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm0[0]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
 ; SSE-NEXT:    pand %xmm6, %xmm0
-; SSE-NEXT:    movdqa %xmm4, %xmm7
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1],xmm7[2],xmm4[2],xmm7[3],xmm4[3],xmm7[4],xmm4[4],xmm7[5],xmm4[5],xmm7[6],xmm4[6],xmm7[7],xmm4[7]
+; SSE-NEXT:    movdqa %xmm5, %xmm7
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm5[0],xmm7[1],xmm5[1],xmm7[2],xmm5[2],xmm7[3],xmm5[3],xmm7[4],xmm5[4],xmm7[5],xmm5[5],xmm7[6],xmm5[6],xmm7[7],xmm5[7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm7[0,0,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,1,1,3]
 ; SSE-NEXT:    pandn %xmm8, %xmm6
 ; SSE-NEXT:    por %xmm0, %xmm6
-; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255]
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm3[0,0,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0]
 ; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm10 = xmm10[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm10[0,2,1,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[0,1,1,0]
-; SSE-NEXT:    movdqa %xmm8, %xmm12
-; SSE-NEXT:    pandn %xmm11, %xmm12
-; SSE-NEXT:    por %xmm0, %xmm12
-; SSE-NEXT:    pand %xmm9, %xmm12
-; SSE-NEXT:    pandn %xmm6, %xmm9
-; SSE-NEXT:    por %xmm12, %xmm9
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm9[0,2,1,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,1,1,0]
+; SSE-NEXT:    movdqa %xmm8, %xmm11
+; SSE-NEXT:    pandn %xmm10, %xmm11
+; SSE-NEXT:    por %xmm0, %xmm11
+; SSE-NEXT:    pand %xmm12, %xmm11
+; SSE-NEXT:    pandn %xmm6, %xmm12
+; SSE-NEXT:    por %xmm11, %xmm12
 ; SSE-NEXT:    pxor %xmm0, %xmm0
-; SSE-NEXT:    movdqa %xmm5, %xmm12
-; SSE-NEXT:    movdqa %xmm5, %xmm15
+; SSE-NEXT:    movdqa %xmm4, %xmm11
+; SSE-NEXT:    movdqa %xmm4, %xmm15
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm15 = xmm15[8],xmm0[8],xmm15[9],xmm0[9],xmm15[10],xmm0[10],xmm15[11],xmm0[11],xmm15[12],xmm0[12],xmm15[13],xmm0[13],xmm15[14],xmm0[14],xmm15[15],xmm0[15]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1],xmm12[2],xmm0[2],xmm12[3],xmm0[3],xmm12[4],xmm0[4],xmm12[5],xmm0[5],xmm12[6],xmm0[6],xmm12[7],xmm0[7]
-; SSE-NEXT:    movdqa %xmm12, %xmm13
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm0[0],xmm11[1],xmm0[1],xmm11[2],xmm0[2],xmm11[3],xmm0[3],xmm11[4],xmm0[4],xmm11[5],xmm0[5],xmm11[6],xmm0[6],xmm11[7],xmm0[7]
+; SSE-NEXT:    movdqa %xmm11, %xmm13
 ; SSE-NEXT:    punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm15[0],xmm13[1],xmm15[1],xmm13[2],xmm15[2],xmm13[3],xmm15[3]
 ; SSE-NEXT:    movdqa %xmm13, %xmm0
 ; SSE-NEXT:    pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
 ; SSE-NEXT:    movdqa %xmm13, %xmm6
 ; SSE-NEXT:    packuswb %xmm0, %xmm6
-; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255]
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm14[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm11, %xmm14
+; SSE-NEXT:    movdqa %xmm10, %xmm14
 ; SSE-NEXT:    pandn %xmm0, %xmm14
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[0,0,2,3]
-; SSE-NEXT:    pand %xmm11, %xmm0
+; SSE-NEXT:    pand %xmm10, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm14
 ; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255]
-; SSE-NEXT:    pand %xmm6, %xmm9
+; SSE-NEXT:    pand %xmm6, %xmm12
 ; SSE-NEXT:    pandn %xmm14, %xmm6
-; SSE-NEXT:    por %xmm9, %xmm6
+; SSE-NEXT:    por %xmm12, %xmm6
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm11, %xmm9
-; SSE-NEXT:    pandn %xmm0, %xmm9
+; SSE-NEXT:    movdqa %xmm10, %xmm12
+; SSE-NEXT:    pandn %xmm0, %xmm12
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[2,1,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    por %xmm9, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255]
-; SSE-NEXT:    movdqa %xmm1, %xmm9
-; SSE-NEXT:    pandn %xmm0, %xmm9
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm10[0,1,2,3,5,6,6,7]
+; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    por %xmm12, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255]
+; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    pandn %xmm0, %xmm1
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm9[0,1,2,3,5,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,2,2,2]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255]
 ; SSE-NEXT:    movdqa %xmm0, %xmm14
@@ -766,85 +766,85 @@ define void @store_i8_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,1,3]
 ; SSE-NEXT:    pand %xmm0, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm14
-; SSE-NEXT:    pand %xmm1, %xmm14
-; SSE-NEXT:    por %xmm9, %xmm14
-; SSE-NEXT:    movdqa %xmm15, %xmm2
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm12[4],xmm2[5],xmm12[5],xmm2[6],xmm12[6],xmm2[7],xmm12[7]
-; SSE-NEXT:    punpckhwd {{.*#+}} xmm12 = xmm12[4],xmm15[4],xmm12[5],xmm15[5],xmm12[6],xmm15[6],xmm12[7],xmm15[7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,5,4]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm12[0,1,2,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,1,2,1]
-; SSE-NEXT:    packuswb %xmm2, %xmm9
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm12, %xmm9
+; SSE-NEXT:    pand %xmm12, %xmm14
+; SSE-NEXT:    por %xmm1, %xmm14
+; SSE-NEXT:    movdqa %xmm15, %xmm1
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm11[4],xmm1[5],xmm11[5],xmm1[6],xmm11[6],xmm1[7],xmm11[7]
+; SSE-NEXT:    punpckhwd {{.*#+}} xmm11 = xmm11[4],xmm15[4],xmm11[5],xmm15[5],xmm11[6],xmm15[6],xmm11[7],xmm15[7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,4]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm11[0,1,2,2,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1]
+; SSE-NEXT:    packuswb %xmm1, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm1, %xmm2
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm15[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
-; SSE-NEXT:    pandn %xmm2, %xmm12
-; SSE-NEXT:    por %xmm9, %xmm12
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,0,0,0,0,255,255,255,0,0,0,0,255,255]
-; SSE-NEXT:    pand %xmm2, %xmm12
-; SSE-NEXT:    pandn %xmm14, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm12
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm13[2,2,3,3]
-; SSE-NEXT:    psrldq {{.*#+}} xmm13 = xmm13[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    packuswb %xmm13, %xmm2
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm15[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,0,0,0]
-; SSE-NEXT:    pandn %xmm9, %xmm0
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm10[0,1,2,2]
-; SSE-NEXT:    movdqa %xmm3, %xmm10
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm3[1,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,0,2,1]
-; SSE-NEXT:    pand %xmm11, %xmm9
-; SSE-NEXT:    pandn %xmm2, %xmm11
-; SSE-NEXT:    por %xmm9, %xmm11
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm3[1,1,2,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,2,1]
-; SSE-NEXT:    pand %xmm8, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[1,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,7,5,6,4]
-; SSE-NEXT:    pandn %xmm7, %xmm8
-; SSE-NEXT:    por %xmm2, %xmm8
-; SSE-NEXT:    pand %xmm1, %xmm8
+; SSE-NEXT:    pshuflw {{.*#+}} xmm11 = xmm15[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm11 = xmm11[0,0,0,0]
 ; SSE-NEXT:    pandn %xmm11, %xmm1
-; SSE-NEXT:    por %xmm8, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255]
-; SSE-NEXT:    pand %xmm2, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3],xmm4[4],xmm3[4],xmm4[5],xmm3[5],xmm4[6],xmm3[6],xmm4[7],xmm3[7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,2,2,2]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm10[0],xmm4[1],xmm10[1],xmm4[2],xmm10[2],xmm4[3],xmm10[3],xmm4[4],xmm10[4],xmm4[5],xmm10[5],xmm4[6],xmm10[6],xmm4[7],xmm10[7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,2,2,2]
-; SSE-NEXT:    pand %xmm0, %xmm4
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [255,255,255,0,0,0,0,255,255,255,0,0,0,0,255,255]
+; SSE-NEXT:    pand %xmm11, %xmm1
+; SSE-NEXT:    pandn %xmm14, %xmm11
+; SSE-NEXT:    por %xmm1, %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[2,2,3,3]
+; SSE-NEXT:    psrldq {{.*#+}} xmm13 = xmm13[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT:    packuswb %xmm13, %xmm1
+; SSE-NEXT:    pand %xmm0, %xmm1
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm15[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
 ; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[3,1,2,3]
-; SSE-NEXT:    por %xmm4, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm15[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm4, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[1,3,2,3,4,5,6,7]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm2[0,1,3,1,4,5,6,7]
-; SSE-NEXT:    pandn %xmm2, %xmm4
-; SSE-NEXT:    por %xmm3, %xmm4
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [0,255,255,255,255,0,0,0,255,255,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm2, %xmm0
-; SSE-NEXT:    pandn %xmm4, %xmm2
-; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    por %xmm1, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[0,1,2,2]
+; SSE-NEXT:    movdqa %xmm3, %xmm9
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm3[1,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,2,1]
+; SSE-NEXT:    pand %xmm10, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm10
+; SSE-NEXT:    por %xmm2, %xmm10
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm3[1,1,2,2,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,2,1]
+; SSE-NEXT:    pand %xmm8, %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[1,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,4]
+; SSE-NEXT:    pandn %xmm2, %xmm8
+; SSE-NEXT:    por %xmm1, %xmm8
+; SSE-NEXT:    pand %xmm12, %xmm8
+; SSE-NEXT:    pandn %xmm10, %xmm12
+; SSE-NEXT:    por %xmm8, %xmm12
+; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255]
+; SSE-NEXT:    pand %xmm7, %xmm12
+; SSE-NEXT:    pandn %xmm0, %xmm7
+; SSE-NEXT:    por %xmm12, %xmm7
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1],xmm5[2],xmm3[2],xmm5[3],xmm3[3],xmm5[4],xmm3[4],xmm5[5],xmm3[5],xmm5[6],xmm3[6],xmm5[7],xmm3[7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm5[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,2,2,2]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm9[0],xmm2[1],xmm9[1],xmm2[2],xmm9[2],xmm2[3],xmm9[3],xmm2[4],xmm9[4],xmm2[5],xmm9[5],xmm2[6],xmm9[6],xmm2[7],xmm9[7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,2,2,2]
+; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[3,1,2,3]
+; SSE-NEXT:    por %xmm2, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm15[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm3, %xmm2
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,1,4,5,6,7]
+; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    por %xmm2, %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [0,255,255,255,255,0,0,0,255,255,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    pandn %xmm3, %xmm1
+; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movq %xmm2, 48(%rax)
-; SSE-NEXT:    movdqa %xmm1, 16(%rax)
-; SSE-NEXT:    movdqa %xmm12, 32(%rax)
+; SSE-NEXT:    movq %xmm1, 48(%rax)
+; SSE-NEXT:    movdqa %xmm7, 16(%rax)
+; SSE-NEXT:    movdqa %xmm11, 32(%rax)
 ; SSE-NEXT:    movdqa %xmm6, (%rax)
 ; SSE-NEXT:    retq
 ;
@@ -1380,113 +1380,110 @@ define void @store_i8_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    movdqa (%rsi), %xmm4
 ; SSE-NEXT:    movdqa (%rdx), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rcx), %xmm5
+; SSE-NEXT:    movdqa (%rcx), %xmm6
 ; SSE-NEXT:    movdqa (%r8), %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%r9), %xmm8
+; SSE-NEXT:    movdqa (%r9), %xmm13
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
-; SSE-NEXT:    pand %xmm13, %xmm0
-; SSE-NEXT:    movdqa %xmm5, %xmm1
-; SSE-NEXT:    movdqa %xmm5, %xmm6
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm5[8],xmm1[9],xmm5[9],xmm1[10],xmm5[10],xmm1[11],xmm5[11],xmm1[12],xmm5[12],xmm1[13],xmm5[13],xmm1[14],xmm5[14],xmm1[15],xmm5[15]
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
+; SSE-NEXT:    pand %xmm12, %xmm0
+; SSE-NEXT:    movdqa %xmm6, %xmm1
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm6[8],xmm1[9],xmm6[9],xmm1[10],xmm6[10],xmm1[11],xmm6[11],xmm1[12],xmm6[12],xmm1[13],xmm6[13],xmm1[14],xmm6[14],xmm1[15],xmm6[15]
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE-NEXT:    movdqa %xmm13, %xmm2
+; SSE-NEXT:    movdqa %xmm12, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
 ; SSE-NEXT:    por %xmm0, %xmm2
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255]
 ; SSE-NEXT:    movdqa %xmm0, %xmm1
 ; SSE-NEXT:    pandn %xmm2, %xmm1
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    movdqa %xmm3, %xmm12
+; SSE-NEXT:    movdqa %xmm3, %xmm11
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[2,2,2,2]
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm10, %xmm3
-; SSE-NEXT:    movdqa %xmm4, %xmm9
-; SSE-NEXT:    movdqa %xmm4, %xmm5
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm8, %xmm3
+; SSE-NEXT:    movdqa %xmm4, %xmm2
+; SSE-NEXT:    movdqa %xmm4, %xmm14
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm4[8],xmm9[9],xmm4[9],xmm9[10],xmm4[10],xmm9[11],xmm4[11],xmm9[12],xmm4[12],xmm9[13],xmm4[13],xmm9[14],xmm4[14],xmm9[15],xmm4[15]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm9[0,1,2,3,4,6,6,7]
-; SSE-NEXT:    movdqa %xmm9, (%rsp) # 16-byte Spill
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm4[8],xmm2[9],xmm4[9],xmm2[10],xmm4[10],xmm2[11],xmm4[11],xmm2[12],xmm4[12],xmm2[13],xmm4[13],xmm2[14],xmm4[14],xmm2[15],xmm4[15]
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm2[0,1,2,3,4,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,1,3,3]
+; SSE-NEXT:    movdqa %xmm8, %xmm10
 ; SSE-NEXT:    pandn %xmm4, %xmm10
 ; SSE-NEXT:    por %xmm3, %xmm10
 ; SSE-NEXT:    pand %xmm0, %xmm10
 ; SSE-NEXT:    por %xmm1, %xmm10
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm7[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255]
-; SSE-NEXT:    movdqa %xmm2, %xmm4
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255]
+; SSE-NEXT:    movdqa %xmm5, %xmm4
 ; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm8, %xmm1
-; SSE-NEXT:    movdqa %xmm8, %xmm3
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm8[8],xmm1[9],xmm8[9],xmm1[10],xmm8[10],xmm1[11],xmm8[11],xmm1[12],xmm8[12],xmm1[13],xmm8[13],xmm1[14],xmm8[14],xmm1[15],xmm8[15]
+; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    movdqa %xmm13, %xmm15
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm13[8],xmm1[9],xmm13[9],xmm1[10],xmm13[10],xmm1[11],xmm13[11],xmm1[12],xmm13[12],xmm1[13],xmm13[13],xmm1[14],xmm13[14],xmm1[15],xmm13[15]
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,3]
-; SSE-NEXT:    pand %xmm2, %xmm1
+; SSE-NEXT:    pand %xmm5, %xmm1
 ; SSE-NEXT:    por %xmm4, %xmm1
 ; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [0,255,255,255,255,255,0,0,255,255,255,255,255,0,0,255]
 ; SSE-NEXT:    pand %xmm4, %xmm10
 ; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    movdqa (%rax), %xmm7
+; SSE-NEXT:    movdqa (%rax), %xmm9
 ; SSE-NEXT:    por %xmm10, %xmm4
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm7[0,1,2,3,6,7,7,7]
-; SSE-NEXT:    movdqa %xmm7, %xmm15
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm9[0,1,2,3,6,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,3,2]
-; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0]
-; SSE-NEXT:    movdqa %xmm11, %xmm7
-; SSE-NEXT:    pandn %xmm1, %xmm7
-; SSE-NEXT:    pand %xmm11, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[2,1,2,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0]
+; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    pand %xmm10, %xmm4
+; SSE-NEXT:    por %xmm4, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[2,1,2,3]
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,0,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm10, %xmm4
+; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm13, %xmm4
 ; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm12[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm12, %xmm5
+; SSE-NEXT:    movdqa %xmm11, %xmm2
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm11[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm10, %xmm1
+; SSE-NEXT:    pand %xmm13, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm4
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[2,1,2,3]
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm12, %xmm14
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm11, %xmm14
 ; SSE-NEXT:    pandn %xmm1, %xmm14
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm8[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm3[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm12, %xmm1
+; SSE-NEXT:    pand %xmm11, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm14
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255]
 ; SSE-NEXT:    pand %xmm1, %xmm4
 ; SSE-NEXT:    pandn %xmm14, %xmm1
 ; SSE-NEXT:    por %xmm4, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[2,1,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm15[2,1,2,3]
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,3,2,0]
-; SSE-NEXT:    movdqa %xmm2, %xmm14
+; SSE-NEXT:    movdqa %xmm5, %xmm14
 ; SSE-NEXT:    pandn %xmm4, %xmm14
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm6[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm7[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm7, %xmm15
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm2, %xmm4
+; SSE-NEXT:    pand %xmm5, %xmm4
 ; SSE-NEXT:    por %xmm4, %xmm14
 ; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255]
 ; SSE-NEXT:    pand %xmm4, %xmm14
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm15[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm15, %xmm3
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm9[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,4,4,4]
 ; SSE-NEXT:    pandn %xmm7, %xmm4
 ; SSE-NEXT:    por %xmm14, %xmm4
@@ -1495,72 +1492,69 @@ define void @store_i8_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    pandn %xmm4, %xmm7
 ; SSE-NEXT:    por %xmm1, %xmm7
 ; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm9[0,1,2,3,4,5,5,7]
+; SSE-NEXT:    pshufhw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,1,2,3,4,5,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE-NEXT:    movdqa %xmm13, %xmm4
+; SSE-NEXT:    movdqa %xmm12, %xmm4
 ; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm5, %xmm15
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm5[0,1,2,3,6,6,6,6]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm2[0,1,2,3,6,6,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
-; SSE-NEXT:    pand %xmm13, %xmm1
+; SSE-NEXT:    pand %xmm12, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm8[0,1,2,3,6,6,6,6]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,6,6,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm2, %xmm7
+; SSE-NEXT:    movdqa %xmm5, %xmm7
 ; SSE-NEXT:    pandn %xmm1, %xmm7
 ; SSE-NEXT:    pshufd $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pand %xmm2, %xmm1
+; SSE-NEXT:    pand %xmm5, %xmm1
 ; SSE-NEXT:    por %xmm7, %xmm1
 ; SSE-NEXT:    pand %xmm0, %xmm1
 ; SSE-NEXT:    pandn %xmm4, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    pshufd $229, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[1,1,2,3]
-; SSE-NEXT:    movdqa %xmm10, %xmm4
+; SSE-NEXT:    movdqa %xmm13, %xmm4
 ; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm6[0,1,2,3,5,6,6,7]
-; SSE-NEXT:    movdqa %xmm6, %xmm9
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm15[0,1,2,3,5,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,3,2]
-; SSE-NEXT:    pand %xmm10, %xmm1
+; SSE-NEXT:    pand %xmm13, %xmm1
 ; SSE-NEXT:    por %xmm4, %xmm1
-; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm3[0,1,2,3,4,5,6,6]
-; SSE-NEXT:    movdqa %xmm3, %xmm6
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm9[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    movdqa %xmm9, %xmm14
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,1,3,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm5, %xmm7
+; SSE-NEXT:    movdqa %xmm8, %xmm7
 ; SSE-NEXT:    pandn %xmm4, %xmm7
-; SSE-NEXT:    pand %xmm5, %xmm1
+; SSE-NEXT:    pand %xmm8, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm7
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0]
 ; SSE-NEXT:    pand %xmm1, %xmm0
 ; SSE-NEXT:    pandn %xmm7, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm1, (%rsp) # 16-byte Spill
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,1,3]
-; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    movdqa %xmm8, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm8[0,0,0,0,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm8, %xmm3
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm3[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    pand %xmm8, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,2,1,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,1,0]
-; SSE-NEXT:    movdqa %xmm11, %xmm7
+; SSE-NEXT:    movdqa %xmm10, %xmm7
 ; SSE-NEXT:    pandn %xmm0, %xmm7
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm15[0,0,2,1,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[0,0,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
-; SSE-NEXT:    pand %xmm11, %xmm0
+; SSE-NEXT:    pand %xmm10, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm7
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255]
 ; SSE-NEXT:    pand %xmm0, %xmm7
@@ -1571,41 +1565,40 @@ define void @store_i8_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[0,1,1,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,2,1]
-; SSE-NEXT:    movdqa %xmm13, %xmm7
+; SSE-NEXT:    movdqa %xmm12, %xmm7
 ; SSE-NEXT:    pandn %xmm1, %xmm7
-; SSE-NEXT:    movdqa %xmm9, %xmm5
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm9[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm15[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
-; SSE-NEXT:    pand %xmm13, %xmm1
+; SSE-NEXT:    pand %xmm12, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm7
-; SSE-NEXT:    movdqa %xmm6, %xmm8
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm6[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm9[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm2, %xmm9
+; SSE-NEXT:    movdqa %xmm5, %xmm9
 ; SSE-NEXT:    pandn %xmm1, %xmm9
-; SSE-NEXT:    pand %xmm2, %xmm7
+; SSE-NEXT:    pand %xmm5, %xmm7
 ; SSE-NEXT:    por %xmm7, %xmm9
-; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255]
-; SSE-NEXT:    pand %xmm14, %xmm0
-; SSE-NEXT:    pandn %xmm9, %xmm14
-; SSE-NEXT:    por %xmm0, %xmm14
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255]
+; SSE-NEXT:    pand %xmm8, %xmm0
+; SSE-NEXT:    pandn %xmm9, %xmm8
+; SSE-NEXT:    por %xmm0, %xmm8
 ; SSE-NEXT:    pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[0,1,1,3]
-; SSE-NEXT:    movdqa %xmm10, %xmm7
+; SSE-NEXT:    movdqa %xmm13, %xmm7
 ; SSE-NEXT:    pandn %xmm0, %xmm7
-; SSE-NEXT:    movdqa %xmm3, %xmm1
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm3[0,1,2,3,4,5,5,7]
+; SSE-NEXT:    movdqa %xmm3, %xmm1
 ; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm0[2,1,3,2]
-; SSE-NEXT:    pand %xmm10, %xmm9
+; SSE-NEXT:    pand %xmm13, %xmm9
 ; SSE-NEXT:    por %xmm7, %xmm9
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm15[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm2[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm2, %xmm7
+; SSE-NEXT:    movdqa %xmm5, %xmm7
 ; SSE-NEXT:    pandn %xmm0, %xmm7
-; SSE-NEXT:    pshuflw $233, (%rsp), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    pshuflw $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[1,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm0[0,0,2,1]
-; SSE-NEXT:    pand %xmm2, %xmm6
+; SSE-NEXT:    pand %xmm5, %xmm6
 ; SSE-NEXT:    por %xmm7, %xmm6
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0]
 ; SSE-NEXT:    pand %xmm0, %xmm6
@@ -1614,29 +1607,28 @@ define void @store_i8_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    pshuflw $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm3 = mem[1,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm12, %xmm6
+; SSE-NEXT:    movdqa %xmm11, %xmm6
 ; SSE-NEXT:    pandn %xmm3, %xmm6
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm5[0,1,2,3,4,4,6,5]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm15[0,1,2,3,4,4,6,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,3,3]
-; SSE-NEXT:    pand %xmm12, %xmm3
+; SSE-NEXT:    pand %xmm11, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm6
-; SSE-NEXT:    pand %xmm13, %xmm6
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm8[0,1,2,3,4,5,5,7]
+; SSE-NEXT:    pand %xmm12, %xmm6
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm14[0,1,2,3,4,5,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
-; SSE-NEXT:    pandn %xmm3, %xmm13
-; SSE-NEXT:    por %xmm6, %xmm13
+; SSE-NEXT:    pandn %xmm3, %xmm12
+; SSE-NEXT:    por %xmm6, %xmm12
 ; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255]
 ; SSE-NEXT:    pand %xmm3, %xmm0
-; SSE-NEXT:    pandn %xmm13, %xmm3
+; SSE-NEXT:    pandn %xmm12, %xmm3
 ; SSE-NEXT:    por %xmm0, %xmm3
 ; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm2, %xmm6
+; SSE-NEXT:    movdqa %xmm5, %xmm6
 ; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    movdqa %xmm1, %xmm13
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[2,1,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
-; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    pand %xmm5, %xmm0
 ; SSE-NEXT:    por %xmm6, %xmm0
 ; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255]
 ; SSE-NEXT:    movdqa %xmm6, %xmm7
@@ -1644,73 +1636,75 @@ define void @store_i8_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,5,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm0[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    movdqa %xmm11, %xmm0
 ; SSE-NEXT:    pandn %xmm9, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm15[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm2[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm2, %xmm15
 ; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,1,1,3]
-; SSE-NEXT:    pand %xmm12, %xmm9
+; SSE-NEXT:    pand %xmm11, %xmm9
 ; SSE-NEXT:    por %xmm9, %xmm0
 ; SSE-NEXT:    pand %xmm6, %xmm0
 ; SSE-NEXT:    por %xmm7, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm5[0,1,2,3,4,6,5,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm4[0,1,2,3,4,6,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[2,1,3,2]
-; SSE-NEXT:    movdqa %xmm11, %xmm9
+; SSE-NEXT:    movdqa %xmm10, %xmm9
 ; SSE-NEXT:    pandn %xmm7, %xmm9
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm4[2,2,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm2[2,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,0,2,1]
-; SSE-NEXT:    pand %xmm11, %xmm7
+; SSE-NEXT:    pand %xmm10, %xmm7
 ; SSE-NEXT:    por %xmm7, %xmm9
-; SSE-NEXT:    pand %xmm10, %xmm9
-; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm8[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pand %xmm13, %xmm9
+; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm14[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,0,0,0]
-; SSE-NEXT:    pandn %xmm7, %xmm10
-; SSE-NEXT:    por %xmm9, %xmm10
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,255,0,0,0,0,255,255,255,0,0,0,0,255,255]
-; SSE-NEXT:    pand %xmm7, %xmm10
-; SSE-NEXT:    pandn %xmm0, %xmm7
-; SSE-NEXT:    por %xmm7, %xmm10
+; SSE-NEXT:    pandn %xmm7, %xmm13
+; SSE-NEXT:    por %xmm9, %xmm13
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,255,0,0,0,0,255,255,255,0,0,0,0,255,255]
+; SSE-NEXT:    pand %xmm12, %xmm13
+; SSE-NEXT:    pandn %xmm0, %xmm12
+; SSE-NEXT:    por %xmm13, %xmm12
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,1,2,2]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm7 = xmm15[1,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm7 = xmm7[0,0,2,1]
-; SSE-NEXT:    pand %xmm2, %xmm7
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    por %xmm7, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm13[1,1,2,2,4,5,6,7]
+; SSE-NEXT:    pand %xmm5, %xmm7
+; SSE-NEXT:    pandn %xmm0, %xmm5
+; SSE-NEXT:    por %xmm7, %xmm5
+; SSE-NEXT:    pshuflw $165, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = mem[1,1,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
-; SSE-NEXT:    pand %xmm11, %xmm0
+; SSE-NEXT:    pand %xmm10, %xmm0
 ; SSE-NEXT:    pshufd $101, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm7 = mem[1,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,7,5,6,4]
-; SSE-NEXT:    pandn %xmm7, %xmm11
-; SSE-NEXT:    por %xmm0, %xmm11
-; SSE-NEXT:    pand %xmm6, %xmm11
-; SSE-NEXT:    pandn %xmm2, %xmm6
-; SSE-NEXT:    por %xmm11, %xmm6
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm4[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pandn %xmm7, %xmm10
+; SSE-NEXT:    por %xmm0, %xmm10
+; SSE-NEXT:    pand %xmm6, %xmm10
+; SSE-NEXT:    pandn %xmm5, %xmm6
+; SSE-NEXT:    por %xmm10, %xmm6
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[1,1,1,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
 ; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm5[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm4[0,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,1,3]
 ; SSE-NEXT:    pandn %xmm2, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm12, %xmm1
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm8[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pand %xmm11, %xmm1
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm14[1,1,1,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    pandn %xmm0, %xmm12
-; SSE-NEXT:    por %xmm1, %xmm12
+; SSE-NEXT:    pandn %xmm0, %xmm11
+; SSE-NEXT:    por %xmm1, %xmm11
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255]
 ; SSE-NEXT:    pand %xmm0, %xmm6
-; SSE-NEXT:    pandn %xmm12, %xmm0
+; SSE-NEXT:    pandn %xmm11, %xmm0
 ; SSE-NEXT:    por %xmm6, %xmm0
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; SSE-NEXT:    movdqa %xmm0, 16(%rax)
-; SSE-NEXT:    movdqa %xmm10, 32(%rax)
+; SSE-NEXT:    movdqa %xmm12, 32(%rax)
 ; SSE-NEXT:    movdqa %xmm3, 64(%rax)
-; SSE-NEXT:    movdqa %xmm14, (%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movdqa %xmm8, (%rax)
+; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 80(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 48(%rax)
@@ -2685,72 +2679,70 @@ define void @store_i8_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 define void @store_i8_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %in.vecptr3, ptr %in.vecptr4, ptr %in.vecptr5, ptr %in.vecptr6, ptr %out.vec) nounwind {
 ; SSE-LABEL: store_i8_stride7_vf32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $360, %rsp # imm = 0x168
-; SSE-NEXT:    movdqa 16(%rdi), %xmm1
+; SSE-NEXT:    subq $344, %rsp # imm = 0x158
+; SSE-NEXT:    movdqa 16(%rdi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 16(%rsi), %xmm4
 ; SSE-NEXT:    movdqa 16(%rdx), %xmm3
-; SSE-NEXT:    movdqa 16(%rcx), %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 16(%r8), %xmm6
+; SSE-NEXT:    movdqa 16(%rcx), %xmm14
+; SSE-NEXT:    movdqa 16(%r8), %xmm7
 ; SSE-NEXT:    movdqa 16(%r9), %xmm5
 ; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,6,6,6,6]
-; SSE-NEXT:    movdqa %xmm1, %xmm15
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,6,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
-; SSE-NEXT:    pand %xmm10, %xmm0
-; SSE-NEXT:    movdqa %xmm4, %xmm8
-; SSE-NEXT:    movdqa %xmm4, %xmm13
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
+; SSE-NEXT:    pand %xmm12, %xmm0
+; SSE-NEXT:    movdqa %xmm4, %xmm9
+; SSE-NEXT:    movdqa %xmm4, %xmm15
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm8 = xmm8[8],xmm4[8],xmm8[9],xmm4[9],xmm8[10],xmm4[10],xmm8[11],xmm4[11],xmm8[12],xmm4[12],xmm8[13],xmm4[13],xmm8[14],xmm4[14],xmm8[15],xmm4[15]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm8[0,1,2,3,4,5,5,7]
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm9 = xmm9[8],xmm4[8],xmm9[9],xmm4[9],xmm9[10],xmm4[10],xmm9[11],xmm4[11],xmm9[12],xmm4[12],xmm9[13],xmm4[13],xmm9[14],xmm4[14],xmm9[15],xmm4[15]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm9[0,1,2,3,4,5,5,7]
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    movdqa %xmm12, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
 ; SSE-NEXT:    por %xmm0, %xmm2
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255]
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm1, %xmm11
+; SSE-NEXT:    movdqa %xmm1, %xmm6
 ; SSE-NEXT:    pandn %xmm2, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm3[0,1,2,3,6,6,6,6]
 ; SSE-NEXT:    movdqa %xmm3, %xmm4
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
-; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255]
-; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255]
+; SSE-NEXT:    movdqa %xmm13, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm7[8],xmm2[9],xmm7[9],xmm2[10],xmm7[10],xmm2[11],xmm7[11],xmm2[12],xmm7[12],xmm2[13],xmm7[13],xmm2[14],xmm7[14],xmm2[15],xmm7[15]
+; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8],xmm14[8],xmm2[9],xmm14[9],xmm2[10],xmm14[10],xmm2[11],xmm14[11],xmm2[12],xmm14[12],xmm2[13],xmm14[13],xmm2[14],xmm14[14],xmm2[15],xmm14[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[2,1,2,3]
-; SSE-NEXT:    movdqa %xmm2, %xmm7
+; SSE-NEXT:    movdqa %xmm2, %xmm8
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pand %xmm9, %xmm1
+; SSE-NEXT:    pand %xmm13, %xmm1
 ; SSE-NEXT:    por %xmm3, %xmm1
-; SSE-NEXT:    pand %xmm11, %xmm1
+; SSE-NEXT:    pand %xmm6, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0]
 ; SSE-NEXT:    pand %xmm11, %xmm1
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm6[0,1,2,3,5,6,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm7[0,1,2,3,5,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,3,2]
 ; SSE-NEXT:    movdqa %xmm11, %xmm3
 ; SSE-NEXT:    pandn %xmm0, %xmm3
 ; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
 ; SSE-NEXT:    movdqa %xmm5, %xmm1
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm5[8],xmm1[9],xmm5[9],xmm1[10],xmm5[10],xmm1[11],xmm5[11],xmm1[12],xmm5[12],xmm1[13],xmm5[13],xmm1[14],xmm5[14],xmm1[15],xmm5[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
 ; SSE-NEXT:    movdqa %xmm1, %xmm5
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    pand %xmm12, %xmm3
+; SSE-NEXT:    pand %xmm2, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm1
-; SSE-NEXT:    movdqa 16(%rax), %xmm14
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm14[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    movdqa 16(%rax), %xmm10
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm10[0,1,2,3,4,5,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,3,3]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
 ; SSE-NEXT:    movdqa %xmm3, %xmm2
@@ -2758,34 +2750,36 @@ define void @store_i8_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    pand %xmm3, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm7[0,1,2,3,6,5,7,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm8[0,1,2,3,6,5,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    movdqa %xmm12, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    movdqa %xmm4, %xmm6
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    pand %xmm12, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255]
 ; SSE-NEXT:    movdqa %xmm2, %xmm0
 ; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm8[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm9[0,1,2,3,4,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,3,3]
 ; SSE-NEXT:    movdqa %xmm3, %xmm4
 ; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm15[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm8[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
 ; SSE-NEXT:    pand %xmm4, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm3
 ; SSE-NEXT:    pand %xmm2, %xmm3
 ; SSE-NEXT:    por %xmm0, %xmm3
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm6[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm7[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    movdqa %xmm13, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm9, %xmm3
+; SSE-NEXT:    pand %xmm13, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm1
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm5[0,1,2,3,5,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,2,2,3]
@@ -2794,114 +2788,116 @@ define void @store_i8_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    pandn %xmm3, %xmm4
 ; SSE-NEXT:    pand %xmm0, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm14[0,1,2,3,6,7,7,7]
+; SSE-NEXT:    movdqa %xmm10, %xmm5
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm10[0,1,2,3,6,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,3,2]
 ; SSE-NEXT:    movdqa %xmm11, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
 ; SSE-NEXT:    pand %xmm11, %xmm4
 ; SSE-NEXT:    por %xmm4, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm13[2,1,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,1,2,3]
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,0,3]
-; SSE-NEXT:    movdqa %xmm12, %xmm4
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm10, %xmm4
 ; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm15[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm8[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm12, %xmm1
+; SSE-NEXT:    pand %xmm10, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm4
-; SSE-NEXT:    pshufd $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[2,1,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm14[2,1,2,3]
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,3,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm15, %xmm7
-; SSE-NEXT:    pandn %xmm1, %xmm7
-; SSE-NEXT:    pshuflw $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm15, %xmm8
+; SSE-NEXT:    pandn %xmm1, %xmm8
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm6[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,4,4,4]
 ; SSE-NEXT:    pand %xmm15, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm7
+; SSE-NEXT:    por %xmm1, %xmm8
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255]
 ; SSE-NEXT:    movdqa %xmm1, %xmm3
-; SSE-NEXT:    pandn %xmm7, %xmm3
+; SSE-NEXT:    pandn %xmm8, %xmm3
 ; SSE-NEXT:    pand %xmm1, %xmm4
 ; SSE-NEXT:    por %xmm4, %xmm3
 ; SSE-NEXT:    pshufd $230, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm4 = mem[2,1,2,3]
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,3,2,0]
-; SSE-NEXT:    movdqa %xmm9, %xmm7
-; SSE-NEXT:    pandn %xmm4, %xmm7
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm6[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm13, %xmm8
+; SSE-NEXT:    pandn %xmm4, %xmm8
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm7[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm9, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm7
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm14[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm13, %xmm4
+; SSE-NEXT:    por %xmm4, %xmm8
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm5[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm0, %xmm8
-; SSE-NEXT:    pandn %xmm4, %xmm8
-; SSE-NEXT:    pand %xmm0, %xmm7
-; SSE-NEXT:    por %xmm7, %xmm8
+; SSE-NEXT:    movdqa %xmm0, %xmm9
+; SSE-NEXT:    pandn %xmm4, %xmm9
+; SSE-NEXT:    pand %xmm0, %xmm8
+; SSE-NEXT:    por %xmm8, %xmm9
 ; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255]
 ; SSE-NEXT:    movdqa %xmm5, %xmm2
-; SSE-NEXT:    pandn %xmm8, %xmm2
+; SSE-NEXT:    pandn %xmm9, %xmm2
 ; SSE-NEXT:    pand %xmm5, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa (%rsi), %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[2,1,2,3]
-; SSE-NEXT:    movdqa %xmm2, %xmm6
+; SSE-NEXT:    movdqa %xmm2, %xmm7
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[3,1,0,3]
-; SSE-NEXT:    movdqa %xmm12, %xmm3
+; SSE-NEXT:    movdqa %xmm10, %xmm3
 ; SSE-NEXT:    pandn %xmm4, %xmm3
-; SSE-NEXT:    movdqa (%rdi), %xmm13
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm13[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%rdi), %xmm2
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm2[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm2, %xmm6
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm12, %xmm4
+; SSE-NEXT:    pand %xmm10, %xmm4
 ; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    movdqa (%rcx), %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm2[2,1,2,3]
-; SSE-NEXT:    movdqa %xmm2, %xmm14
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%rcx), %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm11[2,1,2,3]
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,3,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm15, %xmm7
-; SSE-NEXT:    pandn %xmm4, %xmm7
-; SSE-NEXT:    movdqa (%rdx), %xmm8
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm8[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm8, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm15, %xmm8
+; SSE-NEXT:    pandn %xmm4, %xmm8
+; SSE-NEXT:    movdqa (%rdx), %xmm2
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm2[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm2, %xmm12
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,4,4]
 ; SSE-NEXT:    pand %xmm15, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm7
+; SSE-NEXT:    por %xmm4, %xmm8
 ; SSE-NEXT:    pand %xmm1, %xmm3
-; SSE-NEXT:    pandn %xmm7, %xmm1
+; SSE-NEXT:    pandn %xmm8, %xmm1
 ; SSE-NEXT:    por %xmm3, %xmm1
-; SSE-NEXT:    movdqa (%r9), %xmm11
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[2,1,2,3]
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%r9), %xmm10
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm10[2,1,2,3]
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,2,0]
-; SSE-NEXT:    movdqa %xmm9, %xmm4
+; SSE-NEXT:    movdqa %xmm13, %xmm4
 ; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    movdqa (%r8), %xmm7
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm7[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%r8), %xmm2
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm2[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm2, %xmm8
+; SSE-NEXT:    movdqa %xmm2, (%rsp) # 16-byte Spill
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm9, %xmm3
+; SSE-NEXT:    pand %xmm13, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm4
 ; SSE-NEXT:    pand %xmm0, %xmm4
-; SSE-NEXT:    movdqa (%rax), %xmm10
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm10[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa (%rax), %xmm9
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm9[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4]
 ; SSE-NEXT:    pandn %xmm3, %xmm0
 ; SSE-NEXT:    por %xmm4, %xmm0
@@ -2909,45 +2905,45 @@ define void @store_i8_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    pandn %xmm0, %xmm5
 ; SSE-NEXT:    por %xmm1, %xmm5
 ; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm6[8],xmm0[9],xmm6[9],xmm0[10],xmm6[10],xmm0[11],xmm6[11],xmm0[12],xmm6[12],xmm0[13],xmm6[13],xmm0[14],xmm6[14],xmm0[15],xmm6[15]
+; SSE-NEXT:    movdqa %xmm7, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm7[8],xmm0[9],xmm7[9],xmm0[10],xmm7[10],xmm0[11],xmm7[11],xmm0[12],xmm7[12],xmm0[13],xmm7[13],xmm0[14],xmm7[14],xmm0[15],xmm7[15]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
 ; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm13[0,1,2,3,6,6,6,6]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm6[0,1,2,3,6,6,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
 ; SSE-NEXT:    pand %xmm2, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255]
 ; SSE-NEXT:    movdqa %xmm2, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm8[0,1,2,3,6,6,6,6]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm12[0,1,2,3,6,6,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    movdqa %xmm13, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm14[8],xmm0[9],xmm14[9],xmm0[10],xmm14[10],xmm0[11],xmm14[11],xmm0[12],xmm14[12],xmm0[13],xmm14[13],xmm0[14],xmm14[14],xmm0[15],xmm14[15]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,1,2,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    pand %xmm13, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    pand %xmm2, %xmm0
 ; SSE-NEXT:    por %xmm3, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm11 = xmm11[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[1,1,2,3]
-; SSE-NEXT:    movdqa %xmm12, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm10 = xmm10[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[1,1,2,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm2, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm7[0,1,2,3,5,6,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm8[0,1,2,3,5,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,3,2]
-; SSE-NEXT:    pand %xmm12, %xmm1
+; SSE-NEXT:    pand %xmm2, %xmm1
 ; SSE-NEXT:    por %xmm3, %xmm1
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm10[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm9[0,1,2,3,4,5,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,3,3]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
 ; SSE-NEXT:    movdqa %xmm2, %xmm4
@@ -2959,19 +2955,19 @@ define void @store_i8_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    pandn %xmm4, %xmm1
 ; SSE-NEXT:    por %xmm0, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm11 = xmm11[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm14, %xmm10
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm13, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm7[2,1,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
-; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    pand %xmm13, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255]
-; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255]
+; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
@@ -2985,22 +2981,24 @@ define void @store_i8_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,1,3]
 ; SSE-NEXT:    pand %xmm15, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm3
-; SSE-NEXT:    pand %xmm13, %xmm3
+; SSE-NEXT:    pand %xmm2, %xmm3
 ; SSE-NEXT:    por %xmm1, %xmm3
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,0,0,0,0,255,255,255,0,0,0,0,255,255]
+; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [255,255,255,0,0,0,0,255,255,255,0,0,0,0,255,255]
+; SSE-NEXT:    movdqa %xmm9, %xmm0
 ; SSE-NEXT:    pandn %xmm3, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm8[0,1,2,3,4,6,5,7]
 ; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,3,2]
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0]
-; SSE-NEXT:    movdqa %xmm12, %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0]
+; SSE-NEXT:    movdqa %xmm5, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm10[2,2,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm12[2,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,2,1]
-; SSE-NEXT:    pand %xmm12, %xmm1
+; SSE-NEXT:    pand %xmm5, %xmm1
+; SSE-NEXT:    movdqa %xmm5, %xmm11
 ; SSE-NEXT:    por %xmm1, %xmm3
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm5[2,2,2,2,4,5,6,7]
@@ -3010,37 +3008,38 @@ define void @store_i8_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    pandn %xmm1, %xmm14
 ; SSE-NEXT:    pand %xmm2, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm14
-; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm14
+; SSE-NEXT:    pand %xmm9, %xmm14
 ; SSE-NEXT:    por %xmm0, %xmm14
 ; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[0,1,2,2]
-; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    movdqa %xmm13, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm4[1,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
-; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    pand %xmm13, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm13, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255]
+; SSE-NEXT:    movdqa %xmm9, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[1,1,2,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[1,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm0[0,1,2,3,7,5,6,4]
-; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    movdqa %xmm11, %xmm0
 ; SSE-NEXT:    pandn %xmm3, %xmm0
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm7[1,1,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,0,2,1]
-; SSE-NEXT:    pand %xmm12, %xmm3
-; SSE-NEXT:    movdqa %xmm12, %xmm11
+; SSE-NEXT:    pand %xmm11, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm0
-; SSE-NEXT:    pand %xmm13, %xmm0
+; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm9, %xmm6
 ; SSE-NEXT:    por %xmm1, %xmm0
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm8[0,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,1,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm13, %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm2, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm10[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm12[1,1,1,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
-; SSE-NEXT:    pand %xmm13, %xmm1
+; SSE-NEXT:    pand %xmm2, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm3
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm5[1,1,1,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
@@ -3056,208 +3055,208 @@ define void @store_i8_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[0,1,1,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm12, %xmm1
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    movdqa (%rsp), %xmm8 # 16-byte Reload
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm8[0,1,2,3,4,5,5,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm12[0,1,2,3,4,5,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,3,2]
-; SSE-NEXT:    pand %xmm12, %xmm0
+; SSE-NEXT:    pand %xmm2, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0]
-; SSE-NEXT:    movdqa %xmm6, %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0]
+; SSE-NEXT:    movdqa %xmm1, %xmm3
 ; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm5[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm14[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm9, %xmm4
+; SSE-NEXT:    movdqa %xmm13, %xmm4
 ; SSE-NEXT:    pandn %xmm0, %xmm4
 ; SSE-NEXT:    pshuflw $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[1,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
-; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    pand %xmm13, %xmm0
 ; SSE-NEXT:    por %xmm4, %xmm0
-; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    pand %xmm1, %xmm0
 ; SSE-NEXT:    por %xmm3, %xmm0
 ; SSE-NEXT:    pshuflw $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm3 = mem[1,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,0,0,0]
 ; SSE-NEXT:    movdqa %xmm15, %xmm4
 ; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm13[0,1,2,3,4,4,6,5]
+; SSE-NEXT:    movdqa (%rsp), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm8[0,1,2,3,4,4,6,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,3,3]
 ; SSE-NEXT:    pand %xmm15, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm4
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm12[0,1,2,3,4,5,5,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm10[0,1,2,3,4,5,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
-; SSE-NEXT:    movdqa %xmm1, %xmm10
-; SSE-NEXT:    pandn %xmm3, %xmm10
-; SSE-NEXT:    pand %xmm1, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm10
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255]
-; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    pandn %xmm10, %xmm1
-; SSE-NEXT:    pand %xmm2, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm14 = xmm14[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[0,1,2,2]
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm2, %xmm9
+; SSE-NEXT:    pandn %xmm3, %xmm9
+; SSE-NEXT:    pand %xmm2, %xmm4
+; SSE-NEXT:    por %xmm4, %xmm9
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm3, %xmm2
+; SSE-NEXT:    pandn %xmm9, %xmm2
+; SSE-NEXT:    pand %xmm3, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,1,2,2]
+; SSE-NEXT:    movdqa %xmm2, %xmm7
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm13, %xmm3
 ; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm5[1,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm14[1,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
-; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    pand %xmm13, %xmm0
 ; SSE-NEXT:    por %xmm3, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255]
-; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    movdqa %xmm6, %xmm9
+; SSE-NEXT:    movdqa %xmm6, %xmm3
 ; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm7 = xmm7[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,2,1]
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,2,1]
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,7,5,6,4]
 ; SSE-NEXT:    movdqa %xmm11, %xmm0
 ; SSE-NEXT:    pandn %xmm4, %xmm0
-; SSE-NEXT:    movdqa %xmm8, %xmm1
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm8[1,1,2,2,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm12, %xmm5
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm12[1,1,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,0,2,1]
 ; SSE-NEXT:    pand %xmm11, %xmm4
 ; SSE-NEXT:    por %xmm4, %xmm0
-; SSE-NEXT:    pand %xmm5, %xmm0
+; SSE-NEXT:    pand %xmm9, %xmm0
 ; SSE-NEXT:    por %xmm3, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm8[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm2[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm2, %xmm12
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,1,1,3]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
 ; SSE-NEXT:    movdqa %xmm2, %xmm4
 ; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    movdqa %xmm13, %xmm5
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm13[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm8[1,1,1,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,0,0,0]
 ; SSE-NEXT:    pand %xmm2, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm4
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm12[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm10, %xmm2
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm10[1,1,1,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm15, %xmm10
-; SSE-NEXT:    pandn %xmm3, %xmm10
+; SSE-NEXT:    movdqa %xmm15, %xmm9
+; SSE-NEXT:    pandn %xmm3, %xmm9
 ; SSE-NEXT:    pand %xmm15, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm10
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255]
-; SSE-NEXT:    pand %xmm3, %xmm0
-; SSE-NEXT:    pandn %xmm10, %xmm3
-; SSE-NEXT:    por %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm7[0,0,2,1,4,5,6,7]
+; SSE-NEXT:    por %xmm4, %xmm9
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255]
+; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    pandn %xmm9, %xmm10
+; SSE-NEXT:    por %xmm0, %xmm10
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm6[0,0,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,1,3]
-; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm4, %xmm3
 ; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm5[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    pand %xmm4, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm3
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255]
 ; SSE-NEXT:    movdqa %xmm0, %xmm4
 ; SSE-NEXT:    pandn %xmm3, %xmm4
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm14[0,2,1,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm3[0,1,1,0]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm7[0,2,1,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm3[0,1,1,0]
 ; SSE-NEXT:    movdqa %xmm11, %xmm3
-; SSE-NEXT:    pandn %xmm10, %xmm3
-; SSE-NEXT:    pshuflw $96, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm10 = mem[0,0,2,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,0,2,1]
-; SSE-NEXT:    pand %xmm11, %xmm10
-; SSE-NEXT:    por %xmm10, %xmm3
+; SSE-NEXT:    pandn %xmm9, %xmm3
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm14[0,0,2,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,0,2,1]
+; SSE-NEXT:    pand %xmm11, %xmm9
+; SSE-NEXT:    por %xmm9, %xmm3
 ; SSE-NEXT:    pand %xmm0, %xmm3
 ; SSE-NEXT:    por %xmm4, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm8[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm12[0,1,1,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,0,2,1]
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
-; SSE-NEXT:    movdqa %xmm13, %xmm10
-; SSE-NEXT:    pandn %xmm4, %xmm10
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm5[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm6, %xmm9
+; SSE-NEXT:    pandn %xmm4, %xmm9
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm8[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,0,0,0]
-; SSE-NEXT:    pand %xmm13, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm10
-; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm12[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pand %xmm6, %xmm4
+; SSE-NEXT:    por %xmm4, %xmm9
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm2[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm9, %xmm2
+; SSE-NEXT:    movdqa %xmm13, %xmm2
 ; SSE-NEXT:    pandn %xmm4, %xmm2
-; SSE-NEXT:    pand %xmm9, %xmm10
-; SSE-NEXT:    por %xmm10, %xmm2
-; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255]
-; SSE-NEXT:    movdqa %xmm7, %xmm1
-; SSE-NEXT:    pandn %xmm2, %xmm1
-; SSE-NEXT:    pand %xmm7, %xmm3
-; SSE-NEXT:    por %xmm3, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm13, %xmm9
+; SSE-NEXT:    por %xmm9, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255]
+; SSE-NEXT:    movdqa %xmm4, %xmm12
+; SSE-NEXT:    pandn %xmm2, %xmm12
+; SSE-NEXT:    pand %xmm4, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm12
 ; SSE-NEXT:    pshufd $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = mem[0,1,1,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm5, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm4[0,1,2,3,4,5,5,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm7[0,1,2,3,4,5,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,3,2]
-; SSE-NEXT:    pand %xmm1, %xmm2
+; SSE-NEXT:    pand %xmm5, %xmm2
 ; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm12[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm10[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm9, %xmm5
+; SSE-NEXT:    movdqa %xmm13, %xmm5
 ; SSE-NEXT:    pandn %xmm3, %xmm5
 ; SSE-NEXT:    pshuflw $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm3 = mem[1,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,0,2,1]
-; SSE-NEXT:    pand %xmm9, %xmm3
+; SSE-NEXT:    pand %xmm13, %xmm3
 ; SSE-NEXT:    por %xmm5, %xmm3
-; SSE-NEXT:    pand %xmm6, %xmm3
-; SSE-NEXT:    pandn %xmm2, %xmm6
-; SSE-NEXT:    por %xmm3, %xmm6
+; SSE-NEXT:    pand %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm1
+; SSE-NEXT:    por %xmm3, %xmm1
 ; SSE-NEXT:    pshuflw $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm2 = mem[1,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
 ; SSE-NEXT:    movdqa %xmm15, %xmm3
 ; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm8[0,1,2,3,4,4,6,5]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm9[0,1,2,3,4,4,6,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,3,3]
 ; SSE-NEXT:    pand %xmm15, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm10[0,1,2,3,4,5,5,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm8[0,1,2,3,4,5,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
-; SSE-NEXT:    movdqa %xmm13, %xmm5
+; SSE-NEXT:    movdqa %xmm6, %xmm5
 ; SSE-NEXT:    pandn %xmm2, %xmm5
-; SSE-NEXT:    pand %xmm13, %xmm3
+; SSE-NEXT:    pand %xmm6, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm5
 ; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255]
-; SSE-NEXT:    pand %xmm14, %xmm6
+; SSE-NEXT:    pand %xmm14, %xmm1
 ; SSE-NEXT:    pandn %xmm5, %xmm14
-; SSE-NEXT:    por %xmm6, %xmm14
+; SSE-NEXT:    por %xmm1, %xmm14
 ; SSE-NEXT:    pshuflw $96, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[0,0,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,1,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm3, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm5, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm4[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm7[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
-; SSE-NEXT:    pand %xmm3, %xmm1
-; SSE-NEXT:    movdqa %xmm3, %xmm4
+; SSE-NEXT:    pand %xmm5, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm2
 ; SSE-NEXT:    pshuflw $216, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[0,2,1,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,1,0]
 ; SSE-NEXT:    movdqa %xmm11, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm12[0,0,2,1,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm10[0,0,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,2,1]
 ; SSE-NEXT:    pand %xmm11, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm3
@@ -3267,35 +3266,35 @@ define void @store_i8_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    pshuflw $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[0,1,1,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,2,1]
-; SSE-NEXT:    movdqa %xmm13, %xmm2
+; SSE-NEXT:    movdqa %xmm6, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm8[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm9[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
-; SSE-NEXT:    pand %xmm13, %xmm1
+; SSE-NEXT:    pand %xmm6, %xmm1
 ; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm10[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm8[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    movdqa %xmm13, %xmm3
 ; SSE-NEXT:    pandn %xmm1, %xmm3
-; SSE-NEXT:    pand %xmm9, %xmm2
+; SSE-NEXT:    pand %xmm13, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm3
-; SSE-NEXT:    pand %xmm7, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm7
-; SSE-NEXT:    por %xmm0, %xmm7
-; SSE-NEXT:    movdqa (%rsp), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm5[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pand %xmm4, %xmm0
+; SSE-NEXT:    pandn %xmm3, %xmm4
+; SSE-NEXT:    por %xmm0, %xmm4
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm8[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    pand %xmm13, %xmm0
+; SSE-NEXT:    pand %xmm6, %xmm0
 ; SSE-NEXT:    pshufhw $246, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[0,1,2,3,6,5,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE-NEXT:    pandn %xmm1, %xmm13
-; SSE-NEXT:    por %xmm0, %xmm13
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm6[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pandn %xmm1, %xmm6
+; SSE-NEXT:    por %xmm0, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm7[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm4, %xmm3
-; SSE-NEXT:    pand %xmm4, %xmm0
+; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    pand %xmm5, %xmm0
 ; SSE-NEXT:    pshufhw $232, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[0,1,2,3,4,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,3,3]
@@ -3303,21 +3302,21 @@ define void @store_i8_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    por %xmm0, %xmm3
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255]
 ; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    pandn %xmm13, %xmm0
+; SSE-NEXT:    pandn %xmm6, %xmm0
 ; SSE-NEXT:    por %xmm3, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    movdqa (%rsp), %xmm5 # 16-byte Reload
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm5[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm9, %xmm1
+; SSE-NEXT:    movdqa %xmm13, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
 ; SSE-NEXT:    pshufhw $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm0 = mem[0,1,2,3,5,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,3]
-; SSE-NEXT:    pand %xmm9, %xmm0
+; SSE-NEXT:    pand %xmm13, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm13[0,1,2,3,6,7,7,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm6[0,1,2,3,6,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,3,2]
 ; SSE-NEXT:    movdqa %xmm11, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
@@ -3329,12 +3328,12 @@ define void @store_i8_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    por %xmm3, %xmm0
 ; SSE-NEXT:    pshufd $250, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[2,2,3,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm5[2,1,3,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm8[2,1,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,2,1]
-; SSE-NEXT:    pand %xmm9, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm9
-; SSE-NEXT:    por %xmm2, %xmm9
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm6[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm13, %xmm2
+; SSE-NEXT:    pandn %xmm1, %xmm13
+; SSE-NEXT:    por %xmm2, %xmm13
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm7[0,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,1,3]
 ; SSE-NEXT:    pand %xmm15, %xmm1
 ; SSE-NEXT:    pshufhw $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
@@ -3344,10 +3343,10 @@ define void @store_i8_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    por %xmm1, %xmm15
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255]
 ; SSE-NEXT:    pand %xmm1, %xmm15
-; SSE-NEXT:    pandn %xmm9, %xmm1
+; SSE-NEXT:    pandn %xmm13, %xmm1
 ; SSE-NEXT:    por %xmm15, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm4[2,2,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm5[2,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,2,1]
 ; SSE-NEXT:    pand %xmm11, %xmm1
 ; SSE-NEXT:    pshufhw $216, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
@@ -3355,23 +3354,22 @@ define void @store_i8_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,3,2]
 ; SSE-NEXT:    pandn %xmm2, %xmm11
 ; SSE-NEXT:    por %xmm1, %xmm11
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm4, %xmm11
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm13[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
+; SSE-NEXT:    pand %xmm2, %xmm11
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm6[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
-; SSE-NEXT:    pandn %xmm1, %xmm4
-; SSE-NEXT:    por %xmm11, %xmm4
+; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    por %xmm11, %xmm2
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,0,0,0,0,255,255,255,0,0,0,0,255,255]
-; SSE-NEXT:    pand %xmm1, %xmm4
+; SSE-NEXT:    pand %xmm1, %xmm2
 ; SSE-NEXT:    pandn %xmm3, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm4
+; SSE-NEXT:    por %xmm2, %xmm1
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movdqa %xmm4, 32(%rax)
+; SSE-NEXT:    movdqa %xmm1, 32(%rax)
 ; SSE-NEXT:    movdqa %xmm0, 96(%rax)
-; SSE-NEXT:    movdqa %xmm7, 112(%rax)
+; SSE-NEXT:    movdqa %xmm4, 112(%rax)
 ; SSE-NEXT:    movdqa %xmm14, 176(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, (%rax)
+; SSE-NEXT:    movdqa %xmm12, (%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 16(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -3390,7 +3388,7 @@ define void @store_i8_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    movaps %xmm0, 208(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 192(%rax)
-; SSE-NEXT:    addq $360, %rsp # imm = 0x168
+; SSE-NEXT:    addq $344, %rsp # imm = 0x158
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: store_i8_stride7_vf32:
@@ -5415,193 +5413,192 @@ define void @store_i8_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    subq $648, %rsp # imm = 0x288
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movdqa 48(%rdi), %xmm14
+; SSE-NEXT:    movdqa 48(%rdi), %xmm9
 ; SSE-NEXT:    movdqa 48(%rsi), %xmm2
 ; SSE-NEXT:    movdqa 48(%rdx), %xmm3
-; SSE-NEXT:    movdqa 48(%rcx), %xmm10
+; SSE-NEXT:    movdqa 48(%rcx), %xmm4
 ; SSE-NEXT:    movdqa 48(%r8), %xmm5
 ; SSE-NEXT:    movdqa 48(%r9), %xmm8
 ; SSE-NEXT:    movdqa 48(%rax), %xmm13
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm14[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm9[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
 ; SSE-NEXT:    pand %xmm6, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[2,1,2,3]
-; SSE-NEXT:    movdqa %xmm2, %xmm11
+; SSE-NEXT:    movdqa %xmm2, %xmm12
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,1,0,3]
 ; SSE-NEXT:    movdqa %xmm6, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
 ; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255]
-; SSE-NEXT:    pand %xmm12, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255]
+; SSE-NEXT:    pand %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm0, %xmm7
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm3[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    movdqa %xmm3, %xmm6
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm10[2,1,2,3]
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255]
+; SSE-NEXT:    pand %xmm10, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm4[2,1,2,3]
+; SSE-NEXT:    movdqa %xmm4, %xmm1
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm1, %xmm4
+; SSE-NEXT:    movdqa %xmm10, %xmm4
 ; SSE-NEXT:    pandn %xmm3, %xmm4
 ; SSE-NEXT:    por %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    movdqa %xmm7, %xmm0
 ; SSE-NEXT:    pandn %xmm4, %xmm0
 ; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
-; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
+; SSE-NEXT:    pand %xmm4, %xmm0
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm5[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm5, %xmm9
+; SSE-NEXT:    movdqa %xmm5, %xmm10
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,2,2]
-; SSE-NEXT:    movdqa %xmm1, %xmm3
-; SSE-NEXT:    movdqa %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    movdqa %xmm4, %xmm5
 ; SSE-NEXT:    pandn %xmm2, %xmm3
 ; SSE-NEXT:    por %xmm0, %xmm3
 ; SSE-NEXT:    movdqa {{.*#+}} xmm7 = [255,255,255,255,255,255,0,255,255,255,255,255,255,0,255,255]
 ; SSE-NEXT:    pand %xmm7, %xmm3
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[2,1,2,3]
-; SSE-NEXT:    movdqa %xmm8, %xmm1
+; SSE-NEXT:    movdqa %xmm8, %xmm15
 ; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,0]
 ; SSE-NEXT:    movdqa %xmm7, %xmm4
 ; SSE-NEXT:    pandn %xmm0, %xmm4
 ; SSE-NEXT:    por %xmm3, %xmm4
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255]
-; SSE-NEXT:    pand %xmm2, %xmm4
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [0,255,255,255,255,255,255,0,255,255,255,255,255,255,0,255]
+; SSE-NEXT:    pand %xmm11, %xmm4
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm13[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    movdqa %xmm11, %xmm3
 ; SSE-NEXT:    pandn %xmm0, %xmm3
 ; SSE-NEXT:    por %xmm4, %xmm3
 ; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm11[8],xmm0[9],xmm11[9],xmm0[10],xmm11[10],xmm0[11],xmm11[11],xmm0[12],xmm11[12],xmm0[13],xmm11[13],xmm0[14],xmm11[14],xmm0[15],xmm11[15]
+; SSE-NEXT:    movdqa %xmm12, %xmm0
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm12[8],xmm0[9],xmm12[9],xmm0[10],xmm12[10],xmm0[11],xmm12[11],xmm0[12],xmm12[12],xmm0[13],xmm12[13],xmm0[14],xmm12[14],xmm0[15],xmm12[15]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm0[0,1,2,3,4,5,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
 ; SSE-NEXT:    movdqa %xmm5, %xmm4
-; SSE-NEXT:    pandn %xmm3, %xmm5
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm14[0,1,2,3,6,6,6,6]
+; SSE-NEXT:    pandn %xmm3, %xmm4
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm9[0,1,2,3,6,6,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,2,2,2]
-; SSE-NEXT:    pand %xmm4, %xmm3
-; SSE-NEXT:    por %xmm3, %xmm5
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255]
-; SSE-NEXT:    movdqa %xmm4, %xmm3
-; SSE-NEXT:    pandn %xmm5, %xmm3
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm6[0,1,2,3,6,6,6,6]
-; SSE-NEXT:    movdqa %xmm6, %xmm15
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,2,2,2]
+; SSE-NEXT:    pand %xmm5, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm4
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    movdqa %xmm5, %xmm2
+; SSE-NEXT:    pandn %xmm4, %xmm3
+; SSE-NEXT:    pshufhw {{.*#+}} xmm4 = xmm6[0,1,2,3,6,6,6,6]
+; SSE-NEXT:    movdqa %xmm6, %xmm14
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,2,2,2]
 ; SSE-NEXT:    movdqa %xmm7, %xmm6
-; SSE-NEXT:    pandn %xmm5, %xmm6
-; SSE-NEXT:    movdqa %xmm10, %xmm5
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm10[8],xmm5[9],xmm10[9],xmm5[10],xmm10[10],xmm5[11],xmm10[11],xmm5[12],xmm10[12],xmm5[13],xmm10[13],xmm5[14],xmm10[14],xmm5[15],xmm10[15]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm5[2,1,2,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[3,1,2,0,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    pandn %xmm4, %xmm6
+; SSE-NEXT:    movdqa %xmm1, %xmm5
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm1[8],xmm5[9],xmm1[9],xmm5[10],xmm1[10],xmm5[11],xmm1[11],xmm5[12],xmm1[12],xmm5[13],xmm1[13],xmm5[14],xmm1[14],xmm5[15],xmm1[15]
+; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[2,1,2,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm4 = xmm4[3,1,2,0,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm4[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    pand %xmm7, %xmm8
 ; SSE-NEXT:    por %xmm6, %xmm8
-; SSE-NEXT:    pand %xmm4, %xmm8
+; SSE-NEXT:    pand %xmm2, %xmm8
 ; SSE-NEXT:    por %xmm3, %xmm8
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm9[0,1,2,3,5,6,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm10[0,1,2,3,5,6,6,7]
+; SSE-NEXT:    movdqa %xmm10, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,3,2]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0]
 ; SSE-NEXT:    movdqa %xmm4, %xmm6
 ; SSE-NEXT:    pandn %xmm3, %xmm6
 ; SSE-NEXT:    pand %xmm4, %xmm8
 ; SSE-NEXT:    por %xmm8, %xmm6
-; SSE-NEXT:    movdqa %xmm1, %xmm3
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm1[8],xmm3[9],xmm1[9],xmm3[10],xmm1[10],xmm3[11],xmm1[11],xmm3[12],xmm1[12],xmm3[13],xmm1[13],xmm3[14],xmm1[14],xmm3[15],xmm1[15]
+; SSE-NEXT:    movdqa %xmm15, %xmm3
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm15[8],xmm3[9],xmm15[9],xmm3[10],xmm15[10],xmm3[11],xmm15[11],xmm3[12],xmm15[12],xmm3[13],xmm15[13],xmm3[14],xmm15[14],xmm3[15],xmm15[15]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm3[1,1,2,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm4, %xmm11
-; SSE-NEXT:    pandn %xmm8, %xmm11
-; SSE-NEXT:    pand %xmm4, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm11
-; SSE-NEXT:    movdqa %xmm13, %xmm10
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm1, %xmm12
+; SSE-NEXT:    pandn %xmm8, %xmm12
+; SSE-NEXT:    pand %xmm1, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm12
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm13[0,1,2,3,4,5,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,1,3,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm1, %xmm8
+; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm15, %xmm8
 ; SSE-NEXT:    pandn %xmm6, %xmm8
-; SSE-NEXT:    pand %xmm1, %xmm11
-; SSE-NEXT:    por %xmm11, %xmm8
+; SSE-NEXT:    pand %xmm15, %xmm12
+; SSE-NEXT:    por %xmm12, %xmm8
 ; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[0,1,1,3]
-; SSE-NEXT:    movdqa %xmm4, %xmm8
+; SSE-NEXT:    movdqa %xmm1, %xmm8
 ; SSE-NEXT:    pandn %xmm6, %xmm8
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm15[0,1,2,3,4,5,5,7]
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm14[0,1,2,3,4,5,5,7]
+; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,1,3,2]
-; SSE-NEXT:    pand %xmm4, %xmm6
+; SSE-NEXT:    pand %xmm1, %xmm6
 ; SSE-NEXT:    por %xmm8, %xmm6
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0]
 ; SSE-NEXT:    movdqa %xmm1, %xmm8
 ; SSE-NEXT:    pandn %xmm6, %xmm8
-; SSE-NEXT:    movdqa %xmm14, %xmm13
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm14[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm9[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm7, %xmm11
-; SSE-NEXT:    pandn %xmm6, %xmm11
+; SSE-NEXT:    movdqa %xmm7, %xmm12
+; SSE-NEXT:    pandn %xmm6, %xmm12
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm0[1,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,0,2,1]
 ; SSE-NEXT:    pand %xmm7, %xmm6
-; SSE-NEXT:    por %xmm11, %xmm6
+; SSE-NEXT:    por %xmm12, %xmm6
 ; SSE-NEXT:    pand %xmm1, %xmm6
 ; SSE-NEXT:    por %xmm8, %xmm6
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm9[0,1,2,3,4,4,6,5]
-; SSE-NEXT:    movdqa %xmm9, %xmm1
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm10[0,1,2,3,4,4,6,5]
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[2,1,3,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm4, %xmm11
-; SSE-NEXT:    pandn %xmm8, %xmm11
-; SSE-NEXT:    pand %xmm4, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm11
+; SSE-NEXT:    movdqa %xmm15, %xmm12
+; SSE-NEXT:    pandn %xmm8, %xmm12
+; SSE-NEXT:    pand %xmm15, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm12
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm3[1,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,0,0,0]
-; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm9, %xmm8
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm1, %xmm8
 ; SSE-NEXT:    pandn %xmm6, %xmm8
-; SSE-NEXT:    pand %xmm9, %xmm11
-; SSE-NEXT:    movdqa %xmm9, %xmm14
-; SSE-NEXT:    por %xmm11, %xmm8
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm10[0,1,2,3,4,5,5,7]
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm1, %xmm12
+; SSE-NEXT:    por %xmm12, %xmm8
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm13[0,1,2,3,4,5,5,7]
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
-; SSE-NEXT:    movdqa %xmm11, %xmm9
-; SSE-NEXT:    pandn %xmm6, %xmm9
-; SSE-NEXT:    pand %xmm11, %xmm8
-; SSE-NEXT:    por %xmm8, %xmm9
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm1, %xmm10
+; SSE-NEXT:    pandn %xmm6, %xmm10
+; SSE-NEXT:    pand %xmm1, %xmm8
+; SSE-NEXT:    por %xmm8, %xmm10
+; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,2,3]
-; SSE-NEXT:    movdqa %xmm11, %xmm6
+; SSE-NEXT:    movdqa %xmm1, %xmm6
 ; SSE-NEXT:    pandn %xmm5, %xmm6
-; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm15[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm5 = xmm14[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[2,2,2,2]
-; SSE-NEXT:    pand %xmm11, %xmm5
+; SSE-NEXT:    pand %xmm1, %xmm5
 ; SSE-NEXT:    por %xmm5, %xmm6
-; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255]
-; SSE-NEXT:    movdqa %xmm8, %xmm5
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm1, %xmm5
 ; SSE-NEXT:    pandn %xmm6, %xmm5
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,3,3]
-; SSE-NEXT:    movdqa %xmm4, %xmm6
+; SSE-NEXT:    movdqa %xmm15, %xmm6
 ; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm13[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm9[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    pand %xmm4, %xmm0
+; SSE-NEXT:    pand %xmm15, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm6
-; SSE-NEXT:    pand %xmm8, %xmm6
+; SSE-NEXT:    pand %xmm1, %xmm6
 ; SSE-NEXT:    por %xmm5, %xmm6
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm2[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
 ; SSE-NEXT:    movdqa %xmm7, %xmm5
 ; SSE-NEXT:    pandn %xmm0, %xmm5
@@ -5609,216 +5606,216 @@ define void @store_i8_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    por %xmm6, %xmm5
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm3[0,1,2,3,5,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,3]
-; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    movdqa %xmm11, %xmm3
 ; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    pand %xmm2, %xmm5
+; SSE-NEXT:    pand %xmm11, %xmm5
 ; SSE-NEXT:    por %xmm5, %xmm3
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm10[0,1,2,3,6,7,7,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm13[0,1,2,3,6,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,3,2]
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0]
-; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    movdqa %xmm4, %xmm1
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm5, %xmm3
+; SSE-NEXT:    pand %xmm4, %xmm3
 ; SSE-NEXT:    por %xmm3, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa (%rsi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,0,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm4, %xmm3
-; SSE-NEXT:    pandn %xmm0, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,1,2,3]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[3,1,0,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm9, %xmm6
+; SSE-NEXT:    pandn %xmm3, %xmm6
 ; SSE-NEXT:    movdqa (%rdi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm4, %xmm0
-; SSE-NEXT:    movdqa %xmm4, %xmm11
-; SSE-NEXT:    por %xmm0, %xmm3
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    pand %xmm9, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm6
 ; SSE-NEXT:    movdqa (%rcx), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm14, %xmm5
-; SSE-NEXT:    pandn %xmm0, %xmm5
-; SSE-NEXT:    movdqa (%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm14, %xmm0
-; SSE-NEXT:    movdqa %xmm14, %xmm9
-; SSE-NEXT:    por %xmm0, %xmm5
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    pandn %xmm5, %xmm0
-; SSE-NEXT:    pand %xmm12, %xmm3
-; SSE-NEXT:    por %xmm3, %xmm0
-; SSE-NEXT:    movdqa (%r9), %xmm15
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm15[2,1,2,3]
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,1,2,3]
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,2,0]
-; SSE-NEXT:    movdqa %xmm7, %xmm6
-; SSE-NEXT:    pandn %xmm3, %xmm6
-; SSE-NEXT:    movdqa (%r8), %xmm1
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm1[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm1, %xmm8
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm7, %xmm3
-; SSE-NEXT:    por %xmm3, %xmm6
-; SSE-NEXT:    movdqa (%rax), %xmm4
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm4[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm10, %xmm8
+; SSE-NEXT:    pandn %xmm3, %xmm8
+; SSE-NEXT:    movdqa (%rdx), %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm0, %xmm5
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm2, %xmm14
-; SSE-NEXT:    pandn %xmm3, %xmm14
-; SSE-NEXT:    pand %xmm2, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm14
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255]
-; SSE-NEXT:    movdqa %xmm10, %xmm1
+; SSE-NEXT:    pand %xmm10, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm8
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,0,0,255,255,255,255,255,0,0,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    pandn %xmm8, %xmm3
+; SSE-NEXT:    pand %xmm4, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm3
+; SSE-NEXT:    movdqa (%r9), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm0[2,1,2,3]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,3,2,0]
+; SSE-NEXT:    movdqa %xmm7, %xmm12
+; SSE-NEXT:    pandn %xmm8, %xmm12
+; SSE-NEXT:    movdqa (%r8), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    pand %xmm7, %xmm8
+; SSE-NEXT:    por %xmm8, %xmm12
+; SSE-NEXT:    movdqa (%rax), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    movdqa %xmm11, %xmm14
+; SSE-NEXT:    pandn %xmm8, %xmm14
+; SSE-NEXT:    pand %xmm11, %xmm12
+; SSE-NEXT:    por %xmm12, %xmm14
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [0,255,255,255,255,0,0,0,255,255,255,255,0,0,0,255]
+; SSE-NEXT:    movdqa %xmm2, %xmm1
 ; SSE-NEXT:    pandn %xmm14, %xmm1
-; SSE-NEXT:    pand %xmm10, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm1
+; SSE-NEXT:    pand %xmm2, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movdqa 16(%rsi), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,1,0,3]
-; SSE-NEXT:    movdqa %xmm11, %xmm3
-; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    movdqa 16(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm11, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm3
-; SSE-NEXT:    movdqa 16(%rcx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,0,3]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm9, %xmm6
-; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    movdqa 16(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm9, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm6
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    pandn %xmm6, %xmm0
-; SSE-NEXT:    pand %xmm12, %xmm3
-; SSE-NEXT:    por %xmm3, %xmm0
-; SSE-NEXT:    movdqa 16(%r9), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[2,1,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,1,2,3]
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,2,0]
-; SSE-NEXT:    movdqa %xmm7, %xmm6
-; SSE-NEXT:    pandn %xmm3, %xmm6
-; SSE-NEXT:    movdqa 16(%r8), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm1[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm7, %xmm3
-; SSE-NEXT:    por %xmm3, %xmm6
-; SSE-NEXT:    movdqa 16(%rax), %xmm1
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[3,1,0,3]
+; SSE-NEXT:    movdqa %xmm9, %xmm8
+; SSE-NEXT:    pandn %xmm3, %xmm8
+; SSE-NEXT:    movdqa 16(%rdi), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm1[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm2, %xmm14
-; SSE-NEXT:    pandn %xmm3, %xmm14
-; SSE-NEXT:    pand %xmm2, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm14
-; SSE-NEXT:    movdqa %xmm10, %xmm1
-; SSE-NEXT:    pandn %xmm14, %xmm1
-; SSE-NEXT:    pand %xmm10, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 32(%rsi), %xmm0
+; SSE-NEXT:    pand %xmm9, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm8
+; SSE-NEXT:    movdqa 16(%rcx), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[3,1,0,3]
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    pandn %xmm3, %xmm0
-; SSE-NEXT:    movdqa 32(%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, (%rsp) # 16-byte Spill
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm1[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm11, %xmm3
-; SSE-NEXT:    por %xmm3, %xmm0
-; SSE-NEXT:    movdqa 32(%rcx), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[2,1,2,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,1,2,3]
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,0,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    movdqa %xmm9, %xmm5
-; SSE-NEXT:    movdqa %xmm9, %xmm6
-; SSE-NEXT:    pandn %xmm3, %xmm6
-; SSE-NEXT:    movdqa 32(%rdx), %xmm9
-; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm9[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa %xmm10, %xmm12
+; SSE-NEXT:    pandn %xmm3, %xmm12
+; SSE-NEXT:    movdqa 16(%rdx), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm0[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm5, %xmm3
-; SSE-NEXT:    por %xmm3, %xmm6
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    pandn %xmm6, %xmm12
-; SSE-NEXT:    por %xmm0, %xmm12
-; SSE-NEXT:    movdqa 32(%r9), %xmm0
+; SSE-NEXT:    pand %xmm10, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm12
+; SSE-NEXT:    movdqa %xmm4, %xmm3
+; SSE-NEXT:    pandn %xmm12, %xmm3
+; SSE-NEXT:    pand %xmm4, %xmm8
+; SSE-NEXT:    por %xmm8, %xmm3
+; SSE-NEXT:    movdqa 16(%r9), %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,3,2,0]
-; SSE-NEXT:    movdqa %xmm7, %xmm3
-; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    movdqa 32(%r8), %xmm11
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm11[3,3,3,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pand %xmm7, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm3
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm0[2,1,2,3]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,3,2,0]
+; SSE-NEXT:    movdqa %xmm7, %xmm12
+; SSE-NEXT:    pandn %xmm8, %xmm12
+; SSE-NEXT:    movdqa 16(%r8), %xmm0
+; SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    pand %xmm7, %xmm8
+; SSE-NEXT:    por %xmm8, %xmm12
+; SSE-NEXT:    movdqa 16(%rax), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm0[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    movdqa %xmm11, %xmm14
+; SSE-NEXT:    pandn %xmm8, %xmm14
+; SSE-NEXT:    pand %xmm11, %xmm12
+; SSE-NEXT:    por %xmm12, %xmm14
+; SSE-NEXT:    movdqa %xmm2, %xmm1
+; SSE-NEXT:    pandn %xmm14, %xmm1
 ; SSE-NEXT:    pand %xmm2, %xmm3
-; SSE-NEXT:    movdqa 32(%rax), %xmm13
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm13[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    por %xmm3, %xmm1
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa 32(%rsi), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,1,2,3]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm3[3,1,0,3]
+; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    pandn %xmm8, %xmm3
+; SSE-NEXT:    movdqa 32(%rdi), %xmm1
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm1[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm1, %xmm14
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    pand %xmm9, %xmm8
+; SSE-NEXT:    por %xmm8, %xmm3
+; SSE-NEXT:    movdqa 32(%rcx), %xmm0
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm0[2,1,2,3]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,3,0,3]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    movdqa %xmm10, %xmm12
+; SSE-NEXT:    pandn %xmm8, %xmm12
+; SSE-NEXT:    movdqa 32(%rdx), %xmm6
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm6[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    pand %xmm10, %xmm8
+; SSE-NEXT:    por %xmm8, %xmm12
+; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    pand %xmm4, %xmm3
+; SSE-NEXT:    pandn %xmm12, %xmm0
+; SSE-NEXT:    por %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm0, %xmm9
+; SSE-NEXT:    movdqa 32(%r9), %xmm13
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[2,1,2,3]
 ; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    pand %xmm10, %xmm12
-; SSE-NEXT:    pandn %xmm2, %xmm10
-; SSE-NEXT:    por %xmm12, %xmm10
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,3,2,0]
+; SSE-NEXT:    movdqa %xmm7, %xmm8
+; SSE-NEXT:    pandn %xmm3, %xmm8
+; SSE-NEXT:    movdqa 32(%r8), %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm0[3,3,3,3,4,5,6,7]
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,5,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
-; SSE-NEXT:    movdqa %xmm12, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pshufhw $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[0,1,2,3,6,6,6,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255]
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    pandn %xmm1, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm6[0,1,2,3,6,6,6,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm7, %xmm2
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    pand %xmm7, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm8
+; SSE-NEXT:    pand %xmm11, %xmm8
+; SSE-NEXT:    movdqa 32(%rax), %xmm4
+; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm4[3,3,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4]
+; SSE-NEXT:    pandn %xmm3, %xmm11
+; SSE-NEXT:    por %xmm8, %xmm11
+; SSE-NEXT:    pand %xmm2, %xmm9
+; SSE-NEXT:    pandn %xmm11, %xmm2
+; SSE-NEXT:    por %xmm9, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm10, %xmm2
 ; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    pshufhw $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm1 = mem[0,1,2,3,6,6,6,6]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
+; SSE-NEXT:    pand %xmm10, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm11, %xmm3
+; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm5[0,1,2,3,6,6,6,6]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm7, %xmm8
+; SSE-NEXT:    pandn %xmm1, %xmm8
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; SSE-NEXT:    movdqa %xmm5, %xmm1
 ; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm5[8],xmm1[9],xmm5[9],xmm1[10],xmm5[10],xmm1[11],xmm5[11],xmm1[12],xmm5[12],xmm1[13],xmm5[13],xmm1[14],xmm5[14],xmm1[15],xmm5[15]
@@ -5827,587 +5824,585 @@ define void @store_i8_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    pand %xmm7, %xmm1
-; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    pand %xmm14, %xmm1
-; SSE-NEXT:    por %xmm0, %xmm1
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm15 = xmm15[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm15[1,1,2,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm8[0,1,2,3,5,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,3,2]
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm4[0,1,2,3,4,5,6,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,3,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm4, %xmm15
-; SSE-NEXT:    pandn %xmm3, %xmm15
-; SSE-NEXT:    pand %xmm4, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm15
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0]
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pandn %xmm15, %xmm0
-; SSE-NEXT:    pand %xmm10, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,5,5,7]
+; SSE-NEXT:    por %xmm8, %xmm1
+; SSE-NEXT:    pand %xmm11, %xmm1
+; SSE-NEXT:    por %xmm3, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm2 = xmm2[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,2,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm2, %xmm8
+; SSE-NEXT:    pandn %xmm3, %xmm8
+; SSE-NEXT:    pshufhw $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[0,1,2,3,5,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,3,2]
+; SSE-NEXT:    pand %xmm2, %xmm3
+; SSE-NEXT:    por %xmm8, %xmm3
+; SSE-NEXT:    pshufhw $164, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = mem[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[2,1,3,3]
+; SSE-NEXT:    movdqa %xmm15, %xmm12
+; SSE-NEXT:    pandn %xmm8, %xmm12
+; SSE-NEXT:    pand %xmm15, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm12
+; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [255,0,0,0,255,255,255,255,0,0,0,255,255,255,255,0]
+; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    pandn %xmm12, %xmm3
+; SSE-NEXT:    pand %xmm9, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE-NEXT:    movdqa %xmm12, %xmm8
-; SSE-NEXT:    movdqa %xmm12, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm10, %xmm3
+; SSE-NEXT:    pandn %xmm1, %xmm3
 ; SSE-NEXT:    pshufhw $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[0,1,2,3,6,6,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
-; SSE-NEXT:    pand %xmm12, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm14, %xmm12
-; SSE-NEXT:    movdqa %xmm14, %xmm3
-; SSE-NEXT:    pandn %xmm2, %xmm3
+; SSE-NEXT:    pand %xmm10, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm11, %xmm8
+; SSE-NEXT:    pandn %xmm3, %xmm8
 ; SSE-NEXT:    pshufhw $170, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm1 = mem[0,1,2,3,6,6,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,1,2,3]
+; SSE-NEXT:    movdqa %xmm7, %xmm3
+; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    pand %xmm7, %xmm1
-; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    pand %xmm14, %xmm1
 ; SSE-NEXT:    por %xmm3, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,2,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm14, %xmm3
-; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    pshufhw $233, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = mem[0,1,2,3,5,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,3,2]
-; SSE-NEXT:    pand %xmm14, %xmm2
-; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    pshufhw $164, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = mem[0,1,2,3,4,5,6,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,3,3]
-; SSE-NEXT:    movdqa %xmm4, %xmm15
-; SSE-NEXT:    pandn %xmm3, %xmm15
-; SSE-NEXT:    pand %xmm4, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm15
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    pandn %xmm15, %xmm0
-; SSE-NEXT:    pand %xmm10, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,5,5,7]
+; SSE-NEXT:    pand %xmm11, %xmm1
+; SSE-NEXT:    por %xmm8, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,1,2,3]
+; SSE-NEXT:    movdqa %xmm2, %xmm8
+; SSE-NEXT:    pandn %xmm3, %xmm8
+; SSE-NEXT:    pshufhw $233, (%rsp), %xmm3 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm3 = mem[0,1,2,3,5,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,3,2]
+; SSE-NEXT:    pand %xmm2, %xmm3
+; SSE-NEXT:    por %xmm8, %xmm3
+; SSE-NEXT:    pshufhw $164, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = mem[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[2,1,3,3]
+; SSE-NEXT:    movdqa %xmm15, %xmm12
+; SSE-NEXT:    pandn %xmm8, %xmm12
+; SSE-NEXT:    pand %xmm15, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm12
+; SSE-NEXT:    movdqa %xmm9, %xmm3
+; SSE-NEXT:    pandn %xmm12, %xmm3
+; SSE-NEXT:    pand %xmm9, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
-; SSE-NEXT:    movdqa %xmm8, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pshufhw $170, (%rsp), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[0,1,2,3,6,6,6,6]
+; SSE-NEXT:    movdqa %xmm10, %xmm3
+; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm14[0,1,2,3,6,6,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
-; SSE-NEXT:    pand %xmm8, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm12, %xmm3
-; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm9[0,1,2,3,6,6,6,6]
+; SSE-NEXT:    pand %xmm10, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm3
+; SSE-NEXT:    movdqa %xmm11, %xmm8
+; SSE-NEXT:    pandn %xmm3, %xmm8
+; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm6[0,1,2,3,6,6,6,6]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,1,2,3]
+; SSE-NEXT:    movdqa %xmm7, %xmm3
+; SSE-NEXT:    pandn %xmm1, %xmm3
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm1 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm1[3,1,2,0,4,5,6,7]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    pand %xmm7, %xmm1
-; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    pand %xmm12, %xmm1
 ; SSE-NEXT:    por %xmm3, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    punpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,2,3]
-; SSE-NEXT:    movdqa %xmm14, %xmm3
-; SSE-NEXT:    pandn %xmm2, %xmm3
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm11[0,1,2,3,5,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,3,2]
-; SSE-NEXT:    pand %xmm14, %xmm2
-; SSE-NEXT:    por %xmm3, %xmm2
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm13[0,1,2,3,4,5,6,6]
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,3,3]
-; SSE-NEXT:    movdqa %xmm4, %xmm15
-; SSE-NEXT:    pandn %xmm3, %xmm15
-; SSE-NEXT:    pand %xmm4, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm15
-; SSE-NEXT:    pand %xmm10, %xmm1
-; SSE-NEXT:    pandn %xmm15, %xmm10
-; SSE-NEXT:    por %xmm1, %xmm10
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pand %xmm11, %xmm1
+; SSE-NEXT:    por %xmm8, %xmm1
+; SSE-NEXT:    punpckhbw {{.*#+}} xmm13 = xmm13[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
+; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm13[1,1,2,3]
+; SSE-NEXT:    movdqa %xmm2, %xmm8
+; SSE-NEXT:    pandn %xmm3, %xmm8
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm0[0,1,2,3,5,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,3,2]
+; SSE-NEXT:    pand %xmm2, %xmm3
+; SSE-NEXT:    por %xmm8, %xmm3
+; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm4[0,1,2,3,4,5,6,6]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[2,1,3,3]
+; SSE-NEXT:    movdqa %xmm15, %xmm12
+; SSE-NEXT:    pandn %xmm8, %xmm12
+; SSE-NEXT:    pand %xmm15, %xmm3
+; SSE-NEXT:    por %xmm3, %xmm12
+; SSE-NEXT:    pand %xmm9, %xmm1
+; SSE-NEXT:    pandn %xmm12, %xmm9
+; SSE-NEXT:    por %xmm1, %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm5[0,0,2,1,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm5, %xmm14
+; SSE-NEXT:    movdqa %xmm5, %xmm11
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,1,3]
-; SSE-NEXT:    movdqa %xmm4, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm6, %xmm8
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm6[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm15, %xmm8
+; SSE-NEXT:    pandn %xmm1, %xmm8
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm5[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
-; SSE-NEXT:    pand %xmm4, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm2
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm0, %xmm15
-; SSE-NEXT:    pandn %xmm2, %xmm15
+; SSE-NEXT:    pand %xmm15, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm8
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm1, %xmm12
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    pandn %xmm8, %xmm12
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm0[0,2,1,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm1[0,1,1,0]
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0]
+; SSE-NEXT:    movdqa %xmm3, %xmm1
+; SSE-NEXT:    pandn %xmm8, %xmm1
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm4[0,2,1,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[0,1,1,0]
-; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0]
-; SSE-NEXT:    movdqa %xmm9, %xmm1
-; SSE-NEXT:    pandn %xmm2, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm3[0,0,2,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,2,1]
-; SSE-NEXT:    pand %xmm9, %xmm2
-; SSE-NEXT:    movdqa %xmm9, %xmm6
-; SSE-NEXT:    por %xmm2, %xmm1
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    por %xmm15, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm5[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,2,1]
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
-; SSE-NEXT:    movdqa %xmm0, %xmm15
-; SSE-NEXT:    pandn %xmm2, %xmm15
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm4[0,0,2,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,0,2,1]
+; SSE-NEXT:    pand %xmm3, %xmm8
+; SSE-NEXT:    movdqa %xmm3, %xmm9
+; SSE-NEXT:    por %xmm8, %xmm1
+; SSE-NEXT:    pand %xmm2, %xmm1
+; SSE-NEXT:    por %xmm12, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm6[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,0,2,1]
+; SSE-NEXT:    movdqa %xmm10, %xmm12
+; SSE-NEXT:    pandn %xmm8, %xmm12
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm13[0,0,0,0,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
-; SSE-NEXT:    pand %xmm0, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm15
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm11[0,0,0,0,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm13[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,0,0,0]
+; SSE-NEXT:    pand %xmm10, %xmm8
+; SSE-NEXT:    por %xmm8, %xmm12
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm2[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,0,0,0]
 ; SSE-NEXT:    movdqa %xmm7, %xmm10
-; SSE-NEXT:    pandn %xmm2, %xmm10
-; SSE-NEXT:    pand %xmm7, %xmm15
-; SSE-NEXT:    por %xmm15, %xmm10
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255]
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    pandn %xmm10, %xmm0
-; SSE-NEXT:    pand %xmm2, %xmm1
-; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[0,1,2,2]
-; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    pandn %xmm1, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm3[1,1,2,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm3, %xmm9
+; SSE-NEXT:    pandn %xmm8, %xmm10
+; SSE-NEXT:    pand %xmm7, %xmm12
+; SSE-NEXT:    por %xmm12, %xmm10
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255]
+; SSE-NEXT:    movdqa %xmm3, %xmm8
+; SSE-NEXT:    pandn %xmm10, %xmm8
+; SSE-NEXT:    pand %xmm3, %xmm1
+; SSE-NEXT:    por %xmm1, %xmm8
+; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,1,2,2]
+; SSE-NEXT:    movdqa %xmm7, %xmm8
+; SSE-NEXT:    pandn %xmm1, %xmm8
+; SSE-NEXT:    pshuflw {{.*#+}} xmm1 = xmm4[1,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm1[0,0,2,1]
 ; SSE-NEXT:    pand %xmm7, %xmm10
-; SSE-NEXT:    por %xmm2, %xmm10
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255]
-; SSE-NEXT:    movdqa %xmm0, %xmm15
-; SSE-NEXT:    movdqa %xmm0, %xmm12
-; SSE-NEXT:    pandn %xmm10, %xmm15
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm14[1,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm2[0,1,2,3,7,5,6,4]
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    movdqa %xmm6, %xmm2
-; SSE-NEXT:    pandn %xmm10, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm8[1,1,2,2,4,5,6,7]
+; SSE-NEXT:    por %xmm8, %xmm10
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255]
+; SSE-NEXT:    movdqa %xmm1, %xmm8
+; SSE-NEXT:    movdqa %xmm1, %xmm14
+; SSE-NEXT:    pandn %xmm10, %xmm8
+; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm11[1,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm10[0,1,2,3,7,5,6,4]
+; SSE-NEXT:    movdqa %xmm9, %xmm12
+; SSE-NEXT:    pandn %xmm10, %xmm12
+; SSE-NEXT:    movdqa %xmm5, %xmm3
+; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm5[1,1,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,0,2,1]
-; SSE-NEXT:    pand %xmm0, %xmm10
-; SSE-NEXT:    por %xmm10, %xmm2
-; SSE-NEXT:    pand %xmm12, %xmm2
-; SSE-NEXT:    por %xmm15, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm5[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,1,1,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm0, %xmm15
-; SSE-NEXT:    pandn %xmm10, %xmm15
-; SSE-NEXT:    movdqa %xmm13, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm13[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,0,0,0]
-; SSE-NEXT:    pand %xmm0, %xmm10
-; SSE-NEXT:    por %xmm10, %xmm15
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm11[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,0,0,0]
-; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    pandn %xmm10, %xmm0
-; SSE-NEXT:    pand %xmm13, %xmm15
-; SSE-NEXT:    por %xmm15, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm10 = [255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm10, %xmm1
-; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm10, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm1
+; SSE-NEXT:    pand %xmm9, %xmm10
+; SSE-NEXT:    por %xmm10, %xmm12
+; SSE-NEXT:    pand %xmm1, %xmm12
+; SSE-NEXT:    por %xmm8, %xmm12
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm6[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,1,1,3]
+; SSE-NEXT:    movdqa %xmm15, %xmm10
+; SSE-NEXT:    pandn %xmm8, %xmm10
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm13[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,0,0,0]
+; SSE-NEXT:    pand %xmm15, %xmm8
+; SSE-NEXT:    por %xmm8, %xmm10
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm2[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,0,0,0]
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm1, %xmm9
+; SSE-NEXT:    pandn %xmm8, %xmm9
+; SSE-NEXT:    pand %xmm1, %xmm10
+; SSE-NEXT:    movdqa %xmm1, %xmm8
+; SSE-NEXT:    por %xmm10, %xmm9
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm5, %xmm1
+; SSE-NEXT:    pandn %xmm9, %xmm1
+; SSE-NEXT:    pand %xmm5, %xmm12
+; SSE-NEXT:    por %xmm12, %xmm1
 ; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm14[2,2,3,3]
-; SSE-NEXT:    movdqa %xmm7, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm8[2,1,3,3,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm8, %xmm10
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
-; SSE-NEXT:    pand %xmm7, %xmm0
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm12, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,5,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm11[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm7, %xmm9
+; SSE-NEXT:    pandn %xmm5, %xmm9
+; SSE-NEXT:    pshuflw {{.*#+}} xmm5 = xmm3[2,1,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm3, %xmm11
+; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,0,2,1]
+; SSE-NEXT:    pand %xmm7, %xmm5
+; SSE-NEXT:    por %xmm9, %xmm5
+; SSE-NEXT:    movdqa %xmm14, %xmm9
+; SSE-NEXT:    pandn %xmm5, %xmm9
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm13, %xmm15
-; SSE-NEXT:    movdqa %xmm13, %xmm8
-; SSE-NEXT:    pandn %xmm0, %xmm8
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm9[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm8, %xmm10
+; SSE-NEXT:    pandn %xmm0, %xmm10
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm4[0,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,1,3]
-; SSE-NEXT:    pand %xmm13, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm8
-; SSE-NEXT:    pand %xmm12, %xmm8
-; SSE-NEXT:    por %xmm2, %xmm8
+; SSE-NEXT:    pand %xmm8, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm10
+; SSE-NEXT:    pand %xmm14, %xmm10
+; SSE-NEXT:    por %xmm9, %xmm10
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,0,0,0,0,255,255,255,0,0,0,0,255,255]
 ; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm1, %xmm6
-; SSE-NEXT:    pandn %xmm8, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm5[0,1,2,3,4,6,5,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,3,2]
+; SSE-NEXT:    movdqa %xmm1, %xmm3
+; SSE-NEXT:    pandn %xmm10, %xmm0
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,6,5,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,1,3,2]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0]
-; SSE-NEXT:    movdqa %xmm1, %xmm5
-; SSE-NEXT:    pandn %xmm2, %xmm5
-; SSE-NEXT:    movdqa %xmm3, %xmm4
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm3[2,2,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,2,1]
-; SSE-NEXT:    pand %xmm1, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm5
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm11[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
+; SSE-NEXT:    movdqa %xmm1, %xmm9
+; SSE-NEXT:    pandn %xmm6, %xmm9
+; SSE-NEXT:    movdqa %xmm13, %xmm12
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm13[2,2,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,0,2,1]
+; SSE-NEXT:    pand %xmm1, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm9
+; SSE-NEXT:    movdqa %xmm2, %xmm10
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm2[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,0,0,0]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm1, %xmm8
-; SSE-NEXT:    pandn %xmm2, %xmm8
-; SSE-NEXT:    pand %xmm1, %xmm5
-; SSE-NEXT:    por %xmm5, %xmm8
-; SSE-NEXT:    pand %xmm6, %xmm8
-; SSE-NEXT:    por %xmm0, %xmm8
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[0,1,1,3]
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm10[0,1,2,3,4,5,5,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,3,2]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0]
 ; SSE-NEXT:    movdqa %xmm1, %xmm5
-; SSE-NEXT:    pandn %xmm0, %xmm5
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm9[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm7, %xmm8
-; SSE-NEXT:    pandn %xmm0, %xmm8
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm14[1,2,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,0,2,1]
-; SSE-NEXT:    pand %xmm7, %xmm2
-; SSE-NEXT:    por %xmm8, %xmm2
-; SSE-NEXT:    pand %xmm1, %xmm2
-; SSE-NEXT:    por %xmm5, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm13[1,2,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm15, %xmm5
-; SSE-NEXT:    pandn %xmm0, %xmm5
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm3[0,1,2,3,4,4,6,5]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,3,3]
-; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    pandn %xmm6, %xmm5
+; SSE-NEXT:    pand %xmm1, %xmm9
+; SSE-NEXT:    por %xmm9, %xmm5
+; SSE-NEXT:    pand %xmm3, %xmm5
 ; SSE-NEXT:    por %xmm0, %xmm5
+; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,1,1,3]
+; SSE-NEXT:    movdqa %xmm1, %xmm6
+; SSE-NEXT:    pandn %xmm0, %xmm6
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm11[0,1,2,3,4,5,5,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,3,2]
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    por %xmm6, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0]
+; SSE-NEXT:    movdqa %xmm3, %xmm6
+; SSE-NEXT:    pandn %xmm0, %xmm6
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    movdqa %xmm4, %xmm8
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm7, %xmm9
+; SSE-NEXT:    pandn %xmm0, %xmm9
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[1,2,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
+; SSE-NEXT:    pand %xmm7, %xmm0
+; SSE-NEXT:    por %xmm9, %xmm0
+; SSE-NEXT:    pand %xmm3, %xmm0
+; SSE-NEXT:    por %xmm6, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm3[1,2,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,0,0,0]
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm4, %xmm9
+; SSE-NEXT:    pandn %xmm6, %xmm9
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm13[0,1,2,3,4,4,6,5]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,1,3,3]
+; SSE-NEXT:    pand %xmm4, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm9
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm10[0,1,2,3,4,5,5,7]
+; SSE-NEXT:    movdqa %xmm10, %xmm5
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm13, %xmm10
+; SSE-NEXT:    pandn %xmm6, %xmm10
+; SSE-NEXT:    pand %xmm13, %xmm9
+; SSE-NEXT:    por %xmm9, %xmm10
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm6, %xmm4
+; SSE-NEXT:    pandn %xmm10, %xmm4
+; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm4
+; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm2[0,1,2,3,6,5,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
-; SSE-NEXT:    movdqa %xmm12, %xmm8
-; SSE-NEXT:    pandn %xmm0, %xmm8
-; SSE-NEXT:    pand %xmm12, %xmm5
-; SSE-NEXT:    por %xmm5, %xmm8
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255]
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    pandn %xmm8, %xmm0
-; SSE-NEXT:    pand %xmm3, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm6[0,1,2,3,6,5,7,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    movdqa %xmm12, %xmm2
+; SSE-NEXT:    movdqa %xmm13, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm10[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm11[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    pand %xmm12, %xmm0
+; SSE-NEXT:    pand %xmm13, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255]
-; SSE-NEXT:    movdqa %xmm5, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm4, %xmm6
+; SSE-NEXT:    pandn %xmm2, %xmm6
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,1,3,3]
+; SSE-NEXT:    movdqa %xmm15, %xmm0
 ; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm14[0,1,2,3,4,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm2[2,1,3,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    pandn %xmm6, %xmm2
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm9[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,2,2,2]
-; SSE-NEXT:    pand %xmm3, %xmm6
-; SSE-NEXT:    movdqa %xmm3, %xmm10
-; SSE-NEXT:    por %xmm6, %xmm2
-; SSE-NEXT:    pand %xmm5, %xmm2
-; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,7,7,7,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm7, %xmm6
-; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm13[0,1,2,3,5,6,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,3]
-; SSE-NEXT:    pand %xmm7, %xmm0
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm8[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,2,2,2]
+; SSE-NEXT:    pand %xmm15, %xmm2
+; SSE-NEXT:    por %xmm2, %xmm0
+; SSE-NEXT:    pand %xmm4, %xmm0
 ; SSE-NEXT:    por %xmm6, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm11[0,1,2,3,6,7,7,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm12[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm7, %xmm6
+; SSE-NEXT:    pandn %xmm2, %xmm6
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,5,6,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,2,2,3]
+; SSE-NEXT:    pand %xmm7, %xmm2
+; SSE-NEXT:    por %xmm6, %xmm2
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm5[0,1,2,3,6,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,1,3,2]
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0]
-; SSE-NEXT:    movdqa %xmm1, %xmm8
-; SSE-NEXT:    pandn %xmm6, %xmm8
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm1, %xmm4
-; SSE-NEXT:    por %xmm0, %xmm8
+; SSE-NEXT:    movdqa {{.*#+}} xmm13 = [255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0]
+; SSE-NEXT:    movdqa %xmm13, %xmm9
+; SSE-NEXT:    pandn %xmm6, %xmm9
+; SSE-NEXT:    pand %xmm13, %xmm2
+; SSE-NEXT:    movdqa %xmm13, %xmm8
+; SSE-NEXT:    por %xmm2, %xmm9
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0]
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    pandn %xmm8, %xmm0
-; SSE-NEXT:    pand %xmm1, %xmm2
-; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm9[0,0,2,1,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    pandn %xmm9, %xmm2
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm13[0,0,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,1,3]
-; SSE-NEXT:    movdqa %xmm3, %xmm2
+; SSE-NEXT:    movdqa %xmm15, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm11[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    pand %xmm3, %xmm0
+; SSE-NEXT:    pand %xmm15, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm2
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm3, %xmm6
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm1, %xmm6
 ; SSE-NEXT:    pandn %xmm2, %xmm6
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm5[0,2,1,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[0,1,1,0]
-; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    movdqa %xmm8, %xmm0
 ; SSE-NEXT:    pandn %xmm2, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[0,0,2,1,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm10[0,0,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,2,1]
-; SSE-NEXT:    pand %xmm4, %xmm2
+; SSE-NEXT:    pand %xmm8, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    pand %xmm3, %xmm0
+; SSE-NEXT:    pand %xmm1, %xmm0
 ; SSE-NEXT:    por %xmm6, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm3[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm12[0,1,1,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,2,1]
-; SSE-NEXT:    movdqa %xmm12, %xmm6
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm3, %xmm6
 ; SSE-NEXT:    pandn %xmm2, %xmm6
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm13[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    movdqa (%rsp), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
-; SSE-NEXT:    pand %xmm12, %xmm2
+; SSE-NEXT:    pand %xmm3, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm6
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm14[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm4[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm7, %xmm8
-; SSE-NEXT:    pandn %xmm2, %xmm8
+; SSE-NEXT:    movdqa %xmm7, %xmm9
+; SSE-NEXT:    pandn %xmm2, %xmm9
 ; SSE-NEXT:    pand %xmm7, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm8
+; SSE-NEXT:    por %xmm6, %xmm9
 ; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255]
-; SSE-NEXT:    movdqa %xmm2, %xmm6
-; SSE-NEXT:    pandn %xmm8, %xmm6
+; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    pandn %xmm9, %xmm3
 ; SSE-NEXT:    pand %xmm2, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    por %xmm0, %xmm3
+; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[0,1,2,2]
 ; SSE-NEXT:    movdqa %xmm7, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[1,1,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm10[1,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
 ; SSE-NEXT:    pand %xmm7, %xmm0
 ; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255]
-; SSE-NEXT:    movdqa %xmm12, %xmm2
+; SSE-NEXT:    movdqa %xmm14, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[1,1,2,1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[1,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm0[0,1,2,3,7,5,6,4]
-; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    movdqa %xmm8, %xmm0
 ; SSE-NEXT:    pandn %xmm6, %xmm0
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm11[1,1,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,0,2,1]
-; SSE-NEXT:    pand %xmm4, %xmm6
+; SSE-NEXT:    pand %xmm8, %xmm6
 ; SSE-NEXT:    por %xmm6, %xmm0
-; SSE-NEXT:    pand %xmm12, %xmm0
+; SSE-NEXT:    pand %xmm14, %xmm0
 ; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm3[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm12[0,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,1,3]
-; SSE-NEXT:    movdqa %xmm10, %xmm6
+; SSE-NEXT:    movdqa %xmm15, %xmm6
 ; SSE-NEXT:    pandn %xmm2, %xmm6
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm13[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[1,1,1,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
-; SSE-NEXT:    pand %xmm10, %xmm2
+; SSE-NEXT:    pand %xmm15, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm6
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm14[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm4[1,1,1,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm15, %xmm8
-; SSE-NEXT:    pandn %xmm2, %xmm8
-; SSE-NEXT:    pand %xmm15, %xmm6
-; SSE-NEXT:    movdqa %xmm15, %xmm10
-; SSE-NEXT:    por %xmm6, %xmm8
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm6, %xmm2
-; SSE-NEXT:    pandn %xmm8, %xmm2
-; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm8, %xmm9
+; SSE-NEXT:    pandn %xmm2, %xmm9
+; SSE-NEXT:    pand %xmm8, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm9
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm3, %xmm2
+; SSE-NEXT:    pandn %xmm9, %xmm2
+; SSE-NEXT:    pand %xmm3, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,2,3,3]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,2,3,3]
 ; SSE-NEXT:    movdqa %xmm7, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm11[2,1,3,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm11, %xmm14
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
 ; SSE-NEXT:    pand %xmm7, %xmm0
 ; SSE-NEXT:    por %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm12, %xmm2
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255]
+; SSE-NEXT:    movdqa %xmm3, %xmm2
 ; SSE-NEXT:    pandn %xmm0, %xmm2
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm5[0,1,2,3,5,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm15, %xmm6
+; SSE-NEXT:    movdqa %xmm8, %xmm10
+; SSE-NEXT:    movdqa %xmm8, %xmm6
 ; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm11[0,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,1,3]
-; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    pand %xmm8, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm6
-; SSE-NEXT:    pand %xmm12, %xmm6
+; SSE-NEXT:    pand %xmm3, %xmm6
 ; SSE-NEXT:    por %xmm2, %xmm6
-; SSE-NEXT:    movdqa {{.*#+}} xmm9 = [255,255,255,0,0,0,0,255,255,255,0,0,0,0,255,255]
-; SSE-NEXT:    movdqa %xmm9, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,0,0,0,0,255,255,255,0,0,0,0,255,255]
+; SSE-NEXT:    movdqa %xmm3, %xmm0
 ; SSE-NEXT:    pandn %xmm6, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,4,6,5,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm2 = xmm12[0,1,2,3,4,6,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,1,3,2]
-; SSE-NEXT:    movdqa %xmm4, %xmm12
-; SSE-NEXT:    movdqa %xmm4, %xmm6
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0]
+; SSE-NEXT:    movdqa %xmm5, %xmm6
 ; SSE-NEXT:    pandn %xmm2, %xmm6
-; SSE-NEXT:    movdqa %xmm13, %xmm5
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm13[2,2,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm1, %xmm8
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm1[2,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,0,2,1]
-; SSE-NEXT:    pand %xmm4, %xmm2
+; SSE-NEXT:    pand %xmm5, %xmm2
 ; SSE-NEXT:    por %xmm2, %xmm6
-; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm14[2,2,2,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm2[0,0,0,0]
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm2, %xmm4
-; SSE-NEXT:    pandn %xmm8, %xmm4
-; SSE-NEXT:    pand %xmm2, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm4
-; SSE-NEXT:    pand %xmm9, %xmm4
-; SSE-NEXT:    por %xmm0, %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[0,1,1,3]
-; SSE-NEXT:    movdqa %xmm2, %xmm6
+; SSE-NEXT:    pshuflw {{.*#+}} xmm2 = xmm4[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm2[0,0,0,0]
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    pandn %xmm9, %xmm2
+; SSE-NEXT:    pand %xmm1, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm2
+; SSE-NEXT:    pand %xmm3, %xmm2
+; SSE-NEXT:    por %xmm0, %xmm2
+; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[0,1,1,3]
+; SSE-NEXT:    movdqa %xmm1, %xmm6
 ; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm11[0,1,2,3,4,5,5,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm14[0,1,2,3,4,5,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,3,2]
-; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    pand %xmm1, %xmm0
 ; SSE-NEXT:    por %xmm6, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0]
-; SSE-NEXT:    movdqa %xmm2, %xmm6
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0]
+; SSE-NEXT:    movdqa %xmm1, %xmm6
 ; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm11[0,1,2,3,5,5,5,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm7, %xmm8
-; SSE-NEXT:    pandn %xmm0, %xmm8
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm9[1,2,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm7, %xmm9
+; SSE-NEXT:    pandn %xmm0, %xmm9
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm5[1,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
 ; SSE-NEXT:    pand %xmm7, %xmm0
-; SSE-NEXT:    por %xmm8, %xmm0
-; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    por %xmm9, %xmm0
+; SSE-NEXT:    pand %xmm1, %xmm0
 ; SSE-NEXT:    por %xmm6, %xmm0
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm13[1,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm10, %xmm8
-; SSE-NEXT:    pandn %xmm6, %xmm8
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm5[0,1,2,3,4,4,6,5]
+; SSE-NEXT:    movdqa %xmm10, %xmm9
+; SSE-NEXT:    pandn %xmm6, %xmm9
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm8[0,1,2,3,4,4,6,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,1,3,3]
 ; SSE-NEXT:    pand %xmm10, %xmm6
-; SSE-NEXT:    por %xmm6, %xmm8
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm14[0,1,2,3,4,5,5,7]
-; SSE-NEXT:    movdqa %xmm14, %xmm2
+; SSE-NEXT:    por %xmm6, %xmm9
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm4[0,1,2,3,4,5,5,7]
+; SSE-NEXT:    movdqa %xmm4, %xmm1
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
-; SSE-NEXT:    movdqa %xmm3, %xmm10
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm2, %xmm10
 ; SSE-NEXT:    pandn %xmm6, %xmm10
-; SSE-NEXT:    pand %xmm3, %xmm8
-; SSE-NEXT:    movdqa %xmm3, %xmm6
-; SSE-NEXT:    por %xmm8, %xmm10
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255]
-; SSE-NEXT:    movdqa %xmm3, %xmm4
+; SSE-NEXT:    pand %xmm2, %xmm9
+; SSE-NEXT:    movdqa %xmm2, %xmm12
+; SSE-NEXT:    por %xmm9, %xmm10
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm2, %xmm4
 ; SSE-NEXT:    pandn %xmm10, %xmm4
-; SSE-NEXT:    pand %xmm3, %xmm0
+; SSE-NEXT:    pand %xmm2, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm4
 ; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm15[0,1,2,3,6,5,7,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm3[0,1,2,3,6,5,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    movdqa %xmm6, %xmm3
+; SSE-NEXT:    movdqa %xmm12, %xmm6
 ; SSE-NEXT:    pandn %xmm0, %xmm6
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm11[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm14[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    pand %xmm3, %xmm0
-; SSE-NEXT:    movdqa %xmm3, %xmm14
+; SSE-NEXT:    pand %xmm12, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm6
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255]
-; SSE-NEXT:    movdqa %xmm3, %xmm8
-; SSE-NEXT:    pandn %xmm6, %xmm8
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm9[0,1,2,3,4,6,6,7]
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm2, %xmm9
+; SSE-NEXT:    pandn %xmm6, %xmm9
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm5[0,1,2,3,4,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm0[2,1,3,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm4, %xmm0
+; SSE-NEXT:    movdqa %xmm15, %xmm0
 ; SSE-NEXT:    pandn %xmm6, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm1[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm11[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,2,2,2]
-; SSE-NEXT:    pand %xmm4, %xmm6
-; SSE-NEXT:    movdqa %xmm4, %xmm11
+; SSE-NEXT:    pand %xmm15, %xmm6
 ; SSE-NEXT:    por %xmm6, %xmm0
-; SSE-NEXT:    pand %xmm3, %xmm0
-; SSE-NEXT:    por %xmm8, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm5[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    por %xmm9, %xmm0
+; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm8[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm7, %xmm8
-; SSE-NEXT:    pandn %xmm6, %xmm8
+; SSE-NEXT:    movdqa %xmm7, %xmm9
+; SSE-NEXT:    pandn %xmm6, %xmm9
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm6 = xmm13[0,1,2,3,5,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[2,2,2,3]
 ; SSE-NEXT:    pand %xmm7, %xmm6
-; SSE-NEXT:    por %xmm8, %xmm6
-; SSE-NEXT:    pshufhw {{.*#+}} xmm8 = xmm2[0,1,2,3,6,7,7,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[2,1,3,2]
-; SSE-NEXT:    movdqa %xmm12, %xmm10
-; SSE-NEXT:    pandn %xmm8, %xmm10
-; SSE-NEXT:    pand %xmm12, %xmm6
-; SSE-NEXT:    movdqa %xmm12, %xmm4
+; SSE-NEXT:    por %xmm9, %xmm6
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm1[0,1,2,3,6,7,7,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[2,1,3,2]
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0]
+; SSE-NEXT:    movdqa %xmm4, %xmm10
+; SSE-NEXT:    pandn %xmm9, %xmm10
+; SSE-NEXT:    pand %xmm4, %xmm6
+; SSE-NEXT:    movdqa %xmm4, %xmm11
 ; SSE-NEXT:    por %xmm6, %xmm10
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0]
 ; SSE-NEXT:    movdqa %xmm1, %xmm2
@@ -6415,278 +6410,273 @@ define void @store_i8_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    pand %xmm1, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm9 = xmm9[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm9[0,0,2,1,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm13 = xmm13[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm13[0,0,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,1,3]
-; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    movdqa %xmm11, %xmm8
-; SSE-NEXT:    pandn %xmm0, %xmm8
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm11[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm15, %xmm6
+; SSE-NEXT:    pandn %xmm0, %xmm6
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm8[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm8
-; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm3, %xmm10
-; SSE-NEXT:    pandn %xmm8, %xmm10
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm5[0,2,1,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm0[0,1,1,0]
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    pandn %xmm8, %xmm0
-; SSE-NEXT:    movdqa (%rsp), %xmm13 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm13[0,0,2,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,0,2,1]
-; SSE-NEXT:    pand %xmm12, %xmm8
-; SSE-NEXT:    por %xmm8, %xmm0
-; SSE-NEXT:    pand %xmm3, %xmm0
-; SSE-NEXT:    por %xmm10, %xmm0
+; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm6
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm1, %xmm9
+; SSE-NEXT:    pandn %xmm6, %xmm9
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm3[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,0,2,1]
-; SSE-NEXT:    movdqa %xmm14, %xmm10
-; SSE-NEXT:    pandn %xmm8, %xmm10
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm1[0,0,0,0,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,0,0,0]
-; SSE-NEXT:    pand %xmm14, %xmm8
-; SSE-NEXT:    por %xmm8, %xmm10
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm6[0,0,0,0,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm7, %xmm15
-; SSE-NEXT:    pandn %xmm8, %xmm15
-; SSE-NEXT:    pand %xmm7, %xmm10
-; SSE-NEXT:    por %xmm10, %xmm15
-; SSE-NEXT:    movdqa {{.*#+}} xmm8 = [255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255]
-; SSE-NEXT:    movdqa %xmm8, %xmm10
-; SSE-NEXT:    pandn %xmm15, %xmm10
-; SSE-NEXT:    pand %xmm8, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm10
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm5[0,1,2,2]
-; SSE-NEXT:    movdqa %xmm7, %xmm8
-; SSE-NEXT:    pandn %xmm0, %xmm8
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm13[1,1,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
-; SSE-NEXT:    pand %xmm7, %xmm0
-; SSE-NEXT:    por %xmm8, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255]
-; SSE-NEXT:    movdqa %xmm12, %xmm8
-; SSE-NEXT:    pandn %xmm0, %xmm8
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[1,1,2,1]
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm0[0,1,2,3,7,5,6,4]
-; SSE-NEXT:    movdqa %xmm4, %xmm14
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm3[0,2,1,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm0[0,1,1,0]
 ; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    pandn %xmm10, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm11[1,1,2,2,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,0,2,1]
-; SSE-NEXT:    pand %xmm4, %xmm10
-; SSE-NEXT:    por %xmm10, %xmm0
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    movdqa %xmm12, %xmm4
-; SSE-NEXT:    por %xmm8, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm3[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,1,1,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm2, %xmm10
-; SSE-NEXT:    pandn %xmm8, %xmm10
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm1[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,0,0,0]
-; SSE-NEXT:    pand %xmm2, %xmm8
-; SSE-NEXT:    por %xmm8, %xmm10
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm6[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm8[0,0,0,0]
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm2, %xmm15
-; SSE-NEXT:    pandn %xmm8, %xmm15
-; SSE-NEXT:    pand %xmm2, %xmm10
-; SSE-NEXT:    movdqa %xmm2, %xmm12
-; SSE-NEXT:    por %xmm10, %xmm15
-; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm2, %xmm8
-; SSE-NEXT:    pandn %xmm15, %xmm8
-; SSE-NEXT:    pand %xmm2, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm8
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm9[2,2,3,3]
+; SSE-NEXT:    pandn %xmm6, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm2[0,0,2,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,0,2,1]
+; SSE-NEXT:    pand %xmm4, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm0
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    por %xmm9, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm5[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,0,2,1]
+; SSE-NEXT:    movdqa %xmm12, %xmm9
+; SSE-NEXT:    pandn %xmm6, %xmm9
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm1[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,0,0,0]
+; SSE-NEXT:    pand %xmm12, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm9
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm4[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,0,0,0]
 ; SSE-NEXT:    movdqa %xmm7, %xmm10
-; SSE-NEXT:    pandn %xmm0, %xmm10
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm11[2,1,3,3,4,5,6,7]
+; SSE-NEXT:    pandn %xmm6, %xmm10
+; SSE-NEXT:    pand %xmm7, %xmm9
+; SSE-NEXT:    por %xmm9, %xmm10
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255]
+; SSE-NEXT:    movdqa %xmm6, %xmm9
+; SSE-NEXT:    pandn %xmm10, %xmm9
+; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[0,1,2,2]
+; SSE-NEXT:    movdqa %xmm7, %xmm6
+; SSE-NEXT:    pandn %xmm0, %xmm6
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[1,1,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
 ; SSE-NEXT:    pand %xmm7, %xmm0
-; SSE-NEXT:    por %xmm10, %xmm0
-; SSE-NEXT:    movdqa %xmm4, %xmm10
-; SSE-NEXT:    pandn %xmm0, %xmm10
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm5[0,1,2,3,5,6,6,7]
+; SSE-NEXT:    por %xmm6, %xmm0
+; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255]
+; SSE-NEXT:    movdqa %xmm14, %xmm6
+; SSE-NEXT:    pandn %xmm0, %xmm6
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[1,1,2,1]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm0[0,1,2,3,7,5,6,4]
+; SSE-NEXT:    movdqa %xmm11, %xmm0
+; SSE-NEXT:    pandn %xmm9, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm8[1,1,2,2,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,0,2,1]
+; SSE-NEXT:    pand %xmm11, %xmm9
+; SSE-NEXT:    movdqa %xmm11, %xmm12
+; SSE-NEXT:    por %xmm9, %xmm0
+; SSE-NEXT:    pand %xmm14, %xmm0
+; SSE-NEXT:    por %xmm6, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm5[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,1,1,3]
+; SSE-NEXT:    movdqa %xmm15, %xmm9
+; SSE-NEXT:    pandn %xmm6, %xmm9
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm1[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,0,0,0]
+; SSE-NEXT:    pand %xmm15, %xmm6
+; SSE-NEXT:    por %xmm6, %xmm9
+; SSE-NEXT:    pshuflw {{.*#+}} xmm6 = xmm4[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[0,0,0,0]
+; SSE-NEXT:    movdqa {{.*#+}} xmm11 = [255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm11, %xmm10
+; SSE-NEXT:    pandn %xmm6, %xmm10
+; SSE-NEXT:    pand %xmm11, %xmm9
+; SSE-NEXT:    por %xmm9, %xmm10
+; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm6, %xmm9
+; SSE-NEXT:    pandn %xmm10, %xmm9
+; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm9
+; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm13[2,2,3,3]
+; SSE-NEXT:    movdqa %xmm7, %xmm9
+; SSE-NEXT:    pandn %xmm0, %xmm9
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm8[2,1,3,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
+; SSE-NEXT:    pand %xmm7, %xmm0
+; SSE-NEXT:    por %xmm9, %xmm0
+; SSE-NEXT:    movdqa %xmm14, %xmm9
+; SSE-NEXT:    pandn %xmm0, %xmm9
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm3[0,1,2,3,5,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm12, %xmm15
-; SSE-NEXT:    pandn %xmm0, %xmm15
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm13[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm11, %xmm10
+; SSE-NEXT:    pandn %xmm0, %xmm10
+; SSE-NEXT:    movdqa %xmm2, %xmm3
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[0,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,1,3]
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm15
-; SSE-NEXT:    pand %xmm4, %xmm15
-; SSE-NEXT:    por %xmm10, %xmm15
+; SSE-NEXT:    pand %xmm11, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm10
+; SSE-NEXT:    pand %xmm14, %xmm10
+; SSE-NEXT:    por %xmm9, %xmm10
 ; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,0,0,0,0,255,255,255,0,0,0,0,255,255]
-; SSE-NEXT:    movdqa %xmm2, %xmm10
-; SSE-NEXT:    pandn %xmm15, %xmm10
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm3[0,1,2,3,4,6,5,7]
+; SSE-NEXT:    movdqa %xmm2, %xmm9
+; SSE-NEXT:    pandn %xmm10, %xmm9
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm5[0,1,2,3,4,6,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,3,2]
-; SSE-NEXT:    movdqa %xmm14, %xmm15
-; SSE-NEXT:    pandn %xmm0, %xmm15
-; SSE-NEXT:    movdqa %xmm1, %xmm8
+; SSE-NEXT:    movdqa %xmm12, %xmm10
+; SSE-NEXT:    pandn %xmm0, %xmm10
+; SSE-NEXT:    movdqa %xmm1, %xmm6
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[2,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
-; SSE-NEXT:    pand %xmm14, %xmm0
-; SSE-NEXT:    movdqa %xmm14, %xmm4
-; SSE-NEXT:    por %xmm0, %xmm15
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm6[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pand %xmm12, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm10
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm4[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm1, %xmm5
+; SSE-NEXT:    pandn %xmm0, %xmm5
+; SSE-NEXT:    pand %xmm1, %xmm10
+; SSE-NEXT:    por %xmm10, %xmm5
+; SSE-NEXT:    pand %xmm2, %xmm5
+; SSE-NEXT:    por %xmm9, %xmm5
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm11[0,1,1,3]
 ; SSE-NEXT:    movdqa %xmm1, %xmm9
 ; SSE-NEXT:    pandn %xmm0, %xmm9
-; SSE-NEXT:    pand %xmm1, %xmm15
-; SSE-NEXT:    por %xmm15, %xmm9
-; SSE-NEXT:    pand %xmm2, %xmm9
-; SSE-NEXT:    por %xmm10, %xmm9
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[0,1,1,3]
-; SSE-NEXT:    movdqa %xmm1, %xmm10
-; SSE-NEXT:    pandn %xmm0, %xmm10
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm11[0,1,2,3,4,5,5,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm8[0,1,2,3,4,5,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,3,2]
 ; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    por %xmm10, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm13[0,1,2,3,5,5,5,5]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm7, %xmm15
-; SSE-NEXT:    pandn %xmm10, %xmm15
+; SSE-NEXT:    por %xmm9, %xmm0
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm3[0,1,2,3,5,5,5,5]
+; SSE-NEXT:    movdqa %xmm3, %xmm12
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[2,2,2,2]
+; SSE-NEXT:    movdqa %xmm7, %xmm10
+; SSE-NEXT:    pandn %xmm9, %xmm10
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm3[1,2,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,0,2,1]
-; SSE-NEXT:    pand %xmm7, %xmm10
-; SSE-NEXT:    por %xmm15, %xmm10
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm3[1,2,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,0,2,1]
+; SSE-NEXT:    pand %xmm7, %xmm9
+; SSE-NEXT:    por %xmm10, %xmm9
 ; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,0,0,255,255,255,255,255,0,0,255,255,255,255,255,0]
-; SSE-NEXT:    pand %xmm1, %xmm10
+; SSE-NEXT:    pand %xmm1, %xmm9
 ; SSE-NEXT:    pandn %xmm0, %xmm1
-; SSE-NEXT:    por %xmm10, %xmm1
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm5[1,2,2,3,4,5,6,7]
+; SSE-NEXT:    por %xmm9, %xmm1
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm13[1,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm12, %xmm10
-; SSE-NEXT:    pandn %xmm0, %xmm10
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm8[0,1,2,3,4,4,6,5]
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm2, %xmm9
+; SSE-NEXT:    pandn %xmm0, %xmm9
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm6[0,1,2,3,4,4,6,5]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,3,3]
-; SSE-NEXT:    pand %xmm12, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm10
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm6[0,1,2,3,4,5,5,7]
+; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm9
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,4,5,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
-; SSE-NEXT:    movdqa %xmm14, %xmm15
-; SSE-NEXT:    pandn %xmm0, %xmm15
-; SSE-NEXT:    pand %xmm14, %xmm10
-; SSE-NEXT:    por %xmm10, %xmm15
+; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
+; SSE-NEXT:    movdqa %xmm2, %xmm10
+; SSE-NEXT:    pandn %xmm0, %xmm10
+; SSE-NEXT:    pand %xmm2, %xmm9
+; SSE-NEXT:    por %xmm9, %xmm10
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,0,0,0,255,255,255,255,0,0,0,255,255,255]
 ; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pandn %xmm15, %xmm0
+; SSE-NEXT:    pandn %xmm10, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm2[0,1,2,3,6,5,7,7]
+; SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm11[0,1,2,3,6,5,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; SSE-NEXT:    movdqa %xmm14, %xmm10
-; SSE-NEXT:    pandn %xmm0, %xmm10
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm11[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    movdqa %xmm2, %xmm9
+; SSE-NEXT:    pandn %xmm0, %xmm9
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm8[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    pand %xmm14, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm10
+; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm9
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm3[0,1,2,3,4,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,1,3,3]
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm1, %xmm15
-; SSE-NEXT:    pandn %xmm0, %xmm15
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm13[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    movdqa %xmm15, %xmm10
+; SSE-NEXT:    pandn %xmm0, %xmm10
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm12[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    por %xmm0, %xmm15
+; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm10
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,0,0,255,255,255,255,255,0,0,255,255,255]
-; SSE-NEXT:    pand %xmm0, %xmm15
-; SSE-NEXT:    pandn %xmm10, %xmm0
-; SSE-NEXT:    por %xmm15, %xmm0
+; SSE-NEXT:    pand %xmm0, %xmm10
+; SSE-NEXT:    pandn %xmm9, %xmm0
+; SSE-NEXT:    por %xmm10, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm8[0,1,2,3,7,7,7,7]
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm6[0,1,2,3,7,7,7,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT:    movdqa %xmm7, %xmm10
-; SSE-NEXT:    pandn %xmm0, %xmm10
-; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm5[0,1,2,3,5,6,6,7]
+; SSE-NEXT:    movdqa %xmm7, %xmm9
+; SSE-NEXT:    pandn %xmm0, %xmm9
+; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm13[0,1,2,3,5,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,3]
 ; SSE-NEXT:    pand %xmm7, %xmm0
-; SSE-NEXT:    por %xmm10, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm10 = xmm6[0,1,2,3,6,7,7,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[2,1,3,2]
-; SSE-NEXT:    movdqa %xmm4, %xmm14
-; SSE-NEXT:    movdqa %xmm4, %xmm15
-; SSE-NEXT:    pandn %xmm10, %xmm15
-; SSE-NEXT:    pand %xmm4, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm15
+; SSE-NEXT:    por %xmm9, %xmm0
+; SSE-NEXT:    pshufhw {{.*#+}} xmm9 = xmm4[0,1,2,3,6,7,7,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[2,1,3,2]
+; SSE-NEXT:    movdqa {{.*#+}} xmm12 = [255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0]
+; SSE-NEXT:    movdqa %xmm12, %xmm10
+; SSE-NEXT:    pandn %xmm9, %xmm10
+; SSE-NEXT:    pand %xmm12, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm10
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [0,0,255,255,255,255,0,0,0,255,255,255,255,0,0,0]
 ; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pandn %xmm15, %xmm0
+; SSE-NEXT:    pandn %xmm10, %xmm0
 ; SSE-NEXT:    por %xmm1, %xmm0
-; SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm12 = xmm12[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm12[0,0,2,1,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm6 = xmm6[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm6[0,0,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,1,3]
-; SSE-NEXT:    movdqa %xmm2, %xmm10
-; SSE-NEXT:    pandn %xmm0, %xmm10
+; SSE-NEXT:    movdqa %xmm15, %xmm9
+; SSE-NEXT:    pandn %xmm0, %xmm9
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm11[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    pand %xmm2, %xmm0
-; SSE-NEXT:    movdqa %xmm2, %xmm6
-; SSE-NEXT:    por %xmm0, %xmm10
+; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm9
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; SSE-NEXT:    punpcklbw {{.*#+}} xmm8 = xmm8[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm8[0,2,1,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,1,0]
-; SSE-NEXT:    movdqa %xmm4, %xmm15
-; SSE-NEXT:    pandn %xmm0, %xmm15
+; SSE-NEXT:    movdqa %xmm12, %xmm10
+; SSE-NEXT:    pandn %xmm0, %xmm10
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm13[0,0,2,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
-; SSE-NEXT:    pand %xmm4, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm15
+; SSE-NEXT:    pand %xmm12, %xmm0
+; SSE-NEXT:    por %xmm0, %xmm10
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm0, %xmm15
-; SSE-NEXT:    pandn %xmm10, %xmm0
-; SSE-NEXT:    por %xmm15, %xmm0
+; SSE-NEXT:    pand %xmm0, %xmm10
+; SSE-NEXT:    pandn %xmm9, %xmm0
+; SSE-NEXT:    por %xmm10, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, %xmm3
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[0,0,0,0,4,5,6,7]
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm4[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    movdqa {{.*#+}} xmm15 = [255,255,255,255,255,0,255,255,255,255,255,255,0,255,255,255]
-; SSE-NEXT:    pand %xmm15, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm5 = xmm5[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm5[0,1,1,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,0,2,1]
-; SSE-NEXT:    pandn %xmm10, %xmm15
-; SSE-NEXT:    por %xmm0, %xmm15
+; SSE-NEXT:    movdqa %xmm2, %xmm10
+; SSE-NEXT:    pand %xmm2, %xmm0
+; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm2[0,1,1,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,0,2,1]
+; SSE-NEXT:    pandn %xmm9, %xmm10
+; SSE-NEXT:    por %xmm0, %xmm10
 ; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm7, %xmm10
-; SSE-NEXT:    pandn %xmm0, %xmm10
-; SSE-NEXT:    pand %xmm7, %xmm15
-; SSE-NEXT:    por %xmm15, %xmm10
+; SSE-NEXT:    movdqa %xmm7, %xmm9
+; SSE-NEXT:    pandn %xmm0, %xmm9
+; SSE-NEXT:    pand %xmm7, %xmm10
+; SSE-NEXT:    por %xmm10, %xmm9
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,0,0,0,255,255,255,255,0,0,0,255,255]
 ; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    pandn %xmm10, %xmm0
+; SSE-NEXT:    pandn %xmm9, %xmm0
 ; SSE-NEXT:    por %xmm3, %xmm0
 ; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[0,1,2,2]
@@ -6696,40 +6686,40 @@ define void @store_i8_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
 ; SSE-NEXT:    pand %xmm7, %xmm0
 ; SSE-NEXT:    por %xmm3, %xmm0
-; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255]
-; SSE-NEXT:    movdqa %xmm4, %xmm10
-; SSE-NEXT:    pandn %xmm0, %xmm10
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[1,1,2,1]
+; SSE-NEXT:    movdqa {{.*#+}} xmm14 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255]
+; SSE-NEXT:    movdqa %xmm14, %xmm9
+; SSE-NEXT:    pandn %xmm0, %xmm9
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[1,1,2,1]
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,5,6,4]
-; SSE-NEXT:    movdqa %xmm14, %xmm3
+; SSE-NEXT:    movdqa %xmm12, %xmm3
 ; SSE-NEXT:    pandn %xmm0, %xmm3
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm11[1,1,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
-; SSE-NEXT:    pand %xmm14, %xmm0
+; SSE-NEXT:    pand %xmm12, %xmm0
 ; SSE-NEXT:    por %xmm0, %xmm3
-; SSE-NEXT:    pand %xmm4, %xmm3
-; SSE-NEXT:    por %xmm10, %xmm3
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[1,1,1,1,4,5,6,7]
+; SSE-NEXT:    pand %xmm14, %xmm3
+; SSE-NEXT:    por %xmm9, %xmm3
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm4[1,1,1,1,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    movdqa %xmm6, %xmm4
-; SSE-NEXT:    pand %xmm6, %xmm0
-; SSE-NEXT:    pshuflw {{.*#+}} xmm10 = xmm5[0,2,2,3,4,5,6,7]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm10[0,1,1,3]
-; SSE-NEXT:    pandn %xmm10, %xmm4
-; SSE-NEXT:    por %xmm0, %xmm4
+; SSE-NEXT:    pand %xmm15, %xmm0
+; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm2[0,2,2,3,4,5,6,7]
+; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm9[0,1,1,3]
+; SSE-NEXT:    pandn %xmm9, %xmm15
+; SSE-NEXT:    por %xmm0, %xmm15
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm1[1,1,1,1,4,5,6,7]
-; SSE-NEXT:    movdqa %xmm1, %xmm15
+; SSE-NEXT:    movdqa %xmm1, %xmm10
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
-; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255]
-; SSE-NEXT:    movdqa %xmm6, %xmm10
-; SSE-NEXT:    pandn %xmm0, %xmm10
-; SSE-NEXT:    pand %xmm6, %xmm4
-; SSE-NEXT:    por %xmm4, %xmm10
-; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255]
-; SSE-NEXT:    pand %xmm1, %xmm3
-; SSE-NEXT:    pandn %xmm10, %xmm1
-; SSE-NEXT:    por %xmm3, %xmm1
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[2,2,3,3]
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255]
+; SSE-NEXT:    movdqa %xmm1, %xmm9
+; SSE-NEXT:    pandn %xmm0, %xmm9
+; SSE-NEXT:    pand %xmm1, %xmm15
+; SSE-NEXT:    por %xmm15, %xmm9
+; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,0,0,0,255,255,255,255,0,0,0,255,255,255,255]
+; SSE-NEXT:    pand %xmm0, %xmm3
+; SSE-NEXT:    pandn %xmm9, %xmm0
+; SSE-NEXT:    por %xmm3, %xmm0
+; SSE-NEXT:    movdqa %xmm0, %xmm9
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm6[2,2,3,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm11[2,1,3,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,0,2,1]
 ; SSE-NEXT:    pand %xmm7, %xmm3
@@ -6737,43 +6727,44 @@ define void @store_i8_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
 ; SSE-NEXT:    por %xmm3, %xmm7
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm13[0,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,1,3]
-; SSE-NEXT:    pand %xmm6, %xmm0
+; SSE-NEXT:    movdqa %xmm1, %xmm6
+; SSE-NEXT:    pand %xmm1, %xmm0
 ; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm8[0,1,2,3,5,6,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,2,2,2]
 ; SSE-NEXT:    pandn %xmm3, %xmm6
 ; SSE-NEXT:    por %xmm0, %xmm6
-; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255]
-; SSE-NEXT:    pand %xmm0, %xmm6
+; SSE-NEXT:    movdqa %xmm14, %xmm0
+; SSE-NEXT:    pand %xmm14, %xmm6
 ; SSE-NEXT:    pandn %xmm7, %xmm0
 ; SSE-NEXT:    por %xmm6, %xmm0
-; SSE-NEXT:    movdqa %xmm0, %xmm4
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm2[2,2,2,3,4,5,6,7]
+; SSE-NEXT:    movdqa %xmm0, %xmm6
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm4[2,2,2,3,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,2,1]
-; SSE-NEXT:    pand %xmm14, %xmm0
-; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm5[0,1,2,3,4,6,5,7]
+; SSE-NEXT:    pand %xmm12, %xmm0
+; SSE-NEXT:    pshufhw {{.*#+}} xmm3 = xmm2[0,1,2,3,4,6,5,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,1,3,2]
-; SSE-NEXT:    pandn %xmm3, %xmm14
-; SSE-NEXT:    por %xmm0, %xmm14
+; SSE-NEXT:    pandn %xmm3, %xmm12
+; SSE-NEXT:    por %xmm0, %xmm12
 ; SSE-NEXT:    movdqa {{.*#+}} xmm3 = [255,255,0,255,255,255,255,255,255,0,255,255,255,255,255,255]
-; SSE-NEXT:    pand %xmm3, %xmm14
-; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm15[2,2,2,2,4,5,6,7]
+; SSE-NEXT:    pand %xmm3, %xmm12
+; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm10[2,2,2,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    pandn %xmm0, %xmm3
-; SSE-NEXT:    por %xmm14, %xmm3
+; SSE-NEXT:    por %xmm12, %xmm3
 ; SSE-NEXT:    movdqa {{.*#+}} xmm0 = [255,255,255,0,0,0,0,255,255,255,0,0,0,0,255,255]
 ; SSE-NEXT:    pand %xmm0, %xmm3
-; SSE-NEXT:    pandn %xmm4, %xmm0
-; SSE-NEXT:    por %xmm0, %xmm3
+; SSE-NEXT:    pandn %xmm6, %xmm0
+; SSE-NEXT:    por %xmm3, %xmm0
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movdqa %xmm3, 368(%rax)
-; SSE-NEXT:    movdqa %xmm1, 352(%rax)
+; SSE-NEXT:    movdqa %xmm0, 368(%rax)
+; SSE-NEXT:    movdqa %xmm9, 352(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 336(%rax)
-; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 320(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 320(%rax)
+; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 288(%rax)
-; SSE-NEXT:    movdqa %xmm9, 256(%rax)
+; SSE-NEXT:    movdqa %xmm5, 256(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 240(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
diff --git a/llvm/test/CodeGen/X86/vselect-minmax.ll b/llvm/test/CodeGen/X86/vselect-minmax.ll
index cb0542ca7cea8..d0acfc31a203d 100644
--- a/llvm/test/CodeGen/X86/vselect-minmax.ll
+++ b/llvm/test/CodeGen/X86/vselect-minmax.ll
@@ -10283,47 +10283,46 @@ entry:
 define <8 x i64> @concat_smin_smax(<4 x i64> %a0, <4 x i64> %a1) {
 ; SSE2-LABEL: concat_smin_smax:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movdqa {{.*#+}} xmm6 = [2147483648,2147483648]
+; SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [2147483648,2147483648]
 ; SSE2-NEXT:    movdqa %xmm0, %xmm4
-; SSE2-NEXT:    pxor %xmm6, %xmm4
-; SSE2-NEXT:    movdqa %xmm2, %xmm5
-; SSE2-NEXT:    pxor %xmm6, %xmm5
-; SSE2-NEXT:    movdqa %xmm5, %xmm7
+; SSE2-NEXT:    pxor %xmm5, %xmm4
+; SSE2-NEXT:    movdqa %xmm2, %xmm6
+; SSE2-NEXT:    pxor %xmm5, %xmm6
+; SSE2-NEXT:    movdqa %xmm6, %xmm7
 ; SSE2-NEXT:    pcmpgtd %xmm4, %xmm7
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm8 = xmm7[0,0,2,2]
-; SSE2-NEXT:    pcmpeqd %xmm4, %xmm5
-; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[1,1,3,3]
-; SSE2-NEXT:    pand %xmm8, %xmm4
-; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm7[1,1,3,3]
-; SSE2-NEXT:    por %xmm4, %xmm5
-; SSE2-NEXT:    movdqa %xmm5, %xmm7
-; SSE2-NEXT:    pandn %xmm2, %xmm7
-; SSE2-NEXT:    movdqa %xmm0, %xmm4
-; SSE2-NEXT:    pand %xmm5, %xmm4
-; SSE2-NEXT:    por %xmm7, %xmm4
-; SSE2-NEXT:    movdqa %xmm1, %xmm7
-; SSE2-NEXT:    pxor %xmm6, %xmm7
-; SSE2-NEXT:    pxor %xmm3, %xmm6
-; SSE2-NEXT:    movdqa %xmm6, %xmm8
-; SSE2-NEXT:    pcmpgtd %xmm7, %xmm8
-; SSE2-NEXT:    pshufd {{.*#+}} xmm9 = xmm8[0,0,2,2]
-; SSE2-NEXT:    pcmpeqd %xmm7, %xmm6
+; SSE2-NEXT:    pcmpeqd %xmm4, %xmm6
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm6[1,1,3,3]
-; SSE2-NEXT:    pand %xmm9, %xmm6
-; SSE2-NEXT:    pshufd {{.*#+}} xmm7 = xmm8[1,1,3,3]
-; SSE2-NEXT:    por %xmm6, %xmm7
-; SSE2-NEXT:    movdqa %xmm7, %xmm6
-; SSE2-NEXT:    pandn %xmm3, %xmm6
-; SSE2-NEXT:    movdqa %xmm7, %xmm8
+; SSE2-NEXT:    pand %xmm8, %xmm6
+; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm7[1,1,3,3]
+; SSE2-NEXT:    por %xmm6, %xmm4
+; SSE2-NEXT:    movdqa %xmm4, %xmm6
+; SSE2-NEXT:    pandn %xmm2, %xmm6
+; SSE2-NEXT:    movdqa %xmm4, %xmm7
+; SSE2-NEXT:    pandn %xmm0, %xmm7
+; SSE2-NEXT:    pand %xmm4, %xmm0
+; SSE2-NEXT:    por %xmm6, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm6
+; SSE2-NEXT:    pxor %xmm5, %xmm6
+; SSE2-NEXT:    pxor %xmm3, %xmm5
+; SSE2-NEXT:    movdqa %xmm5, %xmm8
+; SSE2-NEXT:    pcmpgtd %xmm6, %xmm8
+; SSE2-NEXT:    pshufd {{.*#+}} xmm9 = xmm8[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm6, %xmm5
+; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
+; SSE2-NEXT:    pand %xmm9, %xmm5
+; SSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm8[1,1,3,3]
+; SSE2-NEXT:    por %xmm5, %xmm6
+; SSE2-NEXT:    movdqa %xmm6, %xmm5
+; SSE2-NEXT:    pandn %xmm3, %xmm5
+; SSE2-NEXT:    movdqa %xmm6, %xmm8
 ; SSE2-NEXT:    pandn %xmm1, %xmm8
-; SSE2-NEXT:    pand %xmm7, %xmm1
-; SSE2-NEXT:    por %xmm6, %xmm1
-; SSE2-NEXT:    pand %xmm5, %xmm2
-; SSE2-NEXT:    pandn %xmm0, %xmm5
-; SSE2-NEXT:    por %xmm5, %xmm2
-; SSE2-NEXT:    pand %xmm7, %xmm3
+; SSE2-NEXT:    pand %xmm6, %xmm1
+; SSE2-NEXT:    por %xmm5, %xmm1
+; SSE2-NEXT:    pand %xmm4, %xmm2
+; SSE2-NEXT:    por %xmm7, %xmm2
+; SSE2-NEXT:    pand %xmm6, %xmm3
 ; SSE2-NEXT:    por %xmm8, %xmm3
-; SSE2-NEXT:    movdqa %xmm4, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE4-LABEL: concat_smin_smax:
diff --git a/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast_from_memory.ll b/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast_from_memory.ll
index c0fa13f1a3008..75dfec32f6be4 100644
--- a/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast_from_memory.ll
+++ b/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast_from_memory.ll
@@ -5836,6 +5836,7 @@ define void @vec512_i128_widen_to_i256_factor2_broadcast_to_v2i256_factor2(ptr %
 ;
 ; AVX-LABEL: vec512_i128_widen_to_i256_factor2_broadcast_to_v2i256_factor2:
 ; AVX:       # %bb.0:
+; AVX-NEXT:    pushq %r14
 ; AVX-NEXT:    pushq %rbx
 ; AVX-NEXT:    movq (%rdi), %rax
 ; AVX-NEXT:    movq %rax, %rcx
@@ -5847,38 +5848,38 @@ define void @vec512_i128_widen_to_i256_factor2_broadcast_to_v2i256_factor2(ptr %
 ; AVX-NEXT:    vmovd %eax, %xmm0
 ; AVX-NEXT:    shrl $8, %eax
 ; AVX-NEXT:    vpinsrb $1, %eax, %xmm0, %xmm0
+; AVX-NEXT:    movq 8(%rdi), %r14
 ; AVX-NEXT:    shrl $16, %ebx
 ; AVX-NEXT:    vpinsrb $2, %ebx, %xmm0, %xmm0
+; AVX-NEXT:    movq %r14, %rax
 ; AVX-NEXT:    shrl $24, %r11d
 ; AVX-NEXT:    vpinsrb $3, %r11d, %xmm0, %xmm0
+; AVX-NEXT:    movq %r14, %rdi
 ; AVX-NEXT:    shrq $32, %r10
 ; AVX-NEXT:    vpinsrb $4, %r10d, %xmm0, %xmm0
+; AVX-NEXT:    movq %r14, %r10
 ; AVX-NEXT:    shrq $40, %r9
 ; AVX-NEXT:    vpinsrb $5, %r9d, %xmm0, %xmm0
+; AVX-NEXT:    movq %r14, %r9
 ; AVX-NEXT:    shrq $48, %r8
 ; AVX-NEXT:    vpinsrb $6, %r8d, %xmm0, %xmm0
-; AVX-NEXT:    movq 8(%rdi), %rax
+; AVX-NEXT:    movl %r14d, %r8d
 ; AVX-NEXT:    shrq $56, %rcx
 ; AVX-NEXT:    vpinsrb $7, %ecx, %xmm0, %xmm0
-; AVX-NEXT:    movl %eax, %ecx
-; AVX-NEXT:    shrl $8, %ecx
-; AVX-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0
-; AVX-NEXT:    vpinsrb $9, %ecx, %xmm0, %xmm0
-; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    movl %r14d, %ecx
+; AVX-NEXT:    vpinsrb $8, %r14d, %xmm0, %xmm0
+; AVX-NEXT:    shrl $8, %r14d
+; AVX-NEXT:    vpinsrb $9, %r14d, %xmm0, %xmm0
 ; AVX-NEXT:    shrl $16, %ecx
 ; AVX-NEXT:    vpinsrb $10, %ecx, %xmm0, %xmm0
-; AVX-NEXT:    movl %eax, %ecx
-; AVX-NEXT:    shrl $24, %ecx
-; AVX-NEXT:    vpinsrb $11, %ecx, %xmm0, %xmm0
-; AVX-NEXT:    movq %rax, %rcx
-; AVX-NEXT:    shrq $32, %rcx
-; AVX-NEXT:    vpinsrb $12, %ecx, %xmm0, %xmm0
-; AVX-NEXT:    movq %rax, %rcx
-; AVX-NEXT:    shrq $40, %rcx
-; AVX-NEXT:    vpinsrb $13, %ecx, %xmm0, %xmm0
-; AVX-NEXT:    movq %rax, %rcx
-; AVX-NEXT:    shrq $48, %rcx
-; AVX-NEXT:    vpinsrb $14, %ecx, %xmm0, %xmm0
+; AVX-NEXT:    shrl $24, %r8d
+; AVX-NEXT:    vpinsrb $11, %r8d, %xmm0, %xmm0
+; AVX-NEXT:    shrq $32, %r9
+; AVX-NEXT:    vpinsrb $12, %r9d, %xmm0, %xmm0
+; AVX-NEXT:    shrq $40, %r10
+; AVX-NEXT:    vpinsrb $13, %r10d, %xmm0, %xmm0
+; AVX-NEXT:    shrq $48, %rdi
+; AVX-NEXT:    vpinsrb $14, %edi, %xmm0, %xmm0
 ; AVX-NEXT:    shrq $56, %rax
 ; AVX-NEXT:    vpinsrb $15, %eax, %xmm0, %xmm0
 ; AVX-NEXT:    vpaddb 32(%rsi), %xmm0, %xmm1
@@ -5890,60 +5891,63 @@ define void @vec512_i128_widen_to_i256_factor2_broadcast_to_v2i256_factor2(ptr %
 ; AVX-NEXT:    vmovdqa %xmm0, (%rdx)
 ; AVX-NEXT:    vmovdqa %xmm1, 32(%rdx)
 ; AVX-NEXT:    popq %rbx
+; AVX-NEXT:    popq %r14
 ; AVX-NEXT:    retq
 ;
 ; AVX2-LABEL: vec512_i128_widen_to_i256_factor2_broadcast_to_v2i256_factor2:
 ; AVX2:       # %bb.0:
+; AVX2-NEXT:    pushq %r14
 ; AVX2-NEXT:    pushq %rbx
-; AVX2-NEXT:    movq (%rdi), %rax
-; AVX2-NEXT:    movq %rax, %rcx
-; AVX2-NEXT:    movq %rax, %r8
-; AVX2-NEXT:    movq %rax, %r9
-; AVX2-NEXT:    movq %rax, %r10
-; AVX2-NEXT:    movl %eax, %r11d
-; AVX2-NEXT:    movl %eax, %ebx
-; AVX2-NEXT:    vmovd %eax, %xmm0
-; AVX2-NEXT:    shrl $8, %eax
-; AVX2-NEXT:    vpinsrb $1, %eax, %xmm0, %xmm0
+; AVX2-NEXT:    movq (%rdi), %rcx
+; AVX2-NEXT:    movq %rcx, %rax
+; AVX2-NEXT:    movq %rcx, %r8
+; AVX2-NEXT:    movq %rcx, %r9
+; AVX2-NEXT:    movq %rcx, %r10
+; AVX2-NEXT:    movl %ecx, %r11d
+; AVX2-NEXT:    movl %ecx, %ebx
+; AVX2-NEXT:    vmovd %ecx, %xmm0
+; AVX2-NEXT:    shrl $8, %ecx
+; AVX2-NEXT:    vpinsrb $1, %ecx, %xmm0, %xmm0
+; AVX2-NEXT:    movq 8(%rdi), %r14
 ; AVX2-NEXT:    shrl $16, %ebx
 ; AVX2-NEXT:    vpinsrb $2, %ebx, %xmm0, %xmm0
+; AVX2-NEXT:    movq %r14, %rcx
 ; AVX2-NEXT:    shrl $24, %r11d
 ; AVX2-NEXT:    vpinsrb $3, %r11d, %xmm0, %xmm0
+; AVX2-NEXT:    movq %r14, %rdi
 ; AVX2-NEXT:    shrq $32, %r10
 ; AVX2-NEXT:    vpinsrb $4, %r10d, %xmm0, %xmm0
+; AVX2-NEXT:    movq %r14, %r10
 ; AVX2-NEXT:    shrq $40, %r9
 ; AVX2-NEXT:    vpinsrb $5, %r9d, %xmm0, %xmm0
+; AVX2-NEXT:    movq %r14, %r9
 ; AVX2-NEXT:    shrq $48, %r8
 ; AVX2-NEXT:    vpinsrb $6, %r8d, %xmm0, %xmm0
-; AVX2-NEXT:    movq 8(%rdi), %rax
-; AVX2-NEXT:    shrq $56, %rcx
-; AVX2-NEXT:    vpinsrb $7, %ecx, %xmm0, %xmm0
-; AVX2-NEXT:    movl %eax, %ecx
-; AVX2-NEXT:    shrl $8, %ecx
-; AVX2-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0
-; AVX2-NEXT:    vpinsrb $9, %ecx, %xmm0, %xmm0
-; AVX2-NEXT:    movl %eax, %ecx
-; AVX2-NEXT:    shrl $16, %ecx
-; AVX2-NEXT:    vpinsrb $10, %ecx, %xmm0, %xmm0
-; AVX2-NEXT:    movl %eax, %ecx
-; AVX2-NEXT:    shrl $24, %ecx
-; AVX2-NEXT:    vpinsrb $11, %ecx, %xmm0, %xmm0
-; AVX2-NEXT:    movq %rax, %rcx
-; AVX2-NEXT:    shrq $32, %rcx
-; AVX2-NEXT:    vpinsrb $12, %ecx, %xmm0, %xmm0
-; AVX2-NEXT:    movq %rax, %rcx
-; AVX2-NEXT:    shrq $40, %rcx
-; AVX2-NEXT:    vpinsrb $13, %ecx, %xmm0, %xmm0
-; AVX2-NEXT:    movq %rax, %rcx
-; AVX2-NEXT:    shrq $48, %rcx
-; AVX2-NEXT:    vpinsrb $14, %ecx, %xmm0, %xmm0
+; AVX2-NEXT:    movl %r14d, %r8d
 ; AVX2-NEXT:    shrq $56, %rax
-; AVX2-NEXT:    vpinsrb $15, %eax, %xmm0, %xmm0
+; AVX2-NEXT:    vpinsrb $7, %eax, %xmm0, %xmm0
+; AVX2-NEXT:    movl %r14d, %eax
+; AVX2-NEXT:    vpinsrb $8, %r14d, %xmm0, %xmm0
+; AVX2-NEXT:    shrl $8, %r14d
+; AVX2-NEXT:    vpinsrb $9, %r14d, %xmm0, %xmm0
+; AVX2-NEXT:    shrl $16, %eax
+; AVX2-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0
+; AVX2-NEXT:    shrl $24, %r8d
+; AVX2-NEXT:    vpinsrb $11, %r8d, %xmm0, %xmm0
+; AVX2-NEXT:    shrq $32, %r9
+; AVX2-NEXT:    vpinsrb $12, %r9d, %xmm0, %xmm0
+; AVX2-NEXT:    shrq $40, %r10
+; AVX2-NEXT:    vpinsrb $13, %r10d, %xmm0, %xmm0
+; AVX2-NEXT:    shrq $48, %rdi
+; AVX2-NEXT:    vpinsrb $14, %edi, %xmm0, %xmm0
+; AVX2-NEXT:    shrq $56, %rcx
+; AVX2-NEXT:    vpinsrb $15, %ecx, %xmm0, %xmm0
 ; AVX2-NEXT:    vpaddb 32(%rsi), %ymm0, %ymm1
 ; AVX2-NEXT:    vpaddb (%rsi), %ymm0, %ymm0
 ; AVX2-NEXT:    vmovdqa %ymm0, (%rdx)
 ; AVX2-NEXT:    vmovdqa %ymm1, 32(%rdx)
 ; AVX2-NEXT:    popq %rbx
+; AVX2-NEXT:    popq %r14
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;