tstellar · tstellar · Jan 15, 2022 · Oct 8, 2021 · Aug 24, 2021
diff --git a/clang/test/Driver/undefined-libs.cpp b/clang/test/Driver/undefined-libs.cpp
@@ -5,7 +5,7 @@
 // STDLIB: error: invalid library name in argument '-stdlib=nostdlib'
 // STDLIB-EMPTY:
 
-// RUN: not %clangxx --target=i386-unknown-linux -rtlib=nortlib %s 2>&1 | FileCheck --check-prefix=RTLIB %s
+// RUN: not %clangxx --target=i386-unknown-linux -rtlib=nortlib --unwindlib=libgcc %s 2>&1 | FileCheck --check-prefix=RTLIB %s
 // RTLIB: error: invalid runtime library name in argument '-rtlib=nortlib'
 // RTLIB-EMPTY:
 

diff --git a/lld/CMakeLists.txt b/lld/CMakeLists.txt
@@ -176,15 +176,6 @@ if (LLD_DEFAULT_LD_LLD_IS_MINGW)
   add_definitions("-DLLD_DEFAULT_LD_LLD_IS_MINGW=1")
 endif()
 
-option(LLD_DEFAULT_NOSTART_STOP_GC
-  "Default ld.lld to -z nostart-stop-gc. If ON, C identifier name sections are
-  forced retained by __start_/__stop_ references. This may increase output size
-  for many instrumentations, but is compatible with GNU ld newer than 2015-10"
-  ON)
-if (LLD_DEFAULT_NOSTART_STOP_GC)
-  add_definitions("-DLLD_DEFAULT_NOSTART_STOP_GC=1")
-endif()
-
 if (MSVC)
   add_definitions(-wd4530) # Suppress 'warning C4530: C++ exception handler used, but unwind semantics are not enabled.'
   add_definitions(-wd4062) # Suppress 'warning C4062: enumerator X in switch of enum Y is not handled' from system header.

diff --git a/lld/ELF/Driver.cpp b/lld/ELF/Driver.cpp
@@ -1193,17 +1193,8 @@ static void readConfigs(opt::InputArgList &args) {
   config->zSeparate = getZSeparate(args);
   config->zShstk = hasZOption(args, "shstk");
   config->zStackSize = args::getZOptionValue(args, OPT_z, "stack-size", 0);
-#ifdef LLD_DEFAULT_NOSTART_STOP_GC
-  // -z start-stop-gc default matches GNU ld<2015-10 and ld64 section$start
-  // symbols and can decrease file size for many instrumentations.  However,
-  // some users need time to accommodate the -z nostart-stop-gc default, so this
-  // is added as a temporary workaround.
-  config->zStartStopGC =
-      getZFlag(args, "start-stop-gc", "nostart-stop-gc", false);
-#else
   config->zStartStopGC =
       getZFlag(args, "start-stop-gc", "nostart-stop-gc", true);
-#endif
   config->zStartStopVisibility = getZStartStopVisibility(args);
   config->zText = getZFlag(args, "text", "notext", true);
   config->zWxneeded = hasZOption(args, "wxneeded");

diff --git a/lld/test/CMakeLists.txt b/lld/test/CMakeLists.txt
@@ -11,7 +11,6 @@ llvm_canonicalize_cmake_booleans(
   LLVM_ENABLE_ZLIB
   LLVM_ENABLE_LIBXML2
   LLD_DEFAULT_LD_LLD_IS_MINGW
-  LLD_DEFAULT_NOSTART_STOP_GC
   LLVM_HAVE_LIBXAR
   )
 

diff --git a/lld/test/ELF/gc-sections-metadata-startstop.s b/lld/test/ELF/gc-sections-metadata-startstop.s
@@ -1,5 +1,4 @@
 # REQUIRES: x86
-# UNSUPPORTED: default-nostart-stop-gc
 # LINK_ORDER cnamed sections are not kept alive by the __start_* reference.
 
 # RUN: llvm-mc -filetype=obj -triple=x86_64-pc-linux %s -o %t.o

diff --git a/lld/test/ELF/gc-sections-startstop-hint.s b/lld/test/ELF/gc-sections-startstop-hint.s
@@ -1,5 +1,4 @@
 # REQUIRES: x86
-# UNSUPPORTED: default-nostart-stop-gc
 ## Some projects may not work with GNU ld<2015-10 (ld.lld 13.0.0) --gc-sections behavior.
 ## Give a hint.
 

diff --git a/lld/test/ELF/gc-sections-startstop.s b/lld/test/ELF/gc-sections-startstop.s
@@ -1,7 +1,6 @@
 ## Check that group members are retained or discarded as a unit.
 
 # REQUIRES: x86
-# UNSUPPORTED: default-nostart-stop-gc
 
 # RUN: llvm-mc -filetype=obj -triple=x86_64-unknown-linux %s -o %t.o
 # RUN: ld.lld %t.o --gc-sections -o %t

diff --git a/lld/test/lit.cfg.py b/lld/test/lit.cfg.py
@@ -118,6 +118,3 @@
 # ELF tests expect the default target for ld.lld to be ELF.
 if config.ld_lld_default_mingw:
     config.excludes.append('ELF')
-
-if config.ld_lld_default_nostart_stop_gc:
-    config.available_features.add('default-nostart-stop-gc')
diff --git a/lld/test/lit.site.cfg.py.in b/lld/test/lit.site.cfg.py.in
@@ -19,7 +19,6 @@ config.have_libxar = @LLVM_HAVE_LIBXAR@
 config.have_libxml2 = @LLVM_ENABLE_LIBXML2@
 config.sizeof_void_p = @CMAKE_SIZEOF_VOID_P@
 config.ld_lld_default_mingw = @LLD_DEFAULT_LD_LLD_IS_MINGW@
-config.ld_lld_default_nostart_stop_gc = @LLD_DEFAULT_NOSTART_STOP_GC@
 
 # Support substitution of the tools and libs dirs with user parameters. This is
 # used when we can't determine the tool dir at configuration time.

diff --git a/llvm/lib/Target/AArch64/GISel/AArch64InstructionSelector.cpp b/llvm/lib/Target/AArch64/GISel/AArch64InstructionSelector.cpp
@@ -2701,29 +2701,29 @@ bool AArch64InstructionSelector::select(MachineInstr &I) {
   case TargetOpcode::G_ZEXTLOAD:
   case TargetOpcode::G_LOAD:
   case TargetOpcode::G_STORE: {
+    GLoadStore &LdSt = cast<GLoadStore>(I);
     bool IsZExtLoad = I.getOpcode() == TargetOpcode::G_ZEXTLOAD;
-    LLT PtrTy = MRI.getType(I.getOperand(1).getReg());
+    LLT PtrTy = MRI.getType(LdSt.getPointerReg());
 
     if (PtrTy != LLT::pointer(0, 64)) {
       LLVM_DEBUG(dbgs() << "Load/Store pointer has type: " << PtrTy
                         << ", expected: " << LLT::pointer(0, 64) << '\n');
       return false;
     }
 
-    auto &MemOp = **I.memoperands_begin();
-    uint64_t MemSizeInBytes = MemOp.getSize();
-    unsigned MemSizeInBits = MemSizeInBytes * 8;
-    AtomicOrdering Order = MemOp.getSuccessOrdering();
+    uint64_t MemSizeInBytes = LdSt.getMemSize();
+    unsigned MemSizeInBits = LdSt.getMemSizeInBits();
+    AtomicOrdering Order = LdSt.getMMO().getSuccessOrdering();
 
     // Need special instructions for atomics that affect ordering.
     if (Order != AtomicOrdering::NotAtomic &&
         Order != AtomicOrdering::Unordered &&
         Order != AtomicOrdering::Monotonic) {
-      assert(I.getOpcode() != TargetOpcode::G_ZEXTLOAD);
+      assert(!isa<GZExtLoad>(LdSt));
       if (MemSizeInBytes > 64)
         return false;
 
-      if (I.getOpcode() == TargetOpcode::G_LOAD) {
+      if (isa<GLoad>(LdSt)) {
         static unsigned Opcodes[] = {AArch64::LDARB, AArch64::LDARH,
                                      AArch64::LDARW, AArch64::LDARX};
         I.setDesc(TII.get(Opcodes[Log2_32(MemSizeInBytes)]));
@@ -2737,7 +2737,7 @@ bool AArch64InstructionSelector::select(MachineInstr &I) {
     }
 
 #ifndef NDEBUG
-    const Register PtrReg = I.getOperand(1).getReg();
+    const Register PtrReg = LdSt.getPointerReg();
     const RegisterBank &PtrRB = *RBI.getRegBank(PtrReg, MRI, TRI);
     // Sanity-check the pointer register.
     assert(PtrRB.getID() == AArch64::GPRRegBankID &&
@@ -2746,13 +2746,31 @@ bool AArch64InstructionSelector::select(MachineInstr &I) {
            "Load/Store pointer operand isn't a pointer");
 #endif
 
-    const Register ValReg = I.getOperand(0).getReg();
+    const Register ValReg = LdSt.getReg(0);
+    const LLT ValTy = MRI.getType(ValReg);
     const RegisterBank &RB = *RBI.getRegBank(ValReg, MRI, TRI);
 
+    // The code below doesn't support truncating stores, so we need to split it
+    // again.
+    if (isa<GStore>(LdSt) && ValTy.getSizeInBits() > MemSizeInBits) {
+      unsigned SubReg;
+      LLT MemTy = LdSt.getMMO().getMemoryType();
+      auto *RC = getRegClassForTypeOnBank(MemTy, RB, RBI);
+      if (!getSubRegForClass(RC, TRI, SubReg))
+        return false;
+
+      // Generate a subreg copy.
+      auto Copy = MIB.buildInstr(TargetOpcode::COPY, {MemTy}, {})
+                      .addReg(ValReg, 0, SubReg)
+                      .getReg(0);
+      RBI.constrainGenericRegister(Copy, *RC, MRI);
+      LdSt.getOperand(0).setReg(Copy);
+    }
+
     // Helper lambda for partially selecting I. Either returns the original
     // instruction with an updated opcode, or a new instruction.
     auto SelectLoadStoreAddressingMode = [&]() -> MachineInstr * {
-      bool IsStore = I.getOpcode() == TargetOpcode::G_STORE;
+      bool IsStore = isa<GStore>(I);
       const unsigned NewOpc =
           selectLoadStoreUIOp(I.getOpcode(), RB.getID(), MemSizeInBits);
       if (NewOpc == I.getOpcode())
@@ -2769,7 +2787,8 @@ bool AArch64InstructionSelector::select(MachineInstr &I) {
 
       // Folded something. Create a new instruction and return it.
       auto NewInst = MIB.buildInstr(NewOpc, {}, {}, I.getFlags());
-      IsStore ? NewInst.addUse(ValReg) : NewInst.addDef(ValReg);
+      Register CurValReg = I.getOperand(0).getReg();
+      IsStore ? NewInst.addUse(CurValReg) : NewInst.addDef(CurValReg);
       NewInst.cloneMemRefs(I);
       for (auto &Fn : *AddrModeFns)
         Fn(NewInst);

diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/select-store-truncating-float.mir b/llvm/test/CodeGen/AArch64/GlobalISel/select-store-truncating-float.mir
@@ -0,0 +1,115 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=aarch64-- -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s
+--- |
+  define void @truncating_f32(double %x) {
+    %alloca = alloca i32, align 4
+    %bitcast = bitcast double %x to i64
+    %trunc = trunc i64 %bitcast to i32
+    store i32 %trunc, i32* %alloca, align 4
+    ret void
+  }
+
+  define void @truncating_f16(double %x) {
+    %alloca = alloca i16, align 2
+    %bitcast = bitcast double %x to i64
+    %trunc = trunc i64 %bitcast to i16
+    store i16 %trunc, i16* %alloca, align 2
+    ret void
+  }
+
+  define void @truncating_f8(double %x) {
+    %alloca = alloca i8, align 1
+    %bitcast = bitcast double %x to i64
+    %trunc = trunc i64 %bitcast to i8
+    store i8 %trunc, i8* %alloca, align 1
+    ret void
+  }
+
+...
+---
+name:            truncating_f32
+alignment:       4
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+liveins:
+  - { reg: '$d0' }
+frameInfo:
+  maxAlignment:    4
+stack:
+  - { id: 0, name: alloca, size: 4, alignment: 4 }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0
+
+    ; CHECK-LABEL: name: truncating_f32
+    ; CHECK: liveins: $d0
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr32 = COPY [[COPY]].ssub
+    ; CHECK: STRSui [[COPY1]], %stack.0.alloca, 0 :: (store (s32) into %ir.alloca)
+    ; CHECK: RET_ReallyLR
+    %0:fpr(s64) = COPY $d0
+    %1:gpr(p0) = G_FRAME_INDEX %stack.0.alloca
+    G_STORE %0(s64), %1(p0) :: (store (s32) into %ir.alloca)
+    RET_ReallyLR
+
+...
+---
+name:            truncating_f16
+alignment:       4
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+liveins:
+  - { reg: '$d0' }
+frameInfo:
+  maxAlignment:    2
+stack:
+  - { id: 0, name: alloca, size: 2, alignment: 2 }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0
+
+    ; CHECK-LABEL: name: truncating_f16
+    ; CHECK: liveins: $d0
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr16 = COPY [[COPY]].hsub
+    ; CHECK: STRHui [[COPY1]], %stack.0.alloca, 0 :: (store (s16) into %ir.alloca)
+    ; CHECK: RET_ReallyLR
+    %0:fpr(s64) = COPY $d0
+    %1:gpr(p0) = G_FRAME_INDEX %stack.0.alloca
+    G_STORE %0(s64), %1(p0) :: (store (s16) into %ir.alloca)
+    RET_ReallyLR
+
+...
+---
+name:            truncating_f8
+alignment:       4
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+liveins:
+  - { reg: '$d0' }
+frameInfo:
+  maxAlignment:    1
+stack:
+  - { id: 0, name: alloca, size: 1, alignment: 1 }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0
+
+    ; CHECK-LABEL: name: truncating_f8
+    ; CHECK: liveins: $d0
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr8 = COPY [[COPY]].bsub
+    ; CHECK: STRBui [[COPY1]], %stack.0.alloca, 0 :: (store (s8) into %ir.alloca)
+    ; CHECK: RET_ReallyLR
+    %0:fpr(s64) = COPY $d0
+    %1:gpr(p0) = G_FRAME_INDEX %stack.0.alloca
+    G_STORE %0(s64), %1(p0) :: (store (s8) into %ir.alloca)
+    RET_ReallyLR
+
+...
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/select-with-no-legality-check.mir b/llvm/test/CodeGen/AArch64/GlobalISel/select-with-no-legality-check.mir
@@ -278,13 +278,13 @@ body:             |
     ; CHECK-LABEL: name: test_rule96_id2146_at_idx8070
     ; CHECK: liveins: $x0
     ; CHECK: [[COPY:%[0-9]+]]:gpr64sp = COPY $x0
-    ; CHECK: [[LDRBui:%[0-9]+]]:fpr8 = LDRBui [[COPY]], 0 :: (load (s1))
+    ; CHECK: [[LDRBui:%[0-9]+]]:fpr8 = LDRBui [[COPY]], 0 :: (load (s8))
     ; CHECK: [[COPY1:%[0-9]+]]:gpr32 = COPY [[LDRBui]]
-    ; CHECK: [[UBFMWri:%[0-9]+]]:gpr32 = UBFMWri [[COPY1]], 0, 0
+    ; CHECK: [[UBFMWri:%[0-9]+]]:gpr32 = UBFMWri [[COPY1]], 0, 7
     ; CHECK: $noreg = PATCHABLE_RET [[UBFMWri]]
     %2:gpr(p0) = COPY $x0
-    %0:fpr(s1) = G_LOAD %2(p0) :: (load (s1))
-    %1:gpr(s32) = G_ZEXT %0(s1)
+    %0:fpr(s8) = G_LOAD %2(p0) :: (load (s8))
+    %1:gpr(s32) = G_ZEXT %0(s8)
     $noreg = PATCHABLE_RET %1(s32)
 
 ...