[mlir][vector] Decouple unrolling gather and gather to llvm lowering (#132206)

This patch decouples unrolling vector.gather and lowering vector.gather
to llvm.masked.gather.

This is consistent with how vector.load, vector.store,
vector.maskedload, vector.maskedstore lower to LLVM.

Some interesting test changes from this patch:

- 2D vector.gather lowering to llvm tests are deleted. This is
consistent with other memory load/store ops.
- There are still tests for 2D vector.gather, but the constant mask for
these test is modified. This is because with the updated lowering, one
of the unrolled vector.gather disappears because it is masked off (also
demonstrating why this is a better lowering path)

Overall, this makes vector.gather take the same consistent path for
lowering to LLVM as other load/store ops.

Discourse Discussion:
https://discourse.llvm.org/t/rfc-improving-gather-codegen-for-vector-dialect/85011/13

Author: Groverkss

mlir/include/mlir/Dialect/Vector/Transforms/LoweringPatterns.h

@@ -241,16 +241,20 @@ void populateVectorStepLoweringPatterns(RewritePatternSet &patterns,
 
 /// Populate the pattern set with the following patterns:
 ///
-/// [FlattenGather]
-/// Flattens 2 or more dimensional `vector.gather` ops by unrolling the
+/// [UnrollGather]
+/// Unrolls 2 or more dimensional `vector.gather` ops by unrolling the
 /// outermost dimension.
+void populateVectorGatherLoweringPatterns(RewritePatternSet &patterns,
+                                          PatternBenefit benefit = 1);
+
+/// Populate the pattern set with the following patterns:
 ///
 /// [Gather1DToConditionalLoads]
 /// Turns 1-d `vector.gather` into a scalarized sequence of `vector.loads` or
 /// `tensor.extract`s. To avoid out-of-bounds memory accesses, these
 /// loads/extracts are made conditional using `scf.if` ops.
-void populateVectorGatherLoweringPatterns(RewritePatternSet &patterns,
-                                          PatternBenefit benefit = 1);
+void populateVectorGatherToConditionalLoadPatterns(RewritePatternSet &patterns,
+                                                   PatternBenefit benefit = 1);
 
 /// Populates instances of `MaskOpRewritePattern` to lower masked operations
 /// with `vector.mask`. Patterns should rewrite the `vector.mask` operation and

mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVM.cpp

@@ -269,49 +269,30 @@ class VectorGatherOpConversion
     if (failed(isMemRefTypeSupported(memRefType, *this->getTypeConverter())))
       return failure();
 
-    auto loc = gather->getLoc();
+    VectorType vType = gather.getVectorType();
+    if (vType.getRank() > 1)
+      return failure();
+
+    Location loc = gather->getLoc();
 
     // Resolve alignment.
     unsigned align;
     if (failed(getMemRefAlignment(*getTypeConverter(), memRefType, align)))
       return failure();
 
+    // Resolve address.
     Value ptr = getStridedElementPtr(loc, memRefType, adaptor.getBase(),
                                      adaptor.getIndices(), rewriter);
     Value base = adaptor.getBase();
+    Value ptrs =
+        getIndexedPtrs(rewriter, loc, *this->getTypeConverter(), memRefType,
+                       base, ptr, adaptor.getIndexVec(), vType);
 
-    auto llvmNDVectorTy = adaptor.getIndexVec().getType();
-    // Handle the simple case of 1-D vector.
-    if (!isa<LLVM::LLVMArrayType>(llvmNDVectorTy)) {
-      auto vType = gather.getVectorType();
-      // Resolve address.
-      Value ptrs =
-          getIndexedPtrs(rewriter, loc, *this->getTypeConverter(), memRefType,
-                         base, ptr, adaptor.getIndexVec(), vType);
-      // Replace with the gather intrinsic.
-      rewriter.replaceOpWithNewOp<LLVM::masked_gather>(
-          gather, typeConverter->convertType(vType), ptrs, adaptor.getMask(),
-          adaptor.getPassThru(), rewriter.getI32IntegerAttr(align));
-      return success();
-    }
-
-    const LLVMTypeConverter &typeConverter = *this->getTypeConverter();
-    auto callback = [align, memRefType, base, ptr, loc, &rewriter,
-                     &typeConverter](Type llvm1DVectorTy,
-                                     ValueRange vectorOperands) {
-      // Resolve address.
-      Value ptrs = getIndexedPtrs(
-          rewriter, loc, typeConverter, memRefType, base, ptr,
-          /*index=*/vectorOperands[0], cast<VectorType>(llvm1DVectorTy));
-      // Create the gather intrinsic.
-      return rewriter.create<LLVM::masked_gather>(
-          loc, llvm1DVectorTy, ptrs, /*mask=*/vectorOperands[1],
-          /*passThru=*/vectorOperands[2], rewriter.getI32IntegerAttr(align));
-    };
-    SmallVector<Value> vectorOperands = {
-        adaptor.getIndexVec(), adaptor.getMask(), adaptor.getPassThru()};
-    return LLVM::detail::handleMultidimensionalVectors(
-        gather, vectorOperands, *getTypeConverter(), callback, rewriter);
+    // Replace with the gather intrinsic.
+    rewriter.replaceOpWithNewOp<LLVM::masked_gather>(
+        gather, typeConverter->convertType(vType), ptrs, adaptor.getMask(),
+        adaptor.getPassThru(), rewriter.getI32IntegerAttr(align));
+    return success();
   }
 };
 

mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVMPass.cpp

@@ -81,6 +81,7 @@ void ConvertVectorToLLVMPass::runOnOperation() {
     populateVectorInsertExtractStridedSliceTransforms(patterns);
     populateVectorStepLoweringPatterns(patterns);
     populateVectorRankReducingFMAPattern(patterns);
+    populateVectorGatherLoweringPatterns(patterns);
     (void)applyPatternsGreedily(getOperation(), std::move(patterns));
   }
 

mlir/lib/Dialect/Vector/Transforms/LowerVectorGather.cpp

@@ -38,7 +38,7 @@ using namespace mlir;
 using namespace mlir::vector;
 
 namespace {
-/// Flattens 2 or more dimensional `vector.gather` ops by unrolling the
+/// Unrolls 2 or more dimensional `vector.gather` ops by unrolling the
 /// outermost dimension. For example:
 /// ```
 /// %g = vector.gather %base[%c0][%v], %mask, %pass_thru :
@@ -56,14 +56,14 @@ namespace {
 /// When applied exhaustively, this will produce a sequence of 1-d gather ops.
 ///
 /// Supports vector types with a fixed leading dimension.
-struct FlattenGather : OpRewritePattern<vector::GatherOp> {
+struct UnrollGather : OpRewritePattern<vector::GatherOp> {
   using OpRewritePattern::OpRewritePattern;
 
   LogicalResult matchAndRewrite(vector::GatherOp op,
                                 PatternRewriter &rewriter) const override {
     VectorType resultTy = op.getType();
     if (resultTy.getRank() < 2)
-      return rewriter.notifyMatchFailure(op, "already flat");
+      return rewriter.notifyMatchFailure(op, "already 1-D");
 
     // Unrolling doesn't take vscale into account. Pattern is disabled for
     // vectors with leading scalable dim(s).
@@ -107,7 +107,8 @@ struct FlattenGather : OpRewritePattern<vector::GatherOp> {
 /// ```mlir
 ///   %subview = memref.subview %M (...)
 ///     : memref<100x3xf32> to memref<100xf32, strided<[3]>>
-///   %gather = vector.gather %subview[%idxs] (...) : memref<100xf32, strided<[3]>>
+///   %gather = vector.gather %subview[%idxs] (...)
+///     : memref<100xf32, strided<[3]>>
 /// ```
 /// ==>
 /// ```mlir
@@ -269,6 +270,11 @@ struct Gather1DToConditionalLoads : OpRewritePattern<vector::GatherOp> {
 
 void mlir::vector::populateVectorGatherLoweringPatterns(
     RewritePatternSet &patterns, PatternBenefit benefit) {
-  patterns.add<FlattenGather, RemoveStrideFromGatherSource,
-               Gather1DToConditionalLoads>(patterns.getContext(), benefit);
+  patterns.add<UnrollGather>(patterns.getContext(), benefit);
+}
+
+void mlir::vector::populateVectorGatherToConditionalLoadPatterns(
+    RewritePatternSet &patterns, PatternBenefit benefit) {
+  patterns.add<RemoveStrideFromGatherSource, Gather1DToConditionalLoads>(
+      patterns.getContext(), benefit);
 }

mlir/test/Conversion/VectorToLLVM/vector-to-llvm-interface.mlir

@@ -2074,52 +2074,6 @@ func.func @gather_index_scalable(%arg0: memref<?xindex>, %arg1: vector<[3]xindex
 
 // -----
 
-func.func @gather_2d_from_1d(%arg0: memref<?xf32>, %arg1: vector<2x3xi32>, %arg2: vector<2x3xi1>, %arg3: vector<2x3xf32>) -> vector<2x3xf32> {
-  %0 = arith.constant 0: index
-  %1 = vector.gather %arg0[%0][%arg1], %arg2, %arg3 : memref<?xf32>, vector<2x3xi32>, vector<2x3xi1>, vector<2x3xf32> into vector<2x3xf32>
-  return %1 : vector<2x3xf32>
-}
-
-// CHECK-LABEL: func @gather_2d_from_1d
-// CHECK: %[[B:.*]] = llvm.getelementptr %{{.*}} : (!llvm.ptr, i64) -> !llvm.ptr, f32
-// CHECK: %[[I0:.*]] = llvm.extractvalue %{{.*}}[0] : !llvm.array<2 x vector<3xi32>>
-// CHECK: %[[M0:.*]] = llvm.extractvalue %{{.*}}[0] : !llvm.array<2 x vector<3xi1>>
-// CHECK: %[[S0:.*]] = llvm.extractvalue %{{.*}}[0] : !llvm.array<2 x vector<3xf32>>
-// CHECK: %[[P0:.*]] = llvm.getelementptr %[[B]][%[[I0]]] : (!llvm.ptr, vector<3xi32>) -> !llvm.vec<3 x ptr>, f32
-// CHECK: %[[G0:.*]] = llvm.intr.masked.gather %[[P0]], %[[M0]], %[[S0]] {alignment = 4 : i32} : (!llvm.vec<3 x ptr>, vector<3xi1>, vector<3xf32>) -> vector<3xf32>
-// CHECK: %{{.*}} = llvm.insertvalue %[[G0]], %{{.*}}[0] : !llvm.array<2 x vector<3xf32>>
-// CHECK: %[[I1:.*]] = llvm.extractvalue %{{.*}}[1] : !llvm.array<2 x vector<3xi32>>
-// CHECK: %[[M1:.*]] = llvm.extractvalue %{{.*}}[1] : !llvm.array<2 x vector<3xi1>>
-// CHECK: %[[S1:.*]] = llvm.extractvalue %{{.*}}[1] : !llvm.array<2 x vector<3xf32>>
-// CHECK: %[[P1:.*]] = llvm.getelementptr %[[B]][%[[I1]]] : (!llvm.ptr, vector<3xi32>) -> !llvm.vec<3 x ptr>, f32
-// CHECK: %[[G1:.*]] = llvm.intr.masked.gather %[[P1]], %[[M1]], %[[S1]] {alignment = 4 : i32} : (!llvm.vec<3 x ptr>, vector<3xi1>, vector<3xf32>) -> vector<3xf32>
-// CHECK: %{{.*}} = llvm.insertvalue %[[G1]], %{{.*}}[1] : !llvm.array<2 x vector<3xf32>>
-
-// -----
-
-func.func @gather_2d_from_1d_scalable(%arg0: memref<?xf32>, %arg1: vector<2x[3]xi32>, %arg2: vector<2x[3]xi1>, %arg3: vector<2x[3]xf32>) -> vector<2x[3]xf32> {
-  %0 = arith.constant 0: index
-  %1 = vector.gather %arg0[%0][%arg1], %arg2, %arg3 : memref<?xf32>, vector<2x[3]xi32>, vector<2x[3]xi1>, vector<2x[3]xf32> into vector<2x[3]xf32>
-  return %1 : vector<2x[3]xf32>
-}
-
-// CHECK-LABEL: func @gather_2d_from_1d_scalable
-// CHECK: %[[B:.*]] = llvm.getelementptr %{{.*}} : (!llvm.ptr, i64) -> !llvm.ptr, f32
-// CHECK: %[[I0:.*]] = llvm.extractvalue %{{.*}}[0] : !llvm.array<2 x vector<[3]xi32>>
-// CHECK: %[[M0:.*]] = llvm.extractvalue %{{.*}}[0] : !llvm.array<2 x vector<[3]xi1>>
-// CHECK: %[[S0:.*]] = llvm.extractvalue %{{.*}}[0] : !llvm.array<2 x vector<[3]xf32>>
-// CHECK: %[[P0:.*]] = llvm.getelementptr %[[B]][%[[I0]]] : (!llvm.ptr, vector<[3]xi32>) -> !llvm.vec<? x 3 x ptr>, f32
-// CHECK: %[[G0:.*]] = llvm.intr.masked.gather %[[P0]], %[[M0]], %[[S0]] {alignment = 4 : i32} : (!llvm.vec<? x 3 x ptr>, vector<[3]xi1>, vector<[3]xf32>) -> vector<[3]xf32>
-// CHECK: %{{.*}} = llvm.insertvalue %[[G0]], %{{.*}}[0] : !llvm.array<2 x vector<[3]xf32>>
-// CHECK: %[[I1:.*]] = llvm.extractvalue %{{.*}}[1] : !llvm.array<2 x vector<[3]xi32>>
-// CHECK: %[[M1:.*]] = llvm.extractvalue %{{.*}}[1] : !llvm.array<2 x vector<[3]xi1>>
-// CHECK: %[[S1:.*]] = llvm.extractvalue %{{.*}}[1] : !llvm.array<2 x vector<[3]xf32>>
-// CHECK: %[[P1:.*]] = llvm.getelementptr %[[B]][%[[I1]]] : (!llvm.ptr, vector<[3]xi32>) -> !llvm.vec<? x 3 x ptr>, f32
-// CHECK: %[[G1:.*]] = llvm.intr.masked.gather %[[P1]], %[[M1]], %[[S1]] {alignment = 4 : i32} : (!llvm.vec<? x 3 x ptr>, vector<[3]xi1>, vector<[3]xf32>) -> vector<[3]xf32>
-// CHECK: %{{.*}} = llvm.insertvalue %[[G1]], %{{.*}}[1] : !llvm.array<2 x vector<[3]xf32>>
-
-// -----
-
 
 func.func @gather_1d_from_2d(%arg0: memref<4x4xf32>, %arg1: vector<4xi32>, %arg2: vector<4xi1>, %arg3: vector<4xf32>) -> vector<4xf32> {
   %0 = arith.constant 3 : index

mlir/test/Conversion/VectorToLLVM/vector-to-llvm.mlir

@@ -1663,7 +1663,7 @@ func.func @flat_transpose(%arg0: vector<16xf32>) -> vector<16xf32> {
 
 func.func @gather_with_mask(%arg0: memref<?xf32>, %arg1: vector<2x3xi32>, %arg2: vector<2x3xf32>) -> vector<2x3xf32> {
   %0 = arith.constant 0: index
-  %1 = vector.constant_mask [1, 2] : vector<2x3xi1>
+  %1 = vector.constant_mask [2, 2] : vector<2x3xi1>
   %2 = vector.gather %arg0[%0][%arg1], %1, %arg2 : memref<?xf32>, vector<2x3xi32>, vector<2x3xi1>, vector<2x3xf32> into vector<2x3xf32>
   return %2 : vector<2x3xf32>
 }
@@ -1677,9 +1677,9 @@ func.func @gather_with_mask(%arg0: memref<?xf32>, %arg1: vector<2x3xi32>, %arg2:
 func.func @gather_with_mask_scalable(%arg0: memref<?xf32>, %arg1: vector<2x[3]xi32>, %arg2: vector<2x[3]xf32>) -> vector<2x[3]xf32> {
   %0 = arith.constant 0: index
   // vector.constant_mask only supports 'none set' or 'all set' scalable
-  // dimensions, hence [1, 3] rather than [1, 2] as in the example for fixed
+  // dimensions, hence [2, 3] rather than [2, 2] as in the example for fixed
   // width vectors above.
-  %1 = vector.constant_mask [1, 3] : vector<2x[3]xi1>
+  %1 = vector.constant_mask [2, 3] : vector<2x[3]xi1>
   %2 = vector.gather %arg0[%0][%arg1], %1, %arg2 : memref<?xf32>, vector<2x[3]xi32>, vector<2x[3]xi1>, vector<2x[3]xf32> into vector<2x[3]xf32>
   return %2 : vector<2x[3]xf32>
 }

mlir/test/lib/Dialect/Vector/TestVectorTransforms.cpp

@@ -782,6 +782,7 @@ struct TestVectorGatherLowering
   void runOnOperation() override {
     RewritePatternSet patterns(&getContext());
     populateVectorGatherLoweringPatterns(patterns);
+    populateVectorGatherToConditionalLoadPatterns(patterns);
     (void)applyPatternsGreedily(getOperation(), std::move(patterns));
   }
 };