[Mlir-commits] [mlir] [MLIR][Vector] Add Lowering for vector.step (PR #113655)
llvmlistbot at llvm.org
llvmlistbot at llvm.org
Thu Oct 24 22:53:38 PDT 2024
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-mlir-vector
Author: Manupa Karunaratne (manupak)
<details>
<summary>Changes</summary>
Currently, the lowering for vector.step lives
under a folder. This is not ideal if we want
to do transformation on it and defer the
materizaliztion of the constants much later.
This commits adds a rewrite pattern + transform
op to do this instead. Thus enabling more control
on the lowering.
---
Patch is 25.88 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/113655.diff
13 Files Affected:
- (modified) mlir/include/mlir/Dialect/Vector/IR/VectorOps.td (-1)
- (modified) mlir/include/mlir/Dialect/Vector/TransformOps/VectorTransformOps.td (+9)
- (modified) mlir/include/mlir/Dialect/Vector/Transforms/LoweringPatterns.h (+7)
- (modified) mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVM.cpp (+1)
- (modified) mlir/lib/Dialect/SparseTensor/Transforms/SparseVectorization.cpp (+2)
- (modified) mlir/lib/Dialect/Vector/IR/VectorOps.cpp (-14)
- (modified) mlir/lib/Dialect/Vector/TransformOps/VectorTransformOps.cpp (+5)
- (modified) mlir/lib/Dialect/Vector/Transforms/CMakeLists.txt (+1)
- (added) mlir/lib/Dialect/Vector/Transforms/LowerVectorStep.cpp (+64)
- (modified) mlir/test/Dialect/Linalg/vectorization-scalable.mlir (+1)
- (modified) mlir/test/Dialect/Linalg/vectorization-with-patterns.mlir (+15)
- (modified) mlir/test/Dialect/Linalg/vectorize-tensor-extract.mlir (+73-6)
- (modified) mlir/test/Dialect/Vector/canonicalize.mlir (-9)
``````````diff
diff --git a/mlir/include/mlir/Dialect/Vector/IR/VectorOps.td b/mlir/include/mlir/Dialect/Vector/IR/VectorOps.td
index c02b16ea931706..5e7b6659548203 100644
--- a/mlir/include/mlir/Dialect/Vector/IR/VectorOps.td
+++ b/mlir/include/mlir/Dialect/Vector/IR/VectorOps.td
@@ -2940,7 +2940,6 @@ def Vector_StepOp : Vector_Op<"step", [Pure]> {
%1 = vector.step : vector<[4]xindex> ; [0, 1, .., <vscale * 4 - 1>]
```
}];
- let hasFolder = 1;
let results = (outs VectorOfRankAndType<[1], [Index]>:$result);
let assemblyFormat = "attr-dict `:` type($result)";
}
diff --git a/mlir/include/mlir/Dialect/Vector/TransformOps/VectorTransformOps.td b/mlir/include/mlir/Dialect/Vector/TransformOps/VectorTransformOps.td
index c973eca0132a92..3262aa37a81877 100644
--- a/mlir/include/mlir/Dialect/Vector/TransformOps/VectorTransformOps.td
+++ b/mlir/include/mlir/Dialect/Vector/TransformOps/VectorTransformOps.td
@@ -453,4 +453,13 @@ def ApplyVectorReductionToContractPatternsOp : Op<Transform_Dialect,
let assemblyFormat = "attr-dict";
}
+def ApplyLowerStepToArithOps : Op<Transform_Dialect,
+ "apply_patterns.vector.step_to_arith",
+ [DeclareOpInterfaceMethods<PatternDescriptorOpInterface>]> {
+ let description = [{
+ Convert vector.step to arith if not using scalable vectors.
+ }];
+ let assemblyFormat = "attr-dict";
+}
+
#endif // VECTOR_TRANSFORM_OPS
diff --git a/mlir/include/mlir/Dialect/Vector/Transforms/LoweringPatterns.h b/mlir/include/mlir/Dialect/Vector/Transforms/LoweringPatterns.h
index 1976b8399c7f9c..27581443814322 100644
--- a/mlir/include/mlir/Dialect/Vector/Transforms/LoweringPatterns.h
+++ b/mlir/include/mlir/Dialect/Vector/Transforms/LoweringPatterns.h
@@ -235,6 +235,13 @@ void populateVectorTransferPermutationMapLoweringPatterns(
void populateVectorScanLoweringPatterns(RewritePatternSet &patterns,
PatternBenefit benefit = 1);
+/// Populate the pattern set with the following patterns:
+///
+/// [StepToArithOps]
+/// Convert vector.step op into arith ops if not scalable
+void populateVectorStepLoweringPatterns(RewritePatternSet &patterns,
+ PatternBenefit benefit = 1);
+
/// Populate the pattern set with the following patterns:
///
/// [FlattenGather]
diff --git a/mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVM.cpp b/mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVM.cpp
index 984af50a7b0a51..15a545dbb42180 100644
--- a/mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVM.cpp
+++ b/mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVM.cpp
@@ -1886,6 +1886,7 @@ void mlir::populateVectorToLLVMConversionPatterns(
MLIRContext *ctx = converter.getDialect()->getContext();
patterns.add<VectorFMAOpNDRewritePattern>(ctx);
populateVectorInsertExtractStridedSliceTransforms(patterns);
+ populateVectorStepLoweringPatterns(patterns);
patterns.add<VectorReductionOpConversion>(converter, reassociateFPReductions);
patterns.add<VectorCreateMaskOpRewritePattern>(ctx, force32BitVectorIndices);
patterns.add<VectorBitCastOpConversion, VectorShuffleOpConversion,
diff --git a/mlir/lib/Dialect/SparseTensor/Transforms/SparseVectorization.cpp b/mlir/lib/Dialect/SparseTensor/Transforms/SparseVectorization.cpp
index d1c95dabd88a5e..b2eca539194a87 100644
--- a/mlir/lib/Dialect/SparseTensor/Transforms/SparseVectorization.cpp
+++ b/mlir/lib/Dialect/SparseTensor/Transforms/SparseVectorization.cpp
@@ -27,6 +27,7 @@
#include "mlir/Dialect/SCF/IR/SCF.h"
#include "mlir/Dialect/SparseTensor/Transforms/Passes.h"
#include "mlir/Dialect/Vector/IR/VectorOps.h"
+#include "mlir/Dialect/Vector/Transforms/LoweringPatterns.h"
#include "mlir/IR/Matchers.h"
using namespace mlir;
@@ -664,6 +665,7 @@ void mlir::populateSparseVectorizationPatterns(RewritePatternSet &patterns,
bool enableVLAVectorization,
bool enableSIMDIndex32) {
assert(vectorLength > 0);
+ vector::populateVectorStepLoweringPatterns(patterns);
patterns.add<ForOpRewriter>(patterns.getContext(), vectorLength,
enableVLAVectorization, enableSIMDIndex32);
patterns.add<ReducChainRewriter<vector::InsertElementOp>,
diff --git a/mlir/lib/Dialect/Vector/IR/VectorOps.cpp b/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
index a2abe1619454f2..2daf3e8a29ff9c 100644
--- a/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
+++ b/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
@@ -6423,20 +6423,6 @@ OpFoldResult SplatOp::fold(FoldAdaptor adaptor) {
return SplatElementsAttr::get(getType(), {constOperand});
}
-//===----------------------------------------------------------------------===//
-// StepOp
-//===----------------------------------------------------------------------===//
-
-OpFoldResult StepOp::fold(FoldAdaptor adaptor) {
- auto resultType = cast<VectorType>(getType());
- if (resultType.isScalable())
- return nullptr;
- SmallVector<APInt> indices;
- for (unsigned i = 0; i < resultType.getNumElements(); i++)
- indices.push_back(APInt(/*width=*/64, i));
- return DenseElementsAttr::get(resultType, indices);
-}
-
//===----------------------------------------------------------------------===//
// WarpExecuteOnLane0Op
//===----------------------------------------------------------------------===//
diff --git a/mlir/lib/Dialect/Vector/TransformOps/VectorTransformOps.cpp b/mlir/lib/Dialect/Vector/TransformOps/VectorTransformOps.cpp
index 241e83e234d621..b6f49b85c6205a 100644
--- a/mlir/lib/Dialect/Vector/TransformOps/VectorTransformOps.cpp
+++ b/mlir/lib/Dialect/Vector/TransformOps/VectorTransformOps.cpp
@@ -207,6 +207,11 @@ void transform::ApplyTransferToScfPatternsOp::populatePatterns(
populateVectorToSCFConversionPatterns(patterns, vectorTransferToSCFOptions);
}
+void transform::ApplyLowerStepToArithOps::populatePatterns(
+ RewritePatternSet &patterns) {
+ populateVectorStepLoweringPatterns(patterns);
+}
+
//===----------------------------------------------------------------------===//
// Transform op registration
//===----------------------------------------------------------------------===//
diff --git a/mlir/lib/Dialect/Vector/Transforms/CMakeLists.txt b/mlir/lib/Dialect/Vector/Transforms/CMakeLists.txt
index b7e8724c3c2582..9a3bd5d4593d63 100644
--- a/mlir/lib/Dialect/Vector/Transforms/CMakeLists.txt
+++ b/mlir/lib/Dialect/Vector/Transforms/CMakeLists.txt
@@ -9,6 +9,7 @@ add_mlir_dialect_library(MLIRVectorTransforms
LowerVectorMultiReduction.cpp
LowerVectorScan.cpp
LowerVectorShapeCast.cpp
+ LowerVectorStep.cpp
LowerVectorTransfer.cpp
LowerVectorTranspose.cpp
SubsetOpInterfaceImpl.cpp
diff --git a/mlir/lib/Dialect/Vector/Transforms/LowerVectorStep.cpp b/mlir/lib/Dialect/Vector/Transforms/LowerVectorStep.cpp
new file mode 100644
index 00000000000000..fb7a516e4b41c3
--- /dev/null
+++ b/mlir/lib/Dialect/Vector/Transforms/LowerVectorStep.cpp
@@ -0,0 +1,64 @@
+//===- LowerVectorStep.cpp - Lower 'vector.step' operation ----------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements target-independent rewrites and utilities to lower the
+// 'vector.step' operation.
+//
+//===----------------------------------------------------------------------===//
+
+#include "mlir/Dialect/Affine/IR/AffineOps.h"
+#include "mlir/Dialect/Arith/IR/Arith.h"
+#include "mlir/Dialect/Arith/Utils/Utils.h"
+#include "mlir/Dialect/Linalg/IR/Linalg.h"
+#include "mlir/Dialect/MemRef/IR/MemRef.h"
+#include "mlir/Dialect/SCF/IR/SCF.h"
+#include "mlir/Dialect/Tensor/IR/Tensor.h"
+#include "mlir/Dialect/Utils/IndexingUtils.h"
+#include "mlir/Dialect/Utils/StructuredOpsUtils.h"
+#include "mlir/Dialect/Vector/IR/VectorOps.h"
+#include "mlir/Dialect/Vector/Transforms/LoweringPatterns.h"
+#include "mlir/Dialect/Vector/Utils/VectorUtils.h"
+#include "mlir/IR/BuiltinAttributeInterfaces.h"
+#include "mlir/IR/BuiltinTypes.h"
+#include "mlir/IR/ImplicitLocOpBuilder.h"
+#include "mlir/IR/Location.h"
+#include "mlir/IR/Matchers.h"
+#include "mlir/IR/PatternMatch.h"
+#include "mlir/IR/TypeUtilities.h"
+#include "mlir/Interfaces/VectorInterfaces.h"
+
+#define DEBUG_TYPE "vector-step-lowering"
+
+using namespace mlir;
+using namespace mlir::vector;
+
+namespace {
+
+struct StepToArithOps : public OpRewritePattern<vector::StepOp> {
+ using OpRewritePattern::OpRewritePattern;
+
+ LogicalResult matchAndRewrite(vector::StepOp stepOp,
+ PatternRewriter &rewriter) const override {
+ auto resultType = cast<VectorType>(stepOp.getType());
+ if (!resultType.isScalable()) {
+ SmallVector<APInt> indices;
+ for (unsigned i = 0; i < resultType.getNumElements(); i++)
+ indices.push_back(APInt(/*width=*/64, i));
+ rewriter.replaceOpWithNewOp<arith::ConstantOp>(
+ stepOp, DenseElementsAttr::get(resultType, indices));
+ return success();
+ }
+ return failure();
+ }
+};
+} // namespace
+
+void mlir::vector::populateVectorStepLoweringPatterns(
+ RewritePatternSet &patterns, PatternBenefit benefit) {
+ patterns.add<StepToArithOps>(patterns.getContext(), benefit);
+}
diff --git a/mlir/test/Dialect/Linalg/vectorization-scalable.mlir b/mlir/test/Dialect/Linalg/vectorization-scalable.mlir
index c3a30e3ee209e8..96866885df4d49 100644
--- a/mlir/test/Dialect/Linalg/vectorization-scalable.mlir
+++ b/mlir/test/Dialect/Linalg/vectorization-scalable.mlir
@@ -180,6 +180,7 @@ module attributes {transform.with_named_sequence} {
%func = transform.structured.match ops{["func.func"]} in %arg1
: (!transform.any_op) -> !transform.any_op
transform.apply_patterns to %func {
+ transform.apply_patterns.vector.step_to_arith
transform.apply_patterns.canonicalization
transform.apply_patterns.linalg.tiling_canonicalization
} : !transform.any_op
diff --git a/mlir/test/Dialect/Linalg/vectorization-with-patterns.mlir b/mlir/test/Dialect/Linalg/vectorization-with-patterns.mlir
index 1c6a786bfa436d..8a3dbe6765ebd8 100644
--- a/mlir/test/Dialect/Linalg/vectorization-with-patterns.mlir
+++ b/mlir/test/Dialect/Linalg/vectorization-with-patterns.mlir
@@ -346,6 +346,11 @@ module attributes {transform.with_named_sequence} {
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
%2 = transform.structured.vectorize_children_and_apply_patterns %1 { vectorize_nd_extract } : (!transform.any_op) -> !transform.any_op
+ %func = transform.structured.match ops{["func.func"]} in %arg1
+ : (!transform.any_op) -> !transform.any_op
+ transform.apply_patterns to %func {
+ transform.apply_patterns.vector.step_to_arith
+ } : !transform.any_op
transform.yield
}
}
@@ -474,6 +479,11 @@ module attributes {transform.with_named_sequence} {
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
%2 = transform.structured.vectorize_children_and_apply_patterns %1 : (!transform.any_op) -> !transform.any_op
+ %func = transform.structured.match ops{["func.func"]} in %arg1
+ : (!transform.any_op) -> !transform.any_op
+ transform.apply_patterns to %func {
+ transform.apply_patterns.vector.step_to_arith
+ } : !transform.any_op
transform.yield
}
}
@@ -505,6 +515,11 @@ module attributes {transform.with_named_sequence} {
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
%2 = transform.structured.vectorize_children_and_apply_patterns %1 : (!transform.any_op) -> !transform.any_op
+ %func = transform.structured.match ops{["func.func"]} in %arg1
+ : (!transform.any_op) -> !transform.any_op
+ transform.apply_patterns to %func {
+ transform.apply_patterns.vector.step_to_arith
+ } : !transform.any_op
transform.yield
}
}
diff --git a/mlir/test/Dialect/Linalg/vectorize-tensor-extract.mlir b/mlir/test/Dialect/Linalg/vectorize-tensor-extract.mlir
index e611a8e22ee23f..2cf7804264ef8c 100644
--- a/mlir/test/Dialect/Linalg/vectorize-tensor-extract.mlir
+++ b/mlir/test/Dialect/Linalg/vectorize-tensor-extract.mlir
@@ -32,6 +32,11 @@ module attributes {transform.with_named_sequence} {
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
%2 = transform.structured.vectorize_children_and_apply_patterns %1 : (!transform.any_op) -> !transform.any_op
+ %func = transform.structured.match ops{["func.func"]} in %arg1
+ : (!transform.any_op) -> !transform.any_op
+ transform.apply_patterns to %func {
+ transform.apply_patterns.vector.step_to_arith
+ } : !transform.any_op
transform.yield
}
}
@@ -172,6 +177,11 @@ module attributes {transform.with_named_sequence} {
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
%2 = transform.structured.vectorize_children_and_apply_patterns %1 { vectorize_nd_extract } : (!transform.any_op) -> !transform.any_op
+ %func = transform.structured.match ops{["func.func"]} in %arg1
+ : (!transform.any_op) -> !transform.any_op
+ transform.apply_patterns to %func {
+ transform.apply_patterns.vector.step_to_arith
+ } : !transform.any_op
transform.yield
}
}
@@ -207,8 +217,9 @@ func.func @vectorize_nd_tensor_extract_transfer_read_complex(%6: tensor<45x80x16
// CHECK-DAG: %[[VAL_8:.*]] = arith.constant 0.000000e+00 : f32
// CHECK-DAG: %[[VAL_9:.*]] = arith.constant 0 : index
// CHECK-DAG: %[[VAL_10:.*]] = arith.constant 79 : index
-// CHECK: %[[VAL_11:.*]] = arith.addi %[[VAL_1]], %[[VAL_2]] : index
-// CHECK: %[[VAL_12:.*]] = vector.broadcast %[[VAL_11]] : index to vector<1x4xindex>
+// CHECK-DAG: %[[VAL_1_BCAST:.*]] = vector.broadcast %[[VAL_1]] : index to vector<1x4xindex>
+// CHECK-DAG: %[[VAL_2_BCAST:.*]] = vector.broadcast %[[VAL_2]] : index to vector<1x4xindex>
+// CHECK: %[[VAL_12:.*]] = arith.addi %[[VAL_1_BCAST]], %[[VAL_2_BCAST]] : vector<1x4xindex>
// CHECK: %[[VAL_13:.*]] = vector.broadcast %[[VAL_3]] : index to vector<4xindex>
// CHECK: %[[VAL_14:.*]] = arith.addi %[[VAL_13]], %[[VAL_6]] : vector<4xindex>
// CHECK: %[[VAL_15:.*]] = vector.broadcast %[[VAL_4]] : index to vector<4xindex>
@@ -226,6 +237,11 @@ module attributes {transform.with_named_sequence} {
%0 = transform.structured.match ops{["linalg.generic"]} in %arg1 : (!transform.any_op) -> !transform.any_op
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
%2 = transform.structured.vectorize_children_and_apply_patterns %1 { vectorize_nd_extract } : (!transform.any_op) -> !transform.any_op
+ %func = transform.structured.match ops{["func.func"]} in %arg1
+ : (!transform.any_op) -> !transform.any_op
+ transform.apply_patterns to %func {
+ transform.apply_patterns.vector.step_to_arith
+ } : !transform.any_op
transform.yield
}
}
@@ -306,6 +322,11 @@ module attributes {transform.with_named_sequence} {
%0 = transform.structured.match ops{["linalg.generic"]} in %arg0 : (!transform.any_op) -> !transform.any_op
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
%2 = transform.structured.vectorize_children_and_apply_patterns %1 {vectorize_nd_extract} : (!transform.any_op) -> !transform.any_op
+ %func = transform.structured.match ops{["func.func"]} in %arg0
+ : (!transform.any_op) -> !transform.any_op
+ transform.apply_patterns to %func {
+ transform.apply_patterns.vector.step_to_arith
+ } : !transform.any_op
transform.yield
}
}
@@ -321,8 +342,9 @@ module attributes {transform.with_named_sequence} {
// CHECK-DAG: %[[CST_3:.*]] = arith.constant dense<[0, 1, 2, 3, 4, 5, 6, 7]> : vector<8xindex>
// CHECK: %[[EMPTY:.*]] = tensor.empty() : tensor<8x1xf32>
// CHECK: %[[B1:.*]] = vector.broadcast %[[CST_3]] : vector<8xindex> to vector<1x8xindex>
-// CHECK: %[[ADDI_ARG1:.*]] = arith.addi %[[ARG1]], %[[ARG1]] : index
-// CHECK: %[[B2:.*]] = vector.broadcast %[[ADDI_ARG1]] : index to vector<1xindex>
+// CHECK: %[[ARG1_BCAST0:.*]] = vector.broadcast %arg1 : index to vector<1xindex>
+// CHECK: %[[ARG1_BCAST1:.*]] = vector.broadcast %arg1 : index to vector<1xindex>
+// CHECK: %[[B2:.*]] = arith.addi %[[ARG1_BCAST0]], %[[ARG1_BCAST1]] : vector<1xindex>
// CHECK: %[[MULI_1:.*]] = arith.muli %[[B1]], %[[CST_0]] : vector<1x8xindex>
// CHECK: %[[MULI_2:.*]] = arith.muli %[[MULI_1]], %[[CST]] : vector<1x8xindex>
// CHECK: %[[T:.*]] = vector.transpose %[[MULI_2]], [1, 0] : vector<1x8xindex> to vector<8x1xindex>
@@ -357,17 +379,22 @@ module attributes {transform.with_named_sequence} {
%0 = transform.structured.match ops{["linalg.generic"]} in %arg2 : (!transform.any_op) -> !transform.any_op
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
%2 = transform.structured.vectorize_children_and_apply_patterns %1 {vectorize_nd_extract} : (!transform.any_op) -> !transform.any_op
+ %func = transform.structured.match ops{["func.func"]} in %arg2
+ : (!transform.any_op) -> !transform.any_op
+ transform.apply_patterns to %func {
+ transform.apply_patterns.vector.step_to_arith
+ } : !transform.any_op
transform.yield
}
}
// CHECK-LABEL: func.func @index_from_output_column_vector_gather_load(
// CHECK-SAME: %[[SRC:.*]]: tensor<8x128xf32>) -> tensor<8x1xf32> {
+// CHECK: %[[IDX_VEC:.*]] = arith.constant dense<[0, 1, 2, 3, 4, 5, 6, 7]> : vector<8xindex>
// CHECK: %[[C128:.*]] = arith.constant dense<128> : vector<1x8xindex>
// CHECK: %[[C0:.*]] = arith.constant 0 : index
// CHECK: %[[PASS_THRU:.*]] = arith.constant dense<0.000000e+00> : vector<8x1xf32>
// CHECK: %[[MASK:.*]] = arith.constant dense<true> : vector<8x1xi1>
-// CHECK: %[[IDX_VEC:.*]] = arith.constant dense<[0, 1, 2, 3, 4, 5, 6, 7]> : vector<8xindex>
// CHECK: %[[OUT:.*]] = tensor.empty() : tensor<8x1xf32>
// CHECK: %[[B:.*]] = vector.broadcast %[[IDX_VEC]] : vector<8xindex> to vector<1x8xindex>
// CHECK: %[[MUL:.*]] = arith.muli %[[B]], %[[C128]] : vector<1x8xindex>
@@ -404,16 +431,21 @@ module attributes {transform.with_named_sequence} {
%0 = transform.structured.match ops{["linalg.generic"]} in %arg2 : (!transform.any_op) -> !transform.any_op
%1 = transform.get_parent_op %0 {isolated_from_above} : (!transform.any_op) -> !transform.any_op
%2 = transform.structured.vectorize_children_and_apply_patterns %1 {vectorize_nd_extract} : (!transform.any_op) -> !transform.any_op
+ %func = transform.structured.match ops{["func.func"]} in %arg2
+ : (!transform.any_op) -> !transform.any_op
+ transform.apply_patterns to %func {
+ transform.apply_patterns.vector.step_to_arith
+ } : !transform.any_op
transform.yield
}
}
// CHECK-LABEL: func.func @index_from_output_column_vector_contiguous_load(
// CHECK-SAME: %[[SRC:.*]]: tensor<8x128xf32>) -> tensor<8x1xf32> {
+// CHECK: %[[IDX_VEC:.*]] = arith.constant dense<[0, 1, 2, 3, 4, 5, 6, 7]> : vector<8xindex>
// CHECK: %[[C0:.*]] = arith.constant 0 : index
// CHECK: %[[PASS_THRU:.*]] = arith.constant dense<0.000000e+00> : vector<8x1xf32>
// CHECK: %[[MASK:.*]] = arith.constant dense<true> : vector<8x1xi1>
-// CHECK: %[[IDX_VEC:.*]] = arith.constant dense<[0,...
[truncated]
``````````
</details>
https://github.com/llvm/llvm-project/pull/113655
More information about the Mlir-commits
mailing list