[Mlir-commits] [mlir] [mlir][vector] `broadcast(poison)->poison` and `transpose(poison)->poison` (PR #135435)
llvmlistbot at llvm.org
llvmlistbot at llvm.org
Fri Apr 11 13:32:48 PDT 2025
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-mlir
Author: James Newling (newling)
<details>
<summary>Changes</summary>
This is a follow-up after https://github.com/llvm/llvm-project/pull/133988
1) The above added the canonicalizer `shape_cast(poison) -> poison`. This PR adds `broadcast(poison) -> poison` and `transpose(poison) -> poison`
2) [NFC] Reviewer @<!-- -->Groverkss noted that canonicalizers should always be folders where possible (https://github.com/llvm/llvm-project/pull/133988#discussion_r2031455311). This PR moves 2 canonicalizers to folders.
3) [NFC] added missing `-----` between tests in canonicalize.mlir
---
Full diff: https://github.com/llvm/llvm-project/pull/135435.diff
2 Files Affected:
- (modified) mlir/lib/Dialect/Vector/IR/VectorOps.cpp (+65-102)
- (modified) mlir/test/Dialect/Vector/canonicalize.mlir (+40)
``````````diff
diff --git a/mlir/lib/Dialect/Vector/IR/VectorOps.cpp b/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
index 5324e38fa7d25..71c0ccf8fb1ca 100644
--- a/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
+++ b/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
@@ -2590,6 +2590,8 @@ OpFoldResult BroadcastOp::fold(FoldAdaptor adaptor) {
}
if (auto attr = llvm::dyn_cast<SplatElementsAttr>(adaptor.getSource()))
return DenseElementsAttr::get(vectorType, attr.getSplatValue<Attribute>());
+ if (llvm::dyn_cast<ub::PoisonAttr>(adaptor.getSource()))
+ return ub::PoisonAttr::get(getContext());
return {};
}
@@ -3717,6 +3719,59 @@ OpFoldResult ExtractStridedSliceOp::fold(FoldAdaptor adaptor) {
return getVector();
if (succeeded(foldExtractStridedOpFromInsertChain(*this)))
return getResult();
+
+ Attribute foldInput = adaptor.getVector();
+ if (!foldInput) {
+ return {};
+ }
+
+ // rewrite : ExtractStridedSliceOp(splat ConstantOp) -> ConstantOp.
+ if (auto splat = llvm::dyn_cast<SplatElementsAttr>(foldInput))
+ DenseElementsAttr::get(getType(), splat.getSplatValue<Attribute>());
+
+ // rewrite ExtractStridedSliceOp(non-splat ConstantOp) -> ConstantOp.
+ if (auto dense = llvm::dyn_cast<DenseElementsAttr>(foldInput)) {
+ // TODO: Handle non-unit strides when they become available.
+ if (hasNonUnitStrides())
+ return {};
+
+ Value sourceVector = getVector();
+ auto sourceVecTy = llvm::cast<VectorType>(sourceVector.getType());
+ ArrayRef<int64_t> sourceShape = sourceVecTy.getShape();
+ SmallVector<int64_t, 4> sourceStrides = computeStrides(sourceShape);
+
+ VectorType sliceVecTy = getType();
+ ArrayRef<int64_t> sliceShape = sliceVecTy.getShape();
+ int64_t sliceRank = sliceVecTy.getRank();
+
+ // Expand offsets and sizes to match the vector rank.
+ SmallVector<int64_t, 4> offsets(sliceRank, 0);
+ copy(getI64SubArray(getOffsets()), offsets.begin());
+
+ SmallVector<int64_t, 4> sizes(sourceShape);
+ copy(getI64SubArray(getSizes()), sizes.begin());
+
+ // Calculate the slice elements by enumerating all slice positions and
+ // linearizing them. The enumeration order is lexicographic which yields a
+ // sequence of monotonically increasing linearized position indices.
+ auto denseValuesBegin = dense.value_begin<Attribute>();
+ SmallVector<Attribute> sliceValues;
+ sliceValues.reserve(sliceVecTy.getNumElements());
+ SmallVector<int64_t> currSlicePosition(offsets.begin(), offsets.end());
+ do {
+ int64_t linearizedPosition = linearize(currSlicePosition, sourceStrides);
+ assert(linearizedPosition < sourceVecTy.getNumElements() &&
+ "Invalid index");
+ sliceValues.push_back(*(denseValuesBegin + linearizedPosition));
+ } while (
+ succeeded(incSlicePosition(currSlicePosition, sliceShape, offsets)));
+
+ assert(static_cast<int64_t>(sliceValues.size()) ==
+ sliceVecTy.getNumElements() &&
+ "Invalid number of slice elements");
+ return DenseElementsAttr::get(sliceVecTy, sliceValues);
+ }
+
return {};
}
@@ -3781,98 +3836,6 @@ class StridedSliceConstantMaskFolder final
}
};
-// Pattern to rewrite a ExtractStridedSliceOp(splat ConstantOp) -> ConstantOp.
-class StridedSliceSplatConstantFolder final
- : public OpRewritePattern<ExtractStridedSliceOp> {
-public:
- using OpRewritePattern::OpRewritePattern;
-
- LogicalResult matchAndRewrite(ExtractStridedSliceOp extractStridedSliceOp,
- PatternRewriter &rewriter) const override {
- // Return if 'ExtractStridedSliceOp' operand is not defined by a splat
- // ConstantOp.
- Value sourceVector = extractStridedSliceOp.getVector();
- Attribute vectorCst;
- if (!matchPattern(sourceVector, m_Constant(&vectorCst)))
- return failure();
-
- auto splat = llvm::dyn_cast<SplatElementsAttr>(vectorCst);
- if (!splat)
- return failure();
-
- auto newAttr = SplatElementsAttr::get(extractStridedSliceOp.getType(),
- splat.getSplatValue<Attribute>());
- rewriter.replaceOpWithNewOp<arith::ConstantOp>(extractStridedSliceOp,
- newAttr);
- return success();
- }
-};
-
-// Pattern to rewrite a ExtractStridedSliceOp(non-splat ConstantOp) ->
-// ConstantOp.
-class StridedSliceNonSplatConstantFolder final
- : public OpRewritePattern<ExtractStridedSliceOp> {
-public:
- using OpRewritePattern::OpRewritePattern;
-
- LogicalResult matchAndRewrite(ExtractStridedSliceOp extractStridedSliceOp,
- PatternRewriter &rewriter) const override {
- // Return if 'ExtractStridedSliceOp' operand is not defined by a non-splat
- // ConstantOp.
- Value sourceVector = extractStridedSliceOp.getVector();
- Attribute vectorCst;
- if (!matchPattern(sourceVector, m_Constant(&vectorCst)))
- return failure();
-
- // The splat case is handled by `StridedSliceSplatConstantFolder`.
- auto dense = llvm::dyn_cast<DenseElementsAttr>(vectorCst);
- if (!dense || dense.isSplat())
- return failure();
-
- // TODO: Handle non-unit strides when they become available.
- if (extractStridedSliceOp.hasNonUnitStrides())
- return failure();
-
- auto sourceVecTy = llvm::cast<VectorType>(sourceVector.getType());
- ArrayRef<int64_t> sourceShape = sourceVecTy.getShape();
- SmallVector<int64_t, 4> sourceStrides = computeStrides(sourceShape);
-
- VectorType sliceVecTy = extractStridedSliceOp.getType();
- ArrayRef<int64_t> sliceShape = sliceVecTy.getShape();
- int64_t sliceRank = sliceVecTy.getRank();
-
- // Expand offsets and sizes to match the vector rank.
- SmallVector<int64_t, 4> offsets(sliceRank, 0);
- copy(getI64SubArray(extractStridedSliceOp.getOffsets()), offsets.begin());
-
- SmallVector<int64_t, 4> sizes(sourceShape);
- copy(getI64SubArray(extractStridedSliceOp.getSizes()), sizes.begin());
-
- // Calculate the slice elements by enumerating all slice positions and
- // linearizing them. The enumeration order is lexicographic which yields a
- // sequence of monotonically increasing linearized position indices.
- auto denseValuesBegin = dense.value_begin<Attribute>();
- SmallVector<Attribute> sliceValues;
- sliceValues.reserve(sliceVecTy.getNumElements());
- SmallVector<int64_t> currSlicePosition(offsets.begin(), offsets.end());
- do {
- int64_t linearizedPosition = linearize(currSlicePosition, sourceStrides);
- assert(linearizedPosition < sourceVecTy.getNumElements() &&
- "Invalid index");
- sliceValues.push_back(*(denseValuesBegin + linearizedPosition));
- } while (
- succeeded(incSlicePosition(currSlicePosition, sliceShape, offsets)));
-
- assert(static_cast<int64_t>(sliceValues.size()) ==
- sliceVecTy.getNumElements() &&
- "Invalid number of slice elements");
- auto newAttr = DenseElementsAttr::get(sliceVecTy, sliceValues);
- rewriter.replaceOpWithNewOp<arith::ConstantOp>(extractStridedSliceOp,
- newAttr);
- return success();
- }
-};
-
// Pattern to rewrite an ExtractStridedSliceOp(BroadcastOp) to
// BroadcastOp(ExtractStrideSliceOp).
class StridedSliceBroadcast final
@@ -4016,8 +3979,7 @@ void ExtractStridedSliceOp::getCanonicalizationPatterns(
RewritePatternSet &results, MLIRContext *context) {
// Pattern to rewrite a ExtractStridedSliceOp(ConstantMaskOp) ->
// ConstantMaskOp and ExtractStridedSliceOp(ConstantOp) -> ConstantOp.
- results.add<StridedSliceConstantMaskFolder, StridedSliceSplatConstantFolder,
- StridedSliceNonSplatConstantFolder, StridedSliceBroadcast,
+ results.add<StridedSliceConstantMaskFolder, StridedSliceBroadcast,
StridedSliceSplat, ContiguousExtractStridedSliceToExtract>(
context);
}
@@ -5654,10 +5616,8 @@ OpFoldResult ShapeCastOp::fold(FoldAdaptor adaptor) {
// shape_cast(constant) -> constant
if (auto splatAttr =
- llvm::dyn_cast_if_present<SplatElementsAttr>(adaptor.getSource())) {
- return DenseElementsAttr::get(resultType,
- splatAttr.getSplatValue<Attribute>());
- }
+ llvm::dyn_cast_if_present<SplatElementsAttr>(adaptor.getSource()))
+ return splatAttr.reshape(getType());
// shape_cast(poison) -> poison
if (llvm::dyn_cast_if_present<ub::PoisonAttr>(adaptor.getSource())) {
@@ -6001,10 +5961,13 @@ void vector::TransposeOp::build(OpBuilder &builder, OperationState &result,
OpFoldResult vector::TransposeOp::fold(FoldAdaptor adaptor) {
// Eliminate splat constant transpose ops.
- if (auto attr =
- llvm::dyn_cast_if_present<DenseElementsAttr>(adaptor.getVector()))
- if (attr.isSplat())
- return attr.reshape(getResultVectorType());
+ if (auto splat =
+ llvm::dyn_cast_if_present<SplatElementsAttr>(adaptor.getVector()))
+ return splat.reshape(getResultVectorType());
+
+ // Eliminate poison transpose ops.
+ if (llvm::dyn_cast_if_present<ub::PoisonAttr>(adaptor.getVector()))
+ return ub::PoisonAttr::get(getContext());
// Eliminate identity transpose ops. This happens when the dimensions of the
// input vector remain in their original order after the transpose operation.
diff --git a/mlir/test/Dialect/Vector/canonicalize.mlir b/mlir/test/Dialect/Vector/canonicalize.mlir
index a6d82b85777b0..73021a194e5bc 100644
--- a/mlir/test/Dialect/Vector/canonicalize.mlir
+++ b/mlir/test/Dialect/Vector/canonicalize.mlir
@@ -1121,6 +1121,8 @@ func.func @bitcast_folding(%I1: vector<4x8xf32>, %I2: vector<2xi32>) -> (vector<
return %0, %2 : vector<4x8xf32>, vector<2xi32>
}
+// -----
+
// CHECK-LABEL: func @bitcast_f16_to_f32
// bit pattern: 0x40004000
// CHECK-DAG: %[[CST1:.+]] = arith.constant dense<2.00390625> : vector<4xf32>
@@ -1135,6 +1137,8 @@ func.func @bitcast_f16_to_f32() -> (vector<4xf32>, vector<4xf32>) {
return %cast0, %cast1: vector<4xf32>, vector<4xf32>
}
+// -----
+
// CHECK-LABEL: func @bitcast_i8_to_i32
// bit pattern: 0xA0A0A0A0
// CHECK-DAG: %[[CST1:.+]] = arith.constant dense<-1600085856> : vector<4xi32>
@@ -1176,6 +1180,28 @@ func.func @broadcast_folding2() -> vector<4x16xi32> {
// -----
+// CHECK-LABEL: broadcast_poison
+// CHECK: %[[POISON:.*]] = ub.poison : vector<4x6xi8>
+// CHECK: return %[[POISON]] : vector<4x6xi8>
+func.func @broadcast_poison() -> vector<4x6xi8> {
+ %poison = ub.poison : vector<6xi8>
+ %broadcast = vector.broadcast %poison : vector<6xi8> to vector<4x6xi8>
+ return %broadcast : vector<4x6xi8>
+}
+
+// -----
+
+// CHECK-LABEL: broadcast_splat_constant
+// CHECK: %[[CONST:.*]] = arith.constant dense<1> : vector<4x6xi8>
+// CHECK: return %[[CONST]] : vector<4x6xi8>
+func.func @broadcast_splat_constant() -> vector<4x6xi8> {
+ %cst = arith.constant dense<1> : vector<6xi8>
+ %broadcast = vector.broadcast %cst : vector<6xi8> to vector<4x6xi8>
+ return %broadcast : vector<4x6xi8>
+}
+
+// -----
+
// CHECK-LABEL: @fold_consecutive_broadcasts(
// CHECK-SAME: %[[ARG0:.*]]: i32
// CHECK: %[[RESULT:.*]] = vector.broadcast %[[ARG0]] : i32 to vector<4x16xi32>
@@ -1710,6 +1736,7 @@ func.func @vector_multi_reduction_unit_dimensions(%source: vector<5x1x4x1x20xf32
}
// -----
+
// CHECK-LABEL: func.func @vector_multi_reduction_scalable(
// CHECK-SAME: %[[VAL_0:.*]]: vector<1x[4]x1xf32>,
// CHECK-SAME: %[[VAL_1:.*]]: vector<1x[4]xf32>,
@@ -2251,6 +2278,8 @@ func.func @transpose_splat_constant() -> vector<8x4xf32> {
return %0 : vector<8x4xf32>
}
+// -----
+
// CHECK-LABEL: func @transpose_splat2(
// CHECK-SAME: %[[VAL_0:.*]]: f32) -> vector<3x4xf32> {
// CHECK: %[[VAL_1:.*]] = vector.splat %[[VAL_0]] : vector<3x4xf32>
@@ -2264,6 +2293,17 @@ func.func @transpose_splat2(%arg : f32) -> vector<3x4xf32> {
// -----
+// CHECK-LABEL: transpose_poison
+// CHECK: %[[POISON:.*]] = ub.poison : vector<4x6xi8>
+// CHECK: return %[[POISON]] : vector<4x6xi8>
+func.func @transpose_poison() -> vector<4x6xi8> {
+ %poison = ub.poison : vector<6x4xi8>
+ %transpose = vector.transpose %poison, [1, 0] : vector<6x4xi8> to vector<4x6xi8>
+ return %transpose : vector<4x6xi8>
+}
+
+// -----
+
// CHECK-LABEL: func.func @insert_1d_constant
// CHECK-DAG: %[[ACST:.*]] = arith.constant dense<[9, 1, 2]> : vector<3xi32>
// CHECK-DAG: %[[BCST:.*]] = arith.constant dense<[0, 9, 2]> : vector<3xi32>
``````````
</details>
https://github.com/llvm/llvm-project/pull/135435
More information about the Mlir-commits
mailing list