Skip to content

[mlir][vector] Move extract_strided_slice canonicalization to folding #135676

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 4 commits into from
Apr 16, 2025
Merged

[mlir][vector] Move extract_strided_slice canonicalization to folding #135676

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
2e027d8
move canonicalizers to folders
newling Apr 14, 2025
c0cebd2
tidy
newling Apr 15, 2025
788f44c
factorize to function
newling Apr 15, 2025
fc62a9b
prefer static to anonymous namespace in this file
newling Apr 16, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
160 changes: 58 additions & 102 deletions mlir/lib/Dialect/Vector/IR/VectorOps.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3717,6 +3717,58 @@ OpFoldResult ExtractStridedSliceOp::fold(FoldAdaptor adaptor) {
return getVector();
if (succeeded(foldExtractStridedOpFromInsertChain(*this)))
return getResult();

// All subsequent successful folds require a constant input.
Attribute foldInput = adaptor.getVector();
if (!foldInput)
return {};

// ExtractStridedSliceOp(splat ConstantOp) -> ConstantOp.
if (auto splat = llvm::dyn_cast<SplatElementsAttr>(foldInput))
DenseElementsAttr::get(getType(), splat.getSplatValue<Attribute>());

// ExtractStridedSliceOp(non-splat ConstantOp) -> ConstantOp.
if (auto dense = llvm::dyn_cast<DenseElementsAttr>(foldInput)) {
// TODO: Handle non-unit strides when they become available.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: move this long folder to a function and follow the pattern:

if (succeeded(foldXYZ(...)))
  return ...;

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Will do.

if (hasNonUnitStrides())
return {};

VectorType sourceVecTy = getSourceVectorType();
ArrayRef<int64_t> sourceShape = sourceVecTy.getShape();
SmallVector<int64_t, 4> sourceStrides = computeStrides(sourceShape);

VectorType sliceVecTy = getType();
ArrayRef<int64_t> sliceShape = sliceVecTy.getShape();
int64_t rank = sliceVecTy.getRank();

// Expand offsets and sizes to match the vector rank.
SmallVector<int64_t, 4> offsets(rank, 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.
const 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 {};
}

Expand Down Expand Up @@ -3781,98 +3833,6 @@ 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
Expand Down Expand Up @@ -4016,8 +3976,7 @@ 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);
}
Expand Down Expand Up @@ -5657,10 +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()))
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: drop llvm::?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

grep -r " dyn_cast" lib/Dialect/Vector/IR/VectorOps.cpp | wc -l
26
grep -r "llvm::dyn_cast" lib/Dialect/Vector/IR/VectorOps.cpp | wc -l
49
🤷

return splatAttr.reshape(getType());

// shape_cast(poison) -> poison
if (llvm::dyn_cast_if_present<ub::PoisonAttr>(adaptor.getSource())) {
Expand Down Expand Up @@ -6004,10 +5961,9 @@ 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 identity transpose ops. This happens when the dimensions of the
// input vector remain in their original order after the transpose operation.
Expand Down
7 changes: 7 additions & 0 deletions mlir/test/Dialect/Vector/canonicalize.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -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>
Expand All @@ -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>
Expand Down Expand Up @@ -1710,6 +1714,7 @@ 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>,
Expand Down Expand Up @@ -2251,6 +2256,8 @@ 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>
Expand Down
Loading