Skip to content

[MLIR][Vector] Extend elementwise pattern to support unrolling from higher rank to lower rank #162515

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
nbpatel merged 11 commits into from
Oct 15, 2025
Merged

[MLIR][Vector] Extend elementwise pattern to support unrolling from higher rank to lower rank #162515

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
f854b2d
Extend elementwise to support unrolling from higher rank to lower rank
nbpatel Oct 8, 2025
01357d9
remove reshape before insertStride
nbpatel Oct 9, 2025
d59390b
Unused variable
nbpatel Oct 9, 2025
7f567ec
Address feedback
nbpatel Oct 13, 2025
5052b46
Fix CHECK
nbpatel Oct 14, 2025
61c6eda
Rename CHECK variable
nbpatel Oct 14, 2025
bb67a45
Add 4D to 2D test
nbpatel Oct 14, 2025
c203150
Add newline
nbpatel Oct 14, 2025
1765d87
Indent CHECK
nbpatel Oct 14, 2025
14ab24c
Fix CHECKS
nbpatel Oct 14, 2025
dca9d69
Fix test name
nbpatel Oct 15, 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
52 changes: 38 additions & 14 deletions mlir/lib/Dialect/Vector/Transforms/VectorUnroll.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -465,41 +465,65 @@ struct UnrollElementwisePattern : public RewritePattern {
auto targetShape = getTargetShape(options, op);
if (!targetShape)
return failure();
int64_t targetShapeRank = targetShape->size();
auto dstVecType = cast<VectorType>(op->getResult(0).getType());
SmallVector<int64_t> originalSize =
*cast<VectorUnrollOpInterface>(op).getShapeForUnroll();
// Bail-out if rank(source) != rank(target). The main limitation here is the
// fact that `ExtractStridedSlice` requires the rank for the input and
// output to match. If needed, we can relax this later.
if (originalSize.size() != targetShape->size())
return rewriter.notifyMatchFailure(
op, "expected input vector rank to match target shape rank");
int64_t originalShapeRank = originalSize.size();

Location loc = op->getLoc();

// Handle rank mismatch by adding leading unit dimensions to targetShape
SmallVector<int64_t> adjustedTargetShape(originalShapeRank);
int64_t rankDiff = originalShapeRank - targetShapeRank;
std::fill(adjustedTargetShape.begin(),
adjustedTargetShape.begin() + rankDiff, 1);
std::copy(targetShape->begin(), targetShape->end(),
adjustedTargetShape.begin() + rankDiff);

int64_t adjustedTargetShapeRank = adjustedTargetShape.size();
// Prepare the result vector.
Value result = arith::ConstantOp::create(rewriter, loc, dstVecType,
rewriter.getZeroAttr(dstVecType));
SmallVector<int64_t> strides(targetShape->size(), 1);
VectorType newVecType =
SmallVector<int64_t> strides(adjustedTargetShapeRank, 1);
VectorType unrolledVecType =
VectorType::get(*targetShape, dstVecType.getElementType());

// Create the unrolled computation.
for (SmallVector<int64_t> offsets :
StaticTileOffsetRange(originalSize, *targetShape)) {
StaticTileOffsetRange(originalSize, adjustedTargetShape)) {
SmallVector<Value> extractOperands;
for (OpOperand &operand : op->getOpOperands()) {
auto vecType = dyn_cast<VectorType>(operand.get().getType());
if (!vecType) {
extractOperands.push_back(operand.get());
continue;
}
extractOperands.push_back(
rewriter.createOrFold<vector::ExtractStridedSliceOp>(
loc, operand.get(), offsets, *targetShape, strides));
Value extracted = rewriter.createOrFold<vector::ExtractStridedSliceOp>(
loc, operand.get(), offsets, adjustedTargetShape, strides);

// Reshape to remove leading unit dims if needed
if (adjustedTargetShapeRank > targetShapeRank) {
extracted = rewriter.createOrFold<vector::ShapeCastOp>(
loc, VectorType::get(*targetShape, vecType.getElementType()),
extracted);
}
extractOperands.push_back(extracted);
}

Operation *newOp = cloneOpWithOperandsAndTypes(
rewriter, loc, op, extractOperands, newVecType);
rewriter, loc, op, extractOperands, unrolledVecType);

Value computeResult = newOp->getResult(0);

// Use strides sized to targetShape for proper insertion
SmallVector<int64_t> insertStrides =
(adjustedTargetShapeRank > targetShapeRank)
? SmallVector<int64_t>(targetShapeRank, 1)
: strides;

result = rewriter.createOrFold<vector::InsertStridedSliceOp>(
loc, newOp->getResult(0), result, offsets, strides);
loc, computeResult, result, offsets, insertStrides);
}
rewriter.replaceOp(op, result);
return success();
Expand Down
68 changes: 62 additions & 6 deletions mlir/test/Dialect/Vector/vector-unroll-options.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -188,15 +188,38 @@ func.func @vector_fma(%a: vector<4x4xf32>, %b: vector<4x4xf32>, %c: vector<4x4xf
// CHECK-LABEL: func @vector_fma
// CHECK-COUNT-4: vector.fma %{{.+}}, %{{.+}}, %{{.+}} : vector<2x2xf32>

// TODO: We should be able to unroll this like the example above - this will require extending UnrollElementwisePattern.
func.func @negative_vector_fma_3d(%a: vector<3x2x2xf32>) -> vector<3x2x2xf32>{
func.func @vector_fma_3d(%a: vector<3x2x2xf32>) -> vector<3x2x2xf32>{
%0 = vector.fma %a, %a, %a : vector<3x2x2xf32>
return %0 : vector<3x2x2xf32>
}
// CHECK-LABEL: func @negative_vector_fma_3d
// CHECK-NOT: vector.extract_strided_slice
// CHECK: %[[R0:.*]] = vector.fma %{{.+}} : vector<3x2x2xf32>
// CHECK: return
// CHECK-LABEL: func @vector_fma_3d
// CHECK-SAME: (%[[SRC:.*]]: vector<3x2x2xf32>) -> vector<3x2x2xf32> {
// CHECK: %[[CST:.*]] = arith.constant dense<0.000000e+00> : vector<3x2x2xf32>
// CHECK: %[[E_LHS_0:.*]] = vector.extract_strided_slice %[[SRC]] {offsets = [0, 0, 0], sizes = [1, 2, 2], strides = [1, 1, 1]} : vector<3x2x2xf32> to vector<1x2x2xf32>
// CHECK: %[[S_LHS_0:.*]] = vector.shape_cast %[[E_LHS_0]] : vector<1x2x2xf32> to vector<2x2xf32>
// CHECK: %[[E_RHS_0:.*]] = vector.extract_strided_slice %[[SRC]] {offsets = [0, 0, 0], sizes = [1, 2, 2], strides = [1, 1, 1]} : vector<3x2x2xf32> to vector<1x2x2xf32>
// CHECK: %[[S_RHS_0:.*]] = vector.shape_cast %[[E_RHS_0]] : vector<1x2x2xf32> to vector<2x2xf32>
// CHECK: %[[E_OUT_0:.*]] = vector.extract_strided_slice %[[SRC]] {offsets = [0, 0, 0], sizes = [1, 2, 2], strides = [1, 1, 1]} : vector<3x2x2xf32> to vector<1x2x2xf32>
// CHECK: %[[S_OUT_0:.*]] = vector.shape_cast %[[E_OUT_0]] : vector<1x2x2xf32> to vector<2x2xf32>
// CHECK: %[[FMA0:.*]] = vector.fma %[[S_LHS_0]], %[[S_RHS_0]], %[[S_OUT_0]] : vector<2x2xf32>
// CHECK: %[[I0:.*]] = vector.insert_strided_slice %[[FMA0]], %[[CST]] {offsets = [0, 0, 0], strides = [1, 1]} : vector<2x2xf32> into vector<3x2x2xf32>
// CHECK: %[[E_LHS_1:.*]] = vector.extract_strided_slice %[[SRC]] {offsets = [1, 0, 0], sizes = [1, 2, 2], strides = [1, 1, 1]} : vector<3x2x2xf32> to vector<1x2x2xf32>
// CHECK: %[[S_LHS_1:.*]] = vector.shape_cast %[[E_LHS_1]] : vector<1x2x2xf32> to vector<2x2xf32>
// CHECK: %[[E_RHS_1:.*]] = vector.extract_strided_slice %[[SRC]] {offsets = [1, 0, 0], sizes = [1, 2, 2], strides = [1, 1, 1]} : vector<3x2x2xf32> to vector<1x2x2xf32>
// CHECK: %[[S_RHS_1:.*]] = vector.shape_cast %[[E_RHS_1]] : vector<1x2x2xf32> to vector<2x2xf32>
// CHECK: %[[E_OUT_1:.*]] = vector.extract_strided_slice %[[SRC]] {offsets = [1, 0, 0], sizes = [1, 2, 2], strides = [1, 1, 1]} : vector<3x2x2xf32> to vector<1x2x2xf32>
// CHECK: %[[S_OUT_1:.*]] = vector.shape_cast %[[E_OUT_1]] : vector<1x2x2xf32> to vector<2x2xf32>
// CHECK: %[[FMA1:.*]] = vector.fma %[[S_LHS_1]], %[[S_RHS_1]], %[[S_OUT_1]] : vector<2x2xf32>
// CHECK: %[[I1:.*]] = vector.insert_strided_slice %[[FMA1]], %[[I0]] {offsets = [1, 0, 0], strides = [1, 1]} : vector<2x2xf32> into vector<3x2x2xf32>
// CHECK: %[[E_LHS_2:.*]] = vector.extract_strided_slice %[[SRC]] {offsets = [2, 0, 0], sizes = [1, 2, 2], strides = [1, 1, 1]} : vector<3x2x2xf32> to vector<1x2x2xf32>
// CHECK: %[[S_LHS_2:.*]] = vector.shape_cast %[[E_LHS_2]] : vector<1x2x2xf32> to vector<2x2xf32>
// CHECK: %[[E_RHS_2:.*]] = vector.extract_strided_slice %[[SRC]] {offsets = [2, 0, 0], sizes = [1, 2, 2], strides = [1, 1, 1]} : vector<3x2x2xf32> to vector<1x2x2xf32>
// CHECK: %[[S_RHS_2:.*]] = vector.shape_cast %[[E_RHS_2]] : vector<1x2x2xf32> to vector<2x2xf32>
// CHECK: %[[E_OUT_2:.*]] = vector.extract_strided_slice %[[SRC]] {offsets = [2, 0, 0], sizes = [1, 2, 2], strides = [1, 1, 1]} : vector<3x2x2xf32> to vector<1x2x2xf32>
// CHECK: %[[S_OUT_2:.*]] = vector.shape_cast %[[E_OUT_2]] : vector<1x2x2xf32> to vector<2x2xf32>
// CHECK: %[[FMA2:.*]] = vector.fma %[[S_LHS_2]], %[[S_RHS_2]], %[[S_OUT_2]] : vector<2x2xf32>
// CHECK: %[[I2:.*]] = vector.insert_strided_slice %[[FMA2]], %[[I1]] {offsets = [2, 0, 0], strides = [1, 1]} : vector<2x2xf32> into vector<3x2x2xf32>
// CHECK: return %[[I2]] : vector<3x2x2xf32>

func.func @vector_multi_reduction(%v : vector<4x6xf32>, %acc: vector<4xf32>) -> vector<4xf32> {
%0 = vector.multi_reduction #vector.kind<add>, %v, %acc [1] : vector<4x6xf32> to vector<4xf32>
Expand Down Expand Up @@ -440,3 +463,36 @@ func.func @vector_step() -> vector<32xindex> {
// CHECK: %[[ADD3:.*]] = arith.addi %[[STEP]], %[[CST]] : vector<8xindex>
// CHECK: %[[INS3:.*]] = vector.insert_strided_slice %[[ADD3]], %[[INS2]] {offsets = [24], strides = [1]} : vector<8xindex> into vector<32xindex>
// CHECK: return %[[INS3]] : vector<32xindex>


func.func @elementwise_3D_to_2D(%v1: vector<2x2x2xf32>, %v2: vector<2x2x2xf32>) -> vector<2x2x2xf32> {
%0 = arith.addf %v1, %v2 : vector<2x2x2xf32>
return %0 : vector<2x2x2xf32>
}
// CHECK-LABEL: func @elementwise_3D_to_2D
// CHECK-SAME: (%[[ARG0:.*]]: vector<2x2x2xf32>, %[[ARG1:.*]]: vector<2x2x2xf32>) -> vector<2x2x2xf32> {
// CHECK: %[[CST:.*]] = arith.constant dense<0.000000e+00> : vector<2x2x2xf32>
// CHECK: %[[E_LHS_0:.*]] = vector.extract_strided_slice %[[ARG0]] {offsets = [0, 0, 0], sizes = [1, 2, 2], strides = [1, 1, 1]} : vector<2x2x2xf32> to vector<1x2x2xf32>
// CHECK: %[[S_LHS_0:.*]] = vector.shape_cast %[[E_LHS_0]] : vector<1x2x2xf32> to vector<2x2xf32>
// CHECK: %[[E_RHS_0:.*]] = vector.extract_strided_slice %[[ARG1]] {offsets = [0, 0, 0], sizes = [1, 2, 2], strides = [1, 1, 1]} : vector<2x2x2xf32> to vector<1x2x2xf32>
// CHECK: %[[S_RHS_0:.*]] = vector.shape_cast %[[E_RHS_0]] : vector<1x2x2xf32> to vector<2x2xf32>
// CHECK: %[[ADD0:.*]] = arith.addf %[[S_LHS_0]], %[[S_RHS_0]] : vector<2x2xf32>
// CHECK: %[[I0:.*]] = vector.insert_strided_slice %[[ADD0]], %[[CST]] {offsets = [0, 0, 0], strides = [1, 1]} : vector<2x2xf32> into vector<2x2x2xf32>
// CHECK: %[[E_LHS_1:.*]] = vector.extract_strided_slice %[[ARG0]] {offsets = [1, 0, 0], sizes = [1, 2, 2], strides = [1, 1, 1]} : vector<2x2x2xf32> to vector<1x2x2xf32>
// CHECK: %[[S_LHS_1:.*]] = vector.shape_cast %[[E_LHS_1]] : vector<1x2x2xf32> to vector<2x2xf32>
// CHECK: %[[E_RHS_1:.*]] = vector.extract_strided_slice %[[ARG1]] {offsets = [1, 0, 0], sizes = [1, 2, 2], strides = [1, 1, 1]} : vector<2x2x2xf32> to vector<1x2x2xf32>
// CHECK: %[[S_RHS_1:.*]] = vector.shape_cast %[[E_RHS_1]] : vector<1x2x2xf32> to vector<2x2xf32>
// CHECK: %[[ADD1:.*]] = arith.addf %[[S_LHS_1]], %[[S_RHS_1]] : vector<2x2xf32>
// CHECK: %[[I1:.*]] = vector.insert_strided_slice %[[ADD1]], %[[I0]] {offsets = [1, 0, 0], strides = [1, 1]} : vector<2x2xf32> into vector<2x2x2xf32>
// CHECK: return %[[I1]] : vector<2x2x2xf32>


func.func @elementwise_4D_to_2D(%v1: vector<2x2x2x2xf32>, %v2: vector<2x2x2x2xf32>) -> vector<2x2x2x2xf32> {
%0 = arith.addf %v1, %v2 : vector<2x2x2x2xf32>
return %0 : vector<2x2x2x2xf32>
}

// CHECK-LABEL: func @elementwise_4D_to_2D
// CHECK-COUNT-4: arith.addf %{{.*}}, %{{.*}} : vector<2x2xf32>
// CHECK-NOT: arith.addf
// CHECK: return