mirrors/llvm-project
0
0
Fork 0
mirror of https://github.com/llvm/llvm-project.git synced 2025-04-19 01:46:49 +00:00
Code Issues Projects Releases Wiki Activity

[mlir]Add a check to ensure bailing out when reducing to a scalar (#129694)

Browse Source 
Fixes issue #64075
Referencing this comment for more detailed view ->
https://github.com/llvm/llvm-project/issues/64075#issuecomment-2694112594

**Minimal example crashing :** 
```
func.func @multi_reduction(%0: vector<4x2xf32>, %acc1: f32) -> f32 {
  %2 = vector.multi_reduction <add>, %0, %acc1 [0, 1] : vector<4x2xf32> to f32
  return %2 : f32
}
```
This commit is contained in:
Prakhar Dixit 2025-03-08 22:04:01 +05:30 committed by GitHub
parent 46d218d1af
commit 037756242f
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 14 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
mlir/lib/Dialect/Vector/Transforms/VectorUnroll.cpp
View File

@ -355,6 +355,11 @@ struct UnrollMultiReductionPattern
LogicalResult matchAndRewrite(vector::MultiDimReductionOp reductionOp,
PatternRewriter &rewriter) const override {
auto resultType = reductionOp->getResult(0).getType();
if (resultType.isIntOrFloat()) {
return rewriter.notifyMatchFailure(reductionOp,
"Unrolling scalars is not supported");
}
std::optional<SmallVector<int64_t>> targetShape =
getTargetShape(options, reductionOp);
if (!targetShape)

9
mlir/test/Dialect/Vector/vector-unroll-options.mlir
View File

@ -222,6 +222,15 @@ func.func @vector_multi_reduction(%v : vector<4x6xf32>, %acc: vector<4xf32>) ->
// CHECK: %[[V2:.*]] = vector.insert_strided_slice %[[R5]], %[[V1]] {offsets = [2], strides = [1]} : vector<2xf32> into vector<4xf32>
// CHECK: return %[[V2]] : vector<4xf32>
// This is a negative test case to ensure that further unrolling is not performed. Since the vector.multi_reduction
// operation has already been unrolled, attempting additional unrolling should not be allowed.
func.func @negative_vector_multi_reduction(%v: vector<4x2xf32>, %acc: f32) -> f32 {
%0 = vector.multi_reduction #vector.kind<add>, %v, %acc [0, 1] : vector<4x2xf32> to f32
return %0 : f32
}
// CHECK-LABEL: func @negative_vector_multi_reduction
// CHECK-NEXT: %[[R0:.*]] = vector.multi_reduction <add>, %{{.*}}, %{{.*}} [0, 1] : vector<4x2xf32> to f32
// CHECK-NEXT: return %[[R0]] : f32
func.func @vector_reduction(%v : vector<8xf32>) -> f32 {
%0 = vector.reduction <add>, %v : vector<8xf32> into f32