[Mlir-commits] [mlir] [mlir][MemRef] Add position-based matching heuristics for rank-reduction with dynamic strides (PR #184334)

llvmlistbot at llvm.org llvmlistbot at llvm.org
Tue Mar 3 05:00:46 PST 2026


llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-mlir-memref

Author: Abhishek Varma (Abhishek-Varma)

<details>
<summary>Changes</summary>

When multiple source dimensions have multiple unit dimensions, stride-based disambiguation can be wrong with dynamic strides. Add position-based matching: for each result dimension in order, pick the leftmost unmatched source dimension with the same size; unmatched source dims are dropped.

Example: subview from memref<1x8x1x3> to memref<1x8x3>. Both dim 0 and dim 2 have size 1. Stride-based logic cannot distinguish when strides are dynamic. Position-based matching correctly drops dim 2 (middle unit dim) instead of dim 0.

Use position-based matching when multiple dimensions are being dropped, falling back to stride-based logic otherwise.

INPUT :-
```
func.func @<!-- -->fold_rank_reducing_subview_1x8x1x3_to_1x8x3_drop_middle_unit_dim(
    %arg0 : memref<?x?x?x?xf32, strided<[?, ?, ?, ?], offset: ?>>,
    %arg1 : index, %arg2 : index, %arg3 : index, %arg4 : index) -> f32 {
  %c0 = arith.constant 0 : index
  %0 = memref.subview %arg0[0, 0, 0, 0][1, 8, 1, 3][1, 1, 1, 1]
      : memref<?x?x?x?xf32, strided<[?, ?, ?, ?], offset: ?>> to
        memref<1x8x3xf32, strided<[?, ?, ?], offset: ?>>
  %1 = memref.load %0[%c0, %arg1, %arg2] : memref<1x8x3xf32, strided<[?, ?, ?], offset: ?>>
  return %1 : f32
}
```

WITHOUT this patch we get :-
```
memref.load %0[%c0, %c0, %arg1, %arg2]
```

WITH this patch we get :-
```
memref.load %0[%c0, %arg1, %c0, %arg2]
```

Signed-off-by: Abhishek Varma <abhvarma@<!-- -->amd.com>

---
Full diff: https://github.com/llvm/llvm-project/pull/184334.diff


3 Files Affected:

- (modified) mlir/lib/Dialect/MemRef/IR/MemRefOps.cpp (+62) 
- (modified) mlir/test/Dialect/MemRef/canonicalize.mlir (+1-1) 
- (modified) mlir/test/Dialect/MemRef/fold-memref-alias-ops.mlir (+25-4) 


``````````diff
diff --git a/mlir/lib/Dialect/MemRef/IR/MemRefOps.cpp b/mlir/lib/Dialect/MemRef/IR/MemRefOps.cpp
index 844e6183cff06..321bfc27516fc 100644
--- a/mlir/lib/Dialect/MemRef/IR/MemRefOps.cpp
+++ b/mlir/lib/Dialect/MemRef/IR/MemRefOps.cpp
@@ -944,6 +944,56 @@ static std::map<int64_t, unsigned> getNumOccurences(ArrayRef<int64_t> vals) {
   return numOccurences;
 }
 
+/// Returns the set of source dimensions that are dropped in a rank reduction.
+/// For each result dimension in order, matches the leftmost unmatched source
+/// dimension with the same size. Source dimensions not matched are dropped.
+///
+/// Example: memref<1x8x1x3> to memref<1x8x3>. Source sizes [1, 8, 1, 3], result
+/// [1, 8, 3]. Match result[0]=1 -> source dim 0, result[1]=8 -> source dim 1,
+/// result[2]=3 -> source dim 3. Source dim 2 is unmatched and dropped.
+static FailureOr<llvm::SmallBitVector>
+computeMemRefRankReductionMaskByPosition(MemRefType originalType,
+                                         MemRefType reducedType,
+                                         ArrayRef<OpFoldResult> sizes) {
+  int64_t rankReduction = originalType.getRank() - reducedType.getRank();
+  if (rankReduction <= 0)
+    return llvm::SmallBitVector(originalType.getRank());
+
+  // Build source sizes from subview sizes (one per source dim).
+  SmallVector<int64_t> sourceSizes(originalType.getRank());
+  for (const auto &it : llvm::enumerate(sizes)) {
+    if (std::optional<int64_t> cst = getConstantIntValue(it.value()))
+      sourceSizes[it.index()] = *cst;
+    else
+      sourceSizes[it.index()] = ShapedType::kDynamic;
+  }
+
+  ArrayRef<int64_t> resultSizes = reducedType.getShape();
+  llvm::SmallBitVector usedSourceDims(originalType.getRank());
+  for (int64_t resultSize : resultSizes) {
+    bool matched = false;
+    for (int64_t j = 0; j < originalType.getRank(); ++j) {
+      if (usedSourceDims.test(j))
+        continue;
+      if (sourceSizes[j] == resultSize ||
+          (resultSize == ShapedType::kDynamic &&
+           sourceSizes[j] == ShapedType::kDynamic)) {
+        usedSourceDims.set(j);
+        matched = true;
+        break;
+      }
+    }
+    if (!matched)
+      return failure();
+  }
+
+  llvm::SmallBitVector unusedDims(originalType.getRank());
+  for (int64_t i = 0; i < originalType.getRank(); ++i)
+    if (!usedSourceDims.test(i))
+      unusedDims.set(i);
+  return unusedDims;
+}
+
 /// Given the `originalType` and a `candidateReducedType` whose shape is assumed
 /// to be a subset of `originalType` with some `1` entries erased, return the
 /// set of indices that specifies which of the entries of `originalShape` are
@@ -977,6 +1027,18 @@ computeMemRefRankReductionMask(MemRefType originalType, MemRefType reducedType,
           reducedType.getStridesAndOffset(candidateStrides, candidateOffset)))
     return failure();
 
+  // When strides are dynamic and multiple dimensions need to be dropped, we use
+  // position-based matching instead.
+  if (unusedDims.count() > 1 &&
+      (llvm::any_of(originalStrides, ShapedType::isDynamic) ||
+       llvm::any_of(candidateStrides, ShapedType::isDynamic))) {
+    FailureOr<llvm::SmallBitVector> positionBased =
+        computeMemRefRankReductionMaskByPosition(originalType, reducedType,
+                                                 sizes);
+    if (succeeded(positionBased))
+      return *positionBased;
+  }
+
   // For memrefs, a dimension is truly dropped if its corresponding stride is
   // also dropped. This is particularly important when more than one of the dims
   // is 1. Track the number of occurences of the strides in the original type
diff --git a/mlir/test/Dialect/MemRef/canonicalize.mlir b/mlir/test/Dialect/MemRef/canonicalize.mlir
index 3cfea1e8cd961..f0193d066dc37 100644
--- a/mlir/test/Dialect/MemRef/canonicalize.mlir
+++ b/mlir/test/Dialect/MemRef/canonicalize.mlir
@@ -47,7 +47,7 @@ func.func @subview_of_size_memcast(%arg : memref<4x6x16x32xi8>) ->
 //       CHECK: func @subview_of_strides_memcast
 //  CHECK-SAME:   %[[ARG0:.[a-z0-9A-Z_]+]]: memref<1x1x?xf32, strided{{.*}}>
 //       CHECK:   %[[S:.+]] = memref.subview %[[ARG0]][0, 0, 0] [1, 1, 4]
-//  CHECK-SAME:                    to memref<1x4xf32, strided<[7, 1], offset: ?>>
+//  CHECK-SAME:                    to memref<1x4xf32, strided<[35, 1], offset: ?>>
 //       CHECK:   %[[M:.+]] = memref.cast %[[S]]
 //  CHECK-SAME:                    to memref<1x4xf32, strided<[?, ?], offset: ?>>
 //       CHECK:   return %[[M]]
diff --git a/mlir/test/Dialect/MemRef/fold-memref-alias-ops.mlir b/mlir/test/Dialect/MemRef/fold-memref-alias-ops.mlir
index 79156df0ebe1e..9f3e8deb48a4c 100644
--- a/mlir/test/Dialect/MemRef/fold-memref-alias-ops.mlir
+++ b/mlir/test/Dialect/MemRef/fold-memref-alias-ops.mlir
@@ -171,10 +171,31 @@ func.func @fold_rank_reducing_subview_with_load
 // CHECK-SAME:   %[[ARG15:[a-zA-Z0-9_]+]]: index
 // CHECK-SAME:   %[[ARG16:[a-zA-Z0-9_]+]]: index
 //  CHECK-DAG:   %[[I0:.+]] = affine.apply #[[MAP]]()[%[[ARG1]], %[[ARG13]], %[[ARG7]]]
-//  CHECK-DAG:   %[[I2:.+]] = affine.apply #[[MAP]]()[%[[ARG3]], %[[ARG14]], %[[ARG9]]]
-//  CHECK-DAG:   %[[I3:.+]] = affine.apply #[[MAP]]()[%[[ARG4]], %[[ARG15]], %[[ARG10]]]
-//  CHECK-DAG:   %[[I4:.+]] = affine.apply #[[MAP]]()[%[[ARG5]], %[[ARG16]], %[[ARG11]]]
-//      CHECK:   memref.load %[[ARG0]][%[[I0]], %[[ARG2]], %[[I2]], %[[I3]], %[[I4]], %[[ARG6]]]
+//  CHECK-DAG:   %[[I1:.+]] = affine.apply #[[MAP]]()[%[[ARG2]], %[[ARG14]], %[[ARG8]]]
+//  CHECK-DAG:   %[[I2:.+]] = affine.apply #[[MAP]]()[%[[ARG3]], %[[ARG15]], %[[ARG9]]]
+//  CHECK-DAG:   %[[I3:.+]] = affine.apply #[[MAP]]()[%[[ARG4]], %[[ARG16]], %[[ARG10]]]
+//      CHECK:   memref.load %[[ARG0]][%[[I0]], %[[I1]], %[[I2]], %[[I3]], %[[ARG5]], %[[ARG6]]]
+
+// -----
+
+func.func @fold_rank_reducing_subview_1x8x1x3_to_1x8x3_drop_middle_unit_dim(
+    %arg0 : memref<?x?x?x?xf32, strided<[?, ?, ?, ?], offset: ?>>,
+    %arg1 : index, %arg2 : index, %arg3 : index, %arg4 : index) -> f32 {
+  %c0 = arith.constant 0 : index
+  %0 = memref.subview %arg0[0, 0, 0, 0][1, 8, 1, 3][1, 1, 1, 1]
+      : memref<?x?x?x?xf32, strided<[?, ?, ?, ?], offset: ?>> to
+        memref<1x8x3xf32, strided<[?, ?, ?], offset: ?>>
+  %1 = memref.load %0[%c0, %arg1, %arg2] : memref<1x8x3xf32, strided<[?, ?, ?], offset: ?>>
+  return %1 : f32
+}
+//      CHECK: func @fold_rank_reducing_subview_1x8x1x3_to_1x8x3_drop_middle_unit_dim
+// CHECK-SAME:   %[[ARG0:[a-zA-Z0-9_]+]]: memref<?x?x?x?xf32, strided<[?, ?, ?, ?], offset: ?>>
+// CHECK-SAME:   %[[ARG1:[a-zA-Z0-9_]+]]: index
+// CHECK-SAME:   %[[ARG2:[a-zA-Z0-9_]+]]: index
+// CHECK-SAME:   %[[ARG3:[a-zA-Z0-9_]+]]: index
+// CHECK-SAME:   %[[ARG4:[a-zA-Z0-9_]+]]: index
+//  CHECK-DAG:   %[[C0:.+]] = arith.constant 0 : index
+//      CHECK:   memref.load %[[ARG0]][%[[C0]], %[[ARG1]], %[[C0]], %[[ARG2]]]
 
 // -----
 

``````````

</details>


https://github.com/llvm/llvm-project/pull/184334


More information about the Mlir-commits mailing list