[mlir][vector] Fix mask rank extension for TransferWriteNonPermutationLowering #144893
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Groverkss
requested review from
banach-space,
dcaballe,
hanhanW and
nicolasvasilache
as code owners
|
@llvm/pr-subscribers-mlir-vector Author: Kunwar Grover (Groverkss) ChangesAppending new dimensions to the mask is not always correct, infer the correct permutation for the mask and apply it. Full diff: https://github.com/llvm/llvm-project/pull/144893.diff 2 Files Affected:
diff --git a/mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp b/mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp
index 5b81d0d33d484..5863b0b7d45ab 100644
--- a/mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp
+++ b/mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp
@@ -35,8 +35,8 @@ inverseTransposeInBoundsAttr(OpBuilder &builder, ArrayAttr attr,
/// Extend the rank of a vector Value by `addedRanks` by adding outer unit
/// dimensions.
-static Value extendVectorRank(OpBuilder &builder, Location loc, Value vec,
- int64_t addedRank) {
+static TypedValue<VectorType> extendVectorRank(OpBuilder &builder, Location loc,
+ Value vec, int64_t addedRank) {
auto originalVecType = cast<VectorType>(vec.getType());
SmallVector<int64_t> newShape(addedRank, 1);
newShape.append(originalVecType.getShape().begin(),
@@ -53,16 +53,21 @@ static Value extendVectorRank(OpBuilder &builder, Location loc, Value vec,
/// Extend the rank of a vector Value by `addedRanks` by adding inner unit
/// dimensions.
static Value extendMaskRank(OpBuilder &builder, Location loc, Value vec,
- int64_t addedRank) {
- Value broadcasted = extendVectorRank(builder, loc, vec, addedRank);
- SmallVector<int64_t> permutation;
- for (int64_t i = addedRank,
- e = cast<VectorType>(broadcasted.getType()).getRank();
- i < e; ++i)
- permutation.push_back(i);
- for (int64_t i = 0; i < addedRank; ++i)
- permutation.push_back(i);
- return builder.create<vector::TransposeOp>(loc, broadcasted, permutation);
+ ArrayRef<int64_t> missingInnerDims) {
+ TypedValue<VectorType> broadcasted =
+ extendVectorRank(builder, loc, vec, missingInnerDims.size());
+ SmallVector<bool> missing(broadcasted.getType().getRank(), false);
+ SmallVector<int64_t> inversePerm;
+ for (int64_t i : missingInnerDims) {
+ inversePerm.push_back(i);
+ missing[i] = true;
+ }
+ for (auto [i, used] : llvm::enumerate(missing)) {
+ if (!used)
+ inversePerm.push_back(i);
+ }
+ return builder.create<vector::TransposeOp>(
+ loc, broadcasted, invertPermutationVector(inversePerm));
}
//===----------------------------------------------------------------------===//
@@ -268,28 +273,30 @@ struct TransferWriteNonPermutationLowering
for (AffineExpr exp : map.getResults())
foundDim[cast<AffineDimExpr>(exp).getPosition()] = true;
SmallVector<AffineExpr> exprs;
- bool foundFirstDim = false;
+ std::optional<int64_t> firstDim = std::nullopt;
SmallVector<int64_t> missingInnerDim;
for (size_t i = 0; i < foundDim.size(); i++) {
if (foundDim[i]) {
- foundFirstDim = true;
+ if (!firstDim) {
+ firstDim = i;
+ }
continue;
}
- if (!foundFirstDim)
+ if (!firstDim)
continue;
// Once we found one outer dimension existing in the map keep track of all
// the missing dimensions after that.
- missingInnerDim.push_back(i);
+ missingInnerDim.push_back(i - firstDim.value());
exprs.push_back(rewriter.getAffineDimExpr(i));
}
// Vector: add unit dims at the beginning of the shape.
Value newVec = extendVectorRank(rewriter, op.getLoc(), op.getVector(),
missingInnerDim.size());
- // Mask: add unit dims at the end of the shape.
+ // Mask: add unit dims at the positions of the missing dimensions.
Value newMask;
if (op.getMask())
- newMask = extendMaskRank(rewriter, op.getLoc(), op.getMask(),
- missingInnerDim.size());
+ newMask =
+ extendMaskRank(rewriter, op.getLoc(), op.getMask(), missingInnerDim);
exprs.append(map.getResults().begin(), map.getResults().end());
AffineMap newMap =
AffineMap::get(map.getNumDims(), 0, exprs, op.getContext());
diff --git a/mlir/test/Dialect/Vector/vector-transfer-permutation-lowering.mlir b/mlir/test/Dialect/Vector/vector-transfer-permutation-lowering.mlir
index 3ae18835c8367..07e49d7e12e25 100644
--- a/mlir/test/Dialect/Vector/vector-transfer-permutation-lowering.mlir
+++ b/mlir/test/Dialect/Vector/vector-transfer-permutation-lowering.mlir
@@ -160,6 +160,31 @@ func.func @xfer_write_non_minor_identity_with_mask_out_of_bounds(
return
}
+// CHECK-LABEL: func.func @xfer_write_non_minor_identity_with_mask_broadcast(
+// CHECK-SAME: %[[MEM:.*]]: memref<?x?x?xf32>,
+// CHECK-SAME: %[[VEC:.*]]: vector<7x8xf32>,
+// CHECK-SAME: %[[MASK:.*]]: vector<7x8xi1>,
+// CHECK-SAME: %[[IDX_1:.*]]: index, %[[IDX_2:.*]]: index, %[[IDX_3:.*]]: index) {
+// CHECK: %[[BC:.*]] = vector.broadcast %[[VEC]] : vector<7x8xf32> to vector<1x7x8xf32>
+// CHECK: %[[MBC:.*]] = vector.broadcast %[[MASK]] : vector<7x8xi1> to vector<1x7x8xi1>
+// CHECK: %[[MTR:.*]] = vector.transpose %[[MBC]], [1, 0, 2] : vector<1x7x8xi1> to vector<7x1x8xi1>
+// CHECK: %[[TR:.*]] = vector.transpose %[[BC]], [1, 0, 2] : vector<1x7x8xf32> to vector<7x1x8xf32>
+// CHECK: vector.transfer_write %[[TR]], %[[MEM]]{{\[}}%[[IDX_1]], %[[IDX_2]], %[[IDX_3]]], %[[MTR]] {in_bounds = [false, true, false]} : vector<7x1x8xf32>, memref<?x?x?xf32>
+func.func @xfer_write_non_minor_identity_with_mask_broadcast_and_transpose(
+ %mem : memref<?x?x?xf32>,
+ %vec : vector<7x8xf32>,
+ %mask : vector<7x8xi1>,
+ %idx_1 : index,
+ %idx_2 : index,
+ %idx_3 : index) {
+
+ vector.transfer_write %vec, %mem[%idx_1, %idx_2, %idx_3], %mask {
+ permutation_map = affine_map<(d0, d1, d2) -> (d0, d2)>
+ } : vector<7x8xf32>, memref<?x?x?xf32>
+
+ return
+}
+
// CHECK-LABEL: func.func @xfer_write_non_minor_identity_with_mask_scalable(
// CHECK-SAME: %[[VEC:.*]]: vector<4x[8]xi16>,
// CHECK-SAME: %[[MEM:.*]]: memref<1x4x?x1xi16>,
|
|
@llvm/pr-subscribers-mlir Author: Kunwar Grover (Groverkss) ChangesAppending new dimensions to the mask is not always correct, infer the correct permutation for the mask and apply it. Full diff: https://github.com/llvm/llvm-project/pull/144893.diff 2 Files Affected:
diff --git a/mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp b/mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp
index 5b81d0d33d484..5863b0b7d45ab 100644
--- a/mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp
+++ b/mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp
@@ -35,8 +35,8 @@ inverseTransposeInBoundsAttr(OpBuilder &builder, ArrayAttr attr,
/// Extend the rank of a vector Value by `addedRanks` by adding outer unit
/// dimensions.
-static Value extendVectorRank(OpBuilder &builder, Location loc, Value vec,
- int64_t addedRank) {
+static TypedValue<VectorType> extendVectorRank(OpBuilder &builder, Location loc,
+ Value vec, int64_t addedRank) {
auto originalVecType = cast<VectorType>(vec.getType());
SmallVector<int64_t> newShape(addedRank, 1);
newShape.append(originalVecType.getShape().begin(),
@@ -53,16 +53,21 @@ static Value extendVectorRank(OpBuilder &builder, Location loc, Value vec,
/// Extend the rank of a vector Value by `addedRanks` by adding inner unit
/// dimensions.
static Value extendMaskRank(OpBuilder &builder, Location loc, Value vec,
- int64_t addedRank) {
- Value broadcasted = extendVectorRank(builder, loc, vec, addedRank);
- SmallVector<int64_t> permutation;
- for (int64_t i = addedRank,
- e = cast<VectorType>(broadcasted.getType()).getRank();
- i < e; ++i)
- permutation.push_back(i);
- for (int64_t i = 0; i < addedRank; ++i)
- permutation.push_back(i);
- return builder.create<vector::TransposeOp>(loc, broadcasted, permutation);
+ ArrayRef<int64_t> missingInnerDims) {
+ TypedValue<VectorType> broadcasted =
+ extendVectorRank(builder, loc, vec, missingInnerDims.size());
+ SmallVector<bool> missing(broadcasted.getType().getRank(), false);
+ SmallVector<int64_t> inversePerm;
+ for (int64_t i : missingInnerDims) {
+ inversePerm.push_back(i);
+ missing[i] = true;
+ }
+ for (auto [i, used] : llvm::enumerate(missing)) {
+ if (!used)
+ inversePerm.push_back(i);
+ }
+ return builder.create<vector::TransposeOp>(
+ loc, broadcasted, invertPermutationVector(inversePerm));
}
//===----------------------------------------------------------------------===//
@@ -268,28 +273,30 @@ struct TransferWriteNonPermutationLowering
for (AffineExpr exp : map.getResults())
foundDim[cast<AffineDimExpr>(exp).getPosition()] = true;
SmallVector<AffineExpr> exprs;
- bool foundFirstDim = false;
+ std::optional<int64_t> firstDim = std::nullopt;
SmallVector<int64_t> missingInnerDim;
for (size_t i = 0; i < foundDim.size(); i++) {
if (foundDim[i]) {
- foundFirstDim = true;
+ if (!firstDim) {
+ firstDim = i;
+ }
continue;
}
- if (!foundFirstDim)
+ if (!firstDim)
continue;
// Once we found one outer dimension existing in the map keep track of all
// the missing dimensions after that.
- missingInnerDim.push_back(i);
+ missingInnerDim.push_back(i - firstDim.value());
exprs.push_back(rewriter.getAffineDimExpr(i));
}
// Vector: add unit dims at the beginning of the shape.
Value newVec = extendVectorRank(rewriter, op.getLoc(), op.getVector(),
missingInnerDim.size());
- // Mask: add unit dims at the end of the shape.
+ // Mask: add unit dims at the positions of the missing dimensions.
Value newMask;
if (op.getMask())
- newMask = extendMaskRank(rewriter, op.getLoc(), op.getMask(),
- missingInnerDim.size());
+ newMask =
+ extendMaskRank(rewriter, op.getLoc(), op.getMask(), missingInnerDim);
exprs.append(map.getResults().begin(), map.getResults().end());
AffineMap newMap =
AffineMap::get(map.getNumDims(), 0, exprs, op.getContext());
diff --git a/mlir/test/Dialect/Vector/vector-transfer-permutation-lowering.mlir b/mlir/test/Dialect/Vector/vector-transfer-permutation-lowering.mlir
index 3ae18835c8367..07e49d7e12e25 100644
--- a/mlir/test/Dialect/Vector/vector-transfer-permutation-lowering.mlir
+++ b/mlir/test/Dialect/Vector/vector-transfer-permutation-lowering.mlir
@@ -160,6 +160,31 @@ func.func @xfer_write_non_minor_identity_with_mask_out_of_bounds(
return
}
+// CHECK-LABEL: func.func @xfer_write_non_minor_identity_with_mask_broadcast(
+// CHECK-SAME: %[[MEM:.*]]: memref<?x?x?xf32>,
+// CHECK-SAME: %[[VEC:.*]]: vector<7x8xf32>,
+// CHECK-SAME: %[[MASK:.*]]: vector<7x8xi1>,
+// CHECK-SAME: %[[IDX_1:.*]]: index, %[[IDX_2:.*]]: index, %[[IDX_3:.*]]: index) {
+// CHECK: %[[BC:.*]] = vector.broadcast %[[VEC]] : vector<7x8xf32> to vector<1x7x8xf32>
+// CHECK: %[[MBC:.*]] = vector.broadcast %[[MASK]] : vector<7x8xi1> to vector<1x7x8xi1>
+// CHECK: %[[MTR:.*]] = vector.transpose %[[MBC]], [1, 0, 2] : vector<1x7x8xi1> to vector<7x1x8xi1>
+// CHECK: %[[TR:.*]] = vector.transpose %[[BC]], [1, 0, 2] : vector<1x7x8xf32> to vector<7x1x8xf32>
+// CHECK: vector.transfer_write %[[TR]], %[[MEM]]{{\[}}%[[IDX_1]], %[[IDX_2]], %[[IDX_3]]], %[[MTR]] {in_bounds = [false, true, false]} : vector<7x1x8xf32>, memref<?x?x?xf32>
+func.func @xfer_write_non_minor_identity_with_mask_broadcast_and_transpose(
+ %mem : memref<?x?x?xf32>,
+ %vec : vector<7x8xf32>,
+ %mask : vector<7x8xi1>,
+ %idx_1 : index,
+ %idx_2 : index,
+ %idx_3 : index) {
+
+ vector.transfer_write %vec, %mem[%idx_1, %idx_2, %idx_3], %mask {
+ permutation_map = affine_map<(d0, d1, d2) -> (d0, d2)>
+ } : vector<7x8xf32>, memref<?x?x?xf32>
+
+ return
+}
+
// CHECK-LABEL: func.func @xfer_write_non_minor_identity_with_mask_scalable(
// CHECK-SAME: %[[VEC:.*]]: vector<4x[8]xi16>,
// CHECK-SAME: %[[MEM:.*]]: memref<1x4x?x1xi16>,
|
| // CHECK: %[[MTR:.*]] = vector.transpose %[[MBC]], [1, 0, 2] : vector<1x7x8xi1> to vector<7x1x8xi1> | ||
| // CHECK: %[[TR:.*]] = vector.transpose %[[BC]], [1, 0, 2] : vector<1x7x8xf32> to vector<7x1x8xf32> | ||
| // CHECK: vector.transfer_write %[[TR]], %[[MEM]]{{\[}}%[[IDX_1]], %[[IDX_2]], %[[IDX_3]]], %[[MTR]] {in_bounds = [false, true, false]} : vector<7x1x8xf32>, memref<?x?x?xf32> | ||
| func.func @xfer_write_non_minor_identity_with_mask_broadcast_and_transpose( |
There was a problem hiding this comment.
Based on the name (@xfer_write_non_minor_identity_with_mask_broadcast_and_transpose), I would expect to see explicit broadcast when comparing this to @xfer_write_non_minor_identity_with_mask_out_of_bounds. However, the only difference I can see are input ranks.
Don't want to come across as nit-picking, just trying to figure out what cases this PR fixes. Thanks!
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Appending new dimensions to the mask is not always correct, infer the correct permutation for the mask and apply it.