[MLIR][GPUToLLVMSPV] Relax the width check in gpu.shuffle lowering#183445
Merged
Jianhui-Li merged 4 commits intoMar 6, 2026
Conversation
Member
|
@llvm/pr-subscribers-mlir-gpu Author: Jianhui Li (Jianhui-Li) ChangesThis PR modifies gpu.shuffle lowering so only conduct the width check when the subgroupsize attribute is available. Full diff: https://github.com/llvm/llvm-project/pull/183445.diff 2 Files Affected:
diff --git a/mlir/lib/Conversion/GPUToLLVMSPV/GPUToLLVMSPV.cpp b/mlir/lib/Conversion/GPUToLLVMSPV/GPUToLLVMSPV.cpp
index 8c92b7e2b718b..87dca361e81e1 100644
--- a/mlir/lib/Conversion/GPUToLLVMSPV/GPUToLLVMSPV.cpp
+++ b/mlir/lib/Conversion/GPUToLLVMSPV/GPUToLLVMSPV.cpp
@@ -345,7 +345,7 @@ struct GPUShuffleConversion final : ConvertOpToLLVMPattern<gpu::ShuffleOp> {
LogicalResult
matchAndRewrite(gpu::ShuffleOp op, OpAdaptor adaptor,
ConversionPatternRewriter &rewriter) const final {
- if (!hasValidWidth(op))
+ if (getSubgroupSize(op) && !hasValidWidth(op))
return rewriter.notifyMatchFailure(
op, "shuffle width and subgroup size mismatch");
diff --git a/mlir/test/Conversion/GPUToLLVMSPV/gpu-to-llvm-spv.mlir b/mlir/test/Conversion/GPUToLLVMSPV/gpu-to-llvm-spv.mlir
index e56f3117e3b3c..ef5aa023bce51 100644
--- a/mlir/test/Conversion/GPUToLLVMSPV/gpu-to-llvm-spv.mlir
+++ b/mlir/test/Conversion/GPUToLLVMSPV/gpu-to-llvm-spv.mlir
@@ -344,6 +344,25 @@ gpu.module @shuffles {
// -----
+// Check `gpu.shuffle` conversion with no explicit subgroup size.
+
+// CHECK: llvm.func spir_funccc @_Z17sub_group_shuffleij(i32, i32) -> i32 attributes {convergent, no_unwind, will_return}
+// CHECK-LABEL: llvm.func @gpu_shuffles(
+// CHECK-SAME: %[[ARG0:.*]]: i32, %[[ARG1:.*]]: i32
+// CHECK: %[[C16:.*]] = arith.constant 16 : i32
+// CHECK: %[[SHUF:.*]] = llvm.call spir_funccc @_Z17sub_group_shuffleij(%[[ARG0]], %[[ARG1]]) {convergent, no_unwind, will_return} : (i32, i32) -> i32
+// CHECK: %[[TRUE:.*]] = llvm.mlir.constant(true) : i1
+// CHECK: llvm.return
+gpu.module @shuffles_without_intel_reqd_sub_group_size_attribute {
+ llvm.func @gpu_shuffles(%val: i32, %id: i32) {
+ %width = arith.constant 16 : i32
+ %shuffleResult, %valid = gpu.shuffle idx %val, %id, %width : i32
+ llvm.return
+ }
+}
+
+// -----
+
// Cannot convert due to shuffle width and target subgroup size mismatch
gpu.module @shuffles_mismatch {
|
Member
|
@llvm/pr-subscribers-mlir Author: Jianhui Li (Jianhui-Li) ChangesThis PR modifies gpu.shuffle lowering so only conduct the width check when the subgroupsize attribute is available. Full diff: https://github.com/llvm/llvm-project/pull/183445.diff 2 Files Affected:
diff --git a/mlir/lib/Conversion/GPUToLLVMSPV/GPUToLLVMSPV.cpp b/mlir/lib/Conversion/GPUToLLVMSPV/GPUToLLVMSPV.cpp
index 8c92b7e2b718b..87dca361e81e1 100644
--- a/mlir/lib/Conversion/GPUToLLVMSPV/GPUToLLVMSPV.cpp
+++ b/mlir/lib/Conversion/GPUToLLVMSPV/GPUToLLVMSPV.cpp
@@ -345,7 +345,7 @@ struct GPUShuffleConversion final : ConvertOpToLLVMPattern<gpu::ShuffleOp> {
LogicalResult
matchAndRewrite(gpu::ShuffleOp op, OpAdaptor adaptor,
ConversionPatternRewriter &rewriter) const final {
- if (!hasValidWidth(op))
+ if (getSubgroupSize(op) && !hasValidWidth(op))
return rewriter.notifyMatchFailure(
op, "shuffle width and subgroup size mismatch");
diff --git a/mlir/test/Conversion/GPUToLLVMSPV/gpu-to-llvm-spv.mlir b/mlir/test/Conversion/GPUToLLVMSPV/gpu-to-llvm-spv.mlir
index e56f3117e3b3c..ef5aa023bce51 100644
--- a/mlir/test/Conversion/GPUToLLVMSPV/gpu-to-llvm-spv.mlir
+++ b/mlir/test/Conversion/GPUToLLVMSPV/gpu-to-llvm-spv.mlir
@@ -344,6 +344,25 @@ gpu.module @shuffles {
// -----
+// Check `gpu.shuffle` conversion with no explicit subgroup size.
+
+// CHECK: llvm.func spir_funccc @_Z17sub_group_shuffleij(i32, i32) -> i32 attributes {convergent, no_unwind, will_return}
+// CHECK-LABEL: llvm.func @gpu_shuffles(
+// CHECK-SAME: %[[ARG0:.*]]: i32, %[[ARG1:.*]]: i32
+// CHECK: %[[C16:.*]] = arith.constant 16 : i32
+// CHECK: %[[SHUF:.*]] = llvm.call spir_funccc @_Z17sub_group_shuffleij(%[[ARG0]], %[[ARG1]]) {convergent, no_unwind, will_return} : (i32, i32) -> i32
+// CHECK: %[[TRUE:.*]] = llvm.mlir.constant(true) : i1
+// CHECK: llvm.return
+gpu.module @shuffles_without_intel_reqd_sub_group_size_attribute {
+ llvm.func @gpu_shuffles(%val: i32, %id: i32) {
+ %width = arith.constant 16 : i32
+ %shuffleResult, %valid = gpu.shuffle idx %val, %id, %width : i32
+ llvm.return
+ }
+}
+
+// -----
+
// Cannot convert due to shuffle width and target subgroup size mismatch
gpu.module @shuffles_mismatch {
|
silee2
approved these changes
Feb 27, 2026
adam-smnk
approved these changes
Mar 4, 2026
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
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.
This PR modifies gpu.shuffle lowering so only conduct the width check when the subgroupsize attribute is available.