[mlir][vector] Fix attaching write effects on transfer_write's base #142940

simpel01 · 2025-06-05T10:17:31Z

This fixes an issue with TransferWriteOp's implementation of the MemoryEffectOpInterface where the write effect was attached to the stored value rather than the base.

This had the effect that when asking for the memory effects for the input memref buffer using getEffectsOnValue(...), the function would return no-effects (as the effect would have been attached to the stored value rather than the input buffer).

github-actions · 2025-06-05T10:17:51Z

Thank you for submitting a Pull Request (PR) to the LLVM Project!

This PR will be automatically labeled and the relevant teams will be notified.

If you wish to, you can add reviewers by using the "Reviewers" section on this page.

If this is not working for you, it is probably because you do not have write permissions for the repository. In which case you can instead tag reviewers by name in a comment by using @ followed by their GitHub username.

If you have received no comments on your PR for a week, you can request a review by "ping"ing the PR by adding a comment “Ping”. The common courtesy "ping" rate is once a week. Please remember that you are asking for valuable time from other developers.

If you have further questions, they may be answered by the LLVM GitHub User Guide.

You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums.

llvmbot · 2025-06-05T10:18:32Z

@llvm/pr-subscribers-mlir
@llvm/pr-subscribers-mlir-bufferization

@llvm/pr-subscribers-mlir-vector

Author: Simone Pellegrini (simpel01)

Changes

This fixes an issue with TransferWriteOp's implementation of the MemoryEffectOpInterface where the write effect was attached to the stored value rather than the base.

This had the effect that when asking for the memory effects for the input memref buffer using getEffectsOnValue(...), the function would return no-effects (as the effect would have been attached to the stored value rather than the input buffer).

Full diff: https://github.com/llvm/llvm-project/pull/142940.diff

5 Files Affected:

(modified) mlir/lib/Dialect/Vector/IR/VectorOps.cpp (+1-1)
(modified) mlir/test/Dialect/Bufferization/side-effects.mlir (+3-3)
(added) mlir/test/Dialect/Vector/side-effects.mlir (+17)
(modified) mlir/test/IR/test-side-effects.mlir (+4-4)
(modified) mlir/test/lib/IR/TestSideEffects.cpp (+6-6)

diff --git a/mlir/lib/Dialect/Vector/IR/VectorOps.cpp b/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
index fcfb401fd9867..bd926b82cf14f 100644
--- a/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
+++ b/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
@@ -5034,7 +5034,7 @@ void TransferWriteOp::getEffects(
     SmallVectorImpl<SideEffects::EffectInstance<MemoryEffects::Effect>>
         &effects) {
   if (llvm::isa<MemRefType>(getShapedType()))
-    effects.emplace_back(MemoryEffects::Write::get(), &getValueToStoreMutable(),
+    effects.emplace_back(MemoryEffects::Write::get(), &getBaseMutable(),
                          SideEffects::DefaultResource::get());
 }
 
diff --git a/mlir/test/Dialect/Bufferization/side-effects.mlir b/mlir/test/Dialect/Bufferization/side-effects.mlir
index 841490e9f3234..129fc8b32c270 100644
--- a/mlir/test/Dialect/Bufferization/side-effects.mlir
+++ b/mlir/test/Dialect/Bufferization/side-effects.mlir
@@ -1,9 +1,9 @@
 // RUN: mlir-opt %s --test-side-effects --verify-diagnostics
 
 func.func @test_side_effects(%arg0: memref<2xi32>) -> memref<2xi32> {
-  // expected-remark @below {{found an instance of 'read' on a op operand, on resource '<Default>'}}
-  // expected-remark @below {{found an instance of 'write' on a op result, on resource '<Default>'}}
-  // expected-remark @below {{found an instance of 'allocate' on a op result, on resource '<Default>'}}
+  // expected-remark @below {{found an instance of 'read' on op operand 0, on resource '<Default>'}}
+  // expected-remark @below {{found an instance of 'write' on op result 0, on resource '<Default>'}}
+  // expected-remark @below {{found an instance of 'allocate' on op result 0, on resource '<Default>'}}
   %0 = bufferization.clone %arg0 : memref<2xi32> to memref<2xi32>
   return %0 : memref<2xi32>
 }
diff --git a/mlir/test/Dialect/Vector/side-effects.mlir b/mlir/test/Dialect/Vector/side-effects.mlir
new file mode 100644
index 0000000000000..a1c205c2e3228
--- /dev/null
+++ b/mlir/test/Dialect/Vector/side-effects.mlir
@@ -0,0 +1,17 @@
+// RUN: mlir-opt %s --test-side-effects --verify-diagnostics
+
+func.func @test_side_effects(%arg0: memref<8xf32>) {
+  // expected-remark @below {{operation has no memory effects}}
+  %c0 = arith.constant 0 : index
+  // expected-remark @below {{operation has no memory effects}}
+  %c4 = arith.constant 4 : index
+  // expected-remark @below {{operation has no memory effects}}
+  %cst = arith.constant 0.0 : f32
+  // expected-remark @below {{operation has no memory effects}}
+  %vec_cst = arith.constant dense<0.0> : vector<4xf32>
+  // expected-remark @below {{found an instance of 'read' on op operand 0, on resource '<Default>'}}
+  %0 = vector.transfer_read %arg0[%c0], %cst : memref<8xf32>, vector<4xf32>
+  // expected-remark @below {{found an instance of 'write' on op operand 1, on resource '<Default>'}}
+  vector.transfer_write %vec_cst, %arg0[%c4] : vector<4xf32>, memref<8xf32>
+  return
+}
diff --git a/mlir/test/IR/test-side-effects.mlir b/mlir/test/IR/test-side-effects.mlir
index efce4856041a1..b652ecb7dad1d 100644
--- a/mlir/test/IR/test-side-effects.mlir
+++ b/mlir/test/IR/test-side-effects.mlir
@@ -15,7 +15,7 @@ func.func @side_effect(%arg : index) {
     {effect="write", test_resource}
   ]} : () -> i32
   
-  // expected-remark@+1 {{found an instance of 'allocate' on a op result, on resource '<Test>'}}
+  // expected-remark@+1 {{found an instance of 'allocate' on op result 0, on resource '<Test>'}}
   %3 = "test.side_effect_op"() {effects = [
     {effect="allocate", on_result, test_resource}
   ]} : () -> i32
@@ -38,19 +38,19 @@ func.func @side_effect(%arg : index) {
     effect_parameter = affine_map<(i, j) -> (j, i)>
   } : () -> i32
 
-  // expected-remark@+1 {{found an instance of 'allocate' on a op operand, on resource '<Test>'}}
+  // expected-remark@+1 {{found an instance of 'allocate' on op operand 0, on resource '<Test>'}}
   %6 = test.side_effect_with_region_op (%arg) {
   ^bb0(%arg0 : index):
     test.region_yield %arg0 : index 
   } {effects = [ {effect="allocate", on_operand, test_resource} ]} : index -> index
 
-  // expected-remark@+1 {{found an instance of 'allocate' on a op result, on resource '<Test>'}}
+  // expected-remark@+1 {{found an instance of 'allocate' on op result 0, on resource '<Test>'}}
   %7 = test.side_effect_with_region_op (%arg) {
   ^bb0(%arg0 : index):
     test.region_yield %arg0 : index 
   } {effects = [ {effect="allocate", on_result, test_resource} ]} : index -> index
 
-  // expected-remark@+1 {{found an instance of 'allocate' on a block argument, on resource '<Test>'}}
+  // expected-remark@+1 {{found an instance of 'allocate' on block argument 0, on resource '<Test>'}}
   %8 = test.side_effect_with_region_op (%arg) {
   ^bb0(%arg0 : index):
     test.region_yield %arg0 : index 
diff --git a/mlir/test/lib/IR/TestSideEffects.cpp b/mlir/test/lib/IR/TestSideEffects.cpp
index 7e01509d55685..000e7c204fd5f 100644
--- a/mlir/test/lib/IR/TestSideEffects.cpp
+++ b/mlir/test/lib/IR/TestSideEffects.cpp
@@ -52,12 +52,12 @@ struct SideEffectsPass
           diag << "'write'";
 
         if (instance.getValue()) {
-          if (instance.getEffectValue<OpOperand *>())
-            diag << " on a op operand,";
-          else if (instance.getEffectValue<OpResult>())
-            diag << " on a op result,";
-          else if (instance.getEffectValue<BlockArgument>())
-            diag << " on a block argument,";
+          if (auto *opOpd = instance.getEffectValue<OpOperand *>())
+            diag << " on op operand " << opOpd->getOperandNumber() << ",";
+          else if (auto opRes = instance.getEffectValue<OpResult>())
+            diag << " on op result " << opRes.getResultNumber() << ",";
+          else if (auto opBlk = instance.getEffectValue<BlockArgument>())
+            diag << " on block argument " << opBlk.getArgNumber() << ",";
         } else if (SymbolRefAttr symbolRef = instance.getSymbolRef())
           diag << " on a symbol '" << symbolRef << "',";

simpel01 · 2025-06-05T11:03:18Z

mlir/test/Dialect/Vector/side-effects.mlir

+  %vec_cst = arith.constant dense<0.0> : vector<4xf32>
+  // expected-remark @below {{found an instance of 'read' on op operand 0, on resource '<Default>'}}
+  %0 = vector.transfer_read %arg0[%c0], %cst : memref<8xf32>, vector<4xf32>
+  // expected-remark @below {{found an instance of 'write' on op operand 1, on resource '<Default>'}}


Without the changes to the TransferWriteOp::getEffects(...) function the write effect would be attached to op operand 0 rather than on 1.

matthias-springer

It looks like this is breaking a Flang test.

This fixes an issue with `TransferWriteOp`'s implementation of the `MemoryEffectOpInterface` where the write effect was attached to the stored value rather than the base. This had the effect that when asking for the memory effects for the input memref buffer using `getEffectsOnValue(...)`, the function would return no-effects (as the effect would have been attached to the stored value rather than the input buffer).

simpel01 · 2025-06-05T14:29:22Z

It looks like this is breaking a Flang test.

Yes, sorry about that. I didn't expect changes to the mlir module would affect flang tests.
Either way, I pushed an update and tests seems to be passing now.

dcaballe

Thanks!

github-actions · 2025-06-11T11:37:53Z

@simpel01 Congratulations on having your first Pull Request (PR) merged into the LLVM Project!

Your changes will be combined with recent changes from other authors, then tested by our build bots. If there is a problem with a build, you may receive a report in an email or a comment on this PR.

Please check whether problems have been caused by your change specifically, as the builds can include changes from many authors. It is not uncommon for your change to be included in a build that fails due to someone else's changes, or infrastructure issues.

How to do this, and the rest of the post-merge process, is covered in detail here.

If your change does cause a problem, it may be reverted, or you can revert it yourself. This is a normal part of LLVM development. You can fix your changes and open a new PR to merge them again.

If you don't get any reports, no action is required from you. Your changes are working as expected, well done!

…lvm#142940) This fixes an issue with `TransferWriteOp`'s implementation of the `MemoryEffectOpInterface` where the write effect was attached to the stored value rather than the base. This had the effect that when asking for the memory effects for the input memref buffer using `getEffectsOnValue(...)`, the function would return no-effects (as the effect would have been attached to the stored value rather than the input buffer).

simpel01 requested review from banach-space, dcaballe, nicolasvasilache and Groverkss as code owners June 5, 2025 10:17

llvmbot added mlir:vectorops mlir mlir:bufferization Bufferization infrastructure mlir:vector labels Jun 5, 2025

simpel01 commented Jun 5, 2025

View reviewed changes

matthias-springer approved these changes Jun 5, 2025

View reviewed changes

matthias-springer reviewed Jun 5, 2025

View reviewed changes

simpel01 force-pushed the vector-transfer-write-effects branch from 68996d3 to 884beda Compare June 5, 2025 13:10

llvmbot added the flang Flang issues not falling into any other category label Jun 5, 2025

dcaballe approved these changes Jun 9, 2025

View reviewed changes

RoboTux merged commit abbbe4a into llvm:main Jun 11, 2025
12 checks passed

simpel01 deleted the vector-transfer-write-effects branch June 13, 2025 07:24

arjunUpatel mentioned this pull request Jun 24, 2025

[lldb] Revive TestSimulatorPlatform.py (#142244) arjunUpatel/llvm-project#1

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[mlir][vector] Fix attaching write effects on transfer_write's base #142940

[mlir][vector] Fix attaching write effects on transfer_write's base #142940

Uh oh!

simpel01 commented Jun 5, 2025

Uh oh!

github-actions bot commented Jun 5, 2025

Uh oh!

llvmbot commented Jun 5, 2025 •

edited

Loading

Uh oh!

simpel01 Jun 5, 2025

Uh oh!

matthias-springer left a comment

Uh oh!

simpel01 commented Jun 5, 2025

Uh oh!

dcaballe left a comment

Uh oh!

Uh oh!

github-actions bot commented Jun 11, 2025

Uh oh!

Uh oh!

[mlir][vector] Fix attaching write effects on transfer_write's base #142940

[mlir][vector] Fix attaching write effects on transfer_write's base #142940

Uh oh!

Conversation

simpel01 commented Jun 5, 2025

Uh oh!

github-actions bot commented Jun 5, 2025

Uh oh!

llvmbot commented Jun 5, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

simpel01 Jun 5, 2025

Choose a reason for hiding this comment

Uh oh!

matthias-springer left a comment

Choose a reason for hiding this comment

Uh oh!

simpel01 commented Jun 5, 2025

Uh oh!

dcaballe left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

github-actions bot commented Jun 11, 2025

Uh oh!

Uh oh!

llvmbot commented Jun 5, 2025 •

edited

Loading