[mlir] MemRefToSPIRV propagate alignment attributes from MemRef ops. #151723

amd-eochoalo · 2025-08-01T16:24:12Z

This patchset:

propagates alignment attributes from memref operations into the SPIR-V dialect,
fixes an error in the logic which previously propagated alignment attributes but did not add other MemoryAccess attributes.
adds a failure condition in the case where the alignment attribute from the memref dialect (64-bit wide) does not fit in SPIR-V's alignment attribute (specified to be 32-bit wide).

github-actions · 2025-08-01T16:24:32Z

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.

kuhar · 2025-08-01T18:35:09Z

It is unclear to me at the moment whether non-PhysicalStorageBuffers may have an alignment attribute. At the moment I am assuming they won't.

PSB requires alignment, and in the other storage classes alignment is optional.

The alignment attribute added to the SPIR-V operations is 32-bits wide while the alignment attribute in MemRef operations is 64-bits wide. I would recommend changing this to 64-bits per this comment. . However, the current definition constraints alignment to be 32 bits wide.

This is because the SPIR-V spec specifies alignment as a 32-bit power-of-two number.

There was an early return in calculateMemoryRequirements that looked explicitly for alignment and only set the alignment attribute. However, this was not correct for the following reasons: * Alignment was set only if both the alignment and the memory_access attributes were both present in the memref operation, without handling the case when only the alignment was exclusively present. * In the case alignment and memory_access attributes were both present, the memory_access attribute would not be updated to aligned if the memory_access attribute was not marked aligned. * In the case alignment and memory_access attributes were both present, other memory requirements (e.g., non_temporal) would not be added as attributes.

kuhar · 2025-08-01T19:24:10Z

mlir/test/Conversion/MemRefToSPIRV/memref-to-spirv.mlir

+  // CHECK: %[[VAL:.+]] = spirv.Load "StorageBuffer" %[[ADDR]] ["Aligned", 32] : i8
+  // CHECK: %[[ZERO_I8:.+]] = spirv.Constant 0 : i8
+  // CHECK: %[[BOOL:.+]] = spirv.INotEqual %[[VAL]], %[[ZERO_I8]] : i8
+  %0 = memref.load %src[%i] { alignment = 32 } : memref<4xi1, #spirv.storage_class<StorageBuffer>>


Can you also add one non-temporal load to see if both get propagated?

Thanks for the suggestion! I've added it here: 0337e22

llvmbot · 2025-08-01T19:39:02Z

@llvm/pr-subscribers-mlir

Author: Erick Ochoa Lopez (amd-eochoalo)

Changes

Propagating alignment attributes from memref operations into the SPIR-V dialect.

Full diff: https://github.com/llvm/llvm-project/pull/151723.diff

2 Files Affected:

(modified) mlir/lib/Conversion/MemRefToSPIRV/MemRefToSPIRV.cpp (+17-12)
(modified) mlir/test/Conversion/MemRefToSPIRV/memref-to-spirv.mlir (+45)

diff --git a/mlir/lib/Conversion/MemRefToSPIRV/MemRefToSPIRV.cpp b/mlir/lib/Conversion/MemRefToSPIRV/MemRefToSPIRV.cpp
index 7a705336bf11c..e730998f153b0 100644
--- a/mlir/lib/Conversion/MemRefToSPIRV/MemRefToSPIRV.cpp
+++ b/mlir/lib/Conversion/MemRefToSPIRV/MemRefToSPIRV.cpp
@@ -22,6 +22,7 @@
 #include "mlir/IR/MLIRContext.h"
 #include "mlir/IR/Visitors.h"
 #include <cassert>
+#include <limits>
 #include <optional>
 
 #define DEBUG_TYPE "memref-to-spirv-pattern"
@@ -465,7 +466,13 @@ struct MemoryRequirements {
 /// Given an accessed SPIR-V pointer, calculates its alignment requirements, if
 /// any.
 static FailureOr<MemoryRequirements>
-calculateMemoryRequirements(Value accessedPtr, bool isNontemporal) {
+calculateMemoryRequirements(Value accessedPtr, bool isNontemporal,
+                            uint64_t preferredAlignment) {
+
+  if (std::numeric_limits<uint32_t>::max() < preferredAlignment) {
+    return failure();
+  }
+
   MLIRContext *ctx = accessedPtr.getContext();
 
   auto memoryAccess = spirv::MemoryAccess::None;
@@ -474,7 +481,10 @@ calculateMemoryRequirements(Value accessedPtr, bool isNontemporal) {
   }
 
   auto ptrType = cast<spirv::PointerType>(accessedPtr.getType());
-  if (ptrType.getStorageClass() != spirv::StorageClass::PhysicalStorageBuffer) {
+  bool mayOmitAlignment =
+      !preferredAlignment &&
+      ptrType.getStorageClass() != spirv::StorageClass::PhysicalStorageBuffer;
+  if (mayOmitAlignment) {
     if (memoryAccess == spirv::MemoryAccess::None) {
       return MemoryRequirements{spirv::MemoryAccessAttr{}, IntegerAttr{}};
     }
@@ -483,6 +493,7 @@ calculateMemoryRequirements(Value accessedPtr, bool isNontemporal) {
   }
 
   // PhysicalStorageBuffers require the `Aligned` attribute.
+  // Other storage types may show an `Aligned` attribute.
   auto pointeeType = dyn_cast<spirv::ScalarType>(ptrType.getPointeeType());
   if (!pointeeType)
     return failure();
@@ -494,7 +505,8 @@ calculateMemoryRequirements(Value accessedPtr, bool isNontemporal) {
 
   memoryAccess = memoryAccess | spirv::MemoryAccess::Aligned;
   auto memAccessAttr = spirv::MemoryAccessAttr::get(ctx, memoryAccess);
-  auto alignment = IntegerAttr::get(IntegerType::get(ctx, 32), *sizeInBytes);
+  auto alignmentValue = preferredAlignment ? preferredAlignment : *sizeInBytes;
+  auto alignment = IntegerAttr::get(IntegerType::get(ctx, 32), alignmentValue);
   return MemoryRequirements{memAccessAttr, alignment};
 }
 
@@ -508,16 +520,9 @@ calculateMemoryRequirements(Value accessedPtr, LoadOrStoreOp loadOrStoreOp) {
       llvm::is_one_of<LoadOrStoreOp, memref::LoadOp, memref::StoreOp>::value,
       "Must be called on either memref::LoadOp or memref::StoreOp");
 
-  Operation *memrefAccessOp = loadOrStoreOp.getOperation();
-  auto memrefMemAccess = memrefAccessOp->getAttrOfType<spirv::MemoryAccessAttr>(
-      spirv::attributeName<spirv::MemoryAccess>());
-  auto memrefAlignment =
-      memrefAccessOp->getAttrOfType<IntegerAttr>("alignment");
-  if (memrefMemAccess && memrefAlignment)
-    return MemoryRequirements{memrefMemAccess, memrefAlignment};
-
   return calculateMemoryRequirements(accessedPtr,
-                                     loadOrStoreOp.getNontemporal());
+                                     loadOrStoreOp.getNontemporal(),
+                                     loadOrStoreOp.getAlignment().value_or(0));
 }
 
 LogicalResult
diff --git a/mlir/test/Conversion/MemRefToSPIRV/memref-to-spirv.mlir b/mlir/test/Conversion/MemRefToSPIRV/memref-to-spirv.mlir
index d0ddac8cd801c..7c765f70136bb 100644
--- a/mlir/test/Conversion/MemRefToSPIRV/memref-to-spirv.mlir
+++ b/mlir/test/Conversion/MemRefToSPIRV/memref-to-spirv.mlir
@@ -85,6 +85,51 @@ func.func @load_i1(%src: memref<4xi1, #spirv.storage_class<StorageBuffer>>, %i :
   return %0: i1
 }
 
+// CHECK-LABEL: func @load_aligned
+//  CHECK-SAME: (%[[SRC:.+]]: memref<4xi1, #spirv.storage_class<StorageBuffer>>, %[[IDX:.+]]: index)
+func.func @load_aligned(%src: memref<4xi1, #spirv.storage_class<StorageBuffer>>, %i : index) -> i1 {
+  // CHECK-DAG: %[[SRC_CAST:.+]] = builtin.unrealized_conversion_cast %[[SRC]] : memref<4xi1, #spirv.storage_class<StorageBuffer>> to !spirv.ptr<!spirv.struct<(!spirv.array<4 x i8, stride=1> [0])>, StorageBuffer>
+  // CHECK-DAG: %[[IDX_CAST:.+]] = builtin.unrealized_conversion_cast %[[IDX]]
+  // CHECK: %[[ZERO:.*]] = spirv.Constant 0 : i32
+  // CHECK: %[[ADDR:.+]] = spirv.AccessChain %[[SRC_CAST]][%[[ZERO]], %[[IDX_CAST]]]
+  // CHECK: %[[VAL:.+]] = spirv.Load "StorageBuffer" %[[ADDR]] ["Aligned", 32] : i8
+  // CHECK: %[[ZERO_I8:.+]] = spirv.Constant 0 : i8
+  // CHECK: %[[BOOL:.+]] = spirv.INotEqual %[[VAL]], %[[ZERO_I8]] : i8
+  %0 = memref.load %src[%i] { alignment = 32 } : memref<4xi1, #spirv.storage_class<StorageBuffer>>
+  // CHECK: return %[[BOOL]]
+  return %0: i1
+}
+
+// CHECK-LABEL: func @load_aligned_nontemporal
+//  CHECK-SAME: (%[[SRC:.+]]: memref<4xi1, #spirv.storage_class<StorageBuffer>>, %[[IDX:.+]]: index)
+func.func @load_aligned_nontemporal(%src: memref<4xi1, #spirv.storage_class<StorageBuffer>>, %i : index) -> i1 {
+  // CHECK-DAG: %[[SRC_CAST:.+]] = builtin.unrealized_conversion_cast %[[SRC]] : memref<4xi1, #spirv.storage_class<StorageBuffer>> to !spirv.ptr<!spirv.struct<(!spirv.array<4 x i8, stride=1> [0])>, StorageBuffer>
+  // CHECK-DAG: %[[IDX_CAST:.+]] = builtin.unrealized_conversion_cast %[[IDX]]
+  // CHECK: %[[ZERO:.*]] = spirv.Constant 0 : i32
+  // CHECK: %[[ADDR:.+]] = spirv.AccessChain %[[SRC_CAST]][%[[ZERO]], %[[IDX_CAST]]]
+  // CHECK: %[[VAL:.+]] = spirv.Load "StorageBuffer" %[[ADDR]] ["Aligned|Nontemporal", 32] : i8
+  // CHECK: %[[ZERO_I8:.+]] = spirv.Constant 0 : i8
+  // CHECK: %[[BOOL:.+]] = spirv.INotEqual %[[VAL]], %[[ZERO_I8]] : i8
+  %0 = memref.load %src[%i] { alignment = 32, nontemporal = true } : memref<4xi1, #spirv.storage_class<StorageBuffer>>
+  // CHECK: return %[[BOOL]]
+  return %0: i1
+}
+
+// CHECK-LABEL: func @load_aligned_psb
+//  CHECK-SAME: (%[[SRC:.+]]: memref<4xi1, #spirv.storage_class<PhysicalStorageBuffer>>, %[[IDX:.+]]: index)
+func.func @load_aligned_psb(%src: memref<4xi1, #spirv.storage_class<PhysicalStorageBuffer>>, %i : index) -> i1 {
+  // CHECK-DAG: %[[SRC_CAST:.+]] = builtin.unrealized_conversion_cast %[[SRC]] : memref<4xi1, #spirv.storage_class<PhysicalStorageBuffer>> to !spirv.ptr<!spirv.struct<(!spirv.array<4 x i8, stride=1> [0])>, PhysicalStorageBuffer>
+  // CHECK-DAG: %[[IDX_CAST:.+]] = builtin.unrealized_conversion_cast %[[IDX]]
+  // CHECK: %[[ZERO:.*]] = spirv.Constant 0 : i32
+  // CHECK: %[[ADDR:.+]] = spirv.AccessChain %[[SRC_CAST]][%[[ZERO]], %[[IDX_CAST]]]
+  // CHECK: %[[VAL:.+]] = spirv.Load "PhysicalStorageBuffer" %[[ADDR]] ["Aligned", 32] : i8
+  // CHECK: %[[ZERO_I8:.+]] = spirv.Constant 0 : i8
+  // CHECK: %[[BOOL:.+]] = spirv.INotEqual %[[VAL]], %[[ZERO_I8]] : i8
+  %0 = memref.load %src[%i] { alignment = 32 } : memref<4xi1, #spirv.storage_class<PhysicalStorageBuffer>>
+  // CHECK: return %[[BOOL]]
+  return %0: i1
+}
+
 // CHECK-LABEL: func @store_i1
 //  CHECK-SAME: %[[DST:.+]]: memref<4xi1, #spirv.storage_class<StorageBuffer>>,
 //  CHECK-SAME: %[[IDX:.+]]: index

llvmbot · 2025-08-01T19:39:02Z

@llvm/pr-subscribers-mlir-spirv

Author: Erick Ochoa Lopez (amd-eochoalo)

Changes

Propagating alignment attributes from memref operations into the SPIR-V dialect.

Full diff: https://github.com/llvm/llvm-project/pull/151723.diff

2 Files Affected:

(modified) mlir/lib/Conversion/MemRefToSPIRV/MemRefToSPIRV.cpp (+17-12)
(modified) mlir/test/Conversion/MemRefToSPIRV/memref-to-spirv.mlir (+45)

diff --git a/mlir/lib/Conversion/MemRefToSPIRV/MemRefToSPIRV.cpp b/mlir/lib/Conversion/MemRefToSPIRV/MemRefToSPIRV.cpp
index 7a705336bf11c..e730998f153b0 100644
--- a/mlir/lib/Conversion/MemRefToSPIRV/MemRefToSPIRV.cpp
+++ b/mlir/lib/Conversion/MemRefToSPIRV/MemRefToSPIRV.cpp
@@ -22,6 +22,7 @@
 #include "mlir/IR/MLIRContext.h"
 #include "mlir/IR/Visitors.h"
 #include <cassert>
+#include <limits>
 #include <optional>
 
 #define DEBUG_TYPE "memref-to-spirv-pattern"
@@ -465,7 +466,13 @@ struct MemoryRequirements {
 /// Given an accessed SPIR-V pointer, calculates its alignment requirements, if
 /// any.
 static FailureOr<MemoryRequirements>
-calculateMemoryRequirements(Value accessedPtr, bool isNontemporal) {
+calculateMemoryRequirements(Value accessedPtr, bool isNontemporal,
+                            uint64_t preferredAlignment) {
+
+  if (std::numeric_limits<uint32_t>::max() < preferredAlignment) {
+    return failure();
+  }
+
   MLIRContext *ctx = accessedPtr.getContext();
 
   auto memoryAccess = spirv::MemoryAccess::None;
@@ -474,7 +481,10 @@ calculateMemoryRequirements(Value accessedPtr, bool isNontemporal) {
   }
 
   auto ptrType = cast<spirv::PointerType>(accessedPtr.getType());
-  if (ptrType.getStorageClass() != spirv::StorageClass::PhysicalStorageBuffer) {
+  bool mayOmitAlignment =
+      !preferredAlignment &&
+      ptrType.getStorageClass() != spirv::StorageClass::PhysicalStorageBuffer;
+  if (mayOmitAlignment) {
     if (memoryAccess == spirv::MemoryAccess::None) {
       return MemoryRequirements{spirv::MemoryAccessAttr{}, IntegerAttr{}};
     }
@@ -483,6 +493,7 @@ calculateMemoryRequirements(Value accessedPtr, bool isNontemporal) {
   }
 
   // PhysicalStorageBuffers require the `Aligned` attribute.
+  // Other storage types may show an `Aligned` attribute.
   auto pointeeType = dyn_cast<spirv::ScalarType>(ptrType.getPointeeType());
   if (!pointeeType)
     return failure();
@@ -494,7 +505,8 @@ calculateMemoryRequirements(Value accessedPtr, bool isNontemporal) {
 
   memoryAccess = memoryAccess | spirv::MemoryAccess::Aligned;
   auto memAccessAttr = spirv::MemoryAccessAttr::get(ctx, memoryAccess);
-  auto alignment = IntegerAttr::get(IntegerType::get(ctx, 32), *sizeInBytes);
+  auto alignmentValue = preferredAlignment ? preferredAlignment : *sizeInBytes;
+  auto alignment = IntegerAttr::get(IntegerType::get(ctx, 32), alignmentValue);
   return MemoryRequirements{memAccessAttr, alignment};
 }
 
@@ -508,16 +520,9 @@ calculateMemoryRequirements(Value accessedPtr, LoadOrStoreOp loadOrStoreOp) {
       llvm::is_one_of<LoadOrStoreOp, memref::LoadOp, memref::StoreOp>::value,
       "Must be called on either memref::LoadOp or memref::StoreOp");
 
-  Operation *memrefAccessOp = loadOrStoreOp.getOperation();
-  auto memrefMemAccess = memrefAccessOp->getAttrOfType<spirv::MemoryAccessAttr>(
-      spirv::attributeName<spirv::MemoryAccess>());
-  auto memrefAlignment =
-      memrefAccessOp->getAttrOfType<IntegerAttr>("alignment");
-  if (memrefMemAccess && memrefAlignment)
-    return MemoryRequirements{memrefMemAccess, memrefAlignment};
-
   return calculateMemoryRequirements(accessedPtr,
-                                     loadOrStoreOp.getNontemporal());
+                                     loadOrStoreOp.getNontemporal(),
+                                     loadOrStoreOp.getAlignment().value_or(0));
 }
 
 LogicalResult
diff --git a/mlir/test/Conversion/MemRefToSPIRV/memref-to-spirv.mlir b/mlir/test/Conversion/MemRefToSPIRV/memref-to-spirv.mlir
index d0ddac8cd801c..7c765f70136bb 100644
--- a/mlir/test/Conversion/MemRefToSPIRV/memref-to-spirv.mlir
+++ b/mlir/test/Conversion/MemRefToSPIRV/memref-to-spirv.mlir
@@ -85,6 +85,51 @@ func.func @load_i1(%src: memref<4xi1, #spirv.storage_class<StorageBuffer>>, %i :
   return %0: i1
 }
 
+// CHECK-LABEL: func @load_aligned
+//  CHECK-SAME: (%[[SRC:.+]]: memref<4xi1, #spirv.storage_class<StorageBuffer>>, %[[IDX:.+]]: index)
+func.func @load_aligned(%src: memref<4xi1, #spirv.storage_class<StorageBuffer>>, %i : index) -> i1 {
+  // CHECK-DAG: %[[SRC_CAST:.+]] = builtin.unrealized_conversion_cast %[[SRC]] : memref<4xi1, #spirv.storage_class<StorageBuffer>> to !spirv.ptr<!spirv.struct<(!spirv.array<4 x i8, stride=1> [0])>, StorageBuffer>
+  // CHECK-DAG: %[[IDX_CAST:.+]] = builtin.unrealized_conversion_cast %[[IDX]]
+  // CHECK: %[[ZERO:.*]] = spirv.Constant 0 : i32
+  // CHECK: %[[ADDR:.+]] = spirv.AccessChain %[[SRC_CAST]][%[[ZERO]], %[[IDX_CAST]]]
+  // CHECK: %[[VAL:.+]] = spirv.Load "StorageBuffer" %[[ADDR]] ["Aligned", 32] : i8
+  // CHECK: %[[ZERO_I8:.+]] = spirv.Constant 0 : i8
+  // CHECK: %[[BOOL:.+]] = spirv.INotEqual %[[VAL]], %[[ZERO_I8]] : i8
+  %0 = memref.load %src[%i] { alignment = 32 } : memref<4xi1, #spirv.storage_class<StorageBuffer>>
+  // CHECK: return %[[BOOL]]
+  return %0: i1
+}
+
+// CHECK-LABEL: func @load_aligned_nontemporal
+//  CHECK-SAME: (%[[SRC:.+]]: memref<4xi1, #spirv.storage_class<StorageBuffer>>, %[[IDX:.+]]: index)
+func.func @load_aligned_nontemporal(%src: memref<4xi1, #spirv.storage_class<StorageBuffer>>, %i : index) -> i1 {
+  // CHECK-DAG: %[[SRC_CAST:.+]] = builtin.unrealized_conversion_cast %[[SRC]] : memref<4xi1, #spirv.storage_class<StorageBuffer>> to !spirv.ptr<!spirv.struct<(!spirv.array<4 x i8, stride=1> [0])>, StorageBuffer>
+  // CHECK-DAG: %[[IDX_CAST:.+]] = builtin.unrealized_conversion_cast %[[IDX]]
+  // CHECK: %[[ZERO:.*]] = spirv.Constant 0 : i32
+  // CHECK: %[[ADDR:.+]] = spirv.AccessChain %[[SRC_CAST]][%[[ZERO]], %[[IDX_CAST]]]
+  // CHECK: %[[VAL:.+]] = spirv.Load "StorageBuffer" %[[ADDR]] ["Aligned|Nontemporal", 32] : i8
+  // CHECK: %[[ZERO_I8:.+]] = spirv.Constant 0 : i8
+  // CHECK: %[[BOOL:.+]] = spirv.INotEqual %[[VAL]], %[[ZERO_I8]] : i8
+  %0 = memref.load %src[%i] { alignment = 32, nontemporal = true } : memref<4xi1, #spirv.storage_class<StorageBuffer>>
+  // CHECK: return %[[BOOL]]
+  return %0: i1
+}
+
+// CHECK-LABEL: func @load_aligned_psb
+//  CHECK-SAME: (%[[SRC:.+]]: memref<4xi1, #spirv.storage_class<PhysicalStorageBuffer>>, %[[IDX:.+]]: index)
+func.func @load_aligned_psb(%src: memref<4xi1, #spirv.storage_class<PhysicalStorageBuffer>>, %i : index) -> i1 {
+  // CHECK-DAG: %[[SRC_CAST:.+]] = builtin.unrealized_conversion_cast %[[SRC]] : memref<4xi1, #spirv.storage_class<PhysicalStorageBuffer>> to !spirv.ptr<!spirv.struct<(!spirv.array<4 x i8, stride=1> [0])>, PhysicalStorageBuffer>
+  // CHECK-DAG: %[[IDX_CAST:.+]] = builtin.unrealized_conversion_cast %[[IDX]]
+  // CHECK: %[[ZERO:.*]] = spirv.Constant 0 : i32
+  // CHECK: %[[ADDR:.+]] = spirv.AccessChain %[[SRC_CAST]][%[[ZERO]], %[[IDX_CAST]]]
+  // CHECK: %[[VAL:.+]] = spirv.Load "PhysicalStorageBuffer" %[[ADDR]] ["Aligned", 32] : i8
+  // CHECK: %[[ZERO_I8:.+]] = spirv.Constant 0 : i8
+  // CHECK: %[[BOOL:.+]] = spirv.INotEqual %[[VAL]], %[[ZERO_I8]] : i8
+  %0 = memref.load %src[%i] { alignment = 32 } : memref<4xi1, #spirv.storage_class<PhysicalStorageBuffer>>
+  // CHECK: return %[[BOOL]]
+  return %0: i1
+}
+
 // CHECK-LABEL: func @store_i1
 //  CHECK-SAME: %[[DST:.+]]: memref<4xi1, #spirv.storage_class<StorageBuffer>>,
 //  CHECK-SAME: %[[IDX:.+]]: index

amd-eochoalo added 2 commits August 1, 2025 11:51

[mlir] Use memref's alignment attribute directly.

f4664aa

[mlir] MemRefToSPIRV propagate alignment attribute.

1c3455c

kuhar reviewed Aug 1, 2025

View reviewed changes

amd-eochoalo added 2 commits August 1, 2025 15:32

[mlir] Ensure memref's alignment is within SPIRV's alignment bounds.

04a811b

[mlir] Add aligned nontemporal load test to MemRefToSPIRV.

0337e22

amd-eochoalo marked this pull request as ready for review August 1, 2025 19:38

amd-eochoalo requested a review from antiagainst as a code owner August 1, 2025 19:38

llvmbot added mlir:spirv mlir labels Aug 1, 2025

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[mlir] MemRefToSPIRV propagate alignment attributes from MemRef ops. #151723

[mlir] MemRefToSPIRV propagate alignment attributes from MemRef ops. #151723

Uh oh!

amd-eochoalo commented Aug 1, 2025 •

edited

Loading

Uh oh!

github-actions bot commented Aug 1, 2025

Uh oh!

kuhar commented Aug 1, 2025

Uh oh!

kuhar Aug 1, 2025

Uh oh!

amd-eochoalo Aug 1, 2025

Uh oh!

llvmbot commented Aug 1, 2025

Uh oh!

llvmbot commented Aug 1, 2025

Uh oh!

Uh oh!

[mlir] MemRefToSPIRV propagate alignment attributes from MemRef ops. #151723

Are you sure you want to change the base?

[mlir] MemRefToSPIRV propagate alignment attributes from MemRef ops. #151723

Uh oh!

Conversation

amd-eochoalo commented Aug 1, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions bot commented Aug 1, 2025

Uh oh!

kuhar commented Aug 1, 2025

Uh oh!

kuhar Aug 1, 2025

Choose a reason for hiding this comment

Uh oh!

amd-eochoalo Aug 1, 2025

Choose a reason for hiding this comment

Uh oh!

llvmbot commented Aug 1, 2025

Uh oh!

llvmbot commented Aug 1, 2025

Uh oh!

Uh oh!

amd-eochoalo commented Aug 1, 2025 •

edited

Loading