Skip to content

[mlir][SPIR-V] Support floating-point atomic_rmw addf in MemRefToSPIRV #202330

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
aobolensk wants to merge 2 commits into
base: main
Choose a base branch
from
+25 −8
Open

[mlir][SPIR-V] Support floating-point atomic_rmw addf in MemRefToSPIRV #202330

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
c3e9212
[mlir][SPIR-V] Support floating-point atomic_rmw addf in MemRefToSPIRV
aobolensk Jun 8, 2026
f6ac9d2
Merge branch 'main' into mlir-spirv-atomic-fadd
aobolensk Jun 8, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 7 additions & 8 deletions mlir/lib/Conversion/MemRefToSPIRV/MemRefToSPIRV.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -454,10 +454,6 @@ LogicalResult
AtomicRMWOpPattern::matchAndRewrite(memref::AtomicRMWOp atomicOp,
OpAdaptor adaptor,
ConversionPatternRewriter &rewriter) const {
if (isa<FloatType>(atomicOp.getType()))
return rewriter.notifyMatchFailure(atomicOp,
"unimplemented floating-point case");

auto memrefType = cast<MemRefType>(atomicOp.getMemref().getType());
std::optional<spirv::Scope> scope = getAtomicOpScope(memrefType);
if (!scope)
Expand Down Expand Up @@ -488,13 +484,13 @@ AtomicRMWOpPattern::matchAndRewrite(memref::AtomicRMWOp atomicOp,
"failed to convert memref type");

Type pointeeType = pointerType.getPointeeType();
auto dstType = dyn_cast<IntegerType>(
getElementTypeForStoragePointer(pointeeType, typeConverter));
if (!dstType)
Type storageElemType =
getElementTypeForStoragePointer(pointeeType, typeConverter);
if (!storageElemType || !storageElemType.isIntOrFloat())
return rewriter.notifyMatchFailure(
atomicOp, "failed to determine destination element type");

int dstBits = static_cast<int>(dstType.getWidth());
int dstBits = static_cast<int>(storageElemType.getIntOrFloatBitWidth());
assert(dstBits % srcBits == 0);

spirv::MemorySemantics memSem = getAtomicAcqRelMemorySemantics(memrefType);
Expand All @@ -509,6 +505,7 @@ AtomicRMWOpPattern::matchAndRewrite(memref::AtomicRMWOp atomicOp,
break

switch (atomicOp.getKind()) {
ATOMIC_CASE(addf, EXTAtomicFAddOp);
ATOMIC_CASE(addi, AtomicIAddOp);
ATOMIC_CASE(maxs, AtomicSMaxOp);
ATOMIC_CASE(maxu, AtomicUMaxOp);
Expand Down Expand Up @@ -546,6 +543,8 @@ AtomicRMWOpPattern::matchAndRewrite(memref::AtomicRMWOp atomicOp,
atomicOp,
"sub-element-width atomic ops unsupported with Kernel capability");

auto dstType = cast<IntegerType>(storageElemType);

@IgWod IgWod Jun 8, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why is it safe to cast float to int as the dstType?

@aobolensk aobolensk Jun 8, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Integers are handled above


auto accessChainOp = ptr.getDefiningOp<spirv::AccessChainOp>();
if (!accessChainOp)
return failure();
Expand Down
18 changes: 18 additions & 0 deletions mlir/test/Conversion/MemRefToSPIRV/atomic.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -146,3 +146,21 @@ func.func @atomic_andi_i8_storage_buffer(%value: i8, %memref: memref<16xi8, #spi
}

}

// -----

// Floating-point atomic add requires the shader_atomic_float_add extension.

module attributes {spirv.target_env = #spirv.target_env<#spirv.vce<v1.3, [Shader, AtomicFloat32AddEXT], [SPV_EXT_shader_atomic_float_add]>, #spirv.resource_limits<>>} {

// CHECK: func.func @atomic_addf_storage_buffer
// CHECK-SAME: (%[[VAL:.+]]: f32,
func.func @atomic_addf_storage_buffer(%value: f32, %memref: memref<2x3x4xf32, #spirv.storage_class<StorageBuffer>>, %i0: index, %i1: index, %i2: index) -> f32 {
// CHECK: %[[AC:.+]] = spirv.AccessChain
// CHECK: %[[ATOMIC:.+]] = spirv.EXT.AtomicFAdd <Device> <AcquireRelease|UniformMemory> %[[AC]], %[[VAL]] : !spirv.ptr<f32, StorageBuffer>
// CHECK: return %[[ATOMIC]]
%0 = memref.atomic_rmw "addf" %value, %memref[%i0, %i1, %i2] : (f32, memref<2x3x4xf32, #spirv.storage_class<StorageBuffer>>) -> f32
return %0: f32
}

}
Loading