Skip to content

[mlir][amdgpu] Sink op creation in scaled conversion intrinsics (NFC) #168542

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

Merged
merged 3 commits into from
Nov 18, 2025
Merged

[mlir][amdgpu] Sink op creation in scaled conversion intrinsics (NFC) #168542

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
3e2b65c
Sink op creation
amd-eochoalo Nov 18, 2025
d02904e
Move op.emitOpError below notifyMatchFailure
amd-eochoalo Nov 18, 2025
31de997
Update mlir/lib/Conversion/AMDGPUToROCDL/AMDGPUToROCDL.cpp
amd-eochoalo Nov 18, 2025
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
16 changes: 8 additions & 8 deletions mlir/lib/Conversion/AMDGPUToROCDL/AMDGPUToROCDL.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1738,15 +1738,11 @@ LogicalResult ScaledExtPacked816OpLowering::matchAndRewrite(
auto sourceType = cast<VectorType>(op.getSource().getType());
auto srcElemType = cast<FloatType>(sourceType.getElementType());
unsigned bitWidth = srcElemType.getWidth();
int32_t scaleSel =
getScaleSel(blockSize, bitWidth, firstScaleLane, firstScaleByte);

auto targetType = cast<VectorType>(op.getResult().getType());
auto destElemType = cast<FloatType>(targetType.getElementType());
IntegerType i32 = rewriter.getI32Type();
Value castedScale =
LLVM::BitcastOp::create(rewriter, loc, i32, adaptor.getScale());

IntegerType i32 = rewriter.getI32Type();
Value source = adaptor.getSource();
Type llvmResultType = typeConverter->convertType(op.getResult().getType());
Type packedType = nullptr;
Expand All @@ -1767,15 +1763,19 @@ LogicalResult ScaledExtPacked816OpLowering::matchAndRewrite(
return rewriter.notifyMatchFailure(op, "type conversion failed");
}

Value castedSource =
LLVM::BitcastOp::create(rewriter, loc, packedType, source);

std::optional<StringRef> maybeIntrinsic =
scaledExtPacked816ToIntrinsic(srcElemType, destElemType);
if (!maybeIntrinsic.has_value())
return op.emitOpError(
"no intrinsic matching packed scaled conversion on the given chipset");

int32_t scaleSel =
getScaleSel(blockSize, bitWidth, firstScaleLane, firstScaleByte);
Value castedScale =
LLVM::BitcastOp::create(rewriter, loc, i32, adaptor.getScale());
Value castedSource =
LLVM::BitcastOp::create(rewriter, loc, packedType, source);

OperationState loweredOp(loc, *maybeIntrinsic);
loweredOp.addTypes({llvmResultType});
loweredOp.addOperands({castedSource, castedScale});
Expand Down
Loading