Skip to content

[flang][cuda] Avoid stack corruption when setting kernel launch parameters #119469

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 1 commit into from
Dec 11, 2024
Merged

[flang][cuda] Avoid stack corruption when setting kernel launch parameters #119469

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
5 changes: 4 additions & 1 deletion flang/lib/Optimizer/Transforms/CUFGPUToLLVMConversion.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -42,6 +42,8 @@ static mlir::Value createKernelArgArray(mlir::Location loc,
auto structTy = mlir::LLVM::LLVMStructType::getLiteral(ctx, structTypes);
auto ptrTy = mlir::LLVM::LLVMPointerType::get(rewriter.getContext());
mlir::Type i32Ty = rewriter.getI32Type();
auto zero = rewriter.create<mlir::LLVM::ConstantOp>(
loc, i32Ty, rewriter.getIntegerAttr(i32Ty, 0));
auto one = rewriter.create<mlir::LLVM::ConstantOp>(
loc, i32Ty, rewriter.getIntegerAttr(i32Ty, 1));
mlir::Value argStruct =
Expand All @@ -55,7 +57,8 @@ static mlir::Value createKernelArgArray(mlir::Location loc,
auto indice = rewriter.create<mlir::LLVM::ConstantOp>(
loc, i32Ty, rewriter.getIntegerAttr(i32Ty, i));
mlir::Value structMember = rewriter.create<LLVM::GEPOp>(
loc, ptrTy, structTy, argStruct, mlir::ArrayRef<mlir::Value>({indice}));
loc, ptrTy, structTy, argStruct,
mlir::ArrayRef<mlir::Value>({zero, indice}));
rewriter.create<LLVM::StoreOp>(loc, arg, structMember);
mlir::Value arrayMember = rewriter.create<LLVM::GEPOp>(
loc, ptrTy, ptrTy, argArray, mlir::ArrayRef<mlir::Value>({indice}));
Expand Down
2 changes: 1 addition & 1 deletion flang/test/Fir/CUDA/cuda-gpu-launch-func.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -102,7 +102,7 @@ module attributes {dlti.dl_spec = #dlti.dl_spec<#dlti.dl_entry<i1, dense<8> : ve
// CHECK: %[[STRUCT:.*]] = llvm.alloca %{{.*}} x !llvm.struct<(ptr)> : (i32) -> !llvm.ptr
// CHECK: %[[PARAMS:.*]] = llvm.alloca %{{.*}} x !llvm.ptr : (i32) -> !llvm.ptr
// CHECK: %[[ZERO:.*]] = llvm.mlir.constant(0 : i32) : i32
// CHECK: %[[STRUCT_PTR:.*]] = llvm.getelementptr %[[STRUCT]][%[[ZERO]]] : (!llvm.ptr, i32) -> !llvm.ptr, !llvm.struct<(ptr)>
// CHECK: %[[STRUCT_PTR:.*]] = llvm.getelementptr %[[STRUCT]][%{{.*}}, {{.*}}] : (!llvm.ptr, i32) -> !llvm.ptr, !llvm.struct<(ptr)>
// CHECK: llvm.store %{{.*}}, %[[STRUCT_PTR]] : !llvm.ptr, !llvm.ptr
// CHECK: %[[PARAM_PTR:.*]] = llvm.getelementptr %[[PARAMS]][%[[ZERO]]] : (!llvm.ptr, i32) -> !llvm.ptr, !llvm.ptr
// CHECK: llvm.store %[[STRUCT_PTR]], %[[PARAM_PTR]] : !llvm.ptr, !llvm.ptr
Expand Down
Loading