Skip to content

[SYCL] Fix kernel name mangling for CUDA/HIP with Microsoft ABI host #20973

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

Closed
kekaczma wants to merge 1 commit into from
Closed

[SYCL] Fix kernel name mangling for CUDA/HIP with Microsoft ABI host #20973

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
19 changes: 17 additions & 2 deletions clang/lib/Sema/SemaSYCL.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5369,9 +5369,24 @@ void SemaSYCL::copyDeviceKernelAttrs(CXXMethodDecl *CallOperator) {
}
}

static std::unique_ptr<MangleContext>
createSYCLCrossABIMangleContext(ASTContext &Ctx) {
// For SYCL offload compilation with cross-ABI scenarios (e.g., Microsoft
// host + Itanium device on Windows+CUDA), use device mangling context to
// ensure consistent kernel name mangling between host and device.
const TargetInfo *AuxTarget = Ctx.getAuxTargetInfo();
if (AuxTarget && Ctx.getTargetInfo().getCXXABI().isMicrosoft() &&
AuxTarget->getCXXABI().isItaniumFamily()) {
return std::unique_ptr<MangleContext>(
Ctx.createDeviceMangleContext(*AuxTarget));
}
// Same ABI or no offload target: use standard mangling
return std::unique_ptr<MangleContext>(Ctx.createMangleContext());
}

void SemaSYCL::SetSYCLKernelNames() {
std::unique_ptr<MangleContext> MangleCtx(
getASTContext().createMangleContext());
createSYCLCrossABIMangleContext(getASTContext()));
// We assume the list of KernelDescs is the complete list of kernels needing
// to be rewritten.
for (const std::pair<const FunctionDecl *, FunctionDecl *> &Pair :
Expand Down Expand Up @@ -5890,7 +5905,7 @@ void SemaSYCL::finalizeFreeFunctionKernels() {
KernelObjVisitor Visitor{*this};
Visitor.VisitFunctionParameters(kernel, IntHeader);
std::unique_ptr<MangleContext> MangleCtx(
getASTContext().createMangleContext());
createSYCLCrossABIMangleContext(getASTContext()));
std::string Name, MangledName;
std::tie(Name, MangledName) =
constructFreeFunctionKernelName(kernel, *MangleCtx);
Expand Down
2 changes: 0 additions & 2 deletions sycl/test-e2e/Printf/mixed-address-space.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,8 +2,6 @@
// for constant and generic address space can be used in the same module.
//
// UNSUPPORTED: target-amd
// XFAIL: cuda && windows
// XFAIL-TRACKER: https://github.com/intel/llvm/issues/14733
// FIXME: Drop the test once generic AS support is considered stable and the
// dedicated constant AS overload of printf is removed from the library.
//
Expand Down
2 changes: 0 additions & 2 deletions sycl/test-e2e/Printf/percent-symbol.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,8 +5,6 @@
// [1]: https://en.cppreference.com/w/cpp/io/c/fprintf
//
// UNSUPPORTED: target-amd
// XFAIL: cuda && windows
// XFAIL-TRACKER: https://github.com/intel/llvm/issues/14733
// RUN: %{build} -o %t.out
// RUN: %{run} %t.out | FileCheck %s
// FIXME: Remove dedicated constant address space testing once generic AS
Expand Down
Loading