Skip to content

[SYCL][Driver] Enable passing -Xarch_<arch> <option> to SYCL offload compilations. #20917

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

Draft
srividya-sundaram wants to merge 2 commits into
base: sycl
Choose a base branch
from
+53 −7
Draft

[SYCL][Driver] Enable passing -Xarch_<arch> <option> to SYCL offload compilations. #20917

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
1f8b4e8
[SYCL][Driver] Enable passing -Xarch_<arch> <option> to SYCL offload …
srividya-sundaram Dec 16, 2025
802a792
Merge branch 'sycl' into sycl-xarch
srividya-sundaram Dec 16, 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
20 changes: 13 additions & 7 deletions clang/lib/Driver/Driver.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7692,16 +7692,22 @@ Driver::getOffloadArchs(Compilation &C, const llvm::opt::DerivedArgList &Args,
? OffloadArch::Generic
: OffloadArch::HIPDefault));
} else if (Kind == Action::OFK_SYCL) {
// Accept legacy `-march` device arguments for SYCL.
// For SYCL offloading, we need to check the triple for NVPTX or AMDGPU.
// The default arch is set for NVPTX if not provided. For AMDGPU, emit
// an error as the user is responsible to set the arch.
if (TC.getTriple().isNVPTX())
Archs.insert(OffloadArchToString(OffloadArch::SM_75));
else if (TC.getTriple().isAMDGPU())
C.getDriver().Diag(clang::diag::err_drv_sycl_missing_amdgpu_arch)
<< 1 << TC.getTriple().str();
else
Archs.insert(StringRef());
if (auto *Arg = C.getArgsForToolChain(&TC, /*BoundArch=*/"", Kind)
.getLastArg(options::OPT_march_EQ)) {
Archs.insert(Arg->getValue());
} else {
if (TC.getTriple().isNVPTX())
Archs.insert(OffloadArchToString(OffloadArch::SM_50));
else if (TC.getTriple().isAMDGPU())
C.getDriver().Diag(clang::diag::err_drv_sycl_missing_amdgpu_arch)
<< 1 << TC.getTriple().str();
else
Archs.insert(StringRef());
}
} else if (Kind == Action::OFK_OpenMP) {
// Accept legacy `-march` device arguments for OpenMP.
if (auto *Arg = C.getArgsForToolChain(&TC, /*BoundArch=*/"", Kind)
Expand Down
40 changes: 40 additions & 0 deletions clang/test/Driver/sycl-offload-Xarch.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,40 @@
// Test passing of -Xarch_<arch> <option> to SYCL offload compilations.

// RUN: %clang --offload-new-driver -fsycl -fsycl-targets=amdgcn-amd-amdhsa -nogpulib -fno-sycl-libspirv -nogpuinc \
// RUN: -Xarch_amdgcn -march=gfx90a -Xarch_amdgcn -O3 -S -### %s 2>&1 \
// RUN: | FileCheck -check-prefixes=O3ONCE,AMD-ARCH %s
// RUN: %clang --offload-new-driver -fsycl -fsycl-targets=nvptx64-nvidia-cuda -nogpulib -fno-sycl-libspirv -nogpuinc \
// RUN: -Xarch_nvptx64 -march=sm_52 -Xarch_nvptx64 -O3 -S -### %s 2>&1 \
// RUN: | FileCheck -check-prefixes=O3ONCE,NVPTX-ARCH %s


// RUN: %clang --offload-new-driver -fsycl -fsycl-targets=amdgcn-amd-amdhsa -nogpulib -fno-sycl-libspirv -nogpuinc \
// RUN: -Xarch_amdgcn --offload-arch=gfx90a -Xarch_amdgcn -O3 -S -### %s 2>&1 \
// RUN: | FileCheck -check-prefixes=O3ONCE,AMD-ARCH %s
// RUN: %clang --offload-new-driver -fsycl -fsycl-targets=nvptx64-nvidia-cuda -nogpulib -fno-sycl-libspirv -nogpuinc \
// RUN: -Xarch_nvptx64 --offload-arch=sm_52 -Xarch_nvptx64 -O3 -S -### %s 2>&1 \
// RUN: | FileCheck -check-prefixes=O3ONCE,NVPTX-ARCH %s


// RUN: %clang --offload-new-driver -fsycl -fsycl-targets=amdgcn-amd-amdhsa,nvptx64-nvidia-cuda -nogpulib -fno-sycl-libspirv -nogpuinc \
// RUN: -Xarch_amdgcn --offload-arch=gfx90a,gfx906 -Xarch_nvptx64 --offload-arch=sm_52,sm_89 -S -### %s 2>&1 \
// RUN: | FileCheck -check-prefixes=MULTI-ARCH %s

// O3ONCE: "-O3"
// AMD-ARCH: {{"[^"]*llvm-offload-binary[^"]*" "-o".* "--image=file=.*.bc,triple=amdgcn-amd-amdhsa,arch=gfx90a,kind=sycl"}}
// NVPTX-ARCH: {{"[^"]*llvm-offload-binary[^"]*" "-o".* "--image=file=.*.bc,triple=nvptx64-nvidia-cuda,arch=sm_52,kind=sycl"}}
// MULTI-ARCH: {{"[^"]*llvm-offload-binary[^"]*" "-o".* "--image=file=.*.bc,triple=amdgcn-amd-amdhsa,arch=gfx906,kind=sycl"}}
// MULTI-ARCH-SAME: {{"--image=file=.*.bc,triple=amdgcn-amd-amdhsa,arch=gfx90a,kind=sycl"}}
// MULTI-ARCH-SAME: {{"--image=file=.*.bc,triple=nvptx64-nvidia-cuda,arch=sm_52,kind=sycl"}}
// MULTI-ARCH-SAME: {{"--image=file=.*.bc,triple=nvptx64-nvidia-cuda,arch=sm_89,kind=sycl"}}

// Make sure that `-Xarch_amdgcn` forwards libraries to the device linker.
// RUN: %clang --offload-new-driver -fsycl --offload-arch=gfx90a -nogpulib -fno-sycl-libspirv -nogpuinc \
// RUN: --target=x86_64-unknown-linux-gnu -Xarch_amdgcn -Wl,-lfoo -### %s 2>&1 \
// RUN: | FileCheck -check-prefix=LIBS %s
// RUN: %clang --offload-new-driver -fsycl --offload-arch=gfx90a -nogpulib -fno-sycl-libspirv -nogpuinc \
// RUN: -Xoffload-linker-amdgcn-amd-amdhsa -lfoo -### %s 2>&1 \
// RUN: | FileCheck -check-prefix=LIBS %s
// LIBS: "--device-linker=amdgcn-amd-amdhsa=-lfoo"


Loading