Skip to content

[Offload] Handle BoundArchitecture for non-GPU targets #132037

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 2 commits into from
Mar 19, 2025
Merged

[Offload] Handle BoundArchitecture for non-GPU targets #132037

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
181315c
[Offload] Handle `BoundArchitecture` for non-GPU targets
jhuber6 Mar 19, 2025
68916a9
Update clang/lib/Driver/ToolChains/Gnu.cpp
jhuber6 Mar 19, 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
64 changes: 31 additions & 33 deletions clang/lib/Driver/ToolChains/Gnu.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3388,46 +3388,44 @@ Generic_GCC::addLibStdCxxIncludePaths(const llvm::opt::ArgList &DriverArgs,
}

llvm::opt::DerivedArgList *
Generic_GCC::TranslateArgs(const llvm::opt::DerivedArgList &Args, StringRef,
Generic_GCC::TranslateArgs(const llvm::opt::DerivedArgList &Args,
StringRef BoundArch,
Action::OffloadKind DeviceOffloadKind) const {
if (DeviceOffloadKind != Action::OFK_SYCL &&
DeviceOffloadKind != Action::OFK_OpenMP)
return nullptr;

// If this tool chain is used for an OpenMP offloading device we have to make
// sure we always generate a shared library regardless of the commands the
// user passed to the host. This is required because the runtime library
// is required to load the device image dynamically at run time.
DerivedArgList *DAL = new DerivedArgList(Args.getBaseArgs());

// Filter all the arguments we don't care passing to the offloading
// toolchain as they can mess up with the creation of a shared library.
const llvm::DenseSet<unsigned> OpenMPFiltered{
options::OPT_shared, options::OPT_dynamic, options::OPT_static,
options::OPT_fPIE, options::OPT_fno_PIE, options::OPT_fpie,
options::OPT_fno_pie};
for (auto *A : Args)
if (DeviceOffloadKind != Action::OFK_OpenMP ||
!OpenMPFiltered.contains(A->getOption().getID()))
DAL->append(A);

// Request a shared library for CPU offloading. Given that these options
// are decided implicitly, they do not refer to any base argument.
const OptTable &Opts = getDriver().getOpts();
if (DeviceOffloadKind == Action::OFK_OpenMP) {
DerivedArgList *DAL = new DerivedArgList(Args.getBaseArgs());
const OptTable &Opts = getDriver().getOpts();

// Request the shared library. Given that these options are decided
// implicitly, they do not refer to any base argument.
DAL->AddFlagArg(/*BaseArg=*/nullptr, Opts.getOption(options::OPT_shared));
DAL->AddFlagArg(/*BaseArg=*/nullptr, Opts.getOption(options::OPT_fPIC));
}

// Filter all the arguments we don't care passing to the offloading
// toolchain as they can mess up with the creation of a shared library.
for (auto *A : Args) {
switch ((options::ID)A->getOption().getID()) {
default:
DAL->append(A);
break;
case options::OPT_shared:
case options::OPT_dynamic:
case options::OPT_static:
case options::OPT_fPIC:
case options::OPT_fno_PIC:
case options::OPT_fpic:
case options::OPT_fno_pic:
case options::OPT_fPIE:
case options::OPT_fno_PIE:
case options::OPT_fpie:
case options::OPT_fno_pie:
break;
}
}
return DAL;
// Add the bound architecture to the arguments list if present.
if (!BoundArch.empty()) {
options::ID Opt =
getTriple().isARM() || getTriple().isPPC() || getTriple().isAArch64()
? options::OPT_mcpu_EQ
: options::OPT_march_EQ;
DAL->eraseArg(Opt);
DAL->AddJoinedArg(nullptr, Opts.getOption(Opt), BoundArch);
}
return nullptr;
return DAL;
}

void Generic_ELF::anchor() {}
Expand Down
13 changes: 13 additions & 0 deletions clang/test/Driver/openmp-offload.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -218,3 +218,16 @@
// RUN: %clang -### --target=powerpc64le-linux -fopenmp=libomp -fopenmp-targets=powerpc64le-ibm-linux-gnu \
// RUN: -foffload-lto=thin %s 2>&1 | FileCheck -check-prefix=CHK-DEVICE-LTO-THIN %s
// CHK-DEVICE-LTO-THIN: clang-linker-wrapper{{.*}} "--device-compiler=powerpc64le-ibm-linux-gnu=-flto=thin"

//
// Check forwarding architectures to non-GPU targets
//
// RUN: %clang -### --target=aarch64-unknown-linux-gnu -fopenmp=libomp -fopenmp-targets=aarch64-unknown-linux-gnu \
// RUN: --offload-arch=a64fx %s 2>&1 | FileCheck -check-prefix=CHK-CPU-ARCH-A %s
// CHK-CPU-ARCH-A: "-cc1" "-triple" "aarch64-unknown-linux-gnu" {{.*}} "-target-cpu" "generic"
// CHK-CPU-ARCH-A: "-cc1" "-triple" "aarch64-unknown-linux-gnu" {{.*}} "-target-cpu" "a64fx"
//
// RUN: %clang -### --target=x86_64-unknown-linux-gnu -fopenmp=libomp -fopenmp-targets=x86_64-unknown-linux-gnu \
// RUN: --offload-arch=znver4 %s 2>&1 | FileCheck -check-prefix=CHK-CPU-ARCH-X %s
// CHK-CPU-ARCH-X: "-cc1" "-triple" "x86_64-unknown-linux-gnu" {{.*}} "-target-cpu" "x86-64"
// CHK-CPU-ARCH-X: "-cc1" "-triple" "x86_64-unknown-linux-gnu" {{.*}} "-target-cpu" "znver4"