Skip to content

[Clang] Make the GPU toolchains implicitly link -lm and -lc #98170

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
Jul 23, 2024
Merged

[Clang] Make the GPU toolchains implicitly link -lm and -lc #98170

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
69811ec
[Clang] Make the GPU toolchains implicitly link `-lm` and `-lc`
jhuber6 Jul 9, 2024
ea463c1
Remove old method for passing C libraries.
jhuber6 Jul 23, 2024
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
3 changes: 3 additions & 0 deletions clang/lib/Driver/ToolChains/AMDGPU.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -633,6 +633,9 @@ void amdgpu::Linker::ConstructJob(Compilation &C, const JobAction &JA,
else if (Args.hasArg(options::OPT_mcpu_EQ))
CmdArgs.push_back(Args.MakeArgString(
"-plugin-opt=mcpu=" + Args.getLastArgValue(options::OPT_mcpu_EQ)));

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

HIPAMD toolchain has its lld command argument handling in AMDGCN::Linker::constructLldCommand. Also need change there.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

So, I wasn't sure if I should also apply this to HIP / CUDA stuff just yet. HIP doesn't pass LTO bitcode to the linker, so we can't do a full link with the library. CUDA doesn't even invoke the linker either. I think in the future it would be nice to provide these to HIP and CUDA by default however.

addGPULibraries(getToolChain(), Args, CmdArgs);

CmdArgs.push_back("-o");
CmdArgs.push_back(Output.getFilename());
C.addCommand(std::make_unique<Command>(
Expand Down
5 changes: 5 additions & 0 deletions clang/lib/Driver/ToolChains/Clang.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9111,6 +9111,11 @@ void LinkerWrapper::ConstructJob(Compilation &C, const JobAction &JA,
A->claim();
}

// If we disable the GPU C library support it needs to be forwarded to the
// link job.
if (!Args.hasFlag(options::OPT_gpulibc, options::OPT_nogpulibc, true))
CmdArgs.push_back("--device-linker=-nolibc");

// Add the linker arguments to be forwarded by the wrapper.
CmdArgs.push_back(Args.MakeArgString(Twine("--linker-path=") +
LinkCommand->getExecutable()));
Expand Down
56 changes: 17 additions & 39 deletions clang/lib/Driver/ToolChains/CommonArgs.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -509,6 +509,22 @@ void tools::addLinkerCompressDebugSectionsOption(
}
}

void tools::addGPULibraries(const ToolChain &TC, const llvm::opt::ArgList &Args,
llvm::opt::ArgStringList &CmdArgs) {
if (Args.hasArg(options::OPT_nostdlib, options::OPT_r,
options::OPT_nodefaultlibs, options::OPT_nolibc,
options::OPT_nogpulibc))
return;

// If the user's toolchain has the 'include/<triple>/` path, we assume it
// supports the standard C libraries for the GPU and include them.
bool HasLibC = TC.getStdlibIncludePath().has_value();
if (HasLibC) {
CmdArgs.push_back("-lc");
CmdArgs.push_back("-lm");
}
}

void tools::AddTargetFeature(const ArgList &Args,
std::vector<StringRef> &Features,
OptSpecifier OnOpt, OptSpecifier OffOpt,
Expand Down Expand Up @@ -1145,42 +1161,6 @@ void tools::addLTOOptions(const ToolChain &ToolChain, const ArgList &Args,
}
}

/// Adds the '-lcgpu' and '-lmgpu' libraries to the compilation to include the
/// LLVM C library for GPUs.
static void addOpenMPDeviceLibC(const Compilation &C, const ArgList &Args,
ArgStringList &CmdArgs) {
if (Args.hasArg(options::OPT_nogpulib) || Args.hasArg(options::OPT_nolibc))
return;

// Check the resource directory for the LLVM libc GPU declarations. If it's
// found we can assume that LLVM was built with support for the GPU libc.
SmallString<256> LibCDecls(C.getDriver().ResourceDir);
llvm::sys::path::append(LibCDecls, "include", "llvm_libc_wrappers",
"llvm-libc-decls");
bool HasLibC = llvm::sys::fs::exists(LibCDecls) &&
llvm::sys::fs::is_directory(LibCDecls);
if (!Args.hasFlag(options::OPT_gpulibc, options::OPT_nogpulibc, HasLibC))
return;

SmallVector<const ToolChain *> ToolChains;
auto TCRange = C.getOffloadToolChains(Action::OFK_OpenMP);
for (auto TI = TCRange.first, TE = TCRange.second; TI != TE; ++TI)
ToolChains.push_back(TI->second);

if (llvm::any_of(ToolChains, [](const ToolChain *TC) {
return TC->getTriple().isAMDGPU();
})) {
CmdArgs.push_back("-lcgpu-amdgpu");
CmdArgs.push_back("-lmgpu-amdgpu");
}
if (llvm::any_of(ToolChains, [](const ToolChain *TC) {
return TC->getTriple().isNVPTX();
})) {
CmdArgs.push_back("-lcgpu-nvptx");
CmdArgs.push_back("-lmgpu-nvptx");
}
}

void tools::addOpenMPRuntimeLibraryPath(const ToolChain &TC,
const ArgList &Args,
ArgStringList &CmdArgs) {
Expand Down Expand Up @@ -1253,10 +1233,8 @@ bool tools::addOpenMPRuntime(const Compilation &C, ArgStringList &CmdArgs,
if (IsOffloadingHost && !Args.hasArg(options::OPT_nogpulib))
CmdArgs.push_back("-lomptarget.devicertl");

if (IsOffloadingHost)
addOpenMPDeviceLibC(C, Args, CmdArgs);

addArchSpecificRPath(TC, Args, CmdArgs);

addOpenMPRuntimeLibraryPath(TC, Args, CmdArgs);

return true;
Expand Down
3 changes: 3 additions & 0 deletions clang/lib/Driver/ToolChains/CommonArgs.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,6 +35,9 @@ void addLinkerCompressDebugSectionsOption(const ToolChain &TC,
const llvm::opt::ArgList &Args,
llvm::opt::ArgStringList &CmdArgs);

void addGPULibraries(const ToolChain &TC, const llvm::opt::ArgList &Args,
llvm::opt::ArgStringList &CmdArgs);

void claimNoWarnArgs(const llvm::opt::ArgList &Args);

bool addSanitizerRuntimes(const ToolChain &TC, const llvm::opt::ArgList &Args,
Expand Down
2 changes: 2 additions & 0 deletions clang/lib/Driver/ToolChains/Cuda.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -625,6 +625,8 @@ void NVPTX::Linker::ConstructJob(Compilation &C, const JobAction &JA,
addLTOOptions(getToolChain(), Args, CmdArgs, Output, Inputs[0],
C.getDriver().getLTOMode() == LTOK_Thin);

addGPULibraries(getToolChain(), Args, CmdArgs);

// Add paths for the default clang library path.
SmallString<256> DefaultLibPath =
llvm::sys::path::parent_path(TC.getDriver().Dir);
Expand Down
28 changes: 2 additions & 26 deletions clang/test/Driver/openmp-offload-gpu.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -372,33 +372,9 @@
// XARCH-DEVICE: "-cc1" "-triple" "nvptx64-nvidia-cuda"{{.*}}"-O3"
// XARCH-DEVICE-NOT: "-cc1" "-triple" "x86_64-unknown-linux-gnu"{{.*}}"-O3"

//
// Check that `-gpulibc` includes the LLVM C libraries for the GPU.
//
// RUN: %clang -### --target=x86_64-unknown-linux-gnu -fopenmp=libomp \
// RUN: --libomptarget-nvptx-bc-path=%S/Inputs/libomptarget/libomptarget-nvptx-test.bc \
// RUN: --libomptarget-amdgpu-bc-path=%S/Inputs/hip_dev_lib/libomptarget-amdgpu-gfx803.bc \
// RUN: --cuda-path=%S/Inputs/CUDA_102/usr/local/cuda \
// RUN: --rocm-path=%S/Inputs/rocm \
// RUN: --offload-arch=sm_52,gfx803 -gpulibc -nogpuinc %s 2>&1 \
// RUN: | FileCheck --check-prefix=LIBC-GPU %s
// RUN: %clang -### --target=x86_64-unknown-linux-gnu -fopenmp=libomp \
// RUN: --libomptarget-nvptx-bc-path=%S/Inputs/libomptarget/libomptarget-nvptx-test.bc \
// RUN: --libomptarget-amdgpu-bc-path=%S/Inputs/hip_dev_lib/libomptarget-amdgpu-gfx803.bc \
// RUN: --cuda-path=%S/Inputs/CUDA_102/usr/local/cuda \
// RUN: --rocm-path=%S/Inputs/rocm \
// RUN: -Xopenmp-target=nvptx64-nvidia-cuda -march=sm_52 \
// RUN: -Xopenmp-target=amdgcn-amd-amdhsa -march=gfx803 \
// RUN: -fopenmp-targets=nvptx64-nvidia-cuda,amdgcn-amd-amdhsa -gpulibc -nogpuinc %s 2>&1 \
// RUN: | FileCheck --check-prefix=LIBC-GPU %s
// LIBC-GPU-DAG: "-lcgpu-amdgpu"
// LIBC-GPU-DAG: "-lmgpu-amdgpu"
// LIBC-GPU-DAG: "-lcgpu-nvptx"
// LIBC-GPU-DAG: "-lmgpu-nvptx"

// RUN: %clang -### --target=x86_64-unknown-linux-gnu -fopenmp=libomp \
// RUN: --libomptarget-nvptx-bc-path=%S/Inputs/libomptarget/libomptarget-nvptx-test.bc \
// RUN: --cuda-path=%S/Inputs/CUDA_102/usr/local/cuda \
// RUN: --offload-arch=sm_52 -nogpulibc -nogpuinc %s 2>&1 \
// RUN: | FileCheck --check-prefix=NO-LIBC-GPU %s
// NO-LIBC-GPU-NOT: -lmgpu{{.*}}-lcgpu
// RUN: | FileCheck --check-prefix=LIBC-GPU %s
// LIBC-GPU: clang-linker-wrapper{{.*}}"--device-linker=-nolibc"
Loading