diff --git a/clang/include/clang/Options/Options.td b/clang/include/clang/Options/Options.td index 2f7434d8afe11..838da6c270233 100644 --- a/clang/include/clang/Options/Options.td +++ b/clang/include/clang/Options/Options.td @@ -999,6 +999,10 @@ def Xthinlto_distributor_EQ : CommaJoined<["-"], "Xthinlto-distributor=">, "multiple times or with comma-separated values.">, MetaVarName<"">, Group; +def Xoffload_compiler : JoinedAndSeparate<["-"], "Xoffload-compiler">, + Visibility<[ClangOption, FlangOption]>, + HelpText<"Pass to the offload compilers or the ones identified by -">, + MetaVarName<" ">, Group; def Xoffload_linker : JoinedAndSeparate<["-"], "Xoffload-linker">, Visibility<[ClangOption, FlangOption]>, HelpText<"Pass to the offload linkers or the ones identified by -">, diff --git a/clang/lib/Driver/Driver.cpp b/clang/lib/Driver/Driver.cpp index 9fd64d4aac514..fc2433349beee 100644 --- a/clang/lib/Driver/Driver.cpp +++ b/clang/lib/Driver/Driver.cpp @@ -6990,6 +6990,12 @@ const ToolChain &Driver::getToolChain(const ArgList &Args, break; case llvm::Triple::spirv32: case llvm::Triple::spirv64: + if (Target.getOSName() == "hipspv" || + Target.getOSName() == "chipstar") { + TC = std::make_unique(*this, Target, + Args); + break; + } TC = std::make_unique(*this, Target, Args); break; case llvm::Triple::csky: diff --git a/clang/lib/Driver/ToolChains/Clang.cpp b/clang/lib/Driver/ToolChains/Clang.cpp index 80389937ee218..01ae22d6a350e 100644 --- a/clang/lib/Driver/ToolChains/Clang.cpp +++ b/clang/lib/Driver/ToolChains/Clang.cpp @@ -9051,6 +9051,7 @@ void LinkerWrapper::ConstructJob(Compilation &C, const JobAction &JA, OPT_v, OPT_cuda_path_EQ, OPT_rocm_path_EQ, + OPT_hip_path_EQ, OPT_O_Group, OPT_g_Group, OPT_g_flags_Group, @@ -9178,19 +9179,21 @@ void LinkerWrapper::ConstructJob(Compilation &C, const JobAction &JA, Linker->ConstructJob(C, JA, Output, Inputs, Args, LinkingOutput); const auto &LinkCommand = C.getJobs().getJobs().back(); - // Forward -Xoffload-linker<-triple> arguments to the device link job. - for (Arg *A : Args.filtered(options::OPT_Xoffload_linker)) { + for (Arg *A : + Args.filtered(options::OPT_Xoffload_compiler, OPT_Xoffload_linker)) { StringRef Val = A->getValue(0); + bool IsLinkJob = A->getOption().getID() == OPT_Xoffload_linker; + auto WrapperOption = + IsLinkJob ? Twine("--device-linker=") : Twine("--device-compiler="); if (Val.empty()) - CmdArgs.push_back( - Args.MakeArgString(Twine("--device-linker=") + A->getValue(1))); + CmdArgs.push_back(Args.MakeArgString(WrapperOption + A->getValue(1))); else CmdArgs.push_back(Args.MakeArgString( - "--device-linker=" + + WrapperOption + ToolChain::getOpenMPTriple(Val.drop_front()).getTriple() + "=" + A->getValue(1))); } - Args.ClaimAllArgs(options::OPT_Xoffload_linker); + Args.ClaimAllArgs(options::OPT_Xoffload_compiler); // Embed bitcode instead of an object in JIT mode. if (Args.hasFlag(options::OPT_fopenmp_target_jit, diff --git a/clang/lib/Driver/ToolChains/HIPSPV.cpp b/clang/lib/Driver/ToolChains/HIPSPV.cpp index be0f49d8e1497..6a074fe5ebd83 100644 --- a/clang/lib/Driver/ToolChains/HIPSPV.cpp +++ b/clang/lib/Driver/ToolChains/HIPSPV.cpp @@ -93,9 +93,21 @@ void HIPSPV::Linker::constructLinkAndEmitSpirvCommand( } // Emit SPIR-V binary. + llvm::opt::ArgStringList TrArgs; + if (getToolChain().getTriple().getOSName() == "chipstar") { + // chipStar needs 1.2 for supporting warp-level primitivies via sub-group + // extensions. Strictly put we'd need 1.3 for the standard non-extension + // shuffle operations, but it's not supported by any target yet. + TrArgs = {"--spirv-max-version=1.2", + "--spirv-ext=-all" + // Needed for experimental indirect call support. + ",+SPV_INTEL_function_pointers" + // Needed for shuffles below SPIR-V 1.3 + ",+SPV_INTEL_subgroups"}; + } else { + TrArgs = {"--spirv-max-version=1.1", "--spirv-ext=+all"}; + } - llvm::opt::ArgStringList TrArgs{"--spirv-max-version=1.1", - "--spirv-ext=+all"}; InputInfo TrInput = InputInfo(types::TY_LLVM_BC, TempFile, ""); SPIRV::constructTranslateCommand(C, *this, JA, Output, TrInput, TrArgs); } @@ -119,7 +131,16 @@ void HIPSPV::Linker::ConstructJob(Compilation &C, const JobAction &JA, HIPSPVToolChain::HIPSPVToolChain(const Driver &D, const llvm::Triple &Triple, const ToolChain &HostTC, const ArgList &Args) - : ToolChain(D, Triple, Args), HostTC(HostTC) { + : ToolChain(D, Triple, Args), HostTC(&HostTC) { + // Lookup binaries into the driver directory, this is used to + // discover the clang-offload-bundler executable. + getProgramPaths().push_back(getDriver().Dir); +} + +// Non-offloading toolchain. Primaly used by clang-offload-linker. +HIPSPVToolChain::HIPSPVToolChain(const Driver &D, const llvm::Triple &Triple, + const ArgList &Args) + : ToolChain(D, Triple, Args), HostTC(nullptr) { // Lookup binaries into the driver directory, this is used to // discover the clang-offload-bundler executable. getProgramPaths().push_back(getDriver().Dir); @@ -128,7 +149,14 @@ HIPSPVToolChain::HIPSPVToolChain(const Driver &D, const llvm::Triple &Triple, void HIPSPVToolChain::addClangTargetOptions( const llvm::opt::ArgList &DriverArgs, llvm::opt::ArgStringList &CC1Args, Action::OffloadKind DeviceOffloadingKind) const { - HostTC.addClangTargetOptions(DriverArgs, CC1Args, DeviceOffloadingKind); + + if (!HostTC) { + assert(DeviceOffloadingKind == Action::OFK_None && + "Need host toolchain for offloading!"); + return; + } + + HostTC->addClangTargetOptions(DriverArgs, CC1Args, DeviceOffloadingKind); assert(DeviceOffloadingKind == Action::OFK_HIP && "Only HIP offloading kinds are supported for GPUs."); @@ -159,27 +187,37 @@ Tool *HIPSPVToolChain::buildLinker() const { } void HIPSPVToolChain::addClangWarningOptions(ArgStringList &CC1Args) const { - HostTC.addClangWarningOptions(CC1Args); + if (HostTC) + HostTC->addClangWarningOptions(CC1Args); + ToolChain::addClangWarningOptions(CC1Args); } ToolChain::CXXStdlibType HIPSPVToolChain::GetCXXStdlibType(const ArgList &Args) const { - return HostTC.GetCXXStdlibType(Args); + if (HostTC) + return HostTC->GetCXXStdlibType(Args); + return ToolChain::GetCXXStdlibType(Args); } void HIPSPVToolChain::AddClangSystemIncludeArgs(const ArgList &DriverArgs, ArgStringList &CC1Args) const { - HostTC.AddClangSystemIncludeArgs(DriverArgs, CC1Args); + if (HostTC) + HostTC->AddClangSystemIncludeArgs(DriverArgs, CC1Args); + ToolChain::AddClangSystemIncludeArgs(DriverArgs, CC1Args); } void HIPSPVToolChain::AddClangCXXStdlibIncludeArgs( const ArgList &Args, ArgStringList &CC1Args) const { - HostTC.AddClangCXXStdlibIncludeArgs(Args, CC1Args); + if (HostTC) + HostTC->AddClangCXXStdlibIncludeArgs(Args, CC1Args); + ToolChain::AddClangCXXStdlibIncludeArgs(Args, CC1Args); } void HIPSPVToolChain::AddIAMCUIncludeArgs(const ArgList &Args, ArgStringList &CC1Args) const { - HostTC.AddIAMCUIncludeArgs(Args, CC1Args); + if (HostTC) + HostTC->AddIAMCUIncludeArgs(Args, CC1Args); + ToolChain::AddIAMCUIncludeArgs(Args, CC1Args); } void HIPSPVToolChain::AddHIPIncludeArgs(const ArgList &DriverArgs, @@ -273,12 +311,16 @@ SanitizerMask HIPSPVToolChain::getSupportedSanitizers() const { // This behavior is necessary because the host and device toolchains // invocations often share the command line, so the device toolchain must // tolerate flags meant only for the host toolchain. - return HostTC.getSupportedSanitizers(); + if (HostTC) + return HostTC->getSupportedSanitizers(); + return ToolChain::getSupportedSanitizers(); } VersionTuple HIPSPVToolChain::computeMSVCVersion(const Driver *D, const ArgList &Args) const { - return HostTC.computeMSVCVersion(D, Args); + if (HostTC) + return HostTC->computeMSVCVersion(D, Args); + return ToolChain::computeMSVCVersion(D, Args); } void HIPSPVToolChain::adjustDebugInfoKind( diff --git a/clang/lib/Driver/ToolChains/HIPSPV.h b/clang/lib/Driver/ToolChains/HIPSPV.h index caf6924151446..068040ee4f491 100644 --- a/clang/lib/Driver/ToolChains/HIPSPV.h +++ b/clang/lib/Driver/ToolChains/HIPSPV.h @@ -47,9 +47,12 @@ class LLVM_LIBRARY_VISIBILITY HIPSPVToolChain final : public ToolChain { public: HIPSPVToolChain(const Driver &D, const llvm::Triple &Triple, const ToolChain &HostTC, const llvm::opt::ArgList &Args); + HIPSPVToolChain(const Driver &D, const llvm::Triple &Triple, + const llvm::opt::ArgList &Args); const llvm::Triple *getAuxTriple() const override { - return &HostTC.getTriple(); + assert(HostTC); + return &HostTC->getTriple(); } void @@ -90,7 +93,7 @@ class LLVM_LIBRARY_VISIBILITY HIPSPVToolChain final : public ToolChain { bool isPICDefaultForced() const override { return false; } bool SupportsProfiling() const override { return false; } - const ToolChain &HostTC; + const ToolChain *HostTC = nullptr; protected: Tool *buildLinker() const override; diff --git a/clang/tools/clang-linker-wrapper/ClangLinkerWrapper.cpp b/clang/tools/clang-linker-wrapper/ClangLinkerWrapper.cpp index bfeca17d2147e..c708e14b35b5a 100644 --- a/clang/tools/clang-linker-wrapper/ClangLinkerWrapper.cpp +++ b/clang/tools/clang-linker-wrapper/ClangLinkerWrapper.cpp @@ -408,7 +408,7 @@ fatbinary(ArrayRef> InputFiles, namespace amdgcn { Expected -fatbinary(ArrayRef> InputFiles, +fatbinary(ArrayRef> InputFiles, const ArgList &Args) { llvm::TimeTraceScope TimeScope("AMDGPU Fatbinary"); @@ -439,8 +439,8 @@ fatbinary(ArrayRef> InputFiles, Args.MakeArgString(Twine("-compression-level=") + Arg->getValue())); SmallVector Targets = {"-targets=host-x86_64-unknown-linux-gnu"}; - for (const auto &[File, Arch] : InputFiles) - Targets.push_back(Saver.save("hip-amdgcn-amd-amdhsa--" + Arch)); + for (const auto &[File, Triple, Arch] : InputFiles) + Targets.push_back(Saver.save("hip-" + Triple + "-" + Arch)); CmdArgs.push_back(Saver.save(llvm::join(Targets, ","))); #ifdef _WIN32 @@ -448,7 +448,7 @@ fatbinary(ArrayRef> InputFiles, #else CmdArgs.push_back("-input=/dev/null"); #endif - for (const auto &[File, Arch] : InputFiles) + for (const auto &[File, Triple, Arch] : InputFiles) CmdArgs.push_back(Saver.save("-input=" + File)); CmdArgs.push_back(Saver.save("-output=" + *TempFileOrErr)); @@ -812,10 +812,11 @@ bundleCuda(ArrayRef Images, const ArgList &Args) { Expected>> bundleHIP(ArrayRef Images, const ArgList &Args) { - SmallVector, 4> InputFiles; + SmallVector, 4> InputFiles; for (const OffloadingImage &Image : Images) - InputFiles.emplace_back(std::make_pair(Image.Image->getBufferIdentifier(), - Image.StringData.lookup("arch"))); + InputFiles.emplace_back(std::make_tuple(Image.Image->getBufferIdentifier(), + Image.StringData.lookup("triple"), + Image.StringData.lookup("arch"))); auto FileOrErr = amdgcn::fatbinary(InputFiles, Args); if (!FileOrErr)