r291131 - [Driver] Driver changes to support CUDA compilation on Windows.
Reid Kleckner via cfe-commits
cfe-commits at lists.llvm.org
Fri Jan 6 09:46:24 PST 2017
cuda-version-check.cu seems to be failing since this change:
http://lab.llvm.org:8011/builders/clang-x86-windows-msvc2015/builds/1768
The relevant failing part:
$
"C:/b/slave/clang-x86-windows-msvc2015/clang-x86-windows-msvc2015/stage1/./bin/clang.EXE"
"-v" "-###" "--cuda-gpu-arch=sm_60"
"--sysroot=C:\b\slave\clang-x86-windows-msvc2015\clang-x86-windows-msvc2015\llvm\tools\clang\test\Driver/Inputs/CUDA"
"C:\b\slave\clang-x86-windows-msvc2015\clang-x86-windows-msvc2015\llvm\tools\clang\test\Driver\
cuda-version-check.cu"
$
"C:/b/slave/clang-x86-windows-msvc2015/clang-x86-windows-msvc2015/stage1/./bin\FileCheck.EXE"
"C:\b\slave\clang-x86-windows-msvc2015\clang-x86-windows-msvc2015\llvm\tools\clang\test\Driver\
cuda-version-check.cu" "--check-prefix=ERR_SM60"
# command stderr:
C:\b\slave\clang-x86-windows-msvc2015\clang-x86-windows-msvc2015\llvm\tools\clang\test\Driver\cuda-version-check.cu:56:14:
error: expected string not found in input
// ERR_SM60: error: GPU arch sm_60 {{.*}}
^
<stdin>:1:1: note: scanning from here
clang version 4.0.0 (trunk 291142)
^
<stdin>:5:12: note: possible intended match here
clang.EXE: error: cannot find libdevice for sm_60. Provide path to
different CUDA installation via --cuda-path, or pass -nocudalib to build
without linking with libdevice.
^
On Thu, Jan 5, 2017 at 8:52 AM, Justin Lebar via cfe-commits <
cfe-commits at lists.llvm.org> wrote:
> Author: jlebar
> Date: Thu Jan 5 10:52:29 2017
> New Revision: 291131
>
> URL: http://llvm.org/viewvc/llvm-project?rev=291131&view=rev
> Log:
> [Driver] Driver changes to support CUDA compilation on Windows.
>
> Summary:
> For the most part this is straightforward: Just add a CudaInstallation
> object to the MSVC and MinGW toolchains.
>
> CudaToolChain has to override computeMSVCVersion so that
> Clang::constructJob passes the right version flag to cc1. We have to
> modify IsWindowsMSVC and friends in Clang::constructJob to be true when
> compiling CUDA device code on Windows for the same reason.
>
> Depends on: D28319
>
> Reviewers: tra
>
> Subscribers: cfe-commits
>
> Differential Revision: https://reviews.llvm.org/D28320
>
> Added:
> cfe/trunk/test/Driver/Inputs/CUDA-windows/
> cfe/trunk/test/Driver/Inputs/CUDA-windows/Program Files/
> cfe/trunk/test/Driver/Inputs/CUDA-windows/Program Files/NVIDIA GPU
> Computing Toolkit/
> cfe/trunk/test/Driver/Inputs/CUDA-windows/Program Files/NVIDIA GPU
> Computing Toolkit/CUDA/
> cfe/trunk/test/Driver/Inputs/CUDA-windows/Program Files/NVIDIA GPU
> Computing Toolkit/CUDA/v8.0/
> cfe/trunk/test/Driver/Inputs/CUDA-windows/Program Files/NVIDIA GPU
> Computing Toolkit/CUDA/v8.0/bin/
> cfe/trunk/test/Driver/Inputs/CUDA-windows/Program Files/NVIDIA GPU
> Computing Toolkit/CUDA/v8.0/bin/.keep
> cfe/trunk/test/Driver/Inputs/CUDA-windows/Program Files/NVIDIA GPU
> Computing Toolkit/CUDA/v8.0/include/
> cfe/trunk/test/Driver/Inputs/CUDA-windows/Program Files/NVIDIA GPU
> Computing Toolkit/CUDA/v8.0/include/.keep
> cfe/trunk/test/Driver/Inputs/CUDA-windows/Program Files/NVIDIA GPU
> Computing Toolkit/CUDA/v8.0/lib/
> cfe/trunk/test/Driver/Inputs/CUDA-windows/Program Files/NVIDIA GPU
> Computing Toolkit/CUDA/v8.0/lib/.keep
> cfe/trunk/test/Driver/Inputs/CUDA-windows/Program Files/NVIDIA GPU
> Computing Toolkit/CUDA/v8.0/nvvm/
> cfe/trunk/test/Driver/Inputs/CUDA-windows/Program Files/NVIDIA GPU
> Computing Toolkit/CUDA/v8.0/nvvm/libdevice/
> cfe/trunk/test/Driver/Inputs/CUDA-windows/Program Files/NVIDIA GPU
> Computing Toolkit/CUDA/v8.0/nvvm/libdevice/libdevice.compute_30.10.bc
> cfe/trunk/test/Driver/Inputs/CUDA-windows/Program Files/NVIDIA GPU
> Computing Toolkit/CUDA/v8.0/nvvm/libdevice/libdevice.compute_35.10.bc
> cfe/trunk/test/Driver/cuda-windows.cu
> Modified:
> cfe/trunk/include/clang/Driver/ToolChain.h
> cfe/trunk/lib/Driver/MSVCToolChain.cpp
> cfe/trunk/lib/Driver/MinGWToolChain.cpp
> cfe/trunk/lib/Driver/ToolChains.cpp
> cfe/trunk/lib/Driver/ToolChains.h
> cfe/trunk/lib/Driver/Tools.cpp
>
> Modified: cfe/trunk/include/clang/Driver/ToolChain.h
> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/include/
> clang/Driver/ToolChain.h?rev=291131&r1=291130&r2=291131&view=diff
> ============================================================
> ==================
> --- cfe/trunk/include/clang/Driver/ToolChain.h (original)
> +++ cfe/trunk/include/clang/Driver/ToolChain.h Thu Jan 5 10:52:29 2017
> @@ -139,6 +139,13 @@ public:
> vfs::FileSystem &getVFS() const;
> const llvm::Triple &getTriple() const { return Triple; }
>
> + /// Get the toolchain's aux triple, if it has one.
> + ///
> + /// Exactly what the aux triple represents depends on the toolchain,
> but for
> + /// example when compiling CUDA code for the GPU, the triple might be
> NVPTX,
> + /// while the aux triple is the host (CPU) toolchain, e.g.
> x86-linux-gnu.
> + virtual const llvm::Triple *getAuxTriple() const { return nullptr; }
> +
> llvm::Triple::ArchType getArch() const { return Triple.getArch(); }
> StringRef getArchName() const { return Triple.getArchName(); }
> StringRef getPlatform() const { return Triple.getVendorName(); }
>
> Modified: cfe/trunk/lib/Driver/MSVCToolChain.cpp
> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Driver/
> MSVCToolChain.cpp?rev=291131&r1=291130&r2=291131&view=diff
> ============================================================
> ==================
> --- cfe/trunk/lib/Driver/MSVCToolChain.cpp (original)
> +++ cfe/trunk/lib/Driver/MSVCToolChain.cpp Thu Jan 5 10:52:29 2017
> @@ -47,9 +47,9 @@ using namespace clang::driver::toolchain
> using namespace clang;
> using namespace llvm::opt;
>
> -MSVCToolChain::MSVCToolChain(const Driver &D, const llvm::Triple& Triple,
> +MSVCToolChain::MSVCToolChain(const Driver &D, const llvm::Triple &Triple,
> const ArgList &Args)
> - : ToolChain(D, Triple, Args) {
> + : ToolChain(D, Triple, Args), CudaInstallation(D, Triple, Args) {
> getProgramPaths().push_back(getDriver().getInstalledDir());
> if (getDriver().getInstalledDir() != getDriver().Dir)
> getProgramPaths().push_back(getDriver().Dir);
> @@ -94,6 +94,15 @@ bool MSVCToolChain::isPICDefaultForced()
> return getArch() == llvm::Triple::x86_64;
> }
>
> +void MSVCToolChain::AddCudaIncludeArgs(const ArgList &DriverArgs,
> + ArgStringList &CC1Args) const {
> + CudaInstallation.AddCudaIncludeArgs(DriverArgs, CC1Args);
> +}
> +
> +void MSVCToolChain::printVerboseInfo(raw_ostream &OS) const {
> + CudaInstallation.print(OS);
> +}
> +
> #ifdef USE_WIN32
> static bool readFullStringValue(HKEY hkey, const char *valueName,
> std::string &value) {
>
> Modified: cfe/trunk/lib/Driver/MinGWToolChain.cpp
> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Driver/
> MinGWToolChain.cpp?rev=291131&r1=291130&r2=291131&view=diff
> ============================================================
> ==================
> --- cfe/trunk/lib/Driver/MinGWToolChain.cpp (original)
> +++ cfe/trunk/lib/Driver/MinGWToolChain.cpp Thu Jan 5 10:52:29 2017
> @@ -63,7 +63,7 @@ void MinGW::findGccLibDir() {
> }
>
> MinGW::MinGW(const Driver &D, const llvm::Triple &Triple, const ArgList
> &Args)
> - : ToolChain(D, Triple, Args) {
> + : ToolChain(D, Triple, Args), CudaInstallation(D, Triple, Args) {
> getProgramPaths().push_back(getDriver().getInstalledDir());
>
> // In Windows there aren't any standard install locations, we search
> @@ -135,6 +135,15 @@ bool MinGW::UseSEHExceptions() const {
> return getArch() == llvm::Triple::x86_64;
> }
>
> +void MinGW::AddCudaIncludeArgs(const ArgList &DriverArgs,
> + ArgStringList &CC1Args) const {
> + CudaInstallation.AddCudaIncludeArgs(DriverArgs, CC1Args);
> +}
> +
> +void MinGW::printVerboseInfo(raw_ostream &OS) const {
> + CudaInstallation.print(OS);
> +}
> +
> // Include directories for various hosts:
>
> // Windows, mingw.org
>
> Modified: cfe/trunk/lib/Driver/ToolChains.cpp
> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Driver/
> ToolChains.cpp?rev=291131&r1=291130&r2=291131&view=diff
> ============================================================
> ==================
> --- cfe/trunk/lib/Driver/ToolChains.cpp (original)
> +++ cfe/trunk/lib/Driver/ToolChains.cpp Thu Jan 5 10:52:29 2017
> @@ -1810,14 +1810,21 @@ CudaInstallationDetector::CudaInstallati
> : D(D) {
> SmallVector<std::string, 4> CudaPathCandidates;
>
> - if (Args.hasArg(options::OPT_cuda_path_EQ))
> + // In decreasing order so we prefer newer versions to older versions.
> + std::initializer_list<const char *> Versions = {"8.0", "7.5", "7.0"};
> +
> + if (Args.hasArg(options::OPT_cuda_path_EQ)) {
> CudaPathCandidates.push_back(
> Args.getLastArgValue(options::OPT_cuda_path_EQ));
> - else {
> + } else if (HostTriple.isOSWindows()) {
> + for (const char *Ver : Versions)
> + CudaPathCandidates.push_back(
> + D.SysRoot + "/Program Files/NVIDIA GPU Computing
> Toolkit/CUDA/v" +
> + Ver);
> + } else {
> CudaPathCandidates.push_back(D.SysRoot + "/usr/local/cuda");
> - CudaPathCandidates.push_back(D.SysRoot + "/usr/local/cuda-8.0");
> - CudaPathCandidates.push_back(D.SysRoot + "/usr/local/cuda-7.5");
> - CudaPathCandidates.push_back(D.SysRoot + "/usr/local/cuda-7.0");
> + for (const char *Ver : Versions)
> + CudaPathCandidates.push_back(D.SysRoot + "/usr/local/cuda-" + Ver);
> }
>
> for (const auto &CudaPath : CudaPathCandidates) {
> @@ -5021,6 +5028,11 @@ SanitizerMask CudaToolChain::getSupporte
> return HostTC.getSupportedSanitizers();
> }
>
> +VersionTuple CudaToolChain::computeMSVCVersion(const Driver *D,
> + const ArgList &Args) const
> {
> + return HostTC.computeMSVCVersion(D, Args);
> +}
> +
> /// XCore tool chain
> XCoreToolChain::XCoreToolChain(const Driver &D, const llvm::Triple
> &Triple,
> const ArgList &Args)
>
> Modified: cfe/trunk/lib/Driver/ToolChains.h
> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Driver/
> ToolChains.h?rev=291131&r1=291130&r2=291131&view=diff
> ============================================================
> ==================
> --- cfe/trunk/lib/Driver/ToolChains.h (original)
> +++ cfe/trunk/lib/Driver/ToolChains.h Thu Jan 5 10:52:29 2017
> @@ -709,12 +709,19 @@ public:
> const llvm::opt::ArgList &DriverArgs,
> llvm::opt::ArgStringList &CC1Args) const override;
>
> + void AddCudaIncludeArgs(const llvm::opt::ArgList &DriverArgs,
> + llvm::opt::ArgStringList &CC1Args) const
> override;
> +
> + void printVerboseInfo(raw_ostream &OS) const override;
> +
> protected:
> Tool *getTool(Action::ActionClass AC) const override;
> Tool *buildLinker() const override;
> Tool *buildAssembler() const override;
>
> private:
> + CudaInstallationDetector CudaInstallation;
> +
> std::string Base;
> std::string GccLibDir;
> std::string Ver;
> @@ -892,6 +899,10 @@ public:
> CudaToolChain(const Driver &D, const llvm::Triple &Triple,
> const ToolChain &HostTC, const llvm::opt::ArgList &Args);
>
> + virtual const llvm::Triple *getAuxTriple() const override {
> + return &HostTC.getTriple();
> + }
> +
> llvm::opt::DerivedArgList *
> TranslateArgs(const llvm::opt::DerivedArgList &Args, StringRef
> BoundArch,
> Action::OffloadKind DeviceOffloadKind) const override;
> @@ -924,6 +935,10 @@ public:
>
> SanitizerMask getSupportedSanitizers() const override;
>
> + VersionTuple
> + computeMSVCVersion(const Driver *D,
> + const llvm::opt::ArgList &Args) const override;
> +
> const ToolChain &HostTC;
> CudaInstallationDetector CudaInstallation;
>
> @@ -1147,6 +1162,9 @@ public:
> const llvm::opt::ArgList &DriverArgs,
> llvm::opt::ArgStringList &CC1Args) const override;
>
> + void AddCudaIncludeArgs(const llvm::opt::ArgList &DriverArgs,
> + llvm::opt::ArgStringList &CC1Args) const
> override;
> +
> bool getWindowsSDKDir(std::string &path, int &major,
> std::string &windowsSDKIncludeVersion,
> std::string &windowsSDKLibVersion) const;
> @@ -1166,6 +1184,8 @@ public:
> types::ID InputType) const
> override;
> SanitizerMask getSupportedSanitizers() const override;
>
> + void printVerboseInfo(raw_ostream &OS) const override;
> +
> protected:
> void AddSystemIncludeWithSubfolder(const llvm::opt::ArgList
> &DriverArgs,
> llvm::opt::ArgStringList &CC1Args,
> @@ -1179,6 +1199,8 @@ protected:
> private:
> VersionTuple getMSVCVersionFromTriple() const;
> VersionTuple getMSVCVersionFromExe() const;
> +
> + CudaInstallationDetector CudaInstallation;
> };
>
> class LLVM_LIBRARY_VISIBILITY CrossWindowsToolChain : public Generic_GCC {
>
> Modified: cfe/trunk/lib/Driver/Tools.cpp
> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Driver/
> Tools.cpp?rev=291131&r1=291130&r2=291131&view=diff
> ============================================================
> ==================
> --- cfe/trunk/lib/Driver/Tools.cpp (original)
> +++ cfe/trunk/lib/Driver/Tools.cpp Thu Jan 5 10:52:29 2017
> @@ -4086,13 +4086,6 @@ void Clang::ConstructJob(Compilation &C,
> const Driver &D = getToolChain().getDriver();
> ArgStringList CmdArgs;
>
> - bool IsWindowsGNU = getToolChain().getTriple().
> isWindowsGNUEnvironment();
> - bool IsWindowsCygnus =
> - getToolChain().getTriple().isWindowsCygwinEnvironment();
> - bool IsWindowsMSVC = getToolChain().getTriple().
> isWindowsMSVCEnvironment();
> - bool IsPS4CPU = getToolChain().getTriple().isPS4CPU();
> - bool IsIAMCU = getToolChain().getTriple().isOSIAMCU();
> -
> // Check number of inputs for sanity. We need at least one input.
> assert(Inputs.size() >= 1 && "Must have at least one input.");
> const InputInfo &Input = Inputs[0];
> @@ -4106,6 +4099,23 @@ void Clang::ConstructJob(Compilation &C,
> Inputs.size() == 1) &&
> "Unable to handle multiple inputs.");
>
> + bool IsWindowsGNU = getToolChain().getTriple().
> isWindowsGNUEnvironment();
> + bool IsWindowsCygnus =
> + getToolChain().getTriple().isWindowsCygwinEnvironment();
> + bool IsWindowsMSVC = getToolChain().getTriple().
> isWindowsMSVCEnvironment();
> + bool IsPS4CPU = getToolChain().getTriple().isPS4CPU();
> + bool IsIAMCU = getToolChain().getTriple().isOSIAMCU();
> +
> + // Adjust IsWindowsXYZ for CUDA compilations. Even when compiling in
> device
> + // mode (i.e., getToolchain().getTriple() is NVPTX, not Windows), we
> need to
> + // pass Windows-specific flags to cc1.
> + if (IsCuda) {
> + const llvm::Triple *AuxTriple = getToolChain().getAuxTriple();
> + IsWindowsMSVC |= AuxTriple && AuxTriple->isWindowsMSVCEnvironment();
> + IsWindowsGNU |= AuxTriple && AuxTriple->isWindowsGNUEnvironment();
> + IsWindowsCygnus |= AuxTriple && AuxTriple->
> isWindowsCygwinEnvironment();
> + }
> +
> // C++ is not supported for IAMCU.
> if (IsIAMCU && types::isCXX(Input.getType()))
> D.Diag(diag::err_drv_clang_unsupported) << "C++ for IAMCU";
>
> Added: cfe/trunk/test/Driver/Inputs/CUDA-windows/Program Files/NVIDIA GPU
> Computing Toolkit/CUDA/v8.0/bin/.keep
> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/Driver/
> Inputs/CUDA-windows/Program%20Files/NVIDIA%20GPU%
> 20Computing%20Toolkit/CUDA/v8.0/bin/.keep?rev=291131&view=auto
> ============================================================
> ==================
> (empty)
>
> Added: cfe/trunk/test/Driver/Inputs/CUDA-windows/Program Files/NVIDIA GPU
> Computing Toolkit/CUDA/v8.0/include/.keep
> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/Driver/
> Inputs/CUDA-windows/Program%20Files/NVIDIA%20GPU%
> 20Computing%20Toolkit/CUDA/v8.0/include/.keep?rev=291131&view=auto
> ============================================================
> ==================
> (empty)
>
> Added: cfe/trunk/test/Driver/Inputs/CUDA-windows/Program Files/NVIDIA GPU
> Computing Toolkit/CUDA/v8.0/lib/.keep
> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/Driver/
> Inputs/CUDA-windows/Program%20Files/NVIDIA%20GPU%
> 20Computing%20Toolkit/CUDA/v8.0/lib/.keep?rev=291131&view=auto
> ============================================================
> ==================
> (empty)
>
> Added: cfe/trunk/test/Driver/Inputs/CUDA-windows/Program Files/NVIDIA GPU
> Computing Toolkit/CUDA/v8.0/nvvm/libdevice/libdevice.compute_30.10.bc
> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/Driver/
> Inputs/CUDA-windows/Program%20Files/NVIDIA%20GPU%
> 20Computing%20Toolkit/CUDA/v8.0/nvvm/libdevice/libdevice.
> compute_30.10.bc?rev=291131&view=auto
> ============================================================
> ==================
> (empty)
>
> Added: cfe/trunk/test/Driver/Inputs/CUDA-windows/Program Files/NVIDIA GPU
> Computing Toolkit/CUDA/v8.0/nvvm/libdevice/libdevice.compute_35.10.bc
> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/Driver/
> Inputs/CUDA-windows/Program%20Files/NVIDIA%20GPU%
> 20Computing%20Toolkit/CUDA/v8.0/nvvm/libdevice/libdevice.
> compute_35.10.bc?rev=291131&view=auto
> ============================================================
> ==================
> (empty)
>
> Added: cfe/trunk/test/Driver/cuda-windows.cu
> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/Driver/
> cuda-windows.cu?rev=291131&view=auto
> ============================================================
> ==================
> --- cfe/trunk/test/Driver/cuda-windows.cu (added)
> +++ cfe/trunk/test/Driver/cuda-windows.cu Thu Jan 5 10:52:29 2017
> @@ -0,0 +1,14 @@
> +// REQUIRES: clang-driver
> +// REQUIRES: x86-registered-target
> +// REQUIRES: nvptx-registered-target
> +//
> +// RUN: %clang -v --target=i386-pc-windows-msvc \
> +// RUN: --sysroot=%S/Inputs/CUDA-windows 2>&1 %s -### | FileCheck %s
> +// RUN: %clang -v --target=i386-pc-windows-mingw32 \
> +// RUN: --sysroot=%S/Inputs/CUDA-windows 2>&1 %s -### | FileCheck %s
> +
> +// CHECK: Found CUDA installation: {{.*}}/Inputs/CUDA-windows/Program
> Files/NVIDIA GPU Computing Toolkit/CUDA/v8.0
> +// CHECK: "-cc1" "-triple" "nvptx-nvidia-cuda"
> +// CHECK-SAME: "-fms-extensions"
> +// CHECK-SAME: "-fms-compatibility"
> +// CHECK-SAME: "-fms-compatibility-version=
>
>
> _______________________________________________
> cfe-commits mailing list
> cfe-commits at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20170106/ad25e920/attachment-0001.html>
More information about the cfe-commits
mailing list