[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

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@291131 91177308-0d34-0410-b5e6-96231b3b80d8

include/clang/Driver/ToolChain.h

@@ -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(); }

lib/Driver/MSVCToolChain.cpp

@@ -47,9 +47,9 @@ using namespace clang::driver::toolchains;
 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() const {
   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) {

lib/Driver/MinGWToolChain.cpp

@@ -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

lib/Driver/ToolChains.cpp

@@ -1810,14 +1810,21 @@ CudaInstallationDetector::CudaInstallationDetector(
     : 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::getSupportedSanitizers() const {
   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)

lib/Driver/ToolChains.h

@@ -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 {

lib/Driver/Tools.cpp

@@ -4086,13 +4086,6 @@ void Clang::ConstructJob(Compilation &C, const JobAction &JA,
   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, const JobAction &JA,
           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";

test/Driver/cuda-windows.cu

@@ -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=