[Clang] Only emit CUDA version warnings when creating the CUDA toolchain

[jhuber6]

Summary

This warning primarily applies to users of the CUDA langues as there may
be new features we rely on. The other two users of the toolchain are
OpenMP via -fopenmp --offload-arch=sm_70 and a cross-compiled build
via --target=nvptx64-nvida-cuda -march=sm_70. Both of these do not
rely directly on things that would change significantly between CUDA
versions, and the way they are built can sometims make this warning
print many times.

This patch changees the behaiour to only check for the version when
building for CUDA offloading specifically, the other two will not have
this check.

[jhuber6]

Fix wrong comment

[jdoerfert]

I like it

[jdoerfert]

tests?

[jhuber6]

Add test

clang/lib/Driver/Driver.cpp

if (IsCuda) {
     if (!CudaTriple)
       return;
     // Use the CUDA and host triples as the key into the ToolChains map,
     // because the device toolchain we create depends on both.
     auto &CudaTC = ToolChains[CudaTriple->str() + "/" + HostTriple.str()];
     if (!CudaTC) {
       CudaTC = std::make_unique<toolchains::CudaToolChain>(
           *this, *CudaTriple, *HostTC, C.getInputArgs());
+
+      // Emit a warning if the detected CUDA version is too new.
+      CudaInstallationDetector &CudaInstallation =
+          static_cast<toolchains::CudaToolChain &>(*CudaTC).CudaInstallation;
+      if (CudaInstallation.isValid())
+        CudaInstallation.WarnIfUnsupportedVersion();
     }
     C.addOffloadDeviceToolChain(CudaTC.get(), OFK);
   } else if (IsHIP) {
     if (auto *OMPTargetArg =
             C.getInputArgs().getLastArg(options::OPT_fopenmp_targets_EQ)) {
       Diag(clang::diag::err_drv_unsupported_opt_for_language_mode)
           << OMPTargetArg->getSpelling() << "HIP";
       return;

clang/lib/Driver/ToolChains/Cuda.cpp

 /// NVPTX toolchain. Our assembler is ptxas, and our linker is nvlink. This
 /// operates as a stand-alone version of the NVPTX tools without the host
 /// toolchain.
 NVPTXToolChain::NVPTXToolChain(const Driver &D, const llvm::Triple &Triple,
                                const llvm::Triple &HostTriple,
                                const ArgList &Args, bool Freestanding = false)
     : ToolChain(D, Triple, Args), CudaInstallation(D, HostTriple, Args),
       Freestanding(Freestanding) {
-  if (CudaInstallation.isValid()) {
-    CudaInstallation.WarnIfUnsupportedVersion();
+  if (CudaInstallation.isValid())
     getProgramPaths().push_back(std::string(CudaInstallation.getBinPath()));
-  }
   // Lookup binaries into the driver directory, this is used to
   // discover the 'nvptx-arch' executable.
   getProgramPaths().push_back(getDriver().Dir);
 }
 
 /// We only need the host triple to locate the CUDA binary utilities, use the
 /// system's default triple if not provided.
 NVPTXToolChain::NVPTXToolChain(const Driver &D, const llvm::Triple &Triple,

clang/test/Driver/cuda-version-check.cu

 // ERR_SM60: error: GPU arch sm_60 {{.*}}
 // ERR_SM60-NOT: error: GPU arch sm_60
 
 // ERR_SM61: error: GPU arch sm_61 {{.*}}
 // ERR_SM61-NOT: error: GPU arch sm_61
 
 // UNKNOWN_VERSION: CUDA version is newer than the latest{{.*}} supported version
 // UNKNOWN_VERSION_CXX-NOT: unknown CUDA version
+
+// Check to make sure we do not emit these warnings for OpenMP or cross-compilation.
+// RUN: %clang --target=x86_64-linux -v -### -fopenmp -nogpulib --offload-arch=sm_60 --cuda-path=%S/Inputs/CUDA-new/usr/local/cuda 2>&1 -x c %s | \
+// RUN:    FileCheck %s --check-prefix=VERSION
+// RUN: %clang --target=nvptx64-nvidia-cuda -v -### -nogpulib -march=sm_60 --cuda-path=%S/Inputs/CUDA-new/usr/local/cuda 2>&1 -x c %s | \
+// RUN:    FileCheck %s --check-prefix=VERSION
+// VERSION-NOT: CUDA version is newer than the latest{{.*}} supported version
+