[OpenMP] Enable the existing nocudalib flag for OpenMP offloading toolchain.

[gtbercea]

Summary

Enable the -nocudalib flag for the OpenMP device offloading toolchain as well. Currently it can only be used for the CUDA toolchain.

[hfinkel]

Please add a test case.

[gtbercea]

Add test.

[tra]

lib/Driver/ToolChains/Cuda.cpp

255–257

The purpose of the original assert was to catch a programming error and this change negates that purpose.
Perhaps I'm missing something. Could you elaborate on what's the motivation for this particular change?

I don't understand why it would be OK to end up with an unknown GPU architecture if -nocudalib is specified.
You still do want to pass *some* specific GPU arch to ptxas and that has nothing to do with whether you happen to have suitable libdevice.

[gtbercea]

Don't take into account unknown CUDA archs not even for testing purposes.

[tra]

One small nit. LGTM otherwise.

test/Driver/openmp-offload-gpu.c

133

Please split this RUN line further.

[gtbercea]

Split line.

lib/Driver/ToolChains/Cuda.cpp

void NVPTX::Assembler::ConstructJob(Compilation &C, const JobAction &JA,
   // flag or the default value.
   if (JA.isDeviceOffloading(Action::OFK_OpenMP)) {
     GPUArchName = Args.getLastArgValue(options::OPT_march_EQ);
     assert(!GPUArchName.empty() && "Must have an architecture passed in.");
   } else
     GPUArchName = JA.getOffloadingArch();
 
   // Obtain architecture from the action.
   CudaArch gpu_arch = StringToCudaArch(GPUArchName);
-  assert(gpu_arch != CudaArch::UNKNOWN &&
+  assert((gpu_arch != CudaArch::UNKNOWN ||
+          Args.hasArg(options::OPT_nocudalib)) &&
          "Device action expected to have an architecture.");
 
   // Check that our installation's ptxas supports gpu_arch.
   if (!Args.hasArg(options::OPT_no_cuda_version_check)) {
     TC.CudaInstallation.CheckCudaVersionSupportsArch(gpu_arch);
   }
 
   ArgStringList CmdArgs;
if (DeviceOffloadingKind == Action::OFK_Cuda) {
 
     if (DriverArgs.hasFlag(options::OPT_fcuda_flush_denormals_to_zero,
                            options::OPT_fno_cuda_flush_denormals_to_zero, false))
       CC1Args.push_back("-fcuda-flush-denormals-to-zero");
 
     if (DriverArgs.hasFlag(options::OPT_fcuda_approx_transcendentals,
                            options::OPT_fno_cuda_approx_transcendentals, false))
       CC1Args.push_back("-fcuda-approx-transcendentals");
+  }
 
-    if (DriverArgs.hasArg(options::OPT_nocudalib))
-      return;
-  }
 
   std::string LibDeviceFile = CudaInstallation.getLibDeviceFile(GpuArch);
 
   if (LibDeviceFile.empty()) {
     getDriver().Diag(diag::err_drv_no_cuda_libdevice) << GpuArch;
     return;
   }
 

test/Driver/openmp-offload-gpu.c

 /// ###########################################################################
 
 /// PTXAS is passed -c flag by default when offloading to an NVIDIA device using OpenMP
 /// Check that the flag is passed when -fopenmp-relocatable-target is used.
 // RUN:   %clang -### -fopenmp=libomp -fopenmp-targets=nvptx64-nvidia-cuda -fopenmp-relocatable-target -save-temps -no-canonical-prefixes %s 2>&1 \
 // RUN:   | FileCheck -check-prefix=CHK-PTXAS-RELO %s
 
 // CHK-PTXAS-RELO: ptxas{{.*}}" "-c"
+
+/// ###########################################################################
+
+/// Check that error is not thrown by toolchain when no cuda lib flag is used.
+/// Check that the flag is passed when -fopenmp-relocatable-target is used.
+// RUN:   %clang -### -fopenmp=libomp -fopenmp-targets=nvptx64-nvidia-cuda -Xopenmp-target -march=sm_99 -nocudalib -fopenmp-relocatable-target -save-temps -no-canonical-prefixes %s 2>&1 \
+// RUN:   | FileCheck -check-prefix=CHK-FLAG-NOLIBDEVICE %s
+
+// CHK-FLAG-NOLIBDEVICE-NOT: error:{{.*}}sm_99