Enable OPENMP in our CUDA host-configs

[kennyweiss]

I just noticed that OpenMP is disabled in our CUDA host-configs on blueos

My understanding is that they're compatible, and I thought OpenMP was enabled in these configurations.

Did we ever have it enabled? Is there any reason not to enable OpenMP in at least one CUDA host-config?

Apparently, it's been this way for a while (always?):

• https://github.com/LLNL/axom/blame/develop/scripts/spack/specs.json

[white238]

I mistakenly turned this on via a copy/paste error (maybe) but you get the following error when you don't turn it off with xlf:

-- OpenMP Support is ON
-- Found OpenMP_C: -fopenmp=libomp (found version "3.1") 
-- Found OpenMP_CXX: -fopenmp=libomp (found version "3.1") 
CMake Error at /usr/tce/packages/cmake/cmake-3.14.5/share/cmake/Modules/FindPackageHandleStandardArgs.cmake:137 (message):
  Could NOT find OpenMP_Fortran (missing: OpenMP_Fortran_FLAGS
  OpenMP_Fortran_LIB_NAMES)
Call Stack (most recent call first):
  /usr/tce/packages/cmake/cmake-3.14.5/share/cmake/Modules/FindPackageHandleStandardArgs.cmake:378 (_FPHSA_FAILURE_MESSAGE)
  /usr/tce/packages/cmake/cmake-3.14.5/share/cmake/Modules/FindOpenMP.cmake:474 (find_package_handle_standard_args)
  cmake/blt/cmake/thirdparty/SetupOpenMP.cmake:11 (find_package)
  cmake/blt/cmake/thirdparty/SetupThirdParty.cmake:38 (include)
  cmake/blt/SetupBLT.cmake:112 (include)
  CMakeLists.txt:64 (include)
-- Configuring incomplete, errors occurred!

Knee deep in another failure. Will investigate sometime.

[kennyweiss]

To clarify, the intent of this issue was for cuda+openmp with our main compilers on LLNL's blueos clusters: the IBM XL compiler and clang.