LLNL
Remove hard-coded C++11 in CUDA flags
https://github.com/LLNL/blt/blob/develop/cmake/thirdparty/SetupCUDA.cmake#L85
Since CAMP, RAJA, CHAI, Umpire are requiring C++14 in their next release (which should be very soon), this would be great to get in soon.
If it's a simple matter of replacing that hardcoded flag with a cmake variable, I could put in a merge request. Is there a BLT_CUDA_STD variable?
Related logic here:
https://github.com/LLNL/blt/blob/e35f490a8a8b1689e99b5f4308b5251f97eb36cf/cmake/SetupCompilerOptions.cmake#L168-L179
And defining std flags here:
https://github.com/LLNL/blt/blob/e35f490a8a8b1689e99b5f4308b5251f97eb36cf/cmake/SetupCompilerOptions.cmake#L259-L322
Ah, so you set CMAKE_CUDA_STANDARD based off of BLT_CXX_STD. Would it work to use "-std=${BLT_CXX_STD}"? Or is there a complication I'm missing?
I think that should work. Not all compilers use the standardized flag but NVCC uses the normal flags like BLT accepts. Is there a case where you wouldn't want them to be the same? Like C++17 for everything but the CUDA specific code... I'm probably over thinking this and we can solve this later if it comes up.