building with nvcc and clang (or clang based XL) as host compiler: "Kokkos::atomic_fetch_min(volatile int *, int)" has already been defined

[ndellingwood]

Compilation error:

13:19:07 /home/jenkins/white/workspace/Kokkos_develop_white_cuda_92_xl_161_kepler/kokkos/core/src/impl/Kokkos_Atomic_MinMax.hpp(104): error: function "Kokkos::atomic_fetch_min(volatile int *, int)" has already been defined

Reproducer sample:

   #   Load modules:
         module load cmake/3.12.3 cuda/9.2.88 gcc/7.2.0 ibm/xl/16.1.1

   #   Use generate_makefile line below to call cmake which generates makefile for this build:

         <KOKKOS_PATH>/generate_makefile.bash --with-devices=Cuda,OpenMP --arch=Power8,Kepler37 --compiler=<KOKKOS_PATH>/bin/nvcc_wrapper --cxxflags="-O3 -Wall -Wunused-parameter -Wshadow -pedantic -Werror -Wsign-compare -Wtype-limits -Wuninitialized -ccbin xlc++" --cxxstandard="11" --ldflags="" --with-cuda --kokkos-path=<KOKKOS_PATH> --no-examples --with-options=enable_large_mem_tests

[crtrott]

Fix is in #2904

[ndellingwood]

@crtrott the fix from #2904 now allows the failing build to compile, but the openmp.Random_XorShift64 test has now failed two consecutive rounds of testing (very unlikely to occur), should a new issue be opened for that?

https://jenkins-son.sandia.gov/job/Kokkos_develop_white_cuda_92_xl_161_kepler/809/display/redirect

https://jenkins-son.sandia.gov/job/Kokkos_develop_white_cuda_92_xl_161_kepler/810/display/redirect

[crtrott]

Yeah I am looking into those as well as a segfault ...

[masterleinad]

Hmm.. Can't reproduce on Summit.