Trouble in running code with GPU on Jean-Zay
I still have trouble in running code on GPU on Jean-Zay:
I am on version: 7e8441b7
cmake -DCMAKE_BUILD_TYPE=Release -DKokkos_ENABLE_CUDA=On -DKokkos_ARCH_VOLTA70=On -DGENERATE_DOC=Off -DKokkos_ENABLE_OPENMP=On -DKokkos_ENABLE_CUDA_LAMBDA=On ..
to compile:
srun -p compil -c 10 --hint=nomultithread -A oth@cpu make -j 10
then:
srun -C v100-16g --hint=nomultithread -A oth@v100 -c 10 --gres=gpu:1 ~/fargOCA/buildGPU/fargoInit . config.info
output:
[r10i3n1:512100] *** Process received signal ***
[r10i3n1:512100] Signal: Segmentation fault (11)
[r10i3n1:512100] Signal code: Address not mapped (1)
[r10i3n1:512100] Failing at address: 0x7260
[r10i3n1:512100] [ 0] /usr/lib64/libpthread.so.0(+0x12b20)[0x14fa99f6cb20]
[r10i3n1:512100] [ 1] /usr/lib64/libc.so.6(+0x15d627)[0x14fa9950d627]
[r10i3n1:512100] [ 2] /gpfslocalsup/spack_soft/openmpi/4.1.1/gcc-8.3.1-buyiit4vlnfnuq6vgvlsmlkgexrh6myv/lib/libopen-pal.so.40(opal_argv_join+0x39)[0x14fa98560bf9]
[r10i3n1:512100] [ 3] /gpfslocalsup/spack_soft/openmpi/4.1.1/gcc-8.3.1-buyiit4vlnfnuq6vgvlsmlkgexrh6myv/lib/libmpi.so.40(ompi_mpi_init+0xb96)[0x14fa9cd5cb56]
[r10i3n1:512100] [ 4] /gpfslocalsup/spack_soft/openmpi/4.1.1/gcc-8.3.1-buyiit4vlnfnuq6vgvlsmlkgexrh6myv/lib/libmpi.so.40(PMPI_Init_thread+0x55)[0x14fa9cba62c5]
[r10i3n1:512100] [ 5] /gpfslocalsup/spack_soft/boost/1.74.0/gcc-8.4.1-gofcq4cypxij7bu5t5iezakwx32zjz6r/lib/libboost_mpi.so.1.74.0(_ZN5boost3mpi11environmentC1ERiRPPcNS0_9threading5levelEb+0x41)[0x14fa9b4acdd1]