I am preparing for a gpu hackathon this week, and I am trying to get a working case on perlmutter using PrgEnv-nvidia , and on the GPU. I tried using the run_cmake_build_perlmutter_nvidia_nvshmem.sh and batch_script_mpi_runit_perlmutter_3dsolve_nvidia_nvshmem_complex.sh scripts (with the latter script modified to read in my small 30 x 30 sample matrix with 33 nonzeros, and with the part that uses export SUPERLU_ACC_SOLVE=1 and pzdrive3d uncommented, and I am only using one node with 4 gpus), but I am getting a segmentation fault. If you could let me know any advice on how to get these cases to work, that would be great. (Like, if those scripts need to be modified in some way, or if there is something I am not doing). Do I also need to load the nvshem module? (that part is commented out of the run script). Thanks.
I am preparing for a gpu hackathon this week, and I am trying to get a working case on perlmutter using PrgEnv-nvidia , and on the GPU. I tried using the run_cmake_build_perlmutter_nvidia_nvshmem.sh and batch_script_mpi_runit_perlmutter_3dsolve_nvidia_nvshmem_complex.sh scripts (with the latter script modified to read in my small 30 x 30 sample matrix with 33 nonzeros, and with the part that uses export SUPERLU_ACC_SOLVE=1 and pzdrive3d uncommented, and I am only using one node with 4 gpus), but I am getting a segmentation fault. If you could let me know any advice on how to get these cases to work, that would be great. (Like, if those scripts need to be modified in some way, or if there is something I am not doing). Do I also need to load the nvshem module? (that part is commented out of the run script). Thanks.