I have an allocation obtained by $ salloc -N 1 -C knl,snc4,flat
Now I’d like to run a 4-threaded task on CPUs 0,1,2, & 3. Why does this not work? $ export OMP_NUM_THREADS=4 $ srun -n 1 --cpu_bind=map_cpu:0,1,2,3 ./omp_overhead.INTEL Rank 0, thread 0 of 4 on CPU 0, core 0, NUMA node 0 (nid00196). Rank 0, thread 1 of 4 on CPU 0, core 0, NUMA node 0 (nid00196). Rank 0, thread 2 of 4 on CPU 0, core 0, NUMA node 0 (nid00196). Rank 0, thread 3 of 4 on CPU 0, core 0, NUMA node 0 (nid00196). -- David Gunter CCS-7 Los Alamos National Laboratory
