Hi,
I have a 2 socket 20 cores per socket (ntel(R) Xeon(R) Gold 6148 CPU) node .
I wish to launch 1 process per socket and 20 threads per process and if possible - all threads should be pinned to their respective cores.
earlier i used to run intel binaries on cray machine with similar cores , and the syntax was -
aprun –n (mpi tasks) –N (tasks per node) –S (tasks per socket) –d (thread depth) <executable> , example -
OMP_NUM_THREADS=20
aprun -n4 -N2 -S1 -d $OMP_NUM_THREADS ./a.out
node 0 socket 0 process#0 nprocs 4 thread id 0 nthreads 20 core id 0 node 0 socket 0 process#0 nprocs 4 thread id 1 nthreads 20 core id 1 .... node 0 socket 0 process#0 nprocs 4 thread id 19 nthreads 20 core id 19 node 0 socket 1 process#1 nprocs 4 thread id 0 nthreads 20 core id 20 ... node 0 socket 0 process#1 nprocs 4 thread id 19 nthreads 20 core id 39 .... node 1 socket 0 process#1 nprocs 4 thread id 19 nthreads 20 core id 39
How can i achieve the same/equivalent effect using intel's mpirun?