While doing some benchmarking of an application from 32 ranks up to 512 ranks (increments in powers of 2) using an Intel 14 compiler build of OpenMPI 1.6.5, we are finding improved performance if '-bysocket -bind-to-socket' is spec'd to mpirun. The application uses a variety of MPI-1 subroutines.
I am curious if a study was done of the performance of MPI-1 subroutines (operations) under various processor affinity specifications like the one I mentioned. --john
_______________________________________________ users mailing list users@lists.open-mpi.org https://rfd.newmexicoconsortium.org/mailman/listinfo/users
