An efficient MPI/OpenMP parallelization of the Hartree-Fock method for the second generation of Intel Xeon Phi processor

摘要

Modern OpenMP threading techniques are used to convert the MPI-onlyHartree-Fock code in the GAMESS program to a hybrid MPI/OpenMP algorithm. Twoseparate implementations that differ by the sharing or replication of key datastructures among threads are considered, density and Fock matrices. Allimplementations are benchmarked on a super-computer of 3,000 Intel Xeon Phiprocessors. With 64 cores per processor, scaling numbers are reported on up to192,000 cores. The hybrid MPI/OpenMP implementation reduces the memoryfootprint by approximately 200 times compared to the legacy code. TheMPI/OpenMP code was shown to run up to six times faster than the original for arange of molecular system sizes.