On Quantum Chemistry Code Adaptation for RSC PetaStream Architecture

Molecular simulations with quantum chemistry methods consume a large portion of CPU cycles in modern high-performance computing centers. Evolution of modern processors and HPC architectures necessitates adaptation of software to new hardware generations. The present work concentrates on the optimization of the widely used GAMESS code to Intel Xeon Phi architecture and recently devised RSC PetaStream platform. Since improvement in parallelization is required, the most frequently used Hartree-Fock and DFT methods are explored for additional parallelization options. The Xeon Phi requires vectorization that is important for electron-repulsion integrals (ERI) calculations to achieve good performance.