Enabling Fortran Standard Parallelism in GAMESS for Accelerated Quantum Chemistry Calculations

The performance of Fortran 2008 DO CONCURRENT (DC) relativetoOpenACC and OpenMP target offloading (OTO) with different compilersis studied for the GAMESS quantum chemistry application. Specifically,DC and OTO are used to offload the Fock build, which is a computationalbottleneck in most quantum chemistry codes, to GPUs. The DC Fock buildperformance is studied on NVIDIA A100 and V100 accelerators and comparedwith the OTO versions compiled by the NVIDIA HPC, IBM XL, and CrayFortran compilers. The results show that DC can speed up the Fockbuild by 3.0x compared with that of the OTO model. With similaroffloading efforts, DC is a compelling programming model for offloadingFortran applications to GPUs.

Automated summary

The performance of Fortran 2008 DO CONCURRENT (DC) is compared with OpenACC and OpenMP target offloading (OTO) for the GAMESS quantum chemistry application. The study focuses on offloading the Fock build, a computational bottleneck in quantum chemistry codes, to GPUs using DC and OTO. The results show that DC can achieve a 3.0x speedup compared to OTO, making it a compelling programming model for offloading Fortran applications to GPUs.