GPU offloading of a large-scale gyrokinetic particle-in-cell Fortran code: From OpenACC to OpenMP

Workshop:WACCPD 2021: Eighth Workshop on Accelerator Programming Using Directives

Authors: Qiheng Cai, Junyi Cheng, and Yang Chen (University of Colorado, Boulder); Marcus Wagner (Hewlett Packard Enterprise); Christopher Daley (Lawrence Berkeley National Laboratory (LBNL)); Dossay Oryspayev (Brookhaven National Laboratory); and Stefan Tirkas, Sophie Redd, and Scott Parker (University of Colorado, Boulder)

Abstract: GPU offloading of a large-scale gyrokinetic particle-in-cell Fortran code is converted from using OpenACC to using OpenMP. Particle pushing and deposition are completely offloaded to GPU. Performance is compared between CPU and GPU, and between OpenACC and OpenMP. Good weak scaling (increasing particle number with fixed grid number) is obtained. Issues encountered when porting OpenMP GPU offloading are discussed.

Back to WACCPD 2021: Eighth Workshop on Accelerator Programming Using Directives Archive Listing

Back to Full Workshop Archive Listing