A portable OpenCL-based unstructured edge-based Finite Element Navier-Stokes solver on graphics hardware

摘要

The rise of GPUs in modern high-performance systems increases the interest in porting\udportion of codes to such hardware. The current paper aims to explore the performance\udof a portable state-of-the-art FE solver on GPU accelerators. Performance evaluation is\uddone by comparing with an existing highly-optimized OpenMP version of the solver.\udCode portability is ensured by writing the program using the OpenCL 1.1 specifications,\udwhile performance portability is sought through an optimization step performed\udat the beginning of the calculations to find out the optimal parameter set for the solver.\udThe results show that the new implementation can be several times faster than the\udOpenMP version.