Nuclear fusion simulation code optimization on GPU clusters

藤田 典久*; 奴賀 秀男*; 朴 泰祐*; 井戸村 泰宏

Proceedings of 19th IEEE International Conference on Parallel and Distributed Systems (ICPDCS 2013) (USB Flash Drive), 2 Pages, 2013/12

A fusion plasma turbulence simulation code GT5D is optimized for GPU clusters with multiple GPUs on a node. We get 3.37 times faster performance in maximum in function level evaluation, and 2.03 times faster performance in total than the case of CPU-only execution on the HA-PACS GPU cluster. It includes 53% performance gain with overlapping MPI communications and GPU calculations.