Performance Assessment of A Multi-block Incompressible Navier-Stokes Solver using Directive-based GPU Programming in a Cluster Environment

OpenACC, a directive-based GPU programing standard, is emerging as a promis-ing technology for massively-parallel accelerators, such as General-purpose computing on graphics processing units (GPGPU), Accelerated Processing Unit (APU) and Many Inte-grated Core Architecture (MIC). The heterogeneous nature of these accelerators call for careful designs of parallel algorithms and data management, which imposes a great hur-dle for general scientific computation. Similar to OpenMP, the directive-based approach of OpenACC hides many underlying implementation details, thus significantly reduces the programming complexity and increases code portability. However, many challenges remain, due to the very high granularity of GPGPU and the relatively narrow intercon-nection bandwidth among GPUs, which is particularly restrictive when cross-node data exchange is involved in a cluster environment. In our study, a multi-block incompressible Navier-Stokes solver is ported for GPGPU using OpenACC and MVAPICH2. A perfor-mance analysis is carried out based on the profiling of this solver running in a InfiniBand cluster with nVidia GPUs, which helps to identify the potentials of directive-based GPU programming and directions for further improvement.