Performance Portability Evaluation for OpenACC on Intel Knights Corner and Nvidia Kepler
OpenACC is a programming standard designed to simplify heterogeneous parallel programming by using direc- rives Since OpenACC can generate OpenCL and CUDA code lneanwhile running OpenCL on Intel Knight Corner is supported by CAPSHMPP compiler it is attractive to using OpenACC on hardwares with different underlying uricro- architectures This paper studies how realistic it is to use a single OpenACC source code for a set of hardwares with different underlying micro-architectures Iritel Knight Corner and Nvidia Kepler products are the targets in the experiment since they are with the latest architectures and have similar peak performance Meanwhile CAPS OpenACC compilerls usedto compile EPCCO penACC benchulark suite,Stream andMaxFlops of SHOC benchularks to access the pefomlance Tb study the performance portahility.roof line model and relalive performance model arebuilt by the data of experiments This paper shows that at most 82%performance compared with peak performance on Kepler and Knight Comer is achieved by specific benchularks,but as the rise of aritlmletic intensity the average performance is approximately 10%And there is a big performance gap between Intel Knight Corner and Nvidia Kepler on several benchmarks This study confirms that performance portability of OpenACC is related to the arithmetic intensity and a big performance gap still exsits in specific benchmarks between different hardware platfomls.
OpenACC Performance portability GPU MIC
Yichao Wang Qiang Qin Simon Chong Wee SEE James Lin
Center for High Performance Computing,Shanghai Jiao Tong University,Shanghai 200240,China Center for High Performance Computing,Shanghai Jiao Tong University,Shanghai 200240,China;NVIDIA Cor Center for High Performance Computing,Shanghai Jiao Tong University,Shanghai 200240,China;Tokyo Inst
国内会议
桂林
英文
419-426
2013-10-29（万方平台首次上网日期，不代表论文的发表时间）