作者: Jintao Meng , Sangmin Seo , Pavan Balaji , Yanjie Wei , Bingqiang Wang
DOI: 10.1109/ICPP.2016.29
关键词:
摘要: … usage and efficiency, we fix the block size to be 64 MB in the following test. With 16,384 cores, SWAP2 achieves 16X speedup with the optimization of the input parallelization. The I/O …