Efficient and Eventually Consistent Collective Operations

2021 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW)(2021)

引用 1|浏览18
暂无评分
摘要
Collective operations are common features of parallel programming models that are frequently used in High-Performance (HPC) and machine/ deep learning (ML/ DL) applications. In strong scaling scenarios, collective operations can negatively impact the overall application performance: with the increase in core count, the load per rank decreases, while the time spent in collective operations increase...
更多
查看译文
关键词
Deep learning,Distributed processing,Parallel programming,Biological system modeling,Conferences,Computational modeling,Ecosystems
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要