Learning Sparse Matrix Row Permutations for Efficient SpMM on GPU Architectures
2021 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS)(2021)
摘要
Achieving peak performance on sparse operations is challenging. The distribution of the non-zero elements and underlying hardware platform affect the execution efficiency. Given the diversity in workloads and architectures, no unique solution always wins. In this paper, we improve SpMM efficiency on GPUs. We propose several simple, but effective, sparse data permutations on the CSR data structure....
更多查看译文
关键词
Graphics processing units,Predictive models,Performance gain,Data structures,Hardware,Software,Computational efficiency
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