Parallel Algorithm Optimization of Superviscosity Factor on Heterogeneous Multicore Processors

EasyChair Preprint 1121

Abstract

As a widely used earth simulation system, CESM provides an indispensable quantitative basis for the study of earth system changes and the interaction of human activities. In reality, it relies on high performance computing devices such as supercomputers, but there is still a low degree of parallelism in the process of transplanting CESM to heterogeneous supercomputers. Therefore, in order to solve the problem that the finite difference algorithm in CESM has a large amount of redundancy when SW26010 heterogeneous many-core processor reads the memory, different optimization schemes are designed according to the functions of different data structures. As the core application of CESM in high performance computer, the parallel performance of the algorithm has been improved significantly.

Keyphrases: CESM, Heterogeneous Multicore CPU, High Performance Computing, finite difference algorithm, parallel computation

@booklet{EasyChair:1121,
  author    = {Hongbo Chen and Xuezhong Qian and Lin Gan and Jingheng Xu and Xu Liu},
  title     = {Parallel Algorithm Optimization of Superviscosity Factor on Heterogeneous Multicore Processors},
  howpublished = {EasyChair Preprint 1121},
  year      = {EasyChair, 2019}}