A Well-Scaling Parallel FEM-DDM Algorithm for Electromagnetic Simulation on SW26010 Many-core Processor

Autor: Xin-Qing Sheng, Wei-Jia He, Ming-Lin Yang, Rui-Qing Liu, Jian-Zhong Lai, Wu Wang
Rok vydání: 2020
Předmět:
Zdroj: 2020 IEEE MTT-S International Conference on Numerical Electromagnetic and Multiphysics Modeling and Optimization (NEMO).
DOI: 10.1109/nemo49486.2020.9343550
Popis: We present in this paper a well-scaling parallel approach of the finite element-domain decomposition method (FEM-DDM) for solving large scale 3D scattering problems on many-core SW26010 processors. In this parallel approach, the whole solution domain is first decomposed in a two-level hierarchical way to minimize the inter process communication. Then the FEM matrix of each unit-level subdomain to be solved in the iterative solution of the interface problem from FEM-DDM, is solved via another sparse approximate inverse (SPAI) preconditioner accelerated iterative solver. This inner–outer iteration scheme is further accelerated by exploring many-core parallelization, with specially designed sparse matrix vector multiplications to improve the data access efficiency. The performance achievements are investigated in terms of accuracy, speed up, and scalability.
Databáze: OpenAIRE