Heterogeneous Parallel DGTD Calculation Method Based on DSP Clusters
This paper is based on a domestic supercomputer to evaluate the performance of the parallel discontinuous Galerkin time domain (DGTD) method. In conjunction with the MT3000 digital signal processor (DSP) platform, this paper introduces parallel schemes at both coarse-grained process level and fine-grained thread level, and proposes a method for accelerating the DGTD matrix solution on a CPU-DSP heterogeneous platform. The test results indicate that when expanded from 8 processes to 160 processes, the method maintains a parallel efficiency of 50.6% while ensuring accuracy.
Authors
- Huaguang Bao (ORCID: https://orcid.org/0000-0001-9609-247X)
- Dazhi Ding (ORCID: https://orcid.org/0000-0001-5550-1812)
- Run Zhang (ORCID: https://orcid.org/0000-0002-0943-824X)
- Mengmeng Li (ORCID: https://orcid.org/0000-0002-1653-6753)
- Chunyu Li (ORCID: https://orcid.org/0000-0001-8124-2498)
- Tiancheng Zhang (ORCID: https://orcid.org/0000-0001-5592-5893)
Institutions
- Nanjing University of Science and Technology (CN)
Publication Details
- Published
- 2024-11-27
- DOI
- https://doi.org/10.1109/mape62875.2024.10813900
- Primary Topic
- Optical Systems and Laser Technology
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Natural Science Foundation of China
- National Key Research and Development Program of China