A Mixed Precision Direct Electromagnetic Finite Element Solver on GPUs

In this paper, we present a mixed-precision direct electromagnetic finite element solver on GPU machines. In our solver, we propose a mixed-precision factorization algorithm aimed at obtaining results of the same quality as the double-precision algorithm at a much lower cost. In addition, to accelerate the direct solution of finite element sparse linear systems using GPU machines more efficiently, we propose a novel pipelined offloading technique that utilizes CUDA streams to achieve pipeline scheduling with overlapping computations and data transfers. Our numerical experiments show that by combining our proposed mixed-precision algorithm with the pipelined offloading technique, our proposed solver produces exact solutions with tremendous speedup.

Authors

Institutions

Publication Details

Published
2024-08-16
DOI
https://doi.org/10.1109/aces-china62474.2024.10699568
Primary Topic
Electromagnetic Simulation and Numerical Methods
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

A Mixed Precision Direct Electromagnetic Finite Element Solver on GPUs

Xiaojie Dang, Sheng Zuo, Junjie Wang, Zhiming Li et al.
Electromagnetic Simulation and Numerical Methods
article

A Mixed Precision Direct Electromagnetic Finite Element Solver on GPUs

Xiaojie Dang, Sheng Zuo, Junjie Wang, Zhiming Li, Shugang Jiang
article en

Abstract

In this paper, we present a mixed-precision direct electromagnetic finite element solver on GPU machines. In our solver, we propose a mixed-precision factorization algorithm aimed at obtaining results of the same quality as the double-precision algorithm at a much lower cost. In addition, to accelerate the direct solution of finite element sparse linear systems using GPU machines more efficiently, we propose a novel pipelined offloading technique that utilizes CUDA streams to achieve pipeline scheduling with overlapping computations and data transfers. Our numerical experiments show that by combining our proposed mixed-precision algorithm with the pipelined offloading technique, our proposed solver produces exact solutions with tremendous speedup.

Xidian University (CN)
Fundamental Research Funds for the Central Universities, Research and Development
Affordable and clean energy
Openalex Percentile: Top 85%
Electromagnetic Simulation and Numerical Methods
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.