Iterative Krylov Methods for Acoustic Problems on Graphics Processing Unit

TL;DR

This paper explores the implementation of iterative Krylov methods on GPUs for solving complex acoustic problems, demonstrating significant speed-ups and robustness in practical automotive acoustic simulations.

Contribution

It introduces GPU-accelerated iterative Krylov algorithms tailored for complex acoustic matrices, highlighting their performance and robustness improvements over traditional methods.

Findings

Speed-up up to 28x for dot product operations

Speed-up up to 9.8x for sparse matrix-vector products and solvers

Effective handling of acoustic matrices from car interior models

Abstract

This paper deals with linear algebra operations on Graphics Processing Unit (GPU) with complex number arithmetic using double precision. An analysis of their uses within iterative Krylov methods is presented to solve acoustic problems. Numerical experiments performed on a set of acoustic matrices arising from the modelisation of acoustic phenomena inside a car compartment are collected, and outline the performance, robustness and effectiveness of our algorithms, with a speed-up up to 28x for dot product, 9.8x for sparse matrix-vector product and solvers.