Concurrent number cruncher - A GPU implementation of a general sparse linear solver

in: International Journal of Parallel, Emergent and Distributed Systems, 24:3 (205-223)

Abstract

A wide class of numerical methods needs to solve a linear system, where the matrix pattern of non-zero coefficients can be arbitrary. These problems can greatly benefit from highly multithreaded computational power and large memory bandwidth available on GPUs, especially since dedicated general purpose APIs such as CTM (AMD-ATI) and CUDA (NVIDIA) have appeared. CUDA even provides a BLAS implementation, but only for dense matrices (CuBLAS). Other existing linear solvers for the GPU are also limited by their internal matrix representation. This paper describes how to combine recent GPU programming techniques and new GPU dedicated APIs with high performance computing strategies (namely block compressed row storage, register blocking and vectorization), to implement a sparse general-purpose linear solver. Our implementation of the Jacobi-preconditioned Conjugate Gradient algorithm outperforms by up to a factor of 6.0x leading-edge CPU counterparts, making it attractive for applications which are content with single precision.

Download / Links

BibTeX Reference

@article{buatois:inria-00331906,
 abstract = {A wide class of numerical methods needs to solve a linear system, where the matrix pattern of non-zero coefficients can be arbitrary. These problems can greatly benefit from highly multithreaded computational power and large memory bandwidth available on GPUs, especially since dedicated general purpose APIs such as CTM (AMD-ATI) and CUDA (NVIDIA) have appeared. CUDA even provides a BLAS implementation, but only for dense matrices (CuBLAS). Other existing linear solvers for the GPU are also limited by their internal matrix representation. This paper describes how to combine recent GPU programming techniques and new GPU dedicated APIs with high performance computing strategies (namely block compressed row storage, register blocking and vectorization), to implement a sparse general-purpose linear solver. Our implementation of the Jacobi-preconditioned Conjugate Gradient algorithm outperforms by up to a factor of 6.0x leading-edge CPU counterparts, making it attractive for applications which are content with single precision.},
 author = {Buatois, Luc and Caumon, Guillaume and L{\'e}vy, Bruno},
 doi = {10.1080/17445760802337010},
 hal_id = {inria-00331906},
 hal_version = {v1},
 journal = {{International Journal of Parallel, Emergent and Distributed Systems}},
 note = {to appear},
 number = {3},
 pages = {205-223},
 pdf = {https://inria.hal.science/inria-00331906/file/Buatois_et_al_CNC.pdf},
 publisher = {{Taylor \& Francis}},
 title = {{Concurrent number cruncher - A GPU implementation of a general sparse linear solver}},
 url = {https://inria.hal.science/inria-00331906},
 volume = {24},
 year = {2008}
}