Optimal tool for large scale matrix inversion

Accelerated Computing CUDA CUDA Programming and Performance
1

Hello,

I am currently working on a program that involves performing multiple inversion tasks on large matrices ranging in size from 1000x1000 to 10000x10000. (Inversion tasks are necessary, and alternative methods might not be applicable.)

I have been exploring various methods to achieve fast matrix inversion and have come across several tools, including BLAS, LAPACK, MKL, CUBLAS, and CUSOLVER, among others.

I believe that utilizing CUSOLVER is the optimal approach due to the ability of the GPU to leverage numerous CUDA cores efficiently.

However, I am not an expert in this field and would like to inquire if there are any other methods available for performing a large number of matrix inversions.

I understand that my question might be a bit ambiguous, but I would appreciate any insights or guidance you can provide on this topic.

Thank you for your assistance.

2

The usual suggestion is to not use matrix inversion if you can avoid it. There are other solution methods for typical uses such as solution of linear systems.

Otherwise, CUSOLVER is the usual library for such things, when considering GPU based solution methods.