cuSOLVER LU Decomposition with multiple RHS

Accelerated Computing GPU-Accelerated Libraries
1

Good afternoon,

I am new to using cusolver with dense matrices and would like to know if there is information on how to use this library to implement a particular version of LU Decomposition. I would like to be able to pass a matrix and vector to solve using LU Decomposition whereby the vector is actually made up of multiple columns rather than a single column. Is this possible and if so can anyone provide a link/example code to this?

Thank you for any assistance.

2

solution of a dense linear system using LU factorization is supported using the routines cusolverDnXgetrf for factorization followed by cusolverDnXgetrs for the solve. Multiple RHS are supported. A sample code is here. Yes, I’m aware that the sample code doesn’t demonstrate multiple RHS. That should be a fairly trivial adaptation.

1 Like
3

Thanks @Robert_Crovella that is just what I was looking for.

Also, am I correct in the idea that cuSOLVER for dense matrix operations employs column-major format?

Thanks again.

4

From here:

2.3.3. Matrix (Dense) Format
The dense matrices are assumed to be stored in column-major order in memory.

5

Cool, thanks. Just wanted to be sure.

6

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.