SuiteSparse and Magma

Open discussion for MAGMA library (Matrix Algebra on GPU and Multicore Architectures)
Post Reply
Posts: 1
Joined: Tue Jun 02, 2020 5:51 pm

SuiteSparse and Magma

Post by SomeUsername1 » Tue Jun 02, 2020 6:13 pm

Dear MAGMA devs,

MAGMA offers support for both sparse and dense LA in heterogeneous computing environments.
As far as I understood from reading your documentation, SuiteSparse seems to provide a set of capabilities that would fit into the Magma framework perfectly.
For instance the widely used multifront QR solver, CHOLMOD, algorithms for SpMV and SpMM on both CPU and GPU using (as your lib) CUDA and MKL as dependencies.
AFAIK Matlab uses both of your libraries for those specializations: SFMULT, SSMULT, \ come from SuiteSparse, whereas GPU support for dense LA comes from MAGMA.

Would or wouldn't it make sense to integrate SuiteSparse into your framework as a "drop-in" for the sparse parts that SuiteSparse does so well?

In my humble opinion it would be awesome to be able to use the two libraries along with a smooth integration, thus providing a rich set of LA functionality for both sparse and dense cases.

What are your thoughts on this?
Is this feasible, do you welcome contributions in this direction and where is the best place to start?

Kind regards,

Posts: 918
Joined: Fri Jan 06, 2012 2:13 pm

Re: SuiteSparse and Magma

Post by mgates3 » Wed Jun 03, 2020 11:25 am

Matlab uses both dense and sparse components from MAGMA. It also uses components from SuiteSparse.
We do some collaboration with Tim Davis, the primary author of SuiteSparse. Applications can use both, as Matlab does, but there are no current plans to integrate MAGMA and SuiteSparse together.


Post Reply