Feasibility of an ArrayFire project

[Old posts from the commercial version of ArrayFire] Discussion of ArrayFire using CUDA or OpenCL.

Moderator: pavanky

Feasibility of an ArrayFire project

Postby shinagan » Mon Aug 04, 2014 9:36 am

Hi everyone,

I wrote a Matlab code that I would like to accelerate with ArrayFire. Currently it works on my CPU with the PCT enabled, but I foresee speedup with the use of a GPU. Could anyone tell me if my goal could be achievable with ArrayFire before I start working on it ? I would need to :
- Get two matrices A and B from Matlab (approximate size : A is 12000x12000, B is 12000x1)
- Convert A and B from a sparse double precision Matlab format to a sparse single precision format (Maltab do not support sparse single precision currently)
- Solve A\B with ArrayFire on the GPU (Again Matlab does not offer sparse algebra on a GPU)
- Return the result to Matlab

Is that something that looks within the possibilities offered by ArrayFire ?

Thanks in advance,

Alex
shinagan
 
Posts: 1
Joined: Mon Aug 04, 2014 9:26 am

Re: Feasibility of an ArrayFire project

Postby umar456 » Tue Aug 05, 2014 1:56 pm

Hi Alex,

We do not currently offer a solve operation on sparse matrices. It is a difficult operation which does not map well on the GPU. I would suggest looking into the cuSPARSE library. They might have something you could use.

Umar
User avatar
umar456
 
Posts: 7
Joined: Fri Nov 16, 2012 12:24 pm


Return to [archive-commercial] Programming & Development with ArrayFire

cron