using multiple GPU from matlab

General CULA Dense (LAPACK & BLAS) support and troubleshooting. Use this forum if you are having a general problem or have encountered a bug.

using multiple GPU from matlab

Postby mharb » Mon Jun 20, 2011 11:59 am

Is it possible to use the link interface with several GPU in parallel from MATLAB? i.e. invert several matrices at the same time, one on each GPU. If so, how can this be done?
mharb
 
Posts: 9
Joined: Wed Feb 23, 2011 11:26 am

Re: using multiple GPU from matlab

Postby kyle » Mon Jun 20, 2011 1:21 pm

This is not possible with the link interface. You'll have to write some custom code to set the proper GPU device and run the functions.
kyle
Administrator
 
Posts: 301
Joined: Fri Jun 12, 2009 7:47 pm

Re: using multiple GPU from matlab

Postby mharb » Thu Jun 30, 2011 10:10 pm

Thanks. We are currently doing this using mex files. But are there any plans to extend the link interface to multiple GPU in the future?
mharb
 
Posts: 9
Joined: Wed Feb 23, 2011 11:26 am

Re: using multiple GPU from matlab

Postby kyle » Fri Jul 01, 2011 7:56 am

Since the link interface is designed to intercept calls to a traditional CPU based LAPACK library, the method you have described is not possible since LAPACK has no notion of binding to a device.

However, it will be possible, in a future release, to use multiple GPUs to solve one large problem.
kyle
Administrator
 
Posts: 301
Joined: Fri Jun 12, 2009 7:47 pm

Re: using multiple GPU from matlab

Postby john » Fri Jul 01, 2011 8:28 am

That's one of the nice things about the link interface using host memory. In that case, Out-of-Core solvers and multi-gpu solvers are possible, whereas with the Device interface they are not possible (pretty much by definition).
john
Administrator
 
Posts: 587
Joined: Thu Jul 23, 2009 2:31 pm


Return to CULA Dense Support

Who is online

Users browsing this forum: No registered users and 0 guests

cron