Re: [Condor-users] GPU and condor?

[Ioan Raicu <iraicu@xxxxxxxxxxxxxxx>]

I don't have much experience with GPUs either, but as far as I know, a 
GPU cannot be partitioned in order to run different kernels (the 
execution job) at the same time. Its true that a kernel can be specified 
to run on only part of the GPU (e.g. 32 of the 800 cores), but I believe 
the remainder of the cores will be idle, and cannot simply be assigned 
to a different kernel. Getting to the point where a GPU could be treated 
like a collection of separate cores would be great, but I don't think 
its as straight forward as one might think, at least with today's GPU 
hardware. If anyone has more experience with GPUs, and how this might be 
achievable in practice, I'd be curious to learn more.

Cheers,
Ioan

Ian Stokes-Rees wrote:
> On 1/7/10 10:38 AM, Miron Livny wrote:
> > To all GPUers out there,
> >
> > We would be very interested in hearing from you what Condor can do to 
> > help you in managing GPU clusters. So far we did not find much we can 
> > offer in this space. Any guidance you can provide will be most welcomed.
>
> I'm not very involved in GPU computing, but one thing we would imagine 
> wanting to do would be segmentation of a GPU into multiple job slots.  
> One could imagine a GPU-centric pool configuration where a single 
> system with, say, 8 CPU cores, manages a connect GPU system with, say, 
> 32 sets of 32 stream processors each, or 4 sets per CPU.  The one 
> physical system which usually would have 8 condor job slots would 
> instead have 32 job slots.
>
> Ian
>
> _______________________________________________
> Condor-users mailing list
> To unsubscribe, send a message to condor-users-request@xxxxxxxxxxx with a
> subject: Unsubscribe
> You can also unsubscribe by visiting
> https://lists.cs.wisc.edu/mailman/listinfo/condor-users
>
> The archives can be found at:
> https://lists.cs.wisc.edu/archive/condor-users/
>   

--
=================================================================
Ioan Raicu, Ph.D.
NSF/CRA Computing Innovation Fellow
=================================================================
Center for Ultra-scale Computing and Information Security (CUCIS)
Department of Electrical Engineering and Computer Science
Northwestern University
2145 Sheridan Rd, Tech M384 
Evanston, IL 60208-3118
=================================================================
Cel:   1-847-722-0876
Tel:   1-847-491-8163
Email: iraicu@xxxxxxxxxxxxxxxxxxxxx
Web:   http://www.eecs.northwestern.edu/~iraicu/
      https://wiki.cucis.eecs.northwestern.edu/
=================================================================
=================================================================