Objet : Developers list for StarPU
Archives de la liste
- From: Samuel Thibault <samuel.thibault@inria.fr>
- To: Cédric Augonnet <cedric.augonnet@gmail.com>
- Cc: vahuja4@gmail.com, starpu-devel@lists.gforge.inria.fr
- Subject: Re: [Starpu-devel] Contact Form Submission from Vishal
- Date: Tue, 15 Dec 2015 18:03:46 +0100
- List-archive: <http://lists.gforge.inria.fr/pipermail/starpu-devel/>
- List-id: "Developers list. For discussion of new features, code changes, etc." <starpu-devel.lists.gforge.inria.fr>
Hello,
>From Vishal
> I am eager to use StarPU to utilize multiple GPUs over a network. Can you
> please tell me if StarPu works using OpenCL 2.0. Are there some good
> examples
> showing how to use OpenCL in a distributed fashion?
StarPU should work fine with OpenCL 2.0. We don't have distributed
examples using OpenCL, but we have a few distributed examples, Cholesky
and LU matrix decompositions notably. Using OpenCL is then a matter of
adding the OpenCL kernel to the codelet.
Put another way: once you have a working single-node CPU+OpenCL
implementation on top of StarPU, and and working multinode CPU
implementation, the multinode CPU+OpenCL implementation is a mere
combination of the two previous.
Samuel
- Re: [Starpu-devel] Contact Form Submission from Vishal, Cédric Augonnet, 12/12/2015
- Re: [Starpu-devel] Contact Form Submission from Vishal, Samuel Thibault, 15/12/2015
Archives gérées par MHonArc 2.6.19+.