Re: [Starpu-devel] Concurrent CUDA kernel execution

[Keisuke Fukuda <fukuda@matsulab.is.titech.ac.jp>]

Hi Samuel,

I believe this feature is very helpful for many applications that have trade-off issues of task size and load balance.

Are you going to include it in the next release, or will it be experimental for a while?

Thanks,

Keisuke

------------------------------------------
Fukuda, Keisuke<fukuda@matsulab.is.titech.ac.jp>

Graduate Student in Dept. of Math. & Comp. Sciences
Tokyo Institute of Technology

福田圭祐 <fukuda@matsulab.is.titech.ac.jp>
東京工業大学　数理・計算科学専攻　松岡研究室

On Sat, Apr 5, 2014 at 2:28 AM, Samuel Thibault <samuel.thibault@ens-lyon.org> wrote:

Samuel Thibault, le Fri 04 Apr 2014 18:08:14 +0200, a écrit :

> people can already give it a
> try by just setting STARPU_NWORKER_PER_CUDA to 2 or more.

I forgot to mention: you also need to enable async execution, which is
documented, see the STARPU_CUDA_ASYNC flag documentation.

Samuel

_______________________________________________
Starpu-devel mailing list
Starpu-devel@lists.gforge.inria.fr
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/starpu-devel