Accéder au contenu.
Menu Sympa

starpu-devel - Re: [Starpu-devel] Using STARPU_COMMUTE with starpu_mpi_task_build()

Objet : Developers list for StarPU

Archives de la liste

Re: [Starpu-devel] Using STARPU_COMMUTE with starpu_mpi_task_build()


Chronologique Discussions 
  • From: Xavier Lacoste <xl64100@gmail.com>
  • To: Xavier Lacoste <xl64100@gmail.com>
  • Cc: Samuel Thibault <samuel.thibault@ens-lyon.org>, starpu-devel@lists.gforge.inria.fr
  • Subject: Re: [Starpu-devel] Using STARPU_COMMUTE with starpu_mpi_task_build()
  • Date: Tue, 11 Mar 2014 12:30:54 +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>

Indeed, the trunk seems to be slower with r13383 than with r12012 (820s vs 100s on 12 cores). 
I had to set STARPU_HISTORY_MAX_ERROR=2000 to have no error displayed (200 was not enough).
eager seems to be slower than dmda also.... Still waiting for completion...

XL.

Le 11 mars 2014 à 11:20, Xavier Lacoste <xl64100@gmail.com> a écrit :


Le 11 mars 2014 à 11:14, Samuel Thibault <samuel.thibault@ens-lyon.org> a écrit :

Xavier Lacoste, le Tue 11 Mar 2014 11:04:23 +0100, a écrit :
Le 11 mars 2014 à 10:57, Samuel Thibault <samuel.thibault@ens-lyon.org> a écrit :
Xavier Lacoste, le Tue 11 Mar 2014 10:51:02 +0100, a écrit :
I updated my trunk to have the STARPU_COMMUTE ignoring in MPI but now I have a lot of this errors :
[1,0]<stderr>:[starpu][_starpu_update_perfmodel_history] Too big deviation for model DGEMM on cpu0_parallel1_impl0: 1212.959000, vs average 989.918214 over 42 samples (+22.531234%), flushing the performance model. Use the STARPU_HISTORY_MAX_ERROR environement variable to control the threshold (currently 10%)
(Which slows down a lot computation)

Yes, this is a new thing in the trunk, the +-10% error rate may be too
small, perhaps you'd rather set STARPU_HISTORY_MAX_ERROR to 25 or even
more.
Ok, I even have 100% error on some cases.... So should I go with STARPU_HISTORY_MAX_ERROR = 100 ?

Yes.

Are you also deciding which GPU the GEMM should go to?  StarPU still
needs to decide between the cost of moving the data and the computation
cost.  Do you have a good cost function ?

Yes I'm deciding all about GPU mapping,

How are you deciding it?  By prefetching the data or by using
use_a_specific_worker?  If the latter, you could perhaps as well just
use the eager scheduler, which doesn't need a performance model.

Ok I'll do that (I use STARPU_EXECUTE_ON_WORKER, workerid when building a GPU task (and let StarPU decide when it's a CPU task) and can activate prefetch or not).

Samuel

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


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




Archives gérées par MHonArc 2.6.19+.

Haut de le page