Accéder au contenu.
Menu Sympa

starpu-devel - Re: [Starpu-devel] Performance profiling of tile low rank cholesky (HiCMA) with StarPU on distributed memory systems

Objet : Developers list for StarPU

Archives de la liste

Re: [Starpu-devel] Performance profiling of tile low rank cholesky (HiCMA) with StarPU on distributed memory systems


Chronologique Discussions 
  • From: Aniello Esposito <esposito@cray.com>
  • To: Samuel Thibault <samuel.thibault@inria.fr>, Hatem Ltaief <hatem.ltaief@kaust.edu.sa>
  • Cc: Kadir Akbudak <kadir.akbudak@kaust.edu.sa>, "starpu-devel@lists.gforge.inria.fr" <starpu-devel@lists.gforge.inria.fr>, Aleksandr Mikhalev <aleksandr.mikhalev@kaust.edu.sa>, "Sameh M. Abdulah" <sameh.abdulah@kaust.edu.sa>
  • Subject: Re: [Starpu-devel] Performance profiling of tile low rank cholesky (HiCMA) with StarPU on distributed memory systems
  • Date: Wed, 2 May 2018 17:08:48 +0000
  • Accept-language: de-DE, en-US
  • Authentication-results: mail3-smtp-sop.national.inria.fr; spf=None smtp.pra=esposito@cray.com; spf=Pass smtp.mailfrom=prvs=653faafb0=esposito@cray.com; spf=None smtp.helo=postmaster@esa2.cray.iphmx.com
  • Ironport-phdr: 9a23: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
  • 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>
  • Spamdiagnosticmetadata: NSPM
  • Spamdiagnosticoutput: 1:99

Dear all,

Thank you for this intense discussion.
I can confirm the usage of prio and no fxt for the results presented in the
initial mail.
We could repeat the analysis with starpu-1.3 and the ws scheduler to see if
the picture
improves. As Hatem already emphasized, the main point here is the behavior of
threads
in the presence of not enough work. We would have expected no real difference
if even
too many threads are used.

Yours,
Aniello

On 4/26/18, 8:16 PM, "Samuel Thibault" <samuel.thibault@inria.fr> wrote:

Hatem Ltaief, le jeu. 26 avril 2018 16:22:55 +0000, a ecrit:
> We could give it another try with other scheduler policies but it
seemed that eager was giving the best performance.

eager would be very performant with few workers, yes. But with many
workers lws would probably work better.

Samuel





Archives gérées par MHonArc 2.6.19+.

Haut de le page