Accéder au contenu.
Menu Sympa

starpu-devel - Re: [Starpu-devel] About dag.dot

Objet : Developers list for StarPU

Archives de la liste

Re: [Starpu-devel] About dag.dot


Chronologique Discussions 
  • From: Xavier Lacoste <xl64100@gmail.com>
  • To: Samuel Thibault <samuel.thibault@ens-lyon.org>
  • Cc: starpu-devel@lists.gforge.inria.fr
  • Subject: Re: [Starpu-devel] About dag.dot
  • Date: Fri, 14 Mar 2014 10:33:10 +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>


Le 14 mars 2014 à 10:20, Samuel Thibault <samuel.thibault@ens-lyon.org> a
écrit :

> Xavier Lacoste, le Fri 14 Mar 2014 09:13:23 +0100, a écrit :
>> Yes, maybe it comes from the fact I use nested task ?
>
> You mean that you only submit some tasks from the callback of another
> task? In that case StarPU does not know about the dependency, indeed.
> It should however still notice all data dependencies (except those for
> which sequential consistency have been disabled of course).

I even submit tasks inside codelet's functions, at the end, without callback
function, is that bad ?
I submit a HETRF + TRSM only when all GEMM updates on it have been executed
and the unlocked list of GEMM when a HETRF + TRSM finishes.
I did that to reduce the number of task that StarPU has to manage at a time
(but as far as I remember it was not really significant).
Do you think it's better to submit all the graph from the main thread than
inside tasks ?

>
> Samuel






Archives gérées par MHonArc 2.6.19+.

Haut de le page