Accéder au contenu.
Menu Sympa

starpu-devel - Re: [Starpu-devel] Difference between starpu_mpi_cache_flush() and starpu_mpi_cache_flush_all() ?

Veuillez patienter...

starpu-devel@inria.fr

Objet : Developers list for StarPU

Archives de la liste

Re: [Starpu-devel] Difference between starpu_mpi_cache_flush() and starpu_mpi_cache_flush_all() ?


Chronologique Discussions 
  • From: Samuel Thibault <samuel.thibault@ens-lyon.org>
  • To: Benoît Lizé <benoit.lize@gmail.com>
  • Cc: starpu-devel@lists.gforge.inria.fr
  • Subject: Re: [Starpu-devel] Difference between starpu_mpi_cache_flush() and starpu_mpi_cache_flush_all() ?
  • Date: Fri, 11 Jul 2014 17:44:49 +0200
  • 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>

Benoît Lizé, le Fri 11 Jul 2014 15:40:52 +0200, a écrit :
> From the code, it seems to me that there is no difference between:
>
> for (int i = 0; i < nhandles; i++) {
>   starpu_mpi_cache_flush(MPI_COMM_WORLD, starpu_handles[i]);
> }
>
> and
> starpu_mpi_cache_flush_all_data(MPI_COMM_WORLD);
> provided that all the registered handles are in the "handles" array.
>
> Is it true ?

Yes.

> 2.
> From what I understand, and assuming that starpu_mpi_insert_task() doesn't
> block (which AFAIK is true right now), there should not be a difference
> between
> flushing the cache as soon as I know that a given handle won't be touched
> anymore, and using starpu_mpi_cache_flush_all() at the end.
>
> Is it true ?

Yes and no: it also avoids letting the MPI cache hash table getting
filled unnecessarily, which can bring a big difference if there are
really a lot of data.

Samuel




Archives gérées par MHonArc 2.6.19+.

Haut de le page