Accéder au contenu.
Menu Sympa

starpu-devel - Re: [Starpu-devel] About StarPU

Objet : Developers list for StarPU

Archives de la liste

Re: [Starpu-devel] About StarPU


Chronologique Discussions 
  • From: Jeff DiMarco <jdimarco218@gmail.com>
  • To: Samuel Thibault <samuel.thibault@ens-lyon.org>, Jeff DiMarco <jdimarco218@gmail.com>, Sylvain HENRY <sylvain.henry@inria.fr>, "starpu-devel@lists.gforge.inria.fr" <starpu-devel@lists.gforge.inria.fr>
  • Subject: Re: [Starpu-devel] About StarPU
  • Date: Tue, 23 Oct 2012 14:05:56 -0400
  • 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>

Ah I see.  I appreciate all your help!

On Tue, Oct 23, 2012 at 2:04 PM, Samuel Thibault <samuel.thibault@ens-lyon.org> wrote:
Jeff DiMarco, le Tue 23 Oct 2012 14:00:50 -0400, a écrit :
> Thanks for the help, I can reference A from within the cuda kernel.  Why am I
> able to reference it from the cpu_func as opposed to cuda_func that calls the
> kernel?  Does GET_PTR just work that way then?

GET_PTR always provide a pointer for the device where the task is
currently executed. It's one of the main feature of StarPU to be moving
the data automatically to the target device before execution of the
task. The fact that the CUDA codelet execution happens first on a host
CPU does not change that: the data is on the GPU.

Samuel




Archives gérées par MHonArc 2.6.19+.

Haut de le page