Accéder au contenu.
Menu Sympa

starpu-devel - Re: [Starpu-devel] GPU issue with r12137

Objet : Developers list for StarPU

Archives de la liste

Re: [Starpu-devel] GPU issue with r12137


Chronologique Discussions 
  • From: Samuel Thibault <samuel.thibault@ens-lyon.org>
  • To: Xavier Lacoste <xavier.lacoste@inria.fr>
  • Cc: starpu-devel@lists.gforge.inria.fr
  • Subject: Re: [Starpu-devel] GPU issue with r12137
  • Date: Thu, 3 Apr 2014 16:38:21 +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>

Xavier Lacoste, le Thu 03 Apr 2014 14:42:16 +0200, a écrit :
> Yes I have only one GPU with 11 cores.

Could you apply the attached patch and send the resulting backtrace?
That will tell us when it tries to do a CUDA-CUDA transfer.

(I don't have the permissions on starpu.git/loadEnv and test.sh)

Samuel
Index: src/datawizard/coherency.c
===================================================================
--- src/datawizard/coherency.c (révision 12540)
+++ src/datawizard/coherency.c (copie de travail)
@@ -172,6 +172,7 @@
enum starpu_node_kind kind =
starpu_node_get_kind(handling_node);
/* GPUs not always allow direct remote access: if
CUDA4
* is enabled, we allow two CUDA devices to
communicate. */
+ STARPU_ASSERT(kind != STARPU_CUDA_RAM);
return
#if 0
/* CUDA does not seem very safe with
concurrent



Archives gérées par MHonArc 2.6.19+.

Haut de le page