Objet : Developers list for StarPU
Archives de la liste
- From: Miguel Palhas <mpalhas@gmail.com>
- To: starpu-devel@lists.gforge.inria.fr
- Subject: Re: [Starpu-devel] pheft problem?
- Date: Tue, 13 Aug 2013 02:22:19 +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>
I forgot to add some information...
This only happens when i set the codelet for that task to use only STARPU_CUDA (while it depends on data that is ALWAYS executed on a CPU task)
Also, if i switch the task implementation, and make it just an empty function, the problem still persists
This two things lead me to believe this is an issue related with memory transfers within pheft. It seems to be waiting for data to be available. But i guess the fact that peager works fine proves that it is not a dependency problem on my code
On Tue, Aug 13, 2013 at 2:17 AM, Miguel Palhas <mpalhas@gmail.com> wrote:
When using pheft on my algorithm, i have a problem with a specific task, which hangs the scheduler as soon as it's sampling information has 10 executions
As soon as i delete the information for that specific task (~/.starpu/sampling/codelets/task_name), i can once again run the task with no problem, until around 10 executions are stored againAfter a very long bug hunting, i narrowed down the problem to here. If i switch from pheft to peager, everything runs smoothly.Is this a bug with the pheft scheduler?--
CumprimentosMiguel Palhas
Cumprimentos
Miguel Palhas
- [Starpu-devel] pheft problem?, Miguel Palhas, 13/08/2013
- Re: [Starpu-devel] pheft problem?, Miguel Palhas, 13/08/2013
Archives gérées par MHonArc 2.6.19+.