When checking this report, please consider other OSs as well. There have 
been repeated, but inconclusive, reports from Windows users with multiple 
CUDA cards (such as Fred W) of similar behaviour. In Windows, it sounds as 
if this "tasks on all cards exit at the same time" is mainly triggered by 
preemption for EDF, and not on normal task completion as described here.

> Please set <cpu_sched_debug> and send me the resulting message log.
> -- David
>
> Evangelos Katikos wrote:
>> Thanks! I downloaded from trunk, compiled and tested it. GPU assignments
>> seem to be ok.
>> But now there seems to be another problem. I have three GPUs. All three
>> get their tasks ok, but every time one finishes the other two tasks get
>> preempted and a new triplet of tasks start. This repeats every time a
>> task finishes and a new one starts, all three get renewed. I have a
>> message log with <task_debug> that shows this behavior. I've also added
>> <coproc_debug> but didn't give me any messages.


_______________________________________________
boinc_dev mailing list
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Reply via email to