" - client: on startup, if a coproc needed by a job is missing, set a "coproc_missing" flag rather than aborting the job. If use removes a GPU board while there's a large queue of GPU jobs, they'll stay queued (until their deadline passes).
Note: this doesn't fix the situation where user connects via Remote Desktop while GPU jobs are running or queued. We should check for Remote Desktop every minute or so, and stop GPU jobs. " Please, don't introduce any special checks for Remote Desktop itself! It's only _one_ of possible reasons of missing GPU. If you really want to improve BOINC's awareness of coprocessor presence - check this presence itself, not factor that could hide GPU - there are many such factors. BTW, checking for GPU at moment of schedule decision (if not too costly by itself) could have additional benefit (additional to correctly handling RDP and fast user switching). There are some situations when GPU just "disappears". Maybe overheating maybe some another reason, but I see such behavior on one of my nVidia GPUs, 9400GT. This cause CUDA MB to fallback to CPU processing cause it ordered to run on GPU but its own check shows no such GPU. This is worse than it just suspend its work (CPU part of current CUDA MB much worse than standalone opt CPU MB, hence CPU fallback not only doesn't use GPU, it uses CPU in inefficient way). If BOINC could check GPU availability before actual task scheduling it could prevent this situation also. _______________________________________________ boinc_dev mailing list boinc_dev@ssl.berkeley.edu http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev To unsubscribe, visit the above URL and (near bottom of page) enter your email address.
