WARNING: This website is obsolete! Please follow this link to get to the new Albert@Home website!
Intel GPU tasks error out |
Message boards :
Problems and Bug Reports :
Intel GPU tasks error out
Message board moderation
Author | Message |
---|---|
zombie67 [MM] Send message Joined: 10 Oct 06 Posts: 130 Credit: 30,924,459 RAC: 0 |
I have 5 mac minis with HD 4000. They are all erring out. Here is a sample: https://albert.phys.uwm.edu/result.php?resultid=1526988 Dublin, California Team: SETI.USA |
Richard Haselgrove Send message Joined: 10 Dec 05 Posts: 450 Credit: 5,409,572 RAC: 0 |
Ah, the imfamous Exit status 197 (0xc5) EXIT_TIME_LIMIT_EXCEEDED You could try the workround I suggested in the thread of that name, at message 113253. Use the name einsteinbinary_BRP4 in your app_config.xml file. No, scrub that - it's a hsgamma_FGRP3 task, same as the example: I hadn't realised we had FGRP for intel_gpu. Afterthought - not sure if the FGRP utilisation factor control on the preferences page is applied to intel_gpu apps - you'll have to experiment and find out for yourself. If the workround doesn't work, you'll have to do it the old-fashioned way - increasing <rsc_fpops_bound> in client_state.xml by a factor of 100 or so for each task. |
zombie67 [MM] Send message Joined: 10 Oct 06 Posts: 130 Credit: 30,924,459 RAC: 0 |
It worked. Thanks! Do I have to leave this setting indefinitely? If not, at what point can I remove the app_config and change the settings back to normal? Dublin, California Team: SETI.USA |
Richard Haselgrove Send message Joined: 10 Dec 05 Posts: 450 Credit: 5,409,572 RAC: 0 |
It worked. Thanks! Look at the application details page for each of your hosts - e.g. https://albert.phys.uwm.edu/host_app_versions.php?hostid=8551 for the lowest numbered. When the 'Number of tasks completed' for the Gamma-ray pulsar search #3 application reaches 11, that host will be safe to leave to its own devices. But don't remove the app_config.xml file from any of them until you're able to set the utilisation factor in preferences back to 1.0, or all hell will break loose! Note that 'completed' requires that the tasks be validated, not just returned successfully. I see that you've been paired with that dreadful batch of 40-odd intel-gpu hosts with bad OpenCL 1.1 drivers (ID numbers in the low 9xxx range) - that will delay your completions until well-managed hosts come along to sort out the inconclusive validations. |
zombie67 [MM] Send message Joined: 10 Oct 06 Posts: 130 Credit: 30,924,459 RAC: 0 |
Thanks! All are crunching away on the GPU now, except for this one: https://albert.phys.uwm.edu/show_host_detail.php?hostid=8551 It is identical to all the rest, same app_config.xml (copy/pasted the same file from dropbox, so no editing differences), same location, everything. It asks for work, but gets none. Any ideas? Here is one that did get work, for comparison: https://albert.phys.uwm.edu/show_host_detail.php?hostid=8552 Edit: Never mind. Half an hour later, it got some work. Odd. But no longer an issue. Dublin, California Team: SETI.USA |
zombie67 [MM] Send message Joined: 10 Oct 06 Posts: 130 Credit: 30,924,459 RAC: 0 |
Edit: Never mind. Half an hour later, it got some work. Odd. But no longer an issue. Maybe not a one-off quirk after all. My TITAN cannot get any work for half a day now. It asks, and the server says it gets nothing. No explanation why though. Dublin, California Team: SETI.USA |
Claggy Send message Joined: 29 Dec 06 Posts: 78 Credit: 4,040,969 RAC: 0 |
Edit: Never mind. Half an hour later, it got some work. Odd. But no longer an issue. If you look at the task list you'll see the reason why, all it's tasks were aborted via the GUI yesterday, now it's got a small allowance for a few of the app versions, it has managed to pick up 5 tasks today. Claggy |