Welcome to MilkyWay@home

Linux x86_64 1.02 (opencl_amd_ati) - MultiGPU - fglrx asic hang happened (fglrx driver crash)


Advanced search

Message boards : Number crunching : Linux x86_64 1.02 (opencl_amd_ati) - MultiGPU - fglrx asic hang happened (fglrx driver crash)
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile[AF>Libristes>Jip] Elgrande71
Avatar

Send message
Joined: 21 Mar 08
Posts: 20
Credit: 183,766,905
RAC: 0
100 million credit badge13 year member badge
Message 54404 - Posted: 14 May 2012, 10:29:23 UTC

I try to explain what is the problem with my two GPUs configuration ( http://milkyway.cs.rpi.edu/milkyway/show_host_detail.php?hostid=358617 ).
After several hours of calculation, fglrx driver crashed with "fglrx asic hang happened" error message in kernel log file.
This crash also stop my BOINC client (defunct processes related to CPU project).
For more information, you can look at this bug report .
But with the former Linux x86_64 0.82 ati CAL version (with app_info.xml file), I have no problem.
At the beginning, I thought that it would be a driver bug but now I don't know.
Is it a problem with the OpenCL application ?
I have the same driver crash with the Albert@Home project.
ID: 54404 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Matt Arsenault
Volunteer moderator
Project developer
Project tester
Project scientist

Send message
Joined: 8 May 10
Posts: 576
Credit: 15,979,383
RAC: 0
10 million credit badge11 year member badge
Message 54409 - Posted: 14 May 2012, 14:53:12 UTC - in response to Message 54404.  

There are a lot of crashes with Catalyst and the OpenCL driver so probably. There is a horrible defect where the Linux Catalyst driver can't reset the GPU after some error, so even if there is an application error the driver problem makes it much worse since you have to reboot after.
ID: 54409 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile[AF>Libristes>Jip] Elgrande71
Avatar

Send message
Joined: 21 Mar 08
Posts: 20
Credit: 183,766,905
RAC: 0
100 million credit badge13 year member badge
Message 54411 - Posted: 14 May 2012, 15:21:45 UTC

Hopefully, on my computer, I have ssh access so I can reboot easily even if my xorg server is blocking by the driver crash.
The behaviour of the catalyst driver with the linux opencl_amd_ati app is odd because with one GPU, I have no driver crash.
It only happens with two GPUs on the same computer.
Perhaps, the OpenCL multigpu support is only at alpha or beta state not final.
ID: 54411 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote

Message boards : Number crunching : Linux x86_64 1.02 (opencl_amd_ati) - MultiGPU - fglrx asic hang happened (fglrx driver crash)

©2021 Astroinformatics Group