cancel
Showing results for 
Search instead for 
Did you mean: 

Archives Discussions

Meteorhead
Challenger

linux driver problem

after remote setup, things got messed up

Hi!

I seem to have encountered a problem with the linux drivers (I have infact encountered many, but this is a new one). I have followed every step of the "How to run AMD APP applications remotely", and enabled autologin to the machine with my user, and everythings was fine, I could run applications remotely as myself, and as root also (although I know running as root is not neccessary in any case, but there are no otherusers just yet).

Everything was fine until I got a weird error saying (see attached code)

Since I was roughly 1000 kilometers from the computer room, I got a little ****ing angry. I got home, looked at the computer and saw that mnagy is logged into the gui, but I cannot click anywhere (mouse moves, but nothing happens on clicking anywhere on the desktop). I figured something must have gotten messed up with user priviligaes related to the XServer, as that had to be set (according to the manual).

Anyhow, I had no time to figure out the issue (which is now), so I reverted all the settings and issued a reboot. And to my surprise: the same message appeared but GPU desktop accel disappeared alltogether. aticonfig still reports the cards, but desktop don't use them.

So let me ask two things:

1) Could someone tell me how to remotely revert settings that could have gone wrong during the process of modifying the two required files, rebooting twice, running a few apps (for days), and then suddenly, nothing works...   System is SLC 5.7 (Redhat) 64-bit, Catalyst 11.12

2) Machine is configured as single-GPU setup, since multiGPU still don't work. So please let me ask, and I beg of you people: PLEASE OH PLEASE OH PLEASE make those goddamn drivers work properly! This is a joke really. HPC needs headless machines with no crap GUI installed. Not to mention that Ubuntu 12.04 is bringing about XServer alternative. It would be HIGH TIME to finally refactorize the driver. Is there at least any efforts made in this direction?

[gpu001] /home/mnagy > ./AMD/SDK/AMD-APP-SDK-v2.6-RC3-lnx64/bin/x86_64/clinfo X Error of failed request: BadRequest (invalid request code or no such operation) Major opcode of failed request: 143 () Minor opcode of failed request: 19 Serial number of failed request: 8 Current serial number in output stream: 8

0 Likes
28 Replies