Bugzilla – Bug 20537
piglit failures on ATI Mobility M6
Last modified: 2014-01-23 11:04:29 UTC
Created attachment 23653 [details]
piglit main results file
I have run the piglit OpenGL test framework on my Sharp Actius MM20 with an ATI Mobility M6 graphics chipset. I am running Mesa and libDRM from git extracted on 2009/03/07, and Linux kernel 2.6.25, Gentoo distribution.
I will attach results/all/main
All the 'Assertion `exec->ctx->Driver.NeedFlush & 0x2`' errors are fixed in current git. Could you also run piglit against mesa/radeon-rewrite branch?
The mesa/radeon-rewrite branch has failed miserably on my hardware. Any OpenGL application (well I tried two) only displays its first frame, and then freezes. Piglit only got through two tests and hung on the third, which I believe is glean/orthoPosHLines. Attempting to kill piglit with ^C causes it to start the next test, and repeated ^C finally stopped it, but some other test was still around and crashed my machine hard, totally froze, could not even ssh in.
Does it output any error messages while running piglit?
Could you get Xorg log and dmesg output just before total freeze?
No error message was output from piglit, but in the test results/main file there is this:
exception: <type \'exceptions.KeyboardInterrupt\'>
traceback: @@@ File "/import/lightcycle/home/jlm/code/piglit/framework/core.py", line 297, in doRun\n
result = self.run()\n File "/import/lightcycle/home/jlm/code/piglit/framework/gleantest.py", line 6
6, in run\n out, err = glean.communicate()\n File "/usr/lib/python2.5/subprocess.py", line 668, in
communicate\n return self._communicate(input)\n File "/usr/lib/python2.5/subprocess.py", line 1150,
in _communicate\n rlist, wlist, xlist = select.select(read_set, write_set, )\n
I believe that this is merely the result of my pressing ^C, though.
/var/log/messages shows this just prior to my forced reboot:
Mar 20 20:20:03 bit mtrr: no MTRR for f0000000,1000000 found
Mar 20 20:20:15 bit [drm] Setting GART location based on new memory map
Mar 20 20:20:15 bit [drm] writeback test succeeded in 0 usecs
Mar 20 20:20:20 bit (jlm-7640): GConf server is not in use, shutting down.
Mar 20 20:20:21 bit (jlm-7640): Exiting
The Xorg log has already been overwritten. I'll reinstall the radeon-rewrite branch and run the test again.
(In reply to comment #2)
> Any OpenGL application (well I tried two) only displays its first frame, and
> then freezes.
Can you run one of them in gdb and get a backtrace of a freeze? (You may need to do this from a remote login, or you may not be able to interact with gdb when the freeze happens)
Anyway, radeon-rewrite regressions should probably be tracked in a separate report; does the current master branch still give the same failures?
Do you still experience this issue with newer mesa ?
I don't get the freeze with Mesa git from 2009/05/17, but I still have piglit failures.
Created attachment 26016 [details]
Latest piglit main results file
Do you still experience this issue with newer drivers ?
Please check the status of your issue.