From: Michael Ginis (mginis++at++xig.com)
Date: 05/02/2002 08:35:43
Douglas,
I think there are possibly two reasons for the lock ups you are seeing
with the Wildcat and Accelerated-X.
The first reason could stem from a problem that several Accelerated-X
users have reported to us about perfly crashing at initialization. If
performer crashes and during the crash it overwrites a command buffer
there is the potential for a lock up. We were able to trace this to a
problem in Performer that (rarely) occurs on RedHat systems, but more
frequently occurs on other Linux distributions. The temporary fix for
this is to set the 'PFVIDEORATE' environment variable before starting
perfly, e.g:
# export PFVIDEORATE=85
# perfly ...
We have been in contact with Tom Flynn about this problem.
The second reason could be a result of perfly using multiple threads
and using this on SMP boxes. Accelerated-X is thread safe, and I
believe that perfly has worked with a Wildcat on an SMP machine here
in our labs, but we'll try that again. In the meantime, instead of
disabling SMP entirely, you are probably better off running Performer
in single thread mode. I'm not sure how to force that, does anyone
know?
-- Mike Ginis mginis++at++xig.com http://www.xig.com ------------------------------------------------------------------------From: "Douglas B Maxwell" <dmaxwell++at++ait.nrl.navy.mil> Date: Wed, 1 May 2002 09:44:17 -0400 Subject: RE: [info-performer] Perfly lockup on Dual PIII 2.2 GHz Processor
We have a similar situation using a Dell dual P4 Xeon 2.2Ghz. It came witha stock Quadro II card. Our first attempt to solve the problem was to remove the Quadro card and install a GeForce 3 card (tried and true). Machine still locked at "OpenGL Performer" screen.
Next we disabled the second processor in the computer's bios, loaded the single processor Linux kernel and NVidia drivers. It seems to run fine.
We installed a Wildcat 5110 graphics card with XIGraphics drivers and re-instated the second processor. Machine still locked at "OpenGL Performer" screen.
Currently, we are running on a single processor and the machine/Performer work fine.
This is obviously NOT an optimal solution, but it allows us continue working until a more permanent fix is found.
-douglas
___________________________________________ Douglas B Maxwell, MSME US Naval Research Laboratory, Code 5580 4555 Overlook Ave. SW Washington, DC 20375-5337 http://www.ait.nrl.navy.mil/vrlab http://ait.nrl.navy.mil/people/dmaxwell/ Office: 202-767-3901
This archive was generated by hypermail 2b29 : Thu May 02 2002 - 08:36:01 PDT