Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!uwm.edu!bionet!agate!eos!amelia!eagle!news
From: fsset@bach.lerc.nasa.gov (Scott E. Townsend)
Newsgroups: comp.sys.sgi
Subject: Graphics FIFO error, How did I cause it?
Message-ID: <1990Jul12.141406.24366@eagle.lerc.nasa.gov>
Date: 12 Jul 90 14:14:06 GMT
Reply-To: fsset@bach.lerc.nasa.gov (Scott E. Townsend)
Distribution: usa
Organization: NASA/Lewis Research Center, Cleveland
Lines: 33

As the subject line states, I'm getting graphics FIFO errors logged in
/usr/adm/SYSLOG:

Jul 12 07:48:03 bach grcond[1113]: CIO: WARNING: Graphics error: GE PC = 11940
Jul 12 07:48:03 bach grcond[1113]: CIO: 
Jul 12 07:48:03 bach grcond[1113]: CIO: gr1_fifo_intr: TIMEOUT (FIFO still > hal
f full)!

The situation is a heavily loaded system running the usual 4Sight windows
(just wsh's) and a simple surface plotter displaying near real-time
computation data.  The computation is limited to the Iris's graphics speed,
since without the surface plotter running the computation runs a lot quicker.
(The computation is performed on an attached parallel processor, the only
computations on the Iris are for the graphics only)

The system is a Personal Iris, running IRIX 3.2.  The surface program is
using tmeshes on an RGB mode display, nothing amazing.

I get these not on every run, but so frequently that it's getting real
annoying. (The NeWS server goes down and I have to log in again)

Any ideas on what's happening?  Or even better, any ideas on what I can
change to avoid this?  Any suggestions would be appreciated.

(For instance, is there a way to know that I'm about to "overdrive" the
graphics system, and pause for a while?)

--
------------------------------------------------------------------------
Scott Townsend               |   Phone: 216-433-8101
NASA Lewis Research Center   |   Mail Stop: 5-11
Cleveland, Ohio  44135       |   Email: fsset@bach.lerc.nasa.gov
------------------------------------------------------------------------