[info-performer] Performer on Windows Problems

Date view Thread view Subject view Author view

From: Damien Gerhardy (damien.gerhardy++at++sydac.com.au)
Date: 03/15/2006 17:05:40


I've used performer on SGI hardware before (Onyx2's etc) and
frame rate syncing works well, however I've never seen it work
properly on the windows platform. No matter what video card I
use I just can't get it to work.

1) How can we get frame lock working on the windows platform?
   There are a whole bunch of environment variables like
   __GL_SYNC_TO_VBLANK, PF_SYNC_TO_VBLANK, PF_NV_VSYNC_WORKS,
   but none of them has any effect what-so-ever.
   Our vision runs at around 50-150 fps on windows but because
   we cant sync correctly to the refresh rate you see jerks
   whenever the frame rate changes.
   Is pfFrameRate supported at all on Windows?
   Note: If I set the refresh rate to 60Hz then performer
   will max out at 60Hz, but then slowly drop 58, 55, 52 etc
   as the scene gets more complex.
   What I really want to do is run the refresh rate at 85Hz
   and sync correcly in performer to 42.5Hz.
   Have you ever seen this work? Can I have an example of
   what to set __GL_SYNC_TO_VBLANK, PF_SYNC_TO_VBLANK,
   PF_NV_VSYNC_WORKS, PFCLOCKPERIOD, pfFrameRate, pfVideoRate
   etc... to.

2) What is the deal with PFCLOCKPERIOD?
   I've seen a few comments around and they all say different
   things.
   - If I set PFCLOCKPERIOD to the clock period of the
   computer in nanoseconds the performer stats go off the
   charts (ie: flicker between 0 and thousands changing
   every frame). I also get a slightly lower frame rate
   than when setting it to the 'magic' value below.
   - If I set PFCLOCKPERIOD to the 'magic' value of
   2300000000000000 (23 with fourteen zeros) found on some
   web pages the performer stats dont freak out but they
   are slightly out. ie: shows a draw of 16ms when I am
   getting a frame rate of 85Hz (which means draw is less
   than 11ms).
   - Maybe PFCLOCKPERIOD is actually meant to be the
   clock rate in Hz?
   - What's the deal with the 'overflow' of PFCLOCKPERIOD?
   If it is actually the clock period in nanoseconds then
   its not even close to overflowing. eg: 3.2 GHz processor
   has a clock period of 312500 ns.

Thanks.
-Damien
Caution - This e-mail and its contents contain privileged information that
is intended solely for the recipient. If you are not the intended recipient
you are hereby notified that any use, dissemination, distribution or
reproduction of this e-mail is prohibited. If you have received this e-mail
in error please notify the Postmaster at sydac++at++sydac.com.au immediately. Any
views expressed in this e-mail are of the sender and may not necessarily
reflect the views of Sydac Pty Ltd.


Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Wed Mar 15 2006 - 17:05:42 PST