Re: [info-performer] pfSync and NVIDIA driver

Date view Thread view Subject view Author view

From: Allan Schaffer (allan++at++sgi.com)
Date: 02/18/2004 12:33:39


Wunibald Karl wrote:
> - it's related to NVIDIA bug 83999
> - we are considering the problem
> I even have not been notified, that it's filed as a bug and as Brian's
> post shows they are even not aware of the problem! It's as quiet as here
> on the mailing list.
> I've already contacted SGI's european support manager and the guys in
> germany seem to support me also, but I have doubts, that the performer
> team still exists.
> ALLAN IF YOU ARE STILL IN THE COMPANY, IT'S TIME TO MAKE A STATEMENT!!!

Hello Wuni, yes I am definitely here along with the team, please don't let me
falling behind on info-performer emails worry you too much about our
existence. I have about 150 to catch up on -- and thanks to the person who
said I'd better take this one first.. :-)

We've have some good (at least improving) news to report on this vsync
problem with Multi-CPU apps on Linux. The highlight being, a week or two ago
we heard from an engineer at NVIDIA who had been assigned to it. He has made
a change and put it into the latest released driver (53.36) to fix the
problem reported in the problem we reported NVIDIA#83999. We have tested the
fix in the new driver and confirmed that the vsync extension is now
functioning they way it should. However the NVIDIA implementation behaves
slightly differently than the implementation on SGI IRIX, so we will still
need to make changes in Performer to handle these differences. The engineers
are working on this currently. It is fairly complex. There are also some
related issues that we're working on with the nvidia folks.

My hope would be to get this done for the Performer 3.1.1 release. For
reference, this is SCR 878575 in the SGI bugsystem.

Multi-CPU linux systems with ATI cards will continue to suffer from the
problem, although not as bad as with NVIDIA. ATI does not yet provide an
implementation of SGI_video_sync functionality so we fall onto the emulation
path. Single-process apps don't have the problem.

Allan

-- 
Allan Schaffer                                            allan++at++sgi.com
Engr. Manager, Core Rendering                            1-650-933-2160
Silicon Graphics                                     http://www.sgi.com


Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Wed Feb 18 2004 - 12:33:47 PST