From: Tom Flynn (flynnt++at++gmail.com)
Date: 07/22/2004 17:17:00
Unfortunately, I'm sitting in front of a Windows machine and the pf3.1
i have installed on it doesn't have the compositor.C sample or
pfCompositor man page or any perfly code that contains references to
the compositor stuff, or even the pfCompositor man page. This stuff
is probably only shipped on the IRIX version of performer. I also
haven't looked at this kind of thing in a while, so I'll be basing
this post on some fuzzy memory recollection....
First of all, I think you need to setup a hyperpipe for each
compositor you're going to use. For example, if you wish to use 6
pipes, with 2 pipes per compositor, you need to setup 3 hyperpipes
with two pipes each.
i think for each hyperpipe, you'll need a new pfCompositor (i think).
for each pipe in a hyperpipe, you add that pipe to the corresponding
pfCompositor.
i think the answer to the COMP_SOFTWARE question is that it was there
for a feature of the compositors that never came into fruition.
You shouldn't be seeing a crash when enabling or disabling load
balancing. i recommend you create a small sample that recreates the
problem so the pf team and quickly debug and fix the problem for their
next release.
the other questions you have i cannot immediately answer due to lack
of man page or source code that you're referencing.
hope this provides some help,
tom
On Fri, 16 Jul 2004 08:48:21 +1000 (AEST), Bernie Kirby
<bernie++at++ecr.mu.oz.au> wrote:
> Compositor confusion.
>
> Does anyone have a performer3.1.1 app that uses
> multiple compositors?
>
> We have an onyx4 with 3 compositors, but as yet, we've not been
> able to effectively get a single performer app to 'drive' the
> multiple compositors.
>
> One setup is this:
> 2 cards each driving the 3 compositors with a 2x1 vert subdivision.
> that is, 6 cards in total - 6 screens.
> Using the example compositor.C code, and modifying it to use just the
> first two pipes with the 2x1 vert subdivision, it works fine.
> Extending this we tried creating 3 pfCompositors and added the pipes 0,1
> to the first one, 2,3 to the second and 4,5 to the third.... added channels as
> needed, called setScreen as appropriate. Doesn't work. Only the first compositor
> composits, and the others seem to just use the first pipe specified (maybe the
> others - can't see them) and aren't being composited.
>
> This leads to questions:
>
> How do you tell a pfCompositor which actual compositor to use?
> What are the 'rules' for addressing each compositor?
> If I wanted to (for example) just run a graphics program on
> (say) the third compositor using just it's two cards, how can I
> stop performer opening a window on all the 'other' essentially unused pipes?
>
> Also, returning to the basic compositor.C example, we tried to just use
> 2 pipes again, but called setScreen to point to some of the other screens
> on another compositor. Again, didn't work.
>
> Other questions:
> What is a software compositor? (see example perfly C code
> pfCompositorMode(compositor, PFCOMP_SOFTWARE, 1); )
>
> What are the PFCOMP_TYPEs PFCOMP_NxM_VERT and PFCOMP_NxM_HORIZ ? again,
> used in perfly code, but not in man page.
>
> What does pfCompositorAddCompositor do? It's mentioned in the
> man page, but there's no description.
>
> Other problems:
> Calling:
> pfCompositorMode(comp, PFLOAD_BALANCE, PF_OFF /* or PF_ON */);
>
> Gives this message:
> PF Warning: pfCompositor::setMode(), compState is NULL
> and subsequently, a core dump.
>
> --
> Two wrongs don't make a right ... but two wrights make an aeroplane.
>
> -----------------------------------------------------------------------
> List Archives, Info, FAQ: http://www.sgi.com/software/performer/
> Open Development Project: http://oss.sgi.com/projects/performer/
> Submissions: info-performer++at++sgi.com
> Admin. requests: info-performer-request++at++sgi.com
> -----------------------------------------------------------------------
> SIGGRAPH 2004 Friends of Performer BOF:
> Tuesday, August 10, 6pm - 8pm Wilshire Grand Hotel, Sierra Ballroom
> -----------------------------------------------------------------------
>
>
This archive was generated by hypermail 2b29 : Thu Jul 22 2004 - 17:24:36 PDT