From: Scott Herod (herod++at++rtset.com)
Date: 01/19/2000 13:52:03
Hello Anthony,
I've been playing with overlapping many channels on a
single pipe. My experience has been that cost of adding
a channel or two is negligible compared to the cost of
drawing the objects in a channel. I have been drawing
multiple 3D models per channel and have also done a little
testing on the difference between making one depth-buffer
clear and none. There I did see a noticable difference.
I've not tested to compare against post pfDraw() openGL
drawing. I did begin to see a noticable performance hit
on my O2 when I added 10 or more empty ( no scene ) channels
with empty draw callbacks unless I also disabled the channels
using setTravMode(). I still get bus errors when I try to
delete a channel.
I'll add a question for the Performer folk. Is there any
performance advantage to creating channel groups?
Scott
Anthony Bavuso wrote:
>
> Angus,
>
> I figured out what my problem was. Instead of calling:
> pfPipeWindow* pw = p->getPWin(0);
> I called:
> pfPipeWindow* pw = ViewState->masterChan->getPWin()
> and it worked as expected.
>
> But now that I have it working I have another question.
>
> As stated before there seem to be two methods to generate a HUD that I am
> interested in: the first is overlapping channels and the second is openGL in
> the draw callbacks.
>
> What are the performance implications of using overlapping channels compared
> to openGL? Do I loose a lot of performance because I am using a whole
> channel for just 2D drawing? Or is the performance impact negligible?
>
> I am going to use this HUD for a real-time pilot in the loop flight
> simulation. Which method would you recommend?
>
> Thanks.
>
est++at++sgi.com
This archive was generated by hypermail 2b29 : Wed Jan 19 2000 - 13:52:37 PST