Re: [Plib-users] Performance Benchmarks
Brought to you by:
sjbaker
|
From: Steve B. <sjb...@ai...> - 2000-09-06 20:17:06
|
> Aaron Drew wrote:
>
> I'm just wondering if any performance benchmarks have been done with plib?
No - when I first wrote it, the overheads in PLIB were *negligable* compared to the
time consumed in OpenGL doing vertex transforms - so there wasn't much point in leaning
on the optimisation.
Now we have hardware T&L, it would be interesting to see where the cycles are going.
Informal testing I've done at work suggest that even with a GeForce-2, providing you
have an 'equivelently modern' CPU (say a 450MHz or so), PLIB is still not the bottleneck.
> From the website its hard to see if plib comes with any standard visibility algorithm support
> like portals, bsp trees or octrees.
No portals - although it would be moderately easy to add a portal mechanism...we discussed that
a couple of days ago.
No BSP trees - the thing that PLIB implements is better (unless you are trying to
render without a Z buffer - in which case I don't think you should be using PLIB
at all).
Octrees are really just a way of using what PLIB already does.
> If I were to implement an octree using plib, would it typically be a single
> ssgNode object or rather a separate node for each octree volume with some
> sort of a custom intelligent (much more complex than the class I think is
> called ssgSelector) branch class that would look after these issues?
You could either have a single ssgBranch node - with (up to) eight other
ssgNode's beneath it - and allow SSG to do FOV culling...or you could derive
a class from ssgBranch that would do fancier culling given that it has more
information about the structure of the model than ssgBranch does.
--
Steve Baker HomeEmail: <sjb...@ai...>
WorkEmail: <sj...@li...>
HomePage : http://web2.airmail.net/sjbaker1
Projects : http://plib.sourceforge.net
http://tuxaqfh.sourceforge.net
http://tuxkart.sourceforge.net
http://prettypoly.sourceforge.net
|