Sign in to follow this  
nippbit

subdivision of geometry for use in lighting

Recommended Posts

Hello all, Is it truely faster to use light maps or wouldn't it be faster to implement a precomputed geometric subdivision similiar to the way lighting was done in the OLD doom and quake series, only automate it using basic shadow volumes coliding with the faces, then further subdivide the faces and assign an ambient diffuse value to these vertices? Furthermore, you could assign a theta component to subdivide for a falloff period in the "shadow geometry". Combined with shadow maps, I could see this as a real powerful technique because of the raw processing power of the GPU on most systems. You can then reuse the shadow volumes for real time shadow casting using the Z-Pass or Fail routine(Carmack's reverse), which, in all honesty, is something I would like to avoid for now simply because I'm a lowly little developer running a poor graphics card. The major drawback is that you can not support animated meshs(levels) with this, unless you use Z-Fail/Pass algorithm. Has anyone seen any papers on any material like this? -Ken Noland

Share this post


Link to post
Share on other sites
Well, It might work..
But..
If you're fill limited you might want to do what you suggested, but if you're geometry limited you're probably much better off with lightmaps..
It's all about balance..
Some solution might be good in one situation, but horrible in another..
In the end it's better to only bother looking into these kind of things when you need it..
See it as an optimization.. and you know what they say about early optimisations..

Share this post


Link to post
Share on other sites
Well,

This is a special type of optimization that I need to determine early on in the project, I've already started working on it and it should be done within the next week. I'll publish my results when I'm through, but I'll need some help.

I have a severe hardware limitation, simply because my primary development machine didn't survive the trip from New York to Seattle. I could generate the test application, but I don't have any decent hardware to run it on... any help, anyone???

Share this post


Link to post
Share on other sites
Chances are you have atleast a Voodoo2 card (I hope) which means you can do atleast 2 textures in one pass. Given that it would almost certainly be faster, even on a low end card, to use lightmaps rather then subdivide your geometry. If, however, you are working on a software engine then the reverse might be true.
Subdividing geometry MAY give you better looking results depending on your shadow casters and lightmap texture budget.

Share this post


Link to post
Share on other sites
ahh, but sending detailed textures over the bus can cause extreme performance issues. I tested it out on a VooDoo 5, and it was actually faster adding over 500 polys then it was to do dual textures. I've got a new machine on order... I will have a better machine soon to do my testing on.

Thanks :-)

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Sign in to follow this