RE: [Algorithms] per-pixel lighting (was: Dynamic lights with sta tic shadows)

[Tom F. <to...@mu...>]

You could do the two passes of each lightsource to the destination alpha
channel, then add it to the colour channel with a render using a
DESTALPHA:ONE blend, with your source being the colour of the lightsource.

In this case, you can bin the extra pass and do:

-Clear screen
for ( each light )
{
	-Render attenuation stuff to dest alpha channel.
	-Render dot3 bumpmap, use DESTALPHA:ONE.
}
-Steps 2+ as below.


Looks like dest alpha is toast in DX8 though. Why????


Tom Forsyth - purely hypothetical Muckyfoot bloke.

This email is the product of your deranged imagination,
and does not in any way imply existence of the author.



> -----Original Message-----
> From: Charles Bloom [mailto:cb...@cb...]
> Sent: 09 November 2000 22:33
> To: gda...@li...
> Subject: RE: [Algorithms] per-pixel lighting (was: Dynamic lights with
> static shadows)
> 
> 
> 
> Hmmm... I can't see how I could do multiple lights with dot3 mapping
> and per-pixel attenuation.  To summarize, the basic way to do all
> this fancy per-pixel lighting stuff is :
> 
> 1. for each light affecting the model :
>    1a.  set up multi-texturer to compute the lighting per pixel
>         add it into the frame buffer
>         (just set it in the first time)
> 2. render the model with color = ambient lighting,
>         add it into the frame buffer
> 3. render the model with texturing, do a multiply (or multiply 2x)
> 4. render specular and/or environment map pass with additive rop
> 
> This is all very good, the problem is if in pass 1a you want to
> do both per-pixel attenuation and per-pixel dot products.  The
> problem is you need to multiply those together, and then add them
> to the frame buffer.  You can't do this though, because you'd need
> a temp work space to do the multiply (except perhaps with the 
> 3 texture
> combiners in Radeon, you might be able to get it all in one pass).
> 
> Am I right about this?  It seems there's no hope for GF2, except
> with just one light (with one light, you can stuff the attenuation in
> first, then multiply by the dot product, then add on the ambient),
> or if you are willing to use a temp storage space.  If you're just
> doing gray-scale lighting you could store the lighting 
> computation into
> the frame buffer alpha channel.
> 
> 
> At 11:33 AM 11/9/2000 -0800, you wrote:
> >There is a demo on Nvidia's site by Cass Everitt showing two 
> ways to do
> >point light attenuation.
> >One calculates a per pixel ( 1 - d^2 ), the other calculates 
> per-pixel ( e ^
> >-d^2 ).
> >
> >Both of these techniques have the nice property that they 
> work on arbitrary
> >geometry with no
> >CPU work on HW T&L cards, and are per-pixel accurate.
> >
> >See the "Attenuation Maps" portion of Game Programming Gems 
> for my writeup
> >on this idea.
> >
> >-----Original Message-----
> >From: Jason Zisk [mailto:zi...@n-...]
> >Sent: Thursday, November 09, 2000 10:36 AM
> >To: gda...@li...
> >Subject: Re: [Algorithms] Dynamic lights with static shadows
> >
> >
> >> Quake3 is done with a separate pass, which really is a 
> more sensible way
> >to
> >> go.
> >
> >How does the separate pass affect the lightmaps though?  I'd 
> assume you want
> >the globe of light to eliminate any shadows under it, is 
> there a blending
> >mode that would do this?  Somehow keep the texture 
> information and just
> >eliminate the lightmap where there is a texel of light.
> >
> >Also, how is the globe of light calculated?  I can't seem to 
> figure that one
> >out.  Is it just a texture that looks like a circle of light that is
> >projected onto the geometry around the light source?  I 
> can't seem to come
> >up with a way that would work for a complex environment and 
> point lights.
> >
> >Or do you step through each texel on the lightmap 
> calculating distance to
> >the light source?  That would seem slow but easier to 
> program, just need to
> >convert the UV on the lightmap texture to world space to 
> calculate intensity
> >of the light at that point.
> >
> >Thanks,
> >
> >- Jason Zisk
> >- nFusion Interactive LLC
> >
> >_______________________________________________
> >GDAlgorithms-list mailing list
> >GDA...@li...
> >http://lists.sourceforge.net/mailman/listinfo/gdalgorithms-list
> >_______________________________________________
> >GDAlgorithms-list mailing list
> >GDA...@li...
> >http://lists.sourceforge.net/mailman/listinfo/gdalgorithms-list
> >
> >
> --------------------------------------
> Charles Bloom           www.cbloom.com
> 
> _______________________________________________
> GDAlgorithms-list mailing list
> GDA...@li...
> http://lists.sourceforge.net/mailman/listinfo/gdalgorithms-list
>