Combining shaders

re: Combining shaders

Eric — Wed, 07 Apr 2010 19:04:23 GMT

tep, instead of Deferred, you should go with the Light Pre-Pass method by Wolfgang Engel. Not only does it handle trasparency well, it's also compatible with msaa and doesn't limit the different materials you can use :)

re: Combining shaders

n00body — Sun, 23 Aug 2009 09:04:35 GMT

Funny how the flexible nature of shaders encourages us to want to make them generalized, but makes them difficult to generalize.

re: Combining shaders

Shawn Hargreaves - MSFT — Sun, 23 Aug 2009 03:33:31 GMT

tep: I did some early deferred shading research in my previous job, but never tried this using the XNA Framework.

I have mixed feelings about deferred rendering - it has some major advantages, but also significant disadvantages that can't really be worked around (eg. doesn't work at all for alpha blended geometry). I think it really depends on your game whether the tradeoffs are worth it.

I also don't think deferred shading is that great a fit for the 360 GPU hardware (limited EDRAM, predicated tiling, etc).

re: Combining shaders

tep — Sun, 23 Aug 2009 00:51:12 GMT

Yeah, this is a well known shader permutation problem. People usually write uber shader, or generate all variants of the shader or people nowadays are using deferred shading/lighting approach.

Shawn, I wonder if you have experimented with deferred lighting on XNA? and how well does it perform?

I also wonder if you are proponent of deferred approaches (either shading/lighting)?

re: Combining shaders

Chris Vaughan — Wed, 19 Aug 2009 18:32:44 GMT

Helpful and clear as ever Shawn, thank you so much!

Now, where can I start a petition to get this sort of info included in the XNA documentation?? ;¬)

re: Combining shaders

Rex Guo — Tue, 18 Aug 2009 08:08:24 GMT

Shawn, this is a great and timely post. Seems like this is the wild west and everyone has their own scheme of doing things. Will we ever get to a stage where routines and concepts are standardised, like the STL for C++? I don't know, given how fragmented the shader languages are.

And there have been many attempts at solving it using graph-based editors but none of those have yet to become a de-facto standard too.