i dont know, maybe its just nvidia trying to piss me off, but isnt the gf3 supposed to support the matrix palette extension? is it just crappy drivers or nvidia forcing people to do it their way? a nice way to suddenly say "oh, your card got too old, buy a new one"? i remember a dx8 sample that required removing the check for support (which would return "not supported") and blindly use the feature (worked fine of course). no im trying to finally get some skinning done, matrix palette etc. looks perfect and vertex programs are obviously meant to use and support them too. so what exactly is going on and if matrix palette doesnt work, how many other options are there to select the right matrices in the vertex program? at least program matrices work fine, but how would i select them?

Yeah, those skinning hw features weren''t very popular. Vertex shaders seem to be the ideal way to do it on hw and might not involve splitting the mesh into pieces.

hm.. the problem is: vertex programs would still make use of the matrix palette, which isnt supported. so unless im missing the way to pick the right program matrix (maybe they replaced the matrix palette?), im not sure how to go about it.

splitting the mesh is just what i want to avoid by sending each vertex along with one or more matrix indices and weights (for now one index would be enough). problem is that i must somehow miss the proper way to pass them to my vertex program. it works fine, all program matrices are set up.. all i need is a way to pick the right one for the current matrix.

Using vertex shaders:
Just store your matrices in 3 or 4 constant vector registers, and use the index register to select them.

- JQ

hm.. meaning i cant use the program matrices? little unsure what you mean by index register, guess i''ll read the specs once more. except you mean the address register. an old sample for nv''s extension was doing that, but relative addressing doesnt seem to work for program matrices.