RE: [GD-Consoles] Textures vs. Polygons

[Awen L. <ali...@ed...>]

>2.  Jamie/Awen:  Sorry, I am not clear on what you mean by splitting the
geometry matrices, could you explain this in more detail?

Hope i'm going to describe what Jamie had in mind...
Ok: objects verts/polys/etc are sorted and stripped using some criterions
(polys batched on a per-material basis is a common practice). Strip best
size, and neihgbouring choice, is generally choosen using many params
considerations (GPU cache size ?). More specifically on a PS2, if you choose
to use VU to do your mesh mathematics, the VU mem size may be one of these
params. For an object, you need to compute the storage required by the VU to
process one of your vertex (3d coords+normal+uvs+rgbs... etc.), and usually
a matrix or two, to rotate and light your stuff. You have to be smart with
you VU and make some decisions on what could stay for a while, and what must
be quickly discarded to minimize DMA traffic. Matrices may be considered
more 'static' than verts flow.
For an animated object your need to add some skinning consideration to that
point, and configure your mesh to require the least possible matrices
swapping. And it shows naturally the way you choose to strip/sort your mesh
polys.

Ant it's a bit related to your texture details problem, since if you NEED to
split your mesh into chunks because of matrices, why not having many
materials per object, if you're a king of parallelization and swapping
texture buffer, text upload cost stuff may be masked.
Sorry to stay a bit vague, it begins to touch some dungeon secrets, around
here ;)