Original Post
Hi everyone. I''m experimenting with different software-rendering techniques for rendering small-triangle models.
As some already know, the trick for fast software-rendering is often to ''cheat'' or to try some less-obvious approaches. It might help to think of the MDL models in Quake 1, for example, which used affine-mapping (and/or subdivision) for quicker (but less accurate) rendering. The models contain around 400 polygons with a 128x128 skin texture.
The target platform is a Sony-Ericcson P900 smartphone, so the main issues are related to the only-integer rendering pipeline (the phone''s ARM cpu slows down unacceptably with floating-point calcs). In any case, i''m very curious to hear other people''s opinions especially from people with experience in software-rendering from around the quake-era.