|
| |
Originally posted by M3d10n@Aug. 14 2002, 10:21 am
It is possible to convert Saturn 3D models to some standard format. Each face in the saturn model is made of 4 vertices, the converter could simply create 2 triangular faces for each quad face. Of course, in certain cases the result would look quite different... since the saturn's sprite rendered can deal with faces made of non co-planar vertices without trouble - after all, they'll be converted to 4 flat screen coordinates where the distorted sprite corners will be drawn - but when sliced into 2 triangular faces, the volume effect wouldn't be the same.
Also, hardware acceleration is possible. Just grab all the vertices on the scene (2D and 3D, doesn't matter), and use them to build textured polygons, that can be easily manipulated by any 3D hardware. I think the Satourne and the Giri Giri both do that, since they allow the use of HW bilinear filtering, and giri giri has texture perspective correction problems.
|
Er, that's not that immediate. The main problem, actually, is the matrix deformations. Unlike the Dreamcast and N64 (and PSX?) there aren't the so called "display-lists", ie. a list of 3D models that is to be rendered. What are is a list of sprite deformations, in the form of lots of matrix data. That has no correlation with 3D space, but can be turned into 3D polys (like giri giri and satourne do). But, when we get that matrix data, the SH-2 processor has already done all the 3D world to 2D plane conversion, so the resulting acceleration is minimal.
We'd need to find out _when_ the SH-2 are making that conversion, and then use an PC/DC native code. It would be kinda like MDEC decoding por PSX. The problem is that the library (and can be hundreds of diferent solutions for all we know) is not always in the same place in memory, since it's linked at compile time... |