I''ve created a software renderer which supports ps 2.0 shaders. Unlike the reference rasterizer, it is not an interpreter but an optimized JIT compiler using SIMD instructions. So it is dozens of times faster than refrast and beats hardware in flexibility. For one, it is no problem to go beyond the 32 texture instructions limit in ps 2.0 hardware, while still keeping things real-time. You can read the details here: http://www.flipcode.com/cgi-bin/msg.cgi?showThread=COTD-swShader&forum=cotd&id=-1.

Comments welcome!

Refrast is totally NOT optimized. It doesn''t use any MMX/SSE instructions and it doesn''t use branch elimination. It''s an interpreter which means it also wastes time parsing the shader. It''s just written in plain C...

swShader compiles ps 2.0 code into x86 code using lots of MMX and SSE instructions. It uses only the code path for the current render state, so all compare and branch instructions are eliminated. A caching system makes sure that you can switch render modes efficiently. So what you get is THE optimal code for a certain situation. No compromises.