Dear luke, I've been very curious about this, but I've waited until you were stuck in some place before asking so as not to busy you. I'm not arrogant enough to say I'll create the 3D mesa driver or even go beyond theorizing, but I'm very eager to learn.
1. How do we know when it works? Is there HW or an emulator system we can use? 2. How do we know it's fast enough? What do we benchmark it with? 3. What exact APIs do you want implemented? 4. Do you want it in assembly with C and/or C++ bindings or in one or both of C/C++? 5. Are there any special assembly instructions that this will need?
If there's anything else you think would be useful, do tell.
Thanks, David