[Open-graphics] Why "swShader" is not interesting

Nicolas Capens nicolas at capens.net
Thu Dec 9 04:25:37 EST 2004


Hi Kent,

> You completely missed his point. Every instruction the CPU spends
> doing the job of the graphics accelerator is an instruction it is not
> working on actual tasks.

You completely missed the context. He mentioned all the "Zen of" books by 
Abrash. They were all written before graphics cards became popular. It was a 
straight insult for my CPU programming qualities, so I showed him Abrash has 
little to teach me.

> I want to have smooth window compositing
> happening while five things are working in the background. And since
> the fastest machine I own is barely up to the 1.2 GHz Celeron standard
> you consider minimal for swShader, that solution is not a solution.
> TS's card will be.

That 1.2 GHz Celeron is what I consider minimal for running a game like 
Quake 3 at playable framerates. Window compositing doesn't have to be done 
in 40 milliseconds. There's no hard deadline. Furthermore, window 
compositing doesn't require complex texturing operations.

Anyway, I never proposed that swShader is a solution for the Open Graphic's 
goals. I was only showing what performance level should be considered 
minimal, in the context that people were claiming that the raw processing 
power of the FPGA would suffice. It's non-trivial to combine the combined 
performance and flexibility of a TNT2 and swShader. You can't deny this has 
to be a goal for the Open Graphics project. Other than in this context, 
let's not discuss what swShader can and can't ok? I know perfectly where it 
stands.

Cheers,

Nicolas 




More information about the Open-graphics mailing list