Indeed!

I worked last night on the Mesa T&L engine, improving Keith's code a bit. So far I got 10% increase in performance.

However, the changes are not available to windoze users yet, because of the wicked M$ calling convention (d@mnit, why do they have to be different every time?)

Quote:

I worked last night on the Mesa T&L engine, improving Keith's code a bit. So far I got 10% increase in performance.

¡GOOD NEW!  ;D

Regards,
Oscar

Soon  ;)

wrote on 01.04.04 at 19:49:59:

So what's the chance for a Windoze User to see the difference ? And for a Linux User ?

Oh, the windoze users WILL see the difference: the optimized version doesn't compile for "whendoes"  ;D ;D ;D

... yet!

However, I need to make some corrections:
1) The speed gain was reported wrt recently released Mesa 6.1, not Mesa 5.1
2) The speed gain is not fixed. I calculated it based on some synthetic heavy-polygon benchmarks. In a real world application, it might be smaller or might be even greater, depending on the scene being rendered. Your mileage may vary!  :P