ralf willenbacher wrote:
>
> i replaced the transform_v16, matmul4 and matmul34 with simd instruction
> functions and there was no performance increase in quake2 or q3test.
> any good reason to have simd support anyway ? :(
>
Q3test is a big application. The matmul routines don't even register
It is likely that neither of those applications are anywhere near transform
limited. Even on a Voodoo 3 or TNT2 you are probably fill rate limited on
Q2 or Q3test. If you are running something that does 20k triangles per
frame, you will likely see a difference, but only if you are not fill ra
i replaced the transform_v16, matmul4 and matmul34 with simd instruction
functions and there was no performance increase in quake2 or q3test.
any good reason to have simd support anyway ? :(
--
ralf willenbacher ([EMAIL PROTECTED])
___
Mesa-dev maill
