[XviD-devel] MMX/SSE/SEE2 implementation
daniel smith
xvid-devel@xvid.org
Thu, 12 Dec 2002 23:30:06 +0800
> Back last summer I played with various Xvid nasm sad16_sse2
> optimizations and nothing seemed to make much difference for some
> reason, so I never released it.
I remember it took quite a few attempts before I could get an SSE2 SAD16 function faster than the XMM one, but it was replaced in CVS by one that appears slower than XMM. In any case, mine only ran ~15% faster on a P4 than the XMM code, however the specs on the test machine were so strange that it might not be optimal on a decent RDRAM setup anyway.
dan
--
_______________________________________________
Get your free email from http://www.astroboymail.com
Powered by Outblaze