[cairo] [PATCH] SSE2 support to pixman

[Soeren Sandmann]

"André Tupinambá" <andrelrt at gmail.com> writes:

> I'm my machine works fine (software engineer's frequent expressions #2 :)
> 
> This don't make any sense to me. I saw the assembly code generated
> from pixman-sse.c compilation, and it's too much optimized to be
> slower than MMX one.
> 
> Did you change the save128WriteCombining to save128Aligned and ran the
> perf to check this?
> 
> Ok, I didn't make the opaque check, I know. I would run a performance
> test to check which one is faster (look at TODO in the code), but if
> you already test it...

Yes, when I added an opaque check and changed to save128Aligned(), I
got around a 3x speedup over the MMX version. Unfortunately, I
accidentally deleted the patch.



Soren