[FFmpeg-devel] [flamefest-start] A little something on MMX/SSE intrinsics

[Alexander Strange]

On Wed, Feb 27, 2008 at 3:24 PM, Luca Barbato <lu_zero at gentoo.org> wrote:
> Michael Niedermayer wrote:
>  > On Wed, Feb 27, 2008 at 05:42:58PM +0200, Ivan Kalvachev wrote:
>  >> On Wed, Feb 27, 2008 at 4:31 PM, Reimar D?ffinger
>  >> <Reimar.Doeffinger at stud.uni-karlsruhe.de> wrote:
>  >>> Hello,
>  >>>  yet another case of someone relearning "intrinsics are still a mess and
>  >>>  compilers still much too stupid" ;-) :
>  >>>  http://www.virtualdub.org/blog/pivot/entry.php?id=188#body
>  >>>
>  >>>  Let's see if we can make today the flame-heaviest day in FFmpeg history
>  >>>  :-)
>  >> Nice read.
>  >>
>  >> I'd just like to add that there is already architecture in FFmpeg that
>  >> is entirely optimized on intrinsics - AltiVect. There were few
>  >> fundamental problems with it and I'm not sure if they have been
>  >> successfully solved. These include things like generating code that
>  >> picks up processor SIMD dialect at runtime.
>  >
>  > You just found us a new SOC qualification task
>  > rewrite intrinsics to proper asm()
>  >
>
>  And benchmark the result ^^;
>
>  lu

I don't think anyone can get Altivec asm to work better than
intrinsics on more than one CPU - PPC is really, really
scheduling-sensitive, especially the G5 and Cell.

I guess you can always try, though, but don't do anything to
discourage people who know altivec from adding more. There's still a
lot missing from H.264.