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

İsmail Dönmez ismail
Wed Feb 27 23:55:20 CET 2008


On Thu, Feb 28, 2008 at 12:43 AM, Michael Niedermayer <michaelni at gmx.at> wrote:
> On Thu, Feb 28, 2008 at 12:39:20AM +0200, ?smail D?nmez wrote:
>  > That assumes every assembly coder is talented like Leonardo da Vinci,
>  > good luck with that assumption :-)
>  It just assumes that most asm coders who work on ffmpeg are better than gcc.
>  And i claim that this is true.

I know FFmpeg developers are talented, it wouldn't be this fast
otherwise. But the other day I was talking with LLVM [0] guys and they
told me that inline asm is just harder to optimize, and even that LLVM
did better job than gcc in an particular MMX intrinsics case which you
other day sent to mailing list to show intrinsics vs inline asm
problem. So I think one day new tools [1] might be just good enough
for intrinsics.


[0] http://llvm.org
[1] http://clang.llvm.org

Never learn by your mistakes, if you do you may never dare to try again

More information about the ffmpeg-devel mailing list