[Ffmpeg-devel] [PATCH] SSE counterpart of ff_imdct_calc_3dn2

Hervé W. H.O.W.aka.V+ffmpeg
Fri Aug 25 00:53:07 CEST 2006


On 24/08/06, Rich Felker <dalias at aerifal.cx> wrote:
> On Thu, Aug 24, 2006 at 06:35:07PM +0200, Guillaume Poirier wrote:
> > Rich Felker wrote:
> > > On Thu, Aug 24, 2006 at 09:59:37AM +0200, Guillaume POIRIER wrote:
> > >
> > >>>Intrinsics are also gcc4-specific
> > >>
> > >>False, They existed in 3.4 and I think in 3.3 also (I don't know about
> > >>earlier releases, but for sure 2.95 do not support them).
> > >
> > >
> > > Only gcc4 and later have the 3dnow intrinsics.
> >
> > I wasn't specifically talking about 3dnow intrinsics...
> > You probably should back up your claims Rich, or precise that you made
> > smth up.
> > http://gcc.gnu.org/onlinedocs/gcc-3.4.0/gcc/X86-Built-in-Functions.html
>
> The docs are incorrect. If you're RTFml (mailing list) you would know
> that 3dnow intrinsics are actually missing on gcc prior to 4.x.

The BUILTINS for 3dnow exist in gcc 3.4. functions like:
v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)

GCC 4.x introduced a header that declares functions that wrap around
those builtins and (among other things) get rid of those
__builtin_ia32 guards.

Now, can the inline asm be made to work with both gcc and icc (by
someone who cares about icc support)?
if not, then maybe re-introducing the intrinsics versions for
platforms that don't support the inline asm used by ffmpeg would be
acceptable (if someone actually steps up to do the work of course) ?

-V




More information about the ffmpeg-devel mailing list