[FFmpeg-devel] [PATCH] swscale alpha channel support (was: [RFC] Alpha support)

[Michael Niedermayer]

On Tue, Feb 10, 2009 at 11:12:50AM +0100, C?dric Schieli wrote:
[...]
> > > +                    YSCALEYUV2PACKEDX_ACCURATE
> > > +                    YSCALEYUV2RGBX
> > > +                    "movq                      %%mm2, "U_TEMP"(%0)
> >  \n\t"
> > > +                    "movq                      %%mm4, "V_TEMP"(%0)
> >  \n\t"
> > > +                    "movq                      %%mm5, "Y_TEMP"(%0)
> >  \n\t"
> > > +                    YSCALEYUV2PACKEDX_ACCURATE_YA(ALP_MMX_FILTER_OFFSET)
> >
> > > +                    "movq                      %%mm3, %%mm2
> > \n\t"
> > > +                    "movq               "Y_TEMP"(%0), %%mm5
> > \n\t"
> > > +                    "psraw                        $3, %%mm1
> > \n\t"
> > > +                    "psraw                        $3, %%mm7
> > \n\t"
> > > +                    "packuswb                  %%mm7, %%mm1
> > \n\t"
> > > +                    "movq                      %%mm1, %%mm7
> > \n\t"
> > > +                    WRITEBGR32(%4, %5, %%REGa)
> >
> > the 2 register-register moves look avoidable if WRITEBGR32 wouldnt use
> > fixed registers but rather have the used args passed as arguments.
> >
> > also note, a patch changing WRITEBGR32 to take its registers as arguments
> > should be seperate. Also some of the changes you did like spliting macros
> > could be seperate (its trivial and could be commited already)
> >
> 
> That's what I wanted to do in a second round (to help readability of the
> patch), but if you think it should be done first, I'll be happy to issue
> those patches.

i prefer doing that first


> 
> 
> >
> > also iam assuming all the code has been tested?
> > and when already speaking about testing things should be tested on
> > x86_32 and 64
> >
> 
> It has only been tested on x86_32 for the moment. I'll try on a 64bit system
> soon.

thank you

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Complexity theory is the science of finding the exact solution to an
approximation. Benchmarking OTOH is finding an approximation of the exact
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20090210/96907e06/attachment.pgp>