On Sat, 15 Mar 2014 15:46:53 +0100, Diego Biurrun <[email protected]> wrote:
> On Sat, Mar 15, 2014 at 12:06:22PM +0100, Anton Khirnov wrote:
> > On Fri, 14 Mar 2014 05:42:22 -0700, Diego Biurrun <[email protected]> wrote:
> > > --- a/libavcodec/imgconvert.c
> > > +++ b/libavcodec/imgconvert.c
> > > @@ -40,10 +40,6 @@
> > >  #include "libavutil/imgutils.h"
> > >  
> > >  #if HAVE_MMX_EXTERNAL
> > > -#include "x86/dsputil_x86.h"
> > > -#endif
> > > -
> > > -#if HAVE_MMX_EXTERNAL
> > >  #define deinterlace_line_inplace ff_deinterlace_line_inplace_mmx
> > >  #define deinterlace_line         ff_deinterlace_line_mmx
> > >  #else
> > > --- a/libavcodec/imgconvert.h
> > > +++ b/libavcodec/imgconvert.h
> > > @@ -21,6 +21,24 @@
> > >  
> > >  #include <stdint.h>
> > >  
> > > +#include "version.h"
> > > +
> > > +#if FF_API_DEINTERLACE
> > > +
> > > +void ff_deinterlace_line_mmx(uint8_t *dst,
> > > +                             const uint8_t *lum_m4, const uint8_t 
> > > *lum_m3,
> > > +                             const uint8_t *lum_m2, const uint8_t 
> > > *lum_m1,
> > > +                             const uint8_t *lum,
> > > +                             int size);
> > > +
> > > +void ff_deinterlace_line_inplace_mmx(const uint8_t *lum_m4,
> > > +                                     const uint8_t *lum_m3,
> > > +                                     const uint8_t *lum_m2,
> > > +                                     const uint8_t *lum_m1,
> > > +                                     const uint8_t *lum, int size);
> > > +
> > > +#endif /* FF_API_DEINTERLACE */
> > 
> > eew, arch-specific code outside of arch-specific dirs.
> 
> Well, imgconvert is one of the ugliest remaining parts in libavcodec...
> 
> > Why not x86/imgconvert.h?
> 
> I thought about it, but all of this is deprecated and scheduled for removal,
> so I figured it was not worth the trouble to separate it cleanly.  We'd have
> to remember to remove the header as well at the next version bump.
> 
> There is no dsp-like function pointer abstraction going on, it's all being
> set from the non-x86 code anyway.  It's a mess anyway and cleaning it up
> seemed like a pointless distraction...
> 

Oh well, I guess i can live with this for now

-- 
Anton Khirnov
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to