On 2012-10-07 22:27:55 +0200, Diego Biurrun wrote: > On Sun, Oct 07, 2012 at 07:52:54PM +0100, Måns Rullgård wrote: > > Diego Biurrun <[email protected]> writes: > > > On Sun, Oct 07, 2012 at 07:39:09PM +0200, Janne Grunau wrote: > > >> On 2012-10-07 19:22:30 +0200, Diego Biurrun wrote: > > >> > On Sun, Oct 07, 2012 at 06:50:44PM +0200, Janne Grunau wrote: > > >> > > > > >> > > --- /dev/null > > >> > > +++ b/libavcodec/x86/vc1dsp.h > > >> > > @@ -0,0 +1,35 @@ > > >> > > + > > >> > > +#ifndef AVCODEC_X86_VC1DSP_H > > >> > > +#define AVCODEC_X86_VC1DSP_H > > >> > > + > > >> > > +#include "libavcodec/vc1dsp.h" > > >> > > + > > >> > > +void ff_vc1dsp_init_mmx(VC1DSPContext *dsp); > > >> > > +void ff_vc1dsp_init_mmxext(VC1DSPContext *dsp); > > >> > > + > > >> > > +#endif /* AVCODEC_X86_VC1DSP_H */ > > >> > > > >> > Huh? This is already declared in libavcodec/vc1dsp.h ... > > >> > > >> AVCODEC_VC1DSP_H vs. > > >> AVCODEC_X86_VC1DSP_H > > > > > > Not sure what you are trying to say there, let me be more explicit: > > > Move the ff_vc1dsp_init_mmxext declaration to libavcodec/vc1dsp.h. > > > > That would be counter the purpose of this patch. > > ? > > Quoting from libavcodec/vc1dsp.h: > > void ff_vc1dsp_init(VC1DSPContext* c); > void ff_vc1dsp_init_altivec(VC1DSPContext* c); > void ff_vc1dsp_init_mmx(VC1DSPContext* dsp); > > Now you could argue that the arch init functions have no place in the > generic code, but we have a lot of them, so that cleanup would have to > be done separately. Also, the current patch results in two declarations.
no, you missed that the same patch renames ff_vc1dsp_init_mmx() to ff_vc1dsp_init_x86(). ff_vc1dsp_init_mmx() is then reused for the initialization of the inline mmx function pointers. Same for the new ff_vc1dsp_init_mmxext. I don't think it makes sense to touch the altivec initialization in the same patch. Janne _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
