On Wed, Sep 12, 2012 at 12:02:45PM +0100, Måns Rullgård wrote:
> Diego Biurrun <di...@biurrun.de> writes:
> > On Tue, Sep 11, 2012 at 05:17:11PM +0300, Martin Storsjö wrote:
> >> On Tue, 11 Sep 2012, Diego Biurrun wrote:
> >> >On Mon, Sep 03, 2012 at 11:54:20AM +0200, Diego Biurrun wrote:
> >> >>---
> >> >> libavcodec/imgconvert.c | 2 +-
> >> >> libavcodec/x86/ac3dsp_init.c | 2 +-
> >> >> libavcodec/x86/cavsdsp.c | 2 +-
> >> >> libavcodec/x86/{dsputil_mmx.h => dsputil.h} | 6 +++---
> >> >> libavcodec/x86/dsputil_mmx.c | 2 +-
> >> >> libavcodec/x86/dsputilenc_mmx.c | 2 +-
> >> >> libavcodec/x86/h264_qpel.c | 2 +-
> >> >> libavcodec/x86/h264dsp_init.c | 2 +-
> >> >> libavcodec/x86/idct_xvid.c | 2 +-
> >> >> libavcodec/x86/motion_est.c | 2 +-
> >> >> libavcodec/x86/mpegvideo.c | 2 +-
> >> >> libavcodec/x86/mpegvideoenc.c | 2 +-
> >> >> libavcodec/x86/rv40dsp_init.c | 2 +-
> >> >> libavcodec/x86/simple_idct.c | 2 +-
> >> >> libavcodec/x86/snowdsp.c | 2 +-
> >> >> libavcodec/x86/vc1dsp_mmx.c | 2 +-
> >> >> libavfilter/x86/yadif.c | 2 +-
> >> >> 17 files changed, 19 insertions(+), 19 deletions(-)
> >> >> rename libavcodec/x86/{dsputil_mmx.h => dsputil.h} (97%)
> >> >
> >> >ping
> >>
> >> What about naming it dsputil_x86.h instead? While we have a few
> >> ambiguities between headers in libavcodec and libavcodec/<arch>, I
> >> think this might be better off with a less ambiguous name.
> >
> > That would be inconsistent. It would be the only header with a different
> > name in the arch subdirectory. Compare (just for x86):
> >
> > libavcodec/cabac.h ---> libavcodec/x86/cabac.h
> > libavcodec/fft.h ---> libavcodec/x86/fft.h
> > libavutil/cpu.h ---> libavutil/x86/cpu.h
> > libavutil/intreadwrite.h ---> libavutil/x86/intreadwrite.h
> > libavutil/bswap.h ---> libavutil/x86/bswap.h
> > libavutil/timer.h ---> libavutil/x86/timer.h
>
> In that list, all but fft.h are a totally different situation. The
> $arch/foo.h contain arch-specific overrides for things defined in the
> base foo.h. This is not at all the case with fft or dsputil. Those two
> contain but arch-internal declarations and are never included into
> common code.
Not true, cf.
ff_deinterlace_line_*
ff_*_pixels_clamped_*
Then there is some stuff for vc1, rv40 and cavs, but that's to be
expected in that spaghetti code. I'm slowly working to clean it
up anyway, see the patches I will send in a moment.
In any case I don't see how _mmx in the name improves anything. That
just makes things less clear. What's your suggestion?
Diego
_______________________________________________
libav-devel mailing list
libav-devel@libav.org
https://lists.libav.org/mailman/listinfo/libav-devel
