On Sat, Aug 09, 2014 at 04:43:57PM -0700, John Stebbins wrote:
> On 08/09/2014 03:19 PM, John Stebbins wrote:
> > On 08/09/2014 01:42 PM, Diego Biurrun wrote:
> >> On Sat, Aug 09, 2014 at 09:25:36AM -0700, John Stebbins wrote:
> >>> --- a/libavcodec/mpegvideo.c
> >>> +++ b/libavcodec/mpegvideo.c
> >>> @@ -379,7 +379,6 @@ av_cold int ff_dct_common_init(MpegEncContext *s)
> >>> {
> >>> ff_blockdsp_init(&s->bdsp, s->avctx);
> >>> ff_hpeldsp_init(&s->hdsp, s->avctx->flags);
> >>> - ff_idctdsp_init(&s->idsp, s->avctx);
> >>> ff_me_cmp_init(&s->mecc, s->avctx);
> >>> ff_mpegvideodsp_init(&s->mdsp);
> >>> ff_videodsp_init(&s->vdsp, s->avctx->bits_per_raw_sample);
> >> I wonder if some of the places where ff_dct_common_init is called should
> >> be replaced by just ff_idct_common_init now.
> > This would take some research to determine. But I can probably do a quick
> > smoke test by removing the calls and see what
> > breaks in fate.
>
> There is only one place outside of ff_MPV_common_init that ff_dct_common_init
> is called and that is dnxhdenc. I removed
> the call here and it passes fate. So then I did my best inspect the code to
> verify that the things initialized by
> ff_dct_common_init are not used by dnxhdenc. This appears to be true, but
> there is a lot being initialized here and it
> is completely possible I have missed a code path. So it's up to you whether
> you would like to remove this call.
I would like to apply the boyscout rule to this code, otherwise we will
never see the day when it is halfway acceptable. So yes, I think we
should try to disentangle dnxhdenc from that part of mpegvideo.
Diego
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel
