[FFmpeg-devel] [PATCH 2/5] avcodec/nvdec: Add support for decoding HEVC 4:4:4 content

[Philip Langdale]

On Sat, 20 Oct 2018 22:58:34 +0200
Timo Rothenpieler <timo at rothenpieler.org> wrote:

> >   
> > +    // It it semantically incorrect to use AX_PIX_FMT_YUV444P16
> > for either the 10
> > +    // or 12 bit case, but ffmpeg and nvidia disagree on which end
> > the padding
> > +    // bits go at. P16 is unambiguous and matches.  
> 
> This comment seems redundant, since AX_PIX_FMT_YUV444P16 isn't even
> used.

Yeah, that's outdated from the original version with no pix fmts. I'll
remove it.

> > diff --git a/libavutil/hwcontext_cuda.c b/libavutil/hwcontext_cuda.c
> > index 3b1d53e799..43337f14f0 100644
> > --- a/libavutil/hwcontext_cuda.c
> > +++ b/libavutil/hwcontext_cuda.c
> > @@ -38,6 +38,8 @@ static const enum AVPixelFormat
> > supported_formats[] = { AV_PIX_FMT_YUV444P,
> >       AV_PIX_FMT_P010,
> >       AV_PIX_FMT_P016,
> > +    AV_PIX_FMT_YUV444P10_MSB,
> > +    AV_PIX_FMT_YUV444P12_MSB,  
> 
> You are adding these to supported formats, but are not actually
> adding support, so cuda_frames_init and cuda_get_buffer will both run
> into the BUG case. Should be super easy, as they're identical to
> 444P16.

This actually works fine. cuda_frames_init and cuda_get_buffer both use
generic mechanisms that introspect the pix desc information.
 
> >       AV_PIX_FMT_YUV444P16,  
> Technically, this can go now. But I guess removing it is an API
> break, so it gotta stay I guess.

Right. And at the end of the day it works correctly in the context of
the hwcontext. You might feed into some theoretical filter that would
handle it appropriately. Just because nvenc can't actually consume it
fully doesn't make it wrong.

--phil