[FFmpeg-devel] [PATCH v2 4/4] avcodec/nvenc: support for HEVC timecode passthrough

[lance.lmwang at gmail.com]

On Wed, Jun 17, 2020 at 06:41:45PM +0200, Timo Rothenpieler wrote:
> On 17.06.2020 18:07, lance.lmwang at gmail.com wrote:
> > From: Limin Wang <lance.lmwang at gmail.com>
> > 
> > Signed-off-by: Limin Wang <lance.lmwang at gmail.com>
> > ---
> > I failed to find any document about nvenc so that I can update for new option.
> > 
> >   libavcodec/nvenc.c      | 53 +++++++++++++++++++++++++++++++++++++++++++++++++
> >   libavcodec/nvenc.h      |  1 +
> >   libavcodec/nvenc_hevc.c |  1 +
> >   3 files changed, 55 insertions(+)
> > 
> > diff --git a/libavcodec/nvenc.c b/libavcodec/nvenc.c
> > index e4356ce..68b212a 100644
> > --- a/libavcodec/nvenc.c
> > +++ b/libavcodec/nvenc.c
> > @@ -22,6 +22,8 @@
> >   #include "config.h"
> >   #include "nvenc.h"
> > +#include "hevc_sei.h"
> > +#include "put_bits.h"
> >   #include "libavutil/hwcontext_cuda.h"
> >   #include "libavutil/hwcontext.h"
> > @@ -2068,6 +2070,15 @@ static void reconfig_encoder(AVCodecContext *avctx, const AVFrame *frame)
> >       }
> >   }
> > +static unsigned bcd2uint(uint8_t bcd)
> > +{
> > +    unsigned low  = bcd & 0xf;
> > +    unsigned high = bcd >> 4;
> > +    if (low > 9 || high > 9)
> > +        return 0;
> > +    return low + 10*high;
> > +}
> > +
> >   int ff_nvenc_send_frame(AVCodecContext *avctx, const AVFrame *frame)
> >   {
> >       NVENCSTATUS nv_status;
> > @@ -2156,6 +2167,48 @@ int ff_nvenc_send_frame(AVCodecContext *avctx, const AVFrame *frame)
> >               }
> >           }
> > +        if (ctx->tc && av_frame_get_side_data(frame, AV_FRAME_DATA_S12M_TIMECODE)) {
> > +            AVFrameSideData *sd = av_frame_get_side_data(frame, AV_FRAME_DATA_S12M_TIMECODE);
> > +            if (sd) {
> > +                uint32_t *tc =  (uint32_t*)sd->data;
> > +                int m = tc[0] & 3;
> > +                if (tc) {
> > +                    sei_data[sei_count].payloadSize = sizeof(uint32_t) * 4;
> > +                    sei_data[sei_count].payloadType = HEVC_SEI_TYPE_TIME_CODE;
> > +                    sei_data[sei_count].payload     = av_mallocz(sei_data[sei_count].payloadSize);
> > +                    if (sei_data[sei_count].payload) {
> > +                        PutBitContext pb;
> > +
> > +                        init_put_bits(&pb, sei_data[sei_count].payload, sei_data[sei_count].payloadSize);
> > +                        put_bits(&pb, 2, m); // num_clock_ts
> > +
> > +                        for (int j = 1; j <= m; j++) {
> > +                            uint32_t tcsmpte = tc[j];
> > +                            unsigned hh   = bcd2uint(tcsmpte     & 0x3f);    // 6-bit hours
> > +                            unsigned mm   = bcd2uint(tcsmpte>>8  & 0x7f);    // 7-bit minutes
> > +                            unsigned ss   = bcd2uint(tcsmpte>>16 & 0x7f);    // 7-bit seconds
> > +                            unsigned ff   = bcd2uint(tcsmpte>>24 & 0x3f);    // 6-bit frames
> > +                            unsigned drop = tcsmpte & 1<<30 && !0;  // 1-bit drop if not arbitrary bit
> > +
> > +                            put_bits(&pb, 1, 1); // clock_timestamp_flag
> > +                            put_bits(&pb, 1, 1); // units_field_based_flag
> > +                            put_bits(&pb, 5, 0); // counting_type
> > +                            put_bits(&pb, 1, 1); // full_timestamp_flag
> > +                            put_bits(&pb, 1, 0); // discontinuity_flag
> > +                            put_bits(&pb, 1, drop);
> > +                            put_bits(&pb, 9, ff);
> > +                            put_bits(&pb, 6, ss);
> > +                            put_bits(&pb, 6, mm);
> > +                            put_bits(&pb, 5, hh);
> > +                            put_bits(&pb, 5, 0);
> > +                        }
> > +                        flush_put_bits(&pb);
> > +                        sei_count++;
> > +                    }
> > +                }
> > +            }
> > +        }
> > +
> 
> Does all this really have to be in nvenc? It looks very generic to me, and
> better put in some central location to be re-used by multiple codecs.

Yes, I'll add a general function for reuse next version.

> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> 
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe".

-- 
Thanks,
Limin Wang