Re: [FFmpeg-devel] [PATCH v10 6/6] lavc/hevcdec: Parse DOVI RPU NALs

[Andreas Rheinhardt <andreas.rheinhardt@outlook.com>]

Niklas Haas:
> From: Niklas Haas <git@haasn.dev>
> 
> And expose the parsed values as frame side data. Update FATE results to
> match.
> 
> It's worth documenting that this relies on the dovi configuration record
> being present on the first AVPacket fed to the decoder, which in
> practice is the case if if the API user has called something like
> av_format_inject_global_side_data, which is unfortunately not the
> default.
> 
> This commit is not the time and place to change that behavior, though.
> 
> Signed-off-by: Niklas Haas <git@haasn.dev>
> ---
>  configure                  |   2 +-
>  libavcodec/hevcdec.c       |  37 ++++--
>  libavcodec/hevcdec.h       |   2 +
>  tests/ref/fate/hevc-dv-rpu | 224 +++++++++++++++++++++++++++++++++++++
>  4 files changed, 256 insertions(+), 9 deletions(-)
> 
> diff --git a/configure b/configure
> index 8303e1329e..3f71af77db 100755
> --- a/configure
> +++ b/configure
> @@ -2826,7 +2826,7 @@ h264_decoder_suggest="error_resilience"
>  hap_decoder_select="snappy texturedsp"
>  hap_encoder_deps="libsnappy"
>  hap_encoder_select="texturedspenc"
> -hevc_decoder_select="atsc_a53 bswapdsp cabac golomb hevcparse videodsp"
> +hevc_decoder_select="atsc_a53 bswapdsp cabac dovi_rpu golomb hevcparse videodsp"
>  huffyuv_decoder_select="bswapdsp huffyuvdsp llviddsp"
>  huffyuv_encoder_select="bswapdsp huffman huffyuvencdsp llvidencdsp"
>  hymt_decoder_select="huffyuv_decoder"
> diff --git a/libavcodec/hevcdec.c b/libavcodec/hevcdec.c
> index 3aa70e2245..8d7a4f7147 100644
> --- a/libavcodec/hevcdec.c
> +++ b/libavcodec/hevcdec.c
> @@ -2723,6 +2723,7 @@ error:
>  static int set_side_data(HEVCContext *s)
>  {
>      AVFrame *out = s->ref->frame;
> +    int ret;
>  
>      if (s->sei.frame_packing.present &&
>          s->sei.frame_packing.arrangement_type >= 3 &&
> @@ -2977,6 +2978,9 @@ static int set_side_data(HEVCContext *s)
>          s->rpu_buf = NULL;
>      }
>  
> +    if ((ret = ff_dovi_attach_side_data(&s->dovi_ctx, out)) < 0)
> +        return ret;
> +
>      return 0;
>  }
>  
> @@ -3308,16 +3312,23 @@ static int decode_nal_units(HEVCContext *s, const uint8_t *buf, int length)
>      if (s->pkt.nb_nals > 1 && s->pkt.nals[s->pkt.nb_nals - 1].type == HEVC_NAL_UNSPEC62 &&
>          s->pkt.nals[s->pkt.nb_nals - 1].size > 2 && !s->pkt.nals[s->pkt.nb_nals - 1].nuh_layer_id
>          && !s->pkt.nals[s->pkt.nb_nals - 1].temporal_id) {
> +        H2645NAL *nal = &s->pkt.nals[s->pkt.nb_nals - 1];
>          if (s->rpu_buf) {
>              av_buffer_unref(&s->rpu_buf);
>              av_log(s->avctx, AV_LOG_WARNING, "Multiple Dolby Vision RPUs found in one AU. Skipping previous.\n");
>          }
>  
> -        s->rpu_buf = av_buffer_alloc(s->pkt.nals[s->pkt.nb_nals - 1].raw_size - 2);
> +        s->rpu_buf = av_buffer_alloc(nal->raw_size - 2);
>          if (!s->rpu_buf)
>              return AVERROR(ENOMEM);
> +        memcpy(s->rpu_buf->data, nal->raw_data + 2, nal->raw_size - 2);
>  
> -        memcpy(s->rpu_buf->data, s->pkt.nals[s->pkt.nb_nals - 1].raw_data + 2, s->pkt.nals[s->pkt.nb_nals - 1].raw_size - 2);
> +        ret = ff_dovi_rpu_parse(&s->dovi_ctx, nal->data + 2, nal->size - 2);
> +        if (ret < 0) {
> +            av_buffer_unref(&s->rpu_buf);
> +            av_log(s->avctx, AV_LOG_WARNING, "Error parsing DOVI NAL unit.\n");
> +            /* ignore */
> +        }
>      }
>  
>      /* decode the NAL units */
> @@ -3450,8 +3461,8 @@ static int hevc_decode_frame(AVCodecContext *avctx, void *data, int *got_output,
>                               AVPacket *avpkt)
>  {
>      int ret;
> -    size_t new_extradata_size;
> -    uint8_t *new_extradata;
> +    uint8_t *sd;
> +    size_t sd_size;
>      HEVCContext *s = avctx->priv_data;
>  
>      if (!avpkt->size) {
> @@ -3463,14 +3474,17 @@ static int hevc_decode_frame(AVCodecContext *avctx, void *data, int *got_output,
>          return 0;
>      }
>  
> -    new_extradata = av_packet_get_side_data(avpkt, AV_PKT_DATA_NEW_EXTRADATA,
> -                                            &new_extradata_size);
> -    if (new_extradata && new_extradata_size > 0) {
> -        ret = hevc_decode_extradata(s, new_extradata, new_extradata_size, 0);
> +    sd = av_packet_get_side_data(avpkt, AV_PKT_DATA_NEW_EXTRADATA, &sd_size);
> +    if (sd && sd_size > 0) {
> +        ret = hevc_decode_extradata(s, sd, sd_size, 0);
>          if (ret < 0)
>              return ret;
>      }
>  
> +    sd = av_packet_get_side_data(avpkt, AV_PKT_DATA_DOVI_CONF, &sd_size);
> +    if (sd && sd_size > 0)
> +        ff_dovi_update_cfg(&s->dovi_ctx, (AVDOVIDecoderConfigurationRecord *) sd);

Looks good (and so much simpler than the earlier version).

> +
>      s->ref = NULL;
>      ret    = decode_nal_units(s, avpkt->data, avpkt->size);
>      if (ret < 0)
> @@ -3563,6 +3577,7 @@ static av_cold int hevc_decode_free(AVCodecContext *avctx)
>  
>      pic_arrays_free(s);
>  
> +    ff_dovi_ctx_unref(&s->dovi_ctx);
>      av_buffer_unref(&s->rpu_buf);
>  
>      av_freep(&s->md5_ctx);
> @@ -3647,6 +3662,7 @@ static av_cold int hevc_init_context(AVCodecContext *avctx)
>  
>      ff_bswapdsp_init(&s->bdsp);
>  
> +    s->dovi_ctx.logctx = avctx;
>      s->context_initialized = 1;
>      s->eos = 0;
>  
> @@ -3755,6 +3771,10 @@ static int hevc_update_thread_context(AVCodecContext *dst,
>      if (ret < 0)
>          return ret;
>  
> +    ret = ff_dovi_ctx_replace(&s->dovi_ctx, &s0->dovi_ctx);
> +    if (ret < 0)
> +        return ret;
> +
>      s->sei.frame_packing        = s0->sei.frame_packing;
>      s->sei.display_orientation  = s0->sei.display_orientation;
>      s->sei.mastering_display    = s0->sei.mastering_display;
> @@ -3811,6 +3831,7 @@ static void hevc_decode_flush(AVCodecContext *avctx)
>      HEVCContext *s = avctx->priv_data;
>      ff_hevc_flush_dpb(s);
>      ff_hevc_reset_sei(&s->sei);
> +    ff_dovi_ctx_flush(&s->dovi_ctx);
>      av_buffer_unref(&s->rpu_buf);
>      s->max_ra = INT_MAX;
>      s->eos = 1;
> diff --git a/libavcodec/hevcdec.h b/libavcodec/hevcdec.h
> index 870ff178d4..157bc6926f 100644
> --- a/libavcodec/hevcdec.h
> +++ b/libavcodec/hevcdec.h
> @@ -32,6 +32,7 @@
>  #include "avcodec.h"
>  #include "bswapdsp.h"
>  #include "cabac.h"
> +#include "dovi_rpu.h"
>  #include "get_bits.h"
>  #include "hevcpred.h"
>  #include "h2645_parse.h"
> @@ -574,6 +575,7 @@ typedef struct HEVCContext {
>      int nuh_layer_id;
>  
>      AVBufferRef *rpu_buf;       ///< 0 or 1 Dolby Vision RPUs.
> +    DOVIContext dovi_ctx;       ///< Dolby Vision decoding context
>  } HEVCContext;
>  
>  /**
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".