[FFmpeg-devel] [PATCH] avcodec/dovi - disable metadata compression by default

[Niklas Haas]

On Tue, 21 May 2024 04:03:43 +0000 Cosmin Stejerean via ffmpeg-devel <ffmpeg-devel at ffmpeg.org> wrote:
> From: Cosmin Stejerean <cosmin at cosmin.at>
> 
> not all clients support metadata compression, make this an option and off by
> default until we can verify output.
> 
> vdr_dm_metadata_changed = 0 case fails the DV verifier so force this to true
> for now until we can determine the correct output format for this case.

This approach seems reasonable to me, at least until we get those specs.

>
> 
> ---
>  libavcodec/dovi_rpu.h    | 5 +++++
>  libavcodec/dovi_rpuenc.c | 8 ++++++--
>  libavcodec/libaomenc.c   | 1 +
>  libavcodec/libsvtav1.c   | 1 +
>  libavcodec/libx265.c     | 1 +
>  5 files changed, 14 insertions(+), 2 deletions(-)
> 
> diff --git a/libavcodec/dovi_rpu.h b/libavcodec/dovi_rpu.h
> index 8ce0c88e9d..fca30804ae 100644
> --- a/libavcodec/dovi_rpu.h
> +++ b/libavcodec/dovi_rpu.h
> @@ -71,6 +71,11 @@ typedef struct DOVIContext {
>      AVDOVIDmData *ext_blocks;
>      int num_ext_blocks;
>  
> +    /**
> +     * Enable metadata compression in the output. Currently this is experimental.
> +     */
> +    int enable_compression;
> +
>      /**
>       * Private fields internal to dovi_rpu.c
>       */
> diff --git a/libavcodec/dovi_rpuenc.c b/libavcodec/dovi_rpuenc.c
> index 3c3e0f84c0..26ed25733a 100644
> --- a/libavcodec/dovi_rpuenc.c
> +++ b/libavcodec/dovi_rpuenc.c
> @@ -512,8 +512,12 @@ int ff_dovi_rpu_generate(DOVIContext *s, const AVDOVIMetadata *metadata,
>          }
>      }
>  
> -    vdr_dm_metadata_changed = !s->color || memcmp(s->color, color, sizeof(*color));
> -    use_prev_vdr_rpu = !memcmp(&s->vdr[vdr_rpu_id]->mapping, mapping, sizeof(*mapping));
> +    // the output when vdr_dm_metadata_changed is 0 fails the DV verifier
> +    // force it to 1 until we can get some samples or documentation on correct syntax
> +    vdr_dm_metadata_changed = 1; // !s->color || memcmp(s->color, color, sizeof(*color));
> +
> +    // not all clients support metadata compression
> +    use_prev_vdr_rpu = s->enable_compression && !memcmp(&s->vdr[vdr_rpu_id]->mapping, mapping, sizeof(*mapping));
>  
>      buffer_size = 12 /* vdr seq info */ + 5 /* CRC32 + terminator */;
>      buffer_size += num_ext_blocks_v1 * 13;
> diff --git a/libavcodec/libaomenc.c b/libavcodec/libaomenc.c
> index dec74ebecd..c6104f5522 100644
> --- a/libavcodec/libaomenc.c
> +++ b/libavcodec/libaomenc.c
> @@ -1489,6 +1489,7 @@ static const AVOption options[] = {
>      { "still-picture", "Encode in single frame mode (typically used for still AVIF images).", OFFSET(still_picture), AV_OPT_TYPE_BOOL, {.i64 = 0}, -1, 1, VE },
>      { "dolbyvision",     "Enable Dolby Vision RPU coding", OFFSET(dovi.enable), AV_OPT_TYPE_BOOL, {.i64 = FF_DOVI_AUTOMATIC }, -1, 1, VE, .unit = "dovi" },
>      {   "auto", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_DOVI_AUTOMATIC}, .flags = VE, .unit = "dovi" },
> +    { "dv_enable_compression", "Enable Dolby Vision metadata compression", OFFSET(dovi.enable_compression), AV_OPT_TYPE_BOOL, {.i64 = 0 }, 0, 1, VE },
>      { "enable-rect-partitions", "Enable rectangular partitions", OFFSET(enable_rect_partitions), AV_OPT_TYPE_BOOL, {.i64 = -1}, -1, 1, VE},
>      { "enable-1to4-partitions", "Enable 1:4/4:1 partitions",     OFFSET(enable_1to4_partitions), AV_OPT_TYPE_BOOL, {.i64 = -1}, -1, 1, VE},
>      { "enable-ab-partitions",   "Enable ab shape partitions",    OFFSET(enable_ab_partitions),   AV_OPT_TYPE_BOOL, {.i64 = -1}, -1, 1, VE},
> diff --git a/libavcodec/libsvtav1.c b/libavcodec/libsvtav1.c
> index 2fef8c8971..86bb6686dd 100644
> --- a/libavcodec/libsvtav1.c
> +++ b/libavcodec/libsvtav1.c
> @@ -733,6 +733,7 @@ static const AVOption options[] = {
>  
>      { "dolbyvision", "Enable Dolby Vision RPU coding", OFFSET(dovi.enable), AV_OPT_TYPE_BOOL, {.i64 = FF_DOVI_AUTOMATIC }, -1, 1, VE, .unit = "dovi" },
>      {   "auto", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_DOVI_AUTOMATIC}, .flags = VE, .unit = "dovi" },
> +    { "dv_enable_compression", "Enable Dolby Vision metadata compression", OFFSET(dovi.enable_compression), AV_OPT_TYPE_BOOL, {.i64 = 0 }, 0, 1, VE },
>  
>      {NULL},
>  };
> diff --git a/libavcodec/libx265.c b/libavcodec/libx265.c
> index ac1dbc4f97..2a79a5e6da 100644
> --- a/libavcodec/libx265.c
> +++ b/libavcodec/libx265.c
> @@ -953,6 +953,7 @@ static const AVOption options[] = {
>  #if X265_BUILD >= 167
>      { "dolbyvision", "Enable Dolby Vision RPU coding", OFFSET(dovi.enable), AV_OPT_TYPE_BOOL, {.i64 = FF_DOVI_AUTOMATIC }, -1, 1, VE, .unit = "dovi" },
>      {   "auto", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = FF_DOVI_AUTOMATIC}, .flags = VE, .unit = "dovi" },
> +    { "dv_enable_compression", "Enable Dolby Vision metadata compression", OFFSET(dovi.enable_compression), AV_OPT_TYPE_BOOL, {.i64 = 0 }, 0, 1, VE },
>  #endif

Setting up an extra AVClass here seems more hassle than it's worth, but
maybe we could at least hide these options behind a preprocessor
definition so that multiple files can reference them without blatantly
duplicating code?

>      { NULL }
>  };
> -- 
> 2.42.1
> 
> 
> _______________________________________________
> 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".