> On Feb 6, 2025, at 21:25, James Almer <jamr...@gmail.com> wrote:
>
> On 2/6/2025 12:49 AM, Zhao Zhili wrote:
>> From: Zhao Zhili <zhiliz...@tencent.com>
>> Only implementing what's needed for HEVC with alpha.
>> Signed-off-by: Zhao Zhili <zhiliz...@tencent.com>
>> ---
>> libavcodec/hevc/hevc.h | 5 ++
>> libavcodec/hevc/ps.c | 126 +++++++++++++++++++++++++++++------------
>> libavcodec/hevc/ps.h | 4 ++
>> 3 files changed, 98 insertions(+), 37 deletions(-)
>> diff --git a/libavcodec/hevc/hevc.h b/libavcodec/hevc/hevc.h
>> index b2229fda40..710786a89d 100644
>> --- a/libavcodec/hevc/hevc.h
>> +++ b/libavcodec/hevc/hevc.h
>> @@ -170,4 +170,9 @@ enum HEVCScalabilityMask {
>> HEVC_SCALABILITY_MASK_MAX = 0xFFFF,
>> };
>> +enum HEVCAuxId {
>> + HEVC_AUX_ALPHA = 1,
>> + HEVC_AUX_DEPTH = 2,
>> +};
>> +
>> #endif /* AVCODEC_HEVC_HEVC_H */
>> diff --git a/libavcodec/hevc/ps.c b/libavcodec/hevc/ps.c
>> index 861a6bb0a2..4f18cd72e4 100644
>> --- a/libavcodec/hevc/ps.c
>> +++ b/libavcodec/hevc/ps.c
>> @@ -460,14 +460,17 @@ static int decode_vps_ext(GetBitContext *gb,
>> AVCodecContext *avctx, HEVCVPS *vps
>> uint64_t layer1_id_included)
>> {
>> PTL ptl_dummy;
>> - uint8_t max_sub_layers[HEVC_MAX_LAYERS];
>> + uint8_t max_sub_layers[HEVC_MAX_LAYERS] = {1, 1};
>> + uint8_t dimension_id_len[16] = {0};
>> + uint8_t dimension_id[16] = {0};
>> + unsigned n;
>> - int splitting_flag, dimension_id_len, view_id_len, num_add_olss,
>> num_scalability_types,
>> + int splitting_flag, view_id_len, num_add_olss, num_scalability_types,
>> default_output_layer_idc, direct_dep_type_len, direct_dep_type,
>> sub_layers_max_present, sub_layer_flag_info_present_flag, nb_ptl;
>> unsigned non_vui_extension_length;
>> - if (vps->vps_max_layers == 1 || vps->vps_num_layer_sets == 1) {
>> + if (vps->vps_max_layers == 1) {
>> av_log(avctx, AV_LOG_VERBOSE, "Ignoring VPS extensions with a
>> single layer\n");
>> return 0;
>> }
>> @@ -520,7 +523,8 @@ static int decode_vps_ext(GetBitContext *gb,
>> AVCodecContext *avctx, HEVCVPS *vps
>> */
>> vps->nb_layers = 2;
>> - if (parse_ptl(gb, avctx, 0, &ptl_dummy, vps->vps_max_sub_layers) < 0)
>> + if (vps->vps_base_layer_internal_flag &&
>
> What's the point checking for this flag if this code will not be reached if
> it's false?
Here is only meant to follow the spec with little effort. When someone add
support for
vps_base_layer_internal_flag = 0, it’s clear what does the flag can affect.
>
>> + parse_ptl(gb, avctx, 0, &ptl_dummy, vps->vps_max_sub_layers) < 0)
>> return AVERROR_INVALIDDATA;
>> splitting_flag = get_bits1(gb);
>> @@ -529,20 +533,25 @@ static int decode_vps_ext(GetBitContext *gb,
>> AVCodecContext *avctx, HEVCVPS *vps
>> if (!num_scalability_types) {
>> av_log(avctx, AV_LOG_ERROR, "Missing scalability mask\n");
>> return AVERROR_INVALIDDATA;
>> - } else if (num_scalability_types > 1) {
>> - av_log(avctx, AV_LOG_ERROR, "Scalability number %d not supported\n",
>> - num_scalability_types);
>> - return AVERROR_PATCHWELCOME;
>> }
>> - if (!(vps->scalability_mask_flag & HEVC_SCALABILITY_MULTIVIEW)) {
>> + if (!(vps->scalability_mask_flag &
>> + (HEVC_SCALABILITY_MULTIVIEW | HEVC_SCALABILITY_AUXILIARY))) {
>> av_log(avctx, AV_LOG_ERROR, "Scalability type %d not supported\n",
>> 15 - ff_ctz(vps->scalability_mask_flag));
>> return AVERROR_PATCHWELCOME;
>> }
>> + // x265 specify MULTIVIEW when the stream really is alpha video only.
>> + if (num_scalability_types > 1)
>> + av_log(avctx, AV_LOG_WARNING, "Multiple scalability types
>> presented\n");
>> - if (!splitting_flag)
>> - dimension_id_len = get_bits(gb, 3) + 1;
>> + n = 0;
>> + for (int i = 0; i < num_scalability_types - splitting_flag; i++) {
>> + dimension_id_len[i] = get_bits(gb, 3) + 1;
>> + n += dimension_id_len[i];
>> + }
>> + if (splitting_flag)
>> + dimension_id_len[num_scalability_types - 1] = 5 - n;
>> if (get_bits1(gb)) { /* vps_nuh_layer_id_present_flag */
>> int layer_id_in_nuh = get_bits(gb, 6);
>> @@ -559,28 +568,57 @@ static int decode_vps_ext(GetBitContext *gb,
>> AVCodecContext *avctx, HEVCVPS *vps
>> }
>> if (!splitting_flag) {
>> - int view_idx = get_bits(gb, dimension_id_len);
>> - if (view_idx != 1) {
>> - av_log(avctx, AV_LOG_ERROR, "Unexpected ViewOrderIdx: %d\n",
>> view_idx);
>> + int index = 0;
>> +
>> + for (int i = 0; i < num_scalability_types; i++)
>> + dimension_id[i] = get_bits(gb, dimension_id_len[i]);
>> +
>> + if (vps->scalability_mask_flag & HEVC_SCALABILITY_MULTIVIEW)
>> + index++;
>> +
>> + /* AuxId 1 is alpha, 2 is depth. Only support alpha */
>> + if (vps->scalability_mask_flag & HEVC_SCALABILITY_AUXILIARY &&
>> + dimension_id[index] != HEVC_AUX_ALPHA) {
>> + av_log(avctx, AV_LOG_WARNING,
>> + "Unsupported dimension_id %d for
>> HEVC_SCALABILITY_AUXILIARY\n",
>> + dimension_id[index]);
>> return AVERROR_PATCHWELCOME;
>> }
>> }
>> view_id_len = get_bits(gb, 4);
>> - if (view_id_len)
>> - for (int i = 0; i < 2 /* NumViews */; i++)
>> + if (view_id_len) {
>> + n = (vps->scalability_mask_flag & HEVC_SCALABILITY_MULTIVIEW) ? 2 :
>> 1;
>> + for (int i = 0; i < n; i++)
>> vps->view_id[i] = get_bits(gb, view_id_len);
>> + }
>> - if (!get_bits1(gb) /* direct_dependency_flag */) {
>> - av_log(avctx, AV_LOG_WARNING, "Independent output layers not
>> supported\n");
>> - return AVERROR_PATCHWELCOME;
>> + /* direct_dependency_flag */
>> + vps->num_direct_ref_layers[1] = get_bits1(gb);
>> + if (!vps->num_direct_ref_layers[1]) {
>> + vps->num_add_layer_sets = get_ue_golomb(gb);
>> + if (vps->num_add_layer_sets > 1) {
>> + av_log(avctx, AV_LOG_WARNING,
>> + "Unsupported num_add_layer_sets: %d\n",
>> vps->num_add_layer_sets);
>> + return AVERROR_PATCHWELCOME;
>> + }
>> +
>> + if (vps->num_add_layer_sets) {
>> + /* highest_layer_idx_plus1 */
>> + if (!get_bits1(gb))
>> + return AVERROR_PATCHWELCOME;
>> + }
>> }
>> - vps->num_direct_ref_layers[1] = 1;
>> + vps->num_output_layer_sets = vps->vps_num_layer_sets +
>> vps->num_add_layer_sets;
>> + if (vps->num_output_layer_sets != 2)
>> + return AVERROR_INVALIDDATA;
>> sub_layers_max_present = get_bits1(gb); //
>> vps_sub_layers_max_minus1_present_flag
>> - for (int i = 0; i < vps->vps_max_layers; i++)
>> - max_sub_layers[i] = sub_layers_max_present ? get_bits(gb, 3) + 1 :
>> -
>> vps->vps_max_sub_layers;
>> + if (sub_layers_max_present) {
>> + for (int i = 0; i < vps->vps_max_layers; i++)
>> + max_sub_layers[i] = sub_layers_max_present ? get_bits(gb, 3) +
>> 1 :
>> +
>> vps->vps_max_sub_layers;
>> + }
>> if (get_bits1(gb) /* max_tid_ref_present_flag */)
>> skip_bits(gb, 3); // max_tid_il_ref_pics_plus1
>> @@ -613,18 +651,24 @@ static int decode_vps_ext(GetBitContext *gb,
>> AVCodecContext *avctx, HEVCVPS *vps
>> }
>> /* Consequence of established layer dependencies */
>> - if (layer1_id_included != ((1 << vps->layer_id_in_nuh[0]) |
>> + if (layer1_id_included &&
>> + layer1_id_included != ((1 << vps->layer_id_in_nuh[0]) |
>> (1 << vps->layer_id_in_nuh[1]))) {
>> - av_log(avctx, AV_LOG_ERROR, "Dependent layer not included in layer
>> ID?\n");
>> - return AVERROR_PATCHWELCOME;
>> + av_log(avctx, AV_LOG_ERROR,
>> + "Dependent layer not included in layer ID?\n");
>> + return AVERROR_PATCHWELCOME;
>> }
>> + if (!layer1_id_included)
>> + vps->ols[1] = 2;
>> + else
>> + vps->ols[1] = 3;
>> - vps->num_output_layer_sets = 2;
>> - vps->ols[1] = 3;
>> + if (vps->vps_num_layer_sets == 1 || default_output_layer_idc == 2)
>> + skip_bits1(gb);
>> for (int j = 0; j < av_popcount64(vps->ols[1]); j++) {
>> int ptl_idx = get_bits(gb, av_ceil_log2(nb_ptl));
>> - if (ptl_idx < 1 || ptl_idx >= nb_ptl) {
>> + if (ptl_idx >= nb_ptl) {
>> av_log(avctx, AV_LOG_ERROR, "Invalid PTL index: %d\n", ptl_idx);
>> return AVERROR_INVALIDDATA;
>> }
>> @@ -667,6 +711,8 @@ static int decode_vps_ext(GetBitContext *gb,
>> AVCodecContext *avctx, HEVCVPS *vps
>> vps->max_one_active_ref_layer = get_bits1(gb);
>> vps->poc_lsb_aligned = get_bits1(gb);
>> + if (!vps->num_direct_ref_layers[1])
>> + vps->poc_lsb_not_present = get_bits1(gb) << 1;
>
> Huh, so this was being checked in hls_slice_header() but never set before now?
It’s inferred to be 0 and initialized by memset to 0.
>
>> sub_layer_flag_info_present_flag = get_bits1(gb);
>> for (int j = 0; j < FFMAX(max_sub_layers[0], max_sub_layers[1]); j++) {
>> @@ -688,12 +734,14 @@ static int decode_vps_ext(GetBitContext *gb,
>> AVCodecContext *avctx, HEVCVPS *vps
>> return AVERROR_INVALIDDATA;
>> }
>> - skip_bits1(gb); /* direct_depenency_all_layers_flag */
>> - direct_dep_type = get_bits_long(gb, direct_dep_type_len);
>> - if (direct_dep_type > HEVC_DEP_TYPE_BOTH) {
>> - av_log(avctx, AV_LOG_WARNING, "Unsupported direct_dep_type: %d\n",
>> - direct_dep_type);
>> - return AVERROR_PATCHWELCOME;
>> + /* direct_dependency_all_layers_flag */
>> + if (get_bits1(gb) || vps->num_direct_ref_layers[1]) {
>> + direct_dep_type = get_bits_long(gb, direct_dep_type_len);
>> + if (direct_dep_type > HEVC_DEP_TYPE_BOTH) {
>> + av_log(avctx, AV_LOG_WARNING, "Unsupported direct_dep_type:
>> %d\n",
>> + direct_dep_type);
>> + return AVERROR_PATCHWELCOME;
>> + }
>> }
>> non_vui_extension_length = get_ue_golomb(gb);
>> @@ -741,8 +789,12 @@ int ff_hevc_decode_nal_vps(GetBitContext *gb,
>> AVCodecContext *avctx,
>> }
>> vps->vps_id = vps_id;
>> - if (get_bits(gb, 2) != 3) { // vps_reserved_three_2bits
>> - av_log(avctx, AV_LOG_ERROR, "vps_reserved_three_2bits is not
>> three\n");
>> + vps->vps_base_layer_internal_flag = get_bits1(gb);
>> + vps->vps_base_layer_available_flag = get_bits1(gb);
>> + if (!vps->vps_base_layer_internal_flag ||
>> !vps->vps_base_layer_available_flag) {
>> + av_log(avctx, AV_LOG_ERROR,
>> + "vps_base_layer_internal_flag or
>> vps_base_layer_available_flag not set\n");
>> + ret = AVERROR_PATCHWELCOME;
>> goto err;
>> }
>> diff --git a/libavcodec/hevc/ps.h b/libavcodec/hevc/ps.h
>> index d3465e3d27..f706b889e6 100644
>> --- a/libavcodec/hevc/ps.h
>> +++ b/libavcodec/hevc/ps.h
>> @@ -171,6 +171,9 @@ typedef struct RepFormat {
>> typedef struct HEVCVPS {
>> unsigned int vps_id;
>> + uint8_t vps_base_layer_internal_flag;
>> + uint8_t vps_base_layer_available_flag;
>
> These two can be variables local to ff_hevc_decode_nal_vps() instead of being
> in this struct.
It’s used by decode_vps_ext, and meant as a start point to add support for
these flags been zero
in the future.
>
>> +
>> uint8_t vps_temporal_id_nesting_flag;
>> int vps_max_layers;
>> int vps_max_sub_layers; ///< vps_max_temporal_layers_minus1 + 1
>> @@ -237,6 +240,7 @@ typedef struct HEVCVPS {
>> // NumDirectRefLayers[layer_idx]
>> uint8_t num_direct_ref_layers[HEVC_VPS_MAX_LAYERS];
>> + uint8_t num_add_layer_sets;
>> RepFormat rep_format;
>>
>
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
> To unsubscribe, visit link above, or email
> ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
>
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
