lavc/hevcdec: Parse DOVI RPU NALs

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.

Niklas Haas committed 4 years ago

Commit a538df71 parent 08d15184 files · +256 −9Browse files

Unified · Split

4 files

configure +1 −1
libavcodec/hevcdec.c +29 −8
libavcodec/hevcdec.h +2 −0
tests/ref/fate/hevc-dv-rpu +224 −0

configure+1−1

@@ -2826,7 +2826,7 @@ h264_decoder_suggest="error_resilience" · 2825 unmodified lines
2826	2826	hap_decoder_select="snappy texturedsp"
2827	2827	hap_encoder_deps="libsnappy"
2828	2828	hap_encoder_select="texturedspenc"
2829		−hevc_decoder_select="atsc_a53 bswapdsp cabac golomb hevcparse videodsp"
	2829	+hevc_decoder_select="atsc_a53 bswapdsp cabac dovi_rpu golomb hevcparse videodsp"
2830	2830	huffyuv_decoder_select="bswapdsp huffyuvdsp llviddsp"
2831	2831	huffyuv_encoder_select="bswapdsp huffman huffyuvencdsp llvidencdsp"
2832	2832	hymt_decoder_select="huffyuv_decoder"

libavcodec/hevcdec.c+29−8

@@ -2723,6 +2723,7 @@ error: · 2722 unmodified lines
2723	2723	static int set_side_data(HEVCContext *s)
2724	2724	{
2725	2725	AVFrame *out = s->ref->frame;
	2726	+ int ret;
2726	2727
2727	2728	if (s->sei.frame_packing.present &&
2728	2729	s->sei.frame_packing.arrangement_type >= 3 &&
@@ -2977,6 +2978,9 @@ static int set_side_data(HEVCContext *s) · 248 unmodified lines
2977	2978	s->rpu_buf = NULL;
2978	2979	}
2979	2980
	2981	+ if ((ret = ff_dovi_attach_side_data(&s->dovi_ctx, out)) < 0)
	2982	+ return ret;
	2983	+
2980	2984	return 0;
2981	2985	}
2982	2986
@@ -3308,16 +3312,23 @@ static int decode_nal_units(HEVCContext s, const uint8_t buf, int length) · 325 unmodified lines
3308	3312	if (s->pkt.nb_nals > 1 && s->pkt.nals[s->pkt.nb_nals - 1].type == HEVC_NAL_UNSPEC62 &&
3309	3313	s->pkt.nals[s->pkt.nb_nals - 1].size > 2 && !s->pkt.nals[s->pkt.nb_nals - 1].nuh_layer_id
3310	3314	&& !s->pkt.nals[s->pkt.nb_nals - 1].temporal_id) {
	3315	+ H2645NAL *nal = &s->pkt.nals[s->pkt.nb_nals - 1];
3311	3316	if (s->rpu_buf) {
3312	3317	av_buffer_unref(&s->rpu_buf);
3313	3318	av_log(s->avctx, AV_LOG_WARNING, "Multiple Dolby Vision RPUs found in one AU. Skipping previous.\n");
3314	3319	}
3315	3320
3316		− s->rpu_buf = av_buffer_alloc(s->pkt.nals[s->pkt.nb_nals - 1].raw_size - 2);
	3321	+ s->rpu_buf = av_buffer_alloc(nal->raw_size - 2);
3317	3322	if (!s->rpu_buf)
3318	3323	return AVERROR(ENOMEM);
	3324	+ memcpy(s->rpu_buf->data, nal->raw_data + 2, nal->raw_size - 2);
3319	3325
3320		− 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);
	3326	+ ret = ff_dovi_rpu_parse(&s->dovi_ctx, nal->data + 2, nal->size - 2);
	3327	+ if (ret < 0) {
	3328	+ av_buffer_unref(&s->rpu_buf);
	3329	+ av_log(s->avctx, AV_LOG_WARNING, "Error parsing DOVI NAL unit.\n");
	3330	+ /* ignore */
	3331	+ }
3321	3332	}
3322	3333
3323	3334	/* decode the NAL units */
@@ -3450,8 +3461,8 @@ static int hevc_decode_frame(AVCodecContext avctx, void data, int *got_output, · 126 unmodified lines
3450	3461	AVPacket *avpkt)
3451	3462	{
3452	3463	int ret;
3453		− size_t new_extradata_size;
3454		− uint8_t *new_extradata;
	3464	+ uint8_t *sd;
	3465	+ size_t sd_size;
3455	3466	HEVCContext *s = avctx->priv_data;
3456	3467
3457	3468	if (!avpkt->size) {
@@ -3463,14 +3474,17 @@ static int hevc_decode_frame(AVCodecContext avctx, void data, int *got_output, · 5 unmodified lines
3463	3474	return 0;
3464	3475	}
3465	3476
3466		− new_extradata = av_packet_get_side_data(avpkt, AV_PKT_DATA_NEW_EXTRADATA,
3467		− &new_extradata_size);
3468		− if (new_extradata && new_extradata_size > 0) {
3469		− ret = hevc_decode_extradata(s, new_extradata, new_extradata_size, 0);
	3477	+ sd = av_packet_get_side_data(avpkt, AV_PKT_DATA_NEW_EXTRADATA, &sd_size);
	3478	+ if (sd && sd_size > 0) {
	3479	+ ret = hevc_decode_extradata(s, sd, sd_size, 0);
3470	3480	if (ret < 0)
3471	3481	return ret;
3472	3482	}
3473	3483
	3484	+ sd = av_packet_get_side_data(avpkt, AV_PKT_DATA_DOVI_CONF, &sd_size);
	3485	+ if (sd && sd_size > 0)
	3486	+ ff_dovi_update_cfg(&s->dovi_ctx, (AVDOVIDecoderConfigurationRecord *) sd);
	3487	+
3474	3488	s->ref = NULL;
3475	3489	ret = decode_nal_units(s, avpkt->data, avpkt->size);
3476	3490	if (ret < 0)
@@ -3563,6 +3577,7 @@ static av_cold int hevc_decode_free(AVCodecContext *avctx) · 86 unmodified lines
3563	3577
3564	3578	pic_arrays_free(s);
3565	3579
	3580	+ ff_dovi_ctx_unref(&s->dovi_ctx);
3566	3581	av_buffer_unref(&s->rpu_buf);
3567	3582
3568	3583	av_freep(&s->md5_ctx);
@@ -3647,6 +3662,7 @@ static av_cold int hevc_init_context(AVCodecContext *avctx) · 78 unmodified lines
3647	3662
3648	3663	ff_bswapdsp_init(&s->bdsp);
3649	3664
	3665	+ s->dovi_ctx.logctx = avctx;
3650	3666	s->context_initialized = 1;
3651	3667	s->eos = 0;
3652	3668
@@ -3755,6 +3771,10 @@ static int hevc_update_thread_context(AVCodecContext *dst, · 102 unmodified lines
3755	3771	if (ret < 0)
3756	3772	return ret;
3757	3773
	3774	+ ret = ff_dovi_ctx_replace(&s->dovi_ctx, &s0->dovi_ctx);
	3775	+ if (ret < 0)
	3776	+ return ret;
	3777	+
3758	3778	s->sei.frame_packing = s0->sei.frame_packing;
3759	3779	s->sei.display_orientation = s0->sei.display_orientation;
3760	3780	s->sei.mastering_display = s0->sei.mastering_display;
@@ -3811,6 +3831,7 @@ static void hevc_decode_flush(AVCodecContext *avctx) · 50 unmodified lines
3811	3831	HEVCContext *s = avctx->priv_data;
3812	3832	ff_hevc_flush_dpb(s);
3813	3833	ff_hevc_reset_sei(&s->sei);
	3834	+ ff_dovi_ctx_flush(&s->dovi_ctx);
3814	3835	av_buffer_unref(&s->rpu_buf);
3815	3836	s->max_ra = INT_MAX;
3816	3837	s->eos = 1;

libavcodec/hevcdec.h+2−0

@@ -32,6 +32,7 @@ · 31 unmodified lines
32	32	#include "avcodec.h"
33	33	#include "bswapdsp.h"
34	34	#include "cabac.h"
	35	+#include "dovi_rpu.h"
35	36	#include "get_bits.h"
36	37	#include "hevcpred.h"
37	38	#include "h2645_parse.h"
@@ -574,6 +575,7 @@ typedef struct HEVCContext { · 536 unmodified lines
574	575	int nuh_layer_id;
575	576
576	577	AVBufferRef *rpu_buf; ///< 0 or 1 Dolby Vision RPUs.
	578	+ DOVIContext dovi_ctx; ///< Dolby Vision decoding context
577	579	} HEVCContext;
578	580
579	581	/**

tests/ref/fate/hevc-dv-rpu+224−0

@@ -5,9 +5,233 @@ side_data_type=H.26[45] User Data Unregistered SEI message · 4 unmodified lines
5	5	[SIDE_DATA]
6	6	side_data_type=Dolby Vision RPU Data
7	7	[/SIDE_DATA]
	8	+[SIDE_DATA]
	9	+side_data_type=Dolby Vision Metadata
	10	+rpu_type=2
	11	+rpu_format=18
	12	+vdr_rpu_profile=1
	13	+vdr_rpu_level=0
	14	+chroma_resampling_explicit_filter_flag=0
	15	+coef_data_type=0
	16	+coef_log2_denom=23
	17	+vdr_rpu_normalized_idc=1
	18	+bl_video_full_range_flag=0
	19	+bl_bit_depth=10
	20	+el_bit_depth=10
	21	+vdr_bit_depth=12
	22	+spatial_resampling_filter_flag=0
	23	+el_spatial_resampling_filter_flag=0
	24	+disable_residual_flag=1
	25	+vdr_rpu_id=0
	26	+mapping_color_space=0
	27	+mapping_chroma_format_idc=0
	28	+nlq_method_idc=-1
	29	+nlq_method_idc_name=none
	30	+num_x_partitions=1
	31	+num_y_partitions=1
	32	+[COMPONENT]
	33	+pivots=63 132 362 618 874 911 927 935 942
	34	+[SECTION]
	35	+mapping_idc=0
	36	+mapping_idc_name=polynomial
	37	+poly_order=2
	38	+poly_coef=-409680 16721463 -20276640
	39	+[/SECTION]
	40	+[SECTION]
	41	+mapping_idc=0
	42	+mapping_idc_name=polynomial
	43	+poly_order=2
	44	+poly_coef=-119056 13575212 -12867889
	45	+[/SECTION]
	46	+[SECTION]
	47	+mapping_idc=0
	48	+mapping_idc_name=polynomial
	49	+poly_order=2
	50	+poly_coef=1317527 5338528 -948122
	51	+[/SECTION]
	52	+[SECTION]
	53	+mapping_idc=0
	54	+mapping_idc_name=polynomial
	55	+poly_order=2
	56	+poly_coef=2119979 2065496 2288524
	57	+[/SECTION]
	58	+[SECTION]
	59	+mapping_idc=0
	60	+mapping_idc_name=polynomial
	61	+poly_order=2
	62	+poly_coef=7982780 -11367226 9973944
	63	+[/SECTION]
	64	+[SECTION]
	65	+mapping_idc=0
	66	+mapping_idc_name=polynomial
	67	+poly_order=2
	68	+poly_coef=53792084 -114243184 67724328
	69	+[/SECTION]
	70	+[SECTION]
	71	+mapping_idc=0
	72	+mapping_idc_name=polynomial
	73	+poly_order=2
	74	+poly_coef=112973872 -244837568 139764480
	75	+[/SECTION]
	76	+[SECTION]
	77	+mapping_idc=0
	78	+mapping_idc_name=polynomial
	79	+poly_order=2
	80	+poly_coef=236828416 -518849056 291306944
	81	+[/SECTION]
	82	+[/COMPONENT]
	83	+[COMPONENT]
	84	+pivots=0 1023
	85	+[SECTION]
	86	+mapping_idc=1
	87	+mapping_idc_name=mmr
	88	+mmr_order=3
	89	+mmr_constant=-4200453
	90	+mmr_coef=9094176 31944476 739652 -27103344 -3431599 -11015088 22758042 -2028643 -30021218 -906885 26272992 7291404 16488745 -78087176 -1487777 12496543 762460 -4281948 -5768035 -7843163 103636960
	91	+[/SECTION]
	92	+[/COMPONENT]
	93	+[COMPONENT]
	94	+pivots=0 1023
	95	+[SECTION]
	96	+mapping_idc=1
	97	+mapping_idc_name=mmr
	98	+mmr_order=3
	99	+mmr_constant=-10387889
	100	+mmr_coef=29604202 3214133 46206184 -8564340 -53383996 1628407 5426045 -21634202 -5251953 -50812292 19221972 76726656 -425892 -35041240 5917361 2863974 25030554 -14404292 -41230120 1229046 53575140
	101	+[/SECTION]
	102	+[/COMPONENT]
	103	+dm_metadata_id=0
	104	+scene_refresh_flag=1
	105	+ycc_to_rgb_matrix=9574/8192 0/8192 13802/8192 9574/8192 -1540/8192 -5348/8192 9574/8192 17610/8192 0/8192
	106	+ycc_to_rgb_offset=16777216/268435456 134217728/268435456 134217728/268435456
	107	+rgb_to_lms_matrix=7222/16384 8771/16384 390/16384 2654/16384 12430/16384 1300/16384 0/16384 422/16384 15962/16384
	108	+signal_eotf=65535
	109	+signal_eotf_param0=0
	110	+signal_eotf_param1=0
	111	+signal_eotf_param2=0
	112	+signal_bit_depth=12
	113	+signal_color_space=0
	114	+signal_chroma_format=0
	115	+signal_full_range_flag=1
	116	+source_min_pq=0
	117	+source_max_pq=3079
	118	+source_diagonal=42
	119	+[/SIDE_DATA]
8	120	[/FRAME]
9	121	[FRAME]
10	122	[SIDE_DATA]
11	123	side_data_type=Dolby Vision RPU Data
12	124	[/SIDE_DATA]
	125	+[SIDE_DATA]
	126	+side_data_type=Dolby Vision Metadata
	127	+rpu_type=2
	128	+rpu_format=18
	129	+vdr_rpu_profile=1
	130	+vdr_rpu_level=0
	131	+chroma_resampling_explicit_filter_flag=0
	132	+coef_data_type=0
	133	+coef_log2_denom=23
	134	+vdr_rpu_normalized_idc=1
	135	+bl_video_full_range_flag=0
	136	+bl_bit_depth=10
	137	+el_bit_depth=10
	138	+vdr_bit_depth=12
	139	+spatial_resampling_filter_flag=0
	140	+el_spatial_resampling_filter_flag=0
	141	+disable_residual_flag=1
	142	+vdr_rpu_id=0
	143	+mapping_color_space=0
	144	+mapping_chroma_format_idc=0
	145	+nlq_method_idc=-1
	146	+nlq_method_idc_name=none
	147	+num_x_partitions=1
	148	+num_y_partitions=1
	149	+[COMPONENT]
	150	+pivots=63 132 362 618 874 911 927 935 942
	151	+[SECTION]
	152	+mapping_idc=0
	153	+mapping_idc_name=polynomial
	154	+poly_order=2
	155	+poly_coef=-409680 16721463 -20276640
	156	+[/SECTION]
	157	+[SECTION]
	158	+mapping_idc=0
	159	+mapping_idc_name=polynomial
	160	+poly_order=2
	161	+poly_coef=-119056 13575212 -12867889
	162	+[/SECTION]
	163	+[SECTION]
	164	+mapping_idc=0
	165	+mapping_idc_name=polynomial
	166	+poly_order=2
	167	+poly_coef=1317527 5338528 -948122
	168	+[/SECTION]
	169	+[SECTION]
	170	+mapping_idc=0
	171	+mapping_idc_name=polynomial
	172	+poly_order=2
	173	+poly_coef=2119979 2065496 2288524
	174	+[/SECTION]
	175	+[SECTION]
	176	+mapping_idc=0
	177	+mapping_idc_name=polynomial
	178	+poly_order=2
	179	+poly_coef=7982780 -11367226 9973944
	180	+[/SECTION]
	181	+[SECTION]
	182	+mapping_idc=0
	183	+mapping_idc_name=polynomial
	184	+poly_order=2
	185	+poly_coef=53792084 -114243184 67724328
	186	+[/SECTION]
	187	+[SECTION]
	188	+mapping_idc=0
	189	+mapping_idc_name=polynomial
	190	+poly_order=2
	191	+poly_coef=112973872 -244837568 139764480
	192	+[/SECTION]
	193	+[SECTION]
	194	+mapping_idc=0
	195	+mapping_idc_name=polynomial
	196	+poly_order=2
	197	+poly_coef=236828416 -518849056 291306944
	198	+[/SECTION]
	199	+[/COMPONENT]
	200	+[COMPONENT]
	201	+pivots=0 1023
	202	+[SECTION]
	203	+mapping_idc=1
	204	+mapping_idc_name=mmr
	205	+mmr_order=3
	206	+mmr_constant=-4200453
	207	+mmr_coef=9094176 31944476 739652 -27103344 -3431599 -11015088 22758042 -2028643 -30021218 -906885 26272992 7291404 16488745 -78087176 -1487777 12496543 762460 -4281948 -5768035 -7843163 103636960
	208	+[/SECTION]
	209	+[/COMPONENT]
	210	+[COMPONENT]
	211	+pivots=0 1023
	212	+[SECTION]
	213	+mapping_idc=1
	214	+mapping_idc_name=mmr
	215	+mmr_order=3
	216	+mmr_constant=-10387889
	217	+mmr_coef=29604202 3214133 46206184 -8564340 -53383996 1628407 5426045 -21634202 -5251953 -50812292 19221972 76726656 -425892 -35041240 5917361 2863974 25030554 -14404292 -41230120 1229046 53575140
	218	+[/SECTION]
	219	+[/COMPONENT]
	220	+dm_metadata_id=0
	221	+scene_refresh_flag=1
	222	+ycc_to_rgb_matrix=9574/8192 0/8192 13802/8192 9574/8192 -1540/8192 -5348/8192 9574/8192 17610/8192 0/8192
	223	+ycc_to_rgb_offset=16777216/268435456 134217728/268435456 134217728/268435456
	224	+rgb_to_lms_matrix=7222/16384 8771/16384 390/16384 2654/16384 12430/16384 1300/16384 0/16384 422/16384 15962/16384
	225	+signal_eotf=65535
	226	+signal_eotf_param0=0
	227	+signal_eotf_param1=0
	228	+signal_eotf_param2=0
	229	+signal_bit_depth=12
	230	+signal_color_space=0
	231	+signal_chroma_format=0
	232	+signal_full_range_flag=1
	233	+source_min_pq=0
	234	+source_max_pq=3079
	235	+source_diagonal=42
	236	+[/SIDE_DATA]
13	237	[/FRAME]

lavc/hevcdec: Parse DOVI RPU NALs

Keyboard shortcuts

General

Issues and pull requests

Issues, pull requests and Inbox

Repository code