X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fh264.h;h=f551d79209a56456e9247048336bcf062019662e;hb=89ae244e78a8d636fbb8427a2da052b04cf710bd;hp=ba4ee39ab40137120fff38ff502ea38811a7ec35;hpb=c377e04d8aa74d030672e9a4788a700b0695fc14;p=ffmpeg diff --git a/libavcodec/h264.h b/libavcodec/h264.h index ba4ee39ab40..f551d79209a 100644 --- a/libavcodec/h264.h +++ b/libavcodec/h264.h @@ -28,14 +28,20 @@ #ifndef AVCODEC_H264_H #define AVCODEC_H264_H +#include "libavutil/buffer.h" #include "libavutil/intreadwrite.h" +#include "libavutil/thread.h" #include "cabac.h" #include "error_resilience.h" #include "get_bits.h" +#include "h264_parse.h" +#include "h264_sei.h" +#include "h2645_parse.h" #include "h264chroma.h" #include "h264dsp.h" #include "h264pred.h" #include "h264qpel.h" +#include "internal.h" #include "mpegutils.h" #include "parser.h" #include "qpeldsp.h" @@ -88,11 +94,11 @@ #define FIELD_OR_MBAFF_PICTURE(h) (FRAME_MBAFF(h) || FIELD_PICTURE(h)) #ifndef CABAC -#define CABAC(h) h->pps.cabac +#define CABAC(h) h->ps.pps->cabac #endif -#define CHROMA422(h) (h->sps.chroma_format_idc == 2) -#define CHROMA444(h) (h->sps.chroma_format_idc == 3) +#define CHROMA422(h) (h->ps.sps->chroma_format_idc == 2) +#define CHROMA444(h) (h->ps.sps->chroma_format_idc == 3) #define EXTENDED_SAR 255 @@ -122,33 +128,6 @@ enum { NAL_FF_IGNORE = 0xff0f001, }; -/** - * SEI message types - */ -typedef enum { - SEI_TYPE_BUFFERING_PERIOD = 0, ///< buffering period (H.264, D.1.1) - SEI_TYPE_PIC_TIMING = 1, ///< picture timing - SEI_TYPE_USER_DATA_UNREGISTERED = 5, ///< unregistered user data - SEI_TYPE_RECOVERY_POINT = 6, ///< recovery point (frame # to decoder sync) - SEI_TYPE_FRAME_PACKING = 45, ///< frame packing arrangement - SEI_TYPE_DISPLAY_ORIENTATION = 47, ///< display orientation -} SEI_Type; - -/** - * pic_struct in picture timing SEI message - */ -typedef enum { - SEI_PIC_STRUCT_FRAME = 0, ///< 0: %frame - SEI_PIC_STRUCT_TOP_FIELD = 1, ///< 1: top field - SEI_PIC_STRUCT_BOTTOM_FIELD = 2, ///< 2: bottom field - SEI_PIC_STRUCT_TOP_BOTTOM = 3, ///< 3: top field, bottom field, in that order - SEI_PIC_STRUCT_BOTTOM_TOP = 4, ///< 4: bottom field, top field, in that order - SEI_PIC_STRUCT_TOP_BOTTOM_TOP = 5, ///< 5: top field, bottom field, top field repeated, in that order - SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM = 6, ///< 6: bottom field, top field, bottom field repeated, in that order - SEI_PIC_STRUCT_FRAME_DOUBLING = 7, ///< 7: %frame doubling - SEI_PIC_STRUCT_FRAME_TRIPLING = 8 ///< 8: %frame tripling -} SEI_PicStructType; - /** * Sequence parameter set */ @@ -209,7 +188,6 @@ typedef struct SPS { int bit_depth_chroma; ///< bit_depth_chroma_minus8 + 8 int residual_color_transform_flag; ///< residual_colour_transform_flag int constraint_set_flags; ///< constraint_set[0-3]_flag - int new; ///< flag to keep track if the decoder context needs re-init due to changed SPS } SPS; /** @@ -235,8 +213,23 @@ typedef struct PPS { uint8_t scaling_matrix8[6][64]; uint8_t chroma_qp_table[2][64]; ///< pre-scaled (with chroma_qp_index_offset) version of qp_table int chroma_qp_diff; + + uint32_t dequant4_buffer[6][QP_MAX_NUM + 1][16]; + uint32_t dequant8_buffer[6][QP_MAX_NUM + 1][64]; + uint32_t(*dequant4_coeff[6])[16]; + uint32_t(*dequant8_coeff[6])[64]; } PPS; +typedef struct H264ParamSets { + AVBufferRef *sps_list[MAX_SPS_COUNT]; + AVBufferRef *pps_list[MAX_PPS_COUNT]; + + /* currently active parameters sets */ + const PPS *pps; + // FIXME this should properly be const + SPS *sps; +} H264ParamSets; + /** * Memory management control operation opcode. */ @@ -260,7 +253,7 @@ typedef struct MMCO { } MMCO; typedef struct H264Picture { - struct AVFrame f; + AVFrame *f; ThreadFrame tf; AVBufferRef *qscale_table_buf; @@ -291,14 +284,25 @@ typedef struct H264Picture { int mbaff; ///< 1 -> MBAFF frame 0-> not MBAFF int field_picture; ///< whether or not picture was encoded in separate fields - int needs_realloc; ///< picture needs to be reallocated (eg due to a frame size change) int reference; int recovered; ///< picture at IDR or recovery point + recovery count } H264Picture; +typedef struct H264Ref { + uint8_t *data[3]; + int linesize[3]; + + int reference; + int poc; + int pic_id; + + H264Picture *parent; +} H264Ref; + typedef struct H264SliceContext { struct H264Context *h264; GetBitContext gb; + ERContext er; int slice_num; int slice_type; @@ -315,17 +319,7 @@ typedef struct H264SliceContext { int slice_alpha_c0_offset; int slice_beta_offset; - // Weighted pred stuff - int use_weight; - int use_weight_chroma; - int luma_log2_weight_denom; - int chroma_log2_weight_denom; - int luma_weight_flag[2]; ///< 7.4.3.2 luma_weight_lX_flag - int chroma_weight_flag[2]; ///< 7.4.3.2 chroma_weight_lX_flag - // The following 2 can be changed to int8_t but that causes 10cpu cycles speedloss - int luma_weight[48][2][2]; - int chroma_weight[48][2][2][2]; - int implicit_weight[48][48][2]; + H264PredWeightTable pwt; int prev_mb_skipped; int next_mb_skipped; @@ -354,6 +348,7 @@ typedef struct H264SliceContext { unsigned int topright_samples_available; unsigned int left_samples_available; + ptrdiff_t linesize, uvlinesize; ptrdiff_t mb_linesize; ///< may be equal to s->linesize or s->linesize * 2, for mbaff ptrdiff_t mb_uvlinesize; @@ -361,6 +356,8 @@ typedef struct H264SliceContext { int mb_xy; int resync_mb_x; int resync_mb_y; + // index of the first MB of the next slice + int next_slice_idx; int mb_skip_run; int is_complex; @@ -392,12 +389,13 @@ typedef struct H264SliceContext { */ unsigned int ref_count[2]; ///< counts frames or fields, depending on current mb mode unsigned int list_count; - H264Picture ref_list[2][48]; /**< 0..15: frame refs, 16..47: mbaff field refs. + H264Ref ref_list[2][48]; /**< 0..15: frame refs, 16..47: mbaff field refs. * Reordered version of default_ref_list * according to picture reordering in slice header */ int ref2frm[MAX_SLICES][2][64]; ///< reference to frame number lists, used in the loop filter, the first 2 are for -2,-1 const uint8_t *intra_pcm_ptr; + int16_t *dc_val_base; uint8_t *bipred_scratchpad; uint8_t *edge_emu_buffer; @@ -437,32 +435,37 @@ typedef struct H264SliceContext { CABACContext cabac; uint8_t cabac_state[1024]; int cabac_init_idc; + + // rbsp buffer used for this slice + uint8_t *rbsp_buffer; + unsigned int rbsp_buffer_size; } H264SliceContext; /** * H264Context */ typedef struct H264Context { + const AVClass *class; AVCodecContext *avctx; VideoDSPContext vdsp; H264DSPContext h264dsp; H264ChromaContext h264chroma; H264QpelContext h264qpel; GetBitContext gb; - ERContext er; - H264Picture *DPB; + H264Picture DPB[H264_MAX_PICTURE_COUNT]; H264Picture *cur_pic_ptr; H264Picture cur_pic; H264SliceContext *slice_ctx; int nb_slice_ctx; + H2645Packet pkt; + int pixel_shift; ///< 0 for 8-bit H264, 1 for high-bit-depth H264 /* coded dimensions -- 16 * mb w/h */ int width, height; - ptrdiff_t linesize, uvlinesize; int chroma_x_shift, chroma_y_shift; int droppable; @@ -491,14 +494,6 @@ typedef struct H264Context { uint32_t *mb2br_xy; int b_stride; // FIXME use s->b4_stride - SPS sps; ///< current sps - PPS pps; ///< current pps - - uint32_t dequant4_buffer[6][QP_MAX_NUM + 1][16]; // FIXME should these be moved down? - uint32_t dequant8_buffer[6][QP_MAX_NUM + 1][64]; - uint32_t(*dequant4_coeff[6])[16]; - uint32_t(*dequant8_coeff[6])[64]; - uint16_t *slice_table; ///< slice_table_base + 2*mb_stride + 1 // interlacing specific flags @@ -529,8 +524,6 @@ typedef struct H264Context { const uint8_t *field_scan8x8_q0; const uint8_t *field_scan8x8_cavlc_q0; - int x264_build; - int mb_y; int mb_height, mb_width; int mb_stride; @@ -541,8 +534,6 @@ typedef struct H264Context { int nal_ref_idc; int nal_unit_type; - uint8_t *rbsp_buffer; - unsigned int rbsp_buffer_size; /** * Used to parse AVC variant of h264 @@ -553,24 +544,11 @@ typedef struct H264Context { int bit_depth_luma; ///< luma bit depth from sps to detect changes int chroma_format_idc; ///< chroma format from sps to detect changes - SPS *sps_buffers[MAX_SPS_COUNT]; - PPS *pps_buffers[MAX_PPS_COUNT]; - - int dequant_coeff_pps; ///< reinit tables when pps changes + H264ParamSets ps; uint16_t *slice_table_base; - // POC stuff - int poc_lsb; - int poc_msb; - int delta_poc_bottom; - int delta_poc[2]; - int frame_num; - int prev_poc_msb; ///< poc_msb of the last reference pic for POC type 0 - int prev_poc_lsb; ///< poc_lsb of the last reference pic for POC type 0 - int frame_num_offset; ///< for POC type 2 - int prev_frame_num_offset; ///< for POC type 2 - int prev_frame_num; ///< frame_num of the last pic for POC type 1/2 + H264POCContext poc; /** * frame_num for frames or 2 * frame_num + 1 for field pics. @@ -582,13 +560,11 @@ typedef struct H264Context { */ int max_pic_num; - H264Picture default_ref_list[2][32]; ///< base reference list for all slices of a coded picture H264Picture *short_ref[32]; H264Picture *long_ref[32]; H264Picture *delayed_pic[MAX_DELAYED_PIC_COUNT + 2]; // FIXME size? int last_pocs[MAX_DELAYED_PIC_COUNT]; H264Picture *next_output_pic; - int outputed_poc; int next_outputed_poc; /** @@ -605,8 +581,6 @@ typedef struct H264Context { * @name Members for slice based multithreading * @{ */ - struct H264Context *thread_context[H264_MAX_THREADS]; - /** * current slice number, used to initalize slice_num of each thread/context */ @@ -630,14 +604,8 @@ typedef struct H264Context { enum AVPictureType pict_type; - int last_slice_type; /** @} */ - /** - * pic_struct in picture timing SEI message - */ - SEI_PicStructType sei_pic_struct; - /** * Complement sei_pic_struct * SEI_PIC_STRUCT_TOP_BOTTOM and SEI_PIC_STRUCT_BOTTOM_TOP indicate interlaced frames. @@ -646,47 +614,6 @@ typedef struct H264Context { */ int prev_interlaced_frame; - /** - * frame_packing_arrangment SEI message - */ - int sei_frame_packing_present; - int frame_packing_arrangement_type; - int content_interpretation_type; - int quincunx_subsampling; - - /** - * display orientation SEI message - */ - int sei_display_orientation_present; - int sei_anticlockwise_rotation; - int sei_hflip, sei_vflip; - - /** - * Bit set of clock types for fields/frames in picture timing SEI message. - * For each found ct_type, appropriate bit is set (e.g., bit 1 for - * interlaced). - */ - int sei_ct_type; - - /** - * dpb_output_delay in picture timing SEI message, see H.264 C.2.2 - */ - int sei_dpb_output_delay; - - /** - * cpb_removal_delay in picture timing SEI message, see H.264 C.1.2 - */ - int sei_cpb_removal_delay; - - /** - * recovery_frame_cnt from SEI message - * - * Set to -1 if no recovery point SEI message found or to number of frames - * before playback synchronizes. Frames having recovery point are key - * frames. - */ - int sei_recovery_frame_cnt; - /** * recovery_frame is the frame_num at which the next frame should * be fully constructed. @@ -708,13 +635,15 @@ typedef struct H264Context { int frame_recovered; ///< Initial frame has been completely recovered + /* for frame threading, this is set to 1 + * after finish_setup() has been called, so we cannot modify + * some context properties (which are supposed to stay constant between + * slices) anymore */ + int setup_finished; - // Timestamp stuff - int sei_buffering_period_present; ///< Buffering period SEI flag - int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs + int enable_er; - int cur_chroma_format_idc; - int16_t *dc_val_base; + H264SEIContext sei; AVBufferPool *qscale_table_pool; AVBufferPool *mb_type_pool; @@ -726,39 +655,19 @@ typedef struct H264Context { qpel_mc_func (*qpel_avg)[16]; } H264Context; -extern const uint8_t ff_h264_chroma_qp[3][QP_MAX_NUM + 1]; ///< One chroma qp table for each supported bit depth (8, 9, 10). extern const uint16_t ff_h264_mb_sizes[4]; -/** - * Decode SEI - */ -int ff_h264_decode_sei(H264Context *h); - /** * Decode SPS */ -int ff_h264_decode_seq_parameter_set(H264Context *h); - -/** - * compute profile from sps - */ -int ff_h264_get_profile(SPS *sps); +int ff_h264_decode_seq_parameter_set(GetBitContext *gb, AVCodecContext *avctx, + H264ParamSets *ps); /** * Decode PPS */ -int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length); - -/** - * Decode a network abstraction layer unit. - * @param consumed is the number of bytes used as input - * @param length is the length of the array - * @param dst_length is the number of decoded bytes FIXME here - * or a decode rbsp tailing? - * @return decoded bytes, might be src+1 if no escapes - */ -const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src, - int *dst_length, int *consumed, int length); +int ff_h264_decode_picture_parameter_set(GetBitContext *gb, AVCodecContext *avctx, + H264ParamSets *ps, int bit_length); /** * Free any data that may have been allocated in the H264 context @@ -777,13 +686,8 @@ int ff_h264_get_slice_type(const H264SliceContext *sl); */ int ff_h264_alloc_tables(H264Context *h); -/** - * Fill the default_ref_list. - */ -int ff_h264_fill_default_ref_list(H264Context *h, H264SliceContext *sl); - -int ff_h264_decode_ref_pic_list_reordering(H264Context *h, H264SliceContext *sl); -void ff_h264_fill_mbaff_ref_list(H264Context *h, H264SliceContext *sl); +int ff_h264_decode_ref_pic_list_reordering(const H264Context *h, H264SliceContext *sl); +void ff_h264_fill_mbaff_ref_list(H264SliceContext *sl); void ff_h264_remove_all_refs(H264Context *h); /** @@ -796,21 +700,7 @@ int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb, int ff_generate_sliding_window_mmcos(H264Context *h, int first_slice); -/** - * Check if the top & left blocks are available if needed & change the - * dc mode so it only uses the available blocks. - */ -int ff_h264_check_intra4x4_pred_mode(const H264Context *h, H264SliceContext *sl); - -/** - * Check if the top & left blocks are available if needed & change the - * dc mode so it only uses the available blocks. - */ -int ff_h264_check_intra_pred_mode(const H264Context *h, H264SliceContext *sl, - int mode, int is_chroma); - void ff_h264_hl_decode_mb(const H264Context *h, H264SliceContext *sl); -int ff_h264_decode_extradata(H264Context *h); int ff_h264_decode_init(AVCodecContext *avctx); void ff_h264_decode_init_vlc(void); @@ -828,7 +718,7 @@ int ff_h264_decode_mb_cabac(const H264Context *h, H264SliceContext *sl); void ff_h264_init_cabac_states(const H264Context *h, H264SliceContext *sl); -void h264_init_dequant_tables(H264Context *h); +void ff_h264_init_dequant_tables(H264Context *h); void ff_h264_direct_dist_scale_factor(const H264Context *const h, H264SliceContext *sl); void ff_h264_direct_ref_list_init(const H264Context *const h, H264SliceContext *sl); @@ -842,13 +732,6 @@ void ff_h264_filter_mb(const H264Context *h, H264SliceContext *sl, int mb_x, int uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize); -/** - * Reset SEI values at the beginning of the frame. - * - * @param h H.264 context. - */ -void ff_h264_reset_sei(H264Context *h); - /* * o-o o-o * / / / @@ -922,7 +805,7 @@ static av_always_inline uint16_t pack8to16(int a, int b) */ static av_always_inline int get_chroma_qp(const H264Context *h, int t, int qscale) { - return h->pps.chroma_qp_table[t][qscale]; + return h->ps.pps->chroma_qp_table[t][qscale]; } /** @@ -936,7 +819,7 @@ static av_always_inline int pred_intra_mode(const H264Context *h, const int top = sl->intra4x4_pred_mode_cache[index8 - 8]; const int min = FFMIN(left, top); - tprintf(h->avctx, "mode:%d %d min:%d\n", left, top, min); + ff_tlog(h->avctx, "mode:%d %d min:%d\n", left, top, min); if (min < 0) return DC_PRED; @@ -1045,7 +928,7 @@ static av_always_inline void write_back_motion(const H264Context *h, static av_always_inline int get_dct8x8_allowed(const H264Context *h, H264SliceContext *sl) { - if (h->sps.direct_8x8_inference_flag) + if (h->ps.sps->direct_8x8_inference_flag) return !(AV_RN64A(sl->sub_mb_type) & ((MB_TYPE_16x8 | MB_TYPE_8x16 | MB_TYPE_8x8) * 0x0001000100010001ULL)); @@ -1060,21 +943,17 @@ int ff_h264_field_end(H264Context *h, H264SliceContext *sl, int in_setup); int ff_h264_ref_picture(H264Context *h, H264Picture *dst, H264Picture *src); void ff_h264_unref_picture(H264Context *h, H264Picture *pic); -int ff_h264_context_init(H264Context *h); -int ff_h264_set_parameter_from_sps(H264Context *h); +int ff_h264_slice_context_init(H264Context *h, H264SliceContext *sl); void ff_h264_draw_horiz_band(const H264Context *h, H264SliceContext *sl, int y, int height); -int ff_init_poc(H264Context *h, int pic_field_poc[2], int *pic_poc); -int ff_pred_weight_table(H264Context *h, H264SliceContext *sl); -int ff_set_ref_count(H264Context *h, H264SliceContext *sl); -int ff_h264_decode_slice_header(H264Context *h, H264SliceContext *sl, H264Context *h0); +int ff_h264_decode_slice_header(H264Context *h, H264SliceContext *sl); int ff_h264_execute_decode_slices(H264Context *h, unsigned context_count); int ff_h264_update_thread_context(AVCodecContext *dst, const AVCodecContext *src); void ff_h264_flush_change(H264Context *h); -void ff_h264_free_tables(H264Context *h, int free_rbsp); +void ff_h264_free_tables(H264Context *h); #endif /* AVCODEC_H264_H */