* H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
* Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
*
- * This file is part of FFmpeg.
+ * This file is part of Libav.
*
- * FFmpeg is free software; you can redistribute it and/or
+ * Libav is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
- * FFmpeg is distributed in the hope that it will be useful,
+ * Libav is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with FFmpeg; if not, write to the Free Software
+ * License along with Libav; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
/**
- * @file libavcodec/h264.h
+ * @file
* H.264 / AVC / MPEG4 part10 codec.
* @author Michael Niedermayer <michaelni@gmx.at>
*/
#include "dsputil.h"
#include "cabac.h"
#include "mpegvideo.h"
+#include "h264dsp.h"
#include "h264pred.h"
#include "rectangle.h"
#define interlaced_dct interlaced_dct_is_a_bad_name
#define mb_intra mb_intra_is_not_initialized_see_mb_type
-#define LUMA_DC_BLOCK_INDEX 25
-#define CHROMA_DC_BLOCK_INDEX 26
-
#define CHROMA_DC_COEFF_TOKEN_VLC_BITS 8
#define COEFF_TOKEN_VLC_BITS 8
#define TOTAL_ZEROS_VLC_BITS 9
* of progressive decoding by about 2%. */
#define ALLOW_INTERLACE
-#define ALLOW_NOCHROMA
-
#define FMO 0
/**
#endif
#define FIELD_OR_MBAFF_PICTURE (FRAME_MBAFF || FIELD_PICTURE)
-#ifdef ALLOW_NOCHROMA
-#define CHROMA h->sps.chroma_format_idc
-#else
-#define CHROMA 1
-#endif
-
#ifndef CABAC
#define CABAC h->pps.cabac
#endif
+#define CHROMA444 (h->sps.chroma_format_idc == 3)
+
#define EXTENDED_SAR 255
#define MB_TYPE_REF0 MB_TYPE_ACPRED //dirty but it fits in 16 bit
*/
#define DELAYED_PIC_REF 4
+#define QP_MAX_NUM (51 + 2*6) // The maximum supported qp
/* NAL unit types */
enum {
int num_reorder_frames;
int scaling_matrix_present;
uint8_t scaling_matrix4[6][16];
- uint8_t scaling_matrix8[2][64];
+ uint8_t scaling_matrix8[6][64];
int nal_hrd_parameters_present_flag;
int vcl_hrd_parameters_present_flag;
int pic_struct_present_flag;
int bit_depth_luma; ///< bit_depth_luma_minus8 + 8
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
}SPS;
/**
int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
int transform_8x8_mode; ///< transform_8x8_mode_flag
uint8_t scaling_matrix4[6][16];
- uint8_t scaling_matrix8[2][64];
+ 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;
}PPS;
*/
typedef struct H264Context{
MpegEncContext s;
- int nal_ref_idc;
- int nal_unit_type;
- uint8_t *rbsp_buffer[2];
- unsigned int rbsp_buffer_size[2];
-
- /**
- * Used to parse AVC variant of h264
- */
- int is_avc; ///< this flag is != 0 if codec is avc1
- int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
-
+ H264DSPContext h264dsp;
+ int pixel_shift; ///< 0 for 8-bit H264, 1 for high-bit-depth H264
int chroma_qp[2]; //QPc
int qp_thresh; ///< QP threshold to skip loopfilter
unsigned int top_samples_available;
unsigned int topright_samples_available;
unsigned int left_samples_available;
- uint8_t (*top_borders[2])[16+2*8];
- uint8_t left_border[2*(17+2*9)];
+ uint8_t (*top_borders[2])[(16*3)*2];
/**
* non zero coeff count cache.
* is 64 if not available.
*/
- DECLARE_ALIGNED_8(uint8_t, non_zero_count_cache)[6*8];
+ DECLARE_ALIGNED(8, uint8_t, non_zero_count_cache)[15*8];
- /*
- .UU.YYYY
- .UU.YYYY
- .vv.YYYY
- .VV.YYYY
- */
- uint8_t (*non_zero_count)[32];
+ uint8_t (*non_zero_count)[48];
/**
* Motion vector cache.
*/
- DECLARE_ALIGNED_16(int16_t, mv_cache)[2][5*8][2];
- DECLARE_ALIGNED_8(int8_t, ref_cache)[2][5*8];
+ DECLARE_ALIGNED(16, int16_t, mv_cache)[2][5*8][2];
+ DECLARE_ALIGNED(8, int8_t, ref_cache)[2][5*8];
#define LIST_NOT_USED -1 //FIXME rename?
#define PART_NOT_AVAILABLE -2
* block_offset[ 0..23] for frame macroblocks
* block_offset[24..47] for field macroblocks
*/
- int block_offset[2*(16+8)];
+ int block_offset[2*(16*3)];
uint32_t *mb2b_xy; //FIXME are these 4 a good idea?
uint32_t *mb2br_xy;
int emu_edge_width;
int emu_edge_height;
- int halfpel_flag;
- int thirdpel_flag;
-
- int unknown_svq3_flag;
- int next_slice_index;
-
- SPS *sps_buffers[MAX_SPS_COUNT];
SPS sps; ///< current sps
- PPS *pps_buffers[MAX_PPS_COUNT];
/**
* current pps
*/
PPS pps; //FIXME move to Picture perhaps? (->no) do we need that?
- uint32_t dequant4_buffer[6][52][16];
- uint32_t dequant8_buffer[2][52][64];
+ 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[2])[64];
- int dequant_coeff_pps; ///< reinit tables when pps changes
+ uint32_t (*dequant8_coeff[6])[64];
int slice_num;
- uint16_t *slice_table_base;
uint16_t *slice_table; ///< slice_table_base + 2*mb_stride + 1
int slice_type;
int slice_type_nos; ///< S free slice type (SI/SP are remapped to I/P)
int mb_field_decoding_flag;
int mb_mbaff; ///< mb_aff_frame && mb_field_decoding_flag
- DECLARE_ALIGNED_8(uint16_t, sub_mb_type)[4];
-
- //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
-
- /**
- * frame_num for frames or 2*frame_num+1 for field pics.
- */
- int curr_pic_num;
-
- /**
- * max_frame_num or 2*max_frame_num for field pics.
- */
- int max_pic_num;
+ DECLARE_ALIGNED(8, uint16_t, sub_mb_type)[4];
//Weighted pred stuff
int use_weight;
int use_weight_chroma;
int luma_log2_weight_denom;
int chroma_log2_weight_denom;
- int luma_weight[2][48];
- int luma_offset[2][48];
- int chroma_weight[2][48][2];
- int chroma_offset[2][48][2];
- int implicit_weight[48][48];
-
- //deblock
- int deblocking_filter; ///< disable_deblocking_filter_idc with 1<->0
- int slice_alpha_c0_offset;
- int slice_beta_offset;
-
- int redundant_pic_count;
+ //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];
int direct_spatial_mv_pred;
int col_parity;
unsigned int ref_count[2]; ///< counts frames or fields, depending on current mb mode
unsigned int list_count;
uint8_t *list_counts; ///< Array of list_count per MB specifying the slice type
- Picture *short_ref[32];
- Picture *long_ref[32];
- Picture default_ref_list[2][32]; ///< base reference list for all slices of a coded picture
Picture 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
- Picture *delayed_pic[MAX_DELAYED_PIC_COUNT+2]; //FIXME size?
- int outputed_poc;
-
- /**
- * memory management control operations buffer.
- */
- MMCO mmco[MAX_MMCO_COUNT];
- int mmco_index;
-
- int long_ref_count; ///< number of actual long term references
- int short_ref_count; ///< number of actual short term references
//data partitioning
GetBitContext intra_gb;
GetBitContext *intra_gb_ptr;
GetBitContext *inter_gb_ptr;
- DECLARE_ALIGNED_16(DCTELEM, mb)[16*24];
- DCTELEM mb_padding[256]; ///< as mb is addressed by scantable[i] and scantable is uint8_t we can either check that i is not too large or ensure that there is some unused stuff after mb
+ DECLARE_ALIGNED(16, DCTELEM, mb)[16*48*2]; ///< as a dct coeffecient is int32_t in high depth, we need to reserve twice the space.
+ DECLARE_ALIGNED(16, DCTELEM, mb_luma_dc)[3][16*2];
+ DCTELEM mb_padding[256*2]; ///< as mb is addressed by scantable[i] and scantable is uint8_t we can either check that i is not too large or ensure that there is some unused stuff after mb
/**
* Cabac
*/
CABACContext cabac;
- uint8_t cabac_state[460];
- int cabac_init_idc;
+ uint8_t cabac_state[1024];
/* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0,1,2), 0x0? luma_cbp */
uint16_t *cbp_table;
uint8_t *chroma_pred_mode_table;
int last_qscale_diff;
uint8_t (*mvd_table[2])[2];
- DECLARE_ALIGNED_16(uint8_t, mvd_cache)[2][5*8][2];
+ DECLARE_ALIGNED(16, uint8_t, mvd_cache)[2][5*8][2];
uint8_t *direct_table;
uint8_t direct_cache[5*8];
int x264_build;
+ int mb_xy;
+
+ int is_complex;
+
+ //deblock
+ int deblocking_filter; ///< disable_deblocking_filter_idc with 1<->0
+ int slice_alpha_c0_offset;
+ int slice_beta_offset;
+
+//=============================================================
+ //Things below are not used in the MB or more inner code
+
+ int nal_ref_idc;
+ int nal_unit_type;
+ uint8_t *rbsp_buffer[2];
+ unsigned int rbsp_buffer_size[2];
+
+ /**
+ * Used to parse AVC variant of h264
+ */
+ int is_avc; ///< this flag is != 0 if codec is avc1
+ int nal_length_size; ///< Number of bytes used for nal length (1, 2 or 4)
+ int got_first; ///< this flag is != 0 if we've parsed a frame
+
+ SPS *sps_buffers[MAX_SPS_COUNT];
+ PPS *pps_buffers[MAX_PPS_COUNT];
+
+ int dequant_coeff_pps; ///< reinit tables when pps changes
+
+ 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
+
+ /**
+ * frame_num for frames or 2*frame_num+1 for field pics.
+ */
+ int curr_pic_num;
+
+ /**
+ * max_frame_num or 2*max_frame_num for field pics.
+ */
+ int max_pic_num;
+
+ int redundant_pic_count;
+
+ Picture *short_ref[32];
+ Picture *long_ref[32];
+ Picture default_ref_list[2][32]; ///< base reference list for all slices of a coded picture
+ Picture *delayed_pic[MAX_DELAYED_PIC_COUNT+2]; //FIXME size?
+ Picture *next_output_pic;
+ int outputed_poc;
+ int next_outputed_poc;
+
/**
- * @defgroup multithreading Members for slice based multithreading
+ * memory management control operations buffer.
+ */
+ MMCO mmco[MAX_MMCO_COUNT];
+ int mmco_index;
+
+ int long_ref_count; ///< number of actual long term references
+ int short_ref_count; ///< number of actual short term references
+
+ int cabac_init_idc;
+
+ /**
+ * @name Members for slice based multithreading
* @{
*/
struct H264Context *thread_context[MAX_THREADS];
int last_slice_type;
/** @} */
- int mb_xy;
-
- uint32_t svq3_watermark_key;
-
/**
* pic_struct in picture timing SEI message
*/
*/
int sei_recovery_frame_cnt;
- int is_complex;
-
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
}H264Context;
-extern const uint8_t ff_h264_chroma_qp[52];
-
-void ff_svq3_luma_dc_dequant_idct_c(DCTELEM *block, int qp);
-
-void ff_svq3_add_idct_c(uint8_t *dst, DCTELEM *block, int stride, int qp, int dc);
+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).
/**
* Decode SEI
*/
int ff_h264_decode_seq_parameter_set(H264Context *h);
+/**
+ * compute profile from sps
+ */
+int ff_h264_get_profile(SPS *sps);
+
/**
* Decode PPS
*/
int ff_h264_decode_picture_parameter_set(H264Context *h, int bit_length);
/**
- * Decodes a network abstraction layer unit.
+ * 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?
- * @returns decoded bytes, might be src+1 if no escapes
+ * @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);
/**
- * identifies the exact end of the bitstream
- * @return the length of the trailing, or 0 if damaged
- */
-int ff_h264_decode_rbsp_trailing(H264Context *h, const uint8_t *src);
-
-/**
- * frees any data that may have been allocated in the H264 context like SPS, PPS etc.
+ * Free any data that may have been allocated in the H264 context like SPS, PPS etc.
*/
av_cold void ff_h264_free_context(H264Context *h);
/**
- * reconstructs bitstream slice_type.
+ * Reconstruct bitstream slice_type.
*/
int ff_h264_get_slice_type(const H264Context *h);
/**
- * allocates tables.
+ * Allocate tables.
* needs width/height
*/
int ff_h264_alloc_tables(H264Context *h);
/**
- * fills the default_ref_list.
+ * Fill the default_ref_list.
*/
int ff_h264_fill_default_ref_list(H264Context *h);
void ff_h264_remove_all_refs(H264Context *h);
/**
- * Executes the reference picture marking (memory management control operations).
+ * Execute the reference picture marking (memory management control operations).
*/
int ff_h264_execute_ref_pic_marking(H264Context *h, MMCO *mmco, int mmco_count);
int ff_h264_decode_ref_pic_marking(H264Context *h, GetBitContext *gb);
+void ff_generate_sliding_window_mmcos(H264Context *h);
+
/**
- * checks if the top & left blocks are available if needed & changes the dc mode so it only uses the available blocks.
+ * 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(H264Context *h);
/**
- * checks if the top & left blocks are available if needed & changes the dc mode so it only uses the available blocks.
+ * 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(H264Context *h, int mode);
void ff_h264_write_back_intra_pred_mode(H264Context *h);
void ff_h264_hl_decode_mb(H264Context *h);
int ff_h264_frame_start(H264Context *h);
+int ff_h264_decode_extradata(H264Context *h);
av_cold int ff_h264_decode_init(AVCodecContext *avctx);
av_cold int ff_h264_decode_end(AVCodecContext *avctx);
av_cold void ff_h264_decode_init_vlc(void);
/**
- * decodes a macroblock
- * @returns 0 if OK, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
+ * Decode a macroblock
+ * @return 0 if OK, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
*/
int ff_h264_decode_mb_cavlc(H264Context *h);
/**
- * decodes a CABAC coded macroblock
- * @returns 0 if OK, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
+ * Decode a CABAC coded macroblock
+ * @return 0 if OK, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
*/
int ff_h264_decode_mb_cabac(H264Context *h);
/ / /
o-o o-o
*/
+
+/* Scan8 organization:
+ * 0 1 2 3 4 5 6 7
+ * 0 DY y y y y y
+ * 1 y Y Y Y Y
+ * 2 y Y Y Y Y
+ * 3 y Y Y Y Y
+ * 4 y Y Y Y Y
+ * 5 DU u u u u u
+ * 6 u U U U U
+ * 7 u U U U U
+ * 8 u U U U U
+ * 9 u U U U U
+ * 10 DV v v v v v
+ * 11 v V V V V
+ * 12 v V V V V
+ * 13 v V V V V
+ * 14 v V V V V
+ * DY/DU/DV are for luma/chroma DC.
+ */
+
+#define LUMA_DC_BLOCK_INDEX 48
+#define CHROMA_DC_BLOCK_INDEX 49
+
//This table must be here because scan8[constant] must be known at compiletime
-static const uint8_t scan8[16 + 2*4]={
- 4+1*8, 5+1*8, 4+2*8, 5+2*8,
- 6+1*8, 7+1*8, 6+2*8, 7+2*8,
- 4+3*8, 5+3*8, 4+4*8, 5+4*8,
- 6+3*8, 7+3*8, 6+4*8, 7+4*8,
- 1+1*8, 2+1*8,
- 1+2*8, 2+2*8,
- 1+4*8, 2+4*8,
- 1+5*8, 2+5*8,
+static const uint8_t scan8[16*3 + 3]={
+ 4+ 1*8, 5+ 1*8, 4+ 2*8, 5+ 2*8,
+ 6+ 1*8, 7+ 1*8, 6+ 2*8, 7+ 2*8,
+ 4+ 3*8, 5+ 3*8, 4+ 4*8, 5+ 4*8,
+ 6+ 3*8, 7+ 3*8, 6+ 4*8, 7+ 4*8,
+ 4+ 6*8, 5+ 6*8, 4+ 7*8, 5+ 7*8,
+ 6+ 6*8, 7+ 6*8, 6+ 7*8, 7+ 7*8,
+ 4+ 8*8, 5+ 8*8, 4+ 9*8, 5+ 9*8,
+ 6+ 8*8, 7+ 8*8, 6+ 9*8, 7+ 9*8,
+ 4+11*8, 5+11*8, 4+12*8, 5+12*8,
+ 6+11*8, 7+11*8, 6+12*8, 7+12*8,
+ 4+13*8, 5+13*8, 4+14*8, 5+14*8,
+ 6+13*8, 7+13*8, 6+14*8, 7+14*8,
+ 0+ 0*8, 0+ 5*8, 0+10*8
};
static av_always_inline uint32_t pack16to32(int a, int b){
MpegEncContext * const s = &h->s;
const int mb_xy= h->mb_xy;
int topleft_xy, top_xy, topright_xy, left_xy[2];
- static const uint8_t left_block_options[4][16]={
- {0,1,2,3,7,10,8,11,7+0*8, 7+1*8, 7+2*8, 7+3*8, 2+0*8, 2+3*8, 2+1*8, 2+2*8},
- {2,2,3,3,8,11,8,11,7+2*8, 7+2*8, 7+3*8, 7+3*8, 2+1*8, 2+2*8, 2+1*8, 2+2*8},
- {0,0,1,1,7,10,7,10,7+0*8, 7+0*8, 7+1*8, 7+1*8, 2+0*8, 2+3*8, 2+0*8, 2+3*8},
- {0,2,0,2,7,10,7,10,7+0*8, 7+2*8, 7+0*8, 7+2*8, 2+0*8, 2+3*8, 2+0*8, 2+3*8}
+ static const uint8_t left_block_options[4][32]={
+ {0,1,2,3,7,10,8,11,3+0*4, 3+1*4, 3+2*4, 3+3*4, 1+4*4, 1+8*4, 1+5*4, 1+9*4},
+ {2,2,3,3,8,11,8,11,3+2*4, 3+2*4, 3+3*4, 3+3*4, 1+5*4, 1+9*4, 1+5*4, 1+9*4},
+ {0,0,1,1,7,10,7,10,3+0*4, 3+0*4, 3+1*4, 3+1*4, 1+4*4, 1+8*4, 1+4*4, 1+8*4},
+ {0,2,0,2,7,10,7,10,3+0*4, 3+2*4, 3+0*4, 3+2*4, 1+4*4, 1+8*4, 1+4*4, 1+8*4}
};
h->topleft_partition= -1;
h->left_mb_xy[0] = left_xy[0];
h->left_mb_xy[1] = left_xy[1];
//FIXME do we need all in the context?
- h->topleft_type = h->slice_table[topleft_xy ] == h->slice_num ? s->current_picture.mb_type[topleft_xy] : 0;
- h->top_type = h->slice_table[top_xy ] == h->slice_num ? s->current_picture.mb_type[top_xy] : 0;
- h->topright_type= h->slice_table[topright_xy] == h->slice_num ? s->current_picture.mb_type[topright_xy]: 0;
- h->left_type[0] = h->slice_table[left_xy[0] ] == h->slice_num ? s->current_picture.mb_type[left_xy[0]] : 0;
- h->left_type[1] = h->slice_table[left_xy[1] ] == h->slice_num ? s->current_picture.mb_type[left_xy[1]] : 0;
+
+ h->topleft_type = s->current_picture.mb_type[topleft_xy] ;
+ h->top_type = s->current_picture.mb_type[top_xy] ;
+ h->topright_type= s->current_picture.mb_type[topright_xy];
+ h->left_type[0] = s->current_picture.mb_type[left_xy[0]] ;
+ h->left_type[1] = s->current_picture.mb_type[left_xy[1]] ;
+
+ if(FMO){
+ if(h->slice_table[topleft_xy ] != h->slice_num) h->topleft_type = 0;
+ if(h->slice_table[top_xy ] != h->slice_num) h->top_type = 0;
+ if(h->slice_table[left_xy[0] ] != h->slice_num) h->left_type[0] = h->left_type[1] = 0;
+ }else{
+ if(h->slice_table[topleft_xy ] != h->slice_num){
+ h->topleft_type = 0;
+ if(h->slice_table[top_xy ] != h->slice_num) h->top_type = 0;
+ if(h->slice_table[left_xy[0] ] != h->slice_num) h->left_type[0] = h->left_type[1] = 0;
+ }
+ }
+ if(h->slice_table[topright_xy] != h->slice_num) h->topright_type= 0;
}
static void fill_decode_caches(H264Context *h, int mb_type){
*/
//FIXME constraint_intra_pred & partitioning & nnz (let us hope this is just a typo in the spec)
if(top_type){
- AV_COPY32(&h->non_zero_count_cache[4+8*0], &h->non_zero_count[top_xy][4+3*8]);
- h->non_zero_count_cache[1+8*0]= h->non_zero_count[top_xy][1+1*8];
- h->non_zero_count_cache[2+8*0]= h->non_zero_count[top_xy][2+1*8];
-
- h->non_zero_count_cache[1+8*3]= h->non_zero_count[top_xy][1+2*8];
- h->non_zero_count_cache[2+8*3]= h->non_zero_count[top_xy][2+2*8];
- }else {
- h->non_zero_count_cache[1+8*0]=
- h->non_zero_count_cache[2+8*0]=
-
- h->non_zero_count_cache[1+8*3]=
- h->non_zero_count_cache[2+8*3]=
- AV_WN32A(&h->non_zero_count_cache[4+8*0], CABAC && !IS_INTRA(mb_type) ? 0 : 0x40404040);
+ AV_COPY32(&h->non_zero_count_cache[4+8* 0], &h->non_zero_count[top_xy][4*3]);
+ if(CHROMA444){
+ AV_COPY32(&h->non_zero_count_cache[4+8* 5], &h->non_zero_count[top_xy][4* 7]);
+ AV_COPY32(&h->non_zero_count_cache[4+8*10], &h->non_zero_count[top_xy][4*11]);
+ }else{
+ AV_COPY32(&h->non_zero_count_cache[4+8* 5], &h->non_zero_count[top_xy][4* 5]);
+ AV_COPY32(&h->non_zero_count_cache[4+8*10], &h->non_zero_count[top_xy][4* 9]);
+ }
+ }else{
+ uint32_t top_empty = CABAC && !IS_INTRA(mb_type) ? 0 : 0x40404040;
+ AV_WN32A(&h->non_zero_count_cache[4+8* 0], top_empty);
+ AV_WN32A(&h->non_zero_count_cache[4+8* 5], top_empty);
+ AV_WN32A(&h->non_zero_count_cache[4+8*10], top_empty);
}
for (i=0; i<2; i++) {
if(left_type[i]){
- h->non_zero_count_cache[3+8*1 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+0+2*i]];
- h->non_zero_count_cache[3+8*2 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+1+2*i]];
- h->non_zero_count_cache[0+8*1 + 8*i]= h->non_zero_count[left_xy[i]][left_block[8+4+2*i]];
- h->non_zero_count_cache[0+8*4 + 8*i]= h->non_zero_count[left_xy[i]][left_block[8+5+2*i]];
+ h->non_zero_count_cache[3+8* 1 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+0+2*i]];
+ h->non_zero_count_cache[3+8* 2 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+1+2*i]];
+ if(CHROMA444){
+ h->non_zero_count_cache[3+8* 6 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+0+2*i]+4*4];
+ h->non_zero_count_cache[3+8* 7 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+1+2*i]+4*4];
+ h->non_zero_count_cache[3+8*11 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+0+2*i]+8*4];
+ h->non_zero_count_cache[3+8*12 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+1+2*i]+8*4];
+ }else{
+ h->non_zero_count_cache[3+8* 6 + 8*i]= h->non_zero_count[left_xy[i]][left_block[8+4+2*i]];
+ h->non_zero_count_cache[3+8*11 + 8*i]= h->non_zero_count[left_xy[i]][left_block[8+5+2*i]];
+ }
}else{
- h->non_zero_count_cache[3+8*1 + 2*8*i]=
- h->non_zero_count_cache[3+8*2 + 2*8*i]=
- h->non_zero_count_cache[0+8*1 + 8*i]=
- h->non_zero_count_cache[0+8*4 + 8*i]= CABAC && !IS_INTRA(mb_type) ? 0 : 64;
+ h->non_zero_count_cache[3+8* 1 + 2*8*i]=
+ h->non_zero_count_cache[3+8* 2 + 2*8*i]=
+ h->non_zero_count_cache[3+8* 6 + 2*8*i]=
+ h->non_zero_count_cache[3+8* 7 + 2*8*i]=
+ h->non_zero_count_cache[3+8*11 + 2*8*i]=
+ h->non_zero_count_cache[3+8*12 + 2*8*i]= CABAC && !IS_INTRA(mb_type) ? 0 : 64;
}
}
if(top_type) {
h->top_cbp = h->cbp_table[top_xy];
} else {
- h->top_cbp = IS_INTRA(mb_type) ? 0x1CF : 0x00F;
+ h->top_cbp = IS_INTRA(mb_type) ? 0x7CF : 0x00F;
}
// left_cbp
if (left_type[0]) {
- h->left_cbp = h->cbp_table[left_xy[0]] & 0x1f0;
- h->left_cbp |= ((h->cbp_table[left_xy[0]]>>((left_block[0]&(~1))+1))&0x1) << 1;
- h->left_cbp |= ((h->cbp_table[left_xy[1]]>>((left_block[2]&(~1))+1))&0x1) << 3;
+ h->left_cbp = (h->cbp_table[left_xy[0]] & 0x7F0)
+ | ((h->cbp_table[left_xy[0]]>>(left_block[0]&(~1)))&2)
+ | (((h->cbp_table[left_xy[1]]>>(left_block[2]&(~1)))&2) << 2);
} else {
- h->left_cbp = IS_INTRA(mb_type) ? 0x1CF : 0x00F;
+ h->left_cbp = IS_INTRA(mb_type) ? 0x7CF : 0x00F;
}
}
}
-#if 1
if(IS_INTER(mb_type) || (IS_DIRECT(mb_type) && h->direct_spatial_mv_pred)){
int list;
for(list=0; list<h->list_count; list++){
AV_WN32A(&h->ref_cache[list][scan8[0] + 0 - 1*8], ((top_type ? LIST_NOT_USED : PART_NOT_AVAILABLE)&0xFF)*0x01010101);
}
+ if(mb_type & (MB_TYPE_16x8|MB_TYPE_8x8)){
for(i=0; i<2; i++){
int cache_idx = scan8[0] - 1 + i*2*8;
if(USES_LIST(left_type[i], list)){
h->ref_cache[list][cache_idx+8]= (left_type[i]) ? LIST_NOT_USED : PART_NOT_AVAILABLE;
}
}
-
- if(USES_LIST(topleft_type, list)){
- const int b_xy = h->mb2b_xy [topleft_xy] + 3 + h->b_stride + (h->topleft_partition & 2*h->b_stride);
- const int b8_xy= 4*topleft_xy + 1 + (h->topleft_partition & 2);
- AV_COPY32(h->mv_cache[list][scan8[0] - 1 - 1*8], s->current_picture.motion_val[list][b_xy]);
- h->ref_cache[list][scan8[0] - 1 - 1*8]= s->current_picture.ref_index[list][b8_xy];
}else{
- AV_ZERO32(h->mv_cache[list][scan8[0] - 1 - 1*8]);
- h->ref_cache[list][scan8[0] - 1 - 1*8]= topleft_type ? LIST_NOT_USED : PART_NOT_AVAILABLE;
+ if(USES_LIST(left_type[0], list)){
+ const int b_xy= h->mb2b_xy[left_xy[0]] + 3;
+ const int b8_xy= 4*left_xy[0] + 1;
+ AV_COPY32(h->mv_cache[list][scan8[0] - 1], s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[0]]);
+ h->ref_cache[list][scan8[0] - 1]= s->current_picture.ref_index[list][b8_xy + (left_block[0]&~1)];
+ }else{
+ AV_ZERO32(h->mv_cache [list][scan8[0] - 1]);
+ h->ref_cache[list][scan8[0] - 1]= left_type[0] ? LIST_NOT_USED : PART_NOT_AVAILABLE;
+ }
}
if(USES_LIST(topright_type, list)){
AV_ZERO32(h->mv_cache [list][scan8[0] + 4 - 1*8]);
h->ref_cache[list][scan8[0] + 4 - 1*8]= topright_type ? LIST_NOT_USED : PART_NOT_AVAILABLE;
}
+ if(h->ref_cache[list][scan8[0] + 4 - 1*8] < 0){
+ if(USES_LIST(topleft_type, list)){
+ const int b_xy = h->mb2b_xy [topleft_xy] + 3 + h->b_stride + (h->topleft_partition & 2*h->b_stride);
+ const int b8_xy= 4*topleft_xy + 1 + (h->topleft_partition & 2);
+ AV_COPY32(h->mv_cache[list][scan8[0] - 1 - 1*8], s->current_picture.motion_val[list][b_xy]);
+ h->ref_cache[list][scan8[0] - 1 - 1*8]= s->current_picture.ref_index[list][b8_xy];
+ }else{
+ AV_ZERO32(h->mv_cache[list][scan8[0] - 1 - 1*8]);
+ h->ref_cache[list][scan8[0] - 1 - 1*8]= topleft_type ? LIST_NOT_USED : PART_NOT_AVAILABLE;
+ }
+ }
if((mb_type&(MB_TYPE_SKIP|MB_TYPE_DIRECT2)) && !FRAME_MBAFF)
continue;
AV_ZERO16(h->mvd_cache [list][scan8[0] - 1 + 2*8]);
AV_ZERO16(h->mvd_cache [list][scan8[0] - 1 + 3*8]);
}
- AV_ZERO16(h->mvd_cache [list][scan8[5 ]+1]);
- AV_ZERO16(h->mvd_cache [list][scan8[7 ]+1]);
- AV_ZERO16(h->mvd_cache [list][scan8[13]+1]); //FIXME remove past 3 (init somewhere else)
AV_ZERO16(h->mvd_cache [list][scan8[4 ]]);
AV_ZERO16(h->mvd_cache [list][scan8[12]]);
- if(h->slice_type_nos == FF_B_TYPE){
+ if(h->slice_type_nos == AV_PICTURE_TYPE_B){
fill_rectangle(&h->direct_cache[scan8[0]], 4, 4, 8, MB_TYPE_16x16>>1, 1);
if(IS_DIRECT(top_type)){
- AV_WN32A(&h->direct_cache[scan8[0] - 1*8], 0x01010101*(MB_TYPE_DIRECT2>>1));
+ AV_WN32A(&h->direct_cache[scan8[0] - 1*8], 0x01010101u*(MB_TYPE_DIRECT2>>1));
}else if(IS_8X8(top_type)){
int b8_xy = 4*top_xy;
h->direct_cache[scan8[0] + 0 - 1*8]= h->direct_table[b8_xy + 2];
}
}
}
-#endif
h->neighbor_transform_size= !!IS_8x8DCT(top_type) + !!IS_8x8DCT(left_type[0]);
}
-/**
- *
- * @returns non zero if the loop filter can be skiped
- */
-static int fill_filter_caches(H264Context *h, int mb_type){
- MpegEncContext * const s = &h->s;
- const int mb_xy= h->mb_xy;
- int top_xy, left_xy[2];
- int top_type, left_type[2];
-
- top_xy = mb_xy - (s->mb_stride << MB_FIELD);
-
- //FIXME deblocking could skip the intra and nnz parts.
-
- /* Wow, what a mess, why didn't they simplify the interlacing & intra
- * stuff, I can't imagine that these complex rules are worth it. */
-
- left_xy[1] = left_xy[0] = mb_xy-1;
- if(FRAME_MBAFF){
- const int left_mb_field_flag = IS_INTERLACED(s->current_picture.mb_type[mb_xy-1]);
- const int curr_mb_field_flag = IS_INTERLACED(mb_type);
- if(s->mb_y&1){
- if (left_mb_field_flag != curr_mb_field_flag) {
- left_xy[0] -= s->mb_stride;
- }
- }else{
- if(curr_mb_field_flag){
- top_xy += s->mb_stride & (((s->current_picture.mb_type[top_xy ]>>7)&1)-1);
- }
- if (left_mb_field_flag != curr_mb_field_flag) {
- left_xy[1] += s->mb_stride;
- }
- }
- }
-
- h->top_mb_xy = top_xy;
- h->left_mb_xy[0] = left_xy[0];
- h->left_mb_xy[1] = left_xy[1];
- {
- //for sufficiently low qp, filtering wouldn't do anything
- //this is a conservative estimate: could also check beta_offset and more accurate chroma_qp
- int qp_thresh = h->qp_thresh; //FIXME strictly we should store qp_thresh for each mb of a slice
- int qp = s->current_picture.qscale_table[mb_xy];
- if(qp <= qp_thresh
- && (left_xy[0]<0 || ((qp + s->current_picture.qscale_table[left_xy[0]] + 1)>>1) <= qp_thresh)
- && (top_xy < 0 || ((qp + s->current_picture.qscale_table[top_xy ] + 1)>>1) <= qp_thresh)){
- if(!FRAME_MBAFF)
- return 1;
- if( (left_xy[0]< 0 || ((qp + s->current_picture.qscale_table[left_xy[1] ] + 1)>>1) <= qp_thresh)
- && (top_xy < s->mb_stride || ((qp + s->current_picture.qscale_table[top_xy -s->mb_stride] + 1)>>1) <= qp_thresh))
- return 1;
- }
- }
-
- if(h->deblocking_filter == 2){
- h->top_type = top_type = h->slice_table[top_xy ] == h->slice_num ? s->current_picture.mb_type[top_xy] : 0;
- h->left_type[0]= left_type[0] = h->slice_table[left_xy[0] ] == h->slice_num ? s->current_picture.mb_type[left_xy[0]] : 0;
- h->left_type[1]= left_type[1] = h->slice_table[left_xy[1] ] == h->slice_num ? s->current_picture.mb_type[left_xy[1]] : 0;
- }else{
- h->top_type = top_type = h->slice_table[top_xy ] < 0xFFFF ? s->current_picture.mb_type[top_xy] : 0;
- h->left_type[0]= left_type[0] = h->slice_table[left_xy[0] ] < 0xFFFF ? s->current_picture.mb_type[left_xy[0]] : 0;
- h->left_type[1]= left_type[1] = h->slice_table[left_xy[1] ] < 0xFFFF ? s->current_picture.mb_type[left_xy[1]] : 0;
- }
- if(IS_INTRA(mb_type))
- return 0;
-
- AV_COPY64(&h->non_zero_count_cache[0+8*1], &h->non_zero_count[mb_xy][ 0]);
- AV_COPY64(&h->non_zero_count_cache[0+8*2], &h->non_zero_count[mb_xy][ 8]);
- AV_COPY32(&h->non_zero_count_cache[0+8*5], &h->non_zero_count[mb_xy][16]);
- AV_COPY32(&h->non_zero_count_cache[4+8*3], &h->non_zero_count[mb_xy][20]);
- AV_COPY64(&h->non_zero_count_cache[0+8*4], &h->non_zero_count[mb_xy][24]);
-
- h->cbp= h->cbp_table[mb_xy];
-
- {
- int list;
- for(list=0; list<h->list_count; list++){
- int8_t *ref;
- int y, b_stride;
- int16_t (*mv_dst)[2];
- int16_t (*mv_src)[2];
-
- if(!USES_LIST(mb_type, list)){
- fill_rectangle( h->mv_cache[list][scan8[0]], 4, 4, 8, pack16to32(0,0), 4);
- AV_WN32A(&h->ref_cache[list][scan8[ 0]], ((LIST_NOT_USED)&0xFF)*0x01010101u);
- AV_WN32A(&h->ref_cache[list][scan8[ 2]], ((LIST_NOT_USED)&0xFF)*0x01010101u);
- AV_WN32A(&h->ref_cache[list][scan8[ 8]], ((LIST_NOT_USED)&0xFF)*0x01010101u);
- AV_WN32A(&h->ref_cache[list][scan8[10]], ((LIST_NOT_USED)&0xFF)*0x01010101u);
- continue;
- }
-
- ref = &s->current_picture.ref_index[list][4*mb_xy];
- {
- int (*ref2frm)[64] = h->ref2frm[ h->slice_num&(MAX_SLICES-1) ][0] + (MB_MBAFF ? 20 : 2);
- AV_WN32A(&h->ref_cache[list][scan8[ 0]], (pack16to32(ref2frm[list][ref[0]],ref2frm[list][ref[1]])&0x00FF00FF)*0x0101);
- AV_WN32A(&h->ref_cache[list][scan8[ 2]], (pack16to32(ref2frm[list][ref[0]],ref2frm[list][ref[1]])&0x00FF00FF)*0x0101);
- ref += 2;
- AV_WN32A(&h->ref_cache[list][scan8[ 8]], (pack16to32(ref2frm[list][ref[0]],ref2frm[list][ref[1]])&0x00FF00FF)*0x0101);
- AV_WN32A(&h->ref_cache[list][scan8[10]], (pack16to32(ref2frm[list][ref[0]],ref2frm[list][ref[1]])&0x00FF00FF)*0x0101);
- }
-
- b_stride = h->b_stride;
- mv_dst = &h->mv_cache[list][scan8[0]];
- mv_src = &s->current_picture.motion_val[list][4*s->mb_x + 4*s->mb_y*b_stride];
- for(y=0; y<4; y++){
- AV_COPY128(mv_dst + 8*y, mv_src + y*b_stride);
- }
-
- }
- }
-
-
-/*
-0 . T T. T T T T
-1 L . .L . . . .
-2 L . .L . . . .
-3 . T TL . . . .
-4 L . .L . . . .
-5 L . .. . . . .
-*/
-//FIXME constraint_intra_pred & partitioning & nnz (let us hope this is just a typo in the spec)
- if(top_type){
- AV_COPY32(&h->non_zero_count_cache[4+8*0], &h->non_zero_count[top_xy][4+3*8]);
- }
-
- if(left_type[0]){
- h->non_zero_count_cache[3+8*1]= h->non_zero_count[left_xy[0]][7+0*8];
- h->non_zero_count_cache[3+8*2]= h->non_zero_count[left_xy[0]][7+1*8];
- h->non_zero_count_cache[3+8*3]= h->non_zero_count[left_xy[0]][7+2*8];
- h->non_zero_count_cache[3+8*4]= h->non_zero_count[left_xy[0]][7+3*8];
- }
-
- // CAVLC 8x8dct requires NNZ values for residual decoding that differ from what the loop filter needs
- if(!CABAC && h->pps.transform_8x8_mode){
- if(IS_8x8DCT(top_type)){
- h->non_zero_count_cache[4+8*0]=
- h->non_zero_count_cache[5+8*0]= h->cbp_table[top_xy] & 4;
- h->non_zero_count_cache[6+8*0]=
- h->non_zero_count_cache[7+8*0]= h->cbp_table[top_xy] & 8;
- }
- if(IS_8x8DCT(left_type[0])){
- h->non_zero_count_cache[3+8*1]=
- h->non_zero_count_cache[3+8*2]= h->cbp_table[left_xy[0]]&2; //FIXME check MBAFF
- }
- if(IS_8x8DCT(left_type[1])){
- h->non_zero_count_cache[3+8*3]=
- h->non_zero_count_cache[3+8*4]= h->cbp_table[left_xy[1]]&8; //FIXME check MBAFF
- }
-
- if(IS_8x8DCT(mb_type)){
- h->non_zero_count_cache[scan8[0 ]]= h->non_zero_count_cache[scan8[1 ]]=
- h->non_zero_count_cache[scan8[2 ]]= h->non_zero_count_cache[scan8[3 ]]= h->cbp & 1;
-
- h->non_zero_count_cache[scan8[0+ 4]]= h->non_zero_count_cache[scan8[1+ 4]]=
- h->non_zero_count_cache[scan8[2+ 4]]= h->non_zero_count_cache[scan8[3+ 4]]= h->cbp & 2;
-
- h->non_zero_count_cache[scan8[0+ 8]]= h->non_zero_count_cache[scan8[1+ 8]]=
- h->non_zero_count_cache[scan8[2+ 8]]= h->non_zero_count_cache[scan8[3+ 8]]= h->cbp & 4;
-
- h->non_zero_count_cache[scan8[0+12]]= h->non_zero_count_cache[scan8[1+12]]=
- h->non_zero_count_cache[scan8[2+12]]= h->non_zero_count_cache[scan8[3+12]]= h->cbp & 8;
- }
- }
-
- if(IS_INTER(mb_type) || IS_DIRECT(mb_type)){
- int list;
- for(list=0; list<h->list_count; list++){
- if(USES_LIST(top_type, list)){
- const int b_xy= h->mb2b_xy[top_xy] + 3*h->b_stride;
- const int b8_xy= 4*top_xy + 2;
- int (*ref2frm)[64] = h->ref2frm[ h->slice_table[top_xy]&(MAX_SLICES-1) ][0] + (MB_MBAFF ? 20 : 2);
- AV_COPY128(h->mv_cache[list][scan8[0] + 0 - 1*8], s->current_picture.motion_val[list][b_xy + 0]);
- h->ref_cache[list][scan8[0] + 0 - 1*8]=
- h->ref_cache[list][scan8[0] + 1 - 1*8]= ref2frm[list][s->current_picture.ref_index[list][b8_xy + 0]];
- h->ref_cache[list][scan8[0] + 2 - 1*8]=
- h->ref_cache[list][scan8[0] + 3 - 1*8]= ref2frm[list][s->current_picture.ref_index[list][b8_xy + 1]];
- }else{
- AV_ZERO128(h->mv_cache[list][scan8[0] + 0 - 1*8]);
- AV_WN32A(&h->ref_cache[list][scan8[0] + 0 - 1*8], ((LIST_NOT_USED)&0xFF)*0x01010101u);
- }
-
- if(!IS_INTERLACED(mb_type^left_type[0])){
- if(USES_LIST(left_type[0], list)){
- const int b_xy= h->mb2b_xy[left_xy[0]] + 3;
- const int b8_xy= 4*left_xy[0] + 1;
- int (*ref2frm)[64] = h->ref2frm[ h->slice_table[left_xy[0]]&(MAX_SLICES-1) ][0] + (MB_MBAFF ? 20 : 2);
- AV_COPY32(h->mv_cache[list][scan8[0] - 1 + 0 ], s->current_picture.motion_val[list][b_xy + h->b_stride*0]);
- AV_COPY32(h->mv_cache[list][scan8[0] - 1 + 8 ], s->current_picture.motion_val[list][b_xy + h->b_stride*1]);
- AV_COPY32(h->mv_cache[list][scan8[0] - 1 +16 ], s->current_picture.motion_val[list][b_xy + h->b_stride*2]);
- AV_COPY32(h->mv_cache[list][scan8[0] - 1 +24 ], s->current_picture.motion_val[list][b_xy + h->b_stride*3]);
- h->ref_cache[list][scan8[0] - 1 + 0 ]=
- h->ref_cache[list][scan8[0] - 1 + 8 ]= ref2frm[list][s->current_picture.ref_index[list][b8_xy + 2*0]];
- h->ref_cache[list][scan8[0] - 1 +16 ]=
- h->ref_cache[list][scan8[0] - 1 +24 ]= ref2frm[list][s->current_picture.ref_index[list][b8_xy + 2*1]];
- }else{
- AV_ZERO32(h->mv_cache [list][scan8[0] - 1 + 0 ]);
- AV_ZERO32(h->mv_cache [list][scan8[0] - 1 + 8 ]);
- AV_ZERO32(h->mv_cache [list][scan8[0] - 1 +16 ]);
- AV_ZERO32(h->mv_cache [list][scan8[0] - 1 +24 ]);
- h->ref_cache[list][scan8[0] - 1 + 0 ]=
- h->ref_cache[list][scan8[0] - 1 + 8 ]=
- h->ref_cache[list][scan8[0] - 1 + 16 ]=
- h->ref_cache[list][scan8[0] - 1 + 24 ]= LIST_NOT_USED;
- }
- }
- }
- }
-
- return 0;
-}
-
/**
* gets the predicted intra4x4 prediction mode.
*/
static inline void write_back_non_zero_count(H264Context *h){
const int mb_xy= h->mb_xy;
- AV_COPY64(&h->non_zero_count[mb_xy][ 0], &h->non_zero_count_cache[0+8*1]);
- AV_COPY64(&h->non_zero_count[mb_xy][ 8], &h->non_zero_count_cache[0+8*2]);
- AV_COPY32(&h->non_zero_count[mb_xy][16], &h->non_zero_count_cache[0+8*5]);
- AV_COPY32(&h->non_zero_count[mb_xy][20], &h->non_zero_count_cache[4+8*3]);
- AV_COPY64(&h->non_zero_count[mb_xy][24], &h->non_zero_count_cache[0+8*4]);
+ AV_COPY32(&h->non_zero_count[mb_xy][ 0], &h->non_zero_count_cache[4+8* 1]);
+ AV_COPY32(&h->non_zero_count[mb_xy][ 4], &h->non_zero_count_cache[4+8* 2]);
+ AV_COPY32(&h->non_zero_count[mb_xy][ 8], &h->non_zero_count_cache[4+8* 3]);
+ AV_COPY32(&h->non_zero_count[mb_xy][12], &h->non_zero_count_cache[4+8* 4]);
+ AV_COPY32(&h->non_zero_count[mb_xy][16], &h->non_zero_count_cache[4+8* 6]);
+ AV_COPY32(&h->non_zero_count[mb_xy][20], &h->non_zero_count_cache[4+8* 7]);
+ AV_COPY32(&h->non_zero_count[mb_xy][32], &h->non_zero_count_cache[4+8*11]);
+ AV_COPY32(&h->non_zero_count[mb_xy][36], &h->non_zero_count_cache[4+8*12]);
+
+ if(CHROMA444){
+ AV_COPY32(&h->non_zero_count[mb_xy][24], &h->non_zero_count_cache[4+8* 8]);
+ AV_COPY32(&h->non_zero_count[mb_xy][28], &h->non_zero_count_cache[4+8* 9]);
+ AV_COPY32(&h->non_zero_count[mb_xy][40], &h->non_zero_count_cache[4+8*13]);
+ AV_COPY32(&h->non_zero_count[mb_xy][44], &h->non_zero_count_cache[4+8*14]);
+ }
}
static inline void write_back_motion(H264Context *h, int mb_type){
}
}
- if(h->slice_type_nos == FF_B_TYPE && CABAC){
+ if(h->slice_type_nos == AV_PICTURE_TYPE_B && CABAC){
if(IS_8X8(mb_type)){
uint8_t *direct_table = &h->direct_table[4*h->mb_xy];
direct_table[1] = h->sub_mb_type[1]>>1;
/**
* decodes a P_SKIP or B_SKIP macroblock
*/
-static void decode_mb_skip(H264Context *h){
+static void av_unused decode_mb_skip(H264Context *h){
MpegEncContext * const s = &h->s;
const int mb_xy= h->mb_xy;
int mb_type=0;
- memset(h->non_zero_count[mb_xy], 0, 32);
- memset(h->non_zero_count_cache + 8, 0, 8*5); //FIXME ugly, remove pfui
+ memset(h->non_zero_count[mb_xy], 0, 48);
if(MB_FIELD)
mb_type|= MB_TYPE_INTERLACED;
- if( h->slice_type_nos == FF_B_TYPE )
+ if( h->slice_type_nos == AV_PICTURE_TYPE_B )
{
// just for fill_caches. pred_direct_motion will set the real mb_type
mb_type|= MB_TYPE_L0L1|MB_TYPE_DIRECT2|MB_TYPE_SKIP;