ffmpeg/libavcodec/mpeg4videoenc.c

1442 lines
52 KiB
C
Raw Normal View History

/*
* MPEG-4 encoder
* Copyright (c) 2000,2001 Fabrice Bellard
* Copyright (c) 2002-2010 Michael Niedermayer <michaelni@gmx.at>
*
* This file is part of FFmpeg.
*
* FFmpeg 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,
* 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
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "libavutil/attributes.h"
#include "libavutil/log.h"
#include "libavutil/mem.h"
#include "libavutil/opt.h"
#include "libavutil/thread.h"
#include "codec_internal.h"
#include "mpegvideo.h"
#include "h263.h"
#include "h263enc.h"
#include "mpeg4video.h"
#include "mpeg4videodata.h"
#include "mpeg4videodefs.h"
#include "mpeg4videoenc.h"
#include "mpegvideoenc.h"
#include "profiles.h"
#include "version.h"
/**
* Minimal fcode that a motion vector component would need.
*/
static uint8_t fcode_tab[MAX_MV*2+1];
/* The uni_DCtab_* tables below contain unified bits+length tables to encode DC
* differences in MPEG-4. Unified in the sense that the specification specifies
* this encoding in several steps. */
static uint8_t uni_DCtab_lum_len[512];
static uint8_t uni_DCtab_chrom_len[512];
static uint16_t uni_DCtab_lum_bits[512];
static uint16_t uni_DCtab_chrom_bits[512];
/* Unified encoding tables for run length encoding of coefficients.
* Unified in the sense that the specification specifies the encoding in several steps. */
static uint32_t uni_mpeg4_intra_rl_bits[64 * 64 * 2 * 2];
static uint8_t uni_mpeg4_intra_rl_len[64 * 64 * 2 * 2];
static uint32_t uni_mpeg4_inter_rl_bits[64 * 64 * 2 * 2];
static uint8_t uni_mpeg4_inter_rl_len[64 * 64 * 2 * 2];
//#define UNI_MPEG4_ENC_INDEX(last, run, level) ((last) * 128 + (run) * 256 + (level))
//#define UNI_MPEG4_ENC_INDEX(last, run, level) ((last) * 128 * 64 + (run) + (level) * 64)
#define UNI_MPEG4_ENC_INDEX(last, run, level) ((last) * 128 * 64 + (run) * 128 + (level))
/* MPEG-4
* inter
* max level: 24/6
* max run: 53/63
*
* intra
* max level: 53/16
* max run: 29/41
*/
typedef struct Mpeg4EncContext {
MPVMainEncContext m;
/// number of bits to represent the fractional part of time
int time_increment_bits;
} Mpeg4EncContext;
static inline Mpeg4EncContext *mainctx_to_mpeg4(MPVMainEncContext *m)
{
return (Mpeg4EncContext*)m;
}
/**
* Return the number of bits that encoding the 8x8 block in block would need.
* @param[in] block_last_index last index in scantable order that refers to a non zero element in block.
*/
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
static inline int get_block_rate(MPVEncContext *const s, int16_t block[64],
int block_last_index, const uint8_t scantable[64])
{
int last = 0;
int j;
int rate = 0;
for (j = 1; j <= block_last_index; j++) {
const int index = scantable[j];
int level = block[index];
if (level) {
level += 64;
if ((level & (~127)) == 0) {
if (j < block_last_index)
rate += s->intra_ac_vlc_length[UNI_AC_ENC_INDEX(j - last - 1, level)];
else
rate += s->intra_ac_vlc_last_length[UNI_AC_ENC_INDEX(j - last - 1, level)];
} else
rate += s->ac_esc_length;
last = j;
}
}
return rate;
}
/**
* Restore the ac coefficients in block that have been changed by decide_ac_pred().
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
* This function also restores s->c.block_last_index.
* @param[in,out] block MB coefficients, these will be restored
* @param[in] dir ac prediction direction for each 8x8 block
* @param[out] st scantable for each 8x8 block
* @param[in] zigzag_last_index index referring to the last non zero coefficient in zigzag order
*/
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
static inline void restore_ac_coeffs(MPVEncContext *const s, int16_t block[6][64],
const int dir[6], const uint8_t *st[6],
const int zigzag_last_index[6])
{
int i, n;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
memcpy(s->c.block_last_index, zigzag_last_index, sizeof(int) * 6);
for (n = 0; n < 6; n++) {
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
int16_t *ac_val = &s->c.ac_val[0][0][0] + s->c.block_index[n] * 16;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
st[n] = s->c.intra_scantable.permutated;
if (dir[n]) {
/* top prediction */
for (i = 1; i < 8; i++)
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
block[n][s->c.idsp.idct_permutation[i]] = ac_val[i + 8];
} else {
/* left prediction */
for (i = 1; i < 8; i++)
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
block[n][s->c.idsp.idct_permutation[i << 3]] = ac_val[i];
}
}
}
/**
* Return the optimal value (0 or 1) for the ac_pred element for the given MB in MPEG-4.
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
* This function will also update s->c.block_last_index and s->c.ac_val.
* @param[in,out] block MB coefficients, these will be updated if 1 is returned
* @param[in] dir ac prediction direction for each 8x8 block
* @param[out] st scantable for each 8x8 block
* @param[out] zigzag_last_index index referring to the last non zero coefficient in zigzag order
*/
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
static inline int decide_ac_pred(MPVEncContext *const s, int16_t block[6][64],
const int dir[6], const uint8_t *st[6],
int zigzag_last_index[6])
{
int score = 0;
int i, n;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
const int8_t *const qscale_table = s->c.cur_pic.qscale_table;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
memcpy(zigzag_last_index, s->c.block_last_index, sizeof(int) * 6);
for (n = 0; n < 6; n++) {
int16_t *ac_val, *ac_val1;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
score -= get_block_rate(s, block[n], s->c.block_last_index[n],
s->c.intra_scantable.permutated);
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
ac_val = &s->c.ac_val[0][0][0] + s->c.block_index[n] * 16;
ac_val1 = ac_val;
if (dir[n]) {
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
const int xy = s->c.mb_x + s->c.mb_y * s->c.mb_stride - s->c.mb_stride;
/* top prediction */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
ac_val -= s->c.block_wrap[n] * 16;
if (s->c.mb_y == 0 || s->c.qscale == qscale_table[xy] || n == 2 || n == 3) {
/* same qscale */
for (i = 1; i < 8; i++) {
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
const int level = block[n][s->c.idsp.idct_permutation[i]];
block[n][s->c.idsp.idct_permutation[i]] = level - ac_val[i + 8];
ac_val1[i] = block[n][s->c.idsp.idct_permutation[i << 3]];
ac_val1[i + 8] = level;
}
} else {
/* different qscale, we must rescale */
for (i = 1; i < 8; i++) {
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
const int level = block[n][s->c.idsp.idct_permutation[i]];
block[n][s->c.idsp.idct_permutation[i]] = level - ROUNDED_DIV(ac_val[i + 8] * qscale_table[xy], s->c.qscale);
ac_val1[i] = block[n][s->c.idsp.idct_permutation[i << 3]];
ac_val1[i + 8] = level;
}
}
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
st[n] = s->c.permutated_intra_h_scantable;
} else {
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
const int xy = s->c.mb_x - 1 + s->c.mb_y * s->c.mb_stride;
/* left prediction */
ac_val -= 16;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.mb_x == 0 || s->c.qscale == qscale_table[xy] || n == 1 || n == 3) {
/* same qscale */
for (i = 1; i < 8; i++) {
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
const int level = block[n][s->c.idsp.idct_permutation[i << 3]];
block[n][s->c.idsp.idct_permutation[i << 3]] = level - ac_val[i];
ac_val1[i] = level;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
ac_val1[i + 8] = block[n][s->c.idsp.idct_permutation[i]];
}
} else {
/* different qscale, we must rescale */
for (i = 1; i < 8; i++) {
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
const int level = block[n][s->c.idsp.idct_permutation[i << 3]];
block[n][s->c.idsp.idct_permutation[i << 3]] = level - ROUNDED_DIV(ac_val[i] * qscale_table[xy], s->c.qscale);
ac_val1[i] = level;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
ac_val1[i + 8] = block[n][s->c.idsp.idct_permutation[i]];
}
}
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
st[n] = s->c.permutated_intra_v_scantable;
}
for (i = 63; i > 0; i--) // FIXME optimize
if (block[n][st[n][i]])
break;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
s->c.block_last_index[n] = i;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
score += get_block_rate(s, block[n], s->c.block_last_index[n], st[n]);
}
if (score < 0) {
return 1;
} else {
restore_ac_coeffs(s, block, dir, st, zigzag_last_index);
return 0;
}
}
/**
* modify mb_type & qscale so that encoding is actually possible in MPEG-4
*/
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
void ff_clean_mpeg4_qscales(MPVEncContext *const s)
{
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
int8_t *const qscale_table = s->c.cur_pic.qscale_table;
ff_clean_h263_qscales(s);
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.pict_type == AV_PICTURE_TYPE_B) {
int odd = 0;
/* ok, come on, this isn't funny anymore, there's more code for
* handling this MPEG-4 mess than for the actual adaptive quantization */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
for (int i = 0; i < s->c.mb_num; i++) {
int mb_xy = s->c.mb_index2xy[i];
odd += qscale_table[mb_xy] & 1;
}
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (2 * odd > s->c.mb_num)
odd = 1;
else
odd = 0;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
for (int i = 0; i < s->c.mb_num; i++) {
int mb_xy = s->c.mb_index2xy[i];
if ((qscale_table[mb_xy] & 1) != odd)
qscale_table[mb_xy]++;
if (qscale_table[mb_xy] > 31)
qscale_table[mb_xy] = 31;
}
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
for (int i = 1; i < s->c.mb_num; i++) {
int mb_xy = s->c.mb_index2xy[i];
if (qscale_table[mb_xy] != qscale_table[s->c.mb_index2xy[i - 1]] &&
(s->mb_type[mb_xy] & CANDIDATE_MB_TYPE_DIRECT)) {
s->mb_type[mb_xy] |= CANDIDATE_MB_TYPE_BIDIR;
}
}
}
}
/**
* Encode the dc value.
* @param n block index (0-3 are luma, 4-5 are chroma)
*/
static inline void mpeg4_encode_dc(PutBitContext *s, int level, int n)
{
/* DC will overflow if level is outside the [-255,255] range. */
level += 256;
if (n < 4) {
/* luminance */
put_bits(s, uni_DCtab_lum_len[level], uni_DCtab_lum_bits[level]);
} else {
/* chrominance */
put_bits(s, uni_DCtab_chrom_len[level], uni_DCtab_chrom_bits[level]);
}
}
static inline int mpeg4_get_dc_length(int level, int n)
{
if (n < 4)
return uni_DCtab_lum_len[level + 256];
else
return uni_DCtab_chrom_len[level + 256];
}
/**
* Encode an 8x8 block.
* @param n block index (0-3 are luma, 4-5 are chroma)
*/
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
static inline void mpeg4_encode_block(const MPVEncContext *const s,
const int16_t *block, int n, int intra_dc,
const uint8_t *scan_table, PutBitContext *dc_pb,
PutBitContext *ac_pb)
{
int i, last_non_zero;
const uint32_t *bits_tab;
const uint8_t *len_tab;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
const int last_index = s->c.block_last_index[n];
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.mb_intra) { // Note gcc (3.2.1 at least) will optimize this away
/* MPEG-4 based DC predictor */
mpeg4_encode_dc(dc_pb, intra_dc, n);
if (last_index < 1)
return;
i = 1;
bits_tab = uni_mpeg4_intra_rl_bits;
len_tab = uni_mpeg4_intra_rl_len;
} else {
if (last_index < 0)
return;
i = 0;
bits_tab = uni_mpeg4_inter_rl_bits;
len_tab = uni_mpeg4_inter_rl_len;
}
/* AC coefs */
last_non_zero = i - 1;
for (; i < last_index; i++) {
int level = block[scan_table[i]];
if (level) {
int run = i - last_non_zero - 1;
level += 64;
if ((level & (~127)) == 0) {
const int index = UNI_MPEG4_ENC_INDEX(0, run, level);
put_bits(ac_pb, len_tab[index], bits_tab[index]);
} else { // ESC3
put_bits(ac_pb,
7 + 2 + 1 + 6 + 1 + 12 + 1,
(3 << 23) + (3 << 21) + (0 << 20) + (run << 14) +
(1 << 13) + (((level - 64) & 0xfff) << 1) + 1);
}
last_non_zero = i;
}
}
/* if (i <= last_index) */ {
int level = block[scan_table[i]];
int run = i - last_non_zero - 1;
level += 64;
if ((level & (~127)) == 0) {
const int index = UNI_MPEG4_ENC_INDEX(1, run, level);
put_bits(ac_pb, len_tab[index], bits_tab[index]);
} else { // ESC3
put_bits(ac_pb,
7 + 2 + 1 + 6 + 1 + 12 + 1,
(3 << 23) + (3 << 21) + (1 << 20) + (run << 14) +
(1 << 13) + (((level - 64) & 0xfff) << 1) + 1);
}
}
}
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
static int mpeg4_get_block_length(MPVEncContext *const s,
const int16_t *block, int n,
int intra_dc, const uint8_t *scan_table)
{
int i, last_non_zero;
const uint8_t *len_tab;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
const int last_index = s->c.block_last_index[n];
int len = 0;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.mb_intra) { // Note gcc (3.2.1 at least) will optimize this away
/* MPEG-4 based DC predictor */
len += mpeg4_get_dc_length(intra_dc, n);
if (last_index < 1)
return len;
i = 1;
len_tab = uni_mpeg4_intra_rl_len;
} else {
if (last_index < 0)
return 0;
i = 0;
len_tab = uni_mpeg4_inter_rl_len;
}
/* AC coefs */
last_non_zero = i - 1;
for (; i < last_index; i++) {
int level = block[scan_table[i]];
if (level) {
int run = i - last_non_zero - 1;
level += 64;
if ((level & (~127)) == 0) {
const int index = UNI_MPEG4_ENC_INDEX(0, run, level);
len += len_tab[index];
} else { // ESC3
len += 7 + 2 + 1 + 6 + 1 + 12 + 1;
}
last_non_zero = i;
}
}
/* if (i <= last_index) */ {
int level = block[scan_table[i]];
int run = i - last_non_zero - 1;
level += 64;
if ((level & (~127)) == 0) {
const int index = UNI_MPEG4_ENC_INDEX(1, run, level);
len += len_tab[index];
} else { // ESC3
len += 7 + 2 + 1 + 6 + 1 + 12 + 1;
}
}
return len;
}
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
static inline void mpeg4_encode_blocks(MPVEncContext *const s,
const int16_t block[6][64],
const int intra_dc[6],
const uint8_t * const *scan_table,
PutBitContext *dc_pb,
PutBitContext *ac_pb)
{
int i;
if (scan_table) {
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT) {
for (i = 0; i < 6; i++)
skip_put_bits(&s->pb,
mpeg4_get_block_length(s, block[i], i,
intra_dc[i], scan_table[i]));
} else {
/* encode each block */
for (i = 0; i < 6; i++)
mpeg4_encode_block(s, block[i], i,
intra_dc[i], scan_table[i], dc_pb, ac_pb);
}
} else {
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT) {
for (i = 0; i < 6; i++)
skip_put_bits(&s->pb,
mpeg4_get_block_length(s, block[i], i, 0,
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
s->c.intra_scantable.permutated));
} else {
/* encode each block */
for (i = 0; i < 6; i++)
mpeg4_encode_block(s, block[i], i, 0,
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
s->c.intra_scantable.permutated, dc_pb, ac_pb);
}
}
}
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
static inline int get_b_cbp(MPVEncContext *const s, int16_t block[6][64],
int motion_x, int motion_y, int mb_type)
{
int cbp = 0, i;
if (s->mpv_flags & FF_MPV_FLAG_CBP_RD) {
int score = 0;
const int lambda = s->lambda2 >> (FF_LAMBDA_SHIFT - 6);
for (i = 0; i < 6; i++) {
if (s->coded_score[i] < 0) {
score += s->coded_score[i];
cbp |= 1 << (5 - i);
}
}
if (cbp) {
int zero_score = -6;
if ((motion_x | motion_y | s->dquant | mb_type) == 0)
zero_score -= 4; // 2 * MV + mb_type + cbp bit
zero_score *= lambda;
if (zero_score <= score)
cbp = 0;
}
for (i = 0; i < 6; i++) {
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.block_last_index[i] >= 0 && ((cbp >> (5 - i)) & 1) == 0) {
s->c.block_last_index[i] = -1;
s->c.bdsp.clear_block(s->c.block[i]);
}
}
} else {
for (i = 0; i < 6; i++) {
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.block_last_index[i] >= 0)
cbp |= 1 << (5 - i);
}
}
return cbp;
}
// FIXME this is duplicated to h263.c
static const int dquant_code[5] = { 1, 0, 9, 2, 3 };
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
static void mpeg4_encode_mb(MPVEncContext *const s, int16_t block[][64],
int motion_x, int motion_y)
{
int cbpc, cbpy, pred_x, pred_y;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
PutBitContext *const pb2 = s->c.data_partitioning ? &s->pb2 : &s->pb;
PutBitContext *const tex_pb = s->c.data_partitioning && s->c.pict_type != AV_PICTURE_TYPE_B ? &s->tex_pb : &s->pb;
PutBitContext *const dc_pb = s->c.data_partitioning && s->c.pict_type != AV_PICTURE_TYPE_I ? &s->pb2 : &s->pb;
const int interleaved_stats = (s->c.avctx->flags & AV_CODEC_FLAG_PASS1) && !s->c.data_partitioning ? 1 : 0;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (!s->c.mb_intra) {
int i, cbp;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.pict_type == AV_PICTURE_TYPE_B) {
/* convert from mv_dir to type */
static const int mb_type_table[8] = { -1, 3, 2, 1, -1, -1, -1, 0 };
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
int mb_type = mb_type_table[s->c.mv_dir];
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.mb_x == 0) {
for (i = 0; i < 2; i++)
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
s->c.last_mv[i][0][0] =
s->c.last_mv[i][0][1] =
s->c.last_mv[i][1][0] =
s->c.last_mv[i][1][1] = 0;
}
av_assert2(s->dquant >= -2 && s->dquant <= 2);
av_assert2((s->dquant & 1) == 0);
av_assert2(mb_type >= 0);
/* nothing to do if this MB was skipped in the next P-frame */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.next_pic.mbskip_table[s->c.mb_y * s->c.mb_stride + s->c.mb_x]) { // FIXME avoid DCT & ...
s->c.mv[0][0][0] =
s->c.mv[0][0][1] =
s->c.mv[1][0][0] =
s->c.mv[1][0][1] = 0;
s->c.mv_dir = MV_DIR_FORWARD; // doesn't matter
s->c.qscale -= s->dquant;
// s->c.mb_skipped = 1;
return;
}
cbp = get_b_cbp(s, block, motion_x, motion_y, mb_type);
if ((cbp | motion_x | motion_y | mb_type) == 0) {
/* direct MB with MV={0,0} */
av_assert2(s->dquant == 0);
put_bits(&s->pb, 1, 1); /* mb not coded modb1=1 */
if (interleaved_stats) {
s->misc_bits++;
s->last_bits++;
}
return;
}
put_bits(&s->pb, 1, 0); /* mb coded modb1=0 */
put_bits(&s->pb, 1, cbp ? 0 : 1); /* modb2 */ // FIXME merge
put_bits(&s->pb, mb_type + 1, 1); // this table is so simple that we don't need it :)
if (cbp)
put_bits(&s->pb, 6, cbp);
if (cbp && mb_type) {
if (s->dquant)
put_bits(&s->pb, 2, (s->dquant >> 2) + 3);
else
put_bits(&s->pb, 1, 0);
} else
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
s->c.qscale -= s->dquant;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (!s->c.progressive_sequence) {
if (cbp)
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
put_bits(&s->pb, 1, s->c.interlaced_dct);
if (mb_type) // not direct mode
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
put_bits(&s->pb, 1, s->c.mv_type == MV_TYPE_FIELD);
}
if (interleaved_stats)
s->misc_bits += get_bits_diff(s);
if (!mb_type) {
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
av_assert2(s->c.mv_dir & MV_DIRECT);
ff_h263_encode_motion_vector(s, motion_x, motion_y, 1);
} else {
av_assert2(mb_type > 0 && mb_type < 4);
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.mv_type != MV_TYPE_FIELD) {
if (s->c.mv_dir & MV_DIR_FORWARD) {
ff_h263_encode_motion_vector(s,
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
s->c.mv[0][0][0] - s->c.last_mv[0][0][0],
s->c.mv[0][0][1] - s->c.last_mv[0][0][1],
s->c.f_code);
s->c.last_mv[0][0][0] =
s->c.last_mv[0][1][0] = s->c.mv[0][0][0];
s->c.last_mv[0][0][1] =
s->c.last_mv[0][1][1] = s->c.mv[0][0][1];
}
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.mv_dir & MV_DIR_BACKWARD) {
ff_h263_encode_motion_vector(s,
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
s->c.mv[1][0][0] - s->c.last_mv[1][0][0],
s->c.mv[1][0][1] - s->c.last_mv[1][0][1],
s->c.b_code);
s->c.last_mv[1][0][0] =
s->c.last_mv[1][1][0] = s->c.mv[1][0][0];
s->c.last_mv[1][0][1] =
s->c.last_mv[1][1][1] = s->c.mv[1][0][1];
}
} else {
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.mv_dir & MV_DIR_FORWARD) {
put_bits(&s->pb, 1, s->c.field_select[0][0]);
put_bits(&s->pb, 1, s->c.field_select[0][1]);
}
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.mv_dir & MV_DIR_BACKWARD) {
put_bits(&s->pb, 1, s->c.field_select[1][0]);
put_bits(&s->pb, 1, s->c.field_select[1][1]);
}
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.mv_dir & MV_DIR_FORWARD) {
for (i = 0; i < 2; i++) {
ff_h263_encode_motion_vector(s,
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
s->c.mv[0][i][0] - s->c.last_mv[0][i][0],
s->c.mv[0][i][1] - s->c.last_mv[0][i][1] / 2,
s->c.f_code);
s->c.last_mv[0][i][0] = s->c.mv[0][i][0];
s->c.last_mv[0][i][1] = s->c.mv[0][i][1] * 2;
}
}
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.mv_dir & MV_DIR_BACKWARD) {
for (i = 0; i < 2; i++) {
ff_h263_encode_motion_vector(s,
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
s->c.mv[1][i][0] - s->c.last_mv[1][i][0],
s->c.mv[1][i][1] - s->c.last_mv[1][i][1] / 2,
s->c.b_code);
s->c.last_mv[1][i][0] = s->c.mv[1][i][0];
s->c.last_mv[1][i][1] = s->c.mv[1][i][1] * 2;
}
}
}
}
if (interleaved_stats)
s->mv_bits += get_bits_diff(s);
mpeg4_encode_blocks(s, block, NULL, NULL, NULL, &s->pb);
if (interleaved_stats)
s->p_tex_bits += get_bits_diff(s);
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
} else { /* s->c.pict_type == AV_PICTURE_TYPE_B */
cbp = get_p_cbp(s, block, motion_x, motion_y);
if ((cbp | motion_x | motion_y | s->dquant) == 0 &&
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
s->c.mv_type == MV_TYPE_16X16) {
const MPVMainEncContext *const m = slice_to_mainenc(s);
/* Check if the B-frames can skip it too, as we must skip it
* if we skip here why didn't they just compress
* the skip-mb bits instead of reusing them ?! */
if (m->max_b_frames > 0) {
int x, y, offset;
const uint8_t *p_pic;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
x = s->c.mb_x * 16;
y = s->c.mb_y * 16;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
offset = x + y * s->c.linesize;
p_pic = s->new_pic->data[0] + offset;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
s->c.mb_skipped = 1;
for (int i = 0; i < m->max_b_frames; i++) {
const uint8_t *b_pic;
int diff;
const MPVPicture *pic = m->reordered_input_picture[i + 1];
if (!pic || pic->f->pict_type != AV_PICTURE_TYPE_B)
break;
b_pic = pic->f->data[0] + offset;
if (!pic->shared)
b_pic += INPLACE_OFFSET;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (x + 16 > s->c.width || y + 16 > s->c.height) {
int x1, y1;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
int xe = FFMIN(16, s->c.width - x);
int ye = FFMIN(16, s->c.height - y);
diff = 0;
for (y1 = 0; y1 < ye; y1++) {
for (x1 = 0; x1 < xe; x1++) {
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
diff += FFABS(p_pic[x1 + y1 * s->c.linesize] - b_pic[x1 + y1 * s->c.linesize]);
}
}
diff = diff * 256 / (xe * ye);
} else {
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
diff = s->sad_cmp[0](NULL, p_pic, b_pic, s->c.linesize, 16);
}
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (diff > s->c.qscale * 70) { // FIXME check that 70 is optimal
s->c.mb_skipped = 0;
break;
}
}
} else
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
s->c.mb_skipped = 1;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.mb_skipped == 1) {
/* skip macroblock */
put_bits(&s->pb, 1, 1);
if (interleaved_stats) {
s->misc_bits++;
s->last_bits++;
}
return;
}
}
put_bits(&s->pb, 1, 0); /* mb coded */
cbpc = cbp & 3;
cbpy = cbp >> 2;
cbpy ^= 0xf;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.mv_type == MV_TYPE_16X16) {
if (s->dquant)
cbpc += 8;
put_bits(&s->pb,
ff_h263_inter_MCBPC_bits[cbpc],
ff_h263_inter_MCBPC_code[cbpc]);
put_bits(pb2, ff_h263_cbpy_tab[cbpy][1], ff_h263_cbpy_tab[cbpy][0]);
if (s->dquant)
put_bits(pb2, 2, dquant_code[s->dquant + 2]);
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (!s->c.progressive_sequence) {
if (cbp)
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
put_bits(pb2, 1, s->c.interlaced_dct);
put_bits(pb2, 1, 0);
}
if (interleaved_stats)
s->misc_bits += get_bits_diff(s);
/* motion vectors: 16x16 mode */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
ff_h263_pred_motion(&s->c, 0, 0, &pred_x, &pred_y);
ff_h263_encode_motion_vector(s,
motion_x - pred_x,
motion_y - pred_y,
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
s->c.f_code);
} else if (s->c.mv_type == MV_TYPE_FIELD) {
if (s->dquant)
cbpc += 8;
put_bits(&s->pb,
ff_h263_inter_MCBPC_bits[cbpc],
ff_h263_inter_MCBPC_code[cbpc]);
put_bits(pb2, ff_h263_cbpy_tab[cbpy][1], ff_h263_cbpy_tab[cbpy][0]);
if (s->dquant)
put_bits(pb2, 2, dquant_code[s->dquant + 2]);
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
av_assert2(!s->c.progressive_sequence);
if (cbp)
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
put_bits(pb2, 1, s->c.interlaced_dct);
put_bits(pb2, 1, 1);
if (interleaved_stats)
s->misc_bits += get_bits_diff(s);
/* motion vectors: 16x8 interlaced mode */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
ff_h263_pred_motion(&s->c, 0, 0, &pred_x, &pred_y);
pred_y /= 2;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
put_bits(&s->pb, 1, s->c.field_select[0][0]);
put_bits(&s->pb, 1, s->c.field_select[0][1]);
ff_h263_encode_motion_vector(s,
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
s->c.mv[0][0][0] - pred_x,
s->c.mv[0][0][1] - pred_y,
s->c.f_code);
ff_h263_encode_motion_vector(s,
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
s->c.mv[0][1][0] - pred_x,
s->c.mv[0][1][1] - pred_y,
s->c.f_code);
} else {
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
av_assert2(s->c.mv_type == MV_TYPE_8X8);
put_bits(&s->pb,
ff_h263_inter_MCBPC_bits[cbpc + 16],
ff_h263_inter_MCBPC_code[cbpc + 16]);
put_bits(pb2, ff_h263_cbpy_tab[cbpy][1], ff_h263_cbpy_tab[cbpy][0]);
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (!s->c.progressive_sequence && cbp)
put_bits(pb2, 1, s->c.interlaced_dct);
if (interleaved_stats)
s->misc_bits += get_bits_diff(s);
for (i = 0; i < 4; i++) {
/* motion vectors: 8x8 mode*/
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
ff_h263_pred_motion(&s->c, i, 0, &pred_x, &pred_y);
ff_h263_encode_motion_vector(s,
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
s->c.cur_pic.motion_val[0][s->c.block_index[i]][0] - pred_x,
s->c.cur_pic.motion_val[0][s->c.block_index[i]][1] - pred_y,
s->c.f_code);
}
}
if (interleaved_stats)
s->mv_bits += get_bits_diff(s);
mpeg4_encode_blocks(s, block, NULL, NULL, NULL, tex_pb);
if (interleaved_stats)
s->p_tex_bits += get_bits_diff(s);
}
} else {
int cbp;
int dc_diff[6]; // dc values with the dc prediction subtracted
int dir[6]; // prediction direction
int zigzag_last_index[6];
const uint8_t *scan_table[6];
int i;
for (int i = 0; i < 6; i++) {
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
int pred = ff_mpeg4_pred_dc(&s->c, i, &dir[i]);
int scale = i < 4 ? s->c.y_dc_scale : s->c.c_dc_scale;
pred = FASTDIV((pred + (scale >> 1)), scale);
dc_diff[i] = block[i][0] - pred;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
s->c.dc_val[0][s->c.block_index[i]] = av_clip_uintp2(block[i][0] * scale, 11);
}
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.avctx->flags & AV_CODEC_FLAG_AC_PRED) {
s->c.ac_pred = decide_ac_pred(s, block, dir, scan_table, zigzag_last_index);
} else {
for (i = 0; i < 6; i++)
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
scan_table[i] = s->c.intra_scantable.permutated;
}
/* compute cbp */
cbp = 0;
for (i = 0; i < 6; i++)
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.block_last_index[i] >= 1)
cbp |= 1 << (5 - i);
cbpc = cbp & 3;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.pict_type == AV_PICTURE_TYPE_I) {
if (s->dquant)
cbpc += 4;
put_bits(&s->pb,
ff_h263_intra_MCBPC_bits[cbpc],
ff_h263_intra_MCBPC_code[cbpc]);
} else {
if (s->dquant)
cbpc += 8;
put_bits(&s->pb, 1, 0); /* mb coded */
put_bits(&s->pb,
ff_h263_inter_MCBPC_bits[cbpc + 4],
ff_h263_inter_MCBPC_code[cbpc + 4]);
}
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
put_bits(pb2, 1, s->c.ac_pred);
cbpy = cbp >> 2;
put_bits(pb2, ff_h263_cbpy_tab[cbpy][1], ff_h263_cbpy_tab[cbpy][0]);
if (s->dquant)
put_bits(dc_pb, 2, dquant_code[s->dquant + 2]);
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (!s->c.progressive_sequence)
put_bits(dc_pb, 1, s->c.interlaced_dct);
if (interleaved_stats)
s->misc_bits += get_bits_diff(s);
mpeg4_encode_blocks(s, block, dc_diff, scan_table, dc_pb, tex_pb);
if (interleaved_stats)
s->i_tex_bits += get_bits_diff(s);
s->i_count++;
/* restore ac coeffs & last_index stuff
* if we messed them up with the prediction */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.ac_pred)
restore_ac_coeffs(s, block, dir, scan_table, zigzag_last_index);
}
}
/**
* add MPEG-4 stuffing bits (01...1)
*/
void ff_mpeg4_stuffing(PutBitContext *pbc)
{
int length = 8 - (put_bits_count(pbc) & 7);
put_bits(pbc, length, (1 << (length - 1)) - 1);
}
/* must be called before writing the header */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
void ff_set_mpeg4_time(MPVEncContext *const s)
{
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.pict_type == AV_PICTURE_TYPE_B) {
ff_mpeg4_init_direct_mv(&s->c);
} else {
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
s->c.last_time_base = s->c.time_base;
s->c.time_base = FFUDIV(s->c.time, s->c.avctx->time_base.den);
}
}
static void mpeg4_encode_gop_header(MPVMainEncContext *const m)
{
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
MPVEncContext *const s = &m->s;
int64_t hours, minutes, seconds;
int64_t time;
put_bits32(&s->pb, GOP_STARTCODE);
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
time = s->c.cur_pic.ptr->f->pts;
if (m->reordered_input_picture[1])
time = FFMIN(time, m->reordered_input_picture[1]->f->pts);
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
time = time * s->c.avctx->time_base.num;
s->c.last_time_base = FFUDIV(time, s->c.avctx->time_base.den);
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
seconds = FFUDIV(time, s->c.avctx->time_base.den);
minutes = FFUDIV(seconds, 60); seconds = FFUMOD(seconds, 60);
hours = FFUDIV(minutes, 60); minutes = FFUMOD(minutes, 60);
hours = FFUMOD(hours , 24);
put_bits(&s->pb, 5, hours);
put_bits(&s->pb, 6, minutes);
put_bits(&s->pb, 1, 1);
put_bits(&s->pb, 6, seconds);
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
put_bits(&s->pb, 1, !!(s->c.avctx->flags & AV_CODEC_FLAG_CLOSED_GOP));
put_bits(&s->pb, 1, 0); // broken link == NO
ff_mpeg4_stuffing(&s->pb);
}
static void mpeg4_encode_visual_object_header(MPVMainEncContext *const m)
{
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
MPVEncContext *const s = &m->s;
int profile_and_level_indication;
int vo_ver_id;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.avctx->profile != AV_PROFILE_UNKNOWN) {
profile_and_level_indication = s->c.avctx->profile << 4;
} else if (m->max_b_frames || s->c.quarter_sample) {
profile_and_level_indication = 0xF0; // adv simple
} else {
profile_and_level_indication = 0x00; // simple
}
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.avctx->level != AV_LEVEL_UNKNOWN)
profile_and_level_indication |= s->c.avctx->level;
else
profile_and_level_indication |= 1; // level 1
if (profile_and_level_indication >> 4 == 0xF)
vo_ver_id = 5;
else
vo_ver_id = 1;
// FIXME levels
put_bits32(&s->pb, VOS_STARTCODE);
put_bits(&s->pb, 8, profile_and_level_indication);
put_bits32(&s->pb, VISUAL_OBJ_STARTCODE);
put_bits(&s->pb, 1, 1);
put_bits(&s->pb, 4, vo_ver_id);
put_bits(&s->pb, 3, 1); // priority
put_bits(&s->pb, 4, 1); // visual obj type== video obj
put_bits(&s->pb, 1, 0); // video signal type == no clue // FIXME
ff_mpeg4_stuffing(&s->pb);
}
static void mpeg4_encode_vol_header(Mpeg4EncContext *const m4,
int vo_number,
int vol_number)
{
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
MPVEncContext *const s = &m4->m.s;
int vo_ver_id, vo_type, aspect_ratio_info;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (m4->m.max_b_frames || s->c.quarter_sample) {
vo_ver_id = 5;
vo_type = ADV_SIMPLE_VO_TYPE;
} else {
vo_ver_id = 1;
vo_type = SIMPLE_VO_TYPE;
}
put_bits32(&s->pb, 0x100 + vo_number); /* video obj */
put_bits32(&s->pb, 0x120 + vol_number); /* video obj layer */
put_bits(&s->pb, 1, 0); /* random access vol */
put_bits(&s->pb, 8, vo_type); /* video obj type indication */
put_bits(&s->pb, 1, 1); /* is obj layer id= yes */
put_bits(&s->pb, 4, vo_ver_id); /* is obj layer ver id */
put_bits(&s->pb, 3, 1); /* is obj layer priority */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
aspect_ratio_info = ff_h263_aspect_to_info(s->c.avctx->sample_aspect_ratio);
put_bits(&s->pb, 4, aspect_ratio_info); /* aspect ratio info */
if (aspect_ratio_info == FF_ASPECT_EXTENDED) {
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
av_reduce(&s->c.avctx->sample_aspect_ratio.num, &s->c.avctx->sample_aspect_ratio.den,
s->c.avctx->sample_aspect_ratio.num, s->c.avctx->sample_aspect_ratio.den, 255);
put_bits(&s->pb, 8, s->c.avctx->sample_aspect_ratio.num);
put_bits(&s->pb, 8, s->c.avctx->sample_aspect_ratio.den);
}
put_bits(&s->pb, 1, 1); /* vol control parameters= yes */
put_bits(&s->pb, 2, 1); /* chroma format YUV 420/YV12 */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
put_bits(&s->pb, 1, s->c.low_delay);
put_bits(&s->pb, 1, 0); /* vbv parameters= no */
put_bits(&s->pb, 2, RECT_SHAPE); /* vol shape= rectangle */
put_bits(&s->pb, 1, 1); /* marker bit */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
put_bits(&s->pb, 16, s->c.avctx->time_base.den);
if (m4->time_increment_bits < 1)
m4->time_increment_bits = 1;
put_bits(&s->pb, 1, 1); /* marker bit */
put_bits(&s->pb, 1, 0); /* fixed vop rate=no */
put_bits(&s->pb, 1, 1); /* marker bit */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
put_bits(&s->pb, 13, s->c.width); /* vol width */
put_bits(&s->pb, 1, 1); /* marker bit */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
put_bits(&s->pb, 13, s->c.height); /* vol height */
put_bits(&s->pb, 1, 1); /* marker bit */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
put_bits(&s->pb, 1, s->c.progressive_sequence ? 0 : 1);
put_bits(&s->pb, 1, 1); /* obmc disable */
if (vo_ver_id == 1)
put_bits(&s->pb, 1, 0); /* sprite enable */
else
put_bits(&s->pb, 2, 0); /* sprite enable */
put_bits(&s->pb, 1, 0); /* not 8 bit == false */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
put_bits(&s->pb, 1, s->c.mpeg_quant); /* quant type = (0 = H.263 style) */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.mpeg_quant) {
ff_write_quant_matrix(&s->pb, s->c.avctx->intra_matrix);
ff_write_quant_matrix(&s->pb, s->c.avctx->inter_matrix);
}
if (vo_ver_id != 1)
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
put_bits(&s->pb, 1, s->c.quarter_sample);
put_bits(&s->pb, 1, 1); /* complexity estimation disable */
put_bits(&s->pb, 1, s->rtp_mode ? 0 : 1); /* resync marker disable */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
put_bits(&s->pb, 1, s->c.data_partitioning ? 1 : 0);
if (s->c.data_partitioning)
put_bits(&s->pb, 1, 0); /* no rvlc */
if (vo_ver_id != 1) {
put_bits(&s->pb, 1, 0); /* newpred */
put_bits(&s->pb, 1, 0); /* reduced res vop */
}
put_bits(&s->pb, 1, 0); /* scalability */
ff_mpeg4_stuffing(&s->pb);
/* user data */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (!(s->c.avctx->flags & AV_CODEC_FLAG_BITEXACT)) {
put_bits32(&s->pb, USER_DATA_STARTCODE);
ff_put_string(&s->pb, LIBAVCODEC_IDENT, 0);
}
}
/* write MPEG-4 VOP header */
static int mpeg4_encode_picture_header(MPVMainEncContext *const m)
{
Mpeg4EncContext *const m4 = mainctx_to_mpeg4(m);
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
MPVEncContext *const s = &m->s;
uint64_t time_incr;
int64_t time_div, time_mod;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.pict_type == AV_PICTURE_TYPE_I) {
if (!(s->c.avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)) {
if (s->c.avctx->strict_std_compliance < FF_COMPLIANCE_VERY_STRICT) // HACK, the reference sw is buggy
mpeg4_encode_visual_object_header(m);
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.avctx->strict_std_compliance < FF_COMPLIANCE_VERY_STRICT || s->c.picture_number == 0) // HACK, the reference sw is buggy
mpeg4_encode_vol_header(m4, 0, 0);
}
mpeg4_encode_gop_header(m);
}
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
s->c.partitioned_frame = s->c.data_partitioning && s->c.pict_type != AV_PICTURE_TYPE_B;
put_bits32(&s->pb, VOP_STARTCODE); /* vop header */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
put_bits(&s->pb, 2, s->c.pict_type - 1); /* pict type: I = 0 , P = 1 */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
time_div = FFUDIV(s->c.time, s->c.avctx->time_base.den);
time_mod = FFUMOD(s->c.time, s->c.avctx->time_base.den);
time_incr = time_div - s->c.last_time_base;
// This limits the frame duration to max 1 day
if (time_incr > 3600*24) {
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
av_log(s->c.avctx, AV_LOG_ERROR, "time_incr %"PRIu64" too large\n", time_incr);
return AVERROR(EINVAL);
}
while (time_incr--)
put_bits(&s->pb, 1, 1);
put_bits(&s->pb, 1, 0);
put_bits(&s->pb, 1, 1); /* marker */
put_bits(&s->pb, m4->time_increment_bits, time_mod); /* time increment */
put_bits(&s->pb, 1, 1); /* marker */
put_bits(&s->pb, 1, 1); /* vop coded */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.pict_type == AV_PICTURE_TYPE_P) {
put_bits(&s->pb, 1, s->c.no_rounding); /* rounding type */
}
put_bits(&s->pb, 3, 0); /* intra dc VLC threshold */
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (!s->c.progressive_sequence) {
put_bits(&s->pb, 1, !!(s->c.cur_pic.ptr->f->flags & AV_FRAME_FLAG_TOP_FIELD_FIRST));
put_bits(&s->pb, 1, s->c.alternate_scan);
}
// FIXME sprite stuff
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
put_bits(&s->pb, 5, s->c.qscale);
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.pict_type != AV_PICTURE_TYPE_I)
put_bits(&s->pb, 3, s->c.f_code); /* fcode_for */
if (s->c.pict_type == AV_PICTURE_TYPE_B)
put_bits(&s->pb, 3, s->c.b_code); /* fcode_back */
return 0;
}
static av_cold void init_uni_dc_tab(void)
{
int level, uni_code, uni_len;
for (level = -256; level < 256; level++) {
int size, v, l;
/* find number of bits */
size = 0;
v = abs(level);
while (v) {
v >>= 1;
size++;
}
if (level < 0)
l = (-level) ^ ((1 << size) - 1);
else
l = level;
/* luminance */
uni_code = ff_mpeg4_DCtab_lum[size][0];
uni_len = ff_mpeg4_DCtab_lum[size][1];
if (size > 0) {
uni_code <<= size;
uni_code |= l;
uni_len += size;
if (size > 8) {
uni_code <<= 1;
uni_code |= 1;
uni_len++;
}
}
uni_DCtab_lum_bits[level + 256] = uni_code;
uni_DCtab_lum_len[level + 256] = uni_len;
/* chrominance */
uni_code = ff_mpeg4_DCtab_chrom[size][0];
uni_len = ff_mpeg4_DCtab_chrom[size][1];
if (size > 0) {
uni_code <<= size;
uni_code |= l;
uni_len += size;
if (size > 8) {
uni_code <<= 1;
uni_code |= 1;
uni_len++;
}
}
uni_DCtab_chrom_bits[level + 256] = uni_code;
uni_DCtab_chrom_len[level + 256] = uni_len;
}
}
static av_cold void init_uni_mpeg4_rl_tab(RLTable *rl, uint32_t *bits_tab,
uint8_t *len_tab)
{
int slevel, run, last;
av_assert0(MAX_LEVEL >= 64);
av_assert0(MAX_RUN >= 63);
for (slevel = -64; slevel < 64; slevel++) {
if (slevel == 0)
continue;
for (run = 0; run < 64; run++) {
for (last = 0; last <= 1; last++) {
const int index = UNI_MPEG4_ENC_INDEX(last, run, slevel + 64);
int level = slevel < 0 ? -slevel : slevel;
int sign = slevel < 0 ? 1 : 0;
int bits, len, code;
int level1, run1;
len_tab[index] = 100;
/* ESC0 */
code = get_rl_index(rl, last, run, level);
bits = rl->table_vlc[code][0];
len = rl->table_vlc[code][1];
bits = bits * 2 + sign;
len++;
if (code != rl->n && len < len_tab[index]) {
bits_tab[index] = bits;
len_tab[index] = len;
}
/* ESC1 */
bits = rl->table_vlc[rl->n][0];
len = rl->table_vlc[rl->n][1];
bits = bits * 2;
len++; // esc1
level1 = level - rl->max_level[last][run];
if (level1 > 0) {
code = get_rl_index(rl, last, run, level1);
bits <<= rl->table_vlc[code][1];
len += rl->table_vlc[code][1];
bits += rl->table_vlc[code][0];
bits = bits * 2 + sign;
len++;
if (code != rl->n && len < len_tab[index]) {
bits_tab[index] = bits;
len_tab[index] = len;
}
}
/* ESC2 */
bits = rl->table_vlc[rl->n][0];
len = rl->table_vlc[rl->n][1];
bits = bits * 4 + 2;
len += 2; // esc2
run1 = run - rl->max_run[last][level] - 1;
if (run1 >= 0) {
code = get_rl_index(rl, last, run1, level);
bits <<= rl->table_vlc[code][1];
len += rl->table_vlc[code][1];
bits += rl->table_vlc[code][0];
bits = bits * 2 + sign;
len++;
if (code != rl->n && len < len_tab[index]) {
bits_tab[index] = bits;
len_tab[index] = len;
}
}
/* ESC3 */
bits = rl->table_vlc[rl->n][0];
len = rl->table_vlc[rl->n][1];
bits = bits * 4 + 3;
len += 2; // esc3
bits = bits * 2 + last;
len++;
bits = bits * 64 + run;
len += 6;
bits = bits * 2 + 1;
len++; // marker
bits = bits * 4096 + (slevel & 0xfff);
len += 12;
bits = bits * 2 + 1;
len++; // marker
if (len < len_tab[index]) {
bits_tab[index] = bits;
len_tab[index] = len;
}
}
}
}
}
static av_cold void mpeg4_encode_init_static(void)
{
init_uni_dc_tab();
ff_mpeg4_init_rl_intra();
init_uni_mpeg4_rl_tab(&ff_mpeg4_rl_intra, uni_mpeg4_intra_rl_bits, uni_mpeg4_intra_rl_len);
init_uni_mpeg4_rl_tab(&ff_h263_rl_inter, uni_mpeg4_inter_rl_bits, uni_mpeg4_inter_rl_len);
for (int f_code = MAX_FCODE; f_code > 0; f_code--) {
for (int mv = -(16 << f_code); mv < (16 << f_code); mv++)
fcode_tab[mv + MAX_MV] = f_code;
}
}
static av_cold int encode_init(AVCodecContext *avctx)
{
static AVOnce init_static_once = AV_ONCE_INIT;
Mpeg4EncContext *const m4 = avctx->priv_data;
MPVMainEncContext *const m = &m4->m;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
MPVEncContext *const s = &m->s;
int ret;
if (avctx->width >= (1<<13) || avctx->height >= (1<<13)) {
av_log(avctx, AV_LOG_ERROR, "dimensions too large for MPEG-4\n");
return AVERROR(EINVAL);
}
m->encode_picture_header = mpeg4_encode_picture_header;
s->encode_mb = mpeg4_encode_mb;
m->fcode_tab = fcode_tab + MAX_MV;
s->min_qcoeff = -2048;
s->max_qcoeff = 2047;
s->intra_ac_vlc_length = uni_mpeg4_intra_rl_len;
s->intra_ac_vlc_last_length = uni_mpeg4_intra_rl_len + 128 * 64;
s->inter_ac_vlc_length = uni_mpeg4_inter_rl_len;
s->inter_ac_vlc_last_length = uni_mpeg4_inter_rl_len + 128 * 64;
s->luma_dc_vlc_length = uni_DCtab_lum_len;
s->ac_esc_length = 7 + 2 + 1 + 6 + 1 + 12 + 1;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
s->c.y_dc_scale_table = ff_mpeg4_y_dc_scale_table;
s->c.c_dc_scale_table = ff_mpeg4_c_dc_scale_table;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
ff_qpeldsp_init(&s->c.qdsp);
if ((ret = ff_mpv_encode_init(avctx)) < 0)
return ret;
ff_thread_once(&init_static_once, mpeg4_encode_init_static);
if (avctx->time_base.den > (1 << 16) - 1) {
av_log(avctx, AV_LOG_ERROR,
"timebase %d/%d not supported by MPEG 4 standard, "
"the maximum admitted value for the timebase denominator "
"is %d\n", avctx->time_base.num, avctx->time_base.den,
(1 << 16) - 1);
return AVERROR(EINVAL);
}
m4->time_increment_bits = av_log2(avctx->time_base.den - 1) + 1;
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
avctx->extradata = av_malloc(1024);
if (!avctx->extradata)
return AVERROR(ENOMEM);
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
init_put_bits(&s->pb, avctx->extradata, 1024);
mpeg4_encode_visual_object_header(m);
mpeg4_encode_vol_header(m4, 0, 0);
// ff_mpeg4_stuffing(&s->pb); ?
flush_put_bits(&s->pb);
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
avctx->extradata_size = put_bytes_output(&s->pb);
}
return 0;
}
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
void ff_mpeg4_init_partitions(MPVEncContext *const s)
{
uint8_t *start = put_bits_ptr(&s->pb);
uint8_t *end = s->pb.buf_end;
int size = end - start;
int pb_size = (((intptr_t)start + size / 3) & (~3)) - (intptr_t)start;
int tex_size = (size - 2 * pb_size) & (~3);
set_put_bits_buffer_size(&s->pb, pb_size);
init_put_bits(&s->tex_pb, start + pb_size, tex_size);
init_put_bits(&s->pb2, start + pb_size + tex_size, pb_size);
}
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
void ff_mpeg4_merge_partitions(MPVEncContext *const s)
{
const int pb2_len = put_bits_count(&s->pb2);
const int tex_pb_len = put_bits_count(&s->tex_pb);
const int bits = put_bits_count(&s->pb);
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
if (s->c.pict_type == AV_PICTURE_TYPE_I) {
put_bits(&s->pb, 19, DC_MARKER);
s->misc_bits += 19 + pb2_len + bits - s->last_bits;
s->i_tex_bits += tex_pb_len;
} else {
put_bits(&s->pb, 17, MOTION_MARKER);
s->misc_bits += 17 + pb2_len;
s->mv_bits += bits - s->last_bits;
s->p_tex_bits += tex_pb_len;
}
flush_put_bits(&s->pb2);
flush_put_bits(&s->tex_pb);
set_put_bits_buffer_size(&s->pb, s->pb2.buf_end - s->pb.buf);
ff_copy_bits(&s->pb, s->pb2.buf, pb2_len);
ff_copy_bits(&s->pb, s->tex_pb.buf, tex_pb_len);
s->last_bits = put_bits_count(&s->pb);
}
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
void ff_mpeg4_encode_video_packet_header(MPVEncContext *const s)
{
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
int mb_num_bits = av_log2(s->c.mb_num - 1) + 1;
put_bits(&s->pb, ff_mpeg4_get_video_packet_prefix_length(s->c.pict_type, s->c.f_code, s->c.b_code), 0);
put_bits(&s->pb, 1, 1);
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
put_bits(&s->pb, mb_num_bits, s->c.mb_x + s->c.mb_y * s->c.mb_width);
put_bits(&s->pb, 5 /* quant_precision */, s->c.qscale);
put_bits(&s->pb, 1, 0); /* no HEC */
}
avcodec/mpegvideoenc: Add MPVEncContext Many of the fields of MpegEncContext (which is also used by decoders) are actually only used by encoders. Therefore this commit adds a new encoder-only structure and moves all of the encoder-only fields to it except for those which require more explicit synchronisation between the main slice context and the other slice contexts. This synchronisation is currently mainly provided by ff_update_thread_context() which simply copies most of the main slice context over the other slice contexts. Fields which are moved to the new MPVEncContext no longer participate in this (which is desired, because it is horrible and for the fields b) below wasteful) which means that some fields can only be moved when explicit synchronisation code is added in later commits. More explicitly, this commit moves the following fields: a) Fields not copied by ff_update_duplicate_context(): dct_error_sum and dct_count; the former does not need synchronisation, the latter is synchronised in merge_context_after_encode(). b) Fields which do not change after initialisation (these fields could also be put into MPVMainEncContext at the cost of an indirection to access them): lambda_table, adaptive_quant, {luma,chroma}_elim_threshold, new_pic, fdsp, mpvencdsp, pdsp, {p,b_forw,b_back,b_bidir_forw,b_bidir_back,b_direct,b_field}_mv_table, [pb]_field_select_table, mb_{type,var,mean}, mc_mb_var, {min,max}_qcoeff, {inter,intra}_quant_bias, ac_esc_length, the *_vlc_length fields, the q_{intra,inter,chroma_intra}_matrix{,16}, dct_offset, mb_info, mjpeg_ctx, rtp_mode, rtp_payload_size, encode_mb, all function pointers, mpv_flags, quantizer_noise_shaping, frame_reconstruction_bitfield, error_rate and intra_penalty. c) Fields which are already (re)set explicitly: The PutBitContexts pb, tex_pb, pb2; dquant, skipdct, encoding_error, the statistics fields {mv,i_tex,p_tex,misc,last}_bits and i_count; last_mv_dir, esc_pos (reset when writing the header). d) Fields which are only used by encoders not supporting slice threading for which synchronisation doesn't matter: esc3_level_length and the remaining mb_info fields. e) coded_score: This field is only really used when FF_MPV_FLAG_CBP_RD is set (which implies trellis) and even then it is only used for non-intra blocks. For these blocks dct_quantize_trellis_c() either sets coded_score[n] or returns a last_non_zero value of -1 in which case coded_score will be reset in encode_mb_internal(). Therefore no old values are ever used. The MotionEstContext has not been moved yet. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2025-03-19 08:11:01 +01:00
#define OFFSET(x) offsetof(MPVEncContext, c.x)
#define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
static const AVOption options[] = {
{ "data_partitioning", "Use data partitioning.", OFFSET(data_partitioning), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
{ "alternate_scan", "Enable alternate scantable.", OFFSET(alternate_scan), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
{ "mpeg_quant", "Use MPEG quantizers instead of H.263",
OFFSET(mpeg_quant), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, 1, VE },
FF_MPV_COMMON_BFRAME_OPTS
FF_MPV_COMMON_OPTS
FF_MPV_COMMON_MOTION_EST_OPTS
FF_MPEG4_PROFILE_OPTS
{ NULL },
};
static const AVClass mpeg4enc_class = {
.class_name = "MPEG4 encoder",
.item_name = av_default_item_name,
.option = options,
.version = LIBAVUTIL_VERSION_INT,
};
const FFCodec ff_mpeg4_encoder = {
.p.name = "mpeg4",
CODEC_LONG_NAME("MPEG-4 part 2"),
.p.type = AVMEDIA_TYPE_VIDEO,
.p.id = AV_CODEC_ID_MPEG4,
.priv_data_size = sizeof(Mpeg4EncContext),
.init = encode_init,
FF_CODEC_ENCODE_CB(ff_mpv_encode_picture),
.close = ff_mpv_encode_end,
CODEC_PIXFMTS(AV_PIX_FMT_YUV420P),
avcodec/internal: add FFCodec.color_ranges I went through all codecs and put them into five basic categories: 1. JPEG range only 2. MPEG range only 3. Explicitly tagged 4. Broken (codec supports both but encoder ignores tags) 5. N/A (headerless or pseudo-formats) Filters in category 5 remain untouched. The rest gain an explicit assignment of their supported color ranges, with codecs in category 4 being set to MPEG-only for safety. It might be considered redundant to distinguish between 0 (category 5) and MPEG+JPEG (category 3), but in doing so we effectively communicate that we can guarantee that these tags will be encoded, which is distinct from the situation where there are some codecs that simply don't have tagging or implied semantics (e.g. rawvideo). A full list of codecs follows: JPEG range only: - amv - roqvideo MPEG range only: - asv1, asv2 - avui - cfhd - cljr - dnxhd - dvvideo - ffv1 - flv - h261, h263, h263p - {h263,vp8}_v4l2m2m - huffyuv, ffvhuff - jpeg2000 - libopenjpeg - libtheora - libwebp, libwebp_anim - libx262 - libxavs, libxavs2 - libxvid - mpeg1video, mpeg2video - mpeg2_qsv - mpeg2_vaapi - mpeg4, msmpeg4, msmpeg4v2, wmv1, wmv2 - mpeg4_omx - prores, prores_aw, prores_ks - rv10, rv20 - snow - speedhq - svq1 - tiff - utvideo Explicitly tagged (MPEG/JPEG): - {av1,h264,hevc}_nvenc - {av1,h264,hevc}_vaapi - {av1,h264,hevc,vp8,vp9,mpeg4}_mediacodec - {av1,h264,hevc,vp9}_qsv - h264_amf - {h264,hevc,prores}_videotoolbox - libaom-av1 - libkvazaar - libopenh264 - librav1e - libsvtav1 - libvpx, libvpx-vp9 - libx264 - libx265 - ljpeg - mjpeg - vc2 Broken (encoder ignores tags): - {av1,hevc}_amf - {h264,hevc,mpeg4}_v4l2m2m - h264_omx - libxeve - magicyuv - {vp8,vp9,mjpeg}_vaapi N/A: - ayuv, yuv4, y41p, v308, v210, v410, v408 (headerless) - pgmyuv (headerless) - rawvideo, bitpacked (headerless) - vnull, wrapped_avframe (pseudocodecs)
2023-10-11 16:09:33 +02:00
.color_ranges = AVCOL_RANGE_MPEG,
.p.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_DELAY |
AV_CODEC_CAP_SLICE_THREADS |
AV_CODEC_CAP_ENCODER_REORDERED_OPAQUE,
.caps_internal = FF_CODEC_CAP_INIT_CLEANUP,
.p.priv_class = &mpeg4enc_class,
};