/***************************************************-*- coding: iso-8859-1 -*- * ratecontrol.c: h264 encoder library (Rate Control) ***************************************************************************** * Copyright (C) 2005 x264 project * $Id: ratecontrol.c,v 1.1 2004/06/03 19:27:08 fenrir Exp $ * * Authors: Loren Merritt * Michael Niedermayer * Måns Rullgård * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by * the Free Software Foundation; either version 2 of the License, or * (at your option) any later version. * * This program 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 General Public License for more details. * * You should have received a copy of the GNU General Public License * along with this program; if not, write to the Free Software * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA. *****************************************************************************/ #define _ISOC99_SOURCE #undef NDEBUG // always check asserts, the speed effect is far too small to disable them #include #include #include #include "common/common.h" #include "common/cpu.h" #include "ratecontrol.h" typedef struct { int pict_type; int kept_as_ref; float qscale; int mv_bits; int i_tex_bits; int p_tex_bits; int misc_bits; uint64_t expected_bits; float new_qscale; int new_qp; int i_count; int p_count; int s_count; float blurred_complexity; char direct_mode; } ratecontrol_entry_t; typedef struct { double coeff; double count; double decay; } predictor_t; struct x264_ratecontrol_t { /* constants */ int b_abr; int b_2pass; int b_vbv; int b_vbv_min_rate; double fps; double bitrate; double rate_tolerance; int nmb; /* number of macroblocks in a frame */ int qp_constant[5]; /* current frame */ ratecontrol_entry_t *rce; int qp; /* qp for current frame */ int qpm; /* qp for current macroblock */ float qpa; /* average of macroblocks' qp */ int qp_force; /* VBV stuff */ double buffer_size; double buffer_fill_final; /* real buffer as of the last finished frame */ double buffer_fill; /* planned buffer, if all in-progress frames hit their bit budget */ double buffer_rate; /* # of bits added to buffer_fill after each frame */ predictor_t *pred; /* predict frame size from satd */ /* ABR stuff */ int last_satd; double last_rceq; double cplxr_sum; /* sum of bits*qscale/rceq */ double expected_bits_sum; /* sum of qscale2bits after rceq, ratefactor, and overflow */ double wanted_bits_window; /* target bitrate * window */ double cbr_decay; double short_term_cplxsum; double short_term_cplxcount; double rate_factor_constant; double ip_offset; double pb_offset; /* 2pass stuff */ FILE *p_stat_file_out; char *psz_stat_file_tmpname; int num_entries; /* number of ratecontrol_entry_ts */ ratecontrol_entry_t *entry; /* FIXME: copy needed data and free this once init is done */ double last_qscale; double last_qscale_for[5]; /* last qscale for a specific pict type, used for max_diff & ipb factor stuff */ int last_non_b_pict_type; double accum_p_qp; /* for determining I-frame quant */ double accum_p_norm; double last_accum_p_norm; double lmin[5]; /* min qscale by frame type */ double lmax[5]; double lstep; /* max change (multiply) in qscale per frame */ double i_cplx_sum[5]; /* estimated total texture bits in intra MBs at qscale=1 */ double p_cplx_sum[5]; double mv_bits_sum[5]; int frame_count[5]; /* number of frames of each type */ /* MBRC stuff */ double frame_size_planned; predictor_t *row_pred; predictor_t row_preds[5]; predictor_t *pred_b_from_p; /* predict B-frame size from P-frame satd */ int bframes; /* # consecutive B-frames before this P-frame */ int bframe_bits; /* total cost of those frames */ int i_zones; x264_zone_t *zones; x264_zone_t *prev_zone; }; static int parse_zones( x264_t *h ); static int init_pass2(x264_t *); static float rate_estimate_qscale( x264_t *h ); static void update_vbv( x264_t *h, int bits ); static void update_vbv_plan( x264_t *h ); static double predict_size( predictor_t *p, double q, double var ); static void update_predictor( predictor_t *p, double q, double var, double bits ); int x264_rc_analyse_slice( x264_t *h ); /* Terminology: * qp = h.264's quantizer * qscale = linearized quantizer = Lagrange multiplier */ static inline double qp2qscale(double qp) { return 0.85 * pow(2.0, ( qp - 12.0 ) / 6.0); } static inline double qscale2qp(double qscale) { return 12.0 + 6.0 * log(qscale/0.85) / log(2.0); } /* Texture bitrate is not quite inversely proportional to qscale, * probably due the the changing number of SKIP blocks. * MV bits level off at about qp<=12, because the lambda used * for motion estimation is constant there. */ static inline double qscale2bits(ratecontrol_entry_t *rce, double qscale) { if(qscale<0.1) qscale = 0.1; return (rce->i_tex_bits + rce->p_tex_bits + .1) * pow( rce->qscale / qscale, 1.1 ) + rce->mv_bits * pow( X264_MAX(rce->qscale, 1) / X264_MAX(qscale, 1), 0.5 ) + rce->misc_bits; } int x264_ratecontrol_new( x264_t *h ) { x264_ratecontrol_t *rc; int i; x264_cpu_restore( h->param.cpu ); rc = h->rc = x264_malloc( h->param.i_threads * sizeof(x264_ratecontrol_t) ); memset( rc, 0, h->param.i_threads * sizeof(x264_ratecontrol_t) ); rc->b_abr = h->param.rc.i_rc_method != X264_RC_CQP && !h->param.rc.b_stat_read; rc->b_2pass = h->param.rc.i_rc_method == X264_RC_ABR && h->param.rc.b_stat_read; /* FIXME: use integers */ if(h->param.i_fps_num > 0 && h->param.i_fps_den > 0) rc->fps = (float) h->param.i_fps_num / h->param.i_fps_den; else rc->fps = 25.0; rc->bitrate = h->param.rc.i_bitrate * 1000.; rc->rate_tolerance = h->param.rc.f_rate_tolerance; rc->nmb = h->mb.i_mb_count; rc->last_non_b_pict_type = -1; rc->cbr_decay = 1.0; if( h->param.rc.i_rc_method == X264_RC_CRF && h->param.rc.b_stat_read ) { x264_log(h, X264_LOG_ERROR, "constant rate-factor is incompatible with 2pass.\n"); return -1; } if( h->param.rc.i_vbv_buffer_size ) { if( h->param.rc.i_rc_method == X264_RC_CQP ) x264_log(h, X264_LOG_WARNING, "VBV is incompatible with constant QP, ignored.\n"); else if( h->param.rc.i_vbv_max_bitrate == 0 ) { x264_log( h, X264_LOG_DEBUG, "VBV maxrate unspecified, assuming CBR\n" ); h->param.rc.i_vbv_max_bitrate = h->param.rc.i_bitrate; } } if( h->param.rc.i_vbv_max_bitrate < h->param.rc.i_bitrate && h->param.rc.i_vbv_max_bitrate > 0) x264_log(h, X264_LOG_WARNING, "max bitrate less than average bitrate, ignored.\n"); else if( h->param.rc.i_vbv_max_bitrate > 0 && h->param.rc.i_vbv_buffer_size > 0 ) { if( h->param.rc.i_vbv_buffer_size < 3 * h->param.rc.i_vbv_max_bitrate / rc->fps ) { h->param.rc.i_vbv_buffer_size = 3 * h->param.rc.i_vbv_max_bitrate / rc->fps; x264_log( h, X264_LOG_WARNING, "VBV buffer size too small, using %d kbit\n", h->param.rc.i_vbv_buffer_size ); } if( h->param.rc.f_vbv_buffer_init > 1. ) h->param.rc.f_vbv_buffer_init = x264_clip3f( h->param.rc.f_vbv_buffer_init / h->param.rc.i_vbv_buffer_size, 0, 1 ); rc->buffer_rate = h->param.rc.i_vbv_max_bitrate * 1000. / rc->fps; rc->buffer_size = h->param.rc.i_vbv_buffer_size * 1000.; rc->buffer_fill_final = rc->buffer_size * h->param.rc.f_vbv_buffer_init; rc->cbr_decay = 1.0 - rc->buffer_rate / rc->buffer_size * 0.5 * X264_MAX(0, 1.5 - rc->buffer_rate * rc->fps / rc->bitrate); rc->b_vbv = 1; rc->b_vbv_min_rate = !rc->b_2pass && h->param.rc.i_rc_method == X264_RC_ABR && h->param.rc.i_vbv_max_bitrate <= h->param.rc.i_bitrate; } else if( h->param.rc.i_vbv_max_bitrate ) { x264_log(h, X264_LOG_WARNING, "VBV maxrate specified, but no bufsize.\n"); h->param.rc.i_vbv_max_bitrate = 0; } if(rc->rate_tolerance < 0.01) { x264_log(h, X264_LOG_WARNING, "bitrate tolerance too small, using .01\n"); rc->rate_tolerance = 0.01; } h->mb.b_variable_qp = rc->b_vbv && !rc->b_2pass; if( rc->b_abr ) { /* FIXME ABR_INIT_QP is actually used only in CRF */ #define ABR_INIT_QP ( h->param.rc.i_rc_method == X264_RC_CRF ? h->param.rc.f_rf_constant : 24 ) rc->accum_p_norm = .01; rc->accum_p_qp = ABR_INIT_QP * rc->accum_p_norm; /* estimated ratio that produces a reasonable QP for the first I-frame */ rc->cplxr_sum = .01 * pow( 7.0e5, h->param.rc.f_qcompress ) * pow( h->mb.i_mb_count, 0.5 ); rc->wanted_bits_window = 1.0 * rc->bitrate / rc->fps; rc->last_non_b_pict_type = SLICE_TYPE_I; } if( h->param.rc.i_rc_method == X264_RC_CRF ) { /* arbitrary rescaling to make CRF somewhat similar to QP */ double base_cplx = h->mb.i_mb_count * (h->param.i_bframe ? 120 : 80); rc->rate_factor_constant = pow( base_cplx, 1 - h->param.rc.f_qcompress ) / qp2qscale( h->param.rc.f_rf_constant ); } rc->ip_offset = 6.0 * log(h->param.rc.f_ip_factor) / log(2.0); rc->pb_offset = 6.0 * log(h->param.rc.f_pb_factor) / log(2.0); rc->qp_constant[SLICE_TYPE_P] = h->param.rc.i_qp_constant; rc->qp_constant[SLICE_TYPE_I] = x264_clip3( h->param.rc.i_qp_constant - rc->ip_offset + 0.5, 0, 51 ); rc->qp_constant[SLICE_TYPE_B] = x264_clip3( h->param.rc.i_qp_constant + rc->pb_offset + 0.5, 0, 51 ); rc->lstep = pow( 2, h->param.rc.i_qp_step / 6.0 ); rc->last_qscale = qp2qscale(26); rc->pred = x264_malloc( 5*sizeof(predictor_t) ); rc->pred_b_from_p = x264_malloc( sizeof(predictor_t) ); for( i = 0; i < 5; i++ ) { rc->last_qscale_for[i] = qp2qscale( ABR_INIT_QP ); rc->lmin[i] = qp2qscale( h->param.rc.i_qp_min ); rc->lmax[i] = qp2qscale( h->param.rc.i_qp_max ); rc->pred[i].coeff= 2.0; rc->pred[i].count= 1.0; rc->pred[i].decay= 0.5; rc->row_preds[i].coeff= .25; rc->row_preds[i].count= 1.0; rc->row_preds[i].decay= 0.5; } *rc->pred_b_from_p = rc->pred[0]; if( parse_zones( h ) < 0 ) { x264_log( h, X264_LOG_ERROR, "failed to parse zones\n" ); return -1; } /* Load stat file and init 2pass algo */ if( h->param.rc.b_stat_read ) { char *p, *stats_in, *stats_buf; /* read 1st pass stats */ assert( h->param.rc.psz_stat_in ); stats_buf = stats_in = x264_slurp_file( h->param.rc.psz_stat_in ); if( !stats_buf ) { x264_log(h, X264_LOG_ERROR, "ratecontrol_init: can't open stats file\n"); return -1; } /* check whether 1st pass options were compatible with current options */ if( !strncmp( stats_buf, "#options:", 9 ) ) { int i; char *opts = stats_buf; stats_in = strchr( stats_buf, '\n' ); if( !stats_in ) return -1; *stats_in = '\0'; stats_in++; if( ( p = strstr( opts, "bframes=" ) ) && sscanf( p, "bframes=%d", &i ) && h->param.i_bframe != i ) { x264_log( h, X264_LOG_ERROR, "different number of B-frames than 1st pass (%d vs %d)\n", h->param.i_bframe, i ); return -1; } /* since B-adapt doesn't (yet) take into account B-pyramid, * the converse is not a problem */ if( strstr( opts, "b_pyramid=1" ) && !h->param.b_bframe_pyramid ) x264_log( h, X264_LOG_WARNING, "1st pass used B-pyramid, 2nd doesn't\n" ); if( ( p = strstr( opts, "keyint=" ) ) && sscanf( p, "keyint=%d", &i ) && h->param.i_keyint_max != i ) x264_log( h, X264_LOG_WARNING, "different keyint than 1st pass (%d vs %d)\n", h->param.i_keyint_max, i ); if( strstr( opts, "qp=0" ) && h->param.rc.i_rc_method == X264_RC_ABR ) x264_log( h, X264_LOG_WARNING, "1st pass was lossless, bitrate prediction will be inaccurate\n" ); } /* find number of pics */ p = stats_in; for(i=-1; p; i++) p = strchr(p+1, ';'); if(i==0) { x264_log(h, X264_LOG_ERROR, "empty stats file\n"); return -1; } rc->num_entries = i; if( h->param.i_frame_total < rc->num_entries && h->param.i_frame_total > 0 ) { x264_log( h, X264_LOG_WARNING, "2nd pass has fewer frames than 1st pass (%d vs %d)\n", h->param.i_frame_total, rc->num_entries ); } if( h->param.i_frame_total > rc->num_entries + h->param.i_bframe ) { x264_log( h, X264_LOG_ERROR, "2nd pass has more frames than 1st pass (%d vs %d)\n", h->param.i_frame_total, rc->num_entries ); return -1; } /* FIXME: ugly padding because VfW drops delayed B-frames */ rc->num_entries += h->param.i_bframe; rc->entry = (ratecontrol_entry_t*) x264_malloc(rc->num_entries * sizeof(ratecontrol_entry_t)); memset(rc->entry, 0, rc->num_entries * sizeof(ratecontrol_entry_t)); /* init all to skipped p frames */ for(i=0; inum_entries; i++){ ratecontrol_entry_t *rce = &rc->entry[i]; rce->pict_type = SLICE_TYPE_P; rce->qscale = rce->new_qscale = qp2qscale(20); rce->misc_bits = rc->nmb + 10; rce->new_qp = 0; } /* read stats */ p = stats_in; for(i=0; i < rc->num_entries - h->param.i_bframe; i++){ ratecontrol_entry_t *rce; int frame_number; char pict_type; int e; char *next; float qp; next= strchr(p, ';'); if(next){ (*next)=0; //sscanf is unbelievably slow on looong strings next++; } e = sscanf(p, " in:%d ", &frame_number); if(frame_number < 0 || frame_number >= rc->num_entries) { x264_log(h, X264_LOG_ERROR, "bad frame number (%d) at stats line %d\n", frame_number, i); return -1; } rce = &rc->entry[frame_number]; rce->direct_mode = 0; e += sscanf(p, " in:%*d out:%*d type:%c q:%f itex:%d ptex:%d mv:%d misc:%d imb:%d pmb:%d smb:%d d:%c", &pict_type, &qp, &rce->i_tex_bits, &rce->p_tex_bits, &rce->mv_bits, &rce->misc_bits, &rce->i_count, &rce->p_count, &rce->s_count, &rce->direct_mode); switch(pict_type){ case 'I': rce->kept_as_ref = 1; case 'i': rce->pict_type = SLICE_TYPE_I; break; case 'P': rce->pict_type = SLICE_TYPE_P; break; case 'B': rce->kept_as_ref = 1; case 'b': rce->pict_type = SLICE_TYPE_B; break; default: e = -1; break; } if(e < 10){ x264_log(h, X264_LOG_ERROR, "statistics are damaged at line %d, parser out=%d\n", i, e); return -1; } rce->qscale = qp2qscale(qp); p = next; } x264_free(stats_buf); if(h->param.rc.i_rc_method == X264_RC_ABR) { if(init_pass2(h) < 0) return -1; } /* else we're using constant quant, so no need to run the bitrate allocation */ } /* Open output file */ /* If input and output files are the same, output to a temp file * and move it to the real name only when it's complete */ if( h->param.rc.b_stat_write ) { char *p; rc->psz_stat_file_tmpname = x264_malloc( strlen(h->param.rc.psz_stat_out) + 6 ); strcpy( rc->psz_stat_file_tmpname, h->param.rc.psz_stat_out ); strcat( rc->psz_stat_file_tmpname, ".temp" ); rc->p_stat_file_out = fopen( rc->psz_stat_file_tmpname, "wb" ); if( rc->p_stat_file_out == NULL ) { x264_log(h, X264_LOG_ERROR, "ratecontrol_init: can't open stats file\n"); return -1; } p = x264_param2string( &h->param, 1 ); fprintf( rc->p_stat_file_out, "#options: %s\n", p ); x264_free( p ); } for( i=1; iparam.i_threads; i++ ) { h->thread[i]->rc = rc+i; rc[i] = rc[0]; } return 0; } static int parse_zone( x264_t *h, x264_zone_t *z, char *p ) { int len = 0; char *tok, *saveptr; z->param = NULL; z->f_bitrate_factor = 1; if( 3 <= sscanf(p, "%u,%u,q=%u%n", &z->i_start, &z->i_end, &z->i_qp, &len) ) z->b_force_qp = 1; else if( 3 <= sscanf(p, "%u,%u,b=%f%n", &z->i_start, &z->i_end, &z->f_bitrate_factor, &len) ) z->b_force_qp = 0; else if( 2 <= sscanf(p, "%u,%u%n", &z->i_start, &z->i_end, &len) ) z->b_force_qp = 0; else { x264_log( h, X264_LOG_ERROR, "invalid zone: \"%s\"\n", p ); return -1; } p += len; if( !*p ) return 0; z->param = malloc( sizeof(x264_param_t) ); memcpy( z->param, &h->param, sizeof(x264_param_t) ); while( (tok = strtok_r( p, ",", &saveptr )) ) { char *val = strchr( tok, '=' ); if( val ) { *val = '\0'; val++; } if( x264_param_parse( z->param, tok, val ) ) { x264_log( h, X264_LOG_ERROR, "invalid zone param: %s = %s\n", tok, val ); return -1; } p = NULL; } return 0; } static int parse_zones( x264_t *h ) { x264_ratecontrol_t *rc = h->rc; int i; if( h->param.rc.psz_zones && !h->param.rc.i_zones ) { char *p, *tok, *saveptr; char *psz_zones = x264_malloc( strlen(h->param.rc.psz_zones)+1 ); strcpy( psz_zones, h->param.rc.psz_zones ); h->param.rc.i_zones = 1; for( p = psz_zones; *p; p++ ) h->param.rc.i_zones += (*p == '/'); h->param.rc.zones = x264_malloc( h->param.rc.i_zones * sizeof(x264_zone_t) ); p = psz_zones; for( i = 0; i < h->param.rc.i_zones; i++ ) { tok = strtok_r( p, "/", &saveptr ); if( !tok || parse_zone( h, &h->param.rc.zones[i], tok ) ) return -1; p = NULL; } x264_free( psz_zones ); } if( h->param.rc.i_zones > 0 ) { for( i = 0; i < h->param.rc.i_zones; i++ ) { x264_zone_t z = h->param.rc.zones[i]; if( z.i_start < 0 || z.i_start > z.i_end ) { x264_log( h, X264_LOG_ERROR, "invalid zone: start=%d end=%d\n", z.i_start, z.i_end ); return -1; } else if( !z.b_force_qp && z.f_bitrate_factor <= 0 ) { x264_log( h, X264_LOG_ERROR, "invalid zone: bitrate_factor=%f\n", z.f_bitrate_factor ); return -1; } } rc->i_zones = h->param.rc.i_zones + 1; rc->zones = x264_malloc( rc->i_zones * sizeof(x264_zone_t) ); memcpy( rc->zones+1, h->param.rc.zones, (rc->i_zones-1) * sizeof(x264_zone_t) ); // default zone to fall back to if none of the others match rc->zones[0].i_start = 0; rc->zones[0].i_end = INT_MAX; rc->zones[0].b_force_qp = 0; rc->zones[0].f_bitrate_factor = 1; rc->zones[0].param = x264_malloc( sizeof(x264_param_t) ); memcpy( rc->zones[0].param, &h->param, sizeof(x264_param_t) ); for( i = 1; i < rc->i_zones; i++ ) { if( !rc->zones[i].param ) rc->zones[i].param = rc->zones[0].param; } } return 0; } x264_zone_t *get_zone( x264_t *h, int frame_num ) { int i; for( i = h->rc->i_zones-1; i >= 0; i-- ) { x264_zone_t *z = &h->rc->zones[i]; if( frame_num >= z->i_start && frame_num <= z->i_end ) return z; } return NULL; } void x264_ratecontrol_summary( x264_t *h ) { x264_ratecontrol_t *rc = h->rc; if( rc->b_abr && h->param.rc.i_rc_method == X264_RC_ABR && rc->cbr_decay > .9999 ) { double base_cplx = h->mb.i_mb_count * (h->param.i_bframe ? 120 : 80); x264_log( h, X264_LOG_INFO, "final ratefactor: %.2f\n", qscale2qp( pow( base_cplx, 1 - h->param.rc.f_qcompress ) * rc->cplxr_sum / rc->wanted_bits_window ) ); } } void x264_ratecontrol_delete( x264_t *h ) { x264_ratecontrol_t *rc = h->rc; int i; if( rc->p_stat_file_out ) { fclose( rc->p_stat_file_out ); if( h->i_frame >= rc->num_entries - h->param.i_bframe ) if( rename( rc->psz_stat_file_tmpname, h->param.rc.psz_stat_out ) != 0 ) { x264_log( h, X264_LOG_ERROR, "failed to rename \"%s\" to \"%s\"\n", rc->psz_stat_file_tmpname, h->param.rc.psz_stat_out ); } x264_free( rc->psz_stat_file_tmpname ); } x264_free( rc->pred ); x264_free( rc->pred_b_from_p ); x264_free( rc->entry ); if( rc->zones ) { x264_free( rc->zones[0].param ); if( h->param.rc.psz_zones ) for( i=1; ii_zones; i++ ) if( rc->zones[i].param != rc->zones[0].param ) x264_free( rc->zones[i].param ); x264_free( rc->zones ); } x264_free( rc ); } static void accum_p_qp_update( x264_t *h, float qp ) { x264_ratecontrol_t *rc = h->rc; rc->accum_p_qp *= .95; rc->accum_p_norm *= .95; rc->accum_p_norm += 1; if( h->sh.i_type == SLICE_TYPE_I ) rc->accum_p_qp += qp + rc->ip_offset; else rc->accum_p_qp += qp; } /* Before encoding a frame, choose a QP for it */ void x264_ratecontrol_start( x264_t *h, int i_force_qp ) { x264_ratecontrol_t *rc = h->rc; ratecontrol_entry_t *rce = NULL; x264_zone_t *zone = get_zone( h, h->fenc->i_frame ); float q; x264_cpu_restore( h->param.cpu ); if( zone && (!rc->prev_zone || zone->param != rc->prev_zone->param) ) x264_encoder_reconfig( h, zone->param ); rc->prev_zone = zone; rc->qp_force = i_force_qp; if( h->param.rc.b_stat_read ) { int frame = h->fenc->i_frame; assert( frame >= 0 && frame < rc->num_entries ); rce = h->rc->rce = &h->rc->entry[frame]; if( h->sh.i_type == SLICE_TYPE_B && h->param.analyse.i_direct_mv_pred == X264_DIRECT_PRED_AUTO ) { h->sh.b_direct_spatial_mv_pred = ( rce->direct_mode == 's' ); h->mb.b_direct_auto_read = ( rce->direct_mode == 's' || rce->direct_mode == 't' ); } } if( rc->b_vbv ) { memset( h->fdec->i_row_bits, 0, h->sps->i_mb_height * sizeof(int) ); rc->row_pred = &rc->row_preds[h->sh.i_type]; update_vbv_plan( h ); } if( h->sh.i_type != SLICE_TYPE_B ) { rc->bframes = 0; while( h->frames.current[rc->bframes] && IS_X264_TYPE_B(h->frames.current[rc->bframes]->i_type) ) rc->bframes++; } rc->qpa = 0; if( i_force_qp ) { q = i_force_qp - 1; } else if( rc->b_abr ) { q = qscale2qp( rate_estimate_qscale( h ) ); } else if( rc->b_2pass ) { rce->new_qscale = rate_estimate_qscale( h ); q = qscale2qp( rce->new_qscale ); } else /* CQP */ { if( h->sh.i_type == SLICE_TYPE_B && h->fdec->b_kept_as_ref ) q = ( rc->qp_constant[ SLICE_TYPE_B ] + rc->qp_constant[ SLICE_TYPE_P ] ) / 2; else q = rc->qp_constant[ h->sh.i_type ]; if( zone ) { if( zone->b_force_qp ) q += zone->i_qp - rc->qp_constant[SLICE_TYPE_P]; else q -= 6*log(zone->f_bitrate_factor)/log(2); } } h->fdec->f_qp_avg = rc->qpm = rc->qp = x264_clip3( (int)(q + 0.5), 0, 51 ); if( rce ) rce->new_qp = rc->qp; /* accum_p_qp needs to be here so that future frames can benefit from the * data before this frame is done. but this only works because threading * guarantees to not re-encode any frames. so the non-threaded case does * accum_p_qp later. */ if( h->param.i_threads > 1 ) accum_p_qp_update( h, rc->qp ); if( h->sh.i_type != SLICE_TYPE_B ) rc->last_non_b_pict_type = h->sh.i_type; } double predict_row_size( x264_t *h, int y, int qp ) { /* average between two predictors: * absolute SATD, and scaled bit cost of the colocated row in the previous frame */ x264_ratecontrol_t *rc = h->rc; double pred_s = predict_size( rc->row_pred, qp2qscale(qp), h->fdec->i_row_satd[y] ); double pred_t = 0; if( h->sh.i_type != SLICE_TYPE_I && h->fref0[0]->i_type == h->fdec->i_type && h->fref0[0]->i_row_satd[y] > 0 ) { pred_t = h->fref0[0]->i_row_bits[y] * h->fdec->i_row_satd[y] / h->fref0[0]->i_row_satd[y] * qp2qscale(h->fref0[0]->i_row_qp[y]) / qp2qscale(qp); } if( pred_t == 0 ) pred_t = pred_s; return (pred_s + pred_t) / 2; } double predict_row_size_sum( x264_t *h, int y, int qp ) { int i; double bits = 0; for( i = 0; i <= y; i++ ) bits += h->fdec->i_row_bits[i]; for( i = y+1; i < h->sps->i_mb_height; i++ ) bits += predict_row_size( h, i, qp ); return bits; } void x264_ratecontrol_mb( x264_t *h, int bits ) { x264_ratecontrol_t *rc = h->rc; const int y = h->mb.i_mb_y; x264_cpu_restore( h->param.cpu ); h->fdec->i_row_bits[y] += bits; rc->qpa += rc->qpm; if( h->mb.i_mb_x != h->sps->i_mb_width - 1 || !h->mb.b_variable_qp ) return; h->fdec->i_row_qp[y] = rc->qpm; if( h->sh.i_type == SLICE_TYPE_B ) { /* B-frames shouldn't use lower QP than their reference frames */ if( y < h->sps->i_mb_height-1 ) { rc->qpm = X264_MAX( rc->qp, X264_MIN( h->fref0[0]->i_row_qp[y+1], h->fref1[0]->i_row_qp[y+1] )); } } else { update_predictor( rc->row_pred, qp2qscale(rc->qpm), h->fdec->i_row_satd[y], h->fdec->i_row_bits[y] ); /* tweak quality based on difference from predicted size */ if( y < h->sps->i_mb_height-1 && h->stat.i_slice_count[h->sh.i_type] > 0 ) { int prev_row_qp = h->fdec->i_row_qp[y]; int b0 = predict_row_size_sum( h, y, rc->qpm ); int b1 = b0; int i_qp_max = X264_MIN( prev_row_qp + h->param.rc.i_qp_step, h->param.rc.i_qp_max ); int i_qp_min = X264_MAX( prev_row_qp - h->param.rc.i_qp_step, h->param.rc.i_qp_min ); float buffer_left_planned = rc->buffer_fill - rc->frame_size_planned; if( !rc->b_vbv_min_rate ) i_qp_min = X264_MAX( i_qp_min, h->sh.i_qp ); while( rc->qpm < i_qp_max && (b1 > rc->frame_size_planned * 1.15 || (rc->buffer_fill - b1 < buffer_left_planned * 0.5))) { rc->qpm ++; b1 = predict_row_size_sum( h, y, rc->qpm ); } while( rc->qpm > i_qp_min && buffer_left_planned > rc->buffer_size * 0.4 && ((b1 < rc->frame_size_planned * 0.8 && rc->qpm <= prev_row_qp) || b1 < (rc->buffer_fill - rc->buffer_size + rc->buffer_rate) * 1.1) ) { rc->qpm --; b1 = predict_row_size_sum( h, y, rc->qpm ); } } } } int x264_ratecontrol_qp( x264_t *h ) { return h->rc->qpm; } /* In 2pass, force the same frame types as in the 1st pass */ int x264_ratecontrol_slice_type( x264_t *h, int frame_num ) { x264_ratecontrol_t *rc = h->rc; if( h->param.rc.b_stat_read ) { if( frame_num >= rc->num_entries ) { /* We could try to initialize everything required for ABR and * adaptive B-frames, but that would be complicated. * So just calculate the average QP used so far. */ h->param.rc.i_qp_constant = (h->stat.i_slice_count[SLICE_TYPE_P] == 0) ? 24 : 1 + h->stat.i_slice_qp[SLICE_TYPE_P] / h->stat.i_slice_count[SLICE_TYPE_P]; rc->qp_constant[SLICE_TYPE_P] = x264_clip3( h->param.rc.i_qp_constant, 0, 51 ); rc->qp_constant[SLICE_TYPE_I] = x264_clip3( (int)( qscale2qp( qp2qscale( h->param.rc.i_qp_constant ) / fabs( h->param.rc.f_ip_factor )) + 0.5 ), 0, 51 ); rc->qp_constant[SLICE_TYPE_B] = x264_clip3( (int)( qscale2qp( qp2qscale( h->param.rc.i_qp_constant ) * fabs( h->param.rc.f_pb_factor )) + 0.5 ), 0, 51 ); x264_log(h, X264_LOG_ERROR, "2nd pass has more frames than 1st pass (%d)\n", rc->num_entries); x264_log(h, X264_LOG_ERROR, "continuing anyway, at constant QP=%d\n", h->param.rc.i_qp_constant); if( h->param.b_bframe_adaptive ) x264_log(h, X264_LOG_ERROR, "disabling adaptive B-frames\n"); rc->b_abr = 0; rc->b_2pass = 0; h->param.rc.i_rc_method = X264_RC_CQP; h->param.rc.b_stat_read = 0; h->param.b_bframe_adaptive = 0; if( h->param.i_bframe > 1 ) h->param.i_bframe = 1; return X264_TYPE_P; } switch( rc->entry[frame_num].pict_type ) { case SLICE_TYPE_I: return rc->entry[frame_num].kept_as_ref ? X264_TYPE_IDR : X264_TYPE_I; case SLICE_TYPE_B: return rc->entry[frame_num].kept_as_ref ? X264_TYPE_BREF : X264_TYPE_B; case SLICE_TYPE_P: default: return X264_TYPE_P; } } else { return X264_TYPE_AUTO; } } /* After encoding one frame, save stats and update ratecontrol state */ void x264_ratecontrol_end( x264_t *h, int bits ) { x264_ratecontrol_t *rc = h->rc; const int *mbs = h->stat.frame.i_mb_count; int i; x264_cpu_restore( h->param.cpu ); h->stat.frame.i_mb_count_skip = mbs[P_SKIP] + mbs[B_SKIP]; h->stat.frame.i_mb_count_i = mbs[I_16x16] + mbs[I_8x8] + mbs[I_4x4]; h->stat.frame.i_mb_count_p = mbs[P_L0] + mbs[P_8x8]; for( i = B_DIRECT; i < B_8x8; i++ ) h->stat.frame.i_mb_count_p += mbs[i]; if( h->mb.b_variable_qp ) rc->qpa /= h->mb.i_mb_count; else rc->qpa = rc->qp; h->fdec->f_qp_avg = rc->qpa; if( h->param.rc.b_stat_write ) { char c_type = h->sh.i_type==SLICE_TYPE_I ? (h->fenc->i_poc==0 ? 'I' : 'i') : h->sh.i_type==SLICE_TYPE_P ? 'P' : h->fenc->b_kept_as_ref ? 'B' : 'b'; int dir_frame = h->stat.frame.i_direct_score[1] - h->stat.frame.i_direct_score[0]; int dir_avg = h->stat.i_direct_score[1] - h->stat.i_direct_score[0]; char c_direct = h->mb.b_direct_auto_write ? ( dir_frame>0 ? 's' : dir_frame<0 ? 't' : dir_avg>0 ? 's' : dir_avg<0 ? 't' : '-' ) : '-'; fprintf( rc->p_stat_file_out, "in:%d out:%d type:%c q:%.2f itex:%d ptex:%d mv:%d misc:%d imb:%d pmb:%d smb:%d d:%c;\n", h->fenc->i_frame, h->i_frame, c_type, rc->qpa, h->stat.frame.i_itex_bits, h->stat.frame.i_ptex_bits, h->stat.frame.i_hdr_bits, h->stat.frame.i_misc_bits, h->stat.frame.i_mb_count_i, h->stat.frame.i_mb_count_p, h->stat.frame.i_mb_count_skip, c_direct); } if( rc->b_abr ) { if( h->sh.i_type != SLICE_TYPE_B ) rc->cplxr_sum += bits * qp2qscale(rc->qpa) / rc->last_rceq; else { /* Depends on the fact that B-frame's QP is an offset from the following P-frame's. * Not perfectly accurate with B-refs, but good enough. */ rc->cplxr_sum += bits * qp2qscale(rc->qpa) / (rc->last_rceq * fabs(h->param.rc.f_pb_factor)); } rc->cplxr_sum *= rc->cbr_decay; rc->wanted_bits_window += rc->bitrate / rc->fps; rc->wanted_bits_window *= rc->cbr_decay; if( h->param.i_threads == 1 ) accum_p_qp_update( h, rc->qpa ); } if( rc->b_2pass ) { rc->expected_bits_sum += qscale2bits( rc->rce, qp2qscale(rc->rce->new_qp) ); } if( h->mb.b_variable_qp ) { if( h->sh.i_type == SLICE_TYPE_B ) { rc->bframe_bits += bits; if( !h->frames.current[0] || !IS_X264_TYPE_B(h->frames.current[0]->i_type) ) { update_predictor( rc->pred_b_from_p, qp2qscale(rc->qpa), h->fref1[h->i_ref1-1]->i_satd, rc->bframe_bits / rc->bframes ); rc->bframe_bits = 0; } } } update_vbv( h, bits ); } /**************************************************************************** * 2 pass functions ***************************************************************************/ double x264_eval( char *s, double *const_value, const char **const_name, double (**func1)(void *, double), const char **func1_name, double (**func2)(void *, double, double), char **func2_name, void *opaque ); /** * modify the bitrate curve from pass1 for one frame */ static double get_qscale(x264_t *h, ratecontrol_entry_t *rce, double rate_factor, int frame_num) { x264_ratecontrol_t *rcc= h->rc; const int pict_type = rce->pict_type; double q; x264_zone_t *zone = get_zone( h, frame_num ); double const_values[]={ rce->i_tex_bits * rce->qscale, rce->p_tex_bits * rce->qscale, (rce->i_tex_bits + rce->p_tex_bits) * rce->qscale, rce->mv_bits * rce->qscale, (double)rce->i_count / rcc->nmb, (double)rce->p_count / rcc->nmb, (double)rce->s_count / rcc->nmb, rce->pict_type == SLICE_TYPE_I, rce->pict_type == SLICE_TYPE_P, rce->pict_type == SLICE_TYPE_B, h->param.rc.f_qcompress, rcc->i_cplx_sum[SLICE_TYPE_I] / rcc->frame_count[SLICE_TYPE_I], rcc->i_cplx_sum[SLICE_TYPE_P] / rcc->frame_count[SLICE_TYPE_P], rcc->p_cplx_sum[SLICE_TYPE_P] / rcc->frame_count[SLICE_TYPE_P], rcc->p_cplx_sum[SLICE_TYPE_B] / rcc->frame_count[SLICE_TYPE_B], (rcc->i_cplx_sum[pict_type] + rcc->p_cplx_sum[pict_type]) / rcc->frame_count[pict_type], rce->blurred_complexity, 0 }; static const char *const_names[]={ "iTex", "pTex", "tex", "mv", "iCount", "pCount", "sCount", "isI", "isP", "isB", "qComp", "avgIITex", "avgPITex", "avgPPTex", "avgBPTex", "avgTex", "blurCplx", NULL }; static double (*func1[])(void *, double)={ // (void *)bits2qscale, (void *)qscale2bits, NULL }; static const char *func1_names[]={ // "bits2qp", "qp2bits", NULL }; q = x264_eval((char*)h->param.rc.psz_rc_eq, const_values, const_names, func1, func1_names, NULL, NULL, rce); // avoid NaN's in the rc_eq if(!isfinite(q) || rce->i_tex_bits + rce->p_tex_bits + rce->mv_bits == 0) q = rcc->last_qscale; else { rcc->last_rceq = q; q /= rate_factor; rcc->last_qscale = q; } if( zone ) { if( zone->b_force_qp ) q = qp2qscale(zone->i_qp); else q /= zone->f_bitrate_factor; } return q; } static double get_diff_limited_q(x264_t *h, ratecontrol_entry_t *rce, double q) { x264_ratecontrol_t *rcc = h->rc; const int pict_type = rce->pict_type; // force I/B quants as a function of P quants const double last_p_q = rcc->last_qscale_for[SLICE_TYPE_P]; const double last_non_b_q= rcc->last_qscale_for[rcc->last_non_b_pict_type]; if( pict_type == SLICE_TYPE_I ) { double iq = q; double pq = qp2qscale( rcc->accum_p_qp / rcc->accum_p_norm ); double ip_factor = fabs( h->param.rc.f_ip_factor ); /* don't apply ip_factor if the following frame is also I */ if( rcc->accum_p_norm <= 0 ) q = iq; else if( h->param.rc.f_ip_factor < 0 ) q = iq / ip_factor; else if( rcc->accum_p_norm >= 1 ) q = pq / ip_factor; else q = rcc->accum_p_norm * pq / ip_factor + (1 - rcc->accum_p_norm) * iq; } else if( pict_type == SLICE_TYPE_B ) { if( h->param.rc.f_pb_factor > 0 ) q = last_non_b_q; if( !rce->kept_as_ref ) q *= fabs( h->param.rc.f_pb_factor ); } else if( pict_type == SLICE_TYPE_P && rcc->last_non_b_pict_type == SLICE_TYPE_P && rce->i_tex_bits + rce->p_tex_bits == 0 ) { q = last_p_q; } /* last qscale / qdiff stuff */ if(rcc->last_non_b_pict_type==pict_type && (pict_type!=SLICE_TYPE_I || rcc->last_accum_p_norm < 1)) { double last_q = rcc->last_qscale_for[pict_type]; double max_qscale = last_q * rcc->lstep; double min_qscale = last_q / rcc->lstep; if (q > max_qscale) q = max_qscale; else if(q < min_qscale) q = min_qscale; } rcc->last_qscale_for[pict_type] = q; if(pict_type!=SLICE_TYPE_B) rcc->last_non_b_pict_type = pict_type; if(pict_type==SLICE_TYPE_I) { rcc->last_accum_p_norm = rcc->accum_p_norm; rcc->accum_p_norm = 0; rcc->accum_p_qp = 0; } if(pict_type==SLICE_TYPE_P) { float mask = 1 - pow( (float)rce->i_count / rcc->nmb, 2 ); rcc->accum_p_qp = mask * (qscale2qp(q) + rcc->accum_p_qp); rcc->accum_p_norm = mask * (1 + rcc->accum_p_norm); } return q; } static double predict_size( predictor_t *p, double q, double var ) { return p->coeff*var / (q*p->count); } static void update_predictor( predictor_t *p, double q, double var, double bits ) { if( var < 10 ) return; p->count *= p->decay; p->coeff *= p->decay; p->count ++; p->coeff += bits*q / var; } // update VBV after encoding a frame static void update_vbv( x264_t *h, int bits ) { x264_ratecontrol_t *rcc = h->rc; x264_ratecontrol_t *rct = h->thread[0]->rc; if( rcc->last_satd >= h->mb.i_mb_count ) update_predictor( &rct->pred[h->sh.i_type], qp2qscale(rcc->qpa), rcc->last_satd, bits ); if( !rcc->b_vbv ) return; rct->buffer_fill_final += rct->buffer_rate - bits; if( rct->buffer_fill_final < 0 && !rct->b_2pass ) x264_log( h, X264_LOG_WARNING, "VBV underflow (%.0f bits)\n", rct->buffer_fill_final ); rct->buffer_fill_final = x264_clip3f( rct->buffer_fill_final, 0, rct->buffer_size ); } // provisionally update VBV according to the planned size of all frames currently in progress static void update_vbv_plan( x264_t *h ) { x264_ratecontrol_t *rcc = h->rc; rcc->buffer_fill = h->thread[0]->rc->buffer_fill_final; if( h->param.i_threads > 1 ) { int j = h->rc - h->thread[0]->rc; int i; for( i=1; iparam.i_threads; i++ ) { x264_t *t = h->thread[ (j+i)%h->param.i_threads ]; double bits = t->rc->frame_size_planned; if( !t->b_thread_active ) continue; rcc->buffer_fill += rcc->buffer_rate - bits; rcc->buffer_fill = x264_clip3( rcc->buffer_fill, 0, rcc->buffer_size ); } } } // apply VBV constraints and clip qscale to between lmin and lmax static double clip_qscale( x264_t *h, int pict_type, double q ) { x264_ratecontrol_t *rcc = h->rc; double lmin = rcc->lmin[pict_type]; double lmax = rcc->lmax[pict_type]; double q0 = q; /* B-frames are not directly subject to VBV, * since they are controlled by the P-frames' QPs. * FIXME: in 2pass we could modify previous frames' QP too, * instead of waiting for the buffer to fill */ if( rcc->b_vbv && ( pict_type == SLICE_TYPE_P || ( pict_type == SLICE_TYPE_I && rcc->last_non_b_pict_type == SLICE_TYPE_I ) ) ) { if( rcc->buffer_fill/rcc->buffer_size < 0.5 ) q /= x264_clip3f( 2.0*rcc->buffer_fill/rcc->buffer_size, 0.5, 1.0 ); } if( rcc->b_vbv && rcc->last_satd > 0 ) { /* Now a hard threshold to make sure the frame fits in VBV. * This one is mostly for I-frames. */ double bits = predict_size( &rcc->pred[h->sh.i_type], q, rcc->last_satd ); double qf = 1.0; if( bits > rcc->buffer_fill/2 ) qf = x264_clip3f( rcc->buffer_fill/(2*bits), 0.2, 1.0 ); q /= qf; bits *= qf; if( bits < rcc->buffer_rate/2 ) q *= bits*2/rcc->buffer_rate; q = X264_MAX( q0, q ); /* Check B-frame complexity, and use up any bits that would * overflow before the next P-frame. */ if( h->sh.i_type == SLICE_TYPE_P ) { int nb = rcc->bframes; double pbbits = bits; double bbits = predict_size( rcc->pred_b_from_p, q * h->param.rc.f_pb_factor, rcc->last_satd ); double space; if( bbits > rcc->buffer_rate ) nb = 0; pbbits += nb * bbits; space = rcc->buffer_fill + (1+nb)*rcc->buffer_rate - rcc->buffer_size; if( pbbits < space ) { q *= X264_MAX( pbbits / space, bits / (0.5 * rcc->buffer_size) ); } q = X264_MAX( q0-5, q ); } if( !rcc->b_vbv_min_rate ) q = X264_MAX( q0, q ); } if(lmin==lmax) return lmin; else if(rcc->b_2pass) { double min2 = log(lmin); double max2 = log(lmax); q = (log(q) - min2)/(max2-min2) - 0.5; q = 1.0/(1.0 + exp(-4*q)); q = q*(max2-min2) + min2; return exp(q); } else return x264_clip3f(q, lmin, lmax); } // update qscale for 1 frame based on actual bits used so far static float rate_estimate_qscale( x264_t *h ) { float q; x264_ratecontrol_t *rcc = h->rc; ratecontrol_entry_t rce; int pict_type = h->sh.i_type; double lmin = rcc->lmin[pict_type]; double lmax = rcc->lmax[pict_type]; int64_t total_bits = 8*(h->stat.i_slice_size[SLICE_TYPE_I] + h->stat.i_slice_size[SLICE_TYPE_P] + h->stat.i_slice_size[SLICE_TYPE_B]); if( rcc->b_2pass ) { rce = *rcc->rce; if(pict_type != rce.pict_type) { x264_log(h, X264_LOG_ERROR, "slice=%c but 2pass stats say %c\n", slice_type_to_char[pict_type], slice_type_to_char[rce.pict_type]); } } if( pict_type == SLICE_TYPE_B ) { /* B-frames don't have independent ratecontrol, but rather get the * average QP of the two adjacent P-frames + an offset */ int i0 = IS_X264_TYPE_I(h->fref0[0]->i_type); int i1 = IS_X264_TYPE_I(h->fref1[0]->i_type); int dt0 = abs(h->fenc->i_poc - h->fref0[0]->i_poc); int dt1 = abs(h->fenc->i_poc - h->fref1[0]->i_poc); float q0 = h->fref0[0]->f_qp_avg; float q1 = h->fref1[0]->f_qp_avg; if( h->fref0[0]->i_type == X264_TYPE_BREF ) q0 -= rcc->pb_offset/2; if( h->fref1[0]->i_type == X264_TYPE_BREF ) q1 -= rcc->pb_offset/2; if(i0 && i1) q = (q0 + q1) / 2 + rcc->ip_offset; else if(i0) q = q1; else if(i1) q = q0; else q = (q0*dt1 + q1*dt0) / (dt0 + dt1); if(h->fenc->b_kept_as_ref) q += rcc->pb_offset/2; else q += rcc->pb_offset; rcc->frame_size_planned = predict_size( rcc->pred_b_from_p, q, h->fref1[h->i_ref1-1]->i_satd ); rcc->last_satd = 0; return qp2qscale(q); } else { double abr_buffer = 2 * rcc->rate_tolerance * rcc->bitrate; if( rcc->b_2pass ) { //FIXME adjust abr_buffer based on distance to the end of the video int64_t diff = total_bits - (int64_t)rce.expected_bits; q = rce.new_qscale; q /= x264_clip3f((double)(abr_buffer - diff) / abr_buffer, .5, 2); if( h->fenc->i_frame > 30 ) { /* Adjust quant based on the difference between * achieved and expected bitrate so far */ double time = (double)h->fenc->i_frame / rcc->num_entries; double w = x264_clip3f( time*100, 0.0, 1.0 ); q *= pow( (double)total_bits / rcc->expected_bits_sum, w ); } q = x264_clip3f( q, lmin, lmax ); } else /* 1pass ABR */ { /* Calculate the quantizer which would have produced the desired * average bitrate if it had been applied to all frames so far. * Then modulate that quant based on the current frame's complexity * relative to the average complexity so far (using the 2pass RCEQ). * Then bias the quant up or down if total size so far was far from * the target. * Result: Depending on the value of rate_tolerance, there is a * tradeoff between quality and bitrate precision. But at large * tolerances, the bit distribution approaches that of 2pass. */ double wanted_bits, overflow=1, lmin, lmax; rcc->last_satd = x264_rc_analyse_slice( h ); rcc->short_term_cplxsum *= 0.5; rcc->short_term_cplxcount *= 0.5; rcc->short_term_cplxsum += rcc->last_satd; rcc->short_term_cplxcount ++; rce.p_tex_bits = rcc->last_satd; rce.blurred_complexity = rcc->short_term_cplxsum / rcc->short_term_cplxcount; rce.i_tex_bits = 0; rce.mv_bits = 0; rce.p_count = rcc->nmb; rce.i_count = 0; rce.s_count = 0; rce.qscale = 1; rce.pict_type = pict_type; if( h->param.rc.i_rc_method == X264_RC_CRF ) { q = get_qscale( h, &rce, rcc->rate_factor_constant, h->fenc->i_frame ); } else { int i_frame_done = h->fenc->i_frame + 1 - h->param.i_threads; q = get_qscale( h, &rce, rcc->wanted_bits_window / rcc->cplxr_sum, h->fenc->i_frame ); // FIXME is it simpler to keep track of wanted_bits in ratecontrol_end? wanted_bits = i_frame_done * rcc->bitrate / rcc->fps; if( wanted_bits > 0 ) { abr_buffer *= X264_MAX( 1, sqrt(i_frame_done/25) ); overflow = x264_clip3f( 1.0 + (total_bits - wanted_bits) / abr_buffer, .5, 2 ); q *= overflow; } } if( pict_type == SLICE_TYPE_I && h->param.i_keyint_max > 1 /* should test _next_ pict type, but that isn't decided yet */ && rcc->last_non_b_pict_type != SLICE_TYPE_I ) { q = qp2qscale( rcc->accum_p_qp / rcc->accum_p_norm ); q /= fabs( h->param.rc.f_ip_factor ); } else if( h->i_frame > 0 ) { /* Asymmetric clipping, because symmetric would prevent * overflow control in areas of rapidly oscillating complexity */ lmin = rcc->last_qscale_for[pict_type] / rcc->lstep; lmax = rcc->last_qscale_for[pict_type] * rcc->lstep; if( overflow > 1.1 && h->i_frame > 3 ) lmax *= rcc->lstep; else if( overflow < 0.9 ) lmin /= rcc->lstep; q = x264_clip3f(q, lmin, lmax); } else if( h->param.rc.i_rc_method == X264_RC_CRF ) { q = qp2qscale( ABR_INIT_QP ) / fabs( h->param.rc.f_ip_factor ); } //FIXME use get_diff_limited_q() ? q = clip_qscale( h, pict_type, q ); } rcc->last_qscale_for[pict_type] = rcc->last_qscale = q; if( !rcc->b_2pass && h->fenc->i_frame == 0 ) rcc->last_qscale_for[SLICE_TYPE_P] = q; rcc->frame_size_planned = predict_size( &rcc->pred[h->sh.i_type], q, rcc->last_satd ); return q; } } void x264_thread_sync_ratecontrol( x264_t *cur, x264_t *prev, x264_t *next ) { if( cur != prev ) { #define COPY(var) memcpy(&cur->rc->var, &prev->rc->var, sizeof(cur->rc->var)) /* these vars are updated in x264_ratecontrol_start() * so copy them from the context that most recently started (prev) * to the context that's about to start (cur). */ COPY(accum_p_qp); COPY(accum_p_norm); COPY(last_satd); COPY(last_rceq); COPY(last_qscale_for); COPY(last_non_b_pict_type); COPY(short_term_cplxsum); COPY(short_term_cplxcount); COPY(bframes); COPY(prev_zone); #undef COPY } if( cur != next ) { #define COPY(var) next->rc->var = cur->rc->var /* these vars are updated in x264_ratecontrol_end() * so copy them from the context that most recently ended (cur) * to the context that's about to end (next) */ COPY(cplxr_sum); COPY(expected_bits_sum); COPY(wanted_bits_window); COPY(bframe_bits); #undef COPY } //FIXME row_preds[] (not strictly necessary, but would improve prediction) /* the rest of the variables are either constant or thread-local */ } static int init_pass2( x264_t *h ) { x264_ratecontrol_t *rcc = h->rc; uint64_t all_const_bits = 0; uint64_t all_available_bits = (uint64_t)(h->param.rc.i_bitrate * 1000. * rcc->num_entries / rcc->fps); double rate_factor, step, step_mult; double qblur = h->param.rc.f_qblur; double cplxblur = h->param.rc.f_complexity_blur; const int filter_size = (int)(qblur*4) | 1; double expected_bits; double *qscale, *blurred_qscale; int i; /* find total/average complexity & const_bits */ for(i=0; inum_entries; i++){ ratecontrol_entry_t *rce = &rcc->entry[i]; all_const_bits += rce->misc_bits; rcc->i_cplx_sum[rce->pict_type] += rce->i_tex_bits * rce->qscale; rcc->p_cplx_sum[rce->pict_type] += rce->p_tex_bits * rce->qscale; rcc->mv_bits_sum[rce->pict_type] += rce->mv_bits * rce->qscale; rcc->frame_count[rce->pict_type] ++; } if( all_available_bits < all_const_bits) { x264_log(h, X264_LOG_ERROR, "requested bitrate is too low. estimated minimum is %d kbps\n", (int)(all_const_bits * rcc->fps / (rcc->num_entries * 1000.))); return -1; } /* Blur complexities, to reduce local fluctuation of QP. * We don't blur the QPs directly, because then one very simple frame * could drag down the QP of a nearby complex frame and give it more * bits than intended. */ for(i=0; inum_entries; i++){ ratecontrol_entry_t *rce = &rcc->entry[i]; double weight_sum = 0; double cplx_sum = 0; double weight = 1.0; int j; /* weighted average of cplx of future frames */ for(j=1; jnum_entries-i; j++){ ratecontrol_entry_t *rcj = &rcc->entry[i+j]; weight *= 1 - pow( (float)rcj->i_count / rcc->nmb, 2 ); if(weight < .0001) break; weight_sum += weight; cplx_sum += weight * (qscale2bits(rcj, 1) - rcj->misc_bits); } /* weighted average of cplx of past frames */ weight = 1.0; for(j=0; j<=cplxblur*2 && j<=i; j++){ ratecontrol_entry_t *rcj = &rcc->entry[i-j]; weight_sum += weight; cplx_sum += weight * (qscale2bits(rcj, 1) - rcj->misc_bits); weight *= 1 - pow( (float)rcj->i_count / rcc->nmb, 2 ); if(weight < .0001) break; } rce->blurred_complexity = cplx_sum / weight_sum; } qscale = x264_malloc(sizeof(double)*rcc->num_entries); if(filter_size > 1) blurred_qscale = x264_malloc(sizeof(double)*rcc->num_entries); else blurred_qscale = qscale; /* Search for a factor which, when multiplied by the RCEQ values from * each frame, adds up to the desired total size. * There is no exact closed-form solution because of VBV constraints and * because qscale2bits is not invertible, but we can start with the simple * approximation of scaling the 1st pass by the ratio of bitrates. * The search range is probably overkill, but speed doesn't matter here. */ expected_bits = 1; for(i=0; inum_entries; i++) expected_bits += qscale2bits(&rcc->entry[i], get_qscale(h, &rcc->entry[i], 1.0, i)); step_mult = all_available_bits / expected_bits; rate_factor = 0; for(step = 1E4 * step_mult; step > 1E-7 * step_mult; step *= 0.5){ expected_bits = 0; rate_factor += step; rcc->last_non_b_pict_type = -1; rcc->last_accum_p_norm = 1; rcc->accum_p_norm = 0; rcc->buffer_fill = rcc->buffer_size * h->param.rc.f_vbv_buffer_init; /* find qscale */ for(i=0; inum_entries; i++){ qscale[i] = get_qscale(h, &rcc->entry[i], rate_factor, i); } /* fixed I/B qscale relative to P */ for(i=rcc->num_entries-1; i>=0; i--){ qscale[i] = get_diff_limited_q(h, &rcc->entry[i], qscale[i]); assert(qscale[i] >= 0); } /* smooth curve */ if(filter_size > 1){ assert(filter_size%2==1); for(i=0; inum_entries; i++){ ratecontrol_entry_t *rce = &rcc->entry[i]; int j; double q=0.0, sum=0.0; for(j=0; j= rcc->num_entries) continue; if(rce->pict_type != rcc->entry[index].pict_type) continue; q += qscale[index] * coeff; sum += coeff; } blurred_qscale[i] = q/sum; } } /* find expected bits */ for(i=0; inum_entries; i++){ ratecontrol_entry_t *rce = &rcc->entry[i]; double bits; rce->new_qscale = clip_qscale(h, rce->pict_type, blurred_qscale[i]); assert(rce->new_qscale >= 0); bits = qscale2bits(rce, rce->new_qscale); rce->expected_bits = expected_bits; expected_bits += bits; update_vbv(h, bits); rcc->buffer_fill = rcc->buffer_fill_final; } //printf("expected:%llu available:%llu factor:%lf avgQ:%lf\n", (uint64_t)expected_bits, all_available_bits, rate_factor); if(expected_bits > all_available_bits) rate_factor -= step; } x264_free(qscale); if(filter_size > 1) x264_free(blurred_qscale); if(fabs(expected_bits/all_available_bits - 1.0) > 0.01) { double avgq = 0; for(i=0; inum_entries; i++) avgq += rcc->entry[i].new_qscale; avgq = qscale2qp(avgq / rcc->num_entries); x264_log(h, X264_LOG_WARNING, "Error: 2pass curve failed to converge\n"); x264_log(h, X264_LOG_WARNING, "target: %.2f kbit/s, expected: %.2f kbit/s, avg QP: %.4f\n", (float)h->param.rc.i_bitrate, expected_bits * rcc->fps / (rcc->num_entries * 1000.), avgq); if(expected_bits < all_available_bits && avgq < h->param.rc.i_qp_min + 2) { if(h->param.rc.i_qp_min > 0) x264_log(h, X264_LOG_WARNING, "try reducing target bitrate or reducing qp_min (currently %d)\n", h->param.rc.i_qp_min); else x264_log(h, X264_LOG_WARNING, "try reducing target bitrate\n"); } else if(expected_bits > all_available_bits && avgq > h->param.rc.i_qp_max - 2) { if(h->param.rc.i_qp_max < 51) x264_log(h, X264_LOG_WARNING, "try increasing target bitrate or increasing qp_max (currently %d)\n", h->param.rc.i_qp_max); else x264_log(h, X264_LOG_WARNING, "try increasing target bitrate\n"); } else x264_log(h, X264_LOG_WARNING, "internal error\n"); } return 0; }