/*****************************************************************************
* ratecontrol.c: ratecontrol
*****************************************************************************
- * Copyright (C) 2005-2010 x264 project
+ * Copyright (C) 2005-2011 x264 project
*
* Authors: Loren Merritt <lorenm@u.washington.edu>
* Michael Niedermayer <michaelni@gmx.at>
#define _ISOC99_SOURCE
#undef NDEBUG // always check asserts, the speed effect is far too small to disable them
-#include <math.h>
#include "common/common.h"
#include "ratecontrol.h"
int16_t i_weight_denom[2];
int refcount[16];
int refs;
- int i_duration;
- int i_cpb_duration;
+ int64_t i_duration;
+ int64_t i_cpb_duration;
} ratecontrol_entry_t;
typedef struct
{
- double coeff;
- double count;
- double decay;
- double offset;
+ float coeff;
+ float count;
+ float decay;
+ float offset;
} predictor_t;
struct x264_ratecontrol_t
int qp; /* qp for current frame */
float qpm; /* qp for current macroblock: precise float for AQ */
float qpa_rc; /* average of macroblocks' qp before aq */
- float qpa_aq; /* average of macroblocks' qp after aq */
+ int qpa_aq; /* average of macroblocks' qp after aq */
float qp_novbv; /* QP for the current frame if 1-pass VBV was disabled. */
/* VBV stuff */
static float rate_estimate_qscale( x264_t *h );
static int update_vbv( x264_t *h, int bits );
static void update_vbv_plan( x264_t *h, int overhead );
-static double predict_size( predictor_t *p, double q, double var );
-static void update_predictor( predictor_t *p, double q, double var, double bits );
+static float predict_size( predictor_t *p, float q, float var );
+static void update_predictor( predictor_t *p, float q, float var, float bits );
#define CMP_OPT_FIRST_PASS( opt, param_val )\
{\
* qp = h.264's quantizer
* qscale = linearized quantizer = Lagrange multiplier
*/
-static inline double qp2qscale( double qp )
+static inline float qp2qscale( float qp )
{
- return 0.85 * pow( 2.0, ( qp - 12.0 ) / 6.0 );
+ return 0.85f * powf( 2.0f, ( qp - 12.0f ) / 6.0f );
}
-static inline double qscale2qp( double qscale )
+static inline float qscale2qp( float qscale )
{
- return 12.0 + 6.0 * log2( qscale/0.85 );
+ return 12.0f + 6.0f * log2f( qscale/0.85f );
}
/* Texture bitrate is not quite inversely proportional to qscale,
+ rce->misc_bits;
}
-static ALWAYS_INLINE uint32_t ac_energy_var( uint64_t sum_ssd, int shift, x264_frame_t *frame, int i )
+static ALWAYS_INLINE uint32_t ac_energy_var( uint64_t sum_ssd, int shift, x264_frame_t *frame, int i, int b_store )
{
uint32_t sum = sum_ssd;
uint32_t ssd = sum_ssd >> 32;
- frame->i_pixel_sum[i] += sum;
- frame->i_pixel_ssd[i] += ssd;
+ if( b_store )
+ {
+ frame->i_pixel_sum[i] += sum;
+ frame->i_pixel_ssd[i] += ssd;
+ }
return ssd - ((uint64_t)sum * sum >> shift);
}
-static ALWAYS_INLINE uint32_t ac_energy_plane( x264_t *h, int mb_x, int mb_y, x264_frame_t *frame, int i )
+static ALWAYS_INLINE uint32_t ac_energy_plane( x264_t *h, int mb_x, int mb_y, x264_frame_t *frame, int i, int b_chroma, int b_field, int b_store )
{
- int w = i ? 8 : 16;
+ int height = b_chroma ? 16>>CHROMA_V_SHIFT : 16;
int stride = frame->i_stride[i];
- int offset = h->mb.b_interlaced
- ? 16 * mb_x + w * (mb_y&~1) * stride + (mb_y&1) * stride
- : 16 * mb_x + w * mb_y * stride;
- stride <<= h->mb.b_interlaced;
- if( i )
+ int offset = b_field
+ ? 16 * mb_x + height * (mb_y&~1) * stride + (mb_y&1) * stride
+ : 16 * mb_x + height * mb_y * stride;
+ stride <<= b_field;
+ if( b_chroma )
{
- ALIGNED_ARRAY_16( pixel, pix,[FENC_STRIDE*8] );
- h->mc.load_deinterleave_8x8x2_fenc( pix, frame->plane[1] + offset, stride );
- return ac_energy_var( h->pixf.var[PIXEL_8x8]( pix, FENC_STRIDE ), 6, frame, 1 )
- + ac_energy_var( h->pixf.var[PIXEL_8x8]( pix+FENC_STRIDE/2, FENC_STRIDE ), 6, frame, 2 );
+ ALIGNED_ARRAY_16( pixel, pix,[FENC_STRIDE*16] );
+ int chromapix = h->luma2chroma_pixel[PIXEL_16x16];
+ int shift = 7 - CHROMA_V_SHIFT;
+
+ h->mc.load_deinterleave_chroma_fenc( pix, frame->plane[1] + offset, stride, height );
+ return ac_energy_var( h->pixf.var[chromapix]( pix, FENC_STRIDE ), shift, frame, 1, b_store )
+ + ac_energy_var( h->pixf.var[chromapix]( pix+FENC_STRIDE/2, FENC_STRIDE ), shift, frame, 2, b_store );
}
else
- return ac_energy_var( h->pixf.var[PIXEL_16x16]( frame->plane[0] + offset, stride ), 8, frame, 0 );
+ return ac_energy_var( h->pixf.var[PIXEL_16x16]( frame->plane[i] + offset, stride ), 8, frame, i, b_store );
}
// Find the total AC energy of the block in all planes.
* and putting it after floating point ops. As a result, we put the emms at the end of the
* function and make sure that its always called before the float math. Noinline makes
* sure no reordering goes on. */
- uint32_t var = ac_energy_plane( h, mb_x, mb_y, frame, 0 );
- var += ac_energy_plane( h, mb_x, mb_y, frame, 1 );
+ uint32_t var;
+ x264_prefetch_fenc( h, frame, mb_x, mb_y );
+ if( h->mb.b_adaptive_mbaff )
+ {
+ /* We don't know the super-MB mode we're going to pick yet, so
+ * simply try both and pick the lower of the two. */
+ uint32_t var_interlaced, var_progressive;
+ var_interlaced = ac_energy_plane( h, mb_x, mb_y, frame, 0, 0, 1, 1 );
+ var_progressive = ac_energy_plane( h, mb_x, mb_y, frame, 0, 0, 0, 0 );
+ if( CHROMA444 )
+ {
+ var_interlaced += ac_energy_plane( h, mb_x, mb_y, frame, 1, 0, 1, 1 );
+ var_progressive += ac_energy_plane( h, mb_x, mb_y, frame, 1, 0, 0, 0 );
+ var_interlaced += ac_energy_plane( h, mb_x, mb_y, frame, 2, 0, 1, 1 );
+ var_progressive += ac_energy_plane( h, mb_x, mb_y, frame, 2, 0, 0, 0 );
+ }
+ else
+ {
+ var_interlaced += ac_energy_plane( h, mb_x, mb_y, frame, 1, 1, 1, 1 );
+ var_progressive += ac_energy_plane( h, mb_x, mb_y, frame, 1, 1, 0, 0 );
+ }
+ var = X264_MIN( var_interlaced, var_progressive );
+ }
+ else
+ {
+ var = ac_energy_plane( h, mb_x, mb_y, frame, 0, 0, PARAM_INTERLACED, 1 );
+ if( CHROMA444 )
+ {
+ var += ac_energy_plane( h, mb_x, mb_y, frame, 1, 0, PARAM_INTERLACED, 1 );
+ var += ac_energy_plane( h, mb_x, mb_y, frame, 2, 0, PARAM_INTERLACED, 1 );
+ }
+ else
+ var += ac_energy_plane( h, mb_x, mb_y, frame, 1, 1, PARAM_INTERLACED, 1 );
+ }
x264_emms();
return var;
}
{
uint64_t ssd = frame->i_pixel_ssd[i];
uint64_t sum = frame->i_pixel_sum[i];
- int width = h->mb.i_mb_width*16>>!!i;
- int height = h->mb.i_mb_height*16>>!!i;
+ int width = 16*h->mb.i_mb_width >> (i && CHROMA_H_SHIFT);
+ int height = 16*h->mb.i_mb_height >> (i && CHROMA_V_SHIFT);
frame->i_pixel_ssd[i] = ssd - (sum * sum + width * height / 2) / (width * height);
}
}
if( i_type != i_type_actual && rc->qpbuf_pos == 1 )
{
- x264_log(h, X264_LOG_ERROR, "MB-tree frametype %d doesn't match actual frametype %d.\n", i_type, i_type_actual);
+ x264_log( h, X264_LOG_ERROR, "MB-tree frametype %d doesn't match actual frametype %d.\n", i_type, i_type_actual );
return -1;
}
} while( i_type != i_type_actual );
x264_stack_align( x264_adaptive_quant_frame, h, frame, quant_offsets );
return 0;
fail:
- x264_log(h, X264_LOG_ERROR, "Incomplete MB-tree stats file.\n");
+ x264_log( h, X264_LOG_ERROR, "Incomplete MB-tree stats file.\n" );
return -1;
}
x264_weight_t weights[16][3];
int refcount[16];
- if( rce->refs != h->i_ref0 )
+ if( rce->refs != h->i_ref[0] )
return -1;
- memcpy( frames, h->fref0, sizeof(frames) );
+ memcpy( frames, h->fref[0], sizeof(frames) );
memcpy( refcount, rce->refcount, sizeof(refcount) );
memcpy( weights, h->fenc->weight, sizeof(weights) );
memset( &h->fenc->weight[1][0], 0, sizeof(x264_weight_t[15][3]) );
/* For now don't reorder ref 0; it seems to lower quality
in most cases due to skips. */
- for( int ref = 1; ref < h->i_ref0; ref++ )
+ for( int ref = 1; ref < h->i_ref[0]; ref++ )
{
int max = -1;
int bestref = 1;
- for( int i = 1; i < h->i_ref0; i++ )
+ for( int i = 1; i < h->i_ref[0]; i++ )
/* Favor lower POC as a tiebreaker. */
COPY2_IF_GT( max, refcount[i], bestref, i );
* that the optimal ordering doesnt place every duplicate. */
refcount[bestref] = -1;
- h->fref0[ref] = frames[bestref];
+ h->fref[0][ref] = frames[bestref];
memcpy( h->fenc->weight[ref], weights[bestref], sizeof(weights[bestref]) );
}
if( h->param.rc.i_vbv_max_bitrate > 0 && h->param.rc.i_vbv_buffer_size > 0 )
{
+ /* We don't support changing the ABR bitrate right now,
+ so if the stream starts as CBR, keep it CBR. */
+ if( rc->b_vbv_min_rate )
+ h->param.rc.i_vbv_max_bitrate = h->param.rc.i_bitrate;
+
if( h->param.rc.i_vbv_buffer_size < (int)(h->param.rc.i_vbv_max_bitrate / rc->fps) )
{
h->param.rc.i_vbv_buffer_size = h->param.rc.i_vbv_max_bitrate / rc->fps;
h->param.rc.i_vbv_buffer_size );
}
- /* We don't support changing the ABR bitrate right now,
- so if the stream starts as CBR, keep it CBR. */
- if( rc->b_vbv_min_rate )
- h->param.rc.i_vbv_max_bitrate = h->param.rc.i_bitrate;
-
int vbv_buffer_size = h->param.rc.i_vbv_buffer_size * 1000;
int vbv_max_bitrate = h->param.rc.i_vbv_max_bitrate * 1000;
/* Init HRD */
- h->sps->vui.hrd.i_bit_rate_unscaled = vbv_max_bitrate;
- h->sps->vui.hrd.i_cpb_size_unscaled = vbv_buffer_size;
if( h->param.i_nal_hrd && b_init )
{
h->sps->vui.hrd.i_cpb_cnt = 1;
x264_log( h, X264_LOG_WARNING, "VBV parameters cannot be changed when NAL HRD is in use\n" );
return;
}
+ h->sps->vui.hrd.i_bit_rate_unscaled = vbv_max_bitrate;
+ h->sps->vui.hrd.i_cpb_size_unscaled = vbv_buffer_size;
+ if( rc->b_vbv_min_rate )
+ rc->bitrate = h->param.rc.i_bitrate * 1000.;
rc->buffer_rate = vbv_max_bitrate / rc->fps;
rc->vbv_max_rate = vbv_max_bitrate;
rc->buffer_size = vbv_buffer_size;
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");
+ x264_log( h, X264_LOG_ERROR, "constant rate-factor is incompatible with 2pass.\n" );
return -1;
}
if( rc->rate_tolerance < 0.01 )
{
- x264_log(h, X264_LOG_WARNING, "bitrate tolerance too small, using .01\n");
+ x264_log( h, X264_LOG_WARNING, "bitrate tolerance too small, using .01\n" );
rc->rate_tolerance = 0.01;
}
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");
+ x264_log( h, X264_LOG_ERROR, "ratecontrol_init: can't open stats file\n" );
return -1;
}
if( h->param.rc.b_mb_tree )
x264_free( mbtree_stats_in );
if( !rc->p_mbtree_stat_file_in )
{
- x264_log(h, X264_LOG_ERROR, "ratecontrol_init: can't open mbtree stats file\n");
+ x264_log( h, X264_LOG_ERROR, "ratecontrol_init: can't open mbtree stats file\n" );
return -1;
}
}
/* check whether 1st pass options were compatible with current options */
- if( !strncmp( stats_buf, "#options:", 9 ) )
+ if( strncmp( stats_buf, "#options:", 9 ) )
+ {
+ x264_log( h, X264_LOG_ERROR, "options list in stats file not valid\n" );
+ return -1;
+ }
+
+ float res_factor, res_factor_bits;
{
int i, j;
uint32_t k, l;
h->param.i_width, h->param.i_height, i, j );
return -1;
}
+ res_factor = (float)h->param.i_width * h->param.i_height / (i*j);
+ /* Change in bits relative to resolution isn't quite linear on typical sources,
+ * so we'll at least try to roughly approximate this effect. */
+ res_factor_bits = powf( res_factor, 0.7 );
if( ( p = strstr( opts, "timebase=" ) ) && sscanf( p, "timebase=%u/%u", &k, &l ) != 2 )
{
CMP_OPT_FIRST_PASS( "bframes", h->param.i_bframe );
CMP_OPT_FIRST_PASS( "b_pyramid", h->param.i_bframe_pyramid );
CMP_OPT_FIRST_PASS( "intra_refresh", h->param.b_intra_refresh );
- CMP_OPT_FIRST_PASS( "open_gop", h->param.i_open_gop );
+ CMP_OPT_FIRST_PASS( "open_gop", h->param.b_open_gop );
+ CMP_OPT_FIRST_PASS( "bluray_compat", h->param.b_bluray_compat );
+
+ if( (p = strstr( opts, "interlaced=" )) )
+ {
+ char *current = h->param.b_interlaced ? h->param.b_tff ? "tff" : "bff" : h->param.b_fake_interlaced ? "fake" : "0";
+ char buf[5];
+ sscanf( p, "interlaced=%4s", buf );
+ if( strcmp( current, buf ) )
+ {
+ x264_log( h, X264_LOG_ERROR, "different interlaced setting than first pass (%s vs %s)\n", current, buf );
+ return -1;
+ }
+ }
if( (p = strstr( opts, "keyint=" )) )
{
p = strchr( p + 1, ';' );
if( !num_entries )
{
- x264_log(h, X264_LOG_ERROR, "empty stats file\n");
+ x264_log( h, X264_LOG_ERROR, "empty stats file\n" );
return -1;
}
rc->num_entries = num_entries;
rce = &rc->entry[frame_number];
rce->direct_mode = 0;
- e += sscanf( p, " in:%*d out:%*d type:%c dur:%d cpbdur:%d q:%f tex:%d mv:%d misc:%d imb:%d pmb:%d smb:%d d:%c",
+ e += sscanf( p, " in:%*d out:%*d type:%c dur:%"SCNd64" cpbdur:%"SCNd64" q:%f tex:%d mv:%d misc:%d imb:%d pmb:%d smb:%d d:%c",
&pict_type, &rce->i_duration, &rce->i_cpb_duration, &qp, &rce->tex_bits,
&rce->mv_bits, &rce->misc_bits, &rce->i_count, &rce->p_count,
&rce->s_count, &rce->direct_mode );
+ rce->tex_bits *= res_factor_bits;
+ rce->mv_bits *= res_factor_bits;
+ rce->misc_bits *= res_factor_bits;
+ rce->i_count *= res_factor;
+ rce->p_count *= res_factor;
+ rce->s_count *= res_factor;
p = strstr( p, "ref:" );
if( !p )
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");
+ x264_log( h, X264_LOG_ERROR, "ratecontrol_init: can't open stats file\n" );
return -1;
}
rc->p_mbtree_stat_file_out = fopen( rc->psz_mbtree_stat_file_tmpname, "wb" );
if( rc->p_mbtree_stat_file_out == NULL )
{
- x264_log(h, X264_LOG_ERROR, "ratecontrol_init: can't open mbtree stats file\n");
+ x264_log( h, X264_LOG_ERROR, "ratecontrol_init: can't open mbtree stats file\n" );
return -1;
}
}
rc[i] = rc[0];
h->thread[i]->param = h->param;
h->thread[i]->mb.b_variable_qp = h->mb.b_variable_qp;
+ h->thread[i]->mb.ip_offset = h->mb.ip_offset;
}
}
char *tok, UNUSED *saveptr=NULL;
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) )
+ if( 3 <= sscanf(p, "%d,%d,q=%d%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) )
+ else if( 3 <= sscanf(p, "%d,%d,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) )
+ else if( 2 <= sscanf(p, "%d,%d%n", &z->i_start, &z->i_end, &len) )
z->b_force_qp = 0;
else
{
if( rc->b_vbv )
{
memset( h->fdec->i_row_bits, 0, h->mb.i_mb_height * sizeof(int) );
+ memset( h->fdec->f_row_qp, 0, h->mb.i_mb_height * sizeof(float) );
+ memset( h->fdec->f_row_qscale, 0, h->mb.i_mb_height * sizeof(float) );
rc->row_pred = &rc->row_preds[h->sh.i_type];
rc->buffer_rate = h->fenc->i_cpb_duration * rc->vbv_max_rate * h->sps->vui.i_num_units_in_tick / h->sps->vui.i_time_scale;
update_vbv_plan( h, overhead );
int mincr = l->mincr;
- /* Blu-ray requires this */
- if( l->level_idc == 41 && h->param.i_nal_hrd )
+ if( h->param.b_bluray_compat )
mincr = 4;
- /* High 10 doesn't require minCR, so just set the maximum to a large value. */
- if( h->sps->i_profile_idc == PROFILE_HIGH10 )
+ /* Profiles above High don't require minCR, so just set the maximum to a large value. */
+ if( h->sps->i_profile_idc > PROFILE_HIGH )
rc->frame_size_maximum = 1e9;
else
{
if( h->sh.i_type != SLICE_TYPE_B )
rc->bframes = h->fenc->i_bframes;
- if( i_force_qp != X264_QP_AUTO )
- {
- q = i_force_qp - 1;
- }
- else if( rc->b_abr )
+ if( rc->b_abr )
{
q = qscale2qp( rate_estimate_qscale( h ) );
}
q -= 6*log2f( zone->f_bitrate_factor );
}
}
+ if( i_force_qp != X264_QP_AUTO )
+ q = i_force_qp - 1;
q = x264_clip3f( q, h->param.rc.i_qp_min, h->param.rc.i_qp_max );
rc->qpa_rc =
rc->qpa_aq = 0;
- rc->qp = x264_clip3( (int)(q + 0.5), 0, QP_MAX );
+ rc->qp = x264_clip3( q + 0.5f, 0, QP_MAX );
h->fdec->f_qp_avg_rc =
h->fdec->f_qp_avg_aq =
rc->qpm = q;
rc->last_non_b_pict_type = h->sh.i_type;
}
-static double predict_row_size( x264_t *h, int y, double qp )
+static float predict_row_size( x264_t *h, int y, float qscale )
{
/* 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[0], qp2qscale( qp ), h->fdec->i_row_satd[y] );
- double pred_t = 0;
- if( h->sh.i_type == SLICE_TYPE_I || qp >= h->fref0[0]->f_row_qp[y] )
+ float pred_s = predict_size( rc->row_pred[0], qscale, h->fdec->i_row_satd[y] );
+ if( h->sh.i_type == SLICE_TYPE_I || qscale >= h->fref[0][0]->f_row_qscale[y] )
{
if( h->sh.i_type == SLICE_TYPE_P
- && h->fref0[0]->i_type == h->fdec->i_type
- && h->fref0[0]->i_row_satd[y] > 0
- && (abs(h->fref0[0]->i_row_satd[y] - h->fdec->i_row_satd[y]) < h->fdec->i_row_satd[y]/2))
+ && h->fref[0][0]->i_type == h->fdec->i_type
+ && h->fref[0][0]->f_row_qscale[y] > 0
+ && h->fref[0][0]->i_row_satd[y] > 0
+ && (abs(h->fref[0][0]->i_row_satd[y] - h->fdec->i_row_satd[y]) < h->fdec->i_row_satd[y]/2))
{
- 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]->f_row_qp[y] ) / qp2qscale( qp );
+ float pred_t = h->fref[0][0]->i_row_bits[y] * h->fdec->i_row_satd[y] / h->fref[0][0]->i_row_satd[y]
+ * h->fref[0][0]->f_row_qscale[y] / qscale;
+ return (pred_s + pred_t) * 0.5f;
}
- if( pred_t == 0 )
- pred_t = pred_s;
- return (pred_s + pred_t) / 2;
+ return pred_s;
}
/* Our QP is lower than the reference! */
else
{
- double pred_intra = predict_size( rc->row_pred[1], qp2qscale( qp ), h->fdec->i_row_satds[0][0][y] );
+ float pred_intra = predict_size( rc->row_pred[1], qscale, h->fdec->i_row_satds[0][0][y] );
/* Sum: better to overestimate than underestimate by using only one of the two predictors. */
return pred_intra + pred_s;
}
}
-static double row_bits_so_far( x264_t *h, int y )
+static int row_bits_so_far( x264_t *h, int y )
{
- double bits = 0;
+ int bits = 0;
for( int i = h->i_threadslice_start; i <= y; i++ )
bits += h->fdec->i_row_bits[i];
return bits;
}
-static double predict_row_size_sum( x264_t *h, int y, double qp )
+static float predict_row_size_sum( x264_t *h, int y, float qp )
{
- double bits = row_bits_so_far(h, y);
+ float qscale = qp2qscale( qp );
+ float bits = row_bits_so_far( h, y );
for( int i = y+1; i < h->i_threadslice_end; i++ )
- bits += predict_row_size( h, i, qp );
+ bits += predict_row_size( h, i, qscale );
return bits;
}
-
+/* TODO:
+ * eliminate all use of qp in row ratecontrol: make it entirely qscale-based.
+ * make this function stop being needlessly O(N^2)
+ * update more often than once per row? */
void x264_ratecontrol_mb( x264_t *h, int bits )
{
x264_ratecontrol_t *rc = h->rc;
const int y = h->mb.i_mb_y;
- x264_emms();
-
h->fdec->i_row_bits[y] += bits;
- rc->qpa_rc += rc->qpm;
rc->qpa_aq += h->mb.i_qp;
- if( h->mb.i_mb_x != h->mb.i_mb_width - 1 || !rc->b_vbv )
+ if( h->mb.i_mb_x != h->mb.i_mb_width - 1 )
+ return;
+
+ x264_emms();
+ rc->qpa_rc += rc->qpm * h->mb.i_mb_width;
+
+ if( !rc->b_vbv )
return;
+ float qscale = qp2qscale( rc->qpm );
h->fdec->f_row_qp[y] = rc->qpm;
+ h->fdec->f_row_qscale[y] = qscale;
- update_predictor( rc->row_pred[0], qp2qscale( rc->qpm ), h->fdec->i_row_satd[y], h->fdec->i_row_bits[y] );
- if( h->sh.i_type == SLICE_TYPE_P && rc->qpm < h->fref0[0]->f_row_qp[y] )
- update_predictor( rc->row_pred[1], qp2qscale( rc->qpm ), h->fdec->i_row_satds[0][0][y], h->fdec->i_row_bits[y] );
+ update_predictor( rc->row_pred[0], qscale, h->fdec->i_row_satd[y], h->fdec->i_row_bits[y] );
+ if( h->sh.i_type == SLICE_TYPE_P && rc->qpm < h->fref[0][0]->f_row_qp[y] )
+ update_predictor( rc->row_pred[1], qscale, h->fdec->i_row_satds[0][0][y], h->fdec->i_row_bits[y] );
+
+ /* update ratecontrol per-mbpair in MBAFF */
+ if( SLICE_MBAFF && !(y&1) )
+ return;
/* tweak quality based on difference from predicted size */
if( y < h->i_threadslice_end-1 )
if( rc->rate_factor_max_increment )
qp_absolute_max = X264_MIN( qp_absolute_max, rc->qp_novbv + rc->rate_factor_max_increment );
float qp_max = X264_MIN( prev_row_qp + h->param.rc.i_qp_step, qp_absolute_max );
- float step_size = 0.5;
+ float step_size = 0.5f;
/* B-frames shouldn't use lower QP than their reference frames. */
if( h->sh.i_type == SLICE_TYPE_B )
{
- qp_min = X264_MAX( qp_min, X264_MAX( h->fref0[0]->f_row_qp[y+1], h->fref1[0]->f_row_qp[y+1] ) );
+ qp_min = X264_MAX( qp_min, X264_MAX( h->fref[0][0]->f_row_qp[y+1], h->fref[1][0]->f_row_qp[y+1] ) );
rc->qpm = X264_MAX( rc->qpm, qp_min );
}
float buffer_left_planned = rc->buffer_fill - rc->frame_size_planned;
float slice_size_planned = h->param.b_sliced_threads ? rc->slice_size_planned : rc->frame_size_planned;
- float max_frame_error = X264_MAX( 0.05, 1.0 / (h->mb.i_mb_height) );
+ float max_frame_error = X264_MAX( 0.05f, 1.0f / h->mb.i_mb_height );
float size_of_other_slices = 0;
if( h->param.b_sliced_threads )
{
/* More threads means we have to be more cautious in letting ratecontrol use up extra bits. */
float rc_tol = buffer_left_planned / h->param.i_threads * rc->rate_tolerance;
- int b1 = predict_row_size_sum( h, y, rc->qpm ) + size_of_other_slices;
+ float b1 = predict_row_size_sum( h, y, rc->qpm ) + size_of_other_slices;
/* Don't modify the row QPs until a sufficent amount of the bits of the frame have been processed, in case a flat */
/* area at the top of the frame was measured inaccurately. */
- if( row_bits_so_far( h, y ) < 0.05 * slice_size_planned )
+ if( row_bits_so_far( h, y ) < 0.05f * slice_size_planned )
return;
if( h->sh.i_type != SLICE_TYPE_I )
- rc_tol /= 2;
+ rc_tol *= 0.5f;
if( !rc->b_vbv_min_rate )
qp_min = X264_MAX( qp_min, rc->qp_novbv );
while( rc->qpm < qp_max
&& ((b1 > rc->frame_size_planned + rc_tol) ||
- (rc->buffer_fill - b1 < buffer_left_planned * 0.5) ||
+ (rc->buffer_fill - b1 < buffer_left_planned * 0.5f) ||
(b1 > rc->frame_size_planned && rc->qpm < rc->qp_novbv)) )
{
rc->qpm += step_size;
while( rc->qpm > qp_min
&& (rc->qpm > h->fdec->f_row_qp[0] || rc->single_frame_vbv)
- && ((b1 < rc->frame_size_planned * 0.8 && rc->qpm <= prev_row_qp)
- || b1 < (rc->buffer_fill - rc->buffer_size + rc->buffer_rate) * 1.1) )
+ && ((b1 < rc->frame_size_planned * 0.8f && rc->qpm <= prev_row_qp)
+ || b1 < (rc->buffer_fill - rc->buffer_size + rc->buffer_rate) * 1.1f) )
{
rc->qpm -= step_size;
b1 = predict_row_size_sum( h, y, rc->qpm ) + size_of_other_slices;
b1 = predict_row_size_sum( h, y, rc->qpm ) + size_of_other_slices;
}
- h->rc->frame_size_estimated = predict_row_size_sum( h, y, rc->qpm );
+ h->rc->frame_size_estimated = b1 - size_of_other_slices;
}
+ else
+ h->rc->frame_size_estimated = predict_row_size_sum( h, y, rc->qpm );
}
int x264_ratecontrol_qp( x264_t *h )
{
x264_emms();
- return x264_clip3( h->rc->qpm + .5, h->param.rc.i_qp_min, h->param.rc.i_qp_max );
+ return x264_clip3( h->rc->qpm + 0.5f, h->param.rc.i_qp_min, h->param.rc.i_qp_max );
}
int x264_ratecontrol_mb_qp( x264_t *h )
x264_emms();
float qp = h->rc->qpm;
if( h->param.rc.i_aq_mode )
- /* MB-tree currently doesn't adjust quantizers in unreferenced frames. */
- qp += h->fdec->b_kept_as_ref ? h->fenc->f_qp_offset[h->mb.i_mb_xy] : h->fenc->f_qp_offset_aq[h->mb.i_mb_xy];
- return x264_clip3( qp + .5, h->param.rc.i_qp_min, h->param.rc.i_qp_max );
+ {
+ /* MB-tree currently doesn't adjust quantizers in unreferenced frames. */
+ float qp_offset = h->fdec->b_kept_as_ref ? h->fenc->f_qp_offset[h->mb.i_mb_xy] : h->fenc->f_qp_offset_aq[h->mb.i_mb_xy];
+ /* Scale AQ's effect towards zero in emergency mode. */
+ if( qp > QP_MAX_SPEC )
+ qp_offset *= (QP_MAX - qp) / (QP_MAX - QP_MAX_SPEC);
+ qp += qp_offset;
+ }
+ return x264_clip3( qp + 0.5f, h->param.rc.i_qp_min, h->param.rc.i_qp_max );
}
/* In 2pass, force the same frame types as in the 1st pass */
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, QP_MAX );
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, QP_MAX );
- 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);
+ 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.i_bframe_adaptive )
- x264_log(h, X264_LOG_ERROR, "disabling adaptive B-frames\n");
+ x264_log( h, X264_LOG_ERROR, "disabling adaptive B-frames\n" );
for( int i = 0; i < h->param.i_threads; i++ )
{
h->stat.frame.i_mb_count_p += mbs[i];
h->fdec->f_qp_avg_rc = rc->qpa_rc /= h->mb.i_mb_count;
- h->fdec->f_qp_avg_aq = rc->qpa_aq /= h->mb.i_mb_count;
+ h->fdec->f_qp_avg_aq = (float)rc->qpa_aq / h->mb.i_mb_count;
if( h->param.rc.b_stat_write )
{
dir_avg>0 ? 's' : dir_avg<0 ? 't' : '-' )
: '-';
if( fprintf( rc->p_stat_file_out,
- "in:%d out:%d type:%c dur:%d cpbdur:%d q:%.2f tex:%d mv:%d misc:%d imb:%d pmb:%d smb:%d d:%c ref:",
+ "in:%d out:%d type:%c dur:%"PRId64" cpbdur:%"PRId64" q:%.2f tex:%d mv:%d misc:%d imb:%d pmb:%d smb:%d d:%c ref:",
h->fenc->i_frame, h->i_frame,
c_type, h->fenc->i_duration,
h->fenc->i_cpb_duration, rc->qpa_rc,
/* Only write information for reference reordering once. */
int use_old_stats = h->param.rc.b_stat_read && rc->rce->refs > 1;
- for( int i = 0; i < (use_old_stats ? rc->rce->refs : h->i_ref0); i++ )
+ for( int i = 0; i < (use_old_stats ? rc->rce->refs : h->i_ref[0]); i++ )
{
int refcount = use_old_stats ? rc->rce->refcount[i]
- : h->param.b_interlaced ? h->stat.frame.i_mb_count_ref[0][i*2]
+ : PARAM_INTERLACED ? h->stat.frame.i_mb_count_ref[0][i*2]
+ h->stat.frame.i_mb_count_ref[0][i*2+1]
: h->stat.frame.i_mb_count_ref[0][i];
if( fprintf( rc->p_stat_file_out, "%d ", refcount ) < 0 )
goto fail;
if( h->sh.weight[0][1].weightfn || h->sh.weight[0][2].weightfn )
{
- if( fprintf( rc->p_stat_file_out, ",%d,%d,%d,%d,%d\n",
+ if( fprintf( rc->p_stat_file_out, ",%d,%d,%d,%d,%d ",
h->sh.weight[0][1].i_denom, h->sh.weight[0][1].i_scale, h->sh.weight[0][1].i_offset,
h->sh.weight[0][2].i_scale, h->sh.weight[0][2].i_offset ) < 0 )
goto fail;
}
- else if( fprintf( rc->p_stat_file_out, "\n" ) < 0 )
+ else if( fprintf( rc->p_stat_file_out, " " ) < 0 )
goto fail;
}
rc->cplxr_sum += bits * qp2qscale( rc->qpa_rc ) / (rc->last_rceq * fabs( h->param.rc.f_pb_factor ));
}
rc->cplxr_sum *= rc->cbr_decay;
- double frame_duration = (double)h->fenc->i_duration * h->sps->vui.i_num_units_in_tick / h->sps->vui.i_time_scale;
-
- rc->wanted_bits_window += frame_duration * rc->bitrate;
+ rc->wanted_bits_window += h->fenc->f_duration * rc->bitrate;
rc->wanted_bits_window *= rc->cbr_decay;
}
if( h->fenc->b_last_minigop_bframe )
{
update_predictor( rc->pred_b_from_p, qp2qscale( rc->qpa_rc ),
- h->fref1[h->i_ref1-1]->i_satd, rc->bframe_bits / rc->bframes );
+ h->fref[1][h->i_ref[1]-1]->i_satd, rc->bframe_bits / rc->bframes );
rc->bframe_bits = 0;
}
}
return 0;
fail:
- x264_log(h, X264_LOG_ERROR, "ratecontrol_end: stats file could not be written to\n");
+ x264_log( h, X264_LOG_ERROR, "ratecontrol_end: stats file could not be written to\n" );
return -1;
}
{
x264_ratecontrol_t *rcc= h->rc;
x264_zone_t *zone = get_zone( h, frame_num );
- double q = pow( rce->blurred_complexity, 1 - rcc->qcompress );
+ double q;
+ if( h->param.rc.b_mb_tree )
+ {
+ double timescale = (double)h->sps->vui.i_num_units_in_tick / h->sps->vui.i_time_scale;
+ q = pow( BASE_FRAME_DURATION / CLIP_DURATION(rce->i_duration * timescale), 1 - h->param.rc.f_qcompress );
+ }
+ else
+ q = pow( rce->blurred_complexity, 1 - rcc->qcompress );
// avoid NaN's in the rc_eq
if( !isfinite(q) || rce->tex_bits + rce->mv_bits == 0 )
return q;
}
-static double get_diff_limited_q(x264_t *h, ratecontrol_entry_t *rce, double q)
+static double get_diff_limited_q(x264_t *h, ratecontrol_entry_t *rce, double q, int frame_num)
{
x264_ratecontrol_t *rcc = h->rc;
const int pict_type = rce->pict_type;
+ x264_zone_t *zone = get_zone( h, frame_num );
// force I/B quants as a function of P quants
const double last_p_q = rcc->last_qscale_for[SLICE_TYPE_P];
rcc->accum_p_qp = mask * (qscale2qp( q ) + rcc->accum_p_qp);
rcc->accum_p_norm = mask * (1 + rcc->accum_p_norm);
}
+
+ if( zone )
+ {
+ if( zone->b_force_qp )
+ q = qp2qscale( zone->i_qp );
+ else
+ q /= zone->f_bitrate_factor;
+ }
+
return q;
}
-static double predict_size( predictor_t *p, double q, double var )
+static float predict_size( predictor_t *p, float q, float var )
{
- return (p->coeff*var + p->offset) / (q*p->count);
+ return (p->coeff*var + p->offset) / (q*p->count);
}
-static void update_predictor( predictor_t *p, double q, double var, double bits )
+static void update_predictor( predictor_t *p, float q, float var, float bits )
{
- const double range = 1.5;
+ float range = 1.5;
if( var < 10 )
return;
- double old_coeff = p->coeff / p->count;
- double new_coeff = bits*q / var;
- double new_coeff_clipped = x264_clip3f( new_coeff, old_coeff/range, old_coeff*range );
- double new_offset = bits*q - new_coeff_clipped * var;
+ float old_coeff = p->coeff / p->count;
+ float new_coeff = bits*q / var;
+ float new_coeff_clipped = x264_clip3f( new_coeff, old_coeff/range, old_coeff*range );
+ float new_offset = bits*q - new_coeff_clipped * var;
if( new_offset >= 0 )
new_coeff = new_coeff_clipped;
else
if( h->sps->vui.hrd.b_cbr_hrd && rct->buffer_fill_final > buffer_size )
{
- filler = ceil( (rct->buffer_fill_final - buffer_size) / (8. * h->sps->vui.i_time_scale) );
+ int64_t scale = (int64_t)h->sps->vui.i_time_scale * 8;
+ filler = (rct->buffer_fill_final - buffer_size + scale - 1) / scale;
bits = X264_MAX( (FILLER_OVERHEAD - h->param.b_annexb), filler ) * 8;
rct->buffer_fill_final -= (uint64_t)bits * h->sps->vui.i_time_scale;
}
double bits = t->rc->frame_size_planned;
if( !t->b_thread_active )
continue;
- bits = X264_MAX(bits, t->rc->frame_size_estimated);
+ bits = X264_MAX(bits, t->rc->frame_size_estimated);
rcc->buffer_fill -= bits;
rcc->buffer_fill = X264_MAX( rcc->buffer_fill, 0 );
rcc->buffer_fill += t->rc->buffer_rate;
{
float q;
x264_ratecontrol_t *rcc = h->rc;
- ratecontrol_entry_t rce;
+ ratecontrol_entry_t UNINIT(rce);
int pict_type = h->sh.i_type;
int64_t total_bits = 8*(h->stat.i_frame_size[SLICE_TYPE_I]
+ h->stat.i_frame_size[SLICE_TYPE_P]
/* 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_rc;
- float q1 = h->fref1[0]->f_qp_avg_rc;
+ int i0 = IS_X264_TYPE_I(h->fref_nearest[0]->i_type);
+ int i1 = IS_X264_TYPE_I(h->fref_nearest[1]->i_type);
+ int dt0 = abs(h->fenc->i_poc - h->fref_nearest[0]->i_poc);
+ int dt1 = abs(h->fenc->i_poc - h->fref_nearest[1]->i_poc);
+ float q0 = h->fref_nearest[0]->f_qp_avg_rc;
+ float q1 = h->fref_nearest[1]->f_qp_avg_rc;
- if( h->fref0[0]->i_type == X264_TYPE_BREF )
+ if( h->fref_nearest[0]->i_type == X264_TYPE_BREF )
q0 -= rcc->pb_offset/2;
- if( h->fref1[0]->i_type == X264_TYPE_BREF )
+ if( h->fref_nearest[1]->i_type == X264_TYPE_BREF )
q1 -= rcc->pb_offset/2;
if( i0 && i1 )
if( rcc->b_2pass && rcc->b_vbv )
rcc->frame_size_planned = qscale2bits( &rce, qp2qscale( q ) );
else
- rcc->frame_size_planned = predict_size( rcc->pred_b_from_p, qp2qscale( q ), h->fref1[h->i_ref1-1]->i_satd );
+ rcc->frame_size_planned = predict_size( rcc->pred_b_from_p, qp2qscale( q ), h->fref[1][h->i_ref[1]-1]->i_satd );
+ /* Limit planned size by MinCR */
+ if( rcc->b_vbv )
+ rcc->frame_size_planned = X264_MIN( rcc->frame_size_planned, rcc->frame_size_maximum );
h->rc->frame_size_estimated = rcc->frame_size_planned;
/* For row SATDs */
double bits = t->rc->frame_size_planned;
if( !t->b_thread_active )
continue;
- bits = X264_MAX(bits, t->rc->frame_size_estimated);
+ bits = X264_MAX(bits, t->rc->frame_size_estimated);
predicted_bits += (int64_t)bits;
}
}
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_cplxsum += rcc->last_satd / (CLIP_DURATION(h->fenc->f_duration) / BASE_FRAME_DURATION);
rcc->short_term_cplxcount ++;
rce.tex_bits = rcc->last_satd;
rce.s_count = 0;
rce.qscale = 1;
rce.pict_type = pict_type;
+ rce.i_duration = h->fenc->i_duration;
if( h->param.rc.i_rc_method == X264_RC_CRF )
{
/* Always use up the whole VBV in this case. */
if( rcc->single_frame_vbv )
rcc->frame_size_planned = rcc->buffer_rate;
+ /* Limit planned size by MinCR */
+ if( rcc->b_vbv )
+ rcc->frame_size_planned = X264_MIN( rcc->frame_size_planned, rcc->frame_size_maximum );
h->rc->frame_size_estimated = rcc->frame_size_planned;
return q;
}
}
-void x264_threads_normalize_predictors( x264_t *h )
+static void x264_threads_normalize_predictors( x264_t *h )
{
double totalsize = 0;
for( int i = 0; i < h->param.i_threads; i++ )
COPY(prev_zone);
COPY(qpbuf_pos);
/* these vars can be updated by x264_ratecontrol_init_reconfigurable */
- COPY(buffer_rate);
+ COPY(bitrate);
COPY(buffer_size);
+ COPY(buffer_rate);
+ COPY(vbv_max_rate);
COPY(single_frame_vbv);
COPY(cbr_decay);
- COPY(b_vbv_min_rate);
COPY(rate_factor_constant);
- COPY(bitrate);
+ COPY(rate_factor_max_increment);
#undef COPY
}
if( cur != next )
{
x264_ratecontrol_t *rcc = h->rc;
uint64_t all_const_bits = 0;
+ double timescale = (double)h->sps->vui.i_num_units_in_tick / h->sps->vui.i_time_scale;
double duration = 0;
for( int i = 0; i < rcc->num_entries; i++ )
duration += rcc->entry[i].i_duration;
- duration *= (double)h->sps->vui.i_num_units_in_tick / h->sps->vui.i_time_scale;
+ duration *= timescale;
uint64_t all_available_bits = h->param.rc.i_bitrate * 1000. * duration;
double rate_factor, step_mult;
double qblur = h->param.rc.f_qblur;
for( int j = 1; j < cplxblur*2 && j < rcc->num_entries-i; j++ )
{
ratecontrol_entry_t *rcj = &rcc->entry[i+j];
+ double frame_duration = CLIP_DURATION(rcj->i_duration * timescale) / BASE_FRAME_DURATION;
weight *= 1 - pow( (float)rcj->i_count / rcc->nmb, 2 );
if( weight < .0001 )
break;
gaussian_weight = weight * exp( -j*j/200.0 );
weight_sum += gaussian_weight;
- cplx_sum += gaussian_weight * (qscale2bits(rcj, 1) - rcj->misc_bits);
+ cplx_sum += gaussian_weight * (qscale2bits( rcj, 1 ) - rcj->misc_bits) / frame_duration;
}
/* weighted average of cplx of past frames */
weight = 1.0;
for( int j = 0; j <= cplxblur*2 && j <= i; j++ )
{
ratecontrol_entry_t *rcj = &rcc->entry[i-j];
+ double frame_duration = CLIP_DURATION(rcj->i_duration * timescale) / BASE_FRAME_DURATION;
gaussian_weight = weight * exp( -j*j/200.0 );
weight_sum += gaussian_weight;
- cplx_sum += gaussian_weight * (qscale2bits( rcj, 1 ) - rcj->misc_bits);
+ cplx_sum += gaussian_weight * (qscale2bits( rcj, 1 ) - rcj->misc_bits) / frame_duration;
weight *= 1 - pow( (float)rcj->i_count / rcc->nmb, 2 );
if( weight < .0001 )
break;
/* find qscale */
for( int i = 0; i < rcc->num_entries; i++ )
{
- qscale[i] = get_qscale( h, &rcc->entry[i], rate_factor, i );
+ qscale[i] = get_qscale( h, &rcc->entry[i], rate_factor, -1 );
rcc->last_qscale_for[rcc->entry[i].pict_type] = qscale[i];
}
/* fixed I/B qscale relative to P */
for( int i = rcc->num_entries-1; i >= 0; i-- )
{
- qscale[i] = get_diff_limited_q( h, &rcc->entry[i], qscale[i] );
+ qscale[i] = get_diff_limited_q( h, &rcc->entry[i], qscale[i], i );
assert(qscale[i] >= 0);
}
projects / x264 / blobdiff