Use forced key frames when encoding with libxvid.
[ffmpeg.git] / libavcodec / ratecontrol.c
index 53d9005..4261678 100644 (file)
@@ -1,33 +1,38 @@
 /*
  * Rate control for video encoders
  *
- * Copyright (c) 2002-2003 Michael Niedermayer <michaelni@gmx.at>
+ * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
  *
- * This library is free software; you can redistribute it and/or
+ * 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 of the License, or (at your option) any later version.
+ * version 2.1 of the License, or (at your option) any later version.
  *
- * This library is distributed in the hope that it will be useful,
+ * 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 this library; if not, write to the Free Software
- * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
 /**
- * @file ratecontrol.c
+ * @file
  * Rate control for video encoders.
- */ 
+ */
 
+#include "libavutil/intmath.h"
 #include "avcodec.h"
 #include "dsputil.h"
+#include "ratecontrol.h"
 #include "mpegvideo.h"
+#include "libavutil/eval.h"
 
-#undef NDEBUG // allways check asserts, the speed effect is far too small to disable them
+#undef NDEBUG // Always check asserts, the speed effect is far too small to disable them.
 #include <assert.h>
 
 #ifndef M_E
@@ -38,28 +43,85 @@ static int init_pass2(MpegEncContext *s);
 static double get_qscale(MpegEncContext *s, RateControlEntry *rce, double rate_factor, int frame_num);
 
 void ff_write_pass1_stats(MpegEncContext *s){
-    sprintf(s->avctx->stats_out, "in:%d out:%d type:%d q:%d itex:%d ptex:%d mv:%d misc:%d fcode:%d bcode:%d mc-var:%d var:%d icount:%d;\n",
-            s->current_picture_ptr->display_picture_number, s->current_picture_ptr->coded_picture_number, s->pict_type, 
-            s->current_picture.quality, s->i_tex_bits, s->p_tex_bits, s->mv_bits, s->misc_bits, 
-            s->f_code, s->b_code, s->current_picture.mc_mb_var_sum, s->current_picture.mb_var_sum, s->i_count);
+    snprintf(s->avctx->stats_out, 256, "in:%d out:%d type:%d q:%d itex:%d ptex:%d mv:%d misc:%d fcode:%d bcode:%d mc-var:%d var:%d icount:%d skipcount:%d hbits:%d;\n",
+            s->current_picture_ptr->display_picture_number, s->current_picture_ptr->coded_picture_number, s->pict_type,
+            s->current_picture.quality, s->i_tex_bits, s->p_tex_bits, s->mv_bits, s->misc_bits,
+            s->f_code, s->b_code, s->current_picture.mc_mb_var_sum, s->current_picture.mb_var_sum, s->i_count, s->skip_count, s->header_bits);
+}
+
+static inline double qp2bits(RateControlEntry *rce, double qp){
+    if(qp<=0.0){
+        av_log(NULL, AV_LOG_ERROR, "qp<=0.0\n");
+    }
+    return rce->qscale * (double)(rce->i_tex_bits + rce->p_tex_bits+1)/ qp;
+}
+
+static inline double bits2qp(RateControlEntry *rce, double bits){
+    if(bits<0.9){
+        av_log(NULL, AV_LOG_ERROR, "bits<0.9\n");
+    }
+    return rce->qscale * (double)(rce->i_tex_bits + rce->p_tex_bits+1)/ bits;
 }
 
 int ff_rate_control_init(MpegEncContext *s)
 {
     RateControlContext *rcc= &s->rc_context;
-    int i;
+    int i, res;
+    static const char * const const_names[]={
+        "PI",
+        "E",
+        "iTex",
+        "pTex",
+        "tex",
+        "mv",
+        "fCode",
+        "iCount",
+        "mcVar",
+        "var",
+        "isI",
+        "isP",
+        "isB",
+        "avgQP",
+        "qComp",
+/*        "lastIQP",
+        "lastPQP",
+        "lastBQP",
+        "nextNonBQP",*/
+        "avgIITex",
+        "avgPITex",
+        "avgPPTex",
+        "avgBPTex",
+        "avgTex",
+        NULL
+    };
+    static double (* const func1[])(void *, double)={
+        (void *)bits2qp,
+        (void *)qp2bits,
+        NULL
+    };
+    static const char * const func1_names[]={
+        "bits2qp",
+        "qp2bits",
+        NULL
+    };
     emms_c();
 
+    res = av_parse_expr(&rcc->rc_eq_eval, s->avctx->rc_eq ? s->avctx->rc_eq : "tex^qComp", const_names, func1_names, func1, NULL, NULL, 0, s->avctx);
+    if (res < 0) {
+        av_log(s->avctx, AV_LOG_ERROR, "Error parsing rc_eq \"%s\"\n", s->avctx->rc_eq);
+        return res;
+    }
+
     for(i=0; i<5; i++){
         rcc->pred[i].coeff= FF_QP2LAMBDA * 7.0;
         rcc->pred[i].count= 1.0;
-    
+
         rcc->pred[i].decay= 0.4;
         rcc->i_cplx_sum [i]=
         rcc->p_cplx_sum [i]=
         rcc->mv_bits_sum[i]=
         rcc->qscale_sum [i]=
-        rcc->frame_count[i]= 1; // 1 is better cuz of 1/0 and such
+        rcc->frame_count[i]= 1; // 1 is better because of 1/0 and such
         rcc->last_qscale_for[i]=FF_QP2LAMBDA * 5;
     }
     rcc->buffer_index= s->avctx->rc_initial_buffer_occupancy;
@@ -74,18 +136,20 @@ int ff_rate_control_init(MpegEncContext *s)
             p= strchr(p+1, ';');
         }
         i+= s->max_b_frames;
-        rcc->entry = (RateControlEntry*)av_mallocz(i*sizeof(RateControlEntry));
+        if(i<=0 || i>=INT_MAX / sizeof(RateControlEntry))
+            return -1;
+        rcc->entry = av_mallocz(i*sizeof(RateControlEntry));
         rcc->num_entries= i;
-        
-        /* init all to skiped p frames (with b frames we might have a not encoded frame at the end FIXME) */
+
+        /* init all to skipped p frames (with b frames we might have a not encoded frame at the end FIXME) */
         for(i=0; i<rcc->num_entries; i++){
             RateControlEntry *rce= &rcc->entry[i];
-            rce->pict_type= rce->new_pict_type=P_TYPE;
+            rce->pict_type= rce->new_pict_type=FF_P_TYPE;
             rce->qscale= rce->new_qscale=FF_QP2LAMBDA * 2;
             rce->misc_bits= s->mb_num + 10;
             rce->mb_var_sum= s->mb_num*100;
-        }        
-        
+        }
+
         /* read stats */
         p= s->avctx->stats_in;
         for(i=0; i<rcc->num_entries - s->max_b_frames; i++){
@@ -96,7 +160,7 @@ int ff_rate_control_init(MpegEncContext *s)
 
             next= strchr(p, ';');
             if(next){
-                (*next)=0; //sscanf in unbelieavle slow on looong strings //FIXME copy / dont write
+                (*next)=0; //sscanf in unbelievably slow on looong strings //FIXME copy / do not write
                 next++;
             }
             e= sscanf(p, " in:%d ", &picture_number);
@@ -105,37 +169,51 @@ int ff_rate_control_init(MpegEncContext *s)
             assert(picture_number < rcc->num_entries);
             rce= &rcc->entry[picture_number];
 
-            e+=sscanf(p, " in:%*d out:%*d type:%d q:%f itex:%d ptex:%d mv:%d misc:%d fcode:%d bcode:%d mc-var:%d var:%d icount:%d",
-                   &rce->pict_type, &rce->qscale, &rce->i_tex_bits, &rce->p_tex_bits, &rce->mv_bits, &rce->misc_bits, 
-                   &rce->f_code, &rce->b_code, &rce->mc_mb_var_sum, &rce->mb_var_sum, &rce->i_count);
-            if(e!=12){
+            e+=sscanf(p, " in:%*d out:%*d type:%d q:%f itex:%d ptex:%d mv:%d misc:%d fcode:%d bcode:%d mc-var:%d var:%d icount:%d skipcount:%d hbits:%d",
+                   &rce->pict_type, &rce->qscale, &rce->i_tex_bits, &rce->p_tex_bits, &rce->mv_bits, &rce->misc_bits,
+                   &rce->f_code, &rce->b_code, &rce->mc_mb_var_sum, &rce->mb_var_sum, &rce->i_count, &rce->skip_count, &rce->header_bits);
+            if(e!=14){
                 av_log(s->avctx, AV_LOG_ERROR, "statistics are damaged at line %d, parser out=%d\n", i, e);
                 return -1;
             }
+
             p= next;
         }
-        
+
         if(init_pass2(s) < 0) return -1;
+
+        //FIXME maybe move to end
+        if((s->flags&CODEC_FLAG_PASS2) && s->avctx->rc_strategy == FF_RC_STRATEGY_XVID) {
+#if CONFIG_LIBXVID
+            return ff_xvid_rate_control_init(s);
+#else
+            av_log(s->avctx, AV_LOG_ERROR, "Xvid ratecontrol requires libavcodec compiled with Xvid support.\n");
+            return -1;
+#endif
+        }
     }
-     
+
     if(!(s->flags&CODEC_FLAG_PASS2)){
 
         rcc->short_term_qsum=0.001;
         rcc->short_term_qcount=0.001;
-    
+
         rcc->pass1_rc_eq_output_sum= 0.001;
         rcc->pass1_wanted_bits=0.001;
-        
+
+        if(s->avctx->qblur > 1.0){
+            av_log(s->avctx, AV_LOG_ERROR, "qblur too large\n");
+            return -1;
+        }
         /* init stuff with the user specified complexity */
         if(s->avctx->rc_initial_cplx){
             for(i=0; i<60*30; i++){
                 double bits= s->avctx->rc_initial_cplx * (i/10000.0 + 1.0)*s->mb_num;
                 RateControlEntry rce;
-                double q;
-                
-                if     (i%((s->gop_size+3)/4)==0) rce.pict_type= I_TYPE;
-                else if(i%(s->max_b_frames+1))    rce.pict_type= B_TYPE;
-                else                              rce.pict_type= P_TYPE;
+
+                if     (i%((s->gop_size+3)/4)==0) rce.pict_type= FF_I_TYPE;
+                else if(i%(s->max_b_frames+1))    rce.pict_type= FF_B_TYPE;
+                else                              rce.pict_type= FF_P_TYPE;
 
                 rce.new_pict_type= rce.pict_type;
                 rce.mc_mb_var_sum= bits*s->mb_num/100000;
@@ -145,7 +223,7 @@ int ff_rate_control_init(MpegEncContext *s)
                 rce.b_code   = 1;
                 rce.misc_bits= 1;
 
-                if(s->pict_type== I_TYPE){
+                if(s->pict_type== FF_I_TYPE){
                     rce.i_count   = s->mb_num;
                     rce.i_tex_bits= bits;
                     rce.p_tex_bits= 0;
@@ -161,15 +239,13 @@ int ff_rate_control_init(MpegEncContext *s)
                 rcc->mv_bits_sum[rce.pict_type] += rce.mv_bits;
                 rcc->frame_count[rce.pict_type] ++;
 
-                bits= rce.i_tex_bits + rce.p_tex_bits;
-
-                q= get_qscale(s, &rce, rcc->pass1_wanted_bits/rcc->pass1_rc_eq_output_sum, i);
-                rcc->pass1_wanted_bits+= s->bit_rate/(s->avctx->frame_rate / (double)s->avctx->frame_rate_base);
+                get_qscale(s, &rce, rcc->pass1_wanted_bits/rcc->pass1_rc_eq_output_sum, i);
+                rcc->pass1_wanted_bits+= s->bit_rate/(1/av_q2d(s->avctx->time_base)); //FIXME misbehaves a little for variable fps
             }
         }
 
     }
-    
+
     return 0;
 }
 
@@ -178,30 +254,22 @@ void ff_rate_control_uninit(MpegEncContext *s)
     RateControlContext *rcc= &s->rc_context;
     emms_c();
 
+    av_free_expr(rcc->rc_eq_eval);
     av_freep(&rcc->entry);
-}
 
-static inline double qp2bits(RateControlEntry *rce, double qp){
-    if(qp<=0.0){
-        av_log(NULL, AV_LOG_ERROR, "qp<=0.0\n");
-    }
-    return rce->qscale * (double)(rce->i_tex_bits + rce->p_tex_bits+1)/ qp;
+#if CONFIG_LIBXVID
+    if((s->flags&CODEC_FLAG_PASS2) && s->avctx->rc_strategy == FF_RC_STRATEGY_XVID)
+        ff_xvid_rate_control_uninit(s);
+#endif
 }
 
-static inline double bits2qp(RateControlEntry *rce, double bits){
-    if(bits<0.9){
-        av_log(NULL, AV_LOG_ERROR, "bits<0.9\n");
-    }
-    return rce->qscale * (double)(rce->i_tex_bits + rce->p_tex_bits+1)/ bits;
-}
-    
 int ff_vbv_update(MpegEncContext *s, int frame_size){
     RateControlContext *rcc= &s->rc_context;
-    const double fps= (double)s->avctx->frame_rate / (double)s->avctx->frame_rate_base;
+    const double fps= 1/av_q2d(s->avctx->time_base);
     const int buffer_size= s->avctx->rc_buffer_size;
     const double min_rate= s->avctx->rc_min_rate/fps;
     const double max_rate= s->avctx->rc_max_rate/fps;
-    
+
 //printf("%d %f %d %f %f\n", buffer_size, rcc->buffer_index, frame_size, min_rate, max_rate);
     if(buffer_size){
         int left;
@@ -213,15 +281,15 @@ int ff_vbv_update(MpegEncContext *s, int frame_size){
         }
 
         left= buffer_size - rcc->buffer_index - 1;
-        rcc->buffer_index += clip(left, min_rate, max_rate);
+        rcc->buffer_index += av_clip(left, min_rate, max_rate);
 
         if(rcc->buffer_index > buffer_size){
             int stuffing= ceil((rcc->buffer_index - buffer_size)/8);
-            
+
             if(stuffing < 4 && s->codec_id == CODEC_ID_MPEG4)
                 stuffing=4;
             rcc->buffer_index -= 8*stuffing;
-            
+
             if(s->avctx->debug & FF_DEBUG_RC)
                 av_log(s->avctx, AV_LOG_DEBUG, "stuffing %d bytes\n", stuffing);
 
@@ -239,7 +307,7 @@ static double get_qscale(MpegEncContext *s, RateControlEntry *rce, double rate_f
     AVCodecContext *a= s->avctx;
     double q, bits;
     const int pict_type= rce->new_pict_type;
-    const double mb_num= s->mb_num;  
+    const double mb_num= s->mb_num;
     int i;
 
     double const_values[]={
@@ -249,91 +317,59 @@ static double get_qscale(MpegEncContext *s, RateControlEntry *rce, double rate_f
         rce->p_tex_bits*rce->qscale,
         (rce->i_tex_bits + rce->p_tex_bits)*(double)rce->qscale,
         rce->mv_bits/mb_num,
-        rce->pict_type == B_TYPE ? (rce->f_code + rce->b_code)*0.5 : rce->f_code,
+        rce->pict_type == FF_B_TYPE ? (rce->f_code + rce->b_code)*0.5 : rce->f_code,
         rce->i_count/mb_num,
         rce->mc_mb_var_sum/mb_num,
         rce->mb_var_sum/mb_num,
-        rce->pict_type == I_TYPE,
-        rce->pict_type == P_TYPE,
-        rce->pict_type == B_TYPE,
+        rce->pict_type == FF_I_TYPE,
+        rce->pict_type == FF_P_TYPE,
+        rce->pict_type == FF_B_TYPE,
         rcc->qscale_sum[pict_type] / (double)rcc->frame_count[pict_type],
         a->qcompress,
-/*        rcc->last_qscale_for[I_TYPE],
-        rcc->last_qscale_for[P_TYPE],
-        rcc->last_qscale_for[B_TYPE],
+/*        rcc->last_qscale_for[FF_I_TYPE],
+        rcc->last_qscale_for[FF_P_TYPE],
+        rcc->last_qscale_for[FF_B_TYPE],
         rcc->next_non_b_qscale,*/
-        rcc->i_cplx_sum[I_TYPE] / (double)rcc->frame_count[I_TYPE],
-        rcc->i_cplx_sum[P_TYPE] / (double)rcc->frame_count[P_TYPE],
-        rcc->p_cplx_sum[P_TYPE] / (double)rcc->frame_count[P_TYPE],
-        rcc->p_cplx_sum[B_TYPE] / (double)rcc->frame_count[B_TYPE],
+        rcc->i_cplx_sum[FF_I_TYPE] / (double)rcc->frame_count[FF_I_TYPE],
+        rcc->i_cplx_sum[FF_P_TYPE] / (double)rcc->frame_count[FF_P_TYPE],
+        rcc->p_cplx_sum[FF_P_TYPE] / (double)rcc->frame_count[FF_P_TYPE],
+        rcc->p_cplx_sum[FF_B_TYPE] / (double)rcc->frame_count[FF_B_TYPE],
         (rcc->i_cplx_sum[pict_type] + rcc->p_cplx_sum[pict_type]) / (double)rcc->frame_count[pict_type],
         0
     };
-    static const char *const_names[]={
-        "PI",
-        "E",
-        "iTex",
-        "pTex",
-        "tex",
-        "mv",
-        "fCode",
-        "iCount",
-        "mcVar",
-        "var",
-        "isI",
-        "isP",
-        "isB",
-        "avgQP",
-        "qComp",
-/*        "lastIQP",
-        "lastPQP",
-        "lastBQP",
-        "nextNonBQP",*/
-        "avgIITex",
-        "avgPITex",
-        "avgPPTex",
-        "avgBPTex",
-        "avgTex",
-        NULL
-    };
-    static double (*func1[])(void *, double)={
-        (void *)bits2qp,
-        (void *)qp2bits,
-        NULL
-    };
-    static const char *func1_names[]={
-        "bits2qp",
-        "qp2bits",
-        NULL
-    };
 
-    bits= ff_eval(s->avctx->rc_eq, const_values, const_names, func1, func1_names, NULL, NULL, rce);
-    
+    bits = av_eval_expr(rcc->rc_eq_eval, const_values, rce);
+    if (isnan(bits)) {
+        av_log(s->avctx, AV_LOG_ERROR, "Error evaluating rc_eq \"%s\"\n", s->avctx->rc_eq);
+        return -1;
+    }
+
     rcc->pass1_rc_eq_output_sum+= bits;
     bits*=rate_factor;
     if(bits<0.0) bits=0.0;
     bits+= 1.0; //avoid 1/0 issues
-    
+
     /* user override */
     for(i=0; i<s->avctx->rc_override_count; i++){
         RcOverride *rco= s->avctx->rc_override;
         if(rco[i].start_frame > frame_num) continue;
         if(rco[i].end_frame   < frame_num) continue;
-    
-        if(rco[i].qscale) 
+
+        if(rco[i].qscale)
             bits= qp2bits(rce, rco[i].qscale); //FIXME move at end to really force it?
         else
             bits*= rco[i].quality_factor;
     }
 
     q= bits2qp(rce, bits);
-    
+
     /* I/B difference */
-    if     (pict_type==I_TYPE && s->avctx->i_quant_factor<0.0)
+    if     (pict_type==FF_I_TYPE && s->avctx->i_quant_factor<0.0)
         q= -q*s->avctx->i_quant_factor + s->avctx->i_quant_offset;
-    else if(pict_type==B_TYPE && s->avctx->b_quant_factor<0.0)
+    else if(pict_type==FF_B_TYPE && s->avctx->b_quant_factor<0.0)
         q= -q*s->avctx->b_quant_factor + s->avctx->b_quant_offset;
-        
+    if(q<1) q=1;
+
     return q;
 }
 
@@ -341,16 +377,17 @@ static double get_diff_limited_q(MpegEncContext *s, RateControlEntry *rce, doubl
     RateControlContext *rcc= &s->rc_context;
     AVCodecContext *a= s->avctx;
     const int pict_type= rce->new_pict_type;
-    const double last_p_q    = rcc->last_qscale_for[P_TYPE];
+    const double last_p_q    = rcc->last_qscale_for[FF_P_TYPE];
     const double last_non_b_q= rcc->last_qscale_for[rcc->last_non_b_pict_type];
-    
-    if     (pict_type==I_TYPE && (a->i_quant_factor>0.0 || rcc->last_non_b_pict_type==P_TYPE))
-        q= last_p_q    *ABS(a->i_quant_factor) + a->i_quant_offset;
-    else if(pict_type==B_TYPE && a->b_quant_factor>0.0)
+
+    if     (pict_type==FF_I_TYPE && (a->i_quant_factor>0.0 || rcc->last_non_b_pict_type==FF_P_TYPE))
+        q= last_p_q    *FFABS(a->i_quant_factor) + a->i_quant_offset;
+    else if(pict_type==FF_B_TYPE && a->b_quant_factor>0.0)
         q= last_non_b_q*    a->b_quant_factor  + a->b_quant_offset;
+    if(q<1) q=1;
 
     /* last qscale / qdiff stuff */
-    if(rcc->last_non_b_pict_type==pict_type || pict_type!=I_TYPE){
+    if(rcc->last_non_b_pict_type==pict_type || pict_type!=FF_I_TYPE){
         double last_q= rcc->last_qscale_for[pict_type];
         const int maxdiff= FF_QP2LAMBDA * a->max_qdiff;
 
@@ -358,9 +395,9 @@ static double get_diff_limited_q(MpegEncContext *s, RateControlEntry *rce, doubl
         else if(q < last_q - maxdiff) q= last_q - maxdiff;
     }
 
-    rcc->last_qscale_for[pict_type]= q; //Note we cant do that after blurring
-    
-    if(pict_type!=B_TYPE)
+    rcc->last_qscale_for[pict_type]= q; //Note we cannot do that after blurring
+
+    if(pict_type!=FF_B_TYPE)
         rcc->last_non_b_pict_type= pict_type;
 
     return q;
@@ -370,24 +407,24 @@ static double get_diff_limited_q(MpegEncContext *s, RateControlEntry *rce, doubl
  * gets the qmin & qmax for pict_type
  */
 static void get_qminmax(int *qmin_ret, int *qmax_ret, MpegEncContext *s, int pict_type){
-    int qmin= s->avctx->lmin;                                                       
+    int qmin= s->avctx->lmin;
     int qmax= s->avctx->lmax;
-    
+
     assert(qmin <= qmax);
 
-    if(pict_type==B_TYPE){
-        qmin= (int)(qmin*ABS(s->avctx->b_quant_factor)+s->avctx->b_quant_offset + 0.5);
-        qmax= (int)(qmax*ABS(s->avctx->b_quant_factor)+s->avctx->b_quant_offset + 0.5);
-    }else if(pict_type==I_TYPE){
-        qmin= (int)(qmin*ABS(s->avctx->i_quant_factor)+s->avctx->i_quant_offset + 0.5);
-        qmax= (int)(qmax*ABS(s->avctx->i_quant_factor)+s->avctx->i_quant_offset + 0.5);
+    if(pict_type==FF_B_TYPE){
+        qmin= (int)(qmin*FFABS(s->avctx->b_quant_factor)+s->avctx->b_quant_offset + 0.5);
+        qmax= (int)(qmax*FFABS(s->avctx->b_quant_factor)+s->avctx->b_quant_offset + 0.5);
+    }else if(pict_type==FF_I_TYPE){
+        qmin= (int)(qmin*FFABS(s->avctx->i_quant_factor)+s->avctx->i_quant_offset + 0.5);
+        qmax= (int)(qmax*FFABS(s->avctx->i_quant_factor)+s->avctx->i_quant_offset + 0.5);
     }
 
-    qmin= clip(qmin, 1, FF_LAMBDA_MAX);
-    qmax= clip(qmax, 1, FF_LAMBDA_MAX);
+    qmin= av_clip(qmin, 1, FF_LAMBDA_MAX);
+    qmax= av_clip(qmax, 1, FF_LAMBDA_MAX);
 
     if(qmax<qmin) qmax= qmin;
-    
+
     *qmin_ret= qmin;
     *qmax_ret= qmax;
 }
@@ -395,20 +432,18 @@ static void get_qminmax(int *qmin_ret, int *qmax_ret, MpegEncContext *s, int pic
 static double modify_qscale(MpegEncContext *s, RateControlEntry *rce, double q, int frame_num){
     RateControlContext *rcc= &s->rc_context;
     int qmin, qmax;
-    double bits;
     const int pict_type= rce->new_pict_type;
     const double buffer_size= s->avctx->rc_buffer_size;
-    const double fps= (double)s->avctx->frame_rate / (double)s->avctx->frame_rate_base;
+    const double fps= 1/av_q2d(s->avctx->time_base);
     const double min_rate= s->avctx->rc_min_rate / fps;
     const double max_rate= s->avctx->rc_max_rate / fps;
-    
+
     get_qminmax(&qmin, &qmax, s, pict_type);
 
     /* modulation */
-    if(s->avctx->rc_qmod_freq && frame_num%s->avctx->rc_qmod_freq==0 && pict_type==P_TYPE)
+    if(s->avctx->rc_qmod_freq && frame_num%s->avctx->rc_qmod_freq==0 && pict_type==FF_P_TYPE)
         q*= s->avctx->rc_qmod_amp;
 
-    bits= qp2bits(rce, q);
 //printf("q:%f\n", q);
     /* buffer overflow/underflow protection */
     if(buffer_size){
@@ -421,7 +456,7 @@ static double modify_qscale(MpegEncContext *s, RateControlEntry *rce, double q,
             else if(d<0.0001) d=0.0001;
             q*= pow(d, 1.0/s->avctx->rc_buffer_aggressivity);
 
-            q_limit= bits2qp(rce, FFMAX((min_rate - buffer_size + rcc->buffer_index)*3, 1));
+            q_limit= bits2qp(rce, FFMAX((min_rate - buffer_size + rcc->buffer_index) * s->avctx->rc_min_vbv_overflow_use, 1));
             if(q > q_limit){
                 if(s->avctx->debug&FF_DEBUG_RC){
                     av_log(s->avctx, AV_LOG_DEBUG, "limiting QP %f -> %f\n", q, q_limit);
@@ -436,7 +471,7 @@ static double modify_qscale(MpegEncContext *s, RateControlEntry *rce, double q,
             else if(d<0.0001) d=0.0001;
             q/= pow(d, 1.0/s->avctx->rc_buffer_aggressivity);
 
-            q_limit= bits2qp(rce, FFMAX(rcc->buffer_index/3, 1));
+            q_limit= bits2qp(rce, FFMAX(rcc->buffer_index * s->avctx->rc_max_available_vbv_use, 1));
             if(q < q_limit){
                 if(s->avctx->debug&FF_DEBUG_RC){
                     av_log(s->avctx, AV_LOG_DEBUG, "limiting QP %f -> %f\n", q, q_limit);
@@ -452,16 +487,16 @@ static double modify_qscale(MpegEncContext *s, RateControlEntry *rce, double q,
     }else{
         double min2= log(qmin);
         double max2= log(qmax);
-        
+
         q= log(q);
         q= (q - min2)/(max2-min2) - 0.5;
         q*= -4.0;
         q= 1.0/(1.0 + exp(q));
         q= q*(max2-min2) + min2;
-        
+
         q= exp(q);
     }
-    
+
     return q;
 }
 
@@ -499,28 +534,35 @@ static void adaptive_quantization(MpegEncContext *s, double q){
     const float temp_cplx_masking= s->avctx->temporal_cplx_masking;
     const float spatial_cplx_masking = s->avctx->spatial_cplx_masking;
     const float p_masking = s->avctx->p_masking;
+    const float border_masking = s->avctx->border_masking;
     float bits_sum= 0.0;
     float cplx_sum= 0.0;
     float cplx_tab[s->mb_num];
     float bits_tab[s->mb_num];
-    const int qmin= s->avctx->lmin;
-    const int qmax= s->avctx->lmax;
+    const int qmin= s->avctx->mb_lmin;
+    const int qmax= s->avctx->mb_lmax;
     Picture * const pic= &s->current_picture;
-    
+    const int mb_width = s->mb_width;
+    const int mb_height = s->mb_height;
+
     for(i=0; i<s->mb_num; i++){
         const int mb_xy= s->mb_index2xy[i];
         float temp_cplx= sqrt(pic->mc_mb_var[mb_xy]); //FIXME merge in pow()
         float spat_cplx= sqrt(pic->mb_var[mb_xy]);
         const int lumi= pic->mb_mean[mb_xy];
         float bits, cplx, factor;
-#if 0        
+        int mb_x = mb_xy % s->mb_stride;
+        int mb_y = mb_xy / s->mb_stride;
+        int mb_distance;
+        float mb_factor = 0.0;
+#if 0
         if(spat_cplx < q/3) spat_cplx= q/3; //FIXME finetune
         if(temp_cplx < q/3) temp_cplx= q/3; //FIXME finetune
-#endif   
+#endif
         if(spat_cplx < 4) spat_cplx= 4; //FIXME finetune
         if(temp_cplx < 4) temp_cplx= 4; //FIXME finetune
 
-        if((s->mb_type[mb_xy]&CANDIDATE_MB_TYPE_INTRA)){//FIXME hq mode 
+        if((s->mb_type[mb_xy]&CANDIDATE_MB_TYPE_INTRA)){//FIXME hq mode
             cplx= spat_cplx;
             factor= 1.0 + p_masking;
         }else{
@@ -533,9 +575,26 @@ static void adaptive_quantization(MpegEncContext *s, double q){
             factor*= (1.0 - (lumi-128)*(lumi-128)*lumi_masking);
         else
             factor*= (1.0 - (lumi-128)*(lumi-128)*dark_masking);
-        
+
+        if(mb_x < mb_width/5){
+            mb_distance = mb_width/5 - mb_x;
+            mb_factor = (float)mb_distance / (float)(mb_width/5);
+        }else if(mb_x > 4*mb_width/5){
+            mb_distance = mb_x - 4*mb_width/5;
+            mb_factor = (float)mb_distance / (float)(mb_width/5);
+        }
+        if(mb_y < mb_height/5){
+            mb_distance = mb_height/5 - mb_y;
+            mb_factor = FFMAX(mb_factor, (float)mb_distance / (float)(mb_height/5));
+        }else if(mb_y > 4*mb_height/5){
+            mb_distance = mb_y - 4*mb_height/5;
+            mb_factor = FFMAX(mb_factor, (float)mb_distance / (float)(mb_height/5));
+        }
+
+        factor*= 1.0 - border_masking*mb_factor;
+
         if(factor<0.00001) factor= 0.00001;
-        
+
         bits= cplx*factor;
         cplx_sum+= cplx;
         bits_sum+= bits;
@@ -543,11 +602,12 @@ static void adaptive_quantization(MpegEncContext *s, double q){
         bits_tab[i]= bits;
     }
 
-    /* handle qmin/qmax cliping */
+    /* handle qmin/qmax clipping */
     if(s->flags&CODEC_FLAG_NORMALIZE_AQP){
+        float factor= bits_sum/cplx_sum;
         for(i=0; i<s->mb_num; i++){
             float newq= q*cplx_tab[i]/bits_tab[i];
-            newq*= bits_sum/cplx_sum;
+            newq*= factor;
 
             if     (newq > qmax){
                 bits_sum -= bits_tab[i];
@@ -558,8 +618,10 @@ static void adaptive_quantization(MpegEncContext *s, double q){
                 cplx_sum -= cplx_tab[i]*q/qmin;
             }
         }
+        if(bits_sum < 0.001) bits_sum= 0.001;
+        if(cplx_sum < 0.001) cplx_sum= 0.001;
     }
-   
+
     for(i=0; i<s->mb_num; i++){
         const int mb_xy= s->mb_index2xy[i];
         float newq= q*cplx_tab[i]/bits_tab[i];
@@ -578,9 +640,20 @@ static void adaptive_quantization(MpegEncContext *s, double q){
         s->lambda_table[mb_xy]= intq;
     }
 }
+
+void ff_get_2pass_fcode(MpegEncContext *s){
+    RateControlContext *rcc= &s->rc_context;
+    int picture_number= s->picture_number;
+    RateControlEntry *rce;
+
+    rce= &rcc->entry[picture_number];
+    s->f_code= rce->f_code;
+    s->b_code= rce->b_code;
+}
+
 //FIXME rd or at least approx for dquant
 
-float ff_rate_estimate_qscale(MpegEncContext *s)
+float ff_rate_estimate_qscale(MpegEncContext *s, int dry_run)
 {
     float q;
     int qmin, qmax;
@@ -600,13 +673,18 @@ float ff_rate_estimate_qscale(MpegEncContext *s)
     Picture * const pic= &s->current_picture;
     emms_c();
 
+#if CONFIG_LIBXVID
+    if((s->flags&CODEC_FLAG_PASS2) && s->avctx->rc_strategy == FF_RC_STRATEGY_XVID)
+        return ff_xvid_rate_estimate_qscale(s, dry_run);
+#endif
+
     get_qminmax(&qmin, &qmax, s, pict_type);
 
-    fps= (double)s->avctx->frame_rate / (double)s->avctx->frame_rate_base;
+    fps= 1/av_q2d(s->avctx->time_base);
 //printf("input_pic_num:%d pic_num:%d frame_rate:%d\n", s->input_picture_number, s->picture_number, s->frame_rate);
         /* update predictors */
-    if(picture_number>2){
-        const int last_var= s->last_pict_type == I_TYPE ? rcc->last_mb_var_sum : rcc->last_mc_mb_var_sum;
+    if(picture_number>2 && !dry_run){
+        const int last_var= s->last_pict_type == FF_I_TYPE ? rcc->last_mb_var_sum : rcc->last_mc_mb_var_sum;
         update_predictor(&rcc->pred[s->last_pict_type], rcc->last_qscale, sqrt(last_var), s->frame_bits);
     }
 
@@ -616,25 +694,40 @@ float ff_rate_estimate_qscale(MpegEncContext *s)
         rce= &rcc->entry[picture_number];
         wanted_bits= rce->expected_bits;
     }else{
+        Picture *dts_pic;
         rce= &local_rce;
-        wanted_bits= (uint64_t)(s->bit_rate*(double)picture_number/fps);
+
+        //FIXME add a dts field to AVFrame and ensure its set and use it here instead of reordering
+        //but the reordering is simpler for now until h.264 b pyramid must be handeld
+        if(s->pict_type == FF_B_TYPE || s->low_delay)
+            dts_pic= s->current_picture_ptr;
+        else
+            dts_pic= s->last_picture_ptr;
+
+//if(dts_pic)
+//            av_log(NULL, AV_LOG_ERROR, "%Ld %Ld %Ld %d\n", s->current_picture_ptr->pts, s->user_specified_pts, dts_pic->pts, picture_number);
+
+        if(!dts_pic || dts_pic->pts == AV_NOPTS_VALUE)
+            wanted_bits= (uint64_t)(s->bit_rate*(double)picture_number/fps);
+        else
+            wanted_bits= (uint64_t)(s->bit_rate*(double)dts_pic->pts/fps);
     }
 
     diff= s->total_bits - wanted_bits;
     br_compensation= (a->bit_rate_tolerance - diff)/a->bit_rate_tolerance;
     if(br_compensation<=0.0) br_compensation=0.001;
 
-    var= pict_type == I_TYPE ? pic->mb_var_sum : pic->mc_mb_var_sum;
-    
+    var= pict_type == FF_I_TYPE ? pic->mb_var_sum : pic->mc_mb_var_sum;
+
     short_term_q = 0; /* avoid warning */
     if(s->flags&CODEC_FLAG_PASS2){
-        if(pict_type!=I_TYPE)
+        if(pict_type!=FF_I_TYPE)
             assert(pict_type == rce->new_pict_type);
 
         q= rce->new_qscale / br_compensation;
 //printf("%f %f %f last:%d var:%d type:%d//\n", q, rce->new_qscale, br_compensation, s->frame_bits, var, pict_type);
     }else{
-        rce->pict_type= 
+        rce->pict_type=
         rce->new_pict_type= pict_type;
         rce->mc_mb_var_sum= pic->mc_mb_var_sum;
         rce->mb_var_sum   = pic->   mb_var_sum;
@@ -644,7 +737,7 @@ float ff_rate_estimate_qscale(MpegEncContext *s)
         rce->misc_bits= 1;
 
         bits= predict_size(&rcc->pred[pict_type], rce->qscale, sqrt(var));
-        if(pict_type== I_TYPE){
+        if(pict_type== FF_I_TYPE){
             rce->i_count   = s->mb_num;
             rce->i_tex_bits= bits;
             rce->p_tex_bits= 0;
@@ -653,7 +746,7 @@ float ff_rate_estimate_qscale(MpegEncContext *s)
             rce->i_count   = 0; //FIXME we do know this approx
             rce->i_tex_bits= 0;
             rce->p_tex_bits= bits*0.9;
-            
+
             rce->mv_bits= bits*0.1;
         }
         rcc->i_cplx_sum [pict_type] += rce->i_tex_bits*rce->qscale;
@@ -663,8 +756,10 @@ float ff_rate_estimate_qscale(MpegEncContext *s)
 
         bits= rce->i_tex_bits + rce->p_tex_bits;
         rate_factor= rcc->pass1_wanted_bits/rcc->pass1_rc_eq_output_sum * br_compensation;
-    
+
         q= get_qscale(s, rce, rate_factor, picture_number);
+        if (q < 0)
+            return -1;
 
         assert(q>0.0);
 //printf("%f ", q);
@@ -672,7 +767,7 @@ float ff_rate_estimate_qscale(MpegEncContext *s)
 //printf("%f ", q);
         assert(q>0.0);
 
-        if(pict_type==P_TYPE || s->intra_only){ //FIXME type dependant blur like in 2-pass
+        if(pict_type==FF_P_TYPE || s->intra_only){ //FIXME type dependent blur like in 2-pass
             rcc->short_term_qsum*=a->qblur;
             rcc->short_term_qcount*=a->qblur;
 
@@ -683,7 +778,7 @@ float ff_rate_estimate_qscale(MpegEncContext *s)
 //printf("%f ", q);
         }
         assert(q>0.0);
-        
+
         q= modify_qscale(s, rce, q, picture_number);
 
         rcc->pass1_wanted_bits+= s->bit_rate/fps;
@@ -698,17 +793,19 @@ float ff_rate_estimate_qscale(MpegEncContext *s)
         );
     }
 
-    if     (q<qmin) q=qmin; 
+    if     (q<qmin) q=qmin;
     else if(q>qmax) q=qmax;
 
     if(s->adaptive_quant)
         adaptive_quantization(s, q);
     else
         q= (int)(q + 0.5);
-    
-    rcc->last_qscale= q;
-    rcc->last_mc_mb_var_sum= pic->mc_mb_var_sum;
-    rcc->last_mb_var_sum= pic->mb_var_sum;
+
+    if(!dry_run){
+        rcc->last_qscale= q;
+        rcc->last_mc_mb_var_sum= pic->mc_mb_var_sum;
+        rcc->last_mb_var_sum= pic->mb_var_sum;
+    }
 #if 0
 {
     static int mvsum=0, texsum=0;
@@ -727,25 +824,23 @@ static int init_pass2(MpegEncContext *s)
 {
     RateControlContext *rcc= &s->rc_context;
     AVCodecContext *a= s->avctx;
-    int i;
-    double fps= (double)s->avctx->frame_rate / (double)s->avctx->frame_rate_base;
+    int i, toobig;
+    double fps= 1/av_q2d(s->avctx->time_base);
     double complexity[5]={0,0,0,0,0};   // aproximate bits at quant=1
-    double avg_quantizer[5];
-    uint64_t const_bits[5]={0,0,0,0,0}; // quantizer idependant bits
-    uint64_t available_bits[5];
+    uint64_t const_bits[5]={0,0,0,0,0}; // quantizer independent bits
     uint64_t all_const_bits;
     uint64_t all_available_bits= (uint64_t)(s->bit_rate*(double)rcc->num_entries/fps);
     double rate_factor=0;
     double step;
     //int last_i_frame=-10000000;
-    const int filter_size= (int)(a->qblur*4) | 1;  
+    const int filter_size= (int)(a->qblur*4) | 1;
     double expected_bits;
-    double *qscale, *blured_qscale;
+    double *qscale, *blurred_qscale, qscale_sum;
 
     /* find complexity & const_bits & decide the pict_types */
     for(i=0; i<rcc->num_entries; i++){
         RateControlEntry *rce= &rcc->entry[i];
-        
+
         rce->new_pict_type= rce->pict_type;
         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;
@@ -755,45 +850,21 @@ static int init_pass2(MpegEncContext *s)
         complexity[rce->new_pict_type]+= (rce->i_tex_bits+ rce->p_tex_bits)*(double)rce->qscale;
         const_bits[rce->new_pict_type]+= rce->mv_bits + rce->misc_bits;
     }
-    all_const_bits= const_bits[I_TYPE] + const_bits[P_TYPE] + const_bits[B_TYPE];
-    
+    all_const_bits= const_bits[FF_I_TYPE] + const_bits[FF_P_TYPE] + const_bits[FF_B_TYPE];
+
     if(all_available_bits < all_const_bits){
-        av_log(s->avctx, AV_LOG_ERROR, "requested bitrate is to low\n");
+        av_log(s->avctx, AV_LOG_ERROR, "requested bitrate is too low\n");
         return -1;
     }
-    
-    /* find average quantizers */
-    avg_quantizer[P_TYPE]=0;
-    for(step=256*256; step>0.0000001; step*=0.5){
-        double expected_bits=0;
-        avg_quantizer[P_TYPE]+= step;
-        
-        avg_quantizer[I_TYPE]= avg_quantizer[P_TYPE]*ABS(s->avctx->i_quant_factor) + s->avctx->i_quant_offset;
-        avg_quantizer[B_TYPE]= avg_quantizer[P_TYPE]*ABS(s->avctx->b_quant_factor) + s->avctx->b_quant_offset;
-        
-        expected_bits= 
-            + all_const_bits 
-            + complexity[I_TYPE]/avg_quantizer[I_TYPE]
-            + complexity[P_TYPE]/avg_quantizer[P_TYPE]
-            + complexity[B_TYPE]/avg_quantizer[B_TYPE];
-            
-        if(expected_bits < all_available_bits) avg_quantizer[P_TYPE]-= step;
-//printf("%f %lld %f\n", expected_bits, all_available_bits, avg_quantizer[P_TYPE]);
-    }
-//printf("qp_i:%f, qp_p:%f, qp_b:%f\n", avg_quantizer[I_TYPE],avg_quantizer[P_TYPE],avg_quantizer[B_TYPE]);
 
-    for(i=0; i<5; i++){
-        available_bits[i]= const_bits[i] + complexity[i]/avg_quantizer[i];
-    }
-//printf("%lld %lld %lld %lld\n", available_bits[I_TYPE], available_bits[P_TYPE], available_bits[B_TYPE], all_available_bits);
-        
     qscale= av_malloc(sizeof(double)*rcc->num_entries);
-    blured_qscale= av_malloc(sizeof(double)*rcc->num_entries);
+    blurred_qscale= av_malloc(sizeof(double)*rcc->num_entries);
+    toobig = 0;
 
     for(step=256*256; step>0.0000001; step*=0.5){
         expected_bits=0;
         rate_factor+= step;
-        
+
         rcc->buffer_index= s->avctx->rc_buffer_size/2;
 
         /* find qscale */
@@ -805,7 +876,7 @@ static int init_pass2(MpegEncContext *s)
         /* fixed I/B QP relative to P mode */
         for(i=rcc->num_entries-1; i>=0; i--){
             RateControlEntry *rce= &rcc->entry[i];
-            
+
             qscale[i]= get_diff_limited_q(s, rce, qscale[i]);
         }
 
@@ -815,41 +886,73 @@ static int init_pass2(MpegEncContext *s)
             const int pict_type= rce->new_pict_type;
             int j;
             double q=0.0, sum=0.0;
-        
+
             for(j=0; j<filter_size; j++){
                 int index= i+j-filter_size/2;
                 double d= index-i;
                 double coeff= a->qblur==0 ? 1.0 : exp(-d*d/(a->qblur * a->qblur));
-            
+
                 if(index < 0 || index >= rcc->num_entries) continue;
                 if(pict_type != rcc->entry[index].new_pict_type) continue;
                 q+= qscale[index] * coeff;
                 sum+= coeff;
             }
-            blured_qscale[i]= q/sum;
+            blurred_qscale[i]= q/sum;
         }
-    
+
         /* find expected bits */
         for(i=0; i<rcc->num_entries; i++){
             RateControlEntry *rce= &rcc->entry[i];
             double bits;
-            rce->new_qscale= modify_qscale(s, rce, blured_qscale[i], i);
+            rce->new_qscale= modify_qscale(s, rce, blurred_qscale[i], i);
             bits= qp2bits(rce, rce->new_qscale) + rce->mv_bits + rce->misc_bits;
-//printf("%d %f\n", rce->new_bits, blured_qscale[i]);
+//printf("%d %f\n", rce->new_bits, blurred_qscale[i]);
             bits += 8*ff_vbv_update(s, bits);
 
             rce->expected_bits= expected_bits;
             expected_bits += bits;
         }
 
-//        printf("%f %d %f\n", expected_bits, (int)all_available_bits, rate_factor);
-        if(expected_bits > all_available_bits) rate_factor-= step;
+        /*
+        av_log(s->avctx, AV_LOG_INFO,
+            "expected_bits: %f all_available_bits: %d rate_factor: %f\n",
+            expected_bits, (int)all_available_bits, rate_factor);
+        */
+        if(expected_bits > all_available_bits) {
+            rate_factor-= step;
+            ++toobig;
+        }
     }
     av_free(qscale);
-    av_free(blured_qscale);
+    av_free(blurred_qscale);
 
-    if(abs(expected_bits/all_available_bits - 1.0) > 0.01 ){
-        av_log(s->avctx, AV_LOG_ERROR, "Error: 2pass curve failed to converge\n");
+    /* check bitrate calculations and print info */
+    qscale_sum = 0.0;
+    for(i=0; i<rcc->num_entries; i++){
+        /* av_log(s->avctx, AV_LOG_DEBUG, "[lavc rc] entry[%d].new_qscale = %.3f  qp = %.3f\n",
+            i, rcc->entry[i].new_qscale, rcc->entry[i].new_qscale / FF_QP2LAMBDA); */
+        qscale_sum += av_clip(rcc->entry[i].new_qscale / FF_QP2LAMBDA, s->avctx->qmin, s->avctx->qmax);
+    }
+    assert(toobig <= 40);
+    av_log(s->avctx, AV_LOG_DEBUG,
+        "[lavc rc] requested bitrate: %d bps  expected bitrate: %d bps\n",
+        s->bit_rate,
+        (int)(expected_bits / ((double)all_available_bits/s->bit_rate)));
+    av_log(s->avctx, AV_LOG_DEBUG,
+        "[lavc rc] estimated target average qp: %.3f\n",
+        (float)qscale_sum / rcc->num_entries);
+    if (toobig == 0) {
+        av_log(s->avctx, AV_LOG_INFO,
+            "[lavc rc] Using all of requested bitrate is not "
+            "necessary for this video with these parameters.\n");
+    } else if (toobig == 40) {
+        av_log(s->avctx, AV_LOG_ERROR,
+            "[lavc rc] Error: bitrate too low for this video "
+            "with these parameters.\n");
+        return -1;
+    } else if (fabs(expected_bits/all_available_bits - 1.0) > 0.01) {
+        av_log(s->avctx, AV_LOG_ERROR,
+            "[lavc rc] Error: 2pass curve failed to converge\n");
         return -1;
     }