libavcodec/snow.c

   1 /*
   2  * Copyright (C) 2004 Michael Niedermayer <michaelni@gmx.at>
   3  *
   4  * This library is free software; you can redistribute it and/or
   5  * modify it under the terms of the GNU Lesser General Public
   6  * License as published by the Free Software Foundation; either
   7  * version 2 of the License, or (at your option) any later version.
   8  *
   9  * This library is distributed in the hope that it will be useful,
  10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  12  * Lesser General Public License for more details.
  13  *
  14  * You should have received a copy of the GNU Lesser General Public
  15  * License along with this library; if not, write to the Free Software
  16  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  17  */
  18
  19 #include "avcodec.h"
  20 #include "common.h"
  21 #include "dsputil.h"
  22
  23 #include "rangecoder.h"
  24 #define MID_STATE 128
  25
  26 #include "mpegvideo.h"
  27
  28 #undef NDEBUG
  29 #include <assert.h>
  30
  31 #define MAX_DECOMPOSITIONS 8
  32 #define MAX_PLANES 4
  33 #define DWTELEM int
  34 #define QSHIFT 5
  35 #define QROOT (1<<QSHIFT)
  36 #define LOSSLESS_QLOG -128
  37 #define FRAC_BITS 8
  38
  39 static const int8_t quant3[256]={
  40  0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  41  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  42  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  43  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  44  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  45  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  46  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  47  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  48 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  49 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  50 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  51 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  52 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  53 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  54 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  55 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1, 0,
  56 };
  57 static const int8_t quant3b[256]={
  58  0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  59  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  60  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  61  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  62  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  63  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  64  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  65  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  66 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  67 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  68 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  69 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  70 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  71 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  72 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  73 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  74 };
  75 static const int8_t quant3bA[256]={
  76  0, 0, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
  77  1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
  78  1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
  79  1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
  80  1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
  81  1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
  82  1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
  83  1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
  84  1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
  85  1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
  86  1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
  87  1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
  88  1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
  89  1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
  90  1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
  91  1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1, 1,-1,
  92 };
  93 static const int8_t quant5[256]={
  94  0, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  95  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  96  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  97  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  98  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  99  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
 100  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
 101  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
 102 -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
 103 -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
 104 -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
 105 -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
 106 -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
 107 -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
 108 -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
 109 -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-1,-1,-1,
 110 };
 111 static const int8_t quant7[256]={
 112  0, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
 113  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
 114  2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3,
 115  3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
 116  3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
 117  3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
 118  3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
 119  3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
 120 -3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,
 121 -3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,
 122 -3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,
 123 -3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,
 124 -3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,
 125 -3,-3,-3,-3,-3,-3,-3,-3,-3,-2,-2,-2,-2,-2,-2,-2,
 126 -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
 127 -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-1,-1,
 128 };
 129 static const int8_t quant9[256]={
 130  0, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3,
 131  3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
 132  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
 133  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
 134  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
 135  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
 136  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
 137  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
 138 -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
 139 -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
 140 -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
 141 -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
 142 -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
 143 -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
 144 -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-3,-3,-3,-3,
 145 -3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-2,-2,-2,-2,-1,-1,
 146 };
 147 static const int8_t quant11[256]={
 148  0, 1, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4,
 149  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
 150  4, 4, 4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
 151  5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
 152  5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
 153  5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
 154  5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
 155  5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
 156 -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
 157 -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
 158 -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
 159 -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
 160 -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
 161 -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-4,-4,
 162 -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
 163 -4,-4,-4,-4,-4,-3,-3,-3,-3,-3,-3,-3,-2,-2,-2,-1,
 164 };
 165 static const int8_t quant13[256]={
 166  0, 1, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4,
 167  4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
 168  5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
 169  5, 5, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
 170  6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
 171  6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
 172  6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
 173  6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
 174 -6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,
 175 -6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,
 176 -6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,
 177 -6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,
 178 -6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-5,
 179 -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
 180 -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
 181 -4,-4,-4,-4,-4,-4,-4,-4,-4,-3,-3,-3,-3,-2,-2,-1,
 182 };
 183
 184 #define LOG2_OBMC_MAX 6
 185 #define OBMC_MAX (1<<(LOG2_OBMC_MAX))
 186 #if 0 //64*cubic
 187 static const uint8_t obmc32[1024]={
 188  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
 189  0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0,
 190  0, 0, 0, 1, 1, 1, 1, 2, 2, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 3, 3, 3, 2, 2, 1, 1, 1, 1, 0, 0, 0,
 191  0, 0, 1, 1, 2, 2, 3, 4, 4, 5, 6, 6, 7, 7, 8, 8, 8, 8, 7, 7, 6, 6, 5, 4, 4, 3, 2, 2, 1, 1, 0, 0,
 192  0, 0, 1, 2, 2, 3, 4, 6, 7, 8, 9,10,11,12,12,12,12,12,12,11,10, 9, 8, 7, 6, 4, 3, 2, 2, 1, 0, 0,
 193  0, 1, 1, 2, 3, 5, 6, 8,10,11,13,14,15,16,17,18,18,17,16,15,14,13,11,10, 8, 6, 5, 3, 2, 1, 1, 0,
 194  0, 1, 1, 3, 4, 6, 8,10,13,15,17,19,20,22,22,23,23,22,22,20,19,17,15,13,10, 8, 6, 4, 3, 1, 1, 0,
 195  0, 1, 2, 4, 6, 8,10,13,16,19,21,23,25,27,28,29,29,28,27,25,23,21,19,16,13,10, 8, 6, 4, 2, 1, 0,
 196  0, 1, 2, 4, 7,10,13,16,19,22,25,28,31,33,34,35,35,34,33,31,28,25,22,19,16,13,10, 7, 4, 2, 1, 0,
 197  0, 1, 3, 5, 8,11,15,19,22,26,30,33,36,38,40,41,41,40,38,36,33,30,26,22,19,15,11, 8, 5, 3, 1, 0,
 198  0, 1, 3, 6, 9,12,17,21,25,30,34,38,41,44,45,46,46,45,44,41,38,34,30,25,21,17,12, 9, 6, 3, 1, 0,
 199  0, 1, 3, 6,10,14,19,23,28,33,38,42,45,48,51,52,52,51,48,45,42,38,33,28,23,19,14,10, 6, 3, 1, 0,
 200  0, 1, 4, 7,11,15,20,25,31,36,41,45,49,52,55,56,56,55,52,49,45,41,36,31,25,20,15,11, 7, 4, 1, 0,
 201  0, 2, 4, 7,12,16,22,27,33,38,44,48,52,56,58,60,60,58,56,52,48,44,38,33,27,22,16,12, 7, 4, 2, 0,
 202  0, 1, 4, 8,12,17,22,28,34,40,45,51,55,58,61,62,62,61,58,55,51,45,40,34,28,22,17,12, 8, 4, 1, 0,
 203  0, 2, 4, 8,12,18,23,29,35,41,46,52,56,60,62,64,64,62,60,56,52,46,41,35,29,23,18,12, 8, 4, 2, 0,
 204  0, 2, 4, 8,12,18,23,29,35,41,46,52,56,60,62,64,64,62,60,56,52,46,41,35,29,23,18,12, 8, 4, 2, 0,
 205  0, 1, 4, 8,12,17,22,28,34,40,45,51,55,58,61,62,62,61,58,55,51,45,40,34,28,22,17,12, 8, 4, 1, 0,
 206  0, 2, 4, 7,12,16,22,27,33,38,44,48,52,56,58,60,60,58,56,52,48,44,38,33,27,22,16,12, 7, 4, 2, 0,
 207  0, 1, 4, 7,11,15,20,25,31,36,41,45,49,52,55,56,56,55,52,49,45,41,36,31,25,20,15,11, 7, 4, 1, 0,
 208  0, 1, 3, 6,10,14,19,23,28,33,38,42,45,48,51,52,52,51,48,45,42,38,33,28,23,19,14,10, 6, 3, 1, 0,
 209  0, 1, 3, 6, 9,12,17,21,25,30,34,38,41,44,45,46,46,45,44,41,38,34,30,25,21,17,12, 9, 6, 3, 1, 0,
 210  0, 1, 3, 5, 8,11,15,19,22,26,30,33,36,38,40,41,41,40,38,36,33,30,26,22,19,15,11, 8, 5, 3, 1, 0,
 211  0, 1, 2, 4, 7,10,13,16,19,22,25,28,31,33,34,35,35,34,33,31,28,25,22,19,16,13,10, 7, 4, 2, 1, 0,
 212  0, 1, 2, 4, 6, 8,10,13,16,19,21,23,25,27,28,29,29,28,27,25,23,21,19,16,13,10, 8, 6, 4, 2, 1, 0,
 213  0, 1, 1, 3, 4, 6, 8,10,13,15,17,19,20,22,22,23,23,22,22,20,19,17,15,13,10, 8, 6, 4, 3, 1, 1, 0,
 214  0, 1, 1, 2, 3, 5, 6, 8,10,11,13,14,15,16,17,18,18,17,16,15,14,13,11,10, 8, 6, 5, 3, 2, 1, 1, 0,
 215  0, 0, 1, 2, 2, 3, 4, 6, 7, 8, 9,10,11,12,12,12,12,12,12,11,10, 9, 8, 7, 6, 4, 3, 2, 2, 1, 0, 0,
 216  0, 0, 1, 1, 2, 2, 3, 4, 4, 5, 6, 6, 7, 7, 8, 8, 8, 8, 7, 7, 6, 6, 5, 4, 4, 3, 2, 2, 1, 1, 0, 0,
 217  0, 0, 0, 1, 1, 1, 1, 2, 2, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 3, 3, 3, 2, 2, 1, 1, 1, 1, 0, 0, 0,
 218  0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0,
 219  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
 220 //error:0.000022
 221 };
 222 static const uint8_t obmc16[256]={
 223  0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0,
 224  0, 1, 1, 2, 4, 5, 5, 6, 6, 5, 5, 4, 2, 1, 1, 0,
 225  0, 1, 4, 6, 9,11,13,15,15,13,11, 9, 6, 4, 1, 0,
 226  0, 2, 6,11,15,20,24,26,26,24,20,15,11, 6, 2, 0,
 227  0, 4, 9,15,23,29,34,38,38,34,29,23,15, 9, 4, 0,
 228  0, 5,11,20,29,38,45,49,49,45,38,29,20,11, 5, 0,
 229  1, 5,13,24,34,45,53,57,57,53,45,34,24,13, 5, 1,
 230  1, 6,15,26,38,49,57,62,62,57,49,38,26,15, 6, 1,
 231  1, 6,15,26,38,49,57,62,62,57,49,38,26,15, 6, 1,
 232  1, 5,13,24,34,45,53,57,57,53,45,34,24,13, 5, 1,
 233  0, 5,11,20,29,38,45,49,49,45,38,29,20,11, 5, 0,
 234  0, 4, 9,15,23,29,34,38,38,34,29,23,15, 9, 4, 0,
 235  0, 2, 6,11,15,20,24,26,26,24,20,15,11, 6, 2, 0,
 236  0, 1, 4, 6, 9,11,13,15,15,13,11, 9, 6, 4, 1, 0,
 237  0, 1, 1, 2, 4, 5, 5, 6, 6, 5, 5, 4, 2, 1, 1, 0,
 238  0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0,
 239 //error:0.000033
 240 };
 241 #elif 1 // 64*linear
 242 static const uint8_t obmc32[1024]={
 243  0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0,
 244  0, 1, 1, 1, 2, 2, 2, 3, 3, 4, 4, 4, 5, 5, 5, 6, 6, 5, 5, 5, 4, 4, 4, 3, 3, 2, 2, 2, 1, 1, 1, 0,
 245  0, 1, 2, 2, 3, 3, 4, 5, 5, 6, 7, 7, 8, 8, 9,10,10, 9, 8, 8, 7, 7, 6, 5, 5, 4, 3, 3, 2, 2, 1, 0,
 246  0, 1, 2, 3, 4, 5, 6, 7, 7, 8, 9,10,11,12,13,14,14,13,12,11,10, 9, 8, 7, 7, 6, 5, 4, 3, 2, 1, 0,
 247  1, 2, 3, 4, 5, 6, 7, 8,10,11,12,13,14,15,16,17,17,16,15,14,13,12,11,10, 8, 7, 6, 5, 4, 3, 2, 1,
 248  1, 2, 3, 5, 6, 8, 9,10,12,13,14,16,17,19,20,21,21,20,19,17,16,14,13,12,10, 9, 8, 6, 5, 3, 2, 1,
 249  1, 2, 4, 6, 7, 9,11,12,14,15,17,19,20,22,24,25,25,24,22,20,19,17,15,14,12,11, 9, 7, 6, 4, 2, 1,
 250  1, 3, 5, 7, 8,10,12,14,16,18,20,22,23,25,27,29,29,27,25,23,22,20,18,16,14,12,10, 8, 7, 5, 3, 1,
 251  1, 3, 5, 7,10,12,14,16,18,20,22,24,27,29,31,33,33,31,29,27,24,22,20,18,16,14,12,10, 7, 5, 3, 1,
 252  1, 4, 6, 8,11,13,15,18,20,23,25,27,30,32,34,37,37,34,32,30,27,25,23,20,18,15,13,11, 8, 6, 4, 1,
 253  1, 4, 7, 9,12,14,17,20,22,25,28,30,33,35,38,41,41,38,35,33,30,28,25,22,20,17,14,12, 9, 7, 4, 1,
 254  1, 4, 7,10,13,16,19,22,24,27,30,33,36,39,42,45,45,42,39,36,33,30,27,24,22,19,16,13,10, 7, 4, 1,
 255  2, 5, 8,11,14,17,20,23,27,30,33,36,39,42,45,48,48,45,42,39,36,33,30,27,23,20,17,14,11, 8, 5, 2,
 256  2, 5, 8,12,15,19,22,25,29,32,35,39,42,46,49,52,52,49,46,42,39,35,32,29,25,22,19,15,12, 8, 5, 2,
 257  2, 5, 9,13,16,20,24,27,31,34,38,42,45,49,53,56,56,53,49,45,42,38,34,31,27,24,20,16,13, 9, 5, 2,
 258  2, 6,10,14,17,21,25,29,33,37,41,45,48,52,56,60,60,56,52,48,45,41,37,33,29,25,21,17,14,10, 6, 2,
 259  2, 6,10,14,17,21,25,29,33,37,41,45,48,52,56,60,60,56,52,48,45,41,37,33,29,25,21,17,14,10, 6, 2,
 260  2, 5, 9,13,16,20,24,27,31,34,38,42,45,49,53,56,56,53,49,45,42,38,34,31,27,24,20,16,13, 9, 5, 2,
 261  2, 5, 8,12,15,19,22,25,29,32,35,39,42,46,49,52,52,49,46,42,39,35,32,29,25,22,19,15,12, 8, 5, 2,
 262  2, 5, 8,11,14,17,20,23,27,30,33,36,39,42,45,48,48,45,42,39,36,33,30,27,23,20,17,14,11, 8, 5, 2,
 263  1, 4, 7,10,13,16,19,22,24,27,30,33,36,39,42,45,45,42,39,36,33,30,27,24,22,19,16,13,10, 7, 4, 1,
 264  1, 4, 7, 9,12,14,17,20,22,25,28,30,33,35,38,41,41,38,35,33,30,28,25,22,20,17,14,12, 9, 7, 4, 1,
 265  1, 4, 6, 8,11,13,15,18,20,23,25,27,30,32,34,37,37,34,32,30,27,25,23,20,18,15,13,11, 8, 6, 4, 1,
 266  1, 3, 5, 7,10,12,14,16,18,20,22,24,27,29,31,33,33,31,29,27,24,22,20,18,16,14,12,10, 7, 5, 3, 1,
 267  1, 3, 5, 7, 8,10,12,14,16,18,20,22,23,25,27,29,29,27,25,23,22,20,18,16,14,12,10, 8, 7, 5, 3, 1,
 268  1, 2, 4, 6, 7, 9,11,12,14,15,17,19,20,22,24,25,25,24,22,20,19,17,15,14,12,11, 9, 7, 6, 4, 2, 1,
 269  1, 2, 3, 5, 6, 8, 9,10,12,13,14,16,17,19,20,21,21,20,19,17,16,14,13,12,10, 9, 8, 6, 5, 3, 2, 1,
 270  1, 2, 3, 4, 5, 6, 7, 8,10,11,12,13,14,15,16,17,17,16,15,14,13,12,11,10, 8, 7, 6, 5, 4, 3, 2, 1,
 271  0, 1, 2, 3, 4, 5, 6, 7, 7, 8, 9,10,11,12,13,14,14,13,12,11,10, 9, 8, 7, 7, 6, 5, 4, 3, 2, 1, 0,
 272  0, 1, 2, 2, 3, 3, 4, 5, 5, 6, 7, 7, 8, 8, 9,10,10, 9, 8, 8, 7, 7, 6, 5, 5, 4, 3, 3, 2, 2, 1, 0,
 273  0, 1, 1, 1, 2, 2, 2, 3, 3, 4, 4, 4, 5, 5, 5, 6, 6, 5, 5, 5, 4, 4, 4, 3, 3, 2, 2, 2, 1, 1, 1, 0,
 274  0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0,
 275  //error:0.000020
 276 };
 277 static const uint8_t obmc16[256]={
 278  0, 1, 1, 2, 2, 3, 3, 4, 4, 3, 3, 2, 2, 1, 1, 0,
 279  1, 2, 4, 5, 7, 8,10,11,11,10, 8, 7, 5, 4, 2, 1,
 280  1, 4, 6, 9,11,14,16,19,19,16,14,11, 9, 6, 4, 1,
 281  2, 5, 9,12,16,19,23,26,26,23,19,16,12, 9, 5, 2,
 282  2, 7,11,16,20,25,29,34,34,29,25,20,16,11, 7, 2,
 283  3, 8,14,19,25,30,36,41,41,36,30,25,19,14, 8, 3,
 284  3,10,16,23,29,36,42,49,49,42,36,29,23,16,10, 3,
 285  4,11,19,26,34,41,49,56,56,49,41,34,26,19,11, 4,
 286  4,11,19,26,34,41,49,56,56,49,41,34,26,19,11, 4,
 287  3,10,16,23,29,36,42,49,49,42,36,29,23,16,10, 3,
 288  3, 8,14,19,25,30,36,41,41,36,30,25,19,14, 8, 3,
 289  2, 7,11,16,20,25,29,34,34,29,25,20,16,11, 7, 2,
 290  2, 5, 9,12,16,19,23,26,26,23,19,16,12, 9, 5, 2,
 291  1, 4, 6, 9,11,14,16,19,19,16,14,11, 9, 6, 4, 1,
 292  1, 2, 4, 5, 7, 8,10,11,11,10, 8, 7, 5, 4, 2, 1,
 293  0, 1, 1, 2, 2, 3, 3, 4, 4, 3, 3, 2, 2, 1, 1, 0,
 294 //error:0.000015
 295 };
 296 #else //64*cos
 297 static const uint8_t obmc32[1024]={
 298  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
 299  0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0,
 300  0, 0, 0, 1, 1, 1, 1, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 3, 3, 3, 3, 2, 2, 1, 1, 1, 1, 0, 0, 0,
 301  0, 0, 1, 1, 1, 2, 2, 3, 4, 5, 5, 6, 7, 7, 7, 7, 7, 7, 7, 7, 6, 5, 5, 4, 3, 2, 2, 1, 1, 1, 0, 0,
 302  0, 0, 1, 1, 2, 3, 4, 5, 6, 7, 9,10,11,11,12,12,12,12,11,11,10, 9, 7, 6, 5, 4, 3, 2, 1, 1, 0, 0,
 303  0, 0, 1, 2, 3, 5, 6, 8, 9,11,12,14,15,16,17,17,17,17,16,15,14,12,11, 9, 8, 6, 5, 3, 2, 1, 0, 0,
 304  0, 1, 1, 2, 4, 6, 8,10,12,15,17,19,20,21,22,23,23,22,21,20,19,17,15,12,10, 8, 6, 4, 2, 1, 1, 0,
 305  0, 1, 2, 3, 5, 8,10,13,16,19,21,24,26,27,28,29,29,28,27,26,24,21,19,16,13,10, 8, 5, 3, 2, 1, 0,
 306  0, 1, 2, 4, 6, 9,12,16,19,23,26,29,31,33,34,35,35,34,33,31,29,26,23,19,16,12, 9, 6, 4, 2, 1, 0,
 307  0, 1, 3, 5, 7,11,15,19,23,26,30,34,37,39,40,41,41,40,39,37,34,30,26,23,19,15,11, 7, 5, 3, 1, 0,
 308  0, 1, 3, 5, 9,12,17,21,26,30,35,38,42,44,46,47,47,46,44,42,38,35,30,26,21,17,12, 9, 5, 3, 1, 0,
 309  0, 1, 3, 6, 9,14,19,24,29,34,38,43,46,49,51,52,52,51,49,46,43,38,34,29,24,19,14, 9, 6, 3, 1, 0,
 310  0, 1, 3, 6,11,15,20,26,31,37,42,46,50,53,56,57,57,56,53,50,46,42,37,31,26,20,15,11, 6, 3, 1, 0,
 311  0, 1, 3, 7,11,16,21,27,33,39,44,49,53,57,59,60,60,59,57,53,49,44,39,33,27,21,16,11, 7, 3, 1, 0,
 312  0, 1, 4, 7,12,17,22,28,34,40,46,51,56,59,61,63,63,61,59,56,51,46,40,34,28,22,17,12, 7, 4, 1, 0,
 313  0, 1, 4, 7,12,17,23,29,35,41,47,52,57,60,63,64,64,63,60,57,52,47,41,35,29,23,17,12, 7, 4, 1, 0,
 314  0, 1, 4, 7,12,17,23,29,35,41,47,52,57,60,63,64,64,63,60,57,52,47,41,35,29,23,17,12, 7, 4, 1, 0,
 315  0, 1, 4, 7,12,17,22,28,34,40,46,51,56,59,61,63,63,61,59,56,51,46,40,34,28,22,17,12, 7, 4, 1, 0,
 316  0, 1, 3, 7,11,16,21,27,33,39,44,49,53,57,59,60,60,59,57,53,49,44,39,33,27,21,16,11, 7, 3, 1, 0,
 317  0, 1, 3, 6,11,15,20,26,31,37,42,46,50,53,56,57,57,56,53,50,46,42,37,31,26,20,15,11, 6, 3, 1, 0,
 318  0, 1, 3, 6, 9,14,19,24,29,34,38,43,46,49,51,52,52,51,49,46,43,38,34,29,24,19,14, 9, 6, 3, 1, 0,
 319  0, 1, 3, 5, 9,12,17,21,26,30,35,38,42,44,46,47,47,46,44,42,38,35,30,26,21,17,12, 9, 5, 3, 1, 0,
 320  0, 1, 3, 5, 7,11,15,19,23,26,30,34,37,39,40,41,41,40,39,37,34,30,26,23,19,15,11, 7, 5, 3, 1, 0,
 321  0, 1, 2, 4, 6, 9,12,16,19,23,26,29,31,33,34,35,35,34,33,31,29,26,23,19,16,12, 9, 6, 4, 2, 1, 0,
 322  0, 1, 2, 3, 5, 8,10,13,16,19,21,24,26,27,28,29,29,28,27,26,24,21,19,16,13,10, 8, 5, 3, 2, 1, 0,
 323  0, 1, 1, 2, 4, 6, 8,10,12,15,17,19,20,21,22,23,23,22,21,20,19,17,15,12,10, 8, 6, 4, 2, 1, 1, 0,
 324  0, 0, 1, 2, 3, 5, 6, 8, 9,11,12,14,15,16,17,17,17,17,16,15,14,12,11, 9, 8, 6, 5, 3, 2, 1, 0, 0,
 325  0, 0, 1, 1, 2, 3, 4, 5, 6, 7, 9,10,11,11,12,12,12,12,11,11,10, 9, 7, 6, 5, 4, 3, 2, 1, 1, 0, 0,
 326  0, 0, 1, 1, 1, 2, 2, 3, 4, 5, 5, 6, 7, 7, 7, 7, 7, 7, 7, 7, 6, 5, 5, 4, 3, 2, 2, 1, 1, 1, 0, 0,
 327  0, 0, 0, 1, 1, 1, 1, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 3, 3, 3, 3, 2, 2, 1, 1, 1, 1, 0, 0, 0,
 328  0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0,
 329  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
 330 //error:0.000022
 331 };
 332 static const uint8_t obmc16[256]={
 333  0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0,
 334  0, 0, 1, 2, 3, 4, 5, 5, 5, 5, 4, 3, 2, 1, 0, 0,
 335  0, 1, 3, 6, 8,11,13,14,14,13,11, 8, 6, 3, 1, 0,
 336  0, 2, 6,10,15,20,24,26,26,24,20,15,10, 6, 2, 0,
 337  0, 3, 8,16,23,30,35,38,38,35,30,23,16, 8, 3, 0,
 338  1, 4,11,20,30,39,46,49,49,46,39,30,20,11, 4, 1,
 339  1, 5,13,24,35,46,54,58,58,54,46,35,24,13, 5, 1,
 340  0, 5,14,26,38,49,58,63,63,58,49,38,26,14, 5, 0,
 341  0, 5,14,26,38,49,58,63,63,58,49,38,26,14, 5, 0,
 342  1, 5,13,24,35,46,54,58,58,54,46,35,24,13, 5, 1,
 343  1, 4,11,20,30,39,46,49,49,46,39,30,20,11, 4, 1,
 344  0, 3, 8,16,23,30,35,38,38,35,30,23,16, 8, 3, 0,
 345  0, 2, 6,10,15,20,24,26,26,24,20,15,10, 6, 2, 0,
 346  0, 1, 3, 6, 8,11,13,14,14,13,11, 8, 6, 3, 1, 0,
 347  0, 0, 1, 2, 3, 4, 5, 5, 5, 5, 4, 3, 2, 1, 0, 0,
 348  0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0,
 349 //error:0.000022
 350 };
 351 #endif
 352
 353 //linear *64
 354 static const uint8_t obmc8[64]={
 355  1, 3, 5, 7, 7, 5, 3, 1,
 356  3, 9,15,21,21,15, 9, 3,
 357  5,15,25,35,35,25,15, 5,
 358  7,21,35,49,49,35,21, 7,
 359  7,21,35,49,49,35,21, 7,
 360  5,15,25,35,35,25,15, 5,
 361  3, 9,15,21,21,15, 9, 3,
 362  1, 3, 5, 7, 7, 5, 3, 1,
 363 //error:0.000000
 364 };
 365
 366 //linear *64
 367 static const uint8_t obmc4[16]={
 368  4,12,12, 4,
 369 12,36,36,12,
 370 12,36,36,12,
 371  4,12,12, 4,
 372 //error:0.000000
 373 };
 374
 375 static const uint8_t *obmc_tab[4]={
 376     obmc32, obmc16, obmc8, obmc4
 377 };
 378
 379 typedef struct BlockNode{
 380     int16_t mx;
 381     int16_t my;
 382     uint8_t color[3];
 383     uint8_t type;
 384 //#define TYPE_SPLIT    1
 385 #define BLOCK_INTRA   1
 386 //#define TYPE_NOCOLOR  4
 387     uint8_t level; //FIXME merge into type?
 388 }BlockNode;
 389
 390 #define LOG2_MB_SIZE 4
 391 #define MB_SIZE (1<<LOG2_MB_SIZE)
 392
 393 typedef struct x_and_coeff{
 394     int16_t x;
 395     uint16_t coeff;
 396 } x_and_coeff;
 397
 398 typedef struct SubBand{
 399     int level;
 400     int stride;
 401     int width;
 402     int height;
 403     int qlog;                                   ///< log(qscale)/log[2^(1/6)]
 404     DWTELEM *buf;
 405     int buf_x_offset;
 406     int buf_y_offset;
 407     int stride_line; ///< Stride measured in lines, not pixels.
 408     x_and_coeff * x_coeff;
 409     struct SubBand *parent;
 410     uint8_t state[/*7*2*/ 7 + 512][32];
 411 }SubBand;
 412
 413 typedef struct Plane{
 414     int width;
 415     int height;
 416     SubBand band[MAX_DECOMPOSITIONS][4];
 417 }Plane;
 418
 419 /** Used to minimize the amount of memory used in order to optimize cache performance. **/
 420 typedef struct {
 421     DWTELEM * * line; ///< For use by idwt and predict_slices.
 422     DWTELEM * * data_stack; ///< Used for internal purposes.
 423     int data_stack_top;
 424     int line_count;
 425     int line_width;
 426     int data_count;
 427     DWTELEM * base_buffer; ///< Buffer that this structure is caching.
 428 } slice_buffer;
 429
 430 typedef struct SnowContext{
 431 //    MpegEncContext m; // needed for motion estimation, should not be used for anything else, the idea is to make the motion estimation eventually independant of MpegEncContext, so this will be removed then (FIXME/XXX)
 432
 433     AVCodecContext *avctx;
 434     RangeCoder c;
 435     DSPContext dsp;
 436     AVFrame input_picture;
 437     AVFrame current_picture;
 438     AVFrame last_picture;
 439     AVFrame mconly_picture;
 440 //     uint8_t q_context[16];
 441     uint8_t header_state[32];
 442     uint8_t block_state[128 + 32*128];
 443     int keyframe;
 444     int always_reset;
 445     int version;
 446     int spatial_decomposition_type;
 447     int temporal_decomposition_type;
 448     int spatial_decomposition_count;
 449     int temporal_decomposition_count;
 450     DWTELEM *spatial_dwt_buffer;
 451     int colorspace_type;
 452     int chroma_h_shift;
 453     int chroma_v_shift;
 454     int spatial_scalability;
 455     int qlog;
 456     int lambda;
 457     int lambda2;
 458     int mv_scale;
 459     int qbias;
 460 #define QBIAS_SHIFT 3
 461     int b_width;
 462     int b_height;
 463     int block_max_depth;
 464     Plane plane[MAX_PLANES];
 465     BlockNode *block;
 466     slice_buffer sb;
 467
 468     MpegEncContext m; // needed for motion estimation, should not be used for anything else, the idea is to make the motion estimation eventually independant of MpegEncContext, so this will be removed then (FIXME/XXX)
 469 }SnowContext;
 470
 471 typedef struct {
 472     DWTELEM *b0;
 473     DWTELEM *b1;
 474     DWTELEM *b2;
 475     DWTELEM *b3;
 476     int y;
 477 } dwt_compose_t;
 478
 479 #define slice_buffer_get_line(slice_buf, line_num) ((slice_buf)->line[line_num] ? (slice_buf)->line[line_num] : slice_buffer_load_line((slice_buf), (line_num)))
 480 //#define slice_buffer_get_line(slice_buf, line_num) (slice_buffer_load_line((slice_buf), (line_num)))
 481
 482 static void slice_buffer_init(slice_buffer * buf, int line_count, int max_allocated_lines, int line_width, DWTELEM * base_buffer)
 483 {
 484     int i;
 485
 486     buf->base_buffer = base_buffer;
 487     buf->line_count = line_count;
 488     buf->line_width = line_width;
 489     buf->data_count = max_allocated_lines;
 490     buf->line = (DWTELEM * *) av_mallocz (sizeof(DWTELEM *) * line_count);
 491     buf->data_stack = (DWTELEM * *) av_malloc (sizeof(DWTELEM *) * max_allocated_lines);
 492
 493     for (i = 0; i < max_allocated_lines; i++)
 494     {
 495       buf->data_stack[i] = (DWTELEM *) av_malloc (sizeof(DWTELEM) * line_width);
 496     }
 497
 498     buf->data_stack_top = max_allocated_lines - 1;
 499 }
 500
 501 static DWTELEM * slice_buffer_load_line(slice_buffer * buf, int line)
 502 {
 503     int i;
 504     int offset;
 505     DWTELEM * buffer;
 506
 507 //  av_log(NULL, AV_LOG_DEBUG, "Cache hit: %d\n", line);
 508
 509     assert(buf->data_stack_top >= 0);
 510 //  assert(!buf->line[line]);
 511     if (buf->line[line])
 512         return buf->line[line];
 513
 514     offset = buf->line_width * line;
 515     buffer = buf->data_stack[buf->data_stack_top];
 516     buf->data_stack_top--;
 517     buf->line[line] = buffer;
 518
 519 //  av_log(NULL, AV_LOG_DEBUG, "slice_buffer_load_line: line: %d remaining: %d\n", line, buf->data_stack_top + 1);
 520
 521     return buffer;
 522 }
 523
 524 static void slice_buffer_release(slice_buffer * buf, int line)
 525 {
 526     int i;
 527     int offset;
 528     DWTELEM * buffer;
 529
 530     assert(line >= 0 && line < buf->line_count);
 531     assert(buf->line[line]);
 532
 533     offset = buf->line_width * line;
 534     buffer = buf->line[line];
 535     buf->data_stack_top++;
 536     buf->data_stack[buf->data_stack_top] = buffer;
 537     buf->line[line] = NULL;
 538
 539 //  av_log(NULL, AV_LOG_DEBUG, "slice_buffer_release: line: %d remaining: %d\n", line, buf->data_stack_top + 1);
 540 }
 541
 542 static void slice_buffer_flush(slice_buffer * buf)
 543 {
 544     int i;
 545     for (i = 0; i < buf->line_count; i++)
 546     {
 547         if (buf->line[i])
 548         {
 549 //      av_log(NULL, AV_LOG_DEBUG, "slice_buffer_flush: line: %d \n", i);
 550             slice_buffer_release(buf, i);
 551         }
 552     }
 553 }
 554
 555 static void slice_buffer_destroy(slice_buffer * buf)
 556 {
 557     int i;
 558     slice_buffer_flush(buf);
 559
 560     for (i = buf->data_count - 1; i >= 0; i--)
 561     {
 562         assert(buf->data_stack[i]);
 563         av_free(buf->data_stack[i]);
 564     }
 565     assert(buf->data_stack);
 566     av_free(buf->data_stack);
 567     assert(buf->line);
 568     av_free(buf->line);
 569 }
 570
 571 #ifdef  __sgi
 572 // Avoid a name clash on SGI IRIX
 573 #undef  qexp
 574 #endif
 575 #define QEXPSHIFT (7-FRAC_BITS+8) //FIXME try to change this to 0
 576 static uint8_t qexp[QROOT];
 577
 578 static inline int mirror(int v, int m){
 579     if     (v<0) return -v;
 580     else if(v>m) return 2*m-v;
 581     else         return v;
 582 }
 583
 584 static inline void put_symbol(RangeCoder *c, uint8_t *state, int v, int is_signed){
 585     int i;
 586
 587     if(v){
 588         const int a= ABS(v);
 589         const int e= av_log2(a);
 590 #if 1
 591         const int el= FFMIN(e, 10);
 592         put_rac(c, state+0, 0);
 593
 594         for(i=0; i<el; i++){
 595             put_rac(c, state+1+i, 1);  //1..10
 596         }
 597         for(; i<e; i++){
 598             put_rac(c, state+1+9, 1);  //1..10
 599         }
 600         put_rac(c, state+1+FFMIN(i,9), 0);
 601
 602         for(i=e-1; i>=el; i--){
 603             put_rac(c, state+22+9, (a>>i)&1); //22..31
 604         }
 605         for(; i>=0; i--){
 606             put_rac(c, state+22+i, (a>>i)&1); //22..31
 607         }
 608
 609         if(is_signed)
 610             put_rac(c, state+11 + el, v < 0); //11..21
 611 #else
 612
 613         put_rac(c, state+0, 0);
 614         if(e<=9){
 615             for(i=0; i<e; i++){
 616                 put_rac(c, state+1+i, 1);  //1..10
 617             }
 618             put_rac(c, state+1+i, 0);
 619
 620             for(i=e-1; i>=0; i--){
 621                 put_rac(c, state+22+i, (a>>i)&1); //22..31
 622             }
 623
 624             if(is_signed)
 625                 put_rac(c, state+11 + e, v < 0); //11..21
 626         }else{
 627             for(i=0; i<e; i++){
 628                 put_rac(c, state+1+FFMIN(i,9), 1);  //1..10
 629             }
 630             put_rac(c, state+1+FFMIN(i,9), 0);
 631
 632             for(i=e-1; i>=0; i--){
 633                 put_rac(c, state+22+FFMIN(i,9), (a>>i)&1); //22..31
 634             }
 635
 636             if(is_signed)
 637                 put_rac(c, state+11 + FFMIN(e,10), v < 0); //11..21
 638         }
 639 #endif
 640     }else{
 641         put_rac(c, state+0, 1);
 642     }
 643 }
 644
 645 static inline int get_symbol(RangeCoder *c, uint8_t *state, int is_signed){
 646     if(get_rac(c, state+0))
 647         return 0;
 648     else{
 649         int i, e, a;
 650         e= 0;
 651         while(get_rac(c, state+1 + FFMIN(e,9))){ //1..10
 652             e++;
 653         }
 654
 655         a= 1;
 656         for(i=e-1; i>=0; i--){
 657             a += a + get_rac(c, state+22 + FFMIN(i,9)); //22..31
 658         }
 659
 660         if(is_signed && get_rac(c, state+11 + FFMIN(e,10))) //11..21
 661             return -a;
 662         else
 663             return a;
 664     }
 665 }
 666
 667 static inline void put_symbol2(RangeCoder *c, uint8_t *state, int v, int log2){
 668     int i;
 669     int r= log2>=0 ? 1<<log2 : 1;
 670
 671     assert(v>=0);
 672     assert(log2>=-4);
 673
 674     while(v >= r){
 675         put_rac(c, state+4+log2, 1);
 676         v -= r;
 677         log2++;
 678         if(log2>0) r+=r;
 679     }
 680     put_rac(c, state+4+log2, 0);
 681
 682     for(i=log2-1; i>=0; i--){
 683         put_rac(c, state+31-i, (v>>i)&1);
 684     }
 685 }
 686
 687 static inline int get_symbol2(RangeCoder *c, uint8_t *state, int log2){
 688     int i;
 689     int r= log2>=0 ? 1<<log2 : 1;
 690     int v=0;
 691
 692     assert(log2>=-4);
 693
 694     while(get_rac(c, state+4+log2)){
 695         v+= r;
 696         log2++;
 697         if(log2>0) r+=r;
 698     }
 699
 700     for(i=log2-1; i>=0; i--){
 701         v+= get_rac(c, state+31-i)<<i;
 702     }
 703
 704     return v;
 705 }
 706
 707 static always_inline void lift(DWTELEM *dst, DWTELEM *src, DWTELEM *ref, int dst_step, int src_step, int ref_step, int width, int mul, int add, int shift, int highpass, int inverse){
 708     const int mirror_left= !highpass;
 709     const int mirror_right= (width&1) ^ highpass;
 710     const int w= (width>>1) - 1 + (highpass & width);
 711     int i;
 712
 713 #define LIFT(src, ref, inv) ((src) + ((inv) ? - (ref) : + (ref)))
 714     if(mirror_left){
 715         dst[0] = LIFT(src[0], ((mul*2*ref[0]+add)>>shift), inverse);
 716         dst += dst_step;
 717         src += src_step;
 718     }
 719
 720     for(i=0; i<w; i++){
 721         dst[i*dst_step] = LIFT(src[i*src_step], ((mul*(ref[i*ref_step] + ref[(i+1)*ref_step])+add)>>shift), inverse);
 722     }
 723
 724     if(mirror_right){
 725         dst[w*dst_step] = LIFT(src[w*src_step], ((mul*2*ref[w*ref_step]+add)>>shift), inverse);
 726     }
 727 }
 728
 729 static always_inline void lift5(DWTELEM *dst, DWTELEM *src, DWTELEM *ref, int dst_step, int src_step, int ref_step, int width, int mul, int add, int shift, int highpass, int inverse){
 730     const int mirror_left= !highpass;
 731     const int mirror_right= (width&1) ^ highpass;
 732     const int w= (width>>1) - 1 + (highpass & width);
 733     int i;
 734
 735     if(mirror_left){
 736         int r= 3*2*ref[0];
 737         r += r>>4;
 738         r += r>>8;
 739         dst[0] = LIFT(src[0], ((r+add)>>shift), inverse);
 740         dst += dst_step;
 741         src += src_step;
 742     }
 743
 744     for(i=0; i<w; i++){
 745         int r= 3*(ref[i*ref_step] + ref[(i+1)*ref_step]);
 746         r += r>>4;
 747         r += r>>8;
 748         dst[i*dst_step] = LIFT(src[i*src_step], ((r+add)>>shift), inverse);
 749     }
 750
 751     if(mirror_right){
 752         int r= 3*2*ref[w*ref_step];
 753         r += r>>4;
 754         r += r>>8;
 755         dst[w*dst_step] = LIFT(src[w*src_step], ((r+add)>>shift), inverse);
 756     }
 757 }
 758
 759
 760 static void inplace_lift(DWTELEM *dst, int width, int *coeffs, int n, int shift, int start, int inverse){
 761     int x, i;
 762
 763     for(x=start; x<width; x+=2){
 764         int64_t sum=0;
 765
 766         for(i=0; i<n; i++){
 767             int x2= x + 2*i - n + 1;
 768             if     (x2<     0) x2= -x2;
 769             else if(x2>=width) x2= 2*width-x2-2;
 770             sum += coeffs[i]*(int64_t)dst[x2];
 771         }
 772         if(inverse) dst[x] -= (sum + (1<<shift)/2)>>shift;
 773         else        dst[x] += (sum + (1<<shift)/2)>>shift;
 774     }
 775 }
 776
 777 static void inplace_liftV(DWTELEM *dst, int width, int height, int stride, int *coeffs, int n, int shift, int start, int inverse){
 778     int x, y, i;
 779     for(y=start; y<height; y+=2){
 780         for(x=0; x<width; x++){
 781             int64_t sum=0;
 782
 783             for(i=0; i<n; i++){
 784                 int y2= y + 2*i - n + 1;
 785                 if     (y2<      0) y2= -y2;
 786                 else if(y2>=height) y2= 2*height-y2-2;
 787                 sum += coeffs[i]*(int64_t)dst[x + y2*stride];
 788             }
 789             if(inverse) dst[x + y*stride] -= (sum + (1<<shift)/2)>>shift;
 790             else        dst[x + y*stride] += (sum + (1<<shift)/2)>>shift;
 791         }
 792     }
 793 }
 794
 795 #define SCALEX 1
 796 #define LX0 0
 797 #define LX1 1
 798
 799 #if 0 // more accurate 9/7
 800 #define N1 2
 801 #define SHIFT1 14
 802 #define COEFFS1 (int[]){-25987,-25987}
 803 #define N2 2
 804 #define SHIFT2 19
 805 #define COEFFS2 (int[]){-27777,-27777}
 806 #define N3 2
 807 #define SHIFT3 15
 808 #define COEFFS3 (int[]){28931,28931}
 809 #define N4 2
 810 #define SHIFT4 15
 811 #define COEFFS4 (int[]){14533,14533}
 812 #elif 1 // 13/7 CRF
 813 #define N1 4
 814 #define SHIFT1 4
 815 #define COEFFS1 (int[]){1,-9,-9,1}
 816 #define N2 4
 817 #define SHIFT2 4
 818 #define COEFFS2 (int[]){-1,5,5,-1}
 819 #define N3 0
 820 #define SHIFT3 1
 821 #define COEFFS3 NULL
 822 #define N4 0
 823 #define SHIFT4 1
 824 #define COEFFS4 NULL
 825 #elif 1 // 3/5
 826 #define LX0 1
 827 #define LX1 0
 828 #define SCALEX 0.5
 829 #define N1 2
 830 #define SHIFT1 1
 831 #define COEFFS1 (int[]){1,1}
 832 #define N2 2
 833 #define SHIFT2 2
 834 #define COEFFS2 (int[]){-1,-1}
 835 #define N3 0
 836 #define SHIFT3 0
 837 #define COEFFS3 NULL
 838 #define N4 0
 839 #define SHIFT4 0
 840 #define COEFFS4 NULL
 841 #elif 1 // 11/5
 842 #define N1 0
 843 #define SHIFT1 1
 844 #define COEFFS1 NULL
 845 #define N2 2
 846 #define SHIFT2 2
 847 #define COEFFS2 (int[]){-1,-1}
 848 #define N3 2
 849 #define SHIFT3 0
 850 #define COEFFS3 (int[]){-1,-1}
 851 #define N4 4
 852 #define SHIFT4 7
 853 #define COEFFS4 (int[]){-5,29,29,-5}
 854 #define SCALEX 4
 855 #elif 1 // 9/7 CDF
 856 #define N1 2
 857 #define SHIFT1 7
 858 #define COEFFS1 (int[]){-203,-203}
 859 #define N2 2
 860 #define SHIFT2 12
 861 #define COEFFS2 (int[]){-217,-217}
 862 #define N3 2
 863 #define SHIFT3 7
 864 #define COEFFS3 (int[]){113,113}
 865 #define N4 2
 866 #define SHIFT4 9
 867 #define COEFFS4 (int[]){227,227}
 868 #define SCALEX 1
 869 #elif 1 // 7/5 CDF
 870 #define N1 0
 871 #define SHIFT1 1
 872 #define COEFFS1 NULL
 873 #define N2 2
 874 #define SHIFT2 2
 875 #define COEFFS2 (int[]){-1,-1}
 876 #define N3 2
 877 #define SHIFT3 0
 878 #define COEFFS3 (int[]){-1,-1}
 879 #define N4 2
 880 #define SHIFT4 4
 881 #define COEFFS4 (int[]){3,3}
 882 #elif 1 // 9/7 MN
 883 #define N1 4
 884 #define SHIFT1 4
 885 #define COEFFS1 (int[]){1,-9,-9,1}
 886 #define N2 2
 887 #define SHIFT2 2
 888 #define COEFFS2 (int[]){1,1}
 889 #define N3 0
 890 #define SHIFT3 1
 891 #define COEFFS3 NULL
 892 #define N4 0
 893 #define SHIFT4 1
 894 #define COEFFS4 NULL
 895 #else // 13/7 CRF
 896 #define N1 4
 897 #define SHIFT1 4
 898 #define COEFFS1 (int[]){1,-9,-9,1}
 899 #define N2 4
 900 #define SHIFT2 4
 901 #define COEFFS2 (int[]){-1,5,5,-1}
 902 #define N3 0
 903 #define SHIFT3 1
 904 #define COEFFS3 NULL
 905 #define N4 0
 906 #define SHIFT4 1
 907 #define COEFFS4 NULL
 908 #endif
 909 static void horizontal_decomposeX(DWTELEM *b, int width){
 910     DWTELEM temp[width];
 911     const int width2= width>>1;
 912     const int w2= (width+1)>>1;
 913     int A1,A2,A3,A4, x;
 914
 915     inplace_lift(b, width, COEFFS1, N1, SHIFT1, LX1, 0);
 916     inplace_lift(b, width, COEFFS2, N2, SHIFT2, LX0, 0);
 917     inplace_lift(b, width, COEFFS3, N3, SHIFT3, LX1, 0);
 918     inplace_lift(b, width, COEFFS4, N4, SHIFT4, LX0, 0);
 919
 920     for(x=0; x<width2; x++){
 921         temp[x   ]= b[2*x    ];
 922         temp[x+w2]= b[2*x + 1];
 923     }
 924     if(width&1)
 925         temp[x   ]= b[2*x    ];
 926     memcpy(b, temp, width*sizeof(int));
 927 }
 928
 929 static void horizontal_composeX(DWTELEM *b, int width){
 930     DWTELEM temp[width];
 931     const int width2= width>>1;
 932     int A1,A2,A3,A4, x;
 933     const int w2= (width+1)>>1;
 934
 935     memcpy(temp, b, width*sizeof(int));
 936     for(x=0; x<width2; x++){
 937         b[2*x    ]= temp[x   ];
 938         b[2*x + 1]= temp[x+w2];
 939     }
 940     if(width&1)
 941         b[2*x    ]= temp[x   ];
 942
 943     inplace_lift(b, width, COEFFS4, N4, SHIFT4, LX0, 1);
 944     inplace_lift(b, width, COEFFS3, N3, SHIFT3, LX1, 1);
 945     inplace_lift(b, width, COEFFS2, N2, SHIFT2, LX0, 1);
 946     inplace_lift(b, width, COEFFS1, N1, SHIFT1, LX1, 1);
 947 }
 948
 949 static void spatial_decomposeX(DWTELEM *buffer, int width, int height, int stride){
 950     int x, y;
 951
 952     for(y=0; y<height; y++){
 953         for(x=0; x<width; x++){
 954             buffer[y*stride + x] *= SCALEX;
 955         }
 956     }
 957
 958     for(y=0; y<height; y++){
 959         horizontal_decomposeX(buffer + y*stride, width);
 960     }
 961
 962     inplace_liftV(buffer, width, height, stride, COEFFS1, N1, SHIFT1, LX1, 0);
 963     inplace_liftV(buffer, width, height, stride, COEFFS2, N2, SHIFT2, LX0, 0);
 964     inplace_liftV(buffer, width, height, stride, COEFFS3, N3, SHIFT3, LX1, 0);
 965     inplace_liftV(buffer, width, height, stride, COEFFS4, N4, SHIFT4, LX0, 0);
 966 }
 967
 968 static void spatial_composeX(DWTELEM *buffer, int width, int height, int stride){
 969     int x, y;
 970
 971     inplace_liftV(buffer, width, height, stride, COEFFS4, N4, SHIFT4, LX0, 1);
 972     inplace_liftV(buffer, width, height, stride, COEFFS3, N3, SHIFT3, LX1, 1);
 973     inplace_liftV(buffer, width, height, stride, COEFFS2, N2, SHIFT2, LX0, 1);
 974     inplace_liftV(buffer, width, height, stride, COEFFS1, N1, SHIFT1, LX1, 1);
 975
 976     for(y=0; y<height; y++){
 977         horizontal_composeX(buffer + y*stride, width);
 978     }
 979
 980     for(y=0; y<height; y++){
 981         for(x=0; x<width; x++){
 982             buffer[y*stride + x] /= SCALEX;
 983         }
 984     }
 985 }
 986
 987 static void horizontal_decompose53i(DWTELEM *b, int width){
 988     DWTELEM temp[width];
 989     const int width2= width>>1;
 990     int A1,A2,A3,A4, x;
 991     const int w2= (width+1)>>1;
 992
 993     for(x=0; x<width2; x++){
 994         temp[x   ]= b[2*x    ];
 995         temp[x+w2]= b[2*x + 1];
 996     }
 997     if(width&1)
 998         temp[x   ]= b[2*x    ];
 999 #if 0
1000     A2= temp[1       ];
1001     A4= temp[0       ];
1002     A1= temp[0+width2];
1003     A1 -= (A2 + A4)>>1;
1004     A4 += (A1 + 1)>>1;
1005     b[0+width2] = A1;
1006     b[0       ] = A4;
1007     for(x=1; x+1<width2; x+=2){
1008         A3= temp[x+width2];
1009         A4= temp[x+1     ];
1010         A3 -= (A2 + A4)>>1;
1011         A2 += (A1 + A3 + 2)>>2;
1012         b[x+width2] = A3;
1013         b[x       ] = A2;
1014
1015         A1= temp[x+1+width2];
1016         A2= temp[x+2       ];
1017         A1 -= (A2 + A4)>>1;
1018         A4 += (A1 + A3 + 2)>>2;
1019         b[x+1+width2] = A1;
1020         b[x+1       ] = A4;
1021     }
1022     A3= temp[width-1];
1023     A3 -= A2;
1024     A2 += (A1 + A3 + 2)>>2;
1025     b[width -1] = A3;
1026     b[width2-1] = A2;
1027 #else
1028     lift(b+w2, temp+w2, temp, 1, 1, 1, width, -1, 0, 1, 1, 0);
1029     lift(b   , temp   , b+w2, 1, 1, 1, width,  1, 2, 2, 0, 0);
1030 #endif
1031 }
1032
1033 static void vertical_decompose53iH0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
1034     int i;
1035
1036     for(i=0; i<width; i++){
1037         b1[i] -= (b0[i] + b2[i])>>1;
1038     }
1039 }
1040
1041 static void vertical_decompose53iL0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
1042     int i;
1043
1044     for(i=0; i<width; i++){
1045         b1[i] += (b0[i] + b2[i] + 2)>>2;
1046     }
1047 }
1048
1049 static void spatial_decompose53i(DWTELEM *buffer, int width, int height, int stride){
1050     int y;
1051     DWTELEM *b0= buffer + mirror(-2-1, height-1)*stride;
1052     DWTELEM *b1= buffer + mirror(-2  , height-1)*stride;
1053
1054     for(y=-2; y<height; y+=2){
1055         DWTELEM *b2= buffer + mirror(y+1, height-1)*stride;
1056         DWTELEM *b3= buffer + mirror(y+2, height-1)*stride;
1057
1058 {START_TIMER
1059         if(b1 <= b3)     horizontal_decompose53i(b2, width);
1060         if(y+2 < height) horizontal_decompose53i(b3, width);
1061 STOP_TIMER("horizontal_decompose53i")}
1062
1063 {START_TIMER
1064         if(b1 <= b3) vertical_decompose53iH0(b1, b2, b3, width);
1065         if(b0 <= b2) vertical_decompose53iL0(b0, b1, b2, width);
1066 STOP_TIMER("vertical_decompose53i*")}
1067
1068         b0=b2;
1069         b1=b3;
1070     }
1071 }
1072
1073 #define lift5 lift
1074 #if 1
1075 #define W_AM 3
1076 #define W_AO 0
1077 #define W_AS 1
1078
1079 #define W_BM 1
1080 #define W_BO 8
1081 #define W_BS 4
1082
1083 #undef lift5
1084 #define W_CM 9999
1085 #define W_CO 2
1086 #define W_CS 2
1087
1088 #define W_DM 15
1089 #define W_DO 16
1090 #define W_DS 5
1091 #elif 0
1092 #define W_AM 55
1093 #define W_AO 16
1094 #define W_AS 5
1095
1096 #define W_BM 3
1097 #define W_BO 32
1098 #define W_BS 6
1099
1100 #define W_CM 127
1101 #define W_CO 64
1102 #define W_CS 7
1103
1104 #define W_DM 7
1105 #define W_DO 8
1106 #define W_DS 4
1107 #elif 0
1108 #define W_AM 97
1109 #define W_AO 32
1110 #define W_AS 6
1111
1112 #define W_BM 63
1113 #define W_BO 512
1114 #define W_BS 10
1115
1116 #define W_CM 13
1117 #define W_CO 8
1118 #define W_CS 4
1119
1120 #define W_DM 15
1121 #define W_DO 16
1122 #define W_DS 5
1123
1124 #else
1125
1126 #define W_AM 203
1127 #define W_AO 64
1128 #define W_AS 7
1129
1130 #define W_BM 217
1131 #define W_BO 2048
1132 #define W_BS 12
1133
1134 #define W_CM 113
1135 #define W_CO 64
1136 #define W_CS 7
1137
1138 #define W_DM 227
1139 #define W_DO 128
1140 #define W_DS 9
1141 #endif
1142 static void horizontal_decompose97i(DWTELEM *b, int width){
1143     DWTELEM temp[width];
1144     const int w2= (width+1)>>1;
1145
1146     lift (temp+w2, b    +1, b      , 1, 2, 2, width, -W_AM, W_AO, W_AS, 1, 0);
1147     lift (temp   , b      , temp+w2, 1, 2, 1, width, -W_BM, W_BO, W_BS, 0, 0);
1148     lift5(b   +w2, temp+w2, temp   , 1, 1, 1, width,  W_CM, W_CO, W_CS, 1, 0);
1149     lift (b      , temp   , b   +w2, 1, 1, 1, width,  W_DM, W_DO, W_DS, 0, 0);
1150 }
1151
1152
1153 static void vertical_decompose97iH0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
1154     int i;
1155
1156     for(i=0; i<width; i++){
1157         b1[i] -= (W_AM*(b0[i] + b2[i])+W_AO)>>W_AS;
1158     }
1159 }
1160
1161 static void vertical_decompose97iH1(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
1162     int i;
1163
1164     for(i=0; i<width; i++){
1165 #ifdef lift5
1166         b1[i] += (W_CM*(b0[i] + b2[i])+W_CO)>>W_CS;
1167 #else
1168         int r= 3*(b0[i] + b2[i]);
1169         r+= r>>4;
1170         r+= r>>8;
1171         b1[i] += (r+W_CO)>>W_CS;
1172 #endif
1173     }
1174 }
1175
1176 static void vertical_decompose97iL0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
1177     int i;
1178
1179     for(i=0; i<width; i++){
1180         b1[i] -= (W_BM*(b0[i] + b2[i])+W_BO)>>W_BS;
1181     }
1182 }
1183
1184 static void vertical_decompose97iL1(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
1185     int i;
1186
1187     for(i=0; i<width; i++){
1188         b1[i] += (W_DM*(b0[i] + b2[i])+W_DO)>>W_DS;
1189     }
1190 }
1191
1192 static void spatial_decompose97i(DWTELEM *buffer, int width, int height, int stride){
1193     int y;
1194     DWTELEM *b0= buffer + mirror(-4-1, height-1)*stride;
1195     DWTELEM *b1= buffer + mirror(-4  , height-1)*stride;
1196     DWTELEM *b2= buffer + mirror(-4+1, height-1)*stride;
1197     DWTELEM *b3= buffer + mirror(-4+2, height-1)*stride;
1198
1199     for(y=-4; y<height; y+=2){
1200         DWTELEM *b4= buffer + mirror(y+3, height-1)*stride;
1201         DWTELEM *b5= buffer + mirror(y+4, height-1)*stride;
1202
1203 {START_TIMER
1204         if(b3 <= b5)     horizontal_decompose97i(b4, width);
1205         if(y+4 < height) horizontal_decompose97i(b5, width);
1206 if(width>400){
1207 STOP_TIMER("horizontal_decompose97i")
1208 }}
1209
1210 {START_TIMER
1211         if(b3 <= b5) vertical_decompose97iH0(b3, b4, b5, width);
1212         if(b2 <= b4) vertical_decompose97iL0(b2, b3, b4, width);
1213         if(b1 <= b3) vertical_decompose97iH1(b1, b2, b3, width);
1214         if(b0 <= b2) vertical_decompose97iL1(b0, b1, b2, width);
1215
1216 if(width>400){
1217 STOP_TIMER("vertical_decompose97i")
1218 }}
1219
1220         b0=b2;
1221         b1=b3;
1222         b2=b4;
1223         b3=b5;
1224     }
1225 }
1226
1227 void ff_spatial_dwt(DWTELEM *buffer, int width, int height, int stride, int type, int decomposition_count){
1228     int level;
1229
1230     for(level=0; level<decomposition_count; level++){
1231         switch(type){
1232         case 0: spatial_decompose97i(buffer, width>>level, height>>level, stride<<level); break;
1233         case 1: spatial_decompose53i(buffer, width>>level, height>>level, stride<<level); break;
1234         case 2: spatial_decomposeX  (buffer, width>>level, height>>level, stride<<level); break;
1235         }
1236     }
1237 }
1238
1239 static void horizontal_compose53i(DWTELEM *b, int width){
1240     DWTELEM temp[width];
1241     const int width2= width>>1;
1242     const int w2= (width+1)>>1;
1243     int A1,A2,A3,A4, x;
1244
1245 #if 0
1246     A2= temp[1       ];
1247     A4= temp[0       ];
1248     A1= temp[0+width2];
1249     A1 -= (A2 + A4)>>1;
1250     A4 += (A1 + 1)>>1;
1251     b[0+width2] = A1;
1252     b[0       ] = A4;
1253     for(x=1; x+1<width2; x+=2){
1254         A3= temp[x+width2];
1255         A4= temp[x+1     ];
1256         A3 -= (A2 + A4)>>1;
1257         A2 += (A1 + A3 + 2)>>2;
1258         b[x+width2] = A3;
1259         b[x       ] = A2;
1260
1261         A1= temp[x+1+width2];
1262         A2= temp[x+2       ];
1263         A1 -= (A2 + A4)>>1;
1264         A4 += (A1 + A3 + 2)>>2;
1265         b[x+1+width2] = A1;
1266         b[x+1       ] = A4;
1267     }
1268     A3= temp[width-1];
1269     A3 -= A2;
1270     A2 += (A1 + A3 + 2)>>2;
1271     b[width -1] = A3;
1272     b[width2-1] = A2;
1273 #else
1274     lift(temp   , b   , b+w2, 1, 1, 1, width,  1, 2, 2, 0, 1);
1275     lift(temp+w2, b+w2, temp, 1, 1, 1, width, -1, 0, 1, 1, 1);
1276 #endif
1277     for(x=0; x<width2; x++){
1278         b[2*x    ]= temp[x   ];
1279         b[2*x + 1]= temp[x+w2];
1280     }
1281     if(width&1)
1282         b[2*x    ]= temp[x   ];
1283 }
1284
1285 static void vertical_compose53iH0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
1286     int i;
1287
1288     for(i=0; i<width; i++){
1289         b1[i] += (b0[i] + b2[i])>>1;
1290     }
1291 }
1292
1293 static void vertical_compose53iL0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
1294     int i;
1295
1296     for(i=0; i<width; i++){
1297         b1[i] -= (b0[i] + b2[i] + 2)>>2;
1298     }
1299 }
1300
1301 static void spatial_compose53i_buffered_init(dwt_compose_t *cs, slice_buffer * sb, int height, int stride_line){
1302     cs->b0 = slice_buffer_get_line(sb, mirror(-1-1, height-1) * stride_line);
1303     cs->b1 = slice_buffer_get_line(sb, mirror(-1  , height-1) * stride_line);
1304     cs->y = -1;
1305 }
1306
1307 static void spatial_compose53i_init(dwt_compose_t *cs, DWTELEM *buffer, int height, int stride){
1308     cs->b0 = buffer + mirror(-1-1, height-1)*stride;
1309     cs->b1 = buffer + mirror(-1  , height-1)*stride;
1310     cs->y = -1;
1311 }
1312
1313 static void spatial_compose53i_dy_buffered(dwt_compose_t *cs, slice_buffer * sb, int width, int height, int stride_line){
1314     int y= cs->y;
1315     int mirror0 = mirror(y-1, height-1);
1316     int mirror1 = mirror(y  , height-1);
1317     int mirror2 = mirror(y+1, height-1);
1318     int mirror3 = mirror(y+2, height-1);
1319
1320     DWTELEM *b0= cs->b0;
1321     DWTELEM *b1= cs->b1;
1322     DWTELEM *b2= slice_buffer_get_line(sb, mirror2 * stride_line);
1323     DWTELEM *b3= slice_buffer_get_line(sb, mirror3 * stride_line);
1324
1325 {START_TIMER
1326         if(mirror1 <= mirror3) vertical_compose53iL0(b1, b2, b3, width);
1327         if(mirror0 <= mirror2) vertical_compose53iH0(b0, b1, b2, width);
1328 STOP_TIMER("vertical_compose53i*")}
1329
1330 {START_TIMER
1331         if(y-1 >= 0) horizontal_compose53i(b0, width);
1332         if(mirror0 <= mirror2) horizontal_compose53i(b1, width);
1333 STOP_TIMER("horizontal_compose53i")}
1334
1335     cs->b0 = b2;
1336     cs->b1 = b3;
1337     cs->y += 2;
1338 }
1339
1340 static void spatial_compose53i_dy(dwt_compose_t *cs, DWTELEM *buffer, int width, int height, int stride){
1341     int y= cs->y;
1342     DWTELEM *b0= cs->b0;
1343     DWTELEM *b1= cs->b1;
1344     DWTELEM *b2= buffer + mirror(y+1, height-1)*stride;
1345     DWTELEM *b3= buffer + mirror(y+2, height-1)*stride;
1346
1347 {START_TIMER
1348         if(b1 <= b3) vertical_compose53iL0(b1, b2, b3, width);
1349         if(b0 <= b2) vertical_compose53iH0(b0, b1, b2, width);
1350 STOP_TIMER("vertical_compose53i*")}
1351
1352 {START_TIMER
1353         if(y-1 >= 0) horizontal_compose53i(b0, width);
1354         if(b0 <= b2) horizontal_compose53i(b1, width);
1355 STOP_TIMER("horizontal_compose53i")}
1356
1357     cs->b0 = b2;
1358     cs->b1 = b3;
1359     cs->y += 2;
1360 }
1361
1362 static void spatial_compose53i(DWTELEM *buffer, int width, int height, int stride){
1363     dwt_compose_t cs;
1364     spatial_compose53i_init(&cs, buffer, height, stride);
1365     while(cs.y <= height)
1366         spatial_compose53i_dy(&cs, buffer, width, height, stride);
1367 }
1368
1369
1370 static void horizontal_compose97i(DWTELEM *b, int width){
1371     DWTELEM temp[width];
1372     const int w2= (width+1)>>1;
1373
1374     lift (temp   , b      , b   +w2, 1, 1, 1, width,  W_DM, W_DO, W_DS, 0, 1);
1375     lift5(temp+w2, b   +w2, temp   , 1, 1, 1, width,  W_CM, W_CO, W_CS, 1, 1);
1376     lift (b      , temp   , temp+w2, 2, 1, 1, width, -W_BM, W_BO, W_BS, 0, 1);
1377     lift (b+1    , temp+w2, b      , 2, 1, 2, width, -W_AM, W_AO, W_AS, 1, 1);
1378 }
1379
1380 static void vertical_compose97iH0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
1381     int i;
1382
1383     for(i=0; i<width; i++){
1384         b1[i] += (W_AM*(b0[i] + b2[i])+W_AO)>>W_AS;
1385     }
1386 }
1387
1388 static void vertical_compose97iH1(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
1389     int i;
1390
1391     for(i=0; i<width; i++){
1392 #ifdef lift5
1393         b1[i] -= (W_CM*(b0[i] + b2[i])+W_CO)>>W_CS;
1394 #else
1395         int r= 3*(b0[i] + b2[i]);
1396         r+= r>>4;
1397         r+= r>>8;
1398         b1[i] -= (r+W_CO)>>W_CS;
1399 #endif
1400     }
1401 }
1402
1403 static void vertical_compose97iL0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
1404     int i;
1405
1406     for(i=0; i<width; i++){
1407         b1[i] += (W_BM*(b0[i] + b2[i])+W_BO)>>W_BS;
1408     }
1409 }
1410
1411 static void vertical_compose97iL1(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
1412     int i;
1413
1414     for(i=0; i<width; i++){
1415         b1[i] -= (W_DM*(b0[i] + b2[i])+W_DO)>>W_DS;
1416     }
1417 }
1418
1419 static void vertical_compose97i(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, DWTELEM *b3, DWTELEM *b4, DWTELEM *b5, int width){
1420     int i;
1421
1422     for(i=0; i<width; i++){
1423         int r;
1424         b4[i] -= (W_DM*(b3[i] + b5[i])+W_DO)>>W_DS;
1425 #ifdef lift5
1426         b3[i] -= (W_CM*(b2[i] + b4[i])+W_CO)>>W_CS;
1427 #else
1428         r= 3*(b2[i] + b4[i]);
1429         r+= r>>4;
1430         r+= r>>8;
1431         b3[i] -= (r+W_CO)>>W_CS;
1432 #endif
1433         b2[i] += (W_BM*(b1[i] + b3[i])+W_BO)>>W_BS;
1434         b1[i] += (W_AM*(b0[i] + b2[i])+W_AO)>>W_AS;
1435     }
1436 }
1437
1438 static void spatial_compose97i_buffered_init(dwt_compose_t *cs, slice_buffer * sb, int height, int stride_line){
1439     cs->b0 = slice_buffer_get_line(sb, mirror(-3-1, height-1) * stride_line);
1440     cs->b1 = slice_buffer_get_line(sb, mirror(-3  , height-1) * stride_line);
1441     cs->b2 = slice_buffer_get_line(sb, mirror(-3+1, height-1) * stride_line);
1442     cs->b3 = slice_buffer_get_line(sb, mirror(-3+2, height-1) * stride_line);
1443     cs->y = -3;
1444 }
1445
1446 static void spatial_compose97i_init(dwt_compose_t *cs, DWTELEM *buffer, int height, int stride){
1447     cs->b0 = buffer + mirror(-3-1, height-1)*stride;
1448     cs->b1 = buffer + mirror(-3  , height-1)*stride;
1449     cs->b2 = buffer + mirror(-3+1, height-1)*stride;
1450     cs->b3 = buffer + mirror(-3+2, height-1)*stride;
1451     cs->y = -3;
1452 }
1453
1454 static void spatial_compose97i_dy_buffered(dwt_compose_t *cs, slice_buffer * sb, int width, int height, int stride_line){
1455     int y = cs->y;
1456
1457     int mirror0 = mirror(y - 1, height - 1);
1458     int mirror1 = mirror(y + 0, height - 1);
1459     int mirror2 = mirror(y + 1, height - 1);
1460     int mirror3 = mirror(y + 2, height - 1);
1461     int mirror4 = mirror(y + 3, height - 1);
1462     int mirror5 = mirror(y + 4, height - 1);
1463     DWTELEM *b0= cs->b0;
1464     DWTELEM *b1= cs->b1;
1465     DWTELEM *b2= cs->b2;
1466     DWTELEM *b3= cs->b3;
1467     DWTELEM *b4= slice_buffer_get_line(sb, mirror4 * stride_line);
1468     DWTELEM *b5= slice_buffer_get_line(sb, mirror5 * stride_line);
1469
1470 {START_TIMER
1471     if(y>0 && y+4<height){
1472         vertical_compose97i(b0, b1, b2, b3, b4, b5, width);
1473     }else{
1474         if(mirror3 <= mirror5) vertical_compose97iL1(b3, b4, b5, width);
1475         if(mirror2 <= mirror4) vertical_compose97iH1(b2, b3, b4, width);
1476         if(mirror1 <= mirror3) vertical_compose97iL0(b1, b2, b3, width);
1477         if(mirror0 <= mirror2) vertical_compose97iH0(b0, b1, b2, width);
1478     }
1479 if(width>400){
1480 STOP_TIMER("vertical_compose97i")}}
1481
1482 {START_TIMER
1483         if(y-1>=  0) horizontal_compose97i(b0, width);
1484         if(mirror0 <= mirror2) horizontal_compose97i(b1, width);
1485 if(width>400 && mirror0 <= mirror2){
1486 STOP_TIMER("horizontal_compose97i")}}
1487
1488     cs->b0=b2;
1489     cs->b1=b3;
1490     cs->b2=b4;
1491     cs->b3=b5;
1492     cs->y += 2;
1493 }
1494
1495 static void spatial_compose97i_dy(dwt_compose_t *cs, DWTELEM *buffer, int width, int height, int stride){
1496     int y = cs->y;
1497     DWTELEM *b0= cs->b0;
1498     DWTELEM *b1= cs->b1;
1499     DWTELEM *b2= cs->b2;
1500     DWTELEM *b3= cs->b3;
1501     DWTELEM *b4= buffer + mirror(y+3, height-1)*stride;
1502     DWTELEM *b5= buffer + mirror(y+4, height-1)*stride;
1503
1504         if(stride == width && y+4 < height && 0){
1505             int x;
1506             for(x=0; x<width/2; x++)
1507                 b5[x] += 64*2;
1508             for(; x<width; x++)
1509                 b5[x] += 169*2;
1510         }
1511
1512 {START_TIMER
1513         if(b3 <= b5) vertical_compose97iL1(b3, b4, b5, width);
1514         if(b2 <= b4) vertical_compose97iH1(b2, b3, b4, width);
1515         if(b1 <= b3) vertical_compose97iL0(b1, b2, b3, width);
1516         if(b0 <= b2) vertical_compose97iH0(b0, b1, b2, width);
1517 if(width>400){
1518 STOP_TIMER("vertical_compose97i")}}
1519
1520 {START_TIMER
1521         if(y-1>=  0) horizontal_compose97i(b0, width);
1522         if(b0 <= b2) horizontal_compose97i(b1, width);
1523 if(width>400 && b0 <= b2){
1524 STOP_TIMER("horizontal_compose97i")}}
1525
1526     cs->b0=b2;
1527     cs->b1=b3;
1528     cs->b2=b4;
1529     cs->b3=b5;
1530     cs->y += 2;
1531 }
1532
1533 static void spatial_compose97i(DWTELEM *buffer, int width, int height, int stride){
1534     dwt_compose_t cs;
1535     spatial_compose97i_init(&cs, buffer, height, stride);
1536     while(cs.y <= height)
1537         spatial_compose97i_dy(&cs, buffer, width, height, stride);
1538 }
1539
1540 void ff_spatial_idwt_buffered_init(dwt_compose_t *cs, slice_buffer * sb, int width, int height, int stride_line, int type, int decomposition_count){
1541     int level;
1542     for(level=decomposition_count-1; level>=0; level--){
1543         switch(type){
1544         case 0: spatial_compose97i_buffered_init(cs+level, sb, height>>level, stride_line<<level); break;
1545         case 1: spatial_compose53i_buffered_init(cs+level, sb, height>>level, stride_line<<level); break;
1546         /* not slicified yet */
1547         case 2: /*spatial_composeX(buffer, width>>level, height>>level, stride<<level); break;*/
1548           av_log(NULL, AV_LOG_ERROR, "spatial_composeX neither buffered nor slicified yet.\n"); break;
1549         }
1550     }
1551 }
1552
1553 void ff_spatial_idwt_init(dwt_compose_t *cs, DWTELEM *buffer, int width, int height, int stride, int type, int decomposition_count){
1554     int level;
1555     for(level=decomposition_count-1; level>=0; level--){
1556         switch(type){
1557         case 0: spatial_compose97i_init(cs+level, buffer, height>>level, stride<<level); break;
1558         case 1: spatial_compose53i_init(cs+level, buffer, height>>level, stride<<level); break;
1559         /* not slicified yet */
1560         case 2: spatial_composeX(buffer, width>>level, height>>level, stride<<level); break;
1561         }
1562     }
1563 }
1564
1565 void ff_spatial_idwt_slice(dwt_compose_t *cs, DWTELEM *buffer, int width, int height, int stride, int type, int decomposition_count, int y){
1566     const int support = type==1 ? 3 : 5;
1567     int level;
1568     if(type==2) return;
1569
1570     for(level=decomposition_count-1; level>=0; level--){
1571         while(cs[level].y <= FFMIN((y>>level)+support, height>>level)){
1572             switch(type){
1573             case 0: spatial_compose97i_dy(cs+level, buffer, width>>level, height>>level, stride<<level);
1574                     break;
1575             case 1: spatial_compose53i_dy(cs+level, buffer, width>>level, height>>level, stride<<level);
1576                     break;
1577             case 2: break;
1578             }
1579         }
1580     }
1581 }
1582
1583 void ff_spatial_idwt_buffered_slice(dwt_compose_t *cs, slice_buffer * slice_buf, int width, int height, int stride_line, int type, int decomposition_count, int y){
1584     const int support = type==1 ? 3 : 5;
1585     int level;
1586     if(type==2) return;
1587
1588     for(level=decomposition_count-1; level>=0; level--){
1589         while(cs[level].y <= FFMIN((y>>level)+support, height>>level)){
1590             switch(type){
1591             case 0: spatial_compose97i_dy_buffered(cs+level, slice_buf, width>>level, height>>level, stride_line<<level);
1592                     break;
1593             case 1: spatial_compose53i_dy_buffered(cs+level, slice_buf, width>>level, height>>level, stride_line<<level);
1594                     break;
1595             case 2: break;
1596             }
1597         }
1598     }
1599 }
1600
1601 void ff_spatial_idwt(DWTELEM *buffer, int width, int height, int stride, int type, int decomposition_count){
1602     if(type==2){
1603         int level;
1604         for(level=decomposition_count-1; level>=0; level--)
1605             spatial_composeX  (buffer, width>>level, height>>level, stride<<level);
1606     }else{
1607         dwt_compose_t cs[MAX_DECOMPOSITIONS];
1608         int y;
1609         ff_spatial_idwt_init(cs, buffer, width, height, stride, type, decomposition_count);
1610         for(y=0; y<height; y+=4)
1611             ff_spatial_idwt_slice(cs, buffer, width, height, stride, type, decomposition_count, y);
1612     }
1613 }
1614
1615 static int encode_subband_c0run(SnowContext *s, SubBand *b, DWTELEM *src, DWTELEM *parent, int stride, int orientation){
1616     const int w= b->width;
1617     const int h= b->height;
1618     int x, y;
1619
1620     if(1){
1621         int run=0;
1622         int runs[w*h];
1623         int run_index=0;
1624
1625         for(y=0; y<h; y++){
1626             for(x=0; x<w; x++){
1627                 int v, p=0;
1628                 int /*ll=0, */l=0, lt=0, t=0, rt=0;
1629                 v= src[x + y*stride];
1630
1631                 if(y){
1632                     t= src[x + (y-1)*stride];
1633                     if(x){
1634                         lt= src[x - 1 + (y-1)*stride];
1635                     }
1636                     if(x + 1 < w){
1637                         rt= src[x + 1 + (y-1)*stride];
1638                     }
1639                 }
1640                 if(x){
1641                     l= src[x - 1 + y*stride];
1642                     /*if(x > 1){
1643                         if(orientation==1) ll= src[y + (x-2)*stride];
1644                         else               ll= src[x - 2 + y*stride];
1645                     }*/
1646                 }
1647                 if(parent){
1648                     int px= x>>1;
1649                     int py= y>>1;
1650                     if(px<b->parent->width && py<b->parent->height)
1651                         p= parent[px + py*2*stride];
1652                 }
1653                 if(!(/*ll|*/l|lt|t|rt|p)){
1654                     if(v){
1655                         runs[run_index++]= run;
1656                         run=0;
1657                     }else{
1658                         run++;
1659                     }
1660                 }
1661             }
1662         }
1663         runs[run_index++]= run;
1664         run_index=0;
1665         run= runs[run_index++];
1666
1667         put_symbol2(&s->c, b->state[1], run, 3);
1668
1669         for(y=0; y<h; y++){
1670             if(s->c.bytestream_end - s->c.bytestream < w*40){
1671                 av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
1672                 return -1;
1673             }
1674             for(x=0; x<w; x++){
1675                 int v, p=0;
1676                 int /*ll=0, */l=0, lt=0, t=0, rt=0;
1677                 v= src[x + y*stride];
1678
1679                 if(y){
1680                     t= src[x + (y-1)*stride];
1681                     if(x){
1682                         lt= src[x - 1 + (y-1)*stride];
1683                     }
1684                     if(x + 1 < w){
1685                         rt= src[x + 1 + (y-1)*stride];
1686                     }
1687                 }
1688                 if(x){
1689                     l= src[x - 1 + y*stride];
1690                     /*if(x > 1){
1691                         if(orientation==1) ll= src[y + (x-2)*stride];
1692                         else               ll= src[x - 2 + y*stride];
1693                     }*/
1694                 }
1695                 if(parent){
1696                     int px= x>>1;
1697                     int py= y>>1;
1698                     if(px<b->parent->width && py<b->parent->height)
1699                         p= parent[px + py*2*stride];
1700                 }
1701                 if(/*ll|*/l|lt|t|rt|p){
1702                     int context= av_log2(/*ABS(ll) + */3*ABS(l) + ABS(lt) + 2*ABS(t) + ABS(rt) + ABS(p));
1703
1704                     put_rac(&s->c, &b->state[0][context], !!v);
1705                 }else{
1706                     if(!run){
1707                         run= runs[run_index++];
1708
1709                         put_symbol2(&s->c, b->state[1], run, 3);
1710                         assert(v);
1711                     }else{
1712                         run--;
1713                         assert(!v);
1714                     }
1715                 }
1716                 if(v){
1717                     int context= av_log2(/*ABS(ll) + */3*ABS(l) + ABS(lt) + 2*ABS(t) + ABS(rt) + ABS(p));
1718                     int l2= 2*ABS(l) + (l<0);
1719                     int t2= 2*ABS(t) + (t<0);
1720
1721                     put_symbol2(&s->c, b->state[context + 2], ABS(v)-1, context-4);
1722                     put_rac(&s->c, &b->state[0][16 + 1 + 3 + quant3bA[l2&0xFF] + 3*quant3bA[t2&0xFF]], v<0);
1723                 }
1724             }
1725         }
1726     }
1727     return 0;
1728 }
1729
1730 static int encode_subband(SnowContext *s, SubBand *b, DWTELEM *src, DWTELEM *parent, int stride, int orientation){
1731 //    encode_subband_qtree(s, b, src, parent, stride, orientation);
1732 //    encode_subband_z0run(s, b, src, parent, stride, orientation);
1733     return encode_subband_c0run(s, b, src, parent, stride, orientation);
1734 //    encode_subband_dzr(s, b, src, parent, stride, orientation);
1735 }
1736
1737 static inline void unpack_coeffs(SnowContext *s, SubBand *b, SubBand * parent, int orientation){
1738     const int w= b->width;
1739     const int h= b->height;
1740     int x,y;
1741
1742     if(1){
1743         int run;
1744         int index=0;
1745         int prev_index=-1;
1746         int prev2_index=0;
1747         int parent_index= 0;
1748         int prev_parent_index= 0;
1749
1750         run= get_symbol2(&s->c, b->state[1], 3);
1751         for(y=0; y<h; y++){
1752             int v=0;
1753             int lt=0, t=0, rt=0;
1754
1755             if(y && b->x_coeff[prev_index].x == 0){
1756                 rt= b->x_coeff[prev_index].coeff;
1757             }
1758             for(x=0; x<w; x++){
1759                 int p=0;
1760                 const int l= v;
1761
1762                 lt= t; t= rt;
1763
1764                 if(y){
1765                     if(b->x_coeff[prev_index].x <= x)
1766                         prev_index++;
1767                     if(b->x_coeff[prev_index].x == x + 1)
1768                         rt= b->x_coeff[prev_index].coeff;
1769                     else
1770                         rt=0;
1771                 }
1772                 if(parent){
1773                     if(x>>1 > parent->x_coeff[parent_index].x){
1774                         parent_index++;
1775                     }
1776                     if(x>>1 == parent->x_coeff[parent_index].x){
1777                         p= parent->x_coeff[parent_index].coeff;
1778                     }
1779                 }
1780                 if(/*ll|*/l|lt|t|rt|p){
1781                     int context= av_log2(/*ABS(ll) + */3*(l>>1) + (lt>>1) + (t&~1) + (rt>>1) + (p>>1));
1782
1783                     v=get_rac(&s->c, &b->state[0][context]);
1784                 }else{
1785                     if(!run){
1786                         run= get_symbol2(&s->c, b->state[1], 3);
1787                         v=1;
1788                     }else{
1789                         run--;
1790                         v=0;
1791
1792                         if(y && parent){
1793                             int max_run;
1794
1795                             max_run= FFMIN(run, b->x_coeff[prev_index].x - x - 2);
1796                             max_run= FFMIN(max_run, 2*parent->x_coeff[parent_index].x - x - 1);
1797                             x+= max_run;
1798                             run-= max_run;
1799                         }
1800                     }
1801                 }
1802                 if(v){
1803                     int context= av_log2(/*ABS(ll) + */3*(l>>1) + (lt>>1) + (t&~1) + (rt>>1) + (p>>1));
1804                     v= 2*(get_symbol2(&s->c, b->state[context + 2], context-4) + 1);
1805                     v+=get_rac(&s->c, &b->state[0][16 + 1 + 3 + quant3bA[l&0xFF] + 3*quant3bA[t&0xFF]]);
1806
1807                     b->x_coeff[index].x=x;
1808                     b->x_coeff[index++].coeff= v;
1809                 }
1810             }
1811             b->x_coeff[index++].x= w+1; //end marker
1812             prev_index= prev2_index;
1813             prev2_index= index;
1814
1815             if(parent){
1816                 if(y&1){
1817                     while(parent->x_coeff[parent_index].x != parent->width+1)
1818                         parent_index++;
1819                     parent_index++;
1820                     prev_parent_index= parent_index;
1821                 }else{
1822                     parent_index= prev_parent_index;
1823                 }
1824             }
1825         }
1826
1827         b->x_coeff[index++].x= w+1; //end marker
1828     }
1829 }
1830
1831 static inline void decode_subband_slice_buffered(SnowContext *s, SubBand *b, slice_buffer * sb, int start_y, int h, int save_state[1]){
1832     const int w= b->width;
1833     int x,y;
1834     const int qlog= clip(s->qlog + b->qlog, 0, QROOT*16);
1835     int qmul= qexp[qlog&(QROOT-1)]<<(qlog>>QSHIFT);
1836     int qadd= (s->qbias*qmul)>>QBIAS_SHIFT;
1837     int new_index = 0;
1838
1839     START_TIMER
1840
1841     if(b->buf == s->spatial_dwt_buffer || s->qlog == LOSSLESS_QLOG){
1842         qadd= 0;
1843         qmul= 1<<QEXPSHIFT;
1844     }
1845
1846     /* If we are on the second or later slice, restore our index. */
1847     if (start_y != 0)
1848         new_index = save_state[0];
1849
1850
1851     for(y=start_y; y<h; y++){
1852         int x = 0;
1853         int v;
1854         DWTELEM * line = slice_buffer_get_line(sb, y * b->stride_line + b->buf_y_offset) + b->buf_x_offset;
1855         memset(line, 0, b->width*sizeof(DWTELEM));
1856         v = b->x_coeff[new_index].coeff;
1857         x = b->x_coeff[new_index++].x;
1858         while(x < w)
1859         {
1860             register int t= ( (v>>1)*qmul + qadd)>>QEXPSHIFT;
1861             register int u= -(v&1);
1862             line[x] = (t^u) - u;
1863
1864             v = b->x_coeff[new_index].coeff;
1865             x = b->x_coeff[new_index++].x;
1866         }
1867     }
1868     if(w > 200 && start_y != 0/*level+1 == s->spatial_decomposition_count*/){
1869         STOP_TIMER("decode_subband")
1870     }
1871
1872     /* Save our variables for the next slice. */
1873     save_state[0] = new_index;
1874
1875     return;
1876 }
1877
1878 static void reset_contexts(SnowContext *s){
1879     int plane_index, level, orientation;
1880
1881     for(plane_index=0; plane_index<3; plane_index++){
1882         for(level=0; level<s->spatial_decomposition_count; level++){
1883             for(orientation=level ? 1:0; orientation<4; orientation++){
1884                 memset(s->plane[plane_index].band[level][orientation].state, MID_STATE, sizeof(s->plane[plane_index].band[level][orientation].state));
1885             }
1886         }
1887     }
1888     memset(s->header_state, MID_STATE, sizeof(s->header_state));
1889     memset(s->block_state, MID_STATE, sizeof(s->block_state));
1890 }
1891
1892 static int alloc_blocks(SnowContext *s){
1893     int w= -((-s->avctx->width )>>LOG2_MB_SIZE);
1894     int h= -((-s->avctx->height)>>LOG2_MB_SIZE);
1895
1896     s->b_width = w;
1897     s->b_height= h;
1898
1899     s->block= av_mallocz(w * h * sizeof(BlockNode) << (s->block_max_depth*2));
1900     return 0;
1901 }
1902
1903 static inline void copy_rac_state(RangeCoder *d, RangeCoder *s){
1904     uint8_t *bytestream= d->bytestream;
1905     uint8_t *bytestream_start= d->bytestream_start;
1906     *d= *s;
1907     d->bytestream= bytestream;
1908     d->bytestream_start= bytestream_start;
1909 }
1910
1911 //near copy & paste from dsputil, FIXME
1912 static int pix_sum(uint8_t * pix, int line_size, int w)
1913 {
1914     int s, i, j;
1915
1916     s = 0;
1917     for (i = 0; i < w; i++) {
1918         for (j = 0; j < w; j++) {
1919             s += pix[0];
1920             pix ++;
1921         }
1922         pix += line_size - w;
1923     }
1924     return s;
1925 }
1926
1927 //near copy & paste from dsputil, FIXME
1928 static int pix_norm1(uint8_t * pix, int line_size, int w)
1929 {
1930     int s, i, j;
1931     uint32_t *sq = squareTbl + 256;
1932
1933     s = 0;
1934     for (i = 0; i < w; i++) {
1935         for (j = 0; j < w; j ++) {
1936             s += sq[pix[0]];
1937             pix ++;
1938         }
1939         pix += line_size - w;
1940     }
1941     return s;
1942 }
1943
1944 static inline void set_blocks(SnowContext *s, int level, int x, int y, int l, int cb, int cr, int mx, int my, int type){
1945     const int w= s->b_width << s->block_max_depth;
1946     const int rem_depth= s->block_max_depth - level;
1947     const int index= (x + y*w) << rem_depth;
1948     const int block_w= 1<<rem_depth;
1949     BlockNode block;
1950     int i,j;
1951
1952     block.color[0]= l;
1953     block.color[1]= cb;
1954     block.color[2]= cr;
1955     block.mx= mx;
1956     block.my= my;
1957     block.type= type;
1958     block.level= level;
1959
1960     for(j=0; j<block_w; j++){
1961         for(i=0; i<block_w; i++){
1962             s->block[index + i + j*w]= block;
1963         }
1964     }
1965 }
1966
1967 static inline void init_ref(MotionEstContext *c, uint8_t *src[3], uint8_t *ref[3], uint8_t *ref2[3], int x, int y, int ref_index){
1968     const int offset[3]= {
1969           y*c->  stride + x,
1970         ((y*c->uvstride + x)>>1),
1971         ((y*c->uvstride + x)>>1),
1972     };
1973     int i;
1974     for(i=0; i<3; i++){
1975         c->src[0][i]= src [i];
1976         c->ref[0][i]= ref [i] + offset[i];
1977     }
1978     assert(!ref_index);
1979 }
1980
1981 //FIXME copy&paste
1982 #define P_LEFT P[1]
1983 #define P_TOP P[2]
1984 #define P_TOPRIGHT P[3]
1985 #define P_MEDIAN P[4]
1986 #define P_MV1 P[9]
1987 #define FLAG_QPEL   1 //must be 1
1988
1989 static int encode_q_branch(SnowContext *s, int level, int x, int y){
1990     uint8_t p_buffer[1024];
1991     uint8_t i_buffer[1024];
1992     uint8_t p_state[sizeof(s->block_state)];
1993     uint8_t i_state[sizeof(s->block_state)];
1994     RangeCoder pc, ic;
1995     uint8_t *pbbak= s->c.bytestream;
1996     uint8_t *pbbak_start= s->c.bytestream_start;
1997     int score, score2, iscore, i_len, p_len, block_s, sum;
1998     const int w= s->b_width  << s->block_max_depth;
1999     const int h= s->b_height << s->block_max_depth;
2000     const int rem_depth= s->block_max_depth - level;
2001     const int index= (x + y*w) << rem_depth;
2002     const int block_w= 1<<(LOG2_MB_SIZE - level);
2003     static BlockNode null_block= { //FIXME add border maybe
2004         .color= {128,128,128},
2005         .mx= 0,
2006         .my= 0,
2007         .type= 0,
2008         .level= 0,
2009     };
2010     int trx= (x+1)<<rem_depth;
2011     int try= (y+1)<<rem_depth;
2012     BlockNode *left  = x ? &s->block[index-1] : &null_block;
2013     BlockNode *top   = y ? &s->block[index-w] : &null_block;
2014     BlockNode *right = trx<w ? &s->block[index+1] : &null_block;
2015     BlockNode *bottom= try<h ? &s->block[index+w] : &null_block;
2016     BlockNode *tl    = y && x ? &s->block[index-w-1] : left;
2017     BlockNode *tr    = y && trx<w && ((x&1)==0 || level==0) ? &s->block[index-w+(1<<rem_depth)] : tl; //FIXME use lt
2018     int pl = left->color[0];
2019     int pcb= left->color[1];
2020     int pcr= left->color[2];
2021     int pmx= mid_pred(left->mx, top->mx, tr->mx);
2022     int pmy= mid_pred(left->my, top->my, tr->my);
2023     int mx=0, my=0;
2024     int l,cr,cb, i;
2025     const int stride= s->current_picture.linesize[0];
2026     const int uvstride= s->current_picture.linesize[1];
2027     const int instride= s->input_picture.linesize[0];
2028     const int uvinstride= s->input_picture.linesize[1];
2029     uint8_t *new_l = s->input_picture.data[0] + (x + y*  instride)*block_w;
2030     uint8_t *new_cb= s->input_picture.data[1] + (x + y*uvinstride)*block_w/2;
2031     uint8_t *new_cr= s->input_picture.data[2] + (x + y*uvinstride)*block_w/2;
2032     uint8_t current_mb[3][stride*block_w];
2033     uint8_t *current_data[3]= {&current_mb[0][0], &current_mb[1][0], &current_mb[2][0]};
2034     int P[10][2];
2035     int16_t last_mv[3][2];
2036     int qpel= !!(s->avctx->flags & CODEC_FLAG_QPEL); //unused
2037     const int shift= 1+qpel;
2038     MotionEstContext *c= &s->m.me;
2039     int mx_context= av_log2(2*ABS(left->mx - top->mx));
2040     int my_context= av_log2(2*ABS(left->my - top->my));
2041     int s_context= 2*left->level + 2*top->level + tl->level + tr->level;
2042
2043     assert(sizeof(s->block_state) >= 256);
2044     if(s->keyframe){
2045         set_blocks(s, level, x, y, pl, pcb, pcr, pmx, pmy, BLOCK_INTRA);
2046         return 0;
2047     }
2048
2049     //FIXME optimize
2050     for(i=0; i<block_w; i++)
2051         memcpy(&current_mb[0][0] +   stride*i, new_l  +   instride*i, block_w);
2052     for(i=0; i<block_w>>1; i++)
2053         memcpy(&current_mb[1][0] + uvstride*i, new_cb + uvinstride*i, block_w>>1);
2054     for(i=0; i<block_w>>1; i++)
2055         memcpy(&current_mb[2][0] + uvstride*i, new_cr + uvinstride*i, block_w>>1);
2056
2057 //    clip predictors / edge ?
2058
2059     P_LEFT[0]= left->mx;
2060     P_LEFT[1]= left->my;
2061     P_TOP [0]= top->mx;
2062     P_TOP [1]= top->my;
2063     P_TOPRIGHT[0]= tr->mx;
2064     P_TOPRIGHT[1]= tr->my;
2065
2066     last_mv[0][0]= s->block[index].mx;
2067     last_mv[0][1]= s->block[index].my;
2068     last_mv[1][0]= right->mx;
2069     last_mv[1][1]= right->my;
2070     last_mv[2][0]= bottom->mx;
2071     last_mv[2][1]= bottom->my;
2072
2073     s->m.mb_stride=2;
2074     s->m.mb_x=
2075     s->m.mb_y= 0;
2076     s->m.me.skip= 0;
2077
2078     init_ref(c, current_data, s->last_picture.data, NULL, block_w*x, block_w*y, 0);
2079
2080     assert(s->m.me.  stride ==   stride);
2081     assert(s->m.me.uvstride == uvstride);
2082
2083     c->penalty_factor    = get_penalty_factor(s->lambda, s->lambda2, c->avctx->me_cmp);
2084     c->sub_penalty_factor= get_penalty_factor(s->lambda, s->lambda2, c->avctx->me_sub_cmp);
2085     c->mb_penalty_factor = get_penalty_factor(s->lambda, s->lambda2, c->avctx->mb_cmp);
2086     c->current_mv_penalty= c->mv_penalty[s->m.f_code=1] + MAX_MV;
2087
2088     c->xmin = - x*block_w - 16+2;
2089     c->ymin = - y*block_w - 16+2;
2090     c->xmax = - (x+1)*block_w + (w<<(LOG2_MB_SIZE - s->block_max_depth)) + 16-2;
2091     c->ymax