falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
36007 Commits
32 Branches
959MB
Tree: 134aaa79f7
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '134aaa79f7'
${ noResults }
FFmpeg/libavcodec/4xm.c

933 lines
29KB
Raw Blame History 

  1. /*

  2.  * 4XM codec

  3.  * Copyright (c) 2003 Michael Niedermayer

  4.  *

  5.  * This file is part of FFmpeg.

  6.  *

  7.  * FFmpeg is free software; you can redistribute it and/or

  8.  * modify it under the terms of the GNU Lesser General Public

  9.  * License as published by the Free Software Foundation; either

  10.  * version 2.1 of the License, or (at your option) any later version.

  11.  *

  12.  * FFmpeg is distributed in the hope that it will be useful,

  13.  * but WITHOUT ANY WARRANTY; without even the implied warranty of

  14.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU

  15.  * Lesser General Public License for more details.

  16.  *

  17.  * You should have received a copy of the GNU Lesser General Public

  18.  * License along with FFmpeg; if not, write to the Free Software

  19.  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

  20.  */

  21. 

  22. /**

  23.  * @file

  24.  * 4XM codec.

  25.  */

  26. 

  27. #include "libavutil/intreadwrite.h"

  28. #include "avcodec.h"

  29. #include "dsputil.h"

  30. #include "get_bits.h"

  31. #include "bytestream.h"

  32. 

  33. //#undef NDEBUG

  34. //#include <assert.h>

  35. 

  36. #define BLOCK_TYPE_VLC_BITS 5

  37. #define ACDC_VLC_BITS 9

  38. 

  39. #define CFRAME_BUFFER_COUNT 100

  40. 

  41. static const uint8_t block_type_tab[2][4][8][2]={

  42.  {

  43.   {   //{8,4,2}x{8,4,2}

  44.     { 0,1}, { 2,2}, { 6,3}, {14,4}, {30,5}, {31,5}, { 0,0}

  45.   },{ //{8,4}x1

  46.     { 0,1}, { 0,0}, { 2,2}, { 6,3}, {14,4}, {15,4}, { 0,0}

  47.   },{ //1x{8,4}

  48.     { 0,1}, { 2,2}, { 0,0}, { 6,3}, {14,4}, {15,4}, { 0,0}

  49.   },{ //1x2, 2x1

  50.     { 0,1}, { 0,0}, { 0,0}, { 2,2}, { 6,3}, {14,4}, {15,4}

  51.   }

  52.  },{

  53.   {  //{8,4,2}x{8,4,2}

  54.     { 1,2}, { 4,3}, { 5,3}, {0,2}, {6,3}, {7,3}, {0,0}

  55.   },{//{8,4}x1

  56.     { 1,2}, { 0,0}, { 2,2}, {0,2}, {6,3}, {7,3}, {0,0}

  57.   },{//1x{8,4}

  58.     { 1,2}, { 2,2}, { 0,0}, {0,2}, {6,3}, {7,3}, {0,0}

  59.   },{//1x2, 2x1

  60.     { 1,2}, { 0,0}, { 0,0}, {0,2}, {2,2}, {6,3}, {7,3}

  61.   }

  62.  }

  63. };

  64. 

  65. static const uint8_t size2index[4][4]={

  66.   {-1, 3, 1, 1},

  67.   { 3, 0, 0, 0},

  68.   { 2, 0, 0, 0},

  69.   { 2, 0, 0, 0},

  70. };

  71. 

  72. static const int8_t mv[256][2]={

  73. {  0,  0},{  0, -1},{ -1,  0},{  1,  0},{  0,  1},{ -1, -1},{  1, -1},{ -1,  1},

  74. {  1,  1},{  0, -2},{ -2,  0},{  2,  0},{  0,  2},{ -1, -2},{  1, -2},{ -2, -1},

  75. {  2, -1},{ -2,  1},{  2,  1},{ -1,  2},{  1,  2},{ -2, -2},{  2, -2},{ -2,  2},

  76. {  2,  2},{  0, -3},{ -3,  0},{  3,  0},{  0,  3},{ -1, -3},{  1, -3},{ -3, -1},

  77. {  3, -1},{ -3,  1},{  3,  1},{ -1,  3},{  1,  3},{ -2, -3},{  2, -3},{ -3, -2},

  78. {  3, -2},{ -3,  2},{  3,  2},{ -2,  3},{  2,  3},{  0, -4},{ -4,  0},{  4,  0},

  79. {  0,  4},{ -1, -4},{  1, -4},{ -4, -1},{  4, -1},{  4,  1},{ -1,  4},{  1,  4},

  80. { -3, -3},{ -3,  3},{  3,  3},{ -2, -4},{ -4, -2},{  4, -2},{ -4,  2},{ -2,  4},

  81. {  2,  4},{ -3, -4},{  3, -4},{  4, -3},{ -5,  0},{ -4,  3},{ -3,  4},{  3,  4},

  82. { -1, -5},{ -5, -1},{ -5,  1},{ -1,  5},{ -2, -5},{  2, -5},{  5, -2},{  5,  2},

  83. { -4, -4},{ -4,  4},{ -3, -5},{ -5, -3},{ -5,  3},{  3,  5},{ -6,  0},{  0,  6},

  84. { -6, -1},{ -6,  1},{  1,  6},{  2, -6},{ -6,  2},{  2,  6},{ -5, -4},{  5,  4},

  85. {  4,  5},{ -6, -3},{  6,  3},{ -7,  0},{ -1, -7},{  5, -5},{ -7,  1},{ -1,  7},

  86. {  4, -6},{  6,  4},{ -2, -7},{ -7,  2},{ -3, -7},{  7, -3},{  3,  7},{  6, -5},

  87. {  0, -8},{ -1, -8},{ -7, -4},{ -8,  1},{  4,  7},{  2, -8},{ -2,  8},{  6,  6},

  88. { -8,  3},{  5, -7},{ -5,  7},{  8, -4},{  0, -9},{ -9, -1},{  1,  9},{  7, -6},

  89. { -7,  6},{ -5, -8},{ -5,  8},{ -9,  3},{  9, -4},{  7, -7},{  8, -6},{  6,  8},

  90. { 10,  1},{-10,  2},{  9, -5},{ 10, -3},{ -8, -7},{-10, -4},{  6, -9},{-11,  0},

  91. { 11,  1},{-11, -2},{ -2, 11},{  7, -9},{ -7,  9},{ 10,  6},{ -4, 11},{  8, -9},

  92. {  8,  9},{  5, 11},{  7,-10},{ 12, -3},{ 11,  6},{ -9, -9},{  8, 10},{  5, 12},

  93. {-11,  7},{ 13,  2},{  6,-12},{ 10,  9},{-11,  8},{ -7, 12},{  0, 14},{ 14, -2},

  94. { -9, 11},{ -6, 13},{-14, -4},{ -5,-14},{  5, 14},{-15, -1},{-14, -6},{  3,-15},

  95. { 11,-11},{ -7, 14},{ -5, 15},{  8,-14},{ 15,  6},{  3, 16},{  7,-15},{-16,  5},

  96. {  0, 17},{-16, -6},{-10, 14},{-16,  7},{ 12, 13},{-16,  8},{-17,  6},{-18,  3},

  97. { -7, 17},{ 15, 11},{ 16, 10},{  2,-19},{  3,-19},{-11,-16},{-18,  8},{-19, -6},

  98. {  2,-20},{-17,-11},{-10,-18},{  8, 19},{-21, -1},{-20,  7},{ -4, 21},{ 21,  5},

  99. { 15, 16},{  2,-22},{-10,-20},{-22,  5},{ 20,-11},{ -7,-22},{-12, 20},{ 23, -5},

  100. { 13,-20},{ 24, -2},{-15, 19},{-11, 22},{ 16, 19},{ 23,-10},{-18,-18},{ -9,-24},

  101. { 24,-10},{ -3, 26},{-23, 13},{-18,-20},{ 17, 21},{ -4, 27},{ 27,  6},{  1,-28},

  102. {-11, 26},{-17,-23},{  7, 28},{ 11,-27},{ 29,  5},{-23,-19},{-28,-11},{-21, 22},

  103. {-30,  7},{-17, 26},{-27, 16},{ 13, 29},{ 19,-26},{ 10,-31},{-14,-30},{ 20,-27},

  104. {-29, 18},{-16,-31},{-28,-22},{ 21,-30},{-25, 28},{ 26,-29},{ 25,-32},{-32,-32}

  105. };

  106. 

  107. // this is simply the scaled down elementwise product of the standard jpeg quantizer table and the AAN premul table

  108. static const uint8_t dequant_table[64]={

  109.  16, 15, 13, 19, 24, 31, 28, 17,

  110.  17, 23, 25, 31, 36, 63, 45, 21,

  111.  18, 24, 27, 37, 52, 59, 49, 20,

  112.  16, 28, 34, 40, 60, 80, 51, 20,

  113.  18, 31, 48, 66, 68, 86, 56, 21,

  114.  19, 38, 56, 59, 64, 64, 48, 20,

  115.  27, 48, 55, 55, 56, 51, 35, 15,

  116.  20, 35, 34, 32, 31, 22, 15,  8,

  117. };

  118. 

  119. static VLC block_type_vlc[2][4];

  120. 

  121. 

  122. typedef struct CFrameBuffer{

  123.     unsigned int allocated_size;

  124.     unsigned int size;

  125.     int id;

  126.     uint8_t *data;

  127. }CFrameBuffer;

  128. 

  129. typedef struct FourXContext{

  130.     AVCodecContext *avctx;

  131.     DSPContext dsp;

  132.     AVFrame current_picture, last_picture;

  133.     GetBitContext pre_gb;          ///< ac/dc prefix

  134.     GetBitContext gb;

  135.     const uint8_t *bytestream;

  136.     const uint8_t *bytestream_end;

  137.     const uint16_t *wordstream;

  138.     const uint16_t *wordstream_end;

  139.     int mv[256];

  140.     VLC pre_vlc;

  141.     int last_dc;

  142.     DECLARE_ALIGNED(16, DCTELEM, block)[6][64];

  143.     void *bitstream_buffer;

  144.     unsigned int bitstream_buffer_size;

  145.     int version;

  146.     CFrameBuffer cfrm[CFRAME_BUFFER_COUNT];

  147. } FourXContext;

  148. 

  149. 

  150. #define FIX_1_082392200  70936

  151. #define FIX_1_414213562  92682

  152. #define FIX_1_847759065 121095

  153. #define FIX_2_613125930 171254

  154. 

  155. #define MULTIPLY(var,const)  (((var)*(const)) >> 16)

  156. 

  157. static void idct(DCTELEM block[64]){

  158.     int tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;

  159.     int tmp10, tmp11, tmp12, tmp13;

  160.     int z5, z10, z11, z12, z13;

  161.     int i;

  162.     int temp[64];

  163. 

  164.     for(i=0; i<8; i++){

  165.         tmp10 = block[8*0 + i] + block[8*4 + i];

  166.         tmp11 = block[8*0 + i] - block[8*4 + i];

  167. 

  168.         tmp13 =          block[8*2 + i] + block[8*6 + i];

  169.         tmp12 = MULTIPLY(block[8*2 + i] - block[8*6 + i], FIX_1_414213562) - tmp13;

  170. 

  171.         tmp0 = tmp10 + tmp13;

  172.         tmp3 = tmp10 - tmp13;

  173.         tmp1 = tmp11 + tmp12;

  174.         tmp2 = tmp11 - tmp12;

  175. 

  176.         z13 = block[8*5 + i] + block[8*3 + i];

  177.         z10 = block[8*5 + i] - block[8*3 + i];

  178.         z11 = block[8*1 + i] + block[8*7 + i];

  179.         z12 = block[8*1 + i] - block[8*7 + i];

  180. 

  181.         tmp7  =          z11 + z13;

  182.         tmp11 = MULTIPLY(z11 - z13, FIX_1_414213562);

  183. 

  184.         z5    = MULTIPLY(z10 + z12, FIX_1_847759065);

  185.         tmp10 = MULTIPLY(z12, FIX_1_082392200) - z5;

  186.         tmp12 = MULTIPLY(z10, - FIX_2_613125930) + z5;

  187. 

  188.         tmp6 = tmp12 - tmp7;

  189.         tmp5 = tmp11 - tmp6;

  190.         tmp4 = tmp10 + tmp5;

  191. 

  192.         temp[8*0 + i] = tmp0 + tmp7;

  193.         temp[8*7 + i] = tmp0 - tmp7;

  194.         temp[8*1 + i] = tmp1 + tmp6;

  195.         temp[8*6 + i] = tmp1 - tmp6;

  196.         temp[8*2 + i] = tmp2 + tmp5;

  197.         temp[8*5 + i] = tmp2 - tmp5;

  198.         temp[8*4 + i] = tmp3 + tmp4;

  199.         temp[8*3 + i] = tmp3 - tmp4;

  200.     }

  201. 

  202.     for(i=0; i<8*8; i+=8){

  203.         tmp10 = temp[0 + i] + temp[4 + i];

  204.         tmp11 = temp[0 + i] - temp[4 + i];

  205. 

  206.         tmp13 = temp[2 + i] + temp[6 + i];

  207.         tmp12 = MULTIPLY(temp[2 + i] - temp[6 + i], FIX_1_414213562) - tmp13;

  208. 

  209.         tmp0 = tmp10 + tmp13;

  210.         tmp3 = tmp10 - tmp13;

  211.         tmp1 = tmp11 + tmp12;

  212.         tmp2 = tmp11 - tmp12;

  213. 

  214.         z13 = temp[5 + i] + temp[3 + i];

  215.         z10 = temp[5 + i] - temp[3 + i];

  216.         z11 = temp[1 + i] + temp[7 + i];

  217.         z12 = temp[1 + i] - temp[7 + i];

  218. 

  219.         tmp7 = z11 + z13;

  220.         tmp11 = MULTIPLY(z11 - z13, FIX_1_414213562);

  221. 

  222.         z5 = MULTIPLY(z10 + z12, FIX_1_847759065);

  223.         tmp10 = MULTIPLY(z12, FIX_1_082392200) - z5;

  224.         tmp12 = MULTIPLY(z10, - FIX_2_613125930) + z5;

  225. 

  226.         tmp6 = tmp12 - tmp7;

  227.         tmp5 = tmp11 - tmp6;

  228.         tmp4 = tmp10 + tmp5;

  229. 

  230.         block[0 + i] = (tmp0 + tmp7)>>6;

  231.         block[7 + i] = (tmp0 - tmp7)>>6;

  232.         block[1 + i] = (tmp1 + tmp6)>>6;

  233.         block[6 + i] = (tmp1 - tmp6)>>6;

  234.         block[2 + i] = (tmp2 + tmp5)>>6;

  235.         block[5 + i] = (tmp2 - tmp5)>>6;

  236.         block[4 + i] = (tmp3 + tmp4)>>6;

  237.         block[3 + i] = (tmp3 - tmp4)>>6;

  238.     }

  239. }

  240. 

  241. static av_cold void init_vlcs(FourXContext *f){

  242.     static VLC_TYPE table[8][32][2];

  243.     int i;

  244. 

  245.     for(i=0; i<8; i++){

  246.         block_type_vlc[0][i].table= table[i];

  247.         block_type_vlc[0][i].table_allocated= 32;

  248.         init_vlc(&block_type_vlc[0][i], BLOCK_TYPE_VLC_BITS, 7,

  249.                  &block_type_tab[0][i][0][1], 2, 1,

  250.                  &block_type_tab[0][i][0][0], 2, 1, INIT_VLC_USE_NEW_STATIC);

  251.     }

  252. }

  253. 

  254. static void init_mv(FourXContext *f){

  255.     int i;

  256. 

  257.     for(i=0; i<256; i++){

  258.         if(f->version>1)

  259.             f->mv[i] = mv[i][0]   + mv[i][1]  *f->current_picture.linesize[0]/2;

  260.         else

  261.             f->mv[i] = (i&15) - 8 + ((i>>4)-8)*f->current_picture.linesize[0]/2;

  262.     }

  263. }

  264. 

  265. #if HAVE_BIGENDIAN

  266. #define LE_CENTRIC_MUL(dst, src, scale, dc) \

  267.     { \

  268.         unsigned tmpval = AV_RN32(src);                 \

  269.         tmpval = (tmpval <<  16) | (tmpval >>  16);     \

  270.         tmpval = tmpval * (scale) + (dc);               \

  271.         tmpval = (tmpval <<  16) | (tmpval >>  16);     \

  272.         AV_WN32A(dst, tmpval);                          \

  273.     }

  274. #else

  275. #define LE_CENTRIC_MUL(dst, src, scale, dc) \

  276.     { \

  277.         unsigned tmpval = AV_RN32(src) * (scale) + (dc); \

  278.         AV_WN32A(dst, tmpval);                           \

  279.     }

  280. #endif

  281. 

  282. static inline void mcdc(uint16_t *dst, uint16_t *src, int log2w, int h, int stride, int scale, unsigned dc){

  283.    int i;

  284.    dc*= 0x10001;

  285. 

  286.    switch(log2w){

  287.    case 0:

  288.         for(i=0; i<h; i++){

  289.             dst[0] = scale*src[0] + dc;

  290.             if(scale) src += stride;

  291.             dst += stride;

  292.         }

  293.         break;

  294.     case 1:

  295.         for(i=0; i<h; i++){

  296.             LE_CENTRIC_MUL(dst, src, scale, dc);

  297.             if(scale) src += stride;

  298.             dst += stride;

  299.         }

  300.         break;

  301.     case 2:

  302.         for(i=0; i<h; i++){

  303.             LE_CENTRIC_MUL(dst,     src,     scale, dc);

  304.             LE_CENTRIC_MUL(dst + 2, src + 2, scale, dc);

  305.             if(scale) src += stride;

  306.             dst += stride;

  307.         }

  308.         break;

  309.     case 3:

  310.         for(i=0; i<h; i++){

  311.             LE_CENTRIC_MUL(dst,     src,     scale, dc);

  312.             LE_CENTRIC_MUL(dst + 2, src + 2, scale, dc);

  313.             LE_CENTRIC_MUL(dst + 4, src + 4, scale, dc);

  314.             LE_CENTRIC_MUL(dst + 6, src + 6, scale, dc);

  315.             if(scale) src += stride;

  316.             dst += stride;

  317.         }

  318.         break;

  319.     default: assert(0);

  320.     }

  321. }

  322. 

  323. static void decode_p_block(FourXContext *f, uint16_t *dst, uint16_t *src, int log2w, int log2h, int stride){

  324.     const int index= size2index[log2h][log2w];

  325.     const int h= 1<<log2h;

  326.     int code= get_vlc2(&f->gb, block_type_vlc[1-(f->version>1)][index].table, BLOCK_TYPE_VLC_BITS, 1);

  327.     uint16_t *start= (uint16_t*)f->last_picture.data[0];

  328.     uint16_t *end= start + stride*(f->avctx->height-h+1) - (1<<log2w);

  329. 

  330.     assert(code>=0 && code<=6);

  331. 

  332.     if(code == 0){

  333.         if (f->bytestream_end - f->bytestream < 1){

  334.             av_log(f->avctx, AV_LOG_ERROR, "bytestream overread\n");

  335.             return;

  336.         }

  337.         src += f->mv[ *f->bytestream++ ];

  338.         if(start > src || src > end){

  339.             av_log(f->avctx, AV_LOG_ERROR, "mv out of pic\n");

  340.             return;

  341.         }

  342.         mcdc(dst, src, log2w, h, stride, 1, 0);

  343.     }else if(code == 1){

  344.         log2h--;

  345.         decode_p_block(f, dst                  , src                  , log2w, log2h, stride);

  346.         decode_p_block(f, dst + (stride<<log2h), src + (stride<<log2h), log2w, log2h, stride);

  347.     }else if(code == 2){

  348.         log2w--;

  349.         decode_p_block(f, dst             , src             , log2w, log2h, stride);

  350.         decode_p_block(f, dst + (1<<log2w), src + (1<<log2w), log2w, log2h, stride);

  351.     }else if(code == 3 && f->version<2){

  352.         mcdc(dst, src, log2w, h, stride, 1, 0);

  353.     }else if(code == 4){

  354.         if (f->bytestream_end - f->bytestream < 1){

  355.             av_log(f->avctx, AV_LOG_ERROR, "bytestream overread\n");

  356.             return;

  357.         }

  358.         src += f->mv[ *f->bytestream++ ];

  359.         if(start > src || src > end){

  360.             av_log(f->avctx, AV_LOG_ERROR, "mv out of pic\n");

  361.             return;

  362.         }

  363.         if (f->wordstream_end - f->wordstream < 1){

  364.             av_log(f->avctx, AV_LOG_ERROR, "wordstream overread\n");

  365.             return;

  366.         }

  367.         mcdc(dst, src, log2w, h, stride, 1, av_le2ne16(*f->wordstream++));

  368.     }else if(code == 5){

  369.         if (f->wordstream_end - f->wordstream < 1){

  370.             av_log(f->avctx, AV_LOG_ERROR, "wordstream overread\n");

  371.             return;

  372.         }

  373.         mcdc(dst, src, log2w, h, stride, 0, av_le2ne16(*f->wordstream++));

  374.     }else if(code == 6){

  375.         if (f->wordstream_end - f->wordstream < 2){

  376.             av_log(f->avctx, AV_LOG_ERROR, "wordstream overread\n");

  377.             return;

  378.         }

  379.         if(log2w){

  380.             dst[0] = av_le2ne16(*f->wordstream++);

  381.             dst[1] = av_le2ne16(*f->wordstream++);

  382.         }else{

  383.             dst[0     ] = av_le2ne16(*f->wordstream++);

  384.             dst[stride] = av_le2ne16(*f->wordstream++);

  385.         }

  386.     }

  387. }

  388. 

  389. static int decode_p_frame(FourXContext *f, const uint8_t *buf, int length){

  390.     int x, y;

  391.     const int width= f->avctx->width;

  392.     const int height= f->avctx->height;

  393.     uint16_t *src= (uint16_t*)f->last_picture.data[0];

  394.     uint16_t *dst= (uint16_t*)f->current_picture.data[0];

  395.     const int stride= f->current_picture.linesize[0]>>1;

  396.     unsigned int bitstream_size, bytestream_size, wordstream_size, extra;

  397. 

  398.     if(f->version>1){

  399.         extra=20;

  400.         if (length < extra)

  401.             return -1;

  402.         bitstream_size= AV_RL32(buf+8);

  403.         wordstream_size= AV_RL32(buf+12);

  404.         bytestream_size= AV_RL32(buf+16);

  405.     }else{

  406.         extra=0;

  407.         bitstream_size = AV_RL16(buf-4);

  408.         wordstream_size= AV_RL16(buf-2);

  409.         bytestream_size= FFMAX(length - bitstream_size - wordstream_size, 0);

  410.     }

  411. 

  412.     if (bitstream_size > length ||

  413.         bytestream_size > length - bitstream_size ||

  414.         wordstream_size > length - bytestream_size - bitstream_size ||

  415.         extra > length - bytestream_size - bitstream_size - wordstream_size){

  416.         av_log(f->avctx, AV_LOG_ERROR, "lengths %d %d %d %d\n", bitstream_size, bytestream_size, wordstream_size,

  417.         bitstream_size+ bytestream_size+ wordstream_size - length);

  418.         return -1;

  419.     }

  420. 

  421.     av_fast_malloc(&f->bitstream_buffer, &f->bitstream_buffer_size, bitstream_size + FF_INPUT_BUFFER_PADDING_SIZE);

  422.     if (!f->bitstream_buffer)

  423.         return AVERROR(ENOMEM);

  424.     f->dsp.bswap_buf(f->bitstream_buffer, (const uint32_t*)(buf + extra), bitstream_size/4);

  425.     memset((uint8_t*)f->bitstream_buffer + bitstream_size, 0, FF_INPUT_BUFFER_PADDING_SIZE);

  426.     init_get_bits(&f->gb, f->bitstream_buffer, 8*bitstream_size);

  427. 

  428.     f->wordstream= (const uint16_t*)(buf + extra + bitstream_size);

  429.     f->wordstream_end= f->wordstream + wordstream_size/2;

  430.     f->bytestream= buf + extra + bitstream_size + wordstream_size;

  431.     f->bytestream_end = f->bytestream + bytestream_size;

  432. 

  433.     init_mv(f);

  434. 

  435.     for(y=0; y<height; y+=8){

  436.         for(x=0; x<width; x+=8){

  437.             decode_p_block(f, dst + x, src + x, 3, 3, stride);

  438.         }

  439.         src += 8*stride;

  440.         dst += 8*stride;

  441.     }

  442. 

  443.     return 0;

  444. }

  445. 

  446. /**

  447.  * decode block and dequantize.

  448.  * Note this is almost identical to MJPEG.

  449.  */

  450. static int decode_i_block(FourXContext *f, DCTELEM *block){

  451.     int code, i, j, level, val;

  452. 

  453.     if(get_bits_left(&f->gb) < 2){

  454.         av_log(f->avctx, AV_LOG_ERROR, "%d bits left before decode_i_block()\n", get_bits_left(&f->gb));

  455.         return -1;

  456.     }

  457. 

  458.     /* DC coef */

  459.     val = get_vlc2(&f->pre_gb, f->pre_vlc.table, ACDC_VLC_BITS, 3);

  460.     if (val>>4){

  461.         av_log(f->avctx, AV_LOG_ERROR, "error dc run != 0\n");

  462.     }

  463. 

  464.     if(val)

  465.         val = get_xbits(&f->gb, val);

  466. 

  467.     val = val * dequant_table[0] + f->last_dc;

  468.     f->last_dc =

  469.     block[0] = val;

  470.     /* AC coefs */

  471.     i = 1;

  472.     for(;;) {

  473.         code = get_vlc2(&f->pre_gb, f->pre_vlc.table, ACDC_VLC_BITS, 3);

  474. 

  475.         /* EOB */

  476.         if (code == 0)

  477.             break;

  478.         if (code == 0xf0) {

  479.             i += 16;

  480.         } else {

  481.             level = get_xbits(&f->gb, code & 0xf);

  482.             i += code >> 4;

  483.             if (i >= 64) {

  484.                 av_log(f->avctx, AV_LOG_ERROR, "run %d oveflow\n", i);

  485.                 return 0;

  486.             }

  487. 

  488.             j= ff_zigzag_direct[i];

  489.             block[j] = level * dequant_table[j];

  490.             i++;

  491.             if (i >= 64)

  492.                 break;

  493.         }

  494.     }

  495. 

  496.     return 0;

  497. }

  498. 

  499. static inline void idct_put(FourXContext *f, int x, int y){

  500.     DCTELEM (*block)[64]= f->block;

  501.     int stride= f->current_picture.linesize[0]>>1;

  502.     int i;

  503.     uint16_t *dst = ((uint16_t*)f->current_picture.data[0]) + y * stride + x;

  504. 

  505.     for(i=0; i<4; i++){

  506.         block[i][0] += 0x80*8*8;

  507.         idct(block[i]);

  508.     }

  509. 

  510.     if(!(f->avctx->flags&CODEC_FLAG_GRAY)){

  511.         for(i=4; i<6; i++) idct(block[i]);

  512.     }

  513. 

  514. /* Note transform is:

  515. y= ( 1b + 4g + 2r)/14

  516. cb=( 3b - 2g - 1r)/14

  517. cr=(-1b - 4g + 5r)/14

  518. */

  519.     for(y=0; y<8; y++){

  520.         for(x=0; x<8; x++){

  521.             DCTELEM *temp= block[(x>>2) + 2*(y>>2)] + 2*(x&3) + 2*8*(y&3); //FIXME optimize

  522.             int cb= block[4][x + 8*y];

  523.             int cr= block[5][x + 8*y];

  524.             int cg= (cb + cr)>>1;

  525.             int y;

  526. 

  527.             cb+=cb;

  528. 

  529.             y = temp[0];

  530.             dst[0       ]= ((y+cb)>>3) + (((y-cg)&0xFC)<<3) + (((y+cr)&0xF8)<<8);

  531.             y = temp[1];

  532.             dst[1       ]= ((y+cb)>>3) + (((y-cg)&0xFC)<<3) + (((y+cr)&0xF8)<<8);

  533.             y = temp[8];

  534.             dst[  stride]= ((y+cb)>>3) + (((y-cg)&0xFC)<<3) + (((y+cr)&0xF8)<<8);

  535.             y = temp[9];

  536.             dst[1+stride]= ((y+cb)>>3) + (((y-cg)&0xFC)<<3) + (((y+cr)&0xF8)<<8);

  537.             dst += 2;

  538.         }

  539.         dst += 2*stride - 2*8;

  540.     }

  541. }

  542. 

  543. static int decode_i_mb(FourXContext *f){

  544.     int i;

  545. 

  546.     f->dsp.clear_blocks(f->block[0]);

  547. 

  548.     for(i=0; i<6; i++){

  549.         if(decode_i_block(f, f->block[i]) < 0)

  550.             return -1;

  551.     }

  552. 

  553.     return 0;

  554. }

  555. 

  556. static const uint8_t *read_huffman_tables(FourXContext *f, const uint8_t * const buf, int buf_size){

  557.     int frequency[512];

  558.     uint8_t flag[512];

  559.     int up[512];

  560.     uint8_t len_tab[257];

  561.     int bits_tab[257];

  562.     int start, end;

  563.     const uint8_t *ptr= buf;

  564.     const uint8_t *ptr_end = buf + buf_size;

  565.     int j;

  566. 

  567.     memset(frequency, 0, sizeof(frequency));

  568.     memset(up, -1, sizeof(up));

  569. 

  570.     start= *ptr++;

  571.     end= *ptr++;

  572.     for(;;){

  573.         int i;

  574. 

  575.         if (start <= end && ptr_end - ptr < end - start + 1 + 1)

  576.             return NULL;

  577.         for(i=start; i<=end; i++){

  578.             frequency[i]= *ptr++;

  579.         }

  580.         start= *ptr++;

  581.         if(start==0) break;

  582. 

  583.         end= *ptr++;

  584.     }

  585.     frequency[256]=1;

  586. 

  587.     while((ptr - buf)&3) ptr++; // 4byte align

  588. 

  589.     for(j=257; j<512; j++){

  590.         int min_freq[2]= {256*256, 256*256};

  591.         int smallest[2]= {0, 0};

  592.         int i;

  593.         for(i=0; i<j; i++){

  594.             if(frequency[i] == 0) continue;

  595.             if(frequency[i] < min_freq[1]){

  596.                 if(frequency[i] < min_freq[0]){

  597.                     min_freq[1]= min_freq[0]; smallest[1]= smallest[0];

  598.                     min_freq[0]= frequency[i];smallest[0]= i;

  599.                 }else{

  600.                     min_freq[1]= frequency[i];smallest[1]= i;

  601.                 }

  602.             }

  603.         }

  604.         if(min_freq[1] == 256*256) break;

  605. 

  606.         frequency[j]= min_freq[0] + min_freq[1];

  607.         flag[ smallest[0] ]= 0;

  608.         flag[ smallest[1] ]= 1;

  609.         up[ smallest[0] ]=

  610.         up[ smallest[1] ]= j;

  611.         frequency[ smallest[0] ]= frequency[ smallest[1] ]= 0;

  612.     }

  613. 

  614.     for(j=0; j<257; j++){

  615.         int node;

  616.         int len=0;

  617.         int bits=0;

  618. 

  619.         for(node= j; up[node] != -1; node= up[node]){

  620.             bits += flag[node]<<len;

  621.             len++;

  622.             if(len > 31) av_log(f->avctx, AV_LOG_ERROR, "vlc length overflow\n"); //can this happen at all ?

  623.         }

  624. 

  625.         bits_tab[j]= bits;

  626.         len_tab[j]= len;

  627.     }

  628. 

  629.     if (init_vlc(&f->pre_vlc, ACDC_VLC_BITS, 257,

  630.                  len_tab , 1, 1,

  631.                  bits_tab, 4, 4, 0))

  632.         return NULL;

  633. 

  634.     return ptr;

  635. }

  636. 

  637. static int mix(int c0, int c1){

  638.     int blue = 2*(c0&0x001F) + (c1&0x001F);

  639.     int green= (2*(c0&0x03E0) + (c1&0x03E0))>>5;

  640.     int red  = 2*(c0>>10) + (c1>>10);

  641.     return red/3*1024 + green/3*32 + blue/3;

  642. }

  643. 

  644. static int decode_i2_frame(FourXContext *f, const uint8_t *buf, int length){

  645.     int x, y, x2, y2;

  646.     const int width= f->avctx->width;

  647.     const int height= f->avctx->height;

  648.     uint16_t *dst= (uint16_t*)f->current_picture.data[0];

  649.     const int stride= f->current_picture.linesize[0]>>1;

  650.     const uint8_t *buf_end = buf + length;

  651. 

  652.     for(y=0; y<height; y+=16){

  653.         for(x=0; x<width; x+=16){

  654.             unsigned int color[4], bits;

  655.             if (buf_end - buf < 8)

  656.                 return -1;

  657.             memset(color, 0, sizeof(color));

  658. //warning following is purely guessed ...

  659.             color[0]= bytestream_get_le16(&buf);

  660.             color[1]= bytestream_get_le16(&buf);

  661. 

  662.             if(color[0]&0x8000) av_log(NULL, AV_LOG_ERROR, "unk bit 1\n");

  663.             if(color[1]&0x8000) av_log(NULL, AV_LOG_ERROR, "unk bit 2\n");

  664. 

  665.             color[2]= mix(color[0], color[1]);

  666.             color[3]= mix(color[1], color[0]);

  667. 

  668.             bits= bytestream_get_le32(&buf);

  669.             for(y2=0; y2<16; y2++){

  670.                 for(x2=0; x2<16; x2++){

  671.                     int index= 2*(x2>>2) + 8*(y2>>2);

  672.                     dst[y2*stride+x2]= color[(bits>>index)&3];

  673.                 }

  674.             }

  675.             dst+=16;

  676.         }

  677.         dst += 16*stride - width;

  678.     }

  679. 

  680.     return 0;

  681. }

  682. 

  683. static int decode_i_frame(FourXContext *f, const uint8_t *buf, int length){

  684.     int x, y;

  685.     const int width= f->avctx->width;

  686.     const int height= f->avctx->height;

  687.     uint16_t *dst= (uint16_t*)f->current_picture.data[0];

  688.     const int stride= f->current_picture.linesize[0]>>1;

  689.     const unsigned int bitstream_size= AV_RL32(buf);

  690.     unsigned int prestream_size;

  691.     const uint8_t *prestream;

  692. 

  693.     if (bitstream_size > (1<<26) || length < bitstream_size + 12)

  694.         return -1;

  695.     prestream_size = 4*AV_RL32(buf + bitstream_size + 4);

  696.     prestream = buf + bitstream_size + 12;

  697. 

  698.     if (prestream_size > (1<<26) ||

  699.         prestream_size != length - (bitstream_size + 12)){

  700.         av_log(f->avctx, AV_LOG_ERROR, "size mismatch %d %d %d\n", prestream_size, bitstream_size, length);

  701.         return -1;

  702.     }

  703. 

  704.     prestream= read_huffman_tables(f, prestream, buf + length - prestream);

  705.     if (!prestream)

  706.         return -1;

  707. 

  708.     init_get_bits(&f->gb, buf + 4, 8*bitstream_size);

  709. 

  710.     prestream_size= length + buf - prestream;

  711. 

  712.     av_fast_malloc(&f->bitstream_buffer, &f->bitstream_buffer_size, prestream_size + FF_INPUT_BUFFER_PADDING_SIZE);

  713.     if (!f->bitstream_buffer)

  714.         return AVERROR(ENOMEM);

  715.     f->dsp.bswap_buf(f->bitstream_buffer, (const uint32_t*)prestream, prestream_size/4);

  716.     memset((uint8_t*)f->bitstream_buffer + prestream_size, 0, FF_INPUT_BUFFER_PADDING_SIZE);

  717.     init_get_bits(&f->pre_gb, f->bitstream_buffer, 8*prestream_size);

  718. 

  719.     f->last_dc= 0*128*8*8;

  720. 

  721.     for(y=0; y<height; y+=16){

  722.         for(x=0; x<width; x+=16){

  723.             if(decode_i_mb(f) < 0)

  724.                 return -1;

  725. 

  726.             idct_put(f, x, y);

  727.         }

  728.         dst += 16*stride;

  729.     }

  730. 

  731.     if(get_vlc2(&f->pre_gb, f->pre_vlc.table, ACDC_VLC_BITS, 3) != 256)

  732.         av_log(f->avctx, AV_LOG_ERROR, "end mismatch\n");

  733. 

  734.     return 0;

  735. }

  736. 

  737. static int decode_frame(AVCodecContext *avctx,

  738.                         void *data, int *data_size,

  739.                         AVPacket *avpkt)

  740. {

  741.     const uint8_t *buf = avpkt->data;

  742.     int buf_size = avpkt->size;

  743.     FourXContext * const f = avctx->priv_data;

  744.     AVFrame *picture = data;

  745.     AVFrame *p, temp;

  746.     int i, frame_4cc, frame_size;

  747. 

  748.     if (buf_size < 12)

  749.         return AVERROR_INVALIDDATA;

  750.     frame_4cc= AV_RL32(buf);

  751.     if(buf_size != AV_RL32(buf+4)+8 || buf_size < 20){

  752.         av_log(f->avctx, AV_LOG_ERROR, "size mismatch %d %d\n", buf_size, AV_RL32(buf+4));

  753.     }

  754. 

  755.     if(frame_4cc == AV_RL32("cfrm")){

  756.         int free_index=-1;

  757.         const int data_size= buf_size - 20;

  758.         const int id= AV_RL32(buf+12);

  759.         const int whole_size= AV_RL32(buf+16);

  760.         CFrameBuffer *cfrm;

  761. 

  762.         if (data_size < 0 || whole_size < 0){

  763.             av_log(f->avctx, AV_LOG_ERROR, "sizes invalid\n");

  764.             return AVERROR_INVALIDDATA;

  765.         }

  766. 

  767.         for(i=0; i<CFRAME_BUFFER_COUNT; i++){

  768.             if(f->cfrm[i].id && f->cfrm[i].id < avctx->frame_number)

  769.                 av_log(f->avctx, AV_LOG_ERROR, "lost c frame %d\n", f->cfrm[i].id);

  770.         }

  771. 

  772.         for(i=0; i<CFRAME_BUFFER_COUNT; i++){

  773.             if(f->cfrm[i].id   == id) break;

  774.             if(f->cfrm[i].size == 0 ) free_index= i;

  775.         }

  776. 

  777.         if(i>=CFRAME_BUFFER_COUNT){

  778.             i= free_index;

  779.             f->cfrm[i].id= id;

  780.         }

  781.         cfrm= &f->cfrm[i];

  782. 

  783.         if (data_size > UINT_MAX -  cfrm->size - FF_INPUT_BUFFER_PADDING_SIZE)

  784.             return AVERROR_INVALIDDATA;

  785.         cfrm->data= av_fast_realloc(cfrm->data, &cfrm->allocated_size, cfrm->size + data_size + FF_INPUT_BUFFER_PADDING_SIZE);

  786.         if(!cfrm->data){ //explicit check needed as memcpy below might not catch a NULL

  787.             av_log(f->avctx, AV_LOG_ERROR, "realloc falure");

  788.             return -1;

  789.         }

  790. 

  791.         memcpy(cfrm->data + cfrm->size, buf+20, data_size);

  792.         cfrm->size += data_size;

  793. 

  794.         if(cfrm->size >= whole_size){

  795.             buf= cfrm->data;

  796.             frame_size= cfrm->size;

  797. 

  798.             if(id != avctx->frame_number){

  799.                 av_log(f->avctx, AV_LOG_ERROR, "cframe id mismatch %d %d\n", id, avctx->frame_number);

  800.             }

  801. 

  802.             cfrm->size= cfrm->id= 0;

  803.             frame_4cc= AV_RL32("pfrm");

  804.         }else

  805.             return buf_size;

  806.     }else{

  807.         buf= buf + 12;

  808.         frame_size= buf_size - 12;

  809.     }

  810. 

  811.     temp= f->current_picture;

  812.     f->current_picture= f->last_picture;

  813.     f->last_picture= temp;

  814. 

  815.     p= &f->current_picture;

  816.     avctx->coded_frame= p;

  817. 

  818.     avctx->flags |= CODEC_FLAG_EMU_EDGE; // alternatively we would have to use our own buffer management

  819. 

  820.     p->reference= 3;

  821.     if (avctx->reget_buffer(avctx, p) < 0) {

  822.         av_log(avctx, AV_LOG_ERROR, "reget_buffer() failed\n");

  823.         return -1;

  824.     }

  825. 

  826.     if(frame_4cc == AV_RL32("ifr2")){

  827.         p->pict_type= AV_PICTURE_TYPE_I;

  828.         if(decode_i2_frame(f, buf-4, frame_size+4) < 0){

  829.             av_log(f->avctx, AV_LOG_ERROR, "decode i2 frame failed\n");

  830.             return -1;

  831.         }

  832.     }else if(frame_4cc == AV_RL32("ifrm")){

  833.         p->pict_type= AV_PICTURE_TYPE_I;

  834.         if(decode_i_frame(f, buf, frame_size) < 0){

  835.             av_log(f->avctx, AV_LOG_ERROR, "decode i frame failed\n");

  836.             return -1;

  837.         }

  838.     }else if(frame_4cc == AV_RL32("pfrm") || frame_4cc == AV_RL32("pfr2")){

  839.         if(!f->last_picture.data[0]){

  840.             f->last_picture.reference= 3;

  841.             if(avctx->get_buffer(avctx, &f->last_picture) < 0){

  842.                 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");

  843.                 return -1;

  844.             }

  845.         }

  846. 

  847.         p->pict_type= AV_PICTURE_TYPE_P;

  848.         if(decode_p_frame(f, buf, frame_size) < 0){

  849.             av_log(f->avctx, AV_LOG_ERROR, "decode p frame failed\n");

  850.             return -1;

  851.         }

  852.     }else if(frame_4cc == AV_RL32("snd_")){

  853.         av_log(avctx, AV_LOG_ERROR, "ignoring snd_ chunk length:%d\n", buf_size);

  854.     }else{

  855.         av_log(avctx, AV_LOG_ERROR, "ignoring unknown chunk length:%d\n", buf_size);

  856.     }

  857. 

  858.     p->key_frame= p->pict_type == AV_PICTURE_TYPE_I;

  859. 

  860.     *picture= *p;

  861.     *data_size = sizeof(AVPicture);

  862. 

  863.     emms_c();

  864. 

  865.     return buf_size;

  866. }

  867. 

  868. 

  869. static av_cold void common_init(AVCodecContext *avctx){

  870.     FourXContext * const f = avctx->priv_data;

  871. 

  872.     dsputil_init(&f->dsp, avctx);

  873. 

  874.     f->avctx= avctx;

  875. }

  876. 

  877. static av_cold int decode_init(AVCodecContext *avctx){

  878.     FourXContext * const f = avctx->priv_data;

  879. 

  880.     if(avctx->extradata_size != 4 || !avctx->extradata) {

  881.         av_log(avctx, AV_LOG_ERROR, "extradata wrong or missing\n");

  882.         return 1;

  883.     }

  884.     if((avctx->width % 16) || (avctx->height % 16)) {

  885.         av_log(avctx, AV_LOG_ERROR, "unsupported width/height\n");

  886.         return AVERROR_INVALIDDATA;

  887.     }

  888. 

  889.     avcodec_get_frame_defaults(&f->current_picture);

  890.     avcodec_get_frame_defaults(&f->last_picture);

  891.     f->version= AV_RL32(avctx->extradata)>>16;

  892.     common_init(avctx);

  893.     init_vlcs(f);

  894. 

  895.     if(f->version>2) avctx->pix_fmt= PIX_FMT_RGB565;

  896.     else             avctx->pix_fmt= PIX_FMT_BGR555;

  897. 

  898.     return 0;

  899. }

  900. 

  901. 

  902. static av_cold int decode_end(AVCodecContext *avctx){

  903.     FourXContext * const f = avctx->priv_data;

  904.     int i;

  905. 

  906.     av_freep(&f->bitstream_buffer);

  907.     f->bitstream_buffer_size=0;

  908.     for(i=0; i<CFRAME_BUFFER_COUNT; i++){

  909.         av_freep(&f->cfrm[i].data);

  910.         f->cfrm[i].allocated_size= 0;

  911.     }

  912.     free_vlc(&f->pre_vlc);

  913.     if(f->current_picture.data[0])

  914.         avctx->release_buffer(avctx, &f->current_picture);

  915.     if(f->last_picture.data[0])

  916.         avctx->release_buffer(avctx, &f->last_picture);

  917. 

  918.     return 0;

  919. }

  920. 

  921. AVCodec ff_fourxm_decoder = {

  922.     .name           = "4xm",

  923.     .type           = AVMEDIA_TYPE_VIDEO,

  924.     .id             = CODEC_ID_4XM,

  925.     .priv_data_size = sizeof(FourXContext),

  926.     .init           = decode_init,

  927.     .close          = decode_end,

  928.     .decode         = decode_frame,

  929.     .capabilities   = CODEC_CAP_DR1,

  930.     .long_name = NULL_IF_CONFIG_SMALL("4X Movie"),

  931. };