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.
10562 Commits
32 Branches
312MB
Tree: 40802d36bc
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '40802d36bc'
${ noResults }
FFmpeg/libavcodec/4xm.c

829 lines
25KB
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 4xm.c

  24.  * 4XM codec.

  25.  */

  26. 

  27. #include "avcodec.h"

  28. #include "dsputil.h"

  29. #include "mpegvideo.h"

  30. 

  31. //#undef NDEBUG

  32. //#include <assert.h>

  33. 

  34. #define BLOCK_TYPE_VLC_BITS 5

  35. #define ACDC_VLC_BITS 9

  36. 

  37. #define CFRAME_BUFFER_COUNT 100

  38. 

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

  40.  {

  41.   {   //{8,4,2}x{8,4,2}

  42.     { 0,1}, { 2,2}, { 6,3}, {14,4}, {30,5}, {31,5}, { 0,0}

  43.   },{ //{8,4}x1

  44.     { 0,1}, { 0,0}, { 2,2}, { 6,3}, {14,4}, {15,4}, { 0,0}

  45.   },{ //1x{8,4}

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

  47.   },{ //1x2, 2x1

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

  49.   }

  50.  },{

  51.   {  //{8,4,2}x{8,4,2}

  52.     { 1,2}, { 4,3}, { 5,3}, {0,2}, {6,3}, {7,3}, {0,0}

  53.   },{//{8,4}x1

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

  55.   },{//1x{8,4}

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

  57.   },{//1x2, 2x1

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

  59.   }

  60.  }

  61. };

  62. 

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

  64.   {-1, 3, 1, 1},

  65.   { 3, 0, 0, 0},

  66.   { 2, 0, 0, 0},

  67.   { 2, 0, 0, 0},

  68. };

  69. 

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

  71. {  0,  0},{  0, -1},{ -1,  0},{  1,  0},{  0,  1},{ -1, -1},{  1, -1},{ -1,  1},

  72. {  1,  1},{  0, -2},{ -2,  0},{  2,  0},{  0,  2},{ -1, -2},{  1, -2},{ -2, -1},

  73. {  2, -1},{ -2,  1},{  2,  1},{ -1,  2},{  1,  2},{ -2, -2},{  2, -2},{ -2,  2},

  74. {  2,  2},{  0, -3},{ -3,  0},{  3,  0},{  0,  3},{ -1, -3},{  1, -3},{ -3, -1},

  75. {  3, -1},{ -3,  1},{  3,  1},{ -1,  3},{  1,  3},{ -2, -3},{  2, -3},{ -3, -2},

  76. {  3, -2},{ -3,  2},{  3,  2},{ -2,  3},{  2,  3},{  0, -4},{ -4,  0},{  4,  0},

  77. {  0,  4},{ -1, -4},{  1, -4},{ -4, -1},{  4, -1},{  4,  1},{ -1,  4},{  1,  4},

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

  79. {  2,  4},{ -3, -4},{  3, -4},{  4, -3},{ -5,  0},{ -4,  3},{ -3,  4},{  3,  4},

  80. { -1, -5},{ -5, -1},{ -5,  1},{ -1,  5},{ -2, -5},{  2, -5},{  5, -2},{  5,  2},

  81. { -4, -4},{ -4,  4},{ -3, -5},{ -5, -3},{ -5,  3},{  3,  5},{ -6,  0},{  0,  6},

  82. { -6, -1},{ -6,  1},{  1,  6},{  2, -6},{ -6,  2},{  2,  6},{ -5, -4},{  5,  4},

  83. {  4,  5},{ -6, -3},{  6,  3},{ -7,  0},{ -1, -7},{  5, -5},{ -7,  1},{ -1,  7},

  84. {  4, -6},{  6,  4},{ -2, -7},{ -7,  2},{ -3, -7},{  7, -3},{  3,  7},{  6, -5},

  85. {  0, -8},{ -1, -8},{ -7, -4},{ -8,  1},{  4,  7},{  2, -8},{ -2,  8},{  6,  6},

  86. { -8,  3},{  5, -7},{ -5,  7},{  8, -4},{  0, -9},{ -9, -1},{  1,  9},{  7, -6},

  87. { -7,  6},{ -5, -8},{ -5,  8},{ -9,  3},{  9, -4},{  7, -7},{  8, -6},{  6,  8},

  88. { 10,  1},{-10,  2},{  9, -5},{ 10, -3},{ -8, -7},{-10, -4},{  6, -9},{-11,  0},

  89. { 11,  1},{-11, -2},{ -2, 11},{  7, -9},{ -7,  9},{ 10,  6},{ -4, 11},{  8, -9},

  90. {  8,  9},{  5, 11},{  7,-10},{ 12, -3},{ 11,  6},{ -9, -9},{  8, 10},{  5, 12},

  91. {-11,  7},{ 13,  2},{  6,-12},{ 10,  9},{-11,  8},{ -7, 12},{  0, 14},{ 14, -2},

  92. { -9, 11},{ -6, 13},{-14, -4},{ -5,-14},{  5, 14},{-15, -1},{-14, -6},{  3,-15},

  93. { 11,-11},{ -7, 14},{ -5, 15},{  8,-14},{ 15,  6},{  3, 16},{  7,-15},{-16,  5},

  94. {  0, 17},{-16, -6},{-10, 14},{-16,  7},{ 12, 13},{-16,  8},{-17,  6},{-18,  3},

  95. { -7, 17},{ 15, 11},{ 16, 10},{  2,-19},{  3,-19},{-11,-16},{-18,  8},{-19, -6},

  96. {  2,-20},{-17,-11},{-10,-18},{  8, 19},{-21, -1},{-20,  7},{ -4, 21},{ 21,  5},

  97. { 15, 16},{  2,-22},{-10,-20},{-22,  5},{ 20,-11},{ -7,-22},{-12, 20},{ 23, -5},

  98. { 13,-20},{ 24, -2},{-15, 19},{-11, 22},{ 16, 19},{ 23,-10},{-18,-18},{ -9,-24},

  99. { 24,-10},{ -3, 26},{-23, 13},{-18,-20},{ 17, 21},{ -4, 27},{ 27,  6},{  1,-28},

  100. {-11, 26},{-17,-23},{  7, 28},{ 11,-27},{ 29,  5},{-23,-19},{-28,-11},{-21, 22},

  101. {-30,  7},{-17, 26},{-27, 16},{ 13, 29},{ 19,-26},{ 10,-31},{-14,-30},{ 20,-27},

  102. {-29, 18},{-16,-31},{-28,-22},{ 21,-30},{-25, 28},{ 26,-29},{ 25,-32},{-32,-32}

  103. };

  104. 

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

  106. static const uint8_t dequant_table[64]={

  107.  16, 15, 13, 19, 24, 31, 28, 17,

  108.  17, 23, 25, 31, 36, 63, 45, 21,

  109.  18, 24, 27, 37, 52, 59, 49, 20,

  110.  16, 28, 34, 40, 60, 80, 51, 20,

  111.  18, 31, 48, 66, 68, 86, 56, 21,

  112.  19, 38, 56, 59, 64, 64, 48, 20,

  113.  27, 48, 55, 55, 56, 51, 35, 15,

  114.  20, 35, 34, 32, 31, 22, 15,  8,

  115. };

  116. 

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

  118. 

  119. 

  120. typedef struct CFrameBuffer{

  121.     unsigned int allocated_size;

  122.     unsigned int size;

  123.     int id;

  124.     uint8_t *data;

  125. }CFrameBuffer;

  126. 

  127. typedef struct FourXContext{

  128.     AVCodecContext *avctx;

  129.     DSPContext dsp;

  130.     AVFrame current_picture, last_picture;

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

  132.     GetBitContext gb;

  133.     uint8_t *bytestream;

  134.     uint16_t *wordstream;

  135.     int mv[256];

  136.     VLC pre_vlc;

  137.     int last_dc;

  138.     DECLARE_ALIGNED_8(DCTELEM, block[6][64]);

  139.     uint8_t *bitstream_buffer;

  140.     unsigned int bitstream_buffer_size;

  141.     int version;

  142.     CFrameBuffer cfrm[CFRAME_BUFFER_COUNT];

  143. } FourXContext;

  144. 

  145. 

  146. #define FIX_1_082392200  70936

  147. #define FIX_1_414213562  92682

  148. #define FIX_1_847759065 121095

  149. #define FIX_2_613125930 171254

  150. 

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

  152. 

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

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

  155.     int tmp10, tmp11, tmp12, tmp13;

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

  157.     int i;

  158.     int temp[64];

  159. 

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

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

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

  163. 

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

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

  166. 

  167.         tmp0 = tmp10 + tmp13;

  168.         tmp3 = tmp10 - tmp13;

  169.         tmp1 = tmp11 + tmp12;

  170.         tmp2 = tmp11 - tmp12;

  171. 

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

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

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

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

  176. 

  177.         tmp7  =          z11 + z13;

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

  179. 

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

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

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

  183. 

  184.         tmp6 = tmp12 - tmp7;

  185.         tmp5 = tmp11 - tmp6;

  186.         tmp4 = tmp10 + tmp5;

  187. 

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

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

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

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

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

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

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

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

  196.     }

  197. 

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

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

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

  201. 

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

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

  204. 

  205.         tmp0 = tmp10 + tmp13;

  206.         tmp3 = tmp10 - tmp13;

  207.         tmp1 = tmp11 + tmp12;

  208.         tmp2 = tmp11 - tmp12;

  209. 

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

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

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

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

  214. 

  215.         tmp7 = z11 + z13;

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

  217. 

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

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

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

  221. 

  222.         tmp6 = tmp12 - tmp7;

  223.         tmp5 = tmp11 - tmp6;

  224.         tmp4 = tmp10 + tmp5;

  225. 

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

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

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

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

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

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

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

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

  234.     }

  235. }

  236. 

  237. static void init_vlcs(FourXContext *f){

  238.     int i;

  239. 

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

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

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

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

  244.     }

  245. }

  246. 

  247. static void init_mv(FourXContext *f){

  248.     int i;

  249. 

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

  251.         if(f->version)

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

  253.         else

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

  255.     }

  256. }

  257. 

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

  259.    int i;

  260.    dc*= 0x10001;

  261. 

  262.    switch(log2w){

  263.    case 0:

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

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

  266.             if(scale) src += stride;

  267.             dst += stride;

  268.         }

  269.         break;

  270.     case 1:

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

  272.             ((uint32_t*)dst)[0] = scale*((uint32_t*)src)[0] + dc;

  273.             if(scale) src += stride;

  274.             dst += stride;

  275.         }

  276.         break;

  277.     case 2:

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

  279.             ((uint32_t*)dst)[0] = scale*((uint32_t*)src)[0] + dc;

  280.             ((uint32_t*)dst)[1] = scale*((uint32_t*)src)[1] + dc;

  281.             if(scale) src += stride;

  282.             dst += stride;

  283.         }

  284.         break;

  285.     case 3:

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

  287.             ((uint32_t*)dst)[0] = scale*((uint32_t*)src)[0] + dc;

  288.             ((uint32_t*)dst)[1] = scale*((uint32_t*)src)[1] + dc;

  289.             ((uint32_t*)dst)[2] = scale*((uint32_t*)src)[2] + dc;

  290.             ((uint32_t*)dst)[3] = scale*((uint32_t*)src)[3] + dc;

  291.             if(scale) src += stride;

  292.             dst += stride;

  293.         }

  294.         break;

  295.     default: assert(0);

  296.     }

  297. }

  298. 

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

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

  301.     const int h= 1<<log2h;

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

  303. 

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

  305. 

  306.     if(code == 0){

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

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

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

  310.         log2h--;

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

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

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

  314.         log2w--;

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

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

  317.     }else if(code == 3 && f->version==0){

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

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

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

  321.         mcdc(dst, src, log2w, h, stride, 1, le2me_16(*f->wordstream++));

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

  323.         mcdc(dst, src, log2w, h, stride, 0, le2me_16(*f->wordstream++));

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

  325.         if(log2w){

  326.             dst[0] = le2me_16(*f->wordstream++);

  327.             dst[1] = le2me_16(*f->wordstream++);

  328.         }else{

  329.             dst[0     ] = le2me_16(*f->wordstream++);

  330.             dst[stride] = le2me_16(*f->wordstream++);

  331.         }

  332.     }

  333. }

  334. 

  335. static int get32(void *p){

  336.     return le2me_32(*(uint32_t*)p);

  337. }

  338. 

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

  340.     int x, y;

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

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

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

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

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

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

  347. 

  348.     if(f->version){

  349.         extra=20;

  350.         bitstream_size= get32(buf+8);

  351.         wordstream_size= get32(buf+12);

  352.         bytestream_size= get32(buf+16);

  353.     }else{

  354.         extra=0;

  355.         bitstream_size = AV_RL16(buf-4);

  356.         wordstream_size= AV_RL16(buf-2);

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

  358.     }

  359. 

  360.     if(bitstream_size+ bytestream_size+ wordstream_size + extra != length

  361.        || bitstream_size  > (1<<26)

  362.        || bytestream_size > (1<<26)

  363.        || wordstream_size > (1<<26)

  364.        ){

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

  366.         bitstream_size+ bytestream_size+ wordstream_size - length);

  367.         return -1;

  368.     }

  369. 

  370.     f->bitstream_buffer= av_fast_realloc(f->bitstream_buffer, &f->bitstream_buffer_size, bitstream_size + FF_INPUT_BUFFER_PADDING_SIZE);

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

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

  373. 

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

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

  376. 

  377.     init_mv(f);

  378. 

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

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

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

  382.         }

  383.         src += 8*stride;

  384.         dst += 8*stride;

  385.     }

  386. 

  387.     if(   bitstream_size != (get_bits_count(&f->gb)+31)/32*4

  388.        || (((char*)f->wordstream - (char*)buf + 2)&~2) != extra + bitstream_size + wordstream_size

  389.        || (((char*)f->bytestream - (char*)buf + 3)&~3) != extra + bitstream_size + wordstream_size + bytestream_size)

  390.         av_log(f->avctx, AV_LOG_ERROR, " %d %td %td bytes left\n",

  391.             bitstream_size - (get_bits_count(&f->gb)+31)/32*4,

  392.             -(((char*)f->bytestream - (char*)buf + 3)&~3) + (extra + bitstream_size + wordstream_size + bytestream_size),

  393.             -(((char*)f->wordstream - (char*)buf + 2)&~2) + (extra + bitstream_size + wordstream_size)

  394.         );

  395. 

  396.     return 0;

  397. }

  398. 

  399. /**

  400.  * decode block and dequantize.

  401.  * Note this is allmost identical to mjpeg

  402.  */

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

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

  405. 

  406.     /* DC coef */

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

  408.     if (val>>4){

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

  410.     }

  411. 

  412.     if(val)

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

  414. 

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

  416.     f->last_dc =

  417.     block[0] = val;

  418.     /* AC coefs */

  419.     i = 1;

  420.     for(;;) {

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

  422. 

  423.         /* EOB */

  424.         if (code == 0)

  425.             break;

  426.         if (code == 0xf0) {

  427.             i += 16;

  428.         } else {

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

  430.             i += code >> 4;

  431.             if (i >= 64) {

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

  433.                 return 0;

  434.             }

  435. 

  436.             j= ff_zigzag_direct[i];

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

  438.             i++;

  439.             if (i >= 64)

  440.                 break;

  441.         }

  442.     }

  443. 

  444.     return 0;

  445. }

  446. 

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

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

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

  450.     int i;

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

  452. 

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

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

  455.         idct(block[i]);

  456.     }

  457. 

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

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

  460.     }

  461. 

  462. /* Note transform is:

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

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

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

  466. */

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

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

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

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

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

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

  473.             int y;

  474. 

  475.             cb+=cb;

  476. 

  477.             y = temp[0];

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

  479.             y = temp[1];

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

  481.             y = temp[8];

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

  483.             y = temp[9];

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

  485.             dst += 2;

  486.         }

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

  488.     }

  489. }

  490. 

  491. static int decode_i_mb(FourXContext *f){

  492.     int i;

  493. 

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

  495. 

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

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

  498.             return -1;

  499.     }

  500. 

  501.     return 0;

  502. }

  503. 

  504. static uint8_t *read_huffman_tables(FourXContext *f, uint8_t * const buf){

  505.     int frequency[512];

  506.     uint8_t flag[512];

  507.     int up[512];

  508.     uint8_t len_tab[257];

  509.     int bits_tab[257];

  510.     int start, end;

  511.     uint8_t *ptr= buf;

  512.     int j;

  513. 

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

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

  516. 

  517.     start= *ptr++;

  518.     end= *ptr++;

  519.     for(;;){

  520.         int i;

  521. 

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

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

  524.         }

  525.         start= *ptr++;

  526.         if(start==0) break;

  527. 

  528.         end= *ptr++;

  529.     }

  530.     frequency[256]=1;

  531. 

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

  533. 

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

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

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

  537.         int i;

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

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

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

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

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

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

  544.                 }else{

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

  546.                 }

  547.             }

  548.         }

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

  550. 

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

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

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

  554.         up[ smallest[0] ]=

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

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

  557.     }

  558. 

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

  560.         int node;

  561.         int len=0;

  562.         int bits=0;

  563. 

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

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

  566.             len++;

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

  568.         }

  569. 

  570.         bits_tab[j]= bits;

  571.         len_tab[j]= len;

  572.     }

  573. 

  574.     init_vlc(&f->pre_vlc, ACDC_VLC_BITS, 257,

  575.              len_tab , 1, 1,

  576.              bits_tab, 4, 4, 0);

  577. 

  578.     return ptr;

  579. }

  580. 

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

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

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

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

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

  586. }

  587. 

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

  589.     int x, y, x2, y2;

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

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

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

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

  594. 

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

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

  597.             unsigned int color[4], bits;

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

  599. //warning following is purely guessed ...

  600.             color[0]= AV_RN16(buf); buf+=2; //FIXME use bytestream

  601.             color[1]= AV_RN16(buf); buf+=2;

  602. 

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

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

  605. 

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

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

  608. 

  609.             bits= AV_RL32(buf); buf+= 4;

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

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

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

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

  614.                 }

  615.             }

  616.             dst+=16;

  617.         }

  618.         dst += 16*stride - width;

  619.     }

  620. 

  621.     return 0;

  622. }

  623. 

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

  625.     int x, y;

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

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

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

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

  630.     const unsigned int bitstream_size= get32(buf);

  631.     const int token_count av_unused = get32(buf + bitstream_size + 8);

  632.     unsigned int prestream_size= 4*get32(buf + bitstream_size + 4);

  633.     uint8_t *prestream= buf + bitstream_size + 12;

  634. 

  635.     if(prestream_size + bitstream_size + 12 != length

  636.        || bitstream_size > (1<<26)

  637.        || prestream_size > (1<<26)){

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

  639.         return -1;

  640.     }

  641. 

  642.     prestream= read_huffman_tables(f, prestream);

  643. 

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

  645. 

  646.     prestream_size= length + buf - prestream;

  647. 

  648.     f->bitstream_buffer= av_fast_realloc(f->bitstream_buffer, &f->bitstream_buffer_size, prestream_size + FF_INPUT_BUFFER_PADDING_SIZE);

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

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

  651. 

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

  653. 

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

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

  656.             if(decode_i_mb(f) < 0)

  657.                 return -1;

  658. 

  659.             idct_put(f, x, y);

  660.         }

  661.         dst += 16*stride;

  662.     }

  663. 

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

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

  666. 

  667.     return 0;

  668. }

  669. 

  670. static int decode_frame(AVCodecContext *avctx,

  671.                         void *data, int *data_size,

  672.                         uint8_t *buf, int buf_size)

  673. {

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

  675.     AVFrame *picture = data;

  676.     AVFrame *p, temp;

  677.     int i, frame_4cc, frame_size;

  678. 

  679.     frame_4cc= get32(buf);

  680.     if(buf_size != get32(buf+4)+8 || buf_size < 20){

  681.         av_log(f->avctx, AV_LOG_ERROR, "size mismatch %d %d\n", buf_size, get32(buf+4));

  682.     }

  683. 

  684.     if(frame_4cc == ff_get_fourcc("cfrm")){

  685.         int free_index=-1;

  686.         const int data_size= buf_size - 20;

  687.         const int id= get32(buf+12);

  688.         const int whole_size= get32(buf+16);

  689.         CFrameBuffer *cfrm;

  690. 

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

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

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

  694.         }

  695. 

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

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

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

  699.         }

  700. 

  701.         if(i>=CFRAME_BUFFER_COUNT){

  702.             i= free_index;

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

  704.         }

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

  706. 

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

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

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

  710.             return -1;

  711.         }

  712. 

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

  714.         cfrm->size += data_size;

  715. 

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

  717.             buf= cfrm->data;

  718.             frame_size= cfrm->size;

  719. 

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

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

  722.             }

  723. 

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

  725.             frame_4cc= ff_get_fourcc("pfrm");

  726.         }else

  727.             return buf_size;

  728.     }else{

  729.         buf= buf + 12;

  730.         frame_size= buf_size - 12;

  731.     }

  732. 

  733.     temp= f->current_picture;

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

  735.     f->last_picture= temp;

  736. 

  737.     p= &f->current_picture;

  738.     avctx->coded_frame= p;

  739. 

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

  741. 

  742.     if(p->data[0])

  743.         avctx->release_buffer(avctx, p);

  744. 

  745.     p->reference= 1;

  746.     if(avctx->get_buffer(avctx, p) < 0){

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

  748.         return -1;

  749.     }

  750. 

  751.     if(frame_4cc == ff_get_fourcc("ifr2")){

  752.         p->pict_type= I_TYPE;

  753.         if(decode_i2_frame(f, buf-4, frame_size) < 0)

  754.             return -1;

  755.     }else if(frame_4cc == ff_get_fourcc("ifrm")){

  756.         p->pict_type= I_TYPE;

  757.         if(decode_i_frame(f, buf, frame_size) < 0)

  758.             return -1;

  759.     }else if(frame_4cc == ff_get_fourcc("pfrm") || frame_4cc == ff_get_fourcc("pfr2")){

  760.         p->pict_type= P_TYPE;

  761.         if(decode_p_frame(f, buf, frame_size) < 0)

  762.             return -1;

  763.     }else if(frame_4cc == ff_get_fourcc("snd_")){

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

  765.     }else{

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

  767.     }

  768. 

  769.     p->key_frame= p->pict_type == I_TYPE;

  770. 

  771.     *picture= *p;

  772.     *data_size = sizeof(AVPicture);

  773. 

  774.     emms_c();

  775. 

  776.     return buf_size;

  777. }

  778. 

  779. 

  780. static void common_init(AVCodecContext *avctx){

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

  782. 

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

  784. 

  785.     f->avctx= avctx;

  786. }

  787. 

  788. static int decode_init(AVCodecContext *avctx){

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

  790. 

  791.     f->version= avctx->codec_tag == 0x40000;

  792.     common_init(avctx);

  793.     init_vlcs(f);

  794. 

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

  796.     else           avctx->pix_fmt= PIX_FMT_RGB555;

  797. 

  798.     return 0;

  799. }

  800. 

  801. 

  802. static int decode_end(AVCodecContext *avctx){

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

  804.     int i;

  805. 

  806.     av_freep(&f->bitstream_buffer);

  807.     f->bitstream_buffer_size=0;

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

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

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

  811.     }

  812.     free_vlc(&f->pre_vlc);

  813. 

  814.     return 0;

  815. }

  816. 

  817. AVCodec fourxm_decoder = {

  818.     "4xm",

  819.     CODEC_TYPE_VIDEO,

  820.     CODEC_ID_4XM,

  821.     sizeof(FourXContext),

  822.     decode_init,

  823.     NULL,

  824.     decode_end,

  825.     decode_frame,

  826.     /*CODEC_CAP_DR1,*/

  827. };