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.
17508 Commits
32 Branches
283MB
Tree: 68e1794e3a
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '68e1794e3a'
${ noResults }
FFmpeg/libavcodec/4xm.c

843 lines
26KB
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 "libavutil/intreadwrite.h"

  28. #include "avcodec.h"

  29. #include "dsputil.h"

  30. #include "bitstream.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 uint16_t *wordstream;

  137.     int mv[256];

  138.     VLC pre_vlc;

  139.     int last_dc;

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

  141.     uint8_t *bitstream_buffer;

  142.     unsigned int bitstream_buffer_size;

  143.     int version;

  144.     CFrameBuffer cfrm[CFRAME_BUFFER_COUNT];

  145. } FourXContext;

  146. 

  147. 

  148. #define FIX_1_082392200  70936

  149. #define FIX_1_414213562  92682

  150. #define FIX_1_847759065 121095

  151. #define FIX_2_613125930 171254

  152. 

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

  154. 

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

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

  157.     int tmp10, tmp11, tmp12, tmp13;

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

  159.     int i;

  160.     int temp[64];

  161. 

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

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

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

  165. 

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

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

  168. 

  169.         tmp0 = tmp10 + tmp13;

  170.         tmp3 = tmp10 - tmp13;

  171.         tmp1 = tmp11 + tmp12;

  172.         tmp2 = tmp11 - tmp12;

  173. 

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

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

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

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

  178. 

  179.         tmp7  =          z11 + z13;

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

  181. 

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

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

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

  185. 

  186.         tmp6 = tmp12 - tmp7;

  187.         tmp5 = tmp11 - tmp6;

  188.         tmp4 = tmp10 + tmp5;

  189. 

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

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

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

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

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

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

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

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

  198.     }

  199. 

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

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

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

  203. 

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

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

  206. 

  207.         tmp0 = tmp10 + tmp13;

  208.         tmp3 = tmp10 - tmp13;

  209.         tmp1 = tmp11 + tmp12;

  210.         tmp2 = tmp11 - tmp12;

  211. 

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

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

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

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

  216. 

  217.         tmp7 = z11 + z13;

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

  219. 

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

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

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

  223. 

  224.         tmp6 = tmp12 - tmp7;

  225.         tmp5 = tmp11 - tmp6;

  226.         tmp4 = tmp10 + tmp5;

  227. 

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

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

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

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

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

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

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

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

  236.     }

  237. }

  238. 

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

  240.     int i;

  241. 

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

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

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

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

  246.     }

  247. }

  248. 

  249. static void init_mv(FourXContext *f){

  250.     int i;

  251. 

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

  253.         if(f->version>1)

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

  255.         else

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

  257.     }

  258. }

  259. 

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

  261.    int i;

  262.    dc*= 0x10001;

  263. 

  264.    switch(log2w){

  265.    case 0:

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

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

  268.             if(scale) src += stride;

  269.             dst += stride;

  270.         }

  271.         break;

  272.     case 1:

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

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

  275.             if(scale) src += stride;

  276.             dst += stride;

  277.         }

  278.         break;

  279.     case 2:

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

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

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

  283.             if(scale) src += stride;

  284.             dst += stride;

  285.         }

  286.         break;

  287.     case 3:

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

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

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

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

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

  293.             if(scale) src += stride;

  294.             dst += stride;

  295.         }

  296.         break;

  297.     default: assert(0);

  298.     }

  299. }

  300. 

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

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

  303.     const int h= 1<<log2h;

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

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

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

  307. 

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

  309. 

  310.     if(code == 0){

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

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

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

  314.             return;

  315.         }

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

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

  318.         log2h--;

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

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

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

  322.         log2w--;

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

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

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

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

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

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

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

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

  331.             return;

  332.         }

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

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

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

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

  337.         if(log2w){

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

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

  340.         }else{

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

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

  343.         }

  344.     }

  345. }

  346. 

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

  348.     int x, y;

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

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

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

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

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

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

  355. 

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

  357.         extra=20;

  358.         bitstream_size= AV_RL32(buf+8);

  359.         wordstream_size= AV_RL32(buf+12);

  360.         bytestream_size= AV_RL32(buf+16);

  361.     }else{

  362.         extra=0;

  363.         bitstream_size = AV_RL16(buf-4);

  364.         wordstream_size= AV_RL16(buf-2);

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

  366.     }

  367. 

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

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

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

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

  372.        ){

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

  374.         bitstream_size+ bytestream_size+ wordstream_size - length);

  375.         return -1;

  376.     }

  377. 

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

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

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

  381. 

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

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

  384. 

  385.     init_mv(f);

  386. 

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

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

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

  390.         }

  391.         src += 8*stride;

  392.         dst += 8*stride;

  393.     }

  394. 

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

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

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

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

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

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

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

  402.         );

  403. 

  404.     return 0;

  405. }

  406. 

  407. /**

  408.  * decode block and dequantize.

  409.  * Note this is almost identical to MJPEG.

  410.  */

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

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

  413. 

  414.     /* DC coef */

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

  416.     if (val>>4){

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

  418.     }

  419. 

  420.     if(val)

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

  422. 

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

  424.     f->last_dc =

  425.     block[0] = val;

  426.     /* AC coefs */

  427.     i = 1;

  428.     for(;;) {

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

  430. 

  431.         /* EOB */

  432.         if (code == 0)

  433.             break;

  434.         if (code == 0xf0) {

  435.             i += 16;

  436.         } else {

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

  438.             i += code >> 4;

  439.             if (i >= 64) {

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

  441.                 return 0;

  442.             }

  443. 

  444.             j= ff_zigzag_direct[i];

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

  446.             i++;

  447.             if (i >= 64)

  448.                 break;

  449.         }

  450.     }

  451. 

  452.     return 0;

  453. }

  454. 

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

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

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

  458.     int i;

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

  460. 

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

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

  463.         idct(block[i]);

  464.     }

  465. 

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

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

  468.     }

  469. 

  470. /* Note transform is:

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

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

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

  474. */

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

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

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

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

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

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

  481.             int y;

  482. 

  483.             cb+=cb;

  484. 

  485.             y = temp[0];

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

  487.             y = temp[1];

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

  489.             y = temp[8];

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

  491.             y = temp[9];

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

  493.             dst += 2;

  494.         }

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

  496.     }

  497. }

  498. 

  499. static int decode_i_mb(FourXContext *f){

  500.     int i;

  501. 

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

  503. 

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

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

  506.             return -1;

  507.     }

  508. 

  509.     return 0;

  510. }

  511. 

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

  513.     int frequency[512];

  514.     uint8_t flag[512];

  515.     int up[512];

  516.     uint8_t len_tab[257];

  517.     int bits_tab[257];

  518.     int start, end;

  519.     const uint8_t *ptr= buf;

  520.     int j;

  521. 

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

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

  524. 

  525.     start= *ptr++;

  526.     end= *ptr++;

  527.     for(;;){

  528.         int i;

  529. 

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

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

  532.         }

  533.         start= *ptr++;

  534.         if(start==0) break;

  535. 

  536.         end= *ptr++;

  537.     }

  538.     frequency[256]=1;

  539. 

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

  541. 

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

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

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

  545.         int i;

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

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

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

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

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

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

  552.                 }else{

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

  554.                 }

  555.             }

  556.         }

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

  558. 

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

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

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

  562.         up[ smallest[0] ]=

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

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

  565.     }

  566. 

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

  568.         int node;

  569.         int len=0;

  570.         int bits=0;

  571. 

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

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

  574.             len++;

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

  576.         }

  577. 

  578.         bits_tab[j]= bits;

  579.         len_tab[j]= len;

  580.     }

  581. 

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

  583.              len_tab , 1, 1,

  584.              bits_tab, 4, 4, 0);

  585. 

  586.     return ptr;

  587. }

  588. 

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

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

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

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

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

  594. }

  595. 

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

  597.     int x, y, x2, y2;

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

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

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

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

  602. 

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

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

  605.             unsigned int color[4], bits;

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

  607. //warning following is purely guessed ...

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

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

  610. 

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

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

  613. 

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

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

  616. 

  617.             bits= bytestream_get_le32(&buf);

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

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

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

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

  622.                 }

  623.             }

  624.             dst+=16;

  625.         }

  626.         dst += 16*stride - width;

  627.     }

  628. 

  629.     return 0;

  630. }

  631. 

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

  633.     int x, y;

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

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

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

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

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

  639.     const int token_count av_unused = AV_RL32(buf + bitstream_size + 8);

  640.     unsigned int prestream_size= 4*AV_RL32(buf + bitstream_size + 4);

  641.     const uint8_t *prestream= buf + bitstream_size + 12;

  642. 

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

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

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

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

  647.         return -1;

  648.     }

  649. 

  650.     prestream= read_huffman_tables(f, prestream);

  651. 

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

  653. 

  654.     prestream_size= length + buf - prestream;

  655. 

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

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

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

  659. 

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

  661. 

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

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

  664.             if(decode_i_mb(f) < 0)

  665.                 return -1;

  666. 

  667.             idct_put(f, x, y);

  668.         }

  669.         dst += 16*stride;

  670.     }

  671. 

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

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

  674. 

  675.     return 0;

  676. }

  677. 

  678. static int decode_frame(AVCodecContext *avctx,

  679.                         void *data, int *data_size,

  680.                         const uint8_t *buf, int buf_size)

  681. {

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

  683.     AVFrame *picture = data;

  684.     AVFrame *p, temp;

  685.     int i, frame_4cc, frame_size;

  686. 

  687.     frame_4cc= AV_RL32(buf);

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

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

  690.     }

  691. 

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

  693.         int free_index=-1;

  694.         const int data_size= buf_size - 20;

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

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

  697.         CFrameBuffer *cfrm;

  698. 

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

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

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

  702.         }

  703. 

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

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

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

  707.         }

  708. 

  709.         if(i>=CFRAME_BUFFER_COUNT){

  710.             i= free_index;

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

  712.         }

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

  714. 

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

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

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

  718.             return -1;

  719.         }

  720. 

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

  722.         cfrm->size += data_size;

  723. 

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

  725.             buf= cfrm->data;

  726.             frame_size= cfrm->size;

  727. 

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

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

  730.             }

  731. 

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

  733.             frame_4cc= AV_RL32("pfrm");

  734.         }else

  735.             return buf_size;

  736.     }else{

  737.         buf= buf + 12;

  738.         frame_size= buf_size - 12;

  739.     }

  740. 

  741.     temp= f->current_picture;

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

  743.     f->last_picture= temp;

  744. 

  745.     p= &f->current_picture;

  746.     avctx->coded_frame= p;

  747. 

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

  749. 

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

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

  752. 

  753.     p->reference= 1;

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

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

  756.         return -1;

  757.     }

  758. 

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

  760.         p->pict_type= FF_I_TYPE;

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

  762.             return -1;

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

  764.         p->pict_type= FF_I_TYPE;

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

  766.             return -1;

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

  768.         p->pict_type= FF_P_TYPE;

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

  770.             return -1;

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

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

  773.     }else{

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

  775.     }

  776. 

  777.     p->key_frame= p->pict_type == FF_I_TYPE;

  778. 

  779.     *picture= *p;

  780.     *data_size = sizeof(AVPicture);

  781. 

  782.     emms_c();

  783. 

  784.     return buf_size;

  785. }

  786. 

  787. 

  788. static void common_init(AVCodecContext *avctx){

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

  790. 

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

  792. 

  793.     f->avctx= avctx;

  794. }

  795. 

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

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

  798. 

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

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

  801.         return 1;

  802.     }

  803. 

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

  805.     common_init(avctx);

  806.     init_vlcs(f);

  807. 

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

  809.     else             avctx->pix_fmt= PIX_FMT_RGB555;

  810. 

  811.     return 0;

  812. }

  813. 

  814. 

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

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

  817.     int i;

  818. 

  819.     av_freep(&f->bitstream_buffer);

  820.     f->bitstream_buffer_size=0;

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

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

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

  824.     }

  825.     free_vlc(&f->pre_vlc);

  826. 

  827.     return 0;

  828. }

  829. 

  830. AVCodec fourxm_decoder = {

  831.     "4xm",

  832.     CODEC_TYPE_VIDEO,

  833.     CODEC_ID_4XM,

  834.     sizeof(FourXContext),

  835.     decode_init,

  836.     NULL,

  837.     decode_end,

  838.     decode_frame,

  839.     /*CODEC_CAP_DR1,*/

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

  841. };