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.
37311 Commits
32 Branches
1.0GB
Tree: 1fa24d08b2
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '1fa24d08b2'
${ noResults }
FFmpeg/libavcodec/4xm.c

939 lines
30KB
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.     GetByteContext g;

  136.     GetByteContext g2;

  137.     int mv[256];

  138.     VLC pre_vlc;

  139.     int last_dc;

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

  141.     void *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.     static VLC_TYPE table[8][32][2];

  241.     int i;

  242. 

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

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

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

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

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

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

  249.     }

  250. }

  251. 

  252. static void init_mv(FourXContext *f){

  253.     int i;

  254. 

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

  256.         if(f->version>1)

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

  258.         else

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

  260.     }

  261. }

  262. 

  263. #if HAVE_BIGENDIAN

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

  265.     { \

  266.         unsigned tmpval = AV_RN32(src);                 \

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

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

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

  270.         AV_WN32A(dst, tmpval);                          \

  271.     }

  272. #else

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

  274.     { \

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

  276.         AV_WN32A(dst, tmpval);                           \

  277.     }

  278. #endif

  279. 

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

  281.    int i;

  282.    dc*= 0x10001;

  283. 

  284.    switch(log2w){

  285.    case 0:

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

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

  288.             if(scale) src += stride;

  289.             dst += stride;

  290.         }

  291.         break;

  292.     case 1:

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

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

  295.             if(scale) src += stride;

  296.             dst += stride;

  297.         }

  298.         break;

  299.     case 2:

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

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

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

  303.             if(scale) src += stride;

  304.             dst += stride;

  305.         }

  306.         break;

  307.     case 3:

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

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

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

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

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

  313.             if(scale) src += stride;

  314.             dst += stride;

  315.         }

  316.         break;

  317.     default: assert(0);

  318.     }

  319. }

  320. 

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

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

  323.     const int h= 1<<log2h;

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

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

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

  327. 

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

  329. 

  330.     if(code == 0){

  331.         if (f->g.buffer_end - f->g.buffer < 1){

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

  333.             return;

  334.         }

  335.         src += f->mv[ *f->g.buffer++ ];

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

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

  338.             return;

  339.         }

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

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

  342.         log2h--;

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

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

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

  346.         log2w--;

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

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

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

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

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

  352.         if (f->g.buffer_end - f->g.buffer < 1){

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

  354.             return;

  355.         }

  356.         src += f->mv[ *f->g.buffer++ ];

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

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

  359.             return;

  360.         }

  361.         if (f->g2.buffer_end - f->g2.buffer < 1){

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

  363.             return;

  364.         }

  365.         mcdc(dst, src, log2w, h, stride, 1, bytestream2_get_le16(&f->g2));

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

  367.         if (f->g2.buffer_end - f->g2.buffer < 1){

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

  369.             return;

  370.         }

  371.         mcdc(dst, src, log2w, h, stride, 0, bytestream2_get_le16(&f->g2));

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

  373.         if (f->g2.buffer_end - f->g2.buffer < 2){

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

  375.             return;

  376.         }

  377.         if(log2w){

  378.             dst[0] = bytestream2_get_le16(&f->g2);

  379.             dst[1] = bytestream2_get_le16(&f->g2);

  380.         }else{

  381.             dst[0     ] = bytestream2_get_le16(&f->g2);

  382.             dst[stride] = bytestream2_get_le16(&f->g2);

  383.         }

  384.     }

  385. }

  386. 

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

  388.     int x, y;

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

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

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

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

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

  394.     unsigned int bitstream_size, bytestream_size, wordstream_size, extra, bytestream_offset, wordstream_offset;

  395. 

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

  397.         extra=20;

  398.         if (length < extra)

  399.             return -1;

  400.         bitstream_size= AV_RL32(buf+8);

  401.         wordstream_size= AV_RL32(buf+12);

  402.         bytestream_size= AV_RL32(buf+16);

  403.     }else{

  404.         extra=0;

  405.         bitstream_size = AV_RL16(buf-4);

  406.         wordstream_size= AV_RL16(buf-2);

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

  408.     }

  409. 

  410.     if (bitstream_size > length ||

  411.         bytestream_size > length - bitstream_size ||

  412.         wordstream_size > length - bytestream_size - bitstream_size ||

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

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

  415.         bitstream_size+ bytestream_size+ wordstream_size - length);

  416.         return -1;

  417.     }

  418. 

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

  420.     if (!f->bitstream_buffer)

  421.         return AVERROR(ENOMEM);

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

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

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

  425. 

  426.     wordstream_offset = extra + bitstream_size;

  427.     bytestream_offset = extra + bitstream_size + wordstream_size;

  428.     bytestream2_init(&f->g2, buf + wordstream_offset, length - wordstream_offset);

  429.     bytestream2_init(&f->g,  buf + bytestream_offset, length - bytestream_offset);

  430. 

  431.     init_mv(f);

  432. 

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

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

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

  436.         }

  437.         src += 8*stride;

  438.         dst += 8*stride;

  439.     }

  440. 

  441.     return 0;

  442. }

  443. 

  444. /**

  445.  * decode block and dequantize.

  446.  * Note this is almost identical to MJPEG.

  447.  */

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

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

  450. 

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

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

  453.         return -1;

  454.     }

  455. 

  456.     /* DC coef */

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

  458.     if (val>>4){

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

  460.     }

  461. 

  462.     if(val)

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

  464. 

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

  466.     f->last_dc =

  467.     block[0] = val;

  468.     /* AC coefs */

  469.     i = 1;

  470.     for(;;) {

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

  472. 

  473.         /* EOB */

  474.         if (code == 0)

  475.             break;

  476.         if (code == 0xf0) {

  477.             i += 16;

  478.         } else {

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

  480.             i += code >> 4;

  481.             if (i >= 64) {

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

  483.                 return 0;

  484.             }

  485. 

  486.             j= ff_zigzag_direct[i];

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

  488.             i++;

  489.             if (i >= 64)

  490.                 break;

  491.         }

  492.     }

  493. 

  494.     return 0;

  495. }

  496. 

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

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

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

  500.     int i;

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

  502. 

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

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

  505.         idct(block[i]);

  506.     }

  507. 

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

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

  510.     }

  511. 

  512. /* Note transform is:

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

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

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

  516. */

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

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

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

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

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

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

  523.             int y;

  524. 

  525.             cb+=cb;

  526. 

  527.             y = temp[0];

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

  529.             y = temp[1];

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

  531.             y = temp[8];

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

  533.             y = temp[9];

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

  535.             dst += 2;

  536.         }

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

  538.     }

  539. }

  540. 

  541. static int decode_i_mb(FourXContext *f){

  542.     int i;

  543. 

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

  545. 

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

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

  548.             return -1;

  549.     }

  550. 

  551.     return 0;

  552. }

  553. 

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

  555.     int frequency[512];

  556.     uint8_t flag[512];

  557.     int up[512];

  558.     uint8_t len_tab[257];

  559.     int bits_tab[257];

  560.     int start, end;

  561.     const uint8_t *ptr= buf;

  562.     const uint8_t *ptr_end = buf + buf_size;

  563.     int j;

  564. 

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

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

  567. 

  568.     start= *ptr++;

  569.     end= *ptr++;

  570.     for(;;){

  571.         int i;

  572. 

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

  574.             return NULL;

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

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

  577.         }

  578.         start= *ptr++;

  579.         if(start==0) break;

  580. 

  581.         end= *ptr++;

  582.     }

  583.     frequency[256]=1;

  584. 

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

  586. 

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

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

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

  590.         int i;

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

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

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

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

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

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

  597.                 }else{

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

  599.                 }

  600.             }

  601.         }

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

  603. 

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

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

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

  607.         up[ smallest[0] ]=

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

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

  610.     }

  611. 

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

  613.         int node;

  614.         int len=0;

  615.         int bits=0;

  616. 

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

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

  619.             len++;

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

  621.         }

  622. 

  623.         bits_tab[j]= bits;

  624.         len_tab[j]= len;

  625.     }

  626. 

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

  628.                  len_tab , 1, 1,

  629.                  bits_tab, 4, 4, 0))

  630.         return NULL;

  631. 

  632.     return ptr;

  633. }

  634. 

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

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

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

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

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

  640. }

  641. 

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

  643.     int x, y, x2, y2;

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

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

  646.     const int mbs = (FFALIGN(width, 16) >> 4) * (FFALIGN(height, 16) >> 4);

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

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

  649.     const uint8_t *buf_end = buf + length;

  650.     GetByteContext g3;

  651. 

  652.     if(length < mbs * 8) {

  653.         av_log(f->avctx, AV_LOG_ERROR, "packet size too small\n");

  654.         return AVERROR_INVALIDDATA;

  655.     }

  656.     bytestream2_init(&g3, buf, length);

  657. 

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

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

  660.             unsigned int color[4], bits;

  661.             if (buf_end - buf < 8)

  662.                 return -1;

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

  664. //warning following is purely guessed ...

  665.             color[0]= bytestream2_get_le16u(&g3);

  666.             color[1]= bytestream2_get_le16u(&g3);

  667. 

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

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

  670. 

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

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

  673. 

  674.             bits= bytestream2_get_le32u(&g3);

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

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

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

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

  679.                 }

  680.             }

  681.             dst+=16;

  682.         }

  683.         dst += 16 * stride - x;

  684.     }

  685. 

  686.     return 0;

  687. }

  688. 

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

  690.     int x, y;

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

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

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

  694.     unsigned int prestream_size;

  695.     const uint8_t *prestream;

  696. 

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

  698.         av_log(f->avctx, AV_LOG_ERROR, "packet size too small\n");

  699.         return AVERROR_INVALIDDATA;

  700.     }

  701. 

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

  703.     prestream      = buf + bitstream_size + 12;

  704. 

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

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

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

  708.         return -1;

  709.     }

  710. 

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

  712.     if (!prestream)

  713.         return -1;

  714. 

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

  716. 

  717.     prestream_size= length + buf - prestream;

  718. 

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

  720.     if (!f->bitstream_buffer)

  721.         return AVERROR(ENOMEM);

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

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

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

  725. 

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

  727. 

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

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

  730.             if(decode_i_mb(f) < 0)

  731.                 return -1;

  732. 

  733.             idct_put(f, x, y);

  734.         }

  735.     }

  736. 

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

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

  739. 

  740.     return 0;

  741. }

  742. 

  743. static int decode_frame(AVCodecContext *avctx,

  744.                         void *data, int *data_size,

  745.                         AVPacket *avpkt)

  746. {

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

  748.     int buf_size = avpkt->size;

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

  750.     AVFrame *picture = data;

  751.     AVFrame *p, temp;

  752.     int i, frame_4cc, frame_size;

  753. 

  754.     if (buf_size < 12)

  755.         return AVERROR_INVALIDDATA;

  756.     frame_4cc= AV_RL32(buf);

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

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

  759.     }

  760. 

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

  762.         int free_index=-1;

  763.         const int data_size= buf_size - 20;

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

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

  766.         CFrameBuffer *cfrm;

  767. 

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

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

  770.             return AVERROR_INVALIDDATA;

  771.         }

  772. 

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

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

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

  776.         }

  777. 

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

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

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

  781.         }

  782. 

  783.         if(i>=CFRAME_BUFFER_COUNT){

  784.             i= free_index;

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

  786.         }

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

  788. 

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

  790.             return AVERROR_INVALIDDATA;

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

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

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

  794.             return -1;

  795.         }

  796. 

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

  798.         cfrm->size += data_size;

  799. 

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

  801.             buf= cfrm->data;

  802.             frame_size= cfrm->size;

  803. 

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

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

  806.             }

  807. 

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

  809.             frame_4cc= AV_RL32("pfrm");

  810.         }else

  811.             return buf_size;

  812.     }else{

  813.         buf= buf + 12;

  814.         frame_size= buf_size - 12;

  815.     }

  816. 

  817.     temp= f->current_picture;

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

  819.     f->last_picture= temp;

  820. 

  821.     p= &f->current_picture;

  822.     avctx->coded_frame= p;

  823. 

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

  825. 

  826.     p->reference= 3;

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

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

  829.         return -1;

  830.     }

  831. 

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

  833.         p->pict_type= AV_PICTURE_TYPE_I;

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

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

  836.             return -1;

  837.         }

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

  839.         p->pict_type= AV_PICTURE_TYPE_I;

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

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

  842.             return -1;

  843.         }

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

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

  846.             f->last_picture.reference= 3;

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

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

  849.                 return -1;

  850.             }

  851.         }

  852. 

  853.         p->pict_type= AV_PICTURE_TYPE_P;

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

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

  856.             return -1;

  857.         }

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

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

  860.     }else{

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

  862.     }

  863. 

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

  865. 

  866.     *picture= *p;

  867.     *data_size = sizeof(AVPicture);

  868. 

  869.     emms_c();

  870. 

  871.     return buf_size;

  872. }

  873. 

  874. 

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

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

  877. 

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

  879. 

  880.     f->avctx= avctx;

  881. }

  882. 

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

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

  885. 

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

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

  888.         return 1;

  889.     }

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

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

  892.         return AVERROR_INVALIDDATA;

  893.     }

  894. 

  895.     avcodec_get_frame_defaults(&f->current_picture);

  896.     avcodec_get_frame_defaults(&f->last_picture);

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

  898.     common_init(avctx);

  899.     init_vlcs(f);

  900. 

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

  902.     else             avctx->pix_fmt= PIX_FMT_BGR555;

  903. 

  904.     return 0;

  905. }

  906. 

  907. 

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

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

  910.     int i;

  911. 

  912.     av_freep(&f->bitstream_buffer);

  913.     f->bitstream_buffer_size=0;

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

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

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

  917.     }

  918.     free_vlc(&f->pre_vlc);

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

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

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

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

  923. 

  924.     return 0;

  925. }

  926. 

  927. AVCodec ff_fourxm_decoder = {

  928.     .name           = "4xm",

  929.     .type           = AVMEDIA_TYPE_VIDEO,

  930.     .id             = CODEC_ID_4XM,

  931.     .priv_data_size = sizeof(FourXContext),

  932.     .init           = decode_init,

  933.     .close          = decode_end,

  934.     .decode         = decode_frame,

  935.     .capabilities   = CODEC_CAP_DR1,

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

  937. };