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.
39979 Commits
32 Branches
283MB
Tree: 593886b5dd
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '593886b5dd'
${ noResults }
FFmpeg/libavcodec/4xm.c

979 lines
32KB
Raw Blame History 

  1. /*

  2.  * 4XM codec

  3.  * Copyright (c) 2003 Michael Niedermayer

  4.  *

  5.  * This file is part of Libav.

  6.  *

  7.  * Libav 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.  * Libav 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 Libav; 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 <inttypes.h>

  28. 

  29. #include "libavutil/frame.h"

  30. #include "libavutil/imgutils.h"

  31. #include "libavutil/intreadwrite.h"

  32. #include "avcodec.h"

  33. #include "bytestream.h"

  34. #include "dsputil.h"

  35. #include "get_bits.h"

  36. #include "internal.h"

  37. 

  38. #define BLOCK_TYPE_VLC_BITS 5

  39. #define ACDC_VLC_BITS 9

  40. 

  41. #define CFRAME_BUFFER_COUNT 100

  42. 

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

  44.     {

  45.         {    // { 8, 4, 2 } x { 8, 4, 2}

  46.             { 0, 1 }, { 2, 2 }, { 6, 3 }, { 14, 4 }, { 30, 5 }, { 31, 5 }, { 0, 0 }

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

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

  49.         }, { // 1 x { 8, 4 }

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

  51.         }, { // 1 x 2, 2 x 1

  52.             { 0, 1 }, { 0, 0 }, { 0, 0 }, { 2, 2 }, { 6, 3 }, { 14, 4 }, { 15, 4 }

  53.         }

  54.     }, {

  55.         {   // { 8, 4, 2 } x { 8, 4, 2}

  56.             { 1, 2 }, { 4, 3 }, { 5, 3 }, { 0, 2 }, { 6, 3 }, { 7, 3 }, { 0, 0 }

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

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

  59.         }, {// 1 x { 8, 4 }

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

  61.         }, {// 1 x 2, 2 x 1

  62.             { 1, 2 }, { 0, 0 }, { 0, 0 }, { 0, 2 }, { 2, 2 }, { 6, 3 }, { 7, 3 }

  63.       }

  64.     }

  65. };

  66. 

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

  68.     { -1, 3, 1, 1 },

  69.     {  3, 0, 0, 0 },

  70.     {  2, 0, 0, 0 },

  71.     {  2, 0, 0, 0 },

  72. };

  73. 

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

  75.     {   0,   0 }, {   0,  -1 }, {  -1,   0 }, {   1,   0 }, {   0,   1 }, {  -1,  -1 }, {   1,  -1 }, {  -1,   1 },

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

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

  78.     {   2,   2 }, {   0,  -3 }, {  -3,   0 }, {   3,   0 }, {   0,   3 }, {  -1,  -3 }, {   1,  -3 }, {  -3,  -1 },

  79.     {   3,  -1 }, {  -3,   1 }, {   3,   1 }, {  -1,   3 }, {   1,   3 }, {  -2,  -3 }, {   2,  -3 }, {  -3,  -2 },

  80.     {   3,  -2 }, {  -3,   2 }, {   3,   2 }, {  -2,   3 }, {   2,   3 }, {   0,  -4 }, {  -4,   0 }, {   4,   0 },

  81.     {   0,   4 }, {  -1,  -4 }, {   1,  -4 }, {  -4,  -1 }, {   4,  -1 }, {   4,   1 }, {  -1,   4 }, {   1,   4 },

  82.     {  -3,  -3 }, {  -3,   3 }, {   3,   3 }, {  -2,  -4 }, {  -4,  -2 }, {   4,  -2 }, {  -4,   2 }, {  -2,   4 },

  83.     {   2,   4 }, {  -3,  -4 }, {   3,  -4 }, {   4,  -3 }, {  -5,   0 }, {  -4,   3 }, {  -3,   4 }, {   3,   4 },

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

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

  86.     {  -6,  -1 }, {  -6,   1 }, {   1,   6 }, {   2,  -6 }, {  -6,   2 }, {   2,   6 }, {  -5,  -4 }, {   5,   4 },

  87.     {   4,   5 }, {  -6,  -3 }, {   6,   3 }, {  -7,   0 }, {  -1,  -7 }, {   5,  -5 }, {  -7,   1 }, {  -1,   7 },

  88.     {   4,  -6 }, {   6,   4 }, {  -2,  -7 }, {  -7,   2 }, {  -3,  -7 }, {   7,  -3 }, {   3,   7 }, {   6,  -5 },

  89.     {   0,  -8 }, {  -1,  -8 }, {  -7,  -4 }, {  -8,   1 }, {   4,   7 }, {   2,  -8 }, {  -2,   8 }, {   6,   6 },

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

  91.     {  -7,   6 }, {  -5,  -8 }, {  -5,   8 }, {  -9,   3 }, {   9,  -4 }, {   7,  -7 }, {   8,  -6 }, {   6,   8 },

  92.     {  10,   1 }, { -10,   2 }, {   9,  -5 }, {  10,  -3 }, {  -8,  -7 }, { -10,  -4 }, {   6,  -9 }, { -11,   0 },

  93.     {  11,   1 }, { -11,  -2 }, {  -2,  11 }, {   7,  -9 }, {  -7,   9 }, {  10,   6 }, {  -4,  11 }, {   8,  -9 },

  94.     {   8,   9 }, {   5,  11 }, {   7, -10 }, {  12,  -3 }, {  11,   6 }, {  -9,  -9 }, {   8,  10 }, {   5,  12 },

  95.     { -11,   7 }, {  13,   2 }, {   6, -12 }, {  10,   9 }, { -11,   8 }, {  -7,  12 }, {   0,  14 }, {  14,  -2 },

  96.     {  -9,  11 }, {  -6,  13 }, { -14,  -4 }, {  -5, -14 }, {   5,  14 }, { -15,  -1 }, { -14,  -6 }, {   3, -15 },

  97.     {  11, -11 }, {  -7,  14 }, {  -5,  15 }, {   8, -14 }, {  15,   6 }, {   3,  16 }, {   7, -15 }, { -16,   5 },

  98.     {   0,  17 }, { -16,  -6 }, { -10,  14 }, { -16,   7 }, {  12,  13 }, { -16,   8 }, { -17,   6 }, { -18,   3 },

  99.     {  -7,  17 }, {  15,  11 }, {  16,  10 }, {   2, -19 }, {   3, -19 }, { -11, -16 }, { -18,   8 }, { -19,  -6 },

  100.     {   2, -20 }, { -17, -11 }, { -10, -18 }, {   8,  19 }, { -21,  -1 }, { -20,   7 }, {  -4,  21 }, {  21,   5 },

  101.     {  15,  16 }, {   2, -22 }, { -10, -20 }, { -22,   5 }, {  20, -11 }, {  -7, -22 }, { -12,  20 }, {  23,  -5 },

  102.     {  13, -20 }, {  24,  -2 }, { -15,  19 }, { -11,  22 }, {  16,  19 }, {  23, -10 }, { -18, -18 }, {  -9, -24 },

  103.     {  24, -10 }, {  -3,  26 }, { -23,  13 }, { -18, -20 }, {  17,  21 }, {  -4,  27 }, {  27,   6 }, {   1, -28 },

  104.     { -11,  26 }, { -17, -23 }, {   7,  28 }, {  11, -27 }, {  29,   5 }, { -23, -19 }, { -28, -11 }, { -21,  22 },

  105.     { -30,   7 }, { -17,  26 }, { -27,  16 }, {  13,  29 }, {  19, -26 }, {  10, -31 }, { -14, -30 }, {  20, -27 },

  106.     { -29,  18 }, { -16, -31 }, { -28, -22 }, {  21, -30 }, { -25,  28 }, {  26, -29 }, {  25, -32 }, { -32, -32 }

  107. };

  108. 

  109. /* This is simply the scaled down elementwise product of the standard JPEG

  110.  * quantizer table and the AAN premul table. */

  111. static const uint8_t dequant_table[64] = {

  112.     16, 15, 13, 19, 24, 31, 28, 17,

  113.     17, 23, 25, 31, 36, 63, 45, 21,

  114.     18, 24, 27, 37, 52, 59, 49, 20,

  115.     16, 28, 34, 40, 60, 80, 51, 20,

  116.     18, 31, 48, 66, 68, 86, 56, 21,

  117.     19, 38, 56, 59, 64, 64, 48, 20,

  118.     27, 48, 55, 55, 56, 51, 35, 15,

  119.     20, 35, 34, 32, 31, 22, 15,  8,

  120. };

  121. 

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

  123. 

  124. 

  125. typedef struct CFrameBuffer {

  126.     unsigned int allocated_size;

  127.     unsigned int size;

  128.     int id;

  129.     uint8_t *data;

  130. } CFrameBuffer;

  131. 

  132. typedef struct FourXContext {

  133.     AVCodecContext *avctx;

  134.     DSPContext dsp;

  135.     uint16_t *frame_buffer;

  136.     uint16_t *last_frame_buffer;

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

  138.     GetBitContext gb;

  139.     GetByteContext g;

  140.     GetByteContext g2;

  141.     int mv[256];

  142.     VLC pre_vlc;

  143.     int last_dc;

  144.     DECLARE_ALIGNED(16, int16_t, block)[6][64];

  145.     void *bitstream_buffer;

  146.     unsigned int bitstream_buffer_size;

  147.     int version;

  148.     CFrameBuffer cfrm[CFRAME_BUFFER_COUNT];

  149. } FourXContext;

  150. 

  151. 

  152. #define FIX_1_082392200  70936

  153. #define FIX_1_414213562  92682

  154. #define FIX_1_847759065 121095

  155. #define FIX_2_613125930 171254

  156. 

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

  158. 

  159. static void idct(int16_t block[64])

  160. {

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

  162.     int tmp10, tmp11, tmp12, tmp13;

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

  164.     int i;

  165.     int temp[64];

  166. 

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

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

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

  170. 

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

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

  173. 

  174.         tmp0 = tmp10 + tmp13;

  175.         tmp3 = tmp10 - tmp13;

  176.         tmp1 = tmp11 + tmp12;

  177.         tmp2 = tmp11 - tmp12;

  178. 

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

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

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

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

  183. 

  184.         tmp7  =          z11 + z13;

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

  186. 

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

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

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

  190. 

  191.         tmp6 = tmp12 - tmp7;

  192.         tmp5 = tmp11 - tmp6;

  193.         tmp4 = tmp10 + tmp5;

  194. 

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

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

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

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

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

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

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

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

  203.     }

  204. 

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

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

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

  208. 

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

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

  211. 

  212.         tmp0 = tmp10 + tmp13;

  213.         tmp3 = tmp10 - tmp13;

  214.         tmp1 = tmp11 + tmp12;

  215.         tmp2 = tmp11 - tmp12;

  216. 

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

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

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

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

  221. 

  222.         tmp7  = z11 + z13;

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

  224. 

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

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

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

  228. 

  229.         tmp6 = tmp12 - tmp7;

  230.         tmp5 = tmp11 - tmp6;

  231.         tmp4 = tmp10 + tmp5;

  232. 

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

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

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

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

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

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

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

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

  241.     }

  242. }

  243. 

  244. static av_cold void init_vlcs(FourXContext *f)

  245. {

  246.     static VLC_TYPE table[2][4][32][2];

  247.     int i, j;

  248. 

  249.     for (i = 0; i < 2; i++) {

  250.         for (j = 0; j < 4; j++) {

  251.             block_type_vlc[i][j].table           = table[i][j];

  252.             block_type_vlc[i][j].table_allocated = 32;

  253.             init_vlc(&block_type_vlc[i][j], BLOCK_TYPE_VLC_BITS, 7,

  254.                      &block_type_tab[i][j][0][1], 2, 1,

  255.                      &block_type_tab[i][j][0][0], 2, 1,

  256.                      INIT_VLC_USE_NEW_STATIC);

  257.         }

  258.     }

  259. }

  260. 

  261. static void init_mv(FourXContext *f, int linesize)

  262. {

  263.     int i;

  264. 

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

  266.         if (f->version > 1)

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

  268.         else

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

  270.     }

  271. }

  272. 

  273. #if HAVE_BIGENDIAN

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

  275.     {                                                   \

  276.         unsigned tmpval = AV_RN32(src);                 \

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

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

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

  280.         AV_WN32A(dst, tmpval);                          \

  281.     }

  282. #else

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

  284.     {                                                    \

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

  286.         AV_WN32A(dst, tmpval);                           \

  287.     }

  288. #endif

  289. 

  290. static inline void mcdc(uint16_t *dst, uint16_t *src, int log2w,

  291.                         int h, int stride, int scale, unsigned dc)

  292. {

  293.     int i;

  294.     dc *= 0x10001;

  295. 

  296.     switch (log2w) {

  297.     case 0:

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

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

  300.             if (scale)

  301.                 src += stride;

  302.             dst += stride;

  303.         }

  304.         break;

  305.     case 1:

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

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

  308.             if (scale)

  309.                 src += stride;

  310.             dst += stride;

  311.         }

  312.         break;

  313.     case 2:

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

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

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

  317.             if (scale)

  318.                 src += stride;

  319.             dst += stride;

  320.         }

  321.         break;

  322.     case 3:

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

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

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

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

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

  328.             if (scale)

  329.                 src += stride;

  330.             dst += stride;

  331.         }

  332.         break;

  333.     default:

  334.         break;

  335.     }

  336. }

  337. 

  338. static int decode_p_block(FourXContext *f, uint16_t *dst, uint16_t *src,

  339.                           int log2w, int log2h, int stride)

  340. {

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

  342.     const int h     = 1 << log2h;

  343.     int code        = get_vlc2(&f->gb,

  344.                                block_type_vlc[1 - (f->version > 1)][index].table,

  345.                                BLOCK_TYPE_VLC_BITS, 1);

  346.     uint16_t *start = f->last_frame_buffer;

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

  348.     int ret;

  349.     int scale   = 1;

  350.     unsigned dc = 0;

  351. 

  352.     if (code < 0 || code > 6 || log2w < 0)

  353.         return AVERROR_INVALIDDATA;

  354. 

  355.     if (code == 1) {

  356.         log2h--;

  357.         if ((ret = decode_p_block(f, dst, src, log2w, log2h, stride)) < 0)

  358.             return ret;

  359.         return decode_p_block(f, dst + (stride << log2h),

  360.                               src + (stride << log2h),

  361.                               log2w, log2h, stride);

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

  363.         log2w--;

  364.         if ((ret = decode_p_block(f, dst , src, log2w, log2h, stride)) < 0)

  365.             return ret;

  366.         return decode_p_block(f, dst + (1 << log2w),

  367.                               src + (1 << log2w),

  368.                               log2w, log2h, stride);

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

  370.         if (log2w) {

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

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

  373.         } else {

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

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

  376.         }

  377.         return 0;

  378.     }

  379. 

  380.     if (code == 0) {

  381.         src  += f->mv[bytestream2_get_byte(&f->g)];

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

  383.         return 0;

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

  385.         src  += f->mv[bytestream2_get_byte(&f->g)];

  386.         dc    = bytestream2_get_le16(&f->g2);

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

  388.         scale = 0;

  389.         dc    = bytestream2_get_le16(&f->g2);

  390.     }

  391. 

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

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

  394.         return AVERROR_INVALIDDATA;

  395.     }

  396. 

  397.     mcdc(dst, src, log2w, h, stride, scale, dc);

  398. 

  399.     return 0;

  400. }

  401. 

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

  403. {

  404.     int x, y;

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

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

  407.     uint16_t *dst    = f->frame_buffer;

  408.     uint16_t *src;

  409.     unsigned int bitstream_size, bytestream_size, wordstream_size, extra,

  410.                  bytestream_offset, wordstream_offset;

  411.     int ret;

  412. 

  413.     src = f->last_frame_buffer;

  414. 

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

  416.         if (length < 20)

  417.             return AVERROR_INVALIDDATA;

  418.         extra           = 20;

  419.         bitstream_size  = AV_RL32(buf + 8);

  420.         wordstream_size = AV_RL32(buf + 12);

  421.         bytestream_size = AV_RL32(buf + 16);

  422.     } else {

  423.         extra           = 0;

  424.         bitstream_size  = AV_RL16(buf - 4);

  425.         wordstream_size = AV_RL16(buf - 2);

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

  427.     }

  428. 

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

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

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

  432.         || wordstream_size > (1 << 26)) {

  433.         av_log(f->avctx, AV_LOG_ERROR, "lengths %d %d %d %d\n",

  434.                bitstream_size, bytestream_size, wordstream_size,

  435.                bitstream_size + bytestream_size + wordstream_size - length);

  436.         return AVERROR_INVALIDDATA;

  437.     }

  438. 

  439.     av_fast_malloc(&f->bitstream_buffer, &f->bitstream_buffer_size,

  440.                    bitstream_size + FF_INPUT_BUFFER_PADDING_SIZE);

  441.     if (!f->bitstream_buffer)

  442.         return AVERROR(ENOMEM);

  443.     f->dsp.bswap_buf(f->bitstream_buffer, (const uint32_t*)(buf + extra),

  444.                      bitstream_size / 4);

  445.     memset((uint8_t*)f->bitstream_buffer + bitstream_size,

  446.            0, FF_INPUT_BUFFER_PADDING_SIZE);

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

  448. 

  449.     wordstream_offset = extra + bitstream_size;

  450.     bytestream_offset = extra + bitstream_size + wordstream_size;

  451.     bytestream2_init(&f->g2, buf + wordstream_offset,

  452.                      length - wordstream_offset);

  453.     bytestream2_init(&f->g, buf + bytestream_offset,

  454.                      length - bytestream_offset);

  455. 

  456.     init_mv(f, width * 2);

  457. 

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

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

  460.             if ((ret = decode_p_block(f, dst + x, src + x, 3, 3, width)) < 0)

  461.                 return ret;

  462.         src += 8 * width;

  463.         dst += 8 * width;

  464.     }

  465. 

  466.     return 0;

  467. }

  468. 

  469. /**

  470.  * decode block and dequantize.

  471.  * Note this is almost identical to MJPEG.

  472.  */

  473. static int decode_i_block(FourXContext *f, int16_t *block)

  474. {

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

  476. 

  477.     /* DC coef */

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

  479.     if (val >> 4)

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

  481. 

  482.     if (val)

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

  484. 

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

  486.     f->last_dc = block[0] = val;

  487.     /* AC coefs */

  488.     i = 1;

  489.     for (;;) {

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

  491. 

  492.         /* EOB */

  493.         if (code == 0)

  494.             break;

  495.         if (code == 0xf0) {

  496.             i += 16;

  497.         } else {

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

  499.             i    += code >> 4;

  500.             if (i >= 64) {

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

  502.                 return 0;

  503.             }

  504. 

  505.             j = ff_zigzag_direct[i];

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

  507.             i++;

  508.             if (i >= 64)

  509.                 break;

  510.         }

  511.     }

  512. 

  513.     return 0;

  514. }

  515. 

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

  517. {

  518.     int16_t (*block)[64] = f->block;

  519.     int stride           = f->avctx->width;

  520.     int i;

  521.     uint16_t *dst = f->frame_buffer + y * stride + x;

  522. 

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

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

  525.         idct(block[i]);

  526.     }

  527. 

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

  529.         for (i = 4; i < 6; i++)

  530.             idct(block[i]);

  531.     }

  532. 

  533.     /* Note transform is:

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

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

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

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

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

  539.             int16_t *temp = block[(x >> 2) + 2 * (y >> 2)] +

  540.                             2 * (x & 3) + 2 * 8 * (y & 3); // FIXME optimize

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

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

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

  544.             int y;

  545. 

  546.             cb += cb;

  547. 

  548.             y               = temp[0];

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

  550.             y               = temp[1];

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

  552.             y               = temp[8];

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

  554.             y               = temp[9];

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

  556.             dst            += 2;

  557.         }

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

  559.     }

  560. }

  561. 

  562. static int decode_i_mb(FourXContext *f)

  563. {

  564.     int ret;

  565.     int i;

  566. 

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

  568. 

  569.     for (i = 0; i < 6; i++)

  570.         if ((ret = decode_i_block(f, f->block[i])) < 0)

  571.             return ret;

  572. 

  573.     return 0;

  574. }

  575. 

  576. static const uint8_t *read_huffman_tables(FourXContext *f,

  577.                                           const uint8_t * const buf,

  578.                                           int len)

  579. {

  580.     int frequency[512] = { 0 };

  581.     uint8_t flag[512];

  582.     int up[512];

  583.     uint8_t len_tab[257];

  584.     int bits_tab[257];

  585.     int start, end;

  586.     const uint8_t *ptr = buf;

  587.     int j;

  588. 

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

  590. 

  591.     start = *ptr++;

  592.     end   = *ptr++;

  593.     for (;;) {

  594.         int i;

  595. 

  596.         len -= end - start + 1;

  597. 

  598.         if (end < start || len < 0)

  599.             return NULL;

  600. 

  601.         for (i = start; i <= end; i++)

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

  603.         start = *ptr++;

  604.         if (start == 0)

  605.             break;

  606. 

  607.         if (--len < 0)

  608.             return NULL;

  609. 

  610.         end = *ptr++;

  611.     }

  612.     frequency[256] = 1;

  613. 

  614.     while ((ptr - buf) & 3)

  615.         ptr++; // 4byte align

  616. 

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

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

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

  620.         int i;

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

  622.             if (frequency[i] == 0)

  623.                 continue;

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

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

  626.                     min_freq[1] = min_freq[0];

  627.                     smallest[1] = smallest[0];

  628.                     min_freq[0] = frequency[i];

  629.                     smallest[0] = i;

  630.                 } else {

  631.                     min_freq[1] = frequency[i];

  632.                     smallest[1] = i;

  633.                 }

  634.             }

  635.         }

  636.         if (min_freq[1] == 256 * 256)

  637.             break;

  638. 

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

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

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

  642.         up[smallest[0]]        =

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

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

  645.     }

  646. 

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

  648.         int node, len = 0, bits = 0;

  649. 

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

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

  652.             len++;

  653.             if (len > 31)

  654.                 // can this happen at all ?

  655.                 av_log(f->avctx, AV_LOG_ERROR,

  656.                        "vlc length overflow\n");

  657.         }

  658. 

  659.         bits_tab[j] = bits;

  660.         len_tab[j]  = len;

  661.     }

  662. 

  663.     if (init_vlc(&f->pre_vlc, ACDC_VLC_BITS, 257, len_tab, 1, 1,

  664.                  bits_tab, 4, 4, 0))

  665.         return NULL;

  666. 

  667.     return ptr;

  668. }

  669. 

  670. static int mix(int c0, int c1)

  671. {

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

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

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

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

  676. }

  677. 

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

  679. {

  680.     int x, y, x2, y2;

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

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

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

  684.     uint16_t *dst    = f->frame_buffer;

  685.     GetByteContext g3;

  686. 

  687.     if (length < mbs * 8) {

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

  689.         return AVERROR_INVALIDDATA;

  690.     }

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

  692. 

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

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

  695.             unsigned int color[4] = { 0 }, bits;

  696.             // warning following is purely guessed ...

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

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

  699. 

  700.             if (color[0] & 0x8000)

  701.                 av_log(f->avctx, AV_LOG_ERROR, "unk bit 1\n");

  702.             if (color[1] & 0x8000)

  703.                 av_log(f->avctx, AV_LOG_ERROR, "unk bit 2\n");

  704. 

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

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

  707. 

  708.             bits = bytestream2_get_le32u(&g3);

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

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

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

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

  713.                 }

  714.             }

  715.             dst += 16;

  716.         }

  717.         dst += 16 * width - x;

  718.     }

  719. 

  720.     return 0;

  721. }

  722. 

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

  724. {

  725.     int x, y, ret;

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

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

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

  729.     int token_count av_unused;

  730.     unsigned int prestream_size;

  731.     const uint8_t *prestream;

  732. 

  733.     if (bitstream_size > (1 << 26))

  734.         return AVERROR_INVALIDDATA;

  735. 

  736.     if (length < bitstream_size + 12) {

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

  738.         return AVERROR_INVALIDDATA;

  739.     }

  740. 

  741.     token_count    =     AV_RL32(buf + bitstream_size + 8);

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

  743.     prestream      =             buf + bitstream_size + 12;

  744. 

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

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

  747.         av_log(f->avctx, AV_LOG_ERROR, "size mismatch %d %d %d\n",

  748.                prestream_size, bitstream_size, length);

  749.         return AVERROR_INVALIDDATA;

  750.     }

  751. 

  752.     prestream = read_huffman_tables(f, prestream, prestream_size);

  753.     if (!prestream) {

  754.         av_log(f->avctx, AV_LOG_ERROR, "Error reading Huffman tables.\n");

  755.         return AVERROR_INVALIDDATA;

  756.     }

  757. 

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

  759. 

  760.     prestream_size = length + buf - prestream;

  761. 

  762.     av_fast_malloc(&f->bitstream_buffer, &f->bitstream_buffer_size,

  763.                    prestream_size + FF_INPUT_BUFFER_PADDING_SIZE);

  764.     if (!f->bitstream_buffer)

  765.         return AVERROR(ENOMEM);

  766.     f->dsp.bswap_buf(f->bitstream_buffer, (const uint32_t*)prestream,

  767.                      prestream_size / 4);

  768.     memset((uint8_t*)f->bitstream_buffer + prestream_size,

  769.            0, FF_INPUT_BUFFER_PADDING_SIZE);

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

  771. 

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

  773. 

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

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

  776.             if ((ret = decode_i_mb(f)) < 0)

  777.                 return ret;

  778. 

  779.             idct_put(f, x, y);

  780.         }

  781.     }

  782. 

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

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

  785. 

  786.     return 0;

  787. }

  788. 

  789. static int decode_frame(AVCodecContext *avctx, void *data,

  790.                         int *got_frame, AVPacket *avpkt)

  791. {

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

  793.     int buf_size          = avpkt->size;

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

  795.     AVFrame *picture      = data;

  796.     int i, frame_4cc, frame_size, ret;

  797. 

  798.     if (buf_size < 20)

  799.         return AVERROR_INVALIDDATA;

  800. 

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

  802.         av_log(avctx, AV_LOG_ERROR,

  803.                "Dimensions non-multiple of 16 are invalid.\n");

  804.         return AVERROR_INVALIDDATA;

  805.     }

  806. 

  807.     if (buf_size < AV_RL32(buf + 4) + 8) {

  808.         av_log(f->avctx, AV_LOG_ERROR, "size mismatch %d %"PRIu32"\n",

  809.                buf_size, AV_RL32(buf + 4));

  810.         return AVERROR_INVALIDDATA;

  811.     }

  812. 

  813.     frame_4cc = AV_RL32(buf);

  814. 

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

  816.         int free_index       = -1;

  817.         int id, whole_size;

  818.         const int data_size  = buf_size - 20;

  819.         CFrameBuffer *cfrm;

  820. 

  821.         if (data_size < 0)

  822.             return AVERROR_INVALIDDATA;

  823. 

  824.         id         = AV_RL32(buf + 12);

  825.         whole_size = AV_RL32(buf + 16);

  826. 

  827.         for (i = 0; i < CFRAME_BUFFER_COUNT; i++)

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

  829.                 av_log(f->avctx, AV_LOG_ERROR, "lost c frame %d\n",

  830.                        f->cfrm[i].id);

  831. 

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

  833.             if (f->cfrm[i].id == id)

  834.                 break;

  835.             if (f->cfrm[i].size == 0)

  836.                 free_index = i;

  837.         }

  838. 

  839.         if (i >= CFRAME_BUFFER_COUNT) {

  840.             i             = free_index;

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

  842.         }

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

  844. 

  845.         cfrm->data = av_fast_realloc(cfrm->data, &cfrm->allocated_size,

  846.                                      cfrm->size + data_size + FF_INPUT_BUFFER_PADDING_SIZE);

  847.         // explicit check needed as memcpy below might not catch a NULL

  848.         if (!cfrm->data) {

  849.             av_log(f->avctx, AV_LOG_ERROR, "realloc failure");

  850.             return AVERROR(ENOMEM);

  851.         }

  852. 

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

  854.         cfrm->size += data_size;

  855. 

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

  857.             buf        = cfrm->data;

  858.             frame_size = cfrm->size;

  859. 

  860.             if (id != avctx->frame_number)

  861.                 av_log(f->avctx, AV_LOG_ERROR, "cframe id mismatch %d %d\n",

  862.                        id, avctx->frame_number);

  863. 

  864.             if (f->version <= 1)

  865.                 return AVERROR_INVALIDDATA;

  866. 

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

  868.             frame_4cc  = AV_RL32("pfrm");

  869.         } else

  870.             return buf_size;

  871.     } else {

  872.         buf        = buf      + 12;

  873.         frame_size = buf_size - 12;

  874.     }

  875. 

  876. 

  877.     if ((ret = ff_get_buffer(avctx, picture, 0)) < 0) {

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

  879.         return ret;

  880.     }

  881. 

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

  883.         picture->pict_type = AV_PICTURE_TYPE_I;

  884.         if ((ret = decode_i2_frame(f, buf - 4, frame_size + 4)) < 0)

  885.             return ret;

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

  887.         picture->pict_type = AV_PICTURE_TYPE_I;

  888.         if ((ret = decode_i_frame(f, buf, frame_size)) < 0)

  889.             return ret;

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

  891.         picture->pict_type = AV_PICTURE_TYPE_P;

  892.         if ((ret = decode_p_frame(f, buf, frame_size)) < 0)

  893.             return ret;

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

  895.         av_log(avctx, AV_LOG_ERROR, "ignoring snd_ chunk length:%d\n",

  896.                buf_size);

  897.     } else {

  898.         av_log(avctx, AV_LOG_ERROR, "ignoring unknown chunk length:%d\n",

  899.                buf_size);

  900.     }

  901. 

  902.     picture->key_frame = picture->pict_type == AV_PICTURE_TYPE_I;

  903. 

  904.     av_image_copy_plane(picture->data[0], picture->linesize[0],

  905.                         (const uint8_t*)f->frame_buffer,  avctx->width * 2,

  906.                         avctx->width * 2, avctx->height);

  907.     FFSWAP(uint16_t *, f->frame_buffer, f->last_frame_buffer);

  908. 

  909.     *got_frame = 1;

  910. 

  911.     emms_c();

  912. 

  913.     return buf_size;

  914. }

  915. 

  916. static av_cold int decode_end(AVCodecContext *avctx)

  917. {

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

  919.     int i;

  920. 

  921.     av_freep(&f->frame_buffer);

  922.     av_freep(&f->last_frame_buffer);

  923.     av_freep(&f->bitstream_buffer);

  924.     f->bitstream_buffer_size = 0;

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

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

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

  928.     }

  929.     ff_free_vlc(&f->pre_vlc);

  930. 

  931.     return 0;

  932. }

  933. 

  934. static av_cold int decode_init(AVCodecContext *avctx)

  935. {

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

  937.     int ret;

  938. 

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

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

  941.         return AVERROR_INVALIDDATA;

  942.     }

  943. 

  944.     ret = av_image_check_size(avctx->width, avctx->height, 0, avctx);

  945.     if (ret < 0)

  946.         return ret;

  947. 

  948.     f->frame_buffer      = av_mallocz(avctx->width * avctx->height * 2);

  949.     f->last_frame_buffer = av_mallocz(avctx->width * avctx->height * 2);

  950.     if (!f->frame_buffer || !f->last_frame_buffer) {

  951.         decode_end(avctx);

  952.         return AVERROR(ENOMEM);

  953.     }

  954. 

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

  956.     ff_dsputil_init(&f->dsp, avctx);

  957.     f->avctx = avctx;

  958.     init_vlcs(f);

  959. 

  960.     if (f->version > 2)

  961.         avctx->pix_fmt = AV_PIX_FMT_RGB565;

  962.     else

  963.         avctx->pix_fmt = AV_PIX_FMT_BGR555;

  964. 

  965.     return 0;

  966. }

  967. 

  968. AVCodec ff_fourxm_decoder = {

  969.     .name           = "4xm",

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

  971.     .type           = AVMEDIA_TYPE_VIDEO,

  972.     .id             = AV_CODEC_ID_4XM,

  973.     .priv_data_size = sizeof(FourXContext),

  974.     .init           = decode_init,

  975.     .close          = decode_end,

  976.     .decode         = decode_frame,

  977.     .capabilities   = CODEC_CAP_DR1,

  978. };