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.
43415 Commits
32 Branches
824MB
Tree: 82b7525173
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '82b7525173'
${ noResults }
FFmpeg/libavcodec/4xm.c

986 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 "blockdsp.h"

  34. #include "bswapdsp.h"

  35. #include "bytestream.h"

  36. #include "get_bits.h"

  37. #include "internal.h"

  38. 

  39. #define BLOCK_TYPE_VLC_BITS 5

  40. #define ACDC_VLC_BITS 9

  41. 

  42. #define CFRAME_BUFFER_COUNT 100

  43. 

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

  45.     {

  46.         {    // { 8, 4, 2 } x { 8, 4, 2}

  47.             { 0, 1 }, { 2, 2 }, { 6, 3 }, { 14, 4 }, { 30, 5 }, { 31, 5 }, { 0, 0 }

  48.         }, { // { 8, 4 } x 1

  49.             { 0, 1 }, { 0, 0 }, { 2, 2 }, { 6, 3 }, { 14, 4 }, { 15, 4 }, { 0, 0 }

  50.         }, { // 1 x { 8, 4 }

  51.             { 0, 1 }, { 2, 2 }, { 0, 0 }, { 6, 3 }, { 14, 4 }, { 15, 4 }, { 0, 0 }

  52.         }, { // 1 x 2, 2 x 1

  53.             { 0, 1 }, { 0, 0 }, { 0, 0 }, { 2, 2 }, { 6, 3 }, { 14, 4 }, { 15, 4 }

  54.         }

  55.     }, {

  56.         {   // { 8, 4, 2 } x { 8, 4, 2}

  57.             { 1, 2 }, { 4, 3 }, { 5, 3 }, { 0, 2 }, { 6, 3 }, { 7, 3 }, { 0, 0 }

  58.         }, {// { 8, 4 } x 1

  59.             { 1, 2 }, { 0, 0 }, { 2, 2 }, { 0, 2 }, { 6, 3 }, { 7, 3 }, { 0, 0 }

  60.         }, {// 1 x { 8, 4 }

  61.             { 1, 2 }, { 2, 2 }, { 0, 0 }, { 0, 2 }, { 6, 3 }, { 7, 3 }, { 0, 0 }

  62.         }, {// 1 x 2, 2 x 1

  63.             { 1, 2 }, { 0, 0 }, { 0, 0 }, { 0, 2 }, { 2, 2 }, { 6, 3 }, { 7, 3 }

  64.       }

  65.     }

  66. };

  67. 

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

  69.     { -1, 3, 1, 1 },

  70.     {  3, 0, 0, 0 },

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

  72.     {  2, 0, 0, 0 },

  73. };

  74. 

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

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

  77.     {   1,   1 }, {   0,  -2 }, {  -2,   0 }, {   2,   0 }, {   0,   2 }, {  -1,  -2 }, {   1,  -2 }, {  -2,  -1 },

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

  79.     {   2,   2 }, {   0,  -3 }, {  -3,   0 }, {   3,   0 }, {   0,   3 }, {  -1,  -3 }, {   1,  -3 }, {  -3,  -1 },

  80.     {   3,  -1 }, {  -3,   1 }, {   3,   1 }, {  -1,   3 }, {   1,   3 }, {  -2,  -3 }, {   2,  -3 }, {  -3,  -2 },

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

  82.     {   0,   4 }, {  -1,  -4 }, {   1,  -4 }, {  -4,  -1 }, {   4,  -1 }, {   4,   1 }, {  -1,   4 }, {   1,   4 },

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

  84.     {   2,   4 }, {  -3,  -4 }, {   3,  -4 }, {   4,  -3 }, {  -5,   0 }, {  -4,   3 }, {  -3,   4 }, {   3,   4 },

  85.     {  -1,  -5 }, {  -5,  -1 }, {  -5,   1 }, {  -1,   5 }, {  -2,  -5 }, {   2,  -5 }, {   5,  -2 }, {   5,   2 },

  86.     {  -4,  -4 }, {  -4,   4 }, {  -3,  -5 }, {  -5,  -3 }, {  -5,   3 }, {   3,   5 }, {  -6,   0 }, {   0,   6 },

  87.     {  -6,  -1 }, {  -6,   1 }, {   1,   6 }, {   2,  -6 }, {  -6,   2 }, {   2,   6 }, {  -5,  -4 }, {   5,   4 },

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

  89.     {   4,  -6 }, {   6,   4 }, {  -2,  -7 }, {  -7,   2 }, {  -3,  -7 }, {   7,  -3 }, {   3,   7 }, {   6,  -5 },

  90.     {   0,  -8 }, {  -1,  -8 }, {  -7,  -4 }, {  -8,   1 }, {   4,   7 }, {   2,  -8 }, {  -2,   8 }, {   6,   6 },

  91.     {  -8,   3 }, {   5,  -7 }, {  -5,   7 }, {   8,  -4 }, {   0,  -9 }, {  -9,  -1 }, {   1,   9 }, {   7,  -6 },

  92.     {  -7,   6 }, {  -5,  -8 }, {  -5,   8 }, {  -9,   3 }, {   9,  -4 }, {   7,  -7 }, {   8,  -6 }, {   6,   8 },

  93.     {  10,   1 }, { -10,   2 }, {   9,  -5 }, {  10,  -3 }, {  -8,  -7 }, { -10,  -4 }, {   6,  -9 }, { -11,   0 },

  94.     {  11,   1 }, { -11,  -2 }, {  -2,  11 }, {   7,  -9 }, {  -7,   9 }, {  10,   6 }, {  -4,  11 }, {   8,  -9 },

  95.     {   8,   9 }, {   5,  11 }, {   7, -10 }, {  12,  -3 }, {  11,   6 }, {  -9,  -9 }, {   8,  10 }, {   5,  12 },

  96.     { -11,   7 }, {  13,   2 }, {   6, -12 }, {  10,   9 }, { -11,   8 }, {  -7,  12 }, {   0,  14 }, {  14,  -2 },

  97.     {  -9,  11 }, {  -6,  13 }, { -14,  -4 }, {  -5, -14 }, {   5,  14 }, { -15,  -1 }, { -14,  -6 }, {   3, -15 },

  98.     {  11, -11 }, {  -7,  14 }, {  -5,  15 }, {   8, -14 }, {  15,   6 }, {   3,  16 }, {   7, -15 }, { -16,   5 },

  99.     {   0,  17 }, { -16,  -6 }, { -10,  14 }, { -16,   7 }, {  12,  13 }, { -16,   8 }, { -17,   6 }, { -18,   3 },

  100.     {  -7,  17 }, {  15,  11 }, {  16,  10 }, {   2, -19 }, {   3, -19 }, { -11, -16 }, { -18,   8 }, { -19,  -6 },

  101.     {   2, -20 }, { -17, -11 }, { -10, -18 }, {   8,  19 }, { -21,  -1 }, { -20,   7 }, {  -4,  21 }, {  21,   5 },

  102.     {  15,  16 }, {   2, -22 }, { -10, -20 }, { -22,   5 }, {  20, -11 }, {  -7, -22 }, { -12,  20 }, {  23,  -5 },

  103.     {  13, -20 }, {  24,  -2 }, { -15,  19 }, { -11,  22 }, {  16,  19 }, {  23, -10 }, { -18, -18 }, {  -9, -24 },

  104.     {  24, -10 }, {  -3,  26 }, { -23,  13 }, { -18, -20 }, {  17,  21 }, {  -4,  27 }, {  27,   6 }, {   1, -28 },

  105.     { -11,  26 }, { -17, -23 }, {   7,  28 }, {  11, -27 }, {  29,   5 }, { -23, -19 }, { -28, -11 }, { -21,  22 },

  106.     { -30,   7 }, { -17,  26 }, { -27,  16 }, {  13,  29 }, {  19, -26 }, {  10, -31 }, { -14, -30 }, {  20, -27 },

  107.     { -29,  18 }, { -16, -31 }, { -28, -22 }, {  21, -30 }, { -25,  28 }, {  26, -29 }, {  25, -32 }, { -32, -32 }

  108. };

  109. 

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

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

  112. static const uint8_t dequant_table[64] = {

  113.     16, 15, 13, 19, 24, 31, 28, 17,

  114.     17, 23, 25, 31, 36, 63, 45, 21,

  115.     18, 24, 27, 37, 52, 59, 49, 20,

  116.     16, 28, 34, 40, 60, 80, 51, 20,

  117.     18, 31, 48, 66, 68, 86, 56, 21,

  118.     19, 38, 56, 59, 64, 64, 48, 20,

  119.     27, 48, 55, 55, 56, 51, 35, 15,

  120.     20, 35, 34, 32, 31, 22, 15,  8,

  121. };

  122. 

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

  124. 

  125. 

  126. typedef struct CFrameBuffer {

  127.     unsigned int allocated_size;

  128.     unsigned int size;

  129.     int id;

  130.     uint8_t *data;

  131. } CFrameBuffer;

  132. 

  133. typedef struct FourXContext {

  134.     AVCodecContext *avctx;

  135.     BlockDSPContext bdsp;

  136.     BswapDSPContext bbdsp;

  137.     uint16_t *frame_buffer;

  138.     uint16_t *last_frame_buffer;

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

  140.     GetBitContext gb;

  141.     GetByteContext g;

  142.     GetByteContext g2;

  143.     int mv[256];

  144.     VLC pre_vlc;

  145.     int last_dc;

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

  147.     void *bitstream_buffer;

  148.     unsigned int bitstream_buffer_size;

  149.     int version;

  150.     CFrameBuffer cfrm[CFRAME_BUFFER_COUNT];

  151. } FourXContext;

  152. 

  153. 

  154. #define FIX_1_082392200  70936

  155. #define FIX_1_414213562  92682

  156. #define FIX_1_847759065 121095

  157. #define FIX_2_613125930 171254

  158. 

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

  160. 

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

  162. {

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

  164.     int tmp10, tmp11, tmp12, tmp13;

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

  166.     int i;

  167.     int temp[64];

  168. 

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

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

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

  172. 

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

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

  175. 

  176.         tmp0 = tmp10 + tmp13;

  177.         tmp3 = tmp10 - tmp13;

  178.         tmp1 = tmp11 + tmp12;

  179.         tmp2 = tmp11 - tmp12;

  180. 

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

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

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

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

  185. 

  186.         tmp7  =          z11 + z13;

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

  188. 

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

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

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

  192. 

  193.         tmp6 = tmp12 - tmp7;

  194.         tmp5 = tmp11 - tmp6;

  195.         tmp4 = tmp10 + tmp5;

  196. 

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

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

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

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

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

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

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

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

  205.     }

  206. 

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

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

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

  210. 

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

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

  213. 

  214.         tmp0 = tmp10 + tmp13;

  215.         tmp3 = tmp10 - tmp13;

  216.         tmp1 = tmp11 + tmp12;

  217.         tmp2 = tmp11 - tmp12;

  218. 

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

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

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

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

  223. 

  224.         tmp7  = z11 + z13;

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

  226. 

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

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

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

  230. 

  231.         tmp6 = tmp12 - tmp7;

  232.         tmp5 = tmp11 - tmp6;

  233.         tmp4 = tmp10 + tmp5;

  234. 

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

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

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

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

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

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

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

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

  243.     }

  244. }

  245. 

  246. static av_cold void init_vlcs(FourXContext *f)

  247. {

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

  249.     int i, j;

  250. 

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

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

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

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

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

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

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

  258.                      INIT_VLC_USE_NEW_STATIC);

  259.         }

  260.     }

  261. }

  262. 

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

  264. {

  265.     int i;

  266. 

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

  268.         if (f->version > 1)

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

  270.         else

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

  272.     }

  273. }

  274. 

  275. #if HAVE_BIGENDIAN

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

  277.     {                                                   \

  278.         unsigned tmpval = AV_RN32(src);                 \

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

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

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

  282.         AV_WN32A(dst, tmpval);                          \

  283.     }

  284. #else

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

  286.     {                                                    \

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

  288.         AV_WN32A(dst, tmpval);                           \

  289.     }

  290. #endif

  291. 

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

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

  294. {

  295.     int i;

  296.     dc *= 0x10001;

  297. 

  298.     switch (log2w) {

  299.     case 0:

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

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

  302.             if (scale)

  303.                 src += stride;

  304.             dst += stride;

  305.         }

  306.         break;

  307.     case 1:

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

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

  310.             if (scale)

  311.                 src += stride;

  312.             dst += stride;

  313.         }

  314.         break;

  315.     case 2:

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

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

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

  319.             if (scale)

  320.                 src += stride;

  321.             dst += stride;

  322.         }

  323.         break;

  324.     case 3:

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

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

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

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

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

  330.             if (scale)

  331.                 src += stride;

  332.             dst += stride;

  333.         }

  334.         break;

  335.     default:

  336.         break;

  337.     }

  338. }

  339. 

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

  341.                           int log2w, int log2h, int stride)

  342. {

  343.     int index, h, code, ret, scale = 1;

  344.     uint16_t *start, *end;

  345.     unsigned dc = 0;

  346. 

  347.     if (log2h < 0 || log2w < 0)

  348.         return AVERROR_INVALIDDATA;

  349. 

  350.     index = size2index[log2h][log2w];

  351.     if (index < 0)

  352.         return AVERROR_INVALIDDATA;

  353. 

  354.     h     = 1 << log2h;

  355.     code  = get_vlc2(&f->gb, block_type_vlc[1 - (f->version > 1)][index].table,

  356.                      BLOCK_TYPE_VLC_BITS, 1);

  357.     if (code < 0 || code > 6)

  358.         return AVERROR_INVALIDDATA;

  359. 

  360.     start = f->last_frame_buffer;

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

  362. 

  363.     if (code == 1) {

  364.         if (--log2h < 0)

  365.             return AVERROR_INVALIDDATA;

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

  367.             return ret;

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

  369.                               src + (stride << log2h),

  370.                               log2w, log2h, stride);

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

  372.         log2w--;

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

  374.             return ret;

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

  376.                               src + (1 << log2w),

  377.                               log2w, log2h, stride);

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

  379.         if (log2w) {

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

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

  382.         } else {

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

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

  385.         }

  386.         return 0;

  387.     }

  388. 

  389.     if (code == 0) {

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

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

  392.         return 0;

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

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

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

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

  397.         scale = 0;

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

  399.     }

  400. 

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

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

  403.         return AVERROR_INVALIDDATA;

  404.     }

  405. 

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

  407. 

  408.     return 0;

  409. }

  410. 

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

  412. {

  413.     int x, y;

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

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

  416.     uint16_t *dst    = f->frame_buffer;

  417.     uint16_t *src;

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

  419.                  bytestream_offset, wordstream_offset;

  420.     int ret;

  421. 

  422.     src = f->last_frame_buffer;

  423. 

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

  425.         if (length < 20)

  426.             return AVERROR_INVALIDDATA;

  427.         extra           = 20;

  428.         bitstream_size  = AV_RL32(buf + 8);

  429.         wordstream_size = AV_RL32(buf + 12);

  430.         bytestream_size = AV_RL32(buf + 16);

  431.     } else {

  432.         extra           = 0;

  433.         bitstream_size  = AV_RL16(buf - 4);

  434.         wordstream_size = AV_RL16(buf - 2);

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

  436.     }

  437. 

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

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

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

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

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

  443.                bitstream_size, bytestream_size, wordstream_size,

  444.                bitstream_size + bytestream_size + wordstream_size - length);

  445.         return AVERROR_INVALIDDATA;

  446.     }

  447. 

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

  449.                    bitstream_size + AV_INPUT_BUFFER_PADDING_SIZE);

  450.     if (!f->bitstream_buffer)

  451.         return AVERROR(ENOMEM);

  452.     f->bbdsp.bswap_buf(f->bitstream_buffer, (const uint32_t *) (buf + extra),

  453.                        bitstream_size / 4);

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

  455.            0, AV_INPUT_BUFFER_PADDING_SIZE);

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

  457. 

  458.     wordstream_offset = extra + bitstream_size;

  459.     bytestream_offset = extra + bitstream_size + wordstream_size;

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

  461.                      length - wordstream_offset);

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

  463.                      length - bytestream_offset);

  464. 

  465.     init_mv(f, width * 2);

  466. 

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

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

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

  470.                 return ret;

  471.         src += 8 * width;

  472.         dst += 8 * width;

  473.     }

  474. 

  475.     return 0;

  476. }

  477. 

  478. /**

  479.  * decode block and dequantize.

  480.  * Note this is almost identical to MJPEG.

  481.  */

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

  483. {

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

  485. 

  486.     /* DC coef */

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

  488.     if (val >> 4)

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

  490. 

  491.     if (val)

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

  493. 

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

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

  496.     /* AC coefs */

  497.     i = 1;

  498.     for (;;) {

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

  500. 

  501.         /* EOB */

  502.         if (code == 0)

  503.             break;

  504.         if (code == 0xf0) {

  505.             i += 16;

  506.         } else {

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

  508.             i    += code >> 4;

  509.             if (i >= 64) {

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

  511.                 return 0;

  512.             }

  513. 

  514.             j = ff_zigzag_direct[i];

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

  516.             i++;

  517.             if (i >= 64)

  518.                 break;

  519.         }

  520.     }

  521. 

  522.     return 0;

  523. }

  524. 

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

  526. {

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

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

  529.     int i;

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

  531. 

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

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

  534.         idct(block[i]);

  535.     }

  536. 

  537.     if (!(f->avctx->flags & AV_CODEC_FLAG_GRAY)) {

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

  539.             idct(block[i]);

  540.     }

  541. 

  542.     /* Note transform is:

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

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

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

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

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

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

  549.                             2 * (x & 3) + 2 * 8 * (y & 3); // FIXME optimize

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

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

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

  553.             int y;

  554. 

  555.             cb += cb;

  556. 

  557.             y               = temp[0];

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

  559.             y               = temp[1];

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

  561.             y               = temp[8];

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

  563.             y               = temp[9];

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

  565.             dst            += 2;

  566.         }

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

  568.     }

  569. }

  570. 

  571. static int decode_i_mb(FourXContext *f)

  572. {

  573.     int ret;

  574.     int i;

  575. 

  576.     f->bdsp.clear_blocks(f->block[0]);

  577. 

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

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

  580.             return ret;

  581. 

  582.     return 0;

  583. }

  584. 

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

  586.                                           const uint8_t * const buf,

  587.                                           int len)

  588. {

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

  590.     uint8_t flag[512];

  591.     int up[512];

  592.     uint8_t len_tab[257];

  593.     int bits_tab[257];

  594.     int start, end;

  595.     const uint8_t *ptr = buf;

  596.     int j;

  597. 

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

  599. 

  600.     start = *ptr++;

  601.     end   = *ptr++;

  602.     for (;;) {

  603.         int i;

  604. 

  605.         len -= end - start + 1;

  606. 

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

  608.             return NULL;

  609. 

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

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

  612.         start = *ptr++;

  613.         if (start == 0)

  614.             break;

  615. 

  616.         if (--len < 0)

  617.             return NULL;

  618. 

  619.         end = *ptr++;

  620.     }

  621.     frequency[256] = 1;

  622. 

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

  624.         ptr++; // 4byte align

  625. 

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

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

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

  629.         int i;

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

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

  632.                 continue;

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

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

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

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

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

  638.                     smallest[0] = i;

  639.                 } else {

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

  641.                     smallest[1] = i;

  642.                 }

  643.             }

  644.         }

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

  646.             break;

  647. 

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

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

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

  651.         up[smallest[0]]        =

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

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

  654.     }

  655. 

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

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

  658. 

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

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

  661.             len++;

  662.             if (len > 31)

  663.                 // can this happen at all ?

  664.                 av_log(f->avctx, AV_LOG_ERROR,

  665.                        "vlc length overflow\n");

  666.         }

  667. 

  668.         bits_tab[j] = bits;

  669.         len_tab[j]  = len;

  670.     }

  671. 

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

  673.                  bits_tab, 4, 4, 0))

  674.         return NULL;

  675. 

  676.     return ptr;

  677. }

  678. 

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

  680. {

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

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

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

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

  685. }

  686. 

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

  688. {

  689.     int x, y, x2, y2;

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

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

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

  693.     uint16_t *dst    = f->frame_buffer;

  694.     GetByteContext g3;

  695. 

  696.     if (length < mbs * 8) {

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

  698.         return AVERROR_INVALIDDATA;

  699.     }

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

  701. 

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

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

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

  705.             // warning following is purely guessed ...

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

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

  708. 

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

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

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

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

  713. 

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

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

  716. 

  717.             bits = bytestream2_get_le32u(&g3);

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

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

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

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

  722.                 }

  723.             }

  724.             dst += 16;

  725.         }

  726.         dst += 16 * width - x;

  727.     }

  728. 

  729.     return 0;

  730. }

  731. 

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

  733. {

  734.     int x, y, ret;

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

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

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

  738.     int token_count av_unused;

  739.     unsigned int prestream_size;

  740.     const uint8_t *prestream;

  741. 

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

  743.         return AVERROR_INVALIDDATA;

  744. 

  745.     if (length < bitstream_size + 12) {

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

  747.         return AVERROR_INVALIDDATA;

  748.     }

  749. 

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

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

  752.     prestream      =             buf + bitstream_size + 12;

  753. 

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

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

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

  757.                prestream_size, bitstream_size, length);

  758.         return AVERROR_INVALIDDATA;

  759.     }

  760. 

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

  762.     if (!prestream) {

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

  764.         return AVERROR_INVALIDDATA;

  765.     }

  766. 

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

  768. 

  769.     prestream_size = length + buf - prestream;

  770. 

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

  772.                    prestream_size + AV_INPUT_BUFFER_PADDING_SIZE);

  773.     if (!f->bitstream_buffer)

  774.         return AVERROR(ENOMEM);

  775.     f->bbdsp.bswap_buf(f->bitstream_buffer, (const uint32_t *) prestream,

  776.                        prestream_size / 4);

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

  778.            0, AV_INPUT_BUFFER_PADDING_SIZE);

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

  780. 

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

  782. 

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

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

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

  786.                 return ret;

  787. 

  788.             idct_put(f, x, y);

  789.         }

  790.     }

  791. 

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

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

  794. 

  795.     return 0;

  796. }

  797. 

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

  799.                         int *got_frame, AVPacket *avpkt)

  800. {

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

  802.     int buf_size          = avpkt->size;

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

  804.     AVFrame *picture      = data;

  805.     int i, frame_4cc, frame_size, ret;

  806. 

  807.     if (buf_size < 20)

  808.         return AVERROR_INVALIDDATA;

  809. 

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

  811.         av_log(avctx, AV_LOG_ERROR,

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

  813.         return AVERROR_INVALIDDATA;

  814.     }

  815. 

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

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

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

  819.         return AVERROR_INVALIDDATA;

  820.     }

  821. 

  822.     frame_4cc = AV_RL32(buf);

  823. 

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

  825.         int free_index       = -1;

  826.         int id, whole_size;

  827.         const int data_size  = buf_size - 20;

  828.         CFrameBuffer *cfrm;

  829. 

  830.         id         = AV_RL32(buf + 12);

  831.         whole_size = AV_RL32(buf + 16);

  832. 

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

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

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

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

  837. 

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

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

  840.                 break;

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

  842.                 free_index = i;

  843.         }

  844. 

  845.         if (i >= CFRAME_BUFFER_COUNT) {

  846.             i             = free_index;

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

  848.         }

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

  850. 

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

  852.                                      cfrm->size + data_size + AV_INPUT_BUFFER_PADDING_SIZE);

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

  854.         if (!cfrm->data) {

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

  856.             return AVERROR(ENOMEM);

  857.         }

  858. 

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

  860.         cfrm->size += data_size;

  861. 

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

  863.             buf        = cfrm->data;

  864.             frame_size = cfrm->size;

  865. 

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

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

  868.                        id, avctx->frame_number);

  869. 

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

  871.                 return AVERROR_INVALIDDATA;

  872. 

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

  874.             frame_4cc  = AV_RL32("pfrm");

  875.         } else

  876.             return buf_size;

  877.     } else {

  878.         buf        = buf      + 12;

  879.         frame_size = buf_size - 12;

  880.     }

  881. 

  882. 

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

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

  885.         return ret;

  886.     }

  887. 

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

  889.         picture->pict_type = AV_PICTURE_TYPE_I;

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

  891.             return ret;

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

  893.         picture->pict_type = AV_PICTURE_TYPE_I;

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

  895.             return ret;

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

  897.         picture->pict_type = AV_PICTURE_TYPE_P;

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

  899.             return ret;

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

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

  902.                buf_size);

  903.     } else {

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

  905.                buf_size);

  906.     }

  907. 

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

  909. 

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

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

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

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

  914. 

  915.     *got_frame = 1;

  916. 

  917.     emms_c();

  918. 

  919.     return buf_size;

  920. }

  921. 

  922. static av_cold int decode_end(AVCodecContext *avctx)

  923. {

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

  925.     int i;

  926. 

  927.     av_freep(&f->frame_buffer);

  928.     av_freep(&f->last_frame_buffer);

  929.     av_freep(&f->bitstream_buffer);

  930.     f->bitstream_buffer_size = 0;

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

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

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

  934.     }

  935.     ff_free_vlc(&f->pre_vlc);

  936. 

  937.     return 0;

  938. }

  939. 

  940. static av_cold int decode_init(AVCodecContext *avctx)

  941. {

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

  943.     int ret;

  944. 

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

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

  947.         return AVERROR_INVALIDDATA;

  948.     }

  949. 

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

  951.     if (ret < 0)

  952.         return ret;

  953. 

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

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

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

  957.         decode_end(avctx);

  958.         return AVERROR(ENOMEM);

  959.     }

  960. 

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

  962.     ff_blockdsp_init(&f->bdsp, avctx);

  963.     ff_bswapdsp_init(&f->bbdsp);

  964.     f->avctx = avctx;

  965.     init_vlcs(f);

  966. 

  967.     if (f->version > 2)

  968.         avctx->pix_fmt = AV_PIX_FMT_RGB565;

  969.     else

  970.         avctx->pix_fmt = AV_PIX_FMT_BGR555;

  971. 

  972.     return 0;

  973. }

  974. 

  975. AVCodec ff_fourxm_decoder = {

  976.     .name           = "4xm",

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

  978.     .type           = AVMEDIA_TYPE_VIDEO,

  979.     .id             = AV_CODEC_ID_4XM,

  980.     .priv_data_size = sizeof(FourXContext),

  981.     .init           = decode_init,

  982.     .close          = decode_end,

  983.     .decode         = decode_frame,

  984.     .capabilities   = AV_CODEC_CAP_DR1,

  985. };