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.

933 lines
31KB

  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. * @file
  23. * 4XM codec.
  24. */
  25. #include "libavutil/frame.h"
  26. #include "libavutil/intreadwrite.h"
  27. #include "avcodec.h"
  28. #include "bytestream.h"
  29. #include "dsputil.h"
  30. #include "get_bits.h"
  31. #include "internal.h"
  32. //#undef NDEBUG
  33. //#include <assert.h>
  34. #define BLOCK_TYPE_VLC_BITS 5
  35. #define ACDC_VLC_BITS 9
  36. #define CFRAME_BUFFER_COUNT 100
  37. static const uint8_t block_type_tab[2][4][8][2] = {
  38. {
  39. { // { 8, 4, 2 } x { 8, 4, 2}
  40. { 0, 1 }, { 2, 2 }, { 6, 3 }, { 14, 4 }, { 30, 5 }, { 31, 5 }, { 0, 0 }
  41. }, { // { 8, 4 } x 1
  42. { 0, 1 }, { 0, 0 }, { 2, 2 }, { 6, 3 }, { 14, 4 }, { 15, 4 }, { 0, 0 }
  43. }, { // 1 x { 8, 4 }
  44. { 0, 1 }, { 2, 2 }, { 0, 0 }, { 6, 3 }, { 14, 4 }, { 15, 4 }, { 0, 0 }
  45. }, { // 1 x 2, 2 x 1
  46. { 0, 1 }, { 0, 0 }, { 0, 0 }, { 2, 2 }, { 6, 3 }, { 14, 4 }, { 15, 4 }
  47. }
  48. }, {
  49. { // { 8, 4, 2 } x { 8, 4, 2}
  50. { 1, 2 }, { 4, 3 }, { 5, 3 }, { 0, 2 }, { 6, 3 }, { 7, 3 }, { 0, 0 }
  51. }, {// { 8, 4 } x 1
  52. { 1, 2 }, { 0, 0 }, { 2, 2 }, { 0, 2 }, { 6, 3 }, { 7, 3 }, { 0, 0 }
  53. }, {// 1 x { 8, 4 }
  54. { 1, 2 }, { 2, 2 }, { 0, 0 }, { 0, 2 }, { 6, 3 }, { 7, 3 }, { 0, 0 }
  55. }, {// 1 x 2, 2 x 1
  56. { 1, 2 }, { 0, 0 }, { 0, 0 }, { 0, 2 }, { 2, 2 }, { 6, 3 }, { 7, 3 }
  57. }
  58. }
  59. };
  60. static const uint8_t size2index[4][4] = {
  61. { -1, 3, 1, 1 },
  62. { 3, 0, 0, 0 },
  63. { 2, 0, 0, 0 },
  64. { 2, 0, 0, 0 },
  65. };
  66. static const int8_t mv[256][2] = {
  67. { 0, 0 }, { 0, -1 }, { -1, 0 }, { 1, 0 }, { 0, 1 }, { -1, -1 }, { 1, -1 }, { -1, 1 },
  68. { 1, 1 }, { 0, -2 }, { -2, 0 }, { 2, 0 }, { 0, 2 }, { -1, -2 }, { 1, -2 }, { -2, -1 },
  69. { 2, -1 }, { -2, 1 }, { 2, 1 }, { -1, 2 }, { 1, 2 }, { -2, -2 }, { 2, -2 }, { -2, 2 },
  70. { 2, 2 }, { 0, -3 }, { -3, 0 }, { 3, 0 }, { 0, 3 }, { -1, -3 }, { 1, -3 }, { -3, -1 },
  71. { 3, -1 }, { -3, 1 }, { 3, 1 }, { -1, 3 }, { 1, 3 }, { -2, -3 }, { 2, -3 }, { -3, -2 },
  72. { 3, -2 }, { -3, 2 }, { 3, 2 }, { -2, 3 }, { 2, 3 }, { 0, -4 }, { -4, 0 }, { 4, 0 },
  73. { 0, 4 }, { -1, -4 }, { 1, -4 }, { -4, -1 }, { 4, -1 }, { 4, 1 }, { -1, 4 }, { 1, 4 },
  74. { -3, -3 }, { -3, 3 }, { 3, 3 }, { -2, -4 }, { -4, -2 }, { 4, -2 }, { -4, 2 }, { -2, 4 },
  75. { 2, 4 }, { -3, -4 }, { 3, -4 }, { 4, -3 }, { -5, 0 }, { -4, 3 }, { -3, 4 }, { 3, 4 },
  76. { -1, -5 }, { -5, -1 }, { -5, 1 }, { -1, 5 }, { -2, -5 }, { 2, -5 }, { 5, -2 }, { 5, 2 },
  77. { -4, -4 }, { -4, 4 }, { -3, -5 }, { -5, -3 }, { -5, 3 }, { 3, 5 }, { -6, 0 }, { 0, 6 },
  78. { -6, -1 }, { -6, 1 }, { 1, 6 }, { 2, -6 }, { -6, 2 }, { 2, 6 }, { -5, -4 }, { 5, 4 },
  79. { 4, 5 }, { -6, -3 }, { 6, 3 }, { -7, 0 }, { -1, -7 }, { 5, -5 }, { -7, 1 }, { -1, 7 },
  80. { 4, -6 }, { 6, 4 }, { -2, -7 }, { -7, 2 }, { -3, -7 }, { 7, -3 }, { 3, 7 }, { 6, -5 },
  81. { 0, -8 }, { -1, -8 }, { -7, -4 }, { -8, 1 }, { 4, 7 }, { 2, -8 }, { -2, 8 }, { 6, 6 },
  82. { -8, 3 }, { 5, -7 }, { -5, 7 }, { 8, -4 }, { 0, -9 }, { -9, -1 }, { 1, 9 }, { 7, -6 },
  83. { -7, 6 }, { -5, -8 }, { -5, 8 }, { -9, 3 }, { 9, -4 }, { 7, -7 }, { 8, -6 }, { 6, 8 },
  84. { 10, 1 }, { -10, 2 }, { 9, -5 }, { 10, -3 }, { -8, -7 }, { -10, -4 }, { 6, -9 }, { -11, 0 },
  85. { 11, 1 }, { -11, -2 }, { -2, 11 }, { 7, -9 }, { -7, 9 }, { 10, 6 }, { -4, 11 }, { 8, -9 },
  86. { 8, 9 }, { 5, 11 }, { 7, -10 }, { 12, -3 }, { 11, 6 }, { -9, -9 }, { 8, 10 }, { 5, 12 },
  87. { -11, 7 }, { 13, 2 }, { 6, -12 }, { 10, 9 }, { -11, 8 }, { -7, 12 }, { 0, 14 }, { 14, -2 },
  88. { -9, 11 }, { -6, 13 }, { -14, -4 }, { -5, -14 }, { 5, 14 }, { -15, -1 }, { -14, -6 }, { 3, -15 },
  89. { 11, -11 }, { -7, 14 }, { -5, 15 }, { 8, -14 }, { 15, 6 }, { 3, 16 }, { 7, -15 }, { -16, 5 },
  90. { 0, 17 }, { -16, -6 }, { -10, 14 }, { -16, 7 }, { 12, 13 }, { -16, 8 }, { -17, 6 }, { -18, 3 },
  91. { -7, 17 }, { 15, 11 }, { 16, 10 }, { 2, -19 }, { 3, -19 }, { -11, -16 }, { -18, 8 }, { -19, -6 },
  92. { 2, -20 }, { -17, -11 }, { -10, -18 }, { 8, 19 }, { -21, -1 }, { -20, 7 }, { -4, 21 }, { 21, 5 },
  93. { 15, 16 }, { 2, -22 }, { -10, -20 }, { -22, 5 }, { 20, -11 }, { -7, -22 }, { -12, 20 }, { 23, -5 },
  94. { 13, -20 }, { 24, -2 }, { -15, 19 }, { -11, 22 }, { 16, 19 }, { 23, -10 }, { -18, -18 }, { -9, -24 },
  95. { 24, -10 }, { -3, 26 }, { -23, 13 }, { -18, -20 }, { 17, 21 }, { -4, 27 }, { 27, 6 }, { 1, -28 },
  96. { -11, 26 }, { -17, -23 }, { 7, 28 }, { 11, -27 }, { 29, 5 }, { -23, -19 }, { -28, -11 }, { -21, 22 },
  97. { -30, 7 }, { -17, 26 }, { -27, 16 }, { 13, 29 }, { 19, -26 }, { 10, -31 }, { -14, -30 }, { 20, -27 },
  98. { -29, 18 }, { -16, -31 }, { -28, -22 }, { 21, -30 }, { -25, 28 }, { 26, -29 }, { 25, -32 }, { -32, -32 }
  99. };
  100. /* This is simply the scaled down elementwise product of the standard JPEG
  101. * quantizer table and the AAN premul table. */
  102. static const uint8_t dequant_table[64] = {
  103. 16, 15, 13, 19, 24, 31, 28, 17,
  104. 17, 23, 25, 31, 36, 63, 45, 21,
  105. 18, 24, 27, 37, 52, 59, 49, 20,
  106. 16, 28, 34, 40, 60, 80, 51, 20,
  107. 18, 31, 48, 66, 68, 86, 56, 21,
  108. 19, 38, 56, 59, 64, 64, 48, 20,
  109. 27, 48, 55, 55, 56, 51, 35, 15,
  110. 20, 35, 34, 32, 31, 22, 15, 8,
  111. };
  112. static VLC block_type_vlc[2][4];
  113. typedef struct CFrameBuffer {
  114. unsigned int allocated_size;
  115. unsigned int size;
  116. int id;
  117. uint8_t *data;
  118. } CFrameBuffer;
  119. typedef struct FourXContext {
  120. AVCodecContext *avctx;
  121. DSPContext dsp;
  122. AVFrame *last_picture;
  123. GetBitContext pre_gb; ///< ac/dc prefix
  124. GetBitContext gb;
  125. GetByteContext g;
  126. GetByteContext g2;
  127. int mv[256];
  128. VLC pre_vlc;
  129. int last_dc;
  130. DECLARE_ALIGNED(16, int16_t, block)[6][64];
  131. void *bitstream_buffer;
  132. unsigned int bitstream_buffer_size;
  133. int version;
  134. CFrameBuffer cfrm[CFRAME_BUFFER_COUNT];
  135. } FourXContext;
  136. #define FIX_1_082392200 70936
  137. #define FIX_1_414213562 92682
  138. #define FIX_1_847759065 121095
  139. #define FIX_2_613125930 171254
  140. #define MULTIPLY(var, const) (((var) * (const)) >> 16)
  141. static void idct(int16_t block[64])
  142. {
  143. int tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
  144. int tmp10, tmp11, tmp12, tmp13;
  145. int z5, z10, z11, z12, z13;
  146. int i;
  147. int temp[64];
  148. for (i = 0; i < 8; i++) {
  149. tmp10 = block[8 * 0 + i] + block[8 * 4 + i];
  150. tmp11 = block[8 * 0 + i] - block[8 * 4 + i];
  151. tmp13 = block[8 * 2 + i] + block[8 * 6 + i];
  152. tmp12 = MULTIPLY(block[8 * 2 + i] - block[8 * 6 + i], FIX_1_414213562) - tmp13;
  153. tmp0 = tmp10 + tmp13;
  154. tmp3 = tmp10 - tmp13;
  155. tmp1 = tmp11 + tmp12;
  156. tmp2 = tmp11 - tmp12;
  157. z13 = block[8 * 5 + i] + block[8 * 3 + i];
  158. z10 = block[8 * 5 + i] - block[8 * 3 + i];
  159. z11 = block[8 * 1 + i] + block[8 * 7 + i];
  160. z12 = block[8 * 1 + i] - block[8 * 7 + i];
  161. tmp7 = z11 + z13;
  162. tmp11 = MULTIPLY(z11 - z13, FIX_1_414213562);
  163. z5 = MULTIPLY(z10 + z12, FIX_1_847759065);
  164. tmp10 = MULTIPLY(z12, FIX_1_082392200) - z5;
  165. tmp12 = MULTIPLY(z10, -FIX_2_613125930) + z5;
  166. tmp6 = tmp12 - tmp7;
  167. tmp5 = tmp11 - tmp6;
  168. tmp4 = tmp10 + tmp5;
  169. temp[8 * 0 + i] = tmp0 + tmp7;
  170. temp[8 * 7 + i] = tmp0 - tmp7;
  171. temp[8 * 1 + i] = tmp1 + tmp6;
  172. temp[8 * 6 + i] = tmp1 - tmp6;
  173. temp[8 * 2 + i] = tmp2 + tmp5;
  174. temp[8 * 5 + i] = tmp2 - tmp5;
  175. temp[8 * 4 + i] = tmp3 + tmp4;
  176. temp[8 * 3 + i] = tmp3 - tmp4;
  177. }
  178. for (i = 0; i < 8 * 8; i += 8) {
  179. tmp10 = temp[0 + i] + temp[4 + i];
  180. tmp11 = temp[0 + i] - temp[4 + i];
  181. tmp13 = temp[2 + i] + temp[6 + i];
  182. tmp12 = MULTIPLY(temp[2 + i] - temp[6 + i], FIX_1_414213562) - tmp13;
  183. tmp0 = tmp10 + tmp13;
  184. tmp3 = tmp10 - tmp13;
  185. tmp1 = tmp11 + tmp12;
  186. tmp2 = tmp11 - tmp12;
  187. z13 = temp[5 + i] + temp[3 + i];
  188. z10 = temp[5 + i] - temp[3 + i];
  189. z11 = temp[1 + i] + temp[7 + i];
  190. z12 = temp[1 + i] - temp[7 + i];
  191. tmp7 = z11 + z13;
  192. tmp11 = MULTIPLY(z11 - z13, FIX_1_414213562);
  193. z5 = MULTIPLY(z10 + z12, FIX_1_847759065);
  194. tmp10 = MULTIPLY(z12, FIX_1_082392200) - z5;
  195. tmp12 = MULTIPLY(z10, -FIX_2_613125930) + z5;
  196. tmp6 = tmp12 - tmp7;
  197. tmp5 = tmp11 - tmp6;
  198. tmp4 = tmp10 + tmp5;
  199. block[0 + i] = (tmp0 + tmp7) >> 6;
  200. block[7 + i] = (tmp0 - tmp7) >> 6;
  201. block[1 + i] = (tmp1 + tmp6) >> 6;
  202. block[6 + i] = (tmp1 - tmp6) >> 6;
  203. block[2 + i] = (tmp2 + tmp5) >> 6;
  204. block[5 + i] = (tmp2 - tmp5) >> 6;
  205. block[4 + i] = (tmp3 + tmp4) >> 6;
  206. block[3 + i] = (tmp3 - tmp4) >> 6;
  207. }
  208. }
  209. static av_cold void init_vlcs(FourXContext *f)
  210. {
  211. static VLC_TYPE table[2][4][32][2];
  212. int i, j;
  213. for (i = 0; i < 2; i++) {
  214. for (j = 0; j < 4; j++) {
  215. block_type_vlc[i][j].table = table[i][j];
  216. block_type_vlc[i][j].table_allocated = 32;
  217. init_vlc(&block_type_vlc[i][j], BLOCK_TYPE_VLC_BITS, 7,
  218. &block_type_tab[i][j][0][1], 2, 1,
  219. &block_type_tab[i][j][0][0], 2, 1,
  220. INIT_VLC_USE_NEW_STATIC);
  221. }
  222. }
  223. }
  224. static void init_mv(FourXContext *f, int linesize)
  225. {
  226. int i;
  227. for (i = 0; i < 256; i++) {
  228. if (f->version > 1)
  229. f->mv[i] = mv[i][0] + mv[i][1] * linesize / 2;
  230. else
  231. f->mv[i] = (i & 15) - 8 + ((i >> 4) - 8) * linesize / 2;
  232. }
  233. }
  234. #if HAVE_BIGENDIAN
  235. #define LE_CENTRIC_MUL(dst, src, scale, dc) \
  236. { \
  237. unsigned tmpval = AV_RN32(src); \
  238. tmpval = (tmpval << 16) | (tmpval >> 16); \
  239. tmpval = tmpval * (scale) + (dc); \
  240. tmpval = (tmpval << 16) | (tmpval >> 16); \
  241. AV_WN32A(dst, tmpval); \
  242. }
  243. #else
  244. #define LE_CENTRIC_MUL(dst, src, scale, dc) \
  245. { \
  246. unsigned tmpval = AV_RN32(src) * (scale) + (dc); \
  247. AV_WN32A(dst, tmpval); \
  248. }
  249. #endif
  250. static inline void mcdc(uint16_t *dst, uint16_t *src, int log2w,
  251. int h, int stride, int scale, unsigned dc)
  252. {
  253. int i;
  254. dc *= 0x10001;
  255. switch (log2w) {
  256. case 0:
  257. for (i = 0; i < h; i++) {
  258. dst[0] = scale * src[0] + dc;
  259. if (scale)
  260. src += stride;
  261. dst += stride;
  262. }
  263. break;
  264. case 1:
  265. for (i = 0; i < h; i++) {
  266. LE_CENTRIC_MUL(dst, src, scale, dc);
  267. if (scale)
  268. src += stride;
  269. dst += stride;
  270. }
  271. break;
  272. case 2:
  273. for (i = 0; i < h; i++) {
  274. LE_CENTRIC_MUL(dst, src, scale, dc);
  275. LE_CENTRIC_MUL(dst + 2, src + 2, scale, dc);
  276. if (scale)
  277. src += stride;
  278. dst += stride;
  279. }
  280. break;
  281. case 3:
  282. for (i = 0; i < h; i++) {
  283. LE_CENTRIC_MUL(dst, src, scale, dc);
  284. LE_CENTRIC_MUL(dst + 2, src + 2, scale, dc);
  285. LE_CENTRIC_MUL(dst + 4, src + 4, scale, dc);
  286. LE_CENTRIC_MUL(dst + 6, src + 6, scale, dc);
  287. if (scale)
  288. src += stride;
  289. dst += stride;
  290. }
  291. break;
  292. default:
  293. assert(0);
  294. }
  295. }
  296. static void decode_p_block(FourXContext *f, uint16_t *dst, uint16_t *src,
  297. int log2w, int log2h, int stride)
  298. {
  299. const int index = size2index[log2h][log2w];
  300. const int h = 1 << log2h;
  301. int code = get_vlc2(&f->gb,
  302. block_type_vlc[1 - (f->version > 1)][index].table,
  303. BLOCK_TYPE_VLC_BITS, 1);
  304. uint16_t *start = (uint16_t *)f->last_picture->data[0];
  305. uint16_t *end = start + stride * (f->avctx->height - h + 1) - (1 << log2w);
  306. assert(code >= 0 && code <= 6);
  307. if (code == 0) {
  308. src += f->mv[bytestream2_get_byte(&f->g)];
  309. if (start > src || src > end) {
  310. av_log(f->avctx, AV_LOG_ERROR, "mv out of pic\n");
  311. return;
  312. }
  313. mcdc(dst, src, log2w, h, stride, 1, 0);
  314. } else if (code == 1) {
  315. log2h--;
  316. decode_p_block(f, dst, src, log2w, log2h, stride);
  317. decode_p_block(f, dst + (stride << log2h),
  318. src + (stride << log2h), log2w, log2h, stride);
  319. } else if (code == 2) {
  320. log2w--;
  321. decode_p_block(f, dst , src, log2w, log2h, stride);
  322. decode_p_block(f, dst + (1 << log2w),
  323. src + (1 << log2w), log2w, log2h, stride);
  324. } else if (code == 3 && f->version < 2) {
  325. mcdc(dst, src, log2w, h, stride, 1, 0);
  326. } else if (code == 4) {
  327. src += f->mv[bytestream2_get_byte(&f->g)];
  328. if (start > src || src > end) {
  329. av_log(f->avctx, AV_LOG_ERROR, "mv out of pic\n");
  330. return;
  331. }
  332. mcdc(dst, src, log2w, h, stride, 1, bytestream2_get_le16(&f->g2));
  333. } else if (code == 5) {
  334. mcdc(dst, src, log2w, h, stride, 0, bytestream2_get_le16(&f->g2));
  335. } else if (code == 6) {
  336. if (log2w) {
  337. dst[0] = bytestream2_get_le16(&f->g2);
  338. dst[1] = bytestream2_get_le16(&f->g2);
  339. } else {
  340. dst[0] = bytestream2_get_le16(&f->g2);
  341. dst[stride] = bytestream2_get_le16(&f->g2);
  342. }
  343. }
  344. }
  345. static int decode_p_frame(FourXContext *f, AVFrame *frame,
  346. const uint8_t *buf, int length)
  347. {
  348. int x, y;
  349. const int width = f->avctx->width;
  350. const int height = f->avctx->height;
  351. uint16_t *src = (uint16_t *)f->last_picture->data[0];
  352. uint16_t *dst = (uint16_t *)frame->data[0];
  353. const int stride = frame->linesize[0] >> 1;
  354. unsigned int bitstream_size, bytestream_size, wordstream_size, extra,
  355. bytestream_offset, wordstream_offset;
  356. if (f->version > 1) {
  357. extra = 20;
  358. bitstream_size = AV_RL32(buf + 8);
  359. wordstream_size = AV_RL32(buf + 12);
  360. bytestream_size = AV_RL32(buf + 16);
  361. } else {
  362. extra = 0;
  363. bitstream_size = AV_RL16(buf - 4);
  364. wordstream_size = AV_RL16(buf - 2);
  365. bytestream_size = FFMAX(length - bitstream_size - wordstream_size, 0);
  366. }
  367. if (bitstream_size + bytestream_size + wordstream_size + extra != length
  368. || bitstream_size > (1 << 26)
  369. || bytestream_size > (1 << 26)
  370. || wordstream_size > (1 << 26)) {
  371. av_log(f->avctx, AV_LOG_ERROR, "lengths %d %d %d %d\n",
  372. bitstream_size, bytestream_size, wordstream_size,
  373. bitstream_size + bytestream_size + wordstream_size - length);
  374. return AVERROR_INVALIDDATA;
  375. }
  376. av_fast_malloc(&f->bitstream_buffer, &f->bitstream_buffer_size,
  377. bitstream_size + FF_INPUT_BUFFER_PADDING_SIZE);
  378. if (!f->bitstream_buffer)
  379. return AVERROR(ENOMEM);
  380. f->dsp.bswap_buf(f->bitstream_buffer, (const uint32_t*)(buf + extra),
  381. bitstream_size / 4);
  382. memset((uint8_t*)f->bitstream_buffer + bitstream_size,
  383. 0, FF_INPUT_BUFFER_PADDING_SIZE);
  384. init_get_bits(&f->gb, f->bitstream_buffer, 8 * bitstream_size);
  385. wordstream_offset = extra + bitstream_size;
  386. bytestream_offset = extra + bitstream_size + wordstream_size;
  387. bytestream2_init(&f->g2, buf + wordstream_offset,
  388. length - wordstream_offset);
  389. bytestream2_init(&f->g, buf + bytestream_offset,
  390. length - bytestream_offset);
  391. init_mv(f, frame->linesize[0]);
  392. for (y = 0; y < height; y += 8) {
  393. for (x = 0; x < width; x += 8)
  394. decode_p_block(f, dst + x, src + x, 3, 3, stride);
  395. src += 8 * stride;
  396. dst += 8 * stride;
  397. }
  398. return 0;
  399. }
  400. /**
  401. * decode block and dequantize.
  402. * Note this is almost identical to MJPEG.
  403. */
  404. static int decode_i_block(FourXContext *f, int16_t *block)
  405. {
  406. int code, i, j, level, val;
  407. /* DC coef */
  408. val = get_vlc2(&f->pre_gb, f->pre_vlc.table, ACDC_VLC_BITS, 3);
  409. if (val >> 4)
  410. av_log(f->avctx, AV_LOG_ERROR, "error dc run != 0\n");
  411. if (val)
  412. val = get_xbits(&f->gb, val);
  413. val = val * dequant_table[0] + f->last_dc;
  414. f->last_dc = block[0] = val;
  415. /* AC coefs */
  416. i = 1;
  417. for (;;) {
  418. code = get_vlc2(&f->pre_gb, f->pre_vlc.table, ACDC_VLC_BITS, 3);
  419. /* EOB */
  420. if (code == 0)
  421. break;
  422. if (code == 0xf0) {
  423. i += 16;
  424. } else {
  425. level = get_xbits(&f->gb, code & 0xf);
  426. i += code >> 4;
  427. if (i >= 64) {
  428. av_log(f->avctx, AV_LOG_ERROR, "run %d oveflow\n", i);
  429. return 0;
  430. }
  431. j = ff_zigzag_direct[i];
  432. block[j] = level * dequant_table[j];
  433. i++;
  434. if (i >= 64)
  435. break;
  436. }
  437. }
  438. return 0;
  439. }
  440. static inline void idct_put(FourXContext *f, AVFrame *frame, int x, int y)
  441. {
  442. int16_t (*block)[64] = f->block;
  443. int stride = frame->linesize[0] >> 1;
  444. int i;
  445. uint16_t *dst = ((uint16_t*)frame->data[0]) + y * stride + x;
  446. for (i = 0; i < 4; i++) {
  447. block[i][0] += 0x80 * 8 * 8;
  448. idct(block[i]);
  449. }
  450. if (!(f->avctx->flags & CODEC_FLAG_GRAY)) {
  451. for (i = 4; i < 6; i++)
  452. idct(block[i]);
  453. }
  454. /* Note transform is:
  455. * y = ( 1b + 4g + 2r) / 14
  456. * cb = ( 3b - 2g - 1r) / 14
  457. * cr = (-1b - 4g + 5r) / 14 */
  458. for (y = 0; y < 8; y++) {
  459. for (x = 0; x < 8; x++) {
  460. int16_t *temp = block[(x >> 2) + 2 * (y >> 2)] +
  461. 2 * (x & 3) + 2 * 8 * (y & 3); // FIXME optimize
  462. int cb = block[4][x + 8 * y];
  463. int cr = block[5][x + 8 * y];
  464. int cg = (cb + cr) >> 1;
  465. int y;
  466. cb += cb;
  467. y = temp[0];
  468. dst[0] = ((y + cb) >> 3) + (((y - cg) & 0xFC) << 3) + (((y + cr) & 0xF8) << 8);
  469. y = temp[1];
  470. dst[1] = ((y + cb) >> 3) + (((y - cg) & 0xFC) << 3) + (((y + cr) & 0xF8) << 8);
  471. y = temp[8];
  472. dst[stride] = ((y + cb) >> 3) + (((y - cg) & 0xFC) << 3) + (((y + cr) & 0xF8) << 8);
  473. y = temp[9];
  474. dst[1 + stride] = ((y + cb) >> 3) + (((y - cg) & 0xFC) << 3) + (((y + cr) & 0xF8) << 8);
  475. dst += 2;
  476. }
  477. dst += 2 * stride - 2 * 8;
  478. }
  479. }
  480. static int decode_i_mb(FourXContext *f)
  481. {
  482. int ret;
  483. int i;
  484. f->dsp.clear_blocks(f->block[0]);
  485. for (i = 0; i < 6; i++)
  486. if ((ret = decode_i_block(f, f->block[i])) < 0)
  487. return ret;
  488. return 0;
  489. }
  490. static const uint8_t *read_huffman_tables(FourXContext *f,
  491. const uint8_t * const buf)
  492. {
  493. int frequency[512] = { 0 };
  494. uint8_t flag[512];
  495. int up[512];
  496. uint8_t len_tab[257];
  497. int bits_tab[257];
  498. int start, end;
  499. const uint8_t *ptr = buf;
  500. int j;
  501. memset(up, -1, sizeof(up));
  502. start = *ptr++;
  503. end = *ptr++;
  504. for (;;) {
  505. int i;
  506. for (i = start; i <= end; i++)
  507. frequency[i] = *ptr++;
  508. start = *ptr++;
  509. if (start == 0)
  510. break;
  511. end = *ptr++;
  512. }
  513. frequency[256] = 1;
  514. while ((ptr - buf) & 3)
  515. ptr++; // 4byte align
  516. for (j = 257; j < 512; j++) {
  517. int min_freq[2] = { 256 * 256, 256 * 256 };
  518. int smallest[2] = { 0, 0 };
  519. int i;
  520. for (i = 0; i < j; i++) {
  521. if (frequency[i] == 0)
  522. continue;
  523. if (frequency[i] < min_freq[1]) {
  524. if (frequency[i] < min_freq[0]) {
  525. min_freq[1] = min_freq[0];
  526. smallest[1] = smallest[0];
  527. min_freq[0] = frequency[i];
  528. smallest[0] = i;
  529. } else {
  530. min_freq[1] = frequency[i];
  531. smallest[1] = i;
  532. }
  533. }
  534. }
  535. if (min_freq[1] == 256 * 256)
  536. break;
  537. frequency[j] = min_freq[0] + min_freq[1];
  538. flag[smallest[0]] = 0;
  539. flag[smallest[1]] = 1;
  540. up[smallest[0]] =
  541. up[smallest[1]] = j;
  542. frequency[smallest[0]] = frequency[smallest[1]] = 0;
  543. }
  544. for (j = 0; j < 257; j++) {
  545. int node, len = 0, bits = 0;
  546. for (node = j; up[node] != -1; node = up[node]) {
  547. bits += flag[node] << len;
  548. len++;
  549. if (len > 31)
  550. // can this happen at all ?
  551. av_log(f->avctx, AV_LOG_ERROR,
  552. "vlc length overflow\n");
  553. }
  554. bits_tab[j] = bits;
  555. len_tab[j] = len;
  556. }
  557. if (init_vlc(&f->pre_vlc, ACDC_VLC_BITS, 257, len_tab, 1, 1,
  558. bits_tab, 4, 4, 0))
  559. return NULL;
  560. return ptr;
  561. }
  562. static int mix(int c0, int c1)
  563. {
  564. int blue = 2 * (c0 & 0x001F) + (c1 & 0x001F);
  565. int green = (2 * (c0 & 0x03E0) + (c1 & 0x03E0)) >> 5;
  566. int red = 2 * (c0 >> 10) + (c1 >> 10);
  567. return red / 3 * 1024 + green / 3 * 32 + blue / 3;
  568. }
  569. static int decode_i2_frame(FourXContext *f, AVFrame *frame, const uint8_t *buf, int length)
  570. {
  571. int x, y, x2, y2;
  572. const int width = f->avctx->width;
  573. const int height = f->avctx->height;
  574. const int mbs = (FFALIGN(width, 16) >> 4) * (FFALIGN(height, 16) >> 4);
  575. uint16_t *dst = (uint16_t*)frame->data[0];
  576. const int stride = frame->linesize[0]>>1;
  577. GetByteContext g3;
  578. if (length < mbs * 8) {
  579. av_log(f->avctx, AV_LOG_ERROR, "packet size too small\n");
  580. return AVERROR_INVALIDDATA;
  581. }
  582. bytestream2_init(&g3, buf, length);
  583. for (y = 0; y < height; y += 16) {
  584. for (x = 0; x < width; x += 16) {
  585. unsigned int color[4] = { 0 }, bits;
  586. // warning following is purely guessed ...
  587. color[0] = bytestream2_get_le16u(&g3);
  588. color[1] = bytestream2_get_le16u(&g3);
  589. if (color[0] & 0x8000)
  590. av_log(NULL, AV_LOG_ERROR, "unk bit 1\n");
  591. if (color[1] & 0x8000)
  592. av_log(NULL, AV_LOG_ERROR, "unk bit 2\n");
  593. color[2] = mix(color[0], color[1]);
  594. color[3] = mix(color[1], color[0]);
  595. bits = bytestream2_get_le32u(&g3);
  596. for (y2 = 0; y2 < 16; y2++) {
  597. for (x2 = 0; x2 < 16; x2++) {
  598. int index = 2 * (x2 >> 2) + 8 * (y2 >> 2);
  599. dst[y2 * stride + x2] = color[(bits >> index) & 3];
  600. }
  601. }
  602. dst += 16;
  603. }
  604. dst += 16 * stride - x;
  605. }
  606. return 0;
  607. }
  608. static int decode_i_frame(FourXContext *f, AVFrame *frame, const uint8_t *buf, int length)
  609. {
  610. int x, y, ret;
  611. const int width = f->avctx->width;
  612. const int height = f->avctx->height;
  613. const unsigned int bitstream_size = AV_RL32(buf);
  614. int token_count av_unused;
  615. unsigned int prestream_size;
  616. const uint8_t *prestream;
  617. if (length < bitstream_size + 12) {
  618. av_log(f->avctx, AV_LOG_ERROR, "packet size too small\n");
  619. return AVERROR_INVALIDDATA;
  620. }
  621. token_count = AV_RL32(buf + bitstream_size + 8);
  622. prestream_size = 4 * AV_RL32(buf + bitstream_size + 4);
  623. prestream = buf + bitstream_size + 12;
  624. if (prestream_size + bitstream_size + 12 != length
  625. || bitstream_size > (1 << 26)
  626. || prestream_size > (1 << 26)) {
  627. av_log(f->avctx, AV_LOG_ERROR, "size mismatch %d %d %d\n",
  628. prestream_size, bitstream_size, length);
  629. return AVERROR_INVALIDDATA;
  630. }
  631. prestream = read_huffman_tables(f, prestream);
  632. if (!prestream) {
  633. av_log(f->avctx, AV_LOG_ERROR, "Error reading Huffman tables.\n");
  634. return AVERROR_INVALIDDATA;
  635. }
  636. init_get_bits(&f->gb, buf + 4, 8 * bitstream_size);
  637. prestream_size = length + buf - prestream;
  638. av_fast_malloc(&f->bitstream_buffer, &f->bitstream_buffer_size,
  639. prestream_size + FF_INPUT_BUFFER_PADDING_SIZE);
  640. if (!f->bitstream_buffer)
  641. return AVERROR(ENOMEM);
  642. f->dsp.bswap_buf(f->bitstream_buffer, (const uint32_t*)prestream,
  643. prestream_size / 4);
  644. memset((uint8_t*)f->bitstream_buffer + prestream_size,
  645. 0, FF_INPUT_BUFFER_PADDING_SIZE);
  646. init_get_bits(&f->pre_gb, f->bitstream_buffer, 8 * prestream_size);
  647. f->last_dc = 0 * 128 * 8 * 8;
  648. for (y = 0; y < height; y += 16) {
  649. for (x = 0; x < width; x += 16) {
  650. if ((ret = decode_i_mb(f)) < 0)
  651. return ret;
  652. idct_put(f, frame, x, y);
  653. }
  654. }
  655. if (get_vlc2(&f->pre_gb, f->pre_vlc.table, ACDC_VLC_BITS, 3) != 256)
  656. av_log(f->avctx, AV_LOG_ERROR, "end mismatch\n");
  657. return 0;
  658. }
  659. static int decode_frame(AVCodecContext *avctx, void *data,
  660. int *got_frame, AVPacket *avpkt)
  661. {
  662. const uint8_t *buf = avpkt->data;
  663. int buf_size = avpkt->size;
  664. FourXContext *const f = avctx->priv_data;
  665. AVFrame *picture = data;
  666. int i, frame_4cc, frame_size, ret;
  667. frame_4cc = AV_RL32(buf);
  668. if (buf_size != AV_RL32(buf + 4) + 8 || buf_size < 20)
  669. av_log(f->avctx, AV_LOG_ERROR, "size mismatch %d %d\n",
  670. buf_size, AV_RL32(buf + 4));
  671. if (frame_4cc == AV_RL32("cfrm")) {
  672. int free_index = -1;
  673. const int data_size = buf_size - 20;
  674. const int id = AV_RL32(buf + 12);
  675. const int whole_size = AV_RL32(buf + 16);
  676. CFrameBuffer *cfrm;
  677. for (i = 0; i < CFRAME_BUFFER_COUNT; i++)
  678. if (f->cfrm[i].id && f->cfrm[i].id < avctx->frame_number)
  679. av_log(f->avctx, AV_LOG_ERROR, "lost c frame %d\n",
  680. f->cfrm[i].id);
  681. for (i = 0; i < CFRAME_BUFFER_COUNT; i++) {
  682. if (f->cfrm[i].id == id)
  683. break;
  684. if (f->cfrm[i].size == 0)
  685. free_index = i;
  686. }
  687. if (i >= CFRAME_BUFFER_COUNT) {
  688. i = free_index;
  689. f->cfrm[i].id = id;
  690. }
  691. cfrm = &f->cfrm[i];
  692. cfrm->data = av_fast_realloc(cfrm->data, &cfrm->allocated_size,
  693. cfrm->size + data_size + FF_INPUT_BUFFER_PADDING_SIZE);
  694. // explicit check needed as memcpy below might not catch a NULL
  695. if (!cfrm->data) {
  696. av_log(f->avctx, AV_LOG_ERROR, "realloc failure");
  697. return AVERROR(ENOMEM);
  698. }
  699. memcpy(cfrm->data + cfrm->size, buf + 20, data_size);
  700. cfrm->size += data_size;
  701. if (cfrm->size >= whole_size) {
  702. buf = cfrm->data;
  703. frame_size = cfrm->size;
  704. if (id != avctx->frame_number)
  705. av_log(f->avctx, AV_LOG_ERROR, "cframe id mismatch %d %d\n",
  706. id, avctx->frame_number);
  707. cfrm->size = cfrm->id = 0;
  708. frame_4cc = AV_RL32("pfrm");
  709. } else
  710. return buf_size;
  711. } else {
  712. buf = buf + 12;
  713. frame_size = buf_size - 12;
  714. }
  715. // alternatively we would have to use our own buffer management
  716. avctx->flags |= CODEC_FLAG_EMU_EDGE;
  717. if ((ret = ff_get_buffer(avctx, picture, AV_GET_BUFFER_FLAG_REF)) < 0) {
  718. av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
  719. return ret;
  720. }
  721. if (frame_4cc == AV_RL32("ifr2")) {
  722. picture->pict_type = AV_PICTURE_TYPE_I;
  723. if ((ret = decode_i2_frame(f, picture, buf - 4, frame_size + 4)) < 0)
  724. return ret;
  725. } else if (frame_4cc == AV_RL32("ifrm")) {
  726. picture->pict_type = AV_PICTURE_TYPE_I;
  727. if ((ret = decode_i_frame(f, picture, buf, frame_size)) < 0)
  728. return ret;
  729. } else if (frame_4cc == AV_RL32("pfrm") || frame_4cc == AV_RL32("pfr2")) {
  730. if (!f->last_picture->data[0]) {
  731. if ((ret = ff_get_buffer(avctx, f->last_picture,
  732. AV_GET_BUFFER_FLAG_REF)) < 0) {
  733. av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
  734. return ret;
  735. }
  736. memset(f->last_picture->data[0], 0, avctx->height * FFABS(f->last_picture->linesize[0]));
  737. }
  738. picture->pict_type = AV_PICTURE_TYPE_P;
  739. if ((ret = decode_p_frame(f, picture, buf, frame_size)) < 0)
  740. return ret;
  741. } else if (frame_4cc == AV_RL32("snd_")) {
  742. av_log(avctx, AV_LOG_ERROR, "ignoring snd_ chunk length:%d\n",
  743. buf_size);
  744. } else {
  745. av_log(avctx, AV_LOG_ERROR, "ignoring unknown chunk length:%d\n",
  746. buf_size);
  747. }
  748. picture->key_frame = picture->pict_type == AV_PICTURE_TYPE_I;
  749. av_frame_unref(f->last_picture);
  750. if ((ret = av_frame_ref(f->last_picture, picture)) < 0)
  751. return ret;
  752. *got_frame = 1;
  753. emms_c();
  754. return buf_size;
  755. }
  756. static av_cold int decode_init(AVCodecContext *avctx)
  757. {
  758. FourXContext * const f = avctx->priv_data;
  759. if (avctx->extradata_size != 4 || !avctx->extradata) {
  760. av_log(avctx, AV_LOG_ERROR, "extradata wrong or missing\n");
  761. return 1;
  762. }
  763. f->version = AV_RL32(avctx->extradata) >> 16;
  764. ff_dsputil_init(&f->dsp, avctx);
  765. f->avctx = avctx;
  766. init_vlcs(f);
  767. if (f->version > 2)
  768. avctx->pix_fmt = AV_PIX_FMT_RGB565;
  769. else
  770. avctx->pix_fmt = AV_PIX_FMT_BGR555;
  771. f->last_picture = av_frame_alloc();
  772. if (!f->last_picture)
  773. return AVERROR(ENOMEM);
  774. return 0;
  775. }
  776. static av_cold int decode_end(AVCodecContext *avctx)
  777. {
  778. FourXContext * const f = avctx->priv_data;
  779. int i;
  780. av_freep(&f->bitstream_buffer);
  781. f->bitstream_buffer_size = 0;
  782. for (i = 0; i < CFRAME_BUFFER_COUNT; i++) {
  783. av_freep(&f->cfrm[i].data);
  784. f->cfrm[i].allocated_size = 0;
  785. }
  786. ff_free_vlc(&f->pre_vlc);
  787. av_frame_free(&f->last_picture);
  788. return 0;
  789. }
  790. AVCodec ff_fourxm_decoder = {
  791. .name = "4xm",
  792. .type = AVMEDIA_TYPE_VIDEO,
  793. .id = AV_CODEC_ID_4XM,
  794. .priv_data_size = sizeof(FourXContext),
  795. .init = decode_init,
  796. .close = decode_end,
  797. .decode = decode_frame,
  798. .capabilities = CODEC_CAP_DR1,
  799. .long_name = NULL_IF_CONFIG_SMALL("4X Movie"),
  800. };