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.

1183 lines
43KB

  1. /*
  2. * H.26L/H.264/AVC/JVT/14496-10/... cavlc bitstream decoding
  3. * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
  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. * H.264 / AVC / MPEG4 part10 cavlc bitstream decoding.
  24. * @author Michael Niedermayer <michaelni@gmx.at>
  25. */
  26. #define CABAC 0
  27. #include "internal.h"
  28. #include "avcodec.h"
  29. #include "mpegvideo.h"
  30. #include "h264.h"
  31. #include "h264data.h" // FIXME FIXME FIXME
  32. #include "h264_mvpred.h"
  33. #include "golomb.h"
  34. //#undef NDEBUG
  35. #include <assert.h>
  36. static const uint8_t golomb_to_inter_cbp_gray[16]={
  37. 0, 1, 2, 4, 8, 3, 5,10,12,15, 7,11,13,14, 6, 9,
  38. };
  39. static const uint8_t golomb_to_intra4x4_cbp_gray[16]={
  40. 15, 0, 7,11,13,14, 3, 5,10,12, 1, 2, 4, 8, 6, 9,
  41. };
  42. static const uint8_t chroma_dc_coeff_token_len[4*5]={
  43. 2, 0, 0, 0,
  44. 6, 1, 0, 0,
  45. 6, 6, 3, 0,
  46. 6, 7, 7, 6,
  47. 6, 8, 8, 7,
  48. };
  49. static const uint8_t chroma_dc_coeff_token_bits[4*5]={
  50. 1, 0, 0, 0,
  51. 7, 1, 0, 0,
  52. 4, 6, 1, 0,
  53. 3, 3, 2, 5,
  54. 2, 3, 2, 0,
  55. };
  56. static const uint8_t chroma422_dc_coeff_token_len[4*9]={
  57. 1, 0, 0, 0,
  58. 7, 2, 0, 0,
  59. 7, 7, 3, 0,
  60. 9, 7, 7, 5,
  61. 9, 9, 7, 6,
  62. 10, 10, 9, 7,
  63. 11, 11, 10, 7,
  64. 12, 12, 11, 10,
  65. 13, 12, 12, 11,
  66. };
  67. static const uint8_t chroma422_dc_coeff_token_bits[4*9]={
  68. 1, 0, 0, 0,
  69. 15, 1, 0, 0,
  70. 14, 13, 1, 0,
  71. 7, 12, 11, 1,
  72. 6, 5, 10, 1,
  73. 7, 6, 4, 9,
  74. 7, 6, 5, 8,
  75. 7, 6, 5, 4,
  76. 7, 5, 4, 4,
  77. };
  78. static const uint8_t coeff_token_len[4][4*17]={
  79. {
  80. 1, 0, 0, 0,
  81. 6, 2, 0, 0, 8, 6, 3, 0, 9, 8, 7, 5, 10, 9, 8, 6,
  82. 11,10, 9, 7, 13,11,10, 8, 13,13,11, 9, 13,13,13,10,
  83. 14,14,13,11, 14,14,14,13, 15,15,14,14, 15,15,15,14,
  84. 16,15,15,15, 16,16,16,15, 16,16,16,16, 16,16,16,16,
  85. },
  86. {
  87. 2, 0, 0, 0,
  88. 6, 2, 0, 0, 6, 5, 3, 0, 7, 6, 6, 4, 8, 6, 6, 4,
  89. 8, 7, 7, 5, 9, 8, 8, 6, 11, 9, 9, 6, 11,11,11, 7,
  90. 12,11,11, 9, 12,12,12,11, 12,12,12,11, 13,13,13,12,
  91. 13,13,13,13, 13,14,13,13, 14,14,14,13, 14,14,14,14,
  92. },
  93. {
  94. 4, 0, 0, 0,
  95. 6, 4, 0, 0, 6, 5, 4, 0, 6, 5, 5, 4, 7, 5, 5, 4,
  96. 7, 5, 5, 4, 7, 6, 6, 4, 7, 6, 6, 4, 8, 7, 7, 5,
  97. 8, 8, 7, 6, 9, 8, 8, 7, 9, 9, 8, 8, 9, 9, 9, 8,
  98. 10, 9, 9, 9, 10,10,10,10, 10,10,10,10, 10,10,10,10,
  99. },
  100. {
  101. 6, 0, 0, 0,
  102. 6, 6, 0, 0, 6, 6, 6, 0, 6, 6, 6, 6, 6, 6, 6, 6,
  103. 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
  104. 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
  105. 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
  106. }
  107. };
  108. static const uint8_t coeff_token_bits[4][4*17]={
  109. {
  110. 1, 0, 0, 0,
  111. 5, 1, 0, 0, 7, 4, 1, 0, 7, 6, 5, 3, 7, 6, 5, 3,
  112. 7, 6, 5, 4, 15, 6, 5, 4, 11,14, 5, 4, 8,10,13, 4,
  113. 15,14, 9, 4, 11,10,13,12, 15,14, 9,12, 11,10,13, 8,
  114. 15, 1, 9,12, 11,14,13, 8, 7,10, 9,12, 4, 6, 5, 8,
  115. },
  116. {
  117. 3, 0, 0, 0,
  118. 11, 2, 0, 0, 7, 7, 3, 0, 7,10, 9, 5, 7, 6, 5, 4,
  119. 4, 6, 5, 6, 7, 6, 5, 8, 15, 6, 5, 4, 11,14,13, 4,
  120. 15,10, 9, 4, 11,14,13,12, 8,10, 9, 8, 15,14,13,12,
  121. 11,10, 9,12, 7,11, 6, 8, 9, 8,10, 1, 7, 6, 5, 4,
  122. },
  123. {
  124. 15, 0, 0, 0,
  125. 15,14, 0, 0, 11,15,13, 0, 8,12,14,12, 15,10,11,11,
  126. 11, 8, 9,10, 9,14,13, 9, 8,10, 9, 8, 15,14,13,13,
  127. 11,14,10,12, 15,10,13,12, 11,14, 9,12, 8,10,13, 8,
  128. 13, 7, 9,12, 9,12,11,10, 5, 8, 7, 6, 1, 4, 3, 2,
  129. },
  130. {
  131. 3, 0, 0, 0,
  132. 0, 1, 0, 0, 4, 5, 6, 0, 8, 9,10,11, 12,13,14,15,
  133. 16,17,18,19, 20,21,22,23, 24,25,26,27, 28,29,30,31,
  134. 32,33,34,35, 36,37,38,39, 40,41,42,43, 44,45,46,47,
  135. 48,49,50,51, 52,53,54,55, 56,57,58,59, 60,61,62,63,
  136. }
  137. };
  138. static const uint8_t total_zeros_len[16][16]= {
  139. {1,3,3,4,4,5,5,6,6,7,7,8,8,9,9,9},
  140. {3,3,3,3,3,4,4,4,4,5,5,6,6,6,6},
  141. {4,3,3,3,4,4,3,3,4,5,5,6,5,6},
  142. {5,3,4,4,3,3,3,4,3,4,5,5,5},
  143. {4,4,4,3,3,3,3,3,4,5,4,5},
  144. {6,5,3,3,3,3,3,3,4,3,6},
  145. {6,5,3,3,3,2,3,4,3,6},
  146. {6,4,5,3,2,2,3,3,6},
  147. {6,6,4,2,2,3,2,5},
  148. {5,5,3,2,2,2,4},
  149. {4,4,3,3,1,3},
  150. {4,4,2,1,3},
  151. {3,3,1,2},
  152. {2,2,1},
  153. {1,1},
  154. };
  155. static const uint8_t total_zeros_bits[16][16]= {
  156. {1,3,2,3,2,3,2,3,2,3,2,3,2,3,2,1},
  157. {7,6,5,4,3,5,4,3,2,3,2,3,2,1,0},
  158. {5,7,6,5,4,3,4,3,2,3,2,1,1,0},
  159. {3,7,5,4,6,5,4,3,3,2,2,1,0},
  160. {5,4,3,7,6,5,4,3,2,1,1,0},
  161. {1,1,7,6,5,4,3,2,1,1,0},
  162. {1,1,5,4,3,3,2,1,1,0},
  163. {1,1,1,3,3,2,2,1,0},
  164. {1,0,1,3,2,1,1,1},
  165. {1,0,1,3,2,1,1},
  166. {0,1,1,2,1,3},
  167. {0,1,1,1,1},
  168. {0,1,1,1},
  169. {0,1,1},
  170. {0,1},
  171. };
  172. static const uint8_t chroma_dc_total_zeros_len[3][4]= {
  173. { 1, 2, 3, 3,},
  174. { 1, 2, 2, 0,},
  175. { 1, 1, 0, 0,},
  176. };
  177. static const uint8_t chroma_dc_total_zeros_bits[3][4]= {
  178. { 1, 1, 1, 0,},
  179. { 1, 1, 0, 0,},
  180. { 1, 0, 0, 0,},
  181. };
  182. static const uint8_t chroma422_dc_total_zeros_len[7][8]= {
  183. { 1, 3, 3, 4, 4, 4, 5, 5 },
  184. { 3, 2, 3, 3, 3, 3, 3 },
  185. { 3, 3, 2, 2, 3, 3 },
  186. { 3, 2, 2, 2, 3 },
  187. { 2, 2, 2, 2 },
  188. { 2, 2, 1 },
  189. { 1, 1 },
  190. };
  191. static const uint8_t chroma422_dc_total_zeros_bits[7][8]= {
  192. { 1, 2, 3, 2, 3, 1, 1, 0 },
  193. { 0, 1, 1, 4, 5, 6, 7 },
  194. { 0, 1, 1, 2, 6, 7 },
  195. { 6, 0, 1, 2, 7 },
  196. { 0, 1, 2, 3 },
  197. { 0, 1, 1 },
  198. { 0, 1 },
  199. };
  200. static const uint8_t run_len[7][16]={
  201. {1,1},
  202. {1,2,2},
  203. {2,2,2,2},
  204. {2,2,2,3,3},
  205. {2,2,3,3,3,3},
  206. {2,3,3,3,3,3,3},
  207. {3,3,3,3,3,3,3,4,5,6,7,8,9,10,11},
  208. };
  209. static const uint8_t run_bits[7][16]={
  210. {1,0},
  211. {1,1,0},
  212. {3,2,1,0},
  213. {3,2,1,1,0},
  214. {3,2,3,2,1,0},
  215. {3,0,1,3,2,5,4},
  216. {7,6,5,4,3,2,1,1,1,1,1,1,1,1,1},
  217. };
  218. static VLC coeff_token_vlc[4];
  219. static VLC_TYPE coeff_token_vlc_tables[520+332+280+256][2];
  220. static const int coeff_token_vlc_tables_size[4]={520,332,280,256};
  221. static VLC chroma_dc_coeff_token_vlc;
  222. static VLC_TYPE chroma_dc_coeff_token_vlc_table[256][2];
  223. static const int chroma_dc_coeff_token_vlc_table_size = 256;
  224. static VLC chroma422_dc_coeff_token_vlc;
  225. static VLC_TYPE chroma422_dc_coeff_token_vlc_table[8192][2];
  226. static const int chroma422_dc_coeff_token_vlc_table_size = 8192;
  227. static VLC total_zeros_vlc[15];
  228. static VLC_TYPE total_zeros_vlc_tables[15][512][2];
  229. static const int total_zeros_vlc_tables_size = 512;
  230. static VLC chroma_dc_total_zeros_vlc[3];
  231. static VLC_TYPE chroma_dc_total_zeros_vlc_tables[3][8][2];
  232. static const int chroma_dc_total_zeros_vlc_tables_size = 8;
  233. static VLC chroma422_dc_total_zeros_vlc[7];
  234. static VLC_TYPE chroma422_dc_total_zeros_vlc_tables[7][32][2];
  235. static const int chroma422_dc_total_zeros_vlc_tables_size = 32;
  236. static VLC run_vlc[6];
  237. static VLC_TYPE run_vlc_tables[6][8][2];
  238. static const int run_vlc_tables_size = 8;
  239. static VLC run7_vlc;
  240. static VLC_TYPE run7_vlc_table[96][2];
  241. static const int run7_vlc_table_size = 96;
  242. #define LEVEL_TAB_BITS 8
  243. static int8_t cavlc_level_tab[7][1<<LEVEL_TAB_BITS][2];
  244. #define CHROMA_DC_COEFF_TOKEN_VLC_BITS 8
  245. #define CHROMA422_DC_COEFF_TOKEN_VLC_BITS 13
  246. #define COEFF_TOKEN_VLC_BITS 8
  247. #define TOTAL_ZEROS_VLC_BITS 9
  248. #define CHROMA_DC_TOTAL_ZEROS_VLC_BITS 3
  249. #define CHROMA422_DC_TOTAL_ZEROS_VLC_BITS 5
  250. #define RUN_VLC_BITS 3
  251. #define RUN7_VLC_BITS 6
  252. /**
  253. * Get the predicted number of non-zero coefficients.
  254. * @param n block index
  255. */
  256. static inline int pred_non_zero_count(H264Context *h, int n){
  257. const int index8= scan8[n];
  258. const int left= h->non_zero_count_cache[index8 - 1];
  259. const int top = h->non_zero_count_cache[index8 - 8];
  260. int i= left + top;
  261. if(i<64) i= (i+1)>>1;
  262. tprintf(h->s.avctx, "pred_nnz L%X T%X n%d s%d P%X\n", left, top, n, scan8[n], i&31);
  263. return i&31;
  264. }
  265. static av_cold void init_cavlc_level_tab(void){
  266. int suffix_length;
  267. unsigned int i;
  268. for(suffix_length=0; suffix_length<7; suffix_length++){
  269. for(i=0; i<(1<<LEVEL_TAB_BITS); i++){
  270. int prefix= LEVEL_TAB_BITS - av_log2(2*i);
  271. if(prefix + 1 + suffix_length <= LEVEL_TAB_BITS){
  272. int level_code = (prefix << suffix_length) +
  273. (i >> (av_log2(i) - suffix_length)) - (1 << suffix_length);
  274. int mask = -(level_code&1);
  275. level_code = (((2 + level_code) >> 1) ^ mask) - mask;
  276. cavlc_level_tab[suffix_length][i][0]= level_code;
  277. cavlc_level_tab[suffix_length][i][1]= prefix + 1 + suffix_length;
  278. }else if(prefix + 1 <= LEVEL_TAB_BITS){
  279. cavlc_level_tab[suffix_length][i][0]= prefix+100;
  280. cavlc_level_tab[suffix_length][i][1]= prefix + 1;
  281. }else{
  282. cavlc_level_tab[suffix_length][i][0]= LEVEL_TAB_BITS+100;
  283. cavlc_level_tab[suffix_length][i][1]= LEVEL_TAB_BITS;
  284. }
  285. }
  286. }
  287. }
  288. av_cold void ff_h264_decode_init_vlc(void){
  289. static int done = 0;
  290. if (!done) {
  291. int i;
  292. int offset;
  293. done = 1;
  294. chroma_dc_coeff_token_vlc.table = chroma_dc_coeff_token_vlc_table;
  295. chroma_dc_coeff_token_vlc.table_allocated = chroma_dc_coeff_token_vlc_table_size;
  296. init_vlc(&chroma_dc_coeff_token_vlc, CHROMA_DC_COEFF_TOKEN_VLC_BITS, 4*5,
  297. &chroma_dc_coeff_token_len [0], 1, 1,
  298. &chroma_dc_coeff_token_bits[0], 1, 1,
  299. INIT_VLC_USE_NEW_STATIC);
  300. chroma422_dc_coeff_token_vlc.table = chroma422_dc_coeff_token_vlc_table;
  301. chroma422_dc_coeff_token_vlc.table_allocated = chroma422_dc_coeff_token_vlc_table_size;
  302. init_vlc(&chroma422_dc_coeff_token_vlc, CHROMA422_DC_COEFF_TOKEN_VLC_BITS, 4*9,
  303. &chroma422_dc_coeff_token_len [0], 1, 1,
  304. &chroma422_dc_coeff_token_bits[0], 1, 1,
  305. INIT_VLC_USE_NEW_STATIC);
  306. offset = 0;
  307. for(i=0; i<4; i++){
  308. coeff_token_vlc[i].table = coeff_token_vlc_tables+offset;
  309. coeff_token_vlc[i].table_allocated = coeff_token_vlc_tables_size[i];
  310. init_vlc(&coeff_token_vlc[i], COEFF_TOKEN_VLC_BITS, 4*17,
  311. &coeff_token_len [i][0], 1, 1,
  312. &coeff_token_bits[i][0], 1, 1,
  313. INIT_VLC_USE_NEW_STATIC);
  314. offset += coeff_token_vlc_tables_size[i];
  315. }
  316. /*
  317. * This is a one time safety check to make sure that
  318. * the packed static coeff_token_vlc table sizes
  319. * were initialized correctly.
  320. */
  321. assert(offset == FF_ARRAY_ELEMS(coeff_token_vlc_tables));
  322. for(i=0; i<3; i++){
  323. chroma_dc_total_zeros_vlc[i].table = chroma_dc_total_zeros_vlc_tables[i];
  324. chroma_dc_total_zeros_vlc[i].table_allocated = chroma_dc_total_zeros_vlc_tables_size;
  325. init_vlc(&chroma_dc_total_zeros_vlc[i],
  326. CHROMA_DC_TOTAL_ZEROS_VLC_BITS, 4,
  327. &chroma_dc_total_zeros_len [i][0], 1, 1,
  328. &chroma_dc_total_zeros_bits[i][0], 1, 1,
  329. INIT_VLC_USE_NEW_STATIC);
  330. }
  331. for(i=0; i<7; i++){
  332. chroma422_dc_total_zeros_vlc[i].table = chroma422_dc_total_zeros_vlc_tables[i];
  333. chroma422_dc_total_zeros_vlc[i].table_allocated = chroma422_dc_total_zeros_vlc_tables_size;
  334. init_vlc(&chroma422_dc_total_zeros_vlc[i],
  335. CHROMA422_DC_TOTAL_ZEROS_VLC_BITS, 8,
  336. &chroma422_dc_total_zeros_len [i][0], 1, 1,
  337. &chroma422_dc_total_zeros_bits[i][0], 1, 1,
  338. INIT_VLC_USE_NEW_STATIC);
  339. }
  340. for(i=0; i<15; i++){
  341. total_zeros_vlc[i].table = total_zeros_vlc_tables[i];
  342. total_zeros_vlc[i].table_allocated = total_zeros_vlc_tables_size;
  343. init_vlc(&total_zeros_vlc[i],
  344. TOTAL_ZEROS_VLC_BITS, 16,
  345. &total_zeros_len [i][0], 1, 1,
  346. &total_zeros_bits[i][0], 1, 1,
  347. INIT_VLC_USE_NEW_STATIC);
  348. }
  349. for(i=0; i<6; i++){
  350. run_vlc[i].table = run_vlc_tables[i];
  351. run_vlc[i].table_allocated = run_vlc_tables_size;
  352. init_vlc(&run_vlc[i],
  353. RUN_VLC_BITS, 7,
  354. &run_len [i][0], 1, 1,
  355. &run_bits[i][0], 1, 1,
  356. INIT_VLC_USE_NEW_STATIC);
  357. }
  358. run7_vlc.table = run7_vlc_table,
  359. run7_vlc.table_allocated = run7_vlc_table_size;
  360. init_vlc(&run7_vlc, RUN7_VLC_BITS, 16,
  361. &run_len [6][0], 1, 1,
  362. &run_bits[6][0], 1, 1,
  363. INIT_VLC_USE_NEW_STATIC);
  364. init_cavlc_level_tab();
  365. }
  366. }
  367. /**
  368. *
  369. */
  370. static inline int get_level_prefix(GetBitContext *gb){
  371. unsigned int buf;
  372. int log;
  373. OPEN_READER(re, gb);
  374. UPDATE_CACHE(re, gb);
  375. buf=GET_CACHE(re, gb);
  376. log= 32 - av_log2(buf);
  377. #ifdef TRACE
  378. print_bin(buf>>(32-log), log);
  379. av_log(NULL, AV_LOG_DEBUG, "%5d %2d %3d lpr @%5d in %s get_level_prefix\n", buf>>(32-log), log, log-1, get_bits_count(gb), __FILE__);
  380. #endif
  381. LAST_SKIP_BITS(re, gb, log);
  382. CLOSE_READER(re, gb);
  383. return log-1;
  384. }
  385. /**
  386. * Decode a residual block.
  387. * @param n block index
  388. * @param scantable scantable
  389. * @param max_coeff number of coefficients in the block
  390. * @return <0 if an error occurred
  391. */
  392. static int decode_residual(H264Context *h, GetBitContext *gb, DCTELEM *block, int n, const uint8_t *scantable, const uint32_t *qmul, int max_coeff){
  393. MpegEncContext * const s = &h->s;
  394. static const int coeff_token_table_index[17]= {0, 0, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3};
  395. int level[16];
  396. int zeros_left, coeff_token, total_coeff, i, trailing_ones, run_before;
  397. //FIXME put trailing_onex into the context
  398. if(max_coeff <= 8){
  399. if (max_coeff == 4)
  400. coeff_token = get_vlc2(gb, chroma_dc_coeff_token_vlc.table, CHROMA_DC_COEFF_TOKEN_VLC_BITS, 1);
  401. else
  402. coeff_token = get_vlc2(gb, chroma422_dc_coeff_token_vlc.table, CHROMA422_DC_COEFF_TOKEN_VLC_BITS, 1);
  403. total_coeff= coeff_token>>2;
  404. }else{
  405. if(n >= LUMA_DC_BLOCK_INDEX){
  406. total_coeff= pred_non_zero_count(h, (n - LUMA_DC_BLOCK_INDEX)*16);
  407. coeff_token= get_vlc2(gb, coeff_token_vlc[ coeff_token_table_index[total_coeff] ].table, COEFF_TOKEN_VLC_BITS, 2);
  408. total_coeff= coeff_token>>2;
  409. }else{
  410. total_coeff= pred_non_zero_count(h, n);
  411. coeff_token= get_vlc2(gb, coeff_token_vlc[ coeff_token_table_index[total_coeff] ].table, COEFF_TOKEN_VLC_BITS, 2);
  412. total_coeff= coeff_token>>2;
  413. }
  414. }
  415. h->non_zero_count_cache[ scan8[n] ]= total_coeff;
  416. //FIXME set last_non_zero?
  417. if(total_coeff==0)
  418. return 0;
  419. if(total_coeff > (unsigned)max_coeff) {
  420. av_log(h->s.avctx, AV_LOG_ERROR, "corrupted macroblock %d %d (total_coeff=%d)\n", s->mb_x, s->mb_y, total_coeff);
  421. return -1;
  422. }
  423. trailing_ones= coeff_token&3;
  424. tprintf(h->s.avctx, "trailing:%d, total:%d\n", trailing_ones, total_coeff);
  425. assert(total_coeff<=16);
  426. i = show_bits(gb, 3);
  427. skip_bits(gb, trailing_ones);
  428. level[0] = 1-((i&4)>>1);
  429. level[1] = 1-((i&2) );
  430. level[2] = 1-((i&1)<<1);
  431. if(trailing_ones<total_coeff) {
  432. int mask, prefix;
  433. int suffix_length = total_coeff > 10 & trailing_ones < 3;
  434. int bitsi= show_bits(gb, LEVEL_TAB_BITS);
  435. int level_code= cavlc_level_tab[suffix_length][bitsi][0];
  436. skip_bits(gb, cavlc_level_tab[suffix_length][bitsi][1]);
  437. if(level_code >= 100){
  438. prefix= level_code - 100;
  439. if(prefix == LEVEL_TAB_BITS)
  440. prefix += get_level_prefix(gb);
  441. //first coefficient has suffix_length equal to 0 or 1
  442. if(prefix<14){ //FIXME try to build a large unified VLC table for all this
  443. if(suffix_length)
  444. level_code= (prefix<<1) + get_bits1(gb); //part
  445. else
  446. level_code= prefix; //part
  447. }else if(prefix==14){
  448. if(suffix_length)
  449. level_code= (prefix<<1) + get_bits1(gb); //part
  450. else
  451. level_code= prefix + get_bits(gb, 4); //part
  452. }else{
  453. level_code= 30 + get_bits(gb, prefix-3); //part
  454. if(prefix>=16){
  455. if(prefix > 25+3){
  456. av_log(h->s.avctx, AV_LOG_ERROR, "Invalid level prefix\n");
  457. return -1;
  458. }
  459. level_code += (1<<(prefix-3))-4096;
  460. }
  461. }
  462. if(trailing_ones < 3) level_code += 2;
  463. suffix_length = 2;
  464. mask= -(level_code&1);
  465. level[trailing_ones]= (((2+level_code)>>1) ^ mask) - mask;
  466. }else{
  467. level_code += ((level_code>>31)|1) & -(trailing_ones < 3);
  468. suffix_length = 1 + (level_code + 3U > 6U);
  469. level[trailing_ones]= level_code;
  470. }
  471. //remaining coefficients have suffix_length > 0
  472. for(i=trailing_ones+1;i<total_coeff;i++) {
  473. static const unsigned int suffix_limit[7] = {0,3,6,12,24,48,INT_MAX };
  474. int bitsi= show_bits(gb, LEVEL_TAB_BITS);
  475. level_code= cavlc_level_tab[suffix_length][bitsi][0];
  476. skip_bits(gb, cavlc_level_tab[suffix_length][bitsi][1]);
  477. if(level_code >= 100){
  478. prefix= level_code - 100;
  479. if(prefix == LEVEL_TAB_BITS){
  480. prefix += get_level_prefix(gb);
  481. }
  482. if(prefix<15){
  483. level_code = (prefix<<suffix_length) + get_bits(gb, suffix_length);
  484. }else{
  485. level_code = (15<<suffix_length) + get_bits(gb, prefix-3);
  486. if(prefix>=16)
  487. level_code += (1<<(prefix-3))-4096;
  488. }
  489. mask= -(level_code&1);
  490. level_code= (((2+level_code)>>1) ^ mask) - mask;
  491. }
  492. level[i]= level_code;
  493. suffix_length+= suffix_limit[suffix_length] + level_code > 2U*suffix_limit[suffix_length];
  494. }
  495. }
  496. if(total_coeff == max_coeff)
  497. zeros_left=0;
  498. else{
  499. if (max_coeff <= 8) {
  500. if (max_coeff == 4)
  501. zeros_left = get_vlc2(gb, chroma_dc_total_zeros_vlc[total_coeff - 1].table,
  502. CHROMA_DC_TOTAL_ZEROS_VLC_BITS, 1);
  503. else
  504. zeros_left = get_vlc2(gb, chroma422_dc_total_zeros_vlc[total_coeff - 1].table,
  505. CHROMA422_DC_TOTAL_ZEROS_VLC_BITS, 1);
  506. } else {
  507. zeros_left= get_vlc2(gb, total_zeros_vlc[total_coeff - 1].table, TOTAL_ZEROS_VLC_BITS, 1);
  508. }
  509. }
  510. #define STORE_BLOCK(type) \
  511. scantable += zeros_left + total_coeff - 1; \
  512. if(n >= LUMA_DC_BLOCK_INDEX){ \
  513. ((type*)block)[*scantable] = level[0]; \
  514. for(i=1;i<total_coeff && zeros_left > 0;i++) { \
  515. if(zeros_left < 7) \
  516. run_before= get_vlc2(gb, run_vlc[zeros_left - 1].table, RUN_VLC_BITS, 1); \
  517. else \
  518. run_before= get_vlc2(gb, run7_vlc.table, RUN7_VLC_BITS, 2); \
  519. zeros_left -= run_before; \
  520. scantable -= 1 + run_before; \
  521. ((type*)block)[*scantable]= level[i]; \
  522. } \
  523. for(;i<total_coeff;i++) { \
  524. scantable--; \
  525. ((type*)block)[*scantable]= level[i]; \
  526. } \
  527. }else{ \
  528. ((type*)block)[*scantable] = ((int)(level[0] * qmul[*scantable] + 32))>>6; \
  529. for(i=1;i<total_coeff && zeros_left > 0;i++) { \
  530. if(zeros_left < 7) \
  531. run_before= get_vlc2(gb, run_vlc[zeros_left - 1].table, RUN_VLC_BITS, 1); \
  532. else \
  533. run_before= get_vlc2(gb, run7_vlc.table, RUN7_VLC_BITS, 2); \
  534. zeros_left -= run_before; \
  535. scantable -= 1 + run_before; \
  536. ((type*)block)[*scantable]= ((int)(level[i] * qmul[*scantable] + 32))>>6; \
  537. } \
  538. for(;i<total_coeff;i++) { \
  539. scantable--; \
  540. ((type*)block)[*scantable]= ((int)(level[i] * qmul[*scantable] + 32))>>6; \
  541. } \
  542. }
  543. if (h->pixel_shift) {
  544. STORE_BLOCK(int32_t)
  545. } else {
  546. STORE_BLOCK(int16_t)
  547. }
  548. if(zeros_left<0){
  549. av_log(h->s.avctx, AV_LOG_ERROR, "negative number of zero coeffs at %d %d\n", s->mb_x, s->mb_y);
  550. return -1;
  551. }
  552. return 0;
  553. }
  554. static av_always_inline int decode_luma_residual(H264Context *h, GetBitContext *gb, const uint8_t *scan, const uint8_t *scan8x8, int pixel_shift, int mb_type, int cbp, int p){
  555. int i4x4, i8x8;
  556. MpegEncContext * const s = &h->s;
  557. int qscale = p == 0 ? s->qscale : h->chroma_qp[p-1];
  558. if(IS_INTRA16x16(mb_type)){
  559. AV_ZERO128(h->mb_luma_dc[p]+0);
  560. AV_ZERO128(h->mb_luma_dc[p]+8);
  561. AV_ZERO128(h->mb_luma_dc[p]+16);
  562. AV_ZERO128(h->mb_luma_dc[p]+24);
  563. if( decode_residual(h, h->intra_gb_ptr, h->mb_luma_dc[p], LUMA_DC_BLOCK_INDEX+p, scan, NULL, 16) < 0){
  564. return -1; //FIXME continue if partitioned and other return -1 too
  565. }
  566. assert((cbp&15) == 0 || (cbp&15) == 15);
  567. if(cbp&15){
  568. for(i8x8=0; i8x8<4; i8x8++){
  569. for(i4x4=0; i4x4<4; i4x4++){
  570. const int index= i4x4 + 4*i8x8 + p*16;
  571. if( decode_residual(h, h->intra_gb_ptr, h->mb + (16*index << pixel_shift),
  572. index, scan + 1, h->dequant4_coeff[p][qscale], 15) < 0 ){
  573. return -1;
  574. }
  575. }
  576. }
  577. return 0xf;
  578. }else{
  579. fill_rectangle(&h->non_zero_count_cache[scan8[p*16]], 4, 4, 8, 0, 1);
  580. return 0;
  581. }
  582. }else{
  583. int cqm = (IS_INTRA( mb_type ) ? 0:3)+p;
  584. /* For CAVLC 4:4:4, we need to keep track of the luma 8x8 CBP for deblocking nnz purposes. */
  585. int new_cbp = 0;
  586. for(i8x8=0; i8x8<4; i8x8++){
  587. if(cbp & (1<<i8x8)){
  588. if(IS_8x8DCT(mb_type)){
  589. DCTELEM *buf = &h->mb[64*i8x8+256*p << pixel_shift];
  590. uint8_t *nnz;
  591. for(i4x4=0; i4x4<4; i4x4++){
  592. const int index= i4x4 + 4*i8x8 + p*16;
  593. if( decode_residual(h, gb, buf, index, scan8x8+16*i4x4,
  594. h->dequant8_coeff[cqm][qscale], 16) < 0 )
  595. return -1;
  596. }
  597. nnz= &h->non_zero_count_cache[ scan8[4*i8x8+p*16] ];
  598. nnz[0] += nnz[1] + nnz[8] + nnz[9];
  599. new_cbp |= !!nnz[0] << i8x8;
  600. }else{
  601. for(i4x4=0; i4x4<4; i4x4++){
  602. const int index= i4x4 + 4*i8x8 + p*16;
  603. if( decode_residual(h, gb, h->mb + (16*index << pixel_shift), index,
  604. scan, h->dequant4_coeff[cqm][qscale], 16) < 0 ){
  605. return -1;
  606. }
  607. new_cbp |= h->non_zero_count_cache[ scan8[index] ] << i8x8;
  608. }
  609. }
  610. }else{
  611. uint8_t * const nnz= &h->non_zero_count_cache[ scan8[4*i8x8+p*16] ];
  612. nnz[0] = nnz[1] = nnz[8] = nnz[9] = 0;
  613. }
  614. }
  615. return new_cbp;
  616. }
  617. }
  618. int ff_h264_decode_mb_cavlc(H264Context *h){
  619. MpegEncContext * const s = &h->s;
  620. int mb_xy;
  621. int partition_count;
  622. unsigned int mb_type, cbp;
  623. int dct8x8_allowed= h->pps.transform_8x8_mode;
  624. int decode_chroma = h->sps.chroma_format_idc == 1 || h->sps.chroma_format_idc == 2;
  625. const int pixel_shift = h->pixel_shift;
  626. mb_xy = h->mb_xy = s->mb_x + s->mb_y*s->mb_stride;
  627. tprintf(s->avctx, "pic:%d mb:%d/%d\n", h->frame_num, s->mb_x, s->mb_y);
  628. cbp = 0; /* avoid warning. FIXME: find a solution without slowing
  629. down the code */
  630. if(h->slice_type_nos != AV_PICTURE_TYPE_I){
  631. if(s->mb_skip_run==-1)
  632. s->mb_skip_run= get_ue_golomb(&s->gb);
  633. if (s->mb_skip_run--) {
  634. if(FRAME_MBAFF && (s->mb_y&1) == 0){
  635. if(s->mb_skip_run==0)
  636. h->mb_mbaff = h->mb_field_decoding_flag = get_bits1(&s->gb);
  637. }
  638. decode_mb_skip(h);
  639. return 0;
  640. }
  641. }
  642. if(FRAME_MBAFF){
  643. if( (s->mb_y&1) == 0 )
  644. h->mb_mbaff = h->mb_field_decoding_flag = get_bits1(&s->gb);
  645. }
  646. h->prev_mb_skipped= 0;
  647. mb_type= get_ue_golomb(&s->gb);
  648. if(h->slice_type_nos == AV_PICTURE_TYPE_B){
  649. if(mb_type < 23){
  650. partition_count= b_mb_type_info[mb_type].partition_count;
  651. mb_type= b_mb_type_info[mb_type].type;
  652. }else{
  653. mb_type -= 23;
  654. goto decode_intra_mb;
  655. }
  656. }else if(h->slice_type_nos == AV_PICTURE_TYPE_P){
  657. if(mb_type < 5){
  658. partition_count= p_mb_type_info[mb_type].partition_count;
  659. mb_type= p_mb_type_info[mb_type].type;
  660. }else{
  661. mb_type -= 5;
  662. goto decode_intra_mb;
  663. }
  664. }else{
  665. assert(h->slice_type_nos == AV_PICTURE_TYPE_I);
  666. if(h->slice_type == AV_PICTURE_TYPE_SI && mb_type)
  667. mb_type--;
  668. decode_intra_mb:
  669. if(mb_type > 25){
  670. av_log(h->s.avctx, AV_LOG_ERROR, "mb_type %d in %c slice too large at %d %d\n", mb_type, av_get_picture_type_char(h->slice_type), s->mb_x, s->mb_y);
  671. return -1;
  672. }
  673. partition_count=0;
  674. cbp= i_mb_type_info[mb_type].cbp;
  675. h->intra16x16_pred_mode= i_mb_type_info[mb_type].pred_mode;
  676. mb_type= i_mb_type_info[mb_type].type;
  677. }
  678. if(MB_FIELD)
  679. mb_type |= MB_TYPE_INTERLACED;
  680. h->slice_table[ mb_xy ]= h->slice_num;
  681. if(IS_INTRA_PCM(mb_type)){
  682. unsigned int x;
  683. const int mb_size = ff_h264_mb_sizes[h->sps.chroma_format_idc] *
  684. h->sps.bit_depth_luma >> 3;
  685. // We assume these blocks are very rare so we do not optimize it.
  686. align_get_bits(&s->gb);
  687. // The pixels are stored in the same order as levels in h->mb array.
  688. for(x=0; x < mb_size; x++){
  689. ((uint8_t*)h->mb)[x]= get_bits(&s->gb, 8);
  690. }
  691. // In deblocking, the quantizer is 0
  692. s->current_picture.f.qscale_table[mb_xy] = 0;
  693. // All coeffs are present
  694. memset(h->non_zero_count[mb_xy], 16, 48);
  695. s->current_picture.f.mb_type[mb_xy] = mb_type;
  696. return 0;
  697. }
  698. fill_decode_neighbors(h, mb_type);
  699. fill_decode_caches(h, mb_type);
  700. //mb_pred
  701. if(IS_INTRA(mb_type)){
  702. int pred_mode;
  703. // init_top_left_availability(h);
  704. if(IS_INTRA4x4(mb_type)){
  705. int i;
  706. int di = 1;
  707. if(dct8x8_allowed && get_bits1(&s->gb)){
  708. mb_type |= MB_TYPE_8x8DCT;
  709. di = 4;
  710. }
  711. // fill_intra4x4_pred_table(h);
  712. for(i=0; i<16; i+=di){
  713. int mode= pred_intra_mode(h, i);
  714. if(!get_bits1(&s->gb)){
  715. const int rem_mode= get_bits(&s->gb, 3);
  716. mode = rem_mode + (rem_mode >= mode);
  717. }
  718. if(di==4)
  719. fill_rectangle( &h->intra4x4_pred_mode_cache[ scan8[i] ], 2, 2, 8, mode, 1 );
  720. else
  721. h->intra4x4_pred_mode_cache[ scan8[i] ] = mode;
  722. }
  723. write_back_intra_pred_mode(h);
  724. if( ff_h264_check_intra4x4_pred_mode(h) < 0)
  725. return -1;
  726. }else{
  727. h->intra16x16_pred_mode= ff_h264_check_intra_pred_mode(h, h->intra16x16_pred_mode, 0);
  728. if(h->intra16x16_pred_mode < 0)
  729. return -1;
  730. }
  731. if(decode_chroma){
  732. pred_mode= ff_h264_check_intra_pred_mode(h, get_ue_golomb_31(&s->gb), 1);
  733. if(pred_mode < 0)
  734. return -1;
  735. h->chroma_pred_mode= pred_mode;
  736. } else {
  737. h->chroma_pred_mode = DC_128_PRED8x8;
  738. }
  739. }else if(partition_count==4){
  740. int i, j, sub_partition_count[4], list, ref[2][4];
  741. if(h->slice_type_nos == AV_PICTURE_TYPE_B){
  742. for(i=0; i<4; i++){
  743. h->sub_mb_type[i]= get_ue_golomb_31(&s->gb);
  744. if(h->sub_mb_type[i] >=13){
  745. av_log(h->s.avctx, AV_LOG_ERROR, "B sub_mb_type %u out of range at %d %d\n", h->sub_mb_type[i], s->mb_x, s->mb_y);
  746. return -1;
  747. }
  748. sub_partition_count[i]= b_sub_mb_type_info[ h->sub_mb_type[i] ].partition_count;
  749. h->sub_mb_type[i]= b_sub_mb_type_info[ h->sub_mb_type[i] ].type;
  750. }
  751. if( IS_DIRECT(h->sub_mb_type[0]|h->sub_mb_type[1]|h->sub_mb_type[2]|h->sub_mb_type[3])) {
  752. ff_h264_pred_direct_motion(h, &mb_type);
  753. h->ref_cache[0][scan8[4]] =
  754. h->ref_cache[1][scan8[4]] =
  755. h->ref_cache[0][scan8[12]] =
  756. h->ref_cache[1][scan8[12]] = PART_NOT_AVAILABLE;
  757. }
  758. }else{
  759. assert(h->slice_type_nos == AV_PICTURE_TYPE_P); //FIXME SP correct ?
  760. for(i=0; i<4; i++){
  761. h->sub_mb_type[i]= get_ue_golomb_31(&s->gb);
  762. if(h->sub_mb_type[i] >=4){
  763. av_log(h->s.avctx, AV_LOG_ERROR, "P sub_mb_type %u out of range at %d %d\n", h->sub_mb_type[i], s->mb_x, s->mb_y);
  764. return -1;
  765. }
  766. sub_partition_count[i]= p_sub_mb_type_info[ h->sub_mb_type[i] ].partition_count;
  767. h->sub_mb_type[i]= p_sub_mb_type_info[ h->sub_mb_type[i] ].type;
  768. }
  769. }
  770. for(list=0; list<h->list_count; list++){
  771. int ref_count = IS_REF0(mb_type) ? 1 : h->ref_count[list] << MB_MBAFF;
  772. for(i=0; i<4; i++){
  773. if(IS_DIRECT(h->sub_mb_type[i])) continue;
  774. if(IS_DIR(h->sub_mb_type[i], 0, list)){
  775. unsigned int tmp;
  776. if(ref_count == 1){
  777. tmp= 0;
  778. }else if(ref_count == 2){
  779. tmp= get_bits1(&s->gb)^1;
  780. }else{
  781. tmp= get_ue_golomb_31(&s->gb);
  782. if(tmp>=ref_count){
  783. av_log(h->s.avctx, AV_LOG_ERROR, "ref %u overflow\n", tmp);
  784. return -1;
  785. }
  786. }
  787. ref[list][i]= tmp;
  788. }else{
  789. //FIXME
  790. ref[list][i] = -1;
  791. }
  792. }
  793. }
  794. if(dct8x8_allowed)
  795. dct8x8_allowed = get_dct8x8_allowed(h);
  796. for(list=0; list<h->list_count; list++){
  797. for(i=0; i<4; i++){
  798. if(IS_DIRECT(h->sub_mb_type[i])) {
  799. h->ref_cache[list][ scan8[4*i] ] = h->ref_cache[list][ scan8[4*i]+1 ];
  800. continue;
  801. }
  802. h->ref_cache[list][ scan8[4*i] ]=h->ref_cache[list][ scan8[4*i]+1 ]=
  803. h->ref_cache[list][ scan8[4*i]+8 ]=h->ref_cache[list][ scan8[4*i]+9 ]= ref[list][i];
  804. if(IS_DIR(h->sub_mb_type[i], 0, list)){
  805. const int sub_mb_type= h->sub_mb_type[i];
  806. const int block_width= (sub_mb_type & (MB_TYPE_16x16|MB_TYPE_16x8)) ? 2 : 1;
  807. for(j=0; j<sub_partition_count[i]; j++){
  808. int mx, my;
  809. const int index= 4*i + block_width*j;
  810. int16_t (* mv_cache)[2]= &h->mv_cache[list][ scan8[index] ];
  811. pred_motion(h, index, block_width, list, h->ref_cache[list][ scan8[index] ], &mx, &my);
  812. mx += get_se_golomb(&s->gb);
  813. my += get_se_golomb(&s->gb);
  814. tprintf(s->avctx, "final mv:%d %d\n", mx, my);
  815. if(IS_SUB_8X8(sub_mb_type)){
  816. mv_cache[ 1 ][0]=
  817. mv_cache[ 8 ][0]= mv_cache[ 9 ][0]= mx;
  818. mv_cache[ 1 ][1]=
  819. mv_cache[ 8 ][1]= mv_cache[ 9 ][1]= my;
  820. }else if(IS_SUB_8X4(sub_mb_type)){
  821. mv_cache[ 1 ][0]= mx;
  822. mv_cache[ 1 ][1]= my;
  823. }else if(IS_SUB_4X8(sub_mb_type)){
  824. mv_cache[ 8 ][0]= mx;
  825. mv_cache[ 8 ][1]= my;
  826. }
  827. mv_cache[ 0 ][0]= mx;
  828. mv_cache[ 0 ][1]= my;
  829. }
  830. }else{
  831. uint32_t *p= (uint32_t *)&h->mv_cache[list][ scan8[4*i] ][0];
  832. p[0] = p[1]=
  833. p[8] = p[9]= 0;
  834. }
  835. }
  836. }
  837. }else if(IS_DIRECT(mb_type)){
  838. ff_h264_pred_direct_motion(h, &mb_type);
  839. dct8x8_allowed &= h->sps.direct_8x8_inference_flag;
  840. }else{
  841. int list, mx, my, i;
  842. //FIXME we should set ref_idx_l? to 0 if we use that later ...
  843. if(IS_16X16(mb_type)){
  844. for(list=0; list<h->list_count; list++){
  845. unsigned int val;
  846. if(IS_DIR(mb_type, 0, list)){
  847. int rc = h->ref_count[list] << MB_MBAFF;
  848. if (rc == 1) {
  849. val= 0;
  850. } else if (rc == 2) {
  851. val= get_bits1(&s->gb)^1;
  852. }else{
  853. val= get_ue_golomb_31(&s->gb);
  854. if (val >= rc) {
  855. av_log(h->s.avctx, AV_LOG_ERROR, "ref %u overflow\n", val);
  856. return -1;
  857. }
  858. }
  859. fill_rectangle(&h->ref_cache[list][ scan8[0] ], 4, 4, 8, val, 1);
  860. }
  861. }
  862. for(list=0; list<h->list_count; list++){
  863. if(IS_DIR(mb_type, 0, list)){
  864. pred_motion(h, 0, 4, list, h->ref_cache[list][ scan8[0] ], &mx, &my);
  865. mx += get_se_golomb(&s->gb);
  866. my += get_se_golomb(&s->gb);
  867. tprintf(s->avctx, "final mv:%d %d\n", mx, my);
  868. fill_rectangle(h->mv_cache[list][ scan8[0] ], 4, 4, 8, pack16to32(mx,my), 4);
  869. }
  870. }
  871. }
  872. else if(IS_16X8(mb_type)){
  873. for(list=0; list<h->list_count; list++){
  874. for(i=0; i<2; i++){
  875. unsigned int val;
  876. if(IS_DIR(mb_type, i, list)){
  877. int rc = h->ref_count[list] << MB_MBAFF;
  878. if (rc == 1) {
  879. val= 0;
  880. } else if (rc == 2) {
  881. val= get_bits1(&s->gb)^1;
  882. }else{
  883. val= get_ue_golomb_31(&s->gb);
  884. if (val >= rc) {
  885. av_log(h->s.avctx, AV_LOG_ERROR, "ref %u overflow\n", val);
  886. return -1;
  887. }
  888. }
  889. }else
  890. val= LIST_NOT_USED&0xFF;
  891. fill_rectangle(&h->ref_cache[list][ scan8[0] + 16*i ], 4, 2, 8, val, 1);
  892. }
  893. }
  894. for(list=0; list<h->list_count; list++){
  895. for(i=0; i<2; i++){
  896. unsigned int val;
  897. if(IS_DIR(mb_type, i, list)){
  898. pred_16x8_motion(h, 8*i, list, h->ref_cache[list][scan8[0] + 16*i], &mx, &my);
  899. mx += get_se_golomb(&s->gb);
  900. my += get_se_golomb(&s->gb);
  901. tprintf(s->avctx, "final mv:%d %d\n", mx, my);
  902. val= pack16to32(mx,my);
  903. }else
  904. val=0;
  905. fill_rectangle(h->mv_cache[list][ scan8[0] + 16*i ], 4, 2, 8, val, 4);
  906. }
  907. }
  908. }else{
  909. assert(IS_8X16(mb_type));
  910. for(list=0; list<h->list_count; list++){
  911. for(i=0; i<2; i++){
  912. unsigned int val;
  913. if(IS_DIR(mb_type, i, list)){ //FIXME optimize
  914. int rc = h->ref_count[list] << MB_MBAFF;
  915. if (rc == 1) {
  916. val= 0;
  917. } else if (rc == 2) {
  918. val= get_bits1(&s->gb)^1;
  919. }else{
  920. val= get_ue_golomb_31(&s->gb);
  921. if (val >= rc) {
  922. av_log(h->s.avctx, AV_LOG_ERROR, "ref %u overflow\n", val);
  923. return -1;
  924. }
  925. }
  926. }else
  927. val= LIST_NOT_USED&0xFF;
  928. fill_rectangle(&h->ref_cache[list][ scan8[0] + 2*i ], 2, 4, 8, val, 1);
  929. }
  930. }
  931. for(list=0; list<h->list_count; list++){
  932. for(i=0; i<2; i++){
  933. unsigned int val;
  934. if(IS_DIR(mb_type, i, list)){
  935. pred_8x16_motion(h, i*4, list, h->ref_cache[list][ scan8[0] + 2*i ], &mx, &my);
  936. mx += get_se_golomb(&s->gb);
  937. my += get_se_golomb(&s->gb);
  938. tprintf(s->avctx, "final mv:%d %d\n", mx, my);
  939. val= pack16to32(mx,my);
  940. }else
  941. val=0;
  942. fill_rectangle(h->mv_cache[list][ scan8[0] + 2*i ], 2, 4, 8, val, 4);
  943. }
  944. }
  945. }
  946. }
  947. if(IS_INTER(mb_type))
  948. write_back_motion(h, mb_type);
  949. if(!IS_INTRA16x16(mb_type)){
  950. cbp= get_ue_golomb(&s->gb);
  951. if(decode_chroma){
  952. if(cbp > 47){
  953. av_log(h->s.avctx, AV_LOG_ERROR, "cbp too large (%u) at %d %d\n", cbp, s->mb_x, s->mb_y);
  954. return -1;
  955. }
  956. if(IS_INTRA4x4(mb_type)) cbp= golomb_to_intra4x4_cbp[cbp];
  957. else cbp= golomb_to_inter_cbp [cbp];
  958. }else{
  959. if(cbp > 15){
  960. av_log(h->s.avctx, AV_LOG_ERROR, "cbp too large (%u) at %d %d\n", cbp, s->mb_x, s->mb_y);
  961. return -1;
  962. }
  963. if(IS_INTRA4x4(mb_type)) cbp= golomb_to_intra4x4_cbp_gray[cbp];
  964. else cbp= golomb_to_inter_cbp_gray[cbp];
  965. }
  966. }
  967. if(dct8x8_allowed && (cbp&15) && !IS_INTRA(mb_type)){
  968. mb_type |= MB_TYPE_8x8DCT*get_bits1(&s->gb);
  969. }
  970. h->cbp=
  971. h->cbp_table[mb_xy]= cbp;
  972. s->current_picture.f.mb_type[mb_xy] = mb_type;
  973. if(cbp || IS_INTRA16x16(mb_type)){
  974. int i4x4, i8x8, chroma_idx;
  975. int dquant;
  976. int ret;
  977. GetBitContext *gb= IS_INTRA(mb_type) ? h->intra_gb_ptr : h->inter_gb_ptr;
  978. const uint8_t *scan, *scan8x8;
  979. const int max_qp = 51 + 6*(h->sps.bit_depth_luma-8);
  980. if(IS_INTERLACED(mb_type)){
  981. scan8x8= s->qscale ? h->field_scan8x8_cavlc : h->field_scan8x8_cavlc_q0;
  982. scan= s->qscale ? h->field_scan : h->field_scan_q0;
  983. }else{
  984. scan8x8= s->qscale ? h->zigzag_scan8x8_cavlc : h->zigzag_scan8x8_cavlc_q0;
  985. scan= s->qscale ? h->zigzag_scan : h->zigzag_scan_q0;
  986. }
  987. dquant= get_se_golomb(&s->gb);
  988. s->qscale += dquant;
  989. if(((unsigned)s->qscale) > max_qp){
  990. if(s->qscale<0) s->qscale+= max_qp+1;
  991. else s->qscale-= max_qp+1;
  992. if(((unsigned)s->qscale) > max_qp){
  993. av_log(h->s.avctx, AV_LOG_ERROR, "dquant out of range (%d) at %d %d\n", dquant, s->mb_x, s->mb_y);
  994. return -1;
  995. }
  996. }
  997. h->chroma_qp[0]= get_chroma_qp(h, 0, s->qscale);
  998. h->chroma_qp[1]= get_chroma_qp(h, 1, s->qscale);
  999. if( (ret = decode_luma_residual(h, gb, scan, scan8x8, pixel_shift, mb_type, cbp, 0)) < 0 ){
  1000. return -1;
  1001. }
  1002. h->cbp_table[mb_xy] |= ret << 12;
  1003. if(CHROMA444){
  1004. if( decode_luma_residual(h, gb, scan, scan8x8, pixel_shift, mb_type, cbp, 1) < 0 ){
  1005. return -1;
  1006. }
  1007. if( decode_luma_residual(h, gb, scan, scan8x8, pixel_shift, mb_type, cbp, 2) < 0 ){
  1008. return -1;
  1009. }
  1010. } else if (CHROMA422) {
  1011. if(cbp&0x30){
  1012. for(chroma_idx=0; chroma_idx<2; chroma_idx++)
  1013. if (decode_residual(h, gb, h->mb + ((256 + 16*16*chroma_idx) << pixel_shift),
  1014. CHROMA_DC_BLOCK_INDEX+chroma_idx, chroma422_dc_scan,
  1015. NULL, 8) < 0) {
  1016. return -1;
  1017. }
  1018. }
  1019. if(cbp&0x20){
  1020. for(chroma_idx=0; chroma_idx<2; chroma_idx++){
  1021. const uint32_t *qmul = h->dequant4_coeff[chroma_idx+1+(IS_INTRA( mb_type ) ? 0:3)][h->chroma_qp[chroma_idx]];
  1022. DCTELEM *mb = h->mb + (16*(16 + 16*chroma_idx) << pixel_shift);
  1023. for (i8x8 = 0; i8x8 < 2; i8x8++) {
  1024. for (i4x4 = 0; i4x4 < 4; i4x4++) {
  1025. const int index = 16 + 16*chroma_idx + 8*i8x8 + i4x4;
  1026. if (decode_residual(h, gb, mb, index, scan + 1, qmul, 15) < 0)
  1027. return -1;
  1028. mb += 16 << pixel_shift;
  1029. }
  1030. }
  1031. }
  1032. }else{
  1033. fill_rectangle(&h->non_zero_count_cache[scan8[16]], 4, 4, 8, 0, 1);
  1034. fill_rectangle(&h->non_zero_count_cache[scan8[32]], 4, 4, 8, 0, 1);
  1035. }
  1036. } else /* yuv420 */ {
  1037. if(cbp&0x30){
  1038. for(chroma_idx=0; chroma_idx<2; chroma_idx++)
  1039. if( decode_residual(h, gb, h->mb + ((256 + 16*16*chroma_idx) << pixel_shift), CHROMA_DC_BLOCK_INDEX+chroma_idx, chroma_dc_scan, NULL, 4) < 0){
  1040. return -1;
  1041. }
  1042. }
  1043. if(cbp&0x20){
  1044. for(chroma_idx=0; chroma_idx<2; chroma_idx++){
  1045. const uint32_t *qmul = h->dequant4_coeff[chroma_idx+1+(IS_INTRA( mb_type ) ? 0:3)][h->chroma_qp[chroma_idx]];
  1046. for(i4x4=0; i4x4<4; i4x4++){
  1047. const int index= 16 + 16*chroma_idx + i4x4;
  1048. if( decode_residual(h, gb, h->mb + (16*index << pixel_shift), index, scan + 1, qmul, 15) < 0){
  1049. return -1;
  1050. }
  1051. }
  1052. }
  1053. }else{
  1054. fill_rectangle(&h->non_zero_count_cache[scan8[16]], 4, 4, 8, 0, 1);
  1055. fill_rectangle(&h->non_zero_count_cache[scan8[32]], 4, 4, 8, 0, 1);
  1056. }
  1057. }
  1058. }else{
  1059. fill_rectangle(&h->non_zero_count_cache[scan8[ 0]], 4, 4, 8, 0, 1);
  1060. fill_rectangle(&h->non_zero_count_cache[scan8[16]], 4, 4, 8, 0, 1);
  1061. fill_rectangle(&h->non_zero_count_cache[scan8[32]], 4, 4, 8, 0, 1);
  1062. }
  1063. s->current_picture.f.qscale_table[mb_xy] = s->qscale;
  1064. write_back_non_zero_count(h);
  1065. return 0;
  1066. }