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.

1171 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 FFmpeg.
  6. *
  7. * FFmpeg is free software; you can redistribute it and/or
  8. * modify it under the terms of the GNU Lesser General Public
  9. * License as published by the Free Software Foundation; either
  10. * version 2.1 of the License, or (at your option) any later version.
  11. *
  12. * FFmpeg is distributed in the hope that it will be useful,
  13. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  15. * Lesser General Public License for more details.
  16. *
  17. * You should have received a copy of the GNU Lesser General Public
  18. * License along with FFmpeg; if not, write to the Free Software
  19. * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  20. */
  21. /**
  22. * @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. * gets 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, mask;
  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. int level_code= (prefix<<suffix_length) + (i>>(LEVEL_TAB_BITS-prefix-1-suffix_length)) - (1<<suffix_length);
  272. mask= -(level_code&1);
  273. level_code= (((2+level_code)>>1) ^ mask) - mask;
  274. if(prefix + 1 + suffix_length <= LEVEL_TAB_BITS){
  275. cavlc_level_tab[suffix_length][i][0]= level_code;
  276. cavlc_level_tab[suffix_length][i][1]= prefix + 1 + suffix_length;
  277. }else if(prefix + 1 <= LEVEL_TAB_BITS){
  278. cavlc_level_tab[suffix_length][i][0]= prefix+100;
  279. cavlc_level_tab[suffix_length][i][1]= prefix + 1;
  280. }else{
  281. cavlc_level_tab[suffix_length][i][0]= LEVEL_TAB_BITS+100;
  282. cavlc_level_tab[suffix_length][i][1]= LEVEL_TAB_BITS;
  283. }
  284. }
  285. }
  286. }
  287. av_cold void ff_h264_decode_init_vlc(void){
  288. static int done = 0;
  289. if (!done) {
  290. int i;
  291. int offset;
  292. done = 1;
  293. chroma_dc_coeff_token_vlc.table = chroma_dc_coeff_token_vlc_table;
  294. chroma_dc_coeff_token_vlc.table_allocated = chroma_dc_coeff_token_vlc_table_size;
  295. init_vlc(&chroma_dc_coeff_token_vlc, CHROMA_DC_COEFF_TOKEN_VLC_BITS, 4*5,
  296. &chroma_dc_coeff_token_len [0], 1, 1,
  297. &chroma_dc_coeff_token_bits[0], 1, 1,
  298. INIT_VLC_USE_NEW_STATIC);
  299. chroma422_dc_coeff_token_vlc.table = chroma422_dc_coeff_token_vlc_table;
  300. chroma422_dc_coeff_token_vlc.table_allocated = chroma422_dc_coeff_token_vlc_table_size;
  301. init_vlc(&chroma422_dc_coeff_token_vlc, CHROMA422_DC_COEFF_TOKEN_VLC_BITS, 4*9,
  302. &chroma422_dc_coeff_token_len [0], 1, 1,
  303. &chroma422_dc_coeff_token_bits[0], 1, 1,
  304. INIT_VLC_USE_NEW_STATIC);
  305. offset = 0;
  306. for(i=0; i<4; i++){
  307. coeff_token_vlc[i].table = coeff_token_vlc_tables+offset;
  308. coeff_token_vlc[i].table_allocated = coeff_token_vlc_tables_size[i];
  309. init_vlc(&coeff_token_vlc[i], COEFF_TOKEN_VLC_BITS, 4*17,
  310. &coeff_token_len [i][0], 1, 1,
  311. &coeff_token_bits[i][0], 1, 1,
  312. INIT_VLC_USE_NEW_STATIC);
  313. offset += coeff_token_vlc_tables_size[i];
  314. }
  315. /*
  316. * This is a one time safety check to make sure that
  317. * the packed static coeff_token_vlc table sizes
  318. * were initialized correctly.
  319. */
  320. assert(offset == FF_ARRAY_ELEMS(coeff_token_vlc_tables));
  321. for(i=0; i<3; i++){
  322. chroma_dc_total_zeros_vlc[i].table = chroma_dc_total_zeros_vlc_tables[i];
  323. chroma_dc_total_zeros_vlc[i].table_allocated = chroma_dc_total_zeros_vlc_tables_size;
  324. init_vlc(&chroma_dc_total_zeros_vlc[i],
  325. CHROMA_DC_TOTAL_ZEROS_VLC_BITS, 4,
  326. &chroma_dc_total_zeros_len [i][0], 1, 1,
  327. &chroma_dc_total_zeros_bits[i][0], 1, 1,
  328. INIT_VLC_USE_NEW_STATIC);
  329. }
  330. for(i=0; i<7; i++){
  331. chroma422_dc_total_zeros_vlc[i].table = chroma422_dc_total_zeros_vlc_tables[i];
  332. chroma422_dc_total_zeros_vlc[i].table_allocated = chroma422_dc_total_zeros_vlc_tables_size;
  333. init_vlc(&chroma422_dc_total_zeros_vlc[i],
  334. CHROMA422_DC_TOTAL_ZEROS_VLC_BITS, 8,
  335. &chroma422_dc_total_zeros_len [i][0], 1, 1,
  336. &chroma422_dc_total_zeros_bits[i][0], 1, 1,
  337. INIT_VLC_USE_NEW_STATIC);
  338. }
  339. for(i=0; i<15; i++){
  340. total_zeros_vlc[i].table = total_zeros_vlc_tables[i];
  341. total_zeros_vlc[i].table_allocated = total_zeros_vlc_tables_size;
  342. init_vlc(&total_zeros_vlc[i],
  343. TOTAL_ZEROS_VLC_BITS, 16,
  344. &total_zeros_len [i][0], 1, 1,
  345. &total_zeros_bits[i][0], 1, 1,
  346. INIT_VLC_USE_NEW_STATIC);
  347. }
  348. for(i=0; i<6; i++){
  349. run_vlc[i].table = run_vlc_tables[i];
  350. run_vlc[i].table_allocated = run_vlc_tables_size;
  351. init_vlc(&run_vlc[i],
  352. RUN_VLC_BITS, 7,
  353. &run_len [i][0], 1, 1,
  354. &run_bits[i][0], 1, 1,
  355. INIT_VLC_USE_NEW_STATIC);
  356. }
  357. run7_vlc.table = run7_vlc_table,
  358. run7_vlc.table_allocated = run7_vlc_table_size;
  359. init_vlc(&run7_vlc, RUN7_VLC_BITS, 16,
  360. &run_len [6][0], 1, 1,
  361. &run_bits[6][0], 1, 1,
  362. INIT_VLC_USE_NEW_STATIC);
  363. init_cavlc_level_tab();
  364. }
  365. }
  366. /**
  367. *
  368. */
  369. static inline int get_level_prefix(GetBitContext *gb){
  370. unsigned int buf;
  371. int log;
  372. OPEN_READER(re, gb);
  373. UPDATE_CACHE(re, gb);
  374. buf=GET_CACHE(re, gb);
  375. log= 32 - av_log2(buf);
  376. #ifdef TRACE
  377. print_bin(buf>>(32-log), log);
  378. 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__);
  379. #endif
  380. LAST_SKIP_BITS(re, gb, log);
  381. CLOSE_READER(re, gb);
  382. return log-1;
  383. }
  384. /**
  385. * decodes a residual block.
  386. * @param n block index
  387. * @param scantable scantable
  388. * @param max_coeff number of coefficients in the block
  389. * @return <0 if an error occurred
  390. */
  391. static int decode_residual(H264Context *h, GetBitContext *gb, DCTELEM *block, int n, const uint8_t *scantable, const uint32_t *qmul, int max_coeff){
  392. MpegEncContext * const s = &h->s;
  393. 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};
  394. int level[16];
  395. int zeros_left, coeff_token, total_coeff, i, trailing_ones, run_before;
  396. //FIXME put trailing_onex into the context
  397. if(max_coeff <= 8){
  398. if (max_coeff == 4)
  399. coeff_token = get_vlc2(gb, chroma_dc_coeff_token_vlc.table, CHROMA_DC_COEFF_TOKEN_VLC_BITS, 1);
  400. else
  401. coeff_token = get_vlc2(gb, chroma422_dc_coeff_token_vlc.table, CHROMA422_DC_COEFF_TOKEN_VLC_BITS, 1);
  402. total_coeff= coeff_token>>2;
  403. }else{
  404. if(n >= LUMA_DC_BLOCK_INDEX){
  405. total_coeff= pred_non_zero_count(h, (n - LUMA_DC_BLOCK_INDEX)*16);
  406. coeff_token= get_vlc2(gb, coeff_token_vlc[ coeff_token_table_index[total_coeff] ].table, COEFF_TOKEN_VLC_BITS, 2);
  407. total_coeff= coeff_token>>2;
  408. }else{
  409. total_coeff= pred_non_zero_count(h, n);
  410. coeff_token= get_vlc2(gb, coeff_token_vlc[ coeff_token_table_index[total_coeff] ].table, COEFF_TOKEN_VLC_BITS, 2);
  411. total_coeff= coeff_token>>2;
  412. }
  413. }
  414. h->non_zero_count_cache[ scan8[n] ]= total_coeff;
  415. //FIXME set last_non_zero?
  416. if(total_coeff==0)
  417. return 0;
  418. if(total_coeff > (unsigned)max_coeff) {
  419. av_log(h->s.avctx, AV_LOG_ERROR, "corrupted macroblock %d %d (total_coeff=%d)\n", s->mb_x, s->mb_y, total_coeff);
  420. return -1;
  421. }
  422. trailing_ones= coeff_token&3;
  423. tprintf(h->s.avctx, "trailing:%d, total:%d\n", trailing_ones, total_coeff);
  424. assert(total_coeff<=16);
  425. i = show_bits(gb, 3);
  426. skip_bits(gb, trailing_ones);
  427. level[0] = 1-((i&4)>>1);
  428. level[1] = 1-((i&2) );
  429. level[2] = 1-((i&1)<<1);
  430. if(trailing_ones<total_coeff) {
  431. int mask, prefix;
  432. int suffix_length = total_coeff > 10 & trailing_ones < 3;
  433. int bitsi= show_bits(gb, LEVEL_TAB_BITS);
  434. int level_code= cavlc_level_tab[suffix_length][bitsi][0];
  435. skip_bits(gb, cavlc_level_tab[suffix_length][bitsi][1]);
  436. if(level_code >= 100){
  437. prefix= level_code - 100;
  438. if(prefix == LEVEL_TAB_BITS)
  439. prefix += get_level_prefix(gb);
  440. //first coefficient has suffix_length equal to 0 or 1
  441. if(prefix<14){ //FIXME try to build a large unified VLC table for all this
  442. if(suffix_length)
  443. level_code= (prefix<<1) + get_bits1(gb); //part
  444. else
  445. level_code= prefix; //part
  446. }else if(prefix==14){
  447. if(suffix_length)
  448. level_code= (prefix<<1) + get_bits1(gb); //part
  449. else
  450. level_code= prefix + get_bits(gb, 4); //part
  451. }else{
  452. level_code= 30 + get_bits(gb, prefix-3); //part
  453. if(prefix>=16){
  454. if(prefix > 25+3){
  455. av_log(h->s.avctx, AV_LOG_ERROR, "Invalid level prefix\n");
  456. return -1;
  457. }
  458. level_code += (1<<(prefix-3))-4096;
  459. }
  460. }
  461. if(trailing_ones < 3) level_code += 2;
  462. suffix_length = 2;
  463. mask= -(level_code&1);
  464. level[trailing_ones]= (((2+level_code)>>1) ^ mask) - mask;
  465. }else{
  466. level_code += ((level_code>>31)|1) & -(trailing_ones < 3);
  467. suffix_length = 1 + (level_code + 3U > 6U);
  468. level[trailing_ones]= level_code;
  469. }
  470. //remaining coefficients have suffix_length > 0
  471. for(i=trailing_ones+1;i<total_coeff;i++) {
  472. static const unsigned int suffix_limit[7] = {0,3,6,12,24,48,INT_MAX };
  473. int bitsi= show_bits(gb, LEVEL_TAB_BITS);
  474. level_code= cavlc_level_tab[suffix_length][bitsi][0];
  475. skip_bits(gb, cavlc_level_tab[suffix_length][bitsi][1]);
  476. if(level_code >= 100){
  477. prefix= level_code - 100;
  478. if(prefix == LEVEL_TAB_BITS){
  479. prefix += get_level_prefix(gb);
  480. }
  481. if(prefix<15){
  482. level_code = (prefix<<suffix_length) + get_bits(gb, suffix_length);
  483. }else{
  484. level_code = (15<<suffix_length) + get_bits(gb, prefix-3);
  485. if(prefix>=16)
  486. level_code += (1<<(prefix-3))-4096;
  487. }
  488. mask= -(level_code&1);
  489. level_code= (((2+level_code)>>1) ^ mask) - mask;
  490. }
  491. level[i]= level_code;
  492. suffix_length+= suffix_limit[suffix_length] + level_code > 2U*suffix_limit[suffix_length];
  493. }
  494. }
  495. if(total_coeff == max_coeff)
  496. zeros_left=0;
  497. else{
  498. if (max_coeff <= 8) {
  499. if (max_coeff == 4)
  500. zeros_left = get_vlc2(gb, (chroma_dc_total_zeros_vlc-1)[total_coeff].table,
  501. CHROMA_DC_TOTAL_ZEROS_VLC_BITS, 1);
  502. else
  503. zeros_left = get_vlc2(gb, (chroma422_dc_total_zeros_vlc-1)[total_coeff].table,
  504. CHROMA422_DC_TOTAL_ZEROS_VLC_BITS, 1);
  505. } else {
  506. zeros_left= get_vlc2(gb, (total_zeros_vlc-1)[ total_coeff ].table, TOTAL_ZEROS_VLC_BITS, 1);
  507. }
  508. }
  509. #define STORE_BLOCK(type) \
  510. scantable += zeros_left + total_coeff - 1; \
  511. if(n >= LUMA_DC_BLOCK_INDEX){ \
  512. ((type*)block)[*scantable] = level[0]; \
  513. for(i=1;i<total_coeff && zeros_left > 0;i++) { \
  514. if(zeros_left < 7) \
  515. run_before= get_vlc2(gb, (run_vlc-1)[zeros_left].table, RUN_VLC_BITS, 1); \
  516. else \
  517. run_before= get_vlc2(gb, run7_vlc.table, RUN7_VLC_BITS, 2); \
  518. zeros_left -= run_before; \
  519. scantable -= 1 + run_before; \
  520. ((type*)block)[*scantable]= level[i]; \
  521. } \
  522. for(;i<total_coeff;i++) { \
  523. scantable--; \
  524. ((type*)block)[*scantable]= level[i]; \
  525. } \
  526. }else{ \
  527. ((type*)block)[*scantable] = ((int)(level[0] * qmul[*scantable] + 32))>>6; \
  528. for(i=1;i<total_coeff && zeros_left > 0;i++) { \
  529. if(zeros_left < 7) \
  530. run_before= get_vlc2(gb, (run_vlc-1)[zeros_left].table, RUN_VLC_BITS, 1); \
  531. else \
  532. run_before= get_vlc2(gb, run7_vlc.table, RUN7_VLC_BITS, 2); \
  533. zeros_left -= run_before; \
  534. scantable -= 1 + run_before; \
  535. ((type*)block)[*scantable]= ((int)(level[i] * qmul[*scantable] + 32))>>6; \
  536. } \
  537. for(;i<total_coeff;i++) { \
  538. scantable--; \
  539. ((type*)block)[*scantable]= ((int)(level[i] * qmul[*scantable] + 32))>>6; \
  540. } \
  541. }
  542. if (h->pixel_shift) {
  543. STORE_BLOCK(int32_t)
  544. } else {
  545. STORE_BLOCK(int16_t)
  546. }
  547. if(zeros_left<0){
  548. av_log(h->s.avctx, AV_LOG_ERROR, "negative number of zero coeffs at %d %d\n", s->mb_x, s->mb_y);
  549. return -1;
  550. }
  551. return 0;
  552. }
  553. 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){
  554. int i4x4, i8x8;
  555. MpegEncContext * const s = &h->s;
  556. int qscale = p == 0 ? s->qscale : h->chroma_qp[p-1];
  557. if(IS_INTRA16x16(mb_type)){
  558. AV_ZERO128(h->mb_luma_dc[p]+0);
  559. AV_ZERO128(h->mb_luma_dc[p]+8);
  560. AV_ZERO128(h->mb_luma_dc[p]+16);
  561. AV_ZERO128(h->mb_luma_dc[p]+24);
  562. if( decode_residual(h, h->intra_gb_ptr, h->mb_luma_dc[p], LUMA_DC_BLOCK_INDEX+p, scan, NULL, 16) < 0){
  563. return -1; //FIXME continue if partitioned and other return -1 too
  564. }
  565. assert((cbp&15) == 0 || (cbp&15) == 15);
  566. if(cbp&15){
  567. for(i8x8=0; i8x8<4; i8x8++){
  568. for(i4x4=0; i4x4<4; i4x4++){
  569. const int index= i4x4 + 4*i8x8 + p*16;
  570. if( decode_residual(h, h->intra_gb_ptr, h->mb + (16*index << pixel_shift),
  571. index, scan + 1, h->dequant4_coeff[p][qscale], 15) < 0 ){
  572. return -1;
  573. }
  574. }
  575. }
  576. return 0xf;
  577. }else{
  578. fill_rectangle(&h->non_zero_count_cache[scan8[p*16]], 4, 4, 8, 0, 1);
  579. return 0;
  580. }
  581. }else{
  582. int cqm = (IS_INTRA( mb_type ) ? 0:3)+p;
  583. /* For CAVLC 4:4:4, we need to keep track of the luma 8x8 CBP for deblocking nnz purposes. */
  584. int new_cbp = 0;
  585. for(i8x8=0; i8x8<4; i8x8++){
  586. if(cbp & (1<<i8x8)){
  587. if(IS_8x8DCT(mb_type)){
  588. DCTELEM *buf = &h->mb[64*i8x8+256*p << pixel_shift];
  589. uint8_t *nnz;
  590. for(i4x4=0; i4x4<4; i4x4++){
  591. const int index= i4x4 + 4*i8x8 + p*16;
  592. if( decode_residual(h, gb, buf, index, scan8x8+16*i4x4,
  593. h->dequant8_coeff[cqm][qscale], 16) < 0 )
  594. return -1;
  595. }
  596. nnz= &h->non_zero_count_cache[ scan8[4*i8x8+p*16] ];
  597. nnz[0] += nnz[1] + nnz[8] + nnz[9];
  598. new_cbp |= !!nnz[0] << i8x8;
  599. }else{
  600. for(i4x4=0; i4x4<4; i4x4++){
  601. const int index= i4x4 + 4*i8x8 + p*16;
  602. if( decode_residual(h, gb, h->mb + (16*index << pixel_shift), index,
  603. scan, h->dequant4_coeff[cqm][qscale], 16) < 0 ){
  604. return -1;
  605. }
  606. new_cbp |= h->non_zero_count_cache[ scan8[index] ] << i8x8;
  607. }
  608. }
  609. }else{
  610. uint8_t * const nnz= &h->non_zero_count_cache[ scan8[4*i8x8+p*16] ];
  611. nnz[0] = nnz[1] = nnz[8] = nnz[9] = 0;
  612. }
  613. }
  614. return new_cbp;
  615. }
  616. }
  617. int ff_h264_decode_mb_cavlc(H264Context *h){
  618. MpegEncContext * const s = &h->s;
  619. int mb_xy;
  620. int partition_count;
  621. unsigned int mb_type, cbp;
  622. int dct8x8_allowed= h->pps.transform_8x8_mode;
  623. int decode_chroma = h->sps.chroma_format_idc == 1 || h->sps.chroma_format_idc == 2;
  624. const int pixel_shift = h->pixel_shift;
  625. mb_xy = h->mb_xy = s->mb_x + s->mb_y*s->mb_stride;
  626. tprintf(s->avctx, "pic:%d mb:%d/%d\n", h->frame_num, s->mb_x, s->mb_y);
  627. cbp = 0; /* avoid warning. FIXME: find a solution without slowing
  628. down the code */
  629. if(h->slice_type_nos != AV_PICTURE_TYPE_I){
  630. if(s->mb_skip_run==-1)
  631. s->mb_skip_run= get_ue_golomb(&s->gb);
  632. if (s->mb_skip_run--) {
  633. if(FRAME_MBAFF && (s->mb_y&1) == 0){
  634. if(s->mb_skip_run==0)
  635. h->mb_mbaff = h->mb_field_decoding_flag = get_bits1(&s->gb);
  636. }
  637. decode_mb_skip(h);
  638. return 0;
  639. }
  640. }
  641. if(FRAME_MBAFF){
  642. if( (s->mb_y&1) == 0 )
  643. h->mb_mbaff = h->mb_field_decoding_flag = get_bits1(&s->gb);
  644. }
  645. h->prev_mb_skipped= 0;
  646. mb_type= get_ue_golomb(&s->gb);
  647. if(h->slice_type_nos == AV_PICTURE_TYPE_B){
  648. if(mb_type < 23){
  649. partition_count= b_mb_type_info[mb_type].partition_count;
  650. mb_type= b_mb_type_info[mb_type].type;
  651. }else{
  652. mb_type -= 23;
  653. goto decode_intra_mb;
  654. }
  655. }else if(h->slice_type_nos == AV_PICTURE_TYPE_P){
  656. if(mb_type < 5){
  657. partition_count= p_mb_type_info[mb_type].partition_count;
  658. mb_type= p_mb_type_info[mb_type].type;
  659. }else{
  660. mb_type -= 5;
  661. goto decode_intra_mb;
  662. }
  663. }else{
  664. assert(h->slice_type_nos == AV_PICTURE_TYPE_I);
  665. if(h->slice_type == AV_PICTURE_TYPE_SI && mb_type)
  666. mb_type--;
  667. decode_intra_mb:
  668. if(mb_type > 25){
  669. 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);
  670. return -1;
  671. }
  672. partition_count=0;
  673. cbp= i_mb_type_info[mb_type].cbp;
  674. h->intra16x16_pred_mode= i_mb_type_info[mb_type].pred_mode;
  675. mb_type= i_mb_type_info[mb_type].type;
  676. }
  677. if(MB_FIELD)
  678. mb_type |= MB_TYPE_INTERLACED;
  679. h->slice_table[ mb_xy ]= h->slice_num;
  680. if(IS_INTRA_PCM(mb_type)){
  681. unsigned int x;
  682. static const uint16_t mb_sizes[4] = {256,384,512,768};
  683. const int mb_size = mb_sizes[h->sps.chroma_format_idc]*h->sps.bit_depth_luma >> 3;
  684. // We assume these blocks are very rare so we do not optimize it.
  685. align_get_bits(&s->gb);
  686. // The pixels are stored in the same order as levels in h->mb array.
  687. for(x=0; x < mb_size; x++){
  688. ((uint8_t*)h->mb)[x]= get_bits(&s->gb, 8);
  689. }
  690. // In deblocking, the quantizer is 0
  691. s->current_picture.f.qscale_table[mb_xy] = 0;
  692. // All coeffs are present
  693. memset(h->non_zero_count[mb_xy], 16, 48);
  694. s->current_picture.f.mb_type[mb_xy] = mb_type;
  695. return 0;
  696. }
  697. if(MB_MBAFF){
  698. h->ref_count[0] <<= 1;
  699. h->ref_count[1] <<= 1;
  700. }
  701. fill_decode_neighbors(h, mb_type);
  702. fill_decode_caches(h, mb_type);
  703. //mb_pred
  704. if(IS_INTRA(mb_type)){
  705. int pred_mode;
  706. // init_top_left_availability(h);
  707. if(IS_INTRA4x4(mb_type)){
  708. int i;
  709. int di = 1;
  710. if(dct8x8_allowed && get_bits1(&s->gb)){
  711. mb_type |= MB_TYPE_8x8DCT;
  712. di = 4;
  713. }
  714. // fill_intra4x4_pred_table(h);
  715. for(i=0; i<16; i+=di){
  716. int mode= pred_intra_mode(h, i);
  717. if(!get_bits1(&s->gb)){
  718. const int rem_mode= get_bits(&s->gb, 3);
  719. mode = rem_mode + (rem_mode >= mode);
  720. }
  721. if(di==4)
  722. fill_rectangle( &h->intra4x4_pred_mode_cache[ scan8[i] ], 2, 2, 8, mode, 1 );
  723. else
  724. h->intra4x4_pred_mode_cache[ scan8[i] ] = mode;
  725. }
  726. write_back_intra_pred_mode(h);
  727. if( ff_h264_check_intra4x4_pred_mode(h) < 0)
  728. return -1;
  729. }else{
  730. h->intra16x16_pred_mode= ff_h264_check_intra16x16_pred_mode(h, h->intra16x16_pred_mode);
  731. if(h->intra16x16_pred_mode < 0)
  732. return -1;
  733. }
  734. if(decode_chroma){
  735. pred_mode= ff_h264_check_intra_chroma_pred_mode(h, get_ue_golomb_31(&s->gb));
  736. if(pred_mode < 0)
  737. return -1;
  738. h->chroma_pred_mode= pred_mode;
  739. } else {
  740. h->chroma_pred_mode = DC_128_PRED8x8;
  741. }
  742. }else if(partition_count==4){
  743. int i, j, sub_partition_count[4], list, ref[2][4];
  744. if(h->slice_type_nos == AV_PICTURE_TYPE_B){
  745. for(i=0; i<4; i++){
  746. h->sub_mb_type[i]= get_ue_golomb_31(&s->gb);
  747. if(h->sub_mb_type[i] >=13){
  748. 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);
  749. return -1;
  750. }
  751. sub_partition_count[i]= b_sub_mb_type_info[ h->sub_mb_type[i] ].partition_count;
  752. h->sub_mb_type[i]= b_sub_mb_type_info[ h->sub_mb_type[i] ].type;
  753. }
  754. if( IS_DIRECT(h->sub_mb_type[0]|h->sub_mb_type[1]|h->sub_mb_type[2]|h->sub_mb_type[3])) {
  755. ff_h264_pred_direct_motion(h, &mb_type);
  756. h->ref_cache[0][scan8[4]] =
  757. h->ref_cache[1][scan8[4]] =
  758. h->ref_cache[0][scan8[12]] =
  759. h->ref_cache[1][scan8[12]] = PART_NOT_AVAILABLE;
  760. }
  761. }else{
  762. assert(h->slice_type_nos == AV_PICTURE_TYPE_P); //FIXME SP correct ?
  763. for(i=0; i<4; i++){
  764. h->sub_mb_type[i]= get_ue_golomb_31(&s->gb);
  765. if(h->sub_mb_type[i] >=4){
  766. 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);
  767. return -1;
  768. }
  769. sub_partition_count[i]= p_sub_mb_type_info[ h->sub_mb_type[i] ].partition_count;
  770. h->sub_mb_type[i]= p_sub_mb_type_info[ h->sub_mb_type[i] ].type;
  771. }
  772. }
  773. for(list=0; list<h->list_count; list++){
  774. int ref_count= IS_REF0(mb_type) ? 1 : h->ref_count[list];
  775. for(i=0; i<4; i++){
  776. if(IS_DIRECT(h->sub_mb_type[i])) continue;
  777. if(IS_DIR(h->sub_mb_type[i], 0, list)){
  778. unsigned int tmp;
  779. if(ref_count == 1){
  780. tmp= 0;
  781. }else if(ref_count == 2){
  782. tmp= get_bits1(&s->gb)^1;
  783. }else{
  784. tmp= get_ue_golomb_31(&s->gb);
  785. if(tmp>=ref_count){
  786. av_log(h->s.avctx, AV_LOG_ERROR, "ref %u overflow\n", tmp);
  787. return -1;
  788. }
  789. }
  790. ref[list][i]= tmp;
  791. }else{
  792. //FIXME
  793. ref[list][i] = -1;
  794. }
  795. }
  796. }
  797. if(dct8x8_allowed)
  798. dct8x8_allowed = get_dct8x8_allowed(h);
  799. for(list=0; list<h->list_count; list++){
  800. for(i=0; i<4; i++){
  801. if(IS_DIRECT(h->sub_mb_type[i])) {
  802. h->ref_cache[list][ scan8[4*i] ] = h->ref_cache[list][ scan8[4*i]+1 ];
  803. continue;
  804. }
  805. h->ref_cache[list][ scan8[4*i] ]=h->ref_cache[list][ scan8[4*i]+1 ]=
  806. h->ref_cache[list][ scan8[4*i]+8 ]=h->ref_cache[list][ scan8[4*i]+9 ]= ref[list][i];
  807. if(IS_DIR(h->sub_mb_type[i], 0, list)){
  808. const int sub_mb_type= h->sub_mb_type[i];
  809. const int block_width= (sub_mb_type & (MB_TYPE_16x16|MB_TYPE_16x8)) ? 2 : 1;
  810. for(j=0; j<sub_partition_count[i]; j++){
  811. int mx, my;
  812. const int index= 4*i + block_width*j;
  813. int16_t (* mv_cache)[2]= &h->mv_cache[list][ scan8[index] ];
  814. pred_motion(h, index, block_width, list, h->ref_cache[list][ scan8[index] ], &mx, &my);
  815. mx += get_se_golomb(&s->gb);
  816. my += get_se_golomb(&s->gb);
  817. tprintf(s->avctx, "final mv:%d %d\n", mx, my);
  818. if(IS_SUB_8X8(sub_mb_type)){
  819. mv_cache[ 1 ][0]=
  820. mv_cache[ 8 ][0]= mv_cache[ 9 ][0]= mx;
  821. mv_cache[ 1 ][1]=
  822. mv_cache[ 8 ][1]= mv_cache[ 9 ][1]= my;
  823. }else if(IS_SUB_8X4(sub_mb_type)){
  824. mv_cache[ 1 ][0]= mx;
  825. mv_cache[ 1 ][1]= my;
  826. }else if(IS_SUB_4X8(sub_mb_type)){
  827. mv_cache[ 8 ][0]= mx;
  828. mv_cache[ 8 ][1]= my;
  829. }
  830. mv_cache[ 0 ][0]= mx;
  831. mv_cache[ 0 ][1]= my;
  832. }
  833. }else{
  834. uint32_t *p= (uint32_t *)&h->mv_cache[list][ scan8[4*i] ][0];
  835. p[0] = p[1]=
  836. p[8] = p[9]= 0;
  837. }
  838. }
  839. }
  840. }else if(IS_DIRECT(mb_type)){
  841. ff_h264_pred_direct_motion(h, &mb_type);
  842. dct8x8_allowed &= h->sps.direct_8x8_inference_flag;
  843. }else{
  844. int list, mx, my, i;
  845. //FIXME we should set ref_idx_l? to 0 if we use that later ...
  846. if(IS_16X16(mb_type)){
  847. for(list=0; list<h->list_count; list++){
  848. unsigned int val;
  849. if(IS_DIR(mb_type, 0, list)){
  850. if(h->ref_count[list]==1){
  851. val= 0;
  852. }else if(h->ref_count[list]==2){
  853. val= get_bits1(&s->gb)^1;
  854. }else{
  855. val= get_ue_golomb_31(&s->gb);
  856. if(val >= h->ref_count[list]){
  857. av_log(h->s.avctx, AV_LOG_ERROR, "ref %u overflow\n", val);
  858. return -1;
  859. }
  860. }
  861. fill_rectangle(&h->ref_cache[list][ scan8[0] ], 4, 4, 8, val, 1);
  862. }
  863. }
  864. for(list=0; list<h->list_count; list++){
  865. if(IS_DIR(mb_type, 0, list)){
  866. pred_motion(h, 0, 4, list, h->ref_cache[list][ scan8[0] ], &mx, &my);
  867. mx += get_se_golomb(&s->gb);
  868. my += get_se_golomb(&s->gb);
  869. tprintf(s->avctx, "final mv:%d %d\n", mx, my);
  870. fill_rectangle(h->mv_cache[list][ scan8[0] ], 4, 4, 8, pack16to32(mx,my), 4);
  871. }
  872. }
  873. }
  874. else if(IS_16X8(mb_type)){
  875. for(list=0; list<h->list_count; list++){
  876. for(i=0; i<2; i++){
  877. unsigned int val;
  878. if(IS_DIR(mb_type, i, list)){
  879. if(h->ref_count[list] == 1){
  880. val= 0;
  881. }else if(h->ref_count[list] == 2){
  882. val= get_bits1(&s->gb)^1;
  883. }else{
  884. val= get_ue_golomb_31(&s->gb);
  885. if(val >= h->ref_count[list]){
  886. av_log(h->s.avctx, AV_LOG_ERROR, "ref %u overflow\n", val);
  887. return -1;
  888. }
  889. }
  890. }else
  891. val= LIST_NOT_USED&0xFF;
  892. fill_rectangle(&h->ref_cache[list][ scan8[0] + 16*i ], 4, 2, 8, val, 1);
  893. }
  894. }
  895. for(list=0; list<h->list_count; list++){
  896. for(i=0; i<2; i++){
  897. unsigned int val;
  898. if(IS_DIR(mb_type, i, list)){
  899. pred_16x8_motion(h, 8*i, list, h->ref_cache[list][scan8[0] + 16*i], &mx, &my);
  900. mx += get_se_golomb(&s->gb);
  901. my += get_se_golomb(&s->gb);
  902. tprintf(s->avctx, "final mv:%d %d\n", mx, my);
  903. val= pack16to32(mx,my);
  904. }else
  905. val=0;
  906. fill_rectangle(h->mv_cache[list][ scan8[0] + 16*i ], 4, 2, 8, val, 4);
  907. }
  908. }
  909. }else{
  910. assert(IS_8X16(mb_type));
  911. for(list=0; list<h->list_count; list++){
  912. for(i=0; i<2; i++){
  913. unsigned int val;
  914. if(IS_DIR(mb_type, i, list)){ //FIXME optimize
  915. if(h->ref_count[list]==1){
  916. val= 0;
  917. }else if(h->ref_count[list]==2){
  918. val= get_bits1(&s->gb)^1;
  919. }else{
  920. val= get_ue_golomb_31(&s->gb);
  921. if(val >= h->ref_count[list]){
  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 {
  1011. const int num_c8x8 = h->sps.chroma_format_idc;
  1012. if(cbp&0x30){
  1013. for(chroma_idx=0; chroma_idx<2; chroma_idx++)
  1014. if (decode_residual(h, gb, h->mb + ((256 + 16*16*chroma_idx) << pixel_shift),
  1015. CHROMA_DC_BLOCK_INDEX+chroma_idx,
  1016. CHROMA422 ? chroma422_dc_scan : chroma_dc_scan,
  1017. NULL, 4*num_c8x8) < 0) {
  1018. return -1;
  1019. }
  1020. }
  1021. if(cbp&0x20){
  1022. for(chroma_idx=0; chroma_idx<2; chroma_idx++){
  1023. const uint32_t *qmul = h->dequant4_coeff[chroma_idx+1+(IS_INTRA( mb_type ) ? 0:3)][h->chroma_qp[chroma_idx]];
  1024. DCTELEM *mb = h->mb + (16*(16 + 16*chroma_idx) << pixel_shift);
  1025. for (i8x8=0; i8x8<num_c8x8; i8x8++) {
  1026. for (i4x4=0; i4x4<4; i4x4++) {
  1027. const int index= 16 + 16*chroma_idx + 8*i8x8 + i4x4;
  1028. if (decode_residual(h, gb, mb, index, scan + 1, qmul, 15) < 0)
  1029. return -1;
  1030. mb += 16<<pixel_shift;
  1031. }
  1032. }
  1033. }
  1034. }else{
  1035. fill_rectangle(&h->non_zero_count_cache[scan8[16]], 4, 4, 8, 0, 1);
  1036. fill_rectangle(&h->non_zero_count_cache[scan8[32]], 4, 4, 8, 0, 1);
  1037. }
  1038. }
  1039. }else{
  1040. fill_rectangle(&h->non_zero_count_cache[scan8[ 0]], 4, 4, 8, 0, 1);
  1041. fill_rectangle(&h->non_zero_count_cache[scan8[16]], 4, 4, 8, 0, 1);
  1042. fill_rectangle(&h->non_zero_count_cache[scan8[32]], 4, 4, 8, 0, 1);
  1043. }
  1044. s->current_picture.f.qscale_table[mb_xy] = s->qscale;
  1045. write_back_non_zero_count(h);
  1046. if(MB_MBAFF){
  1047. h->ref_count[0] >>= 1;
  1048. h->ref_count[1] >>= 1;
  1049. }
  1050. return 0;
  1051. }