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.

1055 lines
30KB

  1. /*
  2. * FFV1 codec for libavcodec
  3. *
  4. * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
  5. *
  6. * This library is free software; you can redistribute it and/or
  7. * modify it under the terms of the GNU Lesser General Public
  8. * License as published by the Free Software Foundation; either
  9. * version 2 of the License, or (at your option) any later version.
  10. *
  11. * This library is distributed in the hope that it will be useful,
  12. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  14. * Lesser General Public License for more details.
  15. *
  16. * You should have received a copy of the GNU Lesser General Public
  17. * License along with this library; if not, write to the Free Software
  18. * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
  19. *
  20. */
  21. /**
  22. * @file ffv1.c
  23. * FF Video Codec 1 (an experimental lossless codec)
  24. */
  25. #include "common.h"
  26. #include "avcodec.h"
  27. #include "dsputil.h"
  28. #include "cabac.h"
  29. #include "golomb.h"
  30. #define MAX_PLANES 4
  31. #define CONTEXT_SIZE 32
  32. static const int8_t quant3[256]={
  33. 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  34. 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  35. 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  36. 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  37. 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  38. 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  39. 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  40. 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  41. -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  42. -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  43. -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  44. -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  45. -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  46. -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  47. -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  48. -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1, 0,
  49. };
  50. static const int8_t quant5[256]={
  51. 0, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  52. 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  53. 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  54. 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  55. 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  56. 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  57. 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  58. 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  59. -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
  60. -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
  61. -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
  62. -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
  63. -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
  64. -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
  65. -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
  66. -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-1,-1,-1,
  67. };
  68. static const int8_t quant7[256]={
  69. 0, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  70. 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  71. 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3,
  72. 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
  73. 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
  74. 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
  75. 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
  76. 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
  77. -3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,
  78. -3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,
  79. -3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,
  80. -3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,
  81. -3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,
  82. -3,-3,-3,-3,-3,-3,-3,-3,-3,-2,-2,-2,-2,-2,-2,-2,
  83. -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
  84. -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-1,-1,
  85. };
  86. static const int8_t quant9[256]={
  87. 0, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3,
  88. 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
  89. 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
  90. 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
  91. 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
  92. 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
  93. 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
  94. 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
  95. -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
  96. -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
  97. -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
  98. -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
  99. -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
  100. -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
  101. -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-3,-3,-3,-3,
  102. -3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-2,-2,-2,-2,-1,-1,
  103. };
  104. static const int8_t quant11[256]={
  105. 0, 1, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4,
  106. 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
  107. 4, 4, 4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
  108. 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
  109. 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
  110. 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
  111. 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
  112. 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
  113. -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
  114. -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
  115. -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
  116. -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
  117. -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
  118. -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-4,-4,
  119. -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
  120. -4,-4,-4,-4,-4,-3,-3,-3,-3,-3,-3,-3,-2,-2,-2,-1,
  121. };
  122. static const int8_t quant13[256]={
  123. 0, 1, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4,
  124. 4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
  125. 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
  126. 5, 5, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
  127. 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
  128. 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
  129. 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
  130. 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
  131. -6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,
  132. -6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,
  133. -6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,
  134. -6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,
  135. -6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-5,
  136. -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
  137. -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
  138. -4,-4,-4,-4,-4,-4,-4,-4,-4,-3,-3,-3,-3,-2,-2,-1,
  139. };
  140. static const uint8_t log2_run[32]={
  141. 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3,
  142. 4, 4, 5, 5, 6, 6, 7, 7,
  143. 8, 9,10,11,12,13,14,15,
  144. };
  145. typedef struct VlcState{
  146. int16_t drift;
  147. uint16_t error_sum;
  148. int8_t bias;
  149. uint8_t count;
  150. } VlcState;
  151. typedef struct PlaneContext{
  152. int context_count;
  153. uint8_t (*state)[CONTEXT_SIZE];
  154. VlcState *vlc_state;
  155. uint8_t interlace_bit_state[2];
  156. } PlaneContext;
  157. typedef struct FFV1Context{
  158. AVCodecContext *avctx;
  159. CABACContext c;
  160. GetBitContext gb;
  161. PutBitContext pb;
  162. int version;
  163. int width, height;
  164. int chroma_h_shift, chroma_v_shift;
  165. int flags;
  166. int picture_number;
  167. AVFrame picture;
  168. int plane_count;
  169. int ac; ///< 1-> CABAC 0-> golomb rice
  170. PlaneContext plane[MAX_PLANES];
  171. int16_t quant_table[5][256];
  172. int run_index;
  173. int colorspace;
  174. DSPContext dsp;
  175. }FFV1Context;
  176. static always_inline int fold(int diff, int bits){
  177. if(bits==8)
  178. diff= (int8_t)diff;
  179. else{
  180. diff+= 1<<(bits-1);
  181. diff&=(1<<bits)-1;
  182. diff-= 1<<(bits-1);
  183. }
  184. return diff;
  185. }
  186. static inline int predict(int_fast16_t *src, int_fast16_t *last){
  187. const int LT= last[-1];
  188. const int T= last[ 0];
  189. const int L = src[-1];
  190. return mid_pred(L, L + T - LT, T);
  191. }
  192. static inline int get_context(FFV1Context *f, int_fast16_t *src, int_fast16_t *last, int_fast16_t *last2){
  193. const int LT= last[-1];
  194. const int T= last[ 0];
  195. const int RT= last[ 1];
  196. const int L = src[-1];
  197. if(f->quant_table[3][127]){
  198. const int TT= last2[0];
  199. const int LL= src[-2];
  200. return f->quant_table[0][(L-LT) & 0xFF] + f->quant_table[1][(LT-T) & 0xFF] + f->quant_table[2][(T-RT) & 0xFF]
  201. +f->quant_table[3][(LL-L) & 0xFF] + f->quant_table[4][(TT-T) & 0xFF];
  202. }else
  203. return f->quant_table[0][(L-LT) & 0xFF] + f->quant_table[1][(LT-T) & 0xFF] + f->quant_table[2][(T-RT) & 0xFF];
  204. }
  205. /**
  206. * put
  207. */
  208. static inline void put_symbol(CABACContext *c, uint8_t *state, int v, int is_signed, int max_exp){
  209. int i;
  210. if(v){
  211. const int a= ABS(v);
  212. const int e= av_log2(a);
  213. put_cabac(c, state+0, 0);
  214. for(i=0; i<e; i++){
  215. put_cabac(c, state+1+i, 1); //1..8
  216. }
  217. if(e<max_exp){
  218. put_cabac(c, state+1+i, 0); //1..8
  219. for(i=e-1; i>=0; i--){
  220. put_cabac(c, state+16+e+i, (a>>i)&1); //17..29
  221. }
  222. if(is_signed)
  223. put_cabac(c, state+9 + e, v < 0); //9..16
  224. }
  225. }else{
  226. put_cabac(c, state+0, 1);
  227. }
  228. }
  229. static inline int get_symbol(CABACContext *c, uint8_t *state, int is_signed, int max_exp){
  230. if(get_cabac(c, state+0))
  231. return 0;
  232. else{
  233. int i, e;
  234. for(e=0; e<max_exp; e++){
  235. int a= 1<<e;
  236. if(get_cabac(c, state + 1 + e)==0){ // 1..8
  237. for(i=e-1; i>=0; i--){
  238. a += get_cabac(c, state+16+e+i)<<i; //17..29
  239. }
  240. if(is_signed && get_cabac(c, state+9 + e)) //9..16
  241. return -a;
  242. else
  243. return a;
  244. }
  245. }
  246. return -(1<<e);
  247. }
  248. }
  249. static inline void update_vlc_state(VlcState * const state, const int v){
  250. int drift= state->drift;
  251. int count= state->count;
  252. state->error_sum += ABS(v);
  253. drift += v;
  254. if(count == 128){ //FIXME variable
  255. count >>= 1;
  256. drift >>= 1;
  257. state->error_sum >>= 1;
  258. }
  259. count++;
  260. if(drift <= -count){
  261. if(state->bias > -128) state->bias--;
  262. drift += count;
  263. if(drift <= -count)
  264. drift= -count + 1;
  265. }else if(drift > 0){
  266. if(state->bias < 127) state->bias++;
  267. drift -= count;
  268. if(drift > 0)
  269. drift= 0;
  270. }
  271. state->drift= drift;
  272. state->count= count;
  273. }
  274. static inline void put_vlc_symbol(PutBitContext *pb, VlcState * const state, int v, int bits){
  275. int i, k, code;
  276. //printf("final: %d ", v);
  277. v = fold(v - state->bias, bits);
  278. i= state->count;
  279. k=0;
  280. while(i < state->error_sum){ //FIXME optimize
  281. k++;
  282. i += i;
  283. }
  284. assert(k<=8);
  285. #if 0 // JPEG LS
  286. if(k==0 && 2*state->drift <= - state->count) code= v ^ (-1);
  287. else code= v;
  288. #else
  289. code= v ^ ((2*state->drift + state->count)>>31);
  290. #endif
  291. code = -2*code-1;
  292. code^= (code>>31);
  293. //printf("v:%d/%d bias:%d error:%d drift:%d count:%d k:%d\n", v, code, state->bias, state->error_sum, state->drift, state->count, k);
  294. set_ur_golomb(pb, code, k, 12, bits);
  295. update_vlc_state(state, v);
  296. }
  297. static inline int get_vlc_symbol(GetBitContext *gb, VlcState * const state, int bits){
  298. int k, i, v, ret;
  299. i= state->count;
  300. k=0;
  301. while(i < state->error_sum){ //FIXME optimize
  302. k++;
  303. i += i;
  304. }
  305. assert(k<=8);
  306. v= get_ur_golomb(gb, k, 12, bits);
  307. //printf("v:%d bias:%d error:%d drift:%d count:%d k:%d", v, state->bias, state->error_sum, state->drift, state->count, k);
  308. v++;
  309. if(v&1) v= (v>>1);
  310. else v= -(v>>1);
  311. #if 0 // JPEG LS
  312. if(k==0 && 2*state->drift <= - state->count) v ^= (-1);
  313. #else
  314. v ^= ((2*state->drift + state->count)>>31);
  315. #endif
  316. ret= fold(v + state->bias, bits);
  317. update_vlc_state(state, v);
  318. //printf("final: %d\n", ret);
  319. return ret;
  320. }
  321. static always_inline void encode_line(FFV1Context *s, int w, int_fast16_t *sample[2], int plane_index, int bits){
  322. PlaneContext * const p= &s->plane[plane_index];
  323. CABACContext * const c= &s->c;
  324. int x;
  325. int run_index= s->run_index;
  326. int run_count=0;
  327. int run_mode=0;
  328. for(x=0; x<w; x++){
  329. int diff, context;
  330. context= get_context(s, sample[1]+x, sample[0]+x, sample[1]+x);
  331. diff= sample[1][x] - predict(sample[1]+x, sample[0]+x);
  332. if(context < 0){
  333. context = -context;
  334. diff= -diff;
  335. }
  336. diff= fold(diff, bits);
  337. if(s->ac){
  338. put_symbol(c, p->state[context], diff, 1, bits-1);
  339. }else{
  340. if(context == 0) run_mode=1;
  341. if(run_mode){
  342. if(diff){
  343. while(run_count >= 1<<log2_run[run_index]){
  344. run_count -= 1<<log2_run[run_index];
  345. run_index++;
  346. put_bits(&s->pb, 1, 1);
  347. }
  348. put_bits(&s->pb, 1 + log2_run[run_index], run_count);
  349. if(run_index) run_index--;
  350. run_count=0;
  351. run_mode=0;
  352. if(diff>0) diff--;
  353. }else{
  354. run_count++;
  355. }
  356. }
  357. // printf("count:%d index:%d, mode:%d, x:%d y:%d pos:%d\n", run_count, run_index, run_mode, x, y, (int)get_bit_count(&s->pb));
  358. if(run_mode == 0)
  359. put_vlc_symbol(&s->pb, &p->vlc_state[context], diff, bits);
  360. }
  361. }
  362. if(run_mode){
  363. while(run_count >= 1<<log2_run[run_index]){
  364. run_count -= 1<<log2_run[run_index];
  365. run_index++;
  366. put_bits(&s->pb, 1, 1);
  367. }
  368. if(run_count)
  369. put_bits(&s->pb, 1, 1);
  370. }
  371. s->run_index= run_index;
  372. }
  373. static void encode_plane(FFV1Context *s, uint8_t *src, int w, int h, int stride, int plane_index){
  374. int x,y;
  375. int_fast16_t sample_buffer[2][w+6];
  376. int_fast16_t *sample[2]= {sample_buffer[0]+3, sample_buffer[1]+3};
  377. s->run_index=0;
  378. memset(sample_buffer, 0, sizeof(sample_buffer));
  379. for(y=0; y<h; y++){
  380. int_fast16_t *temp= sample[0]; //FIXME try a normal buffer
  381. sample[0]= sample[1];
  382. sample[1]= temp;
  383. sample[1][-1]= sample[0][0 ];
  384. sample[0][ w]= sample[0][w-1];
  385. //{START_TIMER
  386. for(x=0; x<w; x++){
  387. sample[1][x]= src[x + stride*y];
  388. }
  389. encode_line(s, w, sample, plane_index, 8);
  390. //STOP_TIMER("encode line")}
  391. }
  392. }
  393. static void encode_rgb_frame(FFV1Context *s, uint32_t *src, int w, int h, int stride){
  394. int x, y, p;
  395. int_fast16_t sample_buffer[3][2][w+6];
  396. int_fast16_t *sample[3][2]= {
  397. {sample_buffer[0][0]+3, sample_buffer[0][1]+3},
  398. {sample_buffer[1][0]+3, sample_buffer[1][1]+3},
  399. {sample_buffer[2][0]+3, sample_buffer[2][1]+3}};
  400. s->run_index=0;
  401. memset(sample_buffer, 0, sizeof(sample_buffer));
  402. for(y=0; y<h; y++){
  403. for(x=0; x<w; x++){
  404. int v= src[x + stride*y];
  405. int b= v&0xFF;
  406. int g= (v>>8)&0xFF;
  407. int r= (v>>16)&0xFF;
  408. b -= g;
  409. r -= g;
  410. g += (b + r)>>2;
  411. b += 0x100;
  412. r += 0x100;
  413. // assert(g>=0 && b>=0 && r>=0);
  414. // assert(g<256 && b<512 && r<512);
  415. sample[0][0][x]= g;
  416. sample[1][0][x]= b;
  417. sample[2][0][x]= r;
  418. }
  419. for(p=0; p<3; p++){
  420. int_fast16_t *temp= sample[p][0]; //FIXME try a normal buffer
  421. sample[p][0]= sample[p][1];
  422. sample[p][1]= temp;
  423. sample[p][1][-1]= sample[p][0][0 ];
  424. sample[p][0][ w]= sample[p][0][w-1];
  425. encode_line(s, w, sample[p], FFMIN(p, 1), 9);
  426. }
  427. }
  428. }
  429. static void write_quant_table(CABACContext *c, int16_t *quant_table){
  430. int last=0;
  431. int i;
  432. uint8_t state[CONTEXT_SIZE]={0};
  433. for(i=1; i<128 ; i++){
  434. if(quant_table[i] != quant_table[i-1]){
  435. put_symbol(c, state, i-last-1, 0, 7);
  436. last= i;
  437. }
  438. }
  439. put_symbol(c, state, i-last-1, 0, 7);
  440. }
  441. static void write_header(FFV1Context *f){
  442. uint8_t state[CONTEXT_SIZE]={0};
  443. int i;
  444. CABACContext * const c= &f->c;
  445. put_symbol(c, state, f->version, 0, 7);
  446. put_symbol(c, state, f->avctx->coder_type, 0, 7);
  447. put_symbol(c, state, f->colorspace, 0, 7); //YUV cs type
  448. put_cabac(c, state, 1); //chroma planes
  449. put_symbol(c, state, f->chroma_h_shift, 0, 7);
  450. put_symbol(c, state, f->chroma_v_shift, 0, 7);
  451. put_cabac(c, state, 0); //no transparency plane
  452. for(i=0; i<5; i++)
  453. write_quant_table(c, f->quant_table[i]);
  454. }
  455. static int common_init(AVCodecContext *avctx){
  456. FFV1Context *s = avctx->priv_data;
  457. int width, height;
  458. s->avctx= avctx;
  459. s->flags= avctx->flags;
  460. dsputil_init(&s->dsp, avctx);
  461. width= s->width= avctx->width;
  462. height= s->height= avctx->height;
  463. assert(width && height);
  464. return 0;
  465. }
  466. static int encode_init(AVCodecContext *avctx)
  467. {
  468. FFV1Context *s = avctx->priv_data;
  469. int i;
  470. common_init(avctx);
  471. s->version=0;
  472. s->ac= avctx->coder_type;
  473. s->plane_count=2;
  474. for(i=0; i<256; i++){
  475. s->quant_table[0][i]= quant11[i];
  476. s->quant_table[1][i]= 11*quant11[i];
  477. if(avctx->context_model==0){
  478. s->quant_table[2][i]= 11*11*quant11[i];
  479. s->quant_table[3][i]=
  480. s->quant_table[4][i]=0;
  481. }else{
  482. s->quant_table[2][i]= 11*11*quant5 [i];
  483. s->quant_table[3][i]= 5*11*11*quant5 [i];
  484. s->quant_table[4][i]= 5*5*11*11*quant5 [i];
  485. }
  486. }
  487. for(i=0; i<s->plane_count; i++){
  488. PlaneContext * const p= &s->plane[i];
  489. if(avctx->context_model==0){
  490. p->context_count= (11*11*11+1)/2;
  491. }else{
  492. p->context_count= (11*11*5*5*5+1)/2;
  493. }
  494. if(s->ac){
  495. if(!p->state) p->state= av_malloc(CONTEXT_SIZE*p->context_count*sizeof(uint8_t));
  496. }else{
  497. if(!p->vlc_state) p->vlc_state= av_malloc(p->context_count*sizeof(VlcState));
  498. }
  499. }
  500. avctx->coded_frame= &s->picture;
  501. switch(avctx->pix_fmt){
  502. case PIX_FMT_YUV444P:
  503. case PIX_FMT_YUV422P:
  504. case PIX_FMT_YUV420P:
  505. case PIX_FMT_YUV411P:
  506. case PIX_FMT_YUV410P:
  507. s->colorspace= 0;
  508. break;
  509. case PIX_FMT_RGBA32:
  510. s->colorspace= 1;
  511. break;
  512. default:
  513. av_log(avctx, AV_LOG_ERROR, "format not supported\n");
  514. return -1;
  515. }
  516. avcodec_get_chroma_sub_sample(avctx->pix_fmt, &s->chroma_h_shift, &s->chroma_v_shift);
  517. s->picture_number=0;
  518. return 0;
  519. }
  520. static void clear_state(FFV1Context *f){
  521. int i, j;
  522. for(i=0; i<f->plane_count; i++){
  523. PlaneContext *p= &f->plane[i];
  524. p->interlace_bit_state[0]= 0;
  525. p->interlace_bit_state[1]= 0;
  526. for(j=0; j<p->context_count; j++){
  527. if(f->ac){
  528. memset(p->state[j], 0, sizeof(uint8_t)*CONTEXT_SIZE);
  529. p->state[j][7] = 2*62;
  530. }else{
  531. p->vlc_state[j].drift= 0;
  532. p->vlc_state[j].error_sum= 4; //FFMAX((RANGE + 32)/64, 2);
  533. p->vlc_state[j].bias= 0;
  534. p->vlc_state[j].count= 1;
  535. }
  536. }
  537. }
  538. }
  539. static int encode_frame(AVCodecContext *avctx, unsigned char *buf, int buf_size, void *data){
  540. FFV1Context *f = avctx->priv_data;
  541. CABACContext * const c= &f->c;
  542. AVFrame *pict = data;
  543. const int width= f->width;
  544. const int height= f->height;
  545. AVFrame * const p= &f->picture;
  546. int used_count= 0;
  547. if(avctx->strict_std_compliance >= 0){
  548. av_log(avctx, AV_LOG_ERROR, "this codec is under development, files encoded with it wont be decodeable with future versions!!!\n"
  549. "use vstrict=-1 to use it anyway\n");
  550. return -1;
  551. }
  552. ff_init_cabac_encoder(c, buf, buf_size);
  553. ff_init_cabac_states(c, ff_h264_lps_range, ff_h264_mps_state, ff_h264_lps_state, 64);
  554. *p = *pict;
  555. p->pict_type= FF_I_TYPE;
  556. if(avctx->gop_size==0 || f->picture_number % avctx->gop_size == 0){
  557. put_cabac_bypass(c, 1);
  558. p->key_frame= 1;
  559. write_header(f);
  560. clear_state(f);
  561. }else{
  562. put_cabac_bypass(c, 0);
  563. p->key_frame= 0;
  564. }
  565. if(!f->ac){
  566. used_count += put_cabac_terminate(c, 1);
  567. //printf("pos=%d\n", used_count);
  568. init_put_bits(&f->pb, buf + used_count, buf_size - used_count);
  569. }
  570. if(f->colorspace==0){
  571. const int chroma_width = -((-width )>>f->chroma_h_shift);
  572. const int chroma_height= -((-height)>>f->chroma_v_shift);
  573. encode_plane(f, p->data[0], width, height, p->linesize[0], 0);
  574. encode_plane(f, p->data[1], chroma_width, chroma_height, p->linesize[1], 1);
  575. encode_plane(f, p->data[2], chroma_width, chroma_height, p->linesize[2], 1);
  576. }else{
  577. encode_rgb_frame(f, (uint32_t*)(p->data[0]), width, height, p->linesize[0]/4);
  578. }
  579. emms_c();
  580. f->picture_number++;
  581. if(f->ac){
  582. return put_cabac_terminate(c, 1);
  583. }else{
  584. flush_put_bits(&f->pb); //nicer padding FIXME
  585. return used_count + (get_bit_count(&f->pb)+7)/8;
  586. }
  587. }
  588. static void common_end(FFV1Context *s){
  589. int i;
  590. for(i=0; i<s->plane_count; i++){
  591. PlaneContext *p= &s->plane[i];
  592. av_freep(&p->state);
  593. }
  594. }
  595. static int encode_end(AVCodecContext *avctx)
  596. {
  597. FFV1Context *s = avctx->priv_data;
  598. common_end(s);
  599. return 0;
  600. }
  601. static always_inline void decode_line(FFV1Context *s, int w, int_fast16_t *sample[2], int plane_index, int bits){
  602. PlaneContext * const p= &s->plane[plane_index];
  603. CABACContext * const c= &s->c;
  604. int x;
  605. int run_count=0;
  606. int run_mode=0;
  607. int run_index= s->run_index;
  608. for(x=0; x<w; x++){
  609. int diff, context, sign;
  610. context= get_context(s, sample[1] + x, sample[0] + x, sample[1] + x);
  611. if(context < 0){
  612. context= -context;
  613. sign=1;
  614. }else
  615. sign=0;
  616. if(s->ac)
  617. diff= get_symbol(c, p->state[context], 1, bits-1);
  618. else{
  619. if(context == 0 && run_mode==0) run_mode=1;
  620. if(run_mode){
  621. if(run_count==0 && run_mode==1){
  622. if(get_bits1(&s->gb)){
  623. run_count = 1<<log2_run[run_index];
  624. if(x + run_count <= w) run_index++;
  625. }else{
  626. if(log2_run[run_index]) run_count = get_bits(&s->gb, log2_run[run_index]);
  627. else run_count=0;
  628. if(run_index) run_index--;
  629. run_mode=2;
  630. }
  631. }
  632. run_count--;
  633. if(run_count < 0){
  634. run_mode=0;
  635. run_count=0;
  636. diff= get_vlc_symbol(&s->gb, &p->vlc_state[context], bits);
  637. if(diff>=0) diff++;
  638. }else
  639. diff=0;
  640. }else
  641. diff= get_vlc_symbol(&s->gb, &p->vlc_state[context], bits);
  642. // printf("count:%d index:%d, mode:%d, x:%d y:%d pos:%d\n", run_count, run_index, run_mode, x, y, get_bits_count(&s->gb));
  643. }
  644. if(sign) diff= -diff;
  645. sample[1][x]= (predict(sample[1] + x, sample[0] + x) + diff) & ((1<<bits)-1);
  646. }
  647. s->run_index= run_index;
  648. }
  649. static void decode_plane(FFV1Context *s, uint8_t *src, int w, int h, int stride, int plane_index){
  650. int x, y;
  651. int_fast16_t sample_buffer[2][w+6];
  652. int_fast16_t *sample[2]= {sample_buffer[0]+3, sample_buffer[1]+3};
  653. s->run_index=0;
  654. memset(sample_buffer, 0, sizeof(sample_buffer));
  655. for(y=0; y<h; y++){
  656. int_fast16_t *temp= sample[0]; //FIXME try a normal buffer
  657. sample[0]= sample[1];
  658. sample[1]= temp;
  659. sample[1][-1]= sample[0][0 ];
  660. sample[0][ w]= sample[0][w-1];
  661. //{START_TIMER
  662. decode_line(s, w, sample, plane_index, 8);
  663. for(x=0; x<w; x++){
  664. src[x + stride*y]= sample[1][x];
  665. }
  666. //STOP_TIMER("decode-line")}
  667. }
  668. }
  669. static void decode_rgb_frame(FFV1Context *s, uint32_t *src, int w, int h, int stride){
  670. int x, y, p;
  671. int_fast16_t sample_buffer[3][2][w+6];
  672. int_fast16_t *sample[3][2]= {
  673. {sample_buffer[0][0]+3, sample_buffer[0][1]+3},
  674. {sample_buffer[1][0]+3, sample_buffer[1][1]+3},
  675. {sample_buffer[2][0]+3, sample_buffer[2][1]+3}};
  676. s->run_index=0;
  677. memset(sample_buffer, 0, sizeof(sample_buffer));
  678. for(y=0; y<h; y++){
  679. for(p=0; p<3; p++){
  680. int_fast16_t *temp= sample[p][0]; //FIXME try a normal buffer
  681. sample[p][0]= sample[p][1];
  682. sample[p][1]= temp;
  683. sample[p][1][-1]= sample[p][0][0 ];
  684. sample[p][0][ w]= sample[p][0][w-1];
  685. decode_line(s, w, sample[p], FFMIN(p, 1), 9);
  686. }
  687. for(x=0; x<w; x++){
  688. int g= sample[0][1][x];
  689. int b= sample[1][1][x];
  690. int r= sample[2][1][x];
  691. // assert(g>=0 && b>=0 && r>=0);
  692. // assert(g<256 && b<512 && r<512);
  693. b -= 0x100;
  694. r -= 0x100;
  695. g -= (b + r)>>2;
  696. b += g;
  697. r += g;
  698. src[x + stride*y]= b + (g<<8) + (r<<16);
  699. }
  700. }
  701. }
  702. static int read_quant_table(CABACContext *c, int16_t *quant_table, int scale){
  703. int v;
  704. int i=0;
  705. uint8_t state[CONTEXT_SIZE]={0};
  706. for(v=0; i<128 ; v++){
  707. int len= get_symbol(c, state, 0, 7) + 1;
  708. if(len + i > 128) return -1;
  709. while(len--){
  710. quant_table[i] = scale*v;
  711. i++;
  712. //printf("%2d ",v);
  713. //if(i%16==0) printf("\n");
  714. }
  715. }
  716. for(i=1; i<128; i++){
  717. quant_table[256-i]= -quant_table[i];
  718. }
  719. quant_table[128]= -quant_table[127];
  720. return 2*v - 1;
  721. }
  722. static int read_header(FFV1Context *f){
  723. uint8_t state[CONTEXT_SIZE]={0};
  724. int i, context_count;
  725. CABACContext * const c= &f->c;
  726. f->version= get_symbol(c, state, 0, 7);
  727. f->ac= f->avctx->coder_type= get_symbol(c, state, 0, 7);
  728. f->colorspace= get_symbol(c, state, 0, 7); //YUV cs type
  729. get_cabac(c, state); //no chroma = false
  730. f->chroma_h_shift= get_symbol(c, state, 0, 7);
  731. f->chroma_v_shift= get_symbol(c, state, 0, 7);
  732. get_cabac(c, state); //transparency plane
  733. f->plane_count= 2;
  734. if(f->colorspace==0){
  735. switch(16*f->chroma_h_shift + f->chroma_v_shift){
  736. case 0x00: f->avctx->pix_fmt= PIX_FMT_YUV444P; break;
  737. case 0x10: f->avctx->pix_fmt= PIX_FMT_YUV422P; break;
  738. case 0x11: f->avctx->pix_fmt= PIX_FMT_YUV420P; break;
  739. case 0x20: f->avctx->pix_fmt= PIX_FMT_YUV411P; break;
  740. case 0x33: f->avctx->pix_fmt= PIX_FMT_YUV410P; break;
  741. default:
  742. av_log(f->avctx, AV_LOG_ERROR, "format not supported\n");
  743. return -1;
  744. }
  745. }else if(f->colorspace==1){
  746. if(f->chroma_h_shift || f->chroma_v_shift){
  747. av_log(f->avctx, AV_LOG_ERROR, "chroma subsampling not supported in this colorspace\n");
  748. return -1;
  749. }
  750. f->avctx->pix_fmt= PIX_FMT_RGBA32;
  751. }else{
  752. av_log(f->avctx, AV_LOG_ERROR, "colorspace not supported\n");
  753. return -1;
  754. }
  755. //printf("%d %d %d\n", f->chroma_h_shift, f->chroma_v_shift,f->avctx->pix_fmt);
  756. context_count=1;
  757. for(i=0; i<5; i++){
  758. context_count*= read_quant_table(c, f->quant_table[i], context_count);
  759. if(context_count < 0){
  760. av_log(f->avctx, AV_LOG_ERROR, "read_quant_table error\n");
  761. return -1;
  762. }
  763. }
  764. context_count= (context_count+1)/2;
  765. for(i=0; i<f->plane_count; i++){
  766. PlaneContext * const p= &f->plane[i];
  767. p->context_count= context_count;
  768. if(f->ac){
  769. if(!p->state) p->state= av_malloc(CONTEXT_SIZE*p->context_count*sizeof(uint8_t));
  770. }else{
  771. if(!p->vlc_state) p->vlc_state= av_malloc(p->context_count*sizeof(VlcState));
  772. }
  773. }
  774. return 0;
  775. }
  776. static int decode_init(AVCodecContext *avctx)
  777. {
  778. // FFV1Context *s = avctx->priv_data;
  779. common_init(avctx);
  780. return 0;
  781. }
  782. static int decode_frame(AVCodecContext *avctx, void *data, int *data_size, uint8_t *buf, int buf_size){
  783. FFV1Context *f = avctx->priv_data;
  784. CABACContext * const c= &f->c;
  785. const int width= f->width;
  786. const int height= f->height;
  787. AVFrame * const p= &f->picture;
  788. int bytes_read;
  789. AVFrame *picture = data;
  790. *data_size = 0;
  791. /* no supplementary picture */
  792. if (buf_size == 0)
  793. return 0;
  794. ff_init_cabac_decoder(c, buf, buf_size);
  795. ff_init_cabac_states(c, ff_h264_lps_range, ff_h264_mps_state, ff_h264_lps_state, 64);
  796. p->pict_type= FF_I_TYPE; //FIXME I vs. P
  797. if(get_cabac_bypass(c)){
  798. p->key_frame= 1;
  799. read_header(f);
  800. clear_state(f);
  801. }else{
  802. p->key_frame= 0;
  803. }
  804. p->reference= 0;
  805. if(avctx->get_buffer(avctx, p) < 0){
  806. av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
  807. return -1;
  808. }
  809. if(avctx->debug&FF_DEBUG_PICT_INFO)
  810. av_log(avctx, AV_LOG_ERROR, "keyframe:%d coder:%d\n", p->key_frame, f->ac);
  811. if(!f->ac){
  812. bytes_read = get_cabac_terminate(c);
  813. if(bytes_read ==0) av_log(avctx, AV_LOG_ERROR, "error at end of AC stream\n");
  814. //printf("pos=%d\n", bytes_read);
  815. init_get_bits(&f->gb, buf + bytes_read, buf_size - bytes_read);
  816. } else {
  817. bytes_read = 0; /* avoid warning */
  818. }
  819. if(f->colorspace==0){
  820. const int chroma_width = -((-width )>>f->chroma_h_shift);
  821. const int chroma_height= -((-height)>>f->chroma_v_shift);
  822. decode_plane(f, p->data[0], width, height, p->linesize[0], 0);
  823. decode_plane(f, p->data[1], chroma_width, chroma_height, p->linesize[1], 1);
  824. decode_plane(f, p->data[2], chroma_width, chroma_height, p->linesize[2], 1);
  825. }else{
  826. decode_rgb_frame(f, (uint32_t*)p->data[0], width, height, p->linesize[0]/4);
  827. }
  828. emms_c();
  829. f->picture_number++;
  830. *picture= *p;
  831. avctx->release_buffer(avctx, p); //FIXME
  832. *data_size = sizeof(AVFrame);
  833. if(f->ac){
  834. bytes_read= get_cabac_terminate(c);
  835. if(bytes_read ==0) av_log(f->avctx, AV_LOG_ERROR, "error at end of frame\n");
  836. }else{
  837. bytes_read+= (get_bits_count(&f->gb)+7)/8;
  838. }
  839. return bytes_read;
  840. }
  841. static int decode_end(AVCodecContext *avctx)
  842. {
  843. FFV1Context *s = avctx->priv_data;
  844. int i;
  845. if(avctx->get_buffer == avcodec_default_get_buffer){
  846. for(i=0; i<4; i++){
  847. av_freep(&s->picture.base[i]);
  848. s->picture.data[i]= NULL;
  849. }
  850. av_freep(&s->picture.opaque);
  851. }
  852. return 0;
  853. }
  854. AVCodec ffv1_decoder = {
  855. "ffv1",
  856. CODEC_TYPE_VIDEO,
  857. CODEC_ID_FFV1,
  858. sizeof(FFV1Context),
  859. decode_init,
  860. NULL,
  861. decode_end,
  862. decode_frame,
  863. CODEC_CAP_DR1 /*| CODEC_CAP_DRAW_HORIZ_BAND*/,
  864. NULL
  865. };
  866. #ifdef CONFIG_ENCODERS
  867. AVCodec ffv1_encoder = {
  868. "ffv1",
  869. CODEC_TYPE_VIDEO,
  870. CODEC_ID_FFV1,
  871. sizeof(FFV1Context),
  872. encode_init,
  873. encode_frame,
  874. encode_end,
  875. };
  876. #endif