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.

614 lines
25KB

  1. /*
  2. * DSP utils
  3. * Copyright (c) 2000, 2001 Gerard Lantau.
  4. *
  5. * This program is free software; you can redistribute it and/or modify
  6. * it under the terms of the GNU General Public License as published by
  7. * the Free Software Foundation; either version 2 of the License, or
  8. * (at your option) any later version.
  9. *
  10. * This program is distributed in the hope that it will be useful,
  11. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  13. * GNU General Public License for more details.
  14. *
  15. * You should have received a copy of the GNU General Public License
  16. * along with this program; if not, write to the Free Software
  17. * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  18. */
  19. #include <stdlib.h>
  20. #include <stdio.h>
  21. #include <math.h>
  22. #include "avcodec.h"
  23. #include "dsputil.h"
  24. #include "simple_idct.h"
  25. void (*ff_idct)(DCTELEM *block);
  26. void (*get_pixels)(DCTELEM *block, const UINT8 *pixels, int line_size);
  27. void (*put_pixels_clamped)(const DCTELEM *block, UINT8 *pixels, int line_size);
  28. void (*add_pixels_clamped)(const DCTELEM *block, UINT8 *pixels, int line_size);
  29. op_pixels_abs_func pix_abs16x16;
  30. op_pixels_abs_func pix_abs16x16_x2;
  31. op_pixels_abs_func pix_abs16x16_y2;
  32. op_pixels_abs_func pix_abs16x16_xy2;
  33. UINT8 cropTbl[256 + 2 * MAX_NEG_CROP];
  34. UINT32 squareTbl[512];
  35. extern UINT16 default_intra_matrix[64];
  36. extern UINT16 default_non_intra_matrix[64];
  37. UINT8 zigzag_direct[64] = {
  38. 0, 1, 8, 16, 9, 2, 3, 10,
  39. 17, 24, 32, 25, 18, 11, 4, 5,
  40. 12, 19, 26, 33, 40, 48, 41, 34,
  41. 27, 20, 13, 6, 7, 14, 21, 28,
  42. 35, 42, 49, 56, 57, 50, 43, 36,
  43. 29, 22, 15, 23, 30, 37, 44, 51,
  44. 58, 59, 52, 45, 38, 31, 39, 46,
  45. 53, 60, 61, 54, 47, 55, 62, 63
  46. };
  47. /* not permutated inverse zigzag_direct + 1 for MMX quantizer */
  48. UINT16 __align8 inv_zigzag_direct16[64];
  49. /* not permutated zigzag_direct for MMX quantizer */
  50. UINT8 zigzag_direct_noperm[64];
  51. UINT8 ff_alternate_horizontal_scan[64] = {
  52. 0, 1, 2, 3, 8, 9, 16, 17,
  53. 10, 11, 4, 5, 6, 7, 15, 14,
  54. 13, 12, 19, 18, 24, 25, 32, 33,
  55. 26, 27, 20, 21, 22, 23, 28, 29,
  56. 30, 31, 34, 35, 40, 41, 48, 49,
  57. 42, 43, 36, 37, 38, 39, 44, 45,
  58. 46, 47, 50, 51, 56, 57, 58, 59,
  59. 52, 53, 54, 55, 60, 61, 62, 63,
  60. };
  61. UINT8 ff_alternate_vertical_scan[64] = {
  62. 0, 8, 16, 24, 1, 9, 2, 10,
  63. 17, 25, 32, 40, 48, 56, 57, 49,
  64. 41, 33, 26, 18, 3, 11, 4, 12,
  65. 19, 27, 34, 42, 50, 58, 35, 43,
  66. 51, 59, 20, 28, 5, 13, 6, 14,
  67. 21, 29, 36, 44, 52, 60, 37, 45,
  68. 53, 61, 22, 30, 7, 15, 23, 31,
  69. 38, 46, 54, 62, 39, 47, 55, 63,
  70. };
  71. /* Input permutation for the simple_idct_mmx */
  72. static UINT8 simple_mmx_permutation[64]={
  73. 0x00, 0x08, 0x04, 0x09, 0x01, 0x0C, 0x05, 0x0D,
  74. 0x10, 0x18, 0x14, 0x19, 0x11, 0x1C, 0x15, 0x1D,
  75. 0x20, 0x28, 0x24, 0x29, 0x21, 0x2C, 0x25, 0x2D,
  76. 0x12, 0x1A, 0x16, 0x1B, 0x13, 0x1E, 0x17, 0x1F,
  77. 0x02, 0x0A, 0x06, 0x0B, 0x03, 0x0E, 0x07, 0x0F,
  78. 0x30, 0x38, 0x34, 0x39, 0x31, 0x3C, 0x35, 0x3D,
  79. 0x22, 0x2A, 0x26, 0x2B, 0x23, 0x2E, 0x27, 0x2F,
  80. 0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F,
  81. };
  82. /* a*inverse[b]>>32 == a/b for all 0<=a<=65536 && 2<=b<=255 */
  83. UINT32 inverse[256]={
  84. 0, 4294967295U,2147483648U,1431655766, 1073741824, 858993460, 715827883, 613566757,
  85. 536870912, 477218589, 429496730, 390451573, 357913942, 330382100, 306783379, 286331154,
  86. 268435456, 252645136, 238609295, 226050911, 214748365, 204522253, 195225787, 186737709,
  87. 178956971, 171798692, 165191050, 159072863, 153391690, 148102321, 143165577, 138547333,
  88. 134217728, 130150525, 126322568, 122713352, 119304648, 116080198, 113025456, 110127367,
  89. 107374183, 104755300, 102261127, 99882961, 97612894, 95443718, 93368855, 91382283,
  90. 89478486, 87652394, 85899346, 84215046, 82595525, 81037119, 79536432, 78090315,
  91. 76695845, 75350304, 74051161, 72796056, 71582789, 70409300, 69273667, 68174085,
  92. 67108864, 66076420, 65075263, 64103990, 63161284, 62245903, 61356676, 60492498,
  93. 59652324, 58835169, 58040099, 57266231, 56512728, 55778797, 55063684, 54366675,
  94. 53687092, 53024288, 52377650, 51746594, 51130564, 50529028, 49941481, 49367441,
  95. 48806447, 48258060, 47721859, 47197443, 46684428, 46182445, 45691142, 45210183,
  96. 44739243, 44278014, 43826197, 43383509, 42949673, 42524429, 42107523, 41698712,
  97. 41297763, 40904451, 40518560, 40139882, 39768216, 39403370, 39045158, 38693400,
  98. 38347923, 38008561, 37675152, 37347542, 37025581, 36709123, 36398028, 36092163,
  99. 35791395, 35495598, 35204650, 34918434, 34636834, 34359739, 34087043, 33818641,
  100. 33554432, 33294321, 33038210, 32786010, 32537632, 32292988, 32051995, 31814573,
  101. 31580642, 31350127, 31122952, 30899046, 30678338, 30460761, 30246249, 30034737,
  102. 29826162, 29620465, 29417585, 29217465, 29020050, 28825284, 28633116, 28443493,
  103. 28256364, 28071682, 27889399, 27709467, 27531842, 27356480, 27183338, 27012373,
  104. 26843546, 26676816, 26512144, 26349493, 26188825, 26030105, 25873297, 25718368,
  105. 25565282, 25414008, 25264514, 25116768, 24970741, 24826401, 24683721, 24542671,
  106. 24403224, 24265352, 24129030, 23994231, 23860930, 23729102, 23598722, 23469767,
  107. 23342214, 23216040, 23091223, 22967740, 22845571, 22724695, 22605092, 22486740,
  108. 22369622, 22253717, 22139007, 22025474, 21913099, 21801865, 21691755, 21582751,
  109. 21474837, 21367997, 21262215, 21157475, 21053762, 20951060, 20849356, 20748635,
  110. 20648882, 20550083, 20452226, 20355296, 20259280, 20164166, 20069941, 19976593,
  111. 19884108, 19792477, 19701685, 19611723, 19522579, 19434242, 19346700, 19259944,
  112. 19173962, 19088744, 19004281, 18920561, 18837576, 18755316, 18673771, 18592933,
  113. 18512791, 18433337, 18354562, 18276457, 18199014, 18122225, 18046082, 17970575,
  114. 17895698, 17821442, 17747799, 17674763, 17602325, 17530479, 17459217, 17388532,
  115. 17318417, 17248865, 17179870, 17111424, 17043522, 16976156, 16909321, 16843010,
  116. };
  117. /* used to skip zeros at the end */
  118. UINT8 zigzag_end[64];
  119. UINT8 permutation[64];
  120. //UINT8 invPermutation[64];
  121. static void build_zigzag_end()
  122. {
  123. int lastIndex;
  124. int lastIndexAfterPerm=0;
  125. for(lastIndex=0; lastIndex<64; lastIndex++)
  126. {
  127. if(zigzag_direct[lastIndex] > lastIndexAfterPerm)
  128. lastIndexAfterPerm= zigzag_direct[lastIndex];
  129. zigzag_end[lastIndex]= lastIndexAfterPerm + 1;
  130. }
  131. }
  132. void get_pixels_c(DCTELEM *block, const UINT8 *pixels, int line_size)
  133. {
  134. DCTELEM *p;
  135. const UINT8 *pix;
  136. int i;
  137. /* read the pixels */
  138. p = block;
  139. pix = pixels;
  140. for(i=0;i<8;i++) {
  141. p[0] = pix[0];
  142. p[1] = pix[1];
  143. p[2] = pix[2];
  144. p[3] = pix[3];
  145. p[4] = pix[4];
  146. p[5] = pix[5];
  147. p[6] = pix[6];
  148. p[7] = pix[7];
  149. pix += line_size;
  150. p += 8;
  151. }
  152. }
  153. void put_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size)
  154. {
  155. const DCTELEM *p;
  156. UINT8 *pix;
  157. int i;
  158. UINT8 *cm = cropTbl + MAX_NEG_CROP;
  159. /* read the pixels */
  160. p = block;
  161. pix = pixels;
  162. for(i=0;i<8;i++) {
  163. pix[0] = cm[p[0]];
  164. pix[1] = cm[p[1]];
  165. pix[2] = cm[p[2]];
  166. pix[3] = cm[p[3]];
  167. pix[4] = cm[p[4]];
  168. pix[5] = cm[p[5]];
  169. pix[6] = cm[p[6]];
  170. pix[7] = cm[p[7]];
  171. pix += line_size;
  172. p += 8;
  173. }
  174. }
  175. void add_pixels_clamped_c(const DCTELEM *block, UINT8 *pixels, int line_size)
  176. {
  177. const DCTELEM *p;
  178. UINT8 *pix;
  179. int i;
  180. UINT8 *cm = cropTbl + MAX_NEG_CROP;
  181. /* read the pixels */
  182. p = block;
  183. pix = pixels;
  184. for(i=0;i<8;i++) {
  185. pix[0] = cm[pix[0] + p[0]];
  186. pix[1] = cm[pix[1] + p[1]];
  187. pix[2] = cm[pix[2] + p[2]];
  188. pix[3] = cm[pix[3] + p[3]];
  189. pix[4] = cm[pix[4] + p[4]];
  190. pix[5] = cm[pix[5] + p[5]];
  191. pix[6] = cm[pix[6] + p[6]];
  192. pix[7] = cm[pix[7] + p[7]];
  193. pix += line_size;
  194. p += 8;
  195. }
  196. }
  197. #define PIXOP(BTYPE, OPNAME, OP, INCR) \
  198. \
  199. static void OPNAME ## _pixels(BTYPE *block, const UINT8 *pixels, int line_size, int h) \
  200. { \
  201. BTYPE *p; \
  202. const UINT8 *pix; \
  203. \
  204. p = block; \
  205. pix = pixels; \
  206. do { \
  207. OP(p[0], pix[0]); \
  208. OP(p[1], pix[1]); \
  209. OP(p[2], pix[2]); \
  210. OP(p[3], pix[3]); \
  211. OP(p[4], pix[4]); \
  212. OP(p[5], pix[5]); \
  213. OP(p[6], pix[6]); \
  214. OP(p[7], pix[7]); \
  215. pix += line_size; \
  216. p += INCR; \
  217. } while (--h);; \
  218. } \
  219. \
  220. static void OPNAME ## _pixels_x2(BTYPE *block, const UINT8 *pixels, int line_size, int h) \
  221. { \
  222. BTYPE *p; \
  223. const UINT8 *pix; \
  224. \
  225. p = block; \
  226. pix = pixels; \
  227. do { \
  228. OP(p[0], avg2(pix[0], pix[1])); \
  229. OP(p[1], avg2(pix[1], pix[2])); \
  230. OP(p[2], avg2(pix[2], pix[3])); \
  231. OP(p[3], avg2(pix[3], pix[4])); \
  232. OP(p[4], avg2(pix[4], pix[5])); \
  233. OP(p[5], avg2(pix[5], pix[6])); \
  234. OP(p[6], avg2(pix[6], pix[7])); \
  235. OP(p[7], avg2(pix[7], pix[8])); \
  236. pix += line_size; \
  237. p += INCR; \
  238. } while (--h); \
  239. } \
  240. \
  241. static void OPNAME ## _pixels_y2(BTYPE *block, const UINT8 *pixels, int line_size, int h) \
  242. { \
  243. BTYPE *p; \
  244. const UINT8 *pix; \
  245. const UINT8 *pix1; \
  246. \
  247. p = block; \
  248. pix = pixels; \
  249. pix1 = pixels + line_size; \
  250. do { \
  251. OP(p[0], avg2(pix[0], pix1[0])); \
  252. OP(p[1], avg2(pix[1], pix1[1])); \
  253. OP(p[2], avg2(pix[2], pix1[2])); \
  254. OP(p[3], avg2(pix[3], pix1[3])); \
  255. OP(p[4], avg2(pix[4], pix1[4])); \
  256. OP(p[5], avg2(pix[5], pix1[5])); \
  257. OP(p[6], avg2(pix[6], pix1[6])); \
  258. OP(p[7], avg2(pix[7], pix1[7])); \
  259. pix += line_size; \
  260. pix1 += line_size; \
  261. p += INCR; \
  262. } while(--h); \
  263. } \
  264. \
  265. static void OPNAME ## _pixels_xy2(BTYPE *block, const UINT8 *pixels, int line_size, int h) \
  266. { \
  267. BTYPE *p; \
  268. const UINT8 *pix; \
  269. const UINT8 *pix1; \
  270. \
  271. p = block; \
  272. pix = pixels; \
  273. pix1 = pixels + line_size; \
  274. do { \
  275. OP(p[0], avg4(pix[0], pix[1], pix1[0], pix1[1])); \
  276. OP(p[1], avg4(pix[1], pix[2], pix1[1], pix1[2])); \
  277. OP(p[2], avg4(pix[2], pix[3], pix1[2], pix1[3])); \
  278. OP(p[3], avg4(pix[3], pix[4], pix1[3], pix1[4])); \
  279. OP(p[4], avg4(pix[4], pix[5], pix1[4], pix1[5])); \
  280. OP(p[5], avg4(pix[5], pix[6], pix1[5], pix1[6])); \
  281. OP(p[6], avg4(pix[6], pix[7], pix1[6], pix1[7])); \
  282. OP(p[7], avg4(pix[7], pix[8], pix1[7], pix1[8])); \
  283. pix += line_size; \
  284. pix1 += line_size; \
  285. p += INCR; \
  286. } while(--h); \
  287. } \
  288. \
  289. void (*OPNAME ## _pixels_tab[4])(BTYPE *block, const UINT8 *pixels, int line_size, int h) = { \
  290. OPNAME ## _pixels, \
  291. OPNAME ## _pixels_x2, \
  292. OPNAME ## _pixels_y2, \
  293. OPNAME ## _pixels_xy2, \
  294. };
  295. /* rounding primitives */
  296. #define avg2(a,b) ((a+b+1)>>1)
  297. #define avg4(a,b,c,d) ((a+b+c+d+2)>>2)
  298. #define op_put(a, b) a = b
  299. #define op_avg(a, b) a = avg2(a, b)
  300. #define op_sub(a, b) a -= b
  301. PIXOP(UINT8, put, op_put, line_size)
  302. PIXOP(UINT8, avg, op_avg, line_size)
  303. PIXOP(DCTELEM, sub, op_sub, 8)
  304. /* not rounding primitives */
  305. #undef avg2
  306. #undef avg4
  307. #define avg2(a,b) ((a+b)>>1)
  308. #define avg4(a,b,c,d) ((a+b+c+d+1)>>2)
  309. PIXOP(UINT8, put_no_rnd, op_put, line_size)
  310. PIXOP(UINT8, avg_no_rnd, op_avg, line_size)
  311. /* motion estimation */
  312. #undef avg2
  313. #undef avg4
  314. #define avg2(a,b) ((a+b+1)>>1)
  315. #define avg4(a,b,c,d) ((a+b+c+d+2)>>2)
  316. int pix_abs16x16_c(UINT8 *pix1, UINT8 *pix2, int line_size, int h)
  317. {
  318. int s, i;
  319. s = 0;
  320. for(i=0;i<h;i++) {
  321. s += abs(pix1[0] - pix2[0]);
  322. s += abs(pix1[1] - pix2[1]);
  323. s += abs(pix1[2] - pix2[2]);
  324. s += abs(pix1[3] - pix2[3]);
  325. s += abs(pix1[4] - pix2[4]);
  326. s += abs(pix1[5] - pix2[5]);
  327. s += abs(pix1[6] - pix2[6]);
  328. s += abs(pix1[7] - pix2[7]);
  329. s += abs(pix1[8] - pix2[8]);
  330. s += abs(pix1[9] - pix2[9]);
  331. s += abs(pix1[10] - pix2[10]);
  332. s += abs(pix1[11] - pix2[11]);
  333. s += abs(pix1[12] - pix2[12]);
  334. s += abs(pix1[13] - pix2[13]);
  335. s += abs(pix1[14] - pix2[14]);
  336. s += abs(pix1[15] - pix2[15]);
  337. pix1 += line_size;
  338. pix2 += line_size;
  339. }
  340. return s;
  341. }
  342. int pix_abs16x16_x2_c(UINT8 *pix1, UINT8 *pix2, int line_size, int h)
  343. {
  344. int s, i;
  345. s = 0;
  346. for(i=0;i<h;i++) {
  347. s += abs(pix1[0] - avg2(pix2[0], pix2[1]));
  348. s += abs(pix1[1] - avg2(pix2[1], pix2[2]));
  349. s += abs(pix1[2] - avg2(pix2[2], pix2[3]));
  350. s += abs(pix1[3] - avg2(pix2[3], pix2[4]));
  351. s += abs(pix1[4] - avg2(pix2[4], pix2[5]));
  352. s += abs(pix1[5] - avg2(pix2[5], pix2[6]));
  353. s += abs(pix1[6] - avg2(pix2[6], pix2[7]));
  354. s += abs(pix1[7] - avg2(pix2[7], pix2[8]));
  355. s += abs(pix1[8] - avg2(pix2[8], pix2[9]));
  356. s += abs(pix1[9] - avg2(pix2[9], pix2[10]));
  357. s += abs(pix1[10] - avg2(pix2[10], pix2[11]));
  358. s += abs(pix1[11] - avg2(pix2[11], pix2[12]));
  359. s += abs(pix1[12] - avg2(pix2[12], pix2[13]));
  360. s += abs(pix1[13] - avg2(pix2[13], pix2[14]));
  361. s += abs(pix1[14] - avg2(pix2[14], pix2[15]));
  362. s += abs(pix1[15] - avg2(pix2[15], pix2[16]));
  363. pix1 += line_size;
  364. pix2 += line_size;
  365. }
  366. return s;
  367. }
  368. int pix_abs16x16_y2_c(UINT8 *pix1, UINT8 *pix2, int line_size, int h)
  369. {
  370. int s, i;
  371. UINT8 *pix3 = pix2 + line_size;
  372. s = 0;
  373. for(i=0;i<h;i++) {
  374. s += abs(pix1[0] - avg2(pix2[0], pix3[0]));
  375. s += abs(pix1[1] - avg2(pix2[1], pix3[1]));
  376. s += abs(pix1[2] - avg2(pix2[2], pix3[2]));
  377. s += abs(pix1[3] - avg2(pix2[3], pix3[3]));
  378. s += abs(pix1[4] - avg2(pix2[4], pix3[4]));
  379. s += abs(pix1[5] - avg2(pix2[5], pix3[5]));
  380. s += abs(pix1[6] - avg2(pix2[6], pix3[6]));
  381. s += abs(pix1[7] - avg2(pix2[7], pix3[7]));
  382. s += abs(pix1[8] - avg2(pix2[8], pix3[8]));
  383. s += abs(pix1[9] - avg2(pix2[9], pix3[9]));
  384. s += abs(pix1[10] - avg2(pix2[10], pix3[10]));
  385. s += abs(pix1[11] - avg2(pix2[11], pix3[11]));
  386. s += abs(pix1[12] - avg2(pix2[12], pix3[12]));
  387. s += abs(pix1[13] - avg2(pix2[13], pix3[13]));
  388. s += abs(pix1[14] - avg2(pix2[14], pix3[14]));
  389. s += abs(pix1[15] - avg2(pix2[15], pix3[15]));
  390. pix1 += line_size;
  391. pix2 += line_size;
  392. pix3 += line_size;
  393. }
  394. return s;
  395. }
  396. int pix_abs16x16_xy2_c(UINT8 *pix1, UINT8 *pix2, int line_size, int h)
  397. {
  398. int s, i;
  399. UINT8 *pix3 = pix2 + line_size;
  400. s = 0;
  401. for(i=0;i<h;i++) {
  402. s += abs(pix1[0] - avg4(pix2[0], pix2[1], pix3[0], pix3[1]));
  403. s += abs(pix1[1] - avg4(pix2[1], pix2[2], pix3[1], pix3[2]));
  404. s += abs(pix1[2] - avg4(pix2[2], pix2[3], pix3[2], pix3[3]));
  405. s += abs(pix1[3] - avg4(pix2[3], pix2[4], pix3[3], pix3[4]));
  406. s += abs(pix1[4] - avg4(pix2[4], pix2[5], pix3[4], pix3[5]));
  407. s += abs(pix1[5] - avg4(pix2[5], pix2[6], pix3[5], pix3[6]));
  408. s += abs(pix1[6] - avg4(pix2[6], pix2[7], pix3[6], pix3[7]));
  409. s += abs(pix1[7] - avg4(pix2[7], pix2[8], pix3[7], pix3[8]));
  410. s += abs(pix1[8] - avg4(pix2[8], pix2[9], pix3[8], pix3[9]));
  411. s += abs(pix1[9] - avg4(pix2[9], pix2[10], pix3[9], pix3[10]));
  412. s += abs(pix1[10] - avg4(pix2[10], pix2[11], pix3[10], pix3[11]));
  413. s += abs(pix1[11] - avg4(pix2[11], pix2[12], pix3[11], pix3[12]));
  414. s += abs(pix1[12] - avg4(pix2[12], pix2[13], pix3[12], pix3[13]));
  415. s += abs(pix1[13] - avg4(pix2[13], pix2[14], pix3[13], pix3[14]));
  416. s += abs(pix1[14] - avg4(pix2[14], pix2[15], pix3[14], pix3[15]));
  417. s += abs(pix1[15] - avg4(pix2[15], pix2[16], pix3[15], pix3[16]));
  418. pix1 += line_size;
  419. pix2 += line_size;
  420. pix3 += line_size;
  421. }
  422. return s;
  423. }
  424. /* permute block according so that it corresponds to the MMX idct
  425. order */
  426. #ifdef SIMPLE_IDCT
  427. /* general permutation, but perhaps slightly slower */
  428. void block_permute(INT16 *block)
  429. {
  430. int i;
  431. INT16 temp[64];
  432. for(i=0; i<64; i++) temp[ block_permute_op(i) ] = block[i];
  433. for(i=0; i<64; i++) block[i] = temp[i];
  434. }
  435. #else
  436. void block_permute(INT16 *block)
  437. {
  438. int tmp1, tmp2, tmp3, tmp4, tmp5, tmp6;
  439. int i;
  440. for(i=0;i<8;i++) {
  441. tmp1 = block[1];
  442. tmp2 = block[2];
  443. tmp3 = block[3];
  444. tmp4 = block[4];
  445. tmp5 = block[5];
  446. tmp6 = block[6];
  447. block[1] = tmp2;
  448. block[2] = tmp4;
  449. block[3] = tmp6;
  450. block[4] = tmp1;
  451. block[5] = tmp3;
  452. block[6] = tmp5;
  453. block += 8;
  454. }
  455. }
  456. #endif
  457. void dsputil_init(void)
  458. {
  459. int i, j;
  460. int use_permuted_idct;
  461. for(i=0;i<256;i++) cropTbl[i + MAX_NEG_CROP] = i;
  462. for(i=0;i<MAX_NEG_CROP;i++) {
  463. cropTbl[i] = 0;
  464. cropTbl[i + MAX_NEG_CROP + 256] = 255;
  465. }
  466. for(i=0;i<512;i++) {
  467. squareTbl[i] = (i - 256) * (i - 256);
  468. }
  469. #ifdef SIMPLE_IDCT
  470. ff_idct = simple_idct;
  471. #else
  472. ff_idct = j_rev_dct;
  473. #endif
  474. get_pixels = get_pixels_c;
  475. put_pixels_clamped = put_pixels_clamped_c;
  476. add_pixels_clamped = add_pixels_clamped_c;
  477. pix_abs16x16 = pix_abs16x16_c;
  478. pix_abs16x16_x2 = pix_abs16x16_x2_c;
  479. pix_abs16x16_y2 = pix_abs16x16_y2_c;
  480. pix_abs16x16_xy2 = pix_abs16x16_xy2_c;
  481. av_fdct = jpeg_fdct_ifast;
  482. use_permuted_idct = 1;
  483. #ifdef HAVE_MMX
  484. dsputil_init_mmx();
  485. #endif
  486. #ifdef ARCH_ARMV4L
  487. dsputil_init_armv4l();
  488. #endif
  489. #ifdef HAVE_MLIB
  490. dsputil_init_mlib();
  491. use_permuted_idct = 0;
  492. #endif
  493. #ifdef ARCH_ALPHA
  494. dsputil_init_alpha();
  495. use_permuted_idct = 0;
  496. #endif
  497. #ifdef SIMPLE_IDCT
  498. if(ff_idct == simple_idct) use_permuted_idct=0;
  499. #endif
  500. if(use_permuted_idct)
  501. #ifdef SIMPLE_IDCT
  502. for(i=0; i<64; i++) permutation[i]= simple_mmx_permutation[i];
  503. #else
  504. for(i=0; i<64; i++) permutation[i]= (i & 0x38) | ((i & 6) >> 1) | ((i & 1) << 2);
  505. #endif
  506. else
  507. for(i=0; i<64; i++) permutation[i]=i;
  508. for(i=0; i<64; i++) inv_zigzag_direct16[zigzag_direct[i]]= i+1;
  509. for(i=0; i<64; i++) zigzag_direct_noperm[i]= zigzag_direct[i];
  510. if (use_permuted_idct) {
  511. /* permute for IDCT */
  512. for(i=0;i<64;i++) {
  513. j = zigzag_direct[i];
  514. zigzag_direct[i] = block_permute_op(j);
  515. j = ff_alternate_horizontal_scan[i];
  516. ff_alternate_horizontal_scan[i] = block_permute_op(j);
  517. j = ff_alternate_vertical_scan[i];
  518. ff_alternate_vertical_scan[i] = block_permute_op(j);
  519. }
  520. block_permute(default_intra_matrix);
  521. block_permute(default_non_intra_matrix);
  522. }
  523. build_zigzag_end();
  524. }
  525. void get_psnr(UINT8 *orig_image[3], UINT8 *coded_image[3],
  526. int orig_linesize[3], int coded_linesize,
  527. AVCodecContext *avctx)
  528. {
  529. int quad, diff, x, y;
  530. UINT8 *orig, *coded;
  531. UINT32 *sq = squareTbl + 256;
  532. quad = 0;
  533. diff = 0;
  534. /* Luminance */
  535. orig = orig_image[0];
  536. coded = coded_image[0];
  537. for (y=0;y<avctx->height;y++) {
  538. for (x=0;x<avctx->width;x++) {
  539. diff = *(orig + x) - *(coded + x);
  540. quad += sq[diff];
  541. }
  542. orig += orig_linesize[0];
  543. coded += coded_linesize;
  544. }
  545. avctx->psnr_y = (float) quad / (float) (avctx->width * avctx->height);
  546. if (avctx->psnr_y) {
  547. avctx->psnr_y = (float) (255 * 255) / avctx->psnr_y;
  548. avctx->psnr_y = 10 * (float) log10 (avctx->psnr_y);
  549. } else
  550. avctx->psnr_y = 99.99;
  551. }