falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
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.
1131 Commits
32 Branches
580MB
Tree: 050fe8bab5
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '050fe8bab5'
${ noResults }
FFmpeg/libavcodec/dsputil.c

1738 lines
71KB
Raw Blame History 

  1. /*

  2.  * DSP utils

  3.  * Copyright (c) 2000, 2001 Fabrice Bellard.

  4.  *

  5.  * This library is free software; you can redistribute it and/or

  6.  * modify it under the terms of the GNU Lesser General Public

  7.  * License as published by the Free Software Foundation; either

  8.  * version 2 of the License, or (at your option) any later version.

  9.  *

  10.  * This library 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 GNU

  13.  * Lesser General Public License for more details.

  14.  *

  15.  * You should have received a copy of the GNU Lesser General Public

  16.  * License along with this library; if not, write to the Free Software

  17.  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA

  18.  *

  19.  * gmc & q-pel & 32/64 bit based MC by Michael Niedermayer <michaelni@gmx.at>

  20.  */

  21. #include "avcodec.h"

  22. #include "dsputil.h"

  23. #include "simple_idct.h"

  24. 

  25. void (*ff_idct)(DCTELEM *block);

  26. void (*ff_idct_put)(UINT8 *dest, int line_size, DCTELEM *block);

  27. void (*ff_idct_add)(UINT8 *dest, int line_size, DCTELEM *block);

  28. void (*get_pixels)(DCTELEM *block, const UINT8 *pixels, int line_size);

  29. void (*diff_pixels)(DCTELEM *block, const UINT8 *s1, const UINT8 *s2, int stride);

  30. void (*put_pixels_clamped)(const DCTELEM *block, UINT8 *pixels, int line_size);

  31. void (*add_pixels_clamped)(const DCTELEM *block, UINT8 *pixels, int line_size);

  32. void (*gmc1)(UINT8 *dst, UINT8 *src, int srcStride, int h, int x16, int y16, int rounder);

  33. void (*clear_blocks)(DCTELEM *blocks);

  34. int (*pix_sum)(UINT8 * pix, int line_size);

  35. int (*pix_norm1)(UINT8 * pix, int line_size);

  36. 

  37. op_pixels_abs_func pix_abs16x16;

  38. op_pixels_abs_func pix_abs16x16_x2;

  39. op_pixels_abs_func pix_abs16x16_y2;

  40. op_pixels_abs_func pix_abs16x16_xy2;

  41. 

  42. op_pixels_abs_func pix_abs8x8;

  43. op_pixels_abs_func pix_abs8x8_x2;

  44. op_pixels_abs_func pix_abs8x8_y2;

  45. op_pixels_abs_func pix_abs8x8_xy2;

  46. 

  47. UINT8 cropTbl[256 + 2 * MAX_NEG_CROP];

  48. UINT32 squareTbl[512];

  49. 

  50. extern INT16 ff_mpeg1_default_intra_matrix[64];

  51. extern INT16 ff_mpeg1_default_non_intra_matrix[64];

  52. extern INT16 ff_mpeg4_default_intra_matrix[64];

  53. extern INT16 ff_mpeg4_default_non_intra_matrix[64];

  54. 

  55. UINT8 zigzag_direct[64] = {

  56.     0, 1, 8, 16, 9, 2, 3, 10,

  57.     17, 24, 32, 25, 18, 11, 4, 5,

  58.     12, 19, 26, 33, 40, 48, 41, 34,

  59.     27, 20, 13, 6, 7, 14, 21, 28,

  60.     35, 42, 49, 56, 57, 50, 43, 36,

  61.     29, 22, 15, 23, 30, 37, 44, 51,

  62.     58, 59, 52, 45, 38, 31, 39, 46,

  63.     53, 60, 61, 54, 47, 55, 62, 63

  64. };

  65. 

  66. /* not permutated inverse zigzag_direct + 1 for MMX quantizer */

  67. UINT16 __align8 inv_zigzag_direct16[64];

  68. 

  69. /* not permutated zigzag_direct for MMX quantizer */

  70. UINT8 zigzag_direct_noperm[64];

  71. 

  72. UINT8 ff_alternate_horizontal_scan[64] = {

  73.     0,  1,  2,  3,  8,  9, 16, 17, 

  74.     10, 11,  4,  5,  6,  7, 15, 14,

  75.     13, 12, 19, 18, 24, 25, 32, 33, 

  76.     26, 27, 20, 21, 22, 23, 28, 29,

  77.     30, 31, 34, 35, 40, 41, 48, 49, 

  78.     42, 43, 36, 37, 38, 39, 44, 45,

  79.     46, 47, 50, 51, 56, 57, 58, 59, 

  80.     52, 53, 54, 55, 60, 61, 62, 63,

  81. };

  82. 

  83. UINT8 ff_alternate_vertical_scan[64] = {

  84.     0,  8, 16, 24,  1,  9,  2, 10, 

  85.     17, 25, 32, 40, 48, 56, 57, 49,

  86.     41, 33, 26, 18,  3, 11,  4, 12, 

  87.     19, 27, 34, 42, 50, 58, 35, 43,

  88.     51, 59, 20, 28,  5, 13,  6, 14, 

  89.     21, 29, 36, 44, 52, 60, 37, 45,

  90.     53, 61, 22, 30,  7, 15, 23, 31, 

  91.     38, 46, 54, 62, 39, 47, 55, 63,

  92. };

  93. 

  94. #ifdef SIMPLE_IDCT

  95. 

  96. /* Input permutation for the simple_idct_mmx */

  97. static UINT8 simple_mmx_permutation[64]={

  98. 	0x00, 0x08, 0x04, 0x09, 0x01, 0x0C, 0x05, 0x0D, 

  99. 	0x10, 0x18, 0x14, 0x19, 0x11, 0x1C, 0x15, 0x1D, 

  100. 	0x20, 0x28, 0x24, 0x29, 0x21, 0x2C, 0x25, 0x2D, 

  101. 	0x12, 0x1A, 0x16, 0x1B, 0x13, 0x1E, 0x17, 0x1F, 

  102. 	0x02, 0x0A, 0x06, 0x0B, 0x03, 0x0E, 0x07, 0x0F, 

  103. 	0x30, 0x38, 0x34, 0x39, 0x31, 0x3C, 0x35, 0x3D, 

  104. 	0x22, 0x2A, 0x26, 0x2B, 0x23, 0x2E, 0x27, 0x2F, 

  105. 	0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F,

  106. };

  107. #endif

  108. 

  109. /* a*inverse[b]>>32 == a/b for all 0<=a<=65536 && 2<=b<=255 */

  110. UINT32 inverse[256]={

  111.          0, 4294967295U,2147483648U,1431655766, 1073741824,  858993460,  715827883,  613566757, 

  112.  536870912,  477218589,  429496730,  390451573,  357913942,  330382100,  306783379,  286331154, 

  113.  268435456,  252645136,  238609295,  226050911,  214748365,  204522253,  195225787,  186737709, 

  114.  178956971,  171798692,  165191050,  159072863,  153391690,  148102321,  143165577,  138547333, 

  115.  134217728,  130150525,  126322568,  122713352,  119304648,  116080198,  113025456,  110127367, 

  116.  107374183,  104755300,  102261127,   99882961,   97612894,   95443718,   93368855,   91382283, 

  117.   89478486,   87652394,   85899346,   84215046,   82595525,   81037119,   79536432,   78090315, 

  118.   76695845,   75350304,   74051161,   72796056,   71582789,   70409300,   69273667,   68174085, 

  119.   67108864,   66076420,   65075263,   64103990,   63161284,   62245903,   61356676,   60492498, 

  120.   59652324,   58835169,   58040099,   57266231,   56512728,   55778797,   55063684,   54366675, 

  121.   53687092,   53024288,   52377650,   51746594,   51130564,   50529028,   49941481,   49367441, 

  122.   48806447,   48258060,   47721859,   47197443,   46684428,   46182445,   45691142,   45210183, 

  123.   44739243,   44278014,   43826197,   43383509,   42949673,   42524429,   42107523,   41698712, 

  124.   41297763,   40904451,   40518560,   40139882,   39768216,   39403370,   39045158,   38693400, 

  125.   38347923,   38008561,   37675152,   37347542,   37025581,   36709123,   36398028,   36092163, 

  126.   35791395,   35495598,   35204650,   34918434,   34636834,   34359739,   34087043,   33818641, 

  127.   33554432,   33294321,   33038210,   32786010,   32537632,   32292988,   32051995,   31814573, 

  128.   31580642,   31350127,   31122952,   30899046,   30678338,   30460761,   30246249,   30034737, 

  129.   29826162,   29620465,   29417585,   29217465,   29020050,   28825284,   28633116,   28443493, 

  130.   28256364,   28071682,   27889399,   27709467,   27531842,   27356480,   27183338,   27012373, 

  131.   26843546,   26676816,   26512144,   26349493,   26188825,   26030105,   25873297,   25718368, 

  132.   25565282,   25414008,   25264514,   25116768,   24970741,   24826401,   24683721,   24542671, 

  133.   24403224,   24265352,   24129030,   23994231,   23860930,   23729102,   23598722,   23469767, 

  134.   23342214,   23216040,   23091223,   22967740,   22845571,   22724695,   22605092,   22486740, 

  135.   22369622,   22253717,   22139007,   22025474,   21913099,   21801865,   21691755,   21582751, 

  136.   21474837,   21367997,   21262215,   21157475,   21053762,   20951060,   20849356,   20748635, 

  137.   20648882,   20550083,   20452226,   20355296,   20259280,   20164166,   20069941,   19976593, 

  138.   19884108,   19792477,   19701685,   19611723,   19522579,   19434242,   19346700,   19259944, 

  139.   19173962,   19088744,   19004281,   18920561,   18837576,   18755316,   18673771,   18592933, 

  140.   18512791,   18433337,   18354562,   18276457,   18199014,   18122225,   18046082,   17970575, 

  141.   17895698,   17821442,   17747799,   17674763,   17602325,   17530479,   17459217,   17388532, 

  142.   17318417,   17248865,   17179870,   17111424,   17043522,   16976156,   16909321,   16843010,

  143. };

  144. 

  145. /* used to skip zeros at the end */

  146. UINT8 zigzag_end[64];

  147. 

  148. UINT8 permutation[64];

  149. //UINT8 invPermutation[64];

  150. 

  151. static void build_zigzag_end(void)

  152. {

  153.     int lastIndex;

  154.     int lastIndexAfterPerm=0;

  155.     for(lastIndex=0; lastIndex<64; lastIndex++)

  156.     {

  157.         if(zigzag_direct[lastIndex] > lastIndexAfterPerm) 

  158.             lastIndexAfterPerm= zigzag_direct[lastIndex];

  159.         zigzag_end[lastIndex]= lastIndexAfterPerm + 1;

  160.     }

  161. }

  162. 

  163. int pix_sum_c(UINT8 * pix, int line_size)

  164. {

  165.     int s, i, j;

  166. 

  167.     s = 0;

  168.     for (i = 0; i < 16; i++) {

  169. 	for (j = 0; j < 16; j += 8) {

  170. 	    s += pix[0];

  171. 	    s += pix[1];

  172. 	    s += pix[2];

  173. 	    s += pix[3];

  174. 	    s += pix[4];

  175. 	    s += pix[5];

  176. 	    s += pix[6];

  177. 	    s += pix[7];

  178. 	    pix += 8;

  179. 	}

  180. 	pix += line_size - 16;

  181.     }

  182.     return s;

  183. }

  184. 

  185. int pix_norm1_c(UINT8 * pix, int line_size)

  186. {

  187.     int s, i, j;

  188.     UINT32 *sq = squareTbl + 256;

  189. 

  190.     s = 0;

  191.     for (i = 0; i < 16; i++) {

  192. 	for (j = 0; j < 16; j += 8) {

  193. 	    s += sq[pix[0]];

  194. 	    s += sq[pix[1]];

  195. 	    s += sq[pix[2]];

  196. 	    s += sq[pix[3]];

  197. 	    s += sq[pix[4]];

  198. 	    s += sq[pix[5]];

  199. 	    s += sq[pix[6]];

  200. 	    s += sq[pix[7]];

  201. 	    pix += 8;

  202. 	}

  203. 	pix += line_size - 16;

  204.     }

  205.     return s;

  206. }

  207. 

  208. 

  209. void get_pixels_c(DCTELEM *restrict block, const UINT8 *pixels, int line_size)

  210. {

  211.     int i;

  212. 

  213.     /* read the pixels */

  214.     for(i=0;i<8;i++) {

  215.         block[0] = pixels[0];

  216.         block[1] = pixels[1];

  217.         block[2] = pixels[2];

  218.         block[3] = pixels[3];

  219.         block[4] = pixels[4];

  220.         block[5] = pixels[5];

  221.         block[6] = pixels[6];

  222.         block[7] = pixels[7];

  223.         pixels += line_size;

  224.         block += 8;

  225.     }

  226. }

  227. 

  228. void diff_pixels_c(DCTELEM *restrict block, const UINT8 *s1, const UINT8 *s2,

  229. 		   int stride){

  230.     int i;

  231. 

  232.     /* read the pixels */

  233.     for(i=0;i<8;i++) {

  234.         block[0] = s1[0] - s2[0];

  235.         block[1] = s1[1] - s2[1];

  236.         block[2] = s1[2] - s2[2];

  237.         block[3] = s1[3] - s2[3];

  238.         block[4] = s1[4] - s2[4];

  239.         block[5] = s1[5] - s2[5];

  240.         block[6] = s1[6] - s2[6];

  241.         block[7] = s1[7] - s2[7];

  242.         s1 += stride;

  243.         s2 += stride;

  244.         block += 8;

  245.     }

  246. }

  247. 

  248. 

  249. void put_pixels_clamped_c(const DCTELEM *block, UINT8 *restrict pixels,

  250.                           int line_size)

  251. {

  252.     int i;

  253.     UINT8 *cm = cropTbl + MAX_NEG_CROP;

  254.     

  255.     /* read the pixels */

  256.     for(i=0;i<8;i++) {

  257.         pixels[0] = cm[block[0]];

  258.         pixels[1] = cm[block[1]];

  259.         pixels[2] = cm[block[2]];

  260.         pixels[3] = cm[block[3]];

  261.         pixels[4] = cm[block[4]];

  262.         pixels[5] = cm[block[5]];

  263.         pixels[6] = cm[block[6]];

  264.         pixels[7] = cm[block[7]];

  265. 

  266.         pixels += line_size;

  267.         block += 8;

  268.     }

  269. }

  270. 

  271. void add_pixels_clamped_c(const DCTELEM *block, UINT8 *restrict pixels,

  272.                           int line_size)

  273. {

  274.     int i;

  275.     UINT8 *cm = cropTbl + MAX_NEG_CROP;

  276.     

  277.     /* read the pixels */

  278.     for(i=0;i<8;i++) {

  279.         pixels[0] = cm[pixels[0] + block[0]];

  280.         pixels[1] = cm[pixels[1] + block[1]];

  281.         pixels[2] = cm[pixels[2] + block[2]];

  282.         pixels[3] = cm[pixels[3] + block[3]];

  283.         pixels[4] = cm[pixels[4] + block[4]];

  284.         pixels[5] = cm[pixels[5] + block[5]];

  285.         pixels[6] = cm[pixels[6] + block[6]];

  286.         pixels[7] = cm[pixels[7] + block[7]];

  287.         pixels += line_size;

  288.         block += 8;

  289.     }

  290. }

  291. #if 0

  292. 

  293. #define PIXOP2(OPNAME, OP) \

  294. static void OPNAME ## _pixels(uint8_t *block, const uint8_t *pixels, int line_size, int h)\

  295. {\

  296.     int i;\

  297.     for(i=0; i<h; i++){\

  298.         OP(*((uint64_t*)block), LD64(pixels));\

  299.         pixels+=line_size;\

  300.         block +=line_size;\

  301.     }\

  302. }\

  303. \

  304. static void OPNAME ## _no_rnd_pixels_x2(uint8_t *block, const uint8_t *pixels, int line_size, int h)\

  305. {\

  306.     int i;\

  307.     for(i=0; i<h; i++){\

  308.         const uint64_t a= LD64(pixels  );\

  309.         const uint64_t b= LD64(pixels+1);\

  310.         OP(*((uint64_t*)block), (a&b) + (((a^b)&0xFEFEFEFEFEFEFEFEULL)>>1));\

  311.         pixels+=line_size;\

  312.         block +=line_size;\

  313.     }\

  314. }\

  315. \

  316. static void OPNAME ## _pixels_x2(uint8_t *block, const uint8_t *pixels, int line_size, int h)\

  317. {\

  318.     int i;\

  319.     for(i=0; i<h; i++){\

  320.         const uint64_t a= LD64(pixels  );\

  321.         const uint64_t b= LD64(pixels+1);\

  322.         OP(*((uint64_t*)block), (a|b) - (((a^b)&0xFEFEFEFEFEFEFEFEULL)>>1));\

  323.         pixels+=line_size;\

  324.         block +=line_size;\

  325.     }\

  326. }\

  327. \

  328. static void OPNAME ## _no_rnd_pixels_y2(uint8_t *block, const uint8_t *pixels, int line_size, int h)\

  329. {\

  330.     int i;\

  331.     for(i=0; i<h; i++){\

  332.         const uint64_t a= LD64(pixels          );\

  333.         const uint64_t b= LD64(pixels+line_size);\

  334.         OP(*((uint64_t*)block), (a&b) + (((a^b)&0xFEFEFEFEFEFEFEFEULL)>>1));\

  335.         pixels+=line_size;\

  336.         block +=line_size;\

  337.     }\

  338. }\

  339. \

  340. static void OPNAME ## _pixels_y2(uint8_t *block, const uint8_t *pixels, int line_size, int h)\

  341. {\

  342.     int i;\

  343.     for(i=0; i<h; i++){\

  344.         const uint64_t a= LD64(pixels          );\

  345.         const uint64_t b= LD64(pixels+line_size);\

  346.         OP(*((uint64_t*)block), (a|b) - (((a^b)&0xFEFEFEFEFEFEFEFEULL)>>1));\

  347.         pixels+=line_size;\

  348.         block +=line_size;\

  349.     }\

  350. }\

  351. \

  352. static void OPNAME ## _pixels_xy2(uint8_t *block, const uint8_t *pixels, int line_size, int h)\

  353. {\

  354.         int i;\

  355.         const uint64_t a= LD64(pixels  );\

  356.         const uint64_t b= LD64(pixels+1);\

  357.         uint64_t l0=  (a&0x0303030303030303ULL)\

  358.                     + (b&0x0303030303030303ULL)\

  359.                     + 0x0202020202020202ULL;\

  360.         uint64_t h0= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\

  361.                    + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\

  362.         uint64_t l1,h1;\

  363. \

  364.         pixels+=line_size;\

  365.         for(i=0; i<h; i+=2){\

  366.             uint64_t a= LD64(pixels  );\

  367.             uint64_t b= LD64(pixels+1);\

  368.             l1=  (a&0x0303030303030303ULL)\

  369.                + (b&0x0303030303030303ULL);\

  370.             h1= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\

  371.               + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\

  372.             OP(*((uint64_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0F0F0F0F0FULL));\

  373.             pixels+=line_size;\

  374.             block +=line_size;\

  375.             a= LD64(pixels  );\

  376.             b= LD64(pixels+1);\

  377.             l0=  (a&0x0303030303030303ULL)\

  378.                + (b&0x0303030303030303ULL)\

  379.                + 0x0202020202020202ULL;\

  380.             h0= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\

  381.               + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\

  382.             OP(*((uint64_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0F0F0F0F0FULL));\

  383.             pixels+=line_size;\

  384.             block +=line_size;\

  385.         }\

  386. }\

  387. \

  388. static void OPNAME ## _no_rnd_pixels_xy2(uint8_t *block, const uint8_t *pixels, int line_size, int h)\

  389. {\

  390.         int i;\

  391.         const uint64_t a= LD64(pixels  );\

  392.         const uint64_t b= LD64(pixels+1);\

  393.         uint64_t l0=  (a&0x0303030303030303ULL)\

  394.                     + (b&0x0303030303030303ULL)\

  395.                     + 0x0101010101010101ULL;\

  396.         uint64_t h0= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\

  397.                    + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\

  398.         uint64_t l1,h1;\

  399. \

  400.         pixels+=line_size;\

  401.         for(i=0; i<h; i+=2){\

  402.             uint64_t a= LD64(pixels  );\

  403.             uint64_t b= LD64(pixels+1);\

  404.             l1=  (a&0x0303030303030303ULL)\

  405.                + (b&0x0303030303030303ULL);\

  406.             h1= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\

  407.               + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\

  408.             OP(*((uint64_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0F0F0F0F0FULL));\

  409.             pixels+=line_size;\

  410.             block +=line_size;\

  411.             a= LD64(pixels  );\

  412.             b= LD64(pixels+1);\

  413.             l0=  (a&0x0303030303030303ULL)\

  414.                + (b&0x0303030303030303ULL)\

  415.                + 0x0101010101010101ULL;\

  416.             h0= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\

  417.               + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\

  418.             OP(*((uint64_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0F0F0F0F0FULL));\

  419.             pixels+=line_size;\

  420.             block +=line_size;\

  421.         }\

  422. }\

  423. \

  424. CALL_2X_PIXELS(OPNAME ## _pixels16    , OPNAME ## _pixels    , 8)\

  425. CALL_2X_PIXELS(OPNAME ## _pixels16_x2 , OPNAME ## _pixels_x2 , 8)\

  426. CALL_2X_PIXELS(OPNAME ## _pixels16_y2 , OPNAME ## _pixels_y2 , 8)\

  427. CALL_2X_PIXELS(OPNAME ## _pixels16_xy2, OPNAME ## _pixels_xy2, 8)\

  428. CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_x2 , OPNAME ## _no_rnd_pixels_x2 , 8)\

  429. CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_y2 , OPNAME ## _no_rnd_pixels_y2 , 8)\

  430. CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_xy2, OPNAME ## _no_rnd_pixels_xy2, 8)\

  431. \

  432. void (*OPNAME ## _pixels_tab[2][4])(uint8_t *block, const uint8_t *pixels, int line_size, int h) = {\

  433.     {\

  434.         OPNAME ## _pixels,\

  435.         OPNAME ## _pixels_x2,\

  436.         OPNAME ## _pixels_y2,\

  437.         OPNAME ## _pixels_xy2},\

  438.     {\

  439.         OPNAME ## _pixels16,\

  440.         OPNAME ## _pixels16_x2,\

  441.         OPNAME ## _pixels16_y2,\

  442.         OPNAME ## _pixels16_xy2}\

  443. };\

  444. \

  445. void (*OPNAME ## _no_rnd_pixels_tab[2][4])(uint8_t *block, const uint8_t *pixels, int line_size, int h) = {\

  446.     {\

  447.         OPNAME ## _pixels,\

  448.         OPNAME ## _no_rnd_pixels_x2,\

  449.         OPNAME ## _no_rnd_pixels_y2,\

  450.         OPNAME ## _no_rnd_pixels_xy2},\

  451.     {\

  452.         OPNAME ## _pixels16,\

  453.         OPNAME ## _no_rnd_pixels16_x2,\

  454.         OPNAME ## _no_rnd_pixels16_y2,\

  455.         OPNAME ## _no_rnd_pixels16_xy2}\

  456. };

  457. 

  458. #define op_avg(a, b) a = ( ((a)|(b)) - ((((a)^(b))&0xFEFEFEFEFEFEFEFEULL)>>1) )

  459. #else // 64 bit variant

  460. 

  461. #define PIXOP2(OPNAME, OP) \

  462. static void OPNAME ## _pixels8(uint8_t *block, const uint8_t *pixels, int line_size, int h){\

  463.     int i;\

  464.     for(i=0; i<h; i++){\

  465.         OP(*((uint32_t*)(block  )), LD32(pixels  ));\

  466.         OP(*((uint32_t*)(block+4)), LD32(pixels+4));\

  467.         pixels+=line_size;\

  468.         block +=line_size;\

  469.     }\

  470. }\

  471. static inline void OPNAME ## _no_rnd_pixels8(uint8_t *block, const uint8_t *pixels, int line_size, int h){\

  472.     OPNAME ## _pixels8(block, pixels, line_size, h);\

  473. }\

  474. \

  475. static inline void OPNAME ## _no_rnd_pixels8_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \

  476.                                                 int src_stride1, int src_stride2, int h){\

  477.     int i;\

  478.     for(i=0; i<h; i++){\

  479.         uint32_t a,b;\

  480.         a= LD32(&src1[i*src_stride1  ]);\

  481.         b= LD32(&src2[i*src_stride2  ]);\

  482.         OP(*((uint32_t*)&dst[i*dst_stride  ]), (a&b) + (((a^b)&0xFEFEFEFEUL)>>1));\

  483.         a= LD32(&src1[i*src_stride1+4]);\

  484.         b= LD32(&src2[i*src_stride2+4]);\

  485.         OP(*((uint32_t*)&dst[i*dst_stride+4]), (a&b) + (((a^b)&0xFEFEFEFEUL)>>1));\

  486.     }\

  487. }\

  488. \

  489. static inline void OPNAME ## _pixels8_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \

  490.                                                 int src_stride1, int src_stride2, int h){\

  491.     int i;\

  492.     for(i=0; i<h; i++){\

  493.         uint32_t a,b;\

  494.         a= LD32(&src1[i*src_stride1  ]);\

  495.         b= LD32(&src2[i*src_stride2  ]);\

  496.         OP(*((uint32_t*)&dst[i*dst_stride  ]), (a|b) - (((a^b)&0xFEFEFEFEUL)>>1));\

  497.         a= LD32(&src1[i*src_stride1+4]);\

  498.         b= LD32(&src2[i*src_stride2+4]);\

  499.         OP(*((uint32_t*)&dst[i*dst_stride+4]), (a|b) - (((a^b)&0xFEFEFEFEUL)>>1));\

  500.     }\

  501. }\

  502. \

  503. static inline void OPNAME ## _pixels16_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \

  504.                                                 int src_stride1, int src_stride2, int h){\

  505.     OPNAME ## _pixels8_l2(dst  , src1  , src2  , dst_stride, src_stride1, src_stride2, h);\

  506.     OPNAME ## _pixels8_l2(dst+8, src1+8, src2+8, dst_stride, src_stride1, src_stride2, h);\

  507. }\

  508. \

  509. static inline void OPNAME ## _no_rnd_pixels16_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \

  510.                                                 int src_stride1, int src_stride2, int h){\

  511.     OPNAME ## _no_rnd_pixels8_l2(dst  , src1  , src2  , dst_stride, src_stride1, src_stride2, h);\

  512.     OPNAME ## _no_rnd_pixels8_l2(dst+8, src1+8, src2+8, dst_stride, src_stride1, src_stride2, h);\

  513. }\

  514. \

  515. static inline void OPNAME ## _no_rnd_pixels8_x2(uint8_t *block, const uint8_t *pixels, int line_size, int h){\

  516.     OPNAME ## _no_rnd_pixels8_l2(block, pixels, pixels+1, line_size, line_size, line_size, h);\

  517. }\

  518. \

  519. static inline void OPNAME ## _pixels8_x2(uint8_t *block, const uint8_t *pixels, int line_size, int h){\

  520.     OPNAME ## _pixels8_l2(block, pixels, pixels+1, line_size, line_size, line_size, h);\

  521. }\

  522. \

  523. static inline void OPNAME ## _no_rnd_pixels8_y2(uint8_t *block, const uint8_t *pixels, int line_size, int h){\

  524.     OPNAME ## _no_rnd_pixels8_l2(block, pixels, pixels+line_size, line_size, line_size, line_size, h);\

  525. }\

  526. \

  527. static inline void OPNAME ## _pixels8_y2(uint8_t *block, const uint8_t *pixels, int line_size, int h){\

  528.     OPNAME ## _pixels8_l2(block, pixels, pixels+line_size, line_size, line_size, line_size, h);\

  529. }\

  530. \

  531. static inline void OPNAME ## _pixels8_l4(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,\

  532.                  int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\

  533.     int i;\

  534.     for(i=0; i<h; i++){\

  535.         uint32_t a, b, c, d, l0, l1, h0, h1;\

  536.         a= LD32(&src1[i*src_stride1]);\

  537.         b= LD32(&src2[i*src_stride2]);\

  538.         c= LD32(&src3[i*src_stride3]);\

  539.         d= LD32(&src4[i*src_stride4]);\

  540.         l0=  (a&0x03030303UL)\

  541.            + (b&0x03030303UL)\

  542.            + 0x02020202UL;\

  543.         h0= ((a&0xFCFCFCFCUL)>>2)\

  544.           + ((b&0xFCFCFCFCUL)>>2);\

  545.         l1=  (c&0x03030303UL)\

  546.            + (d&0x03030303UL);\

  547.         h1= ((c&0xFCFCFCFCUL)>>2)\

  548.           + ((d&0xFCFCFCFCUL)>>2);\

  549.         OP(*((uint32_t*)&dst[i*dst_stride]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\

  550.         a= LD32(&src1[i*src_stride1+4]);\

  551.         b= LD32(&src2[i*src_stride2+4]);\

  552.         c= LD32(&src3[i*src_stride3+4]);\

  553.         d= LD32(&src4[i*src_stride4+4]);\

  554.         l0=  (a&0x03030303UL)\

  555.            + (b&0x03030303UL)\

  556.            + 0x02020202UL;\

  557.         h0= ((a&0xFCFCFCFCUL)>>2)\

  558.           + ((b&0xFCFCFCFCUL)>>2);\

  559.         l1=  (c&0x03030303UL)\

  560.            + (d&0x03030303UL);\

  561.         h1= ((c&0xFCFCFCFCUL)>>2)\

  562.           + ((d&0xFCFCFCFCUL)>>2);\

  563.         OP(*((uint32_t*)&dst[i*dst_stride+4]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\

  564.     }\

  565. }\

  566. static inline void OPNAME ## _no_rnd_pixels8_l4(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,\

  567.                  int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\

  568.     int i;\

  569.     for(i=0; i<h; i++){\

  570.         uint32_t a, b, c, d, l0, l1, h0, h1;\

  571.         a= LD32(&src1[i*src_stride1]);\

  572.         b= LD32(&src2[i*src_stride2]);\

  573.         c= LD32(&src3[i*src_stride3]);\

  574.         d= LD32(&src4[i*src_stride4]);\

  575.         l0=  (a&0x03030303UL)\

  576.            + (b&0x03030303UL)\

  577.            + 0x01010101UL;\

  578.         h0= ((a&0xFCFCFCFCUL)>>2)\

  579.           + ((b&0xFCFCFCFCUL)>>2);\

  580.         l1=  (c&0x03030303UL)\

  581.            + (d&0x03030303UL);\

  582.         h1= ((c&0xFCFCFCFCUL)>>2)\

  583.           + ((d&0xFCFCFCFCUL)>>2);\

  584.         OP(*((uint32_t*)&dst[i*dst_stride]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\

  585.         a= LD32(&src1[i*src_stride1+4]);\

  586.         b= LD32(&src2[i*src_stride2+4]);\

  587.         c= LD32(&src3[i*src_stride3+4]);\

  588.         d= LD32(&src4[i*src_stride4+4]);\

  589.         l0=  (a&0x03030303UL)\

  590.            + (b&0x03030303UL)\

  591.            + 0x01010101UL;\

  592.         h0= ((a&0xFCFCFCFCUL)>>2)\

  593.           + ((b&0xFCFCFCFCUL)>>2);\

  594.         l1=  (c&0x03030303UL)\

  595.            + (d&0x03030303UL);\

  596.         h1= ((c&0xFCFCFCFCUL)>>2)\

  597.           + ((d&0xFCFCFCFCUL)>>2);\

  598.         OP(*((uint32_t*)&dst[i*dst_stride+4]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\

  599.     }\

  600. }\

  601. static inline void OPNAME ## _pixels16_l4(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,\

  602.                  int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\

  603.     OPNAME ## _pixels8_l4(dst  , src1  , src2  , src3  , src4  , dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\

  604.     OPNAME ## _pixels8_l4(dst+8, src1+8, src2+8, src3+8, src4+8, dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\

  605. }\

  606. static inline void OPNAME ## _no_rnd_pixels16_l4(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,\

  607.                  int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\

  608.     OPNAME ## _no_rnd_pixels8_l4(dst  , src1  , src2  , src3  , src4  , dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\

  609.     OPNAME ## _no_rnd_pixels8_l4(dst+8, src1+8, src2+8, src3+8, src4+8, dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\

  610. }\

  611. \

  612. static inline void OPNAME ## _pixels8_xy2(uint8_t *block, const uint8_t *pixels, int line_size, int h)\

  613. {\

  614.     int j;\

  615.     for(j=0; j<2; j++){\

  616.         int i;\

  617.         const uint32_t a= LD32(pixels  );\

  618.         const uint32_t b= LD32(pixels+1);\

  619.         uint32_t l0=  (a&0x03030303UL)\

  620.                     + (b&0x03030303UL)\

  621.                     + 0x02020202UL;\

  622.         uint32_t h0= ((a&0xFCFCFCFCUL)>>2)\

  623.                    + ((b&0xFCFCFCFCUL)>>2);\

  624.         uint32_t l1,h1;\

  625. \

  626.         pixels+=line_size;\

  627.         for(i=0; i<h; i+=2){\

  628.             uint32_t a= LD32(pixels  );\

  629.             uint32_t b= LD32(pixels+1);\

  630.             l1=  (a&0x03030303UL)\

  631.                + (b&0x03030303UL);\

  632.             h1= ((a&0xFCFCFCFCUL)>>2)\

  633.               + ((b&0xFCFCFCFCUL)>>2);\

  634.             OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\

  635.             pixels+=line_size;\

  636.             block +=line_size;\

  637.             a= LD32(pixels  );\

  638.             b= LD32(pixels+1);\

  639.             l0=  (a&0x03030303UL)\

  640.                + (b&0x03030303UL)\

  641.                + 0x02020202UL;\

  642.             h0= ((a&0xFCFCFCFCUL)>>2)\

  643.               + ((b&0xFCFCFCFCUL)>>2);\

  644.             OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\

  645.             pixels+=line_size;\

  646.             block +=line_size;\

  647.         }\

  648.         pixels+=4-line_size*(h+1);\

  649.         block +=4-line_size*h;\

  650.     }\

  651. }\

  652. \

  653. static inline void OPNAME ## _no_rnd_pixels8_xy2(uint8_t *block, const uint8_t *pixels, int line_size, int h)\

  654. {\

  655.     int j;\

  656.     for(j=0; j<2; j++){\

  657.         int i;\

  658.         const uint32_t a= LD32(pixels  );\

  659.         const uint32_t b= LD32(pixels+1);\

  660.         uint32_t l0=  (a&0x03030303UL)\

  661.                     + (b&0x03030303UL)\

  662.                     + 0x01010101UL;\

  663.         uint32_t h0= ((a&0xFCFCFCFCUL)>>2)\

  664.                    + ((b&0xFCFCFCFCUL)>>2);\

  665.         uint32_t l1,h1;\

  666. \

  667.         pixels+=line_size;\

  668.         for(i=0; i<h; i+=2){\

  669.             uint32_t a= LD32(pixels  );\

  670.             uint32_t b= LD32(pixels+1);\

  671.             l1=  (a&0x03030303UL)\

  672.                + (b&0x03030303UL);\

  673.             h1= ((a&0xFCFCFCFCUL)>>2)\

  674.               + ((b&0xFCFCFCFCUL)>>2);\

  675.             OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\

  676.             pixels+=line_size;\

  677.             block +=line_size;\

  678.             a= LD32(pixels  );\

  679.             b= LD32(pixels+1);\

  680.             l0=  (a&0x03030303UL)\

  681.                + (b&0x03030303UL)\

  682.                + 0x01010101UL;\

  683.             h0= ((a&0xFCFCFCFCUL)>>2)\

  684.               + ((b&0xFCFCFCFCUL)>>2);\

  685.             OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\

  686.             pixels+=line_size;\

  687.             block +=line_size;\

  688.         }\

  689.         pixels+=4-line_size*(h+1);\

  690.         block +=4-line_size*h;\

  691.     }\

  692. }\

  693. \

  694. CALL_2X_PIXELS(OPNAME ## _pixels16    , OPNAME ## _pixels8    , 8)\

  695. CALL_2X_PIXELS(OPNAME ## _pixels16_x2 , OPNAME ## _pixels8_x2 , 8)\

  696. CALL_2X_PIXELS(OPNAME ## _pixels16_y2 , OPNAME ## _pixels8_y2 , 8)\

  697. CALL_2X_PIXELS(OPNAME ## _pixels16_xy2, OPNAME ## _pixels8_xy2, 8)\

  698. CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16    , OPNAME ## _pixels8    , 8)\

  699. CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_x2 , OPNAME ## _no_rnd_pixels8_x2 , 8)\

  700. CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_y2 , OPNAME ## _no_rnd_pixels8_y2 , 8)\

  701. CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_xy2, OPNAME ## _no_rnd_pixels8_xy2, 8)\

  702. \

  703. void (*OPNAME ## _pixels_tab[2][4])(uint8_t *block, const uint8_t *pixels, int line_size, int h) = {\

  704.     {\

  705.         OPNAME ## _pixels16,\

  706.         OPNAME ## _pixels16_x2,\

  707.         OPNAME ## _pixels16_y2,\

  708.         OPNAME ## _pixels16_xy2},\

  709.     {\

  710.         OPNAME ## _pixels8,\

  711.         OPNAME ## _pixels8_x2,\

  712.         OPNAME ## _pixels8_y2,\

  713.         OPNAME ## _pixels8_xy2},\

  714. };\

  715. \

  716. void (*OPNAME ## _no_rnd_pixels_tab[2][4])(uint8_t *block, const uint8_t *pixels, int line_size, int h) = {\

  717.     {\

  718.         OPNAME ## _pixels16,\

  719.         OPNAME ## _no_rnd_pixels16_x2,\

  720.         OPNAME ## _no_rnd_pixels16_y2,\

  721.         OPNAME ## _no_rnd_pixels16_xy2},\

  722.     {\

  723.         OPNAME ## _pixels8,\

  724.         OPNAME ## _no_rnd_pixels8_x2,\

  725.         OPNAME ## _no_rnd_pixels8_y2,\

  726.         OPNAME ## _no_rnd_pixels8_xy2},\

  727. };

  728. 

  729. #define op_avg(a, b) a = ( ((a)|(b)) - ((((a)^(b))&0xFEFEFEFEUL)>>1) )

  730. #endif

  731. #define op_put(a, b) a = b

  732. 

  733. PIXOP2(avg, op_avg)

  734. PIXOP2(put, op_put)

  735. #undef op_avg

  736. #undef op_put

  737. 

  738. #if 0

  739. /* FIXME this stuff could be removed as its ot really used anymore */

  740. #define PIXOP(BTYPE, OPNAME, OP, INCR)                                                   \

  741.                                                                                          \

  742. static void OPNAME ## _pixels(BTYPE *block, const UINT8 *pixels, int line_size, int h)    \

  743. {                                                                                        \

  744.     BTYPE *p;                                                                            \

  745.     const UINT8 *pix;                                                                    \

  746.                                                                                          \

  747.     p = block;                                                                           \

  748.     pix = pixels;                                                                        \

  749.     do {                                                                                 \

  750.         OP(p[0], pix[0]);                                                                  \

  751.         OP(p[1], pix[1]);                                                                  \

  752.         OP(p[2], pix[2]);                                                                  \

  753.         OP(p[3], pix[3]);                                                                  \

  754.         OP(p[4], pix[4]);                                                                  \

  755.         OP(p[5], pix[5]);                                                                  \

  756.         OP(p[6], pix[6]);                                                                  \

  757.         OP(p[7], pix[7]);                                                                  \

  758.         pix += line_size;                                                                \

  759.         p += INCR;                                                                       \

  760.     } while (--h);;                                                                       \

  761. }                                                                                        \

  762.                                                                                          \

  763. static void OPNAME ## _pixels_x2(BTYPE *block, const UINT8 *pixels, int line_size, int h)     \

  764. {                                                                                        \

  765.     BTYPE *p;                                                                          \

  766.     const UINT8 *pix;                                                                    \

  767.                                                                                          \

  768.     p = block;                                                                           \

  769.     pix = pixels;                                                                        \

  770.     do {                                                                   \

  771.         OP(p[0], avg2(pix[0], pix[1]));                                                    \

  772.         OP(p[1], avg2(pix[1], pix[2]));                                                    \

  773.         OP(p[2], avg2(pix[2], pix[3]));                                                    \

  774.         OP(p[3], avg2(pix[3], pix[4]));                                                    \

  775.         OP(p[4], avg2(pix[4], pix[5]));                                                    \

  776.         OP(p[5], avg2(pix[5], pix[6]));                                                    \

  777.         OP(p[6], avg2(pix[6], pix[7]));                                                    \

  778.         OP(p[7], avg2(pix[7], pix[8]));                                                    \

  779.         pix += line_size;                                                                \

  780.         p += INCR;                                                                       \

  781.     } while (--h);                                                                        \

  782. }                                                                                        \

  783.                                                                                          \

  784. static void OPNAME ## _pixels_y2(BTYPE *block, const UINT8 *pixels, int line_size, int h)     \

  785. {                                                                                        \

  786.     BTYPE *p;                                                                          \

  787.     const UINT8 *pix;                                                                    \

  788.     const UINT8 *pix1;                                                                   \

  789.                                                                                          \

  790.     p = block;                                                                           \

  791.     pix = pixels;                                                                        \

  792.     pix1 = pixels + line_size;                                                           \

  793.     do {                                                                                 \

  794.         OP(p[0], avg2(pix[0], pix1[0]));                                                   \

  795.         OP(p[1], avg2(pix[1], pix1[1]));                                                   \

  796.         OP(p[2], avg2(pix[2], pix1[2]));                                                   \

  797.         OP(p[3], avg2(pix[3], pix1[3]));                                                   \

  798.         OP(p[4], avg2(pix[4], pix1[4]));                                                   \

  799.         OP(p[5], avg2(pix[5], pix1[5]));                                                   \

  800.         OP(p[6], avg2(pix[6], pix1[6]));                                                   \

  801.         OP(p[7], avg2(pix[7], pix1[7]));                                                   \

  802.         pix += line_size;                                                                \

  803.         pix1 += line_size;                                                               \

  804.         p += INCR;                                                                       \

  805.     } while(--h);                                                                         \

  806. }                                                                                        \

  807.                                                                                          \

  808. static void OPNAME ## _pixels_xy2(BTYPE *block, const UINT8 *pixels, int line_size, int h)    \

  809. {                                                                                        \

  810.     BTYPE *p;                                                                          \

  811.     const UINT8 *pix;                                                                    \

  812.     const UINT8 *pix1;                                                                   \

  813.                                                                                          \

  814.     p = block;                                                                           \

  815.     pix = pixels;                                                                        \

  816.     pix1 = pixels + line_size;                                                           \

  817.     do {                                                                   \

  818.         OP(p[0], avg4(pix[0], pix[1], pix1[0], pix1[1]));                                  \

  819.         OP(p[1], avg4(pix[1], pix[2], pix1[1], pix1[2]));                                  \

  820.         OP(p[2], avg4(pix[2], pix[3], pix1[2], pix1[3]));                                  \

  821.         OP(p[3], avg4(pix[3], pix[4], pix1[3], pix1[4]));                                  \

  822.         OP(p[4], avg4(pix[4], pix[5], pix1[4], pix1[5]));                                  \

  823.         OP(p[5], avg4(pix[5], pix[6], pix1[5], pix1[6]));                                  \

  824.         OP(p[6], avg4(pix[6], pix[7], pix1[6], pix1[7]));                                  \

  825.         OP(p[7], avg4(pix[7], pix[8], pix1[7], pix1[8]));                                  \

  826.         pix += line_size;                                                                \

  827.         pix1 += line_size;                                                               \

  828.         p += INCR;                                                                       \

  829.     } while(--h);                                                                         \

  830. }                                                                                        \

  831.                                                                                          \

  832. void (*OPNAME ## _pixels_tab[4])(BTYPE *block, const UINT8 *pixels, int line_size, int h) = { \

  833.     OPNAME ## _pixels,                                                                   \

  834.     OPNAME ## _pixels_x2,                                                                \

  835.     OPNAME ## _pixels_y2,                                                                \

  836.     OPNAME ## _pixels_xy2,                                                               \

  837. };

  838. 

  839. /* rounding primitives */

  840. #define avg2(a,b) ((a+b+1)>>1)

  841. #define avg4(a,b,c,d) ((a+b+c+d+2)>>2)

  842. 

  843. #define op_avg(a, b) a = avg2(a, b)

  844. #define op_sub(a, b) a -= b

  845. #define op_put(a, b) a = b

  846. 

  847. PIXOP(DCTELEM, sub, op_sub, 8)

  848. PIXOP(uint8_t, avg, op_avg, line_size)

  849. PIXOP(uint8_t, put, op_put, line_size)

  850. 

  851. /* not rounding primitives */

  852. #undef avg2

  853. #undef avg4

  854. #define avg2(a,b) ((a+b)>>1)

  855. #define avg4(a,b,c,d) ((a+b+c+d+1)>>2)

  856. 

  857. PIXOP(uint8_t, avg_no_rnd, op_avg, line_size)

  858. PIXOP(uint8_t, put_no_rnd, op_put, line_size)

  859. /* motion estimation */

  860. 

  861. #undef avg2

  862. #undef avg4

  863. #endif

  864. 

  865. #define avg2(a,b) ((a+b+1)>>1)

  866. #define avg4(a,b,c,d) ((a+b+c+d+2)>>2)

  867. 

  868. static void gmc1_c(UINT8 *dst, UINT8 *src, int stride, int h, int x16, int y16, int rounder)

  869. {

  870.     const int A=(16-x16)*(16-y16);

  871.     const int B=(   x16)*(16-y16);

  872.     const int C=(16-x16)*(   y16);

  873.     const int D=(   x16)*(   y16);

  874.     int i;

  875.     rounder= 128 - rounder;

  876. 

  877.     for(i=0; i<h; i++)

  878.     {

  879.         dst[0]= (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + rounder)>>8;

  880.         dst[1]= (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + rounder)>>8;

  881.         dst[2]= (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + rounder)>>8;

  882.         dst[3]= (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + rounder)>>8;

  883.         dst[4]= (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + rounder)>>8;

  884.         dst[5]= (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + rounder)>>8;

  885.         dst[6]= (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + rounder)>>8;

  886.         dst[7]= (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + rounder)>>8;

  887.         dst+= stride;

  888.         src+= stride;

  889.     }

  890. }

  891. 

  892. static inline void copy_block17(UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int h)

  893. {

  894.     int i;

  895.     for(i=0; i<h; i++)

  896.     {

  897.         ST32(dst   , LD32(src   ));

  898.         ST32(dst+4 , LD32(src+4 ));

  899.         ST32(dst+8 , LD32(src+8 ));

  900.         ST32(dst+12, LD32(src+12));

  901.         dst[16]= src[16];

  902.         dst+=dstStride;

  903.         src+=srcStride;

  904.     }

  905. }

  906. 

  907. static inline void copy_block9(UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int h)

  908. {

  909.     int i;

  910.     for(i=0; i<h; i++)

  911.     {

  912.         ST32(dst   , LD32(src   ));

  913.         ST32(dst+4 , LD32(src+4 ));

  914.         dst[8]= src[8];

  915.         dst+=dstStride;

  916.         src+=srcStride;

  917.     }

  918. }

  919. 

  920. #define QPEL_MC(r, OPNAME, RND, OP) \

  921. static void OPNAME ## mpeg4_qpel8_h_lowpass(UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int h){\

  922.     UINT8 *cm = cropTbl + MAX_NEG_CROP;\

  923.     int i;\

  924.     for(i=0; i<h; i++)\

  925.     {\

  926.         OP(dst[0], (src[0]+src[1])*20 - (src[0]+src[2])*6 + (src[1]+src[3])*3 - (src[2]+src[4]));\

  927.         OP(dst[1], (src[1]+src[2])*20 - (src[0]+src[3])*6 + (src[0]+src[4])*3 - (src[1]+src[5]));\

  928.         OP(dst[2], (src[2]+src[3])*20 - (src[1]+src[4])*6 + (src[0]+src[5])*3 - (src[0]+src[6]));\

  929.         OP(dst[3], (src[3]+src[4])*20 - (src[2]+src[5])*6 + (src[1]+src[6])*3 - (src[0]+src[7]));\

  930.         OP(dst[4], (src[4]+src[5])*20 - (src[3]+src[6])*6 + (src[2]+src[7])*3 - (src[1]+src[8]));\

  931.         OP(dst[5], (src[5]+src[6])*20 - (src[4]+src[7])*6 + (src[3]+src[8])*3 - (src[2]+src[8]));\

  932.         OP(dst[6], (src[6]+src[7])*20 - (src[5]+src[8])*6 + (src[4]+src[8])*3 - (src[3]+src[7]));\

  933.         OP(dst[7], (src[7]+src[8])*20 - (src[6]+src[8])*6 + (src[5]+src[7])*3 - (src[4]+src[6]));\

  934.         dst+=dstStride;\

  935.         src+=srcStride;\

  936.     }\

  937. }\

  938. \

  939. static void OPNAME ## mpeg4_qpel8_v_lowpass(UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int w){\

  940.     UINT8 *cm = cropTbl + MAX_NEG_CROP;\

  941.     int i;\

  942.     for(i=0; i<w; i++)\

  943.     {\

  944.         const int src0= src[0*srcStride];\

  945.         const int src1= src[1*srcStride];\

  946.         const int src2= src[2*srcStride];\

  947.         const int src3= src[3*srcStride];\

  948.         const int src4= src[4*srcStride];\

  949.         const int src5= src[5*srcStride];\

  950.         const int src6= src[6*srcStride];\

  951.         const int src7= src[7*srcStride];\

  952.         const int src8= src[8*srcStride];\

  953.         OP(dst[0*dstStride], (src0+src1)*20 - (src0+src2)*6 + (src1+src3)*3 - (src2+src4));\

  954.         OP(dst[1*dstStride], (src1+src2)*20 - (src0+src3)*6 + (src0+src4)*3 - (src1+src5));\

  955.         OP(dst[2*dstStride], (src2+src3)*20 - (src1+src4)*6 + (src0+src5)*3 - (src0+src6));\

  956.         OP(dst[3*dstStride], (src3+src4)*20 - (src2+src5)*6 + (src1+src6)*3 - (src0+src7));\

  957.         OP(dst[4*dstStride], (src4+src5)*20 - (src3+src6)*6 + (src2+src7)*3 - (src1+src8));\

  958.         OP(dst[5*dstStride], (src5+src6)*20 - (src4+src7)*6 + (src3+src8)*3 - (src2+src8));\

  959.         OP(dst[6*dstStride], (src6+src7)*20 - (src5+src8)*6 + (src4+src8)*3 - (src3+src7));\

  960.         OP(dst[7*dstStride], (src7+src8)*20 - (src6+src8)*6 + (src5+src7)*3 - (src4+src6));\

  961.         dst++;\

  962.         src++;\

  963.     }\

  964. }\

  965. \

  966. static void OPNAME ## mpeg4_qpel16_h_lowpass(UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int h){\

  967.     UINT8 *cm = cropTbl + MAX_NEG_CROP;\

  968.     int i;\

  969.     for(i=0; i<h; i++)\

  970.     {\

  971.         OP(dst[ 0], (src[ 0]+src[ 1])*20 - (src[ 0]+src[ 2])*6 + (src[ 1]+src[ 3])*3 - (src[ 2]+src[ 4]));\

  972.         OP(dst[ 1], (src[ 1]+src[ 2])*20 - (src[ 0]+src[ 3])*6 + (src[ 0]+src[ 4])*3 - (src[ 1]+src[ 5]));\

  973.         OP(dst[ 2], (src[ 2]+src[ 3])*20 - (src[ 1]+src[ 4])*6 + (src[ 0]+src[ 5])*3 - (src[ 0]+src[ 6]));\

  974.         OP(dst[ 3], (src[ 3]+src[ 4])*20 - (src[ 2]+src[ 5])*6 + (src[ 1]+src[ 6])*3 - (src[ 0]+src[ 7]));\

  975.         OP(dst[ 4], (src[ 4]+src[ 5])*20 - (src[ 3]+src[ 6])*6 + (src[ 2]+src[ 7])*3 - (src[ 1]+src[ 8]));\

  976.         OP(dst[ 5], (src[ 5]+src[ 6])*20 - (src[ 4]+src[ 7])*6 + (src[ 3]+src[ 8])*3 - (src[ 2]+src[ 9]));\

  977.         OP(dst[ 6], (src[ 6]+src[ 7])*20 - (src[ 5]+src[ 8])*6 + (src[ 4]+src[ 9])*3 - (src[ 3]+src[10]));\

  978.         OP(dst[ 7], (src[ 7]+src[ 8])*20 - (src[ 6]+src[ 9])*6 + (src[ 5]+src[10])*3 - (src[ 4]+src[11]));\

  979.         OP(dst[ 8], (src[ 8]+src[ 9])*20 - (src[ 7]+src[10])*6 + (src[ 6]+src[11])*3 - (src[ 5]+src[12]));\

  980.         OP(dst[ 9], (src[ 9]+src[10])*20 - (src[ 8]+src[11])*6 + (src[ 7]+src[12])*3 - (src[ 6]+src[13]));\

  981.         OP(dst[10], (src[10]+src[11])*20 - (src[ 9]+src[12])*6 + (src[ 8]+src[13])*3 - (src[ 7]+src[14]));\

  982.         OP(dst[11], (src[11]+src[12])*20 - (src[10]+src[13])*6 + (src[ 9]+src[14])*3 - (src[ 8]+src[15]));\

  983.         OP(dst[12], (src[12]+src[13])*20 - (src[11]+src[14])*6 + (src[10]+src[15])*3 - (src[ 9]+src[16]));\

  984.         OP(dst[13], (src[13]+src[14])*20 - (src[12]+src[15])*6 + (src[11]+src[16])*3 - (src[10]+src[16]));\

  985.         OP(dst[14], (src[14]+src[15])*20 - (src[13]+src[16])*6 + (src[12]+src[16])*3 - (src[11]+src[15]));\

  986.         OP(dst[15], (src[15]+src[16])*20 - (src[14]+src[16])*6 + (src[13]+src[15])*3 - (src[12]+src[14]));\

  987.         dst+=dstStride;\

  988.         src+=srcStride;\

  989.     }\

  990. }\

  991. \

  992. static void OPNAME ## mpeg4_qpel16_v_lowpass(UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int w){\

  993.     UINT8 *cm = cropTbl + MAX_NEG_CROP;\

  994.     int i;\

  995.     for(i=0; i<w; i++)\

  996.     {\

  997.         const int src0= src[0*srcStride];\

  998.         const int src1= src[1*srcStride];\

  999.         const int src2= src[2*srcStride];\

  1000.         const int src3= src[3*srcStride];\

  1001.         const int src4= src[4*srcStride];\

  1002.         const int src5= src[5*srcStride];\

  1003.         const int src6= src[6*srcStride];\

  1004.         const int src7= src[7*srcStride];\

  1005.         const int src8= src[8*srcStride];\

  1006.         const int src9= src[9*srcStride];\

  1007.         const int src10= src[10*srcStride];\

  1008.         const int src11= src[11*srcStride];\

  1009.         const int src12= src[12*srcStride];\

  1010.         const int src13= src[13*srcStride];\

  1011.         const int src14= src[14*srcStride];\

  1012.         const int src15= src[15*srcStride];\

  1013.         const int src16= src[16*srcStride];\

  1014.         OP(dst[ 0*dstStride], (src0 +src1 )*20 - (src0 +src2 )*6 + (src1 +src3 )*3 - (src2 +src4 ));\

  1015.         OP(dst[ 1*dstStride], (src1 +src2 )*20 - (src0 +src3 )*6 + (src0 +src4 )*3 - (src1 +src5 ));\

  1016.         OP(dst[ 2*dstStride], (src2 +src3 )*20 - (src1 +src4 )*6 + (src0 +src5 )*3 - (src0 +src6 ));\

  1017.         OP(dst[ 3*dstStride], (src3 +src4 )*20 - (src2 +src5 )*6 + (src1 +src6 )*3 - (src0 +src7 ));\

  1018.         OP(dst[ 4*dstStride], (src4 +src5 )*20 - (src3 +src6 )*6 + (src2 +src7 )*3 - (src1 +src8 ));\

  1019.         OP(dst[ 5*dstStride], (src5 +src6 )*20 - (src4 +src7 )*6 + (src3 +src8 )*3 - (src2 +src9 ));\

  1020.         OP(dst[ 6*dstStride], (src6 +src7 )*20 - (src5 +src8 )*6 + (src4 +src9 )*3 - (src3 +src10));\

  1021.         OP(dst[ 7*dstStride], (src7 +src8 )*20 - (src6 +src9 )*6 + (src5 +src10)*3 - (src4 +src11));\

  1022.         OP(dst[ 8*dstStride], (src8 +src9 )*20 - (src7 +src10)*6 + (src6 +src11)*3 - (src5 +src12));\

  1023.         OP(dst[ 9*dstStride], (src9 +src10)*20 - (src8 +src11)*6 + (src7 +src12)*3 - (src6 +src13));\

  1024.         OP(dst[10*dstStride], (src10+src11)*20 - (src9 +src12)*6 + (src8 +src13)*3 - (src7 +src14));\

  1025.         OP(dst[11*dstStride], (src11+src12)*20 - (src10+src13)*6 + (src9 +src14)*3 - (src8 +src15));\

  1026.         OP(dst[12*dstStride], (src12+src13)*20 - (src11+src14)*6 + (src10+src15)*3 - (src9 +src16));\

  1027.         OP(dst[13*dstStride], (src13+src14)*20 - (src12+src15)*6 + (src11+src16)*3 - (src10+src16));\

  1028.         OP(dst[14*dstStride], (src14+src15)*20 - (src13+src16)*6 + (src12+src16)*3 - (src11+src15));\

  1029.         OP(dst[15*dstStride], (src15+src16)*20 - (src14+src16)*6 + (src13+src15)*3 - (src12+src14));\

  1030.         dst++;\

  1031.         src++;\

  1032.     }\

  1033. }\

  1034. \

  1035. static void OPNAME ## qpel8_mc00_c (UINT8 *dst, UINT8 *src, int stride){\

  1036.     OPNAME ## pixels8(dst, src, stride, 8);\

  1037. }\

  1038. \

  1039. static void OPNAME ## qpel8_mc10_c(UINT8 *dst, UINT8 *src, int stride){\

  1040.     UINT8 half[64];\

  1041.     put ## RND ## mpeg4_qpel8_h_lowpass(half, src, 8, stride, 8);\

  1042.     OPNAME ## pixels8_l2(dst, src, half, stride, stride, 8, 8);\

  1043. }\

  1044. \

  1045. static void OPNAME ## qpel8_mc20_c(UINT8 *dst, UINT8 *src, int stride){\

  1046.     OPNAME ## mpeg4_qpel8_h_lowpass(dst, src, stride, stride, 8);\

  1047. }\

  1048. \

  1049. static void OPNAME ## qpel8_mc30_c(UINT8 *dst, UINT8 *src, int stride){\

  1050.     UINT8 half[64];\

  1051.     put ## RND ## mpeg4_qpel8_h_lowpass(half, src, 8, stride, 8);\

  1052.     OPNAME ## pixels8_l2(dst, src+1, half, stride, stride, 8, 8);\

  1053. }\

  1054. \

  1055. static void OPNAME ## qpel8_mc01_c(UINT8 *dst, UINT8 *src, int stride){\

  1056.     UINT8 full[16*9];\

  1057.     UINT8 half[64];\

  1058.     copy_block9(full, src, 16, stride, 9);\

  1059.     put ## RND ## mpeg4_qpel8_v_lowpass(half, full, 8, 16, 8);\

  1060.     OPNAME ## pixels8_l2(dst, full, half, stride, 16, 8, 8);\

  1061. }\

  1062. \

  1063. static void OPNAME ## qpel8_mc02_c(UINT8 *dst, UINT8 *src, int stride){\

  1064.     UINT8 full[16*9];\

  1065.     copy_block9(full, src, 16, stride, 9);\

  1066.     OPNAME ## mpeg4_qpel8_v_lowpass(dst, full, stride, 16, 8);\

  1067. }\

  1068. \

  1069. static void OPNAME ## qpel8_mc03_c(UINT8 *dst, UINT8 *src, int stride){\

  1070.     UINT8 full[16*9];\

  1071.     UINT8 half[64];\

  1072.     copy_block9(full, src, 16, stride, 9);\

  1073.     put ## RND ## mpeg4_qpel8_v_lowpass(half, full, 8, 16, 8);\

  1074.     OPNAME ## pixels8_l2(dst, full+16, half, stride, 16, 8, 8);\

  1075. }\

  1076. static void OPNAME ## qpel8_mc11_c(UINT8 *dst, UINT8 *src, int stride){\

  1077.     UINT8 full[16*9];\

  1078.     UINT8 halfH[72];\

  1079.     UINT8 halfV[64];\

  1080.     UINT8 halfHV[64];\

  1081.     copy_block9(full, src, 16, stride, 9);\

  1082.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\

  1083.     put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full, 8, 16, 8);\

  1084.     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8, 8);\

  1085.     OPNAME ## pixels8_l4(dst, full, halfH, halfV, halfHV, stride, 16, 8, 8, 8, 8);\

  1086. }\

  1087. static void OPNAME ## qpel8_mc31_c(UINT8 *dst, UINT8 *src, int stride){\

  1088.     UINT8 full[16*9];\

  1089.     UINT8 halfH[72];\

  1090.     UINT8 halfV[64];\

  1091.     UINT8 halfHV[64];\

  1092.     copy_block9(full, src, 16, stride, 9);\

  1093.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\

  1094.     put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full+1, 8, 16, 8);\

  1095.     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8, 8);\

  1096.     OPNAME ## pixels8_l4(dst, full+1, halfH, halfV, halfHV, stride, 16, 8, 8, 8, 8);\

  1097. }\

  1098. static void OPNAME ## qpel8_mc13_c(UINT8 *dst, UINT8 *src, int stride){\

  1099.     UINT8 full[16*9];\

  1100.     UINT8 halfH[72];\

  1101.     UINT8 halfV[64];\

  1102.     UINT8 halfHV[64];\

  1103.     copy_block9(full, src, 16, stride, 9);\

  1104.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\

  1105.     put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full, 8, 16, 8);\

  1106.     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8, 8);\

  1107.     OPNAME ## pixels8_l4(dst, full+16, halfH+8, halfV, halfHV, stride, 16, 8, 8, 8, 8);\

  1108. }\

  1109. static void OPNAME ## qpel8_mc33_c(UINT8 *dst, UINT8 *src, int stride){\

  1110.     UINT8 full[16*9];\

  1111.     UINT8 halfH[72];\

  1112.     UINT8 halfV[64];\

  1113.     UINT8 halfHV[64];\

  1114.     copy_block9(full, src, 16, stride, 9);\

  1115.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full  , 8, 16, 9);\

  1116.     put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full+1, 8, 16, 8);\

  1117.     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8, 8);\

  1118.     OPNAME ## pixels8_l4(dst, full+17, halfH+8, halfV, halfHV, stride, 16, 8, 8, 8, 8);\

  1119. }\

  1120. static void OPNAME ## qpel8_mc21_c(UINT8 *dst, UINT8 *src, int stride){\

  1121.     UINT8 halfH[72];\

  1122.     UINT8 halfHV[64];\

  1123.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\

  1124.     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8, 8);\

  1125.     OPNAME ## pixels8_l2(dst, halfH, halfHV, stride, 8, 8, 8);\

  1126. }\

  1127. static void OPNAME ## qpel8_mc23_c(UINT8 *dst, UINT8 *src, int stride){\

  1128.     UINT8 halfH[72];\

  1129.     UINT8 halfHV[64];\

  1130.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\

  1131.     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8, 8);\

  1132.     OPNAME ## pixels8_l2(dst, halfH+8, halfHV, stride, 8, 8, 8);\

  1133. }\

  1134. static void OPNAME ## qpel8_mc12_c(UINT8 *dst, UINT8 *src, int stride){\

  1135.     UINT8 full[16*9];\

  1136.     UINT8 halfH[72];\

  1137.     UINT8 halfV[64];\

  1138.     UINT8 halfHV[64];\

  1139.     copy_block9(full, src, 16, stride, 9);\

  1140.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\

  1141.     put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full, 8, 16, 8);\

  1142.     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8, 8);\

  1143.     OPNAME ## pixels8_l2(dst, halfV, halfHV, stride, 8, 8, 8);\

  1144. }\

  1145. static void OPNAME ## qpel8_mc32_c(UINT8 *dst, UINT8 *src, int stride){\

  1146.     UINT8 full[16*9];\

  1147.     UINT8 halfH[72];\

  1148.     UINT8 halfV[64];\

  1149.     UINT8 halfHV[64];\

  1150.     copy_block9(full, src, 16, stride, 9);\

  1151.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\

  1152.     put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full+1, 8, 16, 8);\

  1153.     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8, 8);\

  1154.     OPNAME ## pixels8_l2(dst, halfV, halfHV, stride, 8, 8, 8);\

  1155. }\

  1156. static void OPNAME ## qpel8_mc22_c(UINT8 *dst, UINT8 *src, int stride){\

  1157.     UINT8 halfH[72];\

  1158.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\

  1159.     OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8, 8);\

  1160. }\

  1161. static void OPNAME ## qpel16_mc00_c (UINT8 *dst, UINT8 *src, int stride){\

  1162.     OPNAME ## pixels16(dst, src, stride, 16);\

  1163. }\

  1164. \

  1165. static void OPNAME ## qpel16_mc10_c(UINT8 *dst, UINT8 *src, int stride){\

  1166.     UINT8 half[256];\

  1167.     put ## RND ## mpeg4_qpel16_h_lowpass(half, src, 16, stride, 16);\

  1168.     OPNAME ## pixels16_l2(dst, src, half, stride, stride, 16, 16);\

  1169. }\

  1170. \

  1171. static void OPNAME ## qpel16_mc20_c(UINT8 *dst, UINT8 *src, int stride){\

  1172.     OPNAME ## mpeg4_qpel16_h_lowpass(dst, src, stride, stride, 16);\

  1173. }\

  1174. \

  1175. static void OPNAME ## qpel16_mc30_c(UINT8 *dst, UINT8 *src, int stride){\

  1176.     UINT8 half[256];\

  1177.     put ## RND ## mpeg4_qpel16_h_lowpass(half, src, 16, stride, 16);\

  1178.     OPNAME ## pixels16_l2(dst, src+1, half, stride, stride, 16, 16);\

  1179. }\

  1180. \

  1181. static void OPNAME ## qpel16_mc01_c(UINT8 *dst, UINT8 *src, int stride){\

  1182.     UINT8 full[24*17];\

  1183.     UINT8 half[256];\

  1184.     copy_block17(full, src, 24, stride, 17);\

  1185.     put ## RND ## mpeg4_qpel16_v_lowpass(half, full, 16, 24, 16);\

  1186.     OPNAME ## pixels16_l2(dst, full, half, stride, 24, 16, 16);\

  1187. }\

  1188. \

  1189. static void OPNAME ## qpel16_mc02_c(UINT8 *dst, UINT8 *src, int stride){\

  1190.     UINT8 full[24*17];\

  1191.     copy_block17(full, src, 24, stride, 17);\

  1192.     OPNAME ## mpeg4_qpel16_v_lowpass(dst, full, stride, 24, 16);\

  1193. }\

  1194. \

  1195. static void OPNAME ## qpel16_mc03_c(UINT8 *dst, UINT8 *src, int stride){\

  1196.     UINT8 full[24*17];\

  1197.     UINT8 half[256];\

  1198.     copy_block17(full, src, 24, stride, 17);\

  1199.     put ## RND ## mpeg4_qpel16_v_lowpass(half, full, 16, 24, 16);\

  1200.     OPNAME ## pixels16_l2(dst, full+24, half, stride, 24, 16, 16);\

  1201. }\

  1202. static void OPNAME ## qpel16_mc11_c(UINT8 *dst, UINT8 *src, int stride){\

  1203.     UINT8 full[24*17];\

  1204.     UINT8 halfH[272];\

  1205.     UINT8 halfV[256];\

  1206.     UINT8 halfHV[256];\

  1207.     copy_block17(full, src, 24, stride, 17);\

  1208.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\

  1209.     put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full, 16, 24, 16);\

  1210.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16, 16);\

  1211.     OPNAME ## pixels16_l4(dst, full, halfH, halfV, halfHV, stride, 24, 16, 16, 16, 16);\

  1212. }\

  1213. static void OPNAME ## qpel16_mc31_c(UINT8 *dst, UINT8 *src, int stride){\

  1214.     UINT8 full[24*17];\

  1215.     UINT8 halfH[272];\

  1216.     UINT8 halfV[256];\

  1217.     UINT8 halfHV[256];\

  1218.     copy_block17(full, src, 24, stride, 17);\

  1219.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\

  1220.     put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full+1, 16, 24, 16);\

  1221.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16, 16);\

  1222.     OPNAME ## pixels16_l4(dst, full+1, halfH, halfV, halfHV, stride, 24, 16, 16, 16, 16);\

  1223. }\

  1224. static void OPNAME ## qpel16_mc13_c(UINT8 *dst, UINT8 *src, int stride){\

  1225.     UINT8 full[24*17];\

  1226.     UINT8 halfH[272];\

  1227.     UINT8 halfV[256];\

  1228.     UINT8 halfHV[256];\

  1229.     copy_block17(full, src, 24, stride, 17);\

  1230.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\

  1231.     put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full, 16, 24, 16);\

  1232.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16, 16);\

  1233.     OPNAME ## pixels16_l4(dst, full+24, halfH+16, halfV, halfHV, stride, 24, 16, 16, 16, 16);\

  1234. }\

  1235. static void OPNAME ## qpel16_mc33_c(UINT8 *dst, UINT8 *src, int stride){\

  1236.     UINT8 full[24*17];\

  1237.     UINT8 halfH[272];\

  1238.     UINT8 halfV[256];\

  1239.     UINT8 halfHV[256];\

  1240.     copy_block17(full, src, 24, stride, 17);\

  1241.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full  , 16, 24, 17);\

  1242.     put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full+1, 16, 24, 16);\

  1243.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16, 16);\

  1244.     OPNAME ## pixels16_l4(dst, full+25, halfH+16, halfV, halfHV, stride, 24, 16, 16, 16, 16);\

  1245. }\

  1246. static void OPNAME ## qpel16_mc21_c(UINT8 *dst, UINT8 *src, int stride){\

  1247.     UINT8 halfH[272];\

  1248.     UINT8 halfHV[256];\

  1249.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\

  1250.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16, 16);\

  1251.     OPNAME ## pixels16_l2(dst, halfH, halfHV, stride, 16, 16, 16);\

  1252. }\

  1253. static void OPNAME ## qpel16_mc23_c(UINT8 *dst, UINT8 *src, int stride){\

  1254.     UINT8 halfH[272];\

  1255.     UINT8 halfHV[256];\

  1256.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\

  1257.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16, 16);\

  1258.     OPNAME ## pixels16_l2(dst, halfH+16, halfHV, stride, 16, 16, 16);\

  1259. }\

  1260. static void OPNAME ## qpel16_mc12_c(UINT8 *dst, UINT8 *src, int stride){\

  1261.     UINT8 full[24*17];\

  1262.     UINT8 halfH[272];\

  1263.     UINT8 halfV[256];\

  1264.     UINT8 halfHV[256];\

  1265.     copy_block17(full, src, 24, stride, 17);\

  1266.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\

  1267.     put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full, 16, 24, 16);\

  1268.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16, 16);\

  1269.     OPNAME ## pixels16_l2(dst, halfV, halfHV, stride, 16, 16, 16);\

  1270. }\

  1271. static void OPNAME ## qpel16_mc32_c(UINT8 *dst, UINT8 *src, int stride){\

  1272.     UINT8 full[24*17];\

  1273.     UINT8 halfH[272];\

  1274.     UINT8 halfV[256];\

  1275.     UINT8 halfHV[256];\

  1276.     copy_block17(full, src, 24, stride, 17);\

  1277.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\

  1278.     put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full+1, 16, 24, 16);\

  1279.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16, 16);\

  1280.     OPNAME ## pixels16_l2(dst, halfV, halfHV, stride, 16, 16, 16);\

  1281. }\

  1282. static void OPNAME ## qpel16_mc22_c(UINT8 *dst, UINT8 *src, int stride){\

  1283.     UINT8 halfH[272];\

  1284.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\

  1285.     OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16, 16);\

  1286. }\

  1287. qpel_mc_func OPNAME ## qpel_pixels_tab[2][16]={ \

  1288.   {\

  1289.     OPNAME ## qpel16_mc00_c,                                                                   \

  1290.     OPNAME ## qpel16_mc10_c,                                                                   \

  1291.     OPNAME ## qpel16_mc20_c,                                                                   \

  1292.     OPNAME ## qpel16_mc30_c,                                                                   \

  1293.     OPNAME ## qpel16_mc01_c,                                                                   \

  1294.     OPNAME ## qpel16_mc11_c,                                                                   \

  1295.     OPNAME ## qpel16_mc21_c,                                                                   \

  1296.     OPNAME ## qpel16_mc31_c,                                                                   \

  1297.     OPNAME ## qpel16_mc02_c,                                                                   \

  1298.     OPNAME ## qpel16_mc12_c,                                                                   \

  1299.     OPNAME ## qpel16_mc22_c,                                                                   \

  1300.     OPNAME ## qpel16_mc32_c,                                                                   \

  1301.     OPNAME ## qpel16_mc03_c,                                                                   \

  1302.     OPNAME ## qpel16_mc13_c,                                                                   \

  1303.     OPNAME ## qpel16_mc23_c,                                                                   \

  1304.     OPNAME ## qpel16_mc33_c,                                                                   \

  1305.   },{\

  1306.     OPNAME ## qpel8_mc00_c,                                                                   \

  1307.     OPNAME ## qpel8_mc10_c,                                                                   \

  1308.     OPNAME ## qpel8_mc20_c,                                                                   \

  1309.     OPNAME ## qpel8_mc30_c,                                                                   \

  1310.     OPNAME ## qpel8_mc01_c,                                                                   \

  1311.     OPNAME ## qpel8_mc11_c,                                                                   \

  1312.     OPNAME ## qpel8_mc21_c,                                                                   \

  1313.     OPNAME ## qpel8_mc31_c,                                                                   \

  1314.     OPNAME ## qpel8_mc02_c,                                                                   \

  1315.     OPNAME ## qpel8_mc12_c,                                                                   \

  1316.     OPNAME ## qpel8_mc22_c,                                                                   \

  1317.     OPNAME ## qpel8_mc32_c,                                                                   \

  1318.     OPNAME ## qpel8_mc03_c,                                                                   \

  1319.     OPNAME ## qpel8_mc13_c,                                                                   \

  1320.     OPNAME ## qpel8_mc23_c,                                                                   \

  1321.     OPNAME ## qpel8_mc33_c,                                                                   \

  1322.   }\

  1323. };

  1324. 

  1325. #define op_avg(a, b) a = (((a)+cm[((b) + 16)>>5]+1)>>1)

  1326. #define op_avg_no_rnd(a, b) a = (((a)+cm[((b) + 15)>>5])>>1)

  1327. #define op_put(a, b) a = cm[((b) + 16)>>5]

  1328. #define op_put_no_rnd(a, b) a = cm[((b) + 15)>>5]

  1329. 

  1330. QPEL_MC(0, put_       , _       , op_put)

  1331. QPEL_MC(1, put_no_rnd_, _no_rnd_, op_put_no_rnd)

  1332. QPEL_MC(0, avg_       , _       , op_avg)

  1333. //QPEL_MC(1, avg_no_rnd , _       , op_avg)

  1334. #undef op_avg

  1335. #undef op_avg_no_rnd

  1336. #undef op_put

  1337. #undef op_put_no_rnd

  1338. 

  1339. int pix_abs16x16_c(UINT8 *pix1, UINT8 *pix2, int line_size)

  1340. {

  1341.     int s, i;

  1342. 

  1343.     s = 0;

  1344.     for(i=0;i<16;i++) {

  1345.         s += abs(pix1[0] - pix2[0]);

  1346.         s += abs(pix1[1] - pix2[1]);

  1347.         s += abs(pix1[2] - pix2[2]);

  1348.         s += abs(pix1[3] - pix2[3]);

  1349.         s += abs(pix1[4] - pix2[4]);

  1350.         s += abs(pix1[5] - pix2[5]);

  1351.         s += abs(pix1[6] - pix2[6]);

  1352.         s += abs(pix1[7] - pix2[7]);

  1353.         s += abs(pix1[8] - pix2[8]);

  1354.         s += abs(pix1[9] - pix2[9]);

  1355.         s += abs(pix1[10] - pix2[10]);

  1356.         s += abs(pix1[11] - pix2[11]);

  1357.         s += abs(pix1[12] - pix2[12]);

  1358.         s += abs(pix1[13] - pix2[13]);

  1359.         s += abs(pix1[14] - pix2[14]);

  1360.         s += abs(pix1[15] - pix2[15]);

  1361.         pix1 += line_size;

  1362.         pix2 += line_size;

  1363.     }

  1364.     return s;

  1365. }

  1366. 

  1367. int pix_abs16x16_x2_c(UINT8 *pix1, UINT8 *pix2, int line_size)

  1368. {

  1369.     int s, i;

  1370. 

  1371.     s = 0;

  1372.     for(i=0;i<16;i++) {

  1373.         s += abs(pix1[0] - avg2(pix2[0], pix2[1]));

  1374.         s += abs(pix1[1] - avg2(pix2[1], pix2[2]));

  1375.         s += abs(pix1[2] - avg2(pix2[2], pix2[3]));

  1376.         s += abs(pix1[3] - avg2(pix2[3], pix2[4]));

  1377.         s += abs(pix1[4] - avg2(pix2[4], pix2[5]));

  1378.         s += abs(pix1[5] - avg2(pix2[5], pix2[6]));

  1379.         s += abs(pix1[6] - avg2(pix2[6], pix2[7]));

  1380.         s += abs(pix1[7] - avg2(pix2[7], pix2[8]));

  1381.         s += abs(pix1[8] - avg2(pix2[8], pix2[9]));

  1382.         s += abs(pix1[9] - avg2(pix2[9], pix2[10]));

  1383.         s += abs(pix1[10] - avg2(pix2[10], pix2[11]));

  1384.         s += abs(pix1[11] - avg2(pix2[11], pix2[12]));

  1385.         s += abs(pix1[12] - avg2(pix2[12], pix2[13]));

  1386.         s += abs(pix1[13] - avg2(pix2[13], pix2[14]));

  1387.         s += abs(pix1[14] - avg2(pix2[14], pix2[15]));

  1388.         s += abs(pix1[15] - avg2(pix2[15], pix2[16]));

  1389.         pix1 += line_size;

  1390.         pix2 += line_size;

  1391.     }

  1392.     return s;

  1393. }

  1394. 

  1395. int pix_abs16x16_y2_c(UINT8 *pix1, UINT8 *pix2, int line_size)

  1396. {

  1397.     int s, i;

  1398.     UINT8 *pix3 = pix2 + line_size;

  1399. 

  1400.     s = 0;

  1401.     for(i=0;i<16;i++) {

  1402.         s += abs(pix1[0] - avg2(pix2[0], pix3[0]));

  1403.         s += abs(pix1[1] - avg2(pix2[1], pix3[1]));

  1404.         s += abs(pix1[2] - avg2(pix2[2], pix3[2]));

  1405.         s += abs(pix1[3] - avg2(pix2[3], pix3[3]));

  1406.         s += abs(pix1[4] - avg2(pix2[4], pix3[4]));

  1407.         s += abs(pix1[5] - avg2(pix2[5], pix3[5]));

  1408.         s += abs(pix1[6] - avg2(pix2[6], pix3[6]));

  1409.         s += abs(pix1[7] - avg2(pix2[7], pix3[7]));

  1410.         s += abs(pix1[8] - avg2(pix2[8], pix3[8]));

  1411.         s += abs(pix1[9] - avg2(pix2[9], pix3[9]));

  1412.         s += abs(pix1[10] - avg2(pix2[10], pix3[10]));

  1413.         s += abs(pix1[11] - avg2(pix2[11], pix3[11]));

  1414.         s += abs(pix1[12] - avg2(pix2[12], pix3[12]));

  1415.         s += abs(pix1[13] - avg2(pix2[13], pix3[13]));

  1416.         s += abs(pix1[14] - avg2(pix2[14], pix3[14]));

  1417.         s += abs(pix1[15] - avg2(pix2[15], pix3[15]));

  1418.         pix1 += line_size;

  1419.         pix2 += line_size;

  1420.         pix3 += line_size;

  1421.     }

  1422.     return s;

  1423. }

  1424. 

  1425. int pix_abs16x16_xy2_c(UINT8 *pix1, UINT8 *pix2, int line_size)

  1426. {

  1427.     int s, i;

  1428.     UINT8 *pix3 = pix2 + line_size;

  1429. 

  1430.     s = 0;

  1431.     for(i=0;i<16;i++) {

  1432.         s += abs(pix1[0] - avg4(pix2[0], pix2[1], pix3[0], pix3[1]));

  1433.         s += abs(pix1[1] - avg4(pix2[1], pix2[2], pix3[1], pix3[2]));

  1434.         s += abs(pix1[2] - avg4(pix2[2], pix2[3], pix3[2], pix3[3]));

  1435.         s += abs(pix1[3] - avg4(pix2[3], pix2[4], pix3[3], pix3[4]));

  1436.         s += abs(pix1[4] - avg4(pix2[4], pix2[5], pix3[4], pix3[5]));

  1437.         s += abs(pix1[5] - avg4(pix2[5], pix2[6], pix3[5], pix3[6]));

  1438.         s += abs(pix1[6] - avg4(pix2[6], pix2[7], pix3[6], pix3[7]));

  1439.         s += abs(pix1[7] - avg4(pix2[7], pix2[8], pix3[7], pix3[8]));

  1440.         s += abs(pix1[8] - avg4(pix2[8], pix2[9], pix3[8], pix3[9]));

  1441.         s += abs(pix1[9] - avg4(pix2[9], pix2[10], pix3[9], pix3[10]));

  1442.         s += abs(pix1[10] - avg4(pix2[10], pix2[11], pix3[10], pix3[11]));

  1443.         s += abs(pix1[11] - avg4(pix2[11], pix2[12], pix3[11], pix3[12]));

  1444.         s += abs(pix1[12] - avg4(pix2[12], pix2[13], pix3[12], pix3[13]));

  1445.         s += abs(pix1[13] - avg4(pix2[13], pix2[14], pix3[13], pix3[14]));

  1446.         s += abs(pix1[14] - avg4(pix2[14], pix2[15], pix3[14], pix3[15]));

  1447.         s += abs(pix1[15] - avg4(pix2[15], pix2[16], pix3[15], pix3[16]));

  1448.         pix1 += line_size;

  1449.         pix2 += line_size;

  1450.         pix3 += line_size;

  1451.     }

  1452.     return s;

  1453. }

  1454. 

  1455. int pix_abs8x8_c(UINT8 *pix1, UINT8 *pix2, int line_size)

  1456. {

  1457.     int s, i;

  1458. 

  1459.     s = 0;

  1460.     for(i=0;i<8;i++) {

  1461.         s += abs(pix1[0] - pix2[0]);

  1462.         s += abs(pix1[1] - pix2[1]);

  1463.         s += abs(pix1[2] - pix2[2]);

  1464.         s += abs(pix1[3] - pix2[3]);

  1465.         s += abs(pix1[4] - pix2[4]);

  1466.         s += abs(pix1[5] - pix2[5]);

  1467.         s += abs(pix1[6] - pix2[6]);

  1468.         s += abs(pix1[7] - pix2[7]);

  1469.         pix1 += line_size;

  1470.         pix2 += line_size;

  1471.     }

  1472.     return s;

  1473. }

  1474. 

  1475. int pix_abs8x8_x2_c(UINT8 *pix1, UINT8 *pix2, int line_size)

  1476. {

  1477.     int s, i;

  1478. 

  1479.     s = 0;

  1480.     for(i=0;i<8;i++) {

  1481.         s += abs(pix1[0] - avg2(pix2[0], pix2[1]));

  1482.         s += abs(pix1[1] - avg2(pix2[1], pix2[2]));

  1483.         s += abs(pix1[2] - avg2(pix2[2], pix2[3]));

  1484.         s += abs(pix1[3] - avg2(pix2[3], pix2[4]));

  1485.         s += abs(pix1[4] - avg2(pix2[4], pix2[5]));

  1486.         s += abs(pix1[5] - avg2(pix2[5], pix2[6]));

  1487.         s += abs(pix1[6] - avg2(pix2[6], pix2[7]));

  1488.         s += abs(pix1[7] - avg2(pix2[7], pix2[8]));

  1489.         pix1 += line_size;

  1490.         pix2 += line_size;

  1491.     }

  1492.     return s;

  1493. }

  1494. 

  1495. int pix_abs8x8_y2_c(UINT8 *pix1, UINT8 *pix2, int line_size)

  1496. {

  1497.     int s, i;

  1498.     UINT8 *pix3 = pix2 + line_size;

  1499. 

  1500.     s = 0;

  1501.     for(i=0;i<8;i++) {

  1502.         s += abs(pix1[0] - avg2(pix2[0], pix3[0]));

  1503.         s += abs(pix1[1] - avg2(pix2[1], pix3[1]));

  1504.         s += abs(pix1[2] - avg2(pix2[2], pix3[2]));

  1505.         s += abs(pix1[3] - avg2(pix2[3], pix3[3]));

  1506.         s += abs(pix1[4] - avg2(pix2[4], pix3[4]));

  1507.         s += abs(pix1[5] - avg2(pix2[5], pix3[5]));

  1508.         s += abs(pix1[6] - avg2(pix2[6], pix3[6]));

  1509.         s += abs(pix1[7] - avg2(pix2[7], pix3[7]));

  1510.         pix1 += line_size;

  1511.         pix2 += line_size;

  1512.         pix3 += line_size;

  1513.     }

  1514.     return s;

  1515. }

  1516. 

  1517. int pix_abs8x8_xy2_c(UINT8 *pix1, UINT8 *pix2, int line_size)

  1518. {

  1519.     int s, i;

  1520.     UINT8 *pix3 = pix2 + line_size;

  1521. 

  1522.     s = 0;

  1523.     for(i=0;i<8;i++) {

  1524.         s += abs(pix1[0] - avg4(pix2[0], pix2[1], pix3[0], pix3[1]));

  1525.         s += abs(pix1[1] - avg4(pix2[1], pix2[2], pix3[1], pix3[2]));

  1526.         s += abs(pix1[2] - avg4(pix2[2], pix2[3], pix3[2], pix3[3]));

  1527.         s += abs(pix1[3] - avg4(pix2[3], pix2[4], pix3[3], pix3[4]));

  1528.         s += abs(pix1[4] - avg4(pix2[4], pix2[5], pix3[4], pix3[5]));

  1529.         s += abs(pix1[5] - avg4(pix2[5], pix2[6], pix3[5], pix3[6]));

  1530.         s += abs(pix1[6] - avg4(pix2[6], pix2[7], pix3[6], pix3[7]));

  1531.         s += abs(pix1[7] - avg4(pix2[7], pix2[8], pix3[7], pix3[8]));

  1532.         pix1 += line_size;

  1533.         pix2 += line_size;

  1534.         pix3 += line_size;

  1535.     }

  1536.     return s;

  1537. }

  1538. 

  1539. /* permute block according so that it corresponds to the MMX idct

  1540.    order */

  1541. #ifdef SIMPLE_IDCT

  1542.  /* general permutation, but perhaps slightly slower */

  1543. void block_permute(INT16 *block)

  1544. {

  1545. 	int i;

  1546. 	INT16 temp[64];

  1547. 

  1548. 	for(i=0; i<64; i++) temp[ block_permute_op(i) ] = block[i];

  1549. 

  1550. 	for(i=0; i<64; i++) block[i] = temp[i];

  1551. }

  1552. #else

  1553. 

  1554. void block_permute(INT16 *block)

  1555. {

  1556.     int tmp1, tmp2, tmp3, tmp4, tmp5, tmp6;

  1557.     int i;

  1558. 

  1559.     for(i=0;i<8;i++) {

  1560.         tmp1 = block[1];

  1561.         tmp2 = block[2];

  1562.         tmp3 = block[3];

  1563.         tmp4 = block[4];

  1564.         tmp5 = block[5];

  1565.         tmp6 = block[6];

  1566.         block[1] = tmp2;

  1567.         block[2] = tmp4;

  1568.         block[3] = tmp6;

  1569.         block[4] = tmp1;

  1570.         block[5] = tmp3;

  1571.         block[6] = tmp5;

  1572.         block += 8;

  1573.     }

  1574. }

  1575. #endif

  1576. 

  1577. void clear_blocks_c(DCTELEM *blocks)

  1578. {

  1579.     memset(blocks, 0, sizeof(DCTELEM)*6*64);

  1580. }

  1581. 

  1582. /* XXX: those functions should be suppressed ASAP when all IDCTs are

  1583.    converted */

  1584. void gen_idct_put(UINT8 *dest, int line_size, DCTELEM *block)

  1585. {

  1586.     ff_idct (block);

  1587.     put_pixels_clamped(block, dest, line_size);

  1588. }

  1589. 

  1590. void gen_idct_add(UINT8 *dest, int line_size, DCTELEM *block)

  1591. {

  1592.     ff_idct (block);

  1593.     add_pixels_clamped(block, dest, line_size);

  1594. }

  1595. 

  1596. void dsputil_init(void)

  1597. {

  1598.     int i, j;

  1599.     int use_permuted_idct;

  1600. 

  1601.     for(i=0;i<256;i++) cropTbl[i + MAX_NEG_CROP] = i;

  1602.     for(i=0;i<MAX_NEG_CROP;i++) {

  1603.         cropTbl[i] = 0;

  1604.         cropTbl[i + MAX_NEG_CROP + 256] = 255;

  1605.     }

  1606. 

  1607.     for(i=0;i<512;i++) {

  1608.         squareTbl[i] = (i - 256) * (i - 256);

  1609.     }

  1610. 

  1611. #ifdef SIMPLE_IDCT

  1612.     ff_idct = NULL;

  1613. #else

  1614.     ff_idct = j_rev_dct;

  1615. #endif

  1616.     get_pixels = get_pixels_c;

  1617.     diff_pixels = diff_pixels_c;

  1618.     put_pixels_clamped = put_pixels_clamped_c;

  1619.     add_pixels_clamped = add_pixels_clamped_c;

  1620.     gmc1= gmc1_c;

  1621.     clear_blocks= clear_blocks_c;

  1622.     pix_sum= pix_sum_c;

  1623.     pix_norm1= pix_norm1_c;

  1624. 

  1625.     pix_abs16x16     = pix_abs16x16_c;

  1626.     pix_abs16x16_x2  = pix_abs16x16_x2_c;

  1627.     pix_abs16x16_y2  = pix_abs16x16_y2_c;

  1628.     pix_abs16x16_xy2 = pix_abs16x16_xy2_c;

  1629.     pix_abs8x8     = pix_abs8x8_c;

  1630.     pix_abs8x8_x2  = pix_abs8x8_x2_c;

  1631.     pix_abs8x8_y2  = pix_abs8x8_y2_c;

  1632.     pix_abs8x8_xy2 = pix_abs8x8_xy2_c;

  1633. 

  1634.     use_permuted_idct = 1;

  1635. 

  1636. #ifdef HAVE_MMX

  1637.     dsputil_init_mmx();

  1638. #endif

  1639. #ifdef ARCH_ARMV4L

  1640.     dsputil_init_armv4l();

  1641. #endif

  1642. #ifdef HAVE_MLIB

  1643.     dsputil_init_mlib();

  1644.     use_permuted_idct = 0;

  1645. #endif

  1646. #ifdef ARCH_ALPHA

  1647.     dsputil_init_alpha();

  1648.     use_permuted_idct = 0;

  1649. #endif

  1650. #ifdef ARCH_POWERPC

  1651.     dsputil_init_ppc();

  1652. #endif

  1653. 

  1654. #ifdef SIMPLE_IDCT

  1655.     if (ff_idct == NULL) {

  1656.         ff_idct_put = simple_idct_put;

  1657.         ff_idct_add = simple_idct_add;

  1658.         use_permuted_idct=0;

  1659.     }

  1660. #endif

  1661.     if(ff_idct != NULL) {

  1662.         ff_idct_put = gen_idct_put;

  1663.         ff_idct_add = gen_idct_add;

  1664.     }

  1665. 

  1666.     if(use_permuted_idct)

  1667. #ifdef SIMPLE_IDCT

  1668.         for(i=0; i<64; i++) permutation[i]= simple_mmx_permutation[i];

  1669. #else

  1670.         for(i=0; i<64; i++) permutation[i]= (i & 0x38) | ((i & 6) >> 1) | ((i & 1) << 2);

  1671. #endif

  1672.     else

  1673.         for(i=0; i<64; i++) permutation[i]=i;

  1674. 

  1675.     for(i=0; i<64; i++) inv_zigzag_direct16[zigzag_direct[i]]= i+1;

  1676.     for(i=0; i<64; i++) zigzag_direct_noperm[i]= zigzag_direct[i];

  1677.     

  1678.     if (use_permuted_idct) {

  1679.         /* permute for IDCT */

  1680.         for(i=0;i<64;i++) {

  1681.             j = zigzag_direct[i];

  1682.             zigzag_direct[i] = block_permute_op(j);

  1683.             j = ff_alternate_horizontal_scan[i];

  1684.             ff_alternate_horizontal_scan[i] = block_permute_op(j);

  1685.             j = ff_alternate_vertical_scan[i];

  1686.             ff_alternate_vertical_scan[i] = block_permute_op(j);

  1687.         }

  1688.         block_permute(ff_mpeg1_default_intra_matrix);

  1689.         block_permute(ff_mpeg1_default_non_intra_matrix);

  1690.         block_permute(ff_mpeg4_default_intra_matrix);

  1691.         block_permute(ff_mpeg4_default_non_intra_matrix);

  1692.     }

  1693.     

  1694.     build_zigzag_end();

  1695. }

  1696. 

  1697. /* remove any non bit exact operation (testing purpose) */

  1698. void avcodec_set_bit_exact(void)

  1699. {

  1700. #ifdef HAVE_MMX

  1701.     dsputil_set_bit_exact_mmx();

  1702. #endif

  1703. }

  1704. 

  1705. void get_psnr(UINT8 *orig_image[3], UINT8 *coded_image[3],

  1706.               int orig_linesize[3], int coded_linesize,

  1707.               AVCodecContext *avctx)

  1708. {

  1709.     int quad, diff, x, y;

  1710.     UINT8 *orig, *coded;

  1711.     UINT32 *sq = squareTbl + 256;

  1712.     

  1713.     quad = 0;

  1714.     diff = 0;

  1715.     

  1716.     /* Luminance */

  1717.     orig = orig_image[0];

  1718.     coded = coded_image[0];

  1719.     

  1720.     for (y=0;y<avctx->height;y++) {

  1721.         for (x=0;x<avctx->width;x++) {

  1722.             diff = *(orig + x) - *(coded + x);

  1723.             quad += sq[diff];

  1724.         }

  1725.         orig += orig_linesize[0];

  1726.         coded += coded_linesize;

  1727.     }

  1728.    

  1729.     avctx->psnr_y = (float) quad / (float) (avctx->width * avctx->height);

  1730.     

  1731.     if (avctx->psnr_y) {

  1732.         avctx->psnr_y = (float) (255 * 255) / avctx->psnr_y;

  1733.         avctx->psnr_y = 10 * (float) log10 (avctx->psnr_y); 

  1734.     } else

  1735.         avctx->psnr_y = 99.99;

  1736. }