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.
64484 Commits
32 Branches
325MB
Tree: 462c6cdb8e
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '462c6cdb8e'
${ noResults }
FFmpeg/libavcodec/dsputil.c

1185 lines
39KB
Raw Blame History 

  1. /*

  2.  * DSP utils

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

  4.  * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>

  5.  *

  6.  * This file is part of FFmpeg.

  7.  *

  8.  * FFmpeg is free software; you can redistribute it and/or

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

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

  11.  * version 2.1 of the License, or (at your option) any later version.

  12.  *

  13.  * FFmpeg is distributed in the hope that it will be useful,

  14.  * but WITHOUT ANY WARRANTY; without even the implied warranty of

  15.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU

  16.  * Lesser General Public License for more details.

  17.  *

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

  19.  * License along with FFmpeg; if not, write to the Free Software

  20.  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

  21.  */

  22. 

  23. /**

  24.  * @file

  25.  * DSP utils

  26.  */

  27. 

  28. #include "libavutil/attributes.h"

  29. #include "libavutil/imgutils.h"

  30. #include "libavutil/internal.h"

  31. #include "avcodec.h"

  32. #include "copy_block.h"

  33. #include "dct.h"

  34. #include "dsputil.h"

  35. #include "simple_idct.h"

  36. #include "faandct.h"

  37. #include "imgconvert.h"

  38. #include "mathops.h"

  39. #include "mpegvideo.h"

  40. #include "config.h"

  41. 

  42. uint32_t ff_square_tab[512] = { 0, };

  43. 

  44. #define BIT_DEPTH 16

  45. #include "dsputilenc_template.c"

  46. #undef BIT_DEPTH

  47. 

  48. #define BIT_DEPTH 8

  49. #include "dsputilenc_template.c"

  50. 

  51. static int pix_sum_c(uint8_t *pix, int line_size)

  52. {

  53.     int s = 0, i, j;

  54. 

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

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

  57.             s   += pix[0];

  58.             s   += pix[1];

  59.             s   += pix[2];

  60.             s   += pix[3];

  61.             s   += pix[4];

  62.             s   += pix[5];

  63.             s   += pix[6];

  64.             s   += pix[7];

  65.             pix += 8;

  66.         }

  67.         pix += line_size - 16;

  68.     }

  69.     return s;

  70. }

  71. 

  72. static int pix_norm1_c(uint8_t *pix, int line_size)

  73. {

  74.     int s = 0, i, j;

  75.     uint32_t *sq = ff_square_tab + 256;

  76. 

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

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

  79. #if 0

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

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

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

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

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

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

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

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

  88. #else

  89. #if HAVE_FAST_64BIT

  90.             register uint64_t x = *(uint64_t *) pix;

  91.             s += sq[x         & 0xff];

  92.             s += sq[(x >>  8) & 0xff];

  93.             s += sq[(x >> 16) & 0xff];

  94.             s += sq[(x >> 24) & 0xff];

  95.             s += sq[(x >> 32) & 0xff];

  96.             s += sq[(x >> 40) & 0xff];

  97.             s += sq[(x >> 48) & 0xff];

  98.             s += sq[(x >> 56) & 0xff];

  99. #else

  100.             register uint32_t x = *(uint32_t *) pix;

  101.             s += sq[x         & 0xff];

  102.             s += sq[(x >>  8) & 0xff];

  103.             s += sq[(x >> 16) & 0xff];

  104.             s += sq[(x >> 24) & 0xff];

  105.             x  = *(uint32_t *) (pix + 4);

  106.             s += sq[x         & 0xff];

  107.             s += sq[(x >>  8) & 0xff];

  108.             s += sq[(x >> 16) & 0xff];

  109.             s += sq[(x >> 24) & 0xff];

  110. #endif

  111. #endif

  112.             pix += 8;

  113.         }

  114.         pix += line_size - 16;

  115.     }

  116.     return s;

  117. }

  118. 

  119. static int sse4_c(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,

  120.                   int line_size, int h)

  121. {

  122.     int s = 0, i;

  123.     uint32_t *sq = ff_square_tab + 256;

  124. 

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

  126.         s    += sq[pix1[0] - pix2[0]];

  127.         s    += sq[pix1[1] - pix2[1]];

  128.         s    += sq[pix1[2] - pix2[2]];

  129.         s    += sq[pix1[3] - pix2[3]];

  130.         pix1 += line_size;

  131.         pix2 += line_size;

  132.     }

  133.     return s;

  134. }

  135. 

  136. static int sse8_c(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,

  137.                   int line_size, int h)

  138. {

  139.     int s = 0, i;

  140.     uint32_t *sq = ff_square_tab + 256;

  141. 

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

  143.         s    += sq[pix1[0] - pix2[0]];

  144.         s    += sq[pix1[1] - pix2[1]];

  145.         s    += sq[pix1[2] - pix2[2]];

  146.         s    += sq[pix1[3] - pix2[3]];

  147.         s    += sq[pix1[4] - pix2[4]];

  148.         s    += sq[pix1[5] - pix2[5]];

  149.         s    += sq[pix1[6] - pix2[6]];

  150.         s    += sq[pix1[7] - pix2[7]];

  151.         pix1 += line_size;

  152.         pix2 += line_size;

  153.     }

  154.     return s;

  155. }

  156. 

  157. static int sse16_c(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,

  158.                    int line_size, int h)

  159. {

  160.     int s = 0, i;

  161.     uint32_t *sq = ff_square_tab + 256;

  162. 

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

  164.         s += sq[pix1[0]  - pix2[0]];

  165.         s += sq[pix1[1]  - pix2[1]];

  166.         s += sq[pix1[2]  - pix2[2]];

  167.         s += sq[pix1[3]  - pix2[3]];

  168.         s += sq[pix1[4]  - pix2[4]];

  169.         s += sq[pix1[5]  - pix2[5]];

  170.         s += sq[pix1[6]  - pix2[6]];

  171.         s += sq[pix1[7]  - pix2[7]];

  172.         s += sq[pix1[8]  - pix2[8]];

  173.         s += sq[pix1[9]  - pix2[9]];

  174.         s += sq[pix1[10] - pix2[10]];

  175.         s += sq[pix1[11] - pix2[11]];

  176.         s += sq[pix1[12] - pix2[12]];

  177.         s += sq[pix1[13] - pix2[13]];

  178.         s += sq[pix1[14] - pix2[14]];

  179.         s += sq[pix1[15] - pix2[15]];

  180. 

  181.         pix1 += line_size;

  182.         pix2 += line_size;

  183.     }

  184.     return s;

  185. }

  186. 

  187. static void diff_pixels_c(int16_t *av_restrict block, const uint8_t *s1,

  188.                           const uint8_t *s2, int stride)

  189. {

  190.     int i;

  191. 

  192.     /* read the pixels */

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

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

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

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

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

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

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

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

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

  202.         s1      += stride;

  203.         s2      += stride;

  204.         block   += 8;

  205.     }

  206. }

  207. 

  208. static int sum_abs_dctelem_c(int16_t *block)

  209. {

  210.     int sum = 0, i;

  211. 

  212.     for (i = 0; i < 64; i++)

  213.         sum += FFABS(block[i]);

  214.     return sum;

  215. }

  216. 

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

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

  219. 

  220. static inline int pix_abs16_c(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,

  221.                               int line_size, int h)

  222. {

  223.     int s = 0, i;

  224. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  242.         pix1 += line_size;

  243.         pix2 += line_size;

  244.     }

  245.     return s;

  246. }

  247. 

  248. static int pix_abs16_x2_c(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,

  249.                           int line_size, int h)

  250. {

  251.     int s = 0, i;

  252. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  270.         pix1 += line_size;

  271.         pix2 += line_size;

  272.     }

  273.     return s;

  274. }

  275. 

  276. static int pix_abs16_y2_c(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,

  277.                           int line_size, int h)

  278. {

  279.     int s = 0, i;

  280.     uint8_t *pix3 = pix2 + line_size;

  281. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  299.         pix1 += line_size;

  300.         pix2 += line_size;

  301.         pix3 += line_size;

  302.     }

  303.     return s;

  304. }

  305. 

  306. static int pix_abs16_xy2_c(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,

  307.                            int line_size, int h)

  308. {

  309.     int s = 0, i;

  310.     uint8_t *pix3 = pix2 + line_size;

  311. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  329.         pix1 += line_size;

  330.         pix2 += line_size;

  331.         pix3 += line_size;

  332.     }

  333.     return s;

  334. }

  335. 

  336. static inline int pix_abs8_c(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,

  337.                              int line_size, int h)

  338. {

  339.     int s = 0, i;

  340. 

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

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

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

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

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

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

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

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

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

  350.         pix1 += line_size;

  351.         pix2 += line_size;

  352.     }

  353.     return s;

  354. }

  355. 

  356. static int pix_abs8_x2_c(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,

  357.                          int line_size, int h)

  358. {

  359.     int s = 0, i;

  360. 

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

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

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

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

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

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

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

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

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

  370.         pix1 += line_size;

  371.         pix2 += line_size;

  372.     }

  373.     return s;

  374. }

  375. 

  376. static int pix_abs8_y2_c(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,

  377.                          int line_size, int h)

  378. {

  379.     int s = 0, i;

  380.     uint8_t *pix3 = pix2 + line_size;

  381. 

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

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

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

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

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

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

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

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

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

  391.         pix1 += line_size;

  392.         pix2 += line_size;

  393.         pix3 += line_size;

  394.     }

  395.     return s;

  396. }

  397. 

  398. static int pix_abs8_xy2_c(MpegEncContext *v, uint8_t *pix1, uint8_t *pix2,

  399.                           int line_size, int h)

  400. {

  401.     int s = 0, i;

  402.     uint8_t *pix3 = pix2 + line_size;

  403. 

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

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

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

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

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

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

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

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

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

  413.         pix1 += line_size;

  414.         pix2 += line_size;

  415.         pix3 += line_size;

  416.     }

  417.     return s;

  418. }

  419. 

  420. static int nsse16_c(MpegEncContext *c, uint8_t *s1, uint8_t *s2, int stride, int h)

  421. {

  422.     int score1 = 0, score2 = 0, x, y;

  423. 

  424.     for (y = 0; y < h; y++) {

  425.         for (x = 0; x < 16; x++)

  426.             score1 += (s1[x] - s2[x]) * (s1[x] - s2[x]);

  427.         if (y + 1 < h) {

  428.             for (x = 0; x < 15; x++)

  429.                 score2 += FFABS(s1[x]     - s1[x + stride] -

  430.                                 s1[x + 1] + s1[x + stride + 1]) -

  431.                           FFABS(s2[x]     - s2[x + stride] -

  432.                                 s2[x + 1] + s2[x + stride + 1]);

  433.         }

  434.         s1 += stride;

  435.         s2 += stride;

  436.     }

  437. 

  438.     if (c)

  439.         return score1 + FFABS(score2) * c->avctx->nsse_weight;

  440.     else

  441.         return score1 + FFABS(score2) * 8;

  442. }

  443. 

  444. static int nsse8_c(MpegEncContext *c, uint8_t *s1, uint8_t *s2, int stride, int h)

  445. {

  446.     int score1 = 0, score2 = 0, x, y;

  447. 

  448.     for (y = 0; y < h; y++) {

  449.         for (x = 0; x < 8; x++)

  450.             score1 += (s1[x] - s2[x]) * (s1[x] - s2[x]);

  451.         if (y + 1 < h) {

  452.             for (x = 0; x < 7; x++)

  453.                 score2 += FFABS(s1[x]     - s1[x + stride] -

  454.                                 s1[x + 1] + s1[x + stride + 1]) -

  455.                           FFABS(s2[x]     - s2[x + stride] -

  456.                                 s2[x + 1] + s2[x + stride + 1]);

  457.         }

  458.         s1 += stride;

  459.         s2 += stride;

  460.     }

  461. 

  462.     if (c)

  463.         return score1 + FFABS(score2) * c->avctx->nsse_weight;

  464.     else

  465.         return score1 + FFABS(score2) * 8;

  466. }

  467. 

  468. static int zero_cmp(MpegEncContext *s, uint8_t *a, uint8_t *b,

  469.                     int stride, int h)

  470. {

  471.     return 0;

  472. }

  473. 

  474. void ff_set_cmp(DSPContext *c, me_cmp_func *cmp, int type)

  475. {

  476.     int i;

  477. 

  478.     memset(cmp, 0, sizeof(void *) * 6);

  479. 

  480.     for (i = 0; i < 6; i++) {

  481.         switch (type & 0xFF) {

  482.         case FF_CMP_SAD:

  483.             cmp[i] = c->sad[i];

  484.             break;

  485.         case FF_CMP_SATD:

  486.             cmp[i] = c->hadamard8_diff[i];

  487.             break;

  488.         case FF_CMP_SSE:

  489.             cmp[i] = c->sse[i];

  490.             break;

  491.         case FF_CMP_DCT:

  492.             cmp[i] = c->dct_sad[i];

  493.             break;

  494.         case FF_CMP_DCT264:

  495.             cmp[i] = c->dct264_sad[i];

  496.             break;

  497.         case FF_CMP_DCTMAX:

  498.             cmp[i] = c->dct_max[i];

  499.             break;

  500.         case FF_CMP_PSNR:

  501.             cmp[i] = c->quant_psnr[i];

  502.             break;

  503.         case FF_CMP_BIT:

  504.             cmp[i] = c->bit[i];

  505.             break;

  506.         case FF_CMP_RD:

  507.             cmp[i] = c->rd[i];

  508.             break;

  509.         case FF_CMP_VSAD:

  510.             cmp[i] = c->vsad[i];

  511.             break;

  512.         case FF_CMP_VSSE:

  513.             cmp[i] = c->vsse[i];

  514.             break;

  515.         case FF_CMP_ZERO:

  516.             cmp[i] = zero_cmp;

  517.             break;

  518.         case FF_CMP_NSSE:

  519.             cmp[i] = c->nsse[i];

  520.             break;

  521. #if CONFIG_DWT

  522.         case FF_CMP_W53:

  523.             cmp[i]= c->w53[i];

  524.             break;

  525.         case FF_CMP_W97:

  526.             cmp[i]= c->w97[i];

  527.             break;

  528. #endif

  529.         default:

  530.             av_log(NULL, AV_LOG_ERROR,

  531.                    "internal error in cmp function selection\n");

  532.         }

  533.     }

  534. }

  535. 

  536. #define BUTTERFLY2(o1, o2, i1, i2)              \

  537.     o1 = (i1) + (i2);                           \

  538.     o2 = (i1) - (i2);

  539. 

  540. #define BUTTERFLY1(x, y)                        \

  541.     {                                           \

  542.         int a, b;                               \

  543.         a = x;                                  \

  544.         b = y;                                  \

  545.         x = a + b;                              \

  546.         y = a - b;                              \

  547.     }

  548. 

  549. #define BUTTERFLYA(x, y) (FFABS((x) + (y)) + FFABS((x) - (y)))

  550. 

  551. static int hadamard8_diff8x8_c(MpegEncContext *s, uint8_t *dst,

  552.                                uint8_t *src, int stride, int h)

  553. {

  554.     int i, temp[64], sum = 0;

  555. 

  556.     av_assert2(h == 8);

  557. 

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

  559.         // FIXME: try pointer walks

  560.         BUTTERFLY2(temp[8 * i + 0], temp[8 * i + 1],

  561.                    src[stride * i + 0] - dst[stride * i + 0],

  562.                    src[stride * i + 1] - dst[stride * i + 1]);

  563.         BUTTERFLY2(temp[8 * i + 2], temp[8 * i + 3],

  564.                    src[stride * i + 2] - dst[stride * i + 2],

  565.                    src[stride * i + 3] - dst[stride * i + 3]);

  566.         BUTTERFLY2(temp[8 * i + 4], temp[8 * i + 5],

  567.                    src[stride * i + 4] - dst[stride * i + 4],

  568.                    src[stride * i + 5] - dst[stride * i + 5]);

  569.         BUTTERFLY2(temp[8 * i + 6], temp[8 * i + 7],

  570.                    src[stride * i + 6] - dst[stride * i + 6],

  571.                    src[stride * i + 7] - dst[stride * i + 7]);

  572. 

  573.         BUTTERFLY1(temp[8 * i + 0], temp[8 * i + 2]);

  574.         BUTTERFLY1(temp[8 * i + 1], temp[8 * i + 3]);

  575.         BUTTERFLY1(temp[8 * i + 4], temp[8 * i + 6]);

  576.         BUTTERFLY1(temp[8 * i + 5], temp[8 * i + 7]);

  577. 

  578.         BUTTERFLY1(temp[8 * i + 0], temp[8 * i + 4]);

  579.         BUTTERFLY1(temp[8 * i + 1], temp[8 * i + 5]);

  580.         BUTTERFLY1(temp[8 * i + 2], temp[8 * i + 6]);

  581.         BUTTERFLY1(temp[8 * i + 3], temp[8 * i + 7]);

  582.     }

  583. 

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

  585.         BUTTERFLY1(temp[8 * 0 + i], temp[8 * 1 + i]);

  586.         BUTTERFLY1(temp[8 * 2 + i], temp[8 * 3 + i]);

  587.         BUTTERFLY1(temp[8 * 4 + i], temp[8 * 5 + i]);

  588.         BUTTERFLY1(temp[8 * 6 + i], temp[8 * 7 + i]);

  589. 

  590.         BUTTERFLY1(temp[8 * 0 + i], temp[8 * 2 + i]);

  591.         BUTTERFLY1(temp[8 * 1 + i], temp[8 * 3 + i]);

  592.         BUTTERFLY1(temp[8 * 4 + i], temp[8 * 6 + i]);

  593.         BUTTERFLY1(temp[8 * 5 + i], temp[8 * 7 + i]);

  594. 

  595.         sum += BUTTERFLYA(temp[8 * 0 + i], temp[8 * 4 + i]) +

  596.                BUTTERFLYA(temp[8 * 1 + i], temp[8 * 5 + i]) +

  597.                BUTTERFLYA(temp[8 * 2 + i], temp[8 * 6 + i]) +

  598.                BUTTERFLYA(temp[8 * 3 + i], temp[8 * 7 + i]);

  599.     }

  600.     return sum;

  601. }

  602. 

  603. static int hadamard8_intra8x8_c(MpegEncContext *s, uint8_t *src,

  604.                                 uint8_t *dummy, int stride, int h)

  605. {

  606.     int i, temp[64], sum = 0;

  607. 

  608.     av_assert2(h == 8);

  609. 

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

  611.         // FIXME: try pointer walks

  612.         BUTTERFLY2(temp[8 * i + 0], temp[8 * i + 1],

  613.                    src[stride * i + 0], src[stride * i + 1]);

  614.         BUTTERFLY2(temp[8 * i + 2], temp[8 * i + 3],

  615.                    src[stride * i + 2], src[stride * i + 3]);

  616.         BUTTERFLY2(temp[8 * i + 4], temp[8 * i + 5],

  617.                    src[stride * i + 4], src[stride * i + 5]);

  618.         BUTTERFLY2(temp[8 * i + 6], temp[8 * i + 7],

  619.                    src[stride * i + 6], src[stride * i + 7]);

  620. 

  621.         BUTTERFLY1(temp[8 * i + 0], temp[8 * i + 2]);

  622.         BUTTERFLY1(temp[8 * i + 1], temp[8 * i + 3]);

  623.         BUTTERFLY1(temp[8 * i + 4], temp[8 * i + 6]);

  624.         BUTTERFLY1(temp[8 * i + 5], temp[8 * i + 7]);

  625. 

  626.         BUTTERFLY1(temp[8 * i + 0], temp[8 * i + 4]);

  627.         BUTTERFLY1(temp[8 * i + 1], temp[8 * i + 5]);

  628.         BUTTERFLY1(temp[8 * i + 2], temp[8 * i + 6]);

  629.         BUTTERFLY1(temp[8 * i + 3], temp[8 * i + 7]);

  630.     }

  631. 

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

  633.         BUTTERFLY1(temp[8 * 0 + i], temp[8 * 1 + i]);

  634.         BUTTERFLY1(temp[8 * 2 + i], temp[8 * 3 + i]);

  635.         BUTTERFLY1(temp[8 * 4 + i], temp[8 * 5 + i]);

  636.         BUTTERFLY1(temp[8 * 6 + i], temp[8 * 7 + i]);

  637. 

  638.         BUTTERFLY1(temp[8 * 0 + i], temp[8 * 2 + i]);

  639.         BUTTERFLY1(temp[8 * 1 + i], temp[8 * 3 + i]);

  640.         BUTTERFLY1(temp[8 * 4 + i], temp[8 * 6 + i]);

  641.         BUTTERFLY1(temp[8 * 5 + i], temp[8 * 7 + i]);

  642. 

  643.         sum +=

  644.             BUTTERFLYA(temp[8 * 0 + i], temp[8 * 4 + i])

  645.             + BUTTERFLYA(temp[8 * 1 + i], temp[8 * 5 + i])

  646.             + BUTTERFLYA(temp[8 * 2 + i], temp[8 * 6 + i])

  647.             + BUTTERFLYA(temp[8 * 3 + i], temp[8 * 7 + i]);

  648.     }

  649. 

  650.     sum -= FFABS(temp[8 * 0] + temp[8 * 4]); // -mean

  651. 

  652.     return sum;

  653. }

  654. 

  655. static int dct_sad8x8_c(MpegEncContext *s, uint8_t *src1,

  656.                         uint8_t *src2, int stride, int h)

  657. {

  658.     LOCAL_ALIGNED_16(int16_t, temp, [64]);

  659. 

  660.     av_assert2(h == 8);

  661. 

  662.     s->dsp.diff_pixels(temp, src1, src2, stride);

  663.     s->dsp.fdct(temp);

  664.     return s->dsp.sum_abs_dctelem(temp);

  665. }

  666. 

  667. #if CONFIG_GPL

  668. #define DCT8_1D                                         \

  669.     {                                                   \

  670.         const int s07 = SRC(0) + SRC(7);                \

  671.         const int s16 = SRC(1) + SRC(6);                \

  672.         const int s25 = SRC(2) + SRC(5);                \

  673.         const int s34 = SRC(3) + SRC(4);                \

  674.         const int a0  = s07 + s34;                      \

  675.         const int a1  = s16 + s25;                      \

  676.         const int a2  = s07 - s34;                      \

  677.         const int a3  = s16 - s25;                      \

  678.         const int d07 = SRC(0) - SRC(7);                \

  679.         const int d16 = SRC(1) - SRC(6);                \

  680.         const int d25 = SRC(2) - SRC(5);                \

  681.         const int d34 = SRC(3) - SRC(4);                \

  682.         const int a4  = d16 + d25 + (d07 + (d07 >> 1)); \

  683.         const int a5  = d07 - d34 - (d25 + (d25 >> 1)); \

  684.         const int a6  = d07 + d34 - (d16 + (d16 >> 1)); \

  685.         const int a7  = d16 - d25 + (d34 + (d34 >> 1)); \

  686.         DST(0, a0 + a1);                                \

  687.         DST(1, a4 + (a7 >> 2));                         \

  688.         DST(2, a2 + (a3 >> 1));                         \

  689.         DST(3, a5 + (a6 >> 2));                         \

  690.         DST(4, a0 - a1);                                \

  691.         DST(5, a6 - (a5 >> 2));                         \

  692.         DST(6, (a2 >> 1) - a3);                         \

  693.         DST(7, (a4 >> 2) - a7);                         \

  694.     }

  695. 

  696. static int dct264_sad8x8_c(MpegEncContext *s, uint8_t *src1,

  697.                            uint8_t *src2, int stride, int h)

  698. {

  699.     int16_t dct[8][8];

  700.     int i, sum = 0;

  701. 

  702.     s->dsp.diff_pixels(dct[0], src1, src2, stride);

  703. 

  704. #define SRC(x) dct[i][x]

  705. #define DST(x, v) dct[i][x] = v

  706.     for (i = 0; i < 8; i++)

  707.         DCT8_1D

  708. #undef SRC

  709. #undef DST

  710. 

  711. #define SRC(x) dct[x][i]

  712. #define DST(x, v) sum += FFABS(v)

  713.         for (i = 0; i < 8; i++)

  714.             DCT8_1D

  715. #undef SRC

  716. #undef DST

  717.             return sum;

  718. }

  719. #endif

  720. 

  721. static int dct_max8x8_c(MpegEncContext *s, uint8_t *src1,

  722.                         uint8_t *src2, int stride, int h)

  723. {

  724.     LOCAL_ALIGNED_16(int16_t, temp, [64]);

  725.     int sum = 0, i;

  726. 

  727.     av_assert2(h == 8);

  728. 

  729.     s->dsp.diff_pixels(temp, src1, src2, stride);

  730.     s->dsp.fdct(temp);

  731. 

  732.     for (i = 0; i < 64; i++)

  733.         sum = FFMAX(sum, FFABS(temp[i]));

  734. 

  735.     return sum;

  736. }

  737. 

  738. static int quant_psnr8x8_c(MpegEncContext *s, uint8_t *src1,

  739.                            uint8_t *src2, int stride, int h)

  740. {

  741.     LOCAL_ALIGNED_16(int16_t, temp, [64 * 2]);

  742.     int16_t *const bak = temp + 64;

  743.     int sum = 0, i;

  744. 

  745.     av_assert2(h == 8);

  746.     s->mb_intra = 0;

  747. 

  748.     s->dsp.diff_pixels(temp, src1, src2, stride);

  749. 

  750.     memcpy(bak, temp, 64 * sizeof(int16_t));

  751. 

  752.     s->block_last_index[0 /* FIXME */] =

  753.         s->fast_dct_quantize(s, temp, 0 /* FIXME */, s->qscale, &i);

  754.     s->dct_unquantize_inter(s, temp, 0, s->qscale);

  755.     ff_simple_idct_8(temp); // FIXME

  756. 

  757.     for (i = 0; i < 64; i++)

  758.         sum += (temp[i] - bak[i]) * (temp[i] - bak[i]);

  759. 

  760.     return sum;

  761. }

  762. 

  763. static int rd8x8_c(MpegEncContext *s, uint8_t *src1, uint8_t *src2,

  764.                    int stride, int h)

  765. {

  766.     const uint8_t *scantable = s->intra_scantable.permutated;

  767.     LOCAL_ALIGNED_16(int16_t, temp, [64]);

  768.     LOCAL_ALIGNED_16(uint8_t, lsrc1, [64]);

  769.     LOCAL_ALIGNED_16(uint8_t, lsrc2, [64]);

  770.     int i, last, run, bits, level, distortion, start_i;

  771.     const int esc_length = s->ac_esc_length;

  772.     uint8_t *length, *last_length;

  773. 

  774.     av_assert2(h == 8);

  775. 

  776.     copy_block8(lsrc1, src1, 8, stride, 8);

  777.     copy_block8(lsrc2, src2, 8, stride, 8);

  778. 

  779.     s->dsp.diff_pixels(temp, lsrc1, lsrc2, 8);

  780. 

  781.     s->block_last_index[0 /* FIXME */] =

  782.     last                               =

  783.         s->fast_dct_quantize(s, temp, 0 /* FIXME */, s->qscale, &i);

  784. 

  785.     bits = 0;

  786. 

  787.     if (s->mb_intra) {

  788.         start_i     = 1;

  789.         length      = s->intra_ac_vlc_length;

  790.         last_length = s->intra_ac_vlc_last_length;

  791.         bits       += s->luma_dc_vlc_length[temp[0] + 256]; // FIXME: chroma

  792.     } else {

  793.         start_i     = 0;

  794.         length      = s->inter_ac_vlc_length;

  795.         last_length = s->inter_ac_vlc_last_length;

  796.     }

  797. 

  798.     if (last >= start_i) {

  799.         run = 0;

  800.         for (i = start_i; i < last; i++) {

  801.             int j = scantable[i];

  802.             level = temp[j];

  803. 

  804.             if (level) {

  805.                 level += 64;

  806.                 if ((level & (~127)) == 0)

  807.                     bits += length[UNI_AC_ENC_INDEX(run, level)];

  808.                 else

  809.                     bits += esc_length;

  810.                 run = 0;

  811.             } else

  812.                 run++;

  813.         }

  814.         i = scantable[last];

  815. 

  816.         level = temp[i] + 64;

  817. 

  818.         av_assert2(level - 64);

  819. 

  820.         if ((level & (~127)) == 0) {

  821.             bits += last_length[UNI_AC_ENC_INDEX(run, level)];

  822.         } else

  823.             bits += esc_length;

  824.     }

  825. 

  826.     if (last >= 0) {

  827.         if (s->mb_intra)

  828.             s->dct_unquantize_intra(s, temp, 0, s->qscale);

  829.         else

  830.             s->dct_unquantize_inter(s, temp, 0, s->qscale);

  831.     }

  832. 

  833.     s->idsp.idct_add(lsrc2, 8, temp);

  834. 

  835.     distortion = s->dsp.sse[1](NULL, lsrc2, lsrc1, 8, 8);

  836. 

  837.     return distortion + ((bits * s->qscale * s->qscale * 109 + 64) >> 7);

  838. }

  839. 

  840. static int bit8x8_c(MpegEncContext *s, uint8_t *src1, uint8_t *src2,

  841.                     int stride, int h)

  842. {

  843.     const uint8_t *scantable = s->intra_scantable.permutated;

  844.     LOCAL_ALIGNED_16(int16_t, temp, [64]);

  845.     int i, last, run, bits, level, start_i;

  846.     const int esc_length = s->ac_esc_length;

  847.     uint8_t *length, *last_length;

  848. 

  849.     av_assert2(h == 8);

  850. 

  851.     s->dsp.diff_pixels(temp, src1, src2, stride);

  852. 

  853.     s->block_last_index[0 /* FIXME */] =

  854.     last                               =

  855.         s->fast_dct_quantize(s, temp, 0 /* FIXME */, s->qscale, &i);

  856. 

  857.     bits = 0;

  858. 

  859.     if (s->mb_intra) {

  860.         start_i     = 1;

  861.         length      = s->intra_ac_vlc_length;

  862.         last_length = s->intra_ac_vlc_last_length;

  863.         bits       += s->luma_dc_vlc_length[temp[0] + 256]; // FIXME: chroma

  864.     } else {

  865.         start_i     = 0;

  866.         length      = s->inter_ac_vlc_length;

  867.         last_length = s->inter_ac_vlc_last_length;

  868.     }

  869. 

  870.     if (last >= start_i) {

  871.         run = 0;

  872.         for (i = start_i; i < last; i++) {

  873.             int j = scantable[i];

  874.             level = temp[j];

  875. 

  876.             if (level) {

  877.                 level += 64;

  878.                 if ((level & (~127)) == 0)

  879.                     bits += length[UNI_AC_ENC_INDEX(run, level)];

  880.                 else

  881.                     bits += esc_length;

  882.                 run = 0;

  883.             } else

  884.                 run++;

  885.         }

  886.         i = scantable[last];

  887. 

  888.         level = temp[i] + 64;

  889. 

  890.         av_assert2(level - 64);

  891. 

  892.         if ((level & (~127)) == 0)

  893.             bits += last_length[UNI_AC_ENC_INDEX(run, level)];

  894.         else

  895.             bits += esc_length;

  896.     }

  897. 

  898.     return bits;

  899. }

  900. 

  901. #define VSAD_INTRA(size)                                                \

  902. static int vsad_intra ## size ## _c(MpegEncContext *c,                  \

  903.                                     uint8_t *s, uint8_t *dummy,         \

  904.                                     int stride, int h)                  \

  905. {                                                                       \

  906.     int score = 0, x, y;                                                \

  907.                                                                         \

  908.     for (y = 1; y < h; y++) {                                           \

  909.         for (x = 0; x < size; x += 4) {                                 \

  910.             score += FFABS(s[x]     - s[x + stride])     +              \

  911.                      FFABS(s[x + 1] - s[x + stride + 1]) +              \

  912.                      FFABS(s[x + 2] - s[x + 2 + stride]) +              \

  913.                      FFABS(s[x + 3] - s[x + 3 + stride]);               \

  914.         }                                                               \

  915.         s += stride;                                                    \

  916.     }                                                                   \

  917.                                                                         \

  918.     return score;                                                       \

  919. }

  920. VSAD_INTRA(8)

  921. VSAD_INTRA(16)

  922. 

  923. #define VSAD(size)                                                             \

  924. static int vsad ## size ## _c(MpegEncContext *c,                               \

  925.                               uint8_t *s1, uint8_t *s2,                        \

  926.                               int stride, int h)                               \

  927. {                                                                              \

  928.     int score = 0, x, y;                                                       \

  929.                                                                                \

  930.     for (y = 1; y < h; y++) {                                                  \

  931.         for (x = 0; x < size; x++)                                             \

  932.             score += FFABS(s1[x] - s2[x] - s1[x + stride] + s2[x + stride]);   \

  933.         s1 += stride;                                                          \

  934.         s2 += stride;                                                          \

  935.     }                                                                          \

  936.                                                                                \

  937.     return score;                                                              \

  938. }

  939. VSAD(8)

  940. VSAD(16)

  941. 

  942. #define SQ(a) ((a) * (a))

  943. #define VSSE_INTRA(size)                                                \

  944. static int vsse_intra ## size ## _c(MpegEncContext *c,                  \

  945.                                     uint8_t *s, uint8_t *dummy,         \

  946.                                     int stride, int h)                  \

  947. {                                                                       \

  948.     int score = 0, x, y;                                                \

  949.                                                                         \

  950.     for (y = 1; y < h; y++) {                                           \

  951.         for (x = 0; x < size; x += 4) {                                 \

  952.             score += SQ(s[x]     - s[x + stride]) +                     \

  953.                      SQ(s[x + 1] - s[x + stride + 1]) +                 \

  954.                      SQ(s[x + 2] - s[x + stride + 2]) +                 \

  955.                      SQ(s[x + 3] - s[x + stride + 3]);                  \

  956.         }                                                               \

  957.         s += stride;                                                    \

  958.     }                                                                   \

  959.                                                                         \

  960.     return score;                                                       \

  961. }

  962. VSSE_INTRA(8)

  963. VSSE_INTRA(16)

  964. 

  965. #define VSSE(size)                                                             \

  966. static int vsse ## size ## _c(MpegEncContext *c, uint8_t *s1, uint8_t *s2,     \

  967.                     int stride, int h)                                         \

  968. {                                                                              \

  969.     int score = 0, x, y;                                                       \

  970.                                                                                \

  971.     for (y = 1; y < h; y++) {                                                  \

  972.         for (x = 0; x < size; x++)                                             \

  973.             score += SQ(s1[x] - s2[x] - s1[x + stride] + s2[x + stride]);      \

  974.         s1 += stride;                                                          \

  975.         s2 += stride;                                                          \

  976.     }                                                                          \

  977.                                                                                \

  978.     return score;                                                              \

  979. }

  980. VSSE(8)

  981. VSSE(16)

  982. 

  983. #define WRAPPER8_16_SQ(name8, name16)                                   \

  984. static int name16(MpegEncContext *s, uint8_t *dst, uint8_t *src,        \

  985.                   int stride, int h)                                    \

  986. {                                                                       \

  987.     int score = 0;                                                      \

  988.                                                                         \

  989.     score += name8(s, dst, src, stride, 8);                             \

  990.     score += name8(s, dst + 8, src + 8, stride, 8);                     \

  991.     if (h == 16) {                                                      \

  992.         dst   += 8 * stride;                                            \

  993.         src   += 8 * stride;                                            \

  994.         score += name8(s, dst, src, stride, 8);                         \

  995.         score += name8(s, dst + 8, src + 8, stride, 8);                 \

  996.     }                                                                   \

  997.     return score;                                                       \

  998. }

  999. 

  1000. WRAPPER8_16_SQ(hadamard8_diff8x8_c, hadamard8_diff16_c)

  1001. WRAPPER8_16_SQ(hadamard8_intra8x8_c, hadamard8_intra16_c)

  1002. WRAPPER8_16_SQ(dct_sad8x8_c, dct_sad16_c)

  1003. #if CONFIG_GPL

  1004. WRAPPER8_16_SQ(dct264_sad8x8_c, dct264_sad16_c)

  1005. #endif

  1006. WRAPPER8_16_SQ(dct_max8x8_c, dct_max16_c)

  1007. WRAPPER8_16_SQ(quant_psnr8x8_c, quant_psnr16_c)

  1008. WRAPPER8_16_SQ(rd8x8_c, rd16_c)

  1009. WRAPPER8_16_SQ(bit8x8_c, bit16_c)

  1010. 

  1011. /* draw the edges of width 'w' of an image of size width, height */

  1012. // FIXME: Check that this is OK for MPEG-4 interlaced.

  1013. static void draw_edges_8_c(uint8_t *buf, int wrap, int width, int height,

  1014.                            int w, int h, int sides)

  1015. {

  1016.     uint8_t *ptr = buf, *last_line;

  1017.     int i;

  1018. 

  1019.     /* left and right */

  1020.     for (i = 0; i < height; i++) {

  1021.         memset(ptr - w, ptr[0], w);

  1022.         memset(ptr + width, ptr[width - 1], w);

  1023.         ptr += wrap;

  1024.     }

  1025. 

  1026.     /* top and bottom + corners */

  1027.     buf -= w;

  1028.     last_line = buf + (height - 1) * wrap;

  1029.     if (sides & EDGE_TOP)

  1030.         for (i = 0; i < h; i++)

  1031.             // top

  1032.             memcpy(buf - (i + 1) * wrap, buf, width + w + w);

  1033.     if (sides & EDGE_BOTTOM)

  1034.         for (i = 0; i < h; i++)

  1035.             // bottom

  1036.             memcpy(last_line + (i + 1) * wrap, last_line, width + w + w);

  1037. }

  1038. 

  1039. /* init static data */

  1040. av_cold void ff_dsputil_static_init(void)

  1041. {

  1042.     int i;

  1043. 

  1044.     for (i = 0; i < 512; i++)

  1045.         ff_square_tab[i] = (i - 256) * (i - 256);

  1046. }

  1047. 

  1048. int ff_check_alignment(void)

  1049. {

  1050.     static int did_fail = 0;

  1051.     LOCAL_ALIGNED_16(int, aligned, [4]);

  1052. 

  1053.     if ((intptr_t)aligned & 15) {

  1054.         if (!did_fail) {

  1055. #if HAVE_MMX || HAVE_ALTIVEC

  1056.             av_log(NULL, AV_LOG_ERROR,

  1057.                 "Compiler did not align stack variables. Libavcodec has been miscompiled\n"

  1058.                 "and may be very slow or crash. This is not a bug in libavcodec,\n"

  1059.                 "but in the compiler. You may try recompiling using gcc >= 4.2.\n"

  1060.                 "Do not report crashes to FFmpeg developers.\n");

  1061. #endif

  1062.             did_fail=1;

  1063.         }

  1064.         return -1;

  1065.     }

  1066.     return 0;

  1067. }

  1068. 

  1069. av_cold void ff_dsputil_init(DSPContext *c, AVCodecContext *avctx)

  1070. {

  1071.     const unsigned high_bit_depth = avctx->bits_per_raw_sample > 8;

  1072. 

  1073.     ff_check_alignment();

  1074. 

  1075. #if CONFIG_ENCODERS

  1076.     if (avctx->bits_per_raw_sample == 10) {

  1077.         c->fdct    = ff_jpeg_fdct_islow_10;

  1078.         c->fdct248 = ff_fdct248_islow_10;

  1079.     } else {

  1080.         if (avctx->dct_algo == FF_DCT_FASTINT) {

  1081.             c->fdct    = ff_fdct_ifast;

  1082.             c->fdct248 = ff_fdct_ifast248;

  1083.         } else if (avctx->dct_algo == FF_DCT_FAAN) {

  1084.             c->fdct    = ff_faandct;

  1085.             c->fdct248 = ff_faandct248;

  1086.         } else {

  1087.             c->fdct    = ff_jpeg_fdct_islow_8; // slow/accurate/default

  1088.             c->fdct248 = ff_fdct248_islow_8;

  1089.         }

  1090.     }

  1091. #endif /* CONFIG_ENCODERS */

  1092. 

  1093.     c->diff_pixels = diff_pixels_c;

  1094. 

  1095.     c->sum_abs_dctelem = sum_abs_dctelem_c;

  1096. 

  1097.     c->pix_sum   = pix_sum_c;

  1098.     c->pix_norm1 = pix_norm1_c;

  1099. 

  1100.     /* TODO [0] 16  [1] 8 */

  1101.     c->pix_abs[0][0] = pix_abs16_c;

  1102.     c->pix_abs[0][1] = pix_abs16_x2_c;

  1103.     c->pix_abs[0][2] = pix_abs16_y2_c;

  1104.     c->pix_abs[0][3] = pix_abs16_xy2_c;

  1105.     c->pix_abs[1][0] = pix_abs8_c;

  1106.     c->pix_abs[1][1] = pix_abs8_x2_c;

  1107.     c->pix_abs[1][2] = pix_abs8_y2_c;

  1108.     c->pix_abs[1][3] = pix_abs8_xy2_c;

  1109. 

  1110. #define SET_CMP_FUNC(name)                      \

  1111.     c->name[0] = name ## 16_c;                  \

  1112.     c->name[1] = name ## 8x8_c;

  1113. 

  1114.     SET_CMP_FUNC(hadamard8_diff)

  1115.     c->hadamard8_diff[4] = hadamard8_intra16_c;

  1116.     c->hadamard8_diff[5] = hadamard8_intra8x8_c;

  1117.     SET_CMP_FUNC(dct_sad)

  1118.     SET_CMP_FUNC(dct_max)

  1119. #if CONFIG_GPL

  1120.     SET_CMP_FUNC(dct264_sad)

  1121. #endif

  1122.     c->sad[0] = pix_abs16_c;

  1123.     c->sad[1] = pix_abs8_c;

  1124.     c->sse[0] = sse16_c;

  1125.     c->sse[1] = sse8_c;

  1126.     c->sse[2] = sse4_c;

  1127.     SET_CMP_FUNC(quant_psnr)

  1128.     SET_CMP_FUNC(rd)

  1129.     SET_CMP_FUNC(bit)

  1130.     c->vsad[0] = vsad16_c;

  1131.     c->vsad[1] = vsad8_c;

  1132.     c->vsad[4] = vsad_intra16_c;

  1133.     c->vsad[5] = vsad_intra8_c;

  1134.     c->vsse[0] = vsse16_c;

  1135.     c->vsse[1] = vsse8_c;

  1136.     c->vsse[4] = vsse_intra16_c;

  1137.     c->vsse[5] = vsse_intra8_c;

  1138.     c->nsse[0] = nsse16_c;

  1139.     c->nsse[1] = nsse8_c;

  1140. #if CONFIG_SNOW_DECODER || CONFIG_SNOW_ENCODER

  1141.     ff_dsputil_init_dwt(c);

  1142. #endif

  1143. 

  1144.     c->shrink[0] = av_image_copy_plane;

  1145.     c->shrink[1] = ff_shrink22;

  1146.     c->shrink[2] = ff_shrink44;

  1147.     c->shrink[3] = ff_shrink88;

  1148. 

  1149.     c->draw_edges = draw_edges_8_c;

  1150. 

  1151.     switch (avctx->bits_per_raw_sample) {

  1152.     case 9:

  1153.     case 10:

  1154.     case 12:

  1155.     case 14:

  1156.         c->get_pixels = get_pixels_16_c;

  1157.         break;

  1158.     default:

  1159.         if (avctx->bits_per_raw_sample<=8 || avctx->codec_type != AVMEDIA_TYPE_VIDEO) {

  1160.             c->get_pixels = get_pixels_8_c;

  1161.         }

  1162.         break;

  1163.     }

  1164. 

  1165. 

  1166.     if (ARCH_ALPHA)

  1167.         ff_dsputil_init_alpha(c, avctx);

  1168.     if (ARCH_ARM)

  1169.         ff_dsputil_init_arm(c, avctx, high_bit_depth);

  1170.     if (ARCH_PPC)

  1171.         ff_dsputil_init_ppc(c, avctx, high_bit_depth);

  1172.     if (ARCH_X86)

  1173.         ff_dsputil_init_x86(c, avctx, high_bit_depth);

  1174. }

  1175. 

  1176. av_cold void dsputil_init(DSPContext* c, AVCodecContext *avctx)

  1177. {

  1178.     ff_dsputil_init(c, avctx);

  1179. }

  1180. 

  1181. av_cold void avpriv_dsputil_init(DSPContext *c, AVCodecContext *avctx)

  1182. {

  1183.     ff_dsputil_init(c, avctx);

  1184. }