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.
64172 Commits
32 Branches
326MB
Tree: 35bb74900b
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '35bb74900b'
${ noResults }
FFmpeg/libavcodec/dsputil.c

1600 lines
53KB
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.  * gmc & q-pel & 32/64 bit based MC by Michael Niedermayer <michaelni@gmx.at>

  7.  *

  8.  * This file is part of FFmpeg.

  9.  *

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

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

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

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

  14.  *

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

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

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

  18.  * Lesser General Public License for more details.

  19.  *

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

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

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

  23.  */

  24. 

  25. /**

  26.  * @file

  27.  * DSP utils

  28.  */

  29. 

  30. #include "libavutil/attributes.h"

  31. #include "libavutil/imgutils.h"

  32. #include "libavutil/internal.h"

  33. #include "avcodec.h"

  34. #include "copy_block.h"

  35. #include "dct.h"

  36. #include "dsputil.h"

  37. #include "simple_idct.h"

  38. #include "faandct.h"

  39. #include "faanidct.h"

  40. #include "imgconvert.h"

  41. #include "mathops.h"

  42. #include "mpegvideo.h"

  43. #include "config.h"

  44. 

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

  46. 

  47. #define BIT_DEPTH 16

  48. #include "dsputilenc_template.c"

  49. #undef BIT_DEPTH

  50. 

  51. #define BIT_DEPTH 8

  52. #include "dsputilenc_template.c"

  53. 

  54. /* Input permutation for the simple_idct_mmx */

  55. static const uint8_t simple_mmx_permutation[64] = {

  56.     0x00, 0x08, 0x04, 0x09, 0x01, 0x0C, 0x05, 0x0D,

  57.     0x10, 0x18, 0x14, 0x19, 0x11, 0x1C, 0x15, 0x1D,

  58.     0x20, 0x28, 0x24, 0x29, 0x21, 0x2C, 0x25, 0x2D,

  59.     0x12, 0x1A, 0x16, 0x1B, 0x13, 0x1E, 0x17, 0x1F,

  60.     0x02, 0x0A, 0x06, 0x0B, 0x03, 0x0E, 0x07, 0x0F,

  61.     0x30, 0x38, 0x34, 0x39, 0x31, 0x3C, 0x35, 0x3D,

  62.     0x22, 0x2A, 0x26, 0x2B, 0x23, 0x2E, 0x27, 0x2F,

  63.     0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F,

  64. };

  65. 

  66. static const uint8_t idct_sse2_row_perm[8] = { 0, 4, 1, 5, 2, 6, 3, 7 };

  67. 

  68. av_cold void ff_init_scantable(uint8_t *permutation, ScanTable *st,

  69.                                const uint8_t *src_scantable)

  70. {

  71.     int i, end;

  72. 

  73.     st->scantable = src_scantable;

  74. 

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

  76.         int j = src_scantable[i];

  77.         st->permutated[i] = permutation[j];

  78.     }

  79. 

  80.     end = -1;

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

  82.         int j = st->permutated[i];

  83.         if (j > end)

  84.             end = j;

  85.         st->raster_end[i] = end;

  86.     }

  87. }

  88. 

  89. av_cold void ff_init_scantable_permutation(uint8_t *idct_permutation,

  90.                                            int idct_permutation_type)

  91. {

  92.     int i;

  93. 

  94.     switch (idct_permutation_type) {

  95.     case FF_NO_IDCT_PERM:

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

  97.             idct_permutation[i] = i;

  98.         break;

  99.     case FF_LIBMPEG2_IDCT_PERM:

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

  101.             idct_permutation[i] = (i & 0x38) | ((i & 6) >> 1) | ((i & 1) << 2);

  102.         break;

  103.     case FF_SIMPLE_IDCT_PERM:

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

  105.             idct_permutation[i] = simple_mmx_permutation[i];

  106.         break;

  107.     case FF_TRANSPOSE_IDCT_PERM:

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

  109.             idct_permutation[i] = ((i & 7) << 3) | (i >> 3);

  110.         break;

  111.     case FF_PARTTRANS_IDCT_PERM:

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

  113.             idct_permutation[i] = (i & 0x24) | ((i & 3) << 3) | ((i >> 3) & 3);

  114.         break;

  115.     case FF_SSE2_IDCT_PERM:

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

  117.             idct_permutation[i] = (i & 0x38) | idct_sse2_row_perm[i & 7];

  118.         break;

  119.     default:

  120.         av_log(NULL, AV_LOG_ERROR,

  121.                "Internal error, IDCT permutation not set\n");

  122.     }

  123. }

  124. 

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

  126. {

  127.     int s = 0, i, j;

  128. 

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

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

  131.             s   += pix[0];

  132.             s   += pix[1];

  133.             s   += pix[2];

  134.             s   += pix[3];

  135.             s   += pix[4];

  136.             s   += pix[5];

  137.             s   += pix[6];

  138.             s   += pix[7];

  139.             pix += 8;

  140.         }

  141.         pix += line_size - 16;

  142.     }

  143.     return s;

  144. }

  145. 

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

  147. {

  148.     int s = 0, i, j;

  149.     uint32_t *sq = ff_square_tab + 256;

  150. 

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

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

  153. #if 0

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

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

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

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

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

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

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

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

  162. #else

  163. #if HAVE_FAST_64BIT

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

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

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

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

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

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

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

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

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

  173. #else

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

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

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

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

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

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

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

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

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

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

  184. #endif

  185. #endif

  186.             pix += 8;

  187.         }

  188.         pix += line_size - 16;

  189.     }

  190.     return s;

  191. }

  192. 

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

  194.                   int line_size, int h)

  195. {

  196.     int s = 0, i;

  197.     uint32_t *sq = ff_square_tab + 256;

  198. 

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

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

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

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

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

  204.         pix1 += line_size;

  205.         pix2 += line_size;

  206.     }

  207.     return s;

  208. }

  209. 

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

  211.                   int line_size, int h)

  212. {

  213.     int s = 0, i;

  214.     uint32_t *sq = ff_square_tab + 256;

  215. 

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

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

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

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

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

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

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

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

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

  225.         pix1 += line_size;

  226.         pix2 += line_size;

  227.     }

  228.     return s;

  229. }

  230. 

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

  232.                    int line_size, int h)

  233. {

  234.     int s = 0, i;

  235.     uint32_t *sq = ff_square_tab + 256;

  236. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  254. 

  255.         pix1 += line_size;

  256.         pix2 += line_size;

  257.     }

  258.     return s;

  259. }

  260. 

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

  262.                           const uint8_t *s2, int stride)

  263. {

  264.     int i;

  265. 

  266.     /* read the pixels */

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

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

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

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

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

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

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

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

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

  276.         s1      += stride;

  277.         s2      += stride;

  278.         block   += 8;

  279.     }

  280. }

  281. 

  282. static void put_pixels_clamped_c(const int16_t *block, uint8_t *av_restrict pixels,

  283.                                  int line_size)

  284. {

  285.     int i;

  286. 

  287.     /* read the pixels */

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

  289.         pixels[0] = av_clip_uint8(block[0]);

  290.         pixels[1] = av_clip_uint8(block[1]);

  291.         pixels[2] = av_clip_uint8(block[2]);

  292.         pixels[3] = av_clip_uint8(block[3]);

  293.         pixels[4] = av_clip_uint8(block[4]);

  294.         pixels[5] = av_clip_uint8(block[5]);

  295.         pixels[6] = av_clip_uint8(block[6]);

  296.         pixels[7] = av_clip_uint8(block[7]);

  297. 

  298.         pixels += line_size;

  299.         block  += 8;

  300.     }

  301. }

  302. 

  303. static void put_pixels_clamped4_c(const int16_t *block, uint8_t *av_restrict pixels,

  304.                                  int line_size)

  305. {

  306.     int i;

  307. 

  308.     /* read the pixels */

  309.     for(i=0;i<4;i++) {

  310.         pixels[0] = av_clip_uint8(block[0]);

  311.         pixels[1] = av_clip_uint8(block[1]);

  312.         pixels[2] = av_clip_uint8(block[2]);

  313.         pixels[3] = av_clip_uint8(block[3]);

  314. 

  315.         pixels += line_size;

  316.         block += 8;

  317.     }

  318. }

  319. 

  320. static void put_pixels_clamped2_c(const int16_t *block, uint8_t *av_restrict pixels,

  321.                                  int line_size)

  322. {

  323.     int i;

  324. 

  325.     /* read the pixels */

  326.     for(i=0;i<2;i++) {

  327.         pixels[0] = av_clip_uint8(block[0]);

  328.         pixels[1] = av_clip_uint8(block[1]);

  329. 

  330.         pixels += line_size;

  331.         block += 8;

  332.     }

  333. }

  334. 

  335. static void put_signed_pixels_clamped_c(const int16_t *block,

  336.                                         uint8_t *av_restrict pixels,

  337.                                         int line_size)

  338. {

  339.     int i, j;

  340. 

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

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

  343.             if (*block < -128)

  344.                 *pixels = 0;

  345.             else if (*block > 127)

  346.                 *pixels = 255;

  347.             else

  348.                 *pixels = (uint8_t) (*block + 128);

  349.             block++;

  350.             pixels++;

  351.         }

  352.         pixels += (line_size - 8);

  353.     }

  354. }

  355. 

  356. static void add_pixels_clamped_c(const int16_t *block, uint8_t *av_restrict pixels,

  357.                                  int line_size)

  358. {

  359.     int i;

  360. 

  361.     /* read the pixels */

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

  363.         pixels[0] = av_clip_uint8(pixels[0] + block[0]);

  364.         pixels[1] = av_clip_uint8(pixels[1] + block[1]);

  365.         pixels[2] = av_clip_uint8(pixels[2] + block[2]);

  366.         pixels[3] = av_clip_uint8(pixels[3] + block[3]);

  367.         pixels[4] = av_clip_uint8(pixels[4] + block[4]);

  368.         pixels[5] = av_clip_uint8(pixels[5] + block[5]);

  369.         pixels[6] = av_clip_uint8(pixels[6] + block[6]);

  370.         pixels[7] = av_clip_uint8(pixels[7] + block[7]);

  371.         pixels   += line_size;

  372.         block    += 8;

  373.     }

  374. }

  375. 

  376. static void add_pixels_clamped4_c(const int16_t *block, uint8_t *av_restrict pixels,

  377.                           int line_size)

  378. {

  379.     int i;

  380. 

  381.     /* read the pixels */

  382.     for(i=0;i<4;i++) {

  383.         pixels[0] = av_clip_uint8(pixels[0] + block[0]);

  384.         pixels[1] = av_clip_uint8(pixels[1] + block[1]);

  385.         pixels[2] = av_clip_uint8(pixels[2] + block[2]);

  386.         pixels[3] = av_clip_uint8(pixels[3] + block[3]);

  387.         pixels += line_size;

  388.         block += 8;

  389.     }

  390. }

  391. 

  392. static void add_pixels_clamped2_c(const int16_t *block, uint8_t *av_restrict pixels,

  393.                           int line_size)

  394. {

  395.     int i;

  396. 

  397.     /* read the pixels */

  398.     for(i=0;i<2;i++) {

  399.         pixels[0] = av_clip_uint8(pixels[0] + block[0]);

  400.         pixels[1] = av_clip_uint8(pixels[1] + block[1]);

  401.         pixels += line_size;

  402.         block += 8;

  403.     }

  404. }

  405. 

  406. static int sum_abs_dctelem_c(int16_t *block)

  407. {

  408.     int sum = 0, i;

  409. 

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

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

  412.     return sum;

  413. }

  414. 

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

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

  417. 

  418. static void gmc1_c(uint8_t *dst, uint8_t *src, int stride, int h,

  419.                    int x16, int y16, int rounder)

  420. {

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

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

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

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

  425.     int i;

  426. 

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

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

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

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

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

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

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

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

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

  436.         dst   += stride;

  437.         src   += stride;

  438.     }

  439. }

  440. 

  441. void ff_gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,

  442.               int dxx, int dxy, int dyx, int dyy, int shift, int r,

  443.               int width, int height)

  444. {

  445.     int y, vx, vy;

  446.     const int s = 1 << shift;

  447. 

  448.     width--;

  449.     height--;

  450. 

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

  452.         int x;

  453. 

  454.         vx = ox;

  455.         vy = oy;

  456.         for (x = 0; x < 8; x++) { // FIXME: optimize

  457.             int index;

  458.             int src_x  = vx >> 16;

  459.             int src_y  = vy >> 16;

  460.             int frac_x = src_x & (s - 1);

  461.             int frac_y = src_y & (s - 1);

  462. 

  463.             src_x >>= shift;

  464.             src_y >>= shift;

  465. 

  466.             if ((unsigned) src_x < width) {

  467.                 if ((unsigned) src_y < height) {

  468.                     index = src_x + src_y * stride;

  469.                     dst[y * stride + x] =

  470.                         ((src[index]                        * (s - frac_x) +

  471.                           src[index + 1]          * frac_x) * (s - frac_y) +

  472.                          (src[index + stride]               * (s - frac_x) +

  473.                           src[index + stride + 1] * frac_x) *      frac_y  +

  474.                          r) >> (shift * 2);

  475.                 } else {

  476.                     index = src_x + av_clip(src_y, 0, height) * stride;

  477.                     dst[y * stride + x] =

  478.                         ((src[index]               * (s - frac_x) +

  479.                           src[index + 1] * frac_x) *  s           +

  480.                          r) >> (shift * 2);

  481.                 }

  482.             } else {

  483.                 if ((unsigned) src_y < height) {

  484.                     index = av_clip(src_x, 0, width) + src_y * stride;

  485.                     dst[y * stride + x] =

  486.                         ((src[index]                    * (s - frac_y) +

  487.                           src[index + stride] * frac_y) *  s           +

  488.                          r) >> (shift * 2);

  489.                 } else {

  490.                     index = av_clip(src_x, 0, width) +

  491.                             av_clip(src_y, 0, height) * stride;

  492.                     dst[y * stride + x] = src[index];

  493.                 }

  494.             }

  495. 

  496.             vx += dxx;

  497.             vy += dyx;

  498.         }

  499.         ox += dxy;

  500.         oy += dyy;

  501.     }

  502. }

  503. 

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

  505.                               int line_size, int h)

  506. {

  507.     int s = 0, i;

  508. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  526.         pix1 += line_size;

  527.         pix2 += line_size;

  528.     }

  529.     return s;

  530. }

  531. 

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

  533.                           int line_size, int h)

  534. {

  535.     int s = 0, i;

  536. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  554.         pix1 += line_size;

  555.         pix2 += line_size;

  556.     }

  557.     return s;

  558. }

  559. 

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

  561.                           int line_size, int h)

  562. {

  563.     int s = 0, i;

  564.     uint8_t *pix3 = pix2 + line_size;

  565. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  583.         pix1 += line_size;

  584.         pix2 += line_size;

  585.         pix3 += line_size;

  586.     }

  587.     return s;

  588. }

  589. 

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

  591.                            int line_size, int h)

  592. {

  593.     int s = 0, i;

  594.     uint8_t *pix3 = pix2 + line_size;

  595. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  613.         pix1 += line_size;

  614.         pix2 += line_size;

  615.         pix3 += line_size;

  616.     }

  617.     return s;

  618. }

  619. 

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

  621.                              int line_size, int h)

  622. {

  623.     int s = 0, i;

  624. 

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

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

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

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

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

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

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

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

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

  634.         pix1 += line_size;

  635.         pix2 += line_size;

  636.     }

  637.     return s;

  638. }

  639. 

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

  641.                          int line_size, int h)

  642. {

  643.     int s = 0, i;

  644. 

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

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

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

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

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

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

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

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

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

  654.         pix1 += line_size;

  655.         pix2 += line_size;

  656.     }

  657.     return s;

  658. }

  659. 

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

  661.                          int line_size, int h)

  662. {

  663.     int s = 0, i;

  664.     uint8_t *pix3 = pix2 + line_size;

  665. 

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

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

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

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

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

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

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

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

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

  675.         pix1 += line_size;

  676.         pix2 += line_size;

  677.         pix3 += line_size;

  678.     }

  679.     return s;

  680. }

  681. 

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

  683.                           int line_size, int h)

  684. {

  685.     int s = 0, i;

  686.     uint8_t *pix3 = pix2 + line_size;

  687. 

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

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

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

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

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

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

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

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

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

  697.         pix1 += line_size;

  698.         pix2 += line_size;

  699.         pix3 += line_size;

  700.     }

  701.     return s;

  702. }

  703. 

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

  705. {

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

  707. 

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

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

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

  711.         if (y + 1 < h) {

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

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

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

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

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

  717.         }

  718.         s1 += stride;

  719.         s2 += stride;

  720.     }

  721. 

  722.     if (c)

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

  724.     else

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

  726. }

  727. 

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

  729. {

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

  731. 

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

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

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

  735.         if (y + 1 < h) {

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

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

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

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

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

  741.         }

  742.         s1 += stride;

  743.         s2 += stride;

  744.     }

  745. 

  746.     if (c)

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

  748.     else

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

  750. }

  751. 

  752. static int try_8x8basis_c(int16_t rem[64], int16_t weight[64],

  753.                           int16_t basis[64], int scale)

  754. {

  755.     int i;

  756.     unsigned int sum = 0;

  757. 

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

  759.         int b = rem[i] + ((basis[i] * scale +

  760.                            (1 << (BASIS_SHIFT - RECON_SHIFT - 1))) >>

  761.                           (BASIS_SHIFT - RECON_SHIFT));

  762.         int w = weight[i];

  763.         b >>= RECON_SHIFT;

  764.         av_assert2(-512 < b && b < 512);

  765. 

  766.         sum += (w * b) * (w * b) >> 4;

  767.     }

  768.     return sum >> 2;

  769. }

  770. 

  771. static void add_8x8basis_c(int16_t rem[64], int16_t basis[64], int scale)

  772. {

  773.     int i;

  774. 

  775.     for (i = 0; i < 8 * 8; i++)

  776.         rem[i] += (basis[i] * scale +

  777.                    (1 << (BASIS_SHIFT - RECON_SHIFT - 1))) >>

  778.                   (BASIS_SHIFT - RECON_SHIFT);

  779. }

  780. 

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

  782.                     int stride, int h)

  783. {

  784.     return 0;

  785. }

  786. 

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

  788. {

  789.     int i;

  790. 

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

  792. 

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

  794.         switch (type & 0xFF) {

  795.         case FF_CMP_SAD:

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

  797.             break;

  798.         case FF_CMP_SATD:

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

  800.             break;

  801.         case FF_CMP_SSE:

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

  803.             break;

  804.         case FF_CMP_DCT:

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

  806.             break;

  807.         case FF_CMP_DCT264:

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

  809.             break;

  810.         case FF_CMP_DCTMAX:

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

  812.             break;

  813.         case FF_CMP_PSNR:

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

  815.             break;

  816.         case FF_CMP_BIT:

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

  818.             break;

  819.         case FF_CMP_RD:

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

  821.             break;

  822.         case FF_CMP_VSAD:

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

  824.             break;

  825.         case FF_CMP_VSSE:

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

  827.             break;

  828.         case FF_CMP_ZERO:

  829.             cmp[i] = zero_cmp;

  830.             break;

  831.         case FF_CMP_NSSE:

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

  833.             break;

  834. #if CONFIG_DWT

  835.         case FF_CMP_W53:

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

  837.             break;

  838.         case FF_CMP_W97:

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

  840.             break;

  841. #endif

  842.         default:

  843.             av_log(NULL, AV_LOG_ERROR,

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

  845.         }

  846.     }

  847. }

  848. 

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

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

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

  852. 

  853. #define BUTTERFLY1(x, y)                        \

  854.     {                                           \

  855.         int a, b;                               \

  856.         a = x;                                  \

  857.         b = y;                                  \

  858.         x = a + b;                              \

  859.         y = a - b;                              \

  860.     }

  861. 

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

  863. 

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

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

  866. {

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

  868. 

  869.     av_assert2(h == 8);

  870. 

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

  872.         // FIXME: try pointer walks

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

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

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

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

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

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

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

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

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

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

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

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

  885. 

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

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

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

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

  890. 

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

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

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

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

  895.     }

  896. 

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

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

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

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

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

  902. 

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

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

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

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

  907. 

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

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

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

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

  912.     }

  913.     return sum;

  914. }

  915. 

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

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

  918. {

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

  920. 

  921.     av_assert2(h == 8);

  922. 

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

  924.         // FIXME: try pointer walks

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

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

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

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

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

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

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

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

  933. 

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

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

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

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

  938. 

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

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

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

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

  943.     }

  944. 

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

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

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

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

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

  950. 

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

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

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

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

  955. 

  956.         sum +=

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

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

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

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

  961.     }

  962. 

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

  964. 

  965.     return sum;

  966. }

  967. 

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

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

  970. {

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

  972. 

  973.     av_assert2(h == 8);

  974. 

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

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

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

  978. }

  979. 

  980. #if CONFIG_GPL

  981. #define DCT8_1D                                         \

  982.     {                                                   \

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

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

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

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

  987.         const int a0  = s07 + s34;                      \

  988.         const int a1  = s16 + s25;                      \

  989.         const int a2  = s07 - s34;                      \

  990.         const int a3  = s16 - s25;                      \

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

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

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

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

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

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

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

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

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

  1000.         DST(1, a4 + (a7 >> 2));                         \

  1001.         DST(2, a2 + (a3 >> 1));                         \

  1002.         DST(3, a5 + (a6 >> 2));                         \

  1003.         DST(4, a0 - a1);                                \

  1004.         DST(5, a6 - (a5 >> 2));                         \

  1005.         DST(6, (a2 >> 1) - a3);                         \

  1006.         DST(7, (a4 >> 2) - a7);                         \

  1007.     }

  1008. 

  1009. static int dct264_sad8x8_c(MpegEncContext *s, uint8_t *src1,

  1010.                            uint8_t *src2, int stride, int h)

  1011. {

  1012.     int16_t dct[8][8];

  1013.     int i, sum = 0;

  1014. 

  1015.     s->dsp.diff_pixels(dct[0], src1, src2, stride);

  1016. 

  1017. #define SRC(x) dct[i][x]

  1018. #define DST(x, v) dct[i][x] = v

  1019.     for (i = 0; i < 8; i++)

  1020.         DCT8_1D

  1021. #undef SRC

  1022. #undef DST

  1023. 

  1024. #define SRC(x) dct[x][i]

  1025. #define DST(x, v) sum += FFABS(v)

  1026.         for (i = 0; i < 8; i++)

  1027.             DCT8_1D

  1028. #undef SRC

  1029. #undef DST

  1030.             return sum;

  1031. }

  1032. #endif

  1033. 

  1034. static int dct_max8x8_c(MpegEncContext *s, uint8_t *src1,

  1035.                         uint8_t *src2, int stride, int h)

  1036. {

  1037.     LOCAL_ALIGNED_16(int16_t, temp, [64]);

  1038.     int sum = 0, i;

  1039. 

  1040.     av_assert2(h == 8);

  1041. 

  1042.     s->dsp.diff_pixels(temp, src1, src2, stride);

  1043.     s->dsp.fdct(temp);

  1044. 

  1045.     for (i = 0; i < 64; i++)

  1046.         sum = FFMAX(sum, FFABS(temp[i]));

  1047. 

  1048.     return sum;

  1049. }

  1050. 

  1051. static int quant_psnr8x8_c(MpegEncContext *s, uint8_t *src1,

  1052.                            uint8_t *src2, int stride, int h)

  1053. {

  1054.     LOCAL_ALIGNED_16(int16_t, temp, [64 * 2]);

  1055.     int16_t *const bak = temp + 64;

  1056.     int sum = 0, i;

  1057. 

  1058.     av_assert2(h == 8);

  1059.     s->mb_intra = 0;

  1060. 

  1061.     s->dsp.diff_pixels(temp, src1, src2, stride);

  1062. 

  1063.     memcpy(bak, temp, 64 * sizeof(int16_t));

  1064. 

  1065.     s->block_last_index[0 /* FIXME */] =

  1066.         s->fast_dct_quantize(s, temp, 0 /* FIXME */, s->qscale, &i);

  1067.     s->dct_unquantize_inter(s, temp, 0, s->qscale);

  1068.     ff_simple_idct_8(temp); // FIXME

  1069. 

  1070.     for (i = 0; i < 64; i++)

  1071.         sum += (temp[i] - bak[i]) * (temp[i] - bak[i]);

  1072. 

  1073.     return sum;

  1074. }

  1075. 

  1076. static int rd8x8_c(MpegEncContext *s, uint8_t *src1, uint8_t *src2,

  1077.                    int stride, int h)

  1078. {

  1079.     const uint8_t *scantable = s->intra_scantable.permutated;

  1080.     LOCAL_ALIGNED_16(int16_t, temp, [64]);

  1081.     LOCAL_ALIGNED_16(uint8_t, lsrc1, [64]);

  1082.     LOCAL_ALIGNED_16(uint8_t, lsrc2, [64]);

  1083.     int i, last, run, bits, level, distortion, start_i;

  1084.     const int esc_length = s->ac_esc_length;

  1085.     uint8_t *length, *last_length;

  1086. 

  1087.     av_assert2(h == 8);

  1088. 

  1089.     copy_block8(lsrc1, src1, 8, stride, 8);

  1090.     copy_block8(lsrc2, src2, 8, stride, 8);

  1091. 

  1092.     s->dsp.diff_pixels(temp, lsrc1, lsrc2, 8);

  1093. 

  1094.     s->block_last_index[0 /* FIXME */] =

  1095.     last                               =

  1096.         s->fast_dct_quantize(s, temp, 0 /* FIXME */, s->qscale, &i);

  1097. 

  1098.     bits = 0;

  1099. 

  1100.     if (s->mb_intra) {

  1101.         start_i     = 1;

  1102.         length      = s->intra_ac_vlc_length;

  1103.         last_length = s->intra_ac_vlc_last_length;

  1104.         bits       += s->luma_dc_vlc_length[temp[0] + 256]; // FIXME: chroma

  1105.     } else {

  1106.         start_i     = 0;

  1107.         length      = s->inter_ac_vlc_length;

  1108.         last_length = s->inter_ac_vlc_last_length;

  1109.     }

  1110. 

  1111.     if (last >= start_i) {

  1112.         run = 0;

  1113.         for (i = start_i; i < last; i++) {

  1114.             int j = scantable[i];

  1115.             level = temp[j];

  1116. 

  1117.             if (level) {

  1118.                 level += 64;

  1119.                 if ((level & (~127)) == 0)

  1120.                     bits += length[UNI_AC_ENC_INDEX(run, level)];

  1121.                 else

  1122.                     bits += esc_length;

  1123.                 run = 0;

  1124.             } else

  1125.                 run++;

  1126.         }

  1127.         i = scantable[last];

  1128. 

  1129.         level = temp[i] + 64;

  1130. 

  1131.         av_assert2(level - 64);

  1132. 

  1133.         if ((level & (~127)) == 0) {

  1134.             bits += last_length[UNI_AC_ENC_INDEX(run, level)];

  1135.         } else

  1136.             bits += esc_length;

  1137.     }

  1138. 

  1139.     if (last >= 0) {

  1140.         if (s->mb_intra)

  1141.             s->dct_unquantize_intra(s, temp, 0, s->qscale);

  1142.         else

  1143.             s->dct_unquantize_inter(s, temp, 0, s->qscale);

  1144.     }

  1145. 

  1146.     s->dsp.idct_add(lsrc2, 8, temp);

  1147. 

  1148.     distortion = s->dsp.sse[1](NULL, lsrc2, lsrc1, 8, 8);

  1149. 

  1150.     return distortion + ((bits * s->qscale * s->qscale * 109 + 64) >> 7);

  1151. }

  1152. 

  1153. static int bit8x8_c(MpegEncContext *s, uint8_t *src1, uint8_t *src2,

  1154.                     int stride, int h)

  1155. {

  1156.     const uint8_t *scantable = s->intra_scantable.permutated;

  1157.     LOCAL_ALIGNED_16(int16_t, temp, [64]);

  1158.     int i, last, run, bits, level, start_i;

  1159.     const int esc_length = s->ac_esc_length;

  1160.     uint8_t *length, *last_length;

  1161. 

  1162.     av_assert2(h == 8);

  1163. 

  1164.     s->dsp.diff_pixels(temp, src1, src2, stride);

  1165. 

  1166.     s->block_last_index[0 /* FIXME */] =

  1167.     last                               =

  1168.         s->fast_dct_quantize(s, temp, 0 /* FIXME */, s->qscale, &i);

  1169. 

  1170.     bits = 0;

  1171. 

  1172.     if (s->mb_intra) {

  1173.         start_i     = 1;

  1174.         length      = s->intra_ac_vlc_length;

  1175.         last_length = s->intra_ac_vlc_last_length;

  1176.         bits       += s->luma_dc_vlc_length[temp[0] + 256]; // FIXME: chroma

  1177.     } else {

  1178.         start_i     = 0;

  1179.         length      = s->inter_ac_vlc_length;

  1180.         last_length = s->inter_ac_vlc_last_length;

  1181.     }

  1182. 

  1183.     if (last >= start_i) {

  1184.         run = 0;

  1185.         for (i = start_i; i < last; i++) {

  1186.             int j = scantable[i];

  1187.             level = temp[j];

  1188. 

  1189.             if (level) {

  1190.                 level += 64;

  1191.                 if ((level & (~127)) == 0)

  1192.                     bits += length[UNI_AC_ENC_INDEX(run, level)];

  1193.                 else

  1194.                     bits += esc_length;

  1195.                 run = 0;

  1196.             } else

  1197.                 run++;

  1198.         }

  1199.         i = scantable[last];

  1200. 

  1201.         level = temp[i] + 64;

  1202. 

  1203.         av_assert2(level - 64);

  1204. 

  1205.         if ((level & (~127)) == 0)

  1206.             bits += last_length[UNI_AC_ENC_INDEX(run, level)];

  1207.         else

  1208.             bits += esc_length;

  1209.     }

  1210. 

  1211.     return bits;

  1212. }

  1213. 

  1214. #define VSAD_INTRA(size)                                                \

  1215. static int vsad_intra ## size ## _c(MpegEncContext *c,                  \

  1216.                                     uint8_t *s, uint8_t *dummy,         \

  1217.                                     int stride, int h)                  \

  1218. {                                                                       \

  1219.     int score = 0, x, y;                                                \

  1220.                                                                         \

  1221.     for (y = 1; y < h; y++) {                                           \

  1222.         for (x = 0; x < size; x += 4) {                                 \

  1223.             score += FFABS(s[x]     - s[x + stride])     +              \

  1224.                      FFABS(s[x + 1] - s[x + stride + 1]) +              \

  1225.                      FFABS(s[x + 2] - s[x + 2 + stride]) +              \

  1226.                      FFABS(s[x + 3] - s[x + 3 + stride]);               \

  1227.         }                                                               \

  1228.         s += stride;                                                    \

  1229.     }                                                                   \

  1230.                                                                         \

  1231.     return score;                                                       \

  1232. }

  1233. VSAD_INTRA(8)

  1234. VSAD_INTRA(16)

  1235. 

  1236. #define VSAD(size)                                                             \

  1237. static int vsad ## size ## _c(MpegEncContext *c,                               \

  1238.                               uint8_t *s1, uint8_t *s2,                        \

  1239.                               int stride, int h)                               \

  1240. {                                                                              \

  1241.     int score = 0, x, y;                                                       \

  1242.                                                                                \

  1243.     for (y = 1; y < h; y++) {                                                  \

  1244.         for (x = 0; x < size; x++)                                             \

  1245.             score += FFABS(s1[x] - s2[x] - s1[x + stride] + s2[x + stride]);   \

  1246.         s1 += stride;                                                          \

  1247.         s2 += stride;                                                          \

  1248.     }                                                                          \

  1249.                                                                                \

  1250.     return score;                                                              \

  1251. }

  1252. VSAD(8)

  1253. VSAD(16)

  1254. 

  1255. #define SQ(a) ((a) * (a))

  1256. #define VSSE_INTRA(size)                                                \

  1257. static int vsse_intra ## size ## _c(MpegEncContext *c,                  \

  1258.                                     uint8_t *s, uint8_t *dummy,         \

  1259.                                     int stride, int h)                  \

  1260. {                                                                       \

  1261.     int score = 0, x, y;                                                \

  1262.                                                                         \

  1263.     for (y = 1; y < h; y++) {                                           \

  1264.         for (x = 0; x < size; x += 4) {                                 \

  1265.             score += SQ(s[x]     - s[x + stride]) +                     \

  1266.                      SQ(s[x + 1] - s[x + stride + 1]) +                 \

  1267.                      SQ(s[x + 2] - s[x + stride + 2]) +                 \

  1268.                      SQ(s[x + 3] - s[x + stride + 3]);                  \

  1269.         }                                                               \

  1270.         s += stride;                                                    \

  1271.     }                                                                   \

  1272.                                                                         \

  1273.     return score;                                                       \

  1274. }

  1275. VSSE_INTRA(8)

  1276. VSSE_INTRA(16)

  1277. 

  1278. #define VSSE(size)                                                             \

  1279. static int vsse ## size ## _c(MpegEncContext *c, uint8_t *s1, uint8_t *s2,     \

  1280.                     int stride, int h)                                         \

  1281. {                                                                              \

  1282.     int score = 0, x, y;                                                       \

  1283.                                                                                \

  1284.     for (y = 1; y < h; y++) {                                                  \

  1285.         for (x = 0; x < size; x++)                                             \

  1286.             score += SQ(s1[x] - s2[x] - s1[x + stride] + s2[x + stride]);      \

  1287.         s1 += stride;                                                          \

  1288.         s2 += stride;                                                          \

  1289.     }                                                                          \

  1290.                                                                                \

  1291.     return score;                                                              \

  1292. }

  1293. VSSE(8)

  1294. VSSE(16)

  1295. 

  1296. #define WRAPPER8_16_SQ(name8, name16)                                   \

  1297. static int name16(MpegEncContext *s, uint8_t *dst, uint8_t *src,        \

  1298.                   int stride, int h)                                    \

  1299. {                                                                       \

  1300.     int score = 0;                                                      \

  1301.                                                                         \

  1302.     score += name8(s, dst, src, stride, 8);                             \

  1303.     score += name8(s, dst + 8, src + 8, stride, 8);                     \

  1304.     if (h == 16) {                                                      \

  1305.         dst   += 8 * stride;                                            \

  1306.         src   += 8 * stride;                                            \

  1307.         score += name8(s, dst, src, stride, 8);                         \

  1308.         score += name8(s, dst + 8, src + 8, stride, 8);                 \

  1309.     }                                                                   \

  1310.     return score;                                                       \

  1311. }

  1312. 

  1313. WRAPPER8_16_SQ(hadamard8_diff8x8_c, hadamard8_diff16_c)

  1314. WRAPPER8_16_SQ(hadamard8_intra8x8_c, hadamard8_intra16_c)

  1315. WRAPPER8_16_SQ(dct_sad8x8_c, dct_sad16_c)

  1316. #if CONFIG_GPL

  1317. WRAPPER8_16_SQ(dct264_sad8x8_c, dct264_sad16_c)

  1318. #endif

  1319. WRAPPER8_16_SQ(dct_max8x8_c, dct_max16_c)

  1320. WRAPPER8_16_SQ(quant_psnr8x8_c, quant_psnr16_c)

  1321. WRAPPER8_16_SQ(rd8x8_c, rd16_c)

  1322. WRAPPER8_16_SQ(bit8x8_c, bit16_c)

  1323. 

  1324. static void jref_idct_put(uint8_t *dest, int line_size, int16_t *block)

  1325. {

  1326.     ff_j_rev_dct(block);

  1327.     put_pixels_clamped_c(block, dest, line_size);

  1328. }

  1329. 

  1330. static void jref_idct_add(uint8_t *dest, int line_size, int16_t *block)

  1331. {

  1332.     ff_j_rev_dct(block);

  1333.     add_pixels_clamped_c(block, dest, line_size);

  1334. }

  1335. 

  1336. static void ff_jref_idct4_put(uint8_t *dest, int line_size, int16_t *block)

  1337. {

  1338.     ff_j_rev_dct4 (block);

  1339.     put_pixels_clamped4_c(block, dest, line_size);

  1340. }

  1341. static void ff_jref_idct4_add(uint8_t *dest, int line_size, int16_t *block)

  1342. {

  1343.     ff_j_rev_dct4 (block);

  1344.     add_pixels_clamped4_c(block, dest, line_size);

  1345. }

  1346. 

  1347. static void ff_jref_idct2_put(uint8_t *dest, int line_size, int16_t *block)

  1348. {

  1349.     ff_j_rev_dct2 (block);

  1350.     put_pixels_clamped2_c(block, dest, line_size);

  1351. }

  1352. static void ff_jref_idct2_add(uint8_t *dest, int line_size, int16_t *block)

  1353. {

  1354.     ff_j_rev_dct2 (block);

  1355.     add_pixels_clamped2_c(block, dest, line_size);

  1356. }

  1357. 

  1358. static void ff_jref_idct1_put(uint8_t *dest, int line_size, int16_t *block)

  1359. {

  1360.     dest[0] = av_clip_uint8((block[0] + 4)>>3);

  1361. }

  1362. static void ff_jref_idct1_add(uint8_t *dest, int line_size, int16_t *block)

  1363. {

  1364.     dest[0] = av_clip_uint8(dest[0] + ((block[0] + 4)>>3));

  1365. }

  1366. 

  1367. /* draw the edges of width 'w' of an image of size width, height */

  1368. // FIXME: Check that this is OK for MPEG-4 interlaced.

  1369. static void draw_edges_8_c(uint8_t *buf, int wrap, int width, int height,

  1370.                            int w, int h, int sides)

  1371. {

  1372.     uint8_t *ptr = buf, *last_line;

  1373.     int i;

  1374. 

  1375.     /* left and right */

  1376.     for (i = 0; i < height; i++) {

  1377.         memset(ptr - w, ptr[0], w);

  1378.         memset(ptr + width, ptr[width - 1], w);

  1379.         ptr += wrap;

  1380.     }

  1381. 

  1382.     /* top and bottom + corners */

  1383.     buf -= w;

  1384.     last_line = buf + (height - 1) * wrap;

  1385.     if (sides & EDGE_TOP)

  1386.         for (i = 0; i < h; i++)

  1387.             // top

  1388.             memcpy(buf - (i + 1) * wrap, buf, width + w + w);

  1389.     if (sides & EDGE_BOTTOM)

  1390.         for (i = 0; i < h; i++)

  1391.             // bottom

  1392.             memcpy(last_line + (i + 1) * wrap, last_line, width + w + w);

  1393. }

  1394. 

  1395. /* init static data */

  1396. av_cold void ff_dsputil_static_init(void)

  1397. {

  1398.     int i;

  1399. 

  1400.     for (i = 0; i < 512; i++)

  1401.         ff_square_tab[i] = (i - 256) * (i - 256);

  1402. }

  1403. 

  1404. int ff_check_alignment(void)

  1405. {

  1406.     static int did_fail = 0;

  1407.     LOCAL_ALIGNED_16(int, aligned, [4]);

  1408. 

  1409.     if ((intptr_t)aligned & 15) {

  1410.         if (!did_fail) {

  1411. #if HAVE_MMX || HAVE_ALTIVEC

  1412.             av_log(NULL, AV_LOG_ERROR,

  1413.                 "Compiler did not align stack variables. Libavcodec has been miscompiled\n"

  1414.                 "and may be very slow or crash. This is not a bug in libavcodec,\n"

  1415.                 "but in the compiler. You may try recompiling using gcc >= 4.2.\n"

  1416.                 "Do not report crashes to FFmpeg developers.\n");

  1417. #endif

  1418.             did_fail=1;

  1419.         }

  1420.         return -1;

  1421.     }

  1422.     return 0;

  1423. }

  1424. 

  1425. av_cold void ff_dsputil_init(DSPContext *c, AVCodecContext *avctx)

  1426. {

  1427.     const unsigned high_bit_depth = avctx->bits_per_raw_sample > 8;

  1428. 

  1429.     ff_check_alignment();

  1430. 

  1431. #if CONFIG_ENCODERS

  1432.     if (avctx->bits_per_raw_sample == 10) {

  1433.         c->fdct    = ff_jpeg_fdct_islow_10;

  1434.         c->fdct248 = ff_fdct248_islow_10;

  1435.     } else {

  1436.         if (avctx->dct_algo == FF_DCT_FASTINT) {

  1437.             c->fdct    = ff_fdct_ifast;

  1438.             c->fdct248 = ff_fdct_ifast248;

  1439.         } else if (avctx->dct_algo == FF_DCT_FAAN) {

  1440.             c->fdct    = ff_faandct;

  1441.             c->fdct248 = ff_faandct248;

  1442.         } else {

  1443.             c->fdct    = ff_jpeg_fdct_islow_8; // slow/accurate/default

  1444.             c->fdct248 = ff_fdct248_islow_8;

  1445.         }

  1446.     }

  1447. #endif /* CONFIG_ENCODERS */

  1448. 

  1449.     if (avctx->lowres==1) {

  1450.         c->idct_put              = ff_jref_idct4_put;

  1451.         c->idct_add              = ff_jref_idct4_add;

  1452.         c->idct                  = ff_j_rev_dct4;

  1453.         c->idct_permutation_type = FF_NO_IDCT_PERM;

  1454.     } else if (avctx->lowres==2) {

  1455.         c->idct_put              =  ff_jref_idct2_put;

  1456.         c->idct_add              =  ff_jref_idct2_add;

  1457.         c->idct                  =  ff_j_rev_dct2;

  1458.         c->idct_permutation_type = FF_NO_IDCT_PERM;

  1459.     } else if (avctx->lowres==3) {

  1460.         c->idct_put              =  ff_jref_idct1_put;

  1461.         c->idct_add              =  ff_jref_idct1_add;

  1462.         c->idct                  =  ff_j_rev_dct1;

  1463.         c->idct_permutation_type = FF_NO_IDCT_PERM;

  1464.     } else {

  1465.         if (avctx->bits_per_raw_sample == 10) {

  1466.             c->idct_put              = ff_simple_idct_put_10;

  1467.             c->idct_add              = ff_simple_idct_add_10;

  1468.             c->idct                  = ff_simple_idct_10;

  1469.             c->idct_permutation_type = FF_NO_IDCT_PERM;

  1470.         } else if (avctx->bits_per_raw_sample == 12) {

  1471.             c->idct_put              = ff_simple_idct_put_12;

  1472.             c->idct_add              = ff_simple_idct_add_12;

  1473.             c->idct                  = ff_simple_idct_12;

  1474.             c->idct_permutation_type = FF_NO_IDCT_PERM;

  1475.         } else {

  1476.         if (avctx->idct_algo == FF_IDCT_INT) {

  1477.             c->idct_put              = jref_idct_put;

  1478.             c->idct_add              = jref_idct_add;

  1479.             c->idct                  = ff_j_rev_dct;

  1480.             c->idct_permutation_type = FF_LIBMPEG2_IDCT_PERM;

  1481.         } else if (avctx->idct_algo == FF_IDCT_FAAN) {

  1482.             c->idct_put              = ff_faanidct_put;

  1483.             c->idct_add              = ff_faanidct_add;

  1484.             c->idct                  = ff_faanidct;

  1485.             c->idct_permutation_type = FF_NO_IDCT_PERM;

  1486.         } else { // accurate/default

  1487.             c->idct_put              = ff_simple_idct_put_8;

  1488.             c->idct_add              = ff_simple_idct_add_8;

  1489.             c->idct                  = ff_simple_idct_8;

  1490.             c->idct_permutation_type = FF_NO_IDCT_PERM;

  1491.         }

  1492.         }

  1493.     }

  1494. 

  1495.     c->diff_pixels = diff_pixels_c;

  1496. 

  1497.     c->put_pixels_clamped        = put_pixels_clamped_c;

  1498.     c->put_signed_pixels_clamped = put_signed_pixels_clamped_c;

  1499.     c->add_pixels_clamped        = add_pixels_clamped_c;

  1500. 

  1501.     c->sum_abs_dctelem = sum_abs_dctelem_c;

  1502. 

  1503.     c->gmc1 = gmc1_c;

  1504.     c->gmc  = ff_gmc_c;

  1505. 

  1506.     c->pix_sum   = pix_sum_c;

  1507.     c->pix_norm1 = pix_norm1_c;

  1508. 

  1509.     /* TODO [0] 16  [1] 8 */

  1510.     c->pix_abs[0][0] = pix_abs16_c;

  1511.     c->pix_abs[0][1] = pix_abs16_x2_c;

  1512.     c->pix_abs[0][2] = pix_abs16_y2_c;

  1513.     c->pix_abs[0][3] = pix_abs16_xy2_c;

  1514.     c->pix_abs[1][0] = pix_abs8_c;

  1515.     c->pix_abs[1][1] = pix_abs8_x2_c;

  1516.     c->pix_abs[1][2] = pix_abs8_y2_c;

  1517.     c->pix_abs[1][3] = pix_abs8_xy2_c;

  1518. 

  1519. #define SET_CMP_FUNC(name)                      \

  1520.     c->name[0] = name ## 16_c;                  \

  1521.     c->name[1] = name ## 8x8_c;

  1522. 

  1523.     SET_CMP_FUNC(hadamard8_diff)

  1524.     c->hadamard8_diff[4] = hadamard8_intra16_c;

  1525.     c->hadamard8_diff[5] = hadamard8_intra8x8_c;

  1526.     SET_CMP_FUNC(dct_sad)

  1527.     SET_CMP_FUNC(dct_max)

  1528. #if CONFIG_GPL

  1529.     SET_CMP_FUNC(dct264_sad)

  1530. #endif

  1531.     c->sad[0] = pix_abs16_c;

  1532.     c->sad[1] = pix_abs8_c;

  1533.     c->sse[0] = sse16_c;

  1534.     c->sse[1] = sse8_c;

  1535.     c->sse[2] = sse4_c;

  1536.     SET_CMP_FUNC(quant_psnr)

  1537.     SET_CMP_FUNC(rd)

  1538.     SET_CMP_FUNC(bit)

  1539.     c->vsad[0] = vsad16_c;

  1540.     c->vsad[1] = vsad8_c;

  1541.     c->vsad[4] = vsad_intra16_c;

  1542.     c->vsad[5] = vsad_intra8_c;

  1543.     c->vsse[0] = vsse16_c;

  1544.     c->vsse[1] = vsse8_c;

  1545.     c->vsse[4] = vsse_intra16_c;

  1546.     c->vsse[5] = vsse_intra8_c;

  1547.     c->nsse[0] = nsse16_c;

  1548.     c->nsse[1] = nsse8_c;

  1549. #if CONFIG_SNOW_DECODER || CONFIG_SNOW_ENCODER

  1550.     ff_dsputil_init_dwt(c);

  1551. #endif

  1552. 

  1553.     c->try_8x8basis = try_8x8basis_c;

  1554.     c->add_8x8basis = add_8x8basis_c;

  1555. 

  1556.     c->shrink[0] = av_image_copy_plane;

  1557.     c->shrink[1] = ff_shrink22;

  1558.     c->shrink[2] = ff_shrink44;

  1559.     c->shrink[3] = ff_shrink88;

  1560. 

  1561.     c->draw_edges = draw_edges_8_c;

  1562. 

  1563.     switch (avctx->bits_per_raw_sample) {

  1564.     case 9:

  1565.     case 10:

  1566.     case 12:

  1567.     case 14:

  1568.         c->get_pixels = get_pixels_16_c;

  1569.         break;

  1570.     default:

  1571.         if (avctx->bits_per_raw_sample<=8 || avctx->codec_type != AVMEDIA_TYPE_VIDEO) {

  1572.             c->get_pixels = get_pixels_8_c;

  1573.         }

  1574.         break;

  1575.     }

  1576. 

  1577. 

  1578.     if (ARCH_ALPHA)

  1579.         ff_dsputil_init_alpha(c, avctx);

  1580.     if (ARCH_ARM)

  1581.         ff_dsputil_init_arm(c, avctx, high_bit_depth);

  1582.     if (ARCH_PPC)

  1583.         ff_dsputil_init_ppc(c, avctx, high_bit_depth);

  1584.     if (ARCH_X86)

  1585.         ff_dsputil_init_x86(c, avctx, high_bit_depth);

  1586. 

  1587.     ff_init_scantable_permutation(c->idct_permutation,

  1588.                                   c->idct_permutation_type);

  1589. }

  1590. 

  1591. av_cold void dsputil_init(DSPContext* c, AVCodecContext *avctx)

  1592. {

  1593.     ff_dsputil_init(c, avctx);

  1594. }

  1595. 

  1596. av_cold void avpriv_dsputil_init(DSPContext *c, AVCodecContext *avctx)

  1597. {

  1598.     ff_dsputil_init(c, avctx);

  1599. }