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.
39631 Commits
32 Branches
941MB
Tree: 53c20f17c7
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '53c20f17c7'
${ noResults }
FFmpeg/libavcodec/vp8dsp.c

569 lines
26KB
Raw Blame History 

  1. /*

  2.  * Copyright (C) 2010 David Conrad

  3.  * Copyright (C) 2010 Ronald S. Bultje

  4.  *

  5.  * This file is part of Libav.

  6.  *

  7.  * Libav is free software; you can redistribute it and/or

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

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

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

  11.  *

  12.  * Libav is distributed in the hope that it will be useful,

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

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

  15.  * Lesser General Public License for more details.

  16.  *

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

  18.  * License along with Libav; if not, write to the Free Software

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

  20.  */

  21. 

  22. /**

  23.  * @file

  24.  * VP8 compatible video decoder

  25.  */

  26. 

  27. #include "libavutil/common.h"

  28. 

  29. #include "mathops.h"

  30. #include "vp8dsp.h"

  31. 

  32. // TODO: Maybe add dequant

  33. static void vp8_luma_dc_wht_c(int16_t block[4][4][16], int16_t dc[16])

  34. {

  35.     int i, t0, t1, t2, t3;

  36. 

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

  38.         t0 = dc[0 * 4 + i] + dc[3 * 4 + i];

  39.         t1 = dc[1 * 4 + i] + dc[2 * 4 + i];

  40.         t2 = dc[1 * 4 + i] - dc[2 * 4 + i];

  41.         t3 = dc[0 * 4 + i] - dc[3 * 4 + i];

  42. 

  43.         dc[0 * 4 + i] = t0 + t1;

  44.         dc[1 * 4 + i] = t3 + t2;

  45.         dc[2 * 4 + i] = t0 - t1;

  46.         dc[3 * 4 + i] = t3 - t2;

  47.     }

  48. 

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

  50.         t0 = dc[i * 4 + 0] + dc[i * 4 + 3] + 3; // rounding

  51.         t1 = dc[i * 4 + 1] + dc[i * 4 + 2];

  52.         t2 = dc[i * 4 + 1] - dc[i * 4 + 2];

  53.         t3 = dc[i * 4 + 0] - dc[i * 4 + 3] + 3; // rounding

  54.         dc[i * 4 + 0] = 0;

  55.         dc[i * 4 + 1] = 0;

  56.         dc[i * 4 + 2] = 0;

  57.         dc[i * 4 + 3] = 0;

  58. 

  59.         block[i][0][0] = (t0 + t1) >> 3;

  60.         block[i][1][0] = (t3 + t2) >> 3;

  61.         block[i][2][0] = (t0 - t1) >> 3;

  62.         block[i][3][0] = (t3 - t2) >> 3;

  63.     }

  64. }

  65. 

  66. static void vp8_luma_dc_wht_dc_c(int16_t block[4][4][16], int16_t dc[16])

  67. {

  68.     int i, val = (dc[0] + 3) >> 3;

  69.     dc[0] = 0;

  70. 

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

  72.         block[i][0][0] = val;

  73.         block[i][1][0] = val;

  74.         block[i][2][0] = val;

  75.         block[i][3][0] = val;

  76.     }

  77. }

  78. 

  79. #define MUL_20091(a) ((((a) * 20091) >> 16) + (a))

  80. #define MUL_35468(a)  (((a) * 35468) >> 16)

  81. 

  82. static void vp8_idct_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)

  83. {

  84.     int i, t0, t1, t2, t3;

  85.     int16_t tmp[16];

  86. 

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

  88.         t0 = block[0 * 4 + i] + block[2 * 4 + i];

  89.         t1 = block[0 * 4 + i] - block[2 * 4 + i];

  90.         t2 = MUL_35468(block[1 * 4 + i]) - MUL_20091(block[3 * 4 + i]);

  91.         t3 = MUL_20091(block[1 * 4 + i]) + MUL_35468(block[3 * 4 + i]);

  92.         block[0 * 4 + i] = 0;

  93.         block[1 * 4 + i] = 0;

  94.         block[2 * 4 + i] = 0;

  95.         block[3 * 4 + i] = 0;

  96. 

  97.         tmp[i * 4 + 0] = t0 + t3;

  98.         tmp[i * 4 + 1] = t1 + t2;

  99.         tmp[i * 4 + 2] = t1 - t2;

  100.         tmp[i * 4 + 3] = t0 - t3;

  101.     }

  102. 

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

  104.         t0 = tmp[0 * 4 + i] + tmp[2 * 4 + i];

  105.         t1 = tmp[0 * 4 + i] - tmp[2 * 4 + i];

  106.         t2 = MUL_35468(tmp[1 * 4 + i]) - MUL_20091(tmp[3 * 4 + i]);

  107.         t3 = MUL_20091(tmp[1 * 4 + i]) + MUL_35468(tmp[3 * 4 + i]);

  108. 

  109.         dst[0] = av_clip_uint8(dst[0] + ((t0 + t3 + 4) >> 3));

  110.         dst[1] = av_clip_uint8(dst[1] + ((t1 + t2 + 4) >> 3));

  111.         dst[2] = av_clip_uint8(dst[2] + ((t1 - t2 + 4) >> 3));

  112.         dst[3] = av_clip_uint8(dst[3] + ((t0 - t3 + 4) >> 3));

  113.         dst   += stride;

  114.     }

  115. }

  116. 

  117. static void vp8_idct_dc_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)

  118. {

  119.     int i, dc = (block[0] + 4) >> 3;

  120.     block[0] = 0;

  121. 

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

  123.         dst[0] = av_clip_uint8(dst[0] + dc);

  124.         dst[1] = av_clip_uint8(dst[1] + dc);

  125.         dst[2] = av_clip_uint8(dst[2] + dc);

  126.         dst[3] = av_clip_uint8(dst[3] + dc);

  127.         dst   += stride;

  128.     }

  129. }

  130. 

  131. static void vp8_idct_dc_add4uv_c(uint8_t *dst, int16_t block[4][16],

  132.                                  ptrdiff_t stride)

  133. {

  134.     vp8_idct_dc_add_c(dst + stride * 0 + 0, block[0], stride);

  135.     vp8_idct_dc_add_c(dst + stride * 0 + 4, block[1], stride);

  136.     vp8_idct_dc_add_c(dst + stride * 4 + 0, block[2], stride);

  137.     vp8_idct_dc_add_c(dst + stride * 4 + 4, block[3], stride);

  138. }

  139. 

  140. static void vp8_idct_dc_add4y_c(uint8_t *dst, int16_t block[4][16],

  141.                                 ptrdiff_t stride)

  142. {

  143.     vp8_idct_dc_add_c(dst +  0, block[0], stride);

  144.     vp8_idct_dc_add_c(dst +  4, block[1], stride);

  145.     vp8_idct_dc_add_c(dst +  8, block[2], stride);

  146.     vp8_idct_dc_add_c(dst + 12, block[3], stride);

  147. }

  148. 

  149. // because I like only having two parameters to pass functions...

  150. #define LOAD_PIXELS                                                           \

  151.     int av_unused p3 = p[-4 * stride];                                        \

  152.     int av_unused p2 = p[-3 * stride];                                        \

  153.     int av_unused p1 = p[-2 * stride];                                        \

  154.     int av_unused p0 = p[-1 * stride];                                        \

  155.     int av_unused q0 = p[ 0 * stride];                                        \

  156.     int av_unused q1 = p[ 1 * stride];                                        \

  157.     int av_unused q2 = p[ 2 * stride];                                        \

  158.     int av_unused q3 = p[ 3 * stride];

  159. 

  160. #define clip_int8(n) (cm[n + 0x80] - 0x80)

  161. 

  162. static av_always_inline void filter_common(uint8_t *p, ptrdiff_t stride,

  163.                                            int is4tap)

  164. {

  165.     LOAD_PIXELS

  166.     int a, f1, f2;

  167.     const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;

  168. 

  169.     a = 3 * (q0 - p0);

  170. 

  171.     if (is4tap)

  172.         a += clip_int8(p1 - q1);

  173. 

  174.     a = clip_int8(a);

  175. 

  176.     // We deviate from the spec here with c(a+3) >> 3

  177.     // since that's what libvpx does.

  178.     f1 = FFMIN(a + 4, 127) >> 3;

  179.     f2 = FFMIN(a + 3, 127) >> 3;

  180. 

  181.     // Despite what the spec says, we do need to clamp here to

  182.     // be bitexact with libvpx.

  183.     p[-1 * stride] = cm[p0 + f2];

  184.     p[ 0 * stride] = cm[q0 - f1];

  185. 

  186.     // only used for _inner on blocks without high edge variance

  187.     if (!is4tap) {

  188.         a = (f1 + 1) >> 1;

  189.         p[-2 * stride] = cm[p1 + a];

  190.         p[ 1 * stride] = cm[q1 - a];

  191.     }

  192. }

  193. 

  194. static av_always_inline int simple_limit(uint8_t *p, ptrdiff_t stride, int flim)

  195. {

  196.     LOAD_PIXELS

  197.     return 2 * FFABS(p0 - q0) + (FFABS(p1 - q1) >> 1) <= flim;

  198. }

  199. 

  200. /**

  201.  * E - limit at the macroblock edge

  202.  * I - limit for interior difference

  203.  */

  204. static av_always_inline int normal_limit(uint8_t *p, ptrdiff_t stride,

  205.                                          int E, int I)

  206. {

  207.     LOAD_PIXELS

  208.     return simple_limit(p, stride, E) &&

  209.            FFABS(p3 - p2) <= I &&

  210.            FFABS(p2 - p1) <= I &&

  211.            FFABS(p1 - p0) <= I &&

  212.            FFABS(q3 - q2) <= I &&

  213.            FFABS(q2 - q1) <= I &&

  214.            FFABS(q1 - q0) <= I;

  215. }

  216. 

  217. // high edge variance

  218. static av_always_inline int hev(uint8_t *p, ptrdiff_t stride, int thresh)

  219. {

  220.     LOAD_PIXELS

  221.     return FFABS(p1 - p0) > thresh || FFABS(q1 - q0) > thresh;

  222. }

  223. 

  224. static av_always_inline void filter_mbedge(uint8_t *p, ptrdiff_t stride)

  225. {

  226.     int a0, a1, a2, w;

  227.     const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;

  228. 

  229.     LOAD_PIXELS

  230. 

  231.     w = clip_int8(p1 - q1);

  232.     w = clip_int8(w + 3 * (q0 - p0));

  233. 

  234.     a0 = (27 * w + 63) >> 7;

  235.     a1 = (18 * w + 63) >> 7;

  236.     a2 =  (9 * w + 63) >> 7;

  237. 

  238.     p[-3 * stride] = cm[p2 + a2];

  239.     p[-2 * stride] = cm[p1 + a1];

  240.     p[-1 * stride] = cm[p0 + a0];

  241.     p[ 0 * stride] = cm[q0 - a0];

  242.     p[ 1 * stride] = cm[q1 - a1];

  243.     p[ 2 * stride] = cm[q2 - a2];

  244. }

  245. 

  246. #define LOOP_FILTER(dir, size, stridea, strideb, maybe_inline)                \

  247. static maybe_inline                                                           \

  248. void vp8_ ## dir ## _loop_filter ## size ## _c(uint8_t *dst,                  \

  249.                                                ptrdiff_t stride,              \

  250.                                                int flim_E, int flim_I,        \

  251.                                                int hev_thresh)                \

  252. {                                                                             \

  253.     int i;                                                                    \

  254.     for (i = 0; i < size; i++)                                                \

  255.         if (normal_limit(dst + i * stridea, strideb, flim_E, flim_I)) {       \

  256.             if (hev(dst + i * stridea, strideb, hev_thresh))                  \

  257.                 filter_common(dst + i * stridea, strideb, 1);                 \

  258.             else                                                              \

  259.                 filter_mbedge(dst + i * stridea, strideb);                    \

  260.         }                                                                     \

  261. }                                                                             \

  262.                                                                               \

  263. static maybe_inline                                                           \

  264. void vp8_ ## dir ## _loop_filter ## size ## _inner_c(uint8_t *dst,            \

  265.                                                      ptrdiff_t stride,        \

  266.                                                      int flim_E, int flim_I,  \

  267.                                                      int hev_thresh)          \

  268. {                                                                             \

  269.     int i;                                                                    \

  270.     for (i = 0; i < size; i++)                                                \

  271.         if (normal_limit(dst + i * stridea, strideb, flim_E, flim_I)) {       \

  272.             int hv = hev(dst + i * stridea, strideb, hev_thresh);             \

  273.             if (hv)                                                           \

  274.                 filter_common(dst + i * stridea, strideb, 1);                 \

  275.             else                                                              \

  276.                 filter_common(dst + i * stridea, strideb, 0);                 \

  277.         }                                                                     \

  278. }

  279. 

  280. LOOP_FILTER(v, 16, 1, stride, )

  281. LOOP_FILTER(h, 16, stride, 1, )

  282. 

  283. #define UV_LOOP_FILTER(dir, stridea, strideb)                                 \

  284. LOOP_FILTER(dir, 8, stridea, strideb, av_always_inline)                       \

  285. static void vp8_ ## dir ## _loop_filter8uv_c(uint8_t *dstU, uint8_t *dstV,    \

  286.                                              ptrdiff_t stride, int fE,        \

  287.                                              int fI, int hev_thresh)          \

  288. {                                                                             \

  289.     vp8_ ## dir ## _loop_filter8_c(dstU, stride, fE, fI, hev_thresh);         \

  290.     vp8_ ## dir ## _loop_filter8_c(dstV, stride, fE, fI, hev_thresh);         \

  291. }                                                                             \

  292.                                                                               \

  293. static void vp8_ ## dir ## _loop_filter8uv_inner_c(uint8_t *dstU,             \

  294.                                                    uint8_t *dstV,             \

  295.                                                    ptrdiff_t stride, int fE,  \

  296.                                                    int fI, int hev_thresh)    \

  297. {                                                                             \

  298.     vp8_ ## dir ## _loop_filter8_inner_c(dstU, stride, fE, fI, hev_thresh);   \

  299.     vp8_ ## dir ## _loop_filter8_inner_c(dstV, stride, fE, fI, hev_thresh);   \

  300. }

  301. 

  302. UV_LOOP_FILTER(v, 1, stride)

  303. UV_LOOP_FILTER(h, stride, 1)

  304. 

  305. static void vp8_v_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, int flim)

  306. {

  307.     int i;

  308. 

  309.     for (i = 0; i < 16; i++)

  310.         if (simple_limit(dst + i, stride, flim))

  311.             filter_common(dst + i, stride, 1);

  312. }

  313. 

  314. static void vp8_h_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, int flim)

  315. {

  316.     int i;

  317. 

  318.     for (i = 0; i < 16; i++)

  319.         if (simple_limit(dst + i * stride, 1, flim))

  320.             filter_common(dst + i * stride, 1, 1);

  321. }

  322. 

  323. static const uint8_t subpel_filters[7][6] = {

  324.     { 0,  6, 123,  12,  1, 0 },

  325.     { 2, 11, 108,  36,  8, 1 },

  326.     { 0,  9,  93,  50,  6, 0 },

  327.     { 3, 16,  77,  77, 16, 3 },

  328.     { 0,  6,  50,  93,  9, 0 },

  329.     { 1,  8,  36, 108, 11, 2 },

  330.     { 0,  1,  12, 123,  6, 0 },

  331. };

  332. 

  333. #define PUT_PIXELS(WIDTH)                                                     \

  334. static void put_vp8_pixels ## WIDTH ## _c(uint8_t *dst, ptrdiff_t dststride,  \

  335.                                           uint8_t *src, ptrdiff_t srcstride,  \

  336.                                           int h, int x, int y)                \

  337. {                                                                             \

  338.     int i;                                                                    \

  339.     for (i = 0; i < h; i++, dst += dststride, src += srcstride)               \

  340.         memcpy(dst, src, WIDTH);                                              \

  341. }

  342. 

  343. PUT_PIXELS(16)

  344. PUT_PIXELS(8)

  345. PUT_PIXELS(4)

  346. 

  347. #define FILTER_6TAP(src, F, stride)                                           \

  348.     cm[(F[2] * src[x + 0 * stride] - F[1] * src[x - 1 * stride] +             \

  349.         F[0] * src[x - 2 * stride] + F[3] * src[x + 1 * stride] -             \

  350.         F[4] * src[x + 2 * stride] + F[5] * src[x + 3 * stride] + 64) >> 7]

  351. 

  352. #define FILTER_4TAP(src, F, stride)                                           \

  353.     cm[(F[2] * src[x + 0 * stride] - F[1] * src[x - 1 * stride] +             \

  354.         F[3] * src[x + 1 * stride] - F[4] * src[x + 2 * stride] + 64) >> 7]

  355. 

  356. #define VP8_EPEL_H(SIZE, TAPS)                                                \

  357. static void put_vp8_epel ## SIZE ## _h ## TAPS ## _c(uint8_t *dst,            \

  358.                                                      ptrdiff_t dststride,     \

  359.                                                      uint8_t *src,            \

  360.                                                      ptrdiff_t srcstride,     \

  361.                                                      int h, int mx, int my)   \

  362. {                                                                             \

  363.     const uint8_t *filter = subpel_filters[mx - 1];                           \

  364.     const uint8_t *cm     = ff_crop_tab + MAX_NEG_CROP;                       \

  365.     int x, y;                                                                 \

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

  367.         for (x = 0; x < SIZE; x++)                                            \

  368.             dst[x] = FILTER_ ## TAPS ## TAP(src, filter, 1);                  \

  369.         dst += dststride;                                                     \

  370.         src += srcstride;                                                     \

  371.     }                                                                         \

  372. }

  373. 

  374. #define VP8_EPEL_V(SIZE, TAPS)                                                \

  375. static void put_vp8_epel ## SIZE ## _v ## TAPS ## _c(uint8_t *dst,            \

  376.                                                      ptrdiff_t dststride,     \

  377.                                                      uint8_t *src,            \

  378.                                                      ptrdiff_t srcstride,     \

  379.                                                      int h, int mx, int my)   \

  380. {                                                                             \

  381.     const uint8_t *filter = subpel_filters[my - 1];                           \

  382.     const uint8_t *cm     = ff_crop_tab + MAX_NEG_CROP;                       \

  383.     int x, y;                                                                 \

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

  385.         for (x = 0; x < SIZE; x++)                                            \

  386.             dst[x] = FILTER_ ## TAPS ## TAP(src, filter, srcstride);          \

  387.         dst += dststride;                                                     \

  388.         src += srcstride;                                                     \

  389.     }                                                                         \

  390. }

  391. 

  392. #define VP8_EPEL_HV(SIZE, HTAPS, VTAPS)                                       \

  393. static void                                                                   \

  394. put_vp8_epel ## SIZE ## _h ## HTAPS ## v ## VTAPS ## _c(uint8_t *dst,         \

  395.                                                         ptrdiff_t dststride,  \

  396.                                                         uint8_t *src,         \

  397.                                                         ptrdiff_t srcstride,  \

  398.                                                         int h, int mx,        \

  399.                                                         int my)               \

  400. {                                                                             \

  401.     const uint8_t *filter = subpel_filters[mx - 1];                           \

  402.     const uint8_t *cm     = ff_crop_tab + MAX_NEG_CROP;                       \

  403.     int x, y;                                                                 \

  404.     uint8_t tmp_array[(2 * SIZE + VTAPS - 1) * SIZE];                         \

  405.     uint8_t *tmp = tmp_array;                                                 \

  406.     src -= (2 - (VTAPS == 4)) * srcstride;                                    \

  407.                                                                               \

  408.     for (y = 0; y < h + VTAPS - 1; y++) {                                     \

  409.         for (x = 0; x < SIZE; x++)                                            \

  410.             tmp[x] = FILTER_ ## HTAPS ## TAP(src, filter, 1);                 \

  411.         tmp += SIZE;                                                          \

  412.         src += srcstride;                                                     \

  413.     }                                                                         \

  414.     tmp    = tmp_array + (2 - (VTAPS == 4)) * SIZE;                           \

  415.     filter = subpel_filters[my - 1];                                          \

  416.                                                                               \

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

  418.         for (x = 0; x < SIZE; x++)                                            \

  419.             dst[x] = FILTER_ ## VTAPS ## TAP(tmp, filter, SIZE);              \

  420.         dst += dststride;                                                     \

  421.         tmp += SIZE;                                                          \

  422.     }                                                                         \

  423. }

  424. 

  425. VP8_EPEL_H(16, 4)

  426. VP8_EPEL_H(8,  4)

  427. VP8_EPEL_H(4,  4)

  428. VP8_EPEL_H(16, 6)

  429. VP8_EPEL_H(8,  6)

  430. VP8_EPEL_H(4,  6)

  431. VP8_EPEL_V(16, 4)

  432. VP8_EPEL_V(8,  4)

  433. VP8_EPEL_V(4,  4)

  434. VP8_EPEL_V(16, 6)

  435. VP8_EPEL_V(8,  6)

  436. VP8_EPEL_V(4,  6)

  437. 

  438. VP8_EPEL_HV(16, 4, 4)

  439. VP8_EPEL_HV(8,  4, 4)

  440. VP8_EPEL_HV(4,  4, 4)

  441. VP8_EPEL_HV(16, 4, 6)

  442. VP8_EPEL_HV(8,  4, 6)

  443. VP8_EPEL_HV(4,  4, 6)

  444. VP8_EPEL_HV(16, 6, 4)

  445. VP8_EPEL_HV(8,  6, 4)

  446. VP8_EPEL_HV(4,  6, 4)

  447. VP8_EPEL_HV(16, 6, 6)

  448. VP8_EPEL_HV(8,  6, 6)

  449. VP8_EPEL_HV(4,  6, 6)

  450. 

  451. #define VP8_BILINEAR(SIZE)                                                    \

  452. static void put_vp8_bilinear ## SIZE ## _h_c(uint8_t *dst, ptrdiff_t dstride, \

  453.                                              uint8_t *src, ptrdiff_t sstride, \

  454.                                              int h, int mx, int my)           \

  455. {                                                                             \

  456.     int a = 8 - mx, b = mx;                                                   \

  457.     int x, y;                                                                 \

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

  459.         for (x = 0; x < SIZE; x++)                                            \

  460.             dst[x] = (a * src[x] + b * src[x + 1] + 4) >> 3;                  \

  461.         dst += dstride;                                                       \

  462.         src += sstride;                                                       \

  463.     }                                                                         \

  464. }                                                                             \

  465.                                                                               \

  466. static void put_vp8_bilinear ## SIZE ## _v_c(uint8_t *dst, ptrdiff_t dstride, \

  467.                                              uint8_t *src, ptrdiff_t sstride, \

  468.                                              int h, int mx, int my)           \

  469. {                                                                             \

  470.     int c = 8 - my, d = my;                                                   \

  471.     int x, y;                                                                 \

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

  473.         for (x = 0; x < SIZE; x++)                                            \

  474.             dst[x] = (c * src[x] + d * src[x + sstride] + 4) >> 3;            \

  475.         dst += dstride;                                                       \

  476.         src += sstride;                                                       \

  477.     }                                                                         \

  478. }                                                                             \

  479.                                                                               \

  480. static void put_vp8_bilinear ## SIZE ## _hv_c(uint8_t *dst,                   \

  481.                                               ptrdiff_t dstride,              \

  482.                                               uint8_t *src,                   \

  483.                                               ptrdiff_t sstride,              \

  484.                                               int h, int mx, int my)          \

  485. {                                                                             \

  486.     int a = 8 - mx, b = mx;                                                   \

  487.     int c = 8 - my, d = my;                                                   \

  488.     int x, y;                                                                 \

  489.     uint8_t tmp_array[(2 * SIZE + 1) * SIZE];                                 \

  490.     uint8_t *tmp = tmp_array;                                                 \

  491.     for (y = 0; y < h + 1; y++) {                                             \

  492.         for (x = 0; x < SIZE; x++)                                            \

  493.             tmp[x] = (a * src[x] + b * src[x + 1] + 4) >> 3;                  \

  494.         tmp += SIZE;                                                          \

  495.         src += sstride;                                                       \

  496.     }                                                                         \

  497.     tmp = tmp_array;                                                          \

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

  499.         for (x = 0; x < SIZE; x++)                                            \

  500.             dst[x] = (c * tmp[x] + d * tmp[x + SIZE] + 4) >> 3;               \

  501.         dst += dstride;                                                       \

  502.         tmp += SIZE;                                                          \

  503.     }                                                                         \

  504. }

  505. 

  506. VP8_BILINEAR(16)

  507. VP8_BILINEAR(8)

  508. VP8_BILINEAR(4)

  509. 

  510. #define VP8_MC_FUNC(IDX, SIZE)                                                \

  511.     dsp->put_vp8_epel_pixels_tab[IDX][0][0] = put_vp8_pixels ## SIZE ## _c;   \

  512.     dsp->put_vp8_epel_pixels_tab[IDX][0][1] = put_vp8_epel ## SIZE ## _h4_c;  \

  513.     dsp->put_vp8_epel_pixels_tab[IDX][0][2] = put_vp8_epel ## SIZE ## _h6_c;  \

  514.     dsp->put_vp8_epel_pixels_tab[IDX][1][0] = put_vp8_epel ## SIZE ## _v4_c;  \

  515.     dsp->put_vp8_epel_pixels_tab[IDX][1][1] = put_vp8_epel ## SIZE ## _h4v4_c; \

  516.     dsp->put_vp8_epel_pixels_tab[IDX][1][2] = put_vp8_epel ## SIZE ## _h6v4_c; \

  517.     dsp->put_vp8_epel_pixels_tab[IDX][2][0] = put_vp8_epel ## SIZE ## _v6_c;  \

  518.     dsp->put_vp8_epel_pixels_tab[IDX][2][1] = put_vp8_epel ## SIZE ## _h4v6_c; \

  519.     dsp->put_vp8_epel_pixels_tab[IDX][2][2] = put_vp8_epel ## SIZE ## _h6v6_c

  520. 

  521. #define VP8_BILINEAR_MC_FUNC(IDX, SIZE)                                       \

  522.     dsp->put_vp8_bilinear_pixels_tab[IDX][0][0] = put_vp8_pixels ## SIZE ## _c; \

  523.     dsp->put_vp8_bilinear_pixels_tab[IDX][0][1] = put_vp8_bilinear ## SIZE ## _h_c; \

  524.     dsp->put_vp8_bilinear_pixels_tab[IDX][0][2] = put_vp8_bilinear ## SIZE ## _h_c; \

  525.     dsp->put_vp8_bilinear_pixels_tab[IDX][1][0] = put_vp8_bilinear ## SIZE ## _v_c; \

  526.     dsp->put_vp8_bilinear_pixels_tab[IDX][1][1] = put_vp8_bilinear ## SIZE ## _hv_c; \

  527.     dsp->put_vp8_bilinear_pixels_tab[IDX][1][2] = put_vp8_bilinear ## SIZE ## _hv_c; \

  528.     dsp->put_vp8_bilinear_pixels_tab[IDX][2][0] = put_vp8_bilinear ## SIZE ## _v_c; \

  529.     dsp->put_vp8_bilinear_pixels_tab[IDX][2][1] = put_vp8_bilinear ## SIZE ## _hv_c; \

  530.     dsp->put_vp8_bilinear_pixels_tab[IDX][2][2] = put_vp8_bilinear ## SIZE ## _hv_c

  531. 

  532. av_cold void ff_vp8dsp_init(VP8DSPContext *dsp)

  533. {

  534.     dsp->vp8_luma_dc_wht    = vp8_luma_dc_wht_c;

  535.     dsp->vp8_luma_dc_wht_dc = vp8_luma_dc_wht_dc_c;

  536.     dsp->vp8_idct_add       = vp8_idct_add_c;

  537.     dsp->vp8_idct_dc_add    = vp8_idct_dc_add_c;

  538.     dsp->vp8_idct_dc_add4y  = vp8_idct_dc_add4y_c;

  539.     dsp->vp8_idct_dc_add4uv = vp8_idct_dc_add4uv_c;

  540. 

  541.     dsp->vp8_v_loop_filter16y = vp8_v_loop_filter16_c;

  542.     dsp->vp8_h_loop_filter16y = vp8_h_loop_filter16_c;

  543.     dsp->vp8_v_loop_filter8uv = vp8_v_loop_filter8uv_c;

  544.     dsp->vp8_h_loop_filter8uv = vp8_h_loop_filter8uv_c;

  545. 

  546.     dsp->vp8_v_loop_filter16y_inner = vp8_v_loop_filter16_inner_c;

  547.     dsp->vp8_h_loop_filter16y_inner = vp8_h_loop_filter16_inner_c;

  548.     dsp->vp8_v_loop_filter8uv_inner = vp8_v_loop_filter8uv_inner_c;

  549.     dsp->vp8_h_loop_filter8uv_inner = vp8_h_loop_filter8uv_inner_c;

  550. 

  551.     dsp->vp8_v_loop_filter_simple = vp8_v_loop_filter_simple_c;

  552.     dsp->vp8_h_loop_filter_simple = vp8_h_loop_filter_simple_c;

  553. 

  554.     VP8_MC_FUNC(0, 16);

  555.     VP8_MC_FUNC(1, 8);

  556.     VP8_MC_FUNC(2, 4);

  557. 

  558.     VP8_BILINEAR_MC_FUNC(0, 16);

  559.     VP8_BILINEAR_MC_FUNC(1, 8);

  560.     VP8_BILINEAR_MC_FUNC(2, 4);

  561. 

  562.     if (ARCH_ARM)

  563.         ff_vp8dsp_init_arm(dsp);

  564.     if (ARCH_PPC)

  565.         ff_vp8dsp_init_ppc(dsp);

  566.     if (ARCH_X86)

  567.         ff_vp8dsp_init_x86(dsp);

  568. }