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.
60901 Commits
32 Branches
340MB
Tree: bef6b27f10
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'bef6b27f10'
${ noResults }
FFmpeg/libavcodec/vp8dsp.c

656 lines
21KB
Raw Blame History 

  1. /*

  2.  * Copyright (C) 2010 David Conrad

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

  4.  * Copyright (C) 2014 Peter Ross

  5.  *

  6.  * This file is part of FFmpeg.

  7.  *

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

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

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

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

  12.  *

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

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

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

  16.  * Lesser General Public License for more details.

  17.  *

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

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

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

  21.  */

  22. 

  23. /**

  24.  * @file

  25.  * VP8 compatible video decoder

  26.  */

  27. 

  28. #include "dsputil.h"

  29. #include "vp8dsp.h"

  30. #include "libavutil/common.h"

  31. 

  32. #define MK_IDCT_DC_ADD4_C(name) \

  33. static void name ## _idct_dc_add4uv_c(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride)\

  34. {\

  35.     name ## _idct_dc_add_c(dst+stride*0+0, block[0], stride);\

  36.     name ## _idct_dc_add_c(dst+stride*0+4, block[1], stride);\

  37.     name ## _idct_dc_add_c(dst+stride*4+0, block[2], stride);\

  38.     name ## _idct_dc_add_c(dst+stride*4+4, block[3], stride);\

  39. }\

  40. \

  41. static void name ## _idct_dc_add4y_c(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride)\

  42. {\

  43.     name ## _idct_dc_add_c(dst+ 0, block[0], stride);\

  44.     name ## _idct_dc_add_c(dst+ 4, block[1], stride);\

  45.     name ## _idct_dc_add_c(dst+ 8, block[2], stride);\

  46.     name ## _idct_dc_add_c(dst+12, block[3], stride);\

  47. }

  48. 

  49. #if CONFIG_VP7_DECODER

  50. static void vp7_luma_dc_wht_c(int16_t block[4][4][16], int16_t dc[16])

  51. {

  52.     int i, a1, b1, c1, d1;

  53.     int16_t tmp[16];

  54. 

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

  56.         a1 = (dc[i*4+0] + dc[i*4+2]) * 23170;

  57.         b1 = (dc[i*4+0] - dc[i*4+2]) * 23170;

  58.         c1 = dc[i*4+1] * 12540 - dc[i*4+3] * 30274;

  59.         d1 = dc[i*4+1] * 30274 + dc[i*4+3] * 12540;

  60.         tmp[i*4+0] = (a1 + d1) >> 14;

  61.         tmp[i*4+3] = (a1 - d1) >> 14;

  62.         tmp[i*4+1] = (b1 + c1) >> 14;

  63.         tmp[i*4+2] = (b1 - c1) >> 14;

  64.     }

  65. 

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

  67.         a1 = (tmp[i + 0] + tmp[i + 8]) * 23170;

  68.         b1 = (tmp[i + 0] - tmp[i + 8]) * 23170;

  69.         c1 = tmp[i + 4] * 12540 - tmp[i + 12] * 30274;

  70.         d1 = tmp[i + 4] * 30274 + tmp[i + 12] * 12540;

  71.         AV_ZERO64(dc + i*4);

  72.         block[0][i][0] = (a1 + d1 + 0x20000) >> 18;

  73.         block[3][i][0] = (a1 - d1 + 0x20000) >> 18;

  74.         block[1][i][0] = (b1 + c1 + 0x20000) >> 18;

  75.         block[2][i][0] = (b1 - c1 + 0x20000) >> 18;

  76.     }

  77. }

  78. 

  79. static void vp7_luma_dc_wht_dc_c(int16_t block[4][4][16], int16_t dc[16])

  80. {

  81.     int i, val = (23170 * (23170 * dc[0] >> 14) + 0x20000) >> 18;

  82.     dc[0] = 0;

  83. 

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

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

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

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

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

  89.     }

  90. }

  91. 

  92. static void vp7_idct_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)

  93. {

  94.     int i, a1, b1, c1, d1;

  95.     int16_t tmp[16];

  96. 

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

  98.         a1 = (block[i*4+0] + block[i*4+2]) * 23170;

  99.         b1 = (block[i*4+0] - block[i*4+2]) * 23170;

  100.         c1 = block[i*4+1] * 12540 - block[i*4+3] * 30274;

  101.         d1 = block[i*4+1] * 30274 + block[i*4+3] * 12540;

  102.         AV_ZERO64(block + i*4);

  103.         tmp[i*4+0] = (a1 + d1) >> 14;

  104.         tmp[i*4+3] = (a1 - d1) >> 14;

  105.         tmp[i*4+1] = (b1 + c1) >> 14;

  106.         tmp[i*4+2] = (b1 - c1) >> 14;

  107.     }

  108. 

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

  110.         a1 = (tmp[i + 0] + tmp[i + 8]) * 23170;

  111.         b1 = (tmp[i + 0] - tmp[i + 8]) * 23170;

  112.         c1 = tmp[i + 4] * 12540 - tmp[i + 12] * 30274;

  113.         d1 = tmp[i + 4] * 30274 + tmp[i + 12] * 12540;

  114.         dst[0*stride+i] = av_clip_uint8(dst[0*stride+i] + ((a1 + d1 + 0x20000) >> 18));

  115.         dst[3*stride+i] = av_clip_uint8(dst[3*stride+i] + ((a1 - d1 + 0x20000) >> 18));

  116.         dst[1*stride+i] = av_clip_uint8(dst[1*stride+i] + ((b1 + c1 + 0x20000) >> 18));

  117.         dst[2*stride+i] = av_clip_uint8(dst[2*stride+i] + ((b1 - c1 + 0x20000) >> 18));

  118.     }

  119. }

  120. 

  121. static void vp7_idct_dc_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)

  122. {

  123.     int i, dc = (23170 * (23170 * block[0] >> 14) + 0x20000) >> 18;

  124.     block[0] = 0;

  125. 

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

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

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

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

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

  131.         dst += stride;

  132.     }

  133. }

  134. 

  135. MK_IDCT_DC_ADD4_C(vp7)

  136. #endif

  137. 

  138. // TODO: Maybe add dequant

  139. #if CONFIG_VP8_DECODER

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

  141. {

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

  143. 

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

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

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

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

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

  149. 

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

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

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

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

  154.     }

  155. 

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

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

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

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

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

  161.         AV_ZERO64(dc + i*4);

  162. 

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

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

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

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

  167.     }

  168. }

  169. 

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

  171. {

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

  173.     dc[0] = 0;

  174. 

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

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

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

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

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

  180.     }

  181. }

  182. 

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

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

  185. 

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

  187. {

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

  189.     int16_t tmp[16];

  190. 

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

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

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

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

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

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

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

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

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

  200. 

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

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

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

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

  205.     }

  206. 

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

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

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

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

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

  212. 

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

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

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

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

  217.         dst += stride;

  218.     }

  219. }

  220. 

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

  222. {

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

  224.     block[0] = 0;

  225. 

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

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

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

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

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

  231.         dst += stride;

  232.     }

  233. }

  234. 

  235. MK_IDCT_DC_ADD4_C(vp8)

  236. #endif

  237. 

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

  239. #define LOAD_PIXELS\

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

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

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

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

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

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

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

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

  248. 

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

  250. 

  251. static av_always_inline void filter_common(uint8_t *p, ptrdiff_t stride, int is4tap, int vpn)

  252. {

  253.     LOAD_PIXELS

  254.     int a, f1, f2;

  255.     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;

  256. 

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

  258. 

  259.     if (is4tap)

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

  261. 

  262.     a = clip_int8(a);

  263. 

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

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

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

  267. 

  268.     if (vpn == 7)

  269.         f2 = f1 - ((a & 7) == 4);

  270.     else

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

  272. 

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

  274.     // be bitexact with libvpx.

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

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

  277. 

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

  279.     if (!is4tap) {

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

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

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

  283.     }

  284. }

  285. 

  286. static av_always_inline int vp7_simple_limit(uint8_t *p, ptrdiff_t stride, int flim)

  287. {

  288.     LOAD_PIXELS

  289.     return FFABS(p0-q0) <= flim;

  290. }

  291. 

  292. static av_always_inline int vp8_simple_limit(uint8_t *p, ptrdiff_t stride, int flim)

  293. {

  294.     LOAD_PIXELS

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

  296. }

  297. 

  298. /**

  299.  * E - limit at the macroblock edge

  300.  * I - limit for interior difference

  301.  */

  302. #define NORMAL_LIMIT(vpn) \

  303. static av_always_inline int vp ## vpn ## _normal_limit(uint8_t *p, ptrdiff_t stride, int E, int I)\

  304. {\

  305.     LOAD_PIXELS\

  306.     return vp ## vpn ## _simple_limit(p, stride, E)\

  307.         && FFABS(p3-p2) <= I && FFABS(p2-p1) <= I && FFABS(p1-p0) <= I\

  308.         && FFABS(q3-q2) <= I && FFABS(q2-q1) <= I && FFABS(q1-q0) <= I;\

  309. }

  310. 

  311. NORMAL_LIMIT(7)

  312. NORMAL_LIMIT(8)

  313. 

  314. // high edge variance

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

  316. {

  317.     LOAD_PIXELS

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

  319. }

  320. 

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

  322. {

  323.     int a0, a1, a2, w;

  324.     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;

  325. 

  326.     LOAD_PIXELS

  327. 

  328.     w = clip_int8(p1-q1);

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

  330. 

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

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

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

  334. 

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

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

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

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

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

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

  341. }

  342. 

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

  344. static maybe_inline void vp ## vpn ## _ ## dir ## _loop_filter ## size ## _c(uint8_t *dst, ptrdiff_t stride,\

  345.                                      int flim_E, int flim_I, int hev_thresh)\

  346. {\

  347.     int i;\

  348. \

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

  350.         if (vp ## vpn ## _normal_limit(dst+i*stridea, strideb, flim_E, flim_I)) {\

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

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

  353.             else\

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

  355.         }\

  356. }\

  357. \

  358. static maybe_inline void vp ## vpn ## _ ## dir ## _loop_filter ## size ## _inner_c(uint8_t *dst, ptrdiff_t stride,\

  359.                                       int flim_E, int flim_I, int hev_thresh)\

  360. {\

  361.     int i;\

  362. \

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

  364.         if (vp ## vpn ## _normal_limit(dst+i*stridea, strideb, flim_E, flim_I)) {\

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

  366.             if (hv) \

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

  368.             else \

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

  370.         }\

  371. }

  372. 

  373. #define UV_LOOP_FILTER(vpn, dir, stridea, strideb) \

  374. LOOP_FILTER(vpn, dir, 8, stridea, strideb, av_always_inline) \

  375. static void vp ## vpn ## _ ## dir ## _loop_filter8uv_c(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,\

  376.                                       int fE, int fI, int hev_thresh)\

  377. {\

  378.   vp ## vpn ## _ ## dir ## _loop_filter8_c(dstU, stride, fE, fI, hev_thresh);\

  379.   vp ## vpn ## _ ## dir ## _loop_filter8_c(dstV, stride, fE, fI, hev_thresh);\

  380. }\

  381. static void vp ## vpn ## _ ## dir ## _loop_filter8uv_inner_c(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,\

  382.                                       int fE, int fI, int hev_thresh)\

  383. {\

  384.   vp ## vpn ## _ ## dir ## _loop_filter8_inner_c(dstU, stride, fE, fI, hev_thresh);\

  385.   vp ## vpn ## _ ## dir ## _loop_filter8_inner_c(dstV, stride, fE, fI, hev_thresh);\

  386. }

  387. 

  388. #define LOOP_FILTER_SIMPLE(vpn) \

  389. static void vp ## vpn ## _v_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, int flim)\

  390. {\

  391.     int i;\

  392. \

  393.     for (i = 0; i < 16; i++)\

  394.         if (vp ## vpn ## _simple_limit(dst+i, stride, flim))\

  395.             filter_common(dst+i, stride, 1, vpn);\

  396. }\

  397. \

  398. static void vp ## vpn ## _h_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, int flim)\

  399. {\

  400.     int i;\

  401. \

  402.     for (i = 0; i < 16; i++)\

  403.         if (vp ## vpn ## _simple_limit(dst+i*stride, 1, flim))\

  404.             filter_common(dst+i*stride, 1, 1, vpn);\

  405. }

  406. 

  407. #if CONFIG_VP7_DECODER

  408. LOOP_FILTER(7, v, 16, 1, stride,)

  409. LOOP_FILTER(7, h, 16, stride, 1,)

  410. UV_LOOP_FILTER(7, v, 1, stride)

  411. UV_LOOP_FILTER(7, h, stride, 1)

  412. LOOP_FILTER_SIMPLE(7)

  413. #endif

  414. 

  415. #if CONFIG_VP8_DECODER

  416. LOOP_FILTER(8, v, 16, 1, stride,)

  417. LOOP_FILTER(8, h, 16, stride, 1,)

  418. UV_LOOP_FILTER(8, v, 1, stride)

  419. UV_LOOP_FILTER(8, h, stride, 1)

  420. LOOP_FILTER_SIMPLE(8)

  421. #endif

  422. 

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

  424.     { 0,   6, 123,  12,   1,   0 },

  425.     { 2,  11, 108,  36,   8,   1 },

  426.     { 0,   9,  93,  50,   6,   0 },

  427.     { 3,  16,  77,  77,  16,   3 },

  428.     { 0,   6,  50,  93,   9,   0 },

  429.     { 1,   8,  36, 108,  11,   2 },

  430.     { 0,   1,  12, 123,   6,   0 },

  431. };

  432. 

  433. #define PUT_PIXELS(WIDTH) \

  434. static void put_vp8_pixels ## WIDTH ##_c(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, ptrdiff_t srcstride, int h, int x, int y) { \

  435.     int i; \

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

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

  438.     } \

  439. }

  440. 

  441. PUT_PIXELS(16)

  442. PUT_PIXELS(8)

  443. PUT_PIXELS(4)

  444. 

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

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

  447.         F[3]*src[x+1*stride] - F[4]*src[x+2*stride] + F[5]*src[x+3*stride] + 64) >> 7]

  448. 

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

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

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

  452. 

  453. #define VP8_EPEL_H(SIZE, TAPS) \

  454. static void put_vp8_epel ## SIZE ## _h ## TAPS ## _c(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my) \

  455. { \

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

  457.     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; \

  458.     int x, y; \

  459. \

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

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

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

  463.         dst += dststride; \

  464.         src += srcstride; \

  465.     } \

  466. }

  467. #define VP8_EPEL_V(SIZE, TAPS) \

  468. static void put_vp8_epel ## SIZE ## _v ## TAPS ## _c(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my) \

  469. { \

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

  471.     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; \

  472.     int x, y; \

  473. \

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

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

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

  477.         dst += dststride; \

  478.         src += srcstride; \

  479.     } \

  480. }

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

  482. static void put_vp8_epel ## SIZE ## _h ## HTAPS ## v ## VTAPS ## _c(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my) \

  483. { \

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

  485.     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; \

  486.     int x, y; \

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

  488.     uint8_t *tmp = tmp_array; \

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

  490. \

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

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

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

  494.         tmp += SIZE; \

  495.         src += srcstride; \

  496.     } \

  497. \

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

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

  500. \

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

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

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

  504.         dst += dststride; \

  505.         tmp += SIZE; \

  506.     } \

  507. }

  508. 

  509. VP8_EPEL_H(16, 4)

  510. VP8_EPEL_H(8,  4)

  511. VP8_EPEL_H(4,  4)

  512. VP8_EPEL_H(16, 6)

  513. VP8_EPEL_H(8,  6)

  514. VP8_EPEL_H(4,  6)

  515. VP8_EPEL_V(16, 4)

  516. VP8_EPEL_V(8,  4)

  517. VP8_EPEL_V(4,  4)

  518. VP8_EPEL_V(16, 6)

  519. VP8_EPEL_V(8,  6)

  520. VP8_EPEL_V(4,  6)

  521. VP8_EPEL_HV(16, 4, 4)

  522. VP8_EPEL_HV(8,  4, 4)

  523. VP8_EPEL_HV(4,  4, 4)

  524. VP8_EPEL_HV(16, 4, 6)

  525. VP8_EPEL_HV(8,  4, 6)

  526. VP8_EPEL_HV(4,  4, 6)

  527. VP8_EPEL_HV(16, 6, 4)

  528. VP8_EPEL_HV(8,  6, 4)

  529. VP8_EPEL_HV(4,  6, 4)

  530. VP8_EPEL_HV(16, 6, 6)

  531. VP8_EPEL_HV(8,  6, 6)

  532. VP8_EPEL_HV(4,  6, 6)

  533. 

  534. #define VP8_BILINEAR(SIZE) \

  535. static void put_vp8_bilinear ## SIZE ## _h_c(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, ptrdiff_t sstride, int h, int mx, int my) \

  536. { \

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

  538.     int x, y; \

  539. \

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

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

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

  543.         dst += dstride; \

  544.         src += sstride; \

  545.     } \

  546. } \

  547. static void put_vp8_bilinear ## SIZE ## _v_c(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, ptrdiff_t sstride, int h, int mx, int my) \

  548. { \

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

  550.     int x, y; \

  551. \

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

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

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

  555.         dst += dstride; \

  556.         src += sstride; \

  557.     } \

  558. } \

  559. \

  560. static void put_vp8_bilinear ## SIZE ## _hv_c(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, ptrdiff_t sstride, int h, int mx, int my) \

  561. { \

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

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

  564.     int x, y; \

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

  566.     uint8_t *tmp = tmp_array; \

  567. \

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

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

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

  571.         tmp += SIZE; \

  572.         src += sstride; \

  573.     } \

  574. \

  575.     tmp = tmp_array; \

  576. \

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

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

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

  580.         dst += dstride; \

  581.         tmp += SIZE; \

  582.     } \

  583. }

  584. 

  585. VP8_BILINEAR(16)

  586. VP8_BILINEAR(8)

  587. VP8_BILINEAR(4)

  588. 

  589. #define VP8_MC_FUNC(IDX, SIZE) \

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

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

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

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

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

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

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

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

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

  599. 

  600. #define VP8_BILINEAR_MC_FUNC(IDX, SIZE) \

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

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

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

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

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

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

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

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

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

  610. 

  611. av_cold void ff_vp8dsp_init(VP8DSPContext *dsp, int vp7)

  612. {

  613. #if CONFIG_VP7_DECODER && CONFIG_VP8_DECODER

  614. #define VPX(f) vp7 ? vp7_ ## f : vp8_ ## f

  615. #elif CONFIG_VP7_DECODER

  616. #define VPX(f) vp7_ ## f

  617. #else // CONFIG_VP8_DECODER

  618. #define VPX(f) vp8_ ## f

  619. #endif

  620. 

  621.     dsp->vp8_luma_dc_wht    = VPX(luma_dc_wht_c);

  622.     dsp->vp8_luma_dc_wht_dc = VPX(luma_dc_wht_dc_c);

  623.     dsp->vp8_idct_add       = VPX(idct_add_c);

  624.     dsp->vp8_idct_dc_add    = VPX(idct_dc_add_c);

  625.     dsp->vp8_idct_dc_add4y  = VPX(idct_dc_add4y_c);

  626.     dsp->vp8_idct_dc_add4uv = VPX(idct_dc_add4uv_c);

  627. 

  628.     dsp->vp8_v_loop_filter16y = VPX(v_loop_filter16_c);

  629.     dsp->vp8_h_loop_filter16y = VPX(h_loop_filter16_c);

  630.     dsp->vp8_v_loop_filter8uv = VPX(v_loop_filter8uv_c);

  631.     dsp->vp8_h_loop_filter8uv = VPX(h_loop_filter8uv_c);

  632. 

  633.     dsp->vp8_v_loop_filter16y_inner = VPX(v_loop_filter16_inner_c);

  634.     dsp->vp8_h_loop_filter16y_inner = VPX(h_loop_filter16_inner_c);

  635.     dsp->vp8_v_loop_filter8uv_inner = VPX(v_loop_filter8uv_inner_c);

  636.     dsp->vp8_h_loop_filter8uv_inner = VPX(h_loop_filter8uv_inner_c);

  637. 

  638.     dsp->vp8_v_loop_filter_simple = VPX(v_loop_filter_simple_c);

  639.     dsp->vp8_h_loop_filter_simple = VPX(h_loop_filter_simple_c);

  640. 

  641.     VP8_MC_FUNC(0, 16);

  642.     VP8_MC_FUNC(1, 8);

  643.     VP8_MC_FUNC(2, 4);

  644. 

  645.     VP8_BILINEAR_MC_FUNC(0, 16);

  646.     VP8_BILINEAR_MC_FUNC(1, 8);

  647.     VP8_BILINEAR_MC_FUNC(2, 4);

  648. 

  649.     if (ARCH_ARM)

  650.         ff_vp8dsp_init_arm(dsp, vp7);

  651.     if (ARCH_PPC)

  652.         ff_vp8dsp_init_ppc(dsp);

  653.     if (ARCH_X86)

  654.         ff_vp8dsp_init_x86(dsp, vp7);

  655. }