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.
61705 Commits
32 Branches
1.2GB
Tree: 672c610d84
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '672c610d84'
${ noResults }
FFmpeg/libavcodec/vp8dsp.c

657 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 "mathops.h"

  29. #include "vp8dsp.h"

  30. #include "libavutil/common.h"

  31. #include "libavutil/intreadwrite.h"

  32. 

  33. #define MK_IDCT_DC_ADD4_C(name) \

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

  35. {\

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

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

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

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

  40. }\

  41. \

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

  43. {\

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

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

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

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

  48. }

  49. 

  50. #if CONFIG_VP7_DECODER

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

  52. {

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

  54.     int16_t tmp[16];

  55. 

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

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

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

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

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

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

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

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

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

  65.     }

  66. 

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

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

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

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

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

  72.         AV_ZERO64(dc + i*4);

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

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

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

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

  77.     }

  78. }

  79. 

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

  81. {

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

  83.     dc[0] = 0;

  84. 

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

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

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

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

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

  90.     }

  91. }

  92. 

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

  94. {

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

  96.     int16_t tmp[16];

  97. 

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

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

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

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

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

  103.         AV_ZERO64(block + i*4);

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

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

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

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

  108.     }

  109. 

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

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

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

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

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

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

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

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

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

  119.     }

  120. }

  121. 

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

  123. {

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

  125.     block[0] = 0;

  126. 

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

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

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

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

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

  132.         dst += stride;

  133.     }

  134. }

  135. 

  136. MK_IDCT_DC_ADD4_C(vp7)

  137. #endif

  138. 

  139. // TODO: Maybe add dequant

  140. #if CONFIG_VP8_DECODER

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

  142. {

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

  144. 

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

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

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

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

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

  150. 

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

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

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

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

  155.     }

  156. 

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

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

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

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

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

  162.         AV_ZERO64(dc + i*4);

  163. 

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

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

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

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

  168.     }

  169. }

  170. 

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

  172. {

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

  174.     dc[0] = 0;

  175. 

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

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

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

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

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

  181.     }

  182. }

  183. 

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

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

  186. 

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

  188. {

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

  190.     int16_t tmp[16];

  191. 

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

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

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

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

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

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

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

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

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

  201. 

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

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

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

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

  206.     }

  207. 

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

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

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

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

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

  213. 

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

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

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

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

  218.         dst += stride;

  219.     }

  220. }

  221. 

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

  223. {

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

  225.     block[0] = 0;

  226. 

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

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

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

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

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

  232.         dst += stride;

  233.     }

  234. }

  235. 

  236. MK_IDCT_DC_ADD4_C(vp8)

  237. #endif

  238. 

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

  240. #define LOAD_PIXELS\

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

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

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

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

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

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

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

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

  249. 

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

  251. 

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

  253. {

  254.     LOAD_PIXELS

  255.     int a, f1, f2;

  256.     const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;

  257. 

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

  259. 

  260.     if (is4tap)

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

  262. 

  263.     a = clip_int8(a);

  264. 

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

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

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

  268. 

  269.     if (vpn == 7)

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

  271.     else

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

  273. 

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

  275.     // be bitexact with libvpx.

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

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

  278. 

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

  280.     if (!is4tap) {

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

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

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

  284.     }

  285. }

  286. 

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

  288. {

  289.     LOAD_PIXELS

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

  291. }

  292. 

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

  294. {

  295.     LOAD_PIXELS

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

  297. }

  298. 

  299. /**

  300.  * E - limit at the macroblock edge

  301.  * I - limit for interior difference

  302.  */

  303. #define NORMAL_LIMIT(vpn) \

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

  305. {\

  306.     LOAD_PIXELS\

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

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

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

  310. }

  311. 

  312. NORMAL_LIMIT(7)

  313. NORMAL_LIMIT(8)

  314. 

  315. // high edge variance

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

  317. {

  318.     LOAD_PIXELS

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

  320. }

  321. 

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

  323. {

  324.     int a0, a1, a2, w;

  325.     const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;

  326. 

  327.     LOAD_PIXELS

  328. 

  329.     w = clip_int8(p1-q1);

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

  331. 

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

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

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

  335. 

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

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

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

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

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

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

  342. }

  343. 

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

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

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

  347. {\

  348.     int i;\

  349. \

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

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

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

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

  354.             else\

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

  356.         }\

  357. }\

  358. \

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

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

  361. {\

  362.     int i;\

  363. \

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

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

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

  367.             if (hv) \

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

  369.             else \

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

  371.         }\

  372. }

  373. 

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

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

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

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

  378. {\

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

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

  381. }\

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

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

  384. {\

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

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

  387. }

  388. 

  389. #define LOOP_FILTER_SIMPLE(vpn) \

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

  391. {\

  392.     int i;\

  393. \

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

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

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

  397. }\

  398. \

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

  400. {\

  401.     int i;\

  402. \

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

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

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

  406. }

  407. 

  408. #if CONFIG_VP7_DECODER

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

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

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

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

  413. LOOP_FILTER_SIMPLE(7)

  414. #endif

  415. 

  416. #if CONFIG_VP8_DECODER

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

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

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

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

  421. LOOP_FILTER_SIMPLE(8)

  422. #endif

  423. 

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

  425.     { 0,   6, 123,  12,   1,   0 },

  426.     { 2,  11, 108,  36,   8,   1 },

  427.     { 0,   9,  93,  50,   6,   0 },

  428.     { 3,  16,  77,  77,  16,   3 },

  429.     { 0,   6,  50,  93,   9,   0 },

  430.     { 1,   8,  36, 108,  11,   2 },

  431.     { 0,   1,  12, 123,   6,   0 },

  432. };

  433. 

  434. #define PUT_PIXELS(WIDTH) \

  435. 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) { \

  436.     int i; \

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

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

  439.     } \

  440. }

  441. 

  442. PUT_PIXELS(16)

  443. PUT_PIXELS(8)

  444. PUT_PIXELS(4)

  445. 

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

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

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

  449. 

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

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

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

  453. 

  454. #define VP8_EPEL_H(SIZE, TAPS) \

  455. 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) \

  456. { \

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

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

  459.     int x, y; \

  460. \

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

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

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

  464.         dst += dststride; \

  465.         src += srcstride; \

  466.     } \

  467. }

  468. #define VP8_EPEL_V(SIZE, TAPS) \

  469. 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) \

  470. { \

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

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

  473.     int x, y; \

  474. \

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

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

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

  478.         dst += dststride; \

  479.         src += srcstride; \

  480.     } \

  481. }

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

  483. 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) \

  484. { \

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

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

  487.     int x, y; \

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

  489.     uint8_t *tmp = tmp_array; \

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

  491. \

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

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

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

  495.         tmp += SIZE; \

  496.         src += srcstride; \

  497.     } \

  498. \

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

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

  501. \

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

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

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

  505.         dst += dststride; \

  506.         tmp += SIZE; \

  507.     } \

  508. }

  509. 

  510. VP8_EPEL_H(16, 4)

  511. VP8_EPEL_H(8,  4)

  512. VP8_EPEL_H(4,  4)

  513. VP8_EPEL_H(16, 6)

  514. VP8_EPEL_H(8,  6)

  515. VP8_EPEL_H(4,  6)

  516. VP8_EPEL_V(16, 4)

  517. VP8_EPEL_V(8,  4)

  518. VP8_EPEL_V(4,  4)

  519. VP8_EPEL_V(16, 6)

  520. VP8_EPEL_V(8,  6)

  521. VP8_EPEL_V(4,  6)

  522. VP8_EPEL_HV(16, 4, 4)

  523. VP8_EPEL_HV(8,  4, 4)

  524. VP8_EPEL_HV(4,  4, 4)

  525. VP8_EPEL_HV(16, 4, 6)

  526. VP8_EPEL_HV(8,  4, 6)

  527. VP8_EPEL_HV(4,  4, 6)

  528. VP8_EPEL_HV(16, 6, 4)

  529. VP8_EPEL_HV(8,  6, 4)

  530. VP8_EPEL_HV(4,  6, 4)

  531. VP8_EPEL_HV(16, 6, 6)

  532. VP8_EPEL_HV(8,  6, 6)

  533. VP8_EPEL_HV(4,  6, 6)

  534. 

  535. #define VP8_BILINEAR(SIZE) \

  536. 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) \

  537. { \

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

  539.     int x, y; \

  540. \

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

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

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

  544.         dst += dstride; \

  545.         src += sstride; \

  546.     } \

  547. } \

  548. 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) \

  549. { \

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

  551.     int x, y; \

  552. \

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

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

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

  556.         dst += dstride; \

  557.         src += sstride; \

  558.     } \

  559. } \

  560. \

  561. 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) \

  562. { \

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

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

  565.     int x, y; \

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

  567.     uint8_t *tmp = tmp_array; \

  568. \

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

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

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

  572.         tmp += SIZE; \

  573.         src += sstride; \

  574.     } \

  575. \

  576.     tmp = tmp_array; \

  577. \

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

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

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

  581.         dst += dstride; \

  582.         tmp += SIZE; \

  583.     } \

  584. }

  585. 

  586. VP8_BILINEAR(16)

  587. VP8_BILINEAR(8)

  588. VP8_BILINEAR(4)

  589. 

  590. #define VP8_MC_FUNC(IDX, SIZE) \

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

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

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

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

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

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

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

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

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

  600. 

  601. #define VP8_BILINEAR_MC_FUNC(IDX, SIZE) \

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

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

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

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

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

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

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

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

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

  611. 

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

  613. {

  614. #if CONFIG_VP7_DECODER && CONFIG_VP8_DECODER

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

  616. #elif CONFIG_VP7_DECODER

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

  618. #else // CONFIG_VP8_DECODER

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

  620. #endif

  621. 

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

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

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

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

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

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

  628. 

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

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

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

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

  633. 

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

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

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

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

  638. 

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

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

  641. 

  642.     VP8_MC_FUNC(0, 16);

  643.     VP8_MC_FUNC(1, 8);

  644.     VP8_MC_FUNC(2, 4);

  645. 

  646.     VP8_BILINEAR_MC_FUNC(0, 16);

  647.     VP8_BILINEAR_MC_FUNC(1, 8);

  648.     VP8_BILINEAR_MC_FUNC(2, 4);

  649. 

  650.     if (ARCH_ARM)

  651.         ff_vp8dsp_init_arm(dsp, vp7);

  652.     if (ARCH_PPC)

  653.         ff_vp8dsp_init_ppc(dsp);

  654.     if (ARCH_X86)

  655.         ff_vp8dsp_init_x86(dsp, vp7);

  656. }