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.
50872 Commits
32 Branches
834MB
Tree: db4e4f766c
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'db4e4f766c'
${ noResults }
FFmpeg/libavcodec/vc1dsp.c

875 lines
29KB
Raw Blame History 

  1. /*

  2.  * VC-1 and WMV3 decoder - DSP functions

  3.  * Copyright (c) 2006 Konstantin Shishkov

  4.  *

  5.  * This file is part of FFmpeg.

  6.  *

  7.  * FFmpeg 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.  * FFmpeg 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 FFmpeg; 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.  * VC-1 and WMV3 decoder

  25.  *

  26.  */

  27. 

  28. #include "libavutil/avassert.h"

  29. #include "libavutil/common.h"

  30. #include "h264chroma.h"

  31. #include "rnd_avg.h"

  32. #include "vc1dsp.h"

  33. 

  34. 

  35. /** Apply overlap transform to horizontal edge

  36. */

  37. static void vc1_v_overlap_c(uint8_t* src, int stride)

  38. {

  39.     int i;

  40.     int a, b, c, d;

  41.     int d1, d2;

  42.     int rnd = 1;

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

  44.         a = src[-2*stride];

  45.         b = src[-stride];

  46.         c = src[0];

  47.         d = src[stride];

  48.         d1 = (a - d + 3 + rnd) >> 3;

  49.         d2 = (a - d + b - c + 4 - rnd) >> 3;

  50. 

  51.         src[-2*stride] = a - d1;

  52.         src[-stride] = av_clip_uint8(b - d2);

  53.         src[0] = av_clip_uint8(c + d2);

  54.         src[stride] = d + d1;

  55.         src++;

  56.         rnd = !rnd;

  57.     }

  58. }

  59. 

  60. /** Apply overlap transform to vertical edge

  61. */

  62. static void vc1_h_overlap_c(uint8_t* src, int stride)

  63. {

  64.     int i;

  65.     int a, b, c, d;

  66.     int d1, d2;

  67.     int rnd = 1;

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

  69.         a = src[-2];

  70.         b = src[-1];

  71.         c = src[0];

  72.         d = src[1];

  73.         d1 = (a - d + 3 + rnd) >> 3;

  74.         d2 = (a - d + b - c + 4 - rnd) >> 3;

  75. 

  76.         src[-2] = a - d1;

  77.         src[-1] = av_clip_uint8(b - d2);

  78.         src[0] = av_clip_uint8(c + d2);

  79.         src[1] = d + d1;

  80.         src += stride;

  81.         rnd = !rnd;

  82.     }

  83. }

  84. 

  85. static void vc1_v_s_overlap_c(int16_t *top,  int16_t *bottom)

  86. {

  87.     int i;

  88.     int a, b, c, d;

  89.     int d1, d2;

  90.     int rnd1 = 4, rnd2 = 3;

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

  92.         a = top[48];

  93.         b = top[56];

  94.         c = bottom[0];

  95.         d = bottom[8];

  96.         d1 = a - d;

  97.         d2 = a - d + b - c;

  98. 

  99.         top[48]   = ((a << 3) - d1 + rnd1) >> 3;

  100.         top[56]   = ((b << 3) - d2 + rnd2) >> 3;

  101.         bottom[0] = ((c << 3) + d2 + rnd1) >> 3;

  102.         bottom[8] = ((d << 3) + d1 + rnd2) >> 3;

  103. 

  104.         bottom++;

  105.         top++;

  106.         rnd2 = 7 - rnd2;

  107.         rnd1 = 7 - rnd1;

  108.     }

  109. }

  110. 

  111. static void vc1_h_s_overlap_c(int16_t *left, int16_t *right)

  112. {

  113.     int i;

  114.     int a, b, c, d;

  115.     int d1, d2;

  116.     int rnd1 = 4, rnd2 = 3;

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

  118.         a = left[6];

  119.         b = left[7];

  120.         c = right[0];

  121.         d = right[1];

  122.         d1 = a - d;

  123.         d2 = a - d + b - c;

  124. 

  125.         left[6]  = ((a << 3) - d1 + rnd1) >> 3;

  126.         left[7]  = ((b << 3) - d2 + rnd2) >> 3;

  127.         right[0] = ((c << 3) + d2 + rnd1) >> 3;

  128.         right[1] = ((d << 3) + d1 + rnd2) >> 3;

  129. 

  130.         right += 8;

  131.         left += 8;

  132.         rnd2 = 7 - rnd2;

  133.         rnd1 = 7 - rnd1;

  134.     }

  135. }

  136. 

  137. /**

  138.  * VC-1 in-loop deblocking filter for one line

  139.  * @param src source block type

  140.  * @param stride block stride

  141.  * @param pq block quantizer

  142.  * @return whether other 3 pairs should be filtered or not

  143.  * @see 8.6

  144.  */

  145. static av_always_inline int vc1_filter_line(uint8_t* src, int stride, int pq){

  146.     int a0 = (2*(src[-2*stride] - src[ 1*stride]) - 5*(src[-1*stride] - src[ 0*stride]) + 4) >> 3;

  147.     int a0_sign = a0 >> 31;        /* Store sign */

  148.     a0 = (a0 ^ a0_sign) - a0_sign; /* a0 = FFABS(a0); */

  149.     if(a0 < pq){

  150.         int a1 = FFABS((2*(src[-4*stride] - src[-1*stride]) - 5*(src[-3*stride] - src[-2*stride]) + 4) >> 3);

  151.         int a2 = FFABS((2*(src[ 0*stride] - src[ 3*stride]) - 5*(src[ 1*stride] - src[ 2*stride]) + 4) >> 3);

  152.         if(a1 < a0 || a2 < a0){

  153.             int clip = src[-1*stride] - src[ 0*stride];

  154.             int clip_sign = clip >> 31;

  155.             clip = ((clip ^ clip_sign) - clip_sign)>>1;

  156.             if(clip){

  157.                 int a3 = FFMIN(a1, a2);

  158.                 int d = 5 * (a3 - a0);

  159.                 int d_sign = (d >> 31);

  160.                 d = ((d ^ d_sign) - d_sign) >> 3;

  161.                 d_sign ^= a0_sign;

  162. 

  163.                 if( d_sign ^ clip_sign )

  164.                     d = 0;

  165.                 else{

  166.                     d = FFMIN(d, clip);

  167.                     d = (d ^ d_sign) - d_sign;          /* Restore sign */

  168.                     src[-1*stride] = av_clip_uint8(src[-1*stride] - d);

  169.                     src[ 0*stride] = av_clip_uint8(src[ 0*stride] + d);

  170.                 }

  171.                 return 1;

  172.             }

  173.         }

  174.     }

  175.     return 0;

  176. }

  177. 

  178. /**

  179.  * VC-1 in-loop deblocking filter

  180.  * @param src source block type

  181.  * @param step distance between horizontally adjacent elements

  182.  * @param stride distance between vertically adjacent elements

  183.  * @param len edge length to filter (4 or 8 pixels)

  184.  * @param pq block quantizer

  185.  * @see 8.6

  186.  */

  187. static inline void vc1_loop_filter(uint8_t* src, int step, int stride, int len, int pq)

  188. {

  189.     int i;

  190.     int filt3;

  191. 

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

  193.         filt3 = vc1_filter_line(src + 2*step, stride, pq);

  194.         if(filt3){

  195.             vc1_filter_line(src + 0*step, stride, pq);

  196.             vc1_filter_line(src + 1*step, stride, pq);

  197.             vc1_filter_line(src + 3*step, stride, pq);

  198.         }

  199.         src += step * 4;

  200.     }

  201. }

  202. 

  203. static void vc1_v_loop_filter4_c(uint8_t *src, int stride, int pq)

  204. {

  205.     vc1_loop_filter(src, 1, stride, 4, pq);

  206. }

  207. 

  208. static void vc1_h_loop_filter4_c(uint8_t *src, int stride, int pq)

  209. {

  210.     vc1_loop_filter(src, stride, 1, 4, pq);

  211. }

  212. 

  213. static void vc1_v_loop_filter8_c(uint8_t *src, int stride, int pq)

  214. {

  215.     vc1_loop_filter(src, 1, stride, 8, pq);

  216. }

  217. 

  218. static void vc1_h_loop_filter8_c(uint8_t *src, int stride, int pq)

  219. {

  220.     vc1_loop_filter(src, stride, 1, 8, pq);

  221. }

  222. 

  223. static void vc1_v_loop_filter16_c(uint8_t *src, int stride, int pq)

  224. {

  225.     vc1_loop_filter(src, 1, stride, 16, pq);

  226. }

  227. 

  228. static void vc1_h_loop_filter16_c(uint8_t *src, int stride, int pq)

  229. {

  230.     vc1_loop_filter(src, stride, 1, 16, pq);

  231. }

  232. 

  233. /** Do inverse transform on 8x8 block

  234. */

  235. static void vc1_inv_trans_8x8_dc_c(uint8_t *dest, int linesize, int16_t *block)

  236. {

  237.     int i;

  238.     int dc = block[0];

  239.     dc = (3 * dc +  1) >> 1;

  240.     dc = (3 * dc + 16) >> 5;

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

  242.         dest[0] = av_clip_uint8(dest[0] + dc);

  243.         dest[1] = av_clip_uint8(dest[1] + dc);

  244.         dest[2] = av_clip_uint8(dest[2] + dc);

  245.         dest[3] = av_clip_uint8(dest[3] + dc);

  246.         dest[4] = av_clip_uint8(dest[4] + dc);

  247.         dest[5] = av_clip_uint8(dest[5] + dc);

  248.         dest[6] = av_clip_uint8(dest[6] + dc);

  249.         dest[7] = av_clip_uint8(dest[7] + dc);

  250.         dest += linesize;

  251.     }

  252. }

  253. 

  254. static void vc1_inv_trans_8x8_c(int16_t block[64])

  255. {

  256.     int i;

  257.     register int t1,t2,t3,t4,t5,t6,t7,t8;

  258.     int16_t *src, *dst, temp[64];

  259. 

  260.     src = block;

  261.     dst = temp;

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

  263.         t1 = 12 * (src[ 0] + src[32]) + 4;

  264.         t2 = 12 * (src[ 0] - src[32]) + 4;

  265.         t3 = 16 * src[16] +  6 * src[48];

  266.         t4 =  6 * src[16] - 16 * src[48];

  267. 

  268.         t5 = t1 + t3;

  269.         t6 = t2 + t4;

  270.         t7 = t2 - t4;

  271.         t8 = t1 - t3;

  272. 

  273.         t1 = 16 * src[ 8] + 15 * src[24] +  9 * src[40] +  4 * src[56];

  274.         t2 = 15 * src[ 8] -  4 * src[24] - 16 * src[40] -  9 * src[56];

  275.         t3 =  9 * src[ 8] - 16 * src[24] +  4 * src[40] + 15 * src[56];

  276.         t4 =  4 * src[ 8] -  9 * src[24] + 15 * src[40] - 16 * src[56];

  277. 

  278.         dst[0] = (t5 + t1) >> 3;

  279.         dst[1] = (t6 + t2) >> 3;

  280.         dst[2] = (t7 + t3) >> 3;

  281.         dst[3] = (t8 + t4) >> 3;

  282.         dst[4] = (t8 - t4) >> 3;

  283.         dst[5] = (t7 - t3) >> 3;

  284.         dst[6] = (t6 - t2) >> 3;

  285.         dst[7] = (t5 - t1) >> 3;

  286. 

  287.         src += 1;

  288.         dst += 8;

  289.     }

  290. 

  291.     src = temp;

  292.     dst = block;

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

  294.         t1 = 12 * (src[ 0] + src[32]) + 64;

  295.         t2 = 12 * (src[ 0] - src[32]) + 64;

  296.         t3 = 16 * src[16] +  6 * src[48];

  297.         t4 =  6 * src[16] - 16 * src[48];

  298. 

  299.         t5 = t1 + t3;

  300.         t6 = t2 + t4;

  301.         t7 = t2 - t4;

  302.         t8 = t1 - t3;

  303. 

  304.         t1 = 16 * src[ 8] + 15 * src[24] +  9 * src[40] +  4 * src[56];

  305.         t2 = 15 * src[ 8] -  4 * src[24] - 16 * src[40] -  9 * src[56];

  306.         t3 =  9 * src[ 8] - 16 * src[24] +  4 * src[40] + 15 * src[56];

  307.         t4 =  4 * src[ 8] -  9 * src[24] + 15 * src[40] - 16 * src[56];

  308. 

  309.         dst[ 0] = (t5 + t1) >> 7;

  310.         dst[ 8] = (t6 + t2) >> 7;

  311.         dst[16] = (t7 + t3) >> 7;

  312.         dst[24] = (t8 + t4) >> 7;

  313.         dst[32] = (t8 - t4 + 1) >> 7;

  314.         dst[40] = (t7 - t3 + 1) >> 7;

  315.         dst[48] = (t6 - t2 + 1) >> 7;

  316.         dst[56] = (t5 - t1 + 1) >> 7;

  317. 

  318.         src++;

  319.         dst++;

  320.     }

  321. }

  322. 

  323. /** Do inverse transform on 8x4 part of block

  324. */

  325. static void vc1_inv_trans_8x4_dc_c(uint8_t *dest, int linesize, int16_t *block)

  326. {

  327.     int i;

  328.     int dc = block[0];

  329.     dc = ( 3 * dc +  1) >> 1;

  330.     dc = (17 * dc + 64) >> 7;

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

  332.         dest[0] = av_clip_uint8(dest[0] + dc);

  333.         dest[1] = av_clip_uint8(dest[1] + dc);

  334.         dest[2] = av_clip_uint8(dest[2] + dc);

  335.         dest[3] = av_clip_uint8(dest[3] + dc);

  336.         dest[4] = av_clip_uint8(dest[4] + dc);

  337.         dest[5] = av_clip_uint8(dest[5] + dc);

  338.         dest[6] = av_clip_uint8(dest[6] + dc);

  339.         dest[7] = av_clip_uint8(dest[7] + dc);

  340.         dest += linesize;

  341.     }

  342. }

  343. 

  344. static void vc1_inv_trans_8x4_c(uint8_t *dest, int linesize, int16_t *block)

  345. {

  346.     int i;

  347.     register int t1,t2,t3,t4,t5,t6,t7,t8;

  348.     int16_t *src, *dst;

  349. 

  350.     src = block;

  351.     dst = block;

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

  353.         t1 = 12 * (src[0] + src[4]) + 4;

  354.         t2 = 12 * (src[0] - src[4]) + 4;

  355.         t3 = 16 * src[2] +  6 * src[6];

  356.         t4 =  6 * src[2] - 16 * src[6];

  357. 

  358.         t5 = t1 + t3;

  359.         t6 = t2 + t4;

  360.         t7 = t2 - t4;

  361.         t8 = t1 - t3;

  362. 

  363.         t1 = 16 * src[1] + 15 * src[3] +  9 * src[5] +  4 * src[7];

  364.         t2 = 15 * src[1] -  4 * src[3] - 16 * src[5] -  9 * src[7];

  365.         t3 =  9 * src[1] - 16 * src[3] +  4 * src[5] + 15 * src[7];

  366.         t4 =  4 * src[1] -  9 * src[3] + 15 * src[5] - 16 * src[7];

  367. 

  368.         dst[0] = (t5 + t1) >> 3;

  369.         dst[1] = (t6 + t2) >> 3;

  370.         dst[2] = (t7 + t3) >> 3;

  371.         dst[3] = (t8 + t4) >> 3;

  372.         dst[4] = (t8 - t4) >> 3;

  373.         dst[5] = (t7 - t3) >> 3;

  374.         dst[6] = (t6 - t2) >> 3;

  375.         dst[7] = (t5 - t1) >> 3;

  376. 

  377.         src += 8;

  378.         dst += 8;

  379.     }

  380. 

  381.     src = block;

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

  383.         t1 = 17 * (src[ 0] + src[16]) + 64;

  384.         t2 = 17 * (src[ 0] - src[16]) + 64;

  385.         t3 = 22 * src[ 8] + 10 * src[24];

  386.         t4 = 22 * src[24] - 10 * src[ 8];

  387. 

  388.         dest[0*linesize] = av_clip_uint8(dest[0*linesize] + ((t1 + t3) >> 7));

  389.         dest[1*linesize] = av_clip_uint8(dest[1*linesize] + ((t2 - t4) >> 7));

  390.         dest[2*linesize] = av_clip_uint8(dest[2*linesize] + ((t2 + t4) >> 7));

  391.         dest[3*linesize] = av_clip_uint8(dest[3*linesize] + ((t1 - t3) >> 7));

  392. 

  393.         src ++;

  394.         dest++;

  395.     }

  396. }

  397. 

  398. /** Do inverse transform on 4x8 parts of block

  399. */

  400. static void vc1_inv_trans_4x8_dc_c(uint8_t *dest, int linesize, int16_t *block)

  401. {

  402.     int i;

  403.     int dc = block[0];

  404.     dc = (17 * dc +  4) >> 3;

  405.     dc = (12 * dc + 64) >> 7;

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

  407.         dest[0] = av_clip_uint8(dest[0] + dc);

  408.         dest[1] = av_clip_uint8(dest[1] + dc);

  409.         dest[2] = av_clip_uint8(dest[2] + dc);

  410.         dest[3] = av_clip_uint8(dest[3] + dc);

  411.         dest += linesize;

  412.     }

  413. }

  414. 

  415. static void vc1_inv_trans_4x8_c(uint8_t *dest, int linesize, int16_t *block)

  416. {

  417.     int i;

  418.     register int t1,t2,t3,t4,t5,t6,t7,t8;

  419.     int16_t *src, *dst;

  420. 

  421.     src = block;

  422.     dst = block;

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

  424.         t1 = 17 * (src[0] + src[2]) + 4;

  425.         t2 = 17 * (src[0] - src[2]) + 4;

  426.         t3 = 22 * src[1] + 10 * src[3];

  427.         t4 = 22 * src[3] - 10 * src[1];

  428. 

  429.         dst[0] = (t1 + t3) >> 3;

  430.         dst[1] = (t2 - t4) >> 3;

  431.         dst[2] = (t2 + t4) >> 3;

  432.         dst[3] = (t1 - t3) >> 3;

  433. 

  434.         src += 8;

  435.         dst += 8;

  436.     }

  437. 

  438.     src = block;

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

  440.         t1 = 12 * (src[ 0] + src[32]) + 64;

  441.         t2 = 12 * (src[ 0] - src[32]) + 64;

  442.         t3 = 16 * src[16] +  6 * src[48];

  443.         t4 =  6 * src[16] - 16 * src[48];

  444. 

  445.         t5 = t1 + t3;

  446.         t6 = t2 + t4;

  447.         t7 = t2 - t4;

  448.         t8 = t1 - t3;

  449. 

  450.         t1 = 16 * src[ 8] + 15 * src[24] +  9 * src[40] +  4 * src[56];

  451.         t2 = 15 * src[ 8] -  4 * src[24] - 16 * src[40] -  9 * src[56];

  452.         t3 =  9 * src[ 8] - 16 * src[24] +  4 * src[40] + 15 * src[56];

  453.         t4 =  4 * src[ 8] -  9 * src[24] + 15 * src[40] - 16 * src[56];

  454. 

  455.         dest[0*linesize] = av_clip_uint8(dest[0*linesize] + ((t5 + t1) >> 7));

  456.         dest[1*linesize] = av_clip_uint8(dest[1*linesize] + ((t6 + t2) >> 7));

  457.         dest[2*linesize] = av_clip_uint8(dest[2*linesize] + ((t7 + t3) >> 7));

  458.         dest[3*linesize] = av_clip_uint8(dest[3*linesize] + ((t8 + t4) >> 7));

  459.         dest[4*linesize] = av_clip_uint8(dest[4*linesize] + ((t8 - t4 + 1) >> 7));

  460.         dest[5*linesize] = av_clip_uint8(dest[5*linesize] + ((t7 - t3 + 1) >> 7));

  461.         dest[6*linesize] = av_clip_uint8(dest[6*linesize] + ((t6 - t2 + 1) >> 7));

  462.         dest[7*linesize] = av_clip_uint8(dest[7*linesize] + ((t5 - t1 + 1) >> 7));

  463. 

  464.         src ++;

  465.         dest++;

  466.     }

  467. }

  468. 

  469. /** Do inverse transform on 4x4 part of block

  470. */

  471. static void vc1_inv_trans_4x4_dc_c(uint8_t *dest, int linesize, int16_t *block)

  472. {

  473.     int i;

  474.     int dc = block[0];

  475.     dc = (17 * dc +  4) >> 3;

  476.     dc = (17 * dc + 64) >> 7;

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

  478.         dest[0] = av_clip_uint8(dest[0] + dc);

  479.         dest[1] = av_clip_uint8(dest[1] + dc);

  480.         dest[2] = av_clip_uint8(dest[2] + dc);

  481.         dest[3] = av_clip_uint8(dest[3] + dc);

  482.         dest += linesize;

  483.     }

  484. }

  485. 

  486. static void vc1_inv_trans_4x4_c(uint8_t *dest, int linesize, int16_t *block)

  487. {

  488.     int i;

  489.     register int t1,t2,t3,t4;

  490.     int16_t *src, *dst;

  491. 

  492.     src = block;

  493.     dst = block;

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

  495.         t1 = 17 * (src[0] + src[2]) + 4;

  496.         t2 = 17 * (src[0] - src[2]) + 4;

  497.         t3 = 22 * src[1] + 10 * src[3];

  498.         t4 = 22 * src[3] - 10 * src[1];

  499. 

  500.         dst[0] = (t1 + t3) >> 3;

  501.         dst[1] = (t2 - t4) >> 3;

  502.         dst[2] = (t2 + t4) >> 3;

  503.         dst[3] = (t1 - t3) >> 3;

  504. 

  505.         src += 8;

  506.         dst += 8;

  507.     }

  508. 

  509.     src = block;

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

  511.         t1 = 17 * (src[ 0] + src[16]) + 64;

  512.         t2 = 17 * (src[ 0] - src[16]) + 64;

  513.         t3 = 22 * src[ 8] + 10 * src[24];

  514.         t4 = 22 * src[24] - 10 * src[ 8];

  515. 

  516.         dest[0*linesize] = av_clip_uint8(dest[0*linesize] + ((t1 + t3) >> 7));

  517.         dest[1*linesize] = av_clip_uint8(dest[1*linesize] + ((t2 - t4) >> 7));

  518.         dest[2*linesize] = av_clip_uint8(dest[2*linesize] + ((t2 + t4) >> 7));

  519.         dest[3*linesize] = av_clip_uint8(dest[3*linesize] + ((t1 - t3) >> 7));

  520. 

  521.         src ++;

  522.         dest++;

  523.     }

  524. }

  525. 

  526. /* motion compensation functions */

  527. /** Filter in case of 2 filters */

  528. #define VC1_MSPEL_FILTER_16B(DIR, TYPE)                                 \

  529. static av_always_inline int vc1_mspel_ ## DIR ## _filter_16bits(const TYPE *src, int stride, int mode) \

  530. {                                                                       \

  531.     switch(mode){                                                       \

  532.     case 0: /* no shift - should not occur */                           \

  533.         return 0;                                                       \

  534.     case 1: /* 1/4 shift */                                             \

  535.         return -4*src[-stride] + 53*src[0] + 18*src[stride] - 3*src[stride*2]; \

  536.     case 2: /* 1/2 shift */                                             \

  537.         return -src[-stride] + 9*src[0] + 9*src[stride] - src[stride*2]; \

  538.     case 3: /* 3/4 shift */                                             \

  539.         return -3*src[-stride] + 18*src[0] + 53*src[stride] - 4*src[stride*2]; \

  540.     }                                                                   \

  541.     return 0; /* should not occur */                                    \

  542. }

  543. 

  544. VC1_MSPEL_FILTER_16B(ver, uint8_t)

  545. VC1_MSPEL_FILTER_16B(hor, int16_t)

  546. 

  547. 

  548. /** Filter used to interpolate fractional pel values

  549.  */

  550. static av_always_inline int vc1_mspel_filter(const uint8_t *src, int stride, int mode, int r)

  551. {

  552.     switch(mode){

  553.     case 0: //no shift

  554.         return src[0];

  555.     case 1: // 1/4 shift

  556.         return (-4*src[-stride] + 53*src[0] + 18*src[stride] - 3*src[stride*2] + 32 - r) >> 6;

  557.     case 2: // 1/2 shift

  558.         return (-src[-stride] + 9*src[0] + 9*src[stride] - src[stride*2] + 8 - r) >> 4;

  559.     case 3: // 3/4 shift

  560.         return (-3*src[-stride] + 18*src[0] + 53*src[stride] - 4*src[stride*2] + 32 - r) >> 6;

  561.     }

  562.     return 0; //should not occur

  563. }

  564. 

  565. /** Function used to do motion compensation with bicubic interpolation

  566.  */

  567. #define VC1_MSPEL_MC(OP, OP4, OPNAME)\

  568. static av_always_inline void OPNAME ## vc1_mspel_mc(uint8_t *dst, const uint8_t *src, int stride, int hmode, int vmode, int rnd)\

  569. {\

  570.     int     i, j;\

  571. \

  572.     if (vmode) { /* Horizontal filter to apply */\

  573.         int r;\

  574. \

  575.         if (hmode) { /* Vertical filter to apply, output to tmp */\

  576.             static const int shift_value[] = { 0, 5, 1, 5 };\

  577.             int              shift = (shift_value[hmode]+shift_value[vmode])>>1;\

  578.             int16_t          tmp[11*8], *tptr = tmp;\

  579. \

  580.             r = (1<<(shift-1)) + rnd-1;\

  581. \

  582.             src -= 1;\

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

  584.                 for(i = 0; i < 11; i++)\

  585.                     tptr[i] = (vc1_mspel_ver_filter_16bits(src + i, stride, vmode)+r)>>shift;\

  586.                 src += stride;\

  587.                 tptr += 11;\

  588.             }\

  589. \

  590.             r = 64-rnd;\

  591.             tptr = tmp+1;\

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

  593.                 for(i = 0; i < 8; i++)\

  594.                     OP(dst[i], (vc1_mspel_hor_filter_16bits(tptr + i, 1, hmode)+r)>>7);\

  595.                 dst += stride;\

  596.                 tptr += 11;\

  597.             }\

  598. \

  599.             return;\

  600.         }\

  601.         else { /* No horizontal filter, output 8 lines to dst */\

  602.             r = 1-rnd;\

  603. \

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

  605.                 for(i = 0; i < 8; i++)\

  606.                     OP(dst[i], vc1_mspel_filter(src + i, stride, vmode, r));\

  607.                 src += stride;\

  608.                 dst += stride;\

  609.             }\

  610.             return;\

  611.         }\

  612.     }\

  613. \

  614.     /* Horizontal mode with no vertical mode */\

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

  616.         for(i = 0; i < 8; i++)\

  617.             OP(dst[i], vc1_mspel_filter(src + i, 1, hmode, rnd));\

  618.         dst += stride;\

  619.         src += stride;\

  620.     }\

  621. }\

  622. static void OPNAME ## pixels8x8_c(uint8_t *block, const uint8_t *pixels, int line_size, int rnd){\

  623.     int i;\

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

  625.         OP4(*(uint32_t*)(block  ), AV_RN32(pixels  ));\

  626.         OP4(*(uint32_t*)(block+4), AV_RN32(pixels+4));\

  627.         pixels+=line_size;\

  628.         block +=line_size;\

  629.     }\

  630. }

  631. 

  632. #define op_put(a, b) a = av_clip_uint8(b)

  633. #define op_avg(a, b) a = (a + av_clip_uint8(b) + 1) >> 1

  634. #define op4_avg(a, b) a = rnd_avg32(a, b)

  635. #define op4_put(a, b) a = b

  636. 

  637. VC1_MSPEL_MC(op_put, op4_put, put_)

  638. VC1_MSPEL_MC(op_avg, op4_avg, avg_)

  639. 

  640. /* pixel functions - really are entry points to vc1_mspel_mc */

  641. 

  642. #define PUT_VC1_MSPEL(a, b)\

  643. static void put_vc1_mspel_mc ## a ## b ##_c(uint8_t *dst,               \

  644.                                             const uint8_t *src,         \

  645.                                             ptrdiff_t stride, int rnd)  \

  646. {                                                                       \

  647.     put_vc1_mspel_mc(dst, src, stride, a, b, rnd);                      \

  648. }                                                                       \

  649. static void avg_vc1_mspel_mc ## a ## b ##_c(uint8_t *dst,               \

  650.                                             const uint8_t *src,         \

  651.                                             ptrdiff_t stride, int rnd)  \

  652. {                                                                       \

  653.     avg_vc1_mspel_mc(dst, src, stride, a, b, rnd);                      \

  654. }

  655. 

  656. PUT_VC1_MSPEL(1, 0)

  657. PUT_VC1_MSPEL(2, 0)

  658. PUT_VC1_MSPEL(3, 0)

  659. 

  660. PUT_VC1_MSPEL(0, 1)

  661. PUT_VC1_MSPEL(1, 1)

  662. PUT_VC1_MSPEL(2, 1)

  663. PUT_VC1_MSPEL(3, 1)

  664. 

  665. PUT_VC1_MSPEL(0, 2)

  666. PUT_VC1_MSPEL(1, 2)

  667. PUT_VC1_MSPEL(2, 2)

  668. PUT_VC1_MSPEL(3, 2)

  669. 

  670. PUT_VC1_MSPEL(0, 3)

  671. PUT_VC1_MSPEL(1, 3)

  672. PUT_VC1_MSPEL(2, 3)

  673. PUT_VC1_MSPEL(3, 3)

  674. 

  675. static void put_no_rnd_vc1_chroma_mc8_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){

  676.     const int A=(8-x)*(8-y);

  677.     const int B=(  x)*(8-y);

  678.     const int C=(8-x)*(  y);

  679.     const int D=(  x)*(  y);

  680.     int i;

  681. 

  682.     av_assert2(x<8 && y<8 && x>=0 && y>=0);

  683. 

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

  685.     {

  686.         dst[0] = (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + 32 - 4) >> 6;

  687.         dst[1] = (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + 32 - 4) >> 6;

  688.         dst[2] = (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + 32 - 4) >> 6;

  689.         dst[3] = (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + 32 - 4) >> 6;

  690.         dst[4] = (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + 32 - 4) >> 6;

  691.         dst[5] = (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + 32 - 4) >> 6;

  692.         dst[6] = (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + 32 - 4) >> 6;

  693.         dst[7] = (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + 32 - 4) >> 6;

  694.         dst+= stride;

  695.         src+= stride;

  696.     }

  697. }

  698. 

  699. static void put_no_rnd_vc1_chroma_mc4_c(uint8_t *dst, uint8_t *src, int stride, int h, int x, int y){

  700.     const int A=(8-x)*(8-y);

  701.     const int B=(  x)*(8-y);

  702.     const int C=(8-x)*(  y);

  703.     const int D=(  x)*(  y);

  704.     int i;

  705. 

  706.     av_assert2(x<8 && y<8 && x>=0 && y>=0);

  707. 

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

  709.     {

  710.         dst[0] = (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + 32 - 4) >> 6;

  711.         dst[1] = (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + 32 - 4) >> 6;

  712.         dst[2] = (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + 32 - 4) >> 6;

  713.         dst[3] = (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + 32 - 4) >> 6;

  714.         dst+= stride;

  715.         src+= stride;

  716.     }

  717. }

  718. 

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

  720. static void avg_no_rnd_vc1_chroma_mc8_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){

  721.     const int A=(8-x)*(8-y);

  722.     const int B=(  x)*(8-y);

  723.     const int C=(8-x)*(  y);

  724.     const int D=(  x)*(  y);

  725.     int i;

  726. 

  727.     av_assert2(x<8 && y<8 && x>=0 && y>=0);

  728. 

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

  730.     {

  731.         dst[0] = avg2(dst[0], ((A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + 32 - 4) >> 6));

  732.         dst[1] = avg2(dst[1], ((A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + 32 - 4) >> 6));

  733.         dst[2] = avg2(dst[2], ((A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + 32 - 4) >> 6));

  734.         dst[3] = avg2(dst[3], ((A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + 32 - 4) >> 6));

  735.         dst[4] = avg2(dst[4], ((A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + 32 - 4) >> 6));

  736.         dst[5] = avg2(dst[5], ((A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + 32 - 4) >> 6));

  737.         dst[6] = avg2(dst[6], ((A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + 32 - 4) >> 6));

  738.         dst[7] = avg2(dst[7], ((A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + 32 - 4) >> 6));

  739.         dst+= stride;

  740.         src+= stride;

  741.     }

  742. }

  743. 

  744. #if CONFIG_WMV3IMAGE_DECODER || CONFIG_VC1IMAGE_DECODER

  745. 

  746. static void sprite_h_c(uint8_t *dst, const uint8_t *src, int offset, int advance, int count)

  747. {

  748.     while (count--) {

  749.         int a = src[(offset >> 16)    ];

  750.         int b = src[(offset >> 16) + 1];

  751.         *dst++ = a + ((b - a) * (offset&0xFFFF) >> 16);

  752.         offset += advance;

  753.     }

  754. }

  755. 

  756. static av_always_inline void sprite_v_template(uint8_t *dst, const uint8_t *src1a, const uint8_t *src1b, int offset1,

  757.                                             int two_sprites, const uint8_t *src2a, const uint8_t *src2b, int offset2,

  758.                                             int alpha, int scaled, int width)

  759. {

  760.     int a1, b1, a2, b2;

  761.     while (width--) {

  762.         a1 = *src1a++;

  763.         if (scaled) {

  764.             b1 = *src1b++;

  765.             a1 = a1 + ((b1 - a1) * offset1 >> 16);

  766.         }

  767.         if (two_sprites) {

  768.             a2 = *src2a++;

  769.             if (scaled > 1) {

  770.                 b2 = *src2b++;

  771.                 a2 = a2 + ((b2 - a2) * offset2 >> 16);

  772.             }

  773.             a1 = a1 + ((a2 - a1) * alpha >> 16);

  774.         }

  775.         *dst++ = a1;

  776.     }

  777. }

  778. 

  779. static void sprite_v_single_c(uint8_t *dst, const uint8_t *src1a, const uint8_t *src1b, int offset, int width)

  780. {

  781.     sprite_v_template(dst, src1a, src1b, offset, 0, NULL, NULL, 0, 0, 1, width);

  782. }

  783. 

  784. static void sprite_v_double_noscale_c(uint8_t *dst, const uint8_t *src1a, const uint8_t *src2a, int alpha, int width)

  785. {

  786.     sprite_v_template(dst, src1a, NULL, 0, 1, src2a, NULL, 0, alpha, 0, width);

  787. }

  788. 

  789. static void sprite_v_double_onescale_c(uint8_t *dst, const uint8_t *src1a, const uint8_t *src1b, int offset1,

  790.                                                      const uint8_t *src2a, int alpha, int width)

  791. {

  792.     sprite_v_template(dst, src1a, src1b, offset1, 1, src2a, NULL, 0, alpha, 1, width);

  793. }

  794. 

  795. static void sprite_v_double_twoscale_c(uint8_t *dst, const uint8_t *src1a, const uint8_t *src1b, int offset1,

  796.                                                      const uint8_t *src2a, const uint8_t *src2b, int offset2,

  797.                                        int alpha, int width)

  798. {

  799.     sprite_v_template(dst, src1a, src1b, offset1, 1, src2a, src2b, offset2, alpha, 2, width);

  800. }

  801. 

  802. #endif

  803. 

  804. av_cold void ff_vc1dsp_init(VC1DSPContext* dsp) {

  805.     dsp->vc1_inv_trans_8x8 = vc1_inv_trans_8x8_c;

  806.     dsp->vc1_inv_trans_4x8 = vc1_inv_trans_4x8_c;

  807.     dsp->vc1_inv_trans_8x4 = vc1_inv_trans_8x4_c;

  808.     dsp->vc1_inv_trans_4x4 = vc1_inv_trans_4x4_c;

  809.     dsp->vc1_inv_trans_8x8_dc = vc1_inv_trans_8x8_dc_c;

  810.     dsp->vc1_inv_trans_4x8_dc = vc1_inv_trans_4x8_dc_c;

  811.     dsp->vc1_inv_trans_8x4_dc = vc1_inv_trans_8x4_dc_c;

  812.     dsp->vc1_inv_trans_4x4_dc = vc1_inv_trans_4x4_dc_c;

  813.     dsp->vc1_h_overlap = vc1_h_overlap_c;

  814.     dsp->vc1_v_overlap = vc1_v_overlap_c;

  815.     dsp->vc1_h_s_overlap = vc1_h_s_overlap_c;

  816.     dsp->vc1_v_s_overlap = vc1_v_s_overlap_c;

  817.     dsp->vc1_v_loop_filter4 = vc1_v_loop_filter4_c;

  818.     dsp->vc1_h_loop_filter4 = vc1_h_loop_filter4_c;

  819.     dsp->vc1_v_loop_filter8 = vc1_v_loop_filter8_c;

  820.     dsp->vc1_h_loop_filter8 = vc1_h_loop_filter8_c;

  821.     dsp->vc1_v_loop_filter16 = vc1_v_loop_filter16_c;

  822.     dsp->vc1_h_loop_filter16 = vc1_h_loop_filter16_c;

  823. 

  824.     dsp->put_vc1_mspel_pixels_tab[ 0] = put_pixels8x8_c;

  825.     dsp->put_vc1_mspel_pixels_tab[ 1] = put_vc1_mspel_mc10_c;

  826.     dsp->put_vc1_mspel_pixels_tab[ 2] = put_vc1_mspel_mc20_c;

  827.     dsp->put_vc1_mspel_pixels_tab[ 3] = put_vc1_mspel_mc30_c;

  828.     dsp->put_vc1_mspel_pixels_tab[ 4] = put_vc1_mspel_mc01_c;

  829.     dsp->put_vc1_mspel_pixels_tab[ 5] = put_vc1_mspel_mc11_c;

  830.     dsp->put_vc1_mspel_pixels_tab[ 6] = put_vc1_mspel_mc21_c;

  831.     dsp->put_vc1_mspel_pixels_tab[ 7] = put_vc1_mspel_mc31_c;

  832.     dsp->put_vc1_mspel_pixels_tab[ 8] = put_vc1_mspel_mc02_c;

  833.     dsp->put_vc1_mspel_pixels_tab[ 9] = put_vc1_mspel_mc12_c;

  834.     dsp->put_vc1_mspel_pixels_tab[10] = put_vc1_mspel_mc22_c;

  835.     dsp->put_vc1_mspel_pixels_tab[11] = put_vc1_mspel_mc32_c;

  836.     dsp->put_vc1_mspel_pixels_tab[12] = put_vc1_mspel_mc03_c;

  837.     dsp->put_vc1_mspel_pixels_tab[13] = put_vc1_mspel_mc13_c;

  838.     dsp->put_vc1_mspel_pixels_tab[14] = put_vc1_mspel_mc23_c;

  839.     dsp->put_vc1_mspel_pixels_tab[15] = put_vc1_mspel_mc33_c;

  840. 

  841.     dsp->avg_vc1_mspel_pixels_tab[ 0] = avg_pixels8x8_c;

  842.     dsp->avg_vc1_mspel_pixels_tab[ 1] = avg_vc1_mspel_mc10_c;

  843.     dsp->avg_vc1_mspel_pixels_tab[ 2] = avg_vc1_mspel_mc20_c;

  844.     dsp->avg_vc1_mspel_pixels_tab[ 3] = avg_vc1_mspel_mc30_c;

  845.     dsp->avg_vc1_mspel_pixels_tab[ 4] = avg_vc1_mspel_mc01_c;

  846.     dsp->avg_vc1_mspel_pixels_tab[ 5] = avg_vc1_mspel_mc11_c;

  847.     dsp->avg_vc1_mspel_pixels_tab[ 6] = avg_vc1_mspel_mc21_c;

  848.     dsp->avg_vc1_mspel_pixels_tab[ 7] = avg_vc1_mspel_mc31_c;

  849.     dsp->avg_vc1_mspel_pixels_tab[ 8] = avg_vc1_mspel_mc02_c;

  850.     dsp->avg_vc1_mspel_pixels_tab[ 9] = avg_vc1_mspel_mc12_c;

  851.     dsp->avg_vc1_mspel_pixels_tab[10] = avg_vc1_mspel_mc22_c;

  852.     dsp->avg_vc1_mspel_pixels_tab[11] = avg_vc1_mspel_mc32_c;

  853.     dsp->avg_vc1_mspel_pixels_tab[12] = avg_vc1_mspel_mc03_c;

  854.     dsp->avg_vc1_mspel_pixels_tab[13] = avg_vc1_mspel_mc13_c;

  855.     dsp->avg_vc1_mspel_pixels_tab[14] = avg_vc1_mspel_mc23_c;

  856.     dsp->avg_vc1_mspel_pixels_tab[15] = avg_vc1_mspel_mc33_c;

  857. 

  858.     dsp->put_no_rnd_vc1_chroma_pixels_tab[0]= put_no_rnd_vc1_chroma_mc8_c;

  859.     dsp->avg_no_rnd_vc1_chroma_pixels_tab[0]= avg_no_rnd_vc1_chroma_mc8_c;

  860.     dsp->put_no_rnd_vc1_chroma_pixels_tab[1] = put_no_rnd_vc1_chroma_mc4_c;

  861. 

  862. #if CONFIG_WMV3IMAGE_DECODER || CONFIG_VC1IMAGE_DECODER

  863.     dsp->sprite_h = sprite_h_c;

  864.     dsp->sprite_v_single = sprite_v_single_c;

  865.     dsp->sprite_v_double_noscale = sprite_v_double_noscale_c;

  866.     dsp->sprite_v_double_onescale = sprite_v_double_onescale_c;

  867.     dsp->sprite_v_double_twoscale = sprite_v_double_twoscale_c;

  868. #endif

  869. 

  870.     if (HAVE_ALTIVEC)

  871.         ff_vc1dsp_init_altivec(dsp);

  872.     if (ARCH_X86)

  873.         ff_vc1dsp_init_x86(dsp);

  874. }