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.
49698 Commits
32 Branches
916MB
Tree: a00c4b4db0
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'a00c4b4db0'
${ noResults }
FFmpeg/libavcodec/vc1dsp.c

868 lines
28KB
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 "vc1dsp.h"

  32. 

  33. 

  34. /** Apply overlap transform to horizontal edge

  35. */

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

  37. {

  38.     int i;

  39.     int a, b, c, d;

  40.     int d1, d2;

  41.     int rnd = 1;

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

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

  44.         b = src[-stride];

  45.         c = src[0];

  46.         d = src[stride];

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

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

  49. 

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

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

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

  53.         src[stride] = d + d1;

  54.         src++;

  55.         rnd = !rnd;

  56.     }

  57. }

  58. 

  59. /** Apply overlap transform to vertical edge

  60. */

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

  62. {

  63.     int i;

  64.     int a, b, c, d;

  65.     int d1, d2;

  66.     int rnd = 1;

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

  68.         a = src[-2];

  69.         b = src[-1];

  70.         c = src[0];

  71.         d = src[1];

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

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

  74. 

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

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

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

  78.         src[1] = d + d1;

  79.         src += stride;

  80.         rnd = !rnd;

  81.     }

  82. }

  83. 

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

  85. {

  86.     int i;

  87.     int a, b, c, d;

  88.     int d1, d2;

  89.     int rnd1 = 4, rnd2 = 3;

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

  91.         a = top[48];

  92.         b = top[56];

  93.         c = bottom[0];

  94.         d = bottom[8];

  95.         d1 = a - d;

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

  97. 

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

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

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

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

  102. 

  103.         bottom++;

  104.         top++;

  105.         rnd2 = 7 - rnd2;

  106.         rnd1 = 7 - rnd1;

  107.     }

  108. }

  109. 

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

  111. {

  112.     int i;

  113.     int a, b, c, d;

  114.     int d1, d2;

  115.     int rnd1 = 4, rnd2 = 3;

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

  117.         a = left[6];

  118.         b = left[7];

  119.         c = right[0];

  120.         d = right[1];

  121.         d1 = a - d;

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

  123. 

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

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

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

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

  128. 

  129.         right += 8;

  130.         left += 8;

  131.         rnd2 = 7 - rnd2;

  132.         rnd1 = 7 - rnd1;

  133.     }

  134. }

  135. 

  136. /**

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

  138.  * @param src source block type

  139.  * @param stride block stride

  140.  * @param pq block quantizer

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

  142.  * @see 8.6

  143.  */

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

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

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

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

  148.     if(a0 < pq){

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

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

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

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

  153.             int clip_sign = clip >> 31;

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

  155.             if(clip){

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

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

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

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

  160.                 d_sign ^= a0_sign;

  161. 

  162.                 if( d_sign ^ clip_sign )

  163.                     d = 0;

  164.                 else{

  165.                     d = FFMIN(d, clip);

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

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

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

  169.                 }

  170.                 return 1;

  171.             }

  172.         }

  173.     }

  174.     return 0;

  175. }

  176. 

  177. /**

  178.  * VC-1 in-loop deblocking filter

  179.  * @param src source block type

  180.  * @param step distance between horizontally adjacent elements

  181.  * @param stride distance between vertically adjacent elements

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

  183.  * @param pq block quantizer

  184.  * @see 8.6

  185.  */

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

  187. {

  188.     int i;

  189.     int filt3;

  190. 

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

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

  193.         if(filt3){

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

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

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

  197.         }

  198.         src += step * 4;

  199.     }

  200. }

  201. 

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

  203. {

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

  205. }

  206. 

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

  208. {

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

  210. }

  211. 

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

  213. {

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

  215. }

  216. 

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

  218. {

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

  220. }

  221. 

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

  223. {

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

  225. }

  226. 

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

  228. {

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

  230. }

  231. 

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

  233. */

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

  235. {

  236.     int i;

  237.     int dc = block[0];

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

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

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

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

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

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

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

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

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

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

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

  249.         dest += linesize;

  250.     }

  251. }

  252. 

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

  254. {

  255.     int i;

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

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

  258. 

  259.     src = block;

  260.     dst = temp;

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

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

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

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

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

  266. 

  267.         t5 = t1 + t3;

  268.         t6 = t2 + t4;

  269.         t7 = t2 - t4;

  270.         t8 = t1 - t3;

  271. 

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

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

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

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

  276. 

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

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

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

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

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

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

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

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

  285. 

  286.         src += 1;

  287.         dst += 8;

  288.     }

  289. 

  290.     src = temp;

  291.     dst = block;

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

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

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

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

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

  297. 

  298.         t5 = t1 + t3;

  299.         t6 = t2 + t4;

  300.         t7 = t2 - t4;

  301.         t8 = t1 - t3;

  302. 

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

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

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

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

  307. 

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

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

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

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

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

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

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

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

  316. 

  317.         src++;

  318.         dst++;

  319.     }

  320. }

  321. 

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

  323. */

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

  325. {

  326.     int i;

  327.     int dc = block[0];

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

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

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

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

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

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

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

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

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

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

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

  339.         dest += linesize;

  340.     }

  341. }

  342. 

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

  344. {

  345.     int i;

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

  347.     int16_t *src, *dst;

  348. 

  349.     src = block;

  350.     dst = block;

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

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

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

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

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

  356. 

  357.         t5 = t1 + t3;

  358.         t6 = t2 + t4;

  359.         t7 = t2 - t4;

  360.         t8 = t1 - t3;

  361. 

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

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

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

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

  366. 

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

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

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

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

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

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

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

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

  375. 

  376.         src += 8;

  377.         dst += 8;

  378.     }

  379. 

  380.     src = block;

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

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

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

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

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

  386. 

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

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

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

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

  391. 

  392.         src ++;

  393.         dest++;

  394.     }

  395. }

  396. 

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

  398. */

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

  400. {

  401.     int i;

  402.     int dc = block[0];

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

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

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

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

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

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

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

  410.         dest += linesize;

  411.     }

  412. }

  413. 

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

  415. {

  416.     int i;

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

  418.     int16_t *src, *dst;

  419. 

  420.     src = block;

  421.     dst = block;

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

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

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

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

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

  427. 

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

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

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

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

  432. 

  433.         src += 8;

  434.         dst += 8;

  435.     }

  436. 

  437.     src = block;

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

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

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

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

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

  443. 

  444.         t5 = t1 + t3;

  445.         t6 = t2 + t4;

  446.         t7 = t2 - t4;

  447.         t8 = t1 - t3;

  448. 

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

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

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

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

  453. 

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

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

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

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

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

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

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

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

  462. 

  463.         src ++;

  464.         dest++;

  465.     }

  466. }

  467. 

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

  469. */

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

  471. {

  472.     int i;

  473.     int dc = block[0];

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

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

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

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

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

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

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

  481.         dest += linesize;

  482.     }

  483. }

  484. 

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

  486. {

  487.     int i;

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

  489.     int16_t *src, *dst;

  490. 

  491.     src = block;

  492.     dst = block;

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

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

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

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

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

  498. 

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

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

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

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

  503. 

  504.         src += 8;

  505.         dst += 8;

  506.     }

  507. 

  508.     src = block;

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

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

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

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

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

  514. 

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

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

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

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

  519. 

  520.         src ++;

  521.         dest++;

  522.     }

  523. }

  524. 

  525. /* motion compensation functions */

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

  527. #define VC1_MSPEL_FILTER_16B(DIR, TYPE)                                 \

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

  529. {                                                                       \

  530.     switch(mode){                                                       \

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

  532.         return 0;                                                       \

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

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

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

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

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

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

  539.     }                                                                   \

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

  541. }

  542. 

  543. VC1_MSPEL_FILTER_16B(ver, uint8_t)

  544. VC1_MSPEL_FILTER_16B(hor, int16_t)

  545. 

  546. 

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

  548.  */

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

  550. {

  551.     switch(mode){

  552.     case 0: //no shift

  553.         return src[0];

  554.     case 1: // 1/4 shift

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

  556.     case 2: // 1/2 shift

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

  558.     case 3: // 3/4 shift

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

  560.     }

  561.     return 0; //should not occur

  562. }

  563. 

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

  565.  */

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

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

  568. {\

  569.     int     i, j;\

  570. \

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

  572.         int r;\

  573. \

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

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

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

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

  578. \

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

  580. \

  581.             src -= 1;\

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

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

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

  585.                 src += stride;\

  586.                 tptr += 11;\

  587.             }\

  588. \

  589.             r = 64-rnd;\

  590.             tptr = tmp+1;\

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

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

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

  594.                 dst += stride;\

  595.                 tptr += 11;\

  596.             }\

  597. \

  598.             return;\

  599.         }\

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

  601.             r = 1-rnd;\

  602. \

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

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

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

  606.                 src += stride;\

  607.                 dst += stride;\

  608.             }\

  609.             return;\

  610.         }\

  611.     }\

  612. \

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

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

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

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

  617.         dst += stride;\

  618.         src += stride;\

  619.     }\

  620. }\

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

  622.     int i;\

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

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

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

  626.         pixels+=line_size;\

  627.         block +=line_size;\

  628.     }\

  629. }

  630. 

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

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

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

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

  635. 

  636. VC1_MSPEL_MC(op_put, op4_put, put_)

  637. VC1_MSPEL_MC(op_avg, op4_avg, avg_)

  638. 

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

  640. 

  641. #define PUT_VC1_MSPEL(a, b)\

  642. static void put_vc1_mspel_mc ## a ## b ##_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) { \

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

  644. }\

  645. static void avg_vc1_mspel_mc ## a ## b ##_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) { \

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

  647. }

  648. 

  649. PUT_VC1_MSPEL(1, 0)

  650. PUT_VC1_MSPEL(2, 0)

  651. PUT_VC1_MSPEL(3, 0)

  652. 

  653. PUT_VC1_MSPEL(0, 1)

  654. PUT_VC1_MSPEL(1, 1)

  655. PUT_VC1_MSPEL(2, 1)

  656. PUT_VC1_MSPEL(3, 1)

  657. 

  658. PUT_VC1_MSPEL(0, 2)

  659. PUT_VC1_MSPEL(1, 2)

  660. PUT_VC1_MSPEL(2, 2)

  661. PUT_VC1_MSPEL(3, 2)

  662. 

  663. PUT_VC1_MSPEL(0, 3)

  664. PUT_VC1_MSPEL(1, 3)

  665. PUT_VC1_MSPEL(2, 3)

  666. PUT_VC1_MSPEL(3, 3)

  667. 

  668. 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){

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

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

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

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

  673.     int i;

  674. 

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

  676. 

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

  678.     {

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

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

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

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

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

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

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

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

  687.         dst+= stride;

  688.         src+= stride;

  689.     }

  690. }

  691. 

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

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

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

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

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

  697.     int i;

  698. 

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

  700. 

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

  702.     {

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

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

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

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

  707.         dst+= stride;

  708.         src+= stride;

  709.     }

  710. }

  711. 

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

  713. 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){

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

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

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

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

  718.     int i;

  719. 

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

  721. 

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

  723.     {

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

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

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

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

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

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

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

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

  732.         dst+= stride;

  733.         src+= stride;

  734.     }

  735. }

  736. 

  737. #if CONFIG_WMV3IMAGE_DECODER || CONFIG_VC1IMAGE_DECODER

  738. 

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

  740. {

  741.     while (count--) {

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

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

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

  745.         offset += advance;

  746.     }

  747. }

  748. 

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

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

  751.                                             int alpha, int scaled, int width)

  752. {

  753.     int a1, b1, a2, b2;

  754.     while (width--) {

  755.         a1 = *src1a++;

  756.         if (scaled) {

  757.             b1 = *src1b++;

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

  759.         }

  760.         if (two_sprites) {

  761.             a2 = *src2a++;

  762.             if (scaled > 1) {

  763.                 b2 = *src2b++;

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

  765.             }

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

  767.         }

  768.         *dst++ = a1;

  769.     }

  770. }

  771. 

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

  773. {

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

  775. }

  776. 

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

  778. {

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

  780. }

  781. 

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

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

  784. {

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

  786. }

  787. 

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

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

  790.                                        int alpha, int width)

  791. {

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

  793. }

  794. 

  795. #endif

  796. 

  797. av_cold void ff_vc1dsp_init(VC1DSPContext* dsp) {

  798.     dsp->vc1_inv_trans_8x8 = vc1_inv_trans_8x8_c;

  799.     dsp->vc1_inv_trans_4x8 = vc1_inv_trans_4x8_c;

  800.     dsp->vc1_inv_trans_8x4 = vc1_inv_trans_8x4_c;

  801.     dsp->vc1_inv_trans_4x4 = vc1_inv_trans_4x4_c;

  802.     dsp->vc1_inv_trans_8x8_dc = vc1_inv_trans_8x8_dc_c;

  803.     dsp->vc1_inv_trans_4x8_dc = vc1_inv_trans_4x8_dc_c;

  804.     dsp->vc1_inv_trans_8x4_dc = vc1_inv_trans_8x4_dc_c;

  805.     dsp->vc1_inv_trans_4x4_dc = vc1_inv_trans_4x4_dc_c;

  806.     dsp->vc1_h_overlap = vc1_h_overlap_c;

  807.     dsp->vc1_v_overlap = vc1_v_overlap_c;

  808.     dsp->vc1_h_s_overlap = vc1_h_s_overlap_c;

  809.     dsp->vc1_v_s_overlap = vc1_v_s_overlap_c;

  810.     dsp->vc1_v_loop_filter4 = vc1_v_loop_filter4_c;

  811.     dsp->vc1_h_loop_filter4 = vc1_h_loop_filter4_c;

  812.     dsp->vc1_v_loop_filter8 = vc1_v_loop_filter8_c;

  813.     dsp->vc1_h_loop_filter8 = vc1_h_loop_filter8_c;

  814.     dsp->vc1_v_loop_filter16 = vc1_v_loop_filter16_c;

  815.     dsp->vc1_h_loop_filter16 = vc1_h_loop_filter16_c;

  816. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  833. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  850. 

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

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

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

  854. 

  855. #if CONFIG_WMV3IMAGE_DECODER || CONFIG_VC1IMAGE_DECODER

  856.     dsp->sprite_h = sprite_h_c;

  857.     dsp->sprite_v_single = sprite_v_single_c;

  858.     dsp->sprite_v_double_noscale = sprite_v_double_noscale_c;

  859.     dsp->sprite_v_double_onescale = sprite_v_double_onescale_c;

  860.     dsp->sprite_v_double_twoscale = sprite_v_double_twoscale_c;

  861. #endif

  862. 

  863.     if (HAVE_ALTIVEC)

  864.         ff_vc1dsp_init_altivec(dsp);

  865.     if (ARCH_X86)

  866.         ff_vc1dsp_init_x86(dsp);

  867. }