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.
30761 Commits
32 Branches
293MB
Tree: c2d9a65bc0
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'c2d9a65bc0'
${ noResults }
FFmpeg/libavcodec/dsputil_template.c

1280 lines
55KB
Raw Blame History 

  1. /*

  2.  * DSP utils

  3.  * Copyright (c) 2000, 2001 Fabrice Bellard

  4.  * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>

  5.  *

  6.  * gmc & q-pel & 32/64 bit based MC by Michael Niedermayer <michaelni@gmx.at>

  7.  *

  8.  * This file is part of Libav.

  9.  *

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

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

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

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

  14.  *

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

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

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

  18.  * Lesser General Public License for more details.

  19.  *

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

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

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

  23.  */

  24. 

  25. /**

  26.  * @file

  27.  * DSP utils

  28.  */

  29. 

  30. #include "bit_depth_template.c"

  31. 

  32. static inline void FUNC(copy_block2)(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)

  33. {

  34.     int i;

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

  36.     {

  37.         AV_WN2P(dst   , AV_RN2P(src   ));

  38.         dst+=dstStride;

  39.         src+=srcStride;

  40.     }

  41. }

  42. 

  43. static inline void FUNC(copy_block4)(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)

  44. {

  45.     int i;

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

  47.     {

  48.         AV_WN4P(dst   , AV_RN4P(src   ));

  49.         dst+=dstStride;

  50.         src+=srcStride;

  51.     }

  52. }

  53. 

  54. static inline void FUNC(copy_block8)(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)

  55. {

  56.     int i;

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

  58.     {

  59.         AV_WN4P(dst                , AV_RN4P(src                ));

  60.         AV_WN4P(dst+4*sizeof(pixel), AV_RN4P(src+4*sizeof(pixel)));

  61.         dst+=dstStride;

  62.         src+=srcStride;

  63.     }

  64. }

  65. 

  66. static inline void FUNC(copy_block16)(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)

  67. {

  68.     int i;

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

  70.     {

  71.         AV_WN4P(dst                 , AV_RN4P(src                 ));

  72.         AV_WN4P(dst+ 4*sizeof(pixel), AV_RN4P(src+ 4*sizeof(pixel)));

  73.         AV_WN4P(dst+ 8*sizeof(pixel), AV_RN4P(src+ 8*sizeof(pixel)));

  74.         AV_WN4P(dst+12*sizeof(pixel), AV_RN4P(src+12*sizeof(pixel)));

  75.         dst+=dstStride;

  76.         src+=srcStride;

  77.     }

  78. }

  79. 

  80. /* draw the edges of width 'w' of an image of size width, height */

  81. //FIXME check that this is ok for mpeg4 interlaced

  82. static void FUNCC(draw_edges)(uint8_t *_buf, int _wrap, int width, int height, int w, int h, int sides)

  83. {

  84.     pixel *buf = (pixel*)_buf;

  85.     int wrap = _wrap / sizeof(pixel);

  86.     pixel *ptr, *last_line;

  87.     int i;

  88. 

  89.     /* left and right */

  90.     ptr = buf;

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

  92. #if BIT_DEPTH > 8

  93.         int j;

  94.         for (j = 0; j < w; j++) {

  95.             ptr[j-w] = ptr[0];

  96.             ptr[j+width] = ptr[width-1];

  97.         }

  98. #else

  99.         memset(ptr - w, ptr[0], w);

  100.         memset(ptr + width, ptr[width-1], w);

  101. #endif

  102.         ptr += wrap;

  103.     }

  104. 

  105.     /* top and bottom + corners */

  106.     buf -= w;

  107.     last_line = buf + (height - 1) * wrap;

  108.     if (sides & EDGE_TOP)

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

  110.             memcpy(buf - (i + 1) * wrap, buf, (width + w + w) * sizeof(pixel)); // top

  111.     if (sides & EDGE_BOTTOM)

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

  113.             memcpy(last_line + (i + 1) * wrap, last_line, (width + w + w) * sizeof(pixel)); // bottom

  114. }

  115. 

  116. /**

  117.  * Copy a rectangular area of samples to a temporary buffer and replicate the border samples.

  118.  * @param buf destination buffer

  119.  * @param src source buffer

  120.  * @param linesize number of bytes between 2 vertically adjacent samples in both the source and destination buffers

  121.  * @param block_w width of block

  122.  * @param block_h height of block

  123.  * @param src_x x coordinate of the top left sample of the block in the source buffer

  124.  * @param src_y y coordinate of the top left sample of the block in the source buffer

  125.  * @param w width of the source buffer

  126.  * @param h height of the source buffer

  127.  */

  128. void FUNC(ff_emulated_edge_mc)(uint8_t *buf, const uint8_t *src, int linesize, int block_w, int block_h,

  129.                                     int src_x, int src_y, int w, int h){

  130.     int x, y;

  131.     int start_y, start_x, end_y, end_x;

  132. 

  133.     if(src_y>= h){

  134.         src+= (h-1-src_y)*linesize;

  135.         src_y=h-1;

  136.     }else if(src_y<=-block_h){

  137.         src+= (1-block_h-src_y)*linesize;

  138.         src_y=1-block_h;

  139.     }

  140.     if(src_x>= w){

  141.         src+= (w-1-src_x)*sizeof(pixel);

  142.         src_x=w-1;

  143.     }else if(src_x<=-block_w){

  144.         src+= (1-block_w-src_x)*sizeof(pixel);

  145.         src_x=1-block_w;

  146.     }

  147. 

  148.     start_y= FFMAX(0, -src_y);

  149.     start_x= FFMAX(0, -src_x);

  150.     end_y= FFMIN(block_h, h-src_y);

  151.     end_x= FFMIN(block_w, w-src_x);

  152.     assert(start_y < end_y && block_h);

  153.     assert(start_x < end_x && block_w);

  154. 

  155.     w    = end_x - start_x;

  156.     src += start_y*linesize + start_x*sizeof(pixel);

  157.     buf += start_x*sizeof(pixel);

  158. 

  159.     //top

  160.     for(y=0; y<start_y; y++){

  161.         memcpy(buf, src, w*sizeof(pixel));

  162.         buf += linesize;

  163.     }

  164. 

  165.     // copy existing part

  166.     for(; y<end_y; y++){

  167.         memcpy(buf, src, w*sizeof(pixel));

  168.         src += linesize;

  169.         buf += linesize;

  170.     }

  171. 

  172.     //bottom

  173.     src -= linesize;

  174.     for(; y<block_h; y++){

  175.         memcpy(buf, src, w*sizeof(pixel));

  176.         buf += linesize;

  177.     }

  178. 

  179.     buf -= block_h * linesize + start_x*sizeof(pixel);

  180.     while (block_h--){

  181.         pixel *bufp = (pixel*)buf;

  182.        //left

  183.         for(x=0; x<start_x; x++){

  184.             bufp[x] = bufp[start_x];

  185.         }

  186. 

  187.        //right

  188.         for(x=end_x; x<block_w; x++){

  189.             bufp[x] = bufp[end_x - 1];

  190.         }

  191.         buf += linesize;

  192.     }

  193. }

  194. 

  195. #define DCTELEM_FUNCS(dctcoef, suffix)                                  \

  196. static void FUNCC(get_pixels ## suffix)(DCTELEM *restrict _block,       \

  197.                                         const uint8_t *_pixels,         \

  198.                                         int line_size)                  \

  199. {                                                                       \

  200.     const pixel *pixels = (const pixel *) _pixels;                      \

  201.     dctcoef *restrict block = (dctcoef *) _block;                       \

  202.     int i;                                                              \

  203.                                                                         \

  204.     /* read the pixels */                                               \

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

  206.         block[0] = pixels[0];                                           \

  207.         block[1] = pixels[1];                                           \

  208.         block[2] = pixels[2];                                           \

  209.         block[3] = pixels[3];                                           \

  210.         block[4] = pixels[4];                                           \

  211.         block[5] = pixels[5];                                           \

  212.         block[6] = pixels[6];                                           \

  213.         block[7] = pixels[7];                                           \

  214.         pixels += line_size / sizeof(pixel);                            \

  215.         block += 8;                                                     \

  216.     }                                                                   \

  217. }                                                                       \

  218.                                                                         \

  219. static void FUNCC(add_pixels8 ## suffix)(uint8_t *restrict _pixels,     \

  220.                                          DCTELEM *_block,               \

  221.                                          int line_size)                 \

  222. {                                                                       \

  223.     int i;                                                              \

  224.     pixel *restrict pixels = (pixel *restrict)_pixels;                  \

  225.     dctcoef *block = (dctcoef*)_block;                                  \

  226.     line_size /= sizeof(pixel);                                         \

  227.                                                                         \

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

  229.         pixels[0] += block[0];                                          \

  230.         pixels[1] += block[1];                                          \

  231.         pixels[2] += block[2];                                          \

  232.         pixels[3] += block[3];                                          \

  233.         pixels[4] += block[4];                                          \

  234.         pixels[5] += block[5];                                          \

  235.         pixels[6] += block[6];                                          \

  236.         pixels[7] += block[7];                                          \

  237.         pixels += line_size;                                            \

  238.         block += 8;                                                     \

  239.     }                                                                   \

  240. }                                                                       \

  241.                                                                         \

  242. static void FUNCC(add_pixels4 ## suffix)(uint8_t *restrict _pixels,     \

  243.                                          DCTELEM *_block,               \

  244.                                          int line_size)                 \

  245. {                                                                       \

  246.     int i;                                                              \

  247.     pixel *restrict pixels = (pixel *restrict)_pixels;                  \

  248.     dctcoef *block = (dctcoef*)_block;                                  \

  249.     line_size /= sizeof(pixel);                                         \

  250.                                                                         \

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

  252.         pixels[0] += block[0];                                          \

  253.         pixels[1] += block[1];                                          \

  254.         pixels[2] += block[2];                                          \

  255.         pixels[3] += block[3];                                          \

  256.         pixels += line_size;                                            \

  257.         block += 4;                                                     \

  258.     }                                                                   \

  259. }                                                                       \

  260.                                                                         \

  261. static void FUNCC(clear_block ## suffix)(DCTELEM *block)                \

  262. {                                                                       \

  263.     memset(block, 0, sizeof(dctcoef)*64);                               \

  264. }                                                                       \

  265.                                                                         \

  266. /**                                                                     \

  267.  * memset(blocks, 0, sizeof(DCTELEM)*6*64)                              \

  268.  */                                                                     \

  269. static void FUNCC(clear_blocks ## suffix)(DCTELEM *blocks)              \

  270. {                                                                       \

  271.     memset(blocks, 0, sizeof(dctcoef)*6*64);                            \

  272. }

  273. 

  274. DCTELEM_FUNCS(DCTELEM, _16)

  275. #if BIT_DEPTH > 8

  276. DCTELEM_FUNCS(dctcoef, _32)

  277. #endif

  278. 

  279. #define PIXOP2(OPNAME, OP) \

  280. static void FUNCC(OPNAME ## _pixels2)(uint8_t *block, const uint8_t *pixels, int line_size, int h){\

  281.     int i;\

  282.     for(i=0; i<h; i++){\

  283.         OP(*((pixel2*)(block  )), AV_RN2P(pixels  ));\

  284.         pixels+=line_size;\

  285.         block +=line_size;\

  286.     }\

  287. }\

  288. static void FUNCC(OPNAME ## _pixels4)(uint8_t *block, const uint8_t *pixels, int line_size, int h){\

  289.     int i;\

  290.     for(i=0; i<h; i++){\

  291.         OP(*((pixel4*)(block  )), AV_RN4P(pixels  ));\

  292.         pixels+=line_size;\

  293.         block +=line_size;\

  294.     }\

  295. }\

  296. static void FUNCC(OPNAME ## _pixels8)(uint8_t *block, const uint8_t *pixels, int line_size, int h){\

  297.     int i;\

  298.     for(i=0; i<h; i++){\

  299.         OP(*((pixel4*)(block                )), AV_RN4P(pixels                ));\

  300.         OP(*((pixel4*)(block+4*sizeof(pixel))), AV_RN4P(pixels+4*sizeof(pixel)));\

  301.         pixels+=line_size;\

  302.         block +=line_size;\

  303.     }\

  304. }\

  305. static inline void FUNCC(OPNAME ## _no_rnd_pixels8)(uint8_t *block, const uint8_t *pixels, int line_size, int h){\

  306.     FUNCC(OPNAME ## _pixels8)(block, pixels, line_size, h);\

  307. }\

  308. \

  309. static inline void FUNC(OPNAME ## _no_rnd_pixels8_l2)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \

  310.                                                 int src_stride1, int src_stride2, int h){\

  311.     int i;\

  312.     for(i=0; i<h; i++){\

  313.         pixel4 a,b;\

  314.         a= AV_RN4P(&src1[i*src_stride1  ]);\

  315.         b= AV_RN4P(&src2[i*src_stride2  ]);\

  316.         OP(*((pixel4*)&dst[i*dst_stride  ]), no_rnd_avg_pixel4(a, b));\

  317.         a= AV_RN4P(&src1[i*src_stride1+4*sizeof(pixel)]);\

  318.         b= AV_RN4P(&src2[i*src_stride2+4*sizeof(pixel)]);\

  319.         OP(*((pixel4*)&dst[i*dst_stride+4*sizeof(pixel)]), no_rnd_avg_pixel4(a, b));\

  320.     }\

  321. }\

  322. \

  323. static inline void FUNC(OPNAME ## _pixels8_l2)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \

  324.                                                 int src_stride1, int src_stride2, int h){\

  325.     int i;\

  326.     for(i=0; i<h; i++){\

  327.         pixel4 a,b;\

  328.         a= AV_RN4P(&src1[i*src_stride1  ]);\

  329.         b= AV_RN4P(&src2[i*src_stride2  ]);\

  330.         OP(*((pixel4*)&dst[i*dst_stride  ]), rnd_avg_pixel4(a, b));\

  331.         a= AV_RN4P(&src1[i*src_stride1+4*sizeof(pixel)]);\

  332.         b= AV_RN4P(&src2[i*src_stride2+4*sizeof(pixel)]);\

  333.         OP(*((pixel4*)&dst[i*dst_stride+4*sizeof(pixel)]), rnd_avg_pixel4(a, b));\

  334.     }\

  335. }\

  336. \

  337. static inline void FUNC(OPNAME ## _pixels4_l2)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \

  338.                                                 int src_stride1, int src_stride2, int h){\

  339.     int i;\

  340.     for(i=0; i<h; i++){\

  341.         pixel4 a,b;\

  342.         a= AV_RN4P(&src1[i*src_stride1  ]);\

  343.         b= AV_RN4P(&src2[i*src_stride2  ]);\

  344.         OP(*((pixel4*)&dst[i*dst_stride  ]), rnd_avg_pixel4(a, b));\

  345.     }\

  346. }\

  347. \

  348. static inline void FUNC(OPNAME ## _pixels2_l2)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \

  349.                                                 int src_stride1, int src_stride2, int h){\

  350.     int i;\

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

  352.         pixel4 a,b;\

  353.         a= AV_RN2P(&src1[i*src_stride1  ]);\

  354.         b= AV_RN2P(&src2[i*src_stride2  ]);\

  355.         OP(*((pixel2*)&dst[i*dst_stride  ]), rnd_avg_pixel4(a, b));\

  356.     }\

  357. }\

  358. \

  359. static inline void FUNC(OPNAME ## _pixels16_l2)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \

  360.                                                 int src_stride1, int src_stride2, int h){\

  361.     FUNC(OPNAME ## _pixels8_l2)(dst  , src1  , src2  , dst_stride, src_stride1, src_stride2, h);\

  362.     FUNC(OPNAME ## _pixels8_l2)(dst+8*sizeof(pixel), src1+8*sizeof(pixel), src2+8*sizeof(pixel), dst_stride, src_stride1, src_stride2, h);\

  363. }\

  364. \

  365. static inline void FUNC(OPNAME ## _no_rnd_pixels16_l2)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \

  366.                                                 int src_stride1, int src_stride2, int h){\

  367.     FUNC(OPNAME ## _no_rnd_pixels8_l2)(dst  , src1  , src2  , dst_stride, src_stride1, src_stride2, h);\

  368.     FUNC(OPNAME ## _no_rnd_pixels8_l2)(dst+8*sizeof(pixel), src1+8*sizeof(pixel), src2+8*sizeof(pixel), dst_stride, src_stride1, src_stride2, h);\

  369. }\

  370. \

  371. static inline void FUNCC(OPNAME ## _no_rnd_pixels8_x2)(uint8_t *block, const uint8_t *pixels, int line_size, int h){\

  372.     FUNC(OPNAME ## _no_rnd_pixels8_l2)(block, pixels, pixels+sizeof(pixel), line_size, line_size, line_size, h);\

  373. }\

  374. \

  375. static inline void FUNCC(OPNAME ## _pixels8_x2)(uint8_t *block, const uint8_t *pixels, int line_size, int h){\

  376.     FUNC(OPNAME ## _pixels8_l2)(block, pixels, pixels+sizeof(pixel), line_size, line_size, line_size, h);\

  377. }\

  378. \

  379. static inline void FUNCC(OPNAME ## _no_rnd_pixels8_y2)(uint8_t *block, const uint8_t *pixels, int line_size, int h){\

  380.     FUNC(OPNAME ## _no_rnd_pixels8_l2)(block, pixels, pixels+line_size, line_size, line_size, line_size, h);\

  381. }\

  382. \

  383. static inline void FUNCC(OPNAME ## _pixels8_y2)(uint8_t *block, const uint8_t *pixels, int line_size, int h){\

  384.     FUNC(OPNAME ## _pixels8_l2)(block, pixels, pixels+line_size, line_size, line_size, line_size, h);\

  385. }\

  386. \

  387. static inline void FUNC(OPNAME ## _pixels8_l4)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, const uint8_t *src3, const uint8_t *src4,\

  388.                  int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\

  389.     /* FIXME HIGH BIT DEPTH */\

  390.     int i;\

  391.     for(i=0; i<h; i++){\

  392.         uint32_t a, b, c, d, l0, l1, h0, h1;\

  393.         a= AV_RN32(&src1[i*src_stride1]);\

  394.         b= AV_RN32(&src2[i*src_stride2]);\

  395.         c= AV_RN32(&src3[i*src_stride3]);\

  396.         d= AV_RN32(&src4[i*src_stride4]);\

  397.         l0=  (a&0x03030303UL)\

  398.            + (b&0x03030303UL)\

  399.            + 0x02020202UL;\

  400.         h0= ((a&0xFCFCFCFCUL)>>2)\

  401.           + ((b&0xFCFCFCFCUL)>>2);\

  402.         l1=  (c&0x03030303UL)\

  403.            + (d&0x03030303UL);\

  404.         h1= ((c&0xFCFCFCFCUL)>>2)\

  405.           + ((d&0xFCFCFCFCUL)>>2);\

  406.         OP(*((uint32_t*)&dst[i*dst_stride]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\

  407.         a= AV_RN32(&src1[i*src_stride1+4]);\

  408.         b= AV_RN32(&src2[i*src_stride2+4]);\

  409.         c= AV_RN32(&src3[i*src_stride3+4]);\

  410.         d= AV_RN32(&src4[i*src_stride4+4]);\

  411.         l0=  (a&0x03030303UL)\

  412.            + (b&0x03030303UL)\

  413.            + 0x02020202UL;\

  414.         h0= ((a&0xFCFCFCFCUL)>>2)\

  415.           + ((b&0xFCFCFCFCUL)>>2);\

  416.         l1=  (c&0x03030303UL)\

  417.            + (d&0x03030303UL);\

  418.         h1= ((c&0xFCFCFCFCUL)>>2)\

  419.           + ((d&0xFCFCFCFCUL)>>2);\

  420.         OP(*((uint32_t*)&dst[i*dst_stride+4]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\

  421.     }\

  422. }\

  423. \

  424. static inline void FUNCC(OPNAME ## _pixels4_x2)(uint8_t *block, const uint8_t *pixels, int line_size, int h){\

  425.     FUNC(OPNAME ## _pixels4_l2)(block, pixels, pixels+sizeof(pixel), line_size, line_size, line_size, h);\

  426. }\

  427. \

  428. static inline void FUNCC(OPNAME ## _pixels4_y2)(uint8_t *block, const uint8_t *pixels, int line_size, int h){\

  429.     FUNC(OPNAME ## _pixels4_l2)(block, pixels, pixels+line_size, line_size, line_size, line_size, h);\

  430. }\

  431. \

  432. static inline void FUNCC(OPNAME ## _pixels2_x2)(uint8_t *block, const uint8_t *pixels, int line_size, int h){\

  433.     FUNC(OPNAME ## _pixels2_l2)(block, pixels, pixels+sizeof(pixel), line_size, line_size, line_size, h);\

  434. }\

  435. \

  436. static inline void FUNCC(OPNAME ## _pixels2_y2)(uint8_t *block, const uint8_t *pixels, int line_size, int h){\

  437.     FUNC(OPNAME ## _pixels2_l2)(block, pixels, pixels+line_size, line_size, line_size, line_size, h);\

  438. }\

  439. \

  440. static inline void FUNC(OPNAME ## _no_rnd_pixels8_l4)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, const uint8_t *src3, const uint8_t *src4,\

  441.                  int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\

  442.     /* FIXME HIGH BIT DEPTH*/\

  443.     int i;\

  444.     for(i=0; i<h; i++){\

  445.         uint32_t a, b, c, d, l0, l1, h0, h1;\

  446.         a= AV_RN32(&src1[i*src_stride1]);\

  447.         b= AV_RN32(&src2[i*src_stride2]);\

  448.         c= AV_RN32(&src3[i*src_stride3]);\

  449.         d= AV_RN32(&src4[i*src_stride4]);\

  450.         l0=  (a&0x03030303UL)\

  451.            + (b&0x03030303UL)\

  452.            + 0x01010101UL;\

  453.         h0= ((a&0xFCFCFCFCUL)>>2)\

  454.           + ((b&0xFCFCFCFCUL)>>2);\

  455.         l1=  (c&0x03030303UL)\

  456.            + (d&0x03030303UL);\

  457.         h1= ((c&0xFCFCFCFCUL)>>2)\

  458.           + ((d&0xFCFCFCFCUL)>>2);\

  459.         OP(*((uint32_t*)&dst[i*dst_stride]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\

  460.         a= AV_RN32(&src1[i*src_stride1+4]);\

  461.         b= AV_RN32(&src2[i*src_stride2+4]);\

  462.         c= AV_RN32(&src3[i*src_stride3+4]);\

  463.         d= AV_RN32(&src4[i*src_stride4+4]);\

  464.         l0=  (a&0x03030303UL)\

  465.            + (b&0x03030303UL)\

  466.            + 0x01010101UL;\

  467.         h0= ((a&0xFCFCFCFCUL)>>2)\

  468.           + ((b&0xFCFCFCFCUL)>>2);\

  469.         l1=  (c&0x03030303UL)\

  470.            + (d&0x03030303UL);\

  471.         h1= ((c&0xFCFCFCFCUL)>>2)\

  472.           + ((d&0xFCFCFCFCUL)>>2);\

  473.         OP(*((uint32_t*)&dst[i*dst_stride+4]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\

  474.     }\

  475. }\

  476. static inline void FUNC(OPNAME ## _pixels16_l4)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, const uint8_t *src3, const uint8_t *src4,\

  477.                  int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\

  478.     FUNC(OPNAME ## _pixels8_l4)(dst  , src1  , src2  , src3  , src4  , dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\

  479.     FUNC(OPNAME ## _pixels8_l4)(dst+8*sizeof(pixel), src1+8*sizeof(pixel), src2+8*sizeof(pixel), src3+8*sizeof(pixel), src4+8*sizeof(pixel), dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\

  480. }\

  481. static inline void FUNC(OPNAME ## _no_rnd_pixels16_l4)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, const uint8_t *src3, const uint8_t *src4,\

  482.                  int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\

  483.     FUNC(OPNAME ## _no_rnd_pixels8_l4)(dst  , src1  , src2  , src3  , src4  , dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\

  484.     FUNC(OPNAME ## _no_rnd_pixels8_l4)(dst+8*sizeof(pixel), src1+8*sizeof(pixel), src2+8*sizeof(pixel), src3+8*sizeof(pixel), src4+8*sizeof(pixel), dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\

  485. }\

  486. \

  487. static inline void FUNCC(OPNAME ## _pixels2_xy2)(uint8_t *_block, const uint8_t *_pixels, int line_size, int h)\

  488. {\

  489.         int i, a0, b0, a1, b1;\

  490.         pixel *block = (pixel*)_block;\

  491.         const pixel *pixels = (const pixel*)_pixels;\

  492.         line_size /= sizeof(pixel);\

  493.         a0= pixels[0];\

  494.         b0= pixels[1] + 2;\

  495.         a0 += b0;\

  496.         b0 += pixels[2];\

  497. \

  498.         pixels+=line_size;\

  499.         for(i=0; i<h; i+=2){\

  500.             a1= pixels[0];\

  501.             b1= pixels[1];\

  502.             a1 += b1;\

  503.             b1 += pixels[2];\

  504. \

  505.             block[0]= (a1+a0)>>2; /* FIXME non put */\

  506.             block[1]= (b1+b0)>>2;\

  507. \

  508.             pixels+=line_size;\

  509.             block +=line_size;\

  510. \

  511.             a0= pixels[0];\

  512.             b0= pixels[1] + 2;\

  513.             a0 += b0;\

  514.             b0 += pixels[2];\

  515. \

  516.             block[0]= (a1+a0)>>2;\

  517.             block[1]= (b1+b0)>>2;\

  518.             pixels+=line_size;\

  519.             block +=line_size;\

  520.         }\

  521. }\

  522. \

  523. static inline void FUNCC(OPNAME ## _pixels4_xy2)(uint8_t *block, const uint8_t *pixels, int line_size, int h)\

  524. {\

  525.         /* FIXME HIGH BIT DEPTH */\

  526.         int i;\

  527.         const uint32_t a= AV_RN32(pixels  );\

  528.         const uint32_t b= AV_RN32(pixels+1);\

  529.         uint32_t l0=  (a&0x03030303UL)\

  530.                     + (b&0x03030303UL)\

  531.                     + 0x02020202UL;\

  532.         uint32_t h0= ((a&0xFCFCFCFCUL)>>2)\

  533.                    + ((b&0xFCFCFCFCUL)>>2);\

  534.         uint32_t l1,h1;\

  535. \

  536.         pixels+=line_size;\

  537.         for(i=0; i<h; i+=2){\

  538.             uint32_t a= AV_RN32(pixels  );\

  539.             uint32_t b= AV_RN32(pixels+1);\

  540.             l1=  (a&0x03030303UL)\

  541.                + (b&0x03030303UL);\

  542.             h1= ((a&0xFCFCFCFCUL)>>2)\

  543.               + ((b&0xFCFCFCFCUL)>>2);\

  544.             OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\

  545.             pixels+=line_size;\

  546.             block +=line_size;\

  547.             a= AV_RN32(pixels  );\

  548.             b= AV_RN32(pixels+1);\

  549.             l0=  (a&0x03030303UL)\

  550.                + (b&0x03030303UL)\

  551.                + 0x02020202UL;\

  552.             h0= ((a&0xFCFCFCFCUL)>>2)\

  553.               + ((b&0xFCFCFCFCUL)>>2);\

  554.             OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\

  555.             pixels+=line_size;\

  556.             block +=line_size;\

  557.         }\

  558. }\

  559. \

  560. static inline void FUNCC(OPNAME ## _pixels8_xy2)(uint8_t *block, const uint8_t *pixels, int line_size, int h)\

  561. {\

  562.     /* FIXME HIGH BIT DEPTH */\

  563.     int j;\

  564.     for(j=0; j<2; j++){\

  565.         int i;\

  566.         const uint32_t a= AV_RN32(pixels  );\

  567.         const uint32_t b= AV_RN32(pixels+1);\

  568.         uint32_t l0=  (a&0x03030303UL)\

  569.                     + (b&0x03030303UL)\

  570.                     + 0x02020202UL;\

  571.         uint32_t h0= ((a&0xFCFCFCFCUL)>>2)\

  572.                    + ((b&0xFCFCFCFCUL)>>2);\

  573.         uint32_t l1,h1;\

  574. \

  575.         pixels+=line_size;\

  576.         for(i=0; i<h; i+=2){\

  577.             uint32_t a= AV_RN32(pixels  );\

  578.             uint32_t b= AV_RN32(pixels+1);\

  579.             l1=  (a&0x03030303UL)\

  580.                + (b&0x03030303UL);\

  581.             h1= ((a&0xFCFCFCFCUL)>>2)\

  582.               + ((b&0xFCFCFCFCUL)>>2);\

  583.             OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\

  584.             pixels+=line_size;\

  585.             block +=line_size;\

  586.             a= AV_RN32(pixels  );\

  587.             b= AV_RN32(pixels+1);\

  588.             l0=  (a&0x03030303UL)\

  589.                + (b&0x03030303UL)\

  590.                + 0x02020202UL;\

  591.             h0= ((a&0xFCFCFCFCUL)>>2)\

  592.               + ((b&0xFCFCFCFCUL)>>2);\

  593.             OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\

  594.             pixels+=line_size;\

  595.             block +=line_size;\

  596.         }\

  597.         pixels+=4-line_size*(h+1);\

  598.         block +=4-line_size*h;\

  599.     }\

  600. }\

  601. \

  602. static inline void FUNCC(OPNAME ## _no_rnd_pixels8_xy2)(uint8_t *block, const uint8_t *pixels, int line_size, int h)\

  603. {\

  604.     /* FIXME HIGH BIT DEPTH */\

  605.     int j;\

  606.     for(j=0; j<2; j++){\

  607.         int i;\

  608.         const uint32_t a= AV_RN32(pixels  );\

  609.         const uint32_t b= AV_RN32(pixels+1);\

  610.         uint32_t l0=  (a&0x03030303UL)\

  611.                     + (b&0x03030303UL)\

  612.                     + 0x01010101UL;\

  613.         uint32_t h0= ((a&0xFCFCFCFCUL)>>2)\

  614.                    + ((b&0xFCFCFCFCUL)>>2);\

  615.         uint32_t l1,h1;\

  616. \

  617.         pixels+=line_size;\

  618.         for(i=0; i<h; i+=2){\

  619.             uint32_t a= AV_RN32(pixels  );\

  620.             uint32_t b= AV_RN32(pixels+1);\

  621.             l1=  (a&0x03030303UL)\

  622.                + (b&0x03030303UL);\

  623.             h1= ((a&0xFCFCFCFCUL)>>2)\

  624.               + ((b&0xFCFCFCFCUL)>>2);\

  625.             OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\

  626.             pixels+=line_size;\

  627.             block +=line_size;\

  628.             a= AV_RN32(pixels  );\

  629.             b= AV_RN32(pixels+1);\

  630.             l0=  (a&0x03030303UL)\

  631.                + (b&0x03030303UL)\

  632.                + 0x01010101UL;\

  633.             h0= ((a&0xFCFCFCFCUL)>>2)\

  634.               + ((b&0xFCFCFCFCUL)>>2);\

  635.             OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\

  636.             pixels+=line_size;\

  637.             block +=line_size;\

  638.         }\

  639.         pixels+=4-line_size*(h+1);\

  640.         block +=4-line_size*h;\

  641.     }\

  642. }\

  643. \

  644. CALL_2X_PIXELS(FUNCC(OPNAME ## _pixels16)    , FUNCC(OPNAME ## _pixels8)    , 8*sizeof(pixel))\

  645. CALL_2X_PIXELS(FUNCC(OPNAME ## _pixels16_x2) , FUNCC(OPNAME ## _pixels8_x2) , 8*sizeof(pixel))\

  646. CALL_2X_PIXELS(FUNCC(OPNAME ## _pixels16_y2) , FUNCC(OPNAME ## _pixels8_y2) , 8*sizeof(pixel))\

  647. CALL_2X_PIXELS(FUNCC(OPNAME ## _pixels16_xy2), FUNCC(OPNAME ## _pixels8_xy2), 8*sizeof(pixel))\

  648. av_unused CALL_2X_PIXELS(FUNCC(OPNAME ## _no_rnd_pixels16)    , FUNCC(OPNAME ## _pixels8) , 8*sizeof(pixel))\

  649. CALL_2X_PIXELS(FUNCC(OPNAME ## _no_rnd_pixels16_x2) , FUNCC(OPNAME ## _no_rnd_pixels8_x2) , 8*sizeof(pixel))\

  650. CALL_2X_PIXELS(FUNCC(OPNAME ## _no_rnd_pixels16_y2) , FUNCC(OPNAME ## _no_rnd_pixels8_y2) , 8*sizeof(pixel))\

  651. CALL_2X_PIXELS(FUNCC(OPNAME ## _no_rnd_pixels16_xy2), FUNCC(OPNAME ## _no_rnd_pixels8_xy2), 8*sizeof(pixel))\

  652. 

  653. #define op_avg(a, b) a = rnd_avg_pixel4(a, b)

  654. #define op_put(a, b) a = b

  655. 

  656. PIXOP2(avg, op_avg)

  657. PIXOP2(put, op_put)

  658. #undef op_avg

  659. #undef op_put

  660. 

  661. #define put_no_rnd_pixels8_c  put_pixels8_c

  662. #define put_no_rnd_pixels16_c put_pixels16_c

  663. 

  664. static void FUNCC(put_no_rnd_pixels16_l2)(uint8_t *dst, const uint8_t *a, const uint8_t *b, int stride, int h){

  665.     FUNC(put_no_rnd_pixels16_l2)(dst, a, b, stride, stride, stride, h);

  666. }

  667. 

  668. static void FUNCC(put_no_rnd_pixels8_l2)(uint8_t *dst, const uint8_t *a, const uint8_t *b, int stride, int h){

  669.     FUNC(put_no_rnd_pixels8_l2)(dst, a, b, stride, stride, stride, h);

  670. }

  671. 

  672. #define H264_CHROMA_MC(OPNAME, OP)\

  673. static void FUNCC(OPNAME ## h264_chroma_mc2)(uint8_t *_dst/*align 8*/, uint8_t *_src/*align 1*/, int stride, int h, int x, int y){\

  674.     pixel *dst = (pixel*)_dst;\

  675.     pixel *src = (pixel*)_src;\

  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.     stride /= sizeof(pixel);\

  682.     \

  683.     assert(x<8 && y<8 && x>=0 && y>=0);\

  684. \

  685.     if(D){\

  686.         for(i=0; i<h; i++){\

  687.             OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1]));\

  688.             OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2]));\

  689.             dst+= stride;\

  690.             src+= stride;\

  691.         }\

  692.     }else{\

  693.         const int E= B+C;\

  694.         const int step= C ? stride : 1;\

  695.         for(i=0; i<h; i++){\

  696.             OP(dst[0], (A*src[0] + E*src[step+0]));\

  697.             OP(dst[1], (A*src[1] + E*src[step+1]));\

  698.             dst+= stride;\

  699.             src+= stride;\

  700.         }\

  701.     }\

  702. }\

  703. \

  704. static void FUNCC(OPNAME ## h264_chroma_mc4)(uint8_t *_dst/*align 8*/, uint8_t *_src/*align 1*/, int stride, int h, int x, int y){\

  705.     pixel *dst = (pixel*)_dst;\

  706.     pixel *src = (pixel*)_src;\

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

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

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

  710.     const int D=(  x)*(  y);\

  711.     int i;\

  712.     stride /= sizeof(pixel);\

  713.     \

  714.     assert(x<8 && y<8 && x>=0 && y>=0);\

  715. \

  716.     if(D){\

  717.         for(i=0; i<h; i++){\

  718.             OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1]));\

  719.             OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2]));\

  720.             OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3]));\

  721.             OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4]));\

  722.             dst+= stride;\

  723.             src+= stride;\

  724.         }\

  725.     }else{\

  726.         const int E= B+C;\

  727.         const int step= C ? stride : 1;\

  728.         for(i=0; i<h; i++){\

  729.             OP(dst[0], (A*src[0] + E*src[step+0]));\

  730.             OP(dst[1], (A*src[1] + E*src[step+1]));\

  731.             OP(dst[2], (A*src[2] + E*src[step+2]));\

  732.             OP(dst[3], (A*src[3] + E*src[step+3]));\

  733.             dst+= stride;\

  734.             src+= stride;\

  735.         }\

  736.     }\

  737. }\

  738. \

  739. static void FUNCC(OPNAME ## h264_chroma_mc8)(uint8_t *_dst/*align 8*/, uint8_t *_src/*align 1*/, int stride, int h, int x, int y){\

  740.     pixel *dst = (pixel*)_dst;\

  741.     pixel *src = (pixel*)_src;\

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

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

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

  745.     const int D=(  x)*(  y);\

  746.     int i;\

  747.     stride /= sizeof(pixel);\

  748.     \

  749.     assert(x<8 && y<8 && x>=0 && y>=0);\

  750. \

  751.     if(D){\

  752.         for(i=0; i<h; i++){\

  753.             OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1]));\

  754.             OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2]));\

  755.             OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3]));\

  756.             OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4]));\

  757.             OP(dst[4], (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5]));\

  758.             OP(dst[5], (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6]));\

  759.             OP(dst[6], (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7]));\

  760.             OP(dst[7], (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8]));\

  761.             dst+= stride;\

  762.             src+= stride;\

  763.         }\

  764.     }else{\

  765.         const int E= B+C;\

  766.         const int step= C ? stride : 1;\

  767.         for(i=0; i<h; i++){\

  768.             OP(dst[0], (A*src[0] + E*src[step+0]));\

  769.             OP(dst[1], (A*src[1] + E*src[step+1]));\

  770.             OP(dst[2], (A*src[2] + E*src[step+2]));\

  771.             OP(dst[3], (A*src[3] + E*src[step+3]));\

  772.             OP(dst[4], (A*src[4] + E*src[step+4]));\

  773.             OP(dst[5], (A*src[5] + E*src[step+5]));\

  774.             OP(dst[6], (A*src[6] + E*src[step+6]));\

  775.             OP(dst[7], (A*src[7] + E*src[step+7]));\

  776.             dst+= stride;\

  777.             src+= stride;\

  778.         }\

  779.     }\

  780. }

  781. 

  782. #define op_avg(a, b) a = (((a)+(((b) + 32)>>6)+1)>>1)

  783. #define op_put(a, b) a = (((b) + 32)>>6)

  784. 

  785. H264_CHROMA_MC(put_       , op_put)

  786. H264_CHROMA_MC(avg_       , op_avg)

  787. #undef op_avg

  788. #undef op_put

  789. 

  790. #define H264_LOWPASS(OPNAME, OP, OP2) \

  791. static av_unused void FUNC(OPNAME ## h264_qpel2_h_lowpass)(uint8_t *_dst, uint8_t *_src, int dstStride, int srcStride){\

  792.     const int h=2;\

  793.     INIT_CLIP\

  794.     int i;\

  795.     pixel *dst = (pixel*)_dst;\

  796.     pixel *src = (pixel*)_src;\

  797.     dstStride /= sizeof(pixel);\

  798.     srcStride /= sizeof(pixel);\

  799.     for(i=0; i<h; i++)\

  800.     {\

  801.         OP(dst[0], (src[0]+src[1])*20 - (src[-1]+src[2])*5 + (src[-2]+src[3]));\

  802.         OP(dst[1], (src[1]+src[2])*20 - (src[0 ]+src[3])*5 + (src[-1]+src[4]));\

  803.         dst+=dstStride;\

  804.         src+=srcStride;\

  805.     }\

  806. }\

  807. \

  808. static av_unused void FUNC(OPNAME ## h264_qpel2_v_lowpass)(uint8_t *_dst, uint8_t *_src, int dstStride, int srcStride){\

  809.     const int w=2;\

  810.     INIT_CLIP\

  811.     int i;\

  812.     pixel *dst = (pixel*)_dst;\

  813.     pixel *src = (pixel*)_src;\

  814.     dstStride /= sizeof(pixel);\

  815.     srcStride /= sizeof(pixel);\

  816.     for(i=0; i<w; i++)\

  817.     {\

  818.         const int srcB= src[-2*srcStride];\

  819.         const int srcA= src[-1*srcStride];\

  820.         const int src0= src[0 *srcStride];\

  821.         const int src1= src[1 *srcStride];\

  822.         const int src2= src[2 *srcStride];\

  823.         const int src3= src[3 *srcStride];\

  824.         const int src4= src[4 *srcStride];\

  825.         OP(dst[0*dstStride], (src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));\

  826.         OP(dst[1*dstStride], (src1+src2)*20 - (src0+src3)*5 + (srcA+src4));\

  827.         dst++;\

  828.         src++;\

  829.     }\

  830. }\

  831. \

  832. static av_unused void FUNC(OPNAME ## h264_qpel2_hv_lowpass)(uint8_t *_dst, int16_t *tmp, uint8_t *_src, int dstStride, int tmpStride, int srcStride){\

  833.     const int h=2;\

  834.     const int w=2;\

  835.     const int pad = (BIT_DEPTH > 9) ? (-10 * ((1<<BIT_DEPTH)-1)) : 0;\

  836.     INIT_CLIP\

  837.     int i;\

  838.     pixel *dst = (pixel*)_dst;\

  839.     pixel *src = (pixel*)_src;\

  840.     dstStride /= sizeof(pixel);\

  841.     srcStride /= sizeof(pixel);\

  842.     src -= 2*srcStride;\

  843.     for(i=0; i<h+5; i++)\

  844.     {\

  845.         tmp[0]= (src[0]+src[1])*20 - (src[-1]+src[2])*5 + (src[-2]+src[3]) + pad;\

  846.         tmp[1]= (src[1]+src[2])*20 - (src[0 ]+src[3])*5 + (src[-1]+src[4]) + pad;\

  847.         tmp+=tmpStride;\

  848.         src+=srcStride;\

  849.     }\

  850.     tmp -= tmpStride*(h+5-2);\

  851.     for(i=0; i<w; i++)\

  852.     {\

  853.         const int tmpB= tmp[-2*tmpStride] - pad;\

  854.         const int tmpA= tmp[-1*tmpStride] - pad;\

  855.         const int tmp0= tmp[0 *tmpStride] - pad;\

  856.         const int tmp1= tmp[1 *tmpStride] - pad;\

  857.         const int tmp2= tmp[2 *tmpStride] - pad;\

  858.         const int tmp3= tmp[3 *tmpStride] - pad;\

  859.         const int tmp4= tmp[4 *tmpStride] - pad;\

  860.         OP2(dst[0*dstStride], (tmp0+tmp1)*20 - (tmpA+tmp2)*5 + (tmpB+tmp3));\

  861.         OP2(dst[1*dstStride], (tmp1+tmp2)*20 - (tmp0+tmp3)*5 + (tmpA+tmp4));\

  862.         dst++;\

  863.         tmp++;\

  864.     }\

  865. }\

  866. static void FUNC(OPNAME ## h264_qpel4_h_lowpass)(uint8_t *_dst, uint8_t *_src, int dstStride, int srcStride){\

  867.     const int h=4;\

  868.     INIT_CLIP\

  869.     int i;\

  870.     pixel *dst = (pixel*)_dst;\

  871.     pixel *src = (pixel*)_src;\

  872.     dstStride /= sizeof(pixel);\

  873.     srcStride /= sizeof(pixel);\

  874.     for(i=0; i<h; i++)\

  875.     {\

  876.         OP(dst[0], (src[0]+src[1])*20 - (src[-1]+src[2])*5 + (src[-2]+src[3]));\

  877.         OP(dst[1], (src[1]+src[2])*20 - (src[0 ]+src[3])*5 + (src[-1]+src[4]));\

  878.         OP(dst[2], (src[2]+src[3])*20 - (src[1 ]+src[4])*5 + (src[0 ]+src[5]));\

  879.         OP(dst[3], (src[3]+src[4])*20 - (src[2 ]+src[5])*5 + (src[1 ]+src[6]));\

  880.         dst+=dstStride;\

  881.         src+=srcStride;\

  882.     }\

  883. }\

  884. \

  885. static void FUNC(OPNAME ## h264_qpel4_v_lowpass)(uint8_t *_dst, uint8_t *_src, int dstStride, int srcStride){\

  886.     const int w=4;\

  887.     INIT_CLIP\

  888.     int i;\

  889.     pixel *dst = (pixel*)_dst;\

  890.     pixel *src = (pixel*)_src;\

  891.     dstStride /= sizeof(pixel);\

  892.     srcStride /= sizeof(pixel);\

  893.     for(i=0; i<w; i++)\

  894.     {\

  895.         const int srcB= src[-2*srcStride];\

  896.         const int srcA= src[-1*srcStride];\

  897.         const int src0= src[0 *srcStride];\

  898.         const int src1= src[1 *srcStride];\

  899.         const int src2= src[2 *srcStride];\

  900.         const int src3= src[3 *srcStride];\

  901.         const int src4= src[4 *srcStride];\

  902.         const int src5= src[5 *srcStride];\

  903.         const int src6= src[6 *srcStride];\

  904.         OP(dst[0*dstStride], (src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));\

  905.         OP(dst[1*dstStride], (src1+src2)*20 - (src0+src3)*5 + (srcA+src4));\

  906.         OP(dst[2*dstStride], (src2+src3)*20 - (src1+src4)*5 + (src0+src5));\

  907.         OP(dst[3*dstStride], (src3+src4)*20 - (src2+src5)*5 + (src1+src6));\

  908.         dst++;\

  909.         src++;\

  910.     }\

  911. }\

  912. \

  913. static void FUNC(OPNAME ## h264_qpel4_hv_lowpass)(uint8_t *_dst, int16_t *tmp, uint8_t *_src, int dstStride, int tmpStride, int srcStride){\

  914.     const int h=4;\

  915.     const int w=4;\

  916.     const int pad = (BIT_DEPTH > 9) ? (-10 * ((1<<BIT_DEPTH)-1)) : 0;\

  917.     INIT_CLIP\

  918.     int i;\

  919.     pixel *dst = (pixel*)_dst;\

  920.     pixel *src = (pixel*)_src;\

  921.     dstStride /= sizeof(pixel);\

  922.     srcStride /= sizeof(pixel);\

  923.     src -= 2*srcStride;\

  924.     for(i=0; i<h+5; i++)\

  925.     {\

  926.         tmp[0]= (src[0]+src[1])*20 - (src[-1]+src[2])*5 + (src[-2]+src[3]) + pad;\

  927.         tmp[1]= (src[1]+src[2])*20 - (src[0 ]+src[3])*5 + (src[-1]+src[4]) + pad;\

  928.         tmp[2]= (src[2]+src[3])*20 - (src[1 ]+src[4])*5 + (src[0 ]+src[5]) + pad;\

  929.         tmp[3]= (src[3]+src[4])*20 - (src[2 ]+src[5])*5 + (src[1 ]+src[6]) + pad;\

  930.         tmp+=tmpStride;\

  931.         src+=srcStride;\

  932.     }\

  933.     tmp -= tmpStride*(h+5-2);\

  934.     for(i=0; i<w; i++)\

  935.     {\

  936.         const int tmpB= tmp[-2*tmpStride] - pad;\

  937.         const int tmpA= tmp[-1*tmpStride] - pad;\

  938.         const int tmp0= tmp[0 *tmpStride] - pad;\

  939.         const int tmp1= tmp[1 *tmpStride] - pad;\

  940.         const int tmp2= tmp[2 *tmpStride] - pad;\

  941.         const int tmp3= tmp[3 *tmpStride] - pad;\

  942.         const int tmp4= tmp[4 *tmpStride] - pad;\

  943.         const int tmp5= tmp[5 *tmpStride] - pad;\

  944.         const int tmp6= tmp[6 *tmpStride] - pad;\

  945.         OP2(dst[0*dstStride], (tmp0+tmp1)*20 - (tmpA+tmp2)*5 + (tmpB+tmp3));\

  946.         OP2(dst[1*dstStride], (tmp1+tmp2)*20 - (tmp0+tmp3)*5 + (tmpA+tmp4));\

  947.         OP2(dst[2*dstStride], (tmp2+tmp3)*20 - (tmp1+tmp4)*5 + (tmp0+tmp5));\

  948.         OP2(dst[3*dstStride], (tmp3+tmp4)*20 - (tmp2+tmp5)*5 + (tmp1+tmp6));\

  949.         dst++;\

  950.         tmp++;\

  951.     }\

  952. }\

  953. \

  954. static void FUNC(OPNAME ## h264_qpel8_h_lowpass)(uint8_t *_dst, uint8_t *_src, int dstStride, int srcStride){\

  955.     const int h=8;\

  956.     INIT_CLIP\

  957.     int i;\

  958.     pixel *dst = (pixel*)_dst;\

  959.     pixel *src = (pixel*)_src;\

  960.     dstStride /= sizeof(pixel);\

  961.     srcStride /= sizeof(pixel);\

  962.     for(i=0; i<h; i++)\

  963.     {\

  964.         OP(dst[0], (src[0]+src[1])*20 - (src[-1]+src[2])*5 + (src[-2]+src[3 ]));\

  965.         OP(dst[1], (src[1]+src[2])*20 - (src[0 ]+src[3])*5 + (src[-1]+src[4 ]));\

  966.         OP(dst[2], (src[2]+src[3])*20 - (src[1 ]+src[4])*5 + (src[0 ]+src[5 ]));\

  967.         OP(dst[3], (src[3]+src[4])*20 - (src[2 ]+src[5])*5 + (src[1 ]+src[6 ]));\

  968.         OP(dst[4], (src[4]+src[5])*20 - (src[3 ]+src[6])*5 + (src[2 ]+src[7 ]));\

  969.         OP(dst[5], (src[5]+src[6])*20 - (src[4 ]+src[7])*5 + (src[3 ]+src[8 ]));\

  970.         OP(dst[6], (src[6]+src[7])*20 - (src[5 ]+src[8])*5 + (src[4 ]+src[9 ]));\

  971.         OP(dst[7], (src[7]+src[8])*20 - (src[6 ]+src[9])*5 + (src[5 ]+src[10]));\

  972.         dst+=dstStride;\

  973.         src+=srcStride;\

  974.     }\

  975. }\

  976. \

  977. static void FUNC(OPNAME ## h264_qpel8_v_lowpass)(uint8_t *_dst, uint8_t *_src, int dstStride, int srcStride){\

  978.     const int w=8;\

  979.     INIT_CLIP\

  980.     int i;\

  981.     pixel *dst = (pixel*)_dst;\

  982.     pixel *src = (pixel*)_src;\

  983.     dstStride /= sizeof(pixel);\

  984.     srcStride /= sizeof(pixel);\

  985.     for(i=0; i<w; i++)\

  986.     {\

  987.         const int srcB= src[-2*srcStride];\

  988.         const int srcA= src[-1*srcStride];\

  989.         const int src0= src[0 *srcStride];\

  990.         const int src1= src[1 *srcStride];\

  991.         const int src2= src[2 *srcStride];\

  992.         const int src3= src[3 *srcStride];\

  993.         const int src4= src[4 *srcStride];\

  994.         const int src5= src[5 *srcStride];\

  995.         const int src6= src[6 *srcStride];\

  996.         const int src7= src[7 *srcStride];\

  997.         const int src8= src[8 *srcStride];\

  998.         const int src9= src[9 *srcStride];\

  999.         const int src10=src[10*srcStride];\

  1000.         OP(dst[0*dstStride], (src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));\

  1001.         OP(dst[1*dstStride], (src1+src2)*20 - (src0+src3)*5 + (srcA+src4));\

  1002.         OP(dst[2*dstStride], (src2+src3)*20 - (src1+src4)*5 + (src0+src5));\

  1003.         OP(dst[3*dstStride], (src3+src4)*20 - (src2+src5)*5 + (src1+src6));\

  1004.         OP(dst[4*dstStride], (src4+src5)*20 - (src3+src6)*5 + (src2+src7));\

  1005.         OP(dst[5*dstStride], (src5+src6)*20 - (src4+src7)*5 + (src3+src8));\

  1006.         OP(dst[6*dstStride], (src6+src7)*20 - (src5+src8)*5 + (src4+src9));\

  1007.         OP(dst[7*dstStride], (src7+src8)*20 - (src6+src9)*5 + (src5+src10));\

  1008.         dst++;\

  1009.         src++;\

  1010.     }\

  1011. }\

  1012. \

  1013. static void FUNC(OPNAME ## h264_qpel8_hv_lowpass)(uint8_t *_dst, int16_t *tmp, uint8_t *_src, int dstStride, int tmpStride, int srcStride){\

  1014.     const int h=8;\

  1015.     const int w=8;\

  1016.     const int pad = (BIT_DEPTH > 9) ? (-10 * ((1<<BIT_DEPTH)-1)) : 0;\

  1017.     INIT_CLIP\

  1018.     int i;\

  1019.     pixel *dst = (pixel*)_dst;\

  1020.     pixel *src = (pixel*)_src;\

  1021.     dstStride /= sizeof(pixel);\

  1022.     srcStride /= sizeof(pixel);\

  1023.     src -= 2*srcStride;\

  1024.     for(i=0; i<h+5; i++)\

  1025.     {\

  1026.         tmp[0]= (src[0]+src[1])*20 - (src[-1]+src[2])*5 + (src[-2]+src[3 ]) + pad;\

  1027.         tmp[1]= (src[1]+src[2])*20 - (src[0 ]+src[3])*5 + (src[-1]+src[4 ]) + pad;\

  1028.         tmp[2]= (src[2]+src[3])*20 - (src[1 ]+src[4])*5 + (src[0 ]+src[5 ]) + pad;\

  1029.         tmp[3]= (src[3]+src[4])*20 - (src[2 ]+src[5])*5 + (src[1 ]+src[6 ]) + pad;\

  1030.         tmp[4]= (src[4]+src[5])*20 - (src[3 ]+src[6])*5 + (src[2 ]+src[7 ]) + pad;\

  1031.         tmp[5]= (src[5]+src[6])*20 - (src[4 ]+src[7])*5 + (src[3 ]+src[8 ]) + pad;\

  1032.         tmp[6]= (src[6]+src[7])*20 - (src[5 ]+src[8])*5 + (src[4 ]+src[9 ]) + pad;\

  1033.         tmp[7]= (src[7]+src[8])*20 - (src[6 ]+src[9])*5 + (src[5 ]+src[10]) + pad;\

  1034.         tmp+=tmpStride;\

  1035.         src+=srcStride;\

  1036.     }\

  1037.     tmp -= tmpStride*(h+5-2);\

  1038.     for(i=0; i<w; i++)\

  1039.     {\

  1040.         const int tmpB= tmp[-2*tmpStride] - pad;\

  1041.         const int tmpA= tmp[-1*tmpStride] - pad;\

  1042.         const int tmp0= tmp[0 *tmpStride] - pad;\

  1043.         const int tmp1= tmp[1 *tmpStride] - pad;\

  1044.         const int tmp2= tmp[2 *tmpStride] - pad;\

  1045.         const int tmp3= tmp[3 *tmpStride] - pad;\

  1046.         const int tmp4= tmp[4 *tmpStride] - pad;\

  1047.         const int tmp5= tmp[5 *tmpStride] - pad;\

  1048.         const int tmp6= tmp[6 *tmpStride] - pad;\

  1049.         const int tmp7= tmp[7 *tmpStride] - pad;\

  1050.         const int tmp8= tmp[8 *tmpStride] - pad;\

  1051.         const int tmp9= tmp[9 *tmpStride] - pad;\

  1052.         const int tmp10=tmp[10*tmpStride] - pad;\

  1053.         OP2(dst[0*dstStride], (tmp0+tmp1)*20 - (tmpA+tmp2)*5 + (tmpB+tmp3));\

  1054.         OP2(dst[1*dstStride], (tmp1+tmp2)*20 - (tmp0+tmp3)*5 + (tmpA+tmp4));\

  1055.         OP2(dst[2*dstStride], (tmp2+tmp3)*20 - (tmp1+tmp4)*5 + (tmp0+tmp5));\

  1056.         OP2(dst[3*dstStride], (tmp3+tmp4)*20 - (tmp2+tmp5)*5 + (tmp1+tmp6));\

  1057.         OP2(dst[4*dstStride], (tmp4+tmp5)*20 - (tmp3+tmp6)*5 + (tmp2+tmp7));\

  1058.         OP2(dst[5*dstStride], (tmp5+tmp6)*20 - (tmp4+tmp7)*5 + (tmp3+tmp8));\

  1059.         OP2(dst[6*dstStride], (tmp6+tmp7)*20 - (tmp5+tmp8)*5 + (tmp4+tmp9));\

  1060.         OP2(dst[7*dstStride], (tmp7+tmp8)*20 - (tmp6+tmp9)*5 + (tmp5+tmp10));\

  1061.         dst++;\

  1062.         tmp++;\

  1063.     }\

  1064. }\

  1065. \

  1066. static void FUNC(OPNAME ## h264_qpel16_v_lowpass)(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1067.     FUNC(OPNAME ## h264_qpel8_v_lowpass)(dst                , src                , dstStride, srcStride);\

  1068.     FUNC(OPNAME ## h264_qpel8_v_lowpass)(dst+8*sizeof(pixel), src+8*sizeof(pixel), dstStride, srcStride);\

  1069.     src += 8*srcStride;\

  1070.     dst += 8*dstStride;\

  1071.     FUNC(OPNAME ## h264_qpel8_v_lowpass)(dst                , src                , dstStride, srcStride);\

  1072.     FUNC(OPNAME ## h264_qpel8_v_lowpass)(dst+8*sizeof(pixel), src+8*sizeof(pixel), dstStride, srcStride);\

  1073. }\

  1074. \

  1075. static void FUNC(OPNAME ## h264_qpel16_h_lowpass)(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1076.     FUNC(OPNAME ## h264_qpel8_h_lowpass)(dst                , src                , dstStride, srcStride);\

  1077.     FUNC(OPNAME ## h264_qpel8_h_lowpass)(dst+8*sizeof(pixel), src+8*sizeof(pixel), dstStride, srcStride);\

  1078.     src += 8*srcStride;\

  1079.     dst += 8*dstStride;\

  1080.     FUNC(OPNAME ## h264_qpel8_h_lowpass)(dst                , src                , dstStride, srcStride);\

  1081.     FUNC(OPNAME ## h264_qpel8_h_lowpass)(dst+8*sizeof(pixel), src+8*sizeof(pixel), dstStride, srcStride);\

  1082. }\

  1083. \

  1084. static void FUNC(OPNAME ## h264_qpel16_hv_lowpass)(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\

  1085.     FUNC(OPNAME ## h264_qpel8_hv_lowpass)(dst                , tmp  , src                , dstStride, tmpStride, srcStride);\

  1086.     FUNC(OPNAME ## h264_qpel8_hv_lowpass)(dst+8*sizeof(pixel), tmp+8, src+8*sizeof(pixel), dstStride, tmpStride, srcStride);\

  1087.     src += 8*srcStride;\

  1088.     dst += 8*dstStride;\

  1089.     FUNC(OPNAME ## h264_qpel8_hv_lowpass)(dst                , tmp  , src                , dstStride, tmpStride, srcStride);\

  1090.     FUNC(OPNAME ## h264_qpel8_hv_lowpass)(dst+8*sizeof(pixel), tmp+8, src+8*sizeof(pixel), dstStride, tmpStride, srcStride);\

  1091. }\

  1092. 

  1093. #define H264_MC(OPNAME, SIZE) \

  1094. static av_unused void FUNCC(OPNAME ## h264_qpel ## SIZE ## _mc00)(uint8_t *dst, uint8_t *src, int stride){\

  1095.     FUNCC(OPNAME ## pixels ## SIZE)(dst, src, stride, SIZE);\

  1096. }\

  1097. \

  1098. static void FUNCC(OPNAME ## h264_qpel ## SIZE ## _mc10)(uint8_t *dst, uint8_t *src, int stride){\

  1099.     uint8_t half[SIZE*SIZE*sizeof(pixel)];\

  1100.     FUNC(put_h264_qpel ## SIZE ## _h_lowpass)(half, src, SIZE*sizeof(pixel), stride);\

  1101.     FUNC(OPNAME ## pixels ## SIZE ## _l2)(dst, src, half, stride, stride, SIZE*sizeof(pixel), SIZE);\

  1102. }\

  1103. \

  1104. static void FUNCC(OPNAME ## h264_qpel ## SIZE ## _mc20)(uint8_t *dst, uint8_t *src, int stride){\

  1105.     FUNC(OPNAME ## h264_qpel ## SIZE ## _h_lowpass)(dst, src, stride, stride);\

  1106. }\

  1107. \

  1108. static void FUNCC(OPNAME ## h264_qpel ## SIZE ## _mc30)(uint8_t *dst, uint8_t *src, int stride){\

  1109.     uint8_t half[SIZE*SIZE*sizeof(pixel)];\

  1110.     FUNC(put_h264_qpel ## SIZE ## _h_lowpass)(half, src, SIZE*sizeof(pixel), stride);\

  1111.     FUNC(OPNAME ## pixels ## SIZE ## _l2)(dst, src+sizeof(pixel), half, stride, stride, SIZE*sizeof(pixel), SIZE);\

  1112. }\

  1113. \

  1114. static void FUNCC(OPNAME ## h264_qpel ## SIZE ## _mc01)(uint8_t *dst, uint8_t *src, int stride){\

  1115.     uint8_t full[SIZE*(SIZE+5)*sizeof(pixel)];\

  1116.     uint8_t * const full_mid= full + SIZE*2*sizeof(pixel);\

  1117.     uint8_t half[SIZE*SIZE*sizeof(pixel)];\

  1118.     FUNC(copy_block ## SIZE )(full, src - stride*2, SIZE*sizeof(pixel),  stride, SIZE + 5);\

  1119.     FUNC(put_h264_qpel ## SIZE ## _v_lowpass)(half, full_mid, SIZE*sizeof(pixel), SIZE*sizeof(pixel));\

  1120.     FUNC(OPNAME ## pixels ## SIZE ## _l2)(dst, full_mid, half, stride, SIZE*sizeof(pixel), SIZE*sizeof(pixel), SIZE);\

  1121. }\

  1122. \

  1123. static void FUNCC(OPNAME ## h264_qpel ## SIZE ## _mc02)(uint8_t *dst, uint8_t *src, int stride){\

  1124.     uint8_t full[SIZE*(SIZE+5)*sizeof(pixel)];\

  1125.     uint8_t * const full_mid= full + SIZE*2*sizeof(pixel);\

  1126.     FUNC(copy_block ## SIZE )(full, src - stride*2, SIZE*sizeof(pixel),  stride, SIZE + 5);\

  1127.     FUNC(OPNAME ## h264_qpel ## SIZE ## _v_lowpass)(dst, full_mid, stride, SIZE*sizeof(pixel));\

  1128. }\

  1129. \

  1130. static void FUNCC(OPNAME ## h264_qpel ## SIZE ## _mc03)(uint8_t *dst, uint8_t *src, int stride){\

  1131.     uint8_t full[SIZE*(SIZE+5)*sizeof(pixel)];\

  1132.     uint8_t * const full_mid= full + SIZE*2*sizeof(pixel);\

  1133.     uint8_t half[SIZE*SIZE*sizeof(pixel)];\

  1134.     FUNC(copy_block ## SIZE )(full, src - stride*2, SIZE*sizeof(pixel),  stride, SIZE + 5);\

  1135.     FUNC(put_h264_qpel ## SIZE ## _v_lowpass)(half, full_mid, SIZE*sizeof(pixel), SIZE*sizeof(pixel));\

  1136.     FUNC(OPNAME ## pixels ## SIZE ## _l2)(dst, full_mid+SIZE*sizeof(pixel), half, stride, SIZE*sizeof(pixel), SIZE*sizeof(pixel), SIZE);\

  1137. }\

  1138. \

  1139. static void FUNCC(OPNAME ## h264_qpel ## SIZE ## _mc11)(uint8_t *dst, uint8_t *src, int stride){\

  1140.     uint8_t full[SIZE*(SIZE+5)*sizeof(pixel)];\

  1141.     uint8_t * const full_mid= full + SIZE*2*sizeof(pixel);\

  1142.     uint8_t halfH[SIZE*SIZE*sizeof(pixel)];\

  1143.     uint8_t halfV[SIZE*SIZE*sizeof(pixel)];\

  1144.     FUNC(put_h264_qpel ## SIZE ## _h_lowpass)(halfH, src, SIZE*sizeof(pixel), stride);\

  1145.     FUNC(copy_block ## SIZE )(full, src - stride*2, SIZE*sizeof(pixel),  stride, SIZE + 5);\

  1146.     FUNC(put_h264_qpel ## SIZE ## _v_lowpass)(halfV, full_mid, SIZE*sizeof(pixel), SIZE*sizeof(pixel));\

  1147.     FUNC(OPNAME ## pixels ## SIZE ## _l2)(dst, halfH, halfV, stride, SIZE*sizeof(pixel), SIZE*sizeof(pixel), SIZE);\

  1148. }\

  1149. \

  1150. static void FUNCC(OPNAME ## h264_qpel ## SIZE ## _mc31)(uint8_t *dst, uint8_t *src, int stride){\

  1151.     uint8_t full[SIZE*(SIZE+5)*sizeof(pixel)];\

  1152.     uint8_t * const full_mid= full + SIZE*2*sizeof(pixel);\

  1153.     uint8_t halfH[SIZE*SIZE*sizeof(pixel)];\

  1154.     uint8_t halfV[SIZE*SIZE*sizeof(pixel)];\

  1155.     FUNC(put_h264_qpel ## SIZE ## _h_lowpass)(halfH, src, SIZE*sizeof(pixel), stride);\

  1156.     FUNC(copy_block ## SIZE )(full, src - stride*2 + sizeof(pixel), SIZE*sizeof(pixel),  stride, SIZE + 5);\

  1157.     FUNC(put_h264_qpel ## SIZE ## _v_lowpass)(halfV, full_mid, SIZE*sizeof(pixel), SIZE*sizeof(pixel));\

  1158.     FUNC(OPNAME ## pixels ## SIZE ## _l2)(dst, halfH, halfV, stride, SIZE*sizeof(pixel), SIZE*sizeof(pixel), SIZE);\

  1159. }\

  1160. \

  1161. static void FUNCC(OPNAME ## h264_qpel ## SIZE ## _mc13)(uint8_t *dst, uint8_t *src, int stride){\

  1162.     uint8_t full[SIZE*(SIZE+5)*sizeof(pixel)];\

  1163.     uint8_t * const full_mid= full + SIZE*2*sizeof(pixel);\

  1164.     uint8_t halfH[SIZE*SIZE*sizeof(pixel)];\

  1165.     uint8_t halfV[SIZE*SIZE*sizeof(pixel)];\

  1166.     FUNC(put_h264_qpel ## SIZE ## _h_lowpass)(halfH, src + stride, SIZE*sizeof(pixel), stride);\

  1167.     FUNC(copy_block ## SIZE )(full, src - stride*2, SIZE*sizeof(pixel),  stride, SIZE + 5);\

  1168.     FUNC(put_h264_qpel ## SIZE ## _v_lowpass)(halfV, full_mid, SIZE*sizeof(pixel), SIZE*sizeof(pixel));\

  1169.     FUNC(OPNAME ## pixels ## SIZE ## _l2)(dst, halfH, halfV, stride, SIZE*sizeof(pixel), SIZE*sizeof(pixel), SIZE);\

  1170. }\

  1171. \

  1172. static void FUNCC(OPNAME ## h264_qpel ## SIZE ## _mc33)(uint8_t *dst, uint8_t *src, int stride){\

  1173.     uint8_t full[SIZE*(SIZE+5)*sizeof(pixel)];\

  1174.     uint8_t * const full_mid= full + SIZE*2*sizeof(pixel);\

  1175.     uint8_t halfH[SIZE*SIZE*sizeof(pixel)];\

  1176.     uint8_t halfV[SIZE*SIZE*sizeof(pixel)];\

  1177.     FUNC(put_h264_qpel ## SIZE ## _h_lowpass)(halfH, src + stride, SIZE*sizeof(pixel), stride);\

  1178.     FUNC(copy_block ## SIZE )(full, src - stride*2 + sizeof(pixel), SIZE*sizeof(pixel),  stride, SIZE + 5);\

  1179.     FUNC(put_h264_qpel ## SIZE ## _v_lowpass)(halfV, full_mid, SIZE*sizeof(pixel), SIZE*sizeof(pixel));\

  1180.     FUNC(OPNAME ## pixels ## SIZE ## _l2)(dst, halfH, halfV, stride, SIZE*sizeof(pixel), SIZE*sizeof(pixel), SIZE);\

  1181. }\

  1182. \

  1183. static void FUNCC(OPNAME ## h264_qpel ## SIZE ## _mc22)(uint8_t *dst, uint8_t *src, int stride){\

  1184.     int16_t tmp[SIZE*(SIZE+5)*sizeof(pixel)];\

  1185.     FUNC(OPNAME ## h264_qpel ## SIZE ## _hv_lowpass)(dst, tmp, src, stride, SIZE*sizeof(pixel), stride);\

  1186. }\

  1187. \

  1188. static void FUNCC(OPNAME ## h264_qpel ## SIZE ## _mc21)(uint8_t *dst, uint8_t *src, int stride){\

  1189.     int16_t tmp[SIZE*(SIZE+5)*sizeof(pixel)];\

  1190.     uint8_t halfH[SIZE*SIZE*sizeof(pixel)];\

  1191.     uint8_t halfHV[SIZE*SIZE*sizeof(pixel)];\

  1192.     FUNC(put_h264_qpel ## SIZE ## _h_lowpass)(halfH, src, SIZE*sizeof(pixel), stride);\

  1193.     FUNC(put_h264_qpel ## SIZE ## _hv_lowpass)(halfHV, tmp, src, SIZE*sizeof(pixel), SIZE*sizeof(pixel), stride);\

  1194.     FUNC(OPNAME ## pixels ## SIZE ## _l2)(dst, halfH, halfHV, stride, SIZE*sizeof(pixel), SIZE*sizeof(pixel), SIZE);\

  1195. }\

  1196. \

  1197. static void FUNCC(OPNAME ## h264_qpel ## SIZE ## _mc23)(uint8_t *dst, uint8_t *src, int stride){\

  1198.     int16_t tmp[SIZE*(SIZE+5)*sizeof(pixel)];\

  1199.     uint8_t halfH[SIZE*SIZE*sizeof(pixel)];\

  1200.     uint8_t halfHV[SIZE*SIZE*sizeof(pixel)];\

  1201.     FUNC(put_h264_qpel ## SIZE ## _h_lowpass)(halfH, src + stride, SIZE*sizeof(pixel), stride);\

  1202.     FUNC(put_h264_qpel ## SIZE ## _hv_lowpass)(halfHV, tmp, src, SIZE*sizeof(pixel), SIZE*sizeof(pixel), stride);\

  1203.     FUNC(OPNAME ## pixels ## SIZE ## _l2)(dst, halfH, halfHV, stride, SIZE*sizeof(pixel), SIZE*sizeof(pixel), SIZE);\

  1204. }\

  1205. \

  1206. static void FUNCC(OPNAME ## h264_qpel ## SIZE ## _mc12)(uint8_t *dst, uint8_t *src, int stride){\

  1207.     uint8_t full[SIZE*(SIZE+5)*sizeof(pixel)];\

  1208.     uint8_t * const full_mid= full + SIZE*2*sizeof(pixel);\

  1209.     int16_t tmp[SIZE*(SIZE+5)*sizeof(pixel)];\

  1210.     uint8_t halfV[SIZE*SIZE*sizeof(pixel)];\

  1211.     uint8_t halfHV[SIZE*SIZE*sizeof(pixel)];\

  1212.     FUNC(copy_block ## SIZE )(full, src - stride*2, SIZE*sizeof(pixel),  stride, SIZE + 5);\

  1213.     FUNC(put_h264_qpel ## SIZE ## _v_lowpass)(halfV, full_mid, SIZE*sizeof(pixel), SIZE*sizeof(pixel));\

  1214.     FUNC(put_h264_qpel ## SIZE ## _hv_lowpass)(halfHV, tmp, src, SIZE*sizeof(pixel), SIZE*sizeof(pixel), stride);\

  1215.     FUNC(OPNAME ## pixels ## SIZE ## _l2)(dst, halfV, halfHV, stride, SIZE*sizeof(pixel), SIZE*sizeof(pixel), SIZE);\

  1216. }\

  1217. \

  1218. static void FUNCC(OPNAME ## h264_qpel ## SIZE ## _mc32)(uint8_t *dst, uint8_t *src, int stride){\

  1219.     uint8_t full[SIZE*(SIZE+5)*sizeof(pixel)];\

  1220.     uint8_t * const full_mid= full + SIZE*2*sizeof(pixel);\

  1221.     int16_t tmp[SIZE*(SIZE+5)*sizeof(pixel)];\

  1222.     uint8_t halfV[SIZE*SIZE*sizeof(pixel)];\

  1223.     uint8_t halfHV[SIZE*SIZE*sizeof(pixel)];\

  1224.     FUNC(copy_block ## SIZE )(full, src - stride*2 + sizeof(pixel), SIZE*sizeof(pixel),  stride, SIZE + 5);\

  1225.     FUNC(put_h264_qpel ## SIZE ## _v_lowpass)(halfV, full_mid, SIZE*sizeof(pixel), SIZE*sizeof(pixel));\

  1226.     FUNC(put_h264_qpel ## SIZE ## _hv_lowpass)(halfHV, tmp, src, SIZE*sizeof(pixel), SIZE*sizeof(pixel), stride);\

  1227.     FUNC(OPNAME ## pixels ## SIZE ## _l2)(dst, halfV, halfHV, stride, SIZE*sizeof(pixel), SIZE*sizeof(pixel), SIZE);\

  1228. }\

  1229. 

  1230. #define op_avg(a, b)  a = (((a)+CLIP(((b) + 16)>>5)+1)>>1)

  1231. //#define op_avg2(a, b) a = (((a)*w1+cm[((b) + 16)>>5]*w2 + o + 64)>>7)

  1232. #define op_put(a, b)  a = CLIP(((b) + 16)>>5)

  1233. #define op2_avg(a, b)  a = (((a)+CLIP(((b) + 512)>>10)+1)>>1)

  1234. #define op2_put(a, b)  a = CLIP(((b) + 512)>>10)

  1235. 

  1236. H264_LOWPASS(put_       , op_put, op2_put)

  1237. H264_LOWPASS(avg_       , op_avg, op2_avg)

  1238. H264_MC(put_, 2)

  1239. H264_MC(put_, 4)

  1240. H264_MC(put_, 8)

  1241. H264_MC(put_, 16)

  1242. H264_MC(avg_, 4)

  1243. H264_MC(avg_, 8)

  1244. H264_MC(avg_, 16)

  1245. 

  1246. #undef op_avg

  1247. #undef op_put

  1248. #undef op2_avg

  1249. #undef op2_put

  1250. 

  1251. #if BIT_DEPTH == 8

  1252. #   define put_h264_qpel8_mc00_8_c  ff_put_pixels8x8_8_c

  1253. #   define avg_h264_qpel8_mc00_8_c  ff_avg_pixels8x8_8_c

  1254. #   define put_h264_qpel16_mc00_8_c ff_put_pixels16x16_8_c

  1255. #   define avg_h264_qpel16_mc00_8_c ff_avg_pixels16x16_8_c

  1256. #elif BIT_DEPTH == 9

  1257. #   define put_h264_qpel8_mc00_9_c  ff_put_pixels8x8_9_c

  1258. #   define avg_h264_qpel8_mc00_9_c  ff_avg_pixels8x8_9_c

  1259. #   define put_h264_qpel16_mc00_9_c ff_put_pixels16x16_9_c

  1260. #   define avg_h264_qpel16_mc00_9_c ff_avg_pixels16x16_9_c

  1261. #elif BIT_DEPTH == 10

  1262. #   define put_h264_qpel8_mc00_10_c  ff_put_pixels8x8_10_c

  1263. #   define avg_h264_qpel8_mc00_10_c  ff_avg_pixels8x8_10_c

  1264. #   define put_h264_qpel16_mc00_10_c ff_put_pixels16x16_10_c

  1265. #   define avg_h264_qpel16_mc00_10_c ff_avg_pixels16x16_10_c

  1266. #endif

  1267. 

  1268. void FUNCC(ff_put_pixels8x8)(uint8_t *dst, uint8_t *src, int stride) {

  1269.     FUNCC(put_pixels8)(dst, src, stride, 8);

  1270. }

  1271. void FUNCC(ff_avg_pixels8x8)(uint8_t *dst, uint8_t *src, int stride) {

  1272.     FUNCC(avg_pixels8)(dst, src, stride, 8);

  1273. }

  1274. void FUNCC(ff_put_pixels16x16)(uint8_t *dst, uint8_t *src, int stride) {

  1275.     FUNCC(put_pixels16)(dst, src, stride, 16);

  1276. }

  1277. void FUNCC(ff_avg_pixels16x16)(uint8_t *dst, uint8_t *src, int stride) {

  1278.     FUNCC(avg_pixels16)(dst, src, stride, 16);

  1279. }