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.
6808 Commits
32 Branches
390MB
Tree: 33592a633f
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '33592a633f'
${ noResults }
FFmpeg/libavcodec/sh4/qpel.c

1666 lines
70KB
Raw Blame History 

  1. /*

  2.  * This is optimized for sh, which have post increment addressing (*p++).

  3.  * Some CPU may be index (p[n]) faster than post increment (*p++).

  4.  *

  5.  * copyright (c) 2001-2003 BERO <bero@geocities.co.jp>

  6.  *

  7.  * This library 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 of the License, or (at your option) any later version.

  11.  *

  12.  * This library 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 this library; if not, write to the Free Software

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

  20.  */

  21. 

  22. #define         LD(adr) *(uint32_t*)(adr)

  23. 

  24. #define PIXOP2(OPNAME, OP) \

  25. /*static inline void OPNAME ## _no_rnd_pixels8_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \

  26. {\

  27.         do {\

  28.                 OP(LP(dst  ),no_rnd_avg32(LD32(src1  ),LD32(src2  )) ); \

  29.                 OP(LP(dst+4),no_rnd_avg32(LD32(src1+4),LD32(src2+4)) ); \

  30.                 src1+=src_stride1; \

  31.                 src2+=src_stride2; \

  32.                 dst+=dst_stride; \

  33.         } while(--h); \

  34. }\

  35. \

  36. static inline void OPNAME ## _pixels8_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \

  37. {\

  38.         do {\

  39.                 OP(LP(dst  ),rnd_avg32(LD32(src1  ),LD32(src2  )) ); \

  40.                 OP(LP(dst+4),rnd_avg32(LD32(src1+4),LD32(src2+4)) ); \

  41.                 src1+=src_stride1; \

  42.                 src2+=src_stride2; \

  43.                 dst+=dst_stride; \

  44.         } while(--h); \

  45. }\

  46. \

  47. static inline void OPNAME ## _pixels4_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \

  48. {\

  49.         do {\

  50.                 OP(LP(dst  ),rnd_avg32(LD32(src1  ),LD32(src2  )) ); \

  51.                 src1+=src_stride1; \

  52.                 src2+=src_stride2; \

  53.                 dst+=dst_stride; \

  54.         } while(--h); \

  55. }\

  56. \

  57. static inline void OPNAME ## _no_rnd_pixels16_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \

  58. {\

  59.         do {\

  60.                 OP(LP(dst  ),no_rnd_avg32(LD32(src1  ),LD32(src2  )) ); \

  61.                 OP(LP(dst+4),no_rnd_avg32(LD32(src1+4),LD32(src2+4)) ); \

  62.                 OP(LP(dst+8),no_rnd_avg32(LD32(src1+8),LD32(src2+8)) ); \

  63.                 OP(LP(dst+12),no_rnd_avg32(LD32(src1+12),LD32(src2+12)) ); \

  64.                 src1+=src_stride1; \

  65.                 src2+=src_stride2; \

  66.                 dst+=dst_stride; \

  67.         } while(--h); \

  68. }\

  69. \

  70. static inline void OPNAME ## _pixels16_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \

  71. {\

  72.         do {\

  73.                 OP(LP(dst  ),rnd_avg32(LD32(src1  ),LD32(src2  )) ); \

  74.                 OP(LP(dst+4),rnd_avg32(LD32(src1+4),LD32(src2+4)) ); \

  75.                 OP(LP(dst+8),rnd_avg32(LD32(src1+8),LD32(src2+8)) ); \

  76.                 OP(LP(dst+12),rnd_avg32(LD32(src1+12),LD32(src2+12)) ); \

  77.                 src1+=src_stride1; \

  78.                 src2+=src_stride2; \

  79.                 dst+=dst_stride; \

  80.         } while(--h); \

  81. }*/\

  82. \

  83. static inline void OPNAME ## _pixels4_l2_aligned(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \

  84. {\

  85.         do {\

  86.                 OP(LP(dst  ),rnd_avg32(LP(src1  ),LP(src2  )) ); \

  87.                 src1+=src_stride1; \

  88.                 src2+=src_stride2; \

  89.                 dst+=dst_stride; \

  90.         } while(--h); \

  91. }\

  92. \

  93. static inline void OPNAME ## _pixels4_l2_aligned2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \

  94. {\

  95.         do {\

  96.                 OP(LP(dst  ),rnd_avg32(LD32(src1  ),LP(src2  )) ); \

  97.                 src1+=src_stride1; \

  98.                 src2+=src_stride2; \

  99.                 dst+=dst_stride; \

  100.         } while(--h); \

  101. }\

  102. \

  103. static inline void OPNAME ## _no_rnd_pixels16_l2_aligned2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \

  104. {\

  105.         do {\

  106.                 OP(LP(dst  ),no_rnd_avg32(LD32(src1  ),LP(src2  )) ); \

  107.                 OP(LP(dst+4),no_rnd_avg32(LD32(src1+4),LP(src2+4)) ); \

  108.                 OP(LP(dst+8),no_rnd_avg32(LD32(src1+8),LP(src2+8)) ); \

  109.                 OP(LP(dst+12),no_rnd_avg32(LD32(src1+12),LP(src2+12)) ); \

  110.                 src1+=src_stride1; \

  111.                 src2+=src_stride2; \

  112.                 dst+=dst_stride; \

  113.         } while(--h); \

  114. }\

  115. \

  116. static inline void OPNAME ## _pixels16_l2_aligned2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \

  117. {\

  118.         do {\

  119.                 OP(LP(dst  ),rnd_avg32(LD32(src1  ),LP(src2  )) ); \

  120.                 OP(LP(dst+4),rnd_avg32(LD32(src1+4),LP(src2+4)) ); \

  121.                 OP(LP(dst+8),rnd_avg32(LD32(src1+8),LP(src2+8)) ); \

  122.                 OP(LP(dst+12),rnd_avg32(LD32(src1+12),LP(src2+12)) ); \

  123.                 src1+=src_stride1; \

  124.                 src2+=src_stride2; \

  125.                 dst+=dst_stride; \

  126.         } while(--h); \

  127. }\

  128. \

  129. static inline void OPNAME ## _no_rnd_pixels8_l2_aligned2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \

  130. {\

  131.         do { /* onlye src2 aligned */\

  132.                 OP(LP(dst  ),no_rnd_avg32(LD32(src1  ),LP(src2  )) ); \

  133.                 OP(LP(dst+4),no_rnd_avg32(LD32(src1+4),LP(src2+4)) ); \

  134.                 src1+=src_stride1; \

  135.                 src2+=src_stride2; \

  136.                 dst+=dst_stride; \

  137.         } while(--h); \

  138. }\

  139. \

  140. static inline void OPNAME ## _pixels8_l2_aligned2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \

  141. {\

  142.         do {\

  143.                 OP(LP(dst  ),rnd_avg32(LD32(src1  ),LP(src2  )) ); \

  144.                 OP(LP(dst+4),rnd_avg32(LD32(src1+4),LP(src2+4)) ); \

  145.                 src1+=src_stride1; \

  146.                 src2+=src_stride2; \

  147.                 dst+=dst_stride; \

  148.         } while(--h); \

  149. }\

  150. \

  151. static inline void OPNAME ## _no_rnd_pixels8_l2_aligned(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \

  152. {\

  153.         do {\

  154.                 OP(LP(dst  ),no_rnd_avg32(LP(src1  ),LP(src2  )) ); \

  155.                 OP(LP(dst+4),no_rnd_avg32(LP(src1+4),LP(src2+4)) ); \

  156.                 src1+=src_stride1; \

  157.                 src2+=src_stride2; \

  158.                 dst+=dst_stride; \

  159.         } while(--h); \

  160. }\

  161. \

  162. static inline void OPNAME ## _pixels8_l2_aligned(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \

  163. {\

  164.         do {\

  165.                 OP(LP(dst  ),rnd_avg32(LP(src1  ),LP(src2  )) ); \

  166.                 OP(LP(dst+4),rnd_avg32(LP(src1+4),LP(src2+4)) ); \

  167.                 src1+=src_stride1; \

  168.                 src2+=src_stride2; \

  169.                 dst+=dst_stride; \

  170.         } while(--h); \

  171. }\

  172. \

  173. static inline void OPNAME ## _no_rnd_pixels16_l2_aligned(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \

  174. {\

  175.         do {\

  176.                 OP(LP(dst  ),no_rnd_avg32(LP(src1  ),LP(src2  )) ); \

  177.                 OP(LP(dst+4),no_rnd_avg32(LP(src1+4),LP(src2+4)) ); \

  178.                 OP(LP(dst+8),no_rnd_avg32(LP(src1+8),LP(src2+8)) ); \

  179.                 OP(LP(dst+12),no_rnd_avg32(LP(src1+12),LP(src2+12)) ); \

  180.                 src1+=src_stride1; \

  181.                 src2+=src_stride2; \

  182.                 dst+=dst_stride; \

  183.         } while(--h); \

  184. }\

  185. \

  186. static inline void OPNAME ## _pixels16_l2_aligned(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \

  187. {\

  188.         do {\

  189.                 OP(LP(dst  ),rnd_avg32(LP(src1  ),LP(src2  )) ); \

  190.                 OP(LP(dst+4),rnd_avg32(LP(src1+4),LP(src2+4)) ); \

  191.                 OP(LP(dst+8),rnd_avg32(LP(src1+8),LP(src2+8)) ); \

  192.                 OP(LP(dst+12),rnd_avg32(LP(src1+12),LP(src2+12)) ); \

  193.                 src1+=src_stride1; \

  194.                 src2+=src_stride2; \

  195.                 dst+=dst_stride; \

  196.         } while(--h); \

  197. }\

  198. \

  199. static inline void OPNAME ## _no_rnd_pixels16_l2_aligned1(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \

  200. { OPNAME ## _no_rnd_pixels16_l2_aligned2(dst,src2,src1,dst_stride,src_stride2,src_stride1,h); } \

  201. \

  202. static inline void OPNAME ## _pixels16_l2_aligned1(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \

  203. { OPNAME ## _pixels16_l2_aligned2(dst,src2,src1,dst_stride,src_stride2,src_stride1,h); } \

  204. \

  205. static inline void OPNAME ## _no_rnd_pixels8_l2_aligned1(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \

  206. { OPNAME ## _no_rnd_pixels8_l2_aligned2(dst,src2,src1,dst_stride,src_stride2,src_stride1,h); } \

  207. \

  208. static inline void OPNAME ## _pixels8_l2_aligned1(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \

  209. { OPNAME ## _pixels8_l2_aligned2(dst,src2,src1,dst_stride,src_stride2,src_stride1,h); } \

  210. \

  211. static inline void OPNAME ## _pixels8_l4_aligned(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\

  212.         do { \

  213.                 uint32_t a0,a1,a2,a3; \

  214.                 UNPACK(a0,a1,LP(src1),LP(src2)); \

  215.                 UNPACK(a2,a3,LP(src3),LP(src4)); \

  216.                 OP(LP(dst),rnd_PACK(a0,a1,a2,a3)); \

  217.                 UNPACK(a0,a1,LP(src1+4),LP(src2+4)); \

  218.                 UNPACK(a2,a3,LP(src3+4),LP(src4+4)); \

  219.                 OP(LP(dst+4),rnd_PACK(a0,a1,a2,a3)); \

  220.                 src1+=src_stride1;\

  221.                 src2+=src_stride2;\

  222.                 src3+=src_stride3;\

  223.                 src4+=src_stride4;\

  224.                 dst+=dst_stride;\

  225.         } while(--h); \

  226. } \

  227. \

  228. static inline void OPNAME ## _no_rnd_pixels8_l4_aligned(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\

  229.         do { \

  230.                 uint32_t a0,a1,a2,a3; \

  231.                 UNPACK(a0,a1,LP(src1),LP(src2)); \

  232.                 UNPACK(a2,a3,LP(src3),LP(src4)); \

  233.                 OP(LP(dst),no_rnd_PACK(a0,a1,a2,a3)); \

  234.                 UNPACK(a0,a1,LP(src1+4),LP(src2+4)); \

  235.                 UNPACK(a2,a3,LP(src3+4),LP(src4+4)); \

  236.                 OP(LP(dst+4),no_rnd_PACK(a0,a1,a2,a3)); \

  237.                 src1+=src_stride1;\

  238.                 src2+=src_stride2;\

  239.                 src3+=src_stride3;\

  240.                 src4+=src_stride4;\

  241.                 dst+=dst_stride;\

  242.         } while(--h); \

  243. } \

  244. \

  245. static inline void OPNAME ## _pixels8_l4_aligned0(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\

  246.         do { \

  247.                 uint32_t a0,a1,a2,a3; /* src1 only not aligned */\

  248.                 UNPACK(a0,a1,LD32(src1),LP(src2)); \

  249.                 UNPACK(a2,a3,LP(src3),LP(src4)); \

  250.                 OP(LP(dst),rnd_PACK(a0,a1,a2,a3)); \

  251.                 UNPACK(a0,a1,LD32(src1+4),LP(src2+4)); \

  252.                 UNPACK(a2,a3,LP(src3+4),LP(src4+4)); \

  253.                 OP(LP(dst+4),rnd_PACK(a0,a1,a2,a3)); \

  254.                 src1+=src_stride1;\

  255.                 src2+=src_stride2;\

  256.                 src3+=src_stride3;\

  257.                 src4+=src_stride4;\

  258.                 dst+=dst_stride;\

  259.         } while(--h); \

  260. } \

  261. \

  262. static inline void OPNAME ## _no_rnd_pixels8_l4_aligned0(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\

  263.         do { \

  264.                 uint32_t a0,a1,a2,a3; \

  265.                 UNPACK(a0,a1,LD32(src1),LP(src2)); \

  266.                 UNPACK(a2,a3,LP(src3),LP(src4)); \

  267.                 OP(LP(dst),no_rnd_PACK(a0,a1,a2,a3)); \

  268.                 UNPACK(a0,a1,LD32(src1+4),LP(src2+4)); \

  269.                 UNPACK(a2,a3,LP(src3+4),LP(src4+4)); \

  270.                 OP(LP(dst+4),no_rnd_PACK(a0,a1,a2,a3)); \

  271.                 src1+=src_stride1;\

  272.                 src2+=src_stride2;\

  273.                 src3+=src_stride3;\

  274.                 src4+=src_stride4;\

  275.                 dst+=dst_stride;\

  276.         } while(--h); \

  277. } \

  278. \

  279. static inline void OPNAME ## _pixels16_l4_aligned(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\

  280.         do { \

  281.                 uint32_t a0,a1,a2,a3; \

  282.                 UNPACK(a0,a1,LP(src1),LP(src2)); \

  283.                 UNPACK(a2,a3,LP(src3),LP(src4)); \

  284.                 OP(LP(dst),rnd_PACK(a0,a1,a2,a3)); \

  285.                 UNPACK(a0,a1,LP(src1+4),LP(src2+4)); \

  286.                 UNPACK(a2,a3,LP(src3+4),LP(src4+4)); \

  287.                 OP(LP(dst+8),rnd_PACK(a0,a1,a2,a3)); \

  288.                 UNPACK(a0,a1,LP(src1+8),LP(src2+8)); \

  289.                 UNPACK(a2,a3,LP(src3+8),LP(src4+8)); \

  290.                 OP(LP(dst+8),rnd_PACK(a0,a1,a2,a3)); \

  291.                 UNPACK(a0,a1,LP(src1+12),LP(src2+12)); \

  292.                 UNPACK(a2,a3,LP(src3+12),LP(src4+12)); \

  293.                 OP(LP(dst+12),rnd_PACK(a0,a1,a2,a3)); \

  294.                 src1+=src_stride1;\

  295.                 src2+=src_stride2;\

  296.                 src3+=src_stride3;\

  297.                 src4+=src_stride4;\

  298.                 dst+=dst_stride;\

  299.         } while(--h); \

  300. } \

  301. \

  302. static inline void OPNAME ## _no_rnd_pixels16_l4_aligned(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\

  303.         do { \

  304.                 uint32_t a0,a1,a2,a3; \

  305.                 UNPACK(a0,a1,LP(src1),LP(src2)); \

  306.                 UNPACK(a2,a3,LP(src3),LP(src4)); \

  307.                 OP(LP(dst),no_rnd_PACK(a0,a1,a2,a3)); \

  308.                 UNPACK(a0,a1,LP(src1+4),LP(src2+4)); \

  309.                 UNPACK(a2,a3,LP(src3+4),LP(src4+4)); \

  310.                 OP(LP(dst+4),no_rnd_PACK(a0,a1,a2,a3)); \

  311.                 UNPACK(a0,a1,LP(src1+8),LP(src2+8)); \

  312.                 UNPACK(a2,a3,LP(src3+8),LP(src4+8)); \

  313.                 OP(LP(dst+8),no_rnd_PACK(a0,a1,a2,a3)); \

  314.                 UNPACK(a0,a1,LP(src1+12),LP(src2+12)); \

  315.                 UNPACK(a2,a3,LP(src3+12),LP(src4+12)); \

  316.                 OP(LP(dst+12),no_rnd_PACK(a0,a1,a2,a3)); \

  317.                 src1+=src_stride1;\

  318.                 src2+=src_stride2;\

  319.                 src3+=src_stride3;\

  320.                 src4+=src_stride4;\

  321.                 dst+=dst_stride;\

  322.         } while(--h); \

  323. } \

  324. \

  325. static inline void OPNAME ## _pixels16_l4_aligned0(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\

  326.         do { /* src1 is unaligned */\

  327.                 uint32_t a0,a1,a2,a3; \

  328.                 UNPACK(a0,a1,LD32(src1),LP(src2)); \

  329.                 UNPACK(a2,a3,LP(src3),LP(src4)); \

  330.                 OP(LP(dst),rnd_PACK(a0,a1,a2,a3)); \

  331.                 UNPACK(a0,a1,LD32(src1+4),LP(src2+4)); \

  332.                 UNPACK(a2,a3,LP(src3+4),LP(src4+4)); \

  333.                 OP(LP(dst+8),rnd_PACK(a0,a1,a2,a3)); \

  334.                 UNPACK(a0,a1,LD32(src1+8),LP(src2+8)); \

  335.                 UNPACK(a2,a3,LP(src3+8),LP(src4+8)); \

  336.                 OP(LP(dst+8),rnd_PACK(a0,a1,a2,a3)); \

  337.                 UNPACK(a0,a1,LD32(src1+12),LP(src2+12)); \

  338.                 UNPACK(a2,a3,LP(src3+12),LP(src4+12)); \

  339.                 OP(LP(dst+12),rnd_PACK(a0,a1,a2,a3)); \

  340.                 src1+=src_stride1;\

  341.                 src2+=src_stride2;\

  342.                 src3+=src_stride3;\

  343.                 src4+=src_stride4;\

  344.                 dst+=dst_stride;\

  345.         } while(--h); \

  346. } \

  347. \

  348. static inline void OPNAME ## _no_rnd_pixels16_l4_aligned0(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\

  349.         do { \

  350.                 uint32_t a0,a1,a2,a3; \

  351.                 UNPACK(a0,a1,LD32(src1),LP(src2)); \

  352.                 UNPACK(a2,a3,LP(src3),LP(src4)); \

  353.                 OP(LP(dst),no_rnd_PACK(a0,a1,a2,a3)); \

  354.                 UNPACK(a0,a1,LD32(src1+4),LP(src2+4)); \

  355.                 UNPACK(a2,a3,LP(src3+4),LP(src4+4)); \

  356.                 OP(LP(dst+4),no_rnd_PACK(a0,a1,a2,a3)); \

  357.                 UNPACK(a0,a1,LD32(src1+8),LP(src2+8)); \

  358.                 UNPACK(a2,a3,LP(src3+8),LP(src4+8)); \

  359.                 OP(LP(dst+8),no_rnd_PACK(a0,a1,a2,a3)); \

  360.                 UNPACK(a0,a1,LD32(src1+12),LP(src2+12)); \

  361.                 UNPACK(a2,a3,LP(src3+12),LP(src4+12)); \

  362.                 OP(LP(dst+12),no_rnd_PACK(a0,a1,a2,a3)); \

  363.                 src1+=src_stride1;\

  364.                 src2+=src_stride2;\

  365.                 src3+=src_stride3;\

  366.                 src4+=src_stride4;\

  367.                 dst+=dst_stride;\

  368.         } while(--h); \

  369. } \

  370. \

  371. 

  372. #define op_avg(a, b) a = rnd_avg32(a,b)

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

  374. 

  375. PIXOP2(avg, op_avg)

  376. PIXOP2(put, op_put)

  377. #undef op_avg

  378. #undef op_put

  379. 

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

  381. #define avg4(a,b,c,d) ((a+b+c+d+2)>>2)

  382. 

  383. 

  384. static void gmc1_c(uint8_t *dst, uint8_t *src, int stride, int h, int x16, int y16, int rounder)

  385. {

  386.     const int A=(16-x16)*(16-y16);

  387.     const int B=(   x16)*(16-y16);

  388.     const int C=(16-x16)*(   y16);

  389.     const int D=(   x16)*(   y16);

  390. 

  391.     do {

  392.         int t0,t1,t2,t3;

  393.         uint8_t *s0 = src;

  394.         uint8_t *s1 = src+stride;

  395.         t0 = *s0++; t2 = *s1++;

  396.         t1 = *s0++; t3 = *s1++;

  397.         dst[0]= (A*t0 + B*t1 + C*t2 + D*t3 + rounder)>>8;

  398.         t0 = *s0++; t2 = *s1++;

  399.         dst[1]= (A*t1 + B*t0 + C*t3 + D*t2 + rounder)>>8;

  400.         t1 = *s0++; t3 = *s1++;

  401.         dst[2]= (A*t0 + B*t1 + C*t2 + D*t3 + rounder)>>8;

  402.         t0 = *s0++; t2 = *s1++;

  403.         dst[3]= (A*t1 + B*t0 + C*t3 + D*t2 + rounder)>>8;

  404.         t1 = *s0++; t3 = *s1++;

  405.         dst[4]= (A*t0 + B*t1 + C*t2 + D*t3 + rounder)>>8;

  406.         t0 = *s0++; t2 = *s1++;

  407.         dst[5]= (A*t1 + B*t0 + C*t3 + D*t2 + rounder)>>8;

  408.         t1 = *s0++; t3 = *s1++;

  409.         dst[6]= (A*t0 + B*t1 + C*t2 + D*t3 + rounder)>>8;

  410.         t0 = *s0++; t2 = *s1++;

  411.         dst[7]= (A*t1 + B*t0 + C*t3 + D*t2 + rounder)>>8;

  412.         dst+= stride;

  413.         src+= stride;

  414.     }while(--h);

  415. }

  416. 

  417. static void gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,

  418.                   int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height)

  419. {

  420.     int y, vx, vy;

  421.     const int s= 1<<shift;

  422. 

  423.     width--;

  424.     height--;

  425. 

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

  427.         int x;

  428. 

  429.         vx= ox;

  430.         vy= oy;

  431.         for(x=0; x<8; x++){ //XXX FIXME optimize

  432.             int src_x, src_y, frac_x, frac_y, index;

  433. 

  434.             src_x= vx>>16;

  435.             src_y= vy>>16;

  436.             frac_x= src_x&(s-1);

  437.             frac_y= src_y&(s-1);

  438.             src_x>>=shift;

  439.             src_y>>=shift;

  440. 

  441.             if((unsigned)src_x < width){

  442.                 if((unsigned)src_y < height){

  443.                     index= src_x + src_y*stride;

  444.                     dst[y*stride + x]= (  (  src[index         ]*(s-frac_x)

  445.                                            + src[index       +1]*   frac_x )*(s-frac_y)

  446.                                         + (  src[index+stride  ]*(s-frac_x)

  447.                                            + src[index+stride+1]*   frac_x )*   frac_y

  448.                                         + r)>>(shift*2);

  449.                 }else{

  450.                     index= src_x + clip(src_y, 0, height)*stride;

  451.                     dst[y*stride + x]= ( (  src[index         ]*(s-frac_x)

  452.                                           + src[index       +1]*   frac_x )*s

  453.                                         + r)>>(shift*2);

  454.                 }

  455.             }else{

  456.                 if((unsigned)src_y < height){

  457.                     index= clip(src_x, 0, width) + src_y*stride;

  458.                     dst[y*stride + x]= (  (  src[index         ]*(s-frac_y)

  459.                                            + src[index+stride  ]*   frac_y )*s

  460.                                         + r)>>(shift*2);

  461.                 }else{

  462.                     index= clip(src_x, 0, width) + clip(src_y, 0, height)*stride;

  463.                     dst[y*stride + x]=    src[index         ];

  464.                 }

  465.             }

  466. 

  467.             vx+= dxx;

  468.             vy+= dyx;

  469.         }

  470.         ox += dxy;

  471.         oy += dyy;

  472.     }

  473. }

  474. #define H264_CHROMA_MC(OPNAME, OP)\

  475. static void OPNAME ## h264_chroma_mc2_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\

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

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

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

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

  480.     \

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

  482. \

  483.     do {\

  484.         int t0,t1,t2,t3; \

  485.         uint8_t *s0 = src; \

  486.         uint8_t *s1 = src+stride; \

  487.         t0 = *s0++; t2 = *s1++; \

  488.         t1 = *s0++; t3 = *s1++; \

  489.         OP(dst[0], (A*t0 + B*t1 + C*t2 + D*t3));\

  490.         t0 = *s0++; t2 = *s1++; \

  491.         OP(dst[1], (A*t1 + B*t0 + C*t3 + D*t2));\

  492.         dst+= stride;\

  493.         src+= stride;\

  494.     }while(--h);\

  495. }\

  496. \

  497. static void OPNAME ## h264_chroma_mc4_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\

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

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

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

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

  502.     \

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

  504. \

  505.     do {\

  506.         int t0,t1,t2,t3; \

  507.         uint8_t *s0 = src; \

  508.         uint8_t *s1 = src+stride; \

  509.         t0 = *s0++; t2 = *s1++; \

  510.         t1 = *s0++; t3 = *s1++; \

  511.         OP(dst[0], (A*t0 + B*t1 + C*t2 + D*t3));\

  512.         t0 = *s0++; t2 = *s1++; \

  513.         OP(dst[1], (A*t1 + B*t0 + C*t3 + D*t2));\

  514.         t1 = *s0++; t3 = *s1++; \

  515.         OP(dst[2], (A*t0 + B*t1 + C*t2 + D*t3));\

  516.         t0 = *s0++; t2 = *s1++; \

  517.         OP(dst[3], (A*t1 + B*t0 + C*t3 + D*t2));\

  518.         dst+= stride;\

  519.         src+= stride;\

  520.     }while(--h);\

  521. }\

  522. \

  523. static void OPNAME ## h264_chroma_mc8_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\

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

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

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

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

  528.     \

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

  530. \

  531.     do {\

  532.         int t0,t1,t2,t3; \

  533.         uint8_t *s0 = src; \

  534.         uint8_t *s1 = src+stride; \

  535.         t0 = *s0++; t2 = *s1++; \

  536.         t1 = *s0++; t3 = *s1++; \

  537.         OP(dst[0], (A*t0 + B*t1 + C*t2 + D*t3));\

  538.         t0 = *s0++; t2 = *s1++; \

  539.         OP(dst[1], (A*t1 + B*t0 + C*t3 + D*t2));\

  540.         t1 = *s0++; t3 = *s1++; \

  541.         OP(dst[2], (A*t0 + B*t1 + C*t2 + D*t3));\

  542.         t0 = *s0++; t2 = *s1++; \

  543.         OP(dst[3], (A*t1 + B*t0 + C*t3 + D*t2));\

  544.         t1 = *s0++; t3 = *s1++; \

  545.         OP(dst[4], (A*t0 + B*t1 + C*t2 + D*t3));\

  546.         t0 = *s0++; t2 = *s1++; \

  547.         OP(dst[5], (A*t1 + B*t0 + C*t3 + D*t2));\

  548.         t1 = *s0++; t3 = *s1++; \

  549.         OP(dst[6], (A*t0 + B*t1 + C*t2 + D*t3));\

  550.         t0 = *s0++; t2 = *s1++; \

  551.         OP(dst[7], (A*t1 + B*t0 + C*t3 + D*t2));\

  552.         dst+= stride;\

  553.         src+= stride;\

  554.     }while(--h);\

  555. }

  556. 

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

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

  559. 

  560. H264_CHROMA_MC(put_       , op_put)

  561. H264_CHROMA_MC(avg_       , op_avg)

  562. #undef op_avg

  563. #undef op_put

  564. 

  565. /* not yet optimized */

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

  567. {

  568.     int i;

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

  570.     {

  571.         ST32(dst   , LD32(src   ));

  572.         dst+=dstStride;

  573.         src+=srcStride;

  574.     }

  575. }

  576. 

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

  578. {

  579.     int i;

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

  581.     {

  582.         ST32(dst   , LD32(src   ));

  583.         ST32(dst+4 , LD32(src+4 ));

  584.         dst+=dstStride;

  585.         src+=srcStride;

  586.     }

  587. }

  588. 

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

  590. {

  591.     int i;

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

  593.     {

  594.         ST32(dst   , LD32(src   ));

  595.         ST32(dst+4 , LD32(src+4 ));

  596.         ST32(dst+8 , LD32(src+8 ));

  597.         ST32(dst+12, LD32(src+12));

  598.         dst+=dstStride;

  599.         src+=srcStride;

  600.     }

  601. }

  602. 

  603. static inline void copy_block17(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)

  604. {

  605.     int i;

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

  607.     {

  608.         ST32(dst   , LD32(src   ));

  609.         ST32(dst+4 , LD32(src+4 ));

  610.         ST32(dst+8 , LD32(src+8 ));

  611.         ST32(dst+12, LD32(src+12));

  612.         dst[16]= src[16];

  613.         dst+=dstStride;

  614.         src+=srcStride;

  615.     }

  616. }

  617. 

  618. static inline void copy_block9(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h)

  619. {

  620.     int i;

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

  622.     {

  623.         ST32(dst   , LD32(src   ));

  624.         ST32(dst+4 , LD32(src+4 ));

  625.         dst[8]= src[8];

  626.         dst+=dstStride;

  627.         src+=srcStride;

  628.     }

  629. }

  630. /* end not optimized */

  631. 

  632. #define QPEL_MC(r, OPNAME, RND, OP) \

  633. static void OPNAME ## mpeg4_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\

  634.     uint8_t *cm = cropTbl + MAX_NEG_CROP;\

  635.     do {\

  636.         uint8_t *s = src; \

  637.         int src0,src1,src2,src3,src4,src5,src6,src7,src8;\

  638.         src0= *s++;\

  639.         src1= *s++;\

  640.         src2= *s++;\

  641.         src3= *s++;\

  642.         src4= *s++;\

  643.         OP(dst[0], (src0+src1)*20 - (src0+src2)*6 + (src1+src3)*3 - (src2+src4));\

  644.         src5= *s++;\

  645.         OP(dst[1], (src1+src2)*20 - (src0+src3)*6 + (src0+src4)*3 - (src1+src5));\

  646.         src6= *s++;\

  647.         OP(dst[2], (src2+src3)*20 - (src1+src4)*6 + (src0+src5)*3 - (src0+src6));\

  648.         src7= *s++;\

  649.         OP(dst[3], (src3+src4)*20 - (src2+src5)*6 + (src1+src6)*3 - (src0+src7));\

  650.         src8= *s++;\

  651.         OP(dst[4], (src4+src5)*20 - (src3+src6)*6 + (src2+src7)*3 - (src1+src8));\

  652.         OP(dst[5], (src5+src6)*20 - (src4+src7)*6 + (src3+src8)*3 - (src2+src8));\

  653.         OP(dst[6], (src6+src7)*20 - (src5+src8)*6 + (src4+src8)*3 - (src3+src7));\

  654.         OP(dst[7], (src7+src8)*20 - (src6+src8)*6 + (src5+src7)*3 - (src4+src6));\

  655.         dst+=dstStride;\

  656.         src+=srcStride;\

  657.     }while(--h);\

  658. }\

  659. \

  660. static void OPNAME ## mpeg4_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  661.     uint8_t *cm = cropTbl + MAX_NEG_CROP;\

  662.     int w=8;\

  663.     do{\

  664.         uint8_t *s = src, *d=dst;\

  665.         int src0,src1,src2,src3,src4,src5,src6,src7,src8;\

  666.         src0 = *s; s+=srcStride; \

  667.         src1 = *s; s+=srcStride; \

  668.         src2 = *s; s+=srcStride; \

  669.         src3 = *s; s+=srcStride; \

  670.         src4 = *s; s+=srcStride; \

  671.         OP(*d, (src0+src1)*20 - (src0+src2)*6 + (src1+src3)*3 - (src2+src4));d+=dstStride;\

  672.         src5 = *s; s+=srcStride; \

  673.         OP(*d, (src1+src2)*20 - (src0+src3)*6 + (src0+src4)*3 - (src1+src5));d+=dstStride;\

  674.         src6 = *s; s+=srcStride; \

  675.         OP(*d, (src2+src3)*20 - (src1+src4)*6 + (src0+src5)*3 - (src0+src6));d+=dstStride;\

  676.         src7 = *s; s+=srcStride; \

  677.         OP(*d, (src3+src4)*20 - (src2+src5)*6 + (src1+src6)*3 - (src0+src7));d+=dstStride;\

  678.         src8 = *s; \

  679.         OP(*d, (src4+src5)*20 - (src3+src6)*6 + (src2+src7)*3 - (src1+src8));d+=dstStride;\

  680.         OP(*d, (src5+src6)*20 - (src4+src7)*6 + (src3+src8)*3 - (src2+src8));d+=dstStride;\

  681.         OP(*d, (src6+src7)*20 - (src5+src8)*6 + (src4+src8)*3 - (src3+src7));d+=dstStride;\

  682.         OP(*d, (src7+src8)*20 - (src6+src8)*6 + (src5+src7)*3 - (src4+src6));\

  683.         dst++;\

  684.         src++;\

  685.     }while(--w);\

  686. }\

  687. \

  688. static void OPNAME ## mpeg4_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\

  689.     uint8_t *cm = cropTbl + MAX_NEG_CROP;\

  690.     do {\

  691.         uint8_t *s = src;\

  692.         int src0,src1,src2,src3,src4,src5,src6,src7,src8;\

  693.         int src9,src10,src11,src12,src13,src14,src15,src16;\

  694.         src0= *s++;\

  695.         src1= *s++;\

  696.         src2= *s++;\

  697.         src3= *s++;\

  698.         src4= *s++;\

  699.         OP(dst[ 0], (src0 +src1 )*20 - (src0 +src2 )*6 + (src1 +src3 )*3 - (src2 +src4 ));\

  700.         src5= *s++;\

  701.         OP(dst[ 1], (src1 +src2 )*20 - (src0 +src3 )*6 + (src0 +src4 )*3 - (src1 +src5 ));\

  702.         src6= *s++;\

  703.         OP(dst[ 2], (src2 +src3 )*20 - (src1 +src4 )*6 + (src0 +src5 )*3 - (src0 +src6 ));\

  704.         src7= *s++;\

  705.         OP(dst[ 3], (src3 +src4 )*20 - (src2 +src5 )*6 + (src1 +src6 )*3 - (src0 +src7 ));\

  706.         src8= *s++;\

  707.         OP(dst[ 4], (src4 +src5 )*20 - (src3 +src6 )*6 + (src2 +src7 )*3 - (src1 +src8 ));\

  708.         src9= *s++;\

  709.         OP(dst[ 5], (src5 +src6 )*20 - (src4 +src7 )*6 + (src3 +src8 )*3 - (src2 +src9 ));\

  710.         src10= *s++;\

  711.         OP(dst[ 6], (src6 +src7 )*20 - (src5 +src8 )*6 + (src4 +src9 )*3 - (src3 +src10));\

  712.         src11= *s++;\

  713.         OP(dst[ 7], (src7 +src8 )*20 - (src6 +src9 )*6 + (src5 +src10)*3 - (src4 +src11));\

  714.         src12= *s++;\

  715.         OP(dst[ 8], (src8 +src9 )*20 - (src7 +src10)*6 + (src6 +src11)*3 - (src5 +src12));\

  716.         src13= *s++;\

  717.         OP(dst[ 9], (src9 +src10)*20 - (src8 +src11)*6 + (src7 +src12)*3 - (src6 +src13));\

  718.         src14= *s++;\

  719.         OP(dst[10], (src10+src11)*20 - (src9 +src12)*6 + (src8 +src13)*3 - (src7 +src14));\

  720.         src15= *s++;\

  721.         OP(dst[11], (src11+src12)*20 - (src10+src13)*6 + (src9 +src14)*3 - (src8 +src15));\

  722.         src16= *s++;\

  723.         OP(dst[12], (src12+src13)*20 - (src11+src14)*6 + (src10+src15)*3 - (src9 +src16));\

  724.         OP(dst[13], (src13+src14)*20 - (src12+src15)*6 + (src11+src16)*3 - (src10+src16));\

  725.         OP(dst[14], (src14+src15)*20 - (src13+src16)*6 + (src12+src16)*3 - (src11+src15));\

  726.         OP(dst[15], (src15+src16)*20 - (src14+src16)*6 + (src13+src15)*3 - (src12+src14));\

  727.         dst+=dstStride;\

  728.         src+=srcStride;\

  729.     }while(--h);\

  730. }\

  731. \

  732. static void OPNAME ## mpeg4_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  733.     uint8_t *cm = cropTbl + MAX_NEG_CROP;\

  734.     int w=16;\

  735.     do {\

  736.         uint8_t *s = src, *d=dst;\

  737.         int src0,src1,src2,src3,src4,src5,src6,src7,src8;\

  738.         int src9,src10,src11,src12,src13,src14,src15,src16;\

  739.         src0 = *s; s+=srcStride; \

  740.         src1 = *s; s+=srcStride; \

  741.         src2 = *s; s+=srcStride; \

  742.         src3 = *s; s+=srcStride; \

  743.         src4 = *s; s+=srcStride; \

  744.         OP(*d, (src0 +src1 )*20 - (src0 +src2 )*6 + (src1 +src3 )*3 - (src2 +src4 ));d+=dstStride;\

  745.         src5 = *s; s+=srcStride; \

  746.         OP(*d, (src1 +src2 )*20 - (src0 +src3 )*6 + (src0 +src4 )*3 - (src1 +src5 ));d+=dstStride;\

  747.         src6 = *s; s+=srcStride; \

  748.         OP(*d, (src2 +src3 )*20 - (src1 +src4 )*6 + (src0 +src5 )*3 - (src0 +src6 ));d+=dstStride;\

  749.         src7 = *s; s+=srcStride; \

  750.         OP(*d, (src3 +src4 )*20 - (src2 +src5 )*6 + (src1 +src6 )*3 - (src0 +src7 ));d+=dstStride;\

  751.         src8 = *s; s+=srcStride; \

  752.         OP(*d, (src4 +src5 )*20 - (src3 +src6 )*6 + (src2 +src7 )*3 - (src1 +src8 ));d+=dstStride;\

  753.         src9 = *s; s+=srcStride; \

  754.         OP(*d, (src5 +src6 )*20 - (src4 +src7 )*6 + (src3 +src8 )*3 - (src2 +src9 ));d+=dstStride;\

  755.         src10 = *s; s+=srcStride; \

  756.         OP(*d, (src6 +src7 )*20 - (src5 +src8 )*6 + (src4 +src9 )*3 - (src3 +src10));d+=dstStride;\

  757.         src11 = *s; s+=srcStride; \

  758.         OP(*d, (src7 +src8 )*20 - (src6 +src9 )*6 + (src5 +src10)*3 - (src4 +src11));d+=dstStride;\

  759.         src12 = *s; s+=srcStride; \

  760.         OP(*d, (src8 +src9 )*20 - (src7 +src10)*6 + (src6 +src11)*3 - (src5 +src12));d+=dstStride;\

  761.         src13 = *s; s+=srcStride; \

  762.         OP(*d, (src9 +src10)*20 - (src8 +src11)*6 + (src7 +src12)*3 - (src6 +src13));d+=dstStride;\

  763.         src14 = *s; s+=srcStride; \

  764.         OP(*d, (src10+src11)*20 - (src9 +src12)*6 + (src8 +src13)*3 - (src7 +src14));d+=dstStride;\

  765.         src15 = *s; s+=srcStride; \

  766.         OP(*d, (src11+src12)*20 - (src10+src13)*6 + (src9 +src14)*3 - (src8 +src15));d+=dstStride;\

  767.         src16 = *s; \

  768.         OP(*d, (src12+src13)*20 - (src11+src14)*6 + (src10+src15)*3 - (src9 +src16));d+=dstStride;\

  769.         OP(*d, (src13+src14)*20 - (src12+src15)*6 + (src11+src16)*3 - (src10+src16));d+=dstStride;\

  770.         OP(*d, (src14+src15)*20 - (src13+src16)*6 + (src12+src16)*3 - (src11+src15));d+=dstStride;\

  771.         OP(*d, (src15+src16)*20 - (src14+src16)*6 + (src13+src15)*3 - (src12+src14));\

  772.         dst++;\

  773.         src++;\

  774.     }while(--w);\

  775. }\

  776. \

  777. static void OPNAME ## qpel8_mc00_c (uint8_t *dst, uint8_t *src, int stride){\

  778.     OPNAME ## pixels8_c(dst, src, stride, 8);\

  779. }\

  780. \

  781. static void OPNAME ## qpel8_mc10_c(uint8_t *dst, uint8_t *src, int stride){\

  782.     uint8_t half[64];\

  783.     put ## RND ## mpeg4_qpel8_h_lowpass(half, src, 8, stride, 8);\

  784.     OPNAME ## pixels8_l2_aligned2(dst, src, half, stride, stride, 8, 8);\

  785. }\

  786. \

  787. static void OPNAME ## qpel8_mc20_c(uint8_t *dst, uint8_t *src, int stride){\

  788.     OPNAME ## mpeg4_qpel8_h_lowpass(dst, src, stride, stride, 8);\

  789. }\

  790. \

  791. static void OPNAME ## qpel8_mc30_c(uint8_t *dst, uint8_t *src, int stride){\

  792.     uint8_t half[64];\

  793.     put ## RND ## mpeg4_qpel8_h_lowpass(half, src, 8, stride, 8);\

  794.     OPNAME ## pixels8_l2_aligned2(dst, src+1, half, stride, stride, 8, 8);\

  795. }\

  796. \

  797. static void OPNAME ## qpel8_mc01_c(uint8_t *dst, uint8_t *src, int stride){\

  798.     uint8_t full[16*9];\

  799.     uint8_t half[64];\

  800.     copy_block9(full, src, 16, stride, 9);\

  801.     put ## RND ## mpeg4_qpel8_v_lowpass(half, full, 8, 16);\

  802.     OPNAME ## pixels8_l2_aligned(dst, full, half, stride, 16, 8, 8);\

  803. }\

  804. \

  805. static void OPNAME ## qpel8_mc02_c(uint8_t *dst, uint8_t *src, int stride){\

  806.     uint8_t full[16*9];\

  807.     copy_block9(full, src, 16, stride, 9);\

  808.     OPNAME ## mpeg4_qpel8_v_lowpass(dst, full, stride, 16);\

  809. }\

  810. \

  811. static void OPNAME ## qpel8_mc03_c(uint8_t *dst, uint8_t *src, int stride){\

  812.     uint8_t full[16*9];\

  813.     uint8_t half[64];\

  814.     copy_block9(full, src, 16, stride, 9);\

  815.     put ## RND ## mpeg4_qpel8_v_lowpass(half, full, 8, 16);\

  816.     OPNAME ## pixels8_l2_aligned(dst, full+16, half, stride, 16, 8, 8);\

  817. }\

  818. static void ff_ ## OPNAME ## qpel8_mc11_old_c(uint8_t *dst, uint8_t *src, int stride){\

  819.     uint8_t full[16*9];\

  820.     uint8_t halfH[72];\

  821.     uint8_t halfV[64];\

  822.     uint8_t halfHV[64];\

  823.     copy_block9(full, src, 16, stride, 9);\

  824.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\

  825.     put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full, 8, 16);\

  826.     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\

  827.     OPNAME ## pixels8_l4_aligned(dst, full, halfH, halfV, halfHV, stride, 16, 8, 8, 8, 8);\

  828. }\

  829. static void OPNAME ## qpel8_mc11_c(uint8_t *dst, uint8_t *src, int stride){\

  830.     uint8_t full[16*9];\

  831.     uint8_t halfH[72];\

  832.     uint8_t halfHV[64];\

  833.     copy_block9(full, src, 16, stride, 9);\

  834.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\

  835.     put ## RND ## pixels8_l2_aligned(halfH, halfH, full, 8, 8, 16, 9);\

  836.     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\

  837.     OPNAME ## pixels8_l2_aligned(dst, halfH, halfHV, stride, 8, 8, 8);\

  838. }\

  839. static void ff_ ## OPNAME ## qpel8_mc31_old_c(uint8_t *dst, uint8_t *src, int stride){\

  840.     uint8_t full[16*9];\

  841.     uint8_t halfH[72];\

  842.     uint8_t halfV[64];\

  843.     uint8_t halfHV[64];\

  844.     copy_block9(full, src, 16, stride, 9);\

  845.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\

  846.     put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full+1, 8, 16);\

  847.     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\

  848.     OPNAME ## pixels8_l4_aligned0(dst, full+1, halfH, halfV, halfHV, stride, 16, 8, 8, 8, 8);\

  849. }\

  850. static void OPNAME ## qpel8_mc31_c(uint8_t *dst, uint8_t *src, int stride){\

  851.     uint8_t full[16*9];\

  852.     uint8_t halfH[72];\

  853.     uint8_t halfHV[64];\

  854.     copy_block9(full, src, 16, stride, 9);\

  855.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\

  856.     put ## RND ## pixels8_l2_aligned1(halfH, halfH, full+1, 8, 8, 16, 9);\

  857.     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\

  858.     OPNAME ## pixels8_l2_aligned(dst, halfH, halfHV, stride, 8, 8, 8);\

  859. }\

  860. static void ff_ ## OPNAME ## qpel8_mc13_old_c(uint8_t *dst, uint8_t *src, int stride){\

  861.     uint8_t full[16*9];\

  862.     uint8_t halfH[72];\

  863.     uint8_t halfV[64];\

  864.     uint8_t halfHV[64];\

  865.     copy_block9(full, src, 16, stride, 9);\

  866.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\

  867.     put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full, 8, 16);\

  868.     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\

  869.     OPNAME ## pixels8_l4_aligned(dst, full+16, halfH+8, halfV, halfHV, stride, 16, 8, 8, 8, 8);\

  870. }\

  871. static void OPNAME ## qpel8_mc13_c(uint8_t *dst, uint8_t *src, int stride){\

  872.     uint8_t full[16*9];\

  873.     uint8_t halfH[72];\

  874.     uint8_t halfHV[64];\

  875.     copy_block9(full, src, 16, stride, 9);\

  876.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\

  877.     put ## RND ## pixels8_l2_aligned(halfH, halfH, full, 8, 8, 16, 9);\

  878.     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\

  879.     OPNAME ## pixels8_l2_aligned(dst, halfH+8, halfHV, stride, 8, 8, 8);\

  880. }\

  881. static void ff_ ## OPNAME ## qpel8_mc33_old_c(uint8_t *dst, uint8_t *src, int stride){\

  882.     uint8_t full[16*9];\

  883.     uint8_t halfH[72];\

  884.     uint8_t halfV[64];\

  885.     uint8_t halfHV[64];\

  886.     copy_block9(full, src, 16, stride, 9);\

  887.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full  , 8, 16, 9);\

  888.     put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full+1, 8, 16);\

  889.     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\

  890.     OPNAME ## pixels8_l4_aligned0(dst, full+17, halfH+8, halfV, halfHV, stride, 16, 8, 8, 8, 8);\

  891. }\

  892. static void OPNAME ## qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride){\

  893.     uint8_t full[16*9];\

  894.     uint8_t halfH[72];\

  895.     uint8_t halfHV[64];\

  896.     copy_block9(full, src, 16, stride, 9);\

  897.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\

  898.     put ## RND ## pixels8_l2_aligned1(halfH, halfH, full+1, 8, 8, 16, 9);\

  899.     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\

  900.     OPNAME ## pixels8_l2_aligned(dst, halfH+8, halfHV, stride, 8, 8, 8);\

  901. }\

  902. static void OPNAME ## qpel8_mc21_c(uint8_t *dst, uint8_t *src, int stride){\

  903.     uint8_t halfH[72];\

  904.     uint8_t halfHV[64];\

  905.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\

  906.     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\

  907.     OPNAME ## pixels8_l2_aligned(dst, halfH, halfHV, stride, 8, 8, 8);\

  908. }\

  909. static void OPNAME ## qpel8_mc23_c(uint8_t *dst, uint8_t *src, int stride){\

  910.     uint8_t halfH[72];\

  911.     uint8_t halfHV[64];\

  912.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\

  913.     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\

  914.     OPNAME ## pixels8_l2_aligned(dst, halfH+8, halfHV, stride, 8, 8, 8);\

  915. }\

  916. static void ff_ ## OPNAME ## qpel8_mc12_old_c(uint8_t *dst, uint8_t *src, int stride){\

  917.     uint8_t full[16*9];\

  918.     uint8_t halfH[72];\

  919.     uint8_t halfV[64];\

  920.     uint8_t halfHV[64];\

  921.     copy_block9(full, src, 16, stride, 9);\

  922.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\

  923.     put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full, 8, 16);\

  924.     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\

  925.     OPNAME ## pixels8_l2_aligned(dst, halfV, halfHV, stride, 8, 8, 8);\

  926. }\

  927. static void OPNAME ## qpel8_mc12_c(uint8_t *dst, uint8_t *src, int stride){\

  928.     uint8_t full[16*9];\

  929.     uint8_t halfH[72];\

  930.     copy_block9(full, src, 16, stride, 9);\

  931.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\

  932.     put ## RND ## pixels8_l2_aligned(halfH, halfH, full, 8, 8, 16, 9);\

  933.     OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\

  934. }\

  935. static void ff_ ## OPNAME ## qpel8_mc32_old_c(uint8_t *dst, uint8_t *src, int stride){\

  936.     uint8_t full[16*9];\

  937.     uint8_t halfH[72];\

  938.     uint8_t halfV[64];\

  939.     uint8_t halfHV[64];\

  940.     copy_block9(full, src, 16, stride, 9);\

  941.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\

  942.     put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full+1, 8, 16);\

  943.     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\

  944.     OPNAME ## pixels8_l2_aligned(dst, halfV, halfHV, stride, 8, 8, 8);\

  945. }\

  946. static void OPNAME ## qpel8_mc32_c(uint8_t *dst, uint8_t *src, int stride){\

  947.     uint8_t full[16*9];\

  948.     uint8_t halfH[72];\

  949.     copy_block9(full, src, 16, stride, 9);\

  950.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\

  951.     put ## RND ## pixels8_l2_aligned1(halfH, halfH, full+1, 8, 8, 16, 9);\

  952.     OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\

  953. }\

  954. static void OPNAME ## qpel8_mc22_c(uint8_t *dst, uint8_t *src, int stride){\

  955.     uint8_t halfH[72];\

  956.     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\

  957.     OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\

  958. }\

  959. static void OPNAME ## qpel16_mc00_c (uint8_t *dst, uint8_t *src, int stride){\

  960.     OPNAME ## pixels16_c(dst, src, stride, 16);\

  961. }\

  962. \

  963. static void OPNAME ## qpel16_mc10_c(uint8_t *dst, uint8_t *src, int stride){\

  964.     uint8_t half[256];\

  965.     put ## RND ## mpeg4_qpel16_h_lowpass(half, src, 16, stride, 16);\

  966.     OPNAME ## pixels16_l2_aligned2(dst, src, half, stride, stride, 16, 16);\

  967. }\

  968. \

  969. static void OPNAME ## qpel16_mc20_c(uint8_t *dst, uint8_t *src, int stride){\

  970.     OPNAME ## mpeg4_qpel16_h_lowpass(dst, src, stride, stride, 16);\

  971. }\

  972. \

  973. static void OPNAME ## qpel16_mc30_c(uint8_t *dst, uint8_t *src, int stride){\

  974.     uint8_t half[256];\

  975.     put ## RND ## mpeg4_qpel16_h_lowpass(half, src, 16, stride, 16);\

  976.     OPNAME ## pixels16_l2_aligned2(dst, src+1, half, stride, stride, 16, 16);\

  977. }\

  978. \

  979. static void OPNAME ## qpel16_mc01_c(uint8_t *dst, uint8_t *src, int stride){\

  980.     uint8_t full[24*17];\

  981.     uint8_t half[256];\

  982.     copy_block17(full, src, 24, stride, 17);\

  983.     put ## RND ## mpeg4_qpel16_v_lowpass(half, full, 16, 24);\

  984.     OPNAME ## pixels16_l2_aligned(dst, full, half, stride, 24, 16, 16);\

  985. }\

  986. \

  987. static void OPNAME ## qpel16_mc02_c(uint8_t *dst, uint8_t *src, int stride){\

  988.     uint8_t full[24*17];\

  989.     copy_block17(full, src, 24, stride, 17);\

  990.     OPNAME ## mpeg4_qpel16_v_lowpass(dst, full, stride, 24);\

  991. }\

  992. \

  993. static void OPNAME ## qpel16_mc03_c(uint8_t *dst, uint8_t *src, int stride){\

  994.     uint8_t full[24*17];\

  995.     uint8_t half[256];\

  996.     copy_block17(full, src, 24, stride, 17);\

  997.     put ## RND ## mpeg4_qpel16_v_lowpass(half, full, 16, 24);\

  998.     OPNAME ## pixels16_l2_aligned(dst, full+24, half, stride, 24, 16, 16);\

  999. }\

  1000. static void ff_ ## OPNAME ## qpel16_mc11_old_c(uint8_t *dst, uint8_t *src, int stride){\

  1001.     uint8_t full[24*17];\

  1002.     uint8_t halfH[272];\

  1003.     uint8_t halfV[256];\

  1004.     uint8_t halfHV[256];\

  1005.     copy_block17(full, src, 24, stride, 17);\

  1006.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\

  1007.     put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full, 16, 24);\

  1008.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\

  1009.     OPNAME ## pixels16_l4_aligned(dst, full, halfH, halfV, halfHV, stride, 24, 16, 16, 16, 16);\

  1010. }\

  1011. static void OPNAME ## qpel16_mc11_c(uint8_t *dst, uint8_t *src, int stride){\

  1012.     uint8_t full[24*17];\

  1013.     uint8_t halfH[272];\

  1014.     uint8_t halfHV[256];\

  1015.     copy_block17(full, src, 24, stride, 17);\

  1016.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\

  1017.     put ## RND ## pixels16_l2_aligned(halfH, halfH, full, 16, 16, 24, 17);\

  1018.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\

  1019.     OPNAME ## pixels16_l2_aligned(dst, halfH, halfHV, stride, 16, 16, 16);\

  1020. }\

  1021. static void ff_ ## OPNAME ## qpel16_mc31_old_c(uint8_t *dst, uint8_t *src, int stride){\

  1022.     uint8_t full[24*17];\

  1023.     uint8_t halfH[272];\

  1024.     uint8_t halfV[256];\

  1025.     uint8_t halfHV[256];\

  1026.     copy_block17(full, src, 24, stride, 17);\

  1027.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\

  1028.     put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full+1, 16, 24);\

  1029.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\

  1030.     OPNAME ## pixels16_l4_aligned0(dst, full+1, halfH, halfV, halfHV, stride, 24, 16, 16, 16, 16);\

  1031. }\

  1032. static void OPNAME ## qpel16_mc31_c(uint8_t *dst, uint8_t *src, int stride){\

  1033.     uint8_t full[24*17];\

  1034.     uint8_t halfH[272];\

  1035.     uint8_t halfHV[256];\

  1036.     copy_block17(full, src, 24, stride, 17);\

  1037.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\

  1038.     put ## RND ## pixels16_l2_aligned1(halfH, halfH, full+1, 16, 16, 24, 17);\

  1039.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\

  1040.     OPNAME ## pixels16_l2_aligned(dst, halfH, halfHV, stride, 16, 16, 16);\

  1041. }\

  1042. static void ff_ ## OPNAME ## qpel16_mc13_old_c(uint8_t *dst, uint8_t *src, int stride){\

  1043.     uint8_t full[24*17];\

  1044.     uint8_t halfH[272];\

  1045.     uint8_t halfV[256];\

  1046.     uint8_t halfHV[256];\

  1047.     copy_block17(full, src, 24, stride, 17);\

  1048.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\

  1049.     put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full, 16, 24);\

  1050.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\

  1051.     OPNAME ## pixels16_l4_aligned(dst, full+24, halfH+16, halfV, halfHV, stride, 24, 16, 16, 16, 16);\

  1052. }\

  1053. static void OPNAME ## qpel16_mc13_c(uint8_t *dst, uint8_t *src, int stride){\

  1054.     uint8_t full[24*17];\

  1055.     uint8_t halfH[272];\

  1056.     uint8_t halfHV[256];\

  1057.     copy_block17(full, src, 24, stride, 17);\

  1058.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\

  1059.     put ## RND ## pixels16_l2_aligned(halfH, halfH, full, 16, 16, 24, 17);\

  1060.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\

  1061.     OPNAME ## pixels16_l2_aligned(dst, halfH+16, halfHV, stride, 16, 16, 16);\

  1062. }\

  1063. static void ff_ ## OPNAME ## qpel16_mc33_old_c(uint8_t *dst, uint8_t *src, int stride){\

  1064.     uint8_t full[24*17];\

  1065.     uint8_t halfH[272];\

  1066.     uint8_t halfV[256];\

  1067.     uint8_t halfHV[256];\

  1068.     copy_block17(full, src, 24, stride, 17);\

  1069.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full  , 16, 24, 17);\

  1070.     put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full+1, 16, 24);\

  1071.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\

  1072.     OPNAME ## pixels16_l4_aligned0(dst, full+25, halfH+16, halfV, halfHV, stride, 24, 16, 16, 16, 16);\

  1073. }\

  1074. static void OPNAME ## qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride){\

  1075.     uint8_t full[24*17];\

  1076.     uint8_t halfH[272];\

  1077.     uint8_t halfHV[256];\

  1078.     copy_block17(full, src, 24, stride, 17);\

  1079.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\

  1080.     put ## RND ## pixels16_l2_aligned1(halfH, halfH, full+1, 16, 16, 24, 17);\

  1081.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\

  1082.     OPNAME ## pixels16_l2_aligned(dst, halfH+16, halfHV, stride, 16, 16, 16);\

  1083. }\

  1084. static void OPNAME ## qpel16_mc21_c(uint8_t *dst, uint8_t *src, int stride){\

  1085.     uint8_t halfH[272];\

  1086.     uint8_t halfHV[256];\

  1087.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\

  1088.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\

  1089.     OPNAME ## pixels16_l2_aligned(dst, halfH, halfHV, stride, 16, 16, 16);\

  1090. }\

  1091. static void OPNAME ## qpel16_mc23_c(uint8_t *dst, uint8_t *src, int stride){\

  1092.     uint8_t halfH[272];\

  1093.     uint8_t halfHV[256];\

  1094.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\

  1095.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\

  1096.     OPNAME ## pixels16_l2_aligned(dst, halfH+16, halfHV, stride, 16, 16, 16);\

  1097. }\

  1098. static void ff_ ## OPNAME ## qpel16_mc12_old_c(uint8_t *dst, uint8_t *src, int stride){\

  1099.     uint8_t full[24*17];\

  1100.     uint8_t halfH[272];\

  1101.     uint8_t halfV[256];\

  1102.     uint8_t halfHV[256];\

  1103.     copy_block17(full, src, 24, stride, 17);\

  1104.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\

  1105.     put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full, 16, 24);\

  1106.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\

  1107.     OPNAME ## pixels16_l2_aligned(dst, halfV, halfHV, stride, 16, 16, 16);\

  1108. }\

  1109. static void OPNAME ## qpel16_mc12_c(uint8_t *dst, uint8_t *src, int stride){\

  1110.     uint8_t full[24*17];\

  1111.     uint8_t halfH[272];\

  1112.     copy_block17(full, src, 24, stride, 17);\

  1113.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\

  1114.     put ## RND ## pixels16_l2_aligned(halfH, halfH, full, 16, 16, 24, 17);\

  1115.     OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\

  1116. }\

  1117. static void ff_ ## OPNAME ## qpel16_mc32_old_c(uint8_t *dst, uint8_t *src, int stride){\

  1118.     uint8_t full[24*17];\

  1119.     uint8_t halfH[272];\

  1120.     uint8_t halfV[256];\

  1121.     uint8_t halfHV[256];\

  1122.     copy_block17(full, src, 24, stride, 17);\

  1123.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\

  1124.     put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full+1, 16, 24);\

  1125.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\

  1126.     OPNAME ## pixels16_l2_aligned(dst, halfV, halfHV, stride, 16, 16, 16);\

  1127. }\

  1128. static void OPNAME ## qpel16_mc32_c(uint8_t *dst, uint8_t *src, int stride){\

  1129.     uint8_t full[24*17];\

  1130.     uint8_t halfH[272];\

  1131.     copy_block17(full, src, 24, stride, 17);\

  1132.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\

  1133.     put ## RND ## pixels16_l2_aligned1(halfH, halfH, full+1, 16, 16, 24, 17);\

  1134.     OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\

  1135. }\

  1136. static void OPNAME ## qpel16_mc22_c(uint8_t *dst, uint8_t *src, int stride){\

  1137.     uint8_t halfH[272];\

  1138.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\

  1139.     OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\

  1140. }

  1141. 

  1142. #define op_avg(a, b) a = (((a)+cm[((b) + 16)>>5]+1)>>1)

  1143. #define op_avg_no_rnd(a, b) a = (((a)+cm[((b) + 15)>>5])>>1)

  1144. #define op_put(a, b) a = cm[((b) + 16)>>5]

  1145. #define op_put_no_rnd(a, b) a = cm[((b) + 15)>>5]

  1146. 

  1147. QPEL_MC(0, put_       , _       , op_put)

  1148. QPEL_MC(1, put_no_rnd_, _no_rnd_, op_put_no_rnd)

  1149. QPEL_MC(0, avg_       , _       , op_avg)

  1150. //QPEL_MC(1, avg_no_rnd , _       , op_avg)

  1151. #undef op_avg

  1152. #undef op_avg_no_rnd

  1153. #undef op_put

  1154. #undef op_put_no_rnd

  1155. 

  1156. #if 1

  1157. #define H264_LOWPASS(OPNAME, OP, OP2) \

  1158. static inline void OPNAME ## h264_qpel_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,int w,int h){\

  1159.     uint8_t *cm = cropTbl + MAX_NEG_CROP;\

  1160.     do {\

  1161.         int srcB,srcA,src0,src1,src2,src3,src4,src5,src6;\

  1162.         uint8_t *s = src-2;\

  1163.         srcB = *s++;\

  1164.         srcA = *s++;\

  1165.         src0 = *s++;\

  1166.         src1 = *s++;\

  1167.         src2 = *s++;\

  1168.         src3 = *s++;\

  1169.         OP(dst[0], (src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));\

  1170.         src4 = *s++;\

  1171.         OP(dst[1], (src1+src2)*20 - (src0+src3)*5 + (srcA+src4));\

  1172.         src5 = *s++;\

  1173.         OP(dst[2], (src2+src3)*20 - (src1+src4)*5 + (src0+src5));\

  1174.         src6 = *s++;\

  1175.         OP(dst[3], (src3+src4)*20 - (src2+src5)*5 + (src1+src6));\

  1176.       if (w>4) { /* it optimized */ \

  1177.         int src7,src8,src9,src10; \

  1178.         src7 = *s++;\

  1179.         OP(dst[4], (src4+src5)*20 - (src3+src6)*5 + (src2+src7));\

  1180.         src8 = *s++;\

  1181.         OP(dst[5], (src5+src6)*20 - (src4+src7)*5 + (src3+src8));\

  1182.         src9 = *s++;\

  1183.         OP(dst[6], (src6+src7)*20 - (src5+src8)*5 + (src4+src9));\

  1184.         src10 = *s++;\

  1185.         OP(dst[7], (src7+src8)*20 - (src6+src9)*5 + (src5+src10));\

  1186.        if (w>8) { \

  1187.         int src11,src12,src13,src14,src15,src16,src17,src18; \

  1188.         src11 = *s++;\

  1189.         OP(dst[8] , (src8 +src9 )*20 - (src7 +src10)*5 + (src6 +src11));\

  1190.         src12 = *s++;\

  1191.         OP(dst[9] , (src9 +src10)*20 - (src8 +src11)*5 + (src7 +src12));\

  1192.         src13 = *s++;\

  1193.         OP(dst[10], (src10+src11)*20 - (src9 +src12)*5 + (src8 +src13));\

  1194.         src14 = *s++;\

  1195.         OP(dst[11], (src11+src12)*20 - (src10+src13)*5 + (src9 +src14));\

  1196.         src15 = *s++;\

  1197.         OP(dst[12], (src12+src13)*20 - (src11+src14)*5 + (src10+src15));\

  1198.         src16 = *s++;\

  1199.         OP(dst[13], (src13+src14)*20 - (src12+src15)*5 + (src11+src16));\

  1200.         src17 = *s++;\

  1201.         OP(dst[14], (src14+src15)*20 - (src13+src16)*5 + (src12+src17));\

  1202.         src18 = *s++;\

  1203.         OP(dst[15], (src15+src16)*20 - (src14+src17)*5 + (src13+src18));\

  1204.        } \

  1205.       } \

  1206.         dst+=dstStride;\

  1207.         src+=srcStride;\

  1208.     }while(--h);\

  1209. }\

  1210. \

  1211. static inline void OPNAME ## h264_qpel_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,int w,int h){\

  1212.     uint8_t *cm = cropTbl + MAX_NEG_CROP;\

  1213.     do{\

  1214.         int srcB,srcA,src0,src1,src2,src3,src4,src5,src6;\

  1215.         uint8_t *s = src-2*srcStride,*d=dst;\

  1216.         srcB = *s; s+=srcStride;\

  1217.         srcA = *s; s+=srcStride;\

  1218.         src0 = *s; s+=srcStride;\

  1219.         src1 = *s; s+=srcStride;\

  1220.         src2 = *s; s+=srcStride;\

  1221.         src3 = *s; s+=srcStride;\

  1222.         OP(*d, (src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));d+=dstStride;\

  1223.         src4 = *s; s+=srcStride;\

  1224.         OP(*d, (src1+src2)*20 - (src0+src3)*5 + (srcA+src4));d+=dstStride;\

  1225.         src5 = *s; s+=srcStride;\

  1226.         OP(*d, (src2+src3)*20 - (src1+src4)*5 + (src0+src5));d+=dstStride;\

  1227.         src6 = *s; s+=srcStride;\

  1228.         OP(*d, (src3+src4)*20 - (src2+src5)*5 + (src1+src6));d+=dstStride;\

  1229.       if (h>4) { \

  1230.         int src7,src8,src9,src10; \

  1231.         src7 = *s; s+=srcStride;\

  1232.         OP(*d, (src4+src5)*20 - (src3+src6)*5 + (src2+src7));d+=dstStride;\

  1233.         src8 = *s; s+=srcStride;\

  1234.         OP(*d, (src5+src6)*20 - (src4+src7)*5 + (src3+src8));d+=dstStride;\

  1235.         src9 = *s; s+=srcStride;\

  1236.         OP(*d, (src6+src7)*20 - (src5+src8)*5 + (src4+src9));d+=dstStride;\

  1237.         src10 = *s; s+=srcStride;\

  1238.         OP(*d, (src7+src8)*20 - (src6+src9)*5 + (src5+src10));d+=dstStride;\

  1239.        if (h>8) { \

  1240.         int src11,src12,src13,src14,src15,src16,src17,src18; \

  1241.         src11 = *s; s+=srcStride;\

  1242.         OP(*d , (src8 +src9 )*20 - (src7 +src10)*5 + (src6 +src11));d+=dstStride;\

  1243.         src12 = *s; s+=srcStride;\

  1244.         OP(*d , (src9 +src10)*20 - (src8 +src11)*5 + (src7 +src12));d+=dstStride;\

  1245.         src13 = *s; s+=srcStride;\

  1246.         OP(*d, (src10+src11)*20 - (src9 +src12)*5 + (src8 +src13));d+=dstStride;\

  1247.         src14 = *s; s+=srcStride;\

  1248.         OP(*d, (src11+src12)*20 - (src10+src13)*5 + (src9 +src14));d+=dstStride;\

  1249.         src15 = *s; s+=srcStride;\

  1250.         OP(*d, (src12+src13)*20 - (src11+src14)*5 + (src10+src15));d+=dstStride;\

  1251.         src16 = *s; s+=srcStride;\

  1252.         OP(*d, (src13+src14)*20 - (src12+src15)*5 + (src11+src16));d+=dstStride;\

  1253.         src17 = *s; s+=srcStride;\

  1254.         OP(*d, (src14+src15)*20 - (src13+src16)*5 + (src12+src17));d+=dstStride;\

  1255.         src18 = *s; s+=srcStride;\

  1256.         OP(*d, (src15+src16)*20 - (src14+src17)*5 + (src13+src18));d+=dstStride;\

  1257.        } \

  1258.       } \

  1259.         dst++;\

  1260.         src++;\

  1261.     }while(--w);\

  1262. }\

  1263. \

  1264. static inline void OPNAME ## h264_qpel_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride,int w,int h){\

  1265.     uint8_t *cm = cropTbl + MAX_NEG_CROP;\

  1266.     int i;\

  1267.     src -= 2*srcStride;\

  1268.     i= h+5; \

  1269.     do {\

  1270.         int srcB,srcA,src0,src1,src2,src3,src4,src5,src6;\

  1271.         uint8_t *s = src-2;\

  1272.         srcB = *s++;\

  1273.         srcA = *s++;\

  1274.         src0 = *s++;\

  1275.         src1 = *s++;\

  1276.         src2 = *s++;\

  1277.         src3 = *s++;\

  1278.         tmp[0] = ((src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));\

  1279.         src4 = *s++;\

  1280.         tmp[1] = ((src1+src2)*20 - (src0+src3)*5 + (srcA+src4));\

  1281.         src5 = *s++;\

  1282.         tmp[2] = ((src2+src3)*20 - (src1+src4)*5 + (src0+src5));\

  1283.         src6 = *s++;\

  1284.         tmp[3] = ((src3+src4)*20 - (src2+src5)*5 + (src1+src6));\

  1285.       if (w>4) { /* it optimized */ \

  1286.         int src7,src8,src9,src10; \

  1287.         src7 = *s++;\

  1288.         tmp[4] = ((src4+src5)*20 - (src3+src6)*5 + (src2+src7));\

  1289.         src8 = *s++;\

  1290.         tmp[5] = ((src5+src6)*20 - (src4+src7)*5 + (src3+src8));\

  1291.         src9 = *s++;\

  1292.         tmp[6] = ((src6+src7)*20 - (src5+src8)*5 + (src4+src9));\

  1293.         src10 = *s++;\

  1294.         tmp[7] = ((src7+src8)*20 - (src6+src9)*5 + (src5+src10));\

  1295.        if (w>8) { \

  1296.         int src11,src12,src13,src14,src15,src16,src17,src18; \

  1297.         src11 = *s++;\

  1298.         tmp[8] = ((src8 +src9 )*20 - (src7 +src10)*5 + (src6 +src11));\

  1299.         src12 = *s++;\

  1300.         tmp[9] = ((src9 +src10)*20 - (src8 +src11)*5 + (src7 +src12));\

  1301.         src13 = *s++;\

  1302.         tmp[10] = ((src10+src11)*20 - (src9 +src12)*5 + (src8 +src13));\

  1303.         src14 = *s++;\

  1304.         tmp[11] = ((src11+src12)*20 - (src10+src13)*5 + (src9 +src14));\

  1305.         src15 = *s++;\

  1306.         tmp[12] = ((src12+src13)*20 - (src11+src14)*5 + (src10+src15));\

  1307.         src16 = *s++;\

  1308.         tmp[13] = ((src13+src14)*20 - (src12+src15)*5 + (src11+src16));\

  1309.         src17 = *s++;\

  1310.         tmp[14] = ((src14+src15)*20 - (src13+src16)*5 + (src12+src17));\

  1311.         src18 = *s++;\

  1312.         tmp[15] = ((src15+src16)*20 - (src14+src17)*5 + (src13+src18));\

  1313.        } \

  1314.       } \

  1315.         tmp+=tmpStride;\

  1316.         src+=srcStride;\

  1317.     }while(--i);\

  1318.     tmp -= tmpStride*(h+5-2);\

  1319.     i = w; \

  1320.     do {\

  1321.         int tmpB,tmpA,tmp0,tmp1,tmp2,tmp3,tmp4,tmp5,tmp6;\

  1322.         int16_t *s = tmp-2*tmpStride; \

  1323.         uint8_t *d=dst;\

  1324.         tmpB = *s; s+=tmpStride;\

  1325.         tmpA = *s; s+=tmpStride;\

  1326.         tmp0 = *s; s+=tmpStride;\

  1327.         tmp1 = *s; s+=tmpStride;\

  1328.         tmp2 = *s; s+=tmpStride;\

  1329.         tmp3 = *s; s+=tmpStride;\

  1330.         OP2(*d, (tmp0+tmp1)*20 - (tmpA+tmp2)*5 + (tmpB+tmp3));d+=dstStride;\

  1331.         tmp4 = *s; s+=tmpStride;\

  1332.         OP2(*d, (tmp1+tmp2)*20 - (tmp0+tmp3)*5 + (tmpA+tmp4));d+=dstStride;\

  1333.         tmp5 = *s; s+=tmpStride;\

  1334.         OP2(*d, (tmp2+tmp3)*20 - (tmp1+tmp4)*5 + (tmp0+tmp5));d+=dstStride;\

  1335.         tmp6 = *s; s+=tmpStride;\

  1336.         OP2(*d, (tmp3+tmp4)*20 - (tmp2+tmp5)*5 + (tmp1+tmp6));d+=dstStride;\

  1337.       if (h>4) { \

  1338.         int tmp7,tmp8,tmp9,tmp10; \

  1339.         tmp7 = *s; s+=tmpStride;\

  1340.         OP2(*d, (tmp4+tmp5)*20 - (tmp3+tmp6)*5 + (tmp2+tmp7));d+=dstStride;\

  1341.         tmp8 = *s; s+=tmpStride;\

  1342.         OP2(*d, (tmp5+tmp6)*20 - (tmp4+tmp7)*5 + (tmp3+tmp8));d+=dstStride;\

  1343.         tmp9 = *s; s+=tmpStride;\

  1344.         OP2(*d, (tmp6+tmp7)*20 - (tmp5+tmp8)*5 + (tmp4+tmp9));d+=dstStride;\

  1345.         tmp10 = *s; s+=tmpStride;\

  1346.         OP2(*d, (tmp7+tmp8)*20 - (tmp6+tmp9)*5 + (tmp5+tmp10));d+=dstStride;\

  1347.        if (h>8) { \

  1348.         int tmp11,tmp12,tmp13,tmp14,tmp15,tmp16,tmp17,tmp18; \

  1349.         tmp11 = *s; s+=tmpStride;\

  1350.         OP2(*d , (tmp8 +tmp9 )*20 - (tmp7 +tmp10)*5 + (tmp6 +tmp11));d+=dstStride;\

  1351.         tmp12 = *s; s+=tmpStride;\

  1352.         OP2(*d , (tmp9 +tmp10)*20 - (tmp8 +tmp11)*5 + (tmp7 +tmp12));d+=dstStride;\

  1353.         tmp13 = *s; s+=tmpStride;\

  1354.         OP2(*d, (tmp10+tmp11)*20 - (tmp9 +tmp12)*5 + (tmp8 +tmp13));d+=dstStride;\

  1355.         tmp14 = *s; s+=tmpStride;\

  1356.         OP2(*d, (tmp11+tmp12)*20 - (tmp10+tmp13)*5 + (tmp9 +tmp14));d+=dstStride;\

  1357.         tmp15 = *s; s+=tmpStride;\

  1358.         OP2(*d, (tmp12+tmp13)*20 - (tmp11+tmp14)*5 + (tmp10+tmp15));d+=dstStride;\

  1359.         tmp16 = *s; s+=tmpStride;\

  1360.         OP2(*d, (tmp13+tmp14)*20 - (tmp12+tmp15)*5 + (tmp11+tmp16));d+=dstStride;\

  1361.         tmp17 = *s; s+=tmpStride;\

  1362.         OP2(*d, (tmp14+tmp15)*20 - (tmp13+tmp16)*5 + (tmp12+tmp17));d+=dstStride;\

  1363.         tmp18 = *s; s+=tmpStride;\

  1364.         OP2(*d, (tmp15+tmp16)*20 - (tmp14+tmp17)*5 + (tmp13+tmp18));d+=dstStride;\

  1365.        } \

  1366.       } \

  1367.         dst++;\

  1368.         tmp++;\

  1369.     }while(--i);\

  1370. }\

  1371. \

  1372. static void OPNAME ## h264_qpel4_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1373.     OPNAME ## h264_qpel_h_lowpass(dst,src,dstStride,srcStride,4,4); \

  1374. }\

  1375. static void OPNAME ## h264_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1376.    OPNAME ## h264_qpel_h_lowpass(dst,src,dstStride,srcStride,8,8); \

  1377. }\

  1378. static void OPNAME ## h264_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1379.    OPNAME ## h264_qpel_h_lowpass(dst,src,dstStride,srcStride,16,16); \

  1380. }\

  1381. \

  1382. static void OPNAME ## h264_qpel4_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1383.    OPNAME ## h264_qpel_v_lowpass(dst,src,dstStride,srcStride,4,4); \

  1384. }\

  1385. static void OPNAME ## h264_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1386.    OPNAME ## h264_qpel_v_lowpass(dst,src,dstStride,srcStride,8,8); \

  1387. }\

  1388. static void OPNAME ## h264_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1389.    OPNAME ## h264_qpel_v_lowpass(dst,src,dstStride,srcStride,16,16); \

  1390. }\

  1391. static void OPNAME ## h264_qpel4_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\

  1392.    OPNAME ## h264_qpel_hv_lowpass(dst,tmp,src,dstStride,tmpStride,srcStride,4,4); \

  1393. }\

  1394. static void OPNAME ## h264_qpel8_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\

  1395.    OPNAME ## h264_qpel_hv_lowpass(dst,tmp,src,dstStride,tmpStride,srcStride,8,8); \

  1396. }\

  1397. static void OPNAME ## h264_qpel16_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\

  1398.    OPNAME ## h264_qpel_hv_lowpass(dst,tmp,src,dstStride,tmpStride,srcStride,16,16); \

  1399. }\

  1400. 

  1401. #define H264_MC(OPNAME, SIZE) \

  1402. static void OPNAME ## h264_qpel ## SIZE ## _mc00_c (uint8_t *dst, uint8_t *src, int stride){\

  1403.     OPNAME ## pixels ## SIZE ## _c(dst, src, stride, SIZE);\

  1404. }\

  1405. \

  1406. static void OPNAME ## h264_qpel ## SIZE ## _mc10_c(uint8_t *dst, uint8_t *src, int stride){\

  1407.     uint8_t half[SIZE*SIZE];\

  1408.     put_h264_qpel ## SIZE ## _h_lowpass(half, src, SIZE, stride);\

  1409.     OPNAME ## pixels ## SIZE ## _l2_aligned2(dst, src, half, stride, stride, SIZE, SIZE);\

  1410. }\

  1411. \

  1412. static void OPNAME ## h264_qpel ## SIZE ## _mc20_c(uint8_t *dst, uint8_t *src, int stride){\

  1413.     OPNAME ## h264_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride);\

  1414. }\

  1415. \

  1416. static void OPNAME ## h264_qpel ## SIZE ## _mc30_c(uint8_t *dst, uint8_t *src, int stride){\

  1417.     uint8_t half[SIZE*SIZE];\

  1418.     put_h264_qpel ## SIZE ## _h_lowpass(half, src, SIZE, stride);\

  1419.     OPNAME ## pixels ## SIZE ## _l2_aligned2(dst, src+1, half, stride, stride, SIZE, SIZE);\

  1420. }\

  1421. \

  1422. static void OPNAME ## h264_qpel ## SIZE ## _mc01_c(uint8_t *dst, uint8_t *src, int stride){\

  1423.     uint8_t full[SIZE*(SIZE+5)];\

  1424.     uint8_t * const full_mid= full + SIZE*2;\

  1425.     uint8_t half[SIZE*SIZE];\

  1426.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1427.     put_h264_qpel ## SIZE ## _v_lowpass(half, full_mid, SIZE, SIZE);\

  1428.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, full_mid, half, stride, SIZE, SIZE, SIZE);\

  1429. }\

  1430. \

  1431. static void OPNAME ## h264_qpel ## SIZE ## _mc02_c(uint8_t *dst, uint8_t *src, int stride){\

  1432.     uint8_t full[SIZE*(SIZE+5)];\

  1433.     uint8_t * const full_mid= full + SIZE*2;\

  1434.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1435.     OPNAME ## h264_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE);\

  1436. }\

  1437. \

  1438. static void OPNAME ## h264_qpel ## SIZE ## _mc03_c(uint8_t *dst, uint8_t *src, int stride){\

  1439.     uint8_t full[SIZE*(SIZE+5)];\

  1440.     uint8_t * const full_mid= full + SIZE*2;\

  1441.     uint8_t half[SIZE*SIZE];\

  1442.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1443.     put_h264_qpel ## SIZE ## _v_lowpass(half, full_mid, SIZE, SIZE);\

  1444.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, full_mid+SIZE, half, stride, SIZE, SIZE, SIZE);\

  1445. }\

  1446. \

  1447. static void OPNAME ## h264_qpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, int stride){\

  1448.     uint8_t full[SIZE*(SIZE+5)];\

  1449.     uint8_t * const full_mid= full + SIZE*2;\

  1450.     uint8_t halfH[SIZE*SIZE];\

  1451.     uint8_t halfV[SIZE*SIZE];\

  1452.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\

  1453.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1454.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1455.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\

  1456. }\

  1457. \

  1458. static void OPNAME ## h264_qpel ## SIZE ## _mc31_c(uint8_t *dst, uint8_t *src, int stride){\

  1459.     uint8_t full[SIZE*(SIZE+5)];\

  1460.     uint8_t * const full_mid= full + SIZE*2;\

  1461.     uint8_t halfH[SIZE*SIZE];\

  1462.     uint8_t halfV[SIZE*SIZE];\

  1463.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\

  1464.     copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\

  1465.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1466.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\

  1467. }\

  1468. \

  1469. static void OPNAME ## h264_qpel ## SIZE ## _mc13_c(uint8_t *dst, uint8_t *src, int stride){\

  1470.     uint8_t full[SIZE*(SIZE+5)];\

  1471.     uint8_t * const full_mid= full + SIZE*2;\

  1472.     uint8_t halfH[SIZE*SIZE];\

  1473.     uint8_t halfV[SIZE*SIZE];\

  1474.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\

  1475.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1476.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1477.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\

  1478. }\

  1479. \

  1480. static void OPNAME ## h264_qpel ## SIZE ## _mc33_c(uint8_t *dst, uint8_t *src, int stride){\

  1481.     uint8_t full[SIZE*(SIZE+5)];\

  1482.     uint8_t * const full_mid= full + SIZE*2;\

  1483.     uint8_t halfH[SIZE*SIZE];\

  1484.     uint8_t halfV[SIZE*SIZE];\

  1485.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\

  1486.     copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\

  1487.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1488.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\

  1489. }\

  1490. \

  1491. static void OPNAME ## h264_qpel ## SIZE ## _mc22_c(uint8_t *dst, uint8_t *src, int stride){\

  1492.     int16_t tmp[SIZE*(SIZE+5)];\

  1493.     OPNAME ## h264_qpel ## SIZE ## _hv_lowpass(dst, tmp, src, stride, SIZE, stride);\

  1494. }\

  1495. \

  1496. static void OPNAME ## h264_qpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, int stride){\

  1497.     int16_t tmp[SIZE*(SIZE+5)];\

  1498.     uint8_t halfH[SIZE*SIZE];\

  1499.     uint8_t halfHV[SIZE*SIZE];\

  1500.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\

  1501.     put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\

  1502.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfHV, stride, SIZE, SIZE, SIZE);\

  1503. }\

  1504. \

  1505. static void OPNAME ## h264_qpel ## SIZE ## _mc23_c(uint8_t *dst, uint8_t *src, int stride){\

  1506.     int16_t tmp[SIZE*(SIZE+5)];\

  1507.     uint8_t halfH[SIZE*SIZE];\

  1508.     uint8_t halfHV[SIZE*SIZE];\

  1509.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\

  1510.     put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\

  1511.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfHV, stride, SIZE, SIZE, SIZE);\

  1512. }\

  1513. \

  1514. static void OPNAME ## h264_qpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, int stride){\

  1515.     uint8_t full[SIZE*(SIZE+5)];\

  1516.     uint8_t * const full_mid= full + SIZE*2;\

  1517.     int16_t tmp[SIZE*(SIZE+5)];\

  1518.     uint8_t halfV[SIZE*SIZE];\

  1519.     uint8_t halfHV[SIZE*SIZE];\

  1520.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1521.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1522.     put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\

  1523.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfV, halfHV, stride, SIZE, SIZE, SIZE);\

  1524. }\

  1525. \

  1526. static void OPNAME ## h264_qpel ## SIZE ## _mc32_c(uint8_t *dst, uint8_t *src, int stride){\

  1527.     uint8_t full[SIZE*(SIZE+5)];\

  1528.     uint8_t * const full_mid= full + SIZE*2;\

  1529.     int16_t tmp[SIZE*(SIZE+5)];\

  1530.     uint8_t halfV[SIZE*SIZE];\

  1531.     uint8_t halfHV[SIZE*SIZE];\

  1532.     copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\

  1533.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1534.     put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\

  1535.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfV, halfHV, stride, SIZE, SIZE, SIZE);\

  1536. }\

  1537. 

  1538. #define op_avg(a, b)  a = (((a)+cm[((b) + 16)>>5]+1)>>1)

  1539. //#define op_avg2(a, b) a = (((a)*w1+cm[((b) + 16)>>5]*w2 + o + 64)>>7)

  1540. #define op_put(a, b)  a = cm[((b) + 16)>>5]

  1541. #define op2_avg(a, b)  a = (((a)+cm[((b) + 512)>>10]+1)>>1)

  1542. #define op2_put(a, b)  a = cm[((b) + 512)>>10]

  1543. 

  1544. H264_LOWPASS(put_       , op_put, op2_put)

  1545. H264_LOWPASS(avg_       , op_avg, op2_avg)

  1546. H264_MC(put_, 4)

  1547. H264_MC(put_, 8)

  1548. H264_MC(put_, 16)

  1549. H264_MC(avg_, 4)

  1550. H264_MC(avg_, 8)

  1551. H264_MC(avg_, 16)

  1552. 

  1553. #undef op_avg

  1554. #undef op_put

  1555. #undef op2_avg

  1556. #undef op2_put

  1557. #endif

  1558. 

  1559. static void wmv2_mspel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){

  1560.     uint8_t *cm = cropTbl + MAX_NEG_CROP;

  1561. 

  1562.     do{

  1563.         int src_1,src0,src1,src2,src3,src4,src5,src6,src7,src8,src9;

  1564.         uint8_t *s = src;

  1565.         src_1 = s[-1];

  1566.         src0 = *s++;

  1567.         src1 = *s++;

  1568.         src2 = *s++;

  1569.         dst[0]= cm[(9*(src0 + src1) - (src_1 + src2) + 8)>>4];

  1570.         src3 = *s++;

  1571.         dst[1]= cm[(9*(src1 + src2) - (src0 + src3) + 8)>>4];

  1572.         src4 = *s++;

  1573.         dst[2]= cm[(9*(src2 + src3) - (src1 + src4) + 8)>>4];

  1574.         src5 = *s++;

  1575.         dst[3]= cm[(9*(src3 + src4) - (src2 + src5) + 8)>>4];

  1576.         src6 = *s++;

  1577.         dst[4]= cm[(9*(src4 + src5) - (src3 + src6) + 8)>>4];

  1578.         src7 = *s++;

  1579.         dst[5]= cm[(9*(src5 + src6) - (src4 + src7) + 8)>>4];

  1580.         src8 = *s++;

  1581.         dst[6]= cm[(9*(src6 + src7) - (src5 + src8) + 8)>>4];

  1582.         src9 = *s++;

  1583.         dst[7]= cm[(9*(src7 + src8) - (src6 + src9) + 8)>>4];

  1584.         dst+=dstStride;

  1585.         src+=srcStride;

  1586.     }while(--h);

  1587. }

  1588. 

  1589. static void wmv2_mspel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int w){

  1590.     uint8_t *cm = cropTbl + MAX_NEG_CROP;

  1591. 

  1592.     do{

  1593.         int src_1,src0,src1,src2,src3,src4,src5,src6,src7,src8,src9;

  1594.         uint8_t *s = src,*d = dst;

  1595.         src_1 = *(s-srcStride);

  1596.         src0 = *s; s+=srcStride;

  1597.         src1 = *s; s+=srcStride;

  1598.         src2 = *s; s+=srcStride;

  1599.         *d= cm[(9*(src0 + src1) - (src_1 + src2) + 8)>>4]; d+=dstStride;

  1600.         src3 = *s; s+=srcStride;

  1601.         *d= cm[(9*(src1 + src2) - (src0  + src3) + 8)>>4]; d+=dstStride;

  1602.         src4 = *s; s+=srcStride;

  1603.         *d= cm[(9*(src2 + src3) - (src1  + src4) + 8)>>4]; d+=dstStride;

  1604.         src5 = *s; s+=srcStride;

  1605.         *d= cm[(9*(src3 + src4) - (src2  + src5) + 8)>>4]; d+=dstStride;

  1606.         src6 = *s; s+=srcStride;

  1607.         *d= cm[(9*(src4 + src5) - (src3  + src6) + 8)>>4]; d+=dstStride;

  1608.         src7 = *s; s+=srcStride;

  1609.         *d= cm[(9*(src5 + src6) - (src4  + src7) + 8)>>4]; d+=dstStride;

  1610.         src8 = *s; s+=srcStride;

  1611.         *d= cm[(9*(src6 + src7) - (src5  + src8) + 8)>>4]; d+=dstStride;

  1612.         src9 = *s;

  1613.         *d= cm[(9*(src7 + src8) - (src6  + src9) + 8)>>4]; d+=dstStride;

  1614.         src++;

  1615.         dst++;

  1616.     }while(--w);

  1617. }

  1618. 

  1619. static void put_mspel8_mc00_c (uint8_t *dst, uint8_t *src, int stride){

  1620.     put_pixels8_c(dst, src, stride, 8);

  1621. }

  1622. 

  1623. static void put_mspel8_mc10_c(uint8_t *dst, uint8_t *src, int stride){

  1624.     uint8_t half[64];

  1625.     wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);

  1626.     put_pixels8_l2_aligned2(dst, src, half, stride, stride, 8, 8);

  1627. }

  1628. 

  1629. static void put_mspel8_mc20_c(uint8_t *dst, uint8_t *src, int stride){

  1630.     wmv2_mspel8_h_lowpass(dst, src, stride, stride, 8);

  1631. }

  1632. 

  1633. static void put_mspel8_mc30_c(uint8_t *dst, uint8_t *src, int stride){

  1634.     uint8_t half[64];

  1635.     wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);

  1636.     put_pixels8_l2_aligned2(dst, src+1, half, stride, stride, 8, 8);

  1637. }

  1638. 

  1639. static void put_mspel8_mc02_c(uint8_t *dst, uint8_t *src, int stride){

  1640.     wmv2_mspel8_v_lowpass(dst, src, stride, stride, 8);

  1641. }

  1642. 

  1643. static void put_mspel8_mc12_c(uint8_t *dst, uint8_t *src, int stride){

  1644.     uint8_t halfH[88];

  1645.     uint8_t halfV[64];

  1646.     uint8_t halfHV[64];

  1647.     wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);

  1648.     wmv2_mspel8_v_lowpass(halfV, src, 8, stride, 8);

  1649.     wmv2_mspel8_v_lowpass(halfHV, halfH+8, 8, 8, 8);

  1650.     put_pixels8_l2_aligned(dst, halfV, halfHV, stride, 8, 8, 8);

  1651. }

  1652. static void put_mspel8_mc32_c(uint8_t *dst, uint8_t *src, int stride){

  1653.     uint8_t halfH[88];

  1654.     uint8_t halfV[64];

  1655.     uint8_t halfHV[64];

  1656.     wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);

  1657.     wmv2_mspel8_v_lowpass(halfV, src+1, 8, stride, 8);

  1658.     wmv2_mspel8_v_lowpass(halfHV, halfH+8, 8, 8, 8);

  1659.     put_pixels8_l2_aligned(dst, halfV, halfHV, stride, 8, 8, 8);

  1660. }

  1661. static void put_mspel8_mc22_c(uint8_t *dst, uint8_t *src, int stride){

  1662.     uint8_t halfH[88];

  1663.     wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);

  1664.     wmv2_mspel8_v_lowpass(dst, halfH+8, stride, 8, 8);

  1665. }