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.
9447 Commits
32 Branches
582MB
Tree: 4440b1188e
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '4440b1188e'
${ noResults }
FFmpeg/libavcodec/sh4/qpel.c

1601 lines
69KB
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 file is part of FFmpeg.

  8.  *

  9.  * FFmpeg is free software; you can redistribute it and/or

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

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

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

  13.  *

  14.  * FFmpeg is distributed in the hope that it will be useful,

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

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

  17.  * Lesser General Public License for more details.

  18.  *

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

  20.  * License along with FFmpeg; if not, write to the Free Software

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

  22.  */

  23. 

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

  25. 

  26. #define PIXOP2(OPNAME, OP) \

  27. /*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) \

  28. {\

  29.         do {\

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

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

  32.                 src1+=src_stride1; \

  33.                 src2+=src_stride2; \

  34.                 dst+=dst_stride; \

  35.         } while(--h); \

  36. }\

  37. \

  38. 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) \

  39. {\

  40.         do {\

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

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

  43.                 src1+=src_stride1; \

  44.                 src2+=src_stride2; \

  45.                 dst+=dst_stride; \

  46.         } while(--h); \

  47. }\

  48. \

  49. 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) \

  50. {\

  51.         do {\

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

  53.                 src1+=src_stride1; \

  54.                 src2+=src_stride2; \

  55.                 dst+=dst_stride; \

  56.         } while(--h); \

  57. }\

  58. \

  59. 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) \

  60. {\

  61.         do {\

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

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

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

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

  66.                 src1+=src_stride1; \

  67.                 src2+=src_stride2; \

  68.                 dst+=dst_stride; \

  69.         } while(--h); \

  70. }\

  71. \

  72. 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) \

  73. {\

  74.         do {\

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

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

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

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

  79.                 src1+=src_stride1; \

  80.                 src2+=src_stride2; \

  81.                 dst+=dst_stride; \

  82.         } while(--h); \

  83. }*/\

  84. \

  85. 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) \

  86. {\

  87.         do {\

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

  89.                 src1+=src_stride1; \

  90.                 src2+=src_stride2; \

  91.                 dst+=dst_stride; \

  92.         } while(--h); \

  93. }\

  94. \

  95. 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) \

  96. {\

  97.         do {\

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

  99.                 src1+=src_stride1; \

  100.                 src2+=src_stride2; \

  101.                 dst+=dst_stride; \

  102.         } while(--h); \

  103. }\

  104. \

  105. 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) \

  106. {\

  107.         do {\

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

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

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

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

  112.                 src1+=src_stride1; \

  113.                 src2+=src_stride2; \

  114.                 dst+=dst_stride; \

  115.         } while(--h); \

  116. }\

  117. \

  118. 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) \

  119. {\

  120.         do {\

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

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

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

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

  125.                 src1+=src_stride1; \

  126.                 src2+=src_stride2; \

  127.                 dst+=dst_stride; \

  128.         } while(--h); \

  129. }\

  130. \

  131. 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) \

  132. {\

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

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

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

  136.                 src1+=src_stride1; \

  137.                 src2+=src_stride2; \

  138.                 dst+=dst_stride; \

  139.         } while(--h); \

  140. }\

  141. \

  142. 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) \

  143. {\

  144.         do {\

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

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

  147.                 src1+=src_stride1; \

  148.                 src2+=src_stride2; \

  149.                 dst+=dst_stride; \

  150.         } while(--h); \

  151. }\

  152. \

  153. 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) \

  154. {\

  155.         do {\

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

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

  158.                 src1+=src_stride1; \

  159.                 src2+=src_stride2; \

  160.                 dst+=dst_stride; \

  161.         } while(--h); \

  162. }\

  163. \

  164. 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) \

  165. {\

  166.         do {\

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

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

  169.                 src1+=src_stride1; \

  170.                 src2+=src_stride2; \

  171.                 dst+=dst_stride; \

  172.         } while(--h); \

  173. }\

  174. \

  175. 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) \

  176. {\

  177.         do {\

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

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

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

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

  182.                 src1+=src_stride1; \

  183.                 src2+=src_stride2; \

  184.                 dst+=dst_stride; \

  185.         } while(--h); \

  186. }\

  187. \

  188. 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) \

  189. {\

  190.         do {\

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

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

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

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

  195.                 src1+=src_stride1; \

  196.                 src2+=src_stride2; \

  197.                 dst+=dst_stride; \

  198.         } while(--h); \

  199. }\

  200. \

  201. 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) \

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

  203. \

  204. 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) \

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

  206. \

  207. 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) \

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

  209. \

  210. 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) \

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

  212. \

  213. 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){\

  214.         do { \

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

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

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

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

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

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

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

  222.                 src1+=src_stride1;\

  223.                 src2+=src_stride2;\

  224.                 src3+=src_stride3;\

  225.                 src4+=src_stride4;\

  226.                 dst+=dst_stride;\

  227.         } while(--h); \

  228. } \

  229. \

  230. 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){\

  231.         do { \

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

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

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

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

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

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

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

  239.                 src1+=src_stride1;\

  240.                 src2+=src_stride2;\

  241.                 src3+=src_stride3;\

  242.                 src4+=src_stride4;\

  243.                 dst+=dst_stride;\

  244.         } while(--h); \

  245. } \

  246. \

  247. 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){\

  248.         do { \

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

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

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

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

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

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

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

  256.                 src1+=src_stride1;\

  257.                 src2+=src_stride2;\

  258.                 src3+=src_stride3;\

  259.                 src4+=src_stride4;\

  260.                 dst+=dst_stride;\

  261.         } while(--h); \

  262. } \

  263. \

  264. 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){\

  265.         do { \

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

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

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

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

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

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

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

  273.                 src1+=src_stride1;\

  274.                 src2+=src_stride2;\

  275.                 src3+=src_stride3;\

  276.                 src4+=src_stride4;\

  277.                 dst+=dst_stride;\

  278.         } while(--h); \

  279. } \

  280. \

  281. 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){\

  282.         do { \

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

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

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

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

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

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

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

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

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

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

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

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

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

  296.                 src1+=src_stride1;\

  297.                 src2+=src_stride2;\

  298.                 src3+=src_stride3;\

  299.                 src4+=src_stride4;\

  300.                 dst+=dst_stride;\

  301.         } while(--h); \

  302. } \

  303. \

  304. 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){\

  305.         do { \

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

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

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

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

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

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

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

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

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

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

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

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

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

  319.                 src1+=src_stride1;\

  320.                 src2+=src_stride2;\

  321.                 src3+=src_stride3;\

  322.                 src4+=src_stride4;\

  323.                 dst+=dst_stride;\

  324.         } while(--h); \

  325. } \

  326. \

  327. 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){\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  342.                 src1+=src_stride1;\

  343.                 src2+=src_stride2;\

  344.                 src3+=src_stride3;\

  345.                 src4+=src_stride4;\

  346.                 dst+=dst_stride;\

  347.         } while(--h); \

  348. } \

  349. \

  350. 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){\

  351.         do { \

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

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

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

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

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

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

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

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

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

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

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

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

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

  365.                 src1+=src_stride1;\

  366.                 src2+=src_stride2;\

  367.                 src3+=src_stride3;\

  368.                 src4+=src_stride4;\

  369.                 dst+=dst_stride;\

  370.         } while(--h); \

  371. } \

  372. \

  373. 

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

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

  376. 

  377. PIXOP2(avg, op_avg)

  378. PIXOP2(put, op_put)

  379. #undef op_avg

  380. #undef op_put

  381. 

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

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

  384. 

  385. 

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

  387. {

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

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

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

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

  392. 

  393.     do {

  394.         int t0,t1,t2,t3;

  395.         uint8_t *s0 = src;

  396.         uint8_t *s1 = src+stride;

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  414.         dst+= stride;

  415.         src+= stride;

  416.     }while(--h);

  417. }

  418. 

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

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

  421. {

  422.     int y, vx, vy;

  423.     const int s= 1<<shift;

  424. 

  425.     width--;

  426.     height--;

  427. 

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

  429.         int x;

  430. 

  431.         vx= ox;

  432.         vy= oy;

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

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

  435. 

  436.             src_x= vx>>16;

  437.             src_y= vy>>16;

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

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

  440.             src_x>>=shift;

  441.             src_y>>=shift;

  442. 

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

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

  445.                     index= src_x + src_y*stride;

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

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

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

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

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

  451.                 }else{

  452.                     index= src_x + av_clip(src_y, 0, height)*stride;

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

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

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

  456.                 }

  457.             }else{

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

  459.                     index= av_clip(src_x, 0, width) + src_y*stride;

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

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

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

  463.                 }else{

  464.                     index= av_clip(src_x, 0, width) + av_clip(src_y, 0, height)*stride;

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

  466.                 }

  467.             }

  468. 

  469.             vx+= dxx;

  470.             vy+= dyx;

  471.         }

  472.         ox += dxy;

  473.         oy += dyy;

  474.     }

  475. }

  476. #define H264_CHROMA_MC(OPNAME, OP)\

  477. 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){\

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

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

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

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

  482.     \

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

  484. \

  485.     do {\

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

  487.         uint8_t *s0 = src; \

  488.         uint8_t *s1 = src+stride; \

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

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

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

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

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

  494.         dst+= stride;\

  495.         src+= stride;\

  496.     }while(--h);\

  497. }\

  498. \

  499. 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){\

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

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

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

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

  504.     \

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

  506. \

  507.     do {\

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

  509.         uint8_t *s0 = src; \

  510.         uint8_t *s1 = src+stride; \

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

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

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

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

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

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

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

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

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

  520.         dst+= stride;\

  521.         src+= stride;\

  522.     }while(--h);\

  523. }\

  524. \

  525. 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){\

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

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

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

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

  530.     \

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

  532. \

  533.     do {\

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

  535.         uint8_t *s0 = src; \

  536.         uint8_t *s1 = src+stride; \

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  554.         dst+= stride;\

  555.         src+= stride;\

  556.     }while(--h);\

  557. }

  558. 

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

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

  561. 

  562. H264_CHROMA_MC(put_       , op_put)

  563. H264_CHROMA_MC(avg_       , op_avg)

  564. #undef op_avg

  565. #undef op_put

  566. 

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

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

  569.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  570.     do {\

  571.         uint8_t *s = src; \

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

  573.         src0= *s++;\

  574.         src1= *s++;\

  575.         src2= *s++;\

  576.         src3= *s++;\

  577.         src4= *s++;\

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

  579.         src5= *s++;\

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

  581.         src6= *s++;\

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

  583.         src7= *s++;\

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

  585.         src8= *s++;\

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

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

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

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

  590.         dst+=dstStride;\

  591.         src+=srcStride;\

  592.     }while(--h);\

  593. }\

  594. \

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

  596.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  597.     int w=8;\

  598.     do{\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  613.         src8 = *s; \

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

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

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

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

  618.         dst++;\

  619.         src++;\

  620.     }while(--w);\

  621. }\

  622. \

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

  624.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  625.     do {\

  626.         uint8_t *s = src;\

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

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

  629.         src0= *s++;\

  630.         src1= *s++;\

  631.         src2= *s++;\

  632.         src3= *s++;\

  633.         src4= *s++;\

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

  635.         src5= *s++;\

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

  637.         src6= *s++;\

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

  639.         src7= *s++;\

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

  641.         src8= *s++;\

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

  643.         src9= *s++;\

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

  645.         src10= *s++;\

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

  647.         src11= *s++;\

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

  649.         src12= *s++;\

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

  651.         src13= *s++;\

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

  653.         src14= *s++;\

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

  655.         src15= *s++;\

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

  657.         src16= *s++;\

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

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

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

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

  662.         dst+=dstStride;\

  663.         src+=srcStride;\

  664.     }while(--h);\

  665. }\

  666. \

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

  668.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  669.     int w=16;\

  670.     do {\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  702.         src16 = *s; \

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

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

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

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

  707.         dst++;\

  708.         src++;\

  709.     }while(--w);\

  710. }\

  711. \

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

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

  714. }\

  715. \

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

  717.     uint8_t half[64];\

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

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

  720. }\

  721. \

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

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

  724. }\

  725. \

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

  727.     uint8_t half[64];\

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

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

  730. }\

  731. \

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

  733.     uint8_t full[16*9];\

  734.     uint8_t half[64];\

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

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

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

  738. }\

  739. \

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

  741.     uint8_t full[16*9];\

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

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

  744. }\

  745. \

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

  747.     uint8_t full[16*9];\

  748.     uint8_t half[64];\

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

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

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

  752. }\

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

  754.     uint8_t full[16*9];\

  755.     uint8_t halfH[72];\

  756.     uint8_t halfV[64];\

  757.     uint8_t halfHV[64];\

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

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

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

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

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

  763. }\

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

  765.     uint8_t full[16*9];\

  766.     uint8_t halfH[72];\

  767.     uint8_t halfHV[64];\

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

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

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

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

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

  773. }\

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

  775.     uint8_t full[16*9];\

  776.     uint8_t halfH[72];\

  777.     uint8_t halfV[64];\

  778.     uint8_t halfHV[64];\

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

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

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

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

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

  784. }\

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

  786.     uint8_t full[16*9];\

  787.     uint8_t halfH[72];\

  788.     uint8_t halfHV[64];\

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

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

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

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

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

  794. }\

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

  796.     uint8_t full[16*9];\

  797.     uint8_t halfH[72];\

  798.     uint8_t halfV[64];\

  799.     uint8_t halfHV[64];\

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

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

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

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

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

  805. }\

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

  807.     uint8_t full[16*9];\

  808.     uint8_t halfH[72];\

  809.     uint8_t halfHV[64];\

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

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

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

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

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

  815. }\

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

  817.     uint8_t full[16*9];\

  818.     uint8_t halfH[72];\

  819.     uint8_t halfV[64];\

  820.     uint8_t halfHV[64];\

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

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

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

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

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

  826. }\

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

  828.     uint8_t full[16*9];\

  829.     uint8_t halfH[72];\

  830.     uint8_t halfHV[64];\

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

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

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

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

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

  836. }\

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

  838.     uint8_t halfH[72];\

  839.     uint8_t halfHV[64];\

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

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

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

  843. }\

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

  845.     uint8_t halfH[72];\

  846.     uint8_t halfHV[64];\

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

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

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

  850. }\

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

  852.     uint8_t full[16*9];\

  853.     uint8_t halfH[72];\

  854.     uint8_t halfV[64];\

  855.     uint8_t halfHV[64];\

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

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

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

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

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

  861. }\

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

  863.     uint8_t full[16*9];\

  864.     uint8_t halfH[72];\

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

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

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

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

  869. }\

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

  871.     uint8_t full[16*9];\

  872.     uint8_t halfH[72];\

  873.     uint8_t halfV[64];\

  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 ## mpeg4_qpel8_v_lowpass(halfV, full+1, 8, 16);\

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

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

  880. }\

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

  882.     uint8_t full[16*9];\

  883.     uint8_t halfH[72];\

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

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

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

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

  888. }\

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

  890.     uint8_t halfH[72];\

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

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

  893. }\

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

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

  896. }\

  897. \

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

  899.     uint8_t half[256];\

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

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

  902. }\

  903. \

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

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

  906. }\

  907. \

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

  909.     uint8_t half[256];\

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

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

  912. }\

  913. \

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

  915.     uint8_t full[24*17];\

  916.     uint8_t half[256];\

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

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

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

  920. }\

  921. \

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

  923.     uint8_t full[24*17];\

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

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

  926. }\

  927. \

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

  929.     uint8_t full[24*17];\

  930.     uint8_t half[256];\

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

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

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

  934. }\

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

  936.     uint8_t full[24*17];\

  937.     uint8_t halfH[272];\

  938.     uint8_t halfV[256];\

  939.     uint8_t halfHV[256];\

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

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

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

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

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

  945. }\

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

  947.     uint8_t full[24*17];\

  948.     uint8_t halfH[272];\

  949.     uint8_t halfHV[256];\

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

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

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

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

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

  955. }\

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

  957.     uint8_t full[24*17];\

  958.     uint8_t halfH[272];\

  959.     uint8_t halfV[256];\

  960.     uint8_t halfHV[256];\

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

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

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

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

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

  966. }\

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

  968.     uint8_t full[24*17];\

  969.     uint8_t halfH[272];\

  970.     uint8_t halfHV[256];\

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

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

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

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

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

  976. }\

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

  978.     uint8_t full[24*17];\

  979.     uint8_t halfH[272];\

  980.     uint8_t halfV[256];\

  981.     uint8_t halfHV[256];\

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

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

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

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

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

  987. }\

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

  989.     uint8_t full[24*17];\

  990.     uint8_t halfH[272];\

  991.     uint8_t halfHV[256];\

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

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

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

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

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

  997. }\

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

  999.     uint8_t full[24*17];\

  1000.     uint8_t halfH[272];\

  1001.     uint8_t halfV[256];\

  1002.     uint8_t halfHV[256];\

  1003.     copy_block17(full, src, 24, stride, 17);\

  1004.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full  , 16, 24, 17);\

  1005.     put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full+1, 16, 24);\

  1006.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\

  1007.     OPNAME ## pixels16_l4_aligned0(dst, full+25, halfH+16, halfV, halfHV, stride, 24, 16, 16, 16, 16);\

  1008. }\

  1009. static void OPNAME ## qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride){\

  1010.     uint8_t full[24*17];\

  1011.     uint8_t halfH[272];\

  1012.     uint8_t halfHV[256];\

  1013.     copy_block17(full, src, 24, stride, 17);\

  1014.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\

  1015.     put ## RND ## pixels16_l2_aligned1(halfH, halfH, full+1, 16, 16, 24, 17);\

  1016.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\

  1017.     OPNAME ## pixels16_l2_aligned(dst, halfH+16, halfHV, stride, 16, 16, 16);\

  1018. }\

  1019. static void OPNAME ## qpel16_mc21_c(uint8_t *dst, uint8_t *src, int stride){\

  1020.     uint8_t halfH[272];\

  1021.     uint8_t halfHV[256];\

  1022.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\

  1023.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\

  1024.     OPNAME ## pixels16_l2_aligned(dst, halfH, halfHV, stride, 16, 16, 16);\

  1025. }\

  1026. static void OPNAME ## qpel16_mc23_c(uint8_t *dst, uint8_t *src, int stride){\

  1027.     uint8_t halfH[272];\

  1028.     uint8_t halfHV[256];\

  1029.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\

  1030.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\

  1031.     OPNAME ## pixels16_l2_aligned(dst, halfH+16, halfHV, stride, 16, 16, 16);\

  1032. }\

  1033. static void ff_ ## OPNAME ## qpel16_mc12_old_c(uint8_t *dst, uint8_t *src, int stride){\

  1034.     uint8_t full[24*17];\

  1035.     uint8_t halfH[272];\

  1036.     uint8_t halfV[256];\

  1037.     uint8_t halfHV[256];\

  1038.     copy_block17(full, src, 24, stride, 17);\

  1039.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\

  1040.     put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full, 16, 24);\

  1041.     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\

  1042.     OPNAME ## pixels16_l2_aligned(dst, halfV, halfHV, stride, 16, 16, 16);\

  1043. }\

  1044. static void OPNAME ## qpel16_mc12_c(uint8_t *dst, uint8_t *src, int stride){\

  1045.     uint8_t full[24*17];\

  1046.     uint8_t halfH[272];\

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

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

  1049.     put ## RND ## pixels16_l2_aligned(halfH, halfH, full, 16, 16, 24, 17);\

  1050.     OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\

  1051. }\

  1052. static void ff_ ## OPNAME ## qpel16_mc32_old_c(uint8_t *dst, uint8_t *src, int stride){\

  1053.     uint8_t full[24*17];\

  1054.     uint8_t halfH[272];\

  1055.     uint8_t halfV[256];\

  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 ## mpeg4_qpel16_v_lowpass(halfV, full+1, 16, 24);\

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

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

  1062. }\

  1063. static void OPNAME ## qpel16_mc32_c(uint8_t *dst, uint8_t *src, int stride){\

  1064.     uint8_t full[24*17];\

  1065.     uint8_t halfH[272];\

  1066.     copy_block17(full, src, 24, stride, 17);\

  1067.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\

  1068.     put ## RND ## pixels16_l2_aligned1(halfH, halfH, full+1, 16, 16, 24, 17);\

  1069.     OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\

  1070. }\

  1071. static void OPNAME ## qpel16_mc22_c(uint8_t *dst, uint8_t *src, int stride){\

  1072.     uint8_t halfH[272];\

  1073.     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\

  1074.     OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\

  1075. }

  1076. 

  1077. #define op_avg(a, b) a = (((a)+cm[((b) + 16)>>5]+1)>>1)

  1078. #define op_avg_no_rnd(a, b) a = (((a)+cm[((b) + 15)>>5])>>1)

  1079. #define op_put(a, b) a = cm[((b) + 16)>>5]

  1080. #define op_put_no_rnd(a, b) a = cm[((b) + 15)>>5]

  1081. 

  1082. QPEL_MC(0, put_       , _       , op_put)

  1083. QPEL_MC(1, put_no_rnd_, _no_rnd_, op_put_no_rnd)

  1084. QPEL_MC(0, avg_       , _       , op_avg)

  1085. //QPEL_MC(1, avg_no_rnd , _       , op_avg)

  1086. #undef op_avg

  1087. #undef op_avg_no_rnd

  1088. #undef op_put

  1089. #undef op_put_no_rnd

  1090. 

  1091. #if 1

  1092. #define H264_LOWPASS(OPNAME, OP, OP2) \

  1093. static inline void OPNAME ## h264_qpel_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,int w,int h){\

  1094.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  1095.     do {\

  1096.         int srcB,srcA,src0,src1,src2,src3,src4,src5,src6;\

  1097.         uint8_t *s = src-2;\

  1098.         srcB = *s++;\

  1099.         srcA = *s++;\

  1100.         src0 = *s++;\

  1101.         src1 = *s++;\

  1102.         src2 = *s++;\

  1103.         src3 = *s++;\

  1104.         OP(dst[0], (src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));\

  1105.         src4 = *s++;\

  1106.         OP(dst[1], (src1+src2)*20 - (src0+src3)*5 + (srcA+src4));\

  1107.         src5 = *s++;\

  1108.         OP(dst[2], (src2+src3)*20 - (src1+src4)*5 + (src0+src5));\

  1109.         src6 = *s++;\

  1110.         OP(dst[3], (src3+src4)*20 - (src2+src5)*5 + (src1+src6));\

  1111.       if (w>4) { /* it optimized */ \

  1112.         int src7,src8,src9,src10; \

  1113.         src7 = *s++;\

  1114.         OP(dst[4], (src4+src5)*20 - (src3+src6)*5 + (src2+src7));\

  1115.         src8 = *s++;\

  1116.         OP(dst[5], (src5+src6)*20 - (src4+src7)*5 + (src3+src8));\

  1117.         src9 = *s++;\

  1118.         OP(dst[6], (src6+src7)*20 - (src5+src8)*5 + (src4+src9));\

  1119.         src10 = *s++;\

  1120.         OP(dst[7], (src7+src8)*20 - (src6+src9)*5 + (src5+src10));\

  1121.        if (w>8) { \

  1122.         int src11,src12,src13,src14,src15,src16,src17,src18; \

  1123.         src11 = *s++;\

  1124.         OP(dst[8] , (src8 +src9 )*20 - (src7 +src10)*5 + (src6 +src11));\

  1125.         src12 = *s++;\

  1126.         OP(dst[9] , (src9 +src10)*20 - (src8 +src11)*5 + (src7 +src12));\

  1127.         src13 = *s++;\

  1128.         OP(dst[10], (src10+src11)*20 - (src9 +src12)*5 + (src8 +src13));\

  1129.         src14 = *s++;\

  1130.         OP(dst[11], (src11+src12)*20 - (src10+src13)*5 + (src9 +src14));\

  1131.         src15 = *s++;\

  1132.         OP(dst[12], (src12+src13)*20 - (src11+src14)*5 + (src10+src15));\

  1133.         src16 = *s++;\

  1134.         OP(dst[13], (src13+src14)*20 - (src12+src15)*5 + (src11+src16));\

  1135.         src17 = *s++;\

  1136.         OP(dst[14], (src14+src15)*20 - (src13+src16)*5 + (src12+src17));\

  1137.         src18 = *s++;\

  1138.         OP(dst[15], (src15+src16)*20 - (src14+src17)*5 + (src13+src18));\

  1139.        } \

  1140.       } \

  1141.         dst+=dstStride;\

  1142.         src+=srcStride;\

  1143.     }while(--h);\

  1144. }\

  1145. \

  1146. static inline void OPNAME ## h264_qpel_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,int w,int h){\

  1147.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  1148.     do{\

  1149.         int srcB,srcA,src0,src1,src2,src3,src4,src5,src6;\

  1150.         uint8_t *s = src-2*srcStride,*d=dst;\

  1151.         srcB = *s; s+=srcStride;\

  1152.         srcA = *s; s+=srcStride;\

  1153.         src0 = *s; s+=srcStride;\

  1154.         src1 = *s; s+=srcStride;\

  1155.         src2 = *s; s+=srcStride;\

  1156.         src3 = *s; s+=srcStride;\

  1157.         OP(*d, (src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));d+=dstStride;\

  1158.         src4 = *s; s+=srcStride;\

  1159.         OP(*d, (src1+src2)*20 - (src0+src3)*5 + (srcA+src4));d+=dstStride;\

  1160.         src5 = *s; s+=srcStride;\

  1161.         OP(*d, (src2+src3)*20 - (src1+src4)*5 + (src0+src5));d+=dstStride;\

  1162.         src6 = *s; s+=srcStride;\

  1163.         OP(*d, (src3+src4)*20 - (src2+src5)*5 + (src1+src6));d+=dstStride;\

  1164.       if (h>4) { \

  1165.         int src7,src8,src9,src10; \

  1166.         src7 = *s; s+=srcStride;\

  1167.         OP(*d, (src4+src5)*20 - (src3+src6)*5 + (src2+src7));d+=dstStride;\

  1168.         src8 = *s; s+=srcStride;\

  1169.         OP(*d, (src5+src6)*20 - (src4+src7)*5 + (src3+src8));d+=dstStride;\

  1170.         src9 = *s; s+=srcStride;\

  1171.         OP(*d, (src6+src7)*20 - (src5+src8)*5 + (src4+src9));d+=dstStride;\

  1172.         src10 = *s; s+=srcStride;\

  1173.         OP(*d, (src7+src8)*20 - (src6+src9)*5 + (src5+src10));d+=dstStride;\

  1174.        if (h>8) { \

  1175.         int src11,src12,src13,src14,src15,src16,src17,src18; \

  1176.         src11 = *s; s+=srcStride;\

  1177.         OP(*d , (src8 +src9 )*20 - (src7 +src10)*5 + (src6 +src11));d+=dstStride;\

  1178.         src12 = *s; s+=srcStride;\

  1179.         OP(*d , (src9 +src10)*20 - (src8 +src11)*5 + (src7 +src12));d+=dstStride;\

  1180.         src13 = *s; s+=srcStride;\

  1181.         OP(*d, (src10+src11)*20 - (src9 +src12)*5 + (src8 +src13));d+=dstStride;\

  1182.         src14 = *s; s+=srcStride;\

  1183.         OP(*d, (src11+src12)*20 - (src10+src13)*5 + (src9 +src14));d+=dstStride;\

  1184.         src15 = *s; s+=srcStride;\

  1185.         OP(*d, (src12+src13)*20 - (src11+src14)*5 + (src10+src15));d+=dstStride;\

  1186.         src16 = *s; s+=srcStride;\

  1187.         OP(*d, (src13+src14)*20 - (src12+src15)*5 + (src11+src16));d+=dstStride;\

  1188.         src17 = *s; s+=srcStride;\

  1189.         OP(*d, (src14+src15)*20 - (src13+src16)*5 + (src12+src17));d+=dstStride;\

  1190.         src18 = *s; s+=srcStride;\

  1191.         OP(*d, (src15+src16)*20 - (src14+src17)*5 + (src13+src18));d+=dstStride;\

  1192.        } \

  1193.       } \

  1194.         dst++;\

  1195.         src++;\

  1196.     }while(--w);\

  1197. }\

  1198. \

  1199. 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){\

  1200.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  1201.     int i;\

  1202.     src -= 2*srcStride;\

  1203.     i= h+5; \

  1204.     do {\

  1205.         int srcB,srcA,src0,src1,src2,src3,src4,src5,src6;\

  1206.         uint8_t *s = src-2;\

  1207.         srcB = *s++;\

  1208.         srcA = *s++;\

  1209.         src0 = *s++;\

  1210.         src1 = *s++;\

  1211.         src2 = *s++;\

  1212.         src3 = *s++;\

  1213.         tmp[0] = ((src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));\

  1214.         src4 = *s++;\

  1215.         tmp[1] = ((src1+src2)*20 - (src0+src3)*5 + (srcA+src4));\

  1216.         src5 = *s++;\

  1217.         tmp[2] = ((src2+src3)*20 - (src1+src4)*5 + (src0+src5));\

  1218.         src6 = *s++;\

  1219.         tmp[3] = ((src3+src4)*20 - (src2+src5)*5 + (src1+src6));\

  1220.       if (w>4) { /* it optimized */ \

  1221.         int src7,src8,src9,src10; \

  1222.         src7 = *s++;\

  1223.         tmp[4] = ((src4+src5)*20 - (src3+src6)*5 + (src2+src7));\

  1224.         src8 = *s++;\

  1225.         tmp[5] = ((src5+src6)*20 - (src4+src7)*5 + (src3+src8));\

  1226.         src9 = *s++;\

  1227.         tmp[6] = ((src6+src7)*20 - (src5+src8)*5 + (src4+src9));\

  1228.         src10 = *s++;\

  1229.         tmp[7] = ((src7+src8)*20 - (src6+src9)*5 + (src5+src10));\

  1230.        if (w>8) { \

  1231.         int src11,src12,src13,src14,src15,src16,src17,src18; \

  1232.         src11 = *s++;\

  1233.         tmp[8] = ((src8 +src9 )*20 - (src7 +src10)*5 + (src6 +src11));\

  1234.         src12 = *s++;\

  1235.         tmp[9] = ((src9 +src10)*20 - (src8 +src11)*5 + (src7 +src12));\

  1236.         src13 = *s++;\

  1237.         tmp[10] = ((src10+src11)*20 - (src9 +src12)*5 + (src8 +src13));\

  1238.         src14 = *s++;\

  1239.         tmp[11] = ((src11+src12)*20 - (src10+src13)*5 + (src9 +src14));\

  1240.         src15 = *s++;\

  1241.         tmp[12] = ((src12+src13)*20 - (src11+src14)*5 + (src10+src15));\

  1242.         src16 = *s++;\

  1243.         tmp[13] = ((src13+src14)*20 - (src12+src15)*5 + (src11+src16));\

  1244.         src17 = *s++;\

  1245.         tmp[14] = ((src14+src15)*20 - (src13+src16)*5 + (src12+src17));\

  1246.         src18 = *s++;\

  1247.         tmp[15] = ((src15+src16)*20 - (src14+src17)*5 + (src13+src18));\

  1248.        } \

  1249.       } \

  1250.         tmp+=tmpStride;\

  1251.         src+=srcStride;\

  1252.     }while(--i);\

  1253.     tmp -= tmpStride*(h+5-2);\

  1254.     i = w; \

  1255.     do {\

  1256.         int tmpB,tmpA,tmp0,tmp1,tmp2,tmp3,tmp4,tmp5,tmp6;\

  1257.         int16_t *s = tmp-2*tmpStride; \

  1258.         uint8_t *d=dst;\

  1259.         tmpB = *s; s+=tmpStride;\

  1260.         tmpA = *s; s+=tmpStride;\

  1261.         tmp0 = *s; s+=tmpStride;\

  1262.         tmp1 = *s; s+=tmpStride;\

  1263.         tmp2 = *s; s+=tmpStride;\

  1264.         tmp3 = *s; s+=tmpStride;\

  1265.         OP2(*d, (tmp0+tmp1)*20 - (tmpA+tmp2)*5 + (tmpB+tmp3));d+=dstStride;\

  1266.         tmp4 = *s; s+=tmpStride;\

  1267.         OP2(*d, (tmp1+tmp2)*20 - (tmp0+tmp3)*5 + (tmpA+tmp4));d+=dstStride;\

  1268.         tmp5 = *s; s+=tmpStride;\

  1269.         OP2(*d, (tmp2+tmp3)*20 - (tmp1+tmp4)*5 + (tmp0+tmp5));d+=dstStride;\

  1270.         tmp6 = *s; s+=tmpStride;\

  1271.         OP2(*d, (tmp3+tmp4)*20 - (tmp2+tmp5)*5 + (tmp1+tmp6));d+=dstStride;\

  1272.       if (h>4) { \

  1273.         int tmp7,tmp8,tmp9,tmp10; \

  1274.         tmp7 = *s; s+=tmpStride;\

  1275.         OP2(*d, (tmp4+tmp5)*20 - (tmp3+tmp6)*5 + (tmp2+tmp7));d+=dstStride;\

  1276.         tmp8 = *s; s+=tmpStride;\

  1277.         OP2(*d, (tmp5+tmp6)*20 - (tmp4+tmp7)*5 + (tmp3+tmp8));d+=dstStride;\

  1278.         tmp9 = *s; s+=tmpStride;\

  1279.         OP2(*d, (tmp6+tmp7)*20 - (tmp5+tmp8)*5 + (tmp4+tmp9));d+=dstStride;\

  1280.         tmp10 = *s; s+=tmpStride;\

  1281.         OP2(*d, (tmp7+tmp8)*20 - (tmp6+tmp9)*5 + (tmp5+tmp10));d+=dstStride;\

  1282.        if (h>8) { \

  1283.         int tmp11,tmp12,tmp13,tmp14,tmp15,tmp16,tmp17,tmp18; \

  1284.         tmp11 = *s; s+=tmpStride;\

  1285.         OP2(*d , (tmp8 +tmp9 )*20 - (tmp7 +tmp10)*5 + (tmp6 +tmp11));d+=dstStride;\

  1286.         tmp12 = *s; s+=tmpStride;\

  1287.         OP2(*d , (tmp9 +tmp10)*20 - (tmp8 +tmp11)*5 + (tmp7 +tmp12));d+=dstStride;\

  1288.         tmp13 = *s; s+=tmpStride;\

  1289.         OP2(*d, (tmp10+tmp11)*20 - (tmp9 +tmp12)*5 + (tmp8 +tmp13));d+=dstStride;\

  1290.         tmp14 = *s; s+=tmpStride;\

  1291.         OP2(*d, (tmp11+tmp12)*20 - (tmp10+tmp13)*5 + (tmp9 +tmp14));d+=dstStride;\

  1292.         tmp15 = *s; s+=tmpStride;\

  1293.         OP2(*d, (tmp12+tmp13)*20 - (tmp11+tmp14)*5 + (tmp10+tmp15));d+=dstStride;\

  1294.         tmp16 = *s; s+=tmpStride;\

  1295.         OP2(*d, (tmp13+tmp14)*20 - (tmp12+tmp15)*5 + (tmp11+tmp16));d+=dstStride;\

  1296.         tmp17 = *s; s+=tmpStride;\

  1297.         OP2(*d, (tmp14+tmp15)*20 - (tmp13+tmp16)*5 + (tmp12+tmp17));d+=dstStride;\

  1298.         tmp18 = *s; s+=tmpStride;\

  1299.         OP2(*d, (tmp15+tmp16)*20 - (tmp14+tmp17)*5 + (tmp13+tmp18));d+=dstStride;\

  1300.        } \

  1301.       } \

  1302.         dst++;\

  1303.         tmp++;\

  1304.     }while(--i);\

  1305. }\

  1306. \

  1307. static void OPNAME ## h264_qpel4_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1308.     OPNAME ## h264_qpel_h_lowpass(dst,src,dstStride,srcStride,4,4); \

  1309. }\

  1310. static void OPNAME ## h264_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1311.    OPNAME ## h264_qpel_h_lowpass(dst,src,dstStride,srcStride,8,8); \

  1312. }\

  1313. static void OPNAME ## h264_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1314.    OPNAME ## h264_qpel_h_lowpass(dst,src,dstStride,srcStride,16,16); \

  1315. }\

  1316. \

  1317. static void OPNAME ## h264_qpel4_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1318.    OPNAME ## h264_qpel_v_lowpass(dst,src,dstStride,srcStride,4,4); \

  1319. }\

  1320. static void OPNAME ## h264_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1321.    OPNAME ## h264_qpel_v_lowpass(dst,src,dstStride,srcStride,8,8); \

  1322. }\

  1323. static void OPNAME ## h264_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1324.    OPNAME ## h264_qpel_v_lowpass(dst,src,dstStride,srcStride,16,16); \

  1325. }\

  1326. static void OPNAME ## h264_qpel4_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\

  1327.    OPNAME ## h264_qpel_hv_lowpass(dst,tmp,src,dstStride,tmpStride,srcStride,4,4); \

  1328. }\

  1329. static void OPNAME ## h264_qpel8_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\

  1330.    OPNAME ## h264_qpel_hv_lowpass(dst,tmp,src,dstStride,tmpStride,srcStride,8,8); \

  1331. }\

  1332. static void OPNAME ## h264_qpel16_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\

  1333.    OPNAME ## h264_qpel_hv_lowpass(dst,tmp,src,dstStride,tmpStride,srcStride,16,16); \

  1334. }\

  1335. 

  1336. #define H264_MC(OPNAME, SIZE) \

  1337. static void OPNAME ## h264_qpel ## SIZE ## _mc00_c (uint8_t *dst, uint8_t *src, int stride){\

  1338.     OPNAME ## pixels ## SIZE ## _c(dst, src, stride, SIZE);\

  1339. }\

  1340. \

  1341. static void OPNAME ## h264_qpel ## SIZE ## _mc10_c(uint8_t *dst, uint8_t *src, int stride){\

  1342.     uint8_t half[SIZE*SIZE];\

  1343.     put_h264_qpel ## SIZE ## _h_lowpass(half, src, SIZE, stride);\

  1344.     OPNAME ## pixels ## SIZE ## _l2_aligned2(dst, src, half, stride, stride, SIZE, SIZE);\

  1345. }\

  1346. \

  1347. static void OPNAME ## h264_qpel ## SIZE ## _mc20_c(uint8_t *dst, uint8_t *src, int stride){\

  1348.     OPNAME ## h264_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride);\

  1349. }\

  1350. \

  1351. static void OPNAME ## h264_qpel ## SIZE ## _mc30_c(uint8_t *dst, uint8_t *src, int stride){\

  1352.     uint8_t half[SIZE*SIZE];\

  1353.     put_h264_qpel ## SIZE ## _h_lowpass(half, src, SIZE, stride);\

  1354.     OPNAME ## pixels ## SIZE ## _l2_aligned2(dst, src+1, half, stride, stride, SIZE, SIZE);\

  1355. }\

  1356. \

  1357. static void OPNAME ## h264_qpel ## SIZE ## _mc01_c(uint8_t *dst, uint8_t *src, int stride){\

  1358.     uint8_t full[SIZE*(SIZE+5)];\

  1359.     uint8_t * const full_mid= full + SIZE*2;\

  1360.     uint8_t half[SIZE*SIZE];\

  1361.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1362.     put_h264_qpel ## SIZE ## _v_lowpass(half, full_mid, SIZE, SIZE);\

  1363.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, full_mid, half, stride, SIZE, SIZE, SIZE);\

  1364. }\

  1365. \

  1366. static void OPNAME ## h264_qpel ## SIZE ## _mc02_c(uint8_t *dst, uint8_t *src, int stride){\

  1367.     uint8_t full[SIZE*(SIZE+5)];\

  1368.     uint8_t * const full_mid= full + SIZE*2;\

  1369.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1370.     OPNAME ## h264_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE);\

  1371. }\

  1372. \

  1373. static void OPNAME ## h264_qpel ## SIZE ## _mc03_c(uint8_t *dst, uint8_t *src, int stride){\

  1374.     uint8_t full[SIZE*(SIZE+5)];\

  1375.     uint8_t * const full_mid= full + SIZE*2;\

  1376.     uint8_t half[SIZE*SIZE];\

  1377.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1378.     put_h264_qpel ## SIZE ## _v_lowpass(half, full_mid, SIZE, SIZE);\

  1379.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, full_mid+SIZE, half, stride, SIZE, SIZE, SIZE);\

  1380. }\

  1381. \

  1382. static void OPNAME ## h264_qpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, int stride){\

  1383.     uint8_t full[SIZE*(SIZE+5)];\

  1384.     uint8_t * const full_mid= full + SIZE*2;\

  1385.     uint8_t halfH[SIZE*SIZE];\

  1386.     uint8_t halfV[SIZE*SIZE];\

  1387.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\

  1388.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1389.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1390.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\

  1391. }\

  1392. \

  1393. static void OPNAME ## h264_qpel ## SIZE ## _mc31_c(uint8_t *dst, uint8_t *src, int stride){\

  1394.     uint8_t full[SIZE*(SIZE+5)];\

  1395.     uint8_t * const full_mid= full + SIZE*2;\

  1396.     uint8_t halfH[SIZE*SIZE];\

  1397.     uint8_t halfV[SIZE*SIZE];\

  1398.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\

  1399.     copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\

  1400.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1401.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\

  1402. }\

  1403. \

  1404. static void OPNAME ## h264_qpel ## SIZE ## _mc13_c(uint8_t *dst, uint8_t *src, int stride){\

  1405.     uint8_t full[SIZE*(SIZE+5)];\

  1406.     uint8_t * const full_mid= full + SIZE*2;\

  1407.     uint8_t halfH[SIZE*SIZE];\

  1408.     uint8_t halfV[SIZE*SIZE];\

  1409.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\

  1410.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1411.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1412.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\

  1413. }\

  1414. \

  1415. static void OPNAME ## h264_qpel ## SIZE ## _mc33_c(uint8_t *dst, uint8_t *src, int stride){\

  1416.     uint8_t full[SIZE*(SIZE+5)];\

  1417.     uint8_t * const full_mid= full + SIZE*2;\

  1418.     uint8_t halfH[SIZE*SIZE];\

  1419.     uint8_t halfV[SIZE*SIZE];\

  1420.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\

  1421.     copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\

  1422.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1423.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\

  1424. }\

  1425. \

  1426. static void OPNAME ## h264_qpel ## SIZE ## _mc22_c(uint8_t *dst, uint8_t *src, int stride){\

  1427.     int16_t tmp[SIZE*(SIZE+5)];\

  1428.     OPNAME ## h264_qpel ## SIZE ## _hv_lowpass(dst, tmp, src, stride, SIZE, stride);\

  1429. }\

  1430. \

  1431. static void OPNAME ## h264_qpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, int stride){\

  1432.     int16_t tmp[SIZE*(SIZE+5)];\

  1433.     uint8_t halfH[SIZE*SIZE];\

  1434.     uint8_t halfHV[SIZE*SIZE];\

  1435.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\

  1436.     put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\

  1437.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfHV, stride, SIZE, SIZE, SIZE);\

  1438. }\

  1439. \

  1440. static void OPNAME ## h264_qpel ## SIZE ## _mc23_c(uint8_t *dst, uint8_t *src, int stride){\

  1441.     int16_t tmp[SIZE*(SIZE+5)];\

  1442.     uint8_t halfH[SIZE*SIZE];\

  1443.     uint8_t halfHV[SIZE*SIZE];\

  1444.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\

  1445.     put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\

  1446.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfHV, stride, SIZE, SIZE, SIZE);\

  1447. }\

  1448. \

  1449. static void OPNAME ## h264_qpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, int stride){\

  1450.     uint8_t full[SIZE*(SIZE+5)];\

  1451.     uint8_t * const full_mid= full + SIZE*2;\

  1452.     int16_t tmp[SIZE*(SIZE+5)];\

  1453.     uint8_t halfV[SIZE*SIZE];\

  1454.     uint8_t halfHV[SIZE*SIZE];\

  1455.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1456.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1457.     put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\

  1458.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfV, halfHV, stride, SIZE, SIZE, SIZE);\

  1459. }\

  1460. \

  1461. static void OPNAME ## h264_qpel ## SIZE ## _mc32_c(uint8_t *dst, uint8_t *src, int stride){\

  1462.     uint8_t full[SIZE*(SIZE+5)];\

  1463.     uint8_t * const full_mid= full + SIZE*2;\

  1464.     int16_t tmp[SIZE*(SIZE+5)];\

  1465.     uint8_t halfV[SIZE*SIZE];\

  1466.     uint8_t halfHV[SIZE*SIZE];\

  1467.     copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\

  1468.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1469.     put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\

  1470.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfV, halfHV, stride, SIZE, SIZE, SIZE);\

  1471. }\

  1472. 

  1473. #define op_avg(a, b)  a = (((a)+cm[((b) + 16)>>5]+1)>>1)

  1474. //#define op_avg2(a, b) a = (((a)*w1+cm[((b) + 16)>>5]*w2 + o + 64)>>7)

  1475. #define op_put(a, b)  a = cm[((b) + 16)>>5]

  1476. #define op2_avg(a, b)  a = (((a)+cm[((b) + 512)>>10]+1)>>1)

  1477. #define op2_put(a, b)  a = cm[((b) + 512)>>10]

  1478. 

  1479. H264_LOWPASS(put_       , op_put, op2_put)

  1480. H264_LOWPASS(avg_       , op_avg, op2_avg)

  1481. H264_MC(put_, 4)

  1482. H264_MC(put_, 8)

  1483. H264_MC(put_, 16)

  1484. H264_MC(avg_, 4)

  1485. H264_MC(avg_, 8)

  1486. H264_MC(avg_, 16)

  1487. 

  1488. #undef op_avg

  1489. #undef op_put

  1490. #undef op2_avg

  1491. #undef op2_put

  1492. #endif

  1493. 

  1494. static void wmv2_mspel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){

  1495.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;

  1496. 

  1497.     do{

  1498.         int src_1,src0,src1,src2,src3,src4,src5,src6,src7,src8,src9;

  1499.         uint8_t *s = src;

  1500.         src_1 = s[-1];

  1501.         src0 = *s++;

  1502.         src1 = *s++;

  1503.         src2 = *s++;

  1504.         dst[0]= cm[(9*(src0 + src1) - (src_1 + src2) + 8)>>4];

  1505.         src3 = *s++;

  1506.         dst[1]= cm[(9*(src1 + src2) - (src0 + src3) + 8)>>4];

  1507.         src4 = *s++;

  1508.         dst[2]= cm[(9*(src2 + src3) - (src1 + src4) + 8)>>4];

  1509.         src5 = *s++;

  1510.         dst[3]= cm[(9*(src3 + src4) - (src2 + src5) + 8)>>4];

  1511.         src6 = *s++;

  1512.         dst[4]= cm[(9*(src4 + src5) - (src3 + src6) + 8)>>4];

  1513.         src7 = *s++;

  1514.         dst[5]= cm[(9*(src5 + src6) - (src4 + src7) + 8)>>4];

  1515.         src8 = *s++;

  1516.         dst[6]= cm[(9*(src6 + src7) - (src5 + src8) + 8)>>4];

  1517.         src9 = *s++;

  1518.         dst[7]= cm[(9*(src7 + src8) - (src6 + src9) + 8)>>4];

  1519.         dst+=dstStride;

  1520.         src+=srcStride;

  1521.     }while(--h);

  1522. }

  1523. 

  1524. static void wmv2_mspel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int w){

  1525.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;

  1526. 

  1527.     do{

  1528.         int src_1,src0,src1,src2,src3,src4,src5,src6,src7,src8,src9;

  1529.         uint8_t *s = src,*d = dst;

  1530.         src_1 = *(s-srcStride);

  1531.         src0 = *s; s+=srcStride;

  1532.         src1 = *s; s+=srcStride;

  1533.         src2 = *s; s+=srcStride;

  1534.         *d= cm[(9*(src0 + src1) - (src_1 + src2) + 8)>>4]; d+=dstStride;

  1535.         src3 = *s; s+=srcStride;

  1536.         *d= cm[(9*(src1 + src2) - (src0  + src3) + 8)>>4]; d+=dstStride;

  1537.         src4 = *s; s+=srcStride;

  1538.         *d= cm[(9*(src2 + src3) - (src1  + src4) + 8)>>4]; d+=dstStride;

  1539.         src5 = *s; s+=srcStride;

  1540.         *d= cm[(9*(src3 + src4) - (src2  + src5) + 8)>>4]; d+=dstStride;

  1541.         src6 = *s; s+=srcStride;

  1542.         *d= cm[(9*(src4 + src5) - (src3  + src6) + 8)>>4]; d+=dstStride;

  1543.         src7 = *s; s+=srcStride;

  1544.         *d= cm[(9*(src5 + src6) - (src4  + src7) + 8)>>4]; d+=dstStride;

  1545.         src8 = *s; s+=srcStride;

  1546.         *d= cm[(9*(src6 + src7) - (src5  + src8) + 8)>>4]; d+=dstStride;

  1547.         src9 = *s;

  1548.         *d= cm[(9*(src7 + src8) - (src6  + src9) + 8)>>4]; d+=dstStride;

  1549.         src++;

  1550.         dst++;

  1551.     }while(--w);

  1552. }

  1553. 

  1554. static void put_mspel8_mc00_c (uint8_t *dst, uint8_t *src, int stride){

  1555.     put_pixels8_c(dst, src, stride, 8);

  1556. }

  1557. 

  1558. static void put_mspel8_mc10_c(uint8_t *dst, uint8_t *src, int stride){

  1559.     uint8_t half[64];

  1560.     wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);

  1561.     put_pixels8_l2_aligned2(dst, src, half, stride, stride, 8, 8);

  1562. }

  1563. 

  1564. static void put_mspel8_mc20_c(uint8_t *dst, uint8_t *src, int stride){

  1565.     wmv2_mspel8_h_lowpass(dst, src, stride, stride, 8);

  1566. }

  1567. 

  1568. static void put_mspel8_mc30_c(uint8_t *dst, uint8_t *src, int stride){

  1569.     uint8_t half[64];

  1570.     wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);

  1571.     put_pixels8_l2_aligned2(dst, src+1, half, stride, stride, 8, 8);

  1572. }

  1573. 

  1574. static void put_mspel8_mc02_c(uint8_t *dst, uint8_t *src, int stride){

  1575.     wmv2_mspel8_v_lowpass(dst, src, stride, stride, 8);

  1576. }

  1577. 

  1578. static void put_mspel8_mc12_c(uint8_t *dst, uint8_t *src, int stride){

  1579.     uint8_t halfH[88];

  1580.     uint8_t halfV[64];

  1581.     uint8_t halfHV[64];

  1582.     wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);

  1583.     wmv2_mspel8_v_lowpass(halfV, src, 8, stride, 8);

  1584.     wmv2_mspel8_v_lowpass(halfHV, halfH+8, 8, 8, 8);

  1585.     put_pixels8_l2_aligned(dst, halfV, halfHV, stride, 8, 8, 8);

  1586. }

  1587. static void put_mspel8_mc32_c(uint8_t *dst, uint8_t *src, int stride){

  1588.     uint8_t halfH[88];

  1589.     uint8_t halfV[64];

  1590.     uint8_t halfHV[64];

  1591.     wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);

  1592.     wmv2_mspel8_v_lowpass(halfV, src+1, 8, stride, 8);

  1593.     wmv2_mspel8_v_lowpass(halfHV, halfH+8, 8, 8, 8);

  1594.     put_pixels8_l2_aligned(dst, halfV, halfHV, stride, 8, 8, 8);

  1595. }

  1596. static void put_mspel8_mc22_c(uint8_t *dst, uint8_t *src, int stride){

  1597.     uint8_t halfH[88];

  1598.     wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);

  1599.     wmv2_mspel8_v_lowpass(dst, halfH+8, stride, 8, 8);

  1600. }