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.
36037 Commits
32 Branches
1.2GB
Tree: c3a76c3dec
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'c3a76c3dec'
${ noResults }
FFmpeg/libavcodec/sh4/qpel.c

1353 lines
59KB
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 Libav.

  8.  *

  9.  * Libav 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.  * Libav 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 Libav; if not, write to the Free Software

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

  22.  */

  23. 

  24. #include "libavutil/common.h"

  25. 

  26. #define PIXOP2(OPNAME, OP) \

  27. \

  28. 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) \

  29. {\

  30.         do {\

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

  32.                 src1+=src_stride1; \

  33.                 src2+=src_stride2; \

  34.                 dst+=dst_stride; \

  35.         } while(--h); \

  36. }\

  37. \

  38. 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) \

  39. {\

  40.         do {\

  41.                 OP(LP(dst  ),rnd_avg32(AV_RN32(src1  ),LPC(src2  )) ); \

  42.                 src1+=src_stride1; \

  43.                 src2+=src_stride2; \

  44.                 dst+=dst_stride; \

  45.         } while(--h); \

  46. }\

  47. \

  48. 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) \

  49. {\

  50.         do {\

  51.                 OP(LP(dst  ),no_rnd_avg32(AV_RN32(src1  ),LPC(src2  )) ); \

  52.                 OP(LP(dst+4),no_rnd_avg32(AV_RN32(src1+4),LPC(src2+4)) ); \

  53.                 OP(LP(dst+8),no_rnd_avg32(AV_RN32(src1+8),LPC(src2+8)) ); \

  54.                 OP(LP(dst+12),no_rnd_avg32(AV_RN32(src1+12),LPC(src2+12)) ); \

  55.                 src1+=src_stride1; \

  56.                 src2+=src_stride2; \

  57.                 dst+=dst_stride; \

  58.         } while(--h); \

  59. }\

  60. \

  61. 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) \

  62. {\

  63.         do {\

  64.                 OP(LP(dst  ),rnd_avg32(AV_RN32(src1  ),LPC(src2  )) ); \

  65.                 OP(LP(dst+4),rnd_avg32(AV_RN32(src1+4),LPC(src2+4)) ); \

  66.                 OP(LP(dst+8),rnd_avg32(AV_RN32(src1+8),LPC(src2+8)) ); \

  67.                 OP(LP(dst+12),rnd_avg32(AV_RN32(src1+12),LPC(src2+12)) ); \

  68.                 src1+=src_stride1; \

  69.                 src2+=src_stride2; \

  70.                 dst+=dst_stride; \

  71.         } while(--h); \

  72. }\

  73. \

  74. 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) \

  75. {\

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

  77.                 OP(LP(dst  ),no_rnd_avg32(AV_RN32(src1  ),LPC(src2  )) ); \

  78.                 OP(LP(dst+4),no_rnd_avg32(AV_RN32(src1+4),LPC(src2+4)) ); \

  79.                 src1+=src_stride1; \

  80.                 src2+=src_stride2; \

  81.                 dst+=dst_stride; \

  82.         } while(--h); \

  83. }\

  84. \

  85. 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) \

  86. {\

  87.         do {\

  88.                 OP(LP(dst  ),rnd_avg32(AV_RN32(src1  ),LPC(src2  )) ); \

  89.                 OP(LP(dst+4),rnd_avg32(AV_RN32(src1+4),LPC(src2+4)) ); \

  90.                 src1+=src_stride1; \

  91.                 src2+=src_stride2; \

  92.                 dst+=dst_stride; \

  93.         } while(--h); \

  94. }\

  95. \

  96. 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) \

  97. {\

  98.         do {\

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

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

  101.                 src1+=src_stride1; \

  102.                 src2+=src_stride2; \

  103.                 dst+=dst_stride; \

  104.         } while(--h); \

  105. }\

  106. \

  107. 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) \

  108. {\

  109.         do {\

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

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

  112.                 src1+=src_stride1; \

  113.                 src2+=src_stride2; \

  114.                 dst+=dst_stride; \

  115.         } while(--h); \

  116. }\

  117. \

  118. 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) \

  119. {\

  120.         do {\

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

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

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

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

  125.                 src1+=src_stride1; \

  126.                 src2+=src_stride2; \

  127.                 dst+=dst_stride; \

  128.         } while(--h); \

  129. }\

  130. \

  131. 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) \

  132. {\

  133.         do {\

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

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

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

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

  138.                 src1+=src_stride1; \

  139.                 src2+=src_stride2; \

  140.                 dst+=dst_stride; \

  141.         } while(--h); \

  142. }\

  143. \

  144. 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) \

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

  146. \

  147. 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) \

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

  149. \

  150. 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) \

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

  152. \

  153. 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) \

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

  155. \

  156. 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){\

  157.         do { \

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

  159.                 UNPACK(a0,a1,LPC(src1),LPC(src2)); \

  160.                 UNPACK(a2,a3,LPC(src3),LPC(src4)); \

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

  162.                 UNPACK(a0,a1,LPC(src1+4),LPC(src2+4)); \

  163.                 UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \

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

  165.                 src1+=src_stride1;\

  166.                 src2+=src_stride2;\

  167.                 src3+=src_stride3;\

  168.                 src4+=src_stride4;\

  169.                 dst+=dst_stride;\

  170.         } while(--h); \

  171. } \

  172. \

  173. 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){\

  174.         do { \

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

  176.                 UNPACK(a0,a1,LPC(src1),LPC(src2)); \

  177.                 UNPACK(a2,a3,LPC(src3),LPC(src4)); \

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

  179.                 UNPACK(a0,a1,LPC(src1+4),LPC(src2+4)); \

  180.                 UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \

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

  182.                 src1+=src_stride1;\

  183.                 src2+=src_stride2;\

  184.                 src3+=src_stride3;\

  185.                 src4+=src_stride4;\

  186.                 dst+=dst_stride;\

  187.         } while(--h); \

  188. } \

  189. \

  190. 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){\

  191.         do { \

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

  193.                 UNPACK(a0,a1,AV_RN32(src1),LPC(src2)); \

  194.                 UNPACK(a2,a3,LPC(src3),LPC(src4)); \

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

  196.                 UNPACK(a0,a1,AV_RN32(src1+4),LPC(src2+4)); \

  197.                 UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \

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

  199.                 src1+=src_stride1;\

  200.                 src2+=src_stride2;\

  201.                 src3+=src_stride3;\

  202.                 src4+=src_stride4;\

  203.                 dst+=dst_stride;\

  204.         } while(--h); \

  205. } \

  206. \

  207. 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){\

  208.         do { \

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

  210.                 UNPACK(a0,a1,AV_RN32(src1),LPC(src2)); \

  211.                 UNPACK(a2,a3,LPC(src3),LPC(src4)); \

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

  213.                 UNPACK(a0,a1,AV_RN32(src1+4),LPC(src2+4)); \

  214.                 UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \

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

  216.                 src1+=src_stride1;\

  217.                 src2+=src_stride2;\

  218.                 src3+=src_stride3;\

  219.                 src4+=src_stride4;\

  220.                 dst+=dst_stride;\

  221.         } while(--h); \

  222. } \

  223. \

  224. 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){\

  225.         do { \

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

  227.                 UNPACK(a0,a1,LPC(src1),LPC(src2)); \

  228.                 UNPACK(a2,a3,LPC(src3),LPC(src4)); \

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

  230.                 UNPACK(a0,a1,LPC(src1+4),LPC(src2+4)); \

  231.                 UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \

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

  233.                 UNPACK(a0,a1,LPC(src1+8),LPC(src2+8)); \

  234.                 UNPACK(a2,a3,LPC(src3+8),LPC(src4+8)); \

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

  236.                 UNPACK(a0,a1,LPC(src1+12),LPC(src2+12)); \

  237.                 UNPACK(a2,a3,LPC(src3+12),LPC(src4+12)); \

  238.                 OP(LP(dst+12),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 ## _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){\

  248.         do { \

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

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

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

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

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

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

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

  256.                 UNPACK(a0,a1,LPC(src1+8),LPC(src2+8)); \

  257.                 UNPACK(a2,a3,LPC(src3+8),LPC(src4+8)); \

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

  259.                 UNPACK(a0,a1,LPC(src1+12),LPC(src2+12)); \

  260.                 UNPACK(a2,a3,LPC(src3+12),LPC(src4+12)); \

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

  262.                 src1+=src_stride1;\

  263.                 src2+=src_stride2;\

  264.                 src3+=src_stride3;\

  265.                 src4+=src_stride4;\

  266.                 dst+=dst_stride;\

  267.         } while(--h); \

  268. } \

  269. \

  270. 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){\

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

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

  273.                 UNPACK(a0,a1,AV_RN32(src1),LPC(src2)); \

  274.                 UNPACK(a2,a3,LPC(src3),LPC(src4)); \

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

  276.                 UNPACK(a0,a1,AV_RN32(src1+4),LPC(src2+4)); \

  277.                 UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \

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

  279.                 UNPACK(a0,a1,AV_RN32(src1+8),LPC(src2+8)); \

  280.                 UNPACK(a2,a3,LPC(src3+8),LPC(src4+8)); \

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

  282.                 UNPACK(a0,a1,AV_RN32(src1+12),LPC(src2+12)); \

  283.                 UNPACK(a2,a3,LPC(src3+12),LPC(src4+12)); \

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

  285.                 src1+=src_stride1;\

  286.                 src2+=src_stride2;\

  287.                 src3+=src_stride3;\

  288.                 src4+=src_stride4;\

  289.                 dst+=dst_stride;\

  290.         } while(--h); \

  291. } \

  292. \

  293. 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){\

  294.         do { \

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

  296.                 UNPACK(a0,a1,AV_RN32(src1),LPC(src2)); \

  297.                 UNPACK(a2,a3,LPC(src3),LPC(src4)); \

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

  299.                 UNPACK(a0,a1,AV_RN32(src1+4),LPC(src2+4)); \

  300.                 UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \

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

  302.                 UNPACK(a0,a1,AV_RN32(src1+8),LPC(src2+8)); \

  303.                 UNPACK(a2,a3,LPC(src3+8),LPC(src4+8)); \

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

  305.                 UNPACK(a0,a1,AV_RN32(src1+12),LPC(src2+12)); \

  306.                 UNPACK(a2,a3,LPC(src3+12),LPC(src4+12)); \

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

  308.                 src1+=src_stride1;\

  309.                 src2+=src_stride2;\

  310.                 src3+=src_stride3;\

  311.                 src4+=src_stride4;\

  312.                 dst+=dst_stride;\

  313.         } while(--h); \

  314. } \

  315. \

  316. 

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

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

  319. 

  320. PIXOP2(avg, op_avg)

  321. PIXOP2(put, op_put)

  322. #undef op_avg

  323. #undef op_put

  324. 

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

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

  327. 

  328. 

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

  330. {

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

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

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

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

  335. 

  336.     do {

  337.         int t0,t1,t2,t3;

  338.         uint8_t *s0 = src;

  339.         uint8_t *s1 = src+stride;

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  357.         dst+= stride;

  358.         src+= stride;

  359.     }while(--h);

  360. }

  361. 

  362. #define H264_CHROMA_MC(OPNAME, OP)\

  363. static void OPNAME ## h264_chroma_mc2_sh4(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\

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

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

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

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

  368.     \

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

  370. \

  371.     do {\

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

  373.         uint8_t *s0 = src; \

  374.         uint8_t *s1 = src+stride; \

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

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

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

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

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

  380.         dst+= stride;\

  381.         src+= stride;\

  382.     }while(--h);\

  383. }\

  384. \

  385. static void OPNAME ## h264_chroma_mc4_sh4(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\

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

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

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

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

  390.     \

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

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

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

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

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

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

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

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

  406.         dst+= stride;\

  407.         src+= stride;\

  408.     }while(--h);\

  409. }\

  410. \

  411. static void OPNAME ## h264_chroma_mc8_sh4(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\

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

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

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

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

  416.     \

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

  418. \

  419.     do {\

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

  421.         uint8_t *s0 = src; \

  422.         uint8_t *s1 = src+stride; \

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  440.         dst+= stride;\

  441.         src+= stride;\

  442.     }while(--h);\

  443. }

  444. 

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

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

  447. 

  448. H264_CHROMA_MC(put_       , op_put)

  449. H264_CHROMA_MC(avg_       , op_avg)

  450. #undef op_avg

  451. #undef op_put

  452. 

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

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

  455.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  456.     do {\

  457.         uint8_t *s = src; \

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

  459.         src0= *s++;\

  460.         src1= *s++;\

  461.         src2= *s++;\

  462.         src3= *s++;\

  463.         src4= *s++;\

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

  465.         src5= *s++;\

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

  467.         src6= *s++;\

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

  469.         src7= *s++;\

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

  471.         src8= *s++;\

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

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

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

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

  476.         dst+=dstStride;\

  477.         src+=srcStride;\

  478.     }while(--h);\

  479. }\

  480. \

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

  482.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  483.     int w=8;\

  484.     do{\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  499.         src8 = *s; \

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

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

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

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

  504.         dst++;\

  505.         src++;\

  506.     }while(--w);\

  507. }\

  508. \

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

  510.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  511.     do {\

  512.         uint8_t *s = src;\

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

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

  515.         src0= *s++;\

  516.         src1= *s++;\

  517.         src2= *s++;\

  518.         src3= *s++;\

  519.         src4= *s++;\

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

  521.         src5= *s++;\

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

  523.         src6= *s++;\

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

  525.         src7= *s++;\

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

  527.         src8= *s++;\

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

  529.         src9= *s++;\

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

  531.         src10= *s++;\

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

  533.         src11= *s++;\

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

  535.         src12= *s++;\

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

  537.         src13= *s++;\

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

  539.         src14= *s++;\

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

  541.         src15= *s++;\

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

  543.         src16= *s++;\

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

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

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

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

  548.         dst+=dstStride;\

  549.         src+=srcStride;\

  550.     }while(--h);\

  551. }\

  552. \

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

  554.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  555.     int w=16;\

  556.     do {\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  588.         src16 = *s; \

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

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

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

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

  593.         dst++;\

  594.         src++;\

  595.     }while(--w);\

  596. }\

  597. \

  598. static void OPNAME ## qpel8_mc00_sh4 (uint8_t *dst, uint8_t *src, int stride){\

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

  600. }\

  601. \

  602. static void OPNAME ## qpel8_mc10_sh4(uint8_t *dst, uint8_t *src, int stride){\

  603.     uint8_t half[64];\

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

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

  606. }\

  607. \

  608. static void OPNAME ## qpel8_mc20_sh4(uint8_t *dst, uint8_t *src, int stride){\

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

  610. }\

  611. \

  612. static void OPNAME ## qpel8_mc30_sh4(uint8_t *dst, uint8_t *src, int stride){\

  613.     uint8_t half[64];\

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

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

  616. }\

  617. \

  618. static void OPNAME ## qpel8_mc01_sh4(uint8_t *dst, uint8_t *src, int stride){\

  619.     uint8_t full[16*9];\

  620.     uint8_t half[64];\

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

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

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

  624. }\

  625. \

  626. static void OPNAME ## qpel8_mc02_sh4(uint8_t *dst, uint8_t *src, int stride){\

  627.     uint8_t full[16*9];\

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

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

  630. }\

  631. \

  632. static void OPNAME ## qpel8_mc03_sh4(uint8_t *dst, uint8_t *src, int stride){\

  633.     uint8_t full[16*9];\

  634.     uint8_t half[64];\

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

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

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

  638. }\

  639. static void OPNAME ## qpel8_mc11_sh4(uint8_t *dst, uint8_t *src, int stride){\

  640.     uint8_t full[16*9];\

  641.     uint8_t halfH[72];\

  642.     uint8_t halfHV[64];\

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

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

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

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

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

  648. }\

  649. static void OPNAME ## qpel8_mc31_sh4(uint8_t *dst, uint8_t *src, int stride){\

  650.     uint8_t full[16*9];\

  651.     uint8_t halfH[72];\

  652.     uint8_t halfHV[64];\

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

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

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

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

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

  658. }\

  659. static void OPNAME ## qpel8_mc13_sh4(uint8_t *dst, uint8_t *src, int stride){\

  660.     uint8_t full[16*9];\

  661.     uint8_t halfH[72];\

  662.     uint8_t halfHV[64];\

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

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

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

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

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

  668. }\

  669. static void OPNAME ## qpel8_mc33_sh4(uint8_t *dst, uint8_t *src, int stride){\

  670.     uint8_t full[16*9];\

  671.     uint8_t halfH[72];\

  672.     uint8_t halfHV[64];\

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

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

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

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

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

  678. }\

  679. static void OPNAME ## qpel8_mc21_sh4(uint8_t *dst, uint8_t *src, int stride){\

  680.     uint8_t halfH[72];\

  681.     uint8_t halfHV[64];\

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

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

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

  685. }\

  686. static void OPNAME ## qpel8_mc23_sh4(uint8_t *dst, uint8_t *src, int stride){\

  687.     uint8_t halfH[72];\

  688.     uint8_t halfHV[64];\

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

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

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

  692. }\

  693. static void OPNAME ## qpel8_mc12_sh4(uint8_t *dst, uint8_t *src, int stride){\

  694.     uint8_t full[16*9];\

  695.     uint8_t halfH[72];\

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

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

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

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

  700. }\

  701. static void OPNAME ## qpel8_mc32_sh4(uint8_t *dst, uint8_t *src, int stride){\

  702.     uint8_t full[16*9];\

  703.     uint8_t halfH[72];\

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

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

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

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

  708. }\

  709. static void OPNAME ## qpel8_mc22_sh4(uint8_t *dst, uint8_t *src, int stride){\

  710.     uint8_t halfH[72];\

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

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

  713. }\

  714. static void OPNAME ## qpel16_mc00_sh4 (uint8_t *dst, uint8_t *src, int stride){\

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

  716. }\

  717. \

  718. static void OPNAME ## qpel16_mc10_sh4(uint8_t *dst, uint8_t *src, int stride){\

  719.     uint8_t half[256];\

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

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

  722. }\

  723. \

  724. static void OPNAME ## qpel16_mc20_sh4(uint8_t *dst, uint8_t *src, int stride){\

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

  726. }\

  727. \

  728. static void OPNAME ## qpel16_mc30_sh4(uint8_t *dst, uint8_t *src, int stride){\

  729.     uint8_t half[256];\

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

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

  732. }\

  733. \

  734. static void OPNAME ## qpel16_mc01_sh4(uint8_t *dst, uint8_t *src, int stride){\

  735.     uint8_t full[24*17];\

  736.     uint8_t half[256];\

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

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

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

  740. }\

  741. \

  742. static void OPNAME ## qpel16_mc02_sh4(uint8_t *dst, uint8_t *src, int stride){\

  743.     uint8_t full[24*17];\

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

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

  746. }\

  747. \

  748. static void OPNAME ## qpel16_mc03_sh4(uint8_t *dst, uint8_t *src, int stride){\

  749.     uint8_t full[24*17];\

  750.     uint8_t half[256];\

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

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

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

  754. }\

  755. static void OPNAME ## qpel16_mc11_sh4(uint8_t *dst, uint8_t *src, int stride){\

  756.     uint8_t full[24*17];\

  757.     uint8_t halfH[272];\

  758.     uint8_t halfHV[256];\

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

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

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

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

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

  764. }\

  765. static void OPNAME ## qpel16_mc31_sh4(uint8_t *dst, uint8_t *src, int stride){\

  766.     uint8_t full[24*17];\

  767.     uint8_t halfH[272];\

  768.     uint8_t halfHV[256];\

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

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

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

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

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

  774. }\

  775. static void OPNAME ## qpel16_mc13_sh4(uint8_t *dst, uint8_t *src, int stride){\

  776.     uint8_t full[24*17];\

  777.     uint8_t halfH[272];\

  778.     uint8_t halfHV[256];\

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

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

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

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

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

  784. }\

  785. static void OPNAME ## qpel16_mc33_sh4(uint8_t *dst, uint8_t *src, int stride){\

  786.     uint8_t full[24*17];\

  787.     uint8_t halfH[272];\

  788.     uint8_t halfHV[256];\

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

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

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

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

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

  794. }\

  795. static void OPNAME ## qpel16_mc21_sh4(uint8_t *dst, uint8_t *src, int stride){\

  796.     uint8_t halfH[272];\

  797.     uint8_t halfHV[256];\

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

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

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

  801. }\

  802. static void OPNAME ## qpel16_mc23_sh4(uint8_t *dst, uint8_t *src, int stride){\

  803.     uint8_t halfH[272];\

  804.     uint8_t halfHV[256];\

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

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

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

  808. }\

  809. static void OPNAME ## qpel16_mc12_sh4(uint8_t *dst, uint8_t *src, int stride){\

  810.     uint8_t full[24*17];\

  811.     uint8_t halfH[272];\

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

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

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

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

  816. }\

  817. static void OPNAME ## qpel16_mc32_sh4(uint8_t *dst, uint8_t *src, int stride){\

  818.     uint8_t full[24*17];\

  819.     uint8_t halfH[272];\

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

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

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

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

  824. }\

  825. static void OPNAME ## qpel16_mc22_sh4(uint8_t *dst, uint8_t *src, int stride){\

  826.     uint8_t halfH[272];\

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

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

  829. }

  830. 

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

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

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

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

  835. 

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

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

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

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

  840. #undef op_avg

  841. #undef op_avg_no_rnd

  842. #undef op_put

  843. #undef op_put_no_rnd

  844. 

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

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

  847.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  848.     do {\

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

  850.         uint8_t *s = src-2;\

  851.         srcB = *s++;\

  852.         srcA = *s++;\

  853.         src0 = *s++;\

  854.         src1 = *s++;\

  855.         src2 = *s++;\

  856.         src3 = *s++;\

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

  858.         src4 = *s++;\

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

  860.         src5 = *s++;\

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

  862.         src6 = *s++;\

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

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

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

  866.         src7 = *s++;\

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

  868.         src8 = *s++;\

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

  870.         src9 = *s++;\

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

  872.         src10 = *s++;\

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

  874.        if (w>8) { \

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

  876.         src11 = *s++;\

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

  878.         src12 = *s++;\

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

  880.         src13 = *s++;\

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

  882.         src14 = *s++;\

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

  884.         src15 = *s++;\

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

  886.         src16 = *s++;\

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

  888.         src17 = *s++;\

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

  890.         src18 = *s++;\

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

  892.        } \

  893.       } \

  894.         dst+=dstStride;\

  895.         src+=srcStride;\

  896.     }while(--h);\

  897. }\

  898. \

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

  900.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  901.     do{\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  917.       if (h>4) { \

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

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

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

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

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

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

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

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

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

  927.        if (h>8) { \

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  945.        } \

  946.       } \

  947.         dst++;\

  948.         src++;\

  949.     }while(--w);\

  950. }\

  951. \

  952. 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){\

  953.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  954.     int i;\

  955.     src -= 2*srcStride;\

  956.     i= h+5; \

  957.     do {\

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

  959.         uint8_t *s = src-2;\

  960.         srcB = *s++;\

  961.         srcA = *s++;\

  962.         src0 = *s++;\

  963.         src1 = *s++;\

  964.         src2 = *s++;\

  965.         src3 = *s++;\

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

  967.         src4 = *s++;\

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

  969.         src5 = *s++;\

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

  971.         src6 = *s++;\

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

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

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

  975.         src7 = *s++;\

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

  977.         src8 = *s++;\

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

  979.         src9 = *s++;\

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

  981.         src10 = *s++;\

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

  983.        if (w>8) { \

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

  985.         src11 = *s++;\

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

  987.         src12 = *s++;\

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

  989.         src13 = *s++;\

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

  991.         src14 = *s++;\

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

  993.         src15 = *s++;\

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

  995.         src16 = *s++;\

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

  997.         src17 = *s++;\

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

  999.         src18 = *s++;\

  1000.         tmp[15] = ((src15+src16)*20 - (src14+src17)*5 + (src13+src18));\

  1001.        } \

  1002.       } \

  1003.         tmp+=tmpStride;\

  1004.         src+=srcStride;\

  1005.     }while(--i);\

  1006.     tmp -= tmpStride*(h+5-2);\

  1007.     i = w; \

  1008.     do {\

  1009.         int tmpB,tmpA,tmp0,tmp1,tmp2,tmp3,tmp4,tmp5,tmp6;\

  1010.         int16_t *s = tmp-2*tmpStride; \

  1011.         uint8_t *d=dst;\

  1012.         tmpB = *s; s+=tmpStride;\

  1013.         tmpA = *s; s+=tmpStride;\

  1014.         tmp0 = *s; s+=tmpStride;\

  1015.         tmp1 = *s; s+=tmpStride;\

  1016.         tmp2 = *s; s+=tmpStride;\

  1017.         tmp3 = *s; s+=tmpStride;\

  1018.         OP2(*d, (tmp0+tmp1)*20 - (tmpA+tmp2)*5 + (tmpB+tmp3));d+=dstStride;\

  1019.         tmp4 = *s; s+=tmpStride;\

  1020.         OP2(*d, (tmp1+tmp2)*20 - (tmp0+tmp3)*5 + (tmpA+tmp4));d+=dstStride;\

  1021.         tmp5 = *s; s+=tmpStride;\

  1022.         OP2(*d, (tmp2+tmp3)*20 - (tmp1+tmp4)*5 + (tmp0+tmp5));d+=dstStride;\

  1023.         tmp6 = *s; s+=tmpStride;\

  1024.         OP2(*d, (tmp3+tmp4)*20 - (tmp2+tmp5)*5 + (tmp1+tmp6));d+=dstStride;\

  1025.       if (h>4) { \

  1026.         int tmp7,tmp8,tmp9,tmp10; \

  1027.         tmp7 = *s; s+=tmpStride;\

  1028.         OP2(*d, (tmp4+tmp5)*20 - (tmp3+tmp6)*5 + (tmp2+tmp7));d+=dstStride;\

  1029.         tmp8 = *s; s+=tmpStride;\

  1030.         OP2(*d, (tmp5+tmp6)*20 - (tmp4+tmp7)*5 + (tmp3+tmp8));d+=dstStride;\

  1031.         tmp9 = *s; s+=tmpStride;\

  1032.         OP2(*d, (tmp6+tmp7)*20 - (tmp5+tmp8)*5 + (tmp4+tmp9));d+=dstStride;\

  1033.         tmp10 = *s; s+=tmpStride;\

  1034.         OP2(*d, (tmp7+tmp8)*20 - (tmp6+tmp9)*5 + (tmp5+tmp10));d+=dstStride;\

  1035.        if (h>8) { \

  1036.         int tmp11,tmp12,tmp13,tmp14,tmp15,tmp16,tmp17,tmp18; \

  1037.         tmp11 = *s; s+=tmpStride;\

  1038.         OP2(*d , (tmp8 +tmp9 )*20 - (tmp7 +tmp10)*5 + (tmp6 +tmp11));d+=dstStride;\

  1039.         tmp12 = *s; s+=tmpStride;\

  1040.         OP2(*d , (tmp9 +tmp10)*20 - (tmp8 +tmp11)*5 + (tmp7 +tmp12));d+=dstStride;\

  1041.         tmp13 = *s; s+=tmpStride;\

  1042.         OP2(*d, (tmp10+tmp11)*20 - (tmp9 +tmp12)*5 + (tmp8 +tmp13));d+=dstStride;\

  1043.         tmp14 = *s; s+=tmpStride;\

  1044.         OP2(*d, (tmp11+tmp12)*20 - (tmp10+tmp13)*5 + (tmp9 +tmp14));d+=dstStride;\

  1045.         tmp15 = *s; s+=tmpStride;\

  1046.         OP2(*d, (tmp12+tmp13)*20 - (tmp11+tmp14)*5 + (tmp10+tmp15));d+=dstStride;\

  1047.         tmp16 = *s; s+=tmpStride;\

  1048.         OP2(*d, (tmp13+tmp14)*20 - (tmp12+tmp15)*5 + (tmp11+tmp16));d+=dstStride;\

  1049.         tmp17 = *s; s+=tmpStride;\

  1050.         OP2(*d, (tmp14+tmp15)*20 - (tmp13+tmp16)*5 + (tmp12+tmp17));d+=dstStride;\

  1051.         tmp18 = *s; s+=tmpStride;\

  1052.         OP2(*d, (tmp15+tmp16)*20 - (tmp14+tmp17)*5 + (tmp13+tmp18));d+=dstStride;\

  1053.        } \

  1054.       } \

  1055.         dst++;\

  1056.         tmp++;\

  1057.     }while(--i);\

  1058. }\

  1059. \

  1060. static void OPNAME ## h264_qpel4_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1061.     OPNAME ## h264_qpel_h_lowpass(dst,src,dstStride,srcStride,4,4); \

  1062. }\

  1063. static void OPNAME ## h264_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1064.    OPNAME ## h264_qpel_h_lowpass(dst,src,dstStride,srcStride,8,8); \

  1065. }\

  1066. static void OPNAME ## h264_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1067.    OPNAME ## h264_qpel_h_lowpass(dst,src,dstStride,srcStride,16,16); \

  1068. }\

  1069. \

  1070. static void OPNAME ## h264_qpel4_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1071.    OPNAME ## h264_qpel_v_lowpass(dst,src,dstStride,srcStride,4,4); \

  1072. }\

  1073. static void OPNAME ## h264_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1074.    OPNAME ## h264_qpel_v_lowpass(dst,src,dstStride,srcStride,8,8); \

  1075. }\

  1076. static void OPNAME ## h264_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1077.    OPNAME ## h264_qpel_v_lowpass(dst,src,dstStride,srcStride,16,16); \

  1078. }\

  1079. static void OPNAME ## h264_qpel4_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\

  1080.    OPNAME ## h264_qpel_hv_lowpass(dst,tmp,src,dstStride,tmpStride,srcStride,4,4); \

  1081. }\

  1082. static void OPNAME ## h264_qpel8_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\

  1083.    OPNAME ## h264_qpel_hv_lowpass(dst,tmp,src,dstStride,tmpStride,srcStride,8,8); \

  1084. }\

  1085. static void OPNAME ## h264_qpel16_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\

  1086.    OPNAME ## h264_qpel_hv_lowpass(dst,tmp,src,dstStride,tmpStride,srcStride,16,16); \

  1087. }\

  1088. 

  1089. #define H264_MC(OPNAME, SIZE) \

  1090. static void OPNAME ## h264_qpel ## SIZE ## _mc00_sh4 (uint8_t *dst, uint8_t *src, int stride){\

  1091.     OPNAME ## pixels ## SIZE ## _c(dst, src, stride, SIZE);\

  1092. }\

  1093. \

  1094. static void OPNAME ## h264_qpel ## SIZE ## _mc10_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1095.     uint8_t half[SIZE*SIZE];\

  1096.     put_h264_qpel ## SIZE ## _h_lowpass(half, src, SIZE, stride);\

  1097.     OPNAME ## pixels ## SIZE ## _l2_aligned2(dst, src, half, stride, stride, SIZE, SIZE);\

  1098. }\

  1099. \

  1100. static void OPNAME ## h264_qpel ## SIZE ## _mc20_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1101.     OPNAME ## h264_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride);\

  1102. }\

  1103. \

  1104. static void OPNAME ## h264_qpel ## SIZE ## _mc30_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1105.     uint8_t half[SIZE*SIZE];\

  1106.     put_h264_qpel ## SIZE ## _h_lowpass(half, src, SIZE, stride);\

  1107.     OPNAME ## pixels ## SIZE ## _l2_aligned2(dst, src+1, half, stride, stride, SIZE, SIZE);\

  1108. }\

  1109. \

  1110. static void OPNAME ## h264_qpel ## SIZE ## _mc01_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1111.     uint8_t full[SIZE*(SIZE+5)];\

  1112.     uint8_t * const full_mid= full + SIZE*2;\

  1113.     uint8_t half[SIZE*SIZE];\

  1114.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1115.     put_h264_qpel ## SIZE ## _v_lowpass(half, full_mid, SIZE, SIZE);\

  1116.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, full_mid, half, stride, SIZE, SIZE, SIZE);\

  1117. }\

  1118. \

  1119. static void OPNAME ## h264_qpel ## SIZE ## _mc02_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1120.     uint8_t full[SIZE*(SIZE+5)];\

  1121.     uint8_t * const full_mid= full + SIZE*2;\

  1122.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1123.     OPNAME ## h264_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE);\

  1124. }\

  1125. \

  1126. static void OPNAME ## h264_qpel ## SIZE ## _mc03_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1127.     uint8_t full[SIZE*(SIZE+5)];\

  1128.     uint8_t * const full_mid= full + SIZE*2;\

  1129.     uint8_t half[SIZE*SIZE];\

  1130.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1131.     put_h264_qpel ## SIZE ## _v_lowpass(half, full_mid, SIZE, SIZE);\

  1132.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, full_mid+SIZE, half, stride, SIZE, SIZE, SIZE);\

  1133. }\

  1134. \

  1135. static void OPNAME ## h264_qpel ## SIZE ## _mc11_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1136.     uint8_t full[SIZE*(SIZE+5)];\

  1137.     uint8_t * const full_mid= full + SIZE*2;\

  1138.     uint8_t halfH[SIZE*SIZE];\

  1139.     uint8_t halfV[SIZE*SIZE];\

  1140.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\

  1141.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1142.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1143.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\

  1144. }\

  1145. \

  1146. static void OPNAME ## h264_qpel ## SIZE ## _mc31_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1147.     uint8_t full[SIZE*(SIZE+5)];\

  1148.     uint8_t * const full_mid= full + SIZE*2;\

  1149.     uint8_t halfH[SIZE*SIZE];\

  1150.     uint8_t halfV[SIZE*SIZE];\

  1151.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\

  1152.     copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\

  1153.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1154.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\

  1155. }\

  1156. \

  1157. static void OPNAME ## h264_qpel ## SIZE ## _mc13_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1158.     uint8_t full[SIZE*(SIZE+5)];\

  1159.     uint8_t * const full_mid= full + SIZE*2;\

  1160.     uint8_t halfH[SIZE*SIZE];\

  1161.     uint8_t halfV[SIZE*SIZE];\

  1162.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\

  1163.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1164.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1165.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\

  1166. }\

  1167. \

  1168. static void OPNAME ## h264_qpel ## SIZE ## _mc33_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1169.     uint8_t full[SIZE*(SIZE+5)];\

  1170.     uint8_t * const full_mid= full + SIZE*2;\

  1171.     uint8_t halfH[SIZE*SIZE];\

  1172.     uint8_t halfV[SIZE*SIZE];\

  1173.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\

  1174.     copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\

  1175.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1176.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\

  1177. }\

  1178. \

  1179. static void OPNAME ## h264_qpel ## SIZE ## _mc22_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1180.     int16_t tmp[SIZE*(SIZE+5)];\

  1181.     OPNAME ## h264_qpel ## SIZE ## _hv_lowpass(dst, tmp, src, stride, SIZE, stride);\

  1182. }\

  1183. \

  1184. static void OPNAME ## h264_qpel ## SIZE ## _mc21_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1185.     int16_t tmp[SIZE*(SIZE+5)];\

  1186.     uint8_t halfH[SIZE*SIZE];\

  1187.     uint8_t halfHV[SIZE*SIZE];\

  1188.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\

  1189.     put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\

  1190.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfHV, stride, SIZE, SIZE, SIZE);\

  1191. }\

  1192. \

  1193. static void OPNAME ## h264_qpel ## SIZE ## _mc23_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1194.     int16_t tmp[SIZE*(SIZE+5)];\

  1195.     uint8_t halfH[SIZE*SIZE];\

  1196.     uint8_t halfHV[SIZE*SIZE];\

  1197.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\

  1198.     put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\

  1199.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfHV, stride, SIZE, SIZE, SIZE);\

  1200. }\

  1201. \

  1202. static void OPNAME ## h264_qpel ## SIZE ## _mc12_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1203.     uint8_t full[SIZE*(SIZE+5)];\

  1204.     uint8_t * const full_mid= full + SIZE*2;\

  1205.     int16_t tmp[SIZE*(SIZE+5)];\

  1206.     uint8_t halfV[SIZE*SIZE];\

  1207.     uint8_t halfHV[SIZE*SIZE];\

  1208.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1209.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1210.     put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\

  1211.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfV, halfHV, stride, SIZE, SIZE, SIZE);\

  1212. }\

  1213. \

  1214. static void OPNAME ## h264_qpel ## SIZE ## _mc32_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1215.     uint8_t full[SIZE*(SIZE+5)];\

  1216.     uint8_t * const full_mid= full + SIZE*2;\

  1217.     int16_t tmp[SIZE*(SIZE+5)];\

  1218.     uint8_t halfV[SIZE*SIZE];\

  1219.     uint8_t halfHV[SIZE*SIZE];\

  1220.     copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\

  1221.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1222.     put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\

  1223.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfV, halfHV, stride, SIZE, SIZE, SIZE);\

  1224. }\

  1225. 

  1226. #define op_avg(a, b)  a = (((a)+cm[((b) + 16)>>5]+1)>>1)

  1227. //#define op_avg2(a, b) a = (((a)*w1+cm[((b) + 16)>>5]*w2 + o + 64)>>7)

  1228. #define op_put(a, b)  a = cm[((b) + 16)>>5]

  1229. #define op2_avg(a, b)  a = (((a)+cm[((b) + 512)>>10]+1)>>1)

  1230. #define op2_put(a, b)  a = cm[((b) + 512)>>10]

  1231. 

  1232. H264_LOWPASS(put_       , op_put, op2_put)

  1233. H264_LOWPASS(avg_       , op_avg, op2_avg)

  1234. H264_MC(put_, 4)

  1235. H264_MC(put_, 8)

  1236. H264_MC(put_, 16)

  1237. H264_MC(avg_, 4)

  1238. H264_MC(avg_, 8)

  1239. H264_MC(avg_, 16)

  1240. 

  1241. #undef op_avg

  1242. #undef op_put

  1243. #undef op2_avg

  1244. #undef op2_put

  1245. 

  1246. static void wmv2_mspel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){

  1247.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;

  1248. 

  1249.     do{

  1250.         int src_1,src0,src1,src2,src3,src4,src5,src6,src7,src8,src9;

  1251.         uint8_t *s = src;

  1252.         src_1 = s[-1];

  1253.         src0 = *s++;

  1254.         src1 = *s++;

  1255.         src2 = *s++;

  1256.         dst[0]= cm[(9*(src0 + src1) - (src_1 + src2) + 8)>>4];

  1257.         src3 = *s++;

  1258.         dst[1]= cm[(9*(src1 + src2) - (src0 + src3) + 8)>>4];

  1259.         src4 = *s++;

  1260.         dst[2]= cm[(9*(src2 + src3) - (src1 + src4) + 8)>>4];

  1261.         src5 = *s++;

  1262.         dst[3]= cm[(9*(src3 + src4) - (src2 + src5) + 8)>>4];

  1263.         src6 = *s++;

  1264.         dst[4]= cm[(9*(src4 + src5) - (src3 + src6) + 8)>>4];

  1265.         src7 = *s++;

  1266.         dst[5]= cm[(9*(src5 + src6) - (src4 + src7) + 8)>>4];

  1267.         src8 = *s++;

  1268.         dst[6]= cm[(9*(src6 + src7) - (src5 + src8) + 8)>>4];

  1269.         src9 = *s++;

  1270.         dst[7]= cm[(9*(src7 + src8) - (src6 + src9) + 8)>>4];

  1271.         dst+=dstStride;

  1272.         src+=srcStride;

  1273.     }while(--h);

  1274. }

  1275. 

  1276. static void wmv2_mspel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int w){

  1277.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;

  1278. 

  1279.     do{

  1280.         int src_1,src0,src1,src2,src3,src4,src5,src6,src7,src8,src9;

  1281.         uint8_t *s = src,*d = dst;

  1282.         src_1 = *(s-srcStride);

  1283.         src0 = *s; s+=srcStride;

  1284.         src1 = *s; s+=srcStride;

  1285.         src2 = *s; s+=srcStride;

  1286.         *d= cm[(9*(src0 + src1) - (src_1 + src2) + 8)>>4]; d+=dstStride;

  1287.         src3 = *s; s+=srcStride;

  1288.         *d= cm[(9*(src1 + src2) - (src0  + src3) + 8)>>4]; d+=dstStride;

  1289.         src4 = *s; s+=srcStride;

  1290.         *d= cm[(9*(src2 + src3) - (src1  + src4) + 8)>>4]; d+=dstStride;

  1291.         src5 = *s; s+=srcStride;

  1292.         *d= cm[(9*(src3 + src4) - (src2  + src5) + 8)>>4]; d+=dstStride;

  1293.         src6 = *s; s+=srcStride;

  1294.         *d= cm[(9*(src4 + src5) - (src3  + src6) + 8)>>4]; d+=dstStride;

  1295.         src7 = *s; s+=srcStride;

  1296.         *d= cm[(9*(src5 + src6) - (src4  + src7) + 8)>>4]; d+=dstStride;

  1297.         src8 = *s; s+=srcStride;

  1298.         *d= cm[(9*(src6 + src7) - (src5  + src8) + 8)>>4]; d+=dstStride;

  1299.         src9 = *s;

  1300.         *d= cm[(9*(src7 + src8) - (src6  + src9) + 8)>>4]; d+=dstStride;

  1301.         src++;

  1302.         dst++;

  1303.     }while(--w);

  1304. }

  1305. 

  1306. static void put_mspel8_mc00_sh4 (uint8_t *dst, uint8_t *src, int stride){

  1307.     put_pixels8_c(dst, src, stride, 8);

  1308. }

  1309. 

  1310. static void put_mspel8_mc10_sh4(uint8_t *dst, uint8_t *src, int stride){

  1311.     uint8_t half[64];

  1312.     wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);

  1313.     put_pixels8_l2_aligned2(dst, src, half, stride, stride, 8, 8);

  1314. }

  1315. 

  1316. static void put_mspel8_mc20_sh4(uint8_t *dst, uint8_t *src, int stride){

  1317.     wmv2_mspel8_h_lowpass(dst, src, stride, stride, 8);

  1318. }

  1319. 

  1320. static void put_mspel8_mc30_sh4(uint8_t *dst, uint8_t *src, int stride){

  1321.     uint8_t half[64];

  1322.     wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);

  1323.     put_pixels8_l2_aligned2(dst, src+1, half, stride, stride, 8, 8);

  1324. }

  1325. 

  1326. static void put_mspel8_mc02_sh4(uint8_t *dst, uint8_t *src, int stride){

  1327.     wmv2_mspel8_v_lowpass(dst, src, stride, stride, 8);

  1328. }

  1329. 

  1330. static void put_mspel8_mc12_sh4(uint8_t *dst, uint8_t *src, int stride){

  1331.     uint8_t halfH[88];

  1332.     uint8_t halfV[64];

  1333.     uint8_t halfHV[64];

  1334.     wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);

  1335.     wmv2_mspel8_v_lowpass(halfV, src, 8, stride, 8);

  1336.     wmv2_mspel8_v_lowpass(halfHV, halfH+8, 8, 8, 8);

  1337.     put_pixels8_l2_aligned(dst, halfV, halfHV, stride, 8, 8, 8);

  1338. }

  1339. static void put_mspel8_mc32_sh4(uint8_t *dst, uint8_t *src, int stride){

  1340.     uint8_t halfH[88];

  1341.     uint8_t halfV[64];

  1342.     uint8_t halfHV[64];

  1343.     wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);

  1344.     wmv2_mspel8_v_lowpass(halfV, src+1, 8, stride, 8);

  1345.     wmv2_mspel8_v_lowpass(halfHV, halfH+8, 8, 8, 8);

  1346.     put_pixels8_l2_aligned(dst, halfV, halfHV, stride, 8, 8, 8);

  1347. }

  1348. static void put_mspel8_mc22_sh4(uint8_t *dst, uint8_t *src, int stride){

  1349.     uint8_t halfH[88];

  1350.     wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);

  1351.     wmv2_mspel8_v_lowpass(dst, halfH+8, stride, 8, 8);

  1352. }