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.
35406 Commits
32 Branches
292MB
Tree: c3e15f7b39
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'c3e15f7b39'
${ noResults }
FFmpeg/libavcodec/sh4/qpel.c

1410 lines
61KB
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. static void gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,

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

  364. {

  365.     int y, vx, vy;

  366.     const int s= 1<<shift;

  367. 

  368.     width--;

  369.     height--;

  370. 

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

  372.         int x;

  373. 

  374.         vx= ox;

  375.         vy= oy;

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

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

  378. 

  379.             src_x= vx>>16;

  380.             src_y= vy>>16;

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

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

  383.             src_x>>=shift;

  384.             src_y>>=shift;

  385. 

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

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

  388.                     index= src_x + src_y*stride;

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

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

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

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

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

  394.                 }else{

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

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

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

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

  399.                 }

  400.             }else{

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

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

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

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

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

  406.                 }else{

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

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

  409.                 }

  410.             }

  411. 

  412.             vx+= dxx;

  413.             vy+= dyx;

  414.         }

  415.         ox += dxy;

  416.         oy += dyy;

  417.     }

  418. }

  419. #define H264_CHROMA_MC(OPNAME, OP)\

  420. 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){\

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

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

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

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

  425.     \

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

  427. \

  428.     do {\

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

  430.         uint8_t *s0 = src; \

  431.         uint8_t *s1 = src+stride; \

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

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

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

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

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

  437.         dst+= stride;\

  438.         src+= stride;\

  439.     }while(--h);\

  440. }\

  441. \

  442. 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){\

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

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

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

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

  447.     \

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

  449. \

  450.     do {\

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

  452.         uint8_t *s0 = src; \

  453.         uint8_t *s1 = src+stride; \

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

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

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

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

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

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

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

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

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

  463.         dst+= stride;\

  464.         src+= stride;\

  465.     }while(--h);\

  466. }\

  467. \

  468. 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){\

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

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

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

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

  473.     \

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

  475. \

  476.     do {\

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

  478.         uint8_t *s0 = src; \

  479.         uint8_t *s1 = src+stride; \

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  497.         dst+= stride;\

  498.         src+= stride;\

  499.     }while(--h);\

  500. }

  501. 

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

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

  504. 

  505. H264_CHROMA_MC(put_       , op_put)

  506. H264_CHROMA_MC(avg_       , op_avg)

  507. #undef op_avg

  508. #undef op_put

  509. 

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

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

  512.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  513.     do {\

  514.         uint8_t *s = src; \

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

  516.         src0= *s++;\

  517.         src1= *s++;\

  518.         src2= *s++;\

  519.         src3= *s++;\

  520.         src4= *s++;\

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

  522.         src5= *s++;\

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

  524.         src6= *s++;\

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

  526.         src7= *s++;\

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

  528.         src8= *s++;\

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

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

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

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

  533.         dst+=dstStride;\

  534.         src+=srcStride;\

  535.     }while(--h);\

  536. }\

  537. \

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

  539.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  540.     int w=8;\

  541.     do{\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  556.         src8 = *s; \

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

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

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

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

  561.         dst++;\

  562.         src++;\

  563.     }while(--w);\

  564. }\

  565. \

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

  567.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  568.     do {\

  569.         uint8_t *s = src;\

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

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

  572.         src0= *s++;\

  573.         src1= *s++;\

  574.         src2= *s++;\

  575.         src3= *s++;\

  576.         src4= *s++;\

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

  578.         src5= *s++;\

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

  580.         src6= *s++;\

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

  582.         src7= *s++;\

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

  584.         src8= *s++;\

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

  586.         src9= *s++;\

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

  588.         src10= *s++;\

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

  590.         src11= *s++;\

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

  592.         src12= *s++;\

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

  594.         src13= *s++;\

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

  596.         src14= *s++;\

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

  598.         src15= *s++;\

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

  600.         src16= *s++;\

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

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

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

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

  605.         dst+=dstStride;\

  606.         src+=srcStride;\

  607.     }while(--h);\

  608. }\

  609. \

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

  611.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  612.     int w=16;\

  613.     do {\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  645.         src16 = *s; \

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

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

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

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

  650.         dst++;\

  651.         src++;\

  652.     }while(--w);\

  653. }\

  654. \

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

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

  657. }\

  658. \

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

  660.     uint8_t half[64];\

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

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

  663. }\

  664. \

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

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

  667. }\

  668. \

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

  670.     uint8_t half[64];\

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

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

  673. }\

  674. \

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

  676.     uint8_t full[16*9];\

  677.     uint8_t half[64];\

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

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

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

  681. }\

  682. \

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

  684.     uint8_t full[16*9];\

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

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

  687. }\

  688. \

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

  690.     uint8_t full[16*9];\

  691.     uint8_t half[64];\

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

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

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

  695. }\

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

  697.     uint8_t full[16*9];\

  698.     uint8_t halfH[72];\

  699.     uint8_t halfHV[64];\

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

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

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

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

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

  705. }\

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

  707.     uint8_t full[16*9];\

  708.     uint8_t halfH[72];\

  709.     uint8_t halfHV[64];\

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

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

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

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

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

  715. }\

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

  717.     uint8_t full[16*9];\

  718.     uint8_t halfH[72];\

  719.     uint8_t halfHV[64];\

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

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

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

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

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

  725. }\

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

  727.     uint8_t full[16*9];\

  728.     uint8_t halfH[72];\

  729.     uint8_t halfHV[64];\

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

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

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

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

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

  735. }\

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

  737.     uint8_t halfH[72];\

  738.     uint8_t halfHV[64];\

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

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

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

  742. }\

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

  744.     uint8_t halfH[72];\

  745.     uint8_t halfHV[64];\

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

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

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

  749. }\

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

  751.     uint8_t full[16*9];\

  752.     uint8_t halfH[72];\

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

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

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

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

  757. }\

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

  759.     uint8_t full[16*9];\

  760.     uint8_t halfH[72];\

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

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

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

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

  765. }\

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

  767.     uint8_t halfH[72];\

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

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

  770. }\

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

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

  773. }\

  774. \

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

  776.     uint8_t half[256];\

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

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

  779. }\

  780. \

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

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

  783. }\

  784. \

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

  786.     uint8_t half[256];\

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

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

  789. }\

  790. \

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

  792.     uint8_t full[24*17];\

  793.     uint8_t half[256];\

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

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

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

  797. }\

  798. \

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

  800.     uint8_t full[24*17];\

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

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

  803. }\

  804. \

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

  806.     uint8_t full[24*17];\

  807.     uint8_t half[256];\

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

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

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

  811. }\

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

  813.     uint8_t full[24*17];\

  814.     uint8_t halfH[272];\

  815.     uint8_t halfHV[256];\

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

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

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

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

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

  821. }\

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

  823.     uint8_t full[24*17];\

  824.     uint8_t halfH[272];\

  825.     uint8_t halfHV[256];\

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

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

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

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

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

  831. }\

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

  833.     uint8_t full[24*17];\

  834.     uint8_t halfH[272];\

  835.     uint8_t halfHV[256];\

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

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

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

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

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

  841. }\

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

  843.     uint8_t full[24*17];\

  844.     uint8_t halfH[272];\

  845.     uint8_t halfHV[256];\

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

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

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

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

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

  851. }\

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

  853.     uint8_t halfH[272];\

  854.     uint8_t halfHV[256];\

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

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

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

  858. }\

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

  860.     uint8_t halfH[272];\

  861.     uint8_t halfHV[256];\

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

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

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

  865. }\

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

  867.     uint8_t full[24*17];\

  868.     uint8_t halfH[272];\

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

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

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

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

  873. }\

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

  875.     uint8_t full[24*17];\

  876.     uint8_t halfH[272];\

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

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

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

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

  881. }\

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

  883.     uint8_t halfH[272];\

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

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

  886. }

  887. 

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

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

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

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

  892. 

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

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

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

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

  897. #undef op_avg

  898. #undef op_avg_no_rnd

  899. #undef op_put

  900. #undef op_put_no_rnd

  901. 

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

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

  904.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  905.     do {\

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

  907.         uint8_t *s = src-2;\

  908.         srcB = *s++;\

  909.         srcA = *s++;\

  910.         src0 = *s++;\

  911.         src1 = *s++;\

  912.         src2 = *s++;\

  913.         src3 = *s++;\

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

  915.         src4 = *s++;\

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

  917.         src5 = *s++;\

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

  919.         src6 = *s++;\

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

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

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

  923.         src7 = *s++;\

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

  925.         src8 = *s++;\

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

  927.         src9 = *s++;\

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

  929.         src10 = *s++;\

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

  931.        if (w>8) { \

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

  933.         src11 = *s++;\

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

  935.         src12 = *s++;\

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

  937.         src13 = *s++;\

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

  939.         src14 = *s++;\

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

  941.         src15 = *s++;\

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

  943.         src16 = *s++;\

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

  945.         src17 = *s++;\

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

  947.         src18 = *s++;\

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

  949.        } \

  950.       } \

  951.         dst+=dstStride;\

  952.         src+=srcStride;\

  953.     }while(--h);\

  954. }\

  955. \

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

  957.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  958.     do{\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  974.       if (h>4) { \

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

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

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

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

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

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

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

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

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

  984.        if (h>8) { \

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  1000.         src18 = *s; s+=srcStride;\

  1001.         OP(*d, (src15+src16)*20 - (src14+src17)*5 + (src13+src18));d+=dstStride;\

  1002.        } \

  1003.       } \

  1004.         dst++;\

  1005.         src++;\

  1006.     }while(--w);\

  1007. }\

  1008. \

  1009. 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){\

  1010.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  1011.     int i;\

  1012.     src -= 2*srcStride;\

  1013.     i= h+5; \

  1014.     do {\

  1015.         int srcB,srcA,src0,src1,src2,src3,src4,src5,src6;\

  1016.         uint8_t *s = src-2;\

  1017.         srcB = *s++;\

  1018.         srcA = *s++;\

  1019.         src0 = *s++;\

  1020.         src1 = *s++;\

  1021.         src2 = *s++;\

  1022.         src3 = *s++;\

  1023.         tmp[0] = ((src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));\

  1024.         src4 = *s++;\

  1025.         tmp[1] = ((src1+src2)*20 - (src0+src3)*5 + (srcA+src4));\

  1026.         src5 = *s++;\

  1027.         tmp[2] = ((src2+src3)*20 - (src1+src4)*5 + (src0+src5));\

  1028.         src6 = *s++;\

  1029.         tmp[3] = ((src3+src4)*20 - (src2+src5)*5 + (src1+src6));\

  1030.       if (w>4) { /* it optimized */ \

  1031.         int src7,src8,src9,src10; \

  1032.         src7 = *s++;\

  1033.         tmp[4] = ((src4+src5)*20 - (src3+src6)*5 + (src2+src7));\

  1034.         src8 = *s++;\

  1035.         tmp[5] = ((src5+src6)*20 - (src4+src7)*5 + (src3+src8));\

  1036.         src9 = *s++;\

  1037.         tmp[6] = ((src6+src7)*20 - (src5+src8)*5 + (src4+src9));\

  1038.         src10 = *s++;\

  1039.         tmp[7] = ((src7+src8)*20 - (src6+src9)*5 + (src5+src10));\

  1040.        if (w>8) { \

  1041.         int src11,src12,src13,src14,src15,src16,src17,src18; \

  1042.         src11 = *s++;\

  1043.         tmp[8] = ((src8 +src9 )*20 - (src7 +src10)*5 + (src6 +src11));\

  1044.         src12 = *s++;\

  1045.         tmp[9] = ((src9 +src10)*20 - (src8 +src11)*5 + (src7 +src12));\

  1046.         src13 = *s++;\

  1047.         tmp[10] = ((src10+src11)*20 - (src9 +src12)*5 + (src8 +src13));\

  1048.         src14 = *s++;\

  1049.         tmp[11] = ((src11+src12)*20 - (src10+src13)*5 + (src9 +src14));\

  1050.         src15 = *s++;\

  1051.         tmp[12] = ((src12+src13)*20 - (src11+src14)*5 + (src10+src15));\

  1052.         src16 = *s++;\

  1053.         tmp[13] = ((src13+src14)*20 - (src12+src15)*5 + (src11+src16));\

  1054.         src17 = *s++;\

  1055.         tmp[14] = ((src14+src15)*20 - (src13+src16)*5 + (src12+src17));\

  1056.         src18 = *s++;\

  1057.         tmp[15] = ((src15+src16)*20 - (src14+src17)*5 + (src13+src18));\

  1058.        } \

  1059.       } \

  1060.         tmp+=tmpStride;\

  1061.         src+=srcStride;\

  1062.     }while(--i);\

  1063.     tmp -= tmpStride*(h+5-2);\

  1064.     i = w; \

  1065.     do {\

  1066.         int tmpB,tmpA,tmp0,tmp1,tmp2,tmp3,tmp4,tmp5,tmp6;\

  1067.         int16_t *s = tmp-2*tmpStride; \

  1068.         uint8_t *d=dst;\

  1069.         tmpB = *s; s+=tmpStride;\

  1070.         tmpA = *s; s+=tmpStride;\

  1071.         tmp0 = *s; s+=tmpStride;\

  1072.         tmp1 = *s; s+=tmpStride;\

  1073.         tmp2 = *s; s+=tmpStride;\

  1074.         tmp3 = *s; s+=tmpStride;\

  1075.         OP2(*d, (tmp0+tmp1)*20 - (tmpA+tmp2)*5 + (tmpB+tmp3));d+=dstStride;\

  1076.         tmp4 = *s; s+=tmpStride;\

  1077.         OP2(*d, (tmp1+tmp2)*20 - (tmp0+tmp3)*5 + (tmpA+tmp4));d+=dstStride;\

  1078.         tmp5 = *s; s+=tmpStride;\

  1079.         OP2(*d, (tmp2+tmp3)*20 - (tmp1+tmp4)*5 + (tmp0+tmp5));d+=dstStride;\

  1080.         tmp6 = *s; s+=tmpStride;\

  1081.         OP2(*d, (tmp3+tmp4)*20 - (tmp2+tmp5)*5 + (tmp1+tmp6));d+=dstStride;\

  1082.       if (h>4) { \

  1083.         int tmp7,tmp8,tmp9,tmp10; \

  1084.         tmp7 = *s; s+=tmpStride;\

  1085.         OP2(*d, (tmp4+tmp5)*20 - (tmp3+tmp6)*5 + (tmp2+tmp7));d+=dstStride;\

  1086.         tmp8 = *s; s+=tmpStride;\

  1087.         OP2(*d, (tmp5+tmp6)*20 - (tmp4+tmp7)*5 + (tmp3+tmp8));d+=dstStride;\

  1088.         tmp9 = *s; s+=tmpStride;\

  1089.         OP2(*d, (tmp6+tmp7)*20 - (tmp5+tmp8)*5 + (tmp4+tmp9));d+=dstStride;\

  1090.         tmp10 = *s; s+=tmpStride;\

  1091.         OP2(*d, (tmp7+tmp8)*20 - (tmp6+tmp9)*5 + (tmp5+tmp10));d+=dstStride;\

  1092.        if (h>8) { \

  1093.         int tmp11,tmp12,tmp13,tmp14,tmp15,tmp16,tmp17,tmp18; \

  1094.         tmp11 = *s; s+=tmpStride;\

  1095.         OP2(*d , (tmp8 +tmp9 )*20 - (tmp7 +tmp10)*5 + (tmp6 +tmp11));d+=dstStride;\

  1096.         tmp12 = *s; s+=tmpStride;\

  1097.         OP2(*d , (tmp9 +tmp10)*20 - (tmp8 +tmp11)*5 + (tmp7 +tmp12));d+=dstStride;\

  1098.         tmp13 = *s; s+=tmpStride;\

  1099.         OP2(*d, (tmp10+tmp11)*20 - (tmp9 +tmp12)*5 + (tmp8 +tmp13));d+=dstStride;\

  1100.         tmp14 = *s; s+=tmpStride;\

  1101.         OP2(*d, (tmp11+tmp12)*20 - (tmp10+tmp13)*5 + (tmp9 +tmp14));d+=dstStride;\

  1102.         tmp15 = *s; s+=tmpStride;\

  1103.         OP2(*d, (tmp12+tmp13)*20 - (tmp11+tmp14)*5 + (tmp10+tmp15));d+=dstStride;\

  1104.         tmp16 = *s; s+=tmpStride;\

  1105.         OP2(*d, (tmp13+tmp14)*20 - (tmp12+tmp15)*5 + (tmp11+tmp16));d+=dstStride;\

  1106.         tmp17 = *s; s+=tmpStride;\

  1107.         OP2(*d, (tmp14+tmp15)*20 - (tmp13+tmp16)*5 + (tmp12+tmp17));d+=dstStride;\

  1108.         tmp18 = *s; s+=tmpStride;\

  1109.         OP2(*d, (tmp15+tmp16)*20 - (tmp14+tmp17)*5 + (tmp13+tmp18));d+=dstStride;\

  1110.        } \

  1111.       } \

  1112.         dst++;\

  1113.         tmp++;\

  1114.     }while(--i);\

  1115. }\

  1116. \

  1117. static void OPNAME ## h264_qpel4_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1118.     OPNAME ## h264_qpel_h_lowpass(dst,src,dstStride,srcStride,4,4); \

  1119. }\

  1120. static void OPNAME ## h264_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1121.    OPNAME ## h264_qpel_h_lowpass(dst,src,dstStride,srcStride,8,8); \

  1122. }\

  1123. static void OPNAME ## h264_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1124.    OPNAME ## h264_qpel_h_lowpass(dst,src,dstStride,srcStride,16,16); \

  1125. }\

  1126. \

  1127. static void OPNAME ## h264_qpel4_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1128.    OPNAME ## h264_qpel_v_lowpass(dst,src,dstStride,srcStride,4,4); \

  1129. }\

  1130. static void OPNAME ## h264_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1131.    OPNAME ## h264_qpel_v_lowpass(dst,src,dstStride,srcStride,8,8); \

  1132. }\

  1133. static void OPNAME ## h264_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\

  1134.    OPNAME ## h264_qpel_v_lowpass(dst,src,dstStride,srcStride,16,16); \

  1135. }\

  1136. static void OPNAME ## h264_qpel4_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\

  1137.    OPNAME ## h264_qpel_hv_lowpass(dst,tmp,src,dstStride,tmpStride,srcStride,4,4); \

  1138. }\

  1139. static void OPNAME ## h264_qpel8_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\

  1140.    OPNAME ## h264_qpel_hv_lowpass(dst,tmp,src,dstStride,tmpStride,srcStride,8,8); \

  1141. }\

  1142. static void OPNAME ## h264_qpel16_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\

  1143.    OPNAME ## h264_qpel_hv_lowpass(dst,tmp,src,dstStride,tmpStride,srcStride,16,16); \

  1144. }\

  1145. 

  1146. #define H264_MC(OPNAME, SIZE) \

  1147. static void OPNAME ## h264_qpel ## SIZE ## _mc00_sh4 (uint8_t *dst, uint8_t *src, int stride){\

  1148.     OPNAME ## pixels ## SIZE ## _c(dst, src, stride, SIZE);\

  1149. }\

  1150. \

  1151. static void OPNAME ## h264_qpel ## SIZE ## _mc10_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1152.     uint8_t half[SIZE*SIZE];\

  1153.     put_h264_qpel ## SIZE ## _h_lowpass(half, src, SIZE, stride);\

  1154.     OPNAME ## pixels ## SIZE ## _l2_aligned2(dst, src, half, stride, stride, SIZE, SIZE);\

  1155. }\

  1156. \

  1157. static void OPNAME ## h264_qpel ## SIZE ## _mc20_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1158.     OPNAME ## h264_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride);\

  1159. }\

  1160. \

  1161. static void OPNAME ## h264_qpel ## SIZE ## _mc30_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1162.     uint8_t half[SIZE*SIZE];\

  1163.     put_h264_qpel ## SIZE ## _h_lowpass(half, src, SIZE, stride);\

  1164.     OPNAME ## pixels ## SIZE ## _l2_aligned2(dst, src+1, half, stride, stride, SIZE, SIZE);\

  1165. }\

  1166. \

  1167. static void OPNAME ## h264_qpel ## SIZE ## _mc01_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1168.     uint8_t full[SIZE*(SIZE+5)];\

  1169.     uint8_t * const full_mid= full + SIZE*2;\

  1170.     uint8_t half[SIZE*SIZE];\

  1171.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1172.     put_h264_qpel ## SIZE ## _v_lowpass(half, full_mid, SIZE, SIZE);\

  1173.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, full_mid, half, stride, SIZE, SIZE, SIZE);\

  1174. }\

  1175. \

  1176. static void OPNAME ## h264_qpel ## SIZE ## _mc02_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1177.     uint8_t full[SIZE*(SIZE+5)];\

  1178.     uint8_t * const full_mid= full + SIZE*2;\

  1179.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1180.     OPNAME ## h264_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE);\

  1181. }\

  1182. \

  1183. static void OPNAME ## h264_qpel ## SIZE ## _mc03_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1184.     uint8_t full[SIZE*(SIZE+5)];\

  1185.     uint8_t * const full_mid= full + SIZE*2;\

  1186.     uint8_t half[SIZE*SIZE];\

  1187.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1188.     put_h264_qpel ## SIZE ## _v_lowpass(half, full_mid, SIZE, SIZE);\

  1189.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, full_mid+SIZE, half, stride, SIZE, SIZE, SIZE);\

  1190. }\

  1191. \

  1192. static void OPNAME ## h264_qpel ## SIZE ## _mc11_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1193.     uint8_t full[SIZE*(SIZE+5)];\

  1194.     uint8_t * const full_mid= full + SIZE*2;\

  1195.     uint8_t halfH[SIZE*SIZE];\

  1196.     uint8_t halfV[SIZE*SIZE];\

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

  1198.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1199.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1200.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\

  1201. }\

  1202. \

  1203. static void OPNAME ## h264_qpel ## SIZE ## _mc31_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1204.     uint8_t full[SIZE*(SIZE+5)];\

  1205.     uint8_t * const full_mid= full + SIZE*2;\

  1206.     uint8_t halfH[SIZE*SIZE];\

  1207.     uint8_t halfV[SIZE*SIZE];\

  1208.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\

  1209.     copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\

  1210.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

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

  1212. }\

  1213. \

  1214. static void OPNAME ## h264_qpel ## SIZE ## _mc13_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.     uint8_t halfH[SIZE*SIZE];\

  1218.     uint8_t halfV[SIZE*SIZE];\

  1219.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\

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

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

  1222.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\

  1223. }\

  1224. \

  1225. static void OPNAME ## h264_qpel ## SIZE ## _mc33_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1226.     uint8_t full[SIZE*(SIZE+5)];\

  1227.     uint8_t * const full_mid= full + SIZE*2;\

  1228.     uint8_t halfH[SIZE*SIZE];\

  1229.     uint8_t halfV[SIZE*SIZE];\

  1230.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\

  1231.     copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\

  1232.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1233.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\

  1234. }\

  1235. \

  1236. static void OPNAME ## h264_qpel ## SIZE ## _mc22_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1237.     int16_t tmp[SIZE*(SIZE+5)];\

  1238.     OPNAME ## h264_qpel ## SIZE ## _hv_lowpass(dst, tmp, src, stride, SIZE, stride);\

  1239. }\

  1240. \

  1241. static void OPNAME ## h264_qpel ## SIZE ## _mc21_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1242.     int16_t tmp[SIZE*(SIZE+5)];\

  1243.     uint8_t halfH[SIZE*SIZE];\

  1244.     uint8_t halfHV[SIZE*SIZE];\

  1245.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\

  1246.     put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\

  1247.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfHV, stride, SIZE, SIZE, SIZE);\

  1248. }\

  1249. \

  1250. static void OPNAME ## h264_qpel ## SIZE ## _mc23_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1251.     int16_t tmp[SIZE*(SIZE+5)];\

  1252.     uint8_t halfH[SIZE*SIZE];\

  1253.     uint8_t halfHV[SIZE*SIZE];\

  1254.     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\

  1255.     put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\

  1256.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfHV, stride, SIZE, SIZE, SIZE);\

  1257. }\

  1258. \

  1259. static void OPNAME ## h264_qpel ## SIZE ## _mc12_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1260.     uint8_t full[SIZE*(SIZE+5)];\

  1261.     uint8_t * const full_mid= full + SIZE*2;\

  1262.     int16_t tmp[SIZE*(SIZE+5)];\

  1263.     uint8_t halfV[SIZE*SIZE];\

  1264.     uint8_t halfHV[SIZE*SIZE];\

  1265.     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\

  1266.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1267.     put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\

  1268.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfV, halfHV, stride, SIZE, SIZE, SIZE);\

  1269. }\

  1270. \

  1271. static void OPNAME ## h264_qpel ## SIZE ## _mc32_sh4(uint8_t *dst, uint8_t *src, int stride){\

  1272.     uint8_t full[SIZE*(SIZE+5)];\

  1273.     uint8_t * const full_mid= full + SIZE*2;\

  1274.     int16_t tmp[SIZE*(SIZE+5)];\

  1275.     uint8_t halfV[SIZE*SIZE];\

  1276.     uint8_t halfHV[SIZE*SIZE];\

  1277.     copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\

  1278.     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\

  1279.     put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\

  1280.     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfV, halfHV, stride, SIZE, SIZE, SIZE);\

  1281. }\

  1282. 

  1283. #define op_avg(a, b)  a = (((a)+cm[((b) + 16)>>5]+1)>>1)

  1284. //#define op_avg2(a, b) a = (((a)*w1+cm[((b) + 16)>>5]*w2 + o + 64)>>7)

  1285. #define op_put(a, b)  a = cm[((b) + 16)>>5]

  1286. #define op2_avg(a, b)  a = (((a)+cm[((b) + 512)>>10]+1)>>1)

  1287. #define op2_put(a, b)  a = cm[((b) + 512)>>10]

  1288. 

  1289. H264_LOWPASS(put_       , op_put, op2_put)

  1290. H264_LOWPASS(avg_       , op_avg, op2_avg)

  1291. H264_MC(put_, 4)

  1292. H264_MC(put_, 8)

  1293. H264_MC(put_, 16)

  1294. H264_MC(avg_, 4)

  1295. H264_MC(avg_, 8)

  1296. H264_MC(avg_, 16)

  1297. 

  1298. #undef op_avg

  1299. #undef op_put

  1300. #undef op2_avg

  1301. #undef op2_put

  1302. 

  1303. static void wmv2_mspel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){

  1304.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;

  1305. 

  1306.     do{

  1307.         int src_1,src0,src1,src2,src3,src4,src5,src6,src7,src8,src9;

  1308.         uint8_t *s = src;

  1309.         src_1 = s[-1];

  1310.         src0 = *s++;

  1311.         src1 = *s++;

  1312.         src2 = *s++;

  1313.         dst[0]= cm[(9*(src0 + src1) - (src_1 + src2) + 8)>>4];

  1314.         src3 = *s++;

  1315.         dst[1]= cm[(9*(src1 + src2) - (src0 + src3) + 8)>>4];

  1316.         src4 = *s++;

  1317.         dst[2]= cm[(9*(src2 + src3) - (src1 + src4) + 8)>>4];

  1318.         src5 = *s++;

  1319.         dst[3]= cm[(9*(src3 + src4) - (src2 + src5) + 8)>>4];

  1320.         src6 = *s++;

  1321.         dst[4]= cm[(9*(src4 + src5) - (src3 + src6) + 8)>>4];

  1322.         src7 = *s++;

  1323.         dst[5]= cm[(9*(src5 + src6) - (src4 + src7) + 8)>>4];

  1324.         src8 = *s++;

  1325.         dst[6]= cm[(9*(src6 + src7) - (src5 + src8) + 8)>>4];

  1326.         src9 = *s++;

  1327.         dst[7]= cm[(9*(src7 + src8) - (src6 + src9) + 8)>>4];

  1328.         dst+=dstStride;

  1329.         src+=srcStride;

  1330.     }while(--h);

  1331. }

  1332. 

  1333. static void wmv2_mspel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int w){

  1334.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;

  1335. 

  1336.     do{

  1337.         int src_1,src0,src1,src2,src3,src4,src5,src6,src7,src8,src9;

  1338.         uint8_t *s = src,*d = dst;

  1339.         src_1 = *(s-srcStride);

  1340.         src0 = *s; s+=srcStride;

  1341.         src1 = *s; s+=srcStride;

  1342.         src2 = *s; s+=srcStride;

  1343.         *d= cm[(9*(src0 + src1) - (src_1 + src2) + 8)>>4]; d+=dstStride;

  1344.         src3 = *s; s+=srcStride;

  1345.         *d= cm[(9*(src1 + src2) - (src0  + src3) + 8)>>4]; d+=dstStride;

  1346.         src4 = *s; s+=srcStride;

  1347.         *d= cm[(9*(src2 + src3) - (src1  + src4) + 8)>>4]; d+=dstStride;

  1348.         src5 = *s; s+=srcStride;

  1349.         *d= cm[(9*(src3 + src4) - (src2  + src5) + 8)>>4]; d+=dstStride;

  1350.         src6 = *s; s+=srcStride;

  1351.         *d= cm[(9*(src4 + src5) - (src3  + src6) + 8)>>4]; d+=dstStride;

  1352.         src7 = *s; s+=srcStride;

  1353.         *d= cm[(9*(src5 + src6) - (src4  + src7) + 8)>>4]; d+=dstStride;

  1354.         src8 = *s; s+=srcStride;

  1355.         *d= cm[(9*(src6 + src7) - (src5  + src8) + 8)>>4]; d+=dstStride;

  1356.         src9 = *s;

  1357.         *d= cm[(9*(src7 + src8) - (src6  + src9) + 8)>>4]; d+=dstStride;

  1358.         src++;

  1359.         dst++;

  1360.     }while(--w);

  1361. }

  1362. 

  1363. static void put_mspel8_mc00_sh4 (uint8_t *dst, uint8_t *src, int stride){

  1364.     put_pixels8_c(dst, src, stride, 8);

  1365. }

  1366. 

  1367. static void put_mspel8_mc10_sh4(uint8_t *dst, uint8_t *src, int stride){

  1368.     uint8_t half[64];

  1369.     wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);

  1370.     put_pixels8_l2_aligned2(dst, src, half, stride, stride, 8, 8);

  1371. }

  1372. 

  1373. static void put_mspel8_mc20_sh4(uint8_t *dst, uint8_t *src, int stride){

  1374.     wmv2_mspel8_h_lowpass(dst, src, stride, stride, 8);

  1375. }

  1376. 

  1377. static void put_mspel8_mc30_sh4(uint8_t *dst, uint8_t *src, int stride){

  1378.     uint8_t half[64];

  1379.     wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);

  1380.     put_pixels8_l2_aligned2(dst, src+1, half, stride, stride, 8, 8);

  1381. }

  1382. 

  1383. static void put_mspel8_mc02_sh4(uint8_t *dst, uint8_t *src, int stride){

  1384.     wmv2_mspel8_v_lowpass(dst, src, stride, stride, 8);

  1385. }

  1386. 

  1387. static void put_mspel8_mc12_sh4(uint8_t *dst, uint8_t *src, int stride){

  1388.     uint8_t halfH[88];

  1389.     uint8_t halfV[64];

  1390.     uint8_t halfHV[64];

  1391.     wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);

  1392.     wmv2_mspel8_v_lowpass(halfV, src, 8, stride, 8);

  1393.     wmv2_mspel8_v_lowpass(halfHV, halfH+8, 8, 8, 8);

  1394.     put_pixels8_l2_aligned(dst, halfV, halfHV, stride, 8, 8, 8);

  1395. }

  1396. static void put_mspel8_mc32_sh4(uint8_t *dst, uint8_t *src, int stride){

  1397.     uint8_t halfH[88];

  1398.     uint8_t halfV[64];

  1399.     uint8_t halfHV[64];

  1400.     wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);

  1401.     wmv2_mspel8_v_lowpass(halfV, src+1, 8, stride, 8);

  1402.     wmv2_mspel8_v_lowpass(halfHV, halfH+8, 8, 8, 8);

  1403.     put_pixels8_l2_aligned(dst, halfV, halfHV, stride, 8, 8, 8);

  1404. }

  1405. static void put_mspel8_mc22_sh4(uint8_t *dst, uint8_t *src, int stride){

  1406.     uint8_t halfH[88];

  1407.     wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);

  1408.     wmv2_mspel8_v_lowpass(dst, halfH+8, stride, 8, 8);

  1409. }