FFmpeg: libavcodec/sh4/qpel.c Source File

00001 /*
00002  * This is optimized for sh, which have post increment addressing (*p++).
00003  * Some CPU may be index (p[n]) faster than post increment (*p++).
00004  *
00005  * copyright (c) 2001-2003 BERO <bero@geocities.co.jp>
00006  *
00007  * This file is part of FFmpeg.
00008  *
00009  * FFmpeg is free software; you can redistribute it and/or
00010  * modify it under the terms of the GNU Lesser General Public
00011  * License as published by the Free Software Foundation; either
00012  * version 2.1 of the License, or (at your option) any later version.
00013  *
00014  * FFmpeg is distributed in the hope that it will be useful,
00015  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00016  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00017  * Lesser General Public License for more details.
00018  *
00019  * You should have received a copy of the GNU Lesser General Public
00020  * License along with FFmpeg; if not, write to the Free Software
00021  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00022  */
00023 
00024 #define PIXOP2(OPNAME, OP) \
00025 \
00026 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) \
00027 {\
00028         do {\
00029                 OP(LP(dst  ),rnd_avg32(LPC(src1  ),LPC(src2  )) ); \
00030                 src1+=src_stride1; \
00031                 src2+=src_stride2; \
00032                 dst+=dst_stride; \
00033         } while(--h); \
00034 }\
00035 \
00036 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) \
00037 {\
00038         do {\
00039                 OP(LP(dst  ),rnd_avg32(AV_RN32(src1  ),LPC(src2  )) ); \
00040                 src1+=src_stride1; \
00041                 src2+=src_stride2; \
00042                 dst+=dst_stride; \
00043         } while(--h); \
00044 }\
00045 \
00046 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) \
00047 {\
00048         do {\
00049                 OP(LP(dst  ),no_rnd_avg32(AV_RN32(src1  ),LPC(src2  )) ); \
00050                 OP(LP(dst+4),no_rnd_avg32(AV_RN32(src1+4),LPC(src2+4)) ); \
00051                 OP(LP(dst+8),no_rnd_avg32(AV_RN32(src1+8),LPC(src2+8)) ); \
00052                 OP(LP(dst+12),no_rnd_avg32(AV_RN32(src1+12),LPC(src2+12)) ); \
00053                 src1+=src_stride1; \
00054                 src2+=src_stride2; \
00055                 dst+=dst_stride; \
00056         } while(--h); \
00057 }\
00058 \
00059 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) \
00060 {\
00061         do {\
00062                 OP(LP(dst  ),rnd_avg32(AV_RN32(src1  ),LPC(src2  )) ); \
00063                 OP(LP(dst+4),rnd_avg32(AV_RN32(src1+4),LPC(src2+4)) ); \
00064                 OP(LP(dst+8),rnd_avg32(AV_RN32(src1+8),LPC(src2+8)) ); \
00065                 OP(LP(dst+12),rnd_avg32(AV_RN32(src1+12),LPC(src2+12)) ); \
00066                 src1+=src_stride1; \
00067                 src2+=src_stride2; \
00068                 dst+=dst_stride; \
00069         } while(--h); \
00070 }\
00071 \
00072 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) \
00073 {\
00074         do { /* onlye src2 aligned */\
00075                 OP(LP(dst  ),no_rnd_avg32(AV_RN32(src1  ),LPC(src2  )) ); \
00076                 OP(LP(dst+4),no_rnd_avg32(AV_RN32(src1+4),LPC(src2+4)) ); \
00077                 src1+=src_stride1; \
00078                 src2+=src_stride2; \
00079                 dst+=dst_stride; \
00080         } while(--h); \
00081 }\
00082 \
00083 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) \
00084 {\
00085         do {\
00086                 OP(LP(dst  ),rnd_avg32(AV_RN32(src1  ),LPC(src2  )) ); \
00087                 OP(LP(dst+4),rnd_avg32(AV_RN32(src1+4),LPC(src2+4)) ); \
00088                 src1+=src_stride1; \
00089                 src2+=src_stride2; \
00090                 dst+=dst_stride; \
00091         } while(--h); \
00092 }\
00093 \
00094 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) \
00095 {\
00096         do {\
00097                 OP(LP(dst  ),no_rnd_avg32(LPC(src1  ),LPC(src2  )) ); \
00098                 OP(LP(dst+4),no_rnd_avg32(LPC(src1+4),LPC(src2+4)) ); \
00099                 src1+=src_stride1; \
00100                 src2+=src_stride2; \
00101                 dst+=dst_stride; \
00102         } while(--h); \
00103 }\
00104 \
00105 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) \
00106 {\
00107         do {\
00108                 OP(LP(dst  ),rnd_avg32(LPC(src1  ),LPC(src2  )) ); \
00109                 OP(LP(dst+4),rnd_avg32(LPC(src1+4),LPC(src2+4)) ); \
00110                 src1+=src_stride1; \
00111                 src2+=src_stride2; \
00112                 dst+=dst_stride; \
00113         } while(--h); \
00114 }\
00115 \
00116 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) \
00117 {\
00118         do {\
00119                 OP(LP(dst  ),no_rnd_avg32(LPC(src1  ),LPC(src2  )) ); \
00120                 OP(LP(dst+4),no_rnd_avg32(LPC(src1+4),LPC(src2+4)) ); \
00121                 OP(LP(dst+8),no_rnd_avg32(LPC(src1+8),LPC(src2+8)) ); \
00122                 OP(LP(dst+12),no_rnd_avg32(LPC(src1+12),LPC(src2+12)) ); \
00123                 src1+=src_stride1; \
00124                 src2+=src_stride2; \
00125                 dst+=dst_stride; \
00126         } while(--h); \
00127 }\
00128 \
00129 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) \
00130 {\
00131         do {\
00132                 OP(LP(dst  ),rnd_avg32(LPC(src1  ),LPC(src2  )) ); \
00133                 OP(LP(dst+4),rnd_avg32(LPC(src1+4),LPC(src2+4)) ); \
00134                 OP(LP(dst+8),rnd_avg32(LPC(src1+8),LPC(src2+8)) ); \
00135                 OP(LP(dst+12),rnd_avg32(LPC(src1+12),LPC(src2+12)) ); \
00136                 src1+=src_stride1; \
00137                 src2+=src_stride2; \
00138                 dst+=dst_stride; \
00139         } while(--h); \
00140 }\
00141 \
00142 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) \
00143 { OPNAME ## _no_rnd_pixels16_l2_aligned2(dst,src2,src1,dst_stride,src_stride2,src_stride1,h); } \
00144 \
00145 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) \
00146 { OPNAME ## _pixels16_l2_aligned2(dst,src2,src1,dst_stride,src_stride2,src_stride1,h); } \
00147 \
00148 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) \
00149 { OPNAME ## _no_rnd_pixels8_l2_aligned2(dst,src2,src1,dst_stride,src_stride2,src_stride1,h); } \
00150 \
00151 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) \
00152 { OPNAME ## _pixels8_l2_aligned2(dst,src2,src1,dst_stride,src_stride2,src_stride1,h); } \
00153 \
00154 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){\
00155         do { \
00156                 uint32_t a0,a1,a2,a3; \
00157                 UNPACK(a0,a1,LPC(src1),LPC(src2)); \
00158                 UNPACK(a2,a3,LPC(src3),LPC(src4)); \
00159                 OP(LP(dst),rnd_PACK(a0,a1,a2,a3)); \
00160                 UNPACK(a0,a1,LPC(src1+4),LPC(src2+4)); \
00161                 UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \
00162                 OP(LP(dst+4),rnd_PACK(a0,a1,a2,a3)); \
00163                 src1+=src_stride1;\
00164                 src2+=src_stride2;\
00165                 src3+=src_stride3;\
00166                 src4+=src_stride4;\
00167                 dst+=dst_stride;\
00168         } while(--h); \
00169 } \
00170 \
00171 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){\
00172         do { \
00173                 uint32_t a0,a1,a2,a3; \
00174                 UNPACK(a0,a1,LPC(src1),LPC(src2)); \
00175                 UNPACK(a2,a3,LPC(src3),LPC(src4)); \
00176                 OP(LP(dst),no_rnd_PACK(a0,a1,a2,a3)); \
00177                 UNPACK(a0,a1,LPC(src1+4),LPC(src2+4)); \
00178                 UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \
00179                 OP(LP(dst+4),no_rnd_PACK(a0,a1,a2,a3)); \
00180                 src1+=src_stride1;\
00181                 src2+=src_stride2;\
00182                 src3+=src_stride3;\
00183                 src4+=src_stride4;\
00184                 dst+=dst_stride;\
00185         } while(--h); \
00186 } \
00187 \
00188 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){\
00189         do { \
00190                 uint32_t a0,a1,a2,a3; /* src1 only not aligned */\
00191                 UNPACK(a0,a1,AV_RN32(src1),LPC(src2)); \
00192                 UNPACK(a2,a3,LPC(src3),LPC(src4)); \
00193                 OP(LP(dst),rnd_PACK(a0,a1,a2,a3)); \
00194                 UNPACK(a0,a1,AV_RN32(src1+4),LPC(src2+4)); \
00195                 UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \
00196                 OP(LP(dst+4),rnd_PACK(a0,a1,a2,a3)); \
00197                 src1+=src_stride1;\
00198                 src2+=src_stride2;\
00199                 src3+=src_stride3;\
00200                 src4+=src_stride4;\
00201                 dst+=dst_stride;\
00202         } while(--h); \
00203 } \
00204 \
00205 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){\
00206         do { \
00207                 uint32_t a0,a1,a2,a3; \
00208                 UNPACK(a0,a1,AV_RN32(src1),LPC(src2)); \
00209                 UNPACK(a2,a3,LPC(src3),LPC(src4)); \
00210                 OP(LP(dst),no_rnd_PACK(a0,a1,a2,a3)); \
00211                 UNPACK(a0,a1,AV_RN32(src1+4),LPC(src2+4)); \
00212                 UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \
00213                 OP(LP(dst+4),no_rnd_PACK(a0,a1,a2,a3)); \
00214                 src1+=src_stride1;\
00215                 src2+=src_stride2;\
00216                 src3+=src_stride3;\
00217                 src4+=src_stride4;\
00218                 dst+=dst_stride;\
00219         } while(--h); \
00220 } \
00221 \
00222 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){\
00223         do { \
00224                 uint32_t a0,a1,a2,a3; \
00225                 UNPACK(a0,a1,LPC(src1),LPC(src2)); \
00226                 UNPACK(a2,a3,LPC(src3),LPC(src4)); \
00227                 OP(LP(dst),rnd_PACK(a0,a1,a2,a3)); \
00228                 UNPACK(a0,a1,LPC(src1+4),LPC(src2+4)); \
00229                 UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \
00230                 OP(LP(dst+8),rnd_PACK(a0,a1,a2,a3)); \
00231                 UNPACK(a0,a1,LPC(src1+8),LPC(src2+8)); \
00232                 UNPACK(a2,a3,LPC(src3+8),LPC(src4+8)); \
00233                 OP(LP(dst+8),rnd_PACK(a0,a1,a2,a3)); \
00234                 UNPACK(a0,a1,LPC(src1+12),LPC(src2+12)); \
00235                 UNPACK(a2,a3,LPC(src3+12),LPC(src4+12)); \
00236                 OP(LP(dst+12),rnd_PACK(a0,a1,a2,a3)); \
00237                 src1+=src_stride1;\
00238                 src2+=src_stride2;\
00239                 src3+=src_stride3;\
00240                 src4+=src_stride4;\
00241                 dst+=dst_stride;\
00242         } while(--h); \
00243 } \
00244 \
00245 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){\
00246         do { \
00247                 uint32_t a0,a1,a2,a3; \
00248                 UNPACK(a0,a1,LPC(src1),LPC(src2)); \
00249                 UNPACK(a2,a3,LPC(src3),LPC(src4)); \
00250                 OP(LP(dst),no_rnd_PACK(a0,a1,a2,a3)); \
00251                 UNPACK(a0,a1,LPC(src1+4),LPC(src2+4)); \
00252                 UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \
00253                 OP(LP(dst+4),no_rnd_PACK(a0,a1,a2,a3)); \
00254                 UNPACK(a0,a1,LPC(src1+8),LPC(src2+8)); \
00255                 UNPACK(a2,a3,LPC(src3+8),LPC(src4+8)); \
00256                 OP(LP(dst+8),no_rnd_PACK(a0,a1,a2,a3)); \
00257                 UNPACK(a0,a1,LPC(src1+12),LPC(src2+12)); \
00258                 UNPACK(a2,a3,LPC(src3+12),LPC(src4+12)); \
00259                 OP(LP(dst+12),no_rnd_PACK(a0,a1,a2,a3)); \
00260                 src1+=src_stride1;\
00261                 src2+=src_stride2;\
00262                 src3+=src_stride3;\
00263                 src4+=src_stride4;\
00264                 dst+=dst_stride;\
00265         } while(--h); \
00266 } \
00267 \
00268 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){\
00269         do { /* src1 is unaligned */\
00270                 uint32_t a0,a1,a2,a3; \
00271                 UNPACK(a0,a1,AV_RN32(src1),LPC(src2)); \
00272                 UNPACK(a2,a3,LPC(src3),LPC(src4)); \
00273                 OP(LP(dst),rnd_PACK(a0,a1,a2,a3)); \
00274                 UNPACK(a0,a1,AV_RN32(src1+4),LPC(src2+4)); \
00275                 UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \
00276                 OP(LP(dst+8),rnd_PACK(a0,a1,a2,a3)); \
00277                 UNPACK(a0,a1,AV_RN32(src1+8),LPC(src2+8)); \
00278                 UNPACK(a2,a3,LPC(src3+8),LPC(src4+8)); \
00279                 OP(LP(dst+8),rnd_PACK(a0,a1,a2,a3)); \
00280                 UNPACK(a0,a1,AV_RN32(src1+12),LPC(src2+12)); \
00281                 UNPACK(a2,a3,LPC(src3+12),LPC(src4+12)); \
00282                 OP(LP(dst+12),rnd_PACK(a0,a1,a2,a3)); \
00283                 src1+=src_stride1;\
00284                 src2+=src_stride2;\
00285                 src3+=src_stride3;\
00286                 src4+=src_stride4;\
00287                 dst+=dst_stride;\
00288         } while(--h); \
00289 } \
00290 \
00291 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){\
00292         do { \
00293                 uint32_t a0,a1,a2,a3; \
00294                 UNPACK(a0,a1,AV_RN32(src1),LPC(src2)); \
00295                 UNPACK(a2,a3,LPC(src3),LPC(src4)); \
00296                 OP(LP(dst),no_rnd_PACK(a0,a1,a2,a3)); \
00297                 UNPACK(a0,a1,AV_RN32(src1+4),LPC(src2+4)); \
00298                 UNPACK(a2,a3,LPC(src3+4),LPC(src4+4)); \
00299                 OP(LP(dst+4),no_rnd_PACK(a0,a1,a2,a3)); \
00300                 UNPACK(a0,a1,AV_RN32(src1+8),LPC(src2+8)); \
00301                 UNPACK(a2,a3,LPC(src3+8),LPC(src4+8)); \
00302                 OP(LP(dst+8),no_rnd_PACK(a0,a1,a2,a3)); \
00303                 UNPACK(a0,a1,AV_RN32(src1+12),LPC(src2+12)); \
00304                 UNPACK(a2,a3,LPC(src3+12),LPC(src4+12)); \
00305                 OP(LP(dst+12),no_rnd_PACK(a0,a1,a2,a3)); \
00306                 src1+=src_stride1;\
00307                 src2+=src_stride2;\
00308                 src3+=src_stride3;\
00309                 src4+=src_stride4;\
00310                 dst+=dst_stride;\
00311         } while(--h); \
00312 } \
00313 \
00314 
00315 #define op_avg(a, b) a = rnd_avg32(a,b)
00316 #define op_put(a, b) a = b
00317 
00318 PIXOP2(avg, op_avg)
00319 PIXOP2(put, op_put)
00320 #undef op_avg
00321 #undef op_put
00322 
00323 #define avg2(a,b) ((a+b+1)>>1)
00324 #define avg4(a,b,c,d) ((a+b+c+d+2)>>2)
00325 
00326 
00327 static void gmc1_c(uint8_t *dst, uint8_t *src, int stride, int h, int x16, int y16, int rounder)
00328 {
00329     const int A=(16-x16)*(16-y16);
00330     const int B=(   x16)*(16-y16);
00331     const int C=(16-x16)*(   y16);
00332     const int D=(   x16)*(   y16);
00333 
00334     do {
00335         int t0,t1,t2,t3;
00336         uint8_t *s0 = src;
00337         uint8_t *s1 = src+stride;
00338         t0 = *s0++; t2 = *s1++;
00339         t1 = *s0++; t3 = *s1++;
00340         dst[0]= (A*t0 + B*t1 + C*t2 + D*t3 + rounder)>>8;
00341         t0 = *s0++; t2 = *s1++;
00342         dst[1]= (A*t1 + B*t0 + C*t3 + D*t2 + rounder)>>8;
00343         t1 = *s0++; t3 = *s1++;
00344         dst[2]= (A*t0 + B*t1 + C*t2 + D*t3 + rounder)>>8;
00345         t0 = *s0++; t2 = *s1++;
00346         dst[3]= (A*t1 + B*t0 + C*t3 + D*t2 + rounder)>>8;
00347         t1 = *s0++; t3 = *s1++;
00348         dst[4]= (A*t0 + B*t1 + C*t2 + D*t3 + rounder)>>8;
00349         t0 = *s0++; t2 = *s1++;
00350         dst[5]= (A*t1 + B*t0 + C*t3 + D*t2 + rounder)>>8;
00351         t1 = *s0++; t3 = *s1++;
00352         dst[6]= (A*t0 + B*t1 + C*t2 + D*t3 + rounder)>>8;
00353         t0 = *s0++; t2 = *s1++;
00354         dst[7]= (A*t1 + B*t0 + C*t3 + D*t2 + rounder)>>8;
00355         dst+= stride;
00356         src+= stride;
00357     }while(--h);
00358 }
00359 
00360 static void gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,
00361                   int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height)
00362 {
00363     int y, vx, vy;
00364     const int s= 1<<shift;
00365 
00366     width--;
00367     height--;
00368 
00369     for(y=0; y<h; y++){
00370         int x;
00371 
00372         vx= ox;
00373         vy= oy;
00374         for(x=0; x<8; x++){ //XXX FIXME optimize
00375             int src_x, src_y, frac_x, frac_y, index;
00376 
00377             src_x= vx>>16;
00378             src_y= vy>>16;
00379             frac_x= src_x&(s-1);
00380             frac_y= src_y&(s-1);
00381             src_x>>=shift;
00382             src_y>>=shift;
00383 
00384             if((unsigned)src_x < width){
00385                 if((unsigned)src_y < height){
00386                     index= src_x + src_y*stride;
00387                     dst[y*stride + x]= (  (  src[index         ]*(s-frac_x)
00388                                            + src[index       +1]*   frac_x )*(s-frac_y)
00389                                         + (  src[index+stride  ]*(s-frac_x)
00390                                            + src[index+stride+1]*   frac_x )*   frac_y
00391                                         + r)>>(shift*2);
00392                 }else{
00393                     index= src_x + av_clip(src_y, 0, height)*stride;
00394                     dst[y*stride + x]= ( (  src[index         ]*(s-frac_x)
00395                                           + src[index       +1]*   frac_x )*s
00396                                         + r)>>(shift*2);
00397                 }
00398             }else{
00399                 if((unsigned)src_y < height){
00400                     index= av_clip(src_x, 0, width) + src_y*stride;
00401                     dst[y*stride + x]= (  (  src[index         ]*(s-frac_y)
00402                                            + src[index+stride  ]*   frac_y )*s
00403                                         + r)>>(shift*2);
00404                 }else{
00405                     index= av_clip(src_x, 0, width) + av_clip(src_y, 0, height)*stride;
00406                     dst[y*stride + x]=    src[index         ];
00407                 }
00408             }
00409 
00410             vx+= dxx;
00411             vy+= dyx;
00412         }
00413         ox += dxy;
00414         oy += dyy;
00415     }
00416 }
00417 #define H264_CHROMA_MC(OPNAME, OP)\
00418 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){\
00419     const int A=(8-x)*(8-y);\
00420     const int B=(  x)*(8-y);\
00421     const int C=(8-x)*(  y);\
00422     const int D=(  x)*(  y);\
00423     \
00424     assert(x<8 && y<8 && x>=0 && y>=0);\
00425 \
00426     do {\
00427         int t0,t1,t2,t3; \
00428         uint8_t *s0 = src; \
00429         uint8_t *s1 = src+stride; \
00430         t0 = *s0++; t2 = *s1++; \
00431         t1 = *s0++; t3 = *s1++; \
00432         OP(dst[0], (A*t0 + B*t1 + C*t2 + D*t3));\
00433         t0 = *s0++; t2 = *s1++; \
00434         OP(dst[1], (A*t1 + B*t0 + C*t3 + D*t2));\
00435         dst+= stride;\
00436         src+= stride;\
00437     }while(--h);\
00438 }\
00439 \
00440 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){\
00441     const int A=(8-x)*(8-y);\
00442     const int B=(  x)*(8-y);\
00443     const int C=(8-x)*(  y);\
00444     const int D=(  x)*(  y);\
00445     \
00446     assert(x<8 && y<8 && x>=0 && y>=0);\
00447 \
00448     do {\
00449         int t0,t1,t2,t3; \
00450         uint8_t *s0 = src; \
00451         uint8_t *s1 = src+stride; \
00452         t0 = *s0++; t2 = *s1++; \
00453         t1 = *s0++; t3 = *s1++; \
00454         OP(dst[0], (A*t0 + B*t1 + C*t2 + D*t3));\
00455         t0 = *s0++; t2 = *s1++; \
00456         OP(dst[1], (A*t1 + B*t0 + C*t3 + D*t2));\
00457         t1 = *s0++; t3 = *s1++; \
00458         OP(dst[2], (A*t0 + B*t1 + C*t2 + D*t3));\
00459         t0 = *s0++; t2 = *s1++; \
00460         OP(dst[3], (A*t1 + B*t0 + C*t3 + D*t2));\
00461         dst+= stride;\
00462         src+= stride;\
00463     }while(--h);\
00464 }\
00465 \
00466 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){\
00467     const int A=(8-x)*(8-y);\
00468     const int B=(  x)*(8-y);\
00469     const int C=(8-x)*(  y);\
00470     const int D=(  x)*(  y);\
00471     \
00472     assert(x<8 && y<8 && x>=0 && y>=0);\
00473 \
00474     do {\
00475         int t0,t1,t2,t3; \
00476         uint8_t *s0 = src; \
00477         uint8_t *s1 = src+stride; \
00478         t0 = *s0++; t2 = *s1++; \
00479         t1 = *s0++; t3 = *s1++; \
00480         OP(dst[0], (A*t0 + B*t1 + C*t2 + D*t3));\
00481         t0 = *s0++; t2 = *s1++; \
00482         OP(dst[1], (A*t1 + B*t0 + C*t3 + D*t2));\
00483         t1 = *s0++; t3 = *s1++; \
00484         OP(dst[2], (A*t0 + B*t1 + C*t2 + D*t3));\
00485         t0 = *s0++; t2 = *s1++; \
00486         OP(dst[3], (A*t1 + B*t0 + C*t3 + D*t2));\
00487         t1 = *s0++; t3 = *s1++; \
00488         OP(dst[4], (A*t0 + B*t1 + C*t2 + D*t3));\
00489         t0 = *s0++; t2 = *s1++; \
00490         OP(dst[5], (A*t1 + B*t0 + C*t3 + D*t2));\
00491         t1 = *s0++; t3 = *s1++; \
00492         OP(dst[6], (A*t0 + B*t1 + C*t2 + D*t3));\
00493         t0 = *s0++; t2 = *s1++; \
00494         OP(dst[7], (A*t1 + B*t0 + C*t3 + D*t2));\
00495         dst+= stride;\
00496         src+= stride;\
00497     }while(--h);\
00498 }
00499 
00500 #define op_avg(a, b) a = (((a)+(((b) + 32)>>6)+1)>>1)
00501 #define op_put(a, b) a = (((b) + 32)>>6)
00502 
00503 H264_CHROMA_MC(put_       , op_put)
00504 H264_CHROMA_MC(avg_       , op_avg)
00505 #undef op_avg
00506 #undef op_put
00507 
00508 #define QPEL_MC(r, OPNAME, RND, OP) \
00509 static void OPNAME ## mpeg4_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
00510     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
00511     do {\
00512         uint8_t *s = src; \
00513         int src0,src1,src2,src3,src4,src5,src6,src7,src8;\
00514         src0= *s++;\
00515         src1= *s++;\
00516         src2= *s++;\
00517         src3= *s++;\
00518         src4= *s++;\
00519         OP(dst[0], (src0+src1)*20 - (src0+src2)*6 + (src1+src3)*3 - (src2+src4));\
00520         src5= *s++;\
00521         OP(dst[1], (src1+src2)*20 - (src0+src3)*6 + (src0+src4)*3 - (src1+src5));\
00522         src6= *s++;\
00523         OP(dst[2], (src2+src3)*20 - (src1+src4)*6 + (src0+src5)*3 - (src0+src6));\
00524         src7= *s++;\
00525         OP(dst[3], (src3+src4)*20 - (src2+src5)*6 + (src1+src6)*3 - (src0+src7));\
00526         src8= *s++;\
00527         OP(dst[4], (src4+src5)*20 - (src3+src6)*6 + (src2+src7)*3 - (src1+src8));\
00528         OP(dst[5], (src5+src6)*20 - (src4+src7)*6 + (src3+src8)*3 - (src2+src8));\
00529         OP(dst[6], (src6+src7)*20 - (src5+src8)*6 + (src4+src8)*3 - (src3+src7));\
00530         OP(dst[7], (src7+src8)*20 - (src6+src8)*6 + (src5+src7)*3 - (src4+src6));\
00531         dst+=dstStride;\
00532         src+=srcStride;\
00533     }while(--h);\
00534 }\
00535 \
00536 static void OPNAME ## mpeg4_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
00537     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
00538     int w=8;\
00539     do{\
00540         uint8_t *s = src, *d=dst;\
00541         int src0,src1,src2,src3,src4,src5,src6,src7,src8;\
00542         src0 = *s; s+=srcStride; \
00543         src1 = *s; s+=srcStride; \
00544         src2 = *s; s+=srcStride; \
00545         src3 = *s; s+=srcStride; \
00546         src4 = *s; s+=srcStride; \
00547         OP(*d, (src0+src1)*20 - (src0+src2)*6 + (src1+src3)*3 - (src2+src4));d+=dstStride;\
00548         src5 = *s; s+=srcStride; \
00549         OP(*d, (src1+src2)*20 - (src0+src3)*6 + (src0+src4)*3 - (src1+src5));d+=dstStride;\
00550         src6 = *s; s+=srcStride; \
00551         OP(*d, (src2+src3)*20 - (src1+src4)*6 + (src0+src5)*3 - (src0+src6));d+=dstStride;\
00552         src7 = *s; s+=srcStride; \
00553         OP(*d, (src3+src4)*20 - (src2+src5)*6 + (src1+src6)*3 - (src0+src7));d+=dstStride;\
00554         src8 = *s; \
00555         OP(*d, (src4+src5)*20 - (src3+src6)*6 + (src2+src7)*3 - (src1+src8));d+=dstStride;\
00556         OP(*d, (src5+src6)*20 - (src4+src7)*6 + (src3+src8)*3 - (src2+src8));d+=dstStride;\
00557         OP(*d, (src6+src7)*20 - (src5+src8)*6 + (src4+src8)*3 - (src3+src7));d+=dstStride;\
00558         OP(*d, (src7+src8)*20 - (src6+src8)*6 + (src5+src7)*3 - (src4+src6));\
00559         dst++;\
00560         src++;\
00561     }while(--w);\
00562 }\
00563 \
00564 static void OPNAME ## mpeg4_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
00565     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
00566     do {\
00567         uint8_t *s = src;\
00568         int src0,src1,src2,src3,src4,src5,src6,src7,src8;\
00569         int src9,src10,src11,src12,src13,src14,src15,src16;\
00570         src0= *s++;\
00571         src1= *s++;\
00572         src2= *s++;\
00573         src3= *s++;\
00574         src4= *s++;\
00575         OP(dst[ 0], (src0 +src1 )*20 - (src0 +src2 )*6 + (src1 +src3 )*3 - (src2 +src4 ));\
00576         src5= *s++;\
00577         OP(dst[ 1], (src1 +src2 )*20 - (src0 +src3 )*6 + (src0 +src4 )*3 - (src1 +src5 ));\
00578         src6= *s++;\
00579         OP(dst[ 2], (src2 +src3 )*20 - (src1 +src4 )*6 + (src0 +src5 )*3 - (src0 +src6 ));\
00580         src7= *s++;\
00581         OP(dst[ 3], (src3 +src4 )*20 - (src2 +src5 )*6 + (src1 +src6 )*3 - (src0 +src7 ));\
00582         src8= *s++;\
00583         OP(dst[ 4], (src4 +src5 )*20 - (src3 +src6 )*6 + (src2 +src7 )*3 - (src1 +src8 ));\
00584         src9= *s++;\
00585         OP(dst[ 5], (src5 +src6 )*20 - (src4 +src7 )*6 + (src3 +src8 )*3 - (src2 +src9 ));\
00586         src10= *s++;\
00587         OP(dst[ 6], (src6 +src7 )*20 - (src5 +src8 )*6 + (src4 +src9 )*3 - (src3 +src10));\
00588         src11= *s++;\
00589         OP(dst[ 7], (src7 +src8 )*20 - (src6 +src9 )*6 + (src5 +src10)*3 - (src4 +src11));\
00590         src12= *s++;\
00591         OP(dst[ 8], (src8 +src9 )*20 - (src7 +src10)*6 + (src6 +src11)*3 - (src5 +src12));\
00592         src13= *s++;\
00593         OP(dst[ 9], (src9 +src10)*20 - (src8 +src11)*6 + (src7 +src12)*3 - (src6 +src13));\
00594         src14= *s++;\
00595         OP(dst[10], (src10+src11)*20 - (src9 +src12)*6 + (src8 +src13)*3 - (src7 +src14));\
00596         src15= *s++;\
00597         OP(dst[11], (src11+src12)*20 - (src10+src13)*6 + (src9 +src14)*3 - (src8 +src15));\
00598         src16= *s++;\
00599         OP(dst[12], (src12+src13)*20 - (src11+src14)*6 + (src10+src15)*3 - (src9 +src16));\
00600         OP(dst[13], (src13+src14)*20 - (src12+src15)*6 + (src11+src16)*3 - (src10+src16));\
00601         OP(dst[14], (src14+src15)*20 - (src13+src16)*6 + (src12+src16)*3 - (src11+src15));\
00602         OP(dst[15], (src15+src16)*20 - (src14+src16)*6 + (src13+src15)*3 - (src12+src14));\
00603         dst+=dstStride;\
00604         src+=srcStride;\
00605     }while(--h);\
00606 }\
00607 \
00608 static void OPNAME ## mpeg4_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
00609     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
00610     int w=16;\
00611     do {\
00612         uint8_t *s = src, *d=dst;\
00613         int src0,src1,src2,src3,src4,src5,src6,src7,src8;\
00614         int src9,src10,src11,src12,src13,src14,src15,src16;\
00615         src0 = *s; s+=srcStride; \
00616         src1 = *s; s+=srcStride; \
00617         src2 = *s; s+=srcStride; \
00618         src3 = *s; s+=srcStride; \
00619         src4 = *s; s+=srcStride; \
00620         OP(*d, (src0 +src1 )*20 - (src0 +src2 )*6 + (src1 +src3 )*3 - (src2 +src4 ));d+=dstStride;\
00621         src5 = *s; s+=srcStride; \
00622         OP(*d, (src1 +src2 )*20 - (src0 +src3 )*6 + (src0 +src4 )*3 - (src1 +src5 ));d+=dstStride;\
00623         src6 = *s; s+=srcStride; \
00624         OP(*d, (src2 +src3 )*20 - (src1 +src4 )*6 + (src0 +src5 )*3 - (src0 +src6 ));d+=dstStride;\
00625         src7 = *s; s+=srcStride; \
00626         OP(*d, (src3 +src4 )*20 - (src2 +src5 )*6 + (src1 +src6 )*3 - (src0 +src7 ));d+=dstStride;\
00627         src8 = *s; s+=srcStride; \
00628         OP(*d, (src4 +src5 )*20 - (src3 +src6 )*6 + (src2 +src7 )*3 - (src1 +src8 ));d+=dstStride;\
00629         src9 = *s; s+=srcStride; \
00630         OP(*d, (src5 +src6 )*20 - (src4 +src7 )*6 + (src3 +src8 )*3 - (src2 +src9 ));d+=dstStride;\
00631         src10 = *s; s+=srcStride; \
00632         OP(*d, (src6 +src7 )*20 - (src5 +src8 )*6 + (src4 +src9 )*3 - (src3 +src10));d+=dstStride;\
00633         src11 = *s; s+=srcStride; \
00634         OP(*d, (src7 +src8 )*20 - (src6 +src9 )*6 + (src5 +src10)*3 - (src4 +src11));d+=dstStride;\
00635         src12 = *s; s+=srcStride; \
00636         OP(*d, (src8 +src9 )*20 - (src7 +src10)*6 + (src6 +src11)*3 - (src5 +src12));d+=dstStride;\
00637         src13 = *s; s+=srcStride; \
00638         OP(*d, (src9 +src10)*20 - (src8 +src11)*6 + (src7 +src12)*3 - (src6 +src13));d+=dstStride;\
00639         src14 = *s; s+=srcStride; \
00640         OP(*d, (src10+src11)*20 - (src9 +src12)*6 + (src8 +src13)*3 - (src7 +src14));d+=dstStride;\
00641         src15 = *s; s+=srcStride; \
00642         OP(*d, (src11+src12)*20 - (src10+src13)*6 + (src9 +src14)*3 - (src8 +src15));d+=dstStride;\
00643         src16 = *s; \
00644         OP(*d, (src12+src13)*20 - (src11+src14)*6 + (src10+src15)*3 - (src9 +src16));d+=dstStride;\
00645         OP(*d, (src13+src14)*20 - (src12+src15)*6 + (src11+src16)*3 - (src10+src16));d+=dstStride;\
00646         OP(*d, (src14+src15)*20 - (src13+src16)*6 + (src12+src16)*3 - (src11+src15));d+=dstStride;\
00647         OP(*d, (src15+src16)*20 - (src14+src16)*6 + (src13+src15)*3 - (src12+src14));\
00648         dst++;\
00649         src++;\
00650     }while(--w);\
00651 }\
00652 \
00653 static void OPNAME ## qpel8_mc00_sh4 (uint8_t *dst, uint8_t *src, int stride){\
00654     OPNAME ## pixels8_c(dst, src, stride, 8);\
00655 }\
00656 \
00657 static void OPNAME ## qpel8_mc10_sh4(uint8_t *dst, uint8_t *src, int stride){\
00658     uint8_t half[64];\
00659     put ## RND ## mpeg4_qpel8_h_lowpass(half, src, 8, stride, 8);\
00660     OPNAME ## pixels8_l2_aligned2(dst, src, half, stride, stride, 8, 8);\
00661 }\
00662 \
00663 static void OPNAME ## qpel8_mc20_sh4(uint8_t *dst, uint8_t *src, int stride){\
00664     OPNAME ## mpeg4_qpel8_h_lowpass(dst, src, stride, stride, 8);\
00665 }\
00666 \
00667 static void OPNAME ## qpel8_mc30_sh4(uint8_t *dst, uint8_t *src, int stride){\
00668     uint8_t half[64];\
00669     put ## RND ## mpeg4_qpel8_h_lowpass(half, src, 8, stride, 8);\
00670     OPNAME ## pixels8_l2_aligned2(dst, src+1, half, stride, stride, 8, 8);\
00671 }\
00672 \
00673 static void OPNAME ## qpel8_mc01_sh4(uint8_t *dst, uint8_t *src, int stride){\
00674     uint8_t full[16*9];\
00675     uint8_t half[64];\
00676     copy_block9(full, src, 16, stride, 9);\
00677     put ## RND ## mpeg4_qpel8_v_lowpass(half, full, 8, 16);\
00678     OPNAME ## pixels8_l2_aligned(dst, full, half, stride, 16, 8, 8);\
00679 }\
00680 \
00681 static void OPNAME ## qpel8_mc02_sh4(uint8_t *dst, uint8_t *src, int stride){\
00682     uint8_t full[16*9];\
00683     copy_block9(full, src, 16, stride, 9);\
00684     OPNAME ## mpeg4_qpel8_v_lowpass(dst, full, stride, 16);\
00685 }\
00686 \
00687 static void OPNAME ## qpel8_mc03_sh4(uint8_t *dst, uint8_t *src, int stride){\
00688     uint8_t full[16*9];\
00689     uint8_t half[64];\
00690     copy_block9(full, src, 16, stride, 9);\
00691     put ## RND ## mpeg4_qpel8_v_lowpass(half, full, 8, 16);\
00692     OPNAME ## pixels8_l2_aligned(dst, full+16, half, stride, 16, 8, 8);\
00693 }\
00694 static void OPNAME ## qpel8_mc11_sh4(uint8_t *dst, uint8_t *src, int stride){\
00695     uint8_t full[16*9];\
00696     uint8_t halfH[72];\
00697     uint8_t halfHV[64];\
00698     copy_block9(full, src, 16, stride, 9);\
00699     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
00700     put ## RND ## pixels8_l2_aligned(halfH, halfH, full, 8, 8, 16, 9);\
00701     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
00702     OPNAME ## pixels8_l2_aligned(dst, halfH, halfHV, stride, 8, 8, 8);\
00703 }\
00704 static void OPNAME ## qpel8_mc31_sh4(uint8_t *dst, uint8_t *src, int stride){\
00705     uint8_t full[16*9];\
00706     uint8_t halfH[72];\
00707     uint8_t halfHV[64];\
00708     copy_block9(full, src, 16, stride, 9);\
00709     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
00710     put ## RND ## pixels8_l2_aligned1(halfH, halfH, full+1, 8, 8, 16, 9);\
00711     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
00712     OPNAME ## pixels8_l2_aligned(dst, halfH, halfHV, stride, 8, 8, 8);\
00713 }\
00714 static void OPNAME ## qpel8_mc13_sh4(uint8_t *dst, uint8_t *src, int stride){\
00715     uint8_t full[16*9];\
00716     uint8_t halfH[72];\
00717     uint8_t halfHV[64];\
00718     copy_block9(full, src, 16, stride, 9);\
00719     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
00720     put ## RND ## pixels8_l2_aligned(halfH, halfH, full, 8, 8, 16, 9);\
00721     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
00722     OPNAME ## pixels8_l2_aligned(dst, halfH+8, halfHV, stride, 8, 8, 8);\
00723 }\
00724 static void OPNAME ## qpel8_mc33_sh4(uint8_t *dst, uint8_t *src, int stride){\
00725     uint8_t full[16*9];\
00726     uint8_t halfH[72];\
00727     uint8_t halfHV[64];\
00728     copy_block9(full, src, 16, stride, 9);\
00729     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
00730     put ## RND ## pixels8_l2_aligned1(halfH, halfH, full+1, 8, 8, 16, 9);\
00731     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
00732     OPNAME ## pixels8_l2_aligned(dst, halfH+8, halfHV, stride, 8, 8, 8);\
00733 }\
00734 static void OPNAME ## qpel8_mc21_sh4(uint8_t *dst, uint8_t *src, int stride){\
00735     uint8_t halfH[72];\
00736     uint8_t halfHV[64];\
00737     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\
00738     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
00739     OPNAME ## pixels8_l2_aligned(dst, halfH, halfHV, stride, 8, 8, 8);\
00740 }\
00741 static void OPNAME ## qpel8_mc23_sh4(uint8_t *dst, uint8_t *src, int stride){\
00742     uint8_t halfH[72];\
00743     uint8_t halfHV[64];\
00744     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\
00745     put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
00746     OPNAME ## pixels8_l2_aligned(dst, halfH+8, halfHV, stride, 8, 8, 8);\
00747 }\
00748 static void OPNAME ## qpel8_mc12_sh4(uint8_t *dst, uint8_t *src, int stride){\
00749     uint8_t full[16*9];\
00750     uint8_t halfH[72];\
00751     copy_block9(full, src, 16, stride, 9);\
00752     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
00753     put ## RND ## pixels8_l2_aligned(halfH, halfH, full, 8, 8, 16, 9);\
00754     OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\
00755 }\
00756 static void OPNAME ## qpel8_mc32_sh4(uint8_t *dst, uint8_t *src, int stride){\
00757     uint8_t full[16*9];\
00758     uint8_t halfH[72];\
00759     copy_block9(full, src, 16, stride, 9);\
00760     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
00761     put ## RND ## pixels8_l2_aligned1(halfH, halfH, full+1, 8, 8, 16, 9);\
00762     OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\
00763 }\
00764 static void OPNAME ## qpel8_mc22_sh4(uint8_t *dst, uint8_t *src, int stride){\
00765     uint8_t halfH[72];\
00766     put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\
00767     OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\
00768 }\
00769 static void OPNAME ## qpel16_mc00_sh4 (uint8_t *dst, uint8_t *src, int stride){\
00770     OPNAME ## pixels16_c(dst, src, stride, 16);\
00771 }\
00772 \
00773 static void OPNAME ## qpel16_mc10_sh4(uint8_t *dst, uint8_t *src, int stride){\
00774     uint8_t half[256];\
00775     put ## RND ## mpeg4_qpel16_h_lowpass(half, src, 16, stride, 16);\
00776     OPNAME ## pixels16_l2_aligned2(dst, src, half, stride, stride, 16, 16);\
00777 }\
00778 \
00779 static void OPNAME ## qpel16_mc20_sh4(uint8_t *dst, uint8_t *src, int stride){\
00780     OPNAME ## mpeg4_qpel16_h_lowpass(dst, src, stride, stride, 16);\
00781 }\
00782 \
00783 static void OPNAME ## qpel16_mc30_sh4(uint8_t *dst, uint8_t *src, int stride){\
00784     uint8_t half[256];\
00785     put ## RND ## mpeg4_qpel16_h_lowpass(half, src, 16, stride, 16);\
00786     OPNAME ## pixels16_l2_aligned2(dst, src+1, half, stride, stride, 16, 16);\
00787 }\
00788 \
00789 static void OPNAME ## qpel16_mc01_sh4(uint8_t *dst, uint8_t *src, int stride){\
00790     uint8_t full[24*17];\
00791     uint8_t half[256];\
00792     copy_block17(full, src, 24, stride, 17);\
00793     put ## RND ## mpeg4_qpel16_v_lowpass(half, full, 16, 24);\
00794     OPNAME ## pixels16_l2_aligned(dst, full, half, stride, 24, 16, 16);\
00795 }\
00796 \
00797 static void OPNAME ## qpel16_mc02_sh4(uint8_t *dst, uint8_t *src, int stride){\
00798     uint8_t full[24*17];\
00799     copy_block17(full, src, 24, stride, 17);\
00800     OPNAME ## mpeg4_qpel16_v_lowpass(dst, full, stride, 24);\
00801 }\
00802 \
00803 static void OPNAME ## qpel16_mc03_sh4(uint8_t *dst, uint8_t *src, int stride){\
00804     uint8_t full[24*17];\
00805     uint8_t half[256];\
00806     copy_block17(full, src, 24, stride, 17);\
00807     put ## RND ## mpeg4_qpel16_v_lowpass(half, full, 16, 24);\
00808     OPNAME ## pixels16_l2_aligned(dst, full+24, half, stride, 24, 16, 16);\
00809 }\
00810 static void OPNAME ## qpel16_mc11_sh4(uint8_t *dst, uint8_t *src, int stride){\
00811     uint8_t full[24*17];\
00812     uint8_t halfH[272];\
00813     uint8_t halfHV[256];\
00814     copy_block17(full, src, 24, stride, 17);\
00815     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
00816     put ## RND ## pixels16_l2_aligned(halfH, halfH, full, 16, 16, 24, 17);\
00817     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
00818     OPNAME ## pixels16_l2_aligned(dst, halfH, halfHV, stride, 16, 16, 16);\
00819 }\
00820 static void OPNAME ## qpel16_mc31_sh4(uint8_t *dst, uint8_t *src, int stride){\
00821     uint8_t full[24*17];\
00822     uint8_t halfH[272];\
00823     uint8_t halfHV[256];\
00824     copy_block17(full, src, 24, stride, 17);\
00825     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
00826     put ## RND ## pixels16_l2_aligned1(halfH, halfH, full+1, 16, 16, 24, 17);\
00827     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
00828     OPNAME ## pixels16_l2_aligned(dst, halfH, halfHV, stride, 16, 16, 16);\
00829 }\
00830 static void OPNAME ## qpel16_mc13_sh4(uint8_t *dst, uint8_t *src, int stride){\
00831     uint8_t full[24*17];\
00832     uint8_t halfH[272];\
00833     uint8_t halfHV[256];\
00834     copy_block17(full, src, 24, stride, 17);\
00835     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
00836     put ## RND ## pixels16_l2_aligned(halfH, halfH, full, 16, 16, 24, 17);\
00837     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
00838     OPNAME ## pixels16_l2_aligned(dst, halfH+16, halfHV, stride, 16, 16, 16);\
00839 }\
00840 static void OPNAME ## qpel16_mc33_sh4(uint8_t *dst, uint8_t *src, int stride){\
00841     uint8_t full[24*17];\
00842     uint8_t halfH[272];\
00843     uint8_t halfHV[256];\
00844     copy_block17(full, src, 24, stride, 17);\
00845     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
00846     put ## RND ## pixels16_l2_aligned1(halfH, halfH, full+1, 16, 16, 24, 17);\
00847     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
00848     OPNAME ## pixels16_l2_aligned(dst, halfH+16, halfHV, stride, 16, 16, 16);\
00849 }\
00850 static void OPNAME ## qpel16_mc21_sh4(uint8_t *dst, uint8_t *src, int stride){\
00851     uint8_t halfH[272];\
00852     uint8_t halfHV[256];\
00853     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\
00854     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
00855     OPNAME ## pixels16_l2_aligned(dst, halfH, halfHV, stride, 16, 16, 16);\
00856 }\
00857 static void OPNAME ## qpel16_mc23_sh4(uint8_t *dst, uint8_t *src, int stride){\
00858     uint8_t halfH[272];\
00859     uint8_t halfHV[256];\
00860     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\
00861     put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
00862     OPNAME ## pixels16_l2_aligned(dst, halfH+16, halfHV, stride, 16, 16, 16);\
00863 }\
00864 static void OPNAME ## qpel16_mc12_sh4(uint8_t *dst, uint8_t *src, int stride){\
00865     uint8_t full[24*17];\
00866     uint8_t halfH[272];\
00867     copy_block17(full, src, 24, stride, 17);\
00868     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
00869     put ## RND ## pixels16_l2_aligned(halfH, halfH, full, 16, 16, 24, 17);\
00870     OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\
00871 }\
00872 static void OPNAME ## qpel16_mc32_sh4(uint8_t *dst, uint8_t *src, int stride){\
00873     uint8_t full[24*17];\
00874     uint8_t halfH[272];\
00875     copy_block17(full, src, 24, stride, 17);\
00876     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
00877     put ## RND ## pixels16_l2_aligned1(halfH, halfH, full+1, 16, 16, 24, 17);\
00878     OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\
00879 }\
00880 static void OPNAME ## qpel16_mc22_sh4(uint8_t *dst, uint8_t *src, int stride){\
00881     uint8_t halfH[272];\
00882     put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\
00883     OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\
00884 }
00885 
00886 #define op_avg(a, b) a = (((a)+cm[((b) + 16)>>5]+1)>>1)
00887 #define op_avg_no_rnd(a, b) a = (((a)+cm[((b) + 15)>>5])>>1)
00888 #define op_put(a, b) a = cm[((b) + 16)>>5]
00889 #define op_put_no_rnd(a, b) a = cm[((b) + 15)>>5]
00890 
00891 QPEL_MC(0, put_       , _       , op_put)
00892 QPEL_MC(1, put_no_rnd_, _no_rnd_, op_put_no_rnd)
00893 QPEL_MC(0, avg_       , _       , op_avg)
00894 //QPEL_MC(1, avg_no_rnd , _       , op_avg)
00895 #undef op_avg
00896 #undef op_avg_no_rnd
00897 #undef op_put
00898 #undef op_put_no_rnd
00899 
00900 #define H264_LOWPASS(OPNAME, OP, OP2) \
00901 static inline void OPNAME ## h264_qpel_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,int w,int h){\
00902     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
00903     do {\
00904         int srcB,srcA,src0,src1,src2,src3,src4,src5,src6;\
00905         uint8_t *s = src-2;\
00906         srcB = *s++;\
00907         srcA = *s++;\
00908         src0 = *s++;\
00909         src1 = *s++;\
00910         src2 = *s++;\
00911         src3 = *s++;\
00912         OP(dst[0], (src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));\
00913         src4 = *s++;\
00914         OP(dst[1], (src1+src2)*20 - (src0+src3)*5 + (srcA+src4));\
00915         src5 = *s++;\
00916         OP(dst[2], (src2+src3)*20 - (src1+src4)*5 + (src0+src5));\
00917         src6 = *s++;\
00918         OP(dst[3], (src3+src4)*20 - (src2+src5)*5 + (src1+src6));\
00919       if (w>4) { /* it optimized */ \
00920         int src7,src8,src9,src10; \
00921         src7 = *s++;\
00922         OP(dst[4], (src4+src5)*20 - (src3+src6)*5 + (src2+src7));\
00923         src8 = *s++;\
00924         OP(dst[5], (src5+src6)*20 - (src4+src7)*5 + (src3+src8));\
00925         src9 = *s++;\
00926         OP(dst[6], (src6+src7)*20 - (src5+src8)*5 + (src4+src9));\
00927         src10 = *s++;\
00928         OP(dst[7], (src7+src8)*20 - (src6+src9)*5 + (src5+src10));\
00929        if (w>8) { \
00930         int src11,src12,src13,src14,src15,src16,src17,src18; \
00931         src11 = *s++;\
00932         OP(dst[8] , (src8 +src9 )*20 - (src7 +src10)*5 + (src6 +src11));\
00933         src12 = *s++;\
00934         OP(dst[9] , (src9 +src10)*20 - (src8 +src11)*5 + (src7 +src12));\
00935         src13 = *s++;\
00936         OP(dst[10], (src10+src11)*20 - (src9 +src12)*5 + (src8 +src13));\
00937         src14 = *s++;\
00938         OP(dst[11], (src11+src12)*20 - (src10+src13)*5 + (src9 +src14));\
00939         src15 = *s++;\
00940         OP(dst[12], (src12+src13)*20 - (src11+src14)*5 + (src10+src15));\
00941         src16 = *s++;\
00942         OP(dst[13], (src13+src14)*20 - (src12+src15)*5 + (src11+src16));\
00943         src17 = *s++;\
00944         OP(dst[14], (src14+src15)*20 - (src13+src16)*5 + (src12+src17));\
00945         src18 = *s++;\
00946         OP(dst[15], (src15+src16)*20 - (src14+src17)*5 + (src13+src18));\
00947        } \
00948       } \
00949         dst+=dstStride;\
00950         src+=srcStride;\
00951     }while(--h);\
00952 }\
00953 \
00954 static inline void OPNAME ## h264_qpel_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,int w,int h){\
00955     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
00956     do{\
00957         int srcB,srcA,src0,src1,src2,src3,src4,src5,src6;\
00958         uint8_t *s = src-2*srcStride,*d=dst;\
00959         srcB = *s; s+=srcStride;\
00960         srcA = *s; s+=srcStride;\
00961         src0 = *s; s+=srcStride;\
00962         src1 = *s; s+=srcStride;\
00963         src2 = *s; s+=srcStride;\
00964         src3 = *s; s+=srcStride;\
00965         OP(*d, (src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));d+=dstStride;\
00966         src4 = *s; s+=srcStride;\
00967         OP(*d, (src1+src2)*20 - (src0+src3)*5 + (srcA+src4));d+=dstStride;\
00968         src5 = *s; s+=srcStride;\
00969         OP(*d, (src2+src3)*20 - (src1+src4)*5 + (src0+src5));d+=dstStride;\
00970         src6 = *s; s+=srcStride;\
00971         OP(*d, (src3+src4)*20 - (src2+src5)*5 + (src1+src6));d+=dstStride;\
00972       if (h>4) { \
00973         int src7,src8,src9,src10; \
00974         src7 = *s; s+=srcStride;\
00975         OP(*d, (src4+src5)*20 - (src3+src6)*5 + (src2+src7));d+=dstStride;\
00976         src8 = *s; s+=srcStride;\
00977         OP(*d, (src5+src6)*20 - (src4+src7)*5 + (src3+src8));d+=dstStride;\
00978         src9 = *s; s+=srcStride;\
00979         OP(*d, (src6+src7)*20 - (src5+src8)*5 + (src4+src9));d+=dstStride;\
00980         src10 = *s; s+=srcStride;\
00981         OP(*d, (src7+src8)*20 - (src6+src9)*5 + (src5+src10));d+=dstStride;\
00982        if (h>8) { \
00983         int src11,src12,src13,src14,src15,src16,src17,src18; \
00984         src11 = *s; s+=srcStride;\
00985         OP(*d , (src8 +src9 )*20 - (src7 +src10)*5 + (src6 +src11));d+=dstStride;\
00986         src12 = *s; s+=srcStride;\
00987         OP(*d , (src9 +src10)*20 - (src8 +src11)*5 + (src7 +src12));d+=dstStride;\
00988         src13 = *s; s+=srcStride;\
00989         OP(*d, (src10+src11)*20 - (src9 +src12)*5 + (src8 +src13));d+=dstStride;\
00990         src14 = *s; s+=srcStride;\
00991         OP(*d, (src11+src12)*20 - (src10+src13)*5 + (src9 +src14));d+=dstStride;\
00992         src15 = *s; s+=srcStride;\
00993         OP(*d, (src12+src13)*20 - (src11+src14)*5 + (src10+src15));d+=dstStride;\
00994         src16 = *s; s+=srcStride;\
00995         OP(*d, (src13+src14)*20 - (src12+src15)*5 + (src11+src16));d+=dstStride;\
00996         src17 = *s; s+=srcStride;\
00997         OP(*d, (src14+src15)*20 - (src13+src16)*5 + (src12+src17));d+=dstStride;\
00998         src18 = *s; s+=srcStride;\
00999         OP(*d, (src15+src16)*20 - (src14+src17)*5 + (src13+src18));d+=dstStride;\
01000        } \
01001       } \
01002         dst++;\
01003         src++;\
01004     }while(--w);\
01005 }\
01006 \
01007 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){\
01008     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
01009     int i;\
01010     src -= 2*srcStride;\
01011     i= h+5; \
01012     do {\
01013         int srcB,srcA,src0,src1,src2,src3,src4,src5,src6;\
01014         uint8_t *s = src-2;\
01015         srcB = *s++;\
01016         srcA = *s++;\
01017         src0 = *s++;\
01018         src1 = *s++;\
01019         src2 = *s++;\
01020         src3 = *s++;\
01021         tmp[0] = ((src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));\
01022         src4 = *s++;\
01023         tmp[1] = ((src1+src2)*20 - (src0+src3)*5 + (srcA+src4));\
01024         src5 = *s++;\
01025         tmp[2] = ((src2+src3)*20 - (src1+src4)*5 + (src0+src5));\
01026         src6 = *s++;\
01027         tmp[3] = ((src3+src4)*20 - (src2+src5)*5 + (src1+src6));\
01028       if (w>4) { /* it optimized */ \
01029         int src7,src8,src9,src10; \
01030         src7 = *s++;\
01031         tmp[4] = ((src4+src5)*20 - (src3+src6)*5 + (src2+src7));\
01032         src8 = *s++;\
01033         tmp[5] = ((src5+src6)*20 - (src4+src7)*5 + (src3+src8));\
01034         src9 = *s++;\
01035         tmp[6] = ((src6+src7)*20 - (src5+src8)*5 + (src4+src9));\
01036         src10 = *s++;\
01037         tmp[7] = ((src7+src8)*20 - (src6+src9)*5 + (src5+src10));\
01038        if (w>8) { \
01039         int src11,src12,src13,src14,src15,src16,src17,src18; \
01040         src11 = *s++;\
01041         tmp[8] = ((src8 +src9 )*20 - (src7 +src10)*5 + (src6 +src11));\
01042         src12 = *s++;\
01043         tmp[9] = ((src9 +src10)*20 - (src8 +src11)*5 + (src7 +src12));\
01044         src13 = *s++;\
01045         tmp[10] = ((src10+src11)*20 - (src9 +src12)*5 + (src8 +src13));\
01046         src14 = *s++;\
01047         tmp[11] = ((src11+src12)*20 - (src10+src13)*5 + (src9 +src14));\
01048         src15 = *s++;\
01049         tmp[12] = ((src12+src13)*20 - (src11+src14)*5 + (src10+src15));\
01050         src16 = *s++;\
01051         tmp[13] = ((src13+src14)*20 - (src12+src15)*5 + (src11+src16));\
01052         src17 = *s++;\
01053         tmp[14] = ((src14+src15)*20 - (src13+src16)*5 + (src12+src17));\
01054         src18 = *s++;\
01055         tmp[15] = ((src15+src16)*20 - (src14+src17)*5 + (src13+src18));\
01056        } \
01057       } \
01058         tmp+=tmpStride;\
01059         src+=srcStride;\
01060     }while(--i);\
01061     tmp -= tmpStride*(h+5-2);\
01062     i = w; \
01063     do {\
01064         int tmpB,tmpA,tmp0,tmp1,tmp2,tmp3,tmp4,tmp5,tmp6;\
01065         int16_t *s = tmp-2*tmpStride; \
01066         uint8_t *d=dst;\
01067         tmpB = *s; s+=tmpStride;\
01068         tmpA = *s; s+=tmpStride;\
01069         tmp0 = *s; s+=tmpStride;\
01070         tmp1 = *s; s+=tmpStride;\
01071         tmp2 = *s; s+=tmpStride;\
01072         tmp3 = *s; s+=tmpStride;\
01073         OP2(*d, (tmp0+tmp1)*20 - (tmpA+tmp2)*5 + (tmpB+tmp3));d+=dstStride;\
01074         tmp4 = *s; s+=tmpStride;\
01075         OP2(*d, (tmp1+tmp2)*20 - (tmp0+tmp3)*5 + (tmpA+tmp4));d+=dstStride;\
01076         tmp5 = *s; s+=tmpStride;\
01077         OP2(*d, (tmp2+tmp3)*20 - (tmp1+tmp4)*5 + (tmp0+tmp5));d+=dstStride;\
01078         tmp6 = *s; s+=tmpStride;\
01079         OP2(*d, (tmp3+tmp4)*20 - (tmp2+tmp5)*5 + (tmp1+tmp6));d+=dstStride;\
01080       if (h>4) { \
01081         int tmp7,tmp8,tmp9,tmp10; \
01082         tmp7 = *s; s+=tmpStride;\
01083         OP2(*d, (tmp4+tmp5)*20 - (tmp3+tmp6)*5 + (tmp2+tmp7));d+=dstStride;\
01084         tmp8 = *s; s+=tmpStride;\
01085         OP2(*d, (tmp5+tmp6)*20 - (tmp4+tmp7)*5 + (tmp3+tmp8));d+=dstStride;\
01086         tmp9 = *s; s+=tmpStride;\
01087         OP2(*d, (tmp6+tmp7)*20 - (tmp5+tmp8)*5 + (tmp4+tmp9));d+=dstStride;\
01088         tmp10 = *s; s+=tmpStride;\
01089         OP2(*d, (tmp7+tmp8)*20 - (tmp6+tmp9)*5 + (tmp5+tmp10));d+=dstStride;\
01090        if (h>8) { \
01091         int tmp11,tmp12,tmp13,tmp14,tmp15,tmp16,tmp17,tmp18; \
01092         tmp11 = *s; s+=tmpStride;\
01093         OP2(*d , (tmp8 +tmp9 )*20 - (tmp7 +tmp10)*5 + (tmp6 +tmp11));d+=dstStride;\
01094         tmp12 = *s; s+=tmpStride;\
01095         OP2(*d , (tmp9 +tmp10)*20 - (tmp8 +tmp11)*5 + (tmp7 +tmp12));d+=dstStride;\
01096         tmp13 = *s; s+=tmpStride;\
01097         OP2(*d, (tmp10+tmp11)*20 - (tmp9 +tmp12)*5 + (tmp8 +tmp13));d+=dstStride;\
01098         tmp14 = *s; s+=tmpStride;\
01099         OP2(*d, (tmp11+tmp12)*20 - (tmp10+tmp13)*5 + (tmp9 +tmp14));d+=dstStride;\
01100         tmp15 = *s; s+=tmpStride;\
01101         OP2(*d, (tmp12+tmp13)*20 - (tmp11+tmp14)*5 + (tmp10+tmp15));d+=dstStride;\
01102         tmp16 = *s; s+=tmpStride;\
01103         OP2(*d, (tmp13+tmp14)*20 - (tmp12+tmp15)*5 + (tmp11+tmp16));d+=dstStride;\
01104         tmp17 = *s; s+=tmpStride;\
01105         OP2(*d, (tmp14+tmp15)*20 - (tmp13+tmp16)*5 + (tmp12+tmp17));d+=dstStride;\
01106         tmp18 = *s; s+=tmpStride;\
01107         OP2(*d, (tmp15+tmp16)*20 - (tmp14+tmp17)*5 + (tmp13+tmp18));d+=dstStride;\
01108        } \
01109       } \
01110         dst++;\
01111         tmp++;\
01112     }while(--i);\
01113 }\
01114 \
01115 static void OPNAME ## h264_qpel4_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
01116     OPNAME ## h264_qpel_h_lowpass(dst,src,dstStride,srcStride,4,4); \
01117 }\
01118 static void OPNAME ## h264_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
01119    OPNAME ## h264_qpel_h_lowpass(dst,src,dstStride,srcStride,8,8); \
01120 }\
01121 static void OPNAME ## h264_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
01122    OPNAME ## h264_qpel_h_lowpass(dst,src,dstStride,srcStride,16,16); \
01123 }\
01124 \
01125 static void OPNAME ## h264_qpel4_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
01126    OPNAME ## h264_qpel_v_lowpass(dst,src,dstStride,srcStride,4,4); \
01127 }\
01128 static void OPNAME ## h264_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
01129    OPNAME ## h264_qpel_v_lowpass(dst,src,dstStride,srcStride,8,8); \
01130 }\
01131 static void OPNAME ## h264_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
01132    OPNAME ## h264_qpel_v_lowpass(dst,src,dstStride,srcStride,16,16); \
01133 }\
01134 static void OPNAME ## h264_qpel4_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
01135    OPNAME ## h264_qpel_hv_lowpass(dst,tmp,src,dstStride,tmpStride,srcStride,4,4); \
01136 }\
01137 static void OPNAME ## h264_qpel8_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
01138    OPNAME ## h264_qpel_hv_lowpass(dst,tmp,src,dstStride,tmpStride,srcStride,8,8); \
01139 }\
01140 static void OPNAME ## h264_qpel16_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
01141    OPNAME ## h264_qpel_hv_lowpass(dst,tmp,src,dstStride,tmpStride,srcStride,16,16); \
01142 }\
01143 
01144 #define H264_MC(OPNAME, SIZE) \
01145 static void OPNAME ## h264_qpel ## SIZE ## _mc00_sh4 (uint8_t *dst, uint8_t *src, int stride){\
01146     OPNAME ## pixels ## SIZE ## _c(dst, src, stride, SIZE);\
01147 }\
01148 \
01149 static void OPNAME ## h264_qpel ## SIZE ## _mc10_sh4(uint8_t *dst, uint8_t *src, int stride){\
01150     uint8_t half[SIZE*SIZE];\
01151     put_h264_qpel ## SIZE ## _h_lowpass(half, src, SIZE, stride);\
01152     OPNAME ## pixels ## SIZE ## _l2_aligned2(dst, src, half, stride, stride, SIZE, SIZE);\
01153 }\
01154 \
01155 static void OPNAME ## h264_qpel ## SIZE ## _mc20_sh4(uint8_t *dst, uint8_t *src, int stride){\
01156     OPNAME ## h264_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride);\
01157 }\
01158 \
01159 static void OPNAME ## h264_qpel ## SIZE ## _mc30_sh4(uint8_t *dst, uint8_t *src, int stride){\
01160     uint8_t half[SIZE*SIZE];\
01161     put_h264_qpel ## SIZE ## _h_lowpass(half, src, SIZE, stride);\
01162     OPNAME ## pixels ## SIZE ## _l2_aligned2(dst, src+1, half, stride, stride, SIZE, SIZE);\
01163 }\
01164 \
01165 static void OPNAME ## h264_qpel ## SIZE ## _mc01_sh4(uint8_t *dst, uint8_t *src, int stride){\
01166     uint8_t full[SIZE*(SIZE+5)];\
01167     uint8_t * const full_mid= full + SIZE*2;\
01168     uint8_t half[SIZE*SIZE];\
01169     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\
01170     put_h264_qpel ## SIZE ## _v_lowpass(half, full_mid, SIZE, SIZE);\
01171     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, full_mid, half, stride, SIZE, SIZE, SIZE);\
01172 }\
01173 \
01174 static void OPNAME ## h264_qpel ## SIZE ## _mc02_sh4(uint8_t *dst, uint8_t *src, int stride){\
01175     uint8_t full[SIZE*(SIZE+5)];\
01176     uint8_t * const full_mid= full + SIZE*2;\
01177     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\
01178     OPNAME ## h264_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE);\
01179 }\
01180 \
01181 static void OPNAME ## h264_qpel ## SIZE ## _mc03_sh4(uint8_t *dst, uint8_t *src, int stride){\
01182     uint8_t full[SIZE*(SIZE+5)];\
01183     uint8_t * const full_mid= full + SIZE*2;\
01184     uint8_t half[SIZE*SIZE];\
01185     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\
01186     put_h264_qpel ## SIZE ## _v_lowpass(half, full_mid, SIZE, SIZE);\
01187     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, full_mid+SIZE, half, stride, SIZE, SIZE, SIZE);\
01188 }\
01189 \
01190 static void OPNAME ## h264_qpel ## SIZE ## _mc11_sh4(uint8_t *dst, uint8_t *src, int stride){\
01191     uint8_t full[SIZE*(SIZE+5)];\
01192     uint8_t * const full_mid= full + SIZE*2;\
01193     uint8_t halfH[SIZE*SIZE];\
01194     uint8_t halfV[SIZE*SIZE];\
01195     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\
01196     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\
01197     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\
01198     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\
01199 }\
01200 \
01201 static void OPNAME ## h264_qpel ## SIZE ## _mc31_sh4(uint8_t *dst, uint8_t *src, int stride){\
01202     uint8_t full[SIZE*(SIZE+5)];\
01203     uint8_t * const full_mid= full + SIZE*2;\
01204     uint8_t halfH[SIZE*SIZE];\
01205     uint8_t halfV[SIZE*SIZE];\
01206     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\
01207     copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\
01208     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\
01209     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\
01210 }\
01211 \
01212 static void OPNAME ## h264_qpel ## SIZE ## _mc13_sh4(uint8_t *dst, uint8_t *src, int stride){\
01213     uint8_t full[SIZE*(SIZE+5)];\
01214     uint8_t * const full_mid= full + SIZE*2;\
01215     uint8_t halfH[SIZE*SIZE];\
01216     uint8_t halfV[SIZE*SIZE];\
01217     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\
01218     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\
01219     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\
01220     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\
01221 }\
01222 \
01223 static void OPNAME ## h264_qpel ## SIZE ## _mc33_sh4(uint8_t *dst, uint8_t *src, int stride){\
01224     uint8_t full[SIZE*(SIZE+5)];\
01225     uint8_t * const full_mid= full + SIZE*2;\
01226     uint8_t halfH[SIZE*SIZE];\
01227     uint8_t halfV[SIZE*SIZE];\
01228     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\
01229     copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\
01230     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\
01231     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\
01232 }\
01233 \
01234 static void OPNAME ## h264_qpel ## SIZE ## _mc22_sh4(uint8_t *dst, uint8_t *src, int stride){\
01235     int16_t tmp[SIZE*(SIZE+5)];\
01236     OPNAME ## h264_qpel ## SIZE ## _hv_lowpass(dst, tmp, src, stride, SIZE, stride);\
01237 }\
01238 \
01239 static void OPNAME ## h264_qpel ## SIZE ## _mc21_sh4(uint8_t *dst, uint8_t *src, int stride){\
01240     int16_t tmp[SIZE*(SIZE+5)];\
01241     uint8_t halfH[SIZE*SIZE];\
01242     uint8_t halfHV[SIZE*SIZE];\
01243     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\
01244     put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\
01245     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfHV, stride, SIZE, SIZE, SIZE);\
01246 }\
01247 \
01248 static void OPNAME ## h264_qpel ## SIZE ## _mc23_sh4(uint8_t *dst, uint8_t *src, int stride){\
01249     int16_t tmp[SIZE*(SIZE+5)];\
01250     uint8_t halfH[SIZE*SIZE];\
01251     uint8_t halfHV[SIZE*SIZE];\
01252     put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\
01253     put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\
01254     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfHV, stride, SIZE, SIZE, SIZE);\
01255 }\
01256 \
01257 static void OPNAME ## h264_qpel ## SIZE ## _mc12_sh4(uint8_t *dst, uint8_t *src, int stride){\
01258     uint8_t full[SIZE*(SIZE+5)];\
01259     uint8_t * const full_mid= full + SIZE*2;\
01260     int16_t tmp[SIZE*(SIZE+5)];\
01261     uint8_t halfV[SIZE*SIZE];\
01262     uint8_t halfHV[SIZE*SIZE];\
01263     copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\
01264     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\
01265     put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\
01266     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfV, halfHV, stride, SIZE, SIZE, SIZE);\
01267 }\
01268 \
01269 static void OPNAME ## h264_qpel ## SIZE ## _mc32_sh4(uint8_t *dst, uint8_t *src, int stride){\
01270     uint8_t full[SIZE*(SIZE+5)];\
01271     uint8_t * const full_mid= full + SIZE*2;\
01272     int16_t tmp[SIZE*(SIZE+5)];\
01273     uint8_t halfV[SIZE*SIZE];\
01274     uint8_t halfHV[SIZE*SIZE];\
01275     copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\
01276     put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\
01277     put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\
01278     OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfV, halfHV, stride, SIZE, SIZE, SIZE);\
01279 }\
01280 
01281 #define op_avg(a, b)  a = (((a)+cm[((b) + 16)>>5]+1)>>1)
01282 //#define op_avg2(a, b) a = (((a)*w1+cm[((b) + 16)>>5]*w2 + o + 64)>>7)
01283 #define op_put(a, b)  a = cm[((b) + 16)>>5]
01284 #define op2_avg(a, b)  a = (((a)+cm[((b) + 512)>>10]+1)>>1)
01285 #define op2_put(a, b)  a = cm[((b) + 512)>>10]
01286 
01287 H264_LOWPASS(put_       , op_put, op2_put)
01288 H264_LOWPASS(avg_       , op_avg, op2_avg)
01289 H264_MC(put_, 4)
01290 H264_MC(put_, 8)
01291 H264_MC(put_, 16)
01292 H264_MC(avg_, 4)
01293 H264_MC(avg_, 8)
01294 H264_MC(avg_, 16)
01295 
01296 #undef op_avg
01297 #undef op_put
01298 #undef op2_avg
01299 #undef op2_put
01300 
01301 static void wmv2_mspel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){
01302     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
01303 
01304     do{
01305         int src_1,src0,src1,src2,src3,src4,src5,src6,src7,src8,src9;
01306         uint8_t *s = src;
01307         src_1 = s[-1];
01308         src0 = *s++;
01309         src1 = *s++;
01310         src2 = *s++;
01311         dst[0]= cm[(9*(src0 + src1) - (src_1 + src2) + 8)>>4];
01312         src3 = *s++;
01313         dst[1]= cm[(9*(src1 + src2) - (src0 + src3) + 8)>>4];
01314         src4 = *s++;
01315         dst[2]= cm[(9*(src2 + src3) - (src1 + src4) + 8)>>4];
01316         src5 = *s++;
01317         dst[3]= cm[(9*(src3 + src4) - (src2 + src5) + 8)>>4];
01318         src6 = *s++;
01319         dst[4]= cm[(9*(src4 + src5) - (src3 + src6) + 8)>>4];
01320         src7 = *s++;
01321         dst[5]= cm[(9*(src5 + src6) - (src4 + src7) + 8)>>4];
01322         src8 = *s++;
01323         dst[6]= cm[(9*(src6 + src7) - (src5 + src8) + 8)>>4];
01324         src9 = *s++;
01325         dst[7]= cm[(9*(src7 + src8) - (src6 + src9) + 8)>>4];
01326         dst+=dstStride;
01327         src+=srcStride;
01328     }while(--h);
01329 }
01330 
01331 static void wmv2_mspel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int w){
01332     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
01333 
01334     do{
01335         int src_1,src0,src1,src2,src3,src4,src5,src6,src7,src8,src9;
01336         uint8_t *s = src,*d = dst;
01337         src_1 = *(s-srcStride);
01338         src0 = *s; s+=srcStride;
01339         src1 = *s; s+=srcStride;
01340         src2 = *s; s+=srcStride;
01341         *d= cm[(9*(src0 + src1) - (src_1 + src2) + 8)>>4]; d+=dstStride;
01342         src3 = *s; s+=srcStride;
01343         *d= cm[(9*(src1 + src2) - (src0  + src3) + 8)>>4]; d+=dstStride;
01344         src4 = *s; s+=srcStride;
01345         *d= cm[(9*(src2 + src3) - (src1  + src4) + 8)>>4]; d+=dstStride;
01346         src5 = *s; s+=srcStride;
01347         *d= cm[(9*(src3 + src4) - (src2  + src5) + 8)>>4]; d+=dstStride;
01348         src6 = *s; s+=srcStride;
01349         *d= cm[(9*(src4 + src5) - (src3  + src6) + 8)>>4]; d+=dstStride;
01350         src7 = *s; s+=srcStride;
01351         *d= cm[(9*(src5 + src6) - (src4  + src7) + 8)>>4]; d+=dstStride;
01352         src8 = *s; s+=srcStride;
01353         *d= cm[(9*(src6 + src7) - (src5  + src8) + 8)>>4]; d+=dstStride;
01354         src9 = *s;
01355         *d= cm[(9*(src7 + src8) - (src6  + src9) + 8)>>4]; d+=dstStride;
01356         src++;
01357         dst++;
01358     }while(--w);
01359 }
01360 
01361 static void put_mspel8_mc00_sh4 (uint8_t *dst, uint8_t *src, int stride){
01362     put_pixels8_c(dst, src, stride, 8);
01363 }
01364 
01365 static void put_mspel8_mc10_sh4(uint8_t *dst, uint8_t *src, int stride){
01366     uint8_t half[64];
01367     wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);
01368     put_pixels8_l2_aligned2(dst, src, half, stride, stride, 8, 8);
01369 }
01370 
01371 static void put_mspel8_mc20_sh4(uint8_t *dst, uint8_t *src, int stride){
01372     wmv2_mspel8_h_lowpass(dst, src, stride, stride, 8);
01373 }
01374 
01375 static void put_mspel8_mc30_sh4(uint8_t *dst, uint8_t *src, int stride){
01376     uint8_t half[64];
01377     wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);
01378     put_pixels8_l2_aligned2(dst, src+1, half, stride, stride, 8, 8);
01379 }
01380 
01381 static void put_mspel8_mc02_sh4(uint8_t *dst, uint8_t *src, int stride){
01382     wmv2_mspel8_v_lowpass(dst, src, stride, stride, 8);
01383 }
01384 
01385 static void put_mspel8_mc12_sh4(uint8_t *dst, uint8_t *src, int stride){
01386     uint8_t halfH[88];
01387     uint8_t halfV[64];
01388     uint8_t halfHV[64];
01389     wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);
01390     wmv2_mspel8_v_lowpass(halfV, src, 8, stride, 8);
01391     wmv2_mspel8_v_lowpass(halfHV, halfH+8, 8, 8, 8);
01392     put_pixels8_l2_aligned(dst, halfV, halfHV, stride, 8, 8, 8);
01393 }
01394 static void put_mspel8_mc32_sh4(uint8_t *dst, uint8_t *src, int stride){
01395     uint8_t halfH[88];
01396     uint8_t halfV[64];
01397     uint8_t halfHV[64];
01398     wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);
01399     wmv2_mspel8_v_lowpass(halfV, src+1, 8, stride, 8);
01400     wmv2_mspel8_v_lowpass(halfHV, halfH+8, 8, 8, 8);
01401     put_pixels8_l2_aligned(dst, halfV, halfHV, stride, 8, 8, 8);
01402 }
01403 static void put_mspel8_mc22_sh4(uint8_t *dst, uint8_t *src, int stride){
01404     uint8_t halfH[88];
01405     wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);
01406     wmv2_mspel8_v_lowpass(dst, halfH+8, stride, 8, 8);
01407 }