Statistics
| Branch: | Revision:

ffmpeg / libavcodec / sh4 / qpel.c @ be449fca

History | View | Annotate | Download (69.1 KB)

1
/*
2
 * This is optimized for sh, which have post increment addressing (*p++).
3
 * Some CPU may be index (p[n]) faster than post increment (*p++).
4
 *
5
 * copyright (c) 2001-2003 BERO <bero@geocities.co.jp>
6
 *
7
 * This file is part of FFmpeg.
8
 *
9
 * FFmpeg is free software; you can redistribute it and/or
10
 * modify it under the terms of the GNU Lesser General Public
11
 * License as published by the Free Software Foundation; either
12
 * version 2.1 of the License, or (at your option) any later version.
13
 *
14
 * FFmpeg is distributed in the hope that it will be useful,
15
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
17
 * Lesser General Public License for more details.
18
 *
19
 * You should have received a copy of the GNU Lesser General Public
20
 * License along with FFmpeg; if not, write to the Free Software
21
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22
 */
23

    
24
#define PIXOP2(OPNAME, OP) \
25
/*static inline void OPNAME ## _no_rnd_pixels8_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
26
{\
27
        do {\
28
                OP(LP(dst  ),no_rnd_avg32(AV_RN32(src1  ),AV_RN32(src2  )) ); \
29
                OP(LP(dst+4),no_rnd_avg32(AV_RN32(src1+4),AV_RN32(src2+4)) ); \
30
                src1+=src_stride1; \
31
                src2+=src_stride2; \
32
                dst+=dst_stride; \
33
        } while(--h); \
34
}\
35
\
36
static inline void OPNAME ## _pixels8_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
37
{\
38
        do {\
39
                OP(LP(dst  ),rnd_avg32(AV_RN32(src1  ),AV_RN32(src2  )) ); \
40
                OP(LP(dst+4),rnd_avg32(AV_RN32(src1+4),AV_RN32(src2+4)) ); \
41
                src1+=src_stride1; \
42
                src2+=src_stride2; \
43
                dst+=dst_stride; \
44
        } while(--h); \
45
}\
46
\
47
static inline void OPNAME ## _pixels4_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
48
{\
49
        do {\
50
                OP(LP(dst  ),rnd_avg32(AV_RN32(src1  ),AV_RN32(src2  )) ); \
51
                src1+=src_stride1; \
52
                src2+=src_stride2; \
53
                dst+=dst_stride; \
54
        } while(--h); \
55
}\
56
\
57
static inline void OPNAME ## _no_rnd_pixels16_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
58
{\
59
        do {\
60
                OP(LP(dst  ),no_rnd_avg32(AV_RN32(src1  ),AV_RN32(src2  )) ); \
61
                OP(LP(dst+4),no_rnd_avg32(AV_RN32(src1+4),AV_RN32(src2+4)) ); \
62
                OP(LP(dst+8),no_rnd_avg32(AV_RN32(src1+8),AV_RN32(src2+8)) ); \
63
                OP(LP(dst+12),no_rnd_avg32(AV_RN32(src1+12),AV_RN32(src2+12)) ); \
64
                src1+=src_stride1; \
65
                src2+=src_stride2; \
66
                dst+=dst_stride; \
67
        } while(--h); \
68
}\
69
\
70
static inline void OPNAME ## _pixels16_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
71
{\
72
        do {\
73
                OP(LP(dst  ),rnd_avg32(AV_RN32(src1  ),AV_RN32(src2  )) ); \
74
                OP(LP(dst+4),rnd_avg32(AV_RN32(src1+4),AV_RN32(src2+4)) ); \
75
                OP(LP(dst+8),rnd_avg32(AV_RN32(src1+8),AV_RN32(src2+8)) ); \
76
                OP(LP(dst+12),rnd_avg32(AV_RN32(src1+12),AV_RN32(src2+12)) ); \
77
                src1+=src_stride1; \
78
                src2+=src_stride2; \
79
                dst+=dst_stride; \
80
        } while(--h); \
81
}*/\
82
\
83
static inline void OPNAME ## _pixels4_l2_aligned(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
84
{\
85
        do {\
86
                OP(LP(dst  ),rnd_avg32(LP(src1  ),LP(src2  )) ); \
87
                src1+=src_stride1; \
88
                src2+=src_stride2; \
89
                dst+=dst_stride; \
90
        } while(--h); \
91
}\
92
\
93
static inline void OPNAME ## _pixels4_l2_aligned2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
94
{\
95
        do {\
96
                OP(LP(dst  ),rnd_avg32(AV_RN32(src1  ),LP(src2  )) ); \
97
                src1+=src_stride1; \
98
                src2+=src_stride2; \
99
                dst+=dst_stride; \
100
        } while(--h); \
101
}\
102
\
103
static inline void OPNAME ## _no_rnd_pixels16_l2_aligned2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
104
{\
105
        do {\
106
                OP(LP(dst  ),no_rnd_avg32(AV_RN32(src1  ),LP(src2  )) ); \
107
                OP(LP(dst+4),no_rnd_avg32(AV_RN32(src1+4),LP(src2+4)) ); \
108
                OP(LP(dst+8),no_rnd_avg32(AV_RN32(src1+8),LP(src2+8)) ); \
109
                OP(LP(dst+12),no_rnd_avg32(AV_RN32(src1+12),LP(src2+12)) ); \
110
                src1+=src_stride1; \
111
                src2+=src_stride2; \
112
                dst+=dst_stride; \
113
        } while(--h); \
114
}\
115
\
116
static inline void OPNAME ## _pixels16_l2_aligned2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
117
{\
118
        do {\
119
                OP(LP(dst  ),rnd_avg32(AV_RN32(src1  ),LP(src2  )) ); \
120
                OP(LP(dst+4),rnd_avg32(AV_RN32(src1+4),LP(src2+4)) ); \
121
                OP(LP(dst+8),rnd_avg32(AV_RN32(src1+8),LP(src2+8)) ); \
122
                OP(LP(dst+12),rnd_avg32(AV_RN32(src1+12),LP(src2+12)) ); \
123
                src1+=src_stride1; \
124
                src2+=src_stride2; \
125
                dst+=dst_stride; \
126
        } while(--h); \
127
}\
128
\
129
static inline void OPNAME ## _no_rnd_pixels8_l2_aligned2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
130
{\
131
        do { /* onlye src2 aligned */\
132
                OP(LP(dst  ),no_rnd_avg32(AV_RN32(src1  ),LP(src2  )) ); \
133
                OP(LP(dst+4),no_rnd_avg32(AV_RN32(src1+4),LP(src2+4)) ); \
134
                src1+=src_stride1; \
135
                src2+=src_stride2; \
136
                dst+=dst_stride; \
137
        } while(--h); \
138
}\
139
\
140
static inline void OPNAME ## _pixels8_l2_aligned2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
141
{\
142
        do {\
143
                OP(LP(dst  ),rnd_avg32(AV_RN32(src1  ),LP(src2  )) ); \
144
                OP(LP(dst+4),rnd_avg32(AV_RN32(src1+4),LP(src2+4)) ); \
145
                src1+=src_stride1; \
146
                src2+=src_stride2; \
147
                dst+=dst_stride; \
148
        } while(--h); \
149
}\
150
\
151
static inline void OPNAME ## _no_rnd_pixels8_l2_aligned(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
152
{\
153
        do {\
154
                OP(LP(dst  ),no_rnd_avg32(LP(src1  ),LP(src2  )) ); \
155
                OP(LP(dst+4),no_rnd_avg32(LP(src1+4),LP(src2+4)) ); \
156
                src1+=src_stride1; \
157
                src2+=src_stride2; \
158
                dst+=dst_stride; \
159
        } while(--h); \
160
}\
161
\
162
static inline void OPNAME ## _pixels8_l2_aligned(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
163
{\
164
        do {\
165
                OP(LP(dst  ),rnd_avg32(LP(src1  ),LP(src2  )) ); \
166
                OP(LP(dst+4),rnd_avg32(LP(src1+4),LP(src2+4)) ); \
167
                src1+=src_stride1; \
168
                src2+=src_stride2; \
169
                dst+=dst_stride; \
170
        } while(--h); \
171
}\
172
\
173
static inline void OPNAME ## _no_rnd_pixels16_l2_aligned(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
174
{\
175
        do {\
176
                OP(LP(dst  ),no_rnd_avg32(LP(src1  ),LP(src2  )) ); \
177
                OP(LP(dst+4),no_rnd_avg32(LP(src1+4),LP(src2+4)) ); \
178
                OP(LP(dst+8),no_rnd_avg32(LP(src1+8),LP(src2+8)) ); \
179
                OP(LP(dst+12),no_rnd_avg32(LP(src1+12),LP(src2+12)) ); \
180
                src1+=src_stride1; \
181
                src2+=src_stride2; \
182
                dst+=dst_stride; \
183
        } while(--h); \
184
}\
185
\
186
static inline void OPNAME ## _pixels16_l2_aligned(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
187
{\
188
        do {\
189
                OP(LP(dst  ),rnd_avg32(LP(src1  ),LP(src2  )) ); \
190
                OP(LP(dst+4),rnd_avg32(LP(src1+4),LP(src2+4)) ); \
191
                OP(LP(dst+8),rnd_avg32(LP(src1+8),LP(src2+8)) ); \
192
                OP(LP(dst+12),rnd_avg32(LP(src1+12),LP(src2+12)) ); \
193
                src1+=src_stride1; \
194
                src2+=src_stride2; \
195
                dst+=dst_stride; \
196
        } while(--h); \
197
}\
198
\
199
static inline void OPNAME ## _no_rnd_pixels16_l2_aligned1(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
200
{ OPNAME ## _no_rnd_pixels16_l2_aligned2(dst,src2,src1,dst_stride,src_stride2,src_stride1,h); } \
201
\
202
static inline void OPNAME ## _pixels16_l2_aligned1(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
203
{ OPNAME ## _pixels16_l2_aligned2(dst,src2,src1,dst_stride,src_stride2,src_stride1,h); } \
204
\
205
static inline void OPNAME ## _no_rnd_pixels8_l2_aligned1(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
206
{ OPNAME ## _no_rnd_pixels8_l2_aligned2(dst,src2,src1,dst_stride,src_stride2,src_stride1,h); } \
207
\
208
static inline void OPNAME ## _pixels8_l2_aligned1(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, int src_stride1, int src_stride2, int h) \
209
{ OPNAME ## _pixels8_l2_aligned2(dst,src2,src1,dst_stride,src_stride2,src_stride1,h); } \
210
\
211
static inline void OPNAME ## _pixels8_l4_aligned(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
212
        do { \
213
                uint32_t a0,a1,a2,a3; \
214
                UNPACK(a0,a1,LP(src1),LP(src2)); \
215
                UNPACK(a2,a3,LP(src3),LP(src4)); \
216
                OP(LP(dst),rnd_PACK(a0,a1,a2,a3)); \
217
                UNPACK(a0,a1,LP(src1+4),LP(src2+4)); \
218
                UNPACK(a2,a3,LP(src3+4),LP(src4+4)); \
219
                OP(LP(dst+4),rnd_PACK(a0,a1,a2,a3)); \
220
                src1+=src_stride1;\
221
                src2+=src_stride2;\
222
                src3+=src_stride3;\
223
                src4+=src_stride4;\
224
                dst+=dst_stride;\
225
        } while(--h); \
226
} \
227
\
228
static inline void OPNAME ## _no_rnd_pixels8_l4_aligned(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
229
        do { \
230
                uint32_t a0,a1,a2,a3; \
231
                UNPACK(a0,a1,LP(src1),LP(src2)); \
232
                UNPACK(a2,a3,LP(src3),LP(src4)); \
233
                OP(LP(dst),no_rnd_PACK(a0,a1,a2,a3)); \
234
                UNPACK(a0,a1,LP(src1+4),LP(src2+4)); \
235
                UNPACK(a2,a3,LP(src3+4),LP(src4+4)); \
236
                OP(LP(dst+4),no_rnd_PACK(a0,a1,a2,a3)); \
237
                src1+=src_stride1;\
238
                src2+=src_stride2;\
239
                src3+=src_stride3;\
240
                src4+=src_stride4;\
241
                dst+=dst_stride;\
242
        } while(--h); \
243
} \
244
\
245
static inline void OPNAME ## _pixels8_l4_aligned0(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
246
        do { \
247
                uint32_t a0,a1,a2,a3; /* src1 only not aligned */\
248
                UNPACK(a0,a1,AV_RN32(src1),LP(src2)); \
249
                UNPACK(a2,a3,LP(src3),LP(src4)); \
250
                OP(LP(dst),rnd_PACK(a0,a1,a2,a3)); \
251
                UNPACK(a0,a1,AV_RN32(src1+4),LP(src2+4)); \
252
                UNPACK(a2,a3,LP(src3+4),LP(src4+4)); \
253
                OP(LP(dst+4),rnd_PACK(a0,a1,a2,a3)); \
254
                src1+=src_stride1;\
255
                src2+=src_stride2;\
256
                src3+=src_stride3;\
257
                src4+=src_stride4;\
258
                dst+=dst_stride;\
259
        } while(--h); \
260
} \
261
\
262
static inline void OPNAME ## _no_rnd_pixels8_l4_aligned0(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
263
        do { \
264
                uint32_t a0,a1,a2,a3; \
265
                UNPACK(a0,a1,AV_RN32(src1),LP(src2)); \
266
                UNPACK(a2,a3,LP(src3),LP(src4)); \
267
                OP(LP(dst),no_rnd_PACK(a0,a1,a2,a3)); \
268
                UNPACK(a0,a1,AV_RN32(src1+4),LP(src2+4)); \
269
                UNPACK(a2,a3,LP(src3+4),LP(src4+4)); \
270
                OP(LP(dst+4),no_rnd_PACK(a0,a1,a2,a3)); \
271
                src1+=src_stride1;\
272
                src2+=src_stride2;\
273
                src3+=src_stride3;\
274
                src4+=src_stride4;\
275
                dst+=dst_stride;\
276
        } while(--h); \
277
} \
278
\
279
static inline void OPNAME ## _pixels16_l4_aligned(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
280
        do { \
281
                uint32_t a0,a1,a2,a3; \
282
                UNPACK(a0,a1,LP(src1),LP(src2)); \
283
                UNPACK(a2,a3,LP(src3),LP(src4)); \
284
                OP(LP(dst),rnd_PACK(a0,a1,a2,a3)); \
285
                UNPACK(a0,a1,LP(src1+4),LP(src2+4)); \
286
                UNPACK(a2,a3,LP(src3+4),LP(src4+4)); \
287
                OP(LP(dst+8),rnd_PACK(a0,a1,a2,a3)); \
288
                UNPACK(a0,a1,LP(src1+8),LP(src2+8)); \
289
                UNPACK(a2,a3,LP(src3+8),LP(src4+8)); \
290
                OP(LP(dst+8),rnd_PACK(a0,a1,a2,a3)); \
291
                UNPACK(a0,a1,LP(src1+12),LP(src2+12)); \
292
                UNPACK(a2,a3,LP(src3+12),LP(src4+12)); \
293
                OP(LP(dst+12),rnd_PACK(a0,a1,a2,a3)); \
294
                src1+=src_stride1;\
295
                src2+=src_stride2;\
296
                src3+=src_stride3;\
297
                src4+=src_stride4;\
298
                dst+=dst_stride;\
299
        } while(--h); \
300
} \
301
\
302
static inline void OPNAME ## _no_rnd_pixels16_l4_aligned(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
303
        do { \
304
                uint32_t a0,a1,a2,a3; \
305
                UNPACK(a0,a1,LP(src1),LP(src2)); \
306
                UNPACK(a2,a3,LP(src3),LP(src4)); \
307
                OP(LP(dst),no_rnd_PACK(a0,a1,a2,a3)); \
308
                UNPACK(a0,a1,LP(src1+4),LP(src2+4)); \
309
                UNPACK(a2,a3,LP(src3+4),LP(src4+4)); \
310
                OP(LP(dst+4),no_rnd_PACK(a0,a1,a2,a3)); \
311
                UNPACK(a0,a1,LP(src1+8),LP(src2+8)); \
312
                UNPACK(a2,a3,LP(src3+8),LP(src4+8)); \
313
                OP(LP(dst+8),no_rnd_PACK(a0,a1,a2,a3)); \
314
                UNPACK(a0,a1,LP(src1+12),LP(src2+12)); \
315
                UNPACK(a2,a3,LP(src3+12),LP(src4+12)); \
316
                OP(LP(dst+12),no_rnd_PACK(a0,a1,a2,a3)); \
317
                src1+=src_stride1;\
318
                src2+=src_stride2;\
319
                src3+=src_stride3;\
320
                src4+=src_stride4;\
321
                dst+=dst_stride;\
322
        } while(--h); \
323
} \
324
\
325
static inline void OPNAME ## _pixels16_l4_aligned0(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
326
        do { /* src1 is unaligned */\
327
                uint32_t a0,a1,a2,a3; \
328
                UNPACK(a0,a1,AV_RN32(src1),LP(src2)); \
329
                UNPACK(a2,a3,LP(src3),LP(src4)); \
330
                OP(LP(dst),rnd_PACK(a0,a1,a2,a3)); \
331
                UNPACK(a0,a1,AV_RN32(src1+4),LP(src2+4)); \
332
                UNPACK(a2,a3,LP(src3+4),LP(src4+4)); \
333
                OP(LP(dst+8),rnd_PACK(a0,a1,a2,a3)); \
334
                UNPACK(a0,a1,AV_RN32(src1+8),LP(src2+8)); \
335
                UNPACK(a2,a3,LP(src3+8),LP(src4+8)); \
336
                OP(LP(dst+8),rnd_PACK(a0,a1,a2,a3)); \
337
                UNPACK(a0,a1,AV_RN32(src1+12),LP(src2+12)); \
338
                UNPACK(a2,a3,LP(src3+12),LP(src4+12)); \
339
                OP(LP(dst+12),rnd_PACK(a0,a1,a2,a3)); \
340
                src1+=src_stride1;\
341
                src2+=src_stride2;\
342
                src3+=src_stride3;\
343
                src4+=src_stride4;\
344
                dst+=dst_stride;\
345
        } while(--h); \
346
} \
347
\
348
static inline void OPNAME ## _no_rnd_pixels16_l4_aligned0(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
349
        do { \
350
                uint32_t a0,a1,a2,a3; \
351
                UNPACK(a0,a1,AV_RN32(src1),LP(src2)); \
352
                UNPACK(a2,a3,LP(src3),LP(src4)); \
353
                OP(LP(dst),no_rnd_PACK(a0,a1,a2,a3)); \
354
                UNPACK(a0,a1,AV_RN32(src1+4),LP(src2+4)); \
355
                UNPACK(a2,a3,LP(src3+4),LP(src4+4)); \
356
                OP(LP(dst+4),no_rnd_PACK(a0,a1,a2,a3)); \
357
                UNPACK(a0,a1,AV_RN32(src1+8),LP(src2+8)); \
358
                UNPACK(a2,a3,LP(src3+8),LP(src4+8)); \
359
                OP(LP(dst+8),no_rnd_PACK(a0,a1,a2,a3)); \
360
                UNPACK(a0,a1,AV_RN32(src1+12),LP(src2+12)); \
361
                UNPACK(a2,a3,LP(src3+12),LP(src4+12)); \
362
                OP(LP(dst+12),no_rnd_PACK(a0,a1,a2,a3)); \
363
                src1+=src_stride1;\
364
                src2+=src_stride2;\
365
                src3+=src_stride3;\
366
                src4+=src_stride4;\
367
                dst+=dst_stride;\
368
        } while(--h); \
369
} \
370
\
371

    
372
#define op_avg(a, b) a = rnd_avg32(a,b)
373
#define op_put(a, b) a = b
374

    
375
PIXOP2(avg, op_avg)
376
PIXOP2(put, op_put)
377
#undef op_avg
378
#undef op_put
379

    
380
#define avg2(a,b) ((a+b+1)>>1)
381
#define avg4(a,b,c,d) ((a+b+c+d+2)>>2)
382

    
383

    
384
static void gmc1_c(uint8_t *dst, uint8_t *src, int stride, int h, int x16, int y16, int rounder)
385
{
386
    const int A=(16-x16)*(16-y16);
387
    const int B=(   x16)*(16-y16);
388
    const int C=(16-x16)*(   y16);
389
    const int D=(   x16)*(   y16);
390

    
391
    do {
392
        int t0,t1,t2,t3;
393
        uint8_t *s0 = src;
394
        uint8_t *s1 = src+stride;
395
        t0 = *s0++; t2 = *s1++;
396
        t1 = *s0++; t3 = *s1++;
397
        dst[0]= (A*t0 + B*t1 + C*t2 + D*t3 + rounder)>>8;
398
        t0 = *s0++; t2 = *s1++;
399
        dst[1]= (A*t1 + B*t0 + C*t3 + D*t2 + rounder)>>8;
400
        t1 = *s0++; t3 = *s1++;
401
        dst[2]= (A*t0 + B*t1 + C*t2 + D*t3 + rounder)>>8;
402
        t0 = *s0++; t2 = *s1++;
403
        dst[3]= (A*t1 + B*t0 + C*t3 + D*t2 + rounder)>>8;
404
        t1 = *s0++; t3 = *s1++;
405
        dst[4]= (A*t0 + B*t1 + C*t2 + D*t3 + rounder)>>8;
406
        t0 = *s0++; t2 = *s1++;
407
        dst[5]= (A*t1 + B*t0 + C*t3 + D*t2 + rounder)>>8;
408
        t1 = *s0++; t3 = *s1++;
409
        dst[6]= (A*t0 + B*t1 + C*t2 + D*t3 + rounder)>>8;
410
        t0 = *s0++; t2 = *s1++;
411
        dst[7]= (A*t1 + B*t0 + C*t3 + D*t2 + rounder)>>8;
412
        dst+= stride;
413
        src+= stride;
414
    }while(--h);
415
}
416

    
417
static void gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,
418
                  int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height)
419
{
420
    int y, vx, vy;
421
    const int s= 1<<shift;
422

    
423
    width--;
424
    height--;
425

    
426
    for(y=0; y<h; y++){
427
        int x;
428

    
429
        vx= ox;
430
        vy= oy;
431
        for(x=0; x<8; x++){ //XXX FIXME optimize
432
            int src_x, src_y, frac_x, frac_y, index;
433

    
434
            src_x= vx>>16;
435
            src_y= vy>>16;
436
            frac_x= src_x&(s-1);
437
            frac_y= src_y&(s-1);
438
            src_x>>=shift;
439
            src_y>>=shift;
440

    
441
            if((unsigned)src_x < width){
442
                if((unsigned)src_y < height){
443
                    index= src_x + src_y*stride;
444
                    dst[y*stride + x]= (  (  src[index         ]*(s-frac_x)
445
                                           + src[index       +1]*   frac_x )*(s-frac_y)
446
                                        + (  src[index+stride  ]*(s-frac_x)
447
                                           + src[index+stride+1]*   frac_x )*   frac_y
448
                                        + r)>>(shift*2);
449
                }else{
450
                    index= src_x + av_clip(src_y, 0, height)*stride;
451
                    dst[y*stride + x]= ( (  src[index         ]*(s-frac_x)
452
                                          + src[index       +1]*   frac_x )*s
453
                                        + r)>>(shift*2);
454
                }
455
            }else{
456
                if((unsigned)src_y < height){
457
                    index= av_clip(src_x, 0, width) + src_y*stride;
458
                    dst[y*stride + x]= (  (  src[index         ]*(s-frac_y)
459
                                           + src[index+stride  ]*   frac_y )*s
460
                                        + r)>>(shift*2);
461
                }else{
462
                    index= av_clip(src_x, 0, width) + av_clip(src_y, 0, height)*stride;
463
                    dst[y*stride + x]=    src[index         ];
464
                }
465
            }
466

    
467
            vx+= dxx;
468
            vy+= dyx;
469
        }
470
        ox += dxy;
471
        oy += dyy;
472
    }
473
}
474
#define H264_CHROMA_MC(OPNAME, OP)\
475
static void OPNAME ## h264_chroma_mc2_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\
476
    const int A=(8-x)*(8-y);\
477
    const int B=(  x)*(8-y);\
478
    const int C=(8-x)*(  y);\
479
    const int D=(  x)*(  y);\
480
    \
481
    assert(x<8 && y<8 && x>=0 && y>=0);\
482
\
483
    do {\
484
        int t0,t1,t2,t3; \
485
        uint8_t *s0 = src; \
486
        uint8_t *s1 = src+stride; \
487
        t0 = *s0++; t2 = *s1++; \
488
        t1 = *s0++; t3 = *s1++; \
489
        OP(dst[0], (A*t0 + B*t1 + C*t2 + D*t3));\
490
        t0 = *s0++; t2 = *s1++; \
491
        OP(dst[1], (A*t1 + B*t0 + C*t3 + D*t2));\
492
        dst+= stride;\
493
        src+= stride;\
494
    }while(--h);\
495
}\
496
\
497
static void OPNAME ## h264_chroma_mc4_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\
498
    const int A=(8-x)*(8-y);\
499
    const int B=(  x)*(8-y);\
500
    const int C=(8-x)*(  y);\
501
    const int D=(  x)*(  y);\
502
    \
503
    assert(x<8 && y<8 && x>=0 && y>=0);\
504
\
505
    do {\
506
        int t0,t1,t2,t3; \
507
        uint8_t *s0 = src; \
508
        uint8_t *s1 = src+stride; \
509
        t0 = *s0++; t2 = *s1++; \
510
        t1 = *s0++; t3 = *s1++; \
511
        OP(dst[0], (A*t0 + B*t1 + C*t2 + D*t3));\
512
        t0 = *s0++; t2 = *s1++; \
513
        OP(dst[1], (A*t1 + B*t0 + C*t3 + D*t2));\
514
        t1 = *s0++; t3 = *s1++; \
515
        OP(dst[2], (A*t0 + B*t1 + C*t2 + D*t3));\
516
        t0 = *s0++; t2 = *s1++; \
517
        OP(dst[3], (A*t1 + B*t0 + C*t3 + D*t2));\
518
        dst+= stride;\
519
        src+= stride;\
520
    }while(--h);\
521
}\
522
\
523
static void OPNAME ## h264_chroma_mc8_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\
524
    const int A=(8-x)*(8-y);\
525
    const int B=(  x)*(8-y);\
526
    const int C=(8-x)*(  y);\
527
    const int D=(  x)*(  y);\
528
    \
529
    assert(x<8 && y<8 && x>=0 && y>=0);\
530
\
531
    do {\
532
        int t0,t1,t2,t3; \
533
        uint8_t *s0 = src; \
534
        uint8_t *s1 = src+stride; \
535
        t0 = *s0++; t2 = *s1++; \
536
        t1 = *s0++; t3 = *s1++; \
537
        OP(dst[0], (A*t0 + B*t1 + C*t2 + D*t3));\
538
        t0 = *s0++; t2 = *s1++; \
539
        OP(dst[1], (A*t1 + B*t0 + C*t3 + D*t2));\
540
        t1 = *s0++; t3 = *s1++; \
541
        OP(dst[2], (A*t0 + B*t1 + C*t2 + D*t3));\
542
        t0 = *s0++; t2 = *s1++; \
543
        OP(dst[3], (A*t1 + B*t0 + C*t3 + D*t2));\
544
        t1 = *s0++; t3 = *s1++; \
545
        OP(dst[4], (A*t0 + B*t1 + C*t2 + D*t3));\
546
        t0 = *s0++; t2 = *s1++; \
547
        OP(dst[5], (A*t1 + B*t0 + C*t3 + D*t2));\
548
        t1 = *s0++; t3 = *s1++; \
549
        OP(dst[6], (A*t0 + B*t1 + C*t2 + D*t3));\
550
        t0 = *s0++; t2 = *s1++; \
551
        OP(dst[7], (A*t1 + B*t0 + C*t3 + D*t2));\
552
        dst+= stride;\
553
        src+= stride;\
554
    }while(--h);\
555
}
556

    
557
#define op_avg(a, b) a = (((a)+(((b) + 32)>>6)+1)>>1)
558
#define op_put(a, b) a = (((b) + 32)>>6)
559

    
560
H264_CHROMA_MC(put_       , op_put)
561
H264_CHROMA_MC(avg_       , op_avg)
562
#undef op_avg
563
#undef op_put
564

    
565
#define QPEL_MC(r, OPNAME, RND, OP) \
566
static void OPNAME ## mpeg4_qpel8_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
        src0= *s++;\
572
        src1= *s++;\
573
        src2= *s++;\
574
        src3= *s++;\
575
        src4= *s++;\
576
        OP(dst[0], (src0+src1)*20 - (src0+src2)*6 + (src1+src3)*3 - (src2+src4));\
577
        src5= *s++;\
578
        OP(dst[1], (src1+src2)*20 - (src0+src3)*6 + (src0+src4)*3 - (src1+src5));\
579
        src6= *s++;\
580
        OP(dst[2], (src2+src3)*20 - (src1+src4)*6 + (src0+src5)*3 - (src0+src6));\
581
        src7= *s++;\
582
        OP(dst[3], (src3+src4)*20 - (src2+src5)*6 + (src1+src6)*3 - (src0+src7));\
583
        src8= *s++;\
584
        OP(dst[4], (src4+src5)*20 - (src3+src6)*6 + (src2+src7)*3 - (src1+src8));\
585
        OP(dst[5], (src5+src6)*20 - (src4+src7)*6 + (src3+src8)*3 - (src2+src8));\
586
        OP(dst[6], (src6+src7)*20 - (src5+src8)*6 + (src4+src8)*3 - (src3+src7));\
587
        OP(dst[7], (src7+src8)*20 - (src6+src8)*6 + (src5+src7)*3 - (src4+src6));\
588
        dst+=dstStride;\
589
        src+=srcStride;\
590
    }while(--h);\
591
}\
592
\
593
static void OPNAME ## mpeg4_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
594
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
595
    int w=8;\
596
    do{\
597
        uint8_t *s = src, *d=dst;\
598
        int src0,src1,src2,src3,src4,src5,src6,src7,src8;\
599
        src0 = *s; s+=srcStride; \
600
        src1 = *s; s+=srcStride; \
601
        src2 = *s; s+=srcStride; \
602
        src3 = *s; s+=srcStride; \
603
        src4 = *s; s+=srcStride; \
604
        OP(*d, (src0+src1)*20 - (src0+src2)*6 + (src1+src3)*3 - (src2+src4));d+=dstStride;\
605
        src5 = *s; s+=srcStride; \
606
        OP(*d, (src1+src2)*20 - (src0+src3)*6 + (src0+src4)*3 - (src1+src5));d+=dstStride;\
607
        src6 = *s; s+=srcStride; \
608
        OP(*d, (src2+src3)*20 - (src1+src4)*6 + (src0+src5)*3 - (src0+src6));d+=dstStride;\
609
        src7 = *s; s+=srcStride; \
610
        OP(*d, (src3+src4)*20 - (src2+src5)*6 + (src1+src6)*3 - (src0+src7));d+=dstStride;\
611
        src8 = *s; \
612
        OP(*d, (src4+src5)*20 - (src3+src6)*6 + (src2+src7)*3 - (src1+src8));d+=dstStride;\
613
        OP(*d, (src5+src6)*20 - (src4+src7)*6 + (src3+src8)*3 - (src2+src8));d+=dstStride;\
614
        OP(*d, (src6+src7)*20 - (src5+src8)*6 + (src4+src8)*3 - (src3+src7));d+=dstStride;\
615
        OP(*d, (src7+src8)*20 - (src6+src8)*6 + (src5+src7)*3 - (src4+src6));\
616
        dst++;\
617
        src++;\
618
    }while(--w);\
619
}\
620
\
621
static void OPNAME ## mpeg4_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
622
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
623
    do {\
624
        uint8_t *s = src;\
625
        int src0,src1,src2,src3,src4,src5,src6,src7,src8;\
626
        int src9,src10,src11,src12,src13,src14,src15,src16;\
627
        src0= *s++;\
628
        src1= *s++;\
629
        src2= *s++;\
630
        src3= *s++;\
631
        src4= *s++;\
632
        OP(dst[ 0], (src0 +src1 )*20 - (src0 +src2 )*6 + (src1 +src3 )*3 - (src2 +src4 ));\
633
        src5= *s++;\
634
        OP(dst[ 1], (src1 +src2 )*20 - (src0 +src3 )*6 + (src0 +src4 )*3 - (src1 +src5 ));\
635
        src6= *s++;\
636
        OP(dst[ 2], (src2 +src3 )*20 - (src1 +src4 )*6 + (src0 +src5 )*3 - (src0 +src6 ));\
637
        src7= *s++;\
638
        OP(dst[ 3], (src3 +src4 )*20 - (src2 +src5 )*6 + (src1 +src6 )*3 - (src0 +src7 ));\
639
        src8= *s++;\
640
        OP(dst[ 4], (src4 +src5 )*20 - (src3 +src6 )*6 + (src2 +src7 )*3 - (src1 +src8 ));\
641
        src9= *s++;\
642
        OP(dst[ 5], (src5 +src6 )*20 - (src4 +src7 )*6 + (src3 +src8 )*3 - (src2 +src9 ));\
643
        src10= *s++;\
644
        OP(dst[ 6], (src6 +src7 )*20 - (src5 +src8 )*6 + (src4 +src9 )*3 - (src3 +src10));\
645
        src11= *s++;\
646
        OP(dst[ 7], (src7 +src8 )*20 - (src6 +src9 )*6 + (src5 +src10)*3 - (src4 +src11));\
647
        src12= *s++;\
648
        OP(dst[ 8], (src8 +src9 )*20 - (src7 +src10)*6 + (src6 +src11)*3 - (src5 +src12));\
649
        src13= *s++;\
650
        OP(dst[ 9], (src9 +src10)*20 - (src8 +src11)*6 + (src7 +src12)*3 - (src6 +src13));\
651
        src14= *s++;\
652
        OP(dst[10], (src10+src11)*20 - (src9 +src12)*6 + (src8 +src13)*3 - (src7 +src14));\
653
        src15= *s++;\
654
        OP(dst[11], (src11+src12)*20 - (src10+src13)*6 + (src9 +src14)*3 - (src8 +src15));\
655
        src16= *s++;\
656
        OP(dst[12], (src12+src13)*20 - (src11+src14)*6 + (src10+src15)*3 - (src9 +src16));\
657
        OP(dst[13], (src13+src14)*20 - (src12+src15)*6 + (src11+src16)*3 - (src10+src16));\
658
        OP(dst[14], (src14+src15)*20 - (src13+src16)*6 + (src12+src16)*3 - (src11+src15));\
659
        OP(dst[15], (src15+src16)*20 - (src14+src16)*6 + (src13+src15)*3 - (src12+src14));\
660
        dst+=dstStride;\
661
        src+=srcStride;\
662
    }while(--h);\
663
}\
664
\
665
static void OPNAME ## mpeg4_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
666
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
667
    int w=16;\
668
    do {\
669
        uint8_t *s = src, *d=dst;\
670
        int src0,src1,src2,src3,src4,src5,src6,src7,src8;\
671
        int src9,src10,src11,src12,src13,src14,src15,src16;\
672
        src0 = *s; s+=srcStride; \
673
        src1 = *s; s+=srcStride; \
674
        src2 = *s; s+=srcStride; \
675
        src3 = *s; s+=srcStride; \
676
        src4 = *s; s+=srcStride; \
677
        OP(*d, (src0 +src1 )*20 - (src0 +src2 )*6 + (src1 +src3 )*3 - (src2 +src4 ));d+=dstStride;\
678
        src5 = *s; s+=srcStride; \
679
        OP(*d, (src1 +src2 )*20 - (src0 +src3 )*6 + (src0 +src4 )*3 - (src1 +src5 ));d+=dstStride;\
680
        src6 = *s; s+=srcStride; \
681
        OP(*d, (src2 +src3 )*20 - (src1 +src4 )*6 + (src0 +src5 )*3 - (src0 +src6 ));d+=dstStride;\
682
        src7 = *s; s+=srcStride; \
683
        OP(*d, (src3 +src4 )*20 - (src2 +src5 )*6 + (src1 +src6 )*3 - (src0 +src7 ));d+=dstStride;\
684
        src8 = *s; s+=srcStride; \
685
        OP(*d, (src4 +src5 )*20 - (src3 +src6 )*6 + (src2 +src7 )*3 - (src1 +src8 ));d+=dstStride;\
686
        src9 = *s; s+=srcStride; \
687
        OP(*d, (src5 +src6 )*20 - (src4 +src7 )*6 + (src3 +src8 )*3 - (src2 +src9 ));d+=dstStride;\
688
        src10 = *s; s+=srcStride; \
689
        OP(*d, (src6 +src7 )*20 - (src5 +src8 )*6 + (src4 +src9 )*3 - (src3 +src10));d+=dstStride;\
690
        src11 = *s; s+=srcStride; \
691
        OP(*d, (src7 +src8 )*20 - (src6 +src9 )*6 + (src5 +src10)*3 - (src4 +src11));d+=dstStride;\
692
        src12 = *s; s+=srcStride; \
693
        OP(*d, (src8 +src9 )*20 - (src7 +src10)*6 + (src6 +src11)*3 - (src5 +src12));d+=dstStride;\
694
        src13 = *s; s+=srcStride; \
695
        OP(*d, (src9 +src10)*20 - (src8 +src11)*6 + (src7 +src12)*3 - (src6 +src13));d+=dstStride;\
696
        src14 = *s; s+=srcStride; \
697
        OP(*d, (src10+src11)*20 - (src9 +src12)*6 + (src8 +src13)*3 - (src7 +src14));d+=dstStride;\
698
        src15 = *s; s+=srcStride; \
699
        OP(*d, (src11+src12)*20 - (src10+src13)*6 + (src9 +src14)*3 - (src8 +src15));d+=dstStride;\
700
        src16 = *s; \
701
        OP(*d, (src12+src13)*20 - (src11+src14)*6 + (src10+src15)*3 - (src9 +src16));d+=dstStride;\
702
        OP(*d, (src13+src14)*20 - (src12+src15)*6 + (src11+src16)*3 - (src10+src16));d+=dstStride;\
703
        OP(*d, (src14+src15)*20 - (src13+src16)*6 + (src12+src16)*3 - (src11+src15));d+=dstStride;\
704
        OP(*d, (src15+src16)*20 - (src14+src16)*6 + (src13+src15)*3 - (src12+src14));\
705
        dst++;\
706
        src++;\
707
    }while(--w);\
708
}\
709
\
710
static void OPNAME ## qpel8_mc00_c (uint8_t *dst, uint8_t *src, int stride){\
711
    OPNAME ## pixels8_c(dst, src, stride, 8);\
712
}\
713
\
714
static void OPNAME ## qpel8_mc10_c(uint8_t *dst, uint8_t *src, int stride){\
715
    uint8_t half[64];\
716
    put ## RND ## mpeg4_qpel8_h_lowpass(half, src, 8, stride, 8);\
717
    OPNAME ## pixels8_l2_aligned2(dst, src, half, stride, stride, 8, 8);\
718
}\
719
\
720
static void OPNAME ## qpel8_mc20_c(uint8_t *dst, uint8_t *src, int stride){\
721
    OPNAME ## mpeg4_qpel8_h_lowpass(dst, src, stride, stride, 8);\
722
}\
723
\
724
static void OPNAME ## qpel8_mc30_c(uint8_t *dst, uint8_t *src, int stride){\
725
    uint8_t half[64];\
726
    put ## RND ## mpeg4_qpel8_h_lowpass(half, src, 8, stride, 8);\
727
    OPNAME ## pixels8_l2_aligned2(dst, src+1, half, stride, stride, 8, 8);\
728
}\
729
\
730
static void OPNAME ## qpel8_mc01_c(uint8_t *dst, uint8_t *src, int stride){\
731
    uint8_t full[16*9];\
732
    uint8_t half[64];\
733
    copy_block9(full, src, 16, stride, 9);\
734
    put ## RND ## mpeg4_qpel8_v_lowpass(half, full, 8, 16);\
735
    OPNAME ## pixels8_l2_aligned(dst, full, half, stride, 16, 8, 8);\
736
}\
737
\
738
static void OPNAME ## qpel8_mc02_c(uint8_t *dst, uint8_t *src, int stride){\
739
    uint8_t full[16*9];\
740
    copy_block9(full, src, 16, stride, 9);\
741
    OPNAME ## mpeg4_qpel8_v_lowpass(dst, full, stride, 16);\
742
}\
743
\
744
static void OPNAME ## qpel8_mc03_c(uint8_t *dst, uint8_t *src, int stride){\
745
    uint8_t full[16*9];\
746
    uint8_t half[64];\
747
    copy_block9(full, src, 16, stride, 9);\
748
    put ## RND ## mpeg4_qpel8_v_lowpass(half, full, 8, 16);\
749
    OPNAME ## pixels8_l2_aligned(dst, full+16, half, stride, 16, 8, 8);\
750
}\
751
static void ff_ ## OPNAME ## qpel8_mc11_old_c(uint8_t *dst, uint8_t *src, int stride){\
752
    uint8_t full[16*9];\
753
    uint8_t halfH[72];\
754
    uint8_t halfV[64];\
755
    uint8_t halfHV[64];\
756
    copy_block9(full, src, 16, stride, 9);\
757
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
758
    put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full, 8, 16);\
759
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
760
    OPNAME ## pixels8_l4_aligned(dst, full, halfH, halfV, halfHV, stride, 16, 8, 8, 8, 8);\
761
}\
762
static void OPNAME ## qpel8_mc11_c(uint8_t *dst, uint8_t *src, int stride){\
763
    uint8_t full[16*9];\
764
    uint8_t halfH[72];\
765
    uint8_t halfHV[64];\
766
    copy_block9(full, src, 16, stride, 9);\
767
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
768
    put ## RND ## pixels8_l2_aligned(halfH, halfH, full, 8, 8, 16, 9);\
769
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
770
    OPNAME ## pixels8_l2_aligned(dst, halfH, halfHV, stride, 8, 8, 8);\
771
}\
772
static void ff_ ## OPNAME ## qpel8_mc31_old_c(uint8_t *dst, uint8_t *src, int stride){\
773
    uint8_t full[16*9];\
774
    uint8_t halfH[72];\
775
    uint8_t halfV[64];\
776
    uint8_t halfHV[64];\
777
    copy_block9(full, src, 16, stride, 9);\
778
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
779
    put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full+1, 8, 16);\
780
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
781
    OPNAME ## pixels8_l4_aligned0(dst, full+1, halfH, halfV, halfHV, stride, 16, 8, 8, 8, 8);\
782
}\
783
static void OPNAME ## qpel8_mc31_c(uint8_t *dst, uint8_t *src, int stride){\
784
    uint8_t full[16*9];\
785
    uint8_t halfH[72];\
786
    uint8_t halfHV[64];\
787
    copy_block9(full, src, 16, stride, 9);\
788
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
789
    put ## RND ## pixels8_l2_aligned1(halfH, halfH, full+1, 8, 8, 16, 9);\
790
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
791
    OPNAME ## pixels8_l2_aligned(dst, halfH, halfHV, stride, 8, 8, 8);\
792
}\
793
static void ff_ ## OPNAME ## qpel8_mc13_old_c(uint8_t *dst, uint8_t *src, int stride){\
794
    uint8_t full[16*9];\
795
    uint8_t halfH[72];\
796
    uint8_t halfV[64];\
797
    uint8_t halfHV[64];\
798
    copy_block9(full, src, 16, stride, 9);\
799
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
800
    put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full, 8, 16);\
801
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
802
    OPNAME ## pixels8_l4_aligned(dst, full+16, halfH+8, halfV, halfHV, stride, 16, 8, 8, 8, 8);\
803
}\
804
static void OPNAME ## qpel8_mc13_c(uint8_t *dst, uint8_t *src, int stride){\
805
    uint8_t full[16*9];\
806
    uint8_t halfH[72];\
807
    uint8_t halfHV[64];\
808
    copy_block9(full, src, 16, stride, 9);\
809
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
810
    put ## RND ## pixels8_l2_aligned(halfH, halfH, full, 8, 8, 16, 9);\
811
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
812
    OPNAME ## pixels8_l2_aligned(dst, halfH+8, halfHV, stride, 8, 8, 8);\
813
}\
814
static void ff_ ## OPNAME ## qpel8_mc33_old_c(uint8_t *dst, uint8_t *src, int stride){\
815
    uint8_t full[16*9];\
816
    uint8_t halfH[72];\
817
    uint8_t halfV[64];\
818
    uint8_t halfHV[64];\
819
    copy_block9(full, src, 16, stride, 9);\
820
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full  , 8, 16, 9);\
821
    put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full+1, 8, 16);\
822
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
823
    OPNAME ## pixels8_l4_aligned0(dst, full+17, halfH+8, halfV, halfHV, stride, 16, 8, 8, 8, 8);\
824
}\
825
static void OPNAME ## qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride){\
826
    uint8_t full[16*9];\
827
    uint8_t halfH[72];\
828
    uint8_t halfHV[64];\
829
    copy_block9(full, src, 16, stride, 9);\
830
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
831
    put ## RND ## pixels8_l2_aligned1(halfH, halfH, full+1, 8, 8, 16, 9);\
832
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
833
    OPNAME ## pixels8_l2_aligned(dst, halfH+8, halfHV, stride, 8, 8, 8);\
834
}\
835
static void OPNAME ## qpel8_mc21_c(uint8_t *dst, uint8_t *src, int stride){\
836
    uint8_t halfH[72];\
837
    uint8_t halfHV[64];\
838
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\
839
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
840
    OPNAME ## pixels8_l2_aligned(dst, halfH, halfHV, stride, 8, 8, 8);\
841
}\
842
static void OPNAME ## qpel8_mc23_c(uint8_t *dst, uint8_t *src, int stride){\
843
    uint8_t halfH[72];\
844
    uint8_t halfHV[64];\
845
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\
846
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
847
    OPNAME ## pixels8_l2_aligned(dst, halfH+8, halfHV, stride, 8, 8, 8);\
848
}\
849
static void ff_ ## OPNAME ## qpel8_mc12_old_c(uint8_t *dst, uint8_t *src, int stride){\
850
    uint8_t full[16*9];\
851
    uint8_t halfH[72];\
852
    uint8_t halfV[64];\
853
    uint8_t halfHV[64];\
854
    copy_block9(full, src, 16, stride, 9);\
855
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
856
    put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full, 8, 16);\
857
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
858
    OPNAME ## pixels8_l2_aligned(dst, halfV, halfHV, stride, 8, 8, 8);\
859
}\
860
static void OPNAME ## qpel8_mc12_c(uint8_t *dst, uint8_t *src, int stride){\
861
    uint8_t full[16*9];\
862
    uint8_t halfH[72];\
863
    copy_block9(full, src, 16, stride, 9);\
864
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
865
    put ## RND ## pixels8_l2_aligned(halfH, halfH, full, 8, 8, 16, 9);\
866
    OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\
867
}\
868
static void ff_ ## OPNAME ## qpel8_mc32_old_c(uint8_t *dst, uint8_t *src, int stride){\
869
    uint8_t full[16*9];\
870
    uint8_t halfH[72];\
871
    uint8_t halfV[64];\
872
    uint8_t halfHV[64];\
873
    copy_block9(full, src, 16, stride, 9);\
874
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
875
    put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full+1, 8, 16);\
876
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
877
    OPNAME ## pixels8_l2_aligned(dst, halfV, halfHV, stride, 8, 8, 8);\
878
}\
879
static void OPNAME ## qpel8_mc32_c(uint8_t *dst, uint8_t *src, int stride){\
880
    uint8_t full[16*9];\
881
    uint8_t halfH[72];\
882
    copy_block9(full, src, 16, stride, 9);\
883
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
884
    put ## RND ## pixels8_l2_aligned1(halfH, halfH, full+1, 8, 8, 16, 9);\
885
    OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\
886
}\
887
static void OPNAME ## qpel8_mc22_c(uint8_t *dst, uint8_t *src, int stride){\
888
    uint8_t halfH[72];\
889
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\
890
    OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\
891
}\
892
static void OPNAME ## qpel16_mc00_c (uint8_t *dst, uint8_t *src, int stride){\
893
    OPNAME ## pixels16_c(dst, src, stride, 16);\
894
}\
895
\
896
static void OPNAME ## qpel16_mc10_c(uint8_t *dst, uint8_t *src, int stride){\
897
    uint8_t half[256];\
898
    put ## RND ## mpeg4_qpel16_h_lowpass(half, src, 16, stride, 16);\
899
    OPNAME ## pixels16_l2_aligned2(dst, src, half, stride, stride, 16, 16);\
900
}\
901
\
902
static void OPNAME ## qpel16_mc20_c(uint8_t *dst, uint8_t *src, int stride){\
903
    OPNAME ## mpeg4_qpel16_h_lowpass(dst, src, stride, stride, 16);\
904
}\
905
\
906
static void OPNAME ## qpel16_mc30_c(uint8_t *dst, uint8_t *src, int stride){\
907
    uint8_t half[256];\
908
    put ## RND ## mpeg4_qpel16_h_lowpass(half, src, 16, stride, 16);\
909
    OPNAME ## pixels16_l2_aligned2(dst, src+1, half, stride, stride, 16, 16);\
910
}\
911
\
912
static void OPNAME ## qpel16_mc01_c(uint8_t *dst, uint8_t *src, int stride){\
913
    uint8_t full[24*17];\
914
    uint8_t half[256];\
915
    copy_block17(full, src, 24, stride, 17);\
916
    put ## RND ## mpeg4_qpel16_v_lowpass(half, full, 16, 24);\
917
    OPNAME ## pixels16_l2_aligned(dst, full, half, stride, 24, 16, 16);\
918
}\
919
\
920
static void OPNAME ## qpel16_mc02_c(uint8_t *dst, uint8_t *src, int stride){\
921
    uint8_t full[24*17];\
922
    copy_block17(full, src, 24, stride, 17);\
923
    OPNAME ## mpeg4_qpel16_v_lowpass(dst, full, stride, 24);\
924
}\
925
\
926
static void OPNAME ## qpel16_mc03_c(uint8_t *dst, uint8_t *src, int stride){\
927
    uint8_t full[24*17];\
928
    uint8_t half[256];\
929
    copy_block17(full, src, 24, stride, 17);\
930
    put ## RND ## mpeg4_qpel16_v_lowpass(half, full, 16, 24);\
931
    OPNAME ## pixels16_l2_aligned(dst, full+24, half, stride, 24, 16, 16);\
932
}\
933
static void ff_ ## OPNAME ## qpel16_mc11_old_c(uint8_t *dst, uint8_t *src, int stride){\
934
    uint8_t full[24*17];\
935
    uint8_t halfH[272];\
936
    uint8_t halfV[256];\
937
    uint8_t halfHV[256];\
938
    copy_block17(full, src, 24, stride, 17);\
939
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
940
    put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full, 16, 24);\
941
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
942
    OPNAME ## pixels16_l4_aligned(dst, full, halfH, halfV, halfHV, stride, 24, 16, 16, 16, 16);\
943
}\
944
static void OPNAME ## qpel16_mc11_c(uint8_t *dst, uint8_t *src, int stride){\
945
    uint8_t full[24*17];\
946
    uint8_t halfH[272];\
947
    uint8_t halfHV[256];\
948
    copy_block17(full, src, 24, stride, 17);\
949
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
950
    put ## RND ## pixels16_l2_aligned(halfH, halfH, full, 16, 16, 24, 17);\
951
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
952
    OPNAME ## pixels16_l2_aligned(dst, halfH, halfHV, stride, 16, 16, 16);\
953
}\
954
static void ff_ ## OPNAME ## qpel16_mc31_old_c(uint8_t *dst, uint8_t *src, int stride){\
955
    uint8_t full[24*17];\
956
    uint8_t halfH[272];\
957
    uint8_t halfV[256];\
958
    uint8_t halfHV[256];\
959
    copy_block17(full, src, 24, stride, 17);\
960
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
961
    put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full+1, 16, 24);\
962
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
963
    OPNAME ## pixels16_l4_aligned0(dst, full+1, halfH, halfV, halfHV, stride, 24, 16, 16, 16, 16);\
964
}\
965
static void OPNAME ## qpel16_mc31_c(uint8_t *dst, uint8_t *src, int stride){\
966
    uint8_t full[24*17];\
967
    uint8_t halfH[272];\
968
    uint8_t halfHV[256];\
969
    copy_block17(full, src, 24, stride, 17);\
970
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
971
    put ## RND ## pixels16_l2_aligned1(halfH, halfH, full+1, 16, 16, 24, 17);\
972
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
973
    OPNAME ## pixels16_l2_aligned(dst, halfH, halfHV, stride, 16, 16, 16);\
974
}\
975
static void ff_ ## OPNAME ## qpel16_mc13_old_c(uint8_t *dst, uint8_t *src, int stride){\
976
    uint8_t full[24*17];\
977
    uint8_t halfH[272];\
978
    uint8_t halfV[256];\
979
    uint8_t halfHV[256];\
980
    copy_block17(full, src, 24, stride, 17);\
981
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
982
    put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full, 16, 24);\
983
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
984
    OPNAME ## pixels16_l4_aligned(dst, full+24, halfH+16, halfV, halfHV, stride, 24, 16, 16, 16, 16);\
985
}\
986
static void OPNAME ## qpel16_mc13_c(uint8_t *dst, uint8_t *src, int stride){\
987
    uint8_t full[24*17];\
988
    uint8_t halfH[272];\
989
    uint8_t halfHV[256];\
990
    copy_block17(full, src, 24, stride, 17);\
991
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
992
    put ## RND ## pixels16_l2_aligned(halfH, halfH, full, 16, 16, 24, 17);\
993
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
994
    OPNAME ## pixels16_l2_aligned(dst, halfH+16, halfHV, stride, 16, 16, 16);\
995
}\
996
static void ff_ ## OPNAME ## qpel16_mc33_old_c(uint8_t *dst, uint8_t *src, int stride){\
997
    uint8_t full[24*17];\
998
    uint8_t halfH[272];\
999
    uint8_t halfV[256];\
1000
    uint8_t halfHV[256];\
1001
    copy_block17(full, src, 24, stride, 17);\
1002
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full  , 16, 24, 17);\
1003
    put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full+1, 16, 24);\
1004
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1005
    OPNAME ## pixels16_l4_aligned0(dst, full+25, halfH+16, halfV, halfHV, stride, 24, 16, 16, 16, 16);\
1006
}\
1007
static void OPNAME ## qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride){\
1008
    uint8_t full[24*17];\
1009
    uint8_t halfH[272];\
1010
    uint8_t halfHV[256];\
1011
    copy_block17(full, src, 24, stride, 17);\
1012
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1013
    put ## RND ## pixels16_l2_aligned1(halfH, halfH, full+1, 16, 16, 24, 17);\
1014
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1015
    OPNAME ## pixels16_l2_aligned(dst, halfH+16, halfHV, stride, 16, 16, 16);\
1016
}\
1017
static void OPNAME ## qpel16_mc21_c(uint8_t *dst, uint8_t *src, int stride){\
1018
    uint8_t halfH[272];\
1019
    uint8_t halfHV[256];\
1020
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\
1021
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1022
    OPNAME ## pixels16_l2_aligned(dst, halfH, halfHV, stride, 16, 16, 16);\
1023
}\
1024
static void OPNAME ## qpel16_mc23_c(uint8_t *dst, uint8_t *src, int stride){\
1025
    uint8_t halfH[272];\
1026
    uint8_t halfHV[256];\
1027
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\
1028
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1029
    OPNAME ## pixels16_l2_aligned(dst, halfH+16, halfHV, stride, 16, 16, 16);\
1030
}\
1031
static void ff_ ## OPNAME ## qpel16_mc12_old_c(uint8_t *dst, uint8_t *src, int stride){\
1032
    uint8_t full[24*17];\
1033
    uint8_t halfH[272];\
1034
    uint8_t halfV[256];\
1035
    uint8_t halfHV[256];\
1036
    copy_block17(full, src, 24, stride, 17);\
1037
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1038
    put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full, 16, 24);\
1039
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1040
    OPNAME ## pixels16_l2_aligned(dst, halfV, halfHV, stride, 16, 16, 16);\
1041
}\
1042
static void OPNAME ## qpel16_mc12_c(uint8_t *dst, uint8_t *src, int stride){\
1043
    uint8_t full[24*17];\
1044
    uint8_t halfH[272];\
1045
    copy_block17(full, src, 24, stride, 17);\
1046
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1047
    put ## RND ## pixels16_l2_aligned(halfH, halfH, full, 16, 16, 24, 17);\
1048
    OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\
1049
}\
1050
static void ff_ ## OPNAME ## qpel16_mc32_old_c(uint8_t *dst, uint8_t *src, int stride){\
1051
    uint8_t full[24*17];\
1052
    uint8_t halfH[272];\
1053
    uint8_t halfV[256];\
1054
    uint8_t halfHV[256];\
1055
    copy_block17(full, src, 24, stride, 17);\
1056
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1057
    put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full+1, 16, 24);\
1058
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
1059
    OPNAME ## pixels16_l2_aligned(dst, halfV, halfHV, stride, 16, 16, 16);\
1060
}\
1061
static void OPNAME ## qpel16_mc32_c(uint8_t *dst, uint8_t *src, int stride){\
1062
    uint8_t full[24*17];\
1063
    uint8_t halfH[272];\
1064
    copy_block17(full, src, 24, stride, 17);\
1065
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
1066
    put ## RND ## pixels16_l2_aligned1(halfH, halfH, full+1, 16, 16, 24, 17);\
1067
    OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\
1068
}\
1069
static void OPNAME ## qpel16_mc22_c(uint8_t *dst, uint8_t *src, int stride){\
1070
    uint8_t halfH[272];\
1071
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\
1072
    OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\
1073
}
1074

    
1075
#define op_avg(a, b) a = (((a)+cm[((b) + 16)>>5]+1)>>1)
1076
#define op_avg_no_rnd(a, b) a = (((a)+cm[((b) + 15)>>5])>>1)
1077
#define op_put(a, b) a = cm[((b) + 16)>>5]
1078
#define op_put_no_rnd(a, b) a = cm[((b) + 15)>>5]
1079

    
1080
QPEL_MC(0, put_       , _       , op_put)
1081
QPEL_MC(1, put_no_rnd_, _no_rnd_, op_put_no_rnd)
1082
QPEL_MC(0, avg_       , _       , op_avg)
1083
//QPEL_MC(1, avg_no_rnd , _       , op_avg)
1084
#undef op_avg
1085
#undef op_avg_no_rnd
1086
#undef op_put
1087
#undef op_put_no_rnd
1088

    
1089
#if 1
1090
#define H264_LOWPASS(OPNAME, OP, OP2) \
1091
static inline void OPNAME ## h264_qpel_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,int w,int h){\
1092
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
1093
    do {\
1094
        int srcB,srcA,src0,src1,src2,src3,src4,src5,src6;\
1095
        uint8_t *s = src-2;\
1096
        srcB = *s++;\
1097
        srcA = *s++;\
1098
        src0 = *s++;\
1099
        src1 = *s++;\
1100
        src2 = *s++;\
1101
        src3 = *s++;\
1102
        OP(dst[0], (src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));\
1103
        src4 = *s++;\
1104
        OP(dst[1], (src1+src2)*20 - (src0+src3)*5 + (srcA+src4));\
1105
        src5 = *s++;\
1106
        OP(dst[2], (src2+src3)*20 - (src1+src4)*5 + (src0+src5));\
1107
        src6 = *s++;\
1108
        OP(dst[3], (src3+src4)*20 - (src2+src5)*5 + (src1+src6));\
1109
      if (w>4) { /* it optimized */ \
1110
        int src7,src8,src9,src10; \
1111
        src7 = *s++;\
1112
        OP(dst[4], (src4+src5)*20 - (src3+src6)*5 + (src2+src7));\
1113
        src8 = *s++;\
1114
        OP(dst[5], (src5+src6)*20 - (src4+src7)*5 + (src3+src8));\
1115
        src9 = *s++;\
1116
        OP(dst[6], (src6+src7)*20 - (src5+src8)*5 + (src4+src9));\
1117
        src10 = *s++;\
1118
        OP(dst[7], (src7+src8)*20 - (src6+src9)*5 + (src5+src10));\
1119
       if (w>8) { \
1120
        int src11,src12,src13,src14,src15,src16,src17,src18; \
1121
        src11 = *s++;\
1122
        OP(dst[8] , (src8 +src9 )*20 - (src7 +src10)*5 + (src6 +src11));\
1123
        src12 = *s++;\
1124
        OP(dst[9] , (src9 +src10)*20 - (src8 +src11)*5 + (src7 +src12));\
1125
        src13 = *s++;\
1126
        OP(dst[10], (src10+src11)*20 - (src9 +src12)*5 + (src8 +src13));\
1127
        src14 = *s++;\
1128
        OP(dst[11], (src11+src12)*20 - (src10+src13)*5 + (src9 +src14));\
1129
        src15 = *s++;\
1130
        OP(dst[12], (src12+src13)*20 - (src11+src14)*5 + (src10+src15));\
1131
        src16 = *s++;\
1132
        OP(dst[13], (src13+src14)*20 - (src12+src15)*5 + (src11+src16));\
1133
        src17 = *s++;\
1134
        OP(dst[14], (src14+src15)*20 - (src13+src16)*5 + (src12+src17));\
1135
        src18 = *s++;\
1136
        OP(dst[15], (src15+src16)*20 - (src14+src17)*5 + (src13+src18));\
1137
       } \
1138
      } \
1139
        dst+=dstStride;\
1140
        src+=srcStride;\
1141
    }while(--h);\
1142
}\
1143
\
1144
static inline void OPNAME ## h264_qpel_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,int w,int h){\
1145
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
1146
    do{\
1147
        int srcB,srcA,src0,src1,src2,src3,src4,src5,src6;\
1148
        uint8_t *s = src-2*srcStride,*d=dst;\
1149
        srcB = *s; s+=srcStride;\
1150
        srcA = *s; s+=srcStride;\
1151
        src0 = *s; s+=srcStride;\
1152
        src1 = *s; s+=srcStride;\
1153
        src2 = *s; s+=srcStride;\
1154
        src3 = *s; s+=srcStride;\
1155
        OP(*d, (src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));d+=dstStride;\
1156
        src4 = *s; s+=srcStride;\
1157
        OP(*d, (src1+src2)*20 - (src0+src3)*5 + (srcA+src4));d+=dstStride;\
1158
        src5 = *s; s+=srcStride;\
1159
        OP(*d, (src2+src3)*20 - (src1+src4)*5 + (src0+src5));d+=dstStride;\
1160
        src6 = *s; s+=srcStride;\
1161
        OP(*d, (src3+src4)*20 - (src2+src5)*5 + (src1+src6));d+=dstStride;\
1162
      if (h>4) { \
1163
        int src7,src8,src9,src10; \
1164
        src7 = *s; s+=srcStride;\
1165
        OP(*d, (src4+src5)*20 - (src3+src6)*5 + (src2+src7));d+=dstStride;\
1166
        src8 = *s; s+=srcStride;\
1167
        OP(*d, (src5+src6)*20 - (src4+src7)*5 + (src3+src8));d+=dstStride;\
1168
        src9 = *s; s+=srcStride;\
1169
        OP(*d, (src6+src7)*20 - (src5+src8)*5 + (src4+src9));d+=dstStride;\
1170
        src10 = *s; s+=srcStride;\
1171
        OP(*d, (src7+src8)*20 - (src6+src9)*5 + (src5+src10));d+=dstStride;\
1172
       if (h>8) { \
1173
        int src11,src12,src13,src14,src15,src16,src17,src18; \
1174
        src11 = *s; s+=srcStride;\
1175
        OP(*d , (src8 +src9 )*20 - (src7 +src10)*5 + (src6 +src11));d+=dstStride;\
1176
        src12 = *s; s+=srcStride;\
1177
        OP(*d , (src9 +src10)*20 - (src8 +src11)*5 + (src7 +src12));d+=dstStride;\
1178
        src13 = *s; s+=srcStride;\
1179
        OP(*d, (src10+src11)*20 - (src9 +src12)*5 + (src8 +src13));d+=dstStride;\
1180
        src14 = *s; s+=srcStride;\
1181
        OP(*d, (src11+src12)*20 - (src10+src13)*5 + (src9 +src14));d+=dstStride;\
1182
        src15 = *s; s+=srcStride;\
1183
        OP(*d, (src12+src13)*20 - (src11+src14)*5 + (src10+src15));d+=dstStride;\
1184
        src16 = *s; s+=srcStride;\
1185
        OP(*d, (src13+src14)*20 - (src12+src15)*5 + (src11+src16));d+=dstStride;\
1186
        src17 = *s; s+=srcStride;\
1187
        OP(*d, (src14+src15)*20 - (src13+src16)*5 + (src12+src17));d+=dstStride;\
1188
        src18 = *s; s+=srcStride;\
1189
        OP(*d, (src15+src16)*20 - (src14+src17)*5 + (src13+src18));d+=dstStride;\
1190
       } \
1191
      } \
1192
        dst++;\
1193
        src++;\
1194
    }while(--w);\
1195
}\
1196
\
1197
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){\
1198
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
1199
    int i;\
1200
    src -= 2*srcStride;\
1201
    i= h+5; \
1202
    do {\
1203
        int srcB,srcA,src0,src1,src2,src3,src4,src5,src6;\
1204
        uint8_t *s = src-2;\
1205
        srcB = *s++;\
1206
        srcA = *s++;\
1207
        src0 = *s++;\
1208
        src1 = *s++;\
1209
        src2 = *s++;\
1210
        src3 = *s++;\
1211
        tmp[0] = ((src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));\
1212
        src4 = *s++;\
1213
        tmp[1] = ((src1+src2)*20 - (src0+src3)*5 + (srcA+src4));\
1214
        src5 = *s++;\
1215
        tmp[2] = ((src2+src3)*20 - (src1+src4)*5 + (src0+src5));\
1216
        src6 = *s++;\
1217
        tmp[3] = ((src3+src4)*20 - (src2+src5)*5 + (src1+src6));\
1218
      if (w>4) { /* it optimized */ \
1219
        int src7,src8,src9,src10; \
1220
        src7 = *s++;\
1221
        tmp[4] = ((src4+src5)*20 - (src3+src6)*5 + (src2+src7));\
1222
        src8 = *s++;\
1223
        tmp[5] = ((src5+src6)*20 - (src4+src7)*5 + (src3+src8));\
1224
        src9 = *s++;\
1225
        tmp[6] = ((src6+src7)*20 - (src5+src8)*5 + (src4+src9));\
1226
        src10 = *s++;\
1227
        tmp[7] = ((src7+src8)*20 - (src6+src9)*5 + (src5+src10));\
1228
       if (w>8) { \
1229
        int src11,src12,src13,src14,src15,src16,src17,src18; \
1230
        src11 = *s++;\
1231
        tmp[8] = ((src8 +src9 )*20 - (src7 +src10)*5 + (src6 +src11));\
1232
        src12 = *s++;\
1233
        tmp[9] = ((src9 +src10)*20 - (src8 +src11)*5 + (src7 +src12));\
1234
        src13 = *s++;\
1235
        tmp[10] = ((src10+src11)*20 - (src9 +src12)*5 + (src8 +src13));\
1236
        src14 = *s++;\
1237
        tmp[11] = ((src11+src12)*20 - (src10+src13)*5 + (src9 +src14));\
1238
        src15 = *s++;\
1239
        tmp[12] = ((src12+src13)*20 - (src11+src14)*5 + (src10+src15));\
1240
        src16 = *s++;\
1241
        tmp[13] = ((src13+src14)*20 - (src12+src15)*5 + (src11+src16));\
1242
        src17 = *s++;\
1243
        tmp[14] = ((src14+src15)*20 - (src13+src16)*5 + (src12+src17));\
1244
        src18 = *s++;\
1245
        tmp[15] = ((src15+src16)*20 - (src14+src17)*5 + (src13+src18));\
1246
       } \
1247
      } \
1248
        tmp+=tmpStride;\
1249
        src+=srcStride;\
1250
    }while(--i);\
1251
    tmp -= tmpStride*(h+5-2);\
1252
    i = w; \
1253
    do {\
1254
        int tmpB,tmpA,tmp0,tmp1,tmp2,tmp3,tmp4,tmp5,tmp6;\
1255
        int16_t *s = tmp-2*tmpStride; \
1256
        uint8_t *d=dst;\
1257
        tmpB = *s; s+=tmpStride;\
1258
        tmpA = *s; s+=tmpStride;\
1259
        tmp0 = *s; s+=tmpStride;\
1260
        tmp1 = *s; s+=tmpStride;\
1261
        tmp2 = *s; s+=tmpStride;\
1262
        tmp3 = *s; s+=tmpStride;\
1263
        OP2(*d, (tmp0+tmp1)*20 - (tmpA+tmp2)*5 + (tmpB+tmp3));d+=dstStride;\
1264
        tmp4 = *s; s+=tmpStride;\
1265
        OP2(*d, (tmp1+tmp2)*20 - (tmp0+tmp3)*5 + (tmpA+tmp4));d+=dstStride;\
1266
        tmp5 = *s; s+=tmpStride;\
1267
        OP2(*d, (tmp2+tmp3)*20 - (tmp1+tmp4)*5 + (tmp0+tmp5));d+=dstStride;\
1268
        tmp6 = *s; s+=tmpStride;\
1269
        OP2(*d, (tmp3+tmp4)*20 - (tmp2+tmp5)*5 + (tmp1+tmp6));d+=dstStride;\
1270
      if (h>4) { \
1271
        int tmp7,tmp8,tmp9,tmp10; \
1272
        tmp7 = *s; s+=tmpStride;\
1273
        OP2(*d, (tmp4+tmp5)*20 - (tmp3+tmp6)*5 + (tmp2+tmp7));d+=dstStride;\
1274
        tmp8 = *s; s+=tmpStride;\
1275
        OP2(*d, (tmp5+tmp6)*20 - (tmp4+tmp7)*5 + (tmp3+tmp8));d+=dstStride;\
1276
        tmp9 = *s; s+=tmpStride;\
1277
        OP2(*d, (tmp6+tmp7)*20 - (tmp5+tmp8)*5 + (tmp4+tmp9));d+=dstStride;\
1278
        tmp10 = *s; s+=tmpStride;\
1279
        OP2(*d, (tmp7+tmp8)*20 - (tmp6+tmp9)*5 + (tmp5+tmp10));d+=dstStride;\
1280
       if (h>8) { \
1281
        int tmp11,tmp12,tmp13,tmp14,tmp15,tmp16,tmp17,tmp18; \
1282
        tmp11 = *s; s+=tmpStride;\
1283
        OP2(*d , (tmp8 +tmp9 )*20 - (tmp7 +tmp10)*5 + (tmp6 +tmp11));d+=dstStride;\
1284
        tmp12 = *s; s+=tmpStride;\
1285
        OP2(*d , (tmp9 +tmp10)*20 - (tmp8 +tmp11)*5 + (tmp7 +tmp12));d+=dstStride;\
1286
        tmp13 = *s; s+=tmpStride;\
1287
        OP2(*d, (tmp10+tmp11)*20 - (tmp9 +tmp12)*5 + (tmp8 +tmp13));d+=dstStride;\
1288
        tmp14 = *s; s+=tmpStride;\
1289
        OP2(*d, (tmp11+tmp12)*20 - (tmp10+tmp13)*5 + (tmp9 +tmp14));d+=dstStride;\
1290
        tmp15 = *s; s+=tmpStride;\
1291
        OP2(*d, (tmp12+tmp13)*20 - (tmp11+tmp14)*5 + (tmp10+tmp15));d+=dstStride;\
1292
        tmp16 = *s; s+=tmpStride;\
1293
        OP2(*d, (tmp13+tmp14)*20 - (tmp12+tmp15)*5 + (tmp11+tmp16));d+=dstStride;\
1294
        tmp17 = *s; s+=tmpStride;\
1295
        OP2(*d, (tmp14+tmp15)*20 - (tmp13+tmp16)*5 + (tmp12+tmp17));d+=dstStride;\
1296
        tmp18 = *s; s+=tmpStride;\
1297
        OP2(*d, (tmp15+tmp16)*20 - (tmp14+tmp17)*5 + (tmp13+tmp18));d+=dstStride;\
1298
       } \
1299
      } \
1300
        dst++;\
1301
        tmp++;\
1302
    }while(--i);\
1303
}\
1304
\
1305
static void OPNAME ## h264_qpel4_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
1306
    OPNAME ## h264_qpel_h_lowpass(dst,src,dstStride,srcStride,4,4); \
1307
}\
1308
static void OPNAME ## h264_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
1309
   OPNAME ## h264_qpel_h_lowpass(dst,src,dstStride,srcStride,8,8); \
1310
}\
1311
static void OPNAME ## h264_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
1312
   OPNAME ## h264_qpel_h_lowpass(dst,src,dstStride,srcStride,16,16); \
1313
}\
1314
\
1315
static void OPNAME ## h264_qpel4_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
1316
   OPNAME ## h264_qpel_v_lowpass(dst,src,dstStride,srcStride,4,4); \
1317
}\
1318
static void OPNAME ## h264_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
1319
   OPNAME ## h264_qpel_v_lowpass(dst,src,dstStride,srcStride,8,8); \
1320
}\
1321
static void OPNAME ## h264_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
1322
   OPNAME ## h264_qpel_v_lowpass(dst,src,dstStride,srcStride,16,16); \
1323
}\
1324
static void OPNAME ## h264_qpel4_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
1325
   OPNAME ## h264_qpel_hv_lowpass(dst,tmp,src,dstStride,tmpStride,srcStride,4,4); \
1326
}\
1327
static void OPNAME ## h264_qpel8_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
1328
   OPNAME ## h264_qpel_hv_lowpass(dst,tmp,src,dstStride,tmpStride,srcStride,8,8); \
1329
}\
1330
static void OPNAME ## h264_qpel16_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
1331
   OPNAME ## h264_qpel_hv_lowpass(dst,tmp,src,dstStride,tmpStride,srcStride,16,16); \
1332
}\
1333

    
1334
#define H264_MC(OPNAME, SIZE) \
1335
static void OPNAME ## h264_qpel ## SIZE ## _mc00_c (uint8_t *dst, uint8_t *src, int stride){\
1336
    OPNAME ## pixels ## SIZE ## _c(dst, src, stride, SIZE);\
1337
}\
1338
\
1339
static void OPNAME ## h264_qpel ## SIZE ## _mc10_c(uint8_t *dst, uint8_t *src, int stride){\
1340
    uint8_t half[SIZE*SIZE];\
1341
    put_h264_qpel ## SIZE ## _h_lowpass(half, src, SIZE, stride);\
1342
    OPNAME ## pixels ## SIZE ## _l2_aligned2(dst, src, half, stride, stride, SIZE, SIZE);\
1343
}\
1344
\
1345
static void OPNAME ## h264_qpel ## SIZE ## _mc20_c(uint8_t *dst, uint8_t *src, int stride){\
1346
    OPNAME ## h264_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride);\
1347
}\
1348
\
1349
static void OPNAME ## h264_qpel ## SIZE ## _mc30_c(uint8_t *dst, uint8_t *src, int stride){\
1350
    uint8_t half[SIZE*SIZE];\
1351
    put_h264_qpel ## SIZE ## _h_lowpass(half, src, SIZE, stride);\
1352
    OPNAME ## pixels ## SIZE ## _l2_aligned2(dst, src+1, half, stride, stride, SIZE, SIZE);\
1353
}\
1354
\
1355
static void OPNAME ## h264_qpel ## SIZE ## _mc01_c(uint8_t *dst, uint8_t *src, int stride){\
1356
    uint8_t full[SIZE*(SIZE+5)];\
1357
    uint8_t * const full_mid= full + SIZE*2;\
1358
    uint8_t half[SIZE*SIZE];\
1359
    copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\
1360
    put_h264_qpel ## SIZE ## _v_lowpass(half, full_mid, SIZE, SIZE);\
1361
    OPNAME ## pixels ## SIZE ## _l2_aligned(dst, full_mid, half, stride, SIZE, SIZE, SIZE);\
1362
}\
1363
\
1364
static void OPNAME ## h264_qpel ## SIZE ## _mc02_c(uint8_t *dst, uint8_t *src, int stride){\
1365
    uint8_t full[SIZE*(SIZE+5)];\
1366
    uint8_t * const full_mid= full + SIZE*2;\
1367
    copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\
1368
    OPNAME ## h264_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE);\
1369
}\
1370
\
1371
static void OPNAME ## h264_qpel ## SIZE ## _mc03_c(uint8_t *dst, uint8_t *src, int stride){\
1372
    uint8_t full[SIZE*(SIZE+5)];\
1373
    uint8_t * const full_mid= full + SIZE*2;\
1374
    uint8_t half[SIZE*SIZE];\
1375
    copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\
1376
    put_h264_qpel ## SIZE ## _v_lowpass(half, full_mid, SIZE, SIZE);\
1377
    OPNAME ## pixels ## SIZE ## _l2_aligned(dst, full_mid+SIZE, half, stride, SIZE, SIZE, SIZE);\
1378
}\
1379
\
1380
static void OPNAME ## h264_qpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, int stride){\
1381
    uint8_t full[SIZE*(SIZE+5)];\
1382
    uint8_t * const full_mid= full + SIZE*2;\
1383
    uint8_t halfH[SIZE*SIZE];\
1384
    uint8_t halfV[SIZE*SIZE];\
1385
    put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\
1386
    copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\
1387
    put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\
1388
    OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\
1389
}\
1390
\
1391
static void OPNAME ## h264_qpel ## SIZE ## _mc31_c(uint8_t *dst, uint8_t *src, int stride){\
1392
    uint8_t full[SIZE*(SIZE+5)];\
1393
    uint8_t * const full_mid= full + SIZE*2;\
1394
    uint8_t halfH[SIZE*SIZE];\
1395
    uint8_t halfV[SIZE*SIZE];\
1396
    put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\
1397
    copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\
1398
    put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\
1399
    OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\
1400
}\
1401
\
1402
static void OPNAME ## h264_qpel ## SIZE ## _mc13_c(uint8_t *dst, uint8_t *src, int stride){\
1403
    uint8_t full[SIZE*(SIZE+5)];\
1404
    uint8_t * const full_mid= full + SIZE*2;\
1405
    uint8_t halfH[SIZE*SIZE];\
1406
    uint8_t halfV[SIZE*SIZE];\
1407
    put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\
1408
    copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\
1409
    put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\
1410
    OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\
1411
}\
1412
\
1413
static void OPNAME ## h264_qpel ## SIZE ## _mc33_c(uint8_t *dst, uint8_t *src, int stride){\
1414
    uint8_t full[SIZE*(SIZE+5)];\
1415
    uint8_t * const full_mid= full + SIZE*2;\
1416
    uint8_t halfH[SIZE*SIZE];\
1417
    uint8_t halfV[SIZE*SIZE];\
1418
    put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\
1419
    copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\
1420
    put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\
1421
    OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\
1422
}\
1423
\
1424
static void OPNAME ## h264_qpel ## SIZE ## _mc22_c(uint8_t *dst, uint8_t *src, int stride){\
1425
    int16_t tmp[SIZE*(SIZE+5)];\
1426
    OPNAME ## h264_qpel ## SIZE ## _hv_lowpass(dst, tmp, src, stride, SIZE, stride);\
1427
}\
1428
\
1429
static void OPNAME ## h264_qpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, int stride){\
1430
    int16_t tmp[SIZE*(SIZE+5)];\
1431
    uint8_t halfH[SIZE*SIZE];\
1432
    uint8_t halfHV[SIZE*SIZE];\
1433
    put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\
1434
    put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\
1435
    OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfHV, stride, SIZE, SIZE, SIZE);\
1436
}\
1437
\
1438
static void OPNAME ## h264_qpel ## SIZE ## _mc23_c(uint8_t *dst, uint8_t *src, int stride){\
1439
    int16_t tmp[SIZE*(SIZE+5)];\
1440
    uint8_t halfH[SIZE*SIZE];\
1441
    uint8_t halfHV[SIZE*SIZE];\
1442
    put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\
1443
    put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\
1444
    OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfH, halfHV, stride, SIZE, SIZE, SIZE);\
1445
}\
1446
\
1447
static void OPNAME ## h264_qpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, int stride){\
1448
    uint8_t full[SIZE*(SIZE+5)];\
1449
    uint8_t * const full_mid= full + SIZE*2;\
1450
    int16_t tmp[SIZE*(SIZE+5)];\
1451
    uint8_t halfV[SIZE*SIZE];\
1452
    uint8_t halfHV[SIZE*SIZE];\
1453
    copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\
1454
    put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\
1455
    put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\
1456
    OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfV, halfHV, stride, SIZE, SIZE, SIZE);\
1457
}\
1458
\
1459
static void OPNAME ## h264_qpel ## SIZE ## _mc32_c(uint8_t *dst, uint8_t *src, int stride){\
1460
    uint8_t full[SIZE*(SIZE+5)];\
1461
    uint8_t * const full_mid= full + SIZE*2;\
1462
    int16_t tmp[SIZE*(SIZE+5)];\
1463
    uint8_t halfV[SIZE*SIZE];\
1464
    uint8_t halfHV[SIZE*SIZE];\
1465
    copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\
1466
    put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\
1467
    put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\
1468
    OPNAME ## pixels ## SIZE ## _l2_aligned(dst, halfV, halfHV, stride, SIZE, SIZE, SIZE);\
1469
}\
1470

    
1471
#define op_avg(a, b)  a = (((a)+cm[((b) + 16)>>5]+1)>>1)
1472
//#define op_avg2(a, b) a = (((a)*w1+cm[((b) + 16)>>5]*w2 + o + 64)>>7)
1473
#define op_put(a, b)  a = cm[((b) + 16)>>5]
1474
#define op2_avg(a, b)  a = (((a)+cm[((b) + 512)>>10]+1)>>1)
1475
#define op2_put(a, b)  a = cm[((b) + 512)>>10]
1476

    
1477
H264_LOWPASS(put_       , op_put, op2_put)
1478
H264_LOWPASS(avg_       , op_avg, op2_avg)
1479
H264_MC(put_, 4)
1480
H264_MC(put_, 8)
1481
H264_MC(put_, 16)
1482
H264_MC(avg_, 4)
1483
H264_MC(avg_, 8)
1484
H264_MC(avg_, 16)
1485

    
1486
#undef op_avg
1487
#undef op_put
1488
#undef op2_avg
1489
#undef op2_put
1490
#endif
1491

    
1492
static void wmv2_mspel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){
1493
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
1494

    
1495
    do{
1496
        int src_1,src0,src1,src2,src3,src4,src5,src6,src7,src8,src9;
1497
        uint8_t *s = src;
1498
        src_1 = s[-1];
1499
        src0 = *s++;
1500
        src1 = *s++;
1501
        src2 = *s++;
1502
        dst[0]= cm[(9*(src0 + src1) - (src_1 + src2) + 8)>>4];
1503
        src3 = *s++;
1504
        dst[1]= cm[(9*(src1 + src2) - (src0 + src3) + 8)>>4];
1505
        src4 = *s++;
1506
        dst[2]= cm[(9*(src2 + src3) - (src1 + src4) + 8)>>4];
1507
        src5 = *s++;
1508
        dst[3]= cm[(9*(src3 + src4) - (src2 + src5) + 8)>>4];
1509
        src6 = *s++;
1510
        dst[4]= cm[(9*(src4 + src5) - (src3 + src6) + 8)>>4];
1511
        src7 = *s++;
1512
        dst[5]= cm[(9*(src5 + src6) - (src4 + src7) + 8)>>4];
1513
        src8 = *s++;
1514
        dst[6]= cm[(9*(src6 + src7) - (src5 + src8) + 8)>>4];
1515
        src9 = *s++;
1516
        dst[7]= cm[(9*(src7 + src8) - (src6 + src9) + 8)>>4];
1517
        dst+=dstStride;
1518
        src+=srcStride;
1519
    }while(--h);
1520
}
1521

    
1522
static void wmv2_mspel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int w){
1523
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
1524

    
1525
    do{
1526
        int src_1,src0,src1,src2,src3,src4,src5,src6,src7,src8,src9;
1527
        uint8_t *s = src,*d = dst;
1528
        src_1 = *(s-srcStride);
1529
        src0 = *s; s+=srcStride;
1530
        src1 = *s; s+=srcStride;
1531
        src2 = *s; s+=srcStride;
1532
        *d= cm[(9*(src0 + src1) - (src_1 + src2) + 8)>>4]; d+=dstStride;
1533
        src3 = *s; s+=srcStride;
1534
        *d= cm[(9*(src1 + src2) - (src0  + src3) + 8)>>4]; d+=dstStride;
1535
        src4 = *s; s+=srcStride;
1536
        *d= cm[(9*(src2 + src3) - (src1  + src4) + 8)>>4]; d+=dstStride;
1537
        src5 = *s; s+=srcStride;
1538
        *d= cm[(9*(src3 + src4) - (src2  + src5) + 8)>>4]; d+=dstStride;
1539
        src6 = *s; s+=srcStride;
1540
        *d= cm[(9*(src4 + src5) - (src3  + src6) + 8)>>4]; d+=dstStride;
1541
        src7 = *s; s+=srcStride;
1542
        *d= cm[(9*(src5 + src6) - (src4  + src7) + 8)>>4]; d+=dstStride;
1543
        src8 = *s; s+=srcStride;
1544
        *d= cm[(9*(src6 + src7) - (src5  + src8) + 8)>>4]; d+=dstStride;
1545
        src9 = *s;
1546
        *d= cm[(9*(src7 + src8) - (src6  + src9) + 8)>>4]; d+=dstStride;
1547
        src++;
1548
        dst++;
1549
    }while(--w);
1550
}
1551

    
1552
static void put_mspel8_mc00_c (uint8_t *dst, uint8_t *src, int stride){
1553
    put_pixels8_c(dst, src, stride, 8);
1554
}
1555

    
1556
static void put_mspel8_mc10_c(uint8_t *dst, uint8_t *src, int stride){
1557
    uint8_t half[64];
1558
    wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);
1559
    put_pixels8_l2_aligned2(dst, src, half, stride, stride, 8, 8);
1560
}
1561

    
1562
static void put_mspel8_mc20_c(uint8_t *dst, uint8_t *src, int stride){
1563
    wmv2_mspel8_h_lowpass(dst, src, stride, stride, 8);
1564
}
1565

    
1566
static void put_mspel8_mc30_c(uint8_t *dst, uint8_t *src, int stride){
1567
    uint8_t half[64];
1568
    wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);
1569
    put_pixels8_l2_aligned2(dst, src+1, half, stride, stride, 8, 8);
1570
}
1571

    
1572
static void put_mspel8_mc02_c(uint8_t *dst, uint8_t *src, int stride){
1573
    wmv2_mspel8_v_lowpass(dst, src, stride, stride, 8);
1574
}
1575

    
1576
static void put_mspel8_mc12_c(uint8_t *dst, uint8_t *src, int stride){
1577
    uint8_t halfH[88];
1578
    uint8_t halfV[64];
1579
    uint8_t halfHV[64];
1580
    wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);
1581
    wmv2_mspel8_v_lowpass(halfV, src, 8, stride, 8);
1582
    wmv2_mspel8_v_lowpass(halfHV, halfH+8, 8, 8, 8);
1583
    put_pixels8_l2_aligned(dst, halfV, halfHV, stride, 8, 8, 8);
1584
}
1585
static void put_mspel8_mc32_c(uint8_t *dst, uint8_t *src, int stride){
1586
    uint8_t halfH[88];
1587
    uint8_t halfV[64];
1588
    uint8_t halfHV[64];
1589
    wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);
1590
    wmv2_mspel8_v_lowpass(halfV, src+1, 8, stride, 8);
1591
    wmv2_mspel8_v_lowpass(halfHV, halfH+8, 8, 8, 8);
1592
    put_pixels8_l2_aligned(dst, halfV, halfHV, stride, 8, 8, 8);
1593
}
1594
static void put_mspel8_mc22_c(uint8_t *dst, uint8_t *src, int stride){
1595
    uint8_t halfH[88];
1596
    wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);
1597
    wmv2_mspel8_v_lowpass(dst, halfH+8, stride, 8, 8);
1598
}