Statistics
| Branch: | Revision:

ffmpeg / libavcodec / armv4l / dsputil_iwmmxt_rnd.h @ 22170ec0

History | View | Annotate | Download (47.3 KB)

1
/*
2
 * iWMMXt optimized DSP utils
3
 * copyright (c) 2004 AGAWA Koji
4
 *
5
 * This file is part of FFmpeg.
6
 *
7
 * FFmpeg is free software; you can redistribute it and/or
8
 * modify it under the terms of the GNU Lesser General Public
9
 * License as published by the Free Software Foundation; either
10
 * version 2.1 of the License, or (at your option) any later version.
11
 *
12
 * FFmpeg is distributed in the hope that it will be useful,
13
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15
 * Lesser General Public License for more details.
16
 *
17
 * You should have received a copy of the GNU Lesser General Public
18
 * License along with FFmpeg; if not, write to the Free Software
19
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20
 */
21

    
22
/* This header intentionally has no multiple inclusion guards. It is meant to
23
 * be included multiple times and generates different code depending on the
24
 * value of certain #defines. */
25

    
26
void DEF(put, pixels8)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
27
{
28
    int stride = line_size;
29
    __asm__ __volatile__ (
30
        "and r12, %[pixels], #7 \n\t"
31
        "bic %[pixels], %[pixels], #7 \n\t"
32
        "tmcr wcgr1, r12 \n\t"
33
        "add r4, %[pixels], %[line_size] \n\t"
34
        "add r5, %[block], %[line_size] \n\t"
35
        "mov %[line_size], %[line_size], lsl #1 \n\t"
36
        "1: \n\t"
37
        "wldrd wr0, [%[pixels]] \n\t"
38
        "subs %[h], %[h], #2 \n\t"
39
        "wldrd wr1, [%[pixels], #8] \n\t"
40
        "add %[pixels], %[pixels], %[line_size] \n\t"
41
        "wldrd wr3, [r4] \n\t"
42
        "pld [%[pixels]] \n\t"
43
        "pld [%[pixels], #32] \n\t"
44
        "wldrd wr4, [r4, #8] \n\t"
45
        "add r4, r4, %[line_size] \n\t"
46
        "walignr1 wr8, wr0, wr1 \n\t"
47
        "pld [r4] \n\t"
48
        "pld [r4, #32] \n\t"
49
        "walignr1 wr10, wr3, wr4 \n\t"
50
        "wstrd wr8, [%[block]] \n\t"
51
        "add %[block], %[block], %[line_size] \n\t"
52
        "wstrd wr10, [r5] \n\t"
53
        "add r5, r5, %[line_size] \n\t"
54
        "bne 1b \n\t"
55
        : [block]"+r"(block), [pixels]"+r"(pixels), [line_size]"+r"(stride), [h]"+r"(h)
56
        :
57
        : "memory", "r4", "r5", "r12");
58
}
59

    
60
void DEF(avg, pixels8)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
61
{
62
    int stride = line_size;
63
    __asm__ __volatile__ (
64
        "and r12, %[pixels], #7 \n\t"
65
        "bic %[pixels], %[pixels], #7 \n\t"
66
        "tmcr wcgr1, r12 \n\t"
67
        "add r4, %[pixels], %[line_size] \n\t"
68
        "add r5, %[block], %[line_size] \n\t"
69
        "mov %[line_size], %[line_size], lsl #1 \n\t"
70
        "1: \n\t"
71
        "wldrd wr0, [%[pixels]] \n\t"
72
        "subs %[h], %[h], #2 \n\t"
73
        "wldrd wr1, [%[pixels], #8] \n\t"
74
        "add %[pixels], %[pixels], %[line_size] \n\t"
75
        "wldrd wr3, [r4] \n\t"
76
        "pld [%[pixels]] \n\t"
77
        "pld [%[pixels], #32] \n\t"
78
        "wldrd wr4, [r4, #8] \n\t"
79
        "add r4, r4, %[line_size] \n\t"
80
        "walignr1 wr8, wr0, wr1 \n\t"
81
        "wldrd wr0, [%[block]] \n\t"
82
        "wldrd wr2, [r5] \n\t"
83
        "pld [r4] \n\t"
84
        "pld [r4, #32] \n\t"
85
        "walignr1 wr10, wr3, wr4 \n\t"
86
        WAVG2B" wr8, wr8, wr0 \n\t"
87
        WAVG2B" wr10, wr10, wr2 \n\t"
88
        "wstrd wr8, [%[block]] \n\t"
89
        "add %[block], %[block], %[line_size] \n\t"
90
        "wstrd wr10, [r5] \n\t"
91
        "pld [%[block]] \n\t"
92
        "pld [%[block], #32] \n\t"
93
        "add r5, r5, %[line_size] \n\t"
94
        "pld [r5] \n\t"
95
        "pld [r5, #32] \n\t"
96
        "bne 1b \n\t"
97
        : [block]"+r"(block), [pixels]"+r"(pixels), [line_size]"+r"(stride), [h]"+r"(h)
98
        :
99
        : "memory", "r4", "r5", "r12");
100
}
101

    
102
void DEF(put, pixels16)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
103
{
104
    int stride = line_size;
105
    __asm__ __volatile__ (
106
        "and r12, %[pixels], #7 \n\t"
107
        "bic %[pixels], %[pixels], #7 \n\t"
108
        "tmcr wcgr1, r12 \n\t"
109
        "add r4, %[pixels], %[line_size] \n\t"
110
        "add r5, %[block], %[line_size] \n\t"
111
        "mov %[line_size], %[line_size], lsl #1 \n\t"
112
        "1: \n\t"
113
        "wldrd wr0, [%[pixels]] \n\t"
114
        "wldrd wr1, [%[pixels], #8] \n\t"
115
        "subs %[h], %[h], #2 \n\t"
116
        "wldrd wr2, [%[pixels], #16] \n\t"
117
        "add %[pixels], %[pixels], %[line_size] \n\t"
118
        "wldrd wr3, [r4] \n\t"
119
        "pld [%[pixels]] \n\t"
120
        "pld [%[pixels], #32] \n\t"
121
        "walignr1 wr8, wr0, wr1 \n\t"
122
        "wldrd wr4, [r4, #8] \n\t"
123
        "walignr1 wr9, wr1, wr2 \n\t"
124
        "wldrd wr5, [r4, #16] \n\t"
125
        "add r4, r4, %[line_size] \n\t"
126
        "pld [r4] \n\t"
127
        "pld [r4, #32] \n\t"
128
        "walignr1 wr10, wr3, wr4 \n\t"
129
        "wstrd wr8, [%[block]] \n\t"
130
        "walignr1 wr11, wr4, wr5 \n\t"
131
        "wstrd wr9, [%[block], #8] \n\t"
132
        "add %[block], %[block], %[line_size] \n\t"
133
        "wstrd wr10, [r5] \n\t"
134
        "wstrd wr11, [r5, #8] \n\t"
135
        "add r5, r5, %[line_size] \n\t"
136
        "bne 1b \n\t"
137
        : [block]"+r"(block), [pixels]"+r"(pixels), [line_size]"+r"(stride), [h]"+r"(h)
138
        :
139
        : "memory", "r4", "r5", "r12");
140
}
141

    
142
void DEF(avg, pixels16)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
143
{
144
    int stride = line_size;
145
    __asm__ __volatile__ (
146
        "pld [%[pixels]]                \n\t"
147
        "pld [%[pixels], #32]           \n\t"
148
        "pld [%[block]]                 \n\t"
149
        "pld [%[block], #32]            \n\t"
150
        "and r12, %[pixels], #7         \n\t"
151
        "bic %[pixels], %[pixels], #7   \n\t"
152
        "tmcr wcgr1, r12                \n\t"
153
        "add r4, %[pixels], %[line_size]\n\t"
154
        "add r5, %[block], %[line_size] \n\t"
155
        "mov %[line_size], %[line_size], lsl #1 \n\t"
156
        "1:                             \n\t"
157
        "wldrd wr0, [%[pixels]]         \n\t"
158
        "wldrd wr1, [%[pixels], #8]     \n\t"
159
        "subs %[h], %[h], #2            \n\t"
160
        "wldrd wr2, [%[pixels], #16]    \n\t"
161
        "add %[pixels], %[pixels], %[line_size] \n\t"
162
        "wldrd wr3, [r4]                \n\t"
163
        "pld [%[pixels]]                \n\t"
164
        "pld [%[pixels], #32]           \n\t"
165
        "walignr1 wr8, wr0, wr1         \n\t"
166
        "wldrd wr4, [r4, #8]            \n\t"
167
        "walignr1 wr9, wr1, wr2         \n\t"
168
        "wldrd wr5, [r4, #16]           \n\t"
169
        "add r4, r4, %[line_size]       \n\t"
170
        "wldrd wr0, [%[block]]          \n\t"
171
        "pld [r4]                       \n\t"
172
        "wldrd wr1, [%[block], #8]      \n\t"
173
        "pld [r4, #32]                  \n\t"
174
        "wldrd wr2, [r5]                \n\t"
175
        "walignr1 wr10, wr3, wr4        \n\t"
176
        "wldrd wr3, [r5, #8]            \n\t"
177
        WAVG2B" wr8, wr8, wr0           \n\t"
178
        WAVG2B" wr9, wr9, wr1           \n\t"
179
        WAVG2B" wr10, wr10, wr2         \n\t"
180
        "wstrd wr8, [%[block]]          \n\t"
181
        "walignr1 wr11, wr4, wr5        \n\t"
182
        WAVG2B" wr11, wr11, wr3         \n\t"
183
        "wstrd wr9, [%[block], #8]      \n\t"
184
        "add %[block], %[block], %[line_size] \n\t"
185
        "wstrd wr10, [r5]               \n\t"
186
        "pld [%[block]]                 \n\t"
187
        "pld [%[block], #32]            \n\t"
188
        "wstrd wr11, [r5, #8]           \n\t"
189
        "add r5, r5, %[line_size]       \n\t"
190
        "pld [r5]                       \n\t"
191
        "pld [r5, #32]                  \n\t"
192
        "bne 1b \n\t"
193
        : [block]"+r"(block), [pixels]"+r"(pixels), [line_size]"+r"(stride), [h]"+r"(h)
194
        :
195
        : "memory", "r4", "r5", "r12");
196
}
197

    
198
void DEF(put, pixels8_x2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
199
{
200
    int stride = line_size;
201
    // [wr0 wr1 wr2 wr3] for previous line
202
    // [wr4 wr5 wr6 wr7] for current line
203
    SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
204
    __asm__ __volatile__(
205
        "pld [%[pixels]]                \n\t"
206
        "pld [%[pixels], #32]           \n\t"
207
        "and r12, %[pixels], #7         \n\t"
208
        "bic %[pixels], %[pixels], #7   \n\t"
209
        "tmcr wcgr1, r12                \n\t"
210
        "add r12, r12, #1               \n\t"
211
        "add r4, %[pixels], %[line_size]\n\t"
212
        "tmcr wcgr2, r12                \n\t"
213
        "add r5, %[block], %[line_size] \n\t"
214
        "mov %[line_size], %[line_size], lsl #1 \n\t"
215

    
216
        "1:                             \n\t"
217
        "wldrd wr10, [%[pixels]]        \n\t"
218
        "cmp r12, #8                    \n\t"
219
        "wldrd wr11, [%[pixels], #8]    \n\t"
220
        "add %[pixels], %[pixels], %[line_size] \n\t"
221
        "wldrd wr13, [r4]               \n\t"
222
        "pld [%[pixels]]                \n\t"
223
        "wldrd wr14, [r4, #8]           \n\t"
224
        "pld [%[pixels], #32]           \n\t"
225
        "add r4, r4, %[line_size]       \n\t"
226
        "walignr1 wr0, wr10, wr11       \n\t"
227
        "pld [r4]                       \n\t"
228
        "pld [r4, #32]                  \n\t"
229
        "walignr1 wr2, wr13, wr14       \n\t"
230
        "wmoveq wr4, wr11               \n\t"
231
        "wmoveq wr6, wr14               \n\t"
232
        "walignr2ne wr4, wr10, wr11     \n\t"
233
        "walignr2ne wr6, wr13, wr14     \n\t"
234
        WAVG2B" wr0, wr0, wr4           \n\t"
235
        WAVG2B" wr2, wr2, wr6           \n\t"
236
        "wstrd wr0, [%[block]]          \n\t"
237
        "subs %[h], %[h], #2            \n\t"
238
        "wstrd wr2, [r5]                \n\t"
239
        "add %[block], %[block], %[line_size]   \n\t"
240
        "add r5, r5, %[line_size]       \n\t"
241
        "bne 1b                         \n\t"
242
        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
243
        :
244
        : "r4", "r5", "r12", "memory");
245
}
246

    
247
void DEF(put, pixels16_x2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
248
{
249
    int stride = line_size;
250
    // [wr0 wr1 wr2 wr3] for previous line
251
    // [wr4 wr5 wr6 wr7] for current line
252
    SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
253
    __asm__ __volatile__(
254
        "pld [%[pixels]]                \n\t"
255
        "pld [%[pixels], #32]           \n\t"
256
        "and r12, %[pixels], #7         \n\t"
257
        "bic %[pixels], %[pixels], #7   \n\t"
258
        "tmcr wcgr1, r12                \n\t"
259
        "add r12, r12, #1               \n\t"
260
        "add r4, %[pixels], %[line_size]\n\t"
261
        "tmcr wcgr2, r12                \n\t"
262
        "add r5, %[block], %[line_size] \n\t"
263
        "mov %[line_size], %[line_size], lsl #1 \n\t"
264

    
265
        "1:                             \n\t"
266
        "wldrd wr10, [%[pixels]]        \n\t"
267
        "cmp r12, #8                    \n\t"
268
        "wldrd wr11, [%[pixels], #8]    \n\t"
269
        "wldrd wr12, [%[pixels], #16]   \n\t"
270
        "add %[pixels], %[pixels], %[line_size] \n\t"
271
        "wldrd wr13, [r4]               \n\t"
272
        "pld [%[pixels]]                \n\t"
273
        "wldrd wr14, [r4, #8]           \n\t"
274
        "pld [%[pixels], #32]           \n\t"
275
        "wldrd wr15, [r4, #16]          \n\t"
276
        "add r4, r4, %[line_size]       \n\t"
277
        "walignr1 wr0, wr10, wr11       \n\t"
278
        "pld [r4]                       \n\t"
279
        "pld [r4, #32]                  \n\t"
280
        "walignr1 wr1, wr11, wr12       \n\t"
281
        "walignr1 wr2, wr13, wr14       \n\t"
282
        "walignr1 wr3, wr14, wr15       \n\t"
283
        "wmoveq wr4, wr11               \n\t"
284
        "wmoveq wr5, wr12               \n\t"
285
        "wmoveq wr6, wr14               \n\t"
286
        "wmoveq wr7, wr15               \n\t"
287
        "walignr2ne wr4, wr10, wr11     \n\t"
288
        "walignr2ne wr5, wr11, wr12     \n\t"
289
        "walignr2ne wr6, wr13, wr14     \n\t"
290
        "walignr2ne wr7, wr14, wr15     \n\t"
291
        WAVG2B" wr0, wr0, wr4           \n\t"
292
        WAVG2B" wr1, wr1, wr5           \n\t"
293
        "wstrd wr0, [%[block]]          \n\t"
294
        WAVG2B" wr2, wr2, wr6           \n\t"
295
        "wstrd wr1, [%[block], #8]      \n\t"
296
        WAVG2B" wr3, wr3, wr7           \n\t"
297
        "add %[block], %[block], %[line_size]   \n\t"
298
        "wstrd wr2, [r5]                \n\t"
299
        "subs %[h], %[h], #2            \n\t"
300
        "wstrd wr3, [r5, #8]            \n\t"
301
        "add r5, r5, %[line_size]       \n\t"
302
        "bne 1b                         \n\t"
303
        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
304
        :
305
        : "r4", "r5", "r12", "memory");
306
}
307

    
308
void DEF(avg, pixels8_x2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
309
{
310
    int stride = line_size;
311
    // [wr0 wr1 wr2 wr3] for previous line
312
    // [wr4 wr5 wr6 wr7] for current line
313
    SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
314
    __asm__ __volatile__(
315
        "pld [%[pixels]]                \n\t"
316
        "pld [%[pixels], #32]           \n\t"
317
        "pld [%[block]]                 \n\t"
318
        "pld [%[block], #32]            \n\t"
319
        "and r12, %[pixels], #7         \n\t"
320
        "bic %[pixels], %[pixels], #7   \n\t"
321
        "tmcr wcgr1, r12                \n\t"
322
        "add r12, r12, #1               \n\t"
323
        "add r4, %[pixels], %[line_size]\n\t"
324
        "tmcr wcgr2, r12                \n\t"
325
        "add r5, %[block], %[line_size] \n\t"
326
        "mov %[line_size], %[line_size], lsl #1 \n\t"
327
        "pld [r5]                       \n\t"
328
        "pld [r5, #32]                  \n\t"
329

    
330
        "1:                             \n\t"
331
        "wldrd wr10, [%[pixels]]        \n\t"
332
        "cmp r12, #8                    \n\t"
333
        "wldrd wr11, [%[pixels], #8]    \n\t"
334
        "add %[pixels], %[pixels], %[line_size] \n\t"
335
        "wldrd wr13, [r4]               \n\t"
336
        "pld [%[pixels]]                \n\t"
337
        "wldrd wr14, [r4, #8]           \n\t"
338
        "pld [%[pixels], #32]           \n\t"
339
        "add r4, r4, %[line_size]       \n\t"
340
        "walignr1 wr0, wr10, wr11       \n\t"
341
        "pld [r4]                       \n\t"
342
        "pld [r4, #32]                  \n\t"
343
        "walignr1 wr2, wr13, wr14       \n\t"
344
        "wmoveq wr4, wr11               \n\t"
345
        "wmoveq wr6, wr14               \n\t"
346
        "walignr2ne wr4, wr10, wr11     \n\t"
347
        "wldrd wr10, [%[block]]         \n\t"
348
        "walignr2ne wr6, wr13, wr14     \n\t"
349
        "wldrd wr12, [r5]               \n\t"
350
        WAVG2B" wr0, wr0, wr4           \n\t"
351
        WAVG2B" wr2, wr2, wr6           \n\t"
352
        WAVG2B" wr0, wr0, wr10          \n\t"
353
        WAVG2B" wr2, wr2, wr12          \n\t"
354
        "wstrd wr0, [%[block]]          \n\t"
355
        "subs %[h], %[h], #2            \n\t"
356
        "wstrd wr2, [r5]                \n\t"
357
        "add %[block], %[block], %[line_size]   \n\t"
358
        "add r5, r5, %[line_size]       \n\t"
359
        "pld [%[block]]                 \n\t"
360
        "pld [%[block], #32]            \n\t"
361
        "pld [r5]                       \n\t"
362
        "pld [r5, #32]                  \n\t"
363
        "bne 1b                         \n\t"
364
        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
365
        :
366
        : "r4", "r5", "r12", "memory");
367
}
368

    
369
void DEF(avg, pixels16_x2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
370
{
371
    int stride = line_size;
372
    // [wr0 wr1 wr2 wr3] for previous line
373
    // [wr4 wr5 wr6 wr7] for current line
374
    SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
375
    __asm__ __volatile__(
376
        "pld [%[pixels]]                \n\t"
377
        "pld [%[pixels], #32]           \n\t"
378
        "pld [%[block]]                 \n\t"
379
        "pld [%[block], #32]            \n\t"
380
        "and r12, %[pixels], #7         \n\t"
381
        "bic %[pixels], %[pixels], #7   \n\t"
382
        "tmcr wcgr1, r12                \n\t"
383
        "add r12, r12, #1               \n\t"
384
        "add r4, %[pixels], %[line_size]\n\t"
385
        "tmcr wcgr2, r12                \n\t"
386
        "add r5, %[block], %[line_size] \n\t"
387
        "mov %[line_size], %[line_size], lsl #1 \n\t"
388
        "pld [r5]                       \n\t"
389
        "pld [r5, #32]                  \n\t"
390

    
391
        "1:                             \n\t"
392
        "wldrd wr10, [%[pixels]]        \n\t"
393
        "cmp r12, #8                    \n\t"
394
        "wldrd wr11, [%[pixels], #8]    \n\t"
395
        "wldrd wr12, [%[pixels], #16]   \n\t"
396
        "add %[pixels], %[pixels], %[line_size] \n\t"
397
        "wldrd wr13, [r4]               \n\t"
398
        "pld [%[pixels]]                \n\t"
399
        "wldrd wr14, [r4, #8]           \n\t"
400
        "pld [%[pixels], #32]           \n\t"
401
        "wldrd wr15, [r4, #16]          \n\t"
402
        "add r4, r4, %[line_size]       \n\t"
403
        "walignr1 wr0, wr10, wr11       \n\t"
404
        "pld [r4]                       \n\t"
405
        "pld [r4, #32]                  \n\t"
406
        "walignr1 wr1, wr11, wr12       \n\t"
407
        "walignr1 wr2, wr13, wr14       \n\t"
408
        "walignr1 wr3, wr14, wr15       \n\t"
409
        "wmoveq wr4, wr11               \n\t"
410
        "wmoveq wr5, wr12               \n\t"
411
        "wmoveq wr6, wr14               \n\t"
412
        "wmoveq wr7, wr15               \n\t"
413
        "walignr2ne wr4, wr10, wr11     \n\t"
414
        "walignr2ne wr5, wr11, wr12     \n\t"
415
        "walignr2ne wr6, wr13, wr14     \n\t"
416
        "walignr2ne wr7, wr14, wr15     \n\t"
417
        "wldrd wr10, [%[block]]         \n\t"
418
        WAVG2B" wr0, wr0, wr4           \n\t"
419
        "wldrd wr11, [%[block], #8]     \n\t"
420
        WAVG2B" wr1, wr1, wr5           \n\t"
421
        "wldrd wr12, [r5]               \n\t"
422
        WAVG2B" wr2, wr2, wr6           \n\t"
423
        "wldrd wr13, [r5, #8]           \n\t"
424
        WAVG2B" wr3, wr3, wr7           \n\t"
425
        WAVG2B" wr0, wr0, wr10          \n\t"
426
        WAVG2B" wr1, wr1, wr11          \n\t"
427
        WAVG2B" wr2, wr2, wr12          \n\t"
428
        WAVG2B" wr3, wr3, wr13          \n\t"
429
        "wstrd wr0, [%[block]]          \n\t"
430
        "subs %[h], %[h], #2            \n\t"
431
        "wstrd wr1, [%[block], #8]      \n\t"
432
        "add %[block], %[block], %[line_size]   \n\t"
433
        "wstrd wr2, [r5]                \n\t"
434
        "pld [%[block]]                 \n\t"
435
        "wstrd wr3, [r5, #8]            \n\t"
436
        "add r5, r5, %[line_size]       \n\t"
437
        "pld [%[block], #32]            \n\t"
438
        "pld [r5]                       \n\t"
439
        "pld [r5, #32]                  \n\t"
440
        "bne 1b                         \n\t"
441
        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
442
        :
443
        :"r4", "r5", "r12", "memory");
444
}
445

    
446
void DEF(avg, pixels8_y2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
447
{
448
    int stride = line_size;
449
    // [wr0 wr1 wr2 wr3] for previous line
450
    // [wr4 wr5 wr6 wr7] for current line
451
    __asm__ __volatile__(
452
        "pld            [%[pixels]]                             \n\t"
453
        "pld            [%[pixels], #32]                        \n\t"
454
        "and            r12, %[pixels], #7                      \n\t"
455
        "tmcr           wcgr1, r12                              \n\t"
456
        "bic            %[pixels], %[pixels], #7                \n\t"
457

    
458
        "wldrd          wr10, [%[pixels]]                       \n\t"
459
        "wldrd          wr11, [%[pixels], #8]                   \n\t"
460
        "pld            [%[block]]                              \n\t"
461
        "add            %[pixels], %[pixels], %[line_size]      \n\t"
462
        "walignr1       wr0, wr10, wr11                         \n\t"
463
        "pld            [%[pixels]]                             \n\t"
464
        "pld            [%[pixels], #32]                        \n\t"
465

    
466
      "1:                                                       \n\t"
467
        "wldrd          wr10, [%[pixels]]                       \n\t"
468
        "wldrd          wr11, [%[pixels], #8]                   \n\t"
469
        "add            %[pixels], %[pixels], %[line_size]      \n\t"
470
        "pld            [%[pixels]]                             \n\t"
471
        "pld            [%[pixels], #32]                        \n\t"
472
        "walignr1       wr4, wr10, wr11                         \n\t"
473
        "wldrd          wr10, [%[block]]                        \n\t"
474
         WAVG2B"        wr8, wr0, wr4                           \n\t"
475
         WAVG2B"        wr8, wr8, wr10                          \n\t"
476
        "wstrd          wr8, [%[block]]                         \n\t"
477
        "add            %[block], %[block], %[line_size]        \n\t"
478

    
479
        "wldrd          wr10, [%[pixels]]                       \n\t"
480
        "wldrd          wr11, [%[pixels], #8]                   \n\t"
481
        "pld            [%[block]]                              \n\t"
482
        "add            %[pixels], %[pixels], %[line_size]      \n\t"
483
        "pld            [%[pixels]]                             \n\t"
484
        "pld            [%[pixels], #32]                        \n\t"
485
        "walignr1       wr0, wr10, wr11                         \n\t"
486
        "wldrd          wr10, [%[block]]                        \n\t"
487
         WAVG2B"        wr8, wr0, wr4                           \n\t"
488
         WAVG2B"        wr8, wr8, wr10                          \n\t"
489
        "wstrd          wr8, [%[block]]                         \n\t"
490
        "add            %[block], %[block], %[line_size]        \n\t"
491

    
492
        "subs           %[h], %[h], #2                          \n\t"
493
        "pld            [%[block]]                              \n\t"
494
        "bne            1b                                      \n\t"
495
        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
496
        :
497
        : "cc", "memory", "r12");
498
}
499

    
500
void DEF(put, pixels16_y2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
501
{
502
    int stride = line_size;
503
    // [wr0 wr1 wr2 wr3] for previous line
504
    // [wr4 wr5 wr6 wr7] for current line
505
    __asm__ __volatile__(
506
        "pld [%[pixels]]                \n\t"
507
        "pld [%[pixels], #32]           \n\t"
508
        "and r12, %[pixels], #7         \n\t"
509
        "tmcr wcgr1, r12                \n\t"
510
        "bic %[pixels], %[pixels], #7   \n\t"
511

    
512
        "wldrd wr10, [%[pixels]]        \n\t"
513
        "wldrd wr11, [%[pixels], #8]    \n\t"
514
        "wldrd wr12, [%[pixels], #16]   \n\t"
515
        "add %[pixels], %[pixels], %[line_size] \n\t"
516
        "pld [%[pixels]]                \n\t"
517
        "pld [%[pixels], #32]           \n\t"
518
        "walignr1 wr0, wr10, wr11       \n\t"
519
        "walignr1 wr1, wr11, wr12       \n\t"
520

    
521
        "1:                             \n\t"
522
        "wldrd wr10, [%[pixels]]        \n\t"
523
        "wldrd wr11, [%[pixels], #8]    \n\t"
524
        "wldrd wr12, [%[pixels], #16]   \n\t"
525
        "add %[pixels], %[pixels], %[line_size] \n\t"
526
        "pld [%[pixels]]                \n\t"
527
        "pld [%[pixels], #32]           \n\t"
528
        "walignr1 wr4, wr10, wr11       \n\t"
529
        "walignr1 wr5, wr11, wr12       \n\t"
530
        WAVG2B" wr8, wr0, wr4           \n\t"
531
        WAVG2B" wr9, wr1, wr5           \n\t"
532
        "wstrd wr8, [%[block]]          \n\t"
533
        "wstrd wr9, [%[block], #8]      \n\t"
534
        "add %[block], %[block], %[line_size]   \n\t"
535

    
536
        "wldrd wr10, [%[pixels]]        \n\t"
537
        "wldrd wr11, [%[pixels], #8]    \n\t"
538
        "wldrd wr12, [%[pixels], #16]   \n\t"
539
        "add %[pixels], %[pixels], %[line_size] \n\t"
540
        "pld [%[pixels]]                \n\t"
541
        "pld [%[pixels], #32]           \n\t"
542
        "walignr1 wr0, wr10, wr11       \n\t"
543
        "walignr1 wr1, wr11, wr12       \n\t"
544
        WAVG2B" wr8, wr0, wr4           \n\t"
545
        WAVG2B" wr9, wr1, wr5           \n\t"
546
        "wstrd wr8, [%[block]]          \n\t"
547
        "wstrd wr9, [%[block], #8]      \n\t"
548
        "add %[block], %[block], %[line_size]   \n\t"
549

    
550
        "subs %[h], %[h], #2            \n\t"
551
        "bne 1b                         \n\t"
552
        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
553
        :
554
        : "r4", "r5", "r12", "memory");
555
}
556

    
557
void DEF(avg, pixels16_y2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
558
{
559
    int stride = line_size;
560
    // [wr0 wr1 wr2 wr3] for previous line
561
    // [wr4 wr5 wr6 wr7] for current line
562
    __asm__ __volatile__(
563
        "pld [%[pixels]]                \n\t"
564
        "pld [%[pixels], #32]           \n\t"
565
        "and r12, %[pixels], #7         \n\t"
566
        "tmcr wcgr1, r12                \n\t"
567
        "bic %[pixels], %[pixels], #7   \n\t"
568

    
569
        "wldrd wr10, [%[pixels]]        \n\t"
570
        "wldrd wr11, [%[pixels], #8]    \n\t"
571
        "pld [%[block]]                 \n\t"
572
        "wldrd wr12, [%[pixels], #16]   \n\t"
573
        "add %[pixels], %[pixels], %[line_size] \n\t"
574
        "pld [%[pixels]]                \n\t"
575
        "pld [%[pixels], #32]           \n\t"
576
        "walignr1 wr0, wr10, wr11       \n\t"
577
        "walignr1 wr1, wr11, wr12       \n\t"
578

    
579
        "1:                             \n\t"
580
        "wldrd wr10, [%[pixels]]        \n\t"
581
        "wldrd wr11, [%[pixels], #8]    \n\t"
582
        "wldrd wr12, [%[pixels], #16]   \n\t"
583
        "add %[pixels], %[pixels], %[line_size] \n\t"
584
        "pld [%[pixels]]                \n\t"
585
        "pld [%[pixels], #32]           \n\t"
586
        "walignr1 wr4, wr10, wr11       \n\t"
587
        "walignr1 wr5, wr11, wr12       \n\t"
588
        "wldrd wr10, [%[block]]         \n\t"
589
        "wldrd wr11, [%[block], #8]     \n\t"
590
        WAVG2B" wr8, wr0, wr4           \n\t"
591
        WAVG2B" wr9, wr1, wr5           \n\t"
592
        WAVG2B" wr8, wr8, wr10          \n\t"
593
        WAVG2B" wr9, wr9, wr11          \n\t"
594
        "wstrd wr8, [%[block]]          \n\t"
595
        "wstrd wr9, [%[block], #8]      \n\t"
596
        "add %[block], %[block], %[line_size]   \n\t"
597

    
598
        "wldrd wr10, [%[pixels]]        \n\t"
599
        "wldrd wr11, [%[pixels], #8]    \n\t"
600
        "pld [%[block]]                 \n\t"
601
        "wldrd wr12, [%[pixels], #16]   \n\t"
602
        "add %[pixels], %[pixels], %[line_size] \n\t"
603
        "pld [%[pixels]]                \n\t"
604
        "pld [%[pixels], #32]           \n\t"
605
        "walignr1 wr0, wr10, wr11       \n\t"
606
        "walignr1 wr1, wr11, wr12       \n\t"
607
        "wldrd wr10, [%[block]]         \n\t"
608
        "wldrd wr11, [%[block], #8]     \n\t"
609
        WAVG2B" wr8, wr0, wr4           \n\t"
610
        WAVG2B" wr9, wr1, wr5           \n\t"
611
        WAVG2B" wr8, wr8, wr10          \n\t"
612
        WAVG2B" wr9, wr9, wr11          \n\t"
613
        "wstrd wr8, [%[block]]          \n\t"
614
        "wstrd wr9, [%[block], #8]      \n\t"
615
        "add %[block], %[block], %[line_size]   \n\t"
616

    
617
        "subs %[h], %[h], #2            \n\t"
618
        "pld [%[block]]                 \n\t"
619
        "bne 1b                         \n\t"
620
        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
621
        :
622
        : "r4", "r5", "r12", "memory");
623
}
624

    
625
void DEF(put, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
626
{
627
    // [wr0 wr1 wr2 wr3] for previous line
628
    // [wr4 wr5 wr6 wr7] for current line
629
    SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
630
    __asm__ __volatile__(
631
        "pld [%[pixels]]                \n\t"
632
        "mov r12, #2                    \n\t"
633
        "pld [%[pixels], #32]           \n\t"
634
        "tmcr wcgr0, r12                \n\t" /* for shift value */
635
        "and r12, %[pixels], #7         \n\t"
636
        "bic %[pixels], %[pixels], #7   \n\t"
637
        "tmcr wcgr1, r12                \n\t"
638

    
639
        // [wr0 wr1 wr2 wr3] <= *
640
        // [wr4 wr5 wr6 wr7]
641
        "wldrd wr12, [%[pixels]]        \n\t"
642
        "add r12, r12, #1               \n\t"
643
        "wldrd wr13, [%[pixels], #8]    \n\t"
644
        "tmcr wcgr2, r12                \n\t"
645
        "add %[pixels], %[pixels], %[line_size] \n\t"
646
        "cmp r12, #8                    \n\t"
647
        "pld [%[pixels]]                \n\t"
648
        "pld [%[pixels], #32]           \n\t"
649
        "walignr1 wr2, wr12, wr13       \n\t"
650
        "wmoveq wr10, wr13              \n\t"
651
        "walignr2ne wr10, wr12, wr13    \n\t"
652
        "wunpckelub wr0, wr2            \n\t"
653
        "wunpckehub wr1, wr2            \n\t"
654
        "wunpckelub wr8, wr10           \n\t"
655
        "wunpckehub wr9, wr10           \n\t"
656
        "waddhus wr0, wr0, wr8          \n\t"
657
        "waddhus wr1, wr1, wr9          \n\t"
658

    
659
        "1:                             \n\t"
660
        // [wr0 wr1 wr2 wr3]
661
        // [wr4 wr5 wr6 wr7] <= *
662
        "wldrd wr12, [%[pixels]]        \n\t"
663
        "cmp r12, #8                    \n\t"
664
        "wldrd wr13, [%[pixels], #8]    \n\t"
665
        "add %[pixels], %[pixels], %[line_size] \n\t"
666
        "walignr1 wr6, wr12, wr13       \n\t"
667
        "pld [%[pixels]]                \n\t"
668
        "pld [%[pixels], #32]           \n\t"
669
        "wmoveq wr10, wr13              \n\t"
670
        "walignr2ne wr10, wr12, wr13    \n\t"
671
        "wunpckelub wr4, wr6            \n\t"
672
        "wunpckehub wr5, wr6            \n\t"
673
        "wunpckelub wr8, wr10           \n\t"
674
        "wunpckehub wr9, wr10           \n\t"
675
        "waddhus wr4, wr4, wr8          \n\t"
676
        "waddhus wr5, wr5, wr9          \n\t"
677
        "waddhus wr8, wr0, wr4          \n\t"
678
        "waddhus wr9, wr1, wr5          \n\t"
679
        "waddhus wr8, wr8, wr15         \n\t"
680
        "waddhus wr9, wr9, wr15         \n\t"
681
        "wsrlhg wr8, wr8, wcgr0         \n\t"
682
        "wsrlhg wr9, wr9, wcgr0         \n\t"
683
        "wpackhus wr8, wr8, wr9         \n\t"
684
        "wstrd wr8, [%[block]]          \n\t"
685
        "add %[block], %[block], %[line_size]   \n\t"
686

    
687
        // [wr0 wr1 wr2 wr3] <= *
688
        // [wr4 wr5 wr6 wr7]
689
        "wldrd wr12, [%[pixels]]        \n\t"
690
        "wldrd wr13, [%[pixels], #8]    \n\t"
691
        "add %[pixels], %[pixels], %[line_size] \n\t"
692
        "walignr1 wr2, wr12, wr13       \n\t"
693
        "pld [%[pixels]]                \n\t"
694
        "pld [%[pixels], #32]           \n\t"
695
        "wmoveq wr10, wr13              \n\t"
696
        "walignr2ne wr10, wr12, wr13    \n\t"
697
        "wunpckelub wr0, wr2            \n\t"
698
        "wunpckehub wr1, wr2            \n\t"
699
        "wunpckelub wr8, wr10           \n\t"
700
        "wunpckehub wr9, wr10           \n\t"
701
        "waddhus wr0, wr0, wr8          \n\t"
702
        "waddhus wr1, wr1, wr9          \n\t"
703
        "waddhus wr8, wr0, wr4          \n\t"
704
        "waddhus wr9, wr1, wr5          \n\t"
705
        "waddhus wr8, wr8, wr15         \n\t"
706
        "waddhus wr9, wr9, wr15         \n\t"
707
        "wsrlhg wr8, wr8, wcgr0         \n\t"
708
        "wsrlhg wr9, wr9, wcgr0         \n\t"
709
        "wpackhus wr8, wr8, wr9         \n\t"
710
        "subs %[h], %[h], #2            \n\t"
711
        "wstrd wr8, [%[block]]          \n\t"
712
        "add %[block], %[block], %[line_size]   \n\t"
713
        "bne 1b                         \n\t"
714
        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block)
715
        : [line_size]"r"(line_size)
716
        : "r12", "memory");
717
}
718

    
719
void DEF(put, pixels16_xy2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
720
{
721
    // [wr0 wr1 wr2 wr3] for previous line
722
    // [wr4 wr5 wr6 wr7] for current line
723
    SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
724
    __asm__ __volatile__(
725
        "pld [%[pixels]]                \n\t"
726
        "mov r12, #2                    \n\t"
727
        "pld [%[pixels], #32]           \n\t"
728
        "tmcr wcgr0, r12                \n\t" /* for shift value */
729
        /* alignment */
730
        "and r12, %[pixels], #7         \n\t"
731
        "bic %[pixels], %[pixels], #7   \n\t"
732
        "tmcr wcgr1, r12                \n\t"
733
        "add r12, r12, #1               \n\t"
734
        "tmcr wcgr2, r12                \n\t"
735

    
736
        // [wr0 wr1 wr2 wr3] <= *
737
        // [wr4 wr5 wr6 wr7]
738
        "wldrd wr12, [%[pixels]]        \n\t"
739
        "cmp r12, #8                    \n\t"
740
        "wldrd wr13, [%[pixels], #8]    \n\t"
741
        "wldrd wr14, [%[pixels], #16]   \n\t"
742
        "add %[pixels], %[pixels], %[line_size] \n\t"
743
        "pld [%[pixels]]                \n\t"
744
        "walignr1 wr2, wr12, wr13       \n\t"
745
        "pld [%[pixels], #32]           \n\t"
746
        "walignr1 wr3, wr13, wr14       \n\t"
747
        "wmoveq wr10, wr13              \n\t"
748
        "wmoveq wr11, wr14              \n\t"
749
        "walignr2ne wr10, wr12, wr13    \n\t"
750
        "walignr2ne wr11, wr13, wr14    \n\t"
751
        "wunpckelub wr0, wr2            \n\t"
752
        "wunpckehub wr1, wr2            \n\t"
753
        "wunpckelub wr2, wr3            \n\t"
754
        "wunpckehub wr3, wr3            \n\t"
755
        "wunpckelub wr8, wr10           \n\t"
756
        "wunpckehub wr9, wr10           \n\t"
757
        "wunpckelub wr10, wr11          \n\t"
758
        "wunpckehub wr11, wr11          \n\t"
759
        "waddhus wr0, wr0, wr8          \n\t"
760
        "waddhus wr1, wr1, wr9          \n\t"
761
        "waddhus wr2, wr2, wr10         \n\t"
762
        "waddhus wr3, wr3, wr11         \n\t"
763

    
764
        "1:                             \n\t"
765
        // [wr0 wr1 wr2 wr3]
766
        // [wr4 wr5 wr6 wr7] <= *
767
        "wldrd wr12, [%[pixels]]        \n\t"
768
        "cmp r12, #8                    \n\t"
769
        "wldrd wr13, [%[pixels], #8]    \n\t"
770
        "wldrd wr14, [%[pixels], #16]   \n\t"
771
        "add %[pixels], %[pixels], %[line_size] \n\t"
772
        "walignr1 wr6, wr12, wr13       \n\t"
773
        "pld [%[pixels]]                \n\t"
774
        "pld [%[pixels], #32]           \n\t"
775
        "walignr1 wr7, wr13, wr14       \n\t"
776
        "wmoveq wr10, wr13              \n\t"
777
        "wmoveq wr11, wr14              \n\t"
778
        "walignr2ne wr10, wr12, wr13    \n\t"
779
        "walignr2ne wr11, wr13, wr14    \n\t"
780
        "wunpckelub wr4, wr6            \n\t"
781
        "wunpckehub wr5, wr6            \n\t"
782
        "wunpckelub wr6, wr7            \n\t"
783
        "wunpckehub wr7, wr7            \n\t"
784
        "wunpckelub wr8, wr10           \n\t"
785
        "wunpckehub wr9, wr10           \n\t"
786
        "wunpckelub wr10, wr11          \n\t"
787
        "wunpckehub wr11, wr11          \n\t"
788
        "waddhus wr4, wr4, wr8          \n\t"
789
        "waddhus wr5, wr5, wr9          \n\t"
790
        "waddhus wr6, wr6, wr10         \n\t"
791
        "waddhus wr7, wr7, wr11         \n\t"
792
        "waddhus wr8, wr0, wr4          \n\t"
793
        "waddhus wr9, wr1, wr5          \n\t"
794
        "waddhus wr10, wr2, wr6         \n\t"
795
        "waddhus wr11, wr3, wr7         \n\t"
796
        "waddhus wr8, wr8, wr15         \n\t"
797
        "waddhus wr9, wr9, wr15         \n\t"
798
        "waddhus wr10, wr10, wr15       \n\t"
799
        "waddhus wr11, wr11, wr15       \n\t"
800
        "wsrlhg wr8, wr8, wcgr0         \n\t"
801
        "wsrlhg wr9, wr9, wcgr0         \n\t"
802
        "wsrlhg wr10, wr10, wcgr0       \n\t"
803
        "wsrlhg wr11, wr11, wcgr0       \n\t"
804
        "wpackhus wr8, wr8, wr9         \n\t"
805
        "wpackhus wr9, wr10, wr11       \n\t"
806
        "wstrd wr8, [%[block]]          \n\t"
807
        "wstrd wr9, [%[block], #8]      \n\t"
808
        "add %[block], %[block], %[line_size]   \n\t"
809

    
810
        // [wr0 wr1 wr2 wr3] <= *
811
        // [wr4 wr5 wr6 wr7]
812
        "wldrd wr12, [%[pixels]]        \n\t"
813
        "wldrd wr13, [%[pixels], #8]    \n\t"
814
        "wldrd wr14, [%[pixels], #16]   \n\t"
815
        "add %[pixels], %[pixels], %[line_size] \n\t"
816
        "walignr1 wr2, wr12, wr13       \n\t"
817
        "pld [%[pixels]]                \n\t"
818
        "pld [%[pixels], #32]           \n\t"
819
        "walignr1 wr3, wr13, wr14       \n\t"
820
        "wmoveq wr10, wr13              \n\t"
821
        "wmoveq wr11, wr14              \n\t"
822
        "walignr2ne wr10, wr12, wr13    \n\t"
823
        "walignr2ne wr11, wr13, wr14    \n\t"
824
        "wunpckelub wr0, wr2            \n\t"
825
        "wunpckehub wr1, wr2            \n\t"
826
        "wunpckelub wr2, wr3            \n\t"
827
        "wunpckehub wr3, wr3            \n\t"
828
        "wunpckelub wr8, wr10           \n\t"
829
        "wunpckehub wr9, wr10           \n\t"
830
        "wunpckelub wr10, wr11          \n\t"
831
        "wunpckehub wr11, wr11          \n\t"
832
        "waddhus wr0, wr0, wr8          \n\t"
833
        "waddhus wr1, wr1, wr9          \n\t"
834
        "waddhus wr2, wr2, wr10         \n\t"
835
        "waddhus wr3, wr3, wr11         \n\t"
836
        "waddhus wr8, wr0, wr4          \n\t"
837
        "waddhus wr9, wr1, wr5          \n\t"
838
        "waddhus wr10, wr2, wr6         \n\t"
839
        "waddhus wr11, wr3, wr7         \n\t"
840
        "waddhus wr8, wr8, wr15         \n\t"
841
        "waddhus wr9, wr9, wr15         \n\t"
842
        "waddhus wr10, wr10, wr15       \n\t"
843
        "waddhus wr11, wr11, wr15       \n\t"
844
        "wsrlhg wr8, wr8, wcgr0         \n\t"
845
        "wsrlhg wr9, wr9, wcgr0         \n\t"
846
        "wsrlhg wr10, wr10, wcgr0       \n\t"
847
        "wsrlhg wr11, wr11, wcgr0       \n\t"
848
        "wpackhus wr8, wr8, wr9         \n\t"
849
        "wpackhus wr9, wr10, wr11       \n\t"
850
        "wstrd wr8, [%[block]]          \n\t"
851
        "wstrd wr9, [%[block], #8]      \n\t"
852
        "add %[block], %[block], %[line_size]   \n\t"
853

    
854
        "subs %[h], %[h], #2            \n\t"
855
        "bne 1b                         \n\t"
856
        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block)
857
        : [line_size]"r"(line_size)
858
        : "r12", "memory");
859
}
860

    
861
void DEF(avg, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
862
{
863
    // [wr0 wr1 wr2 wr3] for previous line
864
    // [wr4 wr5 wr6 wr7] for current line
865
    SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
866
    __asm__ __volatile__(
867
        "pld [%[block]]                 \n\t"
868
        "pld [%[block], #32]            \n\t"
869
        "pld [%[pixels]]                \n\t"
870
        "mov r12, #2                    \n\t"
871
        "pld [%[pixels], #32]           \n\t"
872
        "tmcr wcgr0, r12                \n\t" /* for shift value */
873
        "and r12, %[pixels], #7         \n\t"
874
        "bic %[pixels], %[pixels], #7   \n\t"
875
        "tmcr wcgr1, r12                \n\t"
876

    
877
        // [wr0 wr1 wr2 wr3] <= *
878
        // [wr4 wr5 wr6 wr7]
879
        "wldrd wr12, [%[pixels]]        \n\t"
880
        "add r12, r12, #1               \n\t"
881
        "wldrd wr13, [%[pixels], #8]    \n\t"
882
        "tmcr wcgr2, r12                \n\t"
883
        "add %[pixels], %[pixels], %[line_size] \n\t"
884
        "cmp r12, #8                    \n\t"
885
        "pld [%[pixels]]                \n\t"
886
        "pld [%[pixels], #32]           \n\t"
887
        "walignr1 wr2, wr12, wr13       \n\t"
888
        "wmoveq wr10, wr13              \n\t"
889
        "walignr2ne wr10, wr12, wr13    \n\t"
890
        "wunpckelub wr0, wr2            \n\t"
891
        "wunpckehub wr1, wr2            \n\t"
892
        "wunpckelub wr8, wr10           \n\t"
893
        "wunpckehub wr9, wr10           \n\t"
894
        "waddhus wr0, wr0, wr8          \n\t"
895
        "waddhus wr1, wr1, wr9          \n\t"
896

    
897
        "1:                             \n\t"
898
        // [wr0 wr1 wr2 wr3]
899
        // [wr4 wr5 wr6 wr7] <= *
900
        "wldrd wr12, [%[pixels]]        \n\t"
901
        "cmp r12, #8                    \n\t"
902
        "wldrd wr13, [%[pixels], #8]    \n\t"
903
        "add %[pixels], %[pixels], %[line_size] \n\t"
904
        "walignr1 wr6, wr12, wr13       \n\t"
905
        "pld [%[pixels]]                \n\t"
906
        "pld [%[pixels], #32]           \n\t"
907
        "wmoveq wr10, wr13              \n\t"
908
        "walignr2ne wr10, wr12, wr13    \n\t"
909
        "wunpckelub wr4, wr6            \n\t"
910
        "wunpckehub wr5, wr6            \n\t"
911
        "wunpckelub wr8, wr10           \n\t"
912
        "wunpckehub wr9, wr10           \n\t"
913
        "waddhus wr4, wr4, wr8          \n\t"
914
        "waddhus wr5, wr5, wr9          \n\t"
915
        "waddhus wr8, wr0, wr4          \n\t"
916
        "waddhus wr9, wr1, wr5          \n\t"
917
        "waddhus wr8, wr8, wr15         \n\t"
918
        "waddhus wr9, wr9, wr15         \n\t"
919
        "wldrd wr12, [%[block]]         \n\t"
920
        "wsrlhg wr8, wr8, wcgr0         \n\t"
921
        "wsrlhg wr9, wr9, wcgr0         \n\t"
922
        "wpackhus wr8, wr8, wr9         \n\t"
923
        WAVG2B" wr8, wr8, wr12          \n\t"
924
        "wstrd wr8, [%[block]]          \n\t"
925
        "add %[block], %[block], %[line_size]   \n\t"
926
        "wldrd wr12, [%[pixels]]        \n\t"
927
        "pld [%[block]]                 \n\t"
928
        "pld [%[block], #32]            \n\t"
929

    
930
        // [wr0 wr1 wr2 wr3] <= *
931
        // [wr4 wr5 wr6 wr7]
932
        "wldrd wr13, [%[pixels], #8]    \n\t"
933
        "add %[pixels], %[pixels], %[line_size] \n\t"
934
        "walignr1 wr2, wr12, wr13       \n\t"
935
        "pld [%[pixels]]                \n\t"
936
        "pld [%[pixels], #32]           \n\t"
937
        "wmoveq wr10, wr13              \n\t"
938
        "walignr2ne wr10, wr12, wr13    \n\t"
939
        "wunpckelub wr0, wr2            \n\t"
940
        "wunpckehub wr1, wr2            \n\t"
941
        "wunpckelub wr8, wr10           \n\t"
942
        "wunpckehub wr9, wr10           \n\t"
943
        "waddhus wr0, wr0, wr8          \n\t"
944
        "waddhus wr1, wr1, wr9          \n\t"
945
        "waddhus wr8, wr0, wr4          \n\t"
946
        "waddhus wr9, wr1, wr5          \n\t"
947
        "waddhus wr8, wr8, wr15         \n\t"
948
        "waddhus wr9, wr9, wr15         \n\t"
949
        "wldrd wr12, [%[block]]         \n\t"
950
        "wsrlhg wr8, wr8, wcgr0         \n\t"
951
        "wsrlhg wr9, wr9, wcgr0         \n\t"
952
        "wpackhus wr8, wr8, wr9         \n\t"
953
        "subs %[h], %[h], #2            \n\t"
954
        WAVG2B" wr8, wr8, wr12          \n\t"
955
        "wstrd wr8, [%[block]]          \n\t"
956
        "add %[block], %[block], %[line_size]   \n\t"
957
        "pld [%[block]]                 \n\t"
958
        "pld [%[block], #32]            \n\t"
959
        "bne 1b                         \n\t"
960
        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block)
961
        : [line_size]"r"(line_size)
962
        : "r12", "memory");
963
}
964

    
965
void DEF(avg, pixels16_xy2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
966
{
967
    // [wr0 wr1 wr2 wr3] for previous line
968
    // [wr4 wr5 wr6 wr7] for current line
969
    SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
970
    __asm__ __volatile__(
971
        "pld [%[block]]                 \n\t"
972
        "pld [%[block], #32]            \n\t"
973
        "pld [%[pixels]]                \n\t"
974
        "mov r12, #2                    \n\t"
975
        "pld [%[pixels], #32]           \n\t"
976
        "tmcr wcgr0, r12                \n\t" /* for shift value */
977
        /* alignment */
978
        "and r12, %[pixels], #7         \n\t"
979
        "bic %[pixels], %[pixels], #7           \n\t"
980
        "tmcr wcgr1, r12                \n\t"
981
        "add r12, r12, #1               \n\t"
982
        "tmcr wcgr2, r12                \n\t"
983

    
984
        // [wr0 wr1 wr2 wr3] <= *
985
        // [wr4 wr5 wr6 wr7]
986
        "wldrd wr12, [%[pixels]]        \n\t"
987
        "cmp r12, #8                    \n\t"
988
        "wldrd wr13, [%[pixels], #8]    \n\t"
989
        "wldrd wr14, [%[pixels], #16]   \n\t"
990
        "add %[pixels], %[pixels], %[line_size] \n\t"
991
        "pld [%[pixels]]                \n\t"
992
        "walignr1 wr2, wr12, wr13       \n\t"
993
        "pld [%[pixels], #32]           \n\t"
994
        "walignr1 wr3, wr13, wr14       \n\t"
995
        "wmoveq wr10, wr13              \n\t"
996
        "wmoveq wr11, wr14              \n\t"
997
        "walignr2ne wr10, wr12, wr13    \n\t"
998
        "walignr2ne wr11, wr13, wr14    \n\t"
999
        "wunpckelub wr0, wr2            \n\t"
1000
        "wunpckehub wr1, wr2            \n\t"
1001
        "wunpckelub wr2, wr3            \n\t"
1002
        "wunpckehub wr3, wr3            \n\t"
1003
        "wunpckelub wr8, wr10           \n\t"
1004
        "wunpckehub wr9, wr10           \n\t"
1005
        "wunpckelub wr10, wr11          \n\t"
1006
        "wunpckehub wr11, wr11          \n\t"
1007
        "waddhus wr0, wr0, wr8          \n\t"
1008
        "waddhus wr1, wr1, wr9          \n\t"
1009
        "waddhus wr2, wr2, wr10         \n\t"
1010
        "waddhus wr3, wr3, wr11         \n\t"
1011

    
1012
        "1:                             \n\t"
1013
        // [wr0 wr1 wr2 wr3]
1014
        // [wr4 wr5 wr6 wr7] <= *
1015
        "wldrd wr12, [%[pixels]]        \n\t"
1016
        "cmp r12, #8                    \n\t"
1017
        "wldrd wr13, [%[pixels], #8]    \n\t"
1018
        "wldrd wr14, [%[pixels], #16]   \n\t"
1019
        "add %[pixels], %[pixels], %[line_size] \n\t"
1020
        "walignr1 wr6, wr12, wr13       \n\t"
1021
        "pld [%[pixels]]                \n\t"
1022
        "pld [%[pixels], #32]           \n\t"
1023
        "walignr1 wr7, wr13, wr14       \n\t"
1024
        "wmoveq wr10, wr13              \n\t"
1025
        "wmoveq wr11, wr14              \n\t"
1026
        "walignr2ne wr10, wr12, wr13    \n\t"
1027
        "walignr2ne wr11, wr13, wr14    \n\t"
1028
        "wunpckelub wr4, wr6            \n\t"
1029
        "wunpckehub wr5, wr6            \n\t"
1030
        "wunpckelub wr6, wr7            \n\t"
1031
        "wunpckehub wr7, wr7            \n\t"
1032
        "wunpckelub wr8, wr10           \n\t"
1033
        "wunpckehub wr9, wr10           \n\t"
1034
        "wunpckelub wr10, wr11          \n\t"
1035
        "wunpckehub wr11, wr11          \n\t"
1036
        "waddhus wr4, wr4, wr8          \n\t"
1037
        "waddhus wr5, wr5, wr9          \n\t"
1038
        "waddhus wr6, wr6, wr10         \n\t"
1039
        "waddhus wr7, wr7, wr11         \n\t"
1040
        "waddhus wr8, wr0, wr4          \n\t"
1041
        "waddhus wr9, wr1, wr5          \n\t"
1042
        "waddhus wr10, wr2, wr6         \n\t"
1043
        "waddhus wr11, wr3, wr7         \n\t"
1044
        "waddhus wr8, wr8, wr15         \n\t"
1045
        "waddhus wr9, wr9, wr15         \n\t"
1046
        "waddhus wr10, wr10, wr15       \n\t"
1047
        "waddhus wr11, wr11, wr15       \n\t"
1048
        "wsrlhg wr8, wr8, wcgr0         \n\t"
1049
        "wsrlhg wr9, wr9, wcgr0         \n\t"
1050
        "wldrd wr12, [%[block]]         \n\t"
1051
        "wldrd wr13, [%[block], #8]     \n\t"
1052
        "wsrlhg wr10, wr10, wcgr0       \n\t"
1053
        "wsrlhg wr11, wr11, wcgr0       \n\t"
1054
        "wpackhus wr8, wr8, wr9         \n\t"
1055
        "wpackhus wr9, wr10, wr11       \n\t"
1056
        WAVG2B" wr8, wr8, wr12          \n\t"
1057
        WAVG2B" wr9, wr9, wr13          \n\t"
1058
        "wstrd wr8, [%[block]]          \n\t"
1059
        "wstrd wr9, [%[block], #8]      \n\t"
1060
        "add %[block], %[block], %[line_size]   \n\t"
1061

    
1062
        // [wr0 wr1 wr2 wr3] <= *
1063
        // [wr4 wr5 wr6 wr7]
1064
        "wldrd wr12, [%[pixels]]        \n\t"
1065
        "pld [%[block]]                 \n\t"
1066
        "wldrd wr13, [%[pixels], #8]    \n\t"
1067
        "pld [%[block], #32]            \n\t"
1068
        "wldrd wr14, [%[pixels], #16]   \n\t"
1069
        "add %[pixels], %[pixels], %[line_size] \n\t"
1070
        "walignr1 wr2, wr12, wr13       \n\t"
1071
        "pld [%[pixels]]                \n\t"
1072
        "pld [%[pixels], #32]           \n\t"
1073
        "walignr1 wr3, wr13, wr14       \n\t"
1074
        "wmoveq wr10, wr13              \n\t"
1075
        "wmoveq wr11, wr14              \n\t"
1076
        "walignr2ne wr10, wr12, wr13    \n\t"
1077
        "walignr2ne wr11, wr13, wr14    \n\t"
1078
        "wunpckelub wr0, wr2            \n\t"
1079
        "wunpckehub wr1, wr2            \n\t"
1080
        "wunpckelub wr2, wr3            \n\t"
1081
        "wunpckehub wr3, wr3            \n\t"
1082
        "wunpckelub wr8, wr10           \n\t"
1083
        "wunpckehub wr9, wr10           \n\t"
1084
        "wunpckelub wr10, wr11          \n\t"
1085
        "wunpckehub wr11, wr11          \n\t"
1086
        "waddhus wr0, wr0, wr8          \n\t"
1087
        "waddhus wr1, wr1, wr9          \n\t"
1088
        "waddhus wr2, wr2, wr10         \n\t"
1089
        "waddhus wr3, wr3, wr11         \n\t"
1090
        "waddhus wr8, wr0, wr4          \n\t"
1091
        "waddhus wr9, wr1, wr5          \n\t"
1092
        "waddhus wr10, wr2, wr6         \n\t"
1093
        "waddhus wr11, wr3, wr7         \n\t"
1094
        "waddhus wr8, wr8, wr15         \n\t"
1095
        "waddhus wr9, wr9, wr15         \n\t"
1096
        "waddhus wr10, wr10, wr15       \n\t"
1097
        "waddhus wr11, wr11, wr15       \n\t"
1098
        "wsrlhg wr8, wr8, wcgr0         \n\t"
1099
        "wsrlhg wr9, wr9, wcgr0         \n\t"
1100
        "wldrd wr12, [%[block]]         \n\t"
1101
        "wldrd wr13, [%[block], #8]     \n\t"
1102
        "wsrlhg wr10, wr10, wcgr0       \n\t"
1103
        "wsrlhg wr11, wr11, wcgr0       \n\t"
1104
        "wpackhus wr8, wr8, wr9         \n\t"
1105
        "wpackhus wr9, wr10, wr11       \n\t"
1106
        WAVG2B" wr8, wr8, wr12          \n\t"
1107
        WAVG2B" wr9, wr9, wr13          \n\t"
1108
        "wstrd wr8, [%[block]]          \n\t"
1109
        "wstrd wr9, [%[block], #8]      \n\t"
1110
        "add %[block], %[block], %[line_size]   \n\t"
1111
        "subs %[h], %[h], #2            \n\t"
1112
        "pld [%[block]]                 \n\t"
1113
        "pld [%[block], #32]            \n\t"
1114
        "bne 1b                         \n\t"
1115
        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block)
1116
        : [line_size]"r"(line_size)
1117
        : "r12", "memory");
1118
}