51ba61c47ac038638c13c7d114cf954066b039a2
[libav.git] / libavcodec / armv4l / dsputil_iwmmxt_rnd.h
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 void DEF(put, pixels8)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
23 {
24 int stride = line_size;
25 __asm__ __volatile__ (
26 "and r12, %[pixels], #7 \n\t"
27 "bic %[pixels], %[pixels], #7 \n\t"
28 "tmcr wcgr1, r12 \n\t"
29 "add r4, %[pixels], %[line_size] \n\t"
30 "add r5, %[block], %[line_size] \n\t"
31 "mov %[line_size], %[line_size], lsl #1 \n\t"
32 "1: \n\t"
33 "wldrd wr0, [%[pixels]] \n\t"
34 "subs %[h], %[h], #2 \n\t"
35 "wldrd wr1, [%[pixels], #8] \n\t"
36 "add %[pixels], %[pixels], %[line_size] \n\t"
37 "wldrd wr3, [r4] \n\t"
38 "pld [%[pixels]] \n\t"
39 "pld [%[pixels], #32] \n\t"
40 "wldrd wr4, [r4, #8] \n\t"
41 "add r4, r4, %[line_size] \n\t"
42 "walignr1 wr8, wr0, wr1 \n\t"
43 "pld [r4] \n\t"
44 "pld [r4, #32] \n\t"
45 "walignr1 wr10, wr3, wr4 \n\t"
46 "wstrd wr8, [%[block]] \n\t"
47 "add %[block], %[block], %[line_size] \n\t"
48 "wstrd wr10, [r5] \n\t"
49 "add r5, r5, %[line_size] \n\t"
50 "bne 1b \n\t"
51 : [block]"+r"(block), [pixels]"+r"(pixels), [line_size]"+r"(stride), [h]"+r"(h)
52 :
53 : "memory", "r4", "r5", "r12");
54 }
55
56 void DEF(avg, pixels8)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
57 {
58 int stride = line_size;
59 __asm__ __volatile__ (
60 "and r12, %[pixels], #7 \n\t"
61 "bic %[pixels], %[pixels], #7 \n\t"
62 "tmcr wcgr1, r12 \n\t"
63 "add r4, %[pixels], %[line_size] \n\t"
64 "add r5, %[block], %[line_size] \n\t"
65 "mov %[line_size], %[line_size], lsl #1 \n\t"
66 "1: \n\t"
67 "wldrd wr0, [%[pixels]] \n\t"
68 "subs %[h], %[h], #2 \n\t"
69 "wldrd wr1, [%[pixels], #8] \n\t"
70 "add %[pixels], %[pixels], %[line_size] \n\t"
71 "wldrd wr3, [r4] \n\t"
72 "pld [%[pixels]] \n\t"
73 "pld [%[pixels], #32] \n\t"
74 "wldrd wr4, [r4, #8] \n\t"
75 "add r4, r4, %[line_size] \n\t"
76 "walignr1 wr8, wr0, wr1 \n\t"
77 "wldrd wr0, [%[block]] \n\t"
78 "wldrd wr2, [r5] \n\t"
79 "pld [r4] \n\t"
80 "pld [r4, #32] \n\t"
81 "walignr1 wr10, wr3, wr4 \n\t"
82 WAVG2B" wr8, wr8, wr0 \n\t"
83 WAVG2B" wr10, wr10, wr2 \n\t"
84 "wstrd wr8, [%[block]] \n\t"
85 "add %[block], %[block], %[line_size] \n\t"
86 "wstrd wr10, [r5] \n\t"
87 "pld [%[block]] \n\t"
88 "pld [%[block], #32] \n\t"
89 "add r5, r5, %[line_size] \n\t"
90 "pld [r5] \n\t"
91 "pld [r5, #32] \n\t"
92 "bne 1b \n\t"
93 : [block]"+r"(block), [pixels]"+r"(pixels), [line_size]"+r"(stride), [h]"+r"(h)
94 :
95 : "memory", "r4", "r5", "r12");
96 }
97
98 void DEF(put, pixels16)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
99 {
100 int stride = line_size;
101 __asm__ __volatile__ (
102 "and r12, %[pixels], #7 \n\t"
103 "bic %[pixels], %[pixels], #7 \n\t"
104 "tmcr wcgr1, r12 \n\t"
105 "add r4, %[pixels], %[line_size] \n\t"
106 "add r5, %[block], %[line_size] \n\t"
107 "mov %[line_size], %[line_size], lsl #1 \n\t"
108 "1: \n\t"
109 "wldrd wr0, [%[pixels]] \n\t"
110 "wldrd wr1, [%[pixels], #8] \n\t"
111 "subs %[h], %[h], #2 \n\t"
112 "wldrd wr2, [%[pixels], #16] \n\t"
113 "add %[pixels], %[pixels], %[line_size] \n\t"
114 "wldrd wr3, [r4] \n\t"
115 "pld [%[pixels]] \n\t"
116 "pld [%[pixels], #32] \n\t"
117 "walignr1 wr8, wr0, wr1 \n\t"
118 "wldrd wr4, [r4, #8] \n\t"
119 "walignr1 wr9, wr1, wr2 \n\t"
120 "wldrd wr5, [r4, #16] \n\t"
121 "add r4, r4, %[line_size] \n\t"
122 "pld [r4] \n\t"
123 "pld [r4, #32] \n\t"
124 "walignr1 wr10, wr3, wr4 \n\t"
125 "wstrd wr8, [%[block]] \n\t"
126 "walignr1 wr11, wr4, wr5 \n\t"
127 "wstrd wr9, [%[block], #8] \n\t"
128 "add %[block], %[block], %[line_size] \n\t"
129 "wstrd wr10, [r5] \n\t"
130 "wstrd wr11, [r5, #8] \n\t"
131 "add r5, r5, %[line_size] \n\t"
132 "bne 1b \n\t"
133 : [block]"+r"(block), [pixels]"+r"(pixels), [line_size]"+r"(stride), [h]"+r"(h)
134 :
135 : "memory", "r4", "r5", "r12");
136 }
137
138 void DEF(avg, pixels16)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
139 {
140 int stride = line_size;
141 __asm__ __volatile__ (
142 "pld [%[pixels]] \n\t"
143 "pld [%[pixels], #32] \n\t"
144 "pld [%[block]] \n\t"
145 "pld [%[block], #32] \n\t"
146 "and r12, %[pixels], #7 \n\t"
147 "bic %[pixels], %[pixels], #7 \n\t"
148 "tmcr wcgr1, r12 \n\t"
149 "add r4, %[pixels], %[line_size]\n\t"
150 "add r5, %[block], %[line_size] \n\t"
151 "mov %[line_size], %[line_size], lsl #1 \n\t"
152 "1: \n\t"
153 "wldrd wr0, [%[pixels]] \n\t"
154 "wldrd wr1, [%[pixels], #8] \n\t"
155 "subs %[h], %[h], #2 \n\t"
156 "wldrd wr2, [%[pixels], #16] \n\t"
157 "add %[pixels], %[pixels], %[line_size] \n\t"
158 "wldrd wr3, [r4] \n\t"
159 "pld [%[pixels]] \n\t"
160 "pld [%[pixels], #32] \n\t"
161 "walignr1 wr8, wr0, wr1 \n\t"
162 "wldrd wr4, [r4, #8] \n\t"
163 "walignr1 wr9, wr1, wr2 \n\t"
164 "wldrd wr5, [r4, #16] \n\t"
165 "add r4, r4, %[line_size] \n\t"
166 "wldrd wr0, [%[block]] \n\t"
167 "pld [r4] \n\t"
168 "wldrd wr1, [%[block], #8] \n\t"
169 "pld [r4, #32] \n\t"
170 "wldrd wr2, [r5] \n\t"
171 "walignr1 wr10, wr3, wr4 \n\t"
172 "wldrd wr3, [r5, #8] \n\t"
173 WAVG2B" wr8, wr8, wr0 \n\t"
174 WAVG2B" wr9, wr9, wr1 \n\t"
175 WAVG2B" wr10, wr10, wr2 \n\t"
176 "wstrd wr8, [%[block]] \n\t"
177 "walignr1 wr11, wr4, wr5 \n\t"
178 WAVG2B" wr11, wr11, wr3 \n\t"
179 "wstrd wr9, [%[block], #8] \n\t"
180 "add %[block], %[block], %[line_size] \n\t"
181 "wstrd wr10, [r5] \n\t"
182 "pld [%[block]] \n\t"
183 "pld [%[block], #32] \n\t"
184 "wstrd wr11, [r5, #8] \n\t"
185 "add r5, r5, %[line_size] \n\t"
186 "pld [r5] \n\t"
187 "pld [r5, #32] \n\t"
188 "bne 1b \n\t"
189 : [block]"+r"(block), [pixels]"+r"(pixels), [line_size]"+r"(stride), [h]"+r"(h)
190 :
191 : "memory", "r4", "r5", "r12");
192 }
193
194 void DEF(put, pixels8_x2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
195 {
196 int stride = line_size;
197 // [wr0 wr1 wr2 wr3] for previous line
198 // [wr4 wr5 wr6 wr7] for current line
199 SET_RND(wr15); // =2 for rnd and =1 for no_rnd version
200 __asm__ __volatile__(
201 "pld [%[pixels]] \n\t"
202 "pld [%[pixels], #32] \n\t"
203 "and r12, %[pixels], #7 \n\t"
204 "bic %[pixels], %[pixels], #7 \n\t"
205 "tmcr wcgr1, r12 \n\t"
206 "add r12, r12, #1 \n\t"
207 "add r4, %[pixels], %[line_size]\n\t"
208 "tmcr wcgr2, r12 \n\t"
209 "add r5, %[block], %[line_size] \n\t"
210 "mov %[line_size], %[line_size], lsl #1 \n\t"
211
212 "1: \n\t"
213 "wldrd wr10, [%[pixels]] \n\t"
214 "cmp r12, #8 \n\t"
215 "wldrd wr11, [%[pixels], #8] \n\t"
216 "add %[pixels], %[pixels], %[line_size] \n\t"
217 "wldrd wr13, [r4] \n\t"
218 "pld [%[pixels]] \n\t"
219 "wldrd wr14, [r4, #8] \n\t"
220 "pld [%[pixels], #32] \n\t"
221 "add r4, r4, %[line_size] \n\t"
222 "walignr1 wr0, wr10, wr11 \n\t"
223 "pld [r4] \n\t"
224 "pld [r4, #32] \n\t"
225 "walignr1 wr2, wr13, wr14 \n\t"
226 "wmoveq wr4, wr11 \n\t"
227 "wmoveq wr6, wr14 \n\t"
228 "walignr2ne wr4, wr10, wr11 \n\t"
229 "walignr2ne wr6, wr13, wr14 \n\t"
230 WAVG2B" wr0, wr0, wr4 \n\t"
231 WAVG2B" wr2, wr2, wr6 \n\t"
232 "wstrd wr0, [%[block]] \n\t"
233 "subs %[h], %[h], #2 \n\t"
234 "wstrd wr2, [r5] \n\t"
235 "add %[block], %[block], %[line_size] \n\t"
236 "add r5, r5, %[line_size] \n\t"
237 "bne 1b \n\t"
238 : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
239 :
240 : "r4", "r5", "r12", "memory");
241 }
242
243 void DEF(put, pixels16_x2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
244 {
245 int stride = line_size;
246 // [wr0 wr1 wr2 wr3] for previous line
247 // [wr4 wr5 wr6 wr7] for current line
248 SET_RND(wr15); // =2 for rnd and =1 for no_rnd version
249 __asm__ __volatile__(
250 "pld [%[pixels]] \n\t"
251 "pld [%[pixels], #32] \n\t"
252 "and r12, %[pixels], #7 \n\t"
253 "bic %[pixels], %[pixels], #7 \n\t"
254 "tmcr wcgr1, r12 \n\t"
255 "add r12, r12, #1 \n\t"
256 "add r4, %[pixels], %[line_size]\n\t"
257 "tmcr wcgr2, r12 \n\t"
258 "add r5, %[block], %[line_size] \n\t"
259 "mov %[line_size], %[line_size], lsl #1 \n\t"
260
261 "1: \n\t"
262 "wldrd wr10, [%[pixels]] \n\t"
263 "cmp r12, #8 \n\t"
264 "wldrd wr11, [%[pixels], #8] \n\t"
265 "wldrd wr12, [%[pixels], #16] \n\t"
266 "add %[pixels], %[pixels], %[line_size] \n\t"
267 "wldrd wr13, [r4] \n\t"
268 "pld [%[pixels]] \n\t"
269 "wldrd wr14, [r4, #8] \n\t"
270 "pld [%[pixels], #32] \n\t"
271 "wldrd wr15, [r4, #16] \n\t"
272 "add r4, r4, %[line_size] \n\t"
273 "walignr1 wr0, wr10, wr11 \n\t"
274 "pld [r4] \n\t"
275 "pld [r4, #32] \n\t"
276 "walignr1 wr1, wr11, wr12 \n\t"
277 "walignr1 wr2, wr13, wr14 \n\t"
278 "walignr1 wr3, wr14, wr15 \n\t"
279 "wmoveq wr4, wr11 \n\t"
280 "wmoveq wr5, wr12 \n\t"
281 "wmoveq wr6, wr14 \n\t"
282 "wmoveq wr7, wr15 \n\t"
283 "walignr2ne wr4, wr10, wr11 \n\t"
284 "walignr2ne wr5, wr11, wr12 \n\t"
285 "walignr2ne wr6, wr13, wr14 \n\t"
286 "walignr2ne wr7, wr14, wr15 \n\t"
287 WAVG2B" wr0, wr0, wr4 \n\t"
288 WAVG2B" wr1, wr1, wr5 \n\t"
289 "wstrd wr0, [%[block]] \n\t"
290 WAVG2B" wr2, wr2, wr6 \n\t"
291 "wstrd wr1, [%[block], #8] \n\t"
292 WAVG2B" wr3, wr3, wr7 \n\t"
293 "add %[block], %[block], %[line_size] \n\t"
294 "wstrd wr2, [r5] \n\t"
295 "subs %[h], %[h], #2 \n\t"
296 "wstrd wr3, [r5, #8] \n\t"
297 "add r5, r5, %[line_size] \n\t"
298 "bne 1b \n\t"
299 : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
300 :
301 : "r4", "r5", "r12", "memory");
302 }
303
304 void DEF(avg, pixels8_x2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
305 {
306 int stride = line_size;
307 // [wr0 wr1 wr2 wr3] for previous line
308 // [wr4 wr5 wr6 wr7] for current line
309 SET_RND(wr15); // =2 for rnd and =1 for no_rnd version
310 __asm__ __volatile__(
311 "pld [%[pixels]] \n\t"
312 "pld [%[pixels], #32] \n\t"
313 "pld [%[block]] \n\t"
314 "pld [%[block], #32] \n\t"
315 "and r12, %[pixels], #7 \n\t"
316 "bic %[pixels], %[pixels], #7 \n\t"
317 "tmcr wcgr1, r12 \n\t"
318 "add r12, r12, #1 \n\t"
319 "add r4, %[pixels], %[line_size]\n\t"
320 "tmcr wcgr2, r12 \n\t"
321 "add r5, %[block], %[line_size] \n\t"
322 "mov %[line_size], %[line_size], lsl #1 \n\t"
323 "pld [r5] \n\t"
324 "pld [r5, #32] \n\t"
325
326 "1: \n\t"
327 "wldrd wr10, [%[pixels]] \n\t"
328 "cmp r12, #8 \n\t"
329 "wldrd wr11, [%[pixels], #8] \n\t"
330 "add %[pixels], %[pixels], %[line_size] \n\t"
331 "wldrd wr13, [r4] \n\t"
332 "pld [%[pixels]] \n\t"
333 "wldrd wr14, [r4, #8] \n\t"
334 "pld [%[pixels], #32] \n\t"
335 "add r4, r4, %[line_size] \n\t"
336 "walignr1 wr0, wr10, wr11 \n\t"
337 "pld [r4] \n\t"
338 "pld [r4, #32] \n\t"
339 "walignr1 wr2, wr13, wr14 \n\t"
340 "wmoveq wr4, wr11 \n\t"
341 "wmoveq wr6, wr14 \n\t"
342 "walignr2ne wr4, wr10, wr11 \n\t"
343 "wldrd wr10, [%[block]] \n\t"
344 "walignr2ne wr6, wr13, wr14 \n\t"
345 "wldrd wr12, [r5] \n\t"
346 WAVG2B" wr0, wr0, wr4 \n\t"
347 WAVG2B" wr2, wr2, wr6 \n\t"
348 WAVG2B" wr0, wr0, wr10 \n\t"
349 WAVG2B" wr2, wr2, wr12 \n\t"
350 "wstrd wr0, [%[block]] \n\t"
351 "subs %[h], %[h], #2 \n\t"
352 "wstrd wr2, [r5] \n\t"
353 "add %[block], %[block], %[line_size] \n\t"
354 "add r5, r5, %[line_size] \n\t"
355 "pld [%[block]] \n\t"
356 "pld [%[block], #32] \n\t"
357 "pld [r5] \n\t"
358 "pld [r5, #32] \n\t"
359 "bne 1b \n\t"
360 : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
361 :
362 : "r4", "r5", "r12", "memory");
363 }
364
365 void DEF(avg, pixels16_x2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
366 {
367 int stride = line_size;
368 // [wr0 wr1 wr2 wr3] for previous line
369 // [wr4 wr5 wr6 wr7] for current line
370 SET_RND(wr15); // =2 for rnd and =1 for no_rnd version
371 __asm__ __volatile__(
372 "pld [%[pixels]] \n\t"
373 "pld [%[pixels], #32] \n\t"
374 "pld [%[block]] \n\t"
375 "pld [%[block], #32] \n\t"
376 "and r12, %[pixels], #7 \n\t"
377 "bic %[pixels], %[pixels], #7 \n\t"
378 "tmcr wcgr1, r12 \n\t"
379 "add r12, r12, #1 \n\t"
380 "add r4, %[pixels], %[line_size]\n\t"
381 "tmcr wcgr2, r12 \n\t"
382 "add r5, %[block], %[line_size] \n\t"
383 "mov %[line_size], %[line_size], lsl #1 \n\t"
384 "pld [r5] \n\t"
385 "pld [r5, #32] \n\t"
386
387 "1: \n\t"
388 "wldrd wr10, [%[pixels]] \n\t"
389 "cmp r12, #8 \n\t"
390 "wldrd wr11, [%[pixels], #8] \n\t"
391 "wldrd wr12, [%[pixels], #16] \n\t"
392 "add %[pixels], %[pixels], %[line_size] \n\t"
393 "wldrd wr13, [r4] \n\t"
394 "pld [%[pixels]] \n\t"
395 "wldrd wr14, [r4, #8] \n\t"
396 "pld [%[pixels], #32] \n\t"
397 "wldrd wr15, [r4, #16] \n\t"
398 "add r4, r4, %[line_size] \n\t"
399 "walignr1 wr0, wr10, wr11 \n\t"
400 "pld [r4] \n\t"
401 "pld [r4, #32] \n\t"
402 "walignr1 wr1, wr11, wr12 \n\t"
403 "walignr1 wr2, wr13, wr14 \n\t"
404 "walignr1 wr3, wr14, wr15 \n\t"
405 "wmoveq wr4, wr11 \n\t"
406 "wmoveq wr5, wr12 \n\t"
407 "wmoveq wr6, wr14 \n\t"
408 "wmoveq wr7, wr15 \n\t"
409 "walignr2ne wr4, wr10, wr11 \n\t"
410 "walignr2ne wr5, wr11, wr12 \n\t"
411 "walignr2ne wr6, wr13, wr14 \n\t"
412 "walignr2ne wr7, wr14, wr15 \n\t"
413 "wldrd wr10, [%[block]] \n\t"
414 WAVG2B" wr0, wr0, wr4 \n\t"
415 "wldrd wr11, [%[block], #8] \n\t"
416 WAVG2B" wr1, wr1, wr5 \n\t"
417 "wldrd wr12, [r5] \n\t"
418 WAVG2B" wr2, wr2, wr6 \n\t"
419 "wldrd wr13, [r5, #8] \n\t"
420 WAVG2B" wr3, wr3, wr7 \n\t"
421 WAVG2B" wr0, wr0, wr10 \n\t"
422 WAVG2B" wr1, wr1, wr11 \n\t"
423 WAVG2B" wr2, wr2, wr12 \n\t"
424 WAVG2B" wr3, wr3, wr13 \n\t"
425 "wstrd wr0, [%[block]] \n\t"
426 "subs %[h], %[h], #2 \n\t"
427 "wstrd wr1, [%[block], #8] \n\t"
428 "add %[block], %[block], %[line_size] \n\t"
429 "wstrd wr2, [r5] \n\t"
430 "pld [%[block]] \n\t"
431 "wstrd wr3, [r5, #8] \n\t"
432 "add r5, r5, %[line_size] \n\t"
433 "pld [%[block], #32] \n\t"
434 "pld [r5] \n\t"
435 "pld [r5, #32] \n\t"
436 "bne 1b \n\t"
437 : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
438 :
439 :"r4", "r5", "r12", "memory");
440 }
441
442 void DEF(avg, pixels8_y2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
443 {
444 int stride = line_size;
445 // [wr0 wr1 wr2 wr3] for previous line
446 // [wr4 wr5 wr6 wr7] for current line
447 __asm__ __volatile__(
448 "pld [%[pixels]] \n\t"
449 "pld [%[pixels], #32] \n\t"
450 "and r12, %[pixels], #7 \n\t"
451 "tmcr wcgr1, r12 \n\t"
452 "bic %[pixels], %[pixels], #7 \n\t"
453
454 "wldrd wr10, [%[pixels]] \n\t"
455 "wldrd wr11, [%[pixels], #8] \n\t"
456 "pld [%[block]] \n\t"
457 "add %[pixels], %[pixels], %[line_size] \n\t"
458 "walignr1 wr0, wr10, wr11 \n\t"
459 "pld [%[pixels]] \n\t"
460 "pld [%[pixels], #32] \n\t"
461
462 "1: \n\t"
463 "wldrd wr10, [%[pixels]] \n\t"
464 "wldrd wr11, [%[pixels], #8] \n\t"
465 "add %[pixels], %[pixels], %[line_size] \n\t"
466 "pld [%[pixels]] \n\t"
467 "pld [%[pixels], #32] \n\t"
468 "walignr1 wr4, wr10, wr11 \n\t"
469 "wldrd wr10, [%[block]] \n\t"
470 WAVG2B" wr8, wr0, wr4 \n\t"
471 WAVG2B" wr8, wr8, wr10 \n\t"
472 "wstrd wr8, [%[block]] \n\t"
473 "add %[block], %[block], %[line_size] \n\t"
474
475 "wldrd wr10, [%[pixels]] \n\t"
476 "wldrd wr11, [%[pixels], #8] \n\t"
477 "pld [%[block]] \n\t"
478 "add %[pixels], %[pixels], %[line_size] \n\t"
479 "pld [%[pixels]] \n\t"
480 "pld [%[pixels], #32] \n\t"
481 "walignr1 wr0, wr10, wr11 \n\t"
482 "wldrd wr10, [%[block]] \n\t"
483 WAVG2B" wr8, wr0, wr4 \n\t"
484 WAVG2B" wr8, wr8, wr10 \n\t"
485 "wstrd wr8, [%[block]] \n\t"
486 "add %[block], %[block], %[line_size] \n\t"
487
488 "subs %[h], %[h], #2 \n\t"
489 "pld [%[block]] \n\t"
490 "bne 1b \n\t"
491 : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
492 :
493 : "cc", "memory", "r12");
494 }
495
496 void DEF(put, pixels16_y2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
497 {
498 int stride = line_size;
499 // [wr0 wr1 wr2 wr3] for previous line
500 // [wr4 wr5 wr6 wr7] for current line
501 __asm__ __volatile__(
502 "pld [%[pixels]] \n\t"
503 "pld [%[pixels], #32] \n\t"
504 "and r12, %[pixels], #7 \n\t"
505 "tmcr wcgr1, r12 \n\t"
506 "bic %[pixels], %[pixels], #7 \n\t"
507
508 "wldrd wr10, [%[pixels]] \n\t"
509 "wldrd wr11, [%[pixels], #8] \n\t"
510 "wldrd wr12, [%[pixels], #16] \n\t"
511 "add %[pixels], %[pixels], %[line_size] \n\t"
512 "pld [%[pixels]] \n\t"
513 "pld [%[pixels], #32] \n\t"
514 "walignr1 wr0, wr10, wr11 \n\t"
515 "walignr1 wr1, wr11, wr12 \n\t"
516
517 "1: \n\t"
518 "wldrd wr10, [%[pixels]] \n\t"
519 "wldrd wr11, [%[pixels], #8] \n\t"
520 "wldrd wr12, [%[pixels], #16] \n\t"
521 "add %[pixels], %[pixels], %[line_size] \n\t"
522 "pld [%[pixels]] \n\t"
523 "pld [%[pixels], #32] \n\t"
524 "walignr1 wr4, wr10, wr11 \n\t"
525 "walignr1 wr5, wr11, wr12 \n\t"
526 WAVG2B" wr8, wr0, wr4 \n\t"
527 WAVG2B" wr9, wr1, wr5 \n\t"
528 "wstrd wr8, [%[block]] \n\t"
529 "wstrd wr9, [%[block], #8] \n\t"
530 "add %[block], %[block], %[line_size] \n\t"
531
532 "wldrd wr10, [%[pixels]] \n\t"
533 "wldrd wr11, [%[pixels], #8] \n\t"
534 "wldrd wr12, [%[pixels], #16] \n\t"
535 "add %[pixels], %[pixels], %[line_size] \n\t"
536 "pld [%[pixels]] \n\t"
537 "pld [%[pixels], #32] \n\t"
538 "walignr1 wr0, wr10, wr11 \n\t"
539 "walignr1 wr1, wr11, wr12 \n\t"
540 WAVG2B" wr8, wr0, wr4 \n\t"
541 WAVG2B" wr9, wr1, wr5 \n\t"
542 "wstrd wr8, [%[block]] \n\t"
543 "wstrd wr9, [%[block], #8] \n\t"
544 "add %[block], %[block], %[line_size] \n\t"
545
546 "subs %[h], %[h], #2 \n\t"
547 "bne 1b \n\t"
548 : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
549 :
550 : "r4", "r5", "r12", "memory");
551 }
552
553 void DEF(avg, pixels16_y2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
554 {
555 int stride = line_size;
556 // [wr0 wr1 wr2 wr3] for previous line
557 // [wr4 wr5 wr6 wr7] for current line
558 __asm__ __volatile__(
559 "pld [%[pixels]] \n\t"
560 "pld [%[pixels], #32] \n\t"
561 "and r12, %[pixels], #7 \n\t"
562 "tmcr wcgr1, r12 \n\t"
563 "bic %[pixels], %[pixels], #7 \n\t"
564
565 "wldrd wr10, [%[pixels]] \n\t"
566 "wldrd wr11, [%[pixels], #8] \n\t"
567 "pld [%[block]] \n\t"
568 "wldrd wr12, [%[pixels], #16] \n\t"
569 "add %[pixels], %[pixels], %[line_size] \n\t"
570 "pld [%[pixels]] \n\t"
571 "pld [%[pixels], #32] \n\t"
572 "walignr1 wr0, wr10, wr11 \n\t"
573 "walignr1 wr1, wr11, wr12 \n\t"
574
575 "1: \n\t"
576 "wldrd wr10, [%[pixels]] \n\t"
577 "wldrd wr11, [%[pixels], #8] \n\t"
578 "wldrd wr12, [%[pixels], #16] \n\t"
579 "add %[pixels], %[pixels], %[line_size] \n\t"
580 "pld [%[pixels]] \n\t"
581 "pld [%[pixels], #32] \n\t"
582 "walignr1 wr4, wr10, wr11 \n\t"
583 "walignr1 wr5, wr11, wr12 \n\t"
584 "wldrd wr10, [%[block]] \n\t"
585 "wldrd wr11, [%[block], #8] \n\t"
586 WAVG2B" wr8, wr0, wr4 \n\t"
587 WAVG2B" wr9, wr1, wr5 \n\t"
588 WAVG2B" wr8, wr8, wr10 \n\t"
589 WAVG2B" wr9, wr9, wr11 \n\t"
590 "wstrd wr8, [%[block]] \n\t"
591 "wstrd wr9, [%[block], #8] \n\t"
592 "add %[block], %[block], %[line_size] \n\t"
593
594 "wldrd wr10, [%[pixels]] \n\t"
595 "wldrd wr11, [%[pixels], #8] \n\t"
596 "pld [%[block]] \n\t"
597 "wldrd wr12, [%[pixels], #16] \n\t"
598 "add %[pixels], %[pixels], %[line_size] \n\t"
599 "pld [%[pixels]] \n\t"
600 "pld [%[pixels], #32] \n\t"
601 "walignr1 wr0, wr10, wr11 \n\t"
602 "walignr1 wr1, wr11, wr12 \n\t"
603 "wldrd wr10, [%[block]] \n\t"
604 "wldrd wr11, [%[block], #8] \n\t"
605 WAVG2B" wr8, wr0, wr4 \n\t"
606 WAVG2B" wr9, wr1, wr5 \n\t"
607 WAVG2B" wr8, wr8, wr10 \n\t"
608 WAVG2B" wr9, wr9, wr11 \n\t"
609 "wstrd wr8, [%[block]] \n\t"
610 "wstrd wr9, [%[block], #8] \n\t"
611 "add %[block], %[block], %[line_size] \n\t"
612
613 "subs %[h], %[h], #2 \n\t"
614 "pld [%[block]] \n\t"
615 "bne 1b \n\t"
616 : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
617 :
618 : "r4", "r5", "r12", "memory");
619 }
620
621 void DEF(put, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
622 {
623 // [wr0 wr1 wr2 wr3] for previous line
624 // [wr4 wr5 wr6 wr7] for current line
625 SET_RND(wr15); // =2 for rnd and =1 for no_rnd version
626 __asm__ __volatile__(
627 "pld [%[pixels]] \n\t"
628 "mov r12, #2 \n\t"
629 "pld [%[pixels], #32] \n\t"
630 "tmcr wcgr0, r12 \n\t" /* for shift value */
631 "and r12, %[pixels], #7 \n\t"
632 "bic %[pixels], %[pixels], #7 \n\t"
633 "tmcr wcgr1, r12 \n\t"
634
635 // [wr0 wr1 wr2 wr3] <= *
636 // [wr4 wr5 wr6 wr7]
637 "wldrd wr12, [%[pixels]] \n\t"
638 "add r12, r12, #1 \n\t"
639 "wldrd wr13, [%[pixels], #8] \n\t"
640 "tmcr wcgr2, r12 \n\t"
641 "add %[pixels], %[pixels], %[line_size] \n\t"
642 "cmp r12, #8 \n\t"
643 "pld [%[pixels]] \n\t"
644 "pld [%[pixels], #32] \n\t"
645 "walignr1 wr2, wr12, wr13 \n\t"
646 "wmoveq wr10, wr13 \n\t"
647 "walignr2ne wr10, wr12, wr13 \n\t"
648 "wunpckelub wr0, wr2 \n\t"
649 "wunpckehub wr1, wr2 \n\t"
650 "wunpckelub wr8, wr10 \n\t"
651 "wunpckehub wr9, wr10 \n\t"
652 "waddhus wr0, wr0, wr8 \n\t"
653 "waddhus wr1, wr1, wr9 \n\t"
654
655 "1: \n\t"
656 // [wr0 wr1 wr2 wr3]
657 // [wr4 wr5 wr6 wr7] <= *
658 "wldrd wr12, [%[pixels]] \n\t"
659 "cmp r12, #8 \n\t"
660 "wldrd wr13, [%[pixels], #8] \n\t"
661 "add %[pixels], %[pixels], %[line_size] \n\t"
662 "walignr1 wr6, wr12, wr13 \n\t"
663 "pld [%[pixels]] \n\t"
664 "pld [%[pixels], #32] \n\t"
665 "wmoveq wr10, wr13 \n\t"
666 "walignr2ne wr10, wr12, wr13 \n\t"
667 "wunpckelub wr4, wr6 \n\t"
668 "wunpckehub wr5, wr6 \n\t"
669 "wunpckelub wr8, wr10 \n\t"
670 "wunpckehub wr9, wr10 \n\t"
671 "waddhus wr4, wr4, wr8 \n\t"
672 "waddhus wr5, wr5, wr9 \n\t"
673 "waddhus wr8, wr0, wr4 \n\t"
674 "waddhus wr9, wr1, wr5 \n\t"
675 "waddhus wr8, wr8, wr15 \n\t"
676 "waddhus wr9, wr9, wr15 \n\t"
677 "wsrlhg wr8, wr8, wcgr0 \n\t"
678 "wsrlhg wr9, wr9, wcgr0 \n\t"
679 "wpackhus wr8, wr8, wr9 \n\t"
680 "wstrd wr8, [%[block]] \n\t"
681 "add %[block], %[block], %[line_size] \n\t"
682
683 // [wr0 wr1 wr2 wr3] <= *
684 // [wr4 wr5 wr6 wr7]
685 "wldrd wr12, [%[pixels]] \n\t"
686 "wldrd wr13, [%[pixels], #8] \n\t"
687 "add %[pixels], %[pixels], %[line_size] \n\t"
688 "walignr1 wr2, wr12, wr13 \n\t"
689 "pld [%[pixels]] \n\t"
690 "pld [%[pixels], #32] \n\t"
691 "wmoveq wr10, wr13 \n\t"
692 "walignr2ne wr10, wr12, wr13 \n\t"
693 "wunpckelub wr0, wr2 \n\t"
694 "wunpckehub wr1, wr2 \n\t"
695 "wunpckelub wr8, wr10 \n\t"
696 "wunpckehub wr9, wr10 \n\t"
697 "waddhus wr0, wr0, wr8 \n\t"
698 "waddhus wr1, wr1, wr9 \n\t"
699 "waddhus wr8, wr0, wr4 \n\t"
700 "waddhus wr9, wr1, wr5 \n\t"
701 "waddhus wr8, wr8, wr15 \n\t"
702 "waddhus wr9, wr9, wr15 \n\t"
703 "wsrlhg wr8, wr8, wcgr0 \n\t"
704 "wsrlhg wr9, wr9, wcgr0 \n\t"
705 "wpackhus wr8, wr8, wr9 \n\t"
706 "subs %[h], %[h], #2 \n\t"
707 "wstrd wr8, [%[block]] \n\t"
708 "add %[block], %[block], %[line_size] \n\t"
709 "bne 1b \n\t"
710 : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block)
711 : [line_size]"r"(line_size)
712 : "r12", "memory");
713 }
714
715 void DEF(put, pixels16_xy2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
716 {
717 // [wr0 wr1 wr2 wr3] for previous line
718 // [wr4 wr5 wr6 wr7] for current line
719 SET_RND(wr15); // =2 for rnd and =1 for no_rnd version
720 __asm__ __volatile__(
721 "pld [%[pixels]] \n\t"
722 "mov r12, #2 \n\t"
723 "pld [%[pixels], #32] \n\t"
724 "tmcr wcgr0, r12 \n\t" /* for shift value */
725 /* alignment */
726 "and r12, %[pixels], #7 \n\t"
727 "bic %[pixels], %[pixels], #7 \n\t"
728 "tmcr wcgr1, r12 \n\t"
729 "add r12, r12, #1 \n\t"
730 "tmcr wcgr2, r12 \n\t"
731
732 // [wr0 wr1 wr2 wr3] <= *
733 // [wr4 wr5 wr6 wr7]
734 "wldrd wr12, [%[pixels]] \n\t"
735 "cmp r12, #8 \n\t"
736 "wldrd wr13, [%[pixels], #8] \n\t"
737 "wldrd wr14, [%[pixels], #16] \n\t"
738 "add %[pixels], %[pixels], %[line_size] \n\t"
739 "pld [%[pixels]] \n\t"
740 "walignr1 wr2, wr12, wr13 \n\t"
741 "pld [%[pixels], #32] \n\t"
742 "walignr1 wr3, wr13, wr14 \n\t"
743 "wmoveq wr10, wr13 \n\t"
744 "wmoveq wr11, wr14 \n\t"
745 "walignr2ne wr10, wr12, wr13 \n\t"
746 "walignr2ne wr11, wr13, wr14 \n\t"
747 "wunpckelub wr0, wr2 \n\t"
748 "wunpckehub wr1, wr2 \n\t"
749 "wunpckelub wr2, wr3 \n\t"
750 "wunpckehub wr3, wr3 \n\t"
751 "wunpckelub wr8, wr10 \n\t"
752 "wunpckehub wr9, wr10 \n\t"
753 "wunpckelub wr10, wr11 \n\t"
754 "wunpckehub wr11, wr11 \n\t"
755 "waddhus wr0, wr0, wr8 \n\t"
756 "waddhus wr1, wr1, wr9 \n\t"
757 "waddhus wr2, wr2, wr10 \n\t"
758 "waddhus wr3, wr3, wr11 \n\t"
759
760 "1: \n\t"
761 // [wr0 wr1 wr2 wr3]
762 // [wr4 wr5 wr6 wr7] <= *
763 "wldrd wr12, [%[pixels]] \n\t"
764 "cmp r12, #8 \n\t"
765 "wldrd wr13, [%[pixels], #8] \n\t"
766 "wldrd wr14, [%[pixels], #16] \n\t"
767 "add %[pixels], %[pixels], %[line_size] \n\t"
768 "walignr1 wr6, wr12, wr13 \n\t"
769 "pld [%[pixels]] \n\t"
770 "pld [%[pixels], #32] \n\t"
771 "walignr1 wr7, wr13, wr14 \n\t"
772 "wmoveq wr10, wr13 \n\t"
773 "wmoveq wr11, wr14 \n\t"
774 "walignr2ne wr10, wr12, wr13 \n\t"
775 "walignr2ne wr11, wr13, wr14 \n\t"
776 "wunpckelub wr4, wr6 \n\t"
777 "wunpckehub wr5, wr6 \n\t"
778 "wunpckelub wr6, wr7 \n\t"
779 "wunpckehub wr7, wr7 \n\t"
780 "wunpckelub wr8, wr10 \n\t"
781 "wunpckehub wr9, wr10 \n\t"
782 "wunpckelub wr10, wr11 \n\t"
783 "wunpckehub wr11, wr11 \n\t"
784 "waddhus wr4, wr4, wr8 \n\t"
785 "waddhus wr5, wr5, wr9 \n\t"
786 "waddhus wr6, wr6, wr10 \n\t"
787 "waddhus wr7, wr7, wr11 \n\t"
788 "waddhus wr8, wr0, wr4 \n\t"
789 "waddhus wr9, wr1, wr5 \n\t"
790 "waddhus wr10, wr2, wr6 \n\t"
791 "waddhus wr11, wr3, wr7 \n\t"
792 "waddhus wr8, wr8, wr15 \n\t"
793 "waddhus wr9, wr9, wr15 \n\t"
794 "waddhus wr10, wr10, wr15 \n\t"
795 "waddhus wr11, wr11, wr15 \n\t"
796 "wsrlhg wr8, wr8, wcgr0 \n\t"
797 "wsrlhg wr9, wr9, wcgr0 \n\t"
798 "wsrlhg wr10, wr10, wcgr0 \n\t"
799 "wsrlhg wr11, wr11, wcgr0 \n\t"
800 "wpackhus wr8, wr8, wr9 \n\t"
801 "wpackhus wr9, wr10, wr11 \n\t"
802 "wstrd wr8, [%[block]] \n\t"
803 "wstrd wr9, [%[block], #8] \n\t"
804 "add %[block], %[block], %[line_size] \n\t"
805
806 // [wr0 wr1 wr2 wr3] <= *
807 // [wr4 wr5 wr6 wr7]
808 "wldrd wr12, [%[pixels]] \n\t"
809 "wldrd wr13, [%[pixels], #8] \n\t"
810 "wldrd wr14, [%[pixels], #16] \n\t"
811 "add %[pixels], %[pixels], %[line_size] \n\t"
812 "walignr1 wr2, wr12, wr13 \n\t"
813 "pld [%[pixels]] \n\t"
814 "pld [%[pixels], #32] \n\t"
815 "walignr1 wr3, wr13, wr14 \n\t"
816 "wmoveq wr10, wr13 \n\t"
817 "wmoveq wr11, wr14 \n\t"
818 "walignr2ne wr10, wr12, wr13 \n\t"
819 "walignr2ne wr11, wr13, wr14 \n\t"
820 "wunpckelub wr0, wr2 \n\t"
821 "wunpckehub wr1, wr2 \n\t"
822 "wunpckelub wr2, wr3 \n\t"
823 "wunpckehub wr3, wr3 \n\t"
824 "wunpckelub wr8, wr10 \n\t"
825 "wunpckehub wr9, wr10 \n\t"
826 "wunpckelub wr10, wr11 \n\t"
827 "wunpckehub wr11, wr11 \n\t"
828 "waddhus wr0, wr0, wr8 \n\t"
829 "waddhus wr1, wr1, wr9 \n\t"
830 "waddhus wr2, wr2, wr10 \n\t"
831 "waddhus wr3, wr3, wr11 \n\t"
832 "waddhus wr8, wr0, wr4 \n\t"
833 "waddhus wr9, wr1, wr5 \n\t"
834 "waddhus wr10, wr2, wr6 \n\t"
835 "waddhus wr11, wr3, wr7 \n\t"
836 "waddhus wr8, wr8, wr15 \n\t"
837 "waddhus wr9, wr9, wr15 \n\t"
838 "waddhus wr10, wr10, wr15 \n\t"
839 "waddhus wr11, wr11, wr15 \n\t"
840 "wsrlhg wr8, wr8, wcgr0 \n\t"
841 "wsrlhg wr9, wr9, wcgr0 \n\t"
842 "wsrlhg wr10, wr10, wcgr0 \n\t"
843 "wsrlhg wr11, wr11, wcgr0 \n\t"
844 "wpackhus wr8, wr8, wr9 \n\t"
845 "wpackhus wr9, wr10, wr11 \n\t"
846 "wstrd wr8, [%[block]] \n\t"
847 "wstrd wr9, [%[block], #8] \n\t"
848 "add %[block], %[block], %[line_size] \n\t"
849
850 "subs %[h], %[h], #2 \n\t"
851 "bne 1b \n\t"
852 : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block)
853 : [line_size]"r"(line_size)
854 : "r12", "memory");
855 }
856
857 void DEF(avg, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
858 {
859 // [wr0 wr1 wr2 wr3] for previous line
860 // [wr4 wr5 wr6 wr7] for current line
861 SET_RND(wr15); // =2 for rnd and =1 for no_rnd version
862 __asm__ __volatile__(
863 "pld [%[block]] \n\t"
864 "pld [%[block], #32] \n\t"
865 "pld [%[pixels]] \n\t"
866 "mov r12, #2 \n\t"
867 "pld [%[pixels], #32] \n\t"
868 "tmcr wcgr0, r12 \n\t" /* for shift value */
869 "and r12, %[pixels], #7 \n\t"
870 "bic %[pixels], %[pixels], #7 \n\t"
871 "tmcr wcgr1, r12 \n\t"
872
873 // [wr0 wr1 wr2 wr3] <= *
874 // [wr4 wr5 wr6 wr7]
875 "wldrd wr12, [%[pixels]] \n\t"
876 "add r12, r12, #1 \n\t"
877 "wldrd wr13, [%[pixels], #8] \n\t"
878 "tmcr wcgr2, r12 \n\t"
879 "add %[pixels], %[pixels], %[line_size] \n\t"
880 "cmp r12, #8 \n\t"
881 "pld [%[pixels]] \n\t"
882 "pld [%[pixels], #32] \n\t"
883 "walignr1 wr2, wr12, wr13 \n\t"
884 "wmoveq wr10, wr13 \n\t"
885 "walignr2ne wr10, wr12, wr13 \n\t"
886 "wunpckelub wr0, wr2 \n\t"
887 "wunpckehub wr1, wr2 \n\t"
888 "wunpckelub wr8, wr10 \n\t"
889 "wunpckehub wr9, wr10 \n\t"
890 "waddhus wr0, wr0, wr8 \n\t"
891 "waddhus wr1, wr1, wr9 \n\t"
892
893 "1: \n\t"
894 // [wr0 wr1 wr2 wr3]
895 // [wr4 wr5 wr6 wr7] <= *
896 "wldrd wr12, [%[pixels]] \n\t"
897 "cmp r12, #8 \n\t"
898 "wldrd wr13, [%[pixels], #8] \n\t"
899 "add %[pixels], %[pixels], %[line_size] \n\t"
900 "walignr1 wr6, wr12, wr13 \n\t"
901 "pld [%[pixels]] \n\t"
902 "pld [%[pixels], #32] \n\t"
903 "wmoveq wr10, wr13 \n\t"
904 "walignr2ne wr10, wr12, wr13 \n\t"
905 "wunpckelub wr4, wr6 \n\t"
906 "wunpckehub wr5, wr6 \n\t"
907 "wunpckelub wr8, wr10 \n\t"
908 "wunpckehub wr9, wr10 \n\t"
909 "waddhus wr4, wr4, wr8 \n\t"
910 "waddhus wr5, wr5, wr9 \n\t"
911 "waddhus wr8, wr0, wr4 \n\t"
912 "waddhus wr9, wr1, wr5 \n\t"
913 "waddhus wr8, wr8, wr15 \n\t"
914 "waddhus wr9, wr9, wr15 \n\t"
915 "wldrd wr12, [%[block]] \n\t"
916 "wsrlhg wr8, wr8, wcgr0 \n\t"
917 "wsrlhg wr9, wr9, wcgr0 \n\t"
918 "wpackhus wr8, wr8, wr9 \n\t"
919 WAVG2B" wr8, wr8, wr12 \n\t"
920 "wstrd wr8, [%[block]] \n\t"
921 "add %[block], %[block], %[line_size] \n\t"
922 "wldrd wr12, [%[pixels]] \n\t"
923 "pld [%[block]] \n\t"
924 "pld [%[block], #32] \n\t"
925
926 // [wr0 wr1 wr2 wr3] <= *
927 // [wr4 wr5 wr6 wr7]
928 "wldrd wr13, [%[pixels], #8] \n\t"
929 "add %[pixels], %[pixels], %[line_size] \n\t"
930 "walignr1 wr2, wr12, wr13 \n\t"
931 "pld [%[pixels]] \n\t"
932 "pld [%[pixels], #32] \n\t"
933 "wmoveq wr10, wr13 \n\t"
934 "walignr2ne wr10, wr12, wr13 \n\t"
935 "wunpckelub wr0, wr2 \n\t"
936 "wunpckehub wr1, wr2 \n\t"
937 "wunpckelub wr8, wr10 \n\t"
938 "wunpckehub wr9, wr10 \n\t"
939 "waddhus wr0, wr0, wr8 \n\t"
940 "waddhus wr1, wr1, wr9 \n\t"
941 "waddhus wr8, wr0, wr4 \n\t"
942 "waddhus wr9, wr1, wr5 \n\t"
943 "waddhus wr8, wr8, wr15 \n\t"
944 "waddhus wr9, wr9, wr15 \n\t"
945 "wldrd wr12, [%[block]] \n\t"
946 "wsrlhg wr8, wr8, wcgr0 \n\t"
947 "wsrlhg wr9, wr9, wcgr0 \n\t"
948 "wpackhus wr8, wr8, wr9 \n\t"
949 "subs %[h], %[h], #2 \n\t"
950 WAVG2B" wr8, wr8, wr12 \n\t"
951 "wstrd wr8, [%[block]] \n\t"
952 "add %[block], %[block], %[line_size] \n\t"
953 "pld [%[block]] \n\t"
954 "pld [%[block], #32] \n\t"
955 "bne 1b \n\t"
956 : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block)
957 : [line_size]"r"(line_size)
958 : "r12", "memory");
959 }
960
961 void DEF(avg, pixels16_xy2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
962 {
963 // [wr0 wr1 wr2 wr3] for previous line
964 // [wr4 wr5 wr6 wr7] for current line
965 SET_RND(wr15); // =2 for rnd and =1 for no_rnd version
966 __asm__ __volatile__(
967 "pld [%[block]] \n\t"
968 "pld [%[block], #32] \n\t"
969 "pld [%[pixels]] \n\t"
970 "mov r12, #2 \n\t"
971 "pld [%[pixels], #32] \n\t"
972 "tmcr wcgr0, r12 \n\t" /* for shift value */
973 /* alignment */
974 "and r12, %[pixels], #7 \n\t"
975 "bic %[pixels], %[pixels], #7 \n\t"
976 "tmcr wcgr1, r12 \n\t"
977 "add r12, r12, #1 \n\t"
978 "tmcr wcgr2, r12 \n\t"
979
980 // [wr0 wr1 wr2 wr3] <= *
981 // [wr4 wr5 wr6 wr7]
982 "wldrd wr12, [%[pixels]] \n\t"
983 "cmp r12, #8 \n\t"
984 "wldrd wr13, [%[pixels], #8] \n\t"
985 "wldrd wr14, [%[pixels], #16] \n\t"
986 "add %[pixels], %[pixels], %[line_size] \n\t"
987 "pld [%[pixels]] \n\t"
988 "walignr1 wr2, wr12, wr13 \n\t"
989 "pld [%[pixels], #32] \n\t"
990 "walignr1 wr3, wr13, wr14 \n\t"
991 "wmoveq wr10, wr13 \n\t"
992 "wmoveq wr11, wr14 \n\t"
993 "walignr2ne wr10, wr12, wr13 \n\t"
994 "walignr2ne wr11, wr13, wr14 \n\t"
995 "wunpckelub wr0, wr2 \n\t"
996 "wunpckehub wr1, wr2 \n\t"
997 "wunpckelub wr2, wr3 \n\t"
998 "wunpckehub wr3, wr3 \n\t"
999 "wunpckelub wr8, wr10 \n\t"
1000 "wunpckehub wr9, wr10 \n\t"
1001 "wunpckelub wr10, wr11 \n\t"
1002 "wunpckehub wr11, wr11 \n\t"
1003 "waddhus wr0, wr0, wr8 \n\t"
1004 "waddhus wr1, wr1, wr9 \n\t"
1005 "waddhus wr2, wr2, wr10 \n\t"
1006 "waddhus wr3, wr3, wr11 \n\t"
1007
1008 "1: \n\t"
1009 // [wr0 wr1 wr2 wr3]
1010 // [wr4 wr5 wr6 wr7] <= *
1011 "wldrd wr12, [%[pixels]] \n\t"
1012 "cmp r12, #8 \n\t"
1013 "wldrd wr13, [%[pixels], #8] \n\t"
1014 "wldrd wr14, [%[pixels], #16] \n\t"
1015 "add %[pixels], %[pixels], %[line_size] \n\t"
1016 "walignr1 wr6, wr12, wr13 \n\t"
1017 "pld [%[pixels]] \n\t"
1018 "pld [%[pixels], #32] \n\t"
1019 "walignr1 wr7, wr13, wr14 \n\t"
1020 "wmoveq wr10, wr13 \n\t"
1021 "wmoveq wr11, wr14 \n\t"
1022 "walignr2ne wr10, wr12, wr13 \n\t"
1023 "walignr2ne wr11, wr13, wr14 \n\t"
1024 "wunpckelub wr4, wr6 \n\t"
1025 "wunpckehub wr5, wr6 \n\t"
1026 "wunpckelub wr6, wr7 \n\t"
1027 "wunpckehub wr7, wr7 \n\t"
1028 "wunpckelub wr8, wr10 \n\t"
1029 "wunpckehub wr9, wr10 \n\t"
1030 "wunpckelub wr10, wr11 \n\t"
1031 "wunpckehub wr11, wr11 \n\t"
1032 "waddhus wr4, wr4, wr8 \n\t"
1033 "waddhus wr5, wr5, wr9 \n\t"
1034 "waddhus wr6, wr6, wr10 \n\t"
1035 "waddhus wr7, wr7, wr11 \n\t"
1036 "waddhus wr8, wr0, wr4 \n\t"
1037 "waddhus wr9, wr1, wr5 \n\t"
1038 "waddhus wr10, wr2, wr6 \n\t"
1039 "waddhus wr11, wr3, wr7 \n\t"
1040 "waddhus wr8, wr8, wr15 \n\t"
1041 "waddhus wr9, wr9, wr15 \n\t"
1042 "waddhus wr10, wr10, wr15 \n\t"
1043 "waddhus wr11, wr11, wr15 \n\t"
1044 "wsrlhg wr8, wr8, wcgr0 \n\t"
1045 "wsrlhg wr9, wr9, wcgr0 \n\t"
1046 "wldrd wr12, [%[block]] \n\t"
1047 "wldrd wr13, [%[block], #8] \n\t"
1048 "wsrlhg wr10, wr10, wcgr0 \n\t"
1049 "wsrlhg wr11, wr11, wcgr0 \n\t"
1050 "wpackhus wr8, wr8, wr9 \n\t"
1051 "wpackhus wr9, wr10, wr11 \n\t"
1052 WAVG2B" wr8, wr8, wr12 \n\t"
1053 WAVG2B" wr9, wr9, wr13 \n\t"
1054 "wstrd wr8, [%[block]] \n\t"
1055 "wstrd wr9, [%[block], #8] \n\t"
1056 "add %[block], %[block], %[line_size] \n\t"
1057
1058 // [wr0 wr1 wr2 wr3] <= *
1059 // [wr4 wr5 wr6 wr7]
1060 "wldrd wr12, [%[pixels]] \n\t"
1061 "pld [%[block]] \n\t"
1062 "wldrd wr13, [%[pixels], #8] \n\t"
1063 "pld [%[block], #32] \n\t"
1064 "wldrd wr14, [%[pixels], #16] \n\t"
1065 "add %[pixels], %[pixels], %[line_size] \n\t"
1066 "walignr1 wr2, wr12, wr13 \n\t"
1067 "pld [%[pixels]] \n\t"
1068 "pld [%[pixels], #32] \n\t"
1069 "walignr1 wr3, wr13, wr14 \n\t"
1070 "wmoveq wr10, wr13 \n\t"
1071 "wmoveq wr11, wr14 \n\t"
1072 "walignr2ne wr10, wr12, wr13 \n\t"
1073 "walignr2ne wr11, wr13, wr14 \n\t"
1074 "wunpckelub wr0, wr2 \n\t"
1075 "wunpckehub wr1, wr2 \n\t"
1076 "wunpckelub wr2, wr3 \n\t"
1077 "wunpckehub wr3, wr3 \n\t"
1078 "wunpckelub wr8, wr10 \n\t"
1079 "wunpckehub wr9, wr10 \n\t"
1080 "wunpckelub wr10, wr11 \n\t"
1081 "wunpckehub wr11, wr11 \n\t"
1082 "waddhus wr0, wr0, wr8 \n\t"
1083 "waddhus wr1, wr1, wr9 \n\t"
1084 "waddhus wr2, wr2, wr10 \n\t"
1085 "waddhus wr3, wr3, wr11 \n\t"
1086 "waddhus wr8, wr0, wr4 \n\t"
1087 "waddhus wr9, wr1, wr5 \n\t"
1088 "waddhus wr10, wr2, wr6 \n\t"
1089 "waddhus wr11, wr3, wr7 \n\t"
1090 "waddhus wr8, wr8, wr15 \n\t"
1091 "waddhus wr9, wr9, wr15 \n\t"
1092 "waddhus wr10, wr10, wr15 \n\t"
1093 "waddhus wr11, wr11, wr15 \n\t"
1094 "wsrlhg wr8, wr8, wcgr0 \n\t"
1095 "wsrlhg wr9, wr9, wcgr0 \n\t"
1096 "wldrd wr12, [%[block]] \n\t"
1097 "wldrd wr13, [%[block], #8] \n\t"
1098 "wsrlhg wr10, wr10, wcgr0 \n\t"
1099 "wsrlhg wr11, wr11, wcgr0 \n\t"
1100 "wpackhus wr8, wr8, wr9 \n\t"
1101 "wpackhus wr9, wr10, wr11 \n\t"
1102 WAVG2B" wr8, wr8, wr12 \n\t"
1103 WAVG2B" wr9, wr9, wr13 \n\t"
1104 "wstrd wr8, [%[block]] \n\t"
1105 "wstrd wr9, [%[block], #8] \n\t"
1106 "add %[block], %[block], %[line_size] \n\t"
1107 "subs %[h], %[h], #2 \n\t"
1108 "pld [%[block]] \n\t"
1109 "pld [%[block], #32] \n\t"
1110 "bne 1b \n\t"
1111 : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block)
1112 : [line_size]"r"(line_size)
1113 : "r12", "memory");
1114 }