CoCalc -- predict-a.S

05. Matplotlib / ffmpeg-3.0 / libx264 / common / arm / predict-a.S
⁵²⁸⁶⁷ views
1
/*****************************************************************************
2
 * predict.S: arm intra prediction
3
 *****************************************************************************
4
 * Copyright (C) 2009-2016 x264 project
5
 *
6
 * Authors: David Conrad <[email protected]>
7
 *          Mans Rullgard <[email protected]>
8
 *          Martin Storsjo <[email protected]>
9
 *
10
 * This program is free software; you can redistribute it and/or modify
11
 * it under the terms of the GNU General Public License as published by
12
 * the Free Software Foundation; either version 2 of the License, or
13
 * (at your option) any later version.
14
 *
15
 * This program is distributed in the hope that it will be useful,
16
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
18
 * GNU General Public License for more details.
19
 *
20
 * You should have received a copy of the GNU General Public License
21
 * along with this program; if not, write to the Free Software
22
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.
23
 *
24
 * This program is also available under a commercial proprietary license.
25
 * For more information, contact us at [email protected].
26
 *****************************************************************************/
27

28
#include "asm.S"
29

30
.section .rodata
31
.align 4
32

33
p16weight: .short 1,2,3,4,5,6,7,8
34

35
.text
36

37
.macro ldcol.8  rd,  rs,  rt,  n=8,  hi=0
38
.if \n == 8 || \hi == 0
39
    vld1.8          {\rd[0]}, [\rs], \rt
40
    vld1.8          {\rd[1]}, [\rs], \rt
41
    vld1.8          {\rd[2]}, [\rs], \rt
42
    vld1.8          {\rd[3]}, [\rs], \rt
43
.endif
44
.if \n == 8 || \hi == 1
45
    vld1.8          {\rd[4]}, [\rs], \rt
46
    vld1.8          {\rd[5]}, [\rs], \rt
47
    vld1.8          {\rd[6]}, [\rs], \rt
48
    vld1.8          {\rd[7]}, [\rs], \rt
49
.endif
50
.endm
51

52
.macro ldcol.16  rd1,  rd2,  rs,  rt,  ru
53
    add             \ru, \rs, \rt, lsl #3
54
    vld1.8          {\rd1[0]}, [\rs], \rt
55
    vld1.8          {\rd2[0]}, [\ru], \rt
56
    vld1.8          {\rd1[1]}, [\rs], \rt
57
    vld1.8          {\rd2[1]}, [\ru], \rt
58
    vld1.8          {\rd1[2]}, [\rs], \rt
59
    vld1.8          {\rd2[2]}, [\ru], \rt
60
    vld1.8          {\rd1[3]}, [\rs], \rt
61
    vld1.8          {\rd2[3]}, [\ru], \rt
62
    vld1.8          {\rd1[4]}, [\rs], \rt
63
    vld1.8          {\rd2[4]}, [\ru], \rt
64
    vld1.8          {\rd1[5]}, [\rs], \rt
65
    vld1.8          {\rd2[5]}, [\ru], \rt
66
    vld1.8          {\rd1[6]}, [\rs], \rt
67
    vld1.8          {\rd2[6]}, [\ru], \rt
68
    vld1.8          {\rd1[7]}, [\rs], \rt
69
    vld1.8          {\rd2[7]}, [\ru], \rt
70
.endm
71

72
.macro add16x8  dq,  dl,  dh,  rl,  rh
73
    vaddl.u8        \dq, \rl, \rh
74
    vadd.u16        \dl, \dl, \dh
75
    vpadd.u16       \dl, \dl, \dl
76
    vpadd.u16       \dl, \dl, \dl
77
.endm
78

79

80
// because gcc doesn't believe in using the free shift in add
81
function x264_predict_4x4_h_armv6
82
    ldrb    r1, [r0, #0*FDEC_STRIDE-1]
83
    ldrb    r2, [r0, #1*FDEC_STRIDE-1]
84
    ldrb    r3, [r0, #2*FDEC_STRIDE-1]
85
    ldrb    ip, [r0, #3*FDEC_STRIDE-1]
86
    add     r1, r1, r1, lsl #8
87
    add     r2, r2, r2, lsl #8
88
    add     r3, r3, r3, lsl #8
89
    add     ip, ip, ip, lsl #8
90
    add     r1, r1, r1, lsl #16
91
    str     r1, [r0, #0*FDEC_STRIDE]
92
    add     r2, r2, r2, lsl #16
93
    str     r2, [r0, #1*FDEC_STRIDE]
94
    add     r3, r3, r3, lsl #16
95
    str     r3, [r0, #2*FDEC_STRIDE]
96
    add     ip, ip, ip, lsl #16
97
    str     ip, [r0, #3*FDEC_STRIDE]
98
    bx      lr
99
endfunc
100

101
function x264_predict_4x4_v_armv6
102
    ldr     r1,  [r0, #0 - 1 * FDEC_STRIDE]
103
    str     r1,  [r0, #0 + 0 * FDEC_STRIDE]
104
    str     r1,  [r0, #0 + 1 * FDEC_STRIDE]
105
    str     r1,  [r0, #0 + 2 * FDEC_STRIDE]
106
    str     r1,  [r0, #0 + 3 * FDEC_STRIDE]
107
    bx      lr
108
endfunc
109

110
function x264_predict_4x4_dc_armv6
111
    mov     ip, #0
112
    ldr     r1, [r0, #-FDEC_STRIDE]
113
    ldrb    r2, [r0, #0*FDEC_STRIDE-1]
114
    ldrb    r3, [r0, #1*FDEC_STRIDE-1]
115
    usad8   r1, r1, ip
116
    add     r2, r2, #4
117
    ldrb    ip, [r0, #2*FDEC_STRIDE-1]
118
    add     r2, r2, r3
119
    ldrb    r3, [r0, #3*FDEC_STRIDE-1]
120
    add     r2, r2, ip
121
    add     r2, r2, r3
122
    add     r1, r1, r2
123
    lsr     r1, r1, #3
124
    add     r1, r1, r1, lsl #8
125
    add     r1, r1, r1, lsl #16
126
    str     r1, [r0, #0*FDEC_STRIDE]
127
    str     r1, [r0, #1*FDEC_STRIDE]
128
    str     r1, [r0, #2*FDEC_STRIDE]
129
    str     r1, [r0, #3*FDEC_STRIDE]
130
    bx      lr
131
endfunc
132

133
function x264_predict_4x4_dc_top_neon
134
    mov         r12, #FDEC_STRIDE
135
    sub         r1, r0, #FDEC_STRIDE
136
    vld1.32     d1[], [r1,:32]
137
    vpaddl.u8   d1, d1
138
    vpadd.u16   d1, d1, d1
139
    vrshr.u16   d1, d1, #2
140
    vdup.8      d1, d1[0]
141
    vst1.32     d1[0], [r0,:32], r12
142
    vst1.32     d1[0], [r0,:32], r12
143
    vst1.32     d1[0], [r0,:32], r12
144
    vst1.32     d1[0], [r0,:32], r12
145
    bx          lr
146
endfunc
147

148
// return a1 = (a1+2*b1+c1+2)>>2  a2 = (a2+2*b2+c2+2)>>2
149
.macro PRED4x4_LOWPASS a1 b1 c1 a2 b2 c2 pb_1
150
    uhadd8  \a1, \a1, \c1
151
    uhadd8  \a2, \a2, \c2
152
    uhadd8  \c1, \a1, \b1
153
    uhadd8  \c2, \a2, \b2
154
    eor     \a1, \a1, \b1
155
    eor     \a2, \a2, \b2
156
    and     \a1, \a1, \pb_1
157
    and     \a2, \a2, \pb_1
158
    uadd8   \a1, \a1, \c1
159
    uadd8   \a2, \a2, \c2
160
.endm
161

162
function x264_predict_4x4_ddr_armv6
163
    ldr     r1, [r0, # -FDEC_STRIDE]
164
    ldrb    r2, [r0, # -FDEC_STRIDE-1]
165
    ldrb    r3, [r0, #0*FDEC_STRIDE-1]
166
    push    {r4-r6,lr}
167
    add     r2, r2, r1, lsl #8
168
    ldrb    r4, [r0, #1*FDEC_STRIDE-1]
169
    add     r3, r3, r2, lsl #8
170
    ldrb    r5, [r0, #2*FDEC_STRIDE-1]
171
    ldrb    r6, [r0, #3*FDEC_STRIDE-1]
172
    add     r4, r4, r3, lsl #8
173
    add     r5, r5, r4, lsl #8
174
    add     r6, r6, r5, lsl #8
175
    ldr     ip, =0x01010101
176
    PRED4x4_LOWPASS r1, r2, r3, r4, r5, r6, ip
177
    str     r1, [r0, #0*FDEC_STRIDE]
178
    lsl     r2, r1, #8
179
    lsl     r3, r1, #16
180
    lsl     r4, r4, #8
181
    lsl     r5, r1, #24
182
    add     r2, r2, r4, lsr #24
183
    str     r2, [r0, #1*FDEC_STRIDE]
184
    add     r3, r3, r4, lsr #16
185
    str     r3, [r0, #2*FDEC_STRIDE]
186
    add     r5, r5, r4, lsr #8
187
    str     r5, [r0, #3*FDEC_STRIDE]
188
    pop     {r4-r6,pc}
189
endfunc
190

191
function x264_predict_4x4_ddl_neon
192
    sub         r0, #FDEC_STRIDE
193
    mov         ip, #FDEC_STRIDE
194
    vld1.64     {d0}, [r0], ip
195
    vdup.8      d3, d0[7]
196
    vext.8      d1, d0, d0, #1
197
    vext.8      d2, d0, d3, #2
198
    vhadd.u8    d0, d0, d2
199
    vrhadd.u8   d0, d0, d1
200
    vst1.32     {d0[0]}, [r0,:32], ip
201
    vext.8      d1, d0, d0, #1
202
    vext.8      d2, d0, d0, #2
203
    vst1.32     {d1[0]}, [r0,:32], ip
204
    vext.8      d3, d0, d0, #3
205
    vst1.32     {d2[0]}, [r0,:32], ip
206
    vst1.32     {d3[0]}, [r0,:32], ip
207
    bx          lr
208
endfunc
209

210
function x264_predict_8x8_dc_neon
211
    mov     ip, #0
212
    ldrd    r2, r3, [r1, #8]
213
    push    {r4-r5,lr}
214
    ldrd    r4, r5, [r1, #16]
215
    lsl     r3, r3, #8
216
    ldrb    lr, [r1, #7]
217
    usad8   r2, r2, ip
218
    usad8   r3, r3, ip
219
    usada8  r2, r4, ip, r2
220
    add     lr, lr, #8
221
    usada8  r3, r5, ip, r3
222
    add     r2, r2, lr
223
    mov     ip, #FDEC_STRIDE
224
    add     r2, r2, r3
225
    lsr     r2, r2, #4
226

227
    vdup.8   d0, r2
228
.rept 8
229
    vst1.64 {d0}, [r0,:64], ip
230
.endr
231
    pop    {r4-r5,pc}
232
endfunc
233

234
function x264_predict_8x8_h_neon
235
    add         r1, r1, #7
236
    mov         ip, #FDEC_STRIDE
237
    vld1.64     {d16}, [r1]
238
    vdup.8      d0, d16[7]
239
    vdup.8      d1, d16[6]
240
    vst1.64     {d0}, [r0,:64], ip
241
    vdup.8      d2, d16[5]
242
    vst1.64     {d1}, [r0,:64], ip
243
    vdup.8      d3, d16[4]
244
    vst1.64     {d2}, [r0,:64], ip
245
    vdup.8      d4, d16[3]
246
    vst1.64     {d3}, [r0,:64], ip
247
    vdup.8      d5, d16[2]
248
    vst1.64     {d4}, [r0,:64], ip
249
    vdup.8      d6, d16[1]
250
    vst1.64     {d5}, [r0,:64], ip
251
    vdup.8      d7, d16[0]
252
    vst1.64     {d6}, [r0,:64], ip
253
    vst1.64     {d7}, [r0,:64], ip
254
    bx          lr
255
endfunc
256

257
function x264_predict_8x8_v_neon
258
    add         r1, r1, #16
259
    mov         r12, #FDEC_STRIDE
260
    vld1.8      {d0}, [r1,:64]
261
.rept 8
262
    vst1.8      {d0}, [r0,:64], r12
263
.endr
264
    bx          lr
265
endfunc
266

267
function x264_predict_8x8_ddl_neon
268
    add         r1, #16
269
    vld1.8      {d0, d1}, [r1,:128]
270
    vmov.i8     q3, #0
271
    vrev64.8    d2, d1
272
    vext.8      q8, q3, q0, #15
273
    vext.8      q2, q0, q1, #1
274
    vhadd.u8    q8, q2
275
    mov         r12, #FDEC_STRIDE
276
    vrhadd.u8   q0, q8
277
    vext.8      d2, d0, d1, #1
278
    vext.8      d3, d0, d1, #2
279
    vst1.8      d2, [r0,:64], r12
280
    vext.8      d2, d0, d1, #3
281
    vst1.8      d3, [r0,:64], r12
282
    vext.8      d3, d0, d1, #4
283
    vst1.8      d2, [r0,:64], r12
284
    vext.8      d2, d0, d1, #5
285
    vst1.8      d3, [r0,:64], r12
286
    vext.8      d3, d0, d1, #6
287
    vst1.8      d2, [r0,:64], r12
288
    vext.8      d2, d0, d1, #7
289
    vst1.8      d3, [r0,:64], r12
290
    vst1.8      d2, [r0,:64], r12
291
    vst1.8      d1, [r0,:64], r12
292
    bx          lr
293
endfunc
294

295
function x264_predict_8x8_ddr_neon
296
    vld1.8      {d0-d3}, [r1,:128]
297
    vext.8      q2, q0, q1, #7
298
    vext.8      q3, q0, q1, #9
299

300
    vhadd.u8    q2, q2, q3
301
    vrhadd.u8   d0, d1, d4
302
    vrhadd.u8   d1, d2, d5
303

304
    add         r0, #7*FDEC_STRIDE
305
    mov         r12, #-1*FDEC_STRIDE
306

307
    vext.8      d2, d0, d1, #1
308
    vst1.8      {d0}, [r0,:64], r12
309
    vext.8      d4, d0, d1, #2
310
    vst1.8      {d2}, [r0,:64], r12
311
    vext.8      d5, d0, d1, #3
312
    vst1.8      {d4}, [r0,:64], r12
313
    vext.8      d4, d0, d1, #4
314
    vst1.8      {d5}, [r0,:64], r12
315
    vext.8      d5, d0, d1, #5
316
    vst1.8      {d4}, [r0,:64], r12
317
    vext.8      d4, d0, d1, #6
318
    vst1.8      {d5}, [r0,:64], r12
319
    vext.8      d5, d0, d1, #7
320
    vst1.8      {d4}, [r0,:64], r12
321
    vst1.8      {d5}, [r0,:64], r12
322
    bx          lr
323
endfunc
324

325
function x264_predict_8x8_vl_neon
326
    add         r1, #16
327
    mov         r12, #FDEC_STRIDE
328

329
    vld1.8      {d0, d1}, [r1,:128]
330
    vext.8      q1, q1, q0, #15
331
    vext.8      q2, q0, q2, #1
332

333
    vrhadd.u8   q3, q0, q2
334

335
    vhadd.u8    q1, q1, q2
336
    vrhadd.u8   q0, q0, q1
337

338
    vext.8      d2, d0, d1, #1
339
    vst1.8      {d6}, [r0,:64], r12
340
    vext.8      d3, d6, d7, #1
341
    vst1.8      {d2}, [r0,:64], r12
342
    vext.8      d2, d0, d1, #2
343
    vst1.8      {d3}, [r0,:64], r12
344
    vext.8      d3, d6, d7, #2
345
    vst1.8      {d2}, [r0,:64], r12
346
    vext.8      d2, d0, d1, #3
347
    vst1.8      {d3}, [r0,:64], r12
348
    vext.8      d3, d6, d7, #3
349
    vst1.8      {d2}, [r0,:64], r12
350
    vext.8      d2, d0, d1, #4
351
    vst1.8      {d3}, [r0,:64], r12
352
    vst1.8      {d2}, [r0,:64], r12
353
    bx          lr
354
endfunc
355

356
function x264_predict_8x8_vr_neon
357
    add         r1, #8
358
    mov         r12, #FDEC_STRIDE
359
    vld1.8      {d4,d5}, [r1,:64]
360

361
    vext.8      q1, q2, q2, #14
362
    vext.8      q0, q2, q2, #15
363

364
    vhadd.u8    q3, q2, q1
365
    vrhadd.u8   q2, q2, q0
366
    vrhadd.u8   q0, q0, q3
367

368
    vmov        d2, d0
369

370
    vst1.8      {d5}, [r0,:64], r12
371
    vuzp.8      d2, d0
372
    vst1.8      {d1}, [r0,:64], r12
373
    vext.8      d6, d0, d5, #7
374
    vext.8      d3, d2, d1, #7
375
    vst1.8      {d6}, [r0,:64], r12
376
    vst1.8      {d3}, [r0,:64], r12
377
    vext.8      d6, d0, d5, #6
378
    vext.8      d3, d2, d1, #6
379
    vst1.8      {d6}, [r0,:64], r12
380
    vst1.8      {d3}, [r0,:64], r12
381
    vext.8      d6, d0, d5, #5
382
    vext.8      d3, d2, d1, #5
383
    vst1.8      {d6}, [r0,:64], r12
384
    vst1.8      {d3}, [r0,:64], r12
385
    bx          lr
386
endfunc
387

388
function x264_predict_8x8_hd_neon
389
    mov         r12, #FDEC_STRIDE
390
    add         r1, #7
391

392
    vld1.8      {d2,d3}, [r1]
393
    vext.8      q3, q1, q1, #1
394
    vext.8      q2, q1, q1, #2
395

396
    vrhadd.u8   q8, q1, q3
397

398
    vhadd.u8    q1, q2
399
    vrhadd.u8   q0, q1, q3
400

401
    vzip.8      d16, d0
402

403
    vext.8      d2, d0, d1, #6
404
    vext.8      d3, d0, d1, #4
405
    vst1.8      {d2}, [r0,:64], r12
406
    vext.8      d2, d0, d1, #2
407
    vst1.8      {d3}, [r0,:64], r12
408
    vst1.8      {d2}, [r0,:64], r12
409
    vext.8      d2, d16, d0, #6
410
    vst1.8      {d0}, [r0,:64], r12
411
    vext.8      d3, d16, d0, #4
412
    vst1.8      {d2}, [r0,:64], r12
413
    vext.8      d2, d16, d0, #2
414
    vst1.8      {d3}, [r0,:64], r12
415
    vst1.8      {d2}, [r0,:64], r12
416
    vst1.8      {d16}, [r0,:64], r12
417

418
    bx          lr
419
endfunc
420

421
function x264_predict_8x8_hu_neon
422
    mov         r12, #FDEC_STRIDE
423
    add         r1, #7
424
    vld1.8      {d7}, [r1]
425
    vdup.8      d6, d7[0]
426
    vrev64.8    d7, d7
427

428
    vext.8      d4, d7, d6, #2
429
    vext.8      d2, d7, d6, #1
430

431
    vhadd.u8    d16, d7, d4
432
    vrhadd.u8   d0, d2, d7
433
    vrhadd.u8   d1, d16, d2
434

435
    vzip.8      d0, d1
436

437
    vdup.16     q1, d1[3]
438

439
    vext.8      q2, q0, q1, #2
440
    vext.8      q3, q0, q1, #4
441
    vext.8      q8, q0, q1, #6
442
    vst1.8      {d0}, [r0,:64], r12
443
    vst1.8      {d4}, [r0,:64], r12
444
    vst1.8      {d6}, [r0,:64], r12
445
    vst1.8      {d16}, [r0,:64], r12
446

447
    vst1.8      {d1}, [r0,:64], r12
448
    vst1.8      {d5}, [r0,:64], r12
449
    vst1.8      {d7}, [r0,:64], r12
450
    vst1.8      {d17}, [r0,:64]
451
    bx          lr
452
endfunc
453

454
function x264_predict_8x8c_dc_top_neon
455
    sub         r2,  r0,  #FDEC_STRIDE
456
    mov         r1,  #FDEC_STRIDE
457
    vld1.8      {d0}, [r2,:64]
458
    vpaddl.u8   d0,  d0
459
    vpadd.u16   d0,  d0,  d0
460
    vrshrn.u16  d0,  q0,  #2
461
    vdup.8      d1,  d0[1]
462
    vdup.8      d0,  d0[0]
463
    vtrn.32     d0,  d1
464
    b           pred8x8_dc_end
465
endfunc
466

467
function x264_predict_8x8c_dc_left_neon
468
    mov         r1,  #FDEC_STRIDE
469
    sub         r2,  r0,  #1
470
    ldcol.8     d0,  r2,  r1
471
    vpaddl.u8   d0,  d0
472
    vpadd.u16   d0,  d0,  d0
473
    vrshrn.u16  d0,  q0,  #2
474
    vdup.8      d1,  d0[1]
475
    vdup.8      d0,  d0[0]
476
    b           pred8x8_dc_end
477
endfunc
478

479
function x264_predict_8x8c_dc_neon
480
    sub         r2,  r0,  #FDEC_STRIDE
481
    mov         r1,  #FDEC_STRIDE
482
    vld1.8      {d0}, [r2,:64]
483
    sub         r2,  r0,  #1
484
    ldcol.8     d1,  r2,  r1
485
    vtrn.32     d0,  d1
486
    vpaddl.u8   q0,  q0
487
    vpadd.u16   d0,  d0,  d1
488
    vpadd.u16   d1,  d0,  d0
489
    vrshrn.u16  d2,  q0,  #3
490
    vrshrn.u16  d3,  q0,  #2
491
    vdup.8      d0,  d2[4]
492
    vdup.8      d1,  d3[3]
493
    vdup.8      d4,  d3[2]
494
    vdup.8      d5,  d2[5]
495
    vtrn.32     q0,  q2
496
pred8x8_dc_end:
497
    add         r2,  r0,  r1,  lsl #2
498
.rept 4
499
    vst1.8      {d0}, [r0,:64], r1
500
    vst1.8      {d1}, [r2,:64], r1
501
.endr
502
    bx          lr
503
endfunc
504

505
function x264_predict_8x8c_h_neon
506
    sub         r1, r0, #1
507
    mov         ip, #FDEC_STRIDE
508
.rept 4
509
    vld1.8      {d0[]}, [r1], ip
510
    vld1.8      {d2[]}, [r1], ip
511
    vst1.64     {d0}, [r0,:64], ip
512
    vst1.64     {d2}, [r0,:64], ip
513
.endr
514
    bx          lr
515
endfunc
516

517
function x264_predict_8x8c_v_neon
518
    sub         r0, r0, #FDEC_STRIDE
519
    mov         ip, #FDEC_STRIDE
520
    vld1.64     {d0}, [r0,:64], ip
521
.rept 8
522
    vst1.64     {d0}, [r0,:64], ip
523
.endr
524
    bx          lr
525
endfunc
526

527
function x264_predict_8x8c_p_neon
528
    sub         r3,  r0,  #FDEC_STRIDE
529
    mov         r1,  #FDEC_STRIDE
530
    add         r2,  r3,  #4
531
    sub         r3,  r3,  #1
532
    vld1.32     {d0[0]}, [r3]
533
    vld1.32     {d2[0]}, [r2,:32], r1
534
    ldcol.8     d0,  r3,  r1,  4,  hi=1
535
    add         r3,  r3,  r1
536
    ldcol.8     d3,  r3,  r1,  4
537
    vaddl.u8    q8,  d2,  d3
538
    vrev32.8    d0,  d0
539
    vtrn.32     d2,  d3
540
    vsubl.u8    q2,  d2,  d0
541
    movrel      r3,  p16weight
542
    vld1.16     {q0}, [r3,:128]
543
    vmul.s16    d4,  d4,  d0
544
    vmul.s16    d5,  d5,  d0
545
    vpadd.i16   d4,  d4,  d5
546
    vpaddl.s16  d4,  d4
547
    vshl.i32    d5,  d4,  #4
548
    vadd.s32    d4,  d4,  d5
549
    vrshrn.s32  d4,  q2,  #5
550
    mov         r3,  #0
551
    vtrn.16     d4,  d5
552
    vadd.i16    d2,  d4,  d5
553
    vshl.i16    d3,  d2,  #2
554
    vrev64.16   d16, d16
555
    vsub.i16    d3,  d3,  d2
556
    vadd.i16    d16, d16, d0
557
    vshl.i16    d2,  d16, #4
558
    vsub.i16    d2,  d2,  d3
559
    vext.16     q0,  q0,  q0,  #7
560
    vmov.16     d0[0], r3
561
    vmul.i16    q0,  q0,  d4[0]
562
    vdup.16     q1,  d2[0]
563
    vdup.16     q3,  d5[0]
564
    vadd.i16    q1,  q1,  q0
565
    mov         r3,  #8
566
1:
567
    vqshrun.s16 d0,  q1,  #5
568
    vadd.i16    q1,  q1,  q3
569
    vst1.8      {d0}, [r0,:64], r1
570
    subs        r3,  r3,  #1
571
    bne         1b
572
    bx          lr
573
endfunc
574

575

576
function x264_predict_8x16c_dc_top_neon
577
    sub         r2,  r0,  #FDEC_STRIDE
578
    mov         r1,  #FDEC_STRIDE
579
    vld1.8      {d0}, [r2,:64]
580
    vpaddl.u8   d0,  d0
581
    vpadd.u16   d0,  d0,  d0
582
    vrshrn.u16  d0,  q0,  #2
583
    vdup.8      d1,  d0[1]
584
    vdup.8      d0,  d0[0]
585
    vtrn.32     d0,  d1
586

587
    add         r2,  r0,  r1,  lsl #2
588
.rept 4
589
    vst1.8      {d0}, [r0,:64], r1
590
    vst1.8      {d1}, [r2,:64], r1
591
.endr
592
    add         r2,  r2,  r1,  lsl #2
593
    add         r0,  r0,  r1,  lsl #2
594
.rept 4
595
    vst1.8      {d0}, [r0,:64], r1
596
    vst1.8      {d1}, [r2,:64], r1
597
.endr
598
    bx          lr
599
endfunc
600

601
function x264_predict_8x16c_h_neon
602
    sub         r1, r0, #1
603
    mov         ip, #FDEC_STRIDE
604
.rept 8
605
    vld1.8      {d0[]}, [r1], ip
606
    vld1.8      {d2[]}, [r1], ip
607
    vst1.64     {d0}, [r0,:64], ip
608
    vst1.64     {d2}, [r0,:64], ip
609
.endr
610
    bx          lr
611
endfunc
612

613
function x264_predict_8x16c_p_neon
614
    sub         r3,  r0,  #FDEC_STRIDE
615
    mov         r1,  #FDEC_STRIDE
616
    add         r2,  r3,  #4
617
    sub         r3,  r3,  #1
618
    vld1.32     {d0[0]}, [r3]
619
    vld1.32     {d2[0]}, [r2,:32], r1
620
    ldcol.8     d1,  r3,  r1
621
    add         r3,  r3,  r1
622
    ldcol.8     d3,  r3,  r1
623
    vrev64.32   d16, d3
624
    vaddl.u8    q8,  d2,  d16
625
    vrev32.8    d0,  d0
626
    vsubl.u8    q2,  d2,  d0
627
    vrev64.8    d1,  d1
628
    vsubl.u8    q3,  d3,  d1
629
    movrel      r3,  p16weight
630
    vld1.16     {q0}, [r3,:128]
631
    vmul.s16    d4,  d4,  d0
632
    vmul.s16    q3,  q3,  q0
633
    vpadd.i16   d4,  d4,  d5
634
    vpadd.i16   d6,  d6,  d7
635
    vpaddl.s16  d4,  d4        @ d4[0] = H
636
    vpaddl.s16  d6,  d6
637
    vpadd.s32   d6,  d6        @ d6[0] = V
638
    vshl.i32    d5,  d4,  #4
639
    vadd.s32    d4,  d4,  d5   @ d4[0] = 17*H
640
    vshl.i32    d7,  d6,  #2
641
    vrshrn.s32  d4,  q2,  #5   @ d4[0] = b
642
    vadd.s32    d6,  d6,  d7   @ d6[0] = 5*V
643
    vrshrn.s32  d6,  q3,  #6   @ d6[0] = c
644
    mov         r3,  #0
645
    vshl.i16    d3,  d4,  #2
646
    vsub.i16    d3,  d3,  d4   @ d2[0] = 3 * b
647
    vshl.i16    d2,  d6,  #3
648
    vadd.i16    d3,  d3,  d2   @ d2[0] = 3 * b + 8 * c
649
    vsub.i16    d3,  d3,  d6   @ d2[0] = 3 * b + 7 * c
650
    vrev64.16   d16, d16
651
    vadd.i16    d16, d16, d0   @ d16[0] = src[]+src[] + 1
652
    vshl.i16    d2,  d16, #4   @ d3[0] = a + 16
653
    vsub.i16    d2,  d2,  d3   @ i00
654
    vext.16     q0,  q0,  q0,  #7
655
    vmov.16     d0[0], r3
656
    vmul.i16    q0,  q0,  d4[0]
657
    vdup.16     q1,  d2[0]
658
    vdup.16     q3,  d6[0]
659
    vadd.i16    q1,  q1,  q0
660
    mov         r3,  #16
661
1:
662
    vqshrun.s16 d0,  q1,  #5
663
    vadd.i16    q1,  q1,  q3
664
    vst1.8      {d0}, [r0,:64], r1
665
    subs        r3,  r3,  #1
666
    bne         1b
667
    bx          lr
668
endfunc
669

670

671
function x264_predict_16x16_dc_top_neon
672
    sub         r2,  r0,  #FDEC_STRIDE
673
    mov         r1,  #FDEC_STRIDE
674
    vld1.8      {q0}, [r2,:128]
675
    add16x8     q0,  d0,  d1,  d0,  d1
676
    vrshrn.u16  d0,  q0,  #4
677
    vdup.8      q0,  d0[0]
678
    b           pred16x16_dc_end
679
endfunc
680

681
function x264_predict_16x16_dc_left_neon
682
    mov         r1,  #FDEC_STRIDE
683
    sub         r2,  r0,  #1
684
    ldcol.8     d0,  r2,  r1
685
    ldcol.8     d1,  r2,  r1
686
    add16x8     q0,  d0,  d1,  d0,  d1
687
    vrshrn.u16  d0,  q0,  #4
688
    vdup.8      q0,  d0[0]
689
    b           pred16x16_dc_end
690
endfunc
691

692
function x264_predict_16x16_dc_neon
693
    sub         r3, r0, #FDEC_STRIDE
694
    sub         r0, r0, #1
695
    vld1.64     {d0-d1}, [r3,:128]
696
    ldrb        ip, [r0], #FDEC_STRIDE
697
    vaddl.u8    q0, d0, d1
698
    ldrb        r1, [r0], #FDEC_STRIDE
699
    vadd.u16    d0, d0, d1
700
    vpadd.u16   d0, d0, d0
701
    vpadd.u16   d0, d0, d0
702
.rept 4
703
    ldrb        r2, [r0], #FDEC_STRIDE
704
    add         ip, ip, r1
705
    ldrb        r3, [r0], #FDEC_STRIDE
706
    add         ip, ip, r2
707
    ldrb        r1, [r0], #FDEC_STRIDE
708
    add         ip, ip, r3
709
.endr
710
    ldrb        r2, [r0], #FDEC_STRIDE
711
    add         ip, ip, r1
712
    ldrb        r3, [r0], #FDEC_STRIDE
713
    add         ip, ip, r2
714

715
    sub         r0, r0, #FDEC_STRIDE*16
716
    add         ip, ip, r3
717
    vdup.16     d1, ip
718
    vadd.u16    d0, d0, d1
719
    mov         r1, #FDEC_STRIDE
720
    add         r0, r0, #1
721
    vrshr.u16   d0, d0, #5
722
    vdup.8      q0, d0[0]
723
pred16x16_dc_end:
724
.rept 16
725
    vst1.64     {d0-d1}, [r0,:128], r1
726
.endr
727
    bx          lr
728
endfunc
729

730
function x264_predict_16x16_h_neon
731
    sub         r1, r0, #1
732
    mov         ip, #FDEC_STRIDE
733
.rept 8
734
    vld1.8      {d0[]}, [r1], ip
735
    vmov        d1, d0
736
    vld1.8      {d2[]}, [r1], ip
737
    vmov        d3, d2
738
    vst1.64     {d0-d1}, [r0,:128], ip
739
    vst1.64     {d2-d3}, [r0,:128], ip
740
.endr
741
    bx          lr
742
endfunc
743

744
function x264_predict_16x16_v_neon
745
    sub         r0, r0, #FDEC_STRIDE
746
    mov         ip, #FDEC_STRIDE
747
    vld1.64     {d0-d1}, [r0,:128], ip
748
.rept 16
749
    vst1.64     {d0-d1}, [r0,:128], ip
750
.endr
751
    bx          lr
752
endfunc
753

754
function x264_predict_16x16_p_neon
755
    sub         r3,  r0,  #FDEC_STRIDE
756
    mov         r1,  #FDEC_STRIDE
757
    add         r2,  r3,  #8
758
    sub         r3,  r3,  #1
759
    vld1.8      {d0}, [r3]
760
    vld1.8      {d2}, [r2,:64], r1
761
    ldcol.8     d1,  r3,  r1
762
    add         r3,  r3,  r1
763
    ldcol.8     d3,  r3,  r1
764
    vrev64.8    q0,  q0
765
    vaddl.u8    q8,  d2,  d3
766
    vsubl.u8    q2,  d2,  d0
767
    vsubl.u8    q3,  d3,  d1
768
    movrel      r3,  p16weight
769
    vld1.8      {q0}, [r3,:128]
770
    vmul.s16    q2,  q2,  q0
771
    vmul.s16    q3,  q3,  q0
772
    vadd.i16    d4,  d4,  d5
773
    vadd.i16    d5,  d6,  d7
774
    vpadd.i16   d4,  d4,  d5
775
    vpadd.i16   d4,  d4,  d4
776
    vshll.s16   q3,  d4,  #2
777
    vaddw.s16   q2,  q3,  d4
778
    vrshrn.s32  d4,  q2,  #6
779
    mov         r3,  #0
780
    vtrn.16     d4,  d5
781
    vadd.i16    d2,  d4,  d5
782
    vshl.i16    d3,  d2,  #3
783
    vrev64.16   d16, d17
784
    vsub.i16    d3,  d3,  d2
785
    vadd.i16    d16, d16, d0
786
    vshl.i16    d2,  d16, #4
787
    vsub.i16    d2,  d2,  d3
788
    vshl.i16    d3,  d4,  #4
789
    vext.16     q0,  q0,  q0,  #7
790
    vsub.i16    d6,  d5,  d3
791
    vmov.16     d0[0], r3
792
    vmul.i16    q0,  q0,  d4[0]
793
    vdup.16     q1,  d2[0]
794
    vdup.16     q2,  d4[0]
795
    vdup.16     q3,  d6[0]
796
    vshl.i16    q2,  q2,  #3
797
    vadd.i16    q1,  q1,  q0
798
    vadd.i16    q3,  q3,  q2
799
    mov         r3,  #16
800
1:
801
    vqshrun.s16 d0,  q1,  #5
802
    vadd.i16    q1,  q1,  q2
803
    vqshrun.s16 d1,  q1,  #5
804
    vadd.i16    q1,  q1,  q3
805
    vst1.8      {q0}, [r0,:128], r1
806
    subs        r3,  r3,  #1
807
    bne         1b
808
    bx          lr
809
endfunc
810

811
Product

Resources

Company
