1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
|
;******************************************************************************
;* VP9 loop filter SIMD optimizations
;*
;* Copyright (C) 2013-2014 Clément Bœsch <u pkh me>
;*
;* This file is part of FFmpeg.
;*
;* FFmpeg is free software; you can redistribute it and/or
;* modify it under the terms of the GNU Lesser General Public
;* License as published by the Free Software Foundation; either
;* version 2.1 of the License, or (at your option) any later version.
;*
;* FFmpeg is distributed in the hope that it will be useful,
;* but WITHOUT ANY WARRANTY; without even the implied warranty of
;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
;* Lesser General Public License for more details.
;*
;* You should have received a copy of the GNU Lesser General Public
;* License along with FFmpeg; if not, write to the Free Software
;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
;******************************************************************************
%if ARCH_X86_64
%include "libavutil/x86/x86util.asm"
SECTION_RODATA
cextern pb_3
cextern pb_80
pb_4: times 16 db 0x04
pb_10: times 16 db 0x10
pb_40: times 16 db 0x40
pb_81: times 16 db 0x81
pb_f8: times 16 db 0xf8
pb_fe: times 16 db 0xfe
pw_4: times 8 dw 4
pw_8: times 8 dw 8
SECTION .text
; %1 = abs(%2-%3)
%macro ABSSUB 4 ; dst, src1 (RO), src2 (RO), tmp
psubusb %1, %3, %2
psubusb %4, %2, %3
por %1, %4
%endmacro
; %1 = %1<=%2
%macro CMP_LTE 4 ; src/dst, cmp, tmp, pb_80
pxor %1, %4
pcmpgtb %3, %2, %1 ; cmp > src?
pcmpeqb %1, %2 ; cmp == src? XXX: avoid this with a -1/+1 well placed?
por %1, %3 ; cmp >= src?
%endmacro
; %1 = abs(%2-%3) <= %4
%macro ABSSUB_CMP 6-7 [pb_80]; dst, src1, src2, cmp, tmp1, tmp2, [pb_80]
ABSSUB %1, %2, %3, %6 ; dst = abs(src1-src2)
CMP_LTE %1, %4, %6, %7 ; dst <= cmp
%endmacro
%macro MASK_APPLY 4 ; %1=new_data/dst %2=old_data %3=mask %4=tmp
pand %1, %3 ; new &= mask
pandn %4, %3, %2 ; tmp = ~mask & old
por %1, %4 ; new&mask | old&~mask
%endmacro
%macro FILTER_SUBx2_ADDx2 8 ; %1=dst %2=h/l %3=cache %4=sub1 %5=sub2 %6=add1 %7=add2 %8=rshift
punpck%2bw %3, %4, m0
psubw %1, %3
punpck%2bw %3, %5, m0
psubw %1, %3
punpck%2bw %3, %6, m0
paddw %1, %3
punpck%2bw %3, %7, m0
paddw %1, %3
mova %3, %1
psraw %1, %8
%endmacro
%macro FILTER_INIT 7-8 ; tmp1, tmp2, cacheL, cacheH, dstp, filterid, [source]
FILTER%6_INIT %1, l, %3
FILTER%6_INIT %2, h, %4
packuswb %1, %2
%if %0 == 8
MASK_APPLY %1, %8, %7, %2
%else
MASK_APPLY %1, %5, %7, %2
%endif
mova %5, %1
%endmacro
%macro FILTER_UPDATE 11-12 ; tmp1, tmp2, cacheL, cacheH, dstp, -, -, +, +, rshift, [source]
FILTER_SUBx2_ADDx2 %1, l, %3, %6, %7, %8, %9, %10
FILTER_SUBx2_ADDx2 %2, h, %4, %6, %7, %8, %9, %10
packuswb %1, %2
%if %0 == 12
MASK_APPLY %1, %12, %11, %2
%else
MASK_APPLY %1, %5, %11, %2
%endif
mova %5, %1
%endmacro
%macro SRSHIFT3B_2X 4 ; reg1, reg2, [pb_10], tmp
mova %4, [pb_f8]
pand %1, %4
pand %2, %4
psrlq %1, 3
psrlq %2, 3
pxor %1, %3
pxor %2, %3
psubb %1, %3
psubb %2, %3
%endmacro
%macro EXTRACT_POS_NEG 3 ; i8, neg, pos
pxor %3, %3
pxor %2, %2
pcmpgtb %3, %1 ; i8 < 0 mask
psubb %2, %1 ; neg values (only the originally - will be kept)
pand %2, %3 ; negative values of i8 (but stored as +)
pandn %3, %1 ; positive values of i8
%endmacro
; clip_u8(u8 + i8)
%macro SIGN_ADD 5 ; dst, u8, i8, tmp1, tmp2
EXTRACT_POS_NEG %3, %4, %5
psubusb %1, %2, %4 ; sub the negatives
paddusb %1, %5 ; add the positives
%endmacro
; clip_u8(u8 - i8)
%macro SIGN_SUB 5 ; dst, u8, i8, tmp1, tmp2
EXTRACT_POS_NEG %3, %4, %5
psubusb %1, %2, %5 ; sub the positives
paddusb %1, %4 ; add the negatives
%endmacro
%macro FILTER6_INIT 3 ; %1=dst %2=h/l %3=cache
punpck%2bw %3, m14, m0 ; p3: B->W
mova %1, %3 ; p3
paddw %1, %3 ; p3*2
paddw %1, %3 ; p3*3
punpck%2bw %3, m15, m0 ; p2: B->W
paddw %1, %3 ; p3*3 + p2
paddw %1, %3 ; p3*3 + p2*2
punpck%2bw %3, m10, m0 ; p1: B->W
paddw %1, %3 ; p3*3 + p2*2 + p1
punpck%2bw %3, m11, m0 ; p0: B->W
paddw %1, %3 ; p3*3 + p2*2 + p1 + p0
punpck%2bw %3, m12, m0 ; q0: B->W
paddw %1, %3 ; p3*3 + p2*2 + p1 + p0 + q0
paddw %1, [pw_4] ; p3*3 + p2*2 + p1 + p0 + q0 + 4
mova %3, %1 ; base for next line (cache)
psraw %1, 3 ; (p3*3 + p2*2 + p1 + p0 + q0 + 4) >> 3
%endmacro
%macro FILTER14_INIT 3 ; %1=dst %2=h/l %3=cache
punpck%2bw %1, m2, m0 ; p7: B->W
mova %3, %1
psllw %1, 3 ; p7*8
psubw %1, %3 ; p7*7
punpck%2bw %3, m3, m0 ; p6: B->W
paddw %1, %3 ; p7*7 + p6
paddw %1, %3 ; p7*7 + p6*2
punpck%2bw %3, m8, m0 ; p5: B->W
paddw %1, %3 ; p7*7 + p6*2 + p5
punpck%2bw %3, m9, m0 ; p4: B->W
paddw %1, %3 ; p7*7 + p6*2 + p5 + p4
punpck%2bw %3, m14, m0 ; p3: B->W
paddw %1, %3 ; p7*7 + p6*2 + p5 + p4 + p3
punpck%2bw %3, m15, m0 ; p2: B->W
paddw %1, %3 ; p7*7 + p6*2 + p5 + .. + p2
punpck%2bw %3, m10, m0 ; p1: B->W
paddw %1, %3 ; p7*7 + p6*2 + p5 + .. + p1
punpck%2bw %3, m11, m0 ; p0: B->W
paddw %1, %3 ; p7*7 + p6*2 + p5 + .. + p0
punpck%2bw %3, m12, m0 ; q0: B->W
paddw %1, %3 ; p7*7 + p6*2 + p5 + .. + p0 + q0
paddw %1, [pw_8] ; p7*7 + p6*2 + p5 + .. + p0 + q0 + 8
mova %3, %1 ; base for next line (cache)
psraw %1, 4 ; (p7*7 + p6*2 + p5 + .. + p0 + q0 + 8) >> 4
%endmacro
%macro TRANSPOSE16x16B 17
mova %17, m%16
SBUTTERFLY bw, %1, %2, %16
SBUTTERFLY bw, %3, %4, %16
SBUTTERFLY bw, %5, %6, %16
SBUTTERFLY bw, %7, %8, %16
SBUTTERFLY bw, %9, %10, %16
SBUTTERFLY bw, %11, %12, %16
SBUTTERFLY bw, %13, %14, %16
mova m%16, %17
mova %17, m%14
SBUTTERFLY bw, %15, %16, %14
SBUTTERFLY wd, %1, %3, %14
SBUTTERFLY wd, %2, %4, %14
SBUTTERFLY wd, %5, %7, %14
SBUTTERFLY wd, %6, %8, %14
SBUTTERFLY wd, %9, %11, %14
SBUTTERFLY wd, %10, %12, %14
SBUTTERFLY wd, %13, %15, %14
mova m%14, %17
mova %17, m%12
SBUTTERFLY wd, %14, %16, %12
SBUTTERFLY dq, %1, %5, %12
SBUTTERFLY dq, %2, %6, %12
SBUTTERFLY dq, %3, %7, %12
SBUTTERFLY dq, %4, %8, %12
SBUTTERFLY dq, %9, %13, %12
SBUTTERFLY dq, %10, %14, %12
SBUTTERFLY dq, %11, %15, %12
mova m%12, %17
mova %17, m%8
SBUTTERFLY dq, %12, %16, %8
SBUTTERFLY qdq, %1, %9, %8
SBUTTERFLY qdq, %2, %10, %8
SBUTTERFLY qdq, %3, %11, %8
SBUTTERFLY qdq, %4, %12, %8
SBUTTERFLY qdq, %5, %13, %8
SBUTTERFLY qdq, %6, %14, %8
SBUTTERFLY qdq, %7, %15, %8
mova m%8, %17
mova %17, m%1
SBUTTERFLY qdq, %8, %16, %1
mova m%1, %17
SWAP %2, %9
SWAP %3, %5
SWAP %4, %13
SWAP %6, %11
SWAP %8, %15
SWAP %12, %14
%endmacro
%macro LPF_16_16 1
lea mstrideq, [strideq]
neg mstrideq
lea stride3q, [strideq+2*strideq]
mov mstride3q, stride3q
neg mstride3q
%ifidn %1, h
lea dstq, [dstq + 8*strideq - 8] ; go from top center (h pos) to center left (v pos)
%endif
lea dst1q, [dstq + 2*mstride3q] ; dst1q = &dst[stride * -6]
lea dst2q, [dstq + 2* stride3q] ; dst2q = &dst[stride * +6]
%define P7 dst1q + 2*mstrideq
%define P6 dst1q + mstrideq
%define P5 dst1q
%define P4 dst1q + strideq
%define P3 dstq + 4*mstrideq
%define P2 dstq + mstride3q
%define P1 dstq + 2*mstrideq
%define P0 dstq + mstrideq
%define Q0 dstq
%define Q1 dstq + strideq
%define Q2 dstq + 2*strideq
%define Q3 dstq + stride3q
%define Q4 dstq + 4*strideq
%define Q5 dst2q + mstrideq
%define Q6 dst2q
%define Q7 dst2q + strideq
%ifidn %1, h
movu m0, [P7]
movu m1, [P6]
movu m2, [P5]
movu m3, [P4]
movu m4, [P3]
movu m5, [P2]
movu m6, [P1]
movu m7, [P0]
movu m8, [Q0]
movu m9, [Q1]
movu m10, [Q2]
movu m11, [Q3]
movu m12, [Q4]
movu m13, [Q5]
movu m14, [Q6]
movu m15, [Q7]
TRANSPOSE16x16B 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, [rsp]
mova [rsp + 0], m0 ; dst[stride * -8] (p7)
mova [rsp + 16], m1 ; dst[stride * -7] (p6)
mova [rsp + 32], m2 ; dst[stride * -6] (p5)
mova [rsp + 48], m3 ; dst[stride * -5] (p4)
mova [rsp + 64], m4 ; dst[stride * -4] (p3)
mova [rsp + 80], m5 ; dst[stride * -3] (p2)
mova [rsp + 96], m6 ; dst[stride * -2] (p1)
mova [rsp + 112], m7 ; dst[stride * -1] (p0)
mova [rsp + 128], m8 ; dst[stride * +0] (q0)
mova [rsp + 144], m9 ; dst[stride * +1] (q1)
mova [rsp + 160], m10 ; dst[stride * +2] (q2)
mova [rsp + 176], m11 ; dst[stride * +3] (q3)
mova [rsp + 192], m12 ; dst[stride * +4] (q4)
mova [rsp + 208], m13 ; dst[stride * +5] (q5)
mova [rsp + 224], m14 ; dst[stride * +6] (q6)
mova [rsp + 240], m15 ; dst[stride * +7] (q7)
%endif
; calc fm mask
%if cpuflag(ssse3)
pxor m0, m0
%endif
SPLATB_REG m2, I, m0 ; I I I I ...
SPLATB_REG m3, E, m0 ; E E E E ...
mova m0, [pb_80]
pxor m2, m0
pxor m3, m0
%ifidn %1, v
mova m8, [P3]
mova m9, [P2]
mova m10, [P1]
mova m11, [P0]
mova m12, [Q0]
mova m13, [Q1]
mova m14, [Q2]
mova m15, [Q3]
%else
SWAP 8, 4, 12
SWAP 9, 5, 13
SWAP 10, 6, 14
SWAP 11, 7, 15
%endif
ABSSUB_CMP m5, m8, m9, m2, m6, m7, m0 ; m5 = abs(p3-p2) <= I
ABSSUB_CMP m1, m9, m10, m2, m6, m7, m0 ; m1 = abs(p2-p1) <= I
pand m5, m1
ABSSUB_CMP m1, m10, m11, m2, m6, m7, m0 ; m1 = abs(p1-p0) <= I
pand m5, m1
ABSSUB_CMP m1, m12, m13, m2, m6, m7, m0 ; m1 = abs(q1-q0) <= I
pand m5, m1
ABSSUB_CMP m1, m13, m14, m2, m6, m7, m0 ; m1 = abs(q2-q1) <= I
pand m5, m1
ABSSUB_CMP m1, m14, m15, m2, m6, m7, m0 ; m1 = abs(q3-q2) <= I
pand m5, m1
ABSSUB m1, m11, m12, m7 ; abs(p0-q0)
paddusb m1, m1 ; abs(p0-q0) * 2
ABSSUB m2, m10, m13, m7 ; abs(p1-q1)
pand m2, [pb_fe] ; drop lsb so shift can work
psrlq m2, 1 ; abs(p1-q1)/2
paddusb m1, m2 ; abs(p0-q0)*2 + abs(p1-q1)/2
pxor m1, m0
pcmpgtb m4, m3, m1 ; E > X?
pcmpeqb m3, m1 ; E == X?
por m3, m4 ; E >= X?
pand m3, m5 ; fm final value
; (m3: fm, m8..15: p3 p2 p1 p0 q0 q1 q2 q3)
; calc flat8in and hev masks
mova m6, [pb_81] ; [1 1 1 1 ...] ^ 0x80
ABSSUB_CMP m2, m8, m11, m6, m4, m5 ; abs(p3 - p0) <= 1
mova m8, [pb_80]
ABSSUB_CMP m1, m9, m11, m6, m4, m5, m8 ; abs(p2 - p0) <= 1
pand m2, m1
ABSSUB m4, m10, m11, m5 ; abs(p1 - p0)
%if cpuflag(ssse3)
pxor m0, m0
%endif
SPLATB_REG m7, H, m0 ; H H H H ...
pxor m7, m8
pxor m4, m8
pcmpgtb m0, m4, m7 ; abs(p1 - p0) > H (1/2 hev condition)
pxor m4, m8
mova m1, m4
CMP_LTE m1, m6, m5, m8 ; abs(p1 - p0) <= 1
pand m2, m1 ; (flat8in)
ABSSUB m4, m13, m12, m1 ; abs(q1 - q0)
pxor m4, m8
pcmpgtb m5, m4, m7 ; abs(q1 - q0) > H (2/2 hev condition)
pxor m4, m8
por m0, m5 ; hev final value
mova m1, m4
CMP_LTE m1, m6, m5, m8 ; abs(q1 - q0) <= 1
pand m2, m1 ; (flat8in)
ABSSUB_CMP m1, m14, m12, m6, m4, m5, m8 ; abs(q2 - q0) <= 1
pand m2, m1
ABSSUB_CMP m1, m15, m12, m6, m4, m5, m8 ; abs(q3 - q0) <= 1
pand m2, m1 ; flat8in final value
; (m0: hev, m2: flat8in, m3: fm, m6: pb_81, m9..15: p2 p1 p0 q0 q1 q2 q3)
; calc flat8out mask
%ifidn %1, v
mova m8, [P7]
mova m9, [P6]
%else
mova m8, [rsp + 0] ; m8 = p7
mova m9, [rsp + 16] ; m9 = p6
%endif
ABSSUB_CMP m1, m8, m11, m6, m4, m5 ; abs(p7 - p0) <= 1
ABSSUB_CMP m7, m9, m11, m6, m4, m5 ; abs(p6 - p0) <= 1
pand m1, m7
%ifidn %1, v
mova m8, [P5]
mova m9, [P4]
%else
mova m8, [rsp + 32] ; m8 = p5
mova m9, [rsp + 48] ; m9 = p4
%endif
ABSSUB_CMP m7, m8, m11, m6, m4, m5 ; abs(p5 - p0) <= 1
pand m1, m7
ABSSUB_CMP m7, m9, m11, m6, m4, m5 ; abs(p4 - p0) <= 1
pand m1, m7
%ifidn %1, v
mova m14, [Q4]
mova m15, [Q5]
%else
mova m14, [rsp + 192] ; m14 = q4
mova m15, [rsp + 208] ; m15 = q5
%endif
ABSSUB_CMP m7, m14, m12, m6, m4, m5 ; abs(q4 - q0) <= 1
pand m1, m7
ABSSUB_CMP m7, m15, m12, m6, m4, m5 ; abs(q5 - q0) <= 1
pand m1, m7
%ifidn %1, v
mova m14, [Q6]
mova m15, [Q7]
%else
mova m14, [rsp + 224] ; m14 = q6
mova m15, [rsp + 240] ; m15 = q7
%endif
ABSSUB_CMP m7, m14, m12, m6, m4, m5 ; abs(q4 - q0) <= 1
pand m1, m7
ABSSUB_CMP m7, m15, m12, m6, m4, m5 ; abs(q5 - q0) <= 1
pand m1, m7 ; flat8out final value
; if (fm) {
; if (out && in) filter_14()
; else if (in) filter_6()
; else if (hev) filter_2()
; else filter_4()
; }
;
; f14: fm & out & in
; f6: fm & ~f14 & in => fm & ~(out & in) & in => fm & ~out & in
; f2: fm & ~f14 & ~f6 & hev => fm & ~(out & in) & ~(~out & in) & hev => fm & ~in & hev
; f4: fm & ~f14 & ~f6 & ~f2 => fm & ~(out & in) & ~(~out & in) & ~(~in & hev) => fm & ~in & ~hev
; (m0: hev, m1: flat8out, m2: flat8in, m3: fm, m8..15: p5 p4 p1 p0 q0 q1 q6 q7)
; filter2()
mova m6, [pb_80]
pxor m15, m12, m6 ; q0 ^ 0x80
pxor m14, m11, m6 ; p0 ^ 0x80
psubsb m15, m14 ; (signed) q0 - p0
pxor m4, m10, m6 ; p1 ^ 0x80
pxor m5, m13, m6 ; q1 ^ 0x80
psubsb m4, m5 ; (signed) p1 - q1
paddsb m4, m15 ; (q0 - p0) + (p1 - q1)
paddsb m4, m15 ; 2*(q0 - p0) + (p1 - q1)
paddsb m4, m15 ; 3*(q0 - p0) + (p1 - q1)
paddsb m6, m4, [pb_4] ; m6: f1 = clip(f + 4, 127)
paddsb m4, [pb_3] ; m4: f2 = clip(f + 3, 127)
mova m14, [pb_10] ; will be reused in filter4()
SRSHIFT3B_2X m6, m4, m14, m7 ; f1 and f2 sign byte shift by 3
SIGN_SUB m7, m12, m6, m5, m9 ; m7 = q0 - f1
SIGN_ADD m8, m11, m4, m5, m9 ; m8 = p0 + f2
pandn m6, m2, m3 ; ~mask(in) & mask(fm)
pand m6, m0 ; (~mask(in) & mask(fm)) & mask(hev)
MASK_APPLY m7, m12, m6, m5 ; m7 = filter2(q0) & mask / we write it in filter4()
MASK_APPLY m8, m11, m6, m5 ; m8 = filter2(p0) & mask / we write it in filter4()
; (m0: hev, m1: flat8out, m2: flat8in, m3: fm, m7..m8: q0' p0', m10..13: p1 p0 q0 q1, m14: pb_10, m15: q0-p0)
; filter4()
mova m4, m15
paddsb m15, m4 ; 2 * (q0 - p0)
paddsb m15, m4 ; 3 * (q0 - p0)
paddsb m6, m15, [pb_4] ; m6: f1 = clip(f + 4, 127)
paddsb m15, [pb_3] ; m15: f2 = clip(f + 3, 127)
SRSHIFT3B_2X m6, m15, m14, m9 ; f1 and f2 sign byte shift by 3
pandn m5, m2, m3 ; ~mask(in) & mask(fm)
pandn m0, m5 ; ~mask(hev) & (~mask(in) & mask(fm))
SIGN_SUB m9, m12, m6, m4, m14 ; q0 - f1
MASK_APPLY m9, m7, m0, m5 ; m9 = filter4(q0) & mask
%ifidn %1, v
mova [dstq], m9 ; update q0
%else
mova [rsp + 128], m9 ; update q0
%endif
SIGN_ADD m7, m11, m15, m4, m14 ; p0 + f2
MASK_APPLY m7, m8, m0, m5 ; m7 = filter4(p0) & mask
%ifidn %1, v
mova [dstq + 1*mstrideq], m7 ; update p0
%else
mova [rsp + 112], m7 ; update p0
%endif
paddb m6, [pb_80] ;
pxor m8, m8 ; f=(f1+1)>>1
pavgb m6, m8 ;
psubb m6, [pb_40] ;
SIGN_ADD m7, m10, m6, m8, m9 ; p1 + f
SIGN_SUB m4, m13, m6, m8, m9 ; q1 - f
MASK_APPLY m7, m10, m0, m14 ; m7 = filter4(p1)
MASK_APPLY m4, m13, m0, m14 ; m4 = filter4(q1)
%ifidn %1, v
mova [dstq + 2*mstrideq], m7 ; update p1
mova [dstq + 1* strideq], m4 ; update q1
%else
mova [rsp + 96], m7 ; update p1
mova [rsp + 144], m4 ; update q1
%endif
; (m1: flat8out, m2: flat8in, m3: fm, m10..13: p1 p0 q0 q1)
; filter6()
pxor m0, m0
pand m2, m3 ; mask(fm) & mask(in)
pandn m3, m1, m2 ; ~mask(out) & (mask(fm) & mask(in))
%ifidn %1, v
mova m14, [P3]
mova m15, [P2]
mova m8, [Q2]
mova m9, [Q3]
FILTER_INIT m4, m5, m6, m7, [dstq + mstride3q], 6, m3, m15 ; [p2]
FILTER_UPDATE m6, m7, m4, m5, [dstq +2*mstrideq ], m14, m15, m10, m13, 3, m3 ; [p1] -p3 -p2 +p1 +q1
FILTER_UPDATE m4, m5, m6, m7, [dstq + mstrideq ], m14, m10, m11, m8, 3, m3 ; [p0] -p3 -p1 +p0 +q2
FILTER_UPDATE m6, m7, m4, m5, [dstq ], m14, m11, m12, m9, 3, m3 ; [q0] -p3 -p0 +q0 +q3
FILTER_UPDATE m4, m5, m6, m7, [dstq + 1*strideq], m15, m12, m13, m9, 3, m3 ; [q1] -p2 -q0 +q1 +q3
FILTER_UPDATE m6, m7, m4, m5, [dstq + 2*strideq], m10, m13, m8, m9, 3, m3, m8 ; [q2] -p1 -q1 +q2 +q3
%else
mova m14, [rsp + 64] ; m14 = p3
mova m15, [rsp + 80] ; m15 = p2
mova m8, [rsp + 160] ; m8 = q2
mova m9, [rsp + 176] ; m9 = q3
FILTER_INIT m4, m5, m6, m7, [rsp + 80], 6, m3, m15 ; [p2]
FILTER_UPDATE m6, m7, m4, m5, [rsp + 96], m14, m15, m10, m13, 3, m3 ; [p1] -p3 -p2 +p1 +q1
FILTER_UPDATE m4, m5, m6, m7, [rsp + 112], m14, m10, m11, m8, 3, m3 ; [p0] -p3 -p1 +p0 +q2
FILTER_UPDATE m6, m7, m4, m5, [rsp + 128], m14, m11, m12, m9, 3, m3 ; [q0] -p3 -p0 +q0 +q3
FILTER_UPDATE m4, m5, m6, m7, [rsp + 144], m15, m12, m13, m9, 3, m3 ; [q1] -p2 -q0 +q1 +q3
FILTER_UPDATE m6, m7, m4, m5, [rsp + 160], m10, m13, m8, m9, 3, m3, m8 ; [q2] -p1 -q1 +q2 +q3
%endif
; (m0: 0, m1: flat8out, m2: fm & flat8in, m8..15: q2 q3 p1 p0 q0 q1 p3 p2)
; filter14()
;
; m2 m3 m8 m9 m14 m15 m10 m11 m12 m13
;
; q2 q3 p3 p2 p1 p0 q0 q1
; p6 -7 p7 p6 p5 p4 . . . . .
; p5 -6 -p7 -p6 +p5 +q1 . . . .
; p4 -5 -p7 -p5 +p4 +q2 . . . q2
; p3 -4 -p7 -p4 +p3 +q3 . . . q3
; p2 -3 -p7 -p3 +p2 +q4 . . . q4
; p1 -2 -p7 -p2 +p1 +q5 . . . q5
; p0 -1 -p7 -p1 +p0 +q6 . . . q6
; q0 +0 -p7 -p0 +q0 +q7 . . . q7
; q1 +1 -p6 -q0 +q1 +q7 q1 . . .
; q2 +2 -p5 -q1 +q2 +q7 . q2 . .
; q3 +3 -p4 -q2 +q3 +q7 . q3 . .
; q4 +4 -p3 -q3 +q4 +q7 . q4 . .
; q5 +5 -p2 -q4 +q5 +q7 . q5 . .
; q6 +6 -p1 -q5 +q6 +q7 . q6 . .
pand m1, m2 ; mask(out) & (mask(fm) & mask(in))
%ifidn %1, v
mova m2, [P7]
mova m3, [P6]
mova m8, [P5]
mova m9, [P4]
FILTER_INIT m4, m5, m6, m7, [P6], 14, m1, m3
FILTER_UPDATE m6, m7, m4, m5, [P5], m2, m3, m8, m13, 4, m1, m8 ; [p5] -p7 -p6 +p5 +q1
mova m13, [Q2]
FILTER_UPDATE m4, m5, m6, m7, [P4], m2, m8, m9, m13, 4, m1, m9 ; [p4] -p7 -p5 +p4 +q2
mova m13, [Q3]
FILTER_UPDATE m6, m7, m4, m5, [P3], m2, m9, m14, m13, 4, m1, m14 ; [p3] -p7 -p4 +p3 +q3
mova m13, [Q4]
FILTER_UPDATE m4, m5, m6, m7, [P2], m2, m14, m15, m13, 4, m1 ; [p2] -p7 -p3 +p2 +q4
mova m13, [Q5]
FILTER_UPDATE m6, m7, m4, m5, [P1], m2, m15, m10, m13, 4, m1 ; [p1] -p7 -p2 +p1 +q5
mova m13, [Q6]
FILTER_UPDATE m4, m5, m6, m7, [P0], m2, m10, m11, m13, 4, m1 ; [p0] -p7 -p1 +p0 +q6
mova m13, [Q7]
FILTER_UPDATE m6, m7, m4, m5, [Q0], m2, m11, m12, m13, 4, m1 ; [q0] -p7 -p0 +q0 +q7
mova m2, [Q1]
FILTER_UPDATE m4, m5, m6, m7, [Q1], m3, m12, m2, m13, 4, m1 ; [q1] -p6 -q0 +q1 +q7
mova m3, [Q2]
FILTER_UPDATE m6, m7, m4, m5, [Q2], m8, m2, m3, m13, 4, m1 ; [q2] -p5 -q1 +q2 +q7
mova m8, [Q3]
FILTER_UPDATE m4, m5, m6, m7, [Q3], m9, m3, m8, m13, 4, m1, m8 ; [q3] -p4 -q2 +q3 +q7
mova m9, [Q4]
FILTER_UPDATE m6, m7, m4, m5, [Q4], m14, m8, m9, m13, 4, m1, m9 ; [q4] -p3 -q3 +q4 +q7
mova m14, [Q5]
FILTER_UPDATE m4, m5, m6, m7, [Q5], m15, m9, m14, m13, 4, m1, m14 ; [q5] -p2 -q4 +q5 +q7
mova m15, [Q6]
FILTER_UPDATE m6, m7, m4, m5, [Q6], m10, m14, m15, m13, 4, m1, m15 ; [q6] -p1 -q5 +q6 +q7
%else
mova m2, [rsp + 0] ; m2 = p7
mova m3, [rsp + 16] ; m3 = p6
mova m8, [rsp + 32] ; m8 = p5
mova m9, [rsp + 48] ; m9 = p4
FILTER_INIT m4, m5, m6, m7, [rsp + 16], 14, m1, m3 ; [p6]
FILTER_UPDATE m6, m7, m4, m5, [rsp + 32], m2, m3, m8, m13, 4, m1, m8 ; [p5] -p7 -p6 +p5 +q1
mova m13, [rsp + 160] ; m13 = q2
FILTER_UPDATE m4, m5, m6, m7, [rsp + 48], m2, m8, m9, m13, 4, m1, m9 ; [p4] -p7 -p5 +p4 +q2
mova m13, [rsp + 176] ; m13 = q3
FILTER_UPDATE m6, m7, m4, m5, [rsp + 64], m2, m9, m14, m13, 4, m1, m14 ; [p3] -p7 -p4 +p3 +q3
mova m13, [rsp + 192] ; m13 = q4
FILTER_UPDATE m4, m5, m6, m7, [rsp + 80], m2, m14, m15, m13, 4, m1 ; [p2] -p7 -p3 +p2 +q4
mova m13, [rsp + 208] ; m13 = q5
FILTER_UPDATE m6, m7, m4, m5, [rsp + 96], m2, m15, m10, m13, 4, m1 ; [p1] -p7 -p2 +p1 +q5
mova m13, [rsp + 224] ; m13 = q6
FILTER_UPDATE m4, m5, m6, m7, [rsp + 112], m2, m10, m11, m13, 4, m1 ; [p0] -p7 -p1 +p0 +q6
mova m13, [rsp + 240] ; m13 = q7
FILTER_UPDATE m6, m7, m4, m5, [rsp + 128], m2, m11, m12, m13, 4, m1 ; [q0] -p7 -p0 +q0 +q7
mova m2, [rsp + 144] ; m2 = q1
FILTER_UPDATE m4, m5, m6, m7, [rsp + 144], m3, m12, m2, m13, 4, m1 ; [q1] -p6 -q0 +q1 +q7
mova m3, [rsp + 160] ; m3 = q2
FILTER_UPDATE m6, m7, m4, m5, [rsp + 160], m8, m2, m3, m13, 4, m1 ; [q2] -p5 -q1 +q2 +q7
mova m8, [rsp + 176] ; m8 = q3
FILTER_UPDATE m4, m5, m6, m7, [rsp + 176], m9, m3, m8, m13, 4, m1, m8 ; [q3] -p4 -q2 +q3 +q7
mova m9, [rsp + 192] ; m9 = q4
FILTER_UPDATE m6, m7, m4, m5, [rsp + 192], m14, m8, m9, m13, 4, m1, m9 ; [q4] -p3 -q3 +q4 +q7
mova m14, [rsp + 208] ; m14 = q5
FILTER_UPDATE m4, m5, m6, m7, [rsp + 208], m15, m9, m14, m13, 4, m1, m14 ; [q5] -p2 -q4 +q5 +q7
mova m15, [rsp + 224] ; m15 = q6
FILTER_UPDATE m6, m7, m4, m5, [rsp + 224], m10, m14, m15, m13, 4, m1, m15 ; [q6] -p1 -q5 +q6 +q7
%endif
%ifidn %1, h
mova m0, [rsp + 0] ; dst[stride * -8] (p7)
mova m1, [rsp + 16] ; dst[stride * -7] (p6)
mova m2, [rsp + 32] ; dst[stride * -6] (p5)
mova m3, [rsp + 48] ; dst[stride * -5] (p4)
mova m4, [rsp + 64] ; dst[stride * -4] (p3)
mova m5, [rsp + 80] ; dst[stride * -3] (p2)
mova m6, [rsp + 96] ; dst[stride * -2] (p1)
mova m7, [rsp + 112] ; dst[stride * -1] (p0)
mova m8, [rsp + 128] ; dst[stride * +0] (q0)
mova m9, [rsp + 144] ; dst[stride * +1] (q1)
mova m10, [rsp + 160] ; dst[stride * +2] (q2)
mova m11, [rsp + 176] ; dst[stride * +3] (q3)
mova m12, [rsp + 192] ; dst[stride * +4] (q4)
mova m13, [rsp + 208] ; dst[stride * +5] (q5)
mova m14, [rsp + 224] ; dst[stride * +6] (q6)
mova m15, [rsp + 240] ; dst[stride * +7] (q7)
TRANSPOSE16x16B 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, [rsp]
movu [P7], m0
movu [P6], m1
movu [P5], m2
movu [P4], m3
movu [P3], m4
movu [P2], m5
movu [P1], m6
movu [P0], m7
movu [Q0], m8
movu [Q1], m9
movu [Q2], m10
movu [Q3], m11
movu [Q4], m12
movu [Q5], m13
movu [Q6], m14
movu [Q7], m15
%endif
%endmacro
%macro LPF_16_16_VH 1
INIT_XMM %1
cglobal vp9_loop_filter_v_16_16, 5,10,16, dst, stride, E, I, H, mstride, dst1, dst2, stride3, mstride3
LPF_16_16 v
RET
cglobal vp9_loop_filter_h_16_16, 5,10,16, 256, dst, stride, E, I, H, mstride, dst1, dst2, stride3, mstride3
LPF_16_16 h
RET
%endmacro
LPF_16_16_VH sse2
LPF_16_16_VH ssse3
LPF_16_16_VH avx
%endif ; x86-64
|
