aboutsummaryrefslogtreecommitdiffstats
path: root/libavcodec/aarch64/vp8dsp_neon.S
blob: 490b0ca9dcdf113e6f0fab2be2ca642c3e7a9a33 (plain) (blame)
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
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
/*
 * VP8 NEON optimisations
 *
 * Copyright (c) 2010 Rob Clark <rob@ti.com>
 * Copyright (c) 2011 Mans Rullgard <mans@mansr.com>
 * Copyright (c) 2018 Magnus Röös <mla2.roos@gmail.com>
 *
 * 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
 */

#include "libavutil/aarch64/asm.S"
#include "neon.S"

function ff_vp8_idct_add_neon, export=1
        ld1             {v0.8b - v3.8b},  [x1]
        mov             w4,  #20091
        movk            w4,  #35468/2, lsl 16
        dup             v4.2s, w4

        smull           v26.4s, v1.4h,  v4.4h[0]
        smull           v27.4s, v3.4h,  v4.4h[0]
        sqdmulh         v20.4h, v1.4h,  v4.4h[1]
        sqdmulh         v23.4h, v3.4h,  v4.4h[1]
        sqshrn          v21.4h, v26.4s, #16
        sqshrn          v22.4h, v27.4s, #16
        add             v21.4h, v21.4h, v1.4h
        add             v22.4h, v22.4h, v3.4h

        add             v16.4h,  v0.4h,   v2.4h
        sub             v17.4h,  v0.4h,   v2.4h

        add             v18.4h,  v21.4h,  v23.4h
        sub             v19.4h,  v20.4h,  v22.4h

        add             v0.4h,   v16.4h,  v18.4h
        add             v1.4h,   v17.4h,  v19.4h
        sub             v3.4h,   v16.4h,  v18.4h
        sub             v2.4h,   v17.4h,  v19.4h

        transpose_4x4H  v0, v1, v2, v3, v24, v5, v6, v7

        movi            v29.8h, #0
        smull           v26.4s,     v1.4h,  v4.4h[0]
        st1             {v29.8h},   [x1],   #16
        smull           v27.4s,     v3.4h,  v4.4h[0]
        st1             {v29.16b},  [x1]
        sqdmulh         v21.4h,     v1.4h,  v4.4h[1]
        sqdmulh         v23.4h,     v3.4h,  v4.4h[1]
        sqshrn          v20.4h,     v26.4s, #16
        sqshrn          v22.4h,     v27.4s, #16
        add             v20.4h,     v20.4h, v1.4h
        add             v22.4h,     v22.4h, v3.4h
        add             v16.4h,     v0.4h,  v2.4h
        sub             v17.4h,     v0.4h,  v2.4h

        add             v18.4h,     v20.4h, v23.4h
        ld1             {v24.d}[0], [x0],   x2
        zip1            v16.2d,     v16.2d, v17.2d
        sub             v19.4h,     v21.4h, v22.4h
        ld1             {v25.d}[0], [x0],   x2
        zip1            v18.2d,     v18.2d, v19.2d
        add             v0.8h,      v16.8h, v18.8h
        ld1             {v25.d}[1], [x0],   x2
        sub             v1.8h,      v16.8h, v18.8h
        ld1             {v24.d}[1], [x0],   x2
        srshr           v0.8h,      v0.8h,  #3
        trn1            v24.4s,     v24.4s, v25.4s
        srshr           v1.8h,      v1.8h,  #3
        sub             x0,  x0,  x2,  lsl #2

        ext             v1.16b, v1.16b, v1.16b, #8
        trn1            v3.2d,  v0.2d,  v1.2d
        trn2            v0.2d,  v0.2d,  v1.2d
        trn1            v1.8h,  v3.8h,  v0.8h
        trn2            v3.8h,  v3.8h,  v0.8h
        uzp1            v0.4s,  v1.4s,  v3.4s
        uzp2            v1.4s,  v3.4s,  v1.4s

        uaddw           v0.8h,  v0.8h, v24.8b
        uaddw2          v1.8h,  v1.8h, v24.16b
        sqxtun          v0.8b,  v0.8h
        sqxtun2         v0.16b, v1.8h
        st1             {v0.s}[0],  [x0], x2
        st1             {v0.s}[1],  [x0], x2
        st1             {v0.s}[3],  [x0], x2
        st1             {v0.s}[2],  [x0], x2

        ret
endfunc

function ff_vp8_idct_dc_add4y_neon, export=1
        movi            v0.16b,  #0
        mov             x3,  #32
        ld1r            {v16.4h},    [x1]
        st1             {v0.h}[0],   [x1], x3
        ld1r            {v17.4h},    [x1]
        st1             {v0.h}[0],   [x1], x3
        zip1            v16.2d,      v16.2d, v17.2d
        ld1r            {v18.4h},    [x1]
        st1             {v0.h}[0],   [x1], x3
        ld1r            {v19.4h},    [x1]
        st1             {v0.h}[0],   [x1], x3
        zip1            v18.2d,      v18.2d, v19.2d
        srshr           v16.8h,      v16.8h,  #3            // dc >>= 3
        ld1             {v0.16b},     [x0], x2
        srshr           v18.8h,       v18.8h,  #3
        ld1             {v1.16b},     [x0], x2
        uaddw           v20.8h,       v16.8h,  v0.8b
        ld1             {v2.16b},     [x0], x2
        uaddw2          v0.8h,        v18.8h,   v0.16b
        ld1             {v3.16b},     [x0], x2
        uaddw           v21.8h, v16.8h,  v1.8b
        uaddw2          v1.8h,  v18.8h,  v1.16b
        uaddw           v22.8h, v16.8h,  v2.8b
        uaddw2          v2.8h,  v18.8h,  v2.16b
        uaddw           v23.8h, v16.8h,  v3.8b
        uaddw2          v3.8h,  v18.8h,  v3.16b
        sub             x0,  x0,  x2,  lsl #2
        sqxtun          v20.8b,  v20.8h
        sqxtun2         v20.16b, v0.8h
        sqxtun          v21.8b,  v21.8h
        sqxtun2         v21.16b, v1.8h
        sqxtun          v22.8b,  v22.8h
        st1             {v20.16b},    [x0], x2
        sqxtun2         v22.16b, v2.8h
        st1             {v21.16b},    [x0], x2
        sqxtun          v23.8b,  v23.8h
        st1             {v22.16b},    [x0], x2
        sqxtun2         v23.16b, v3.8h
        st1             {v23.16b},    [x0], x2

        ret
endfunc

function ff_vp8_idct_dc_add_neon, export=1
        mov             w3,       #0
        ld1r            {v2.8h},  [x1]
        strh            w3,       [x1]
        srshr           v2.8h,  v2.8h,  #3
        ld1             {v0.s}[0],  [x0], x2
        ld1             {v0.s}[1],  [x0], x2
        uaddw           v3.8h,  v2.8h,  v0.8b
        ld1             {v1.s}[0],  [x0], x2
        ld1             {v1.s}[1],  [x0], x2
        uaddw           v4.8h,  v2.8h,  v1.8b
        sqxtun          v0.8b,  v3.8h
        sqxtun          v1.8b,  v4.8h
        sub             x0,  x0,  x2, lsl #2
        st1             {v0.s}[0],  [x0], x2
        st1             {v0.s}[1],  [x0], x2
        st1             {v1.s}[0],  [x0], x2
        st1             {v1.s}[1],  [x0], x2
        ret
endfunc

// Register layout:
//   P3..Q3 -> v0..v7
//   flim_E -> v22
//   flim_I -> v23
//   hev_thresh -> x5
//
.macro  vp8_loop_filter, inner=0, simple=0, hev_thresh
    .if \simple
        uabd            v17.16b, v3.16b,  v4.16b      // abs(P0-Q0)
        uabd            v23.16b, v2.16b,  v5.16b      // abs(P1-Q1)
        uqadd           v17.16b, v17.16b, v17.16b     // abs(P0-Q0) * 2
        ushr            v18.16b, v23.16b, #1          // abs(P1-Q1) / 2
        uqadd           v19.16b, v17.16b,  v18.16b    // (abs(P0-Q0)*2) + (abs(P1-Q1)/2)
        movi            v21.16b, #0x80
        cmhs            v16.16b, v22.16b, v19.16b    // (abs(P0-Q0)*2) + (abs(P1-Q1)/2) <= flim
    .else
        // calculate hev and normal_limit:
        uabd            v20.16b, v2.16b,  v3.16b      // abs(P1-P0)
        uabd            v21.16b, v5.16b,  v4.16b      // abs(Q1-Q0)
        uabd            v18.16b, v0.16b,  v1.16b      // abs(P3-P2)
        uabd            v19.16b, v1.16b,  v2.16b      // abs(P2-P1)
        cmhs            v16.16b, v23.16b, v20.16b     // abs(P1-P0) <= flim_I
        cmhs            v17.16b, v23.16b, v21.16b     // abs(Q1-Q0) <= flim_I
        cmhs            v18.16b, v23.16b, v18.16b     // abs(P3-P2) <= flim_I
        cmhs            v19.16b, v23.16b, v19.16b     // abs(P2-P1) <= flim_I
        and             v16.16b, v17.16b, v16.16b
        uabd            v17.16b, v7.16b,  v6.16b      // abs(Q3-Q2)
        and             v16.16b, v16.16b, v19.16b
        uabd            v19.16b, v6.16b,  v5.16b      // abs(Q2-Q1)
        and             v16.16b, v16.16b, v18.16b
        cmhs            v18.16b, v23.16b, v17.16b     // abs(Q3-Q2) <= flim_I
        cmhs            v19.16b, v23.16b, v19.16b     // abs(Q2-Q1) <= flim_I
        uabd            v17.16b, v3.16b,  v4.16b      // abs(P0-Q0)
        uabd            v23.16b, v2.16b,  v5.16b      // abs(P1-Q1)
        and             v16.16b, v16.16b, v18.16b
        uqadd           v17.16b, v17.16b, v17.16b     // abs(P0-Q0) * 2
        and             v16.16b, v16.16b, v19.16b
        ushr            v18.16b, v23.16b, #1          // abs(P1-Q1) / 2
        dup             v23.16b, \hev_thresh          // hev_thresh
        uqadd           v19.16b, v17.16b, v18.16b     // (abs(P0-Q0)*2) + (abs(P1-Q1)/2)
        cmhi            v20.16b, v20.16b, v23.16b     // abs(P1-P0) > hev_thresh
        cmhs            v19.16b, v22.16b, v19.16b     // (abs(P0-Q0)*2) + (abs(P1-Q1)/2) <= flim_E
        cmhi            v22.16b, v21.16b, v23.16b     // abs(Q1-Q0) > hev_thresh
        and             v16.16b, v16.16b, v19.16b
        movi            v21.16b, #0x80
        orr             v17.16b, v20.16b, v22.16b
    .endif

        // at this point:
        //   v16: normal_limit
        //   v17: hev

        // convert to signed value:
        eor            v3.16b, v3.16b, v21.16b           // PS0 = P0 ^ 0x80
        eor            v4.16b, v4.16b, v21.16b           // QS0 = Q0 ^ 0x80

        movi           v20.8h, #3
        ssubl          v18.8h, v4.8b,  v3.8b             // QS0 - PS0
        ssubl2         v19.8h, v4.16b, v3.16b            //   (widened to 16bit)
        eor            v2.16b, v2.16b, v21.16b           // PS1 = P1 ^ 0x80
        eor            v5.16b, v5.16b, v21.16b           // QS1 = Q1 ^ 0x80
        mul            v18.8h, v18.8h, v20.8h            // w = 3 * (QS0 - PS0)
        mul            v19.8h, v19.8h, v20.8h

        sqsub          v20.16b, v2.16b, v5.16b           // clamp(PS1-QS1)
        movi           v22.16b, #4
        movi           v23.16b, #3
    .if \inner
        and            v20.16b, v20.16b, v17.16b         // if(hev) w += clamp(PS1-QS1)
    .endif
        saddw          v18.8h,  v18.8h, v20.8b           // w += clamp(PS1-QS1)
        saddw2         v19.8h,  v19.8h, v20.16b
        sqxtn          v18.8b,  v18.8h                   // narrow result back into v18
        sqxtn2         v18.16b, v19.8h
    .if !\inner && !\simple
        eor            v1.16b,  v1.16b,  v21.16b         // PS2 = P2 ^ 0x80
        eor            v6.16b,  v6.16b,  v21.16b         // QS2 = Q2 ^ 0x80
    .endif
        and            v18.16b, v18.16b, v16.16b         // w &= normal_limit

        // registers used at this point..
        //   v0 -> P3  (don't corrupt)
        //   v1-v6 -> PS2-QS2
        //   v7 -> Q3  (don't corrupt)
        //   v17 -> hev
        //   v18 -> w
        //   v21 -> #0x80
        //   v22 -> #4
        //   v23 -> #3
        //   v16, v19, v29 -> unused
        //
        // filter_common:   is4tap==1
        //   c1 = clamp(w + 4) >> 3;
        //   c2 = clamp(w + 3) >> 3;
        //   Q0 = s2u(QS0 - c1);
        //   P0 = s2u(PS0 + c2);

    .if \simple
        sqadd          v19.16b, v18.16b, v22.16b           // c1 = clamp((w&hev)+4)
        sqadd          v20.16b, v18.16b, v23.16b           // c2 = clamp((w&hev)+3)
        sshr           v19.16b, v19.16b, #3                // c1 >>= 3
        sshr           v20.16b, v20.16b, #3                // c2 >>= 3
        sqsub          v4.16b,  v4.16b,  v19.16b           // QS0 = clamp(QS0-c1)
        sqadd          v3.16b,  v3.16b,  v20.16b           // PS0 = clamp(PS0+c2)
        eor            v4.16b,  v4.16b,  v21.16b           // Q0 = QS0 ^ 0x80
        eor            v3.16b,  v3.16b,  v21.16b           // P0 = PS0 ^ 0x80
        eor            v5.16b,  v5.16b,  v21.16b           // Q1 = QS1 ^ 0x80
        eor            v2.16b,  v2.16b,  v21.16b           // P1 = PS1 ^ 0x80
    .elseif \inner
        // the !is4tap case of filter_common, only used for inner blocks
        //   c3 = ((c1&~hev) + 1) >> 1;
        //   Q1 = s2u(QS1 - c3);
        //   P1 = s2u(PS1 + c3);
        sqadd          v19.16b, v18.16b, v22.16b           // c1 = clamp((w&hev)+4)
        sqadd          v20.16b, v18.16b, v23.16b           // c2 = clamp((w&hev)+3)
        sshr           v19.16b, v19.16b, #3                // c1 >>= 3
        sshr           v20.16b, v20.16b, #3                // c2 >>= 3
        sqsub          v4.16b,  v4.16b,  v19.16b           // QS0 = clamp(QS0-c1)
        sqadd          v3.16b,  v3.16b,  v20.16b           // PS0 = clamp(PS0+c2)
        bic            v19.16b, v19.16b, v17.16b           // c1 & ~hev
        eor            v4.16b,  v4.16b,  v21.16b           // Q0 = QS0 ^ 0x80
        srshr          v19.16b, v19.16b, #1                // c3 >>= 1
        eor            v3.16b,  v3.16b,  v21.16b           // P0 = PS0 ^ 0x80
        sqsub          v5.16b,  v5.16b,  v19.16b           // QS1 = clamp(QS1-c3)
        sqadd          v2.16b,  v2.16b,  v19.16b           // PS1 = clamp(PS1+c3)
        eor            v5.16b,  v5.16b,  v21.16b           // Q1 = QS1 ^ 0x80
        eor            v2.16b,  v2.16b,  v21.16b           // P1 = PS1 ^ 0x80
    .else
        and            v20.16b, v18.16b, v17.16b           // w & hev
        sqadd          v19.16b, v20.16b, v22.16b           // c1 = clamp((w&hev)+4)
        sqadd          v20.16b, v20.16b, v23.16b           // c2 = clamp((w&hev)+3)
        sshr           v19.16b, v19.16b, #3                // c1 >>= 3
        sshr           v20.16b, v20.16b, #3                // c2 >>= 3
        bic            v18.16b, v18.16b, v17.16b           // w &= ~hev
        sqsub          v4.16b,  v4.16b,  v19.16b           // QS0 = clamp(QS0-c1)
        sqadd          v3.16b,  v3.16b,  v20.16b           // PS0 = clamp(PS0+c2)

        // filter_mbedge:
        //   a = clamp((27*w + 63) >> 7);
        //   Q0 = s2u(QS0 - a);
        //   P0 = s2u(PS0 + a);
        //   a = clamp((18*w + 63) >> 7);
        //   Q1 = s2u(QS1 - a);
        //   P1 = s2u(PS1 + a);
        //   a = clamp((9*w + 63) >> 7);
        //   Q2 = s2u(QS2 - a);
        //   P2 = s2u(PS2 + a);
        movi           v17.8h,  #63
        sshll          v22.8h,  v18.8b, #3
        sshll2         v23.8h,  v18.16b, #3
        saddw          v22.8h,  v22.8h, v18.8b
        saddw2         v23.8h,  v23.8h, v18.16b
        add            v16.8h,  v17.8h, v22.8h
        add            v17.8h,  v17.8h, v23.8h           //  9*w + 63
        add            v19.8h,  v16.8h, v22.8h
        add            v20.8h,  v17.8h, v23.8h           // 18*w + 63
        add            v22.8h,  v19.8h, v22.8h
        add            v23.8h,  v20.8h, v23.8h           // 27*w + 63
        sqshrn         v16.8b,  v16.8h,  #7
        sqshrn2        v16.16b, v17.8h, #7              // clamp(( 9*w + 63)>>7)
        sqshrn         v19.8b,  v19.8h, #7
        sqshrn2        v19.16b, v20.8h, #7              // clamp((18*w + 63)>>7)
        sqshrn         v22.8b,  v22.8h, #7
        sqshrn2        v22.16b, v23.8h, #7              // clamp((27*w + 63)>>7)
        sqadd          v1.16b,  v1.16b,  v16.16b        // PS2 = clamp(PS2+a)
        sqsub          v6.16b,  v6.16b,  v16.16b        // QS2 = clamp(QS2-a)
        sqadd          v2.16b,  v2.16b,  v19.16b        // PS1 = clamp(PS1+a)
        sqsub          v5.16b,  v5.16b,  v19.16b        // QS1 = clamp(QS1-a)
        sqadd          v3.16b,  v3.16b,  v22.16b        // PS0 = clamp(PS0+a)
        sqsub          v4.16b,  v4.16b,  v22.16b        // QS0 = clamp(QS0-a)
        eor            v3.16b,  v3.16b,  v21.16b        // P0 = PS0 ^ 0x80
        eor            v4.16b,  v4.16b,  v21.16b        // Q0 = QS0 ^ 0x80
        eor            v2.16b,  v2.16b,  v21.16b        // P1 = PS1 ^ 0x80
        eor            v5.16b,  v5.16b,  v21.16b        // Q1 = QS1 ^ 0x80
        eor            v1.16b,  v1.16b,  v21.16b        // P2 = PS2 ^ 0x80
        eor            v6.16b,  v6.16b,  v21.16b        // Q2 = QS2 ^ 0x80
    .endif
.endm

.macro  vp8_v_loop_filter16 name, inner=0, simple=0
function ff_vp8_v_loop_filter16\name\()_neon, export=1
        sub             x0,  x0,  x1,  lsl #1+!\simple

        // Load pixels:
    .if !\simple
        ld1             {v0.16b},     [x0], x1 // P3
        ld1             {v1.16b},     [x0], x1 // P2
    .endif
        ld1             {v2.16b},     [x0], x1 // P1
        ld1             {v3.16b},     [x0], x1 // P0
        ld1             {v4.16b},     [x0], x1 // Q0
        ld1             {v5.16b},     [x0], x1 // Q1
    .if !\simple
        ld1             {v6.16b},     [x0], x1 // Q2
        ld1             {v7.16b},     [x0]     // Q3
        dup             v23.16b, w3                 // flim_I
    .endif
        dup             v22.16b, w2                 // flim_E

        vp8_loop_filter inner=\inner, simple=\simple, hev_thresh=w4

        // back up to P2:  dst -= stride * 6
        sub             x0,  x0,  x1,  lsl #2
    .if !\simple
        sub             x0,  x0,  x1,  lsl #1

        // Store pixels:
        st1             {v1.16b},     [x0], x1 // P2
    .endif
        st1             {v2.16b},     [x0], x1 // P1
        st1             {v3.16b},     [x0], x1 // P0
        st1             {v4.16b},     [x0], x1 // Q0
        st1             {v5.16b},     [x0], x1 // Q1
    .if !\simple
        st1             {v6.16b},     [x0]     // Q2
    .endif

        ret
endfunc
.endm

vp8_v_loop_filter16
vp8_v_loop_filter16 _inner,  inner=1
vp8_v_loop_filter16 _simple, simple=1

.macro  vp8_v_loop_filter8uv name, inner=0
function ff_vp8_v_loop_filter8uv\name\()_neon, export=1
        sub             x0,  x0,  x2,  lsl #2
        sub             x1,  x1,  x2,  lsl #2
        // Load pixels:
        ld1          {v0.d}[0],     [x0], x2  // P3
        ld1          {v0.d}[1],     [x1], x2  // P3
        ld1          {v1.d}[0],     [x0], x2  // P2
        ld1          {v1.d}[1],     [x1], x2  // P2
        ld1          {v2.d}[0],     [x0], x2  // P1
        ld1          {v2.d}[1],     [x1], x2  // P1
        ld1          {v3.d}[0],     [x0], x2  // P0
        ld1          {v3.d}[1],     [x1], x2  // P0
        ld1          {v4.d}[0],     [x0], x2  // Q0
        ld1          {v4.d}[1],     [x1], x2  // Q0
        ld1          {v5.d}[0],     [x0], x2  // Q1
        ld1          {v5.d}[1],     [x1], x2  // Q1
        ld1          {v6.d}[0],     [x0], x2  // Q2
        ld1          {v6.d}[1],     [x1], x2  // Q2
        ld1          {v7.d}[0],     [x0]      // Q3
        ld1          {v7.d}[1],     [x1]      // Q3

        dup          v22.16b, w3                 // flim_E
        dup          v23.16b, w4                 // flim_I

        vp8_loop_filter inner=\inner, hev_thresh=w5

        // back up to P2:  u,v -= stride * 6
        sub          x0,  x0,  x2,  lsl #2
        sub          x1,  x1,  x2,  lsl #2
        sub          x0,  x0,  x2,  lsl #1
        sub          x1,  x1,  x2,  lsl #1

        // Store pixels:

        st1          {v1.d}[0],     [x0], x2  // P2
        st1          {v1.d}[1],     [x1], x2  // P2
        st1          {v2.d}[0],     [x0], x2  // P1
        st1          {v2.d}[1],     [x1], x2  // P1
        st1          {v3.d}[0],     [x0], x2  // P0
        st1          {v3.d}[1],     [x1], x2  // P0
        st1          {v4.d}[0],     [x0], x2  // Q0
        st1          {v4.d}[1],     [x1], x2  // Q0
        st1          {v5.d}[0],     [x0], x2  // Q1
        st1          {v5.d}[1],     [x1], x2  // Q1
        st1          {v6.d}[0],     [x0]      // Q2
        st1          {v6.d}[1],     [x1]      // Q2

        ret
endfunc
.endm

vp8_v_loop_filter8uv
vp8_v_loop_filter8uv _inner, inner=1

.macro  vp8_h_loop_filter16 name, inner=0, simple=0
function ff_vp8_h_loop_filter16\name\()_neon, export=1

        sub             x0,  x0,  #4
        // Load pixels:
        ld1             {v0.d}[0], [x0], x1
        ld1             {v1.d}[0], [x0], x1
        ld1             {v2.d}[0], [x0], x1
        ld1             {v3.d}[0], [x0], x1
        ld1             {v4.d}[0], [x0], x1
        ld1             {v5.d}[0], [x0], x1
        ld1             {v6.d}[0], [x0], x1
        ld1             {v7.d}[0], [x0], x1
        ld1             {v0.d}[1], [x0], x1
        ld1             {v1.d}[1], [x0], x1
        ld1             {v2.d}[1], [x0], x1
        ld1             {v3.d}[1], [x0], x1
        ld1             {v4.d}[1], [x0], x1
        ld1             {v5.d}[1], [x0], x1
        ld1             {v6.d}[1], [x0], x1
        ld1             {v7.d}[1], [x0], x1

        transpose_8x16b   v0,  v1,  v2,  v3,  v4,  v5,  v6,  v7, v30, v31

        dup             v22.16b, w2                 // flim_E
    .if !\simple
        dup             v23.16b, w3                 // flim_I
    .endif

        vp8_loop_filter inner=\inner, simple=\simple, hev_thresh=w4

        sub             x0,  x0,  x1, lsl #4    // backup 16 rows

        transpose_8x16b   v0,  v1,  v2,  v3,  v4,  v5,  v6,  v7, v30, v31

        // Store pixels:
        st1             {v0.d}[0], [x0], x1
        st1             {v1.d}[0], [x0], x1
        st1             {v2.d}[0], [x0], x1
        st1             {v3.d}[0], [x0], x1
        st1             {v4.d}[0], [x0], x1
        st1             {v5.d}[0], [x0], x1
        st1             {v6.d}[0], [x0], x1
        st1             {v7.d}[0], [x0], x1
        st1             {v0.d}[1], [x0], x1
        st1             {v1.d}[1], [x0], x1
        st1             {v2.d}[1], [x0], x1
        st1             {v3.d}[1], [x0], x1
        st1             {v4.d}[1], [x0], x1
        st1             {v5.d}[1], [x0], x1
        st1             {v6.d}[1], [x0], x1
        st1             {v7.d}[1], [x0]

        ret
endfunc
.endm

vp8_h_loop_filter16
vp8_h_loop_filter16 _inner,  inner=1
vp8_h_loop_filter16 _simple, simple=1

.macro  vp8_h_loop_filter8uv name, inner=0
function ff_vp8_h_loop_filter8uv\name\()_neon, export=1
        sub             x0,  x0,  #4
        sub             x1,  x1,  #4

        // Load pixels:
        ld1          {v0.d}[0],     [x0], x2 // load u
        ld1          {v0.d}[1],     [x1], x2 // load v
        ld1          {v1.d}[0],     [x0], x2
        ld1          {v1.d}[1],     [x1], x2
        ld1          {v2.d}[0],     [x0], x2
        ld1          {v2.d}[1],     [x1], x2
        ld1          {v3.d}[0],     [x0], x2
        ld1          {v3.d}[1],     [x1], x2
        ld1          {v4.d}[0],     [x0], x2
        ld1          {v4.d}[1],     [x1], x2
        ld1          {v5.d}[0],     [x0], x2
        ld1          {v5.d}[1],     [x1], x2
        ld1          {v6.d}[0],     [x0], x2
        ld1          {v6.d}[1],     [x1], x2
        ld1          {v7.d}[0],     [x0], x2
        ld1          {v7.d}[1],     [x1], x2

        transpose_8x16b   v0,  v1,  v2,  v3,  v4,  v5,  v6,  v7, v30, v31

        dup             v22.16b, w3                 // flim_E
        dup             v23.16b, w4                 // flim_I

        vp8_loop_filter inner=\inner, hev_thresh=w5

        sub             x0,  x0,  x2, lsl #3    // backup u 8 rows
        sub             x1,  x1,  x2, lsl #3    // backup v 8 rows

        transpose_8x16b   v0,  v1,  v2,  v3,  v4,  v5,  v6,  v7, v30, v31

        // Store pixels:
        st1          {v0.d}[0],     [x0], x2 // load u
        st1          {v0.d}[1],     [x1], x2 // load v
        st1          {v1.d}[0],     [x0], x2
        st1          {v1.d}[1],     [x1], x2
        st1          {v2.d}[0],     [x0], x2
        st1          {v2.d}[1],     [x1], x2
        st1          {v3.d}[0],     [x0], x2
        st1          {v3.d}[1],     [x1], x2
        st1          {v4.d}[0],     [x0], x2
        st1          {v4.d}[1],     [x1], x2
        st1          {v5.d}[0],     [x0], x2
        st1          {v5.d}[1],     [x1], x2
        st1          {v6.d}[0],     [x0], x2
        st1          {v6.d}[1],     [x1], x2
        st1          {v7.d}[0],     [x0]
        st1          {v7.d}[1],     [x1]

        ret

endfunc
.endm

vp8_h_loop_filter8uv
vp8_h_loop_filter8uv _inner, inner=1


function ff_put_vp8_pixels16_neon, export=1
1:
        subs            w4, w4, #4
        ld1             {v0.16b},     [x2], x3
        ld1             {v1.16b},     [x2], x3
        ld1             {v2.16b},     [x2], x3
        ld1             {v3.16b},     [x2], x3
        st1             {v0.16b},     [x0], x1
        st1             {v1.16b},     [x0], x1
        st1             {v2.16b},     [x0], x1
        st1             {v3.16b},     [x0], x1
        bgt             1b
        ret
endfunc

function ff_put_vp8_pixels8_neon, export=1
1:
        subs            w4, w4, #4
        ld1             {v0.8b},   [x2], x3
        ld1             {v0.d}[1], [x2], x3
        ld1             {v1.8b},   [x2], x3
        ld1             {v1.d}[1], [x2], x3
        st1             {v0.8b},   [x0], x1
        st1             {v0.d}[1], [x0], x1
        st1             {v1.8b},   [x0], x1
        st1             {v1.d}[1], [x0], x1
        bgt             1b
        ret
endfunc

/* 4/6-tap 8th-pel MC */

.macro  vp8_epel8_h6    d,   s0,   s1
        ext             v22.8b, \s0\().8b,  \s1\().8b,  #1
        uxtl            v18.8h, \s0\().8b
        ext             v23.8b, \s0\().8b,  \s1\().8b,  #2
        uxtl            v19.8h, v22.8b
        ext             v24.8b, \s0\().8b,  \s1\().8b,  #3
        uxtl            v21.8h, v23.8b
        ext             v25.8b, \s0\().8b,  \s1\().8b,  #4
        uxtl            v22.8h, v24.8b
        ext             v26.8b, \s0\().8b,  \s1\().8b,  #5
        uxtl            v25.8h, v25.8b
        mul             v21.8h, v21.8h, v0.8h[2]
        uxtl            v26.8h, v26.8b
        mul             v22.8h, v22.8h, v0.8h[3]
        mls             v21.8h, v19.8h, v0.8h[1]
        mls             v22.8h, v25.8h, v0.8h[4]
        mla             v21.8h, v18.8h, v0.8h[0]
        mla             v22.8h, v26.8h, v0.8h[5]
        sqadd           v22.8h, v21.8h, v22.8h
        sqrshrun        \d\().8b, v22.8h, #7
.endm

.macro  vp8_epel16_h6   d0,  v0,  v1
        ext             v22.16b, \v0\().16b, \v1\().16b, #3
        ext             v23.16b, \v0\().16b, \v1\().16b, #4
        uxtl            v19.8h,  v22.8b
        uxtl2           v22.8h,  v22.16b
        ext             v3.16b,  \v0\().16b, \v1\().16b, #2
        uxtl            v20.8h,  v23.8b
        uxtl2           v23.8h,  v23.16b
        ext             v16.16b, \v0\().16b, \v1\().16b, #1
        uxtl            v18.8h,  v3.8b
        uxtl2           v3.8h,   v3.16b
        ext             v2.16b,  \v0\().16b, \v1\().16b, #5
        uxtl            v21.8h,  v2.8b
        uxtl2           v2.8h,   v2.16b
        uxtl            v17.8h,  v16.8b
        uxtl2           v16.8h,  v16.16b
        mul             v19.8h,  v19.8h, v0.8h[3]
        mul             v18.8h,  v18.8h, v0.8h[2]
        mul             v3.8h,   v3.8h,  v0.8h[2]
        mul             v22.8h,  v22.8h, v0.8h[3]
        mls             v19.8h,  v20.8h, v0.8h[4]
        uxtl            v20.8h,  \v0\().8b
        uxtl2           v1.8h,   \v0\().16b
        mls             v18.8h,  v17.8h, v0.8h[1]
        mls             v3.8h,   v16.8h, v0.8h[1]
        mls             v22.8h,  v23.8h, v0.8h[4]
        mla             v18.8h,  v20.8h, v0.8h[0]
        mla             v19.8h,  v21.8h, v0.8h[5]
        mla             v3.8h,   v1.8h,  v0.8h[0]
        mla             v22.8h,  v2.8h,  v0.8h[5]
        sqadd           v19.8h,  v18.8h, v19.8h
        sqadd           v22.8h,  v3.8h,  v22.8h
        sqrshrun        \d0\().8b,  v19.8h, #7
        sqrshrun2       \d0\().16b, v22.8h, #7
.endm

.macro  vp8_epel8_v6    d0,  s0,  s1,  s2, s3, s4, s5
        uxtl            \s2\().8h, \s2\().8b
        uxtl            \s3\().8h, \s3\().8b
        uxtl            \s1\().8h, \s1\().8b
        uxtl            \s4\().8h, \s4\().8b
        uxtl            \s0\().8h, \s0\().8b
        uxtl            \s5\().8h, \s5\().8b
        mul             \s2\().8h, \s2\().8h, v0.8h[2]
        mul             \s3\().8h, \s3\().8h, v0.8h[3]
        mls             \s2\().8h, \s1\().8h, v0.8h[1]
        mls             \s3\().8h, \s4\().8h, v0.8h[4]
        mla             \s2\().8h, \s0\().8h, v0.8h[0]
        mla             \s3\().8h, \s5\().8h, v0.8h[5]
        sqadd           \s3\().8h, \s2\().8h, \s3\().8h
        sqrshrun        \d0\().8b, \s3\().8h, #7
.endm

.macro  vp8_epel8_v6_y2 d0, d1, s0, s1, s2, s3, s4, s5, s6
        uxtl            \s0\().8h, \s0\().8b
        uxtl            \s3\().8h, \s3\().8b
        uxtl            \s6\().8h, \s6\().8b
        uxtl            \s1\().8h, \s1\().8b
        uxtl            \s4\().8h, \s4\().8b
        uxtl            \s2\().8h, \s2\().8b
        uxtl            \s5\().8h, \s5\().8b
        mul             \s0\().8h, \s0\().8h, v0.8h[0]
        mul             v31.8h   , \s3\().8h, v0.8h[3]
        mul             \s3\().8h, \s3\().8h, v0.8h[2]
        mul             \s6\().8h, \s6\().8h, v0.8h[5]

        mls             \s0\().8h, \s1\().8h, v0.8h[1]
        mls             v31.8h   , \s4\().8h, v0.8h[4]
        mls             \s3\().8h, \s2\().8h, v0.8h[1]
        mls             \s6\().8h, \s5\().8h, v0.8h[4]

        mla             \s0\().8h, \s2\().8h, v0.8h[2]
        mla             v31.8h   , \s5\().8h, v0.8h[5]
        mla             \s3\().8h, \s1\().8h, v0.8h[0]
        mla             \s6\().8h, \s4\().8h, v0.8h[3]
        sqadd           v31.8h   , \s0\().8h, v31.8h
        sqadd           \s6\().8h, \s3\().8h, \s6\().8h
        sqrshrun        \d0\().8b, v31.8h,    #7
        sqrshrun        \d1\().8b, \s6\().8h, #7
.endm

.macro  vp8_epel8_h4    d,   v0,   v1
        ext             v22.8b, \v0\().8b,  \v1\().8b,  #1
        uxtl            v19.8h, \v0\().8b
        ext             v23.8b, \v0\().8b,  \v1\().8b,  #2
        uxtl            v20.8h, v22.8b
        ext             v25.8b, \v0\().8b,  \v1\().8b,  #3
        uxtl            v22.8h, v23.8b
        uxtl            v25.8h, v25.8b
        mul             v20.8h, v20.8h, v0.8h[2]
        mul             v22.8h, v22.8h, v0.8h[3]
        mls             v20.8h, v19.8h, v0.8h[1]
        mls             v22.8h, v25.8h, v0.8h[4]
        sqadd           v22.8h, v20.8h, v22.8h
        sqrshrun        \d\().8b, v22.8h, #7
.endm

.macro  vp8_epel8_v4_y2 d0, s0, s1, s2, s3, s4
        uxtl            \s0\().8h,  \s0\().8b
        uxtl            \s1\().8h,  \s1\().8b
        uxtl            \s2\().8h,  \s2\().8b
        uxtl            \s3\().8h,  \s3\().8b
        uxtl            \s4\().8h,  \s4\().8b
        mul             v21.8h,     \s1\().8h, v0.8h[2]
        mul             v23.8h,     \s2\().8h, v0.8h[3]
        mul             \s2\().8h,  \s2\().8h, v0.8h[2]
        mul             v22.8h,     \s3\().8h, v0.8h[3]
        mls             v21.8h,     \s0\().8h, v0.8h[1]
        mls             v23.8h,     \s3\().8h, v0.8h[4]
        mls             \s2\().8h,  \s1\().8h, v0.8h[1]
        mls             v22.8h,     \s4\().8h, v0.8h[4]
        sqadd           v21.8h,     v21.8h,    v23.8h
        sqadd           \s2\().8h,  \s2\().8h, v22.8h
        sqrshrun        \d0\().8b,  v21.8h,    #7
        sqrshrun2       \d0\().16b, \s2\().8h, #7
.endm


// note: worst case sum of all 6-tap filter values * 255 is 0x7f80 so 16 bit
// arithmatic can be used to apply filters
const   subpel_filters, align=4
        .short     0,   6, 123,  12,   1,   0,   0,   0
        .short     2,  11, 108,  36,   8,   1,   0,   0
        .short     0,   9,  93,  50,   6,   0,   0,   0
        .short     3,  16,  77,  77,  16,   3,   0,   0
        .short     0,   6,  50,  93,   9,   0,   0,   0
        .short     1,   8,  36, 108,  11,   2,   0,   0
        .short     0,   1,  12, 123,   6,   0,   0,   0
endconst

function ff_put_vp8_epel16_v6_neon, export=1
        sub             x2,  x2,  x3,  lsl #1

        sxtw            x4,  w4
        sxtw            x6,  w6
        movrel          x17,  subpel_filters-16
        add             x6,  x17,  x6, lsl #4  // y
        ld1             {v0.8h},     [x6]
1:
        ld1             {v1.1d - v2.1d},    [x2], x3
        ld1             {v3.1d - v4.1d},    [x2], x3
        ld1             {v16.1d - v17.1d},  [x2], x3
        ld1             {v18.1d - v19.1d},  [x2], x3
        ld1             {v20.1d - v21.1d},  [x2], x3
        ld1             {v22.1d - v23.1d},  [x2], x3
        ld1             {v24.1d - v25.1d},  [x2]
        sub             x2,  x2,  x3, lsl #2

        vp8_epel8_v6_y2 v1, v3, v1, v3, v16, v18, v20, v22, v24
        vp8_epel8_v6_y2 v2, v4, v2, v4, v17, v19, v21, v23, v25

        st1             {v1.1d - v2.1d}, [x0], x1
        st1             {v3.1d - v4.1d}, [x0], x1
        subs            x4, x4, #2
        bne             1b

        ret
endfunc

function ff_put_vp8_epel16_h6_neon, export=1
        sub             x2,  x2,  #2
        sxtw            x5,  w5 // x

        // first pass (horizontal):
        movrel          x17,  subpel_filters-16
        add             x5,  x17,  x5, lsl #4 // x
        ld1             {v0.8h},  [x5]
1:
        ld1             {v1.16b, v2.16b}, [x2], x3
        vp8_epel16_h6   v1, v1, v2
        st1             {v1.16b}, [x0], x1

        subs            w4, w4, #1
        bne             1b
        ret
endfunc


function ff_put_vp8_epel16_h6v6_neon, export=1
        sub             x2,  x2,  x3,  lsl #1
        sub             x2,  x2,  #2

        // first pass (horizontal):
        movrel          x17,  subpel_filters-16
        sxtw            x5,  w5 // x
        add             x16,  x17,  x5, lsl #4 // x
        sub             sp,  sp,  #336+16
        ld1             {v0.8h},  [x16]
        add             x7,  sp,  #15
        sxtw            x4,  w4
        add             x16, x4, #5   // h
        bic             x7,  x7,  #15
1:
        ld1             {v1.16b, v2.16b}, [x2], x3
        vp8_epel16_h6   v1, v1, v2
        st1             {v1.16b}, [x7], #16
        subs            x16, x16, #1
        bne             1b


        // second pass (vertical):
        sxtw            x6,  w6
        add             x6,  x17,  x6, lsl #4  // y
        add             x7,  sp,  #15
        ld1             {v0.8h},     [x6]
        bic             x7,  x7,  #15
2:
        ld1             {v1.8b - v4.8b},    [x7], #32
        ld1             {v16.8b - v19.8b},  [x7], #32
        ld1             {v20.8b - v23.8b},  [x7]
        sub             x7,  x7,  #48

        vp8_epel8_v6    v5, v1, v3, v16, v18, v20, v22
        vp8_epel8_v6    v2, v2, v4, v17, v19, v21, v23
        trn1            v2.2d, v5.2d, v2.2d

        st1             {v2.16b}, [x0], x1
        subs            x4, x4, #1
        bne             2b

        add             sp,  sp,  #336+16
        ret
endfunc

function ff_put_vp8_epel8_h6v6_neon, export=1
        sub             x2,  x2,  x3,  lsl #1
        sub             x2,  x2,  #2
        sxtw            x4,  w4

        // first pass (horizontal):
        movrel          x17,  subpel_filters-16
        sxtw            x5,  w5
        add             x5,  x17,  x5, lsl #4 // x
        sub             sp,  sp,  #168+16
        ld1             {v0.8h},  [x5]
        add             x7,  sp,  #15
        add             x16, x4,  #5   // h
        bic             x7,  x7,  #15
1:
        ld1             {v1.8b, v2.8b}, [x2], x3

        vp8_epel8_h6    v1, v1, v2

        st1             {v1.8b}, [x7], #8
        subs            x16, x16, #1
        bne             1b

        // second pass (vertical):
        sxtw            x6,  w6
        add             x6,  x17,  x6, lsl #4  // y
        add             x7,  sp,   #15
        ld1             {v0.8h},   [x6]
        bic             x7,  x7,   #15
2:
        ld1             {v1.8b - v4.8b}, [x7], #32
        ld1             {v5.8b - v7.8b}, [x7]

        sub             x7,  x7,  #16

        vp8_epel8_v6_y2 v1, v2, v1, v2, v3, v4, v5, v6, v7

        st1             {v1.8b}, [x0], x1
        st1             {v2.8b}, [x0], x1
        subs            x4, x4, #2
        bne             2b

        add             sp,  sp,  #168+16
        ret
endfunc

function ff_put_vp8_epel8_h4v6_neon, export=1
        sub             x2,  x2,  x3,  lsl #1
        sub             x2,  x2,  #1
        sxtw            x4,  w4

        // first pass (horizontal):
        movrel          x17,  subpel_filters-16
        sxtw            x5,  w5
        add             x5,  x17,  x5, lsl #4 // x
        sub             sp,  sp,  #168+16
        ld1             {v0.8h},  [x5]
        add             x7,  sp,  #15
        add             x16, x4, #5   // h
        bic             x7,  x7,  #15
1:
        ld1             {v1.8b, v2.8b}, [x2], x3

        vp8_epel8_h4    v1, v1, v2

        st1             {v1.8b}, [x7], #8
        subs            x16, x16, #1
        bne             1b

        // second pass (vertical):
        sxtw            x6,  w6
        add             x6,  x17,  x6, lsl #4  // y
        add             x7,  sp,   #15
        ld1             {v0.8h},   [x6]
        bic             x7,  x7,   #15
2:
        ld1             {v1.8b - v4.8b}, [x7], #32
        ld1             {v5.8b - v7.8b}, [x7]

        sub             x7,  x7,  #16

        vp8_epel8_v6_y2 v1, v2, v1, v2, v3, v4, v5, v6, v7

        st1             {v1.8b}, [x0], x1
        st1             {v2.8b}, [x0], x1
        subs            x4, x4, #2
        bne             2b

        add             sp,  sp,  #168+16
        ret
endfunc

function ff_put_vp8_epel8_h4v4_neon, export=1
        sub             x2,  x2,  x3
        sub             x2,  x2,  #1
        sxtw            x4,  w4


        // first pass (horizontal):
        movrel          x17,  subpel_filters-16
        sxtw            x5,  w5
        add             x5,  x17,  x5, lsl #4 // x
        sub             sp,  sp,  #168+16
        ld1             {v0.8h},  [x5]
        add             x7,  sp,  #15
        add             x16, x4, #3   // h
        bic             x7,  x7,  #15
1:
        ld1             {v1.8b, v2.8b}, [x2], x3

        vp8_epel8_h4    v1, v1, v2

        st1             {v1.8b}, [x7], #8
        subs            x16, x16, #1
        bne             1b

        // second pass (vertical):
        sxtw            x6,  w6
        add             x6,  x17,  x6, lsl #4  // y
        add             x7,  sp,   #15
        ld1             {v0.8h},   [x6]
        bic             x7,  x7,   #15
2:
        ld1             {v1.8b - v2.8b}, [x7], #16
        ld1             {v3.8b - v5.8b}, [x7]

        vp8_epel8_v4_y2 v1, v1, v2, v3, v4, v5

        st1             {v1.d}[0], [x0], x1
        st1             {v1.d}[1], [x0], x1
        subs            x4, x4, #2
        bne             2b

        add             sp,  sp,  #168+16
        ret
endfunc

function ff_put_vp8_epel8_h6v4_neon, export=1
        sub             x2,  x2,  x3
        sub             x2,  x2,  #2
        sxtw            x4,  w4


        // first pass (horizontal):
        movrel          x17,  subpel_filters-16
        sxtw            x5,  w5
        add             x5,  x17,  x5, lsl #4 // x
        sub             sp,  sp,  #168+16
        ld1             {v0.8h},  [x5]
        add             x7,  sp,  #15
        add             x16, x4, #3   // h
        bic             x7,  x7,  #15
1:
        ld1             {v1.8b, v2.8b}, [x2], x3

        vp8_epel8_h6    v1, v1, v2

        st1             {v1.8b}, [x7], #8
        subs            x16, x16, #1
        bne             1b

        // second pass (vertical):
        sxtw            x6,  w6
        add             x6,  x17,  x6, lsl #4  // y
        add             x7,  sp,   #15
        ld1             {v0.8h},   [x6]
        bic             x7,  x7,   #15
2:
        ld1             {v1.8b - v2.8b}, [x7], #16
        ld1             {v3.8b - v5.8b}, [x7]

        vp8_epel8_v4_y2 v1, v1, v2, v3, v4, v5

        st1             {v1.d}[0], [x0], x1
        st1             {v1.d}[1], [x0], x1
        subs            x4, x4, #2
        bne             2b

        add             sp,  sp,  #168+16
        ret
endfunc