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
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
|
@/*****************************************************************************
@*
@* Copyright (C) 2012 Ittiam Systems Pvt Ltd, Bangalore
@*
@* Licensed under the Apache License, Version 2.0 (the "License");
@* you may not use this file except in compliance with the License.
@* You may obtain a copy of the License at:
@*
@* http://www.apache.org/licenses/LICENSE-2.0
@*
@* Unless required by applicable law or agreed to in writing, software
@* distributed under the License is distributed on an "AS IS" BASIS,
@* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
@* See the License for the specific language governing permissions and
@* limitations under the License.
@*
@*****************************************************************************/
@/**
@*******************************************************************************
@* ,:file
@* ihevc_sao_edge_offset_class3_chroma.s
@*
@* ,:brief
@* Contains function definitions for inter prediction interpolation.
@* Functions are coded using NEON intrinsics and can be compiled using@ ARM
@* RVCT
@*
@* ,:author
@* Parthiban V
@*
@* ,:par List of Functions:
@*
@*
@* ,:remarks
@* None
@*
@*******************************************************************************
@*/
@void ihevc_sao_edge_offset_class3_chroma(UWORD8 *pu1_src,
@ WORD32 src_strd,
@ UWORD8 *pu1_src_left,
@ UWORD8 *pu1_src_top,
@ UWORD8 *pu1_src_top_left,
@ UWORD8 *pu1_src_top_right,
@ UWORD8 *pu1_src_bot_left,
@ UWORD8 *pu1_avail,
@ WORD8 *pi1_sao_offset_u,
@ WORD8 *pi1_sao_offset_v,
@ WORD32 wd,
@ WORD32 ht)
@**************Variables Vs Registers*****************************************
@r0 => *pu1_src
@r1 => src_strd
@r2 => *pu1_src_left
@r3 => *pu1_src_top
@r4 => *pu1_src_top_left
@r5 => *pu1_avail
@r6 => *pi1_sao_offset_u
@r9 => *pi1_sao_offset_v
@r7 => wd
@r8=> ht
.equ pu1_src_top_left_offset, 328
.equ pu1_src_top_right_offset, 332
.equ pu1_src_bot_left_offset, 336
.equ pu1_avail_offset, 340
.equ pi1_sao_u_offset, 344
.equ pi1_sao_v_offset, 348
.equ wd_offset, 352
.equ ht_offset, 356
.text
.syntax unified
.p2align 2
.extern gi1_table_edge_idx
.globl ihevc_sao_edge_offset_class3_chroma_a9q
gi1_table_edge_idx_addr_1:
.long gi1_table_edge_idx - ulbl1 - 8
gi1_table_edge_idx_addr_2:
.long gi1_table_edge_idx - ulbl2 - 8
gi1_table_edge_idx_addr_3:
.long gi1_table_edge_idx - ulbl3 - 8
gi1_table_edge_idx_addr_4:
.long gi1_table_edge_idx - ulbl4 - 8
gi1_table_edge_idx_addr_5:
.long gi1_table_edge_idx - ulbl5 - 8
ihevc_sao_edge_offset_class3_chroma_a9q:
STMFD sp!,{r4-r12,r14} @stack stores the values of the arguments
vpush {d8 - d15}
SUB sp,sp,#224 @Decrement the stack pointer to store some temp arr values
LDR r7,[sp,#wd_offset] @Loads wd
LDR r8,[sp,#ht_offset] @Loads ht
SUB r9,r7,#2 @wd - 2
LDR r4,[sp,#pu1_src_top_left_offset] @Loads pu1_src_top_left
LDRH r10,[r3,r9] @pu1_src_top[wd - 2]
MOV r9,r7 @Move width to r9 for loop count
LDR r5,[sp,#pu1_avail_offset] @Loads pu1_avail
LDR r6,[sp,#pi1_sao_u_offset] @Loads pi1_sao_offset_u
STR r3,[sp,#220] @Store pu1_src_top in sp
STRH r10,[sp] @u1_src_top_left_tmp = pu1_src_top[wd - 2]
SUB r10,r8,#1 @ht-1
MLA r11,r10,r1,r0 @pu1_src[(ht - 1) * src_strd + col]
ADD r12,sp,#10 @temp array
AU1_SRC_TOP_LOOP:
VLD1.8 D0,[r11]! @pu1_src[(ht - 1) * src_strd + col]
SUBS r9,r9,#8 @Decrement the loop count by 8
VST1.8 D0,[r12]! @au1_src_top_tmp[col] = pu1_src[(ht - 1) * src_strd + col]
BNE AU1_SRC_TOP_LOOP
PU1_AVAIL_5_LOOP_U:
LDRB r9,[r5,#5] @pu1_avail[5]
CMP r9,#0
SUB r14,r7,#2 @[wd - 2]
LDRB r9,[r0,r14] @u1_pos_0_0_tmp_u = pu1_src[wd - 2]
SUB r11,r7,#1 @[wd - 1]
LDRB r10,[r0,r11] @u1_pos_0_0_tmp_v = pu1_src[wd - 1]
BEQ PU1_AVAIL_6_LOOP_U
LDR r11,[sp,#pu1_src_top_right_offset] @Load pu1_src_top_right from sp
LDRB r11,[r11] @pu1_src_top_right[0]
SUB r12,r9,r11 @pu1_src[wd - 2] - pu1_src_top_right[0]
CMP r12,#0
MVNLT r12,#0
MOVGT r12,#1 @SIGN(pu1_src[wd - 2] - pu1_src_top_right[0])
ADD r11,r0,r1 @pu1_src + src_strd
SUB r14,r14,#2 @[wd - 2 - 2]
LDRB r14,[r11,r14] @pu1_src[wd - 2 - 2 + src_strd]
SUB r11,r9,r14 @pu1_src[wd - 2] - pu1_src[wd - 2 - 2 + src_strd]
CMP r11,#0
MVNLT r11,#0
MOVGT r11,#1 @SIGN(pu1_src[wd - 2] - pu1_src[wd - 2 - 2 + src_strd])
ADD r11,r12,r11 @SIGN(pu1_src[wd - 2] - pu1_src_top_right[0]) + SIGN(pu1_src[wd - 2] - pu1_src[wd - 2 - 2 + src_strd])
ADD r11,r11,#2 @edge_idx
LDR r14, gi1_table_edge_idx_addr_1 @table pointer
ulbl1:
add r14,r14,pc
LDRSB r12,[r14,r11] @edge_idx = gi1_table_edge_idx[edge_idx]
CMP r12,#0 @0 != edge_idx
BEQ PU1_AVAIL_5_LOOP_V
LDRSB r11,[r6,r12] @pi1_sao_offset_u[edge_idx]
ADD r9,r9,r11 @pu1_src[wd - 2] + pi1_sao_offset_u[edge_idx]
USAT r9,#8,r9 @u1_pos_0_0_tmp_u = CLIP3(pu1_src[wd - 2] + pi1_sao_offset[edge_idx], 0, (1 << bit_depth) - 1)
PU1_AVAIL_5_LOOP_V:
LDR r11,[sp,#pu1_src_top_right_offset] @Load pu1_src_top_right from sp
LDRB r11,[r11,#1] @pu1_src_top_right[1]
SUB r12,r10,r11 @pu1_src[wd - 1] - pu1_src_top_right[1]
CMP r12,#0
MVNLT r12,#0
MOVGT r12,#1 @SIGN(pu1_src[wd - 1] - pu1_src_top_right[1])
ADD r11,r0,r1 @pu1_src + src_strd
SUB r14,r7,#3 @[wd - 1 - 2]
LDRB r14,[r11,r14] @pu1_src[wd - 1 - 2 + src_strd]
SUB r11,r10,r14 @pu1_src[wd - 1] - pu1_src[wd - 1 - 2 + src_strd]
CMP r11,#0
MVNLT r11,#0
MOVGT r11,#1 @SIGN(pu1_src[wd - 1] - pu1_src[wd - 1 - 2 + src_strd])
ADD r11,r12,r11 @SIGN(pu1_src[wd - 1] - pu1_src_top_right[1]) + SIGN(pu1_src[wd - 1] - pu1_src[wd - 1 - 2 + src_strd])
ADD r11,r11,#2 @edge_idx
LDR r14, gi1_table_edge_idx_addr_2 @table pointer
ulbl2:
add r14,r14,pc
LDRSB r12,[r14,r11] @edge_idx = gi1_table_edge_idx[edge_idx]
CMP r12,#0 @0 != edge_idx
BEQ PU1_AVAIL_6_LOOP_U
LDR r11,[sp,#pi1_sao_v_offset] @Loads pi1_sao_offset_v
LDRSB r11,[r11,r12] @pi1_sao_offset_v[edge_idx]
ADD r10,r10,r11 @pu1_src[wd - 1] + pi1_sao_offset_v[edge_idx]
USAT r10,#8,r10 @u1_pos_0_0_tmp_v = CLIP3(pu1_src[wd - 1] + pi1_sao_offset_v[edge_idx], 0, (1 << bit_depth) - 1)
PU1_AVAIL_6_LOOP_U:
STRB r9,[sp,#6]
STRB r10,[sp,#7]
STR r0,[sp,#212] @Store pu1_src in sp
LDRB r10,[r5,#6] @pu1_avail[6]
CMP r10,#0
SUB r11,r8,#1 @ht - 1
MLA r12,r11,r1,r0 @pu1_src[(ht - 1) * src_strd]
LDRB r10,[r12] @u1_pos_wd_ht_tmp_u = pu1_src[(ht - 1) * src_strd]
LDRB r9,[r12,#1] @u1_pos_wd_ht_tmp_v = pu1_src[(ht - 1) * src_strd + 1]
BEQ PU1_AVAIL_3_LOOP
SUB r11,r12,r1 @pu1_src[(ht - 1) * src_strd - src_strd]
ADD r11,r11,#2 @pu1_src[(ht - 1) * src_strd + 2 - src_strd]
LDRB r11,[r11] @Load pu1_src[(ht - 1) * src_strd + 2 - src_strd]
SUB r11,r10,r11 @pu1_src[(ht - 1) * src_strd] - pu1_src[(ht - 1) * src_strd + 2 - src_strd]
CMP r11,#0
MVNLT r11,#0
MOVGT r11,#1 @SIGN(pu1_src[(ht - 1) * src_strd] - pu1_src[(ht - 1) * src_strd + 2 - src_strd])
LDR r14,[sp,#pu1_src_bot_left_offset] @Load pu1_src_bot_left from sp
LDRB r14,[r14] @Load pu1_src_bot_left[0]
SUB r14,r10,r14 @pu1_src[(ht - 1) * src_strd] - pu1_src_bot_left[0]
CMP r14,#0
MVNLT r14,#0
MOVGT r14,#1 @SIGN(pu1_src[(ht - 1) * src_strd] - pu1_src_bot_left[0])
ADD r11,r11,r14 @Add 2 sign value
ADD r11,r11,#2 @edge_idx
LDR r14, gi1_table_edge_idx_addr_3 @table pointer
ulbl3:
add r14,r14,pc
LDRSB r14,[r14,r11] @edge_idx = gi1_table_edge_idx[edge_idx]
CMP r14,#0
BEQ PU1_AVAIL_6_LOOP_V
LDRSB r11,[r6,r14] @pi1_sao_offset_u[edge_idx]
ADD r10,r10,r11 @pu1_src[(ht - 1) * src_strd] + pi1_sao_offset[edge_idx]
USAT r10,#8,r10 @u1_pos_wd_ht_tmp = CLIP3(pu1_src[(ht - 1) * src_strd] + pi1_sao_offset[edge_idx], 0, (1 << bit_depth) - 1)
PU1_AVAIL_6_LOOP_V:
ADD r12,r12,#1 @pu1_src[(ht - 1) * src_strd + 1]
SUB r11,r12,r1 @pu1_src[(ht - 1) * src_strd + 1) - src_strd]
ADD r11,r11,#2 @pu1_src[(ht - 1) * src_strd + 2 - src_strd]
LDRB r11,[r11] @Load pu1_src[(ht - 1) * src_strd + 2 - src_strd]
SUB r11,r9,r11 @pu1_src[(ht - 1) * src_strd + 1] - pu1_src[(ht - 1) * src_strd + 1 + 2 - src_strd]
CMP r11,#0
MVNLT r11,#0
MOVGT r11,#1 @SIGN(pu1_src[(ht - 1) * src_strd + 1] - pu1_src[(ht - 1) * src_strd + 1 + 2 - src_strd])
LDR r14,[sp,#pu1_src_bot_left_offset] @Load pu1_src_bot_left from sp
LDRB r14,[r14,#1] @Load pu1_src_bot_left[1]
SUB r14,r9,r14 @pu1_src[(ht - 1) * src_strd + 1] - pu1_src_bot_left[1]
CMP r14,#0
MVNLT r14,#0
MOVGT r14,#1 @SIGN(pu1_src[(ht - 1) * src_strd + 1] - pu1_src_bot_left[1])
ADD r11,r11,r14 @Add 2 sign value
ADD r11,r11,#2 @edge_idx
LDR r14, gi1_table_edge_idx_addr_4 @table pointer
ulbl4:
add r14,r14,pc
LDRSB r12,[r14,r11] @edge_idx = gi1_table_edge_idx[edge_idx]
CMP r12,#0
BEQ PU1_AVAIL_3_LOOP
LDR r14,[sp,#pi1_sao_v_offset] @Loads pi1_sao_offset_v
LDRSB r11,[r14,r12] @pi1_sao_offset_v[edge_idx]
ADD r9,r9,r11 @pu1_src[(ht - 1) * src_strd] + pi1_sao_offset[edge_idx]
USAT r9,#8,r9 @u1_pos_wd_ht_tmp_v = CLIP3(pu1_src[(ht - 1) * src_strd] + pi1_sao_offset[edge_idx], 0, (1 << bit_depth) - 1)
PU1_AVAIL_3_LOOP:
STRB r10,[sp,#8]
STRB r9,[sp,#9]
STR r2,[sp,#216] @Store pu1_src_left in sp
MOV r12,r8 @Move ht
MOV r14,r2 @Move pu1_src_left to pu1_src_left_cpy
LDRB r11,[r5,#3] @pu1_avail[3]
CMP r11,#0
BNE PU1_AVAIL_2_LOOP
SUB r12,r12,#1 @ht_tmp--
PU1_AVAIL_2_LOOP:
LDRB r5,[r5,#2] @pu1_avail[2]
CMP r5,#0
BNE PU1_AVAIL_2_LOOP_END
ADD r0,r0,r1 @pu1_src += src_strd
SUB r12,r12,#1 @ht_tmp--
ADD r14,r14,#2 @pu1_src_left_cpy += 2
PU1_AVAIL_2_LOOP_END:
STR r0,[sp,#2] @Store pu1_src in sp
VMOV.I8 Q0,#2 @const_2 = vdupq_n_s8(2)
VMOV.I16 Q1,#0 @const_min_clip = vdupq_n_s16(0)
VMOV.I16 Q2,#255 @const_max_clip = vdupq_n_u16((1 << bit_depth) - 1)
VLD1.8 D6,[r6] @offset_tbl_u = vld1_s8(pi1_sao_offset_u)
LDR r6,[sp,#pi1_sao_v_offset] @Loads pi1_sao_offset_v
VLD1.8 D7,[r6] @offset_tbl_v = vld1_s8(pi1_sao_offset_v)
LDR r2, gi1_table_edge_idx_addr_5 @table pointer
ulbl5:
add r2,r2,pc
@VLD1.8 D6,[r6] @edge_idx_tbl = vld1_s8(gi1_table_edge_idx)
VMOV.S8 Q4,#0xFF @au1_mask = vdupq_n_s8(-1)
MOV r6,r7 @move wd to r6 loop_count
CMP r7,#16 @Compare wd with 16
BLT WIDTH_RESIDUE @If not jump to WIDTH_RESIDUE where loop is unrolled for 8 case
CMP r8,#4 @Compare ht with 4
BLE WD_16_HT_4_LOOP @If jump to WD_16_HT_4_LOOP
WIDTH_LOOP_16:
LDR r7,[sp,#wd_offset] @Loads wd
CMP r6,r7 @col == wd
LDR r5,[sp,#pu1_avail_offset] @Loads pu1_avail
LDRBEQ r8,[r5] @pu1_avail[0]
MOVNE r8,#-1
VMOV.8 D8[0],r8 @au1_mask = vsetq_lane_s8(-1, au1_mask, 0)
LDRB r11,[r5,#2] @pu1_avail[2]
CMP r6,#16 @if(col == 16)
VMOV.8 D8[1],r8 @au1_mask = vsetq_lane_s8(-1, au1_mask, 0)
BNE SKIP_AU1_MASK_VAL
LDRB r8,[r5,#1] @pu1_avail[1]
VMOV.8 D9[6],r8 @au1_mask = vsetq_lane_s8(pu1_avail[1], au1_mask, 15)
VMOV.8 D9[7],r8 @au1_mask = vsetq_lane_s8(pu1_avail[1], au1_mask, 15)
SKIP_AU1_MASK_VAL:
CMP r11,#0
VLD1.8 D12,[r0]! @pu1_cur_row = vld1q_u8(pu1_src)
VLD1.8 D13,[r0] @pu1_cur_row = vld1q_u8(pu1_src)
SUB r0,#8
ADD r5,sp,#75 @*au1_src_left_tmp
SUBEQ r8,r0,r1 @pu1_src - src_strd
VMOV.I8 Q9,#0
MOVNE r8,r3
ADD r8,r8,#2 @pu1_src - src_strd + 2
VLD1.8 D10,[r8]! @pu1_top_row = vld1q_u8(pu1_src - src_strd + 2)
VLD1.8 D11,[r8] @pu1_top_row = vld1q_u8(pu1_src - src_strd + 2)
SUB r8,#8
ADD r3,r3,#16
LDR r4,[sp,#ht_offset] @Loads ht
VCGT.U8 Q7,Q6,Q5 @vcgtq_u8(pu1_cur_row, pu1_top_row)
LDR r7,[sp,#wd_offset] @Loads wd
SUB r7,r7,r6 @(wd - col)
VCLT.U8 Q8,Q6,Q5 @vcltq_u8(pu1_cur_row, pu1_top_row)
ADD r7,r7,#14 @15 + (wd - col)
LDR r8,[sp,#212] @Loads *pu1_src
VSUB.U8 Q7,Q8,Q7 @sign_up = vreinterpretq_s8_u8(vsubq_u8(cmp_lt, cmp_gt))
ADD r7,r8,r7 @pu1_src[0 * src_strd + 15 + (wd - col)]
AU1_SRC_LEFT_LOOP:
LDRH r8,[r7] @load the value and increment by src_strd
SUBS r4,r4,#1 @decrement the loop count
STRH r8,[r5],#2 @store it in the stack pointer
ADD r7,r7,r1
BNE AU1_SRC_LEFT_LOOP
MOV r7,r12 @row count, move ht_tmp to r7
VMOV.I8 Q9,#0 @I
ADD r11,r0,r1 @I *pu1_src + src_strd
SUB r5,r12,r7 @I ht_tmp - row
VLD1.8 D16,[r11]! @I pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd)
VLD1.8 D17,[r11] @I pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd)
SUB r11,#8
ADD r8,r14,r5,LSL #1 @I pu1_src_left_cpy[(ht_tmp - row) * 2]
LDRH r5,[r8,#2] @I
VMOV.16 D19[3],r5 @I vsetq_lane_u8
LDR r11,[sp,#pu1_avail_offset] @I Loads pu1_avail
LDRB r11,[r11,#2] @I pu1_avail[2]
VEXT.8 Q9,Q9,Q8,#14 @I pu1_next_row_tmp = vextq_u8(pu1_next_row_tmp, pu1_next_row, 14)
CMP r11,#0 @I
BNE SIGN_UP_CHANGE_DONE @I
LDRB r8,[r0,#14] @I pu1_src_cpy[14]
SUB r5,r0,r1 @I
LDRB r11,[r5,#16] @I load the value pu1_src_cpy[16 - src_strd]
LDRB r9,[r0,#15] @I pu1_src_cpy[15]
SUB r8,r8,r11 @I pu1_src_cpy[14] - pu1_src_cpy[16 - src_strd]
LDRB r10,[r5,#17] @I load the value pu1_src_cpy[17 - src_strd]
CMP r8,#0 @I
MVNLT r8,#0 @I
SUB r9,r9,r10 @I pu1_src_cpy[15] - pu1_src_cpy[17 - src_strd]
MOVGT r8,#1 @I SIGN(pu1_src_cpy[14] - pu1_src_cpy[16 - src_strd])
CMP r9,#0 @I
MVNLT r9,#0 @I
VMOV.8 D15[6],r8 @I sign_up = sign_up = vsetq_lane_s8(SIGN(pu1_src_cpy[14] -pu1_src_cpy[16 - src_strd]), sign_up, 0)
MOVGT r9,#1 @I SIGN(pu1_src_cpy[15] - pu1_src_cpy[17 - src_strd]
VMOV.8 D15[7],r9 @I sign_up = vsetq_lane_s8(SIGN(pu1_src_cpy[15] -pu1_src_cpy[17 - src_strd]), sign_up, 1)
SIGN_UP_CHANGE_DONE:
VLD1.8 D28,[r2] @edge_idx_tbl = vld1_s8(gi1_table_edge_idx)
VCGT.U8 Q10,Q6,Q9 @I vcgtq_u8(pu1_cur_row, pu1_next_row_tmp)
VCLT.U8 Q11,Q6,Q9 @I vcltq_u8(pu1_cur_row, pu1_next_row_tmp)
VSUB.U8 Q11,Q11,Q10 @I sign_down = vreinterpretq_s8_u8(vsubq_u8(cmp_lt, cmp_gt))
VADD.I8 Q9,Q0,Q7 @I edge_idx = vaddq_s8(const_2, sign_up)
VADD.I8 Q9,Q9,Q11 @I edge_idx = vaddq_s8(edge_idx, sign_down)
VTBL.8 D18,{D28},D18 @I vtbl1_s8(edge_idx_tbl, vget_low_s8(edge_idx))
VNEG.S8 Q7,Q11 @I sign_up = vnegq_s8(sign_down)
VTBL.8 D19,{D28},D19 @I vtbl1_s8(edge_idx_tbl, vget_high_s8(edge_idx))
VEXT.8 Q7,Q7,Q7,#2 @I sign_up = vextq_s8(sign_up, sign_up, 2)
VMOVL.U8 Q10,D12 @I pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_cur_row)))
VAND Q9,Q9,Q4 @I edge_idx = vandq_s8(edge_idx, au1_mask)
VUZP.8 D18,D19 @I
VTBL.8 D22,{D6},D18 @I
VTBL.8 D23,{D7},D19 @I
VZIP.8 D22,D23 @I
VMOVL.U8 Q9,D13 @I pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(pu1_cur_row)))
VADDW.S8 Q10,Q10,D22 @I pi2_tmp_cur_row.val[0] = vaddw_s8(pi2_tmp_cur_row.val[0], offset)
VMAX.S16 Q10,Q10,Q1 @I pi2_tmp_cur_row.val[0] = vmaxq_s16(pi2_tmp_cur_row.val[0], const_min_clip)
VMIN.U16 Q10,Q10,Q2 @I pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[0]), const_max_clip))
VMOV Q6,Q8 @I pu1_cur_row = pu1_next_row
VADDW.S8 Q9,Q9,D23 @I pi2_tmp_cur_row.val[1] = vaddw_s8(pi2_tmp_cur_row.val[1], offset)
SUB r7,r7,#1 @I Decrement the ht_tmp loop count by 1
VMAX.S16 Q9,Q9,Q1 @I pi2_tmp_cur_row.val[1] = vmaxq_s16(pi2_tmp_cur_row.val[1], const_min_clip)
VMIN.U16 Q9,Q9,Q2 @I pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[1]), const_max_clip))
PU1_SRC_LOOP:
ADD r11,r0,r1,LSL #1 @II *pu1_src + src_strd
VMOVN.I16 D20,Q10 @I vmovn_s16(pi2_tmp_cur_row.val[0])
SUB r5,r12,r7 @II ht_tmp - row
ADD r4,r0,r1 @III *pu1_src + src_strd
VMOVN.I16 D21,Q9 @I vmovn_s16(pi2_tmp_cur_row.val[1])
ADD r8,r14,r5,LSL #1 @II pu1_src_left_cpy[(ht_tmp - row) * 2]
LDRH r9,[r8,#2]
VLD1.8 D16,[r11]! @II pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd)
VLD1.8 D17,[r11] @II pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd)
SUB r11,#8
LDRB r10,[r4,#14] @II pu1_src_cpy[14]
LDRB r8,[r4,#15] @II pu1_src_cpy[15]
VMOV.16 D29[3],r9 @II vsetq_lane_u8
ADD r4,r11,r1 @III *pu1_src + src_strd
LDRB r5,[r0,#17] @II load the value pu1_src_cpy[17 - src_strd]
VLD1.8 D30,[r4]! @III pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd)
VLD1.8 D31,[r4] @III pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd)
SUB r4,#8
LDRB r11,[r0,#16] @II load the value pu1_src_cpy[16 - src_strd]
SUB r7,r7,#1 @II Decrement the ht_tmp loop count by 1
VST1.8 {Q10},[r0],r1 @I vst1q_u8(pu1_src_cpy, pu1_cur_row)
SUB r10,r10,r11 @II pu1_src_cpy[14] - pu1_src_cpy[16 - src_strd]
CMP r10,#0 @II
VEXT.8 Q14,Q14,Q8,#14 @II pu1_next_row_tmp = vextq_u8(pu1_next_row_tmp, pu1_next_row, 14)
SUB r8,r8,r5 @II pu1_src_cpy[15] - pu1_src_cpy[17 - src_strd]
MVNLT r10,#0 @II
VLD1.8 D21,[r2] @edge_idx_tbl = vld1_s8(gi1_table_edge_idx)
MOVGT r10,#1 @II SIGN(pu1_src_cpy[14] - pu1_src_cpy[16 - src_strd])
CMP r8,#0 @II
VMOV.8 D15[6],r10 @II sign_up = sign_up = vsetq_lane_s8(SIGN(pu1_src_cpy[14] -pu1_src_cpy[16 - src_strd]), sign_up, 0)
MVNLT r8,#0 @II
MOVGT r8,#1 @II SIGN(pu1_src_cpy[15] - pu1_src_cpy[17 - src_strd]
SUB r10,r12,r7 @III ht_tmp - row
VMOV.8 D15[7],r8 @II sign_up = vsetq_lane_s8(SIGN(pu1_src_cpy[15] -pu1_src_cpy[17 - src_strd]), sign_up, 1)
ADD r11,r14,r10,LSL #1 @III pu1_src_left_cpy[(ht_tmp - row) * 2]
CMP r7,#1 @III
VCGT.U8 Q11,Q6,Q14 @II vcgtq_u8(pu1_cur_row, pu1_next_row_tmp)
BNE NEXT_ROW_POINTER_ASSIGNED_2 @III
LDR r5,[sp,#pu1_avail_offset] @III Loads pu1_avail
LDRB r5,[r5,#3] @III pu1_avail[3]
CMP r5,#0 @III
SUBNE r11,r4,#4 @III pu1_src[src_strd - 2]
NEXT_ROW_POINTER_ASSIGNED_2:
LDRH r5,[r11,#2] @III
VCLT.U8 Q12,Q6,Q14 @II vcltq_u8(pu1_cur_row, pu1_next_row_tmp)
ADD r11,r0,r1 @III
LDRB r9,[r11,#14] @III pu1_src_cpy[14]
VMOV.16 D19[3],r5 @III vsetq_lane_u8
LDRB r8,[r11,#15] @III pu1_src_cpy[15]
LDRB r11,[r0,#16] @III load the value pu1_src_cpy[16 - src_strd]
VSUB.U8 Q12,Q12,Q11 @II sign_down = vreinterpretq_s8_u8(vsubq_u8(cmp_lt, cmp_gt))
LDRB r10,[r0,#17] @III load the value pu1_src_cpy[17 - src_strd]
SUB r9,r9,r11 @III pu1_src_cpy[14] - pu1_src_cpy[16 - src_strd]
VEXT.8 Q9,Q9,Q15,#14 @III pu1_next_row_tmp = vextq_u8(pu1_next_row_tmp, pu1_next_row, 14)
SUB r10,r8,r10 @III pu1_src_cpy[15] - pu1_src_cpy[17 - src_strd]
CMP r9,#0 @III
VADD.I8 Q13,Q0,Q7 @II edge_idx = vaddq_s8(const_2, sign_up)
MVNLT r9,#0 @III
MOVGT r9,#1 @III SIGN(pu1_src_cpy[14] - pu1_src_cpy[16 - src_strd])
VADD.I8 Q13,Q13,Q12 @II edge_idx = vaddq_s8(edge_idx, sign_down)
CMP r10,#0 @III
VNEG.S8 Q7,Q12 @II sign_up = vnegq_s8(sign_down)
VTBL.8 D26,{D21},D26 @II vtbl1_s8(edge_idx_tbl, vget_low_s8(edge_idx))
MVNLT r10,#0 @III
MOVGT r10,#1 @III SIGN(pu1_src_cpy[15] - pu1_src_cpy[17 - src_strd]
VEXT.8 Q7,Q7,Q7,#2 @II sign_up = vextq_s8(sign_up, sign_up, 2)
VTBL.8 D27,{D21},D27 @II vtbl1_s8(edge_idx_tbl, vget_high_s8(edge_idx))
VCGT.U8 Q11,Q8,Q9 @III vcgtq_u8(pu1_cur_row, pu1_next_row_tmp)
VMOV.8 D15[6],r9 @III sign_up = sign_up = vsetq_lane_s8(SIGN(pu1_src_cpy[14] -pu1_src_cpy[16 - src_strd]), sign_up, 0)
VAND Q13,Q13,Q4 @II edge_idx = vandq_s8(edge_idx, au1_mask)
VMOV.8 D15[7],r10 @III sign_up = vsetq_lane_s8(SIGN(pu1_src_cpy[15] -pu1_src_cpy[17 - src_strd]), sign_up, 1)
VUZP.8 D26,D27 @II
VCLT.U8 Q10,Q8,Q9 @III vcltq_u8(pu1_cur_row, pu1_next_row_tmp)
VTBL.8 D24,{D6},D26 @II
VSUB.U8 Q11,Q10,Q11 @III sign_down = vreinterpretq_s8_u8(vsubq_u8(cmp_lt, cmp_gt))
VADD.I8 Q9,Q0,Q7 @III edge_idx = vaddq_s8(const_2, sign_up)
VTBL.8 D25,{D7},D27 @II
VADD.I8 Q9,Q9,Q11 @III edge_idx = vaddq_s8(edge_idx, sign_down)
VLD1.8 D20,[r2] @edge_idx_tbl = vld1_s8(gi1_table_edge_idx)
VZIP.8 D24,D25 @II
VMOVL.U8 Q14,D12 @II pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_cur_row)))
VTBL.8 D18,{D20},D18 @III vtbl1_s8(edge_idx_tbl, vget_low_s8(edge_idx))
VNEG.S8 Q7,Q11 @III sign_up = vnegq_s8(sign_down)
VADDW.S8 Q14,Q14,D24 @II pi2_tmp_cur_row.val[0] = vaddw_s8(pi2_tmp_cur_row.val[0], offset)
VTBL.8 D19,{D20},D19 @III vtbl1_s8(edge_idx_tbl, vget_high_s8(edge_idx))
VEXT.8 Q7,Q7,Q7,#2 @III sign_up = vextq_s8(sign_up, sign_up, 2)
VMOVL.U8 Q13,D13 @II pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(pu1_cur_row)))
VAND Q9,Q9,Q4 @III edge_idx = vandq_s8(edge_idx, au1_mask)
VMOVL.U8 Q10,D16 @III pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_cur_row)))
VUZP.8 D18,D19 @III
VMAX.S16 Q14,Q14,Q1 @II pi2_tmp_cur_row.val[0] = vmaxq_s16(pi2_tmp_cur_row.val[0], const_min_clip)
VTBL.8 D22,{D6},D18 @III
VMIN.U16 Q14,Q14,Q2 @II pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[0]), const_max_clip))
VADDW.S8 Q13,Q13,D25 @II pi2_tmp_cur_row.val[1] = vaddw_s8(pi2_tmp_cur_row.val[1], offset)
VTBL.8 D23,{D7},D19 @III
VMAX.S16 Q13,Q13,Q1 @II pi2_tmp_cur_row.val[1] = vmaxq_s16(pi2_tmp_cur_row.val[1], const_min_clip)
VMOVL.U8 Q9,D17 @III pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(pu1_cur_row)))
VZIP.8 D22,D23 @III
VMOVN.I16 D28,Q14 @II vmovn_s16(pi2_tmp_cur_row.val[0])
VADDW.S8 Q10,Q10,D22 @III pi2_tmp_cur_row.val[0] = vaddw_s8(pi2_tmp_cur_row.val[0], offset)
VMOV Q6,Q15 @III pu1_cur_row = pu1_next_row
VMIN.U16 Q13,Q13,Q2 @II pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[1]), const_max_clip))
SUB r7,r7,#1 @III Decrement the ht_tmp loop count by 1
VMAX.S16 Q10,Q10,Q1 @III pi2_tmp_cur_row.val[0] = vmaxq_s16(pi2_tmp_cur_row.val[0], const_min_clip)
CMP r7,#1 @III
VMOVN.I16 D29,Q13 @II vmovn_s16(pi2_tmp_cur_row.val[1])
VMIN.U16 Q10,Q10,Q2 @III pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[0]), const_max_clip))
VADDW.S8 Q9,Q9,D23 @III pi2_tmp_cur_row.val[1] = vaddw_s8(pi2_tmp_cur_row.val[1], offset)
VMAX.S16 Q9,Q9,Q1 @III pi2_tmp_cur_row.val[1] = vmaxq_s16(pi2_tmp_cur_row.val[1], const_min_clip)
VST1.8 {Q14},[r0],r1 @II vst1q_u8(pu1_src_cpy, pu1_cur_row)
VMIN.U16 Q9,Q9,Q2 @III pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[1]), const_max_clip))
BGT PU1_SRC_LOOP @If not equal jump to PU1_SRC_LOOP
BLT INNER_LOOP_DONE
ADD r11,r0,r1,LSL #1 @*pu1_src + src_strd
VMOVN.I16 D20,Q10 @III vmovn_s16(pi2_tmp_cur_row.val[0])
SUB r5,r12,r7 @ht_tmp - row
ADD r8,r14,r5,LSL #1 @pu1_src_left_cpy[(ht_tmp - row) * 2]
VMOVN.I16 D21,Q9 @III vmovn_s16(pi2_tmp_cur_row.val[1])
CMP r7,#1
LDRB r4,[r0,#16] @load the value pu1_src_cpy[16 - src_strd]
VLD1.8 D16,[r11]! @pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd)
VLD1.8 D17,[r11] @pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd)
SUB r11,#8
LDRB r9,[r0,#17] @load the value pu1_src_cpy[17 - src_strd]
BNE NEXT_ROW_POINTER_ASSIGNED_3
LDR r5,[sp,#pu1_avail_offset] @Loads pu1_avail
LDRB r5,[r5,#3] @pu1_avail[3]
CMP r5,#0
SUBNE r8,r11,#4 @pu1_src[src_strd - 2]
NEXT_ROW_POINTER_ASSIGNED_3:
LDRH r5,[r8,#2]
VST1.8 {Q10},[r0],r1 @III vst1q_u8(pu1_src_cpy, pu1_cur_row)
LDRB r8,[r0,#14] @pu1_src_cpy[14]
SUB r8,r8,r4 @pu1_src_cpy[14] - pu1_src_cpy[16 - src_strd]
VMOV.16 D19[3],r5 @vsetq_lane_u8
LDRB r10,[r0,#15] @pu1_src_cpy[15]
CMP r8,#0
VEXT.8 Q9,Q9,Q8,#14 @pu1_next_row_tmp = vextq_u8(pu1_next_row_tmp, pu1_next_row, 14)
SUB r10,r10,r9 @pu1_src_cpy[15] - pu1_src_cpy[17 - src_strd]
MVNLT r8,#0
VLD1.8 D28,[r2] @edge_idx_tbl = vld1_s8(gi1_table_edge_idx)
MOVGT r8,#1 @SIGN(pu1_src_cpy[14] - pu1_src_cpy[16 - src_strd])
CMP r10,#0
VMOV.8 D15[6],r8 @sign_up = sign_up = vsetq_lane_s8(SIGN(pu1_src_cpy[14] -pu1_src_cpy[16 - src_strd]), sign_up, 0)
MVNLT r10,#0
MOVGT r10,#1 @SIGN(pu1_src_cpy[15] - pu1_src_cpy[17 - src_strd]
VMOV.8 D15[7],r10 @sign_up = vsetq_lane_s8(SIGN(pu1_src_cpy[15] -pu1_src_cpy[17 - src_strd]), sign_up, 1)
VCGT.U8 Q10,Q6,Q9 @vcgtq_u8(pu1_cur_row, pu1_next_row_tmp)
VCLT.U8 Q11,Q6,Q9 @vcltq_u8(pu1_cur_row, pu1_next_row_tmp)
VSUB.U8 Q11,Q11,Q10 @sign_down = vreinterpretq_s8_u8(vsubq_u8(cmp_lt, cmp_gt))
VADD.I8 Q9,Q0,Q7 @edge_idx = vaddq_s8(const_2, sign_up)
VADD.I8 Q9,Q9,Q11 @edge_idx = vaddq_s8(edge_idx, sign_down)
VTBL.8 D18,{D28},D18 @vtbl1_s8(edge_idx_tbl, vget_low_s8(edge_idx))
VTBL.8 D19,{D28},D19 @vtbl1_s8(edge_idx_tbl, vget_high_s8(edge_idx))
VAND Q9,Q9,Q4 @edge_idx = vandq_s8(edge_idx, au1_mask)
VMOVL.U8 Q10,D12 @pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_cur_row)))
VUZP.8 D18,D19
VTBL.8 D22,{D6},D18
VTBL.8 D23,{D7},D19
VMOVL.U8 Q9,D13 @pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(pu1_cur_row)))
VZIP.8 D22,D23
VADDW.S8 Q10,Q10,D22 @pi2_tmp_cur_row.val[0] = vaddw_s8(pi2_tmp_cur_row.val[0], offset)
VMAX.S16 Q10,Q10,Q1 @pi2_tmp_cur_row.val[0] = vmaxq_s16(pi2_tmp_cur_row.val[0], const_min_clip)
VMIN.U16 Q10,Q10,Q2 @pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[0]), const_max_clip))
VADDW.S8 Q9,Q9,D23 @pi2_tmp_cur_row.val[1] = vaddw_s8(pi2_tmp_cur_row.val[1], offset)
VMAX.S16 Q9,Q9,Q1 @pi2_tmp_cur_row.val[1] = vmaxq_s16(pi2_tmp_cur_row.val[1], const_min_clip)
VMIN.U16 Q9,Q9,Q2 @pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[1]), const_max_clip))
INNER_LOOP_DONE:
LDR r8,[sp,#ht_offset] @Loads ht
VMOVN.I16 D20,Q10 @III vmovn_s16(pi2_tmp_cur_row.val[0])
ADD r5,sp,#75 @*au1_src_left_tmp
LSL r8,r8,#1
VMOVN.I16 D21,Q9 @III vmovn_s16(pi2_tmp_cur_row.val[1])
LDR r11,[sp,#216] @Loads *pu1_src_left
SRC_LEFT_LOOP:
LDR r7,[r5],#4 @au1_src_left_tmp[row]
SUBS r8,r8,#4
STR r7,[r11],#4 @pu1_src_left[row] = au1_src_left_tmp[row]
BNE SRC_LEFT_LOOP
SUBS r6,r6,#16 @Decrement the wd loop count by 16
VST1.8 {Q10},[r0],r1 @III vst1q_u8(pu1_src_cpy, pu1_cur_row)
CMP r6,#8 @Check whether residue remains
BLT RE_ASSINING_LOOP @Jump to re-assigning loop
LDR r7,[sp,#wd_offset] @Loads wd
LDR r0,[sp,#0x02] @Loads *pu1_src
SUB r7,r7,r6
ADD r0,r0,r7
BGT WIDTH_LOOP_16 @If not equal jump to width_loop
BEQ WIDTH_RESIDUE @If residue remains jump to residue loop
WD_16_HT_4_LOOP:
LDR r7,[sp,#wd_offset] @Loads wd
LDR r5,[sp,#pu1_avail_offset] @Loads pu1_avail
CMP r6,r7 @col == wd
LDRBEQ r8,[r5] @pu1_avail[0]
MOVNE r8,#-1
VMOV.8 D8[0],r8 @au1_mask = vsetq_lane_s8(-1, au1_mask, 0)
CMP r6,#16 @if(col == 16)
VMOV.8 D8[1],r8 @au1_mask = vsetq_lane_s8(-1, au1_mask, 0)
BNE SKIP_AU1_MASK_VAL_WD_16_HT_4
LDRB r8,[r5,#1] @pu1_avail[1]
VMOV.8 D9[6],r8 @au1_mask = vsetq_lane_s8(pu1_avail[1], au1_mask, 15)
VMOV.8 D9[7],r8 @au1_mask = vsetq_lane_s8(pu1_avail[1], au1_mask, 15)
SKIP_AU1_MASK_VAL_WD_16_HT_4:
LDRB r11,[r5,#2] @pu1_avail[2]
CMP r11,#0
SUBEQ r8,r0,r1 @pu1_src - src_strd
MOVNE r8,r3
VLD1.8 D12,[r0]! @pu1_cur_row = vld1q_u8(pu1_src)
VLD1.8 D13,[r0] @pu1_cur_row = vld1q_u8(pu1_src)
SUB r0,#8
ADD r8,r8,#2 @pu1_src - src_strd + 2
ADD r3,r3,#16
VLD1.8 D10,[r8]! @pu1_top_row = vld1q_u8(pu1_src - src_strd + 2)
VLD1.8 D11,[r8] @pu1_top_row = vld1q_u8(pu1_src - src_strd + 2)
SUB r8,#8
ADD r5,sp,#75 @*au1_src_left_tmp
LDR r4,[sp,#ht_offset] @Loads ht
VCGT.U8 Q7,Q6,Q5 @vcgtq_u8(pu1_cur_row, pu1_top_row)
LDR r7,[sp,#wd_offset] @Loads wd
SUB r7,r7,r6 @(wd - col)
VCLT.U8 Q8,Q6,Q5 @vcltq_u8(pu1_cur_row, pu1_top_row)
ADD r7,r7,#14 @15 + (wd - col)
LDR r8,[sp,#212] @Loads *pu1_src
VSUB.U8 Q7,Q8,Q7 @sign_up = vreinterpretq_s8_u8(vsubq_u8(cmp_lt, cmp_gt))
ADD r7,r8,r7 @pu1_src[0 * src_strd + 15 + (wd - col)]
AU1_SRC_LEFT_LOOP_WD_16_HT_4:
LDRH r8,[r7] @load the value and increment by src_strd
SUBS r4,r4,#1 @decrement the loop count
STRH r8,[r5],#2 @store it in the stack pointer
ADD r7,r7,r1
BNE AU1_SRC_LEFT_LOOP_WD_16_HT_4
VMOV.I8 Q9,#0
MOV r7,r12 @row count, move ht_tmp to r7
PU1_SRC_LOOP_WD_16_HT_4:
ADD r9,r0,r1 @*pu1_src + src_strd
LDR r5,[sp,#pu1_avail_offset] @Loads pu1_avail
VLD1.8 D16,[r9]! @pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd)
VLD1.8 D17,[r9] @pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd)
SUB r9,#8
LDRB r5,[r5,#3] @pu1_avail[3]
SUB r11,r12,r7 @ht_tmp - row
ADD r8,r14,r11,LSL #1 @pu1_src_left_cpy[(ht_tmp - row) * 2]
ADD r8,r8,#2 @pu1_src_left_cpy[(ht_tmp - row + 1) * 2]
CMP r5,#0
BEQ NEXT_ROW_POINTER_ASSIGNED_WD_16_HT_4
CMP r7,#1
SUBEQ r8,r9,#2 @pu1_src[src_strd - 2]
NEXT_ROW_POINTER_ASSIGNED_WD_16_HT_4:
LDRH r5,[r8]
VMOV.16 D19[3],r5 @vsetq_lane_u8
VEXT.8 Q9,Q9,Q8,#14 @pu1_next_row_tmp = vextq_u8(pu1_next_row_tmp, pu1_next_row, 14)
CMP r7,r12
BLT SIGN_UP_CHANGE_WD_16_HT_4
LDR r5,[sp,#pu1_avail_offset] @Loads pu1_avail
LDRB r5,[r5,#2] @pu1_avail[2]
CMP r5,#0
BNE SIGN_UP_CHANGE_DONE_WD_16_HT_4
SIGN_UP_CHANGE_WD_16_HT_4:
LDRB r8,[r0,#14] @pu1_src_cpy[14]
SUB r9,r0,r1
LDRB r5,[r9,#16] @load the value pu1_src_cpy[16 - src_strd]
LDRB r10,[r0,#15] @pu1_src_cpy[15]
SUB r8,r8,r5 @pu1_src_cpy[14] - pu1_src_cpy[16 - src_strd]
LDRB r11,[r9,#17] @load the value pu1_src_cpy[17 - src_strd]
CMP r8,#0
MVNLT r8,#0
SUB r10,r10,r11 @pu1_src_cpy[15] - pu1_src_cpy[17 - src_strd]
MOVGT r8,#1 @SIGN(pu1_src_cpy[14] - pu1_src_cpy[16 - src_strd])
CMP r10,#0
VMOV.8 D15[6],r8 @sign_up = sign_up = vsetq_lane_s8(SIGN(pu1_src_cpy[14] -pu1_src_cpy[16 - src_strd]), sign_up, 0)
MVNLT r10,#0
MOVGT r10,#1 @SIGN(pu1_src_cpy[15] - pu1_src_cpy[17 - src_strd]
VMOV.8 D15[7],r10 @sign_up = vsetq_lane_s8(SIGN(pu1_src_cpy[15] -pu1_src_cpy[17 - src_strd]), sign_up, 1)
SIGN_UP_CHANGE_DONE_WD_16_HT_4:
VLD1.8 D20,[r2] @edge_idx_tbl = vld1_s8(gi1_table_edge_idx)
VCGT.U8 Q11,Q6,Q9 @vcgtq_u8(pu1_cur_row, pu1_next_row_tmp)
VCLT.U8 Q12,Q6,Q9 @vcltq_u8(pu1_cur_row, pu1_next_row_tmp)
VSUB.U8 Q12,Q12,Q11 @sign_down = vreinterpretq_s8_u8(vsubq_u8(cmp_lt, cmp_gt))
VADD.I8 Q13,Q0,Q7 @edge_idx = vaddq_s8(const_2, sign_up)
VADD.I8 Q13,Q13,Q12 @edge_idx = vaddq_s8(edge_idx, sign_down)
VNEG.S8 Q7,Q12 @sign_up = vnegq_s8(sign_down)
VTBL.8 D26,{D20},D26 @vtbl1_s8(edge_idx_tbl, vget_low_s8(edge_idx))
VTBL.8 D27,{D20},D27 @vtbl1_s8(edge_idx_tbl, vget_high_s8(edge_idx))
VEXT.8 Q7,Q7,Q7,#2 @sign_up = vextq_s8(sign_up, sign_up, 2)
VMOVL.U8 Q14,D12 @pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_cur_row)))
VAND Q13,Q13,Q4 @edge_idx = vandq_s8(edge_idx, au1_mask)
VUZP.8 D26,D27
VTBL.8 D24,{D6},D26
VTBL.8 D25,{D7},D27
VZIP.8 D24,D25
VMOVL.U8 Q15,D13 @pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(pu1_cur_row)))
VADDW.S8 Q14,Q14,D24 @pi2_tmp_cur_row.val[0] = vaddw_s8(pi2_tmp_cur_row.val[0], offset)
VMAX.S16 Q14,Q14,Q1 @pi2_tmp_cur_row.val[0] = vmaxq_s16(pi2_tmp_cur_row.val[0], const_min_clip)
VMIN.U16 Q14,Q14,Q2 @pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[0]), const_max_clip))
VMOV Q6,Q8 @pu1_cur_row = pu1_next_row
VADDW.S8 Q15,Q15,D25 @pi2_tmp_cur_row.val[1] = vaddw_s8(pi2_tmp_cur_row.val[1], offset)
VMAX.S16 Q15,Q15,Q1 @pi2_tmp_cur_row.val[1] = vmaxq_s16(pi2_tmp_cur_row.val[1], const_min_clip)
VMIN.U16 Q15,Q15,Q2 @pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[1]), const_max_clip))
VMOVN.I16 D28,Q14 @vmovn_s16(pi2_tmp_cur_row.val[0])
VMOVN.I16 D29,Q15 @vmovn_s16(pi2_tmp_cur_row.val[1])
SUBS r7,r7,#1 @Decrement the ht_tmp loop count by 1
VST1.8 {Q14},[r0],r1 @vst1q_u8(pu1_src_cpy, pu1_cur_row)
BNE PU1_SRC_LOOP_WD_16_HT_4 @If not equal jump to PU1_SRC_LOOP_WD_16_HT_4
LDR r8,[sp,#ht_offset] @Loads ht
ADD r5,sp,#75 @*au1_src_left_tmp
LDR r11,[sp,#216] @Loads *pu1_src_left
SRC_LEFT_LOOP_WD_16_HT_4:
LDR r7,[r5],#4 @au1_src_left_tmp[row]
SUBS r8,r8,#2
STR r7,[r11],#4 @pu1_src_left[row] = au1_src_left_tmp[row]
BNE SRC_LEFT_LOOP_WD_16_HT_4
SUBS r6,r6,#16 @Decrement the wd loop count by 16
BLE RE_ASSINING_LOOP @Jump to re-assigning loop
LDR r7,[sp,#wd_offset] @Loads wd
LDR r0,[sp,#0x02] @Loads *pu1_src
SUB r7,r7,r6
ADD r0,r0,r7
BGT WD_16_HT_4_LOOP @If not equal jump to width_loop
WIDTH_RESIDUE:
LDR r7,[sp,#wd_offset] @Loads wd
LDR r5,[sp,#pu1_avail_offset] @Loads pu1_avail
CMP r6,r7 @wd_residue == wd
LDRBEQ r8,[r5] @pu1_avail[0]
MOVNE r8,#-1
LDRB r11,[r5,#1] @pu1_avail[1]
LDRB r9,[r5,#2] @pu1_avail[2]
VMOV.8 d8[0],r8 @au1_mask = vsetq_lane_s8(-1, au1_mask, 0)
CMP r9,#0
SUBEQ r10,r0,r1 @pu1_src - src_strd
VMOV.8 d8[1],r8 @au1_mask = vsetq_lane_s8(-1, au1_mask, 0)
MOVNE r10,r3
ADD r10,r10,#2 @pu1_src - src_strd + 2
VMOV.8 d8[6],r11 @au1_mask = vsetq_lane_s8(pu1_avail[1], au1_mask, 15)
ADD r5,sp,#75 @*au1_src_left_tmp
LDR r4,[sp,#ht_offset] @Loads ht
VMOV.8 d8[7],r11 @au1_mask = vsetq_lane_s8(pu1_avail[1], au1_mask, 15)
LDR r7,[sp,#wd_offset] @Loads wd
LDR r8,[sp,#212] @Loads *pu1_src
VLD1.8 D10,[r10]! @pu1_top_row = vld1q_u8(pu1_src - src_strd + 2)
VLD1.8 D11,[r10] @pu1_top_row = vld1q_u8(pu1_src - src_strd + 2)
SUB r10,#8
SUB r7,r7,#2 @(wd - 2)
ADD r7,r8,r7 @pu1_src[0 * src_strd + (wd - 2)]
AU1_SRC_LEFT_LOOP_RESIDUE:
LDRH r8,[r7] @load the value and increment by src_strd
ADD r7,r7,r1
STRH r8,[r5],#2 @store it in the stack pointer
SUBS r4,r4,#1 @decrement the loop count
BNE AU1_SRC_LEFT_LOOP_RESIDUE
VLD1.8 D12,[r0]! @pu1_cur_row = vld1q_u8(pu1_src)
VLD1.8 D13,[r0] @pu1_cur_row = vld1q_u8(pu1_src)
SUB r0,#8
VMOV.I8 Q9,#0
VCGT.U8 Q7,Q6,Q5 @vcgtq_u8(pu1_cur_row, pu1_top_row)
VCLT.U8 Q8,Q6,Q5 @vcltq_u8(pu1_cur_row, pu1_top_row)
VSUB.U8 Q7,Q8,Q7 @sign_up = vreinterpretq_s8_u8(vsubq_u8(cmp_lt, cmp_gt))
MOV r7,r12 @row count, move ht_tmp to r7
PU1_SRC_LOOP_RESIDUE:
ADD r9,r0,r1 @*pu1_src + src_strd
SUB r11,r12,r7 @ht_tmp - row
VLD1.8 D16,[r9]! @pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd)
VLD1.8 D17,[r9] @pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd)
SUB r9,#8
LDR r5,[sp,#pu1_avail_offset] @Loads pu1_avail
LDRB r5,[r5,#3] @pu1_avail[3]
ADD r8,r14,r11,LSL #1 @pu1_src_left_cpy[(ht_tmp - row) * 2]
CMP r5,#0
ADD r8,r8,#2 @pu1_src_left_cpy[(ht_tmp - row + 1) * 2]
BEQ NEXT_ROW_POINTER_ASSIGNED_RESIDUE
CMP r7,#1
SUBEQ r8,r9,#2 @pu1_src[src_strd - 2]
NEXT_ROW_POINTER_ASSIGNED_RESIDUE:
LDRB r5,[r8]
LDRB r8,[r8,#1]
VMOV.8 D19[6],r5 @vsetq_lane_u8
CMP r7,r12
VMOV.8 D19[7],r8 @vsetq_lane_u8
VEXT.8 Q9,Q9,Q8,#14 @pu1_next_row_tmp = vextq_u8(pu1_next_row_tmp, pu1_next_row, 14)
BLT SIGN_UP_CHANGE_RESIDUE
LDR r5,[sp,#pu1_avail_offset] @Loads pu1_avail
LDRB r5,[r5,#2] @pu1_avail[2]
CMP r5,#0
BNE SIGN_UP_CHANGE_DONE_RESIDUE
SIGN_UP_CHANGE_RESIDUE:
LDRB r8,[r0,#14] @pu1_src_cpy[14]
SUB r9,r0,r1
LDRB r5,[r9,#16] @load the value pu1_src_cpy[16 - src_strd]
LDRB r10,[r0,#15] @pu1_src_cpy[15]
SUB r8,r8,r5 @pu1_src_cpy[14] - pu1_src_cpy[16 - src_strd]
LDRB r11,[r9,#17] @load the value pu1_src_cpy[17 - src_strd]
CMP r8,#0
MVNLT r8,#0
SUB r10,r10,r11 @pu1_src_cpy[15] - pu1_src_cpy[17 - src_strd]
MOVGT r8,#1 @SIGN(pu1_src_cpy[14] - pu1_src_cpy[16 - src_strd])
CMP r10,#0
VMOV.8 D15[6],r8 @sign_up = sign_up = vsetq_lane_s8(SIGN(pu1_src_cpy[14] -pu1_src_cpy[16 - src_strd]), sign_up, 0)
MVNLT r10,#0
MOVGT r10,#1 @SIGN(pu1_src_cpy[15] - pu1_src_cpy[17 - src_strd]
VMOV.8 D15[7],r10 @sign_up = vsetq_lane_s8(SIGN(pu1_src_cpy[15] -pu1_src_cpy[17 - src_strd]), sign_up, 1)
SIGN_UP_CHANGE_DONE_RESIDUE:
VLD1.8 D20,[r2] @edge_idx_tbl = vld1_s8(gi1_table_edge_idx)
VCGT.U8 Q11,Q6,Q9 @vcgtq_u8(pu1_cur_row, pu1_next_row_tmp)
VCLT.U8 Q12,Q6,Q9 @vcltq_u8(pu1_cur_row, pu1_next_row_tmp)
VSUB.U8 Q12,Q12,Q11 @sign_down = vreinterpretq_s8_u8(vsubq_u8(cmp_lt, cmp_gt))
VADD.I8 Q13,Q0,Q7 @edge_idx = vaddq_s8(const_2, sign_up)
VADD.I8 Q13,Q13,Q12 @edge_idx = vaddq_s8(edge_idx, sign_down)
VNEG.S8 Q7,Q12 @sign_up = vnegq_s8(sign_down)
VTBL.8 D26,{D20},D26 @vtbl1_s8(edge_idx_tbl, vget_low_s8(edge_idx))
VTBL.8 D27,{D20},D27 @vtbl1_s8(edge_idx_tbl, vget_high_s8(edge_idx))
VEXT.8 Q7,Q7,Q7,#2 @sign_up = vextq_s8(sign_up, sign_up, 14)
VMOVL.U8 Q14,D12 @pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_cur_row)))
VAND Q13,Q13,Q4 @edge_idx = vandq_s8(edge_idx, au1_mask)
VUZP.8 D26,D27
VTBL.8 D24,{D6},D26
VTBL.8 D25,{D7},D27
VZIP.8 D24,D25
VMOV Q6,Q8 @pu1_cur_row = pu1_next_row
VADDW.S8 Q14,Q14,D24 @pi2_tmp_cur_row.val[0] = vaddw_s8(pi2_tmp_cur_row.val[0], offset)
VMAX.S16 Q14,Q14,Q1 @pi2_tmp_cur_row.val[0] = vmaxq_s16(pi2_tmp_cur_row.val[0], const_min_clip)
VMIN.U16 Q14,Q14,Q2 @pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[0]), const_max_clip))
SUBS r7,r7,#1 @Decrement the ht_tmp loop count by 1
VMOVN.I16 D30,Q14 @vmovn_s16(pi2_tmp_cur_row.val[0])
VST1.8 {D30},[r0],r1 @vst1q_u8(pu1_src_cpy, pu1_cur_row)
BNE PU1_SRC_LOOP_RESIDUE @If not equal jump to PU1_SRC_LOOP
LDR r8,[sp,#ht_offset] @Loads ht
ADD r5,sp,#75 @*au1_src_left_tmp
LDR r11,[sp,#216] @Loads *pu1_src_left
SRC_LEFT_LOOP_RESIDUE:
LDR r7,[r5],#4 @au1_src_left_tmp[row]
SUBS r8,r8,#2
STR r7,[r11],#4 @pu1_src_left[row] = au1_src_left_tmp[row]
BNE SRC_LEFT_LOOP_RESIDUE
RE_ASSINING_LOOP:
LDR r7,[sp,#wd_offset] @Loads wd
LDR r8,[sp,#ht_offset] @Loads ht
LDR r0,[sp,#212] @Loads *pu1_src
SUB r10,r7,#2 @wd - 2
LDRH r9,[sp,#6]
SUB r8,r8,#1 @ht - 1
STRH r9,[r0,r10] @pu1_src_org[0] = u1_pos_0_0_tmp
MLA r6,r8,r1,r0 @pu1_src[(ht - 1) * src_strd]
LDR r4,[sp,#pu1_src_top_left_offset] @Loads pu1_src_top_left
LDRH r9,[sp,#8]
ADD r12,sp,#10
STRH r9,[r6] @pu1_src_org[(ht - 1) * src_strd] = u1_pos_wd_ht_tmp_u
LDRH r10,[sp] @load u1_src_top_left_tmp from stack pointer
STRH r10,[r4] @*pu1_src_top_left = u1_src_top_left_tmp
LDR r3,[sp,#220] @Loads pu1_src_top
SRC_TOP_LOOP:
VLD1.8 D0,[r12]! @pu1_src_top[col] = au1_src_top_tmp[col]
SUBS r7,r7,#8 @Decrement the width
VST1.8 D0,[r3]! @pu1_src_top[col] = au1_src_top_tmp[col]
BNE SRC_TOP_LOOP
END_LOOPS:
ADD sp,sp,#224
vpop {d8 - d15}
LDMFD sp!,{r4-r12,r15} @Reload the registers from SP
|
