summaryrefslogtreecommitdiffstats
path: root/third_party/aom/av1/common/resize.c
blob: f6fdec76941d1c3d93faf3f7a894ec882a7bf30c (plain)
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
/*
 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
 *
 * This source code is subject to the terms of the BSD 2 Clause License and
 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
 * was not distributed with this source code in the LICENSE file, you can
 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
 * Media Patent License 1.0 was not distributed with this source code in the
 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
 */

#include <assert.h>
#include <limits.h>
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>

#include "./aom_config.h"
#if CONFIG_HIGHBITDEPTH
#include "aom_dsp/aom_dsp_common.h"
#endif  // CONFIG_HIGHBITDEPTH
#include "aom_ports/mem.h"
#include "aom_scale/aom_scale.h"
#include "av1/common/common.h"
#include "av1/common/resize.h"

#include "./aom_scale_rtcd.h"

#define FILTER_BITS 7

#define INTERP_TAPS 8
#define SUBPEL_BITS_RS 5
#define SUBPEL_MASK_RS ((1 << SUBPEL_BITS_RS) - 1)
#define INTERP_PRECISION_BITS 32

typedef int16_t interp_kernel[INTERP_TAPS];

// Filters for interpolation (0.5-band) - note this also filters integer pels.
static const interp_kernel filteredinterp_filters500[(1 << SUBPEL_BITS_RS)] = {
  { -3, 0, 35, 64, 35, 0, -3, 0 },    { -3, -1, 34, 64, 36, 1, -3, 0 },
  { -3, -1, 32, 64, 38, 1, -3, 0 },   { -2, -2, 31, 63, 39, 2, -3, 0 },
  { -2, -2, 29, 63, 41, 2, -3, 0 },   { -2, -2, 28, 63, 42, 3, -4, 0 },
  { -2, -3, 27, 63, 43, 4, -4, 0 },   { -2, -3, 25, 62, 45, 5, -4, 0 },
  { -2, -3, 24, 62, 46, 5, -4, 0 },   { -2, -3, 23, 61, 47, 6, -4, 0 },
  { -2, -3, 21, 60, 49, 7, -4, 0 },   { -1, -4, 20, 60, 50, 8, -4, -1 },
  { -1, -4, 19, 59, 51, 9, -4, -1 },  { -1, -4, 17, 58, 52, 10, -4, 0 },
  { -1, -4, 16, 57, 53, 12, -4, -1 }, { -1, -4, 15, 56, 54, 13, -4, -1 },
  { -1, -4, 14, 55, 55, 14, -4, -1 }, { -1, -4, 13, 54, 56, 15, -4, -1 },
  { -1, -4, 12, 53, 57, 16, -4, -1 }, { 0, -4, 10, 52, 58, 17, -4, -1 },
  { -1, -4, 9, 51, 59, 19, -4, -1 },  { -1, -4, 8, 50, 60, 20, -4, -1 },
  { 0, -4, 7, 49, 60, 21, -3, -2 },   { 0, -4, 6, 47, 61, 23, -3, -2 },
  { 0, -4, 5, 46, 62, 24, -3, -2 },   { 0, -4, 5, 45, 62, 25, -3, -2 },
  { 0, -4, 4, 43, 63, 27, -3, -2 },   { 0, -4, 3, 42, 63, 28, -2, -2 },
  { 0, -3, 2, 41, 63, 29, -2, -2 },   { 0, -3, 2, 39, 63, 31, -2, -2 },
  { 0, -3, 1, 38, 64, 32, -1, -3 },   { 0, -3, 1, 36, 64, 34, -1, -3 }
};

// Filters for interpolation (0.625-band) - note this also filters integer pels.
static const interp_kernel filteredinterp_filters625[(1 << SUBPEL_BITS_RS)] = {
  { -1, -8, 33, 80, 33, -8, -1, 0 }, { -1, -8, 30, 80, 35, -8, -1, 1 },
  { -1, -8, 28, 80, 37, -7, -2, 1 }, { 0, -8, 26, 79, 39, -7, -2, 1 },
  { 0, -8, 24, 79, 41, -7, -2, 1 },  { 0, -8, 22, 78, 43, -6, -2, 1 },
  { 0, -8, 20, 78, 45, -5, -3, 1 },  { 0, -8, 18, 77, 48, -5, -3, 1 },
  { 0, -8, 16, 76, 50, -4, -3, 1 },  { 0, -8, 15, 75, 52, -3, -4, 1 },
  { 0, -7, 13, 74, 54, -3, -4, 1 },  { 0, -7, 11, 73, 56, -2, -4, 1 },
  { 0, -7, 10, 71, 58, -1, -4, 1 },  { 1, -7, 8, 70, 60, 0, -5, 1 },
  { 1, -6, 6, 68, 62, 1, -5, 1 },    { 1, -6, 5, 67, 63, 2, -5, 1 },
  { 1, -6, 4, 65, 65, 4, -6, 1 },    { 1, -5, 2, 63, 67, 5, -6, 1 },
  { 1, -5, 1, 62, 68, 6, -6, 1 },    { 1, -5, 0, 60, 70, 8, -7, 1 },
  { 1, -4, -1, 58, 71, 10, -7, 0 },  { 1, -4, -2, 56, 73, 11, -7, 0 },
  { 1, -4, -3, 54, 74, 13, -7, 0 },  { 1, -4, -3, 52, 75, 15, -8, 0 },
  { 1, -3, -4, 50, 76, 16, -8, 0 },  { 1, -3, -5, 48, 77, 18, -8, 0 },
  { 1, -3, -5, 45, 78, 20, -8, 0 },  { 1, -2, -6, 43, 78, 22, -8, 0 },
  { 1, -2, -7, 41, 79, 24, -8, 0 },  { 1, -2, -7, 39, 79, 26, -8, 0 },
  { 1, -2, -7, 37, 80, 28, -8, -1 }, { 1, -1, -8, 35, 80, 30, -8, -1 },
};

// Filters for interpolation (0.75-band) - note this also filters integer pels.
static const interp_kernel filteredinterp_filters750[(1 << SUBPEL_BITS_RS)] = {
  { 2, -11, 25, 96, 25, -11, 2, 0 }, { 2, -11, 22, 96, 28, -11, 2, 0 },
  { 2, -10, 19, 95, 31, -11, 2, 0 }, { 2, -10, 17, 95, 34, -12, 2, 0 },
  { 2, -9, 14, 94, 37, -12, 2, 0 },  { 2, -8, 12, 93, 40, -12, 1, 0 },
  { 2, -8, 9, 92, 43, -12, 1, 1 },   { 2, -7, 7, 91, 46, -12, 1, 0 },
  { 2, -7, 5, 90, 49, -12, 1, 0 },   { 2, -6, 3, 88, 52, -12, 0, 1 },
  { 2, -5, 1, 86, 55, -12, 0, 1 },   { 2, -5, -1, 84, 58, -11, 0, 1 },
  { 2, -4, -2, 82, 61, -11, -1, 1 }, { 2, -4, -4, 80, 64, -10, -1, 1 },
  { 1, -3, -5, 77, 67, -9, -1, 1 },  { 1, -3, -6, 75, 70, -8, -2, 1 },
  { 1, -2, -7, 72, 72, -7, -2, 1 },  { 1, -2, -8, 70, 75, -6, -3, 1 },
  { 1, -1, -9, 67, 77, -5, -3, 1 },  { 1, -1, -10, 64, 80, -4, -4, 2 },
  { 1, -1, -11, 61, 82, -2, -4, 2 }, { 1, 0, -11, 58, 84, -1, -5, 2 },
  { 1, 0, -12, 55, 86, 1, -5, 2 },   { 1, 0, -12, 52, 88, 3, -6, 2 },
  { 0, 1, -12, 49, 90, 5, -7, 2 },   { 0, 1, -12, 46, 91, 7, -7, 2 },
  { 1, 1, -12, 43, 92, 9, -8, 2 },   { 0, 1, -12, 40, 93, 12, -8, 2 },
  { 0, 2, -12, 37, 94, 14, -9, 2 },  { 0, 2, -12, 34, 95, 17, -10, 2 },
  { 0, 2, -11, 31, 95, 19, -10, 2 }, { 0, 2, -11, 28, 96, 22, -11, 2 }
};

// Filters for interpolation (0.875-band) - note this also filters integer pels.
static const interp_kernel filteredinterp_filters875[(1 << SUBPEL_BITS_RS)] = {
  { 3, -8, 13, 112, 13, -8, 3, 0 },   { 3, -7, 10, 112, 17, -9, 3, -1 },
  { 2, -6, 7, 111, 21, -9, 3, -1 },   { 2, -5, 4, 111, 24, -10, 3, -1 },
  { 2, -4, 1, 110, 28, -11, 3, -1 },  { 1, -3, -1, 108, 32, -12, 4, -1 },
  { 1, -2, -3, 106, 36, -13, 4, -1 }, { 1, -1, -6, 105, 40, -14, 4, -1 },
  { 1, -1, -7, 102, 44, -14, 4, -1 }, { 1, 0, -9, 100, 48, -15, 4, -1 },
  { 1, 1, -11, 97, 53, -16, 4, -1 },  { 0, 1, -12, 95, 57, -16, 4, -1 },
  { 0, 2, -13, 91, 61, -16, 4, -1 },  { 0, 2, -14, 88, 65, -16, 4, -1 },
  { 0, 3, -15, 84, 69, -17, 4, 0 },   { 0, 3, -16, 81, 73, -16, 3, 0 },
  { 0, 3, -16, 77, 77, -16, 3, 0 },   { 0, 3, -16, 73, 81, -16, 3, 0 },
  { 0, 4, -17, 69, 84, -15, 3, 0 },   { -1, 4, -16, 65, 88, -14, 2, 0 },
  { -1, 4, -16, 61, 91, -13, 2, 0 },  { -1, 4, -16, 57, 95, -12, 1, 0 },
  { -1, 4, -16, 53, 97, -11, 1, 1 },  { -1, 4, -15, 48, 100, -9, 0, 1 },
  { -1, 4, -14, 44, 102, -7, -1, 1 }, { -1, 4, -14, 40, 105, -6, -1, 1 },
  { -1, 4, -13, 36, 106, -3, -2, 1 }, { -1, 4, -12, 32, 108, -1, -3, 1 },
  { -1, 3, -11, 28, 110, 1, -4, 2 },  { -1, 3, -10, 24, 111, 4, -5, 2 },
  { -1, 3, -9, 21, 111, 7, -6, 2 },   { -1, 3, -9, 17, 112, 10, -7, 3 }
};

// Filters for interpolation (full-band) - no filtering for integer pixels
static const interp_kernel filteredinterp_filters1000[(1 << SUBPEL_BITS_RS)] = {
  { 0, 0, 0, 128, 0, 0, 0, 0 },        { 0, 1, -3, 128, 3, -1, 0, 0 },
  { -1, 2, -6, 127, 7, -2, 1, 0 },     { -1, 3, -9, 126, 12, -4, 1, 0 },
  { -1, 4, -12, 125, 16, -5, 1, 0 },   { -1, 4, -14, 123, 20, -6, 2, 0 },
  { -1, 5, -15, 120, 25, -8, 2, 0 },   { -1, 5, -17, 118, 30, -9, 3, -1 },
  { -1, 6, -18, 114, 35, -10, 3, -1 }, { -1, 6, -19, 111, 41, -12, 3, -1 },
  { -1, 6, -20, 107, 46, -13, 4, -1 }, { -1, 6, -21, 103, 52, -14, 4, -1 },
  { -1, 6, -21, 99, 57, -16, 5, -1 },  { -1, 6, -21, 94, 63, -17, 5, -1 },
  { -1, 6, -20, 89, 68, -18, 5, -1 },  { -1, 6, -20, 84, 73, -19, 6, -1 },
  { -1, 6, -20, 79, 79, -20, 6, -1 },  { -1, 6, -19, 73, 84, -20, 6, -1 },
  { -1, 5, -18, 68, 89, -20, 6, -1 },  { -1, 5, -17, 63, 94, -21, 6, -1 },
  { -1, 5, -16, 57, 99, -21, 6, -1 },  { -1, 4, -14, 52, 103, -21, 6, -1 },
  { -1, 4, -13, 46, 107, -20, 6, -1 }, { -1, 3, -12, 41, 111, -19, 6, -1 },
  { -1, 3, -10, 35, 114, -18, 6, -1 }, { -1, 3, -9, 30, 118, -17, 5, -1 },
  { 0, 2, -8, 25, 120, -15, 5, -1 },   { 0, 2, -6, 20, 123, -14, 4, -1 },
  { 0, 1, -5, 16, 125, -12, 4, -1 },   { 0, 1, -4, 12, 126, -9, 3, -1 },
  { 0, 1, -2, 7, 127, -6, 2, -1 },     { 0, 0, -1, 3, 128, -3, 1, 0 }
};

// Filters for factor of 2 downsampling.
static const int16_t av1_down2_symeven_half_filter[] = { 56, 12, -3, -1 };
static const int16_t av1_down2_symodd_half_filter[] = { 64, 35, 0, -3 };

static const interp_kernel *choose_interp_filter(int inlength, int outlength) {
  int outlength16 = outlength * 16;
  if (outlength16 >= inlength * 16)
    return filteredinterp_filters1000;
  else if (outlength16 >= inlength * 13)
    return filteredinterp_filters875;
  else if (outlength16 >= inlength * 11)
    return filteredinterp_filters750;
  else if (outlength16 >= inlength * 9)
    return filteredinterp_filters625;
  else
    return filteredinterp_filters500;
}

static void interpolate(const uint8_t *const input, int inlength,
                        uint8_t *output, int outlength) {
  const int64_t delta =
      (((uint64_t)inlength << 32) + outlength / 2) / outlength;
  const int64_t offset =
      inlength > outlength
          ? (((int64_t)(inlength - outlength) << 31) + outlength / 2) /
                outlength
          : -(((int64_t)(outlength - inlength) << 31) + outlength / 2) /
                outlength;
  uint8_t *optr = output;
  int x, x1, x2, sum, k, int_pel, sub_pel;
  int64_t y;

  const interp_kernel *interp_filters =
      choose_interp_filter(inlength, outlength);

  x = 0;
  y = offset;
  while ((y >> INTERP_PRECISION_BITS) < (INTERP_TAPS / 2 - 1)) {
    x++;
    y += delta;
  }
  x1 = x;
  x = outlength - 1;
  y = delta * x + offset;
  while ((y >> INTERP_PRECISION_BITS) + (int64_t)(INTERP_TAPS / 2) >=
         inlength) {
    x--;
    y -= delta;
  }
  x2 = x;
  if (x1 > x2) {
    for (x = 0, y = offset; x < outlength; ++x, y += delta) {
      const int16_t *filter;
      int_pel = y >> INTERP_PRECISION_BITS;
      sub_pel =
          (y >> (INTERP_PRECISION_BITS - SUBPEL_BITS_RS)) & SUBPEL_MASK_RS;
      filter = interp_filters[sub_pel];
      sum = 0;
      for (k = 0; k < INTERP_TAPS; ++k) {
        const int pk = int_pel - INTERP_TAPS / 2 + 1 + k;
        sum += filter[k] * input[AOMMAX(AOMMIN(pk, inlength - 1), 0)];
      }
      *optr++ = clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS));
    }
  } else {
    // Initial part.
    for (x = 0, y = offset; x < x1; ++x, y += delta) {
      const int16_t *filter;
      int_pel = y >> INTERP_PRECISION_BITS;
      sub_pel =
          (y >> (INTERP_PRECISION_BITS - SUBPEL_BITS_RS)) & SUBPEL_MASK_RS;
      filter = interp_filters[sub_pel];
      sum = 0;
      for (k = 0; k < INTERP_TAPS; ++k)
        sum += filter[k] * input[AOMMAX(int_pel - INTERP_TAPS / 2 + 1 + k, 0)];
      *optr++ = clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS));
    }
    // Middle part.
    for (; x <= x2; ++x, y += delta) {
      const int16_t *filter;
      int_pel = y >> INTERP_PRECISION_BITS;
      sub_pel =
          (y >> (INTERP_PRECISION_BITS - SUBPEL_BITS_RS)) & SUBPEL_MASK_RS;
      filter = interp_filters[sub_pel];
      sum = 0;
      for (k = 0; k < INTERP_TAPS; ++k)
        sum += filter[k] * input[int_pel - INTERP_TAPS / 2 + 1 + k];
      *optr++ = clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS));
    }
    // End part.
    for (; x < outlength; ++x, y += delta) {
      const int16_t *filter;
      int_pel = y >> INTERP_PRECISION_BITS;
      sub_pel =
          (y >> (INTERP_PRECISION_BITS - SUBPEL_BITS_RS)) & SUBPEL_MASK_RS;
      filter = interp_filters[sub_pel];
      sum = 0;
      for (k = 0; k < INTERP_TAPS; ++k)
        sum += filter[k] *
               input[AOMMIN(int_pel - INTERP_TAPS / 2 + 1 + k, inlength - 1)];
      *optr++ = clip_pixel(ROUND_POWER_OF_TWO(sum, FILTER_BITS));
    }
  }
}

#ifndef __clang_analyzer__
static void down2_symeven(const uint8_t *const input, int length,
                          uint8_t *output) {
  // Actual filter len = 2 * filter_len_half.
  const int16_t *filter = av1_down2_symeven_half_filter;
  const int filter_len_half = sizeof(av1_down2_symeven_half_filter) / 2;
  int i, j;
  uint8_t *optr = output;
  int l1 = filter_len_half;
  int l2 = (length - filter_len_half);
  l1 += (l1 & 1);
  l2 += (l2 & 1);
  if (l1 > l2) {
    // Short input length.
    for (i = 0; i < length; i += 2) {
      int sum = (1 << (FILTER_BITS - 1));
      for (j = 0; j < filter_len_half; ++j) {
        sum +=
            (input[AOMMAX(i - j, 0)] + input[AOMMIN(i + 1 + j, length - 1)]) *
            filter[j];
      }
      sum >>= FILTER_BITS;
      *optr++ = clip_pixel(sum);
    }
  } else {
    // Initial part.
    for (i = 0; i < l1; i += 2) {
      int sum = (1 << (FILTER_BITS - 1));
      for (j = 0; j < filter_len_half; ++j) {
        sum += (input[AOMMAX(i - j, 0)] + input[i + 1 + j]) * filter[j];
      }
      sum >>= FILTER_BITS;
      *optr++ = clip_pixel(sum);
    }
    // Middle part.
    for (; i < l2; i += 2) {
      int sum = (1 << (FILTER_BITS - 1));
      for (j = 0; j < filter_len_half; ++j) {
        sum += (input[i - j] + input[i + 1 + j]) * filter[j];
      }
      sum >>= FILTER_BITS;
      *optr++ = clip_pixel(sum);
    }
    // End part.
    for (; i < length; i += 2) {
      int sum = (1 << (FILTER_BITS - 1));
      for (j = 0; j < filter_len_half; ++j) {
        sum +=
            (input[i - j] + input[AOMMIN(i + 1 + j, length - 1)]) * filter[j];
      }
      sum >>= FILTER_BITS;
      *optr++ = clip_pixel(sum);
    }
  }
}
#endif

static void down2_symodd(const uint8_t *const input, int length,
                         uint8_t *output) {
  // Actual filter len = 2 * filter_len_half - 1.
  const int16_t *filter = av1_down2_symodd_half_filter;
  const int filter_len_half = sizeof(av1_down2_symodd_half_filter) / 2;
  int i, j;
  uint8_t *optr = output;
  int l1 = filter_len_half - 1;
  int l2 = (length - filter_len_half + 1);
  l1 += (l1 & 1);
  l2 += (l2 & 1);
  if (l1 > l2) {
    // Short input length.
    for (i = 0; i < length; i += 2) {
      int sum = (1 << (FILTER_BITS - 1)) + input[i] * filter[0];
      for (j = 1; j < filter_len_half; ++j) {
        sum += (input[(i - j < 0 ? 0 : i - j)] +
                input[(i + j >= length ? length - 1 : i + j)]) *
               filter[j];
      }
      sum >>= FILTER_BITS;
      *optr++ = clip_pixel(sum);
    }
  } else {
    // Initial part.
    for (i = 0; i < l1; i += 2) {
      int sum = (1 << (FILTER_BITS - 1)) + input[i] * filter[0];
      for (j = 1; j < filter_len_half; ++j) {
        sum += (input[(i - j < 0 ? 0 : i - j)] + input[i + j]) * filter[j];
      }
      sum >>= FILTER_BITS;
      *optr++ = clip_pixel(sum);
    }
    // Middle part.
    for (; i < l2; i += 2) {
      int sum = (1 << (FILTER_BITS - 1)) + input[i] * filter[0];
      for (j = 1; j < filter_len_half; ++j) {
        sum += (input[i - j] + input[i + j]) * filter[j];
      }
      sum >>= FILTER_BITS;
      *optr++ = clip_pixel(sum);
    }
    // End part.
    for (; i < length; i += 2) {
      int sum = (1 << (FILTER_BITS - 1)) + input[i] * filter[0];
      for (j = 1; j < filter_len_half; ++j) {
        sum += (input[i - j] + input[(i + j >= length ? length - 1 : i + j)]) *
               filter[j];
      }
      sum >>= FILTER_BITS;
      *optr++ = clip_pixel(sum);
    }
  }
}

static int get_down2_length(int length, int steps) {
  int s;
  for (s = 0; s < steps; ++s) length = (length + 1) >> 1;
  return length;
}

static int get_down2_steps(int in_length, int out_length) {
  int steps = 0;
  int proj_in_length;
  while ((proj_in_length = get_down2_length(in_length, 1)) >= out_length) {
    ++steps;
    in_length = proj_in_length;
  }
  return steps;
}

static void resize_multistep(const uint8_t *const input, int length,
                             uint8_t *output, int olength, uint8_t *otmp) {
  if (length == olength) {
    memcpy(output, input, sizeof(output[0]) * length);
    return;
  }
  const int steps = get_down2_steps(length, olength);

  if (steps > 0) {
    uint8_t *out = NULL;
    int filteredlength = length;

    assert(otmp != NULL);
    uint8_t *otmp2 = otmp + get_down2_length(length, 1);
    for (int s = 0; s < steps; ++s) {
      const int proj_filteredlength = get_down2_length(filteredlength, 1);
      const uint8_t *const in = (s == 0 ? input : out);
      if (s == steps - 1 && proj_filteredlength == olength)
        out = output;
      else
        out = (s & 1 ? otmp2 : otmp);
      if (filteredlength & 1)
        down2_symodd(in, filteredlength, out);
      else
        down2_symeven(in, filteredlength, out);
      filteredlength = proj_filteredlength;
    }
    if (filteredlength != olength) {
      interpolate(out, filteredlength, output, olength);
    }
  } else {
    interpolate(input, length, output, olength);
  }
}

static void fill_col_to_arr(uint8_t *img, int stride, int len, uint8_t *arr) {
  int i;
  uint8_t *iptr = img;
  uint8_t *aptr = arr;
  for (i = 0; i < len; ++i, iptr += stride) {
    *aptr++ = *iptr;
  }
}

static void fill_arr_to_col(uint8_t *img, int stride, int len, uint8_t *arr) {
  int i;
  uint8_t *iptr = img;
  uint8_t *aptr = arr;
  for (i = 0; i < len; ++i, iptr += stride) {
    *iptr = *aptr++;
  }
}

void av1_resize_plane(const uint8_t *const input, int height, int width,
                      int in_stride, uint8_t *output, int height2, int width2,
                      int out_stride) {
  int i;
  uint8_t *intbuf = (uint8_t *)malloc(sizeof(uint8_t) * width2 * height);
  uint8_t *tmpbuf = (uint8_t *)malloc(sizeof(uint8_t) * AOMMAX(width, height));
  uint8_t *arrbuf = (uint8_t *)malloc(sizeof(uint8_t) * height);
  uint8_t *arrbuf2 = (uint8_t *)malloc(sizeof(uint8_t) * height2);
  if (intbuf == NULL || tmpbuf == NULL || arrbuf == NULL || arrbuf2 == NULL)
    goto Error;
  assert(width > 0);
  assert(height > 0);
  assert(width2 > 0);
  assert(height2 > 0);
  for (i = 0; i < height; ++i)
    resize_multistep(input + in_stride * i, width, intbuf + width2 * i, width2,
                     tmpbuf);
  for (i = 0; i < width2; ++i) {
    fill_col_to_arr(intbuf + i, width2, height, arrbuf);
    resize_multistep(arrbuf, height, arrbuf2, height2, tmpbuf);
    fill_arr_to_col(output + i, out_stride, height2, arrbuf2);
  }

Error:
  free(intbuf);
  free(tmpbuf);
  free(arrbuf);
  free(arrbuf2);
}

#if CONFIG_HIGHBITDEPTH
static void highbd_interpolate(const uint16_t *const input, int inlength,
                               uint16_t *output, int outlength, int bd) {
  const int64_t delta =
      (((uint64_t)inlength << 32) + outlength / 2) / outlength;
  const int64_t offset =
      inlength > outlength
          ? (((int64_t)(inlength - outlength) << 31) + outlength / 2) /
                outlength
          : -(((int64_t)(outlength - inlength) << 31) + outlength / 2) /
                outlength;
  uint16_t *optr = output;
  int x, x1, x2, sum, k, int_pel, sub_pel;
  int64_t y;

  const interp_kernel *interp_filters =
      choose_interp_filter(inlength, outlength);

  x = 0;
  y = offset;
  while ((y >> INTERP_PRECISION_BITS) < (INTERP_TAPS / 2 - 1)) {
    x++;
    y += delta;
  }
  x1 = x;
  x = outlength - 1;
  y = delta * x + offset;
  while ((y >> INTERP_PRECISION_BITS) + (int64_t)(INTERP_TAPS / 2) >=
         inlength) {
    x--;
    y -= delta;
  }
  x2 = x;
  if (x1 > x2) {
    for (x = 0, y = offset; x < outlength; ++x, y += delta) {
      const int16_t *filter;
      int_pel = y >> INTERP_PRECISION_BITS;
      sub_pel =
          (y >> (INTERP_PRECISION_BITS - SUBPEL_BITS_RS)) & SUBPEL_MASK_RS;
      filter = interp_filters[sub_pel];
      sum = 0;
      for (k = 0; k < INTERP_TAPS; ++k) {
        const int pk = int_pel - INTERP_TAPS / 2 + 1 + k;
        sum += filter[k] * input[AOMMAX(AOMMIN(pk, inlength - 1), 0)];
      }
      *optr++ = clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd);
    }
  } else {
    // Initial part.
    for (x = 0, y = offset; x < x1; ++x, y += delta) {
      const int16_t *filter;
      int_pel = y >> INTERP_PRECISION_BITS;
      sub_pel =
          (y >> (INTERP_PRECISION_BITS - SUBPEL_BITS_RS)) & SUBPEL_MASK_RS;
      filter = interp_filters[sub_pel];
      sum = 0;
      for (k = 0; k < INTERP_TAPS; ++k)
        sum += filter[k] * input[AOMMAX(int_pel - INTERP_TAPS / 2 + 1 + k, 0)];
      *optr++ = clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd);
    }
    // Middle part.
    for (; x <= x2; ++x, y += delta) {
      const int16_t *filter;
      int_pel = y >> INTERP_PRECISION_BITS;
      sub_pel =
          (y >> (INTERP_PRECISION_BITS - SUBPEL_BITS_RS)) & SUBPEL_MASK_RS;
      filter = interp_filters[sub_pel];
      sum = 0;
      for (k = 0; k < INTERP_TAPS; ++k)
        sum += filter[k] * input[int_pel - INTERP_TAPS / 2 + 1 + k];
      *optr++ = clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd);
    }
    // End part.
    for (; x < outlength; ++x, y += delta) {
      const int16_t *filter;
      int_pel = y >> INTERP_PRECISION_BITS;
      sub_pel =
          (y >> (INTERP_PRECISION_BITS - SUBPEL_BITS_RS)) & SUBPEL_MASK_RS;
      filter = interp_filters[sub_pel];
      sum = 0;
      for (k = 0; k < INTERP_TAPS; ++k)
        sum += filter[k] *
               input[AOMMIN(int_pel - INTERP_TAPS / 2 + 1 + k, inlength - 1)];
      *optr++ = clip_pixel_highbd(ROUND_POWER_OF_TWO(sum, FILTER_BITS), bd);
    }
  }
}
#ifndef __clang_analyzer__
static void highbd_down2_symeven(const uint16_t *const input, int length,
                                 uint16_t *output, int bd) {
  // Actual filter len = 2 * filter_len_half.
  static const int16_t *filter = av1_down2_symeven_half_filter;
  const int filter_len_half = sizeof(av1_down2_symeven_half_filter) / 2;
  int i, j;
  uint16_t *optr = output;
  int l1 = filter_len_half;
  int l2 = (length - filter_len_half);
  l1 += (l1 & 1);
  l2 += (l2 & 1);
  if (l1 > l2) {
    // Short input length.
    for (i = 0; i < length; i += 2) {
      int sum = (1 << (FILTER_BITS - 1));
      for (j = 0; j < filter_len_half; ++j) {
        sum +=
            (input[AOMMAX(0, i - j)] + input[AOMMIN(i + 1 + j, length - 1)]) *
            filter[j];
      }
      sum >>= FILTER_BITS;
      *optr++ = clip_pixel_highbd(sum, bd);
    }
  } else {
    // Initial part.
    for (i = 0; i < l1; i += 2) {
      int sum = (1 << (FILTER_BITS - 1));
      for (j = 0; j < filter_len_half; ++j) {
        sum += (input[AOMMAX(0, i - j)] + input[i + 1 + j]) * filter[j];
      }
      sum >>= FILTER_BITS;
      *optr++ = clip_pixel_highbd(sum, bd);
    }
    // Middle part.
    for (; i < l2; i += 2) {
      int sum = (1 << (FILTER_BITS - 1));
      for (j = 0; j < filter_len_half; ++j) {
        sum += (input[i - j] + input[i + 1 + j]) * filter[j];
      }
      sum >>= FILTER_BITS;
      *optr++ = clip_pixel_highbd(sum, bd);
    }
    // End part.
    for (; i < length; i += 2) {
      int sum = (1 << (FILTER_BITS - 1));
      for (j = 0; j < filter_len_half; ++j) {
        sum +=
            (input[i - j] + input[AOMMIN(i + 1 + j, length - 1)]) * filter[j];
      }
      sum >>= FILTER_BITS;
      *optr++ = clip_pixel_highbd(sum, bd);
    }
  }
}

static void highbd_down2_symodd(const uint16_t *const input, int length,
                                uint16_t *output, int bd) {
  // Actual filter len = 2 * filter_len_half - 1.
  static const int16_t *filter = av1_down2_symodd_half_filter;
  const int filter_len_half = sizeof(av1_down2_symodd_half_filter) / 2;
  int i, j;
  uint16_t *optr = output;
  int l1 = filter_len_half - 1;
  int l2 = (length - filter_len_half + 1);
  l1 += (l1 & 1);
  l2 += (l2 & 1);
  if (l1 > l2) {
    // Short input length.
    for (i = 0; i < length; i += 2) {
      int sum = (1 << (FILTER_BITS - 1)) + input[i] * filter[0];
      for (j = 1; j < filter_len_half; ++j) {
        sum += (input[AOMMAX(i - j, 0)] + input[AOMMIN(i + j, length - 1)]) *
               filter[j];
      }
      sum >>= FILTER_BITS;
      *optr++ = clip_pixel_highbd(sum, bd);
    }
  } else {
    // Initial part.
    for (i = 0; i < l1; i += 2) {
      int sum = (1 << (FILTER_BITS - 1)) + input[i] * filter[0];
      for (j = 1; j < filter_len_half; ++j) {
        sum += (input[AOMMAX(i - j, 0)] + input[i + j]) * filter[j];
      }
      sum >>= FILTER_BITS;
      *optr++ = clip_pixel_highbd(sum, bd);
    }
    // Middle part.
    for (; i < l2; i += 2) {
      int sum = (1 << (FILTER_BITS - 1)) + input[i] * filter[0];
      for (j = 1; j < filter_len_half; ++j) {
        sum += (input[i - j] + input[i + j]) * filter[j];
      }
      sum >>= FILTER_BITS;
      *optr++ = clip_pixel_highbd(sum, bd);
    }
    // End part.
    for (; i < length; i += 2) {
      int sum = (1 << (FILTER_BITS - 1)) + input[i] * filter[0];
      for (j = 1; j < filter_len_half; ++j) {
        sum += (input[i - j] + input[AOMMIN(i + j, length - 1)]) * filter[j];
      }
      sum >>= FILTER_BITS;
      *optr++ = clip_pixel_highbd(sum, bd);
    }
  }
}
#endif

static void highbd_resize_multistep(const uint16_t *const input, int length,
                                    uint16_t *output, int olength,
                                    uint16_t *otmp, int bd) {
  if (length == olength) {
    memcpy(output, input, sizeof(output[0]) * length);
    return;
  }
  const int steps = get_down2_steps(length, olength);

  if (steps > 0) {
    uint16_t *out = NULL;
    int filteredlength = length;

    assert(otmp != NULL);
    uint16_t *otmp2 = otmp + get_down2_length(length, 1);
    for (int s = 0; s < steps; ++s) {
      const int proj_filteredlength = get_down2_length(filteredlength, 1);
      const uint16_t *const in = (s == 0 ? input : out);
      if (s == steps - 1 && proj_filteredlength == olength)
        out = output;
      else
        out = (s & 1 ? otmp2 : otmp);
      if (filteredlength & 1)
        highbd_down2_symodd(in, filteredlength, out, bd);
      else
        highbd_down2_symeven(in, filteredlength, out, bd);
      filteredlength = proj_filteredlength;
    }
    if (filteredlength != olength) {
      highbd_interpolate(out, filteredlength, output, olength, bd);
    }
  } else {
    highbd_interpolate(input, length, output, olength, bd);
  }
}

static void highbd_fill_col_to_arr(uint16_t *img, int stride, int len,
                                   uint16_t *arr) {
  int i;
  uint16_t *iptr = img;
  uint16_t *aptr = arr;
  for (i = 0; i < len; ++i, iptr += stride) {
    *aptr++ = *iptr;
  }
}

static void highbd_fill_arr_to_col(uint16_t *img, int stride, int len,
                                   uint16_t *arr) {
  int i;
  uint16_t *iptr = img;
  uint16_t *aptr = arr;
  for (i = 0; i < len; ++i, iptr += stride) {
    *iptr = *aptr++;
  }
}

void av1_highbd_resize_plane(const uint8_t *const input, int height, int width,
                             int in_stride, uint8_t *output, int height2,
                             int width2, int out_stride, int bd) {
  int i;
  uint16_t *intbuf = (uint16_t *)malloc(sizeof(uint16_t) * width2 * height);
  uint16_t *tmpbuf =
      (uint16_t *)malloc(sizeof(uint16_t) * AOMMAX(width, height));
  uint16_t *arrbuf = (uint16_t *)malloc(sizeof(uint16_t) * height);
  uint16_t *arrbuf2 = (uint16_t *)malloc(sizeof(uint16_t) * height2);
  if (intbuf == NULL || tmpbuf == NULL || arrbuf == NULL || arrbuf2 == NULL)
    goto Error;
  for (i = 0; i < height; ++i) {
    highbd_resize_multistep(CONVERT_TO_SHORTPTR(input + in_stride * i), width,
                            intbuf + width2 * i, width2, tmpbuf, bd);
  }
  for (i = 0; i < width2; ++i) {
    highbd_fill_col_to_arr(intbuf + i, width2, height, arrbuf);
    highbd_resize_multistep(arrbuf, height, arrbuf2, height2, tmpbuf, bd);
    highbd_fill_arr_to_col(CONVERT_TO_SHORTPTR(output + i), out_stride, height2,
                           arrbuf2);
  }

Error:
  free(intbuf);
  free(tmpbuf);
  free(arrbuf);
  free(arrbuf2);
}
#endif  // CONFIG_HIGHBITDEPTH

void av1_resize_frame420(const uint8_t *const y, int y_stride,
                         const uint8_t *const u, const uint8_t *const v,
                         int uv_stride, int height, int width, uint8_t *oy,
                         int oy_stride, uint8_t *ou, uint8_t *ov,
                         int ouv_stride, int oheight, int owidth) {
  av1_resize_plane(y, height, width, y_stride, oy, oheight, owidth, oy_stride);
  av1_resize_plane(u, height / 2, width / 2, uv_stride, ou, oheight / 2,
                   owidth / 2, ouv_stride);
  av1_resize_plane(v, height / 2, width / 2, uv_stride, ov, oheight / 2,
                   owidth / 2, ouv_stride);
}

void av1_resize_frame422(const uint8_t *const y, int y_stride,
                         const uint8_t *const u, const uint8_t *const v,
                         int uv_stride, int height, int width, uint8_t *oy,
                         int oy_stride, uint8_t *ou, uint8_t *ov,
                         int ouv_stride, int oheight, int owidth) {
  av1_resize_plane(y, height, width, y_stride, oy, oheight, owidth, oy_stride);
  av1_resize_plane(u, height, width / 2, uv_stride, ou, oheight, owidth / 2,
                   ouv_stride);
  av1_resize_plane(v, height, width / 2, uv_stride, ov, oheight, owidth / 2,
                   ouv_stride);
}

void av1_resize_frame444(const uint8_t *const y, int y_stride,
                         const uint8_t *const u, const uint8_t *const v,
                         int uv_stride, int height, int width, uint8_t *oy,
                         int oy_stride, uint8_t *ou, uint8_t *ov,
                         int ouv_stride, int oheight, int owidth) {
  av1_resize_plane(y, height, width, y_stride, oy, oheight, owidth, oy_stride);
  av1_resize_plane(u, height, width, uv_stride, ou, oheight, owidth,
                   ouv_stride);
  av1_resize_plane(v, height, width, uv_stride, ov, oheight, owidth,
                   ouv_stride);
}

#if CONFIG_HIGHBITDEPTH
void av1_highbd_resize_frame420(const uint8_t *const y, int y_stride,
                                const uint8_t *const u, const uint8_t *const v,
                                int uv_stride, int height, int width,
                                uint8_t *oy, int oy_stride, uint8_t *ou,
                                uint8_t *ov, int ouv_stride, int oheight,
                                int owidth, int bd) {
  av1_highbd_resize_plane(y, height, width, y_stride, oy, oheight, owidth,
                          oy_stride, bd);
  av1_highbd_resize_plane(u, height / 2, width / 2, uv_stride, ou, oheight / 2,
                          owidth / 2, ouv_stride, bd);
  av1_highbd_resize_plane(v, height / 2, width / 2, uv_stride, ov, oheight / 2,
                          owidth / 2, ouv_stride, bd);
}

void av1_highbd_resize_frame422(const uint8_t *const y, int y_stride,
                                const uint8_t *const u, const uint8_t *const v,
                                int uv_stride, int height, int width,
                                uint8_t *oy, int oy_stride, uint8_t *ou,
                                uint8_t *ov, int ouv_stride, int oheight,
                                int owidth, int bd) {
  av1_highbd_resize_plane(y, height, width, y_stride, oy, oheight, owidth,
                          oy_stride, bd);
  av1_highbd_resize_plane(u, height, width / 2, uv_stride, ou, oheight,
                          owidth / 2, ouv_stride, bd);
  av1_highbd_resize_plane(v, height, width / 2, uv_stride, ov, oheight,
                          owidth / 2, ouv_stride, bd);
}

void av1_highbd_resize_frame444(const uint8_t *const y, int y_stride,
                                const uint8_t *const u, const uint8_t *const v,
                                int uv_stride, int height, int width,
                                uint8_t *oy, int oy_stride, uint8_t *ou,
                                uint8_t *ov, int ouv_stride, int oheight,
                                int owidth, int bd) {
  av1_highbd_resize_plane(y, height, width, y_stride, oy, oheight, owidth,
                          oy_stride, bd);
  av1_highbd_resize_plane(u, height, width, uv_stride, ou, oheight, owidth,
                          ouv_stride, bd);
  av1_highbd_resize_plane(v, height, width, uv_stride, ov, oheight, owidth,
                          ouv_stride, bd);
}
#endif  // CONFIG_HIGHBITDEPTH

#if CONFIG_HIGHBITDEPTH
static void resize_and_extend_frame(const YV12_BUFFER_CONFIG *src,
                                    YV12_BUFFER_CONFIG *dst, int bd) {
#else
static void resize_and_extend_frame(const YV12_BUFFER_CONFIG *src,
                                    YV12_BUFFER_CONFIG *dst) {
#endif  // CONFIG_HIGHBITDEPTH
  // TODO(dkovalev): replace YV12_BUFFER_CONFIG with aom_image_t
  int i;
  const uint8_t *const srcs[3] = { src->y_buffer, src->u_buffer,
                                   src->v_buffer };
  const int src_strides[3] = { src->y_stride, src->uv_stride, src->uv_stride };
  const int src_widths[3] = { src->y_crop_width, src->uv_crop_width,
                              src->uv_crop_width };
  const int src_heights[3] = { src->y_crop_height, src->uv_crop_height,
                               src->uv_crop_height };
  uint8_t *const dsts[3] = { dst->y_buffer, dst->u_buffer, dst->v_buffer };
  const int dst_strides[3] = { dst->y_stride, dst->uv_stride, dst->uv_stride };
  const int dst_widths[3] = { dst->y_crop_width, dst->uv_crop_width,
                              dst->uv_crop_width };
  const int dst_heights[3] = { dst->y_crop_height, dst->uv_crop_height,
                               dst->uv_crop_height };

  for (i = 0; i < MAX_MB_PLANE; ++i) {
#if CONFIG_HIGHBITDEPTH
    if (src->flags & YV12_FLAG_HIGHBITDEPTH)
      av1_highbd_resize_plane(srcs[i], src_heights[i], src_widths[i],
                              src_strides[i], dsts[i], dst_heights[i],
                              dst_widths[i], dst_strides[i], bd);
    else
#endif  // CONFIG_HIGHBITDEPTH
      av1_resize_plane(srcs[i], src_heights[i], src_widths[i], src_strides[i],
                       dsts[i], dst_heights[i], dst_widths[i], dst_strides[i]);
  }
  aom_extend_frame_borders(dst);
}

YV12_BUFFER_CONFIG *av1_scale_if_required_fast(AV1_COMMON *cm,
                                               YV12_BUFFER_CONFIG *unscaled,
                                               YV12_BUFFER_CONFIG *scaled) {
  if (cm->mi_cols * MI_SIZE != unscaled->y_width ||
      cm->mi_rows * MI_SIZE != unscaled->y_height) {
    // For 2x2 scaling down.
    aom_scale_frame(unscaled, scaled, unscaled->y_buffer, 9, 2, 1, 2, 1, 0);
    aom_extend_frame_borders(scaled);
    return scaled;
  } else {
    return unscaled;
  }
}

YV12_BUFFER_CONFIG *av1_scale_if_required(AV1_COMMON *cm,
                                          YV12_BUFFER_CONFIG *unscaled,
                                          YV12_BUFFER_CONFIG *scaled) {
  if (cm->width != unscaled->y_width || cm->height != unscaled->y_height) {
#if CONFIG_HIGHBITDEPTH
    resize_and_extend_frame(unscaled, scaled, (int)cm->bit_depth);
#else
    resize_and_extend_frame(unscaled, scaled);
#endif  // CONFIG_HIGHBITDEPTH
    return scaled;
  } else {
    return unscaled;
  }
}