summaryrefslogtreecommitdiffstats
path: root/third_party/dav1d/src/arm/64/mc16.S
blob: 1bfb12ebb380bb317a1e48059af37c60cb69c20d (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
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
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
/*
 * Copyright © 2018, VideoLAN and dav1d authors
 * Copyright © 2018, Janne Grunau
 * Copyright © 2020, Martin Storsjo
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are met:
 *
 * 1. Redistributions of source code must retain the above copyright notice, this
 *    list of conditions and the following disclaimer.
 *
 * 2. Redistributions in binary form must reproduce the above copyright notice,
 *    this list of conditions and the following disclaimer in the documentation
 *    and/or other materials provided with the distribution.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
 * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 */

#include "src/arm/asm.S"
#include "util.S"

#define PREP_BIAS 8192

.macro avg d0, d1, t0, t1, t2, t3
        ld1             {\t0\().8h,\t1\().8h},  [x2],  32
        ld1             {\t2\().8h,\t3\().8h},  [x3],  32
        sqadd           \t0\().8h,  \t0\().8h,  \t2\().8h
        sqadd           \t1\().8h,  \t1\().8h,  \t3\().8h
        smax            \t0\().8h,  \t0\().8h,  v28.8h // -2*PREP_BIAS - 1 << intermediate_bits
        smax            \t1\().8h,  \t1\().8h,  v28.8h // -2*PREP_BIAS - 1 << intermediate_bits
        sqsub           \t0\().8h,  \t0\().8h,  v28.8h // -2*PREP_BIAS - 1 << intermediate_bits
        sqsub           \t1\().8h,  \t1\().8h,  v28.8h // -2*PREP_BIAS - 1 << intermediate_bits
        sshl            \d0\().8h,  \t0\().8h,  v29.8h // -(intermediate_bits+1)
        sshl            \d1\().8h,  \t1\().8h,  v29.8h // -(intermediate_bits+1)
.endm

.macro w_avg d0, d1, t0, t1, t2, t3
        ld1             {\t0\().8h,\t1\().8h},  [x2],  32
        ld1             {\t2\().8h,\t3\().8h},  [x3],  32
        // This difference requires a 17 bit range, and all bits are
        // significant for the following multiplication.
        ssubl           \d0\().4s,  \t2\().4h,  \t0\().4h
        ssubl2          \t0\().4s,  \t2\().8h,  \t0\().8h
        ssubl           \d1\().4s,  \t3\().4h,  \t1\().4h
        ssubl2          \t1\().4s,  \t3\().8h,  \t1\().8h
        mul             \d0\().4s,  \d0\().4s,  v27.4s
        mul             \t0\().4s,  \t0\().4s,  v27.4s
        mul             \d1\().4s,  \d1\().4s,  v27.4s
        mul             \t1\().4s,  \t1\().4s,  v27.4s
        sshr            \d0\().4s,  \d0\().4s,  #4
        sshr            \t0\().4s,  \t0\().4s,  #4
        sshr            \d1\().4s,  \d1\().4s,  #4
        sshr            \t1\().4s,  \t1\().4s,  #4
        saddw           \d0\().4s,  \d0\().4s,  \t2\().4h
        saddw2          \t0\().4s,  \t0\().4s,  \t2\().8h
        saddw           \d1\().4s,  \d1\().4s,  \t3\().4h
        saddw2          \t1\().4s,  \t1\().4s,  \t3\().8h
        uzp1            \d0\().8h,  \d0\().8h,  \t0\().8h // Same as xtn, xtn2
        uzp1            \d1\().8h,  \d1\().8h,  \t1\().8h // Ditto
        srshl           \d0\().8h,  \d0\().8h,  v29.8h // -intermediate_bits
        srshl           \d1\().8h,  \d1\().8h,  v29.8h // -intermediate_bits
        add             \d0\().8h,  \d0\().8h,  v28.8h // PREP_BIAS >> intermediate_bits
        add             \d1\().8h,  \d1\().8h,  v28.8h // PREP_BIAS >> intermediate_bits
        smin            \d0\().8h,  \d0\().8h,  v31.8h // bitdepth_max
        smin            \d1\().8h,  \d1\().8h,  v31.8h // bitdepth_max
        smax            \d0\().8h,  \d0\().8h,  v30.8h // 0
        smax            \d1\().8h,  \d1\().8h,  v30.8h // 0
.endm

.macro mask d0, d1, t0, t1, t2, t3
        ld1             {v27.16b}, [x6],  16
        ld1             {\t0\().8h,\t1\().8h},  [x2],  32
        neg             v27.16b, v27.16b
        ld1             {\t2\().8h,\t3\().8h},  [x3],  32
        sxtl            v26.8h,  v27.8b
        sxtl2           v27.8h,  v27.16b
        sxtl            v24.4s,  v26.4h
        sxtl2           v25.4s,  v26.8h
        sxtl            v26.4s,  v27.4h
        sxtl2           v27.4s,  v27.8h
        ssubl           \d0\().4s,  \t2\().4h,  \t0\().4h
        ssubl2          \t0\().4s,  \t2\().8h,  \t0\().8h
        ssubl           \d1\().4s,  \t3\().4h,  \t1\().4h
        ssubl2          \t1\().4s,  \t3\().8h,  \t1\().8h
        mul             \d0\().4s,  \d0\().4s,  v24.4s
        mul             \t0\().4s,  \t0\().4s,  v25.4s
        mul             \d1\().4s,  \d1\().4s,  v26.4s
        mul             \t1\().4s,  \t1\().4s,  v27.4s
        sshr            \d0\().4s,  \d0\().4s,  #6
        sshr            \t0\().4s,  \t0\().4s,  #6
        sshr            \d1\().4s,  \d1\().4s,  #6
        sshr            \t1\().4s,  \t1\().4s,  #6
        saddw           \d0\().4s,  \d0\().4s,  \t2\().4h
        saddw2          \t0\().4s,  \t0\().4s,  \t2\().8h
        saddw           \d1\().4s,  \d1\().4s,  \t3\().4h
        saddw2          \t1\().4s,  \t1\().4s,  \t3\().8h
        uzp1            \d0\().8h,  \d0\().8h,  \t0\().8h  // Same as xtn, xtn2
        uzp1            \d1\().8h,  \d1\().8h,  \t1\().8h  // Ditto
        srshl           \d0\().8h,  \d0\().8h,  v29.8h // -intermediate_bits
        srshl           \d1\().8h,  \d1\().8h,  v29.8h // -intermediate_bits
        add             \d0\().8h,  \d0\().8h,  v28.8h // PREP_BIAS >> intermediate_bits
        add             \d1\().8h,  \d1\().8h,  v28.8h // PREP_BIAS >> intermediate_bits
        smin            \d0\().8h,  \d0\().8h,  v31.8h // bitdepth_max
        smin            \d1\().8h,  \d1\().8h,  v31.8h // bitdepth_max
        smax            \d0\().8h,  \d0\().8h,  v30.8h // 0
        smax            \d1\().8h,  \d1\().8h,  v30.8h // 0
.endm

.macro bidir_fn type, bdmax
function \type\()_16bpc_neon, export=1
        clz             w4,  w4
.ifnc \type, avg
        dup             v31.8h,  \bdmax // bitdepth_max
        movi            v30.8h,  #0
.endif
        clz             w7,  \bdmax
        sub             w7,  w7,  #18   // intermediate_bits = clz(bitdepth_max) - 18
.ifc \type, avg
        mov             w9,  #1
        mov             w8,  #-2*PREP_BIAS
        lsl             w9,  w9,  w7    // 1 << intermediate_bits
        add             w7,  w7,  #1
        sub             w8,  w8,  w9    // -2*PREP_BIAS - 1 << intermediate_bits
        neg             w7,  w7         // -(intermediate_bits+1)
        dup             v28.8h,   w8    // -2*PREP_BIAS - 1 << intermediate_bits
        dup             v29.8h,   w7    // -(intermediate_bits+1)
.else
        mov             w8,  #PREP_BIAS
        lsr             w8,  w8,  w7    // PREP_BIAS >> intermediate_bits
        neg             w7,  w7         // -intermediate_bits
        dup             v28.8h,  w8     // PREP_BIAS >> intermediate_bits
        dup             v29.8h,  w7     // -intermediate_bits
.endif
.ifc \type, w_avg
        dup             v27.4s,  w6
        neg             v27.4s,  v27.4s
.endif
        adr             x7,  L(\type\()_tbl)
        sub             w4,  w4,  #24
        \type           v4,  v5,  v0,  v1,  v2,  v3
        ldrh            w4,  [x7, x4, lsl #1]
        sub             x7,  x7,  w4, uxtw
        br              x7
40:
        AARCH64_VALID_JUMP_TARGET
        add             x7,  x0,  x1
        lsl             x1,  x1,  #1
4:
        subs            w5,  w5,  #4
        st1             {v4.d}[0],  [x0], x1
        st1             {v4.d}[1],  [x7], x1
        st1             {v5.d}[0],  [x0], x1
        st1             {v5.d}[1],  [x7], x1
        b.le            0f
        \type           v4,  v5,  v0,  v1,  v2,  v3
        b               4b
80:
        AARCH64_VALID_JUMP_TARGET
        add             x7,  x0,  x1
        lsl             x1,  x1,  #1
8:
        st1             {v4.8h},  [x0], x1
        subs            w5,  w5,  #2
        st1             {v5.8h},  [x7], x1
        b.le            0f
        \type           v4,  v5,  v0,  v1,  v2,  v3
        b               8b
16:
        AARCH64_VALID_JUMP_TARGET
        \type           v6,  v7,  v0,  v1,  v2,  v3
        st1             {v4.8h, v5.8h}, [x0], x1
        subs            w5,  w5,  #2
        st1             {v6.8h, v7.8h}, [x0], x1
        b.le            0f
        \type           v4,  v5,  v0,  v1,  v2,  v3
        b               16b
32:
        AARCH64_VALID_JUMP_TARGET
        \type           v6,  v7,  v0,  v1,  v2,  v3
        subs            w5,  w5,  #1
        st1             {v4.8h, v5.8h, v6.8h, v7.8h},  [x0], x1
        b.le            0f
        \type           v4,  v5,  v0,  v1,  v2,  v3
        b               32b
640:
        AARCH64_VALID_JUMP_TARGET
        add             x7,  x0,  #64
64:
        \type           v6,  v7,  v0,  v1,  v2,  v3
        \type           v16, v17, v0,  v1,  v2,  v3
        st1             {v4.8h, v5.8h, v6.8h, v7.8h},  [x0], x1
        \type           v18, v19, v0,  v1,  v2,  v3
        subs            w5,  w5,  #1
        st1             {v16.8h,v17.8h,v18.8h,v19.8h}, [x7], x1
        b.le            0f
        \type           v4,  v5,  v0,  v1,  v2,  v3
        b               64b
1280:
        AARCH64_VALID_JUMP_TARGET
        add             x7,  x0,  #64
        mov             x8,  #128
        sub             x1,  x1,  #128
128:
        \type           v6,  v7,  v0,  v1,  v2,  v3
        \type           v16, v17, v0,  v1,  v2,  v3
        st1             {v4.8h, v5.8h, v6.8h, v7.8h},  [x0], x8
        \type           v18, v19, v0,  v1,  v2,  v3
        st1             {v16.8h,v17.8h,v18.8h,v19.8h}, [x7], x8
        \type           v4,  v5,  v0,  v1,  v2,  v3
        \type           v6,  v7,  v0,  v1,  v2,  v3
        \type           v16, v17, v0,  v1,  v2,  v3
        subs            w5,  w5,  #1
        st1             {v4.8h, v5.8h, v6.8h, v7.8h},  [x0], x1
        \type           v18, v19, v0,  v1,  v2,  v3
        st1             {v16.8h,v17.8h,v18.8h,v19.8h}, [x7], x1
        b.le            0f
        \type           v4,  v5,  v0,  v1,  v2,  v3
        b               128b
0:
        ret
L(\type\()_tbl):
        .hword L(\type\()_tbl) - 1280b
        .hword L(\type\()_tbl) -  640b
        .hword L(\type\()_tbl) -   32b
        .hword L(\type\()_tbl) -   16b
        .hword L(\type\()_tbl) -   80b
        .hword L(\type\()_tbl) -   40b
endfunc
.endm

bidir_fn avg, w6
bidir_fn w_avg, w7
bidir_fn mask, w7


.macro w_mask_fn type
function w_mask_\type\()_16bpc_neon, export=1
        ldr             w8,  [sp]
        clz             w9,  w4
        adr             x10, L(w_mask_\type\()_tbl)
        dup             v31.8h,  w8   // bitdepth_max
        sub             w9,  w9,  #24
        clz             w8,  w8       // clz(bitdepth_max)
        ldrh            w9,  [x10,  x9,  lsl #1]
        sub             x10, x10, w9,  uxtw
        sub             w8,  w8,  #12 // sh = intermediate_bits + 6 = clz(bitdepth_max) - 12
        mov             w9,  #PREP_BIAS*64
        neg             w8,  w8       // -sh
        mov             w11, #27615   // (64 + 1 - 38)<<mask_sh - 1 - mask_rnd
        dup             v30.4s,  w9   // PREP_BIAS*64
        dup             v29.4s,  w8   // -sh
        dup             v0.8h,   w11
.if \type == 444
        movi            v1.16b,  #64
.elseif \type == 422
        dup             v2.8b,   w7
        movi            v3.8b,   #129
        sub             v3.8b,   v3.8b,   v2.8b
.elseif \type == 420
        dup             v2.8h,   w7
        movi            v3.8h,   #1, lsl #8
        sub             v3.8h,   v3.8h,   v2.8h
.endif
        add             x12,  x0,  x1
        lsl             x1,   x1,  #1
        br              x10
4:
        AARCH64_VALID_JUMP_TARGET
        ld1             {v4.8h, v5.8h}, [x2], #32 // tmp1 (four rows at once)
        ld1             {v6.8h, v7.8h}, [x3], #32 // tmp2 (four rows at once)
        subs            w5,  w5,  #4
        sabd            v20.8h,  v4.8h,   v6.8h   // abs(tmp1 - tmp2)
        sabd            v21.8h,  v5.8h,   v7.8h
        ssubl           v16.4s,  v6.4h,   v4.4h   // tmp2 - tmp1 (requires 17 bit)
        ssubl2          v17.4s,  v6.8h,   v4.8h
        ssubl           v18.4s,  v7.4h,   v5.4h
        ssubl2          v19.4s,  v7.8h,   v5.8h
        uqsub           v20.8h,  v0.8h,   v20.8h  // 27615 - abs()
        uqsub           v21.8h,  v0.8h,   v21.8h
        sshll2          v7.4s,   v5.8h,   #6      // tmp1 << 6
        sshll           v6.4s,   v5.4h,   #6
        sshll2          v5.4s,   v4.8h,   #6
        sshll           v4.4s,   v4.4h,   #6
        ushr            v20.8h,  v20.8h,  #10     // 64-m = (27615 - abs()) >> mask_sh
        ushr            v21.8h,  v21.8h,  #10
        add             v4.4s,   v4.4s,   v30.4s  // += PREP_BIAS*64
        add             v5.4s,   v5.4s,   v30.4s
        add             v6.4s,   v6.4s,   v30.4s
        add             v7.4s,   v7.4s,   v30.4s
        uxtl            v22.4s,  v20.4h
        uxtl2           v23.4s,  v20.8h
        uxtl            v24.4s,  v21.4h
        uxtl2           v25.4s,  v21.8h
        mla             v4.4s,   v16.4s,  v22.4s  // (tmp2-tmp1)*(64-m)
        mla             v5.4s,   v17.4s,  v23.4s
        mla             v6.4s,   v18.4s,  v24.4s
        mla             v7.4s,   v19.4s,  v25.4s
        srshl           v4.4s,   v4.4s,   v29.4s  // (tmp1<<6 + (tmp2-tmp1)*(64-m) + (1 << (sh-1)) + PREP_BIAS*64) >> sh
        srshl           v5.4s,   v5.4s,   v29.4s
        srshl           v6.4s,   v6.4s,   v29.4s
        srshl           v7.4s,   v7.4s,   v29.4s
        sqxtun          v4.4h,   v4.4s            // iclip_pixel
        sqxtun2         v4.8h,   v5.4s
        sqxtun          v5.4h,   v6.4s
        sqxtun2         v5.8h,   v7.4s
        umin            v4.8h,   v4.8h,   v31.8h  // iclip_pixel
        umin            v5.8h,   v5.8h,   v31.8h
.if \type == 444
        uzp1            v20.16b, v20.16b, v21.16b // 64 - m
        sub             v20.16b, v1.16b,  v20.16b // m
        st1             {v20.16b}, [x6], #16
.elseif \type == 422
        addp            v20.8h,  v20.8h,  v21.8h  // (64 - m) + (64 - n) (column wise addition)
        xtn             v20.8b,  v20.8h
        uhsub           v20.8b,  v3.8b,   v20.8b  // ((129 - sign) - ((64 - m) + (64 - n)) >> 1
        st1             {v20.8b}, [x6], #8
.elseif \type == 420
        trn1            v24.2d,  v20.2d,  v21.2d
        trn2            v25.2d,  v20.2d,  v21.2d
        add             v24.8h,  v24.8h,  v25.8h  // (64 - my1) + (64 - my2) (row wise addition)
        addp            v20.8h,  v24.8h,  v24.8h  // (128 - m) + (128 - n) (column wise addition)
        sub             v20.4h,  v3.4h,   v20.4h  // (256 - sign) - ((128 - m) + (128 - n))
        rshrn           v20.8b,  v20.8h,  #2      // ((256 - sign) - ((128 - m) + (128 - n)) + 2) >> 2
        st1             {v20.s}[0], [x6], #4
.endif
        st1             {v4.d}[0],  [x0],  x1
        st1             {v4.d}[1],  [x12], x1
        st1             {v5.d}[0],  [x0],  x1
        st1             {v5.d}[1],  [x12], x1
        b.gt            4b
        ret
8:
        AARCH64_VALID_JUMP_TARGET
        ld1             {v4.8h, v5.8h}, [x2], #32 // tmp1
        ld1             {v6.8h, v7.8h}, [x3], #32 // tmp2
        subs            w5,  w5,  #2
        sabd            v20.8h,  v4.8h,   v6.8h   // abs(tmp1 - tmp2)
        sabd            v21.8h,  v5.8h,   v7.8h
        ssubl           v16.4s,  v6.4h,   v4.4h   // tmp2 - tmp1 (requires 17 bit)
        ssubl2          v17.4s,  v6.8h,   v4.8h
        ssubl           v18.4s,  v7.4h,   v5.4h
        ssubl2          v19.4s,  v7.8h,   v5.8h
        uqsub           v20.8h,  v0.8h,   v20.8h  // 27615 - abs()
        uqsub           v21.8h,  v0.8h,   v21.8h
        sshll2          v7.4s,   v5.8h,   #6      // tmp1 << 6
        sshll           v6.4s,   v5.4h,   #6
        sshll2          v5.4s,   v4.8h,   #6
        sshll           v4.4s,   v4.4h,   #6
        ushr            v20.8h,  v20.8h,  #10     // 64-m = (27615 - abs()) >> mask_sh
        ushr            v21.8h,  v21.8h,  #10
        add             v4.4s,   v4.4s,   v30.4s  // += PREP_BIAS*64
        add             v5.4s,   v5.4s,   v30.4s
        add             v6.4s,   v6.4s,   v30.4s
        add             v7.4s,   v7.4s,   v30.4s
        uxtl            v22.4s,  v20.4h
        uxtl2           v23.4s,  v20.8h
        uxtl            v24.4s,  v21.4h
        uxtl2           v25.4s,  v21.8h
        mla             v4.4s,   v16.4s,  v22.4s  // (tmp2-tmp1)*(64-m)
        mla             v5.4s,   v17.4s,  v23.4s
        mla             v6.4s,   v18.4s,  v24.4s
        mla             v7.4s,   v19.4s,  v25.4s
        srshl           v4.4s,   v4.4s,   v29.4s  // (tmp1<<6 + (tmp2-tmp1)*(64-m) + (1 << (sh-1)) + PREP_BIAS*64) >> sh
        srshl           v5.4s,   v5.4s,   v29.4s
        srshl           v6.4s,   v6.4s,   v29.4s
        srshl           v7.4s,   v7.4s,   v29.4s
        sqxtun          v4.4h,   v4.4s            // iclip_pixel
        sqxtun2         v4.8h,   v5.4s
        sqxtun          v5.4h,   v6.4s
        sqxtun2         v5.8h,   v7.4s
        umin            v4.8h,   v4.8h,   v31.8h  // iclip_pixel
        umin            v5.8h,   v5.8h,   v31.8h
.if \type == 444
        uzp1            v20.16b, v20.16b, v21.16b // 64 - m
        sub             v20.16b, v1.16b,  v20.16b // m
        st1             {v20.16b}, [x6], #16
.elseif \type == 422
        addp            v20.8h,  v20.8h,  v21.8h  // (64 - m) + (64 - n) (column wise addition)
        xtn             v20.8b,  v20.8h
        uhsub           v20.8b,  v3.8b,   v20.8b  // ((129 - sign) - ((64 - m) + (64 - n)) >> 1
        st1             {v20.8b}, [x6], #8
.elseif \type == 420
        add             v20.8h,  v20.8h,  v21.8h  // (64 - my1) + (64 - my2) (row wise addition)
        addp            v20.8h,  v20.8h,  v20.8h  // (128 - m) + (128 - n) (column wise addition)
        sub             v20.4h,  v3.4h,   v20.4h  // (256 - sign) - ((128 - m) + (128 - n))
        rshrn           v20.8b,  v20.8h,  #2      // ((256 - sign) - ((128 - m) + (128 - n)) + 2) >> 2
        st1             {v20.s}[0], [x6], #4
.endif
        st1             {v4.8h}, [x0],  x1
        st1             {v5.8h}, [x12], x1
        b.gt            8b
        ret
1280:
640:
320:
160:
        AARCH64_VALID_JUMP_TARGET
        mov             w11, w4
        sub             x1,  x1,  w4,  uxtw #1
.if \type == 444
        add             x10, x6,  w4,  uxtw
.elseif \type == 422
        add             x10, x6,  x11, lsr #1
.endif
        add             x9,  x3,  w4,  uxtw #1
        add             x7,  x2,  w4,  uxtw #1
161:
        mov             w8,  w4
16:
        ld1             {v4.8h,   v5.8h},  [x2], #32 // tmp1
        ld1             {v16.8h,  v17.8h}, [x3], #32 // tmp2
        ld1             {v6.8h,   v7.8h},  [x7], #32
        ld1             {v18.8h,  v19.8h}, [x9], #32
        subs            w8,  w8,  #16
        sabd            v20.8h,  v4.8h,   v16.8h  // abs(tmp1 - tmp2)
        sabd            v21.8h,  v5.8h,   v17.8h
        ssubl           v22.4s,  v16.4h,  v4.4h   // tmp2 - tmp1 (requires 17 bit)
        ssubl2          v23.4s,  v16.8h,  v4.8h
        ssubl           v24.4s,  v17.4h,  v5.4h
        ssubl2          v25.4s,  v17.8h,  v5.8h
        uqsub           v20.8h,  v0.8h,   v20.8h  // 27615 - abs()
        uqsub           v21.8h,  v0.8h,   v21.8h
        sshll2          v27.4s,  v5.8h,   #6      // tmp1 << 6
        sshll           v26.4s,  v5.4h,   #6
        sshll2          v5.4s,   v4.8h,   #6
        sshll           v4.4s,   v4.4h,   #6
        ushr            v20.8h,  v20.8h,  #10     // 64-m = (27615 - abs()) >> mask_sh
        ushr            v21.8h,  v21.8h,  #10
        add             v4.4s,   v4.4s,   v30.4s  // += PREP_BIAS*64
        add             v5.4s,   v5.4s,   v30.4s
        add             v26.4s,  v26.4s,  v30.4s
        add             v27.4s,  v27.4s,  v30.4s
        uxtl            v16.4s,  v20.4h
        uxtl2           v17.4s,  v20.8h
        uxtl            v28.4s,  v21.4h
        mla             v4.4s,   v22.4s,  v16.4s  // (tmp2-tmp1)*(64-m)
        uxtl2           v16.4s,  v21.8h
        mla             v5.4s,   v23.4s,  v17.4s
        mla             v26.4s,  v24.4s,  v28.4s
        mla             v27.4s,  v25.4s,  v16.4s
        srshl           v4.4s,   v4.4s,   v29.4s  // (tmp1<<6 + (tmp2-tmp1)*(64-m) + (1 << (sh-1)) + PREP_BIAS*64) >> sh
        srshl           v5.4s,   v5.4s,   v29.4s
        srshl           v26.4s,  v26.4s,  v29.4s
        srshl           v27.4s,  v27.4s,  v29.4s
        sqxtun          v4.4h,   v4.4s            // iclip_pixel
        sqxtun2         v4.8h,   v5.4s
        sqxtun          v5.4h,   v26.4s
        sqxtun2         v5.8h,   v27.4s

        // Start of other half
        sabd            v22.8h,  v6.8h,   v18.8h  // abs(tmp1 - tmp2)
        sabd            v23.8h,  v7.8h,   v19.8h

        umin            v4.8h,   v4.8h,   v31.8h  // iclip_pixel
        umin            v5.8h,   v5.8h,   v31.8h

        ssubl           v16.4s,  v18.4h,  v6.4h   // tmp2 - tmp1 (requires 17 bit)
        ssubl2          v17.4s,  v18.8h,  v6.8h
        ssubl           v18.4s,  v19.4h,  v7.4h
        ssubl2          v19.4s,  v19.8h,  v7.8h
        uqsub           v22.8h,  v0.8h,   v22.8h  // 27615 - abs()
        uqsub           v23.8h,  v0.8h,   v23.8h
        sshll           v24.4s,  v6.4h,   #6      // tmp1 << 6
        sshll2          v25.4s,  v6.8h,   #6
        sshll           v26.4s,  v7.4h,   #6
        sshll2          v27.4s,  v7.8h,   #6
        ushr            v22.8h,  v22.8h,  #10     // 64-m = (27615 - abs()) >> mask_sh
        ushr            v23.8h,  v23.8h,  #10
        add             v24.4s,  v24.4s,  v30.4s  // += PREP_BIAS*64
        add             v25.4s,  v25.4s,  v30.4s
        add             v26.4s,  v26.4s,  v30.4s
        add             v27.4s,  v27.4s,  v30.4s
        uxtl            v6.4s,   v22.4h
        uxtl2           v7.4s,   v22.8h
        uxtl            v28.4s,  v23.4h
        mla             v24.4s,  v16.4s,  v6.4s   // (tmp2-tmp1)*(64-m)
        uxtl2           v6.4s,   v23.8h
        mla             v25.4s,  v17.4s,  v7.4s
        mla             v26.4s,  v18.4s,  v28.4s
        mla             v27.4s,  v19.4s,  v6.4s
        srshl           v24.4s,  v24.4s,  v29.4s  // (tmp1<<6 + (tmp2-tmp1)*(64-m) + (1 << (sh-1)) + PREP_BIAS*64) >> sh
        srshl           v25.4s,  v25.4s,  v29.4s
        srshl           v26.4s,  v26.4s,  v29.4s
        srshl           v27.4s,  v27.4s,  v29.4s
        sqxtun          v6.4h,   v24.4s           // iclip_pixel
        sqxtun2         v6.8h,   v25.4s
        sqxtun          v7.4h,   v26.4s
        sqxtun2         v7.8h,   v27.4s
        umin            v6.8h,   v6.8h,   v31.8h  // iclip_pixel
        umin            v7.8h,   v7.8h,   v31.8h
.if \type == 444
        uzp1            v20.16b, v20.16b, v21.16b // 64 - m
        uzp1            v21.16b, v22.16b, v23.16b
        sub             v20.16b, v1.16b,  v20.16b // m
        sub             v21.16b, v1.16b,  v21.16b
        st1             {v20.16b}, [x6],  #16
        st1             {v21.16b}, [x10], #16
.elseif \type == 422
        addp            v20.8h,  v20.8h,  v21.8h  // (64 - m) + (64 - n) (column wise addition)
        addp            v21.8h,  v22.8h,  v23.8h
        xtn             v20.8b,  v20.8h
        xtn             v21.8b,  v21.8h
        uhsub           v20.8b,  v3.8b,   v20.8b  // ((129 - sign) - ((64 - m) + (64 - n)) >> 1
        uhsub           v21.8b,  v3.8b,   v21.8b
        st1             {v20.8b}, [x6],  #8
        st1             {v21.8b}, [x10], #8
.elseif \type == 420
        add             v20.8h,  v20.8h,  v22.8h  // (64 - my1) + (64 - my2) (row wise addition)
        add             v21.8h,  v21.8h,  v23.8h
        addp            v20.8h,  v20.8h,  v21.8h  // (128 - m) + (128 - n) (column wise addition)
        sub             v20.8h,  v3.8h,   v20.8h  // (256 - sign) - ((128 - m) + (128 - n))
        rshrn           v20.8b,  v20.8h,  #2      // ((256 - sign) - ((128 - m) + (128 - n)) + 2) >> 2
        st1             {v20.8b}, [x6], #8
.endif
        st1             {v4.8h, v5.8h}, [x0],  #32
        st1             {v6.8h, v7.8h}, [x12], #32
        b.gt            16b
        subs            w5,  w5,  #2
        add             x2,  x2,  w4,  uxtw #1
        add             x3,  x3,  w4,  uxtw #1
        add             x7,  x7,  w4,  uxtw #1
        add             x9,  x9,  w4,  uxtw #1
.if \type == 444
        add             x6,  x6,  w4,  uxtw
        add             x10, x10, w4,  uxtw
.elseif \type == 422
        add             x6,  x6,  x11, lsr #1
        add             x10, x10, x11, lsr #1
.endif
        add             x0,  x0,  x1
        add             x12, x12, x1
        b.gt            161b
        ret
L(w_mask_\type\()_tbl):
        .hword L(w_mask_\type\()_tbl) - 1280b
        .hword L(w_mask_\type\()_tbl) -  640b
        .hword L(w_mask_\type\()_tbl) -  320b
        .hword L(w_mask_\type\()_tbl) -  160b
        .hword L(w_mask_\type\()_tbl) -    8b
        .hword L(w_mask_\type\()_tbl) -    4b
endfunc
.endm

w_mask_fn 444
w_mask_fn 422
w_mask_fn 420


function blend_16bpc_neon, export=1
        adr             x6,  L(blend_tbl)
        clz             w3,  w3
        sub             w3,  w3,  #26
        ldrh            w3,  [x6,  x3,  lsl #1]
        sub             x6,  x6,  w3,  uxtw
        add             x8,  x0,  x1
        br              x6
40:
        AARCH64_VALID_JUMP_TARGET
        lsl             x1,  x1,  #1
4:
        ld1             {v2.8b},   [x5], #8
        ld1             {v1.8h},   [x2], #16
        ld1             {v0.d}[0], [x0]
        neg             v2.8b,   v2.8b            // -m
        subs            w4,  w4,  #2
        ld1             {v0.d}[1], [x8]
        sxtl            v2.8h,   v2.8b
        shl             v2.8h,   v2.8h,   #9      // -m << 9
        sub             v1.8h,   v0.8h,   v1.8h   // a - b
        sqrdmulh        v1.8h,   v1.8h,   v2.8h   // ((a-b)*-m + 32) >> 6
        add             v0.8h,   v0.8h,   v1.8h
        st1             {v0.d}[0], [x0], x1
        st1             {v0.d}[1], [x8], x1
        b.gt            4b
        ret
80:
        AARCH64_VALID_JUMP_TARGET
        lsl             x1,  x1,  #1
8:
        ld1             {v4.16b},       [x5], #16
        ld1             {v2.8h, v3.8h}, [x2], #32
        neg             v5.16b,  v4.16b           // -m
        ld1             {v0.8h},   [x0]
        ld1             {v1.8h},   [x8]
        sxtl            v4.8h,   v5.8b
        sxtl2           v5.8h,   v5.16b
        shl             v4.8h,   v4.8h,   #9      // -m << 9
        shl             v5.8h,   v5.8h,   #9
        sub             v2.8h,   v0.8h,   v2.8h   // a - b
        sub             v3.8h,   v1.8h,   v3.8h
        subs            w4,  w4,  #2
        sqrdmulh        v2.8h,   v2.8h,   v4.8h   // ((a-b)*-m + 32) >> 6
        sqrdmulh        v3.8h,   v3.8h,   v5.8h
        add             v0.8h,   v0.8h,   v2.8h
        add             v1.8h,   v1.8h,   v3.8h
        st1             {v0.8h}, [x0], x1
        st1             {v1.8h}, [x8], x1
        b.gt            8b
        ret
160:
        AARCH64_VALID_JUMP_TARGET
        lsl             x1,  x1,  #1
16:
        ld1             {v16.16b, v17.16b},           [x5], #32
        ld1             {v4.8h, v5.8h, v6.8h, v7.8h}, [x2], #64
        subs            w4,  w4,  #2
        neg             v18.16b, v16.16b          // -m
        neg             v19.16b, v17.16b
        ld1             {v0.8h, v1.8h}, [x0]
        sxtl            v16.8h,  v18.8b
        sxtl2           v17.8h,  v18.16b
        sxtl            v18.8h,  v19.8b
        sxtl2           v19.8h,  v19.16b
        ld1             {v2.8h, v3.8h}, [x8]
        shl             v16.8h,  v16.8h,  #9      // -m << 9
        shl             v17.8h,  v17.8h,  #9
        shl             v18.8h,  v18.8h,  #9
        shl             v19.8h,  v19.8h,  #9
        sub             v4.8h,   v0.8h,   v4.8h   // a - b
        sub             v5.8h,   v1.8h,   v5.8h
        sub             v6.8h,   v2.8h,   v6.8h
        sub             v7.8h,   v3.8h,   v7.8h
        sqrdmulh        v4.8h,   v4.8h,   v16.8h  // ((a-b)*-m + 32) >> 6
        sqrdmulh        v5.8h,   v5.8h,   v17.8h
        sqrdmulh        v6.8h,   v6.8h,   v18.8h
        sqrdmulh        v7.8h,   v7.8h,   v19.8h
        add             v0.8h,   v0.8h,   v4.8h
        add             v1.8h,   v1.8h,   v5.8h
        add             v2.8h,   v2.8h,   v6.8h
        add             v3.8h,   v3.8h,   v7.8h
        st1             {v0.8h, v1.8h}, [x0], x1
        st1             {v2.8h, v3.8h}, [x8], x1
        b.gt            16b
        ret
32:
        AARCH64_VALID_JUMP_TARGET
        ld1             {v16.16b, v17.16b},           [x5], #32
        ld1             {v4.8h, v5.8h, v6.8h, v7.8h}, [x2], #64
        subs            w4,  w4,  #1
        neg             v18.16b, v16.16b          // -m
        neg             v19.16b, v17.16b
        sxtl            v16.8h,  v18.8b
        sxtl2           v17.8h,  v18.16b
        sxtl            v18.8h,  v19.8b
        sxtl2           v19.8h,  v19.16b
        ld1             {v0.8h, v1.8h, v2.8h, v3.8h}, [x0]
        shl             v16.8h,  v16.8h,  #9      // -m << 9
        shl             v17.8h,  v17.8h,  #9
        shl             v18.8h,  v18.8h,  #9
        shl             v19.8h,  v19.8h,  #9
        sub             v4.8h,   v0.8h,   v4.8h   // a - b
        sub             v5.8h,   v1.8h,   v5.8h
        sub             v6.8h,   v2.8h,   v6.8h
        sub             v7.8h,   v3.8h,   v7.8h
        sqrdmulh        v4.8h,   v4.8h,   v16.8h  // ((a-b)*-m + 32) >> 6
        sqrdmulh        v5.8h,   v5.8h,   v17.8h
        sqrdmulh        v6.8h,   v6.8h,   v18.8h
        sqrdmulh        v7.8h,   v7.8h,   v19.8h
        add             v0.8h,   v0.8h,   v4.8h
        add             v1.8h,   v1.8h,   v5.8h
        add             v2.8h,   v2.8h,   v6.8h
        add             v3.8h,   v3.8h,   v7.8h
        st1             {v0.8h, v1.8h, v2.8h, v3.8h}, [x0], x1
        b.gt            32b
        ret
L(blend_tbl):
        .hword L(blend_tbl) -  32b
        .hword L(blend_tbl) - 160b
        .hword L(blend_tbl) -  80b
        .hword L(blend_tbl) -  40b
endfunc

function blend_h_16bpc_neon, export=1
        adr             x6,  L(blend_h_tbl)
        movrel          x5,  X(obmc_masks)
        add             x5,  x5,  w4,  uxtw
        sub             w4,  w4,  w4,  lsr #2
        clz             w7,  w3
        add             x8,  x0,  x1
        lsl             x1,  x1,  #1
        sub             w7,  w7,  #24
        ldrh            w7,  [x6,  x7,  lsl #1]
        sub             x6,  x6,  w7, uxtw
        br              x6
2:
        AARCH64_VALID_JUMP_TARGET
        ld2r            {v2.8b, v3.8b}, [x5], #2
        ld1             {v1.4h},        [x2], #8
        ext             v2.8b,   v2.8b,   v3.8b,   #6
        subs            w4,  w4,  #2
        neg             v2.8b,   v2.8b            // -m
        ld1             {v0.s}[0], [x0]
        ld1             {v0.s}[1], [x8]
        sxtl            v2.8h,   v2.8b
        shl             v2.4h,   v2.4h,   #9      // -m << 9
        sub             v1.4h,   v0.4h,   v1.4h   // a - b
        sqrdmulh        v1.4h,   v1.4h,   v2.4h   // ((a-b)*-m + 32) >> 6
        add             v0.4h,   v0.4h,   v1.4h
        st1             {v0.s}[0], [x0], x1
        st1             {v0.s}[1], [x8], x1
        b.gt            2b
        ret
4:
        AARCH64_VALID_JUMP_TARGET
        ld2r            {v2.8b, v3.8b}, [x5], #2
        ld1             {v1.8h},        [x2], #16
        ext             v2.8b,   v2.8b,   v3.8b,   #4
        subs            w4,  w4,  #2
        neg             v2.8b,   v2.8b            // -m
        ld1             {v0.d}[0],   [x0]
        ld1             {v0.d}[1],   [x8]
        sxtl            v2.8h,   v2.8b
        shl             v2.8h,   v2.8h,   #9      // -m << 9
        sub             v1.8h,   v0.8h,   v1.8h   // a - b
        sqrdmulh        v1.8h,   v1.8h,   v2.8h   // ((a-b)*-m + 32) >> 6
        add             v0.8h,   v0.8h,   v1.8h
        st1             {v0.d}[0], [x0], x1
        st1             {v0.d}[1], [x8], x1
        b.gt            4b
        ret
8:
        AARCH64_VALID_JUMP_TARGET
        ld2r            {v4.8b, v5.8b}, [x5], #2
        ld1             {v2.8h, v3.8h}, [x2], #32
        neg             v4.8b,   v4.8b            // -m
        neg             v5.8b,   v5.8b
        ld1             {v0.8h}, [x0]
        subs            w4,  w4,  #2
        sxtl            v4.8h,   v4.8b
        sxtl            v5.8h,   v5.8b
        ld1             {v1.8h}, [x8]
        shl             v4.8h,   v4.8h,   #9      // -m << 9
        shl             v5.8h,   v5.8h,   #9
        sub             v2.8h,   v0.8h,   v2.8h   // a - b
        sub             v3.8h,   v1.8h,   v3.8h
        sqrdmulh        v2.8h,   v2.8h,   v4.8h   // ((a-b)*-m + 32) >> 6
        sqrdmulh        v3.8h,   v3.8h,   v5.8h
        add             v0.8h,   v0.8h,   v2.8h
        add             v1.8h,   v1.8h,   v3.8h
        st1             {v0.8h}, [x0], x1
        st1             {v1.8h}, [x8], x1
        b.gt            8b
        ret
16:
        AARCH64_VALID_JUMP_TARGET
        ld2r            {v16.8b, v17.8b}, [x5], #2
        ld1             {v4.8h, v5.8h, v6.8h, v7.8h}, [x2], #64
        neg             v16.8b,  v16.8b           // -m
        neg             v17.8b,  v17.8b
        ld1             {v0.8h, v1.8h},  [x0]
        ld1             {v2.8h, v3.8h},  [x8]
        subs            w4,  w4,  #2
        sxtl            v16.8h,  v16.8b
        sxtl            v17.8h,  v17.8b
        shl             v16.8h,  v16.8h,  #9      // -m << 9
        shl             v17.8h,  v17.8h,  #9
        sub             v4.8h,   v0.8h,   v4.8h   // a - b
        sub             v5.8h,   v1.8h,   v5.8h
        sub             v6.8h,   v2.8h,   v6.8h
        sub             v7.8h,   v3.8h,   v7.8h
        sqrdmulh        v4.8h,   v4.8h,   v16.8h  // ((a-b)*-m + 32) >> 6
        sqrdmulh        v5.8h,   v5.8h,   v16.8h
        sqrdmulh        v6.8h,   v6.8h,   v17.8h
        sqrdmulh        v7.8h,   v7.8h,   v17.8h
        add             v0.8h,   v0.8h,   v4.8h
        add             v1.8h,   v1.8h,   v5.8h
        add             v2.8h,   v2.8h,   v6.8h
        add             v3.8h,   v3.8h,   v7.8h
        st1             {v0.8h, v1.8h}, [x0], x1
        st1             {v2.8h, v3.8h}, [x8], x1
        b.gt            16b
        ret
1280:
640:
320:
        AARCH64_VALID_JUMP_TARGET
        sub             x1,  x1,  w3,  uxtw #1
        add             x7,  x2,  w3,  uxtw #1
321:
        ld2r            {v24.8b, v25.8b}, [x5], #2
        mov             w6,  w3
        neg             v24.8b,  v24.8b           // -m
        neg             v25.8b,  v25.8b
        sxtl            v24.8h,  v24.8b
        sxtl            v25.8h,  v25.8b
        shl             v24.8h,  v24.8h,  #9      // -m << 9
        shl             v25.8h,  v25.8h,  #9
32:
        ld1             {v16.8h, v17.8h, v18.8h, v19.8h}, [x2], #64
        ld1             {v0.8h,  v1.8h,  v2.8h,  v3.8h},  [x0]
        subs            w6,  w6,  #32
        sub             v16.8h,  v0.8h,   v16.8h  // a - b
        sub             v17.8h,  v1.8h,   v17.8h
        sub             v18.8h,  v2.8h,   v18.8h
        sub             v19.8h,  v3.8h,   v19.8h
        ld1             {v20.8h, v21.8h, v22.8h, v23.8h}, [x7], #64
        ld1             {v4.8h,  v5.8h,  v6.8h,  v7.8h},  [x8]
        sqrdmulh        v16.8h,  v16.8h,  v24.8h  // ((a-b)*-m + 32) >> 6
        sqrdmulh        v17.8h,  v17.8h,  v24.8h
        sqrdmulh        v18.8h,  v18.8h,  v24.8h
        sqrdmulh        v19.8h,  v19.8h,  v24.8h
        sub             v20.8h,  v4.8h,   v20.8h  // a - b
        sub             v21.8h,  v5.8h,   v21.8h
        sub             v22.8h,  v6.8h,   v22.8h
        sub             v23.8h,  v7.8h,   v23.8h
        add             v0.8h,   v0.8h,   v16.8h
        add             v1.8h,   v1.8h,   v17.8h
        add             v2.8h,   v2.8h,   v18.8h
        add             v3.8h,   v3.8h,   v19.8h
        sqrdmulh        v20.8h,  v20.8h,  v25.8h  // ((a-b)*-m + 32) >> 6
        sqrdmulh        v21.8h,  v21.8h,  v25.8h
        sqrdmulh        v22.8h,  v22.8h,  v25.8h
        sqrdmulh        v23.8h,  v23.8h,  v25.8h
        st1             {v0.8h,  v1.8h,  v2.8h,  v3.8h},  [x0], #64
        add             v4.8h,   v4.8h,   v20.8h
        add             v5.8h,   v5.8h,   v21.8h
        add             v6.8h,   v6.8h,   v22.8h
        add             v7.8h,   v7.8h,   v23.8h
        st1             {v4.8h,  v5.8h,  v6.8h,  v7.8h},  [x8], #64
        b.gt            32b
        subs            w4,  w4,  #2
        add             x0,  x0,  x1
        add             x8,  x8,  x1
        add             x2,  x2,  w3,  uxtw #1
        add             x7,  x7,  w3,  uxtw #1
        b.gt            321b
        ret
L(blend_h_tbl):
        .hword L(blend_h_tbl) - 1280b
        .hword L(blend_h_tbl) -  640b
        .hword L(blend_h_tbl) -  320b
        .hword L(blend_h_tbl) -   16b
        .hword L(blend_h_tbl) -    8b
        .hword L(blend_h_tbl) -    4b
        .hword L(blend_h_tbl) -    2b
endfunc

function blend_v_16bpc_neon, export=1
        adr             x6,  L(blend_v_tbl)
        movrel          x5,  X(obmc_masks)
        add             x5,  x5,  w3,  uxtw
        clz             w3,  w3
        add             x8,  x0,  x1
        lsl             x1,  x1,  #1
        sub             w3,  w3,  #26
        ldrh            w3,  [x6,  x3,  lsl #1]
        sub             x6,  x6,  w3,  uxtw
        br              x6
20:
        AARCH64_VALID_JUMP_TARGET
        ld1r            {v2.8b}, [x5]
        neg             v2.8b,   v2.8b            // -m
        sxtl            v2.8h,   v2.8b
        shl             v2.4h,   v2.4h,   #9      // -m << 9
2:
        ld1             {v1.s}[0], [x2], #4
        ld1             {v0.h}[0], [x0]
        subs            w4,  w4,  #2
        ld1             {v1.h}[1], [x2]
        ld1             {v0.h}[1], [x8]
        add             x2,  x2,  #4
        sub             v1.4h,   v0.4h,   v1.4h   // a - b
        sqrdmulh        v1.4h,   v1.4h,   v2.4h   // ((a-b)*-m + 32) >> 6
        add             v0.4h,   v0.4h,   v1.4h
        st1             {v0.h}[0], [x0],  x1
        st1             {v0.h}[1], [x8],  x1
        b.gt            2b
        ret
40:
        AARCH64_VALID_JUMP_TARGET
        ld1r            {v2.2s}, [x5]
        sub             x1,  x1,  #4
        neg             v2.8b,   v2.8b            // -m
        sxtl            v2.8h,   v2.8b
        shl             v2.8h,   v2.8h,   #9      // -m << 9
4:
        ld1             {v1.8h},   [x2], #16
        ld1             {v0.d}[0], [x0]
        ld1             {v0.d}[1], [x8]
        subs            w4,  w4,  #2
        sub             v1.8h,   v0.8h,   v1.8h   // a - b
        sqrdmulh        v1.8h,   v1.8h,   v2.8h   // ((a-b)*-m + 32) >> 6
        add             v0.8h,   v0.8h,   v1.8h
        st1             {v0.s}[0], [x0], #4
        st1             {v0.s}[2], [x8], #4
        st1             {v0.h}[2], [x0], x1
        st1             {v0.h}[6], [x8], x1
        b.gt            4b
        ret
80:
        AARCH64_VALID_JUMP_TARGET
        ld1             {v4.8b}, [x5]
        sub             x1,  x1,  #8
        neg             v4.8b,   v4.8b            // -m
        sxtl            v4.8h,   v4.8b
        shl             v4.8h,   v4.8h,   #9      // -m << 9
8:
        ld1             {v2.8h, v3.8h}, [x2], #32
        ld1             {v0.8h}, [x0]
        ld1             {v1.8h}, [x8]
        subs            w4,  w4,  #2
        sub             v2.8h,   v0.8h,   v2.8h   // a - b
        sub             v3.8h,   v1.8h,   v3.8h
        sqrdmulh        v2.8h,   v2.8h,   v4.8h   // ((a-b)*-m + 32) >> 6
        sqrdmulh        v3.8h,   v3.8h,   v4.8h
        add             v0.8h,   v0.8h,   v2.8h
        add             v1.8h,   v1.8h,   v3.8h
        st1             {v0.d}[0], [x0], #8
        st1             {v1.d}[0], [x8], #8
        st1             {v0.s}[2], [x0], x1
        st1             {v1.s}[2], [x8], x1
        b.gt            8b
        ret
160:
        AARCH64_VALID_JUMP_TARGET
        ld1             {v16.16b}, [x5]
        sub             x1,  x1,  #16
        neg             v17.16b, v16.16b          // -m
        sxtl            v16.8h,  v17.8b
        sxtl2           v17.8h,  v17.16b
        shl             v16.8h,  v16.8h,  #9      // -m << 9
        shl             v17.4h,  v17.4h,  #9
16:
        ld1             {v4.8h, v5.8h, v6.8h, v7.8h}, [x2], #64
        ld1             {v0.8h, v1.8h}, [x0]
        subs            w4,  w4,  #2
        ld1             {v2.8h, v3.8h}, [x8]
        sub             v4.8h,   v0.8h,   v4.8h   // a - b
        sub             v5.4h,   v1.4h,   v5.4h
        sub             v6.8h,   v2.8h,   v6.8h
        sub             v7.4h,   v3.4h,   v7.4h
        sqrdmulh        v4.8h,   v4.8h,   v16.8h  // ((a-b)*-m + 32) >> 6
        sqrdmulh        v5.4h,   v5.4h,   v17.4h
        sqrdmulh        v6.8h,   v6.8h,   v16.8h
        sqrdmulh        v7.4h,   v7.4h,   v17.4h
        add             v0.8h,   v0.8h,   v4.8h
        add             v1.4h,   v1.4h,   v5.4h
        add             v2.8h,   v2.8h,   v6.8h
        add             v3.4h,   v3.4h,   v7.4h
        st1             {v0.8h}, [x0], #16
        st1             {v2.8h}, [x8], #16
        st1             {v1.4h}, [x0], x1
        st1             {v3.4h}, [x8], x1
        b.gt            16b
        ret
320:
        AARCH64_VALID_JUMP_TARGET
        ld1             {v24.16b, v25.16b},  [x5]
        neg             v26.16b, v24.16b          // -m
        neg             v27.8b,  v25.8b
        sxtl            v24.8h,  v26.8b
        sxtl2           v25.8h,  v26.16b
        sxtl            v26.8h,  v27.8b
        shl             v24.8h,  v24.8h,  #9      // -m << 9
        shl             v25.8h,  v25.8h,  #9
        shl             v26.8h,  v26.8h,  #9
32:
        ld1             {v16.8h, v17.8h, v18.8h, v19.8h}, [x2], #64
        ld1             {v0.8h, v1.8h, v2.8h}, [x0]
        ld1             {v20.8h, v21.8h, v22.8h, v23.8h}, [x2], #64
        ld1             {v4.8h, v5.8h, v6.8h}, [x8]
        subs            w4,  w4,  #2
        sub             v16.8h,  v0.8h,   v16.8h  // a - b
        sub             v17.8h,  v1.8h,   v17.8h
        sub             v18.8h,  v2.8h,   v18.8h
        sub             v20.8h,  v4.8h,   v20.8h
        sub             v21.8h,  v5.8h,   v21.8h
        sub             v22.8h,  v6.8h,   v22.8h
        sqrdmulh        v16.8h,  v16.8h,  v24.8h  // ((a-b)*-m + 32) >> 6
        sqrdmulh        v17.8h,  v17.8h,  v25.8h
        sqrdmulh        v18.8h,  v18.8h,  v26.8h
        sqrdmulh        v20.8h,  v20.8h,  v24.8h
        sqrdmulh        v21.8h,  v21.8h,  v25.8h
        sqrdmulh        v22.8h,  v22.8h,  v26.8h
        add             v0.8h,   v0.8h,   v16.8h
        add             v1.8h,   v1.8h,   v17.8h
        add             v2.8h,   v2.8h,   v18.8h
        add             v4.8h,   v4.8h,   v20.8h
        add             v5.8h,   v5.8h,   v21.8h
        add             v6.8h,   v6.8h,   v22.8h
        st1             {v0.8h, v1.8h, v2.8h}, [x0], x1
        st1             {v4.8h, v5.8h, v6.8h}, [x8], x1
        b.gt            32b
        ret
L(blend_v_tbl):
        .hword L(blend_v_tbl) - 320b
        .hword L(blend_v_tbl) - 160b
        .hword L(blend_v_tbl) -  80b
        .hword L(blend_v_tbl) -  40b
        .hword L(blend_v_tbl) -  20b
endfunc


// This has got the same signature as the put_8tap functions,
// and assumes that x9 is set to (clz(w)-24).
function put_neon
        adr             x10, L(put_tbl)
        ldrh            w9, [x10, x9, lsl #1]
        sub             x10, x10, w9, uxtw
        br              x10

2:
        AARCH64_VALID_JUMP_TARGET
        ld1             {v0.s}[0], [x2], x3
        ld1             {v1.s}[0], [x2], x3
        subs            w5,  w5,  #2
        st1             {v0.s}[0], [x0], x1
        st1             {v1.s}[0], [x0], x1
        b.gt            2b
        ret
4:
        AARCH64_VALID_JUMP_TARGET
        ld1             {v0.4h}, [x2], x3
        ld1             {v1.4h}, [x2], x3
        subs            w5,  w5,  #2
        st1             {v0.4h}, [x0], x1
        st1             {v1.4h}, [x0], x1
        b.gt            4b
        ret
80:
        AARCH64_VALID_JUMP_TARGET
        add             x8,  x0,  x1
        lsl             x1,  x1,  #1
        add             x9,  x2,  x3
        lsl             x3,  x3,  #1
8:
        ld1             {v0.8h}, [x2], x3
        ld1             {v1.8h}, [x9], x3
        subs            w5,  w5,  #2
        st1             {v0.8h}, [x0], x1
        st1             {v1.8h}, [x8], x1
        b.gt            8b
        ret
16:
        AARCH64_VALID_JUMP_TARGET
        ldp             x6,  x7,  [x2]
        ldp             x8,  x9,  [x2, #16]
        stp             x6,  x7,  [x0]
        subs            w5,  w5,  #1
        stp             x8,  x9,  [x0, #16]
        add             x2,  x2,  x3
        add             x0,  x0,  x1
        b.gt            16b
        ret
32:
        AARCH64_VALID_JUMP_TARGET
        ldp             x6,  x7,  [x2]
        ldp             x8,  x9,  [x2, #16]
        stp             x6,  x7,  [x0]
        ldp             x10, x11, [x2, #32]
        stp             x8,  x9,  [x0, #16]
        subs            w5,  w5,  #1
        ldp             x12, x13, [x2, #48]
        stp             x10, x11, [x0, #32]
        stp             x12, x13, [x0, #48]
        add             x2,  x2,  x3
        add             x0,  x0,  x1
        b.gt            32b
        ret
64:
        AARCH64_VALID_JUMP_TARGET
        ldp             q0,  q1,  [x2]
        ldp             q2,  q3,  [x2, #32]
        stp             q0,  q1,  [x0]
        ldp             q4,  q5,  [x2, #64]
        stp             q2,  q3,  [x0, #32]
        ldp             q6,  q7,  [x2, #96]
        subs            w5,  w5,  #1
        stp             q4,  q5,  [x0, #64]
        stp             q6,  q7,  [x0, #96]
        add             x2,  x2,  x3
        add             x0,  x0,  x1
        b.gt            64b
        ret
128:
        AARCH64_VALID_JUMP_TARGET
        ldp             q0,  q1,  [x2]
        ldp             q2,  q3,  [x2, #32]
        stp             q0,  q1,  [x0]
        ldp             q4,  q5,  [x2, #64]
        stp             q2,  q3,  [x0, #32]
        ldp             q6,  q7,  [x2, #96]
        subs            w5,  w5,  #1
        stp             q4,  q5,  [x0, #64]
        ldp             q16, q17, [x2, #128]
        stp             q6,  q7,  [x0, #96]
        ldp             q18, q19, [x2, #160]
        stp             q16, q17, [x0, #128]
        ldp             q20, q21, [x2, #192]
        stp             q18, q19, [x0, #160]
        ldp             q22, q23, [x2, #224]
        stp             q20, q21, [x0, #192]
        stp             q22, q23, [x0, #224]
        add             x2,  x2,  x3
        add             x0,  x0,  x1
        b.gt            128b
        ret

L(put_tbl):
        .hword L(put_tbl) - 128b
        .hword L(put_tbl) -  64b
        .hword L(put_tbl) -  32b
        .hword L(put_tbl) -  16b
        .hword L(put_tbl) -  80b
        .hword L(put_tbl) -   4b
        .hword L(put_tbl) -   2b
endfunc


// This has got the same signature as the prep_8tap functions,
// and assumes that x9 is set to (clz(w)-24), w7 to intermediate_bits and
// x8 to w*2.
function prep_neon
        adr             x10, L(prep_tbl)
        ldrh            w9, [x10, x9, lsl #1]
        dup             v31.8h,  w7   // intermediate_bits
        movi            v30.8h,  #(PREP_BIAS >> 8), lsl #8
        sub             x10, x10, w9, uxtw
        br              x10

40:
        AARCH64_VALID_JUMP_TARGET
        add             x9,  x1,  x2
        lsl             x2,  x2,  #1
4:
        ld1             {v0.d}[0], [x1], x2
        ld1             {v0.d}[1], [x9], x2
        subs            w4,  w4,  #2
        sshl            v0.8h,   v0.8h,   v31.8h
        sub             v0.8h,   v0.8h,   v30.8h
        st1             {v0.8h}, [x0], #16
        b.gt            4b
        ret
80:
        AARCH64_VALID_JUMP_TARGET
        add             x9,  x1,  x2
        lsl             x2,  x2,  #1
8:
        ld1             {v0.8h}, [x1], x2
        ld1             {v1.8h}, [x9], x2
        subs            w4,  w4,  #2
        sshl            v0.8h,   v0.8h,   v31.8h
        sshl            v1.8h,   v1.8h,   v31.8h
        sub             v0.8h,   v0.8h,   v30.8h
        sub             v1.8h,   v1.8h,   v30.8h
        st1             {v0.8h, v1.8h}, [x0], #32
        b.gt            8b
        ret
16:
        AARCH64_VALID_JUMP_TARGET
        ldp             q0,  q1,  [x1]
        add             x1,  x1,  x2
        sshl            v0.8h,   v0.8h,   v31.8h
        ldp             q2,  q3,  [x1]
        add             x1,  x1,  x2
        subs            w4,  w4,  #2
        sshl            v1.8h,   v1.8h,   v31.8h
        sshl            v2.8h,   v2.8h,   v31.8h
        sshl            v3.8h,   v3.8h,   v31.8h
        sub             v0.8h,   v0.8h,   v30.8h
        sub             v1.8h,   v1.8h,   v30.8h
        sub             v2.8h,   v2.8h,   v30.8h
        sub             v3.8h,   v3.8h,   v30.8h
        st1             {v0.8h, v1.8h, v2.8h, v3.8h}, [x0], #64
        b.gt            16b
        ret
32:
        AARCH64_VALID_JUMP_TARGET
        ldp             q0,  q1,  [x1]
        sshl            v0.8h,   v0.8h,   v31.8h
        ldp             q2,  q3,  [x1, #32]
        add             x1,  x1,  x2
        sshl            v1.8h,   v1.8h,   v31.8h
        sshl            v2.8h,   v2.8h,   v31.8h
        sshl            v3.8h,   v3.8h,   v31.8h
        subs            w4,  w4,  #1
        sub             v0.8h,   v0.8h,   v30.8h
        sub             v1.8h,   v1.8h,   v30.8h
        sub             v2.8h,   v2.8h,   v30.8h
        sub             v3.8h,   v3.8h,   v30.8h
        st1             {v0.8h, v1.8h, v2.8h, v3.8h}, [x0], #64
        b.gt            32b
        ret
64:
        AARCH64_VALID_JUMP_TARGET
        ldp             q0,  q1,  [x1]
        subs            w4,  w4,  #1
        sshl            v0.8h,   v0.8h,   v31.8h
        ldp             q2,  q3,  [x1, #32]
        sshl            v1.8h,   v1.8h,   v31.8h
        ldp             q4,  q5,  [x1, #64]
        sshl            v2.8h,   v2.8h,   v31.8h
        sshl            v3.8h,   v3.8h,   v31.8h
        ldp             q6,  q7,  [x1, #96]
        add             x1,  x1,  x2
        sshl            v4.8h,   v4.8h,   v31.8h
        sshl            v5.8h,   v5.8h,   v31.8h
        sshl            v6.8h,   v6.8h,   v31.8h
        sshl            v7.8h,   v7.8h,   v31.8h
        sub             v0.8h,   v0.8h,   v30.8h
        sub             v1.8h,   v1.8h,   v30.8h
        sub             v2.8h,   v2.8h,   v30.8h
        sub             v3.8h,   v3.8h,   v30.8h
        stp             q0,  q1,  [x0]
        sub             v4.8h,   v4.8h,   v30.8h
        sub             v5.8h,   v5.8h,   v30.8h
        stp             q2,  q3,  [x0, #32]
        sub             v6.8h,   v6.8h,   v30.8h
        sub             v7.8h,   v7.8h,   v30.8h
        stp             q4,  q5,  [x0, #64]
        stp             q6,  q7,  [x0, #96]
        add             x0,  x0,  x8
        b.gt            64b
        ret
128:
        AARCH64_VALID_JUMP_TARGET
        ldp             q0,  q1,  [x1]
        subs            w4,  w4,  #1
        sshl            v0.8h,   v0.8h,   v31.8h
        ldp             q2,  q3,  [x1, #32]
        sshl            v1.8h,   v1.8h,   v31.8h
        ldp             q4,  q5,  [x1, #64]
        sshl            v2.8h,   v2.8h,   v31.8h
        sshl            v3.8h,   v3.8h,   v31.8h
        ldp             q6,  q7,  [x1, #96]
        sshl            v4.8h,   v4.8h,   v31.8h
        sshl            v5.8h,   v5.8h,   v31.8h
        ldp             q16, q17, [x1, #128]
        sshl            v6.8h,   v6.8h,   v31.8h
        sshl            v7.8h,   v7.8h,   v31.8h
        ldp             q18, q19, [x1, #160]
        sshl            v16.8h,  v16.8h,  v31.8h
        sshl            v17.8h,  v17.8h,  v31.8h
        ldp             q20, q21, [x1, #192]
        sshl            v18.8h,  v18.8h,  v31.8h
        sshl            v19.8h,  v19.8h,  v31.8h
        ldp             q22, q23, [x1, #224]
        add             x1,  x1,  x2
        sshl            v20.8h,  v20.8h,  v31.8h
        sshl            v21.8h,  v21.8h,  v31.8h
        sshl            v22.8h,  v22.8h,  v31.8h
        sshl            v23.8h,  v23.8h,  v31.8h
        sub             v0.8h,   v0.8h,   v30.8h
        sub             v1.8h,   v1.8h,   v30.8h
        sub             v2.8h,   v2.8h,   v30.8h
        sub             v3.8h,   v3.8h,   v30.8h
        stp             q0,  q1,  [x0]
        sub             v4.8h,   v4.8h,   v30.8h
        sub             v5.8h,   v5.8h,   v30.8h
        stp             q2,  q3,  [x0, #32]
        sub             v6.8h,   v6.8h,   v30.8h
        sub             v7.8h,   v7.8h,   v30.8h
        stp             q4,  q5,  [x0, #64]
        sub             v16.8h,  v16.8h,  v30.8h
        sub             v17.8h,  v17.8h,  v30.8h
        stp             q6,  q7,  [x0, #96]
        sub             v18.8h,  v18.8h,  v30.8h
        sub             v19.8h,  v19.8h,  v30.8h
        stp             q16, q17, [x0, #128]
        sub             v20.8h,  v20.8h,  v30.8h
        sub             v21.8h,  v21.8h,  v30.8h
        stp             q18, q19, [x0, #160]
        sub             v22.8h,  v22.8h,  v30.8h
        sub             v23.8h,  v23.8h,  v30.8h
        stp             q20, q21, [x0, #192]
        stp             q22, q23, [x0, #224]
        add             x0,  x0,  x8
        b.gt            128b
        ret

L(prep_tbl):
        .hword L(prep_tbl) - 128b
        .hword L(prep_tbl) -  64b
        .hword L(prep_tbl) -  32b
        .hword L(prep_tbl) -  16b
        .hword L(prep_tbl) -  80b
        .hword L(prep_tbl) -  40b
endfunc


.macro load_slice s0, s1, strd, wd, d0, d1, d2, d3, d4, d5, d6
        ld1             {\d0\wd}[0], [\s0], \strd
        ld1             {\d1\wd}[0], [\s1], \strd
.ifnb \d2
        ld1             {\d2\wd}[0], [\s0], \strd
        ld1             {\d3\wd}[0], [\s1], \strd
.endif
.ifnb \d4
        ld1             {\d4\wd}[0], [\s0], \strd
.endif
.ifnb \d5
        ld1             {\d5\wd}[0], [\s1], \strd
.endif
.ifnb \d6
        ld1             {\d6\wd}[0], [\s0], \strd
.endif
.endm
.macro load_reg s0, s1, strd, wd, d0, d1, d2, d3, d4, d5, d6
        ld1             {\d0\wd}, [\s0], \strd
        ld1             {\d1\wd}, [\s1], \strd
.ifnb \d2
        ld1             {\d2\wd}, [\s0], \strd
        ld1             {\d3\wd}, [\s1], \strd
.endif
.ifnb \d4
        ld1             {\d4\wd}, [\s0], \strd
.endif
.ifnb \d5
        ld1             {\d5\wd}, [\s1], \strd
.endif
.ifnb \d6
        ld1             {\d6\wd}, [\s0], \strd
.endif
.endm
.macro load_regpair s0, s1, strd, wd, d0, d1, d2, d3, d4, d5
        ld1             {\d0\wd, \d1\wd}, [\s0], \strd
.ifnb \d2
        ld1             {\d2\wd, \d3\wd}, [\s1], \strd
.endif
.ifnb \d4
        ld1             {\d4\wd, \d5\wd}, [\s0], \strd
.endif
.endm
.macro load_s s0, s1, strd, d0, d1, d2, d3, d4, d5, d6
        load_slice      \s0, \s1, \strd, .s, \d0, \d1, \d2, \d3, \d4, \d5, \d6
.endm
.macro load_4h s0, s1, strd, d0, d1, d2, d3, d4, d5, d6
        load_reg        \s0, \s1, \strd, .4h, \d0, \d1, \d2, \d3, \d4, \d5, \d6
.endm
.macro load_8h s0, s1, strd, d0, d1, d2, d3, d4, d5, d6
        load_reg        \s0, \s1, \strd, .8h, \d0, \d1, \d2, \d3, \d4, \d5, \d6
.endm
.macro load_16h s0, s1, strd, d0, d1, d2, d3, d4, d5
        load_regpair    \s0, \s1, \strd, .8h, \d0, \d1, \d2, \d3, \d4, \d5
.endm
.macro interleave_1 wd, r0, r1, r2, r3, r4
        trn1            \r0\wd, \r0\wd, \r1\wd
        trn1            \r1\wd, \r1\wd, \r2\wd
.ifnb \r3
        trn1            \r2\wd, \r2\wd, \r3\wd
        trn1            \r3\wd, \r3\wd, \r4\wd
.endif
.endm
.macro interleave_1_s r0, r1, r2, r3, r4
        interleave_1    .2s, \r0, \r1, \r2, \r3, \r4
.endm
.macro umin_h c, wd, r0, r1, r2, r3
        umin            \r0\wd,  \r0\wd,  \c\wd
.ifnb \r1
        umin            \r1\wd,  \r1\wd,  \c\wd
.endif
.ifnb \r2
        umin            \r2\wd,  \r2\wd,  \c\wd
        umin            \r3\wd,  \r3\wd,  \c\wd
.endif
.endm
.macro sub_h c, wd, r0, r1, r2, r3
        sub             \r0\wd,  \r0\wd,  \c\wd
.ifnb \r1
        sub             \r1\wd,  \r1\wd,  \c\wd
.endif
.ifnb \r2
        sub             \r2\wd,  \r2\wd,  \c\wd
        sub             \r3\wd,  \r3\wd,  \c\wd
.endif
.endm
.macro smull_smlal_4 d, s0, s1, s2, s3
        smull           \d\().4s,  \s0\().4h,  v0.h[0]
        smlal           \d\().4s,  \s1\().4h,  v0.h[1]
        smlal           \d\().4s,  \s2\().4h,  v0.h[2]
        smlal           \d\().4s,  \s3\().4h,  v0.h[3]
.endm
.macro smull2_smlal2_4 d, s0, s1, s2, s3
        smull2          \d\().4s,  \s0\().8h,  v0.h[0]
        smlal2          \d\().4s,  \s1\().8h,  v0.h[1]
        smlal2          \d\().4s,  \s2\().8h,  v0.h[2]
        smlal2          \d\().4s,  \s3\().8h,  v0.h[3]
.endm
.macro smull_smlal_8 d, s0, s1, s2, s3, s4, s5, s6, s7
        smull           \d\().4s,  \s0\().4h,  v0.h[0]
        smlal           \d\().4s,  \s1\().4h,  v0.h[1]
        smlal           \d\().4s,  \s2\().4h,  v0.h[2]
        smlal           \d\().4s,  \s3\().4h,  v0.h[3]
        smlal           \d\().4s,  \s4\().4h,  v0.h[4]
        smlal           \d\().4s,  \s5\().4h,  v0.h[5]
        smlal           \d\().4s,  \s6\().4h,  v0.h[6]
        smlal           \d\().4s,  \s7\().4h,  v0.h[7]
.endm
.macro smull2_smlal2_8 d, s0, s1, s2, s3, s4, s5, s6, s7
        smull2          \d\().4s,  \s0\().8h,  v0.h[0]
        smlal2          \d\().4s,  \s1\().8h,  v0.h[1]
        smlal2          \d\().4s,  \s2\().8h,  v0.h[2]
        smlal2          \d\().4s,  \s3\().8h,  v0.h[3]
        smlal2          \d\().4s,  \s4\().8h,  v0.h[4]
        smlal2          \d\().4s,  \s5\().8h,  v0.h[5]
        smlal2          \d\().4s,  \s6\().8h,  v0.h[6]
        smlal2          \d\().4s,  \s7\().8h,  v0.h[7]
.endm
.macro sqrshrun_h shift, r0, r1, r2, r3
        sqrshrun        \r0\().4h, \r0\().4s,  #\shift
.ifnb \r1
        sqrshrun2       \r0\().8h, \r1\().4s,  #\shift
.endif
.ifnb \r2
        sqrshrun        \r2\().4h, \r2\().4s,  #\shift
        sqrshrun2       \r2\().8h, \r3\().4s,  #\shift
.endif
.endm
.macro xtn_h r0, r1, r2, r3
        uzp1            \r0\().8h,  \r0\().8h,  \r1\().8h // Same as xtn, xtn2
.ifnb \r2
        uzp1            \r2\().8h,  \r2\().8h,  \r3\().8h // Ditto
.endif
.endm
.macro srshl_s shift, r0, r1, r2, r3
        srshl           \r0\().4s,  \r0\().4s,  \shift\().4s
        srshl           \r1\().4s,  \r1\().4s,  \shift\().4s
.ifnb \r2
        srshl           \r2\().4s,  \r2\().4s,  \shift\().4s
        srshl           \r3\().4s,  \r3\().4s,  \shift\().4s
.endif
.endm
.macro st_s strd, reg, lanes
        st1             {\reg\().s}[0], [x0], \strd
        st1             {\reg\().s}[1], [x9], \strd
.if \lanes > 2
        st1             {\reg\().s}[2], [x0], \strd
        st1             {\reg\().s}[3], [x9], \strd
.endif
.endm
.macro st_d strd, r0, r1
        st1             {\r0\().d}[0], [x0], \strd
        st1             {\r0\().d}[1], [x9], \strd
.ifnb \r1
        st1             {\r1\().d}[0], [x0], \strd
        st1             {\r1\().d}[1], [x9], \strd
.endif
.endm
.macro shift_store_4 type, strd, r0, r1, r2, r3
.ifc \type, put
        sqrshrun_h      6,   \r0, \r1, \r2, \r3
        umin_h          v31, .8h, \r0, \r2
.else
        srshl_s         v30, \r0, \r1, \r2, \r3  // -(6-intermediate_bits)
        xtn_h           \r0, \r1, \r2, \r3
        sub_h           v29, .8h, \r0, \r2       // PREP_BIAS
.endif
        st_d            \strd, \r0, \r2
.endm
.macro st_reg strd, wd, r0, r1, r2, r3, r4, r5, r6, r7
        st1             {\r0\wd}, [x0], \strd
        st1             {\r1\wd}, [x9], \strd
.ifnb \r2
        st1             {\r2\wd}, [x0], \strd
        st1             {\r3\wd}, [x9], \strd
.endif
.ifnb \r4
        st1             {\r4\wd}, [x0], \strd
        st1             {\r5\wd}, [x9], \strd
        st1             {\r6\wd}, [x0], \strd
        st1             {\r7\wd}, [x9], \strd
.endif
.endm
.macro st_8h strd, r0, r1, r2, r3, r4, r5, r6, r7
        st_reg          \strd, .8h, \r0, \r1, \r2, \r3, \r4, \r5, \r6, \r7
.endm
.macro shift_store_8 type, strd, r0, r1, r2, r3
.ifc \type, put
        sqrshrun_h      6,   \r0, \r1, \r2, \r3
        umin_h          v31, .8h, \r0, \r2
.else
        srshl_s         v30, \r0, \r1, \r2, \r3  // -(6-intermediate_bits)
        xtn_h           \r0, \r1, \r2, \r3
        sub_h           v29, .8h, \r0, \r2       // PREP_BIAS
.endif
        st_8h           \strd, \r0, \r2
.endm
.macro shift_store_16 type, strd, dst, r0, r1, r2, r3
.ifc \type, put
        sqrshrun_h      6,   \r0, \r1, \r2, \r3
        umin            \r0\().8h, \r0\().8h, v31.8h
        umin            \r1\().8h, \r2\().8h, v31.8h
.else
        srshl_s         v30, \r0, \r1, \r2, \r3  // -(6-intermediate_bits)
        xtn_h           \r0, \r1, \r2, \r3
        sub             \r0\().8h, \r0\().8h, v29.8h
        sub             \r1\().8h, \r2\().8h, v29.8h
.endif
        st1             {\r0\().8h, \r1\().8h}, [\dst], \strd
.endm

.macro make_8tap_fn op, type, type_h, type_v
function \op\()_8tap_\type\()_16bpc_neon, export=1
        mov             w9,  \type_h
        mov             w10, \type_v
        b               \op\()_8tap_neon
endfunc
.endm

// No spaces in these expressions, due to gas-preprocessor.
#define REGULAR ((0*15<<7)|3*15)
#define SMOOTH  ((1*15<<7)|4*15)
#define SHARP   ((2*15<<7)|3*15)

.macro filter_fn type, dst, d_strd, src, s_strd, w, h, mx, xmx, my, xmy, bdmax, ds2, sr2
make_8tap_fn \type, regular,        REGULAR, REGULAR
make_8tap_fn \type, regular_smooth, REGULAR, SMOOTH
make_8tap_fn \type, regular_sharp,  REGULAR, SHARP
make_8tap_fn \type, smooth,         SMOOTH,  SMOOTH
make_8tap_fn \type, smooth_regular, SMOOTH,  REGULAR
make_8tap_fn \type, smooth_sharp,   SMOOTH,  SHARP
make_8tap_fn \type, sharp,          SHARP,   SHARP
make_8tap_fn \type, sharp_regular,  SHARP,   REGULAR
make_8tap_fn \type, sharp_smooth,   SHARP,   SMOOTH

function \type\()_8tap_neon
.ifc \bdmax, w8
        ldr             w8,  [sp]
.endif
        mov             w11,  #0x4081  // (1 << 14) | (1 << 7) | (1 << 0)
        mul             \mx,  \mx, w11
        mul             \my,  \my, w11
        add             \mx,  \mx, w9  // mx, 8tap_h, 4tap_h
        add             \my,  \my, w10 // my, 8tap_v, 4tap_v
.ifc \type, prep
        uxtw            \d_strd, \w
        lsl             \d_strd, \d_strd, #1
.endif

        dup             v31.8h,  \bdmax        // bitdepth_max
        clz             \bdmax,  \bdmax
        clz             w9,  \w
        sub             \bdmax,  \bdmax,  #18  // intermediate_bits = clz(bitdepth_max) - 18
        mov             w12, #6
        tst             \mx, #(0x7f << 14)
        sub             w9,  w9,  #24
        add             w13, w12, \bdmax       // 6 + intermediate_bits
        sub             w12, w12, \bdmax       // 6 - intermediate_bits
        movrel          x11, X(mc_subpel_filters), -8
        b.ne            L(\type\()_8tap_h)
        tst             \my, #(0x7f << 14)
        b.ne            L(\type\()_8tap_v)
        b               \type\()_neon

L(\type\()_8tap_h):
        cmp             \w,   #4
        ubfx            w10,  \mx, #7, #7
        and             \mx,  \mx, #0x7f
        b.le            4f
        mov             \mx,  w10
4:
        tst             \my,  #(0x7f << 14)
        add             \xmx, x11, \mx, uxtw #3
        b.ne            L(\type\()_8tap_hv)

        adr             x10, L(\type\()_8tap_h_tbl)
        dup             v30.4s,  w12           // 6 - intermediate_bits
        ldrh            w9,  [x10, x9, lsl #1]
        neg             v30.4s,  v30.4s        // -(6-intermediate_bits)
.ifc \type, put
        dup             v29.8h,  \bdmax        // intermediate_bits
.else
        movi            v28.8h,  #(PREP_BIAS >> 8), lsl #8
.endif
        sub             x10, x10, w9, uxtw
.ifc \type, put
        neg             v29.8h,  v29.8h        // -intermediate_bits
.endif
        br              x10

20:     // 2xN h
        AARCH64_VALID_JUMP_TARGET
.ifc \type, put
        add             \xmx,  \xmx,  #2
        ld1             {v0.s}[0], [\xmx]
        sub             \src,  \src,  #2
        add             \ds2,  \dst,  \d_strd
        add             \sr2,  \src,  \s_strd
        lsl             \d_strd,  \d_strd,  #1
        lsl             \s_strd,  \s_strd,  #1
        sxtl            v0.8h,   v0.8b
2:
        ld1             {v4.8h},  [\src], \s_strd
        ld1             {v6.8h},  [\sr2], \s_strd
        ext             v5.16b,  v4.16b,  v4.16b,  #2
        ext             v7.16b,  v6.16b,  v6.16b,  #2
        subs            \h,  \h,  #2
        trn1            v3.2s,   v4.2s,   v6.2s
        trn2            v6.2s,   v4.2s,   v6.2s
        trn1            v4.2s,   v5.2s,   v7.2s
        trn2            v7.2s,   v5.2s,   v7.2s
        smull           v3.4s,   v3.4h,   v0.h[0]
        smlal           v3.4s,   v4.4h,   v0.h[1]
        smlal           v3.4s,   v6.4h,   v0.h[2]
        smlal           v3.4s,   v7.4h,   v0.h[3]
        srshl           v3.4s,   v3.4s,   v30.4s // -(6-intermediate_bits)
        sqxtun          v3.4h,   v3.4s
        srshl           v3.4h,   v3.4h,   v29.4h // -intermediate_bits
        umin            v3.4h,   v3.4h,   v31.4h
        st1             {v3.s}[0], [\dst], \d_strd
        st1             {v3.s}[1], [\ds2], \d_strd
        b.gt            2b
        ret
.endif

40:     // 4xN h
        AARCH64_VALID_JUMP_TARGET
        add             \xmx,  \xmx,  #2
        ld1             {v0.s}[0], [\xmx]
        sub             \src,  \src,  #2
        add             \ds2,  \dst,  \d_strd
        add             \sr2,  \src,  \s_strd
        lsl             \d_strd,  \d_strd,  #1
        lsl             \s_strd,  \s_strd,  #1
        sxtl            v0.8h,   v0.8b
4:
        ld1             {v16.8h}, [\src], \s_strd
        ld1             {v20.8h}, [\sr2], \s_strd
        ext             v17.16b, v16.16b, v16.16b, #2
        ext             v18.16b, v16.16b, v16.16b, #4
        ext             v19.16b, v16.16b, v16.16b, #6
        ext             v21.16b, v20.16b, v20.16b, #2
        ext             v22.16b, v20.16b, v20.16b, #4
        ext             v23.16b, v20.16b, v20.16b, #6
        subs            \h,  \h,  #2
        smull           v16.4s,  v16.4h,  v0.h[0]
        smlal           v16.4s,  v17.4h,  v0.h[1]
        smlal           v16.4s,  v18.4h,  v0.h[2]
        smlal           v16.4s,  v19.4h,  v0.h[3]
        smull           v20.4s,  v20.4h,  v0.h[0]
        smlal           v20.4s,  v21.4h,  v0.h[1]
        smlal           v20.4s,  v22.4h,  v0.h[2]
        smlal           v20.4s,  v23.4h,  v0.h[3]
        srshl           v16.4s,  v16.4s,  v30.4s // -(6-intermediate_bits)
        srshl           v20.4s,  v20.4s,  v30.4s // -(6-intermediate_bits)
.ifc \type, put
        sqxtun          v16.4h,  v16.4s
        sqxtun2         v16.8h,  v20.4s
        srshl           v16.8h,  v16.8h,  v29.8h // -intermediate_bits
        umin            v16.8h,  v16.8h,  v31.8h
.else
        uzp1            v16.8h,  v16.8h,  v20.8h // Same as xtn, xtn2
        sub             v16.8h,  v16.8h,  v28.8h // PREP_BIAS
.endif
        st1             {v16.d}[0], [\dst], \d_strd
        st1             {v16.d}[1], [\ds2], \d_strd
        b.gt            4b
        ret

80:
160:
320:
640:
1280:   // 8xN, 16xN, 32xN, ... h
        AARCH64_VALID_JUMP_TARGET
        ld1             {v0.8b}, [\xmx]
        sub             \src,  \src,  #6
        add             \ds2,  \dst,  \d_strd
        add             \sr2,  \src,  \s_strd
        lsl             \s_strd,  \s_strd,  #1
        sxtl            v0.8h,   v0.8b

        sub             \s_strd,  \s_strd,  \w, uxtw #1
        sub             \s_strd,  \s_strd,  #16
.ifc \type, put
        lsl             \d_strd,  \d_strd,  #1
        sub             \d_strd,  \d_strd,  \w, uxtw #1
.endif
81:
        ld1             {v16.8h, v17.8h},  [\src], #32
        ld1             {v20.8h, v21.8h},  [\sr2], #32
        mov             \mx, \w

8:
        smull           v18.4s,  v16.4h,  v0.h[0]
        smull2          v19.4s,  v16.8h,  v0.h[0]
        smull           v22.4s,  v20.4h,  v0.h[0]
        smull2          v23.4s,  v20.8h,  v0.h[0]
.irpc i, 1234567
        ext             v24.16b, v16.16b, v17.16b, #(2*\i)
        ext             v25.16b, v20.16b, v21.16b, #(2*\i)
        smlal           v18.4s,  v24.4h,  v0.h[\i]
        smlal2          v19.4s,  v24.8h,  v0.h[\i]
        smlal           v22.4s,  v25.4h,  v0.h[\i]
        smlal2          v23.4s,  v25.8h,  v0.h[\i]
.endr
        subs            \mx, \mx, #8
        srshl           v18.4s,  v18.4s,  v30.4s // -(6-intermediate_bits)
        srshl           v19.4s,  v19.4s,  v30.4s // -(6-intermediate_bits)
        srshl           v22.4s,  v22.4s,  v30.4s // -(6-intermediate_bits)
        srshl           v23.4s,  v23.4s,  v30.4s // -(6-intermediate_bits)
.ifc \type, put
        sqxtun          v18.4h,  v18.4s
        sqxtun2         v18.8h,  v19.4s
        sqxtun          v22.4h,  v22.4s
        sqxtun2         v22.8h,  v23.4s
        srshl           v18.8h,  v18.8h,  v29.8h // -intermediate_bits
        srshl           v22.8h,  v22.8h,  v29.8h // -intermediate_bits
        umin            v18.8h,  v18.8h,  v31.8h
        umin            v22.8h,  v22.8h,  v31.8h
.else
        uzp1            v18.8h,  v18.8h,  v19.8h // Same as xtn, xtn2
        uzp1            v22.8h,  v22.8h,  v23.8h // Ditto
        sub             v18.8h,  v18.8h,  v28.8h // PREP_BIAS
        sub             v22.8h,  v22.8h,  v28.8h // PREP_BIAS
.endif
        st1             {v18.8h}, [\dst], #16
        st1             {v22.8h}, [\ds2], #16
        b.le            9f

        mov             v16.16b, v17.16b
        mov             v20.16b, v21.16b
        ld1             {v17.8h}, [\src], #16
        ld1             {v21.8h}, [\sr2], #16
        b               8b

9:
        add             \dst,  \dst,  \d_strd
        add             \ds2,  \ds2,  \d_strd
        add             \src,  \src,  \s_strd
        add             \sr2,  \sr2,  \s_strd

        subs            \h,  \h,  #2
        b.gt            81b
        ret

L(\type\()_8tap_h_tbl):
        .hword L(\type\()_8tap_h_tbl) - 1280b
        .hword L(\type\()_8tap_h_tbl) -  640b
        .hword L(\type\()_8tap_h_tbl) -  320b
        .hword L(\type\()_8tap_h_tbl) -  160b
        .hword L(\type\()_8tap_h_tbl) -   80b
        .hword L(\type\()_8tap_h_tbl) -   40b
        .hword L(\type\()_8tap_h_tbl) -   20b
        .hword 0


L(\type\()_8tap_v):
        cmp             \h,  #4
        ubfx            w10, \my, #7, #7
        and             \my, \my, #0x7f
        b.le            4f
        mov             \my, w10
4:
        add             \xmy, x11, \my, uxtw #3

.ifc \type, prep
        dup             v30.4s,  w12           // 6 - intermediate_bits
        movi            v29.8h,  #(PREP_BIAS >> 8), lsl #8
.endif
        adr             x10, L(\type\()_8tap_v_tbl)
        ldrh            w9,  [x10, x9, lsl #1]
.ifc \type, prep
        neg             v30.4s,  v30.4s        // -(6-intermediate_bits)
.endif
        sub             x10, x10, w9, uxtw
        br              x10

20:     // 2xN v
        AARCH64_VALID_JUMP_TARGET
.ifc \type, put
        b.gt            28f

        cmp             \h,  #2
        add             \xmy, \xmy, #2
        ld1             {v0.s}[0], [\xmy]
        sub             \src,  \src,  \s_strd
        add             \ds2,  \dst,  \d_strd
        add             \sr2,  \src,  \s_strd
        lsl             \s_strd,  \s_strd,  #1
        lsl             \d_strd,  \d_strd,  #1
        sxtl            v0.8h,   v0.8b

        // 2x2 v
        load_s          \src, \sr2, \s_strd, v1, v2, v3, v4, v5
        interleave_1_s  v1,  v2,  v3,  v4,  v5
        b.gt            24f
        smull_smlal_4   v6,  v1,  v2,  v3,  v4
        sqrshrun_h      6,   v6
        umin_h          v31, .8h, v6
        st_s            \d_strd, v6, 2
        ret

24:     // 2x4 v
        load_s          \sr2, \src, \s_strd, v6, v7
        interleave_1_s  v5,  v6,  v7
        smull_smlal_4   v16, v1,  v2,  v3,  v4
        smull_smlal_4   v17, v3,  v4,  v5,  v6
        sqrshrun_h      6,   v16, v17
        umin_h          v31, .8h, v16
        st_s            \d_strd, v16, 4
        ret

28:     // 2x6, 2x8, 2x12, 2x16 v
        ld1             {v0.8b}, [\xmy]
        sub             \sr2,  \src,  \s_strd, lsl #1
        add             \ds2,  \dst,  \d_strd
        sub             \src,  \sr2,  \s_strd
        lsl             \d_strd,  \d_strd,  #1
        lsl             \s_strd,  \s_strd,  #1
        sxtl            v0.8h,   v0.8b

        load_s          \src, \sr2, \s_strd, v1,  v2,  v3,  v4, v5, v6, v7
        interleave_1_s  v1,  v2,  v3,  v4,  v5
        interleave_1_s  v5,  v6,  v7
216:
        subs            \h,  \h,  #4
        load_s          \sr2, \src, \s_strd, v16, v17, v18, v19
        interleave_1_s  v7,  v16, v17, v18, v19
        smull_smlal_8   v24, v1,  v2,  v3,  v4,  v5,  v6,  v7,  v16
        smull_smlal_8   v25, v3,  v4,  v5,  v6,  v7,  v16, v17, v18
        sqrshrun_h      6,   v24, v25
        umin_h          v31, .8h, v24
        st_s            \d_strd, v24, 4
        b.le            0f
        cmp             \h,  #2
        mov             v1.16b,  v5.16b
        mov             v2.16b,  v6.16b
        mov             v3.16b,  v7.16b
        mov             v4.16b,  v16.16b
        mov             v5.16b,  v17.16b
        mov             v6.16b,  v18.16b
        mov             v7.16b,  v19.16b
        b.eq            26f
        b               216b
26:
        load_s          \sr2, \src, \s_strd, v16, v17
        interleave_1_s  v7,  v16, v17
        smull_smlal_8   v24, v1,  v2,  v3,  v4,  v5,  v6,  v7,  v16
        sqrshrun_h      6,   v24
        umin_h          v31, .4h, v24
        st_s            \d_strd, v24, 2
0:
        ret
.endif

40:
        AARCH64_VALID_JUMP_TARGET
        b.gt            480f

        // 4x2, 4x4 v
        cmp             \h,  #2
        add             \xmy, \xmy, #2
        ld1             {v0.s}[0], [\xmy]
        sub             \src, \src, \s_strd
        add             \ds2, \dst, \d_strd
        add             \sr2, \src, \s_strd
        lsl             \s_strd, \s_strd, #1
        lsl             \d_strd, \d_strd, #1
        sxtl            v0.8h,   v0.8b

        load_4h         \src, \sr2, \s_strd, v1, v2, v3, v4, v5
        smull_smlal_4   v6,  v1,  v2,  v3,  v4
        smull_smlal_4   v7,  v2,  v3,  v4,  v5
        shift_store_4   \type, \d_strd, v6, v7
        b.le            0f
        load_4h         \sr2, \src, \s_strd, v6, v7
        smull_smlal_4   v1,  v3,  v4,  v5,  v6
        smull_smlal_4   v2,  v4,  v5,  v6,  v7
        shift_store_4   \type, \d_strd, v1, v2
0:
        ret

480:    // 4x6, 4x8, 4x12, 4x16 v
        ld1             {v0.8b}, [\xmy]
        sub             \sr2, \src, \s_strd, lsl #1
        add             \ds2, \dst, \d_strd
        sub             \src, \sr2, \s_strd
        lsl             \s_strd, \s_strd, #1
        lsl             \d_strd, \d_strd, #1
        sxtl            v0.8h,   v0.8b

        load_4h         \src, \sr2, \s_strd, v16, v17, v18, v19, v20, v21, v22

48:
        subs            \h,  \h,  #4
        load_4h         \sr2, \src, \s_strd, v23, v24, v25, v26
        smull_smlal_8   v1,  v16, v17, v18, v19, v20, v21, v22, v23
        smull_smlal_8   v2,  v17, v18, v19, v20, v21, v22, v23, v24
        smull_smlal_8   v3,  v18, v19, v20, v21, v22, v23, v24, v25
        smull_smlal_8   v4,  v19, v20, v21, v22, v23, v24, v25, v26
        shift_store_4   \type, \d_strd, v1, v2, v3, v4
        b.le            0f
        cmp             \h,  #2
        mov             v16.8b,  v20.8b
        mov             v17.8b,  v21.8b
        mov             v18.8b,  v22.8b
        mov             v19.8b,  v23.8b
        mov             v20.8b,  v24.8b
        mov             v21.8b,  v25.8b
        mov             v22.8b,  v26.8b
        b.eq            46f
        b               48b
46:
        load_4h         \sr2, \src, \s_strd, v23, v24
        smull_smlal_8   v1,  v16, v17, v18, v19, v20, v21, v22, v23
        smull_smlal_8   v2,  v17, v18, v19, v20, v21, v22, v23, v24
        shift_store_4   \type, \d_strd, v1, v2
0:
        ret

80:
        AARCH64_VALID_JUMP_TARGET
        b.gt            880f

        // 8x2, 8x4 v
        cmp             \h,  #2
        add             \xmy, \xmy, #2
        ld1             {v0.s}[0], [\xmy]
        sub             \src, \src, \s_strd
        add             \ds2, \dst, \d_strd
        add             \sr2, \src, \s_strd
        lsl             \s_strd, \s_strd, #1
        lsl             \d_strd, \d_strd, #1
        sxtl            v0.8h,   v0.8b

        load_8h         \src, \sr2, \s_strd, v1, v2, v3, v4, v5
        smull_smlal_4   v16, v1,  v2,  v3,  v4
        smull2_smlal2_4 v17, v1,  v2,  v3,  v4
        smull_smlal_4   v18, v2,  v3,  v4,  v5
        smull2_smlal2_4 v19, v2,  v3,  v4,  v5
        shift_store_8   \type, \d_strd, v16, v17, v18, v19
        b.le            0f
        load_8h         \sr2, \src, \s_strd, v6, v7
        smull_smlal_4   v16, v3,  v4,  v5,  v6
        smull2_smlal2_4 v17, v3,  v4,  v5,  v6
        smull_smlal_4   v18, v4,  v5,  v6,  v7
        smull2_smlal2_4 v19, v4,  v5,  v6,  v7
        shift_store_8   \type, \d_strd, v16, v17, v18, v19
0:
        ret

880:    // 8x6, 8x8, 8x16, 8x32 v
1680:   // 16x8, 16x16, ...
320:    // 32x8, 32x16, ...
640:
1280:
        AARCH64_VALID_JUMP_TARGET
        ld1             {v0.8b}, [\xmy]
        sub             \src, \src, \s_strd
        sub             \src, \src, \s_strd, lsl #1
        sxtl            v0.8h,   v0.8b
        mov             \my,  \h
168:
        add             \ds2, \dst, \d_strd
        add             \sr2, \src, \s_strd
        lsl             \s_strd, \s_strd, #1
        lsl             \d_strd, \d_strd, #1

        load_8h         \src, \sr2, \s_strd, v16, v17, v18, v19, v20, v21, v22

88:
        subs            \h,  \h,  #2
        load_8h         \sr2, \src, \s_strd, v23, v24
        smull_smlal_8   v1,  v16, v17, v18, v19, v20, v21, v22, v23
        smull2_smlal2_8 v2,  v16, v17, v18, v19, v20, v21, v22, v23
        smull_smlal_8   v3,  v17, v18, v19, v20, v21, v22, v23, v24
        smull2_smlal2_8 v4,  v17, v18, v19, v20, v21, v22, v23, v24
        shift_store_8   \type, \d_strd, v1, v2, v3, v4
        b.le            9f
        subs            \h,  \h,  #2
        load_8h         \sr2, \src, \s_strd, v25, v26
        smull_smlal_8   v1,  v18, v19, v20, v21, v22, v23, v24, v25
        smull2_smlal2_8 v2,  v18, v19, v20, v21, v22, v23, v24, v25
        smull_smlal_8   v3,  v19, v20, v21, v22, v23, v24, v25, v26
        smull2_smlal2_8 v4,  v19, v20, v21, v22, v23, v24, v25, v26
        shift_store_8   \type, \d_strd, v1, v2, v3, v4
        b.le            9f
        mov             v16.16b, v20.16b
        mov             v17.16b, v21.16b
        mov             v18.16b, v22.16b
        mov             v19.16b, v23.16b
        mov             v20.16b, v24.16b
        mov             v21.16b, v25.16b
        mov             v22.16b, v26.16b
        b               88b
9:
        subs            \w,  \w,  #8
        b.le            0f
        asr             \s_strd, \s_strd, #1
        asr             \d_strd, \d_strd, #1
        msub            \src, \s_strd, \xmy, \src
        msub            \dst, \d_strd, \xmy, \dst
        sub             \src, \src, \s_strd, lsl #3
        mov             \h,  \my
        add             \src, \src, #16
        add             \dst, \dst, #16
        b               168b
0:
        ret

160:
        AARCH64_VALID_JUMP_TARGET
        b.gt            1680b

        // 16x2, 16x4 v
        add             \xmy, \xmy, #2
        ld1             {v0.s}[0], [\xmy]
        sub             \src, \src, \s_strd
        sxtl            v0.8h,   v0.8b

        load_16h        \src, \src, \s_strd, v16, v17, v18, v19, v20, v21
16:
        load_16h        \src, \src, \s_strd, v22, v23
        subs            \h,  \h,  #1
        smull_smlal_4   v1,  v16, v18, v20, v22
        smull2_smlal2_4 v2,  v16, v18, v20, v22
        smull_smlal_4   v3,  v17, v19, v21, v23
        smull2_smlal2_4 v4,  v17, v19, v21, v23
        shift_store_16  \type, \d_strd, x0, v1, v2, v3, v4
        b.le            0f
        mov             v16.16b, v18.16b
        mov             v17.16b, v19.16b
        mov             v18.16b, v20.16b
        mov             v19.16b, v21.16b
        mov             v20.16b, v22.16b
        mov             v21.16b, v23.16b
        b               16b
0:
        ret

L(\type\()_8tap_v_tbl):
        .hword L(\type\()_8tap_v_tbl) - 1280b
        .hword L(\type\()_8tap_v_tbl) -  640b
        .hword L(\type\()_8tap_v_tbl) -  320b
        .hword L(\type\()_8tap_v_tbl) -  160b
        .hword L(\type\()_8tap_v_tbl) -   80b
        .hword L(\type\()_8tap_v_tbl) -   40b
        .hword L(\type\()_8tap_v_tbl) -   20b
        .hword 0

L(\type\()_8tap_hv):
        cmp             \h,  #4
        ubfx            w10, \my, #7, #7
        and             \my, \my, #0x7f
        b.le            4f
        mov             \my,  w10
4:
        add             \xmy, x11, \my, uxtw #3

        adr             x10, L(\type\()_8tap_hv_tbl)
        dup             v30.4s,  w12           // 6 - intermediate_bits
        ldrh            w9,  [x10, x9, lsl #1]
        neg             v30.4s,  v30.4s        // -(6-intermediate_bits)
.ifc \type, put
        dup             v29.4s,  w13           // 6 + intermediate_bits
.else
        movi            v29.8h,  #(PREP_BIAS >> 8), lsl #8
.endif
        sub             x10, x10, w9, uxtw
.ifc \type, put
        neg             v29.4s,  v29.4s        // -(6+intermediate_bits)
.endif
        br              x10

20:
        AARCH64_VALID_JUMP_TARGET
.ifc \type, put
        add             \xmx,  \xmx,  #2
        ld1             {v0.s}[0],  [\xmx]
        b.gt            280f
        add             \xmy,  \xmy,  #2
        ld1             {v1.s}[0],  [\xmy]

        // 2x2, 2x4 hv
        sub             \sr2, \src, #2
        sub             \src, \sr2, \s_strd
        add             \ds2, \dst, \d_strd
        lsl             \s_strd, \s_strd, #1
        lsl             \d_strd, \d_strd, #1
        sxtl            v0.8h,   v0.8b
        sxtl            v1.8h,   v1.8b
        mov             x15, x30

        ld1             {v27.8h}, [\src], \s_strd
        ext             v28.16b, v27.16b, v27.16b, #2
        smull           v27.4s,  v27.4h,  v0.4h
        smull           v28.4s,  v28.4h,  v0.4h
        addp            v27.4s,  v27.4s,  v28.4s
        addp            v16.4s,  v27.4s,  v27.4s
        srshl           v16.2s,  v16.2s,  v30.2s // -(6-intermediate_bits)
        bl              L(\type\()_8tap_filter_2)
        // The intermediates from the horizontal pass fit in 16 bit without
        // any bias; we could just as well keep them as .4s, but narrowing
        // them to .4h gives a significant speedup on out of order cores
        // (at the cost of a smaller slowdown on in-order cores such as A53).
        xtn             v16.4h,  v16.4s

        trn1            v16.2s,  v16.2s,  v24.2s
        mov             v17.8b,  v24.8b

2:
        bl              L(\type\()_8tap_filter_2)

        ext             v18.8b,  v17.8b,  v24.8b,  #4
        smull           v2.4s,   v16.4h,  v1.h[0]
        smlal           v2.4s,   v17.4h,  v1.h[1]
        smlal           v2.4s,   v18.4h,  v1.h[2]
        smlal           v2.4s,   v24.4h,  v1.h[3]

        srshl           v2.4s,   v2.4s,   v29.4s // -(6+intermediate_bits)
        sqxtun          v2.4h,   v2.4s
        umin            v2.4h,   v2.4h,   v31.4h
        subs            \h,  \h,  #2
        st1             {v2.s}[0], [\dst], \d_strd
        st1             {v2.s}[1], [\ds2], \d_strd
        b.le            0f
        mov             v16.8b,  v18.8b
        mov             v17.8b,  v24.8b
        b               2b

280:    // 2x8, 2x16, 2x32 hv
        ld1             {v1.8b},  [\xmy]
        sub             \src, \src, #2
        sub             \sr2, \src, \s_strd, lsl #1
        sub             \src, \sr2, \s_strd
        add             \ds2, \dst, \d_strd
        lsl             \s_strd, \s_strd, #1
        lsl             \d_strd, \d_strd, #1
        sxtl            v0.8h,   v0.8b
        sxtl            v1.8h,   v1.8b
        mov             x15, x30

        ld1             {v27.8h}, [\src], \s_strd
        ext             v28.16b, v27.16b, v27.16b, #2
        smull           v27.4s,  v27.4h,  v0.4h
        smull           v28.4s,  v28.4h,  v0.4h
        addp            v27.4s,  v27.4s,  v28.4s
        addp            v16.4s,  v27.4s,  v27.4s
        srshl           v16.2s,  v16.2s,  v30.2s // -(6-intermediate_bits)
        // The intermediates from the horizontal pass fit in 16 bit without
        // any bias; we could just as well keep them as .4s, but narrowing
        // them to .4h gives a significant speedup on out of order cores
        // (at the cost of a smaller slowdown on in-order cores such as A53).

        bl              L(\type\()_8tap_filter_2)
        xtn             v16.4h,  v16.4s
        trn1            v16.2s,  v16.2s,  v24.2s
        mov             v17.8b,  v24.8b
        bl              L(\type\()_8tap_filter_2)
        ext             v18.8b,  v17.8b,  v24.8b,  #4
        mov             v19.8b,  v24.8b
        bl              L(\type\()_8tap_filter_2)
        ext             v20.8b,  v19.8b,  v24.8b,  #4
        mov             v21.8b,  v24.8b

28:
        bl              L(\type\()_8tap_filter_2)
        ext             v22.8b,  v21.8b,  v24.8b,  #4
        smull           v3.4s,   v16.4h,  v1.h[0]
        smlal           v3.4s,   v17.4h,  v1.h[1]
        smlal           v3.4s,   v18.4h,  v1.h[2]
        smlal           v3.4s,   v19.4h,  v1.h[3]
        smlal           v3.4s,   v20.4h,  v1.h[4]
        smlal           v3.4s,   v21.4h,  v1.h[5]
        smlal           v3.4s,   v22.4h,  v1.h[6]
        smlal           v3.4s,   v24.4h,  v1.h[7]

        srshl           v3.4s,   v3.4s,   v29.4s // -(6+intermediate_bits)
        sqxtun          v3.4h,   v3.4s
        umin            v3.4h,   v3.4h,   v31.4h
        subs            \h,  \h,  #2
        st1             {v3.s}[0], [\dst], \d_strd
        st1             {v3.s}[1], [\ds2], \d_strd
        b.le            0f
        mov             v16.8b,  v18.8b
        mov             v17.8b,  v19.8b
        mov             v18.8b,  v20.8b
        mov             v19.8b,  v21.8b
        mov             v20.8b,  v22.8b
        mov             v21.8b,  v24.8b
        b               28b

0:
        ret             x15

L(\type\()_8tap_filter_2):
        ld1             {v25.8h},  [\sr2], \s_strd
        ld1             {v27.8h},  [\src], \s_strd
        ext             v26.16b, v25.16b, v25.16b, #2
        ext             v28.16b, v27.16b, v27.16b, #2
        trn1            v24.2s,  v25.2s,  v27.2s
        trn2            v27.2s,  v25.2s,  v27.2s
        trn1            v25.2s,  v26.2s,  v28.2s
        trn2            v28.2s,  v26.2s,  v28.2s
        smull           v24.4s,  v24.4h,  v0.h[0]
        smlal           v24.4s,  v25.4h,  v0.h[1]
        smlal           v24.4s,  v27.4h,  v0.h[2]
        smlal           v24.4s,  v28.4h,  v0.h[3]
        srshl           v24.4s,  v24.4s,  v30.4s // -(6-intermediate_bits)
        xtn             v24.4h,  v24.4s
        ret
.endif

40:
        AARCH64_VALID_JUMP_TARGET
        add             \xmx, \xmx, #2
        ld1             {v0.s}[0],  [\xmx]
        b.gt            480f
        add             \xmy, \xmy,  #2
        ld1             {v1.s}[0],  [\xmy]
        sub             \sr2, \src, #2
        sub             \src, \sr2, \s_strd
        add             \ds2, \dst, \d_strd
        lsl             \s_strd, \s_strd, #1
        lsl             \d_strd, \d_strd, #1
        sxtl            v0.8h,   v0.8b
        sxtl            v1.8h,   v1.8b
        mov             x15, x30

        // 4x2, 4x4 hv
        ld1             {v25.8h}, [\src], \s_strd
        ext             v26.16b, v25.16b, v25.16b, #2
        ext             v27.16b, v25.16b, v25.16b, #4
        ext             v28.16b, v25.16b, v25.16b, #6
        smull           v25.4s,  v25.4h,  v0.h[0]
        smlal           v25.4s,  v26.4h,  v0.h[1]
        smlal           v25.4s,  v27.4h,  v0.h[2]
        smlal           v25.4s,  v28.4h,  v0.h[3]
        srshl           v16.4s,  v25.4s,  v30.4s // -(6-intermediate_bits)
        // The intermediates from the horizontal pass fit in 16 bit without
        // any bias; we could just as well keep them as .4s, but narrowing
        // them to .4h gives a significant speedup on out of order cores
        // (at the cost of a smaller slowdown on in-order cores such as A53).
        xtn             v16.4h,  v16.4s

        bl              L(\type\()_8tap_filter_4)
        mov             v17.8b,  v24.8b
        mov             v18.8b,  v25.8b

4:
        bl              L(\type\()_8tap_filter_4)
        smull           v2.4s,   v16.4h,  v1.h[0]
        smlal           v2.4s,   v17.4h,  v1.h[1]
        smlal           v2.4s,   v18.4h,  v1.h[2]
        smlal           v2.4s,   v24.4h,  v1.h[3]
        smull           v3.4s,   v17.4h,  v1.h[0]
        smlal           v3.4s,   v18.4h,  v1.h[1]
        smlal           v3.4s,   v24.4h,  v1.h[2]
        smlal           v3.4s,   v25.4h,  v1.h[3]
.ifc \type, put
        srshl           v2.4s,   v2.4s,   v29.4s // -(6+intermediate_bits)
        srshl           v3.4s,   v3.4s,   v29.4s // -(6+intermediate_bits)
        sqxtun          v2.4h,   v2.4s
        sqxtun2         v2.8h,   v3.4s
        umin            v2.8h,   v2.8h,   v31.8h
.else
        rshrn           v2.4h,   v2.4s,   #6
        rshrn2          v2.8h,   v3.4s,   #6
        sub             v2.8h,   v2.8h,   v29.8h // PREP_BIAS
.endif
        subs            \h,  \h,  #2

        st1             {v2.d}[0], [\dst], \d_strd
        st1             {v2.d}[1], [\ds2], \d_strd
        b.le            0f
        mov             v16.8b,  v18.8b
        mov             v17.8b,  v24.8b
        mov             v18.8b,  v25.8b
        b               4b

480:    // 4x8, 4x16, 4x32 hv
        ld1             {v1.8b},  [\xmy]
        sub             \src, \src, #2
        sub             \sr2, \src, \s_strd, lsl #1
        sub             \src, \sr2, \s_strd
        add             \ds2, \dst, \d_strd
        lsl             \s_strd, \s_strd, #1
        lsl             \d_strd, \d_strd, #1
        sxtl            v0.8h,   v0.8b
        sxtl            v1.8h,   v1.8b
        mov             x15, x30

        ld1             {v25.8h}, [\src], \s_strd
        ext             v26.16b, v25.16b, v25.16b, #2
        ext             v27.16b, v25.16b, v25.16b, #4
        ext             v28.16b, v25.16b, v25.16b, #6
        smull           v25.4s,  v25.4h,  v0.h[0]
        smlal           v25.4s,  v26.4h,  v0.h[1]
        smlal           v25.4s,  v27.4h,  v0.h[2]
        smlal           v25.4s,  v28.4h,  v0.h[3]
        srshl           v16.4s,  v25.4s,  v30.4s // -(6-intermediate_bits)
        // The intermediates from the horizontal pass fit in 16 bit without
        // any bias; we could just as well keep them as .4s, but narrowing
        // them to .4h gives a significant speedup on out of order cores
        // (at the cost of a smaller slowdown on in-order cores such as A53).
        xtn             v16.4h,  v16.4s

        bl              L(\type\()_8tap_filter_4)
        mov             v17.8b,  v24.8b
        mov             v18.8b,  v25.8b
        bl              L(\type\()_8tap_filter_4)
        mov             v19.8b,  v24.8b
        mov             v20.8b,  v25.8b
        bl              L(\type\()_8tap_filter_4)
        mov             v21.8b,  v24.8b
        mov             v22.8b,  v25.8b

48:
        bl              L(\type\()_8tap_filter_4)
        smull           v3.4s,   v16.4h,  v1.h[0]
        smlal           v3.4s,   v17.4h,  v1.h[1]
        smlal           v3.4s,   v18.4h,  v1.h[2]
        smlal           v3.4s,   v19.4h,  v1.h[3]
        smlal           v3.4s,   v20.4h,  v1.h[4]
        smlal           v3.4s,   v21.4h,  v1.h[5]
        smlal           v3.4s,   v22.4h,  v1.h[6]
        smlal           v3.4s,   v24.4h,  v1.h[7]
        smull           v4.4s,   v17.4h,  v1.h[0]
        smlal           v4.4s,   v18.4h,  v1.h[1]
        smlal           v4.4s,   v19.4h,  v1.h[2]
        smlal           v4.4s,   v20.4h,  v1.h[3]
        smlal           v4.4s,   v21.4h,  v1.h[4]
        smlal           v4.4s,   v22.4h,  v1.h[5]
        smlal           v4.4s,   v24.4h,  v1.h[6]
        smlal           v4.4s,   v25.4h,  v1.h[7]
.ifc \type, put
        srshl           v3.4s,   v3.4s,   v29.4s // -(6+intermediate_bits)
        srshl           v4.4s,   v4.4s,   v29.4s // -(6+intermediate_bits)
        sqxtun          v3.4h,   v3.4s
        sqxtun2         v3.8h,   v4.4s
        umin            v3.8h,   v3.8h,   v31.8h
.else
        rshrn           v3.4h,   v3.4s,   #6
        rshrn2          v3.8h,   v4.4s,   #6
        sub             v3.8h,   v3.8h,   v29.8h // PREP_BIAS
.endif
        subs            \h,  \h,  #2
        st1             {v3.d}[0], [\dst], \d_strd
        st1             {v3.d}[1], [\ds2], \d_strd
        b.le            0f
        mov             v16.8b,  v18.8b
        mov             v17.8b,  v19.8b
        mov             v18.8b,  v20.8b
        mov             v19.8b,  v21.8b
        mov             v20.8b,  v22.8b
        mov             v21.8b,  v24.8b
        mov             v22.8b,  v25.8b
        b               48b
0:
        ret             x15

L(\type\()_8tap_filter_4):
        ld1             {v24.8h}, [\sr2], \s_strd
        ld1             {v25.8h}, [\src], \s_strd
        ext             v26.16b, v24.16b, v24.16b, #2
        ext             v27.16b, v24.16b, v24.16b, #4
        ext             v28.16b, v24.16b, v24.16b, #6
        smull           v24.4s,  v24.4h,  v0.h[0]
        smlal           v24.4s,  v26.4h,  v0.h[1]
        smlal           v24.4s,  v27.4h,  v0.h[2]
        smlal           v24.4s,  v28.4h,  v0.h[3]
        ext             v26.16b, v25.16b, v25.16b, #2
        ext             v27.16b, v25.16b, v25.16b, #4
        ext             v28.16b, v25.16b, v25.16b, #6
        smull           v25.4s,  v25.4h,  v0.h[0]
        smlal           v25.4s,  v26.4h,  v0.h[1]
        smlal           v25.4s,  v27.4h,  v0.h[2]
        smlal           v25.4s,  v28.4h,  v0.h[3]
        srshl           v24.4s,  v24.4s,  v30.4s // -(6-intermediate_bits)
        srshl           v25.4s,  v25.4s,  v30.4s // -(6-intermediate_bits)
        xtn             v24.4h,  v24.4s
        xtn             v25.4h,  v25.4s
        ret

80:
160:
320:
        AARCH64_VALID_JUMP_TARGET
        b.gt            880f
        add             \xmy,  \xmy,  #2
        ld1             {v0.8b},  [\xmx]
        ld1             {v1.s}[0],  [\xmy]
        sub             \src,  \src,  #6
        sub             \src,  \src,  \s_strd
        sxtl            v0.8h,   v0.8b
        sxtl            v1.8h,   v1.8b
        mov             x15, x30
        mov             \my, \h

164:    // 8x2, 8x4, 16x2, 16x4, 32x2, 32x4 hv
        add             \ds2,  \dst,  \d_strd
        add             \sr2,  \src,  \s_strd
        lsl             \d_strd, \d_strd, #1
        lsl             \s_strd, \s_strd, #1

        ld1             {v27.8h, v28.8h},  [\src], \s_strd
        smull           v24.4s,  v27.4h,  v0.h[0]
        smull2          v25.4s,  v27.8h,  v0.h[0]
.irpc i, 1234567
        ext             v26.16b, v27.16b, v28.16b, #(2*\i)
        smlal           v24.4s,  v26.4h,  v0.h[\i]
        smlal2          v25.4s,  v26.8h,  v0.h[\i]
.endr
        srshl           v24.4s,  v24.4s,  v30.4s // -(6-intermediate_bits)
        srshl           v25.4s,  v25.4s,  v30.4s // -(6-intermediate_bits)
        // The intermediates from the horizontal pass fit in 16 bit without
        // any bias; we could just as well keep them as .4s, but narrowing
        // them to .4h gives a significant speedup on out of order cores
        // (at the cost of a smaller slowdown on in-order cores such as A53),
        // and conserves register space (no need to clobber v8-v15).
        uzp1            v16.8h,  v24.8h,  v25.8h // Same as xtn, xtn2

        bl              L(\type\()_8tap_filter_8)
        mov             v17.16b, v23.16b
        mov             v18.16b, v24.16b

8:
        smull           v2.4s,   v16.4h,  v1.h[0]
        smull2          v3.4s,   v16.8h,  v1.h[0]
        bl              L(\type\()_8tap_filter_8)
        smull           v4.4s,   v17.4h,  v1.h[0]
        smull2          v5.4s,   v17.8h,  v1.h[0]
        smlal           v2.4s,   v17.4h,  v1.h[1]
        smlal2          v3.4s,   v17.8h,  v1.h[1]
        smlal           v4.4s,   v18.4h,  v1.h[1]
        smlal2          v5.4s,   v18.8h,  v1.h[1]
        smlal           v2.4s,   v18.4h,  v1.h[2]
        smlal2          v3.4s,   v18.8h,  v1.h[2]
        smlal           v4.4s,   v23.4h,  v1.h[2]
        smlal2          v5.4s,   v23.8h,  v1.h[2]
        smlal           v2.4s,   v23.4h,  v1.h[3]
        smlal2          v3.4s,   v23.8h,  v1.h[3]
        smlal           v4.4s,   v24.4h,  v1.h[3]
        smlal2          v5.4s,   v24.8h,  v1.h[3]
.ifc \type, put
        srshl           v2.4s,   v2.4s,   v29.4s // -(6+intermediate_bits)
        srshl           v3.4s,   v3.4s,   v29.4s // -(6+intermediate_bits)
        srshl           v4.4s,   v4.4s,   v29.4s // -(6+intermediate_bits)
        srshl           v5.4s,   v5.4s,   v29.4s // -(6+intermediate_bits)
        sqxtun          v2.4h,   v2.4s
        sqxtun2         v2.8h,   v3.4s
        sqxtun          v3.4h,   v4.4s
        sqxtun2         v3.8h,   v5.4s
        umin            v2.8h,   v2.8h,   v31.8h
        umin            v3.8h,   v3.8h,   v31.8h
.else
        rshrn           v2.4h,   v2.4s,   #6
        rshrn2          v2.8h,   v3.4s,   #6
        rshrn           v3.4h,   v4.4s,   #6
        rshrn2          v3.8h,   v5.4s,   #6
        sub             v2.8h,   v2.8h,   v29.8h // PREP_BIAS
        sub             v3.8h,   v3.8h,   v29.8h // PREP_BIAS
.endif
        subs            \h,  \h,  #2
        st1             {v2.8h}, [\dst], \d_strd
        st1             {v3.8h}, [\ds2], \d_strd
        b.le            9f
        mov             v16.16b, v18.16b
        mov             v17.16b, v23.16b
        mov             v18.16b, v24.16b
        b               8b
9:
        subs            \w,  \w,  #8
        b.le            0f
        asr             \s_strd,  \s_strd,  #1
        asr             \d_strd,  \d_strd,  #1
        msub            \src,  \s_strd,  \xmy,  \src
        msub            \dst,  \d_strd,  \xmy,  \dst
        sub             \src,  \src,  \s_strd,  lsl #2
        mov             \h,  \my
        add             \src,  \src,  #16
        add             \dst,  \dst,  #16
        b               164b

880:    // 8x8, 8x16, ..., 16x8, ..., 32x8, ... hv
640:
1280:
        AARCH64_VALID_JUMP_TARGET
        ld1             {v0.8b},  [\xmx]
        ld1             {v1.8b},  [\xmy]
        sub             \src,  \src,  #6
        sub             \src,  \src,  \s_strd
        sub             \src,  \src,  \s_strd, lsl #1
        sxtl            v0.8h,   v0.8b
        sxtl            v1.8h,   v1.8b
        mov             x15, x30
        mov             \my, \h

168:
        add             \ds2,  \dst,  \d_strd
        add             \sr2,  \src,  \s_strd
        lsl             \d_strd, \d_strd, #1
        lsl             \s_strd, \s_strd, #1

        ld1             {v27.8h, v28.8h},  [\src], \s_strd
        smull           v24.4s,  v27.4h,  v0.h[0]
        smull2          v25.4s,  v27.8h,  v0.h[0]
.irpc i, 1234567
        ext             v26.16b, v27.16b, v28.16b, #(2*\i)
        smlal           v24.4s,  v26.4h,  v0.h[\i]
        smlal2          v25.4s,  v26.8h,  v0.h[\i]
.endr
        srshl           v24.4s,  v24.4s,  v30.4s // -(6-intermediate_bits)
        srshl           v25.4s,  v25.4s,  v30.4s // -(6-intermediate_bits)
        // The intermediates from the horizontal pass fit in 16 bit without
        // any bias; we could just as well keep them as .4s, but narrowing
        // them to .4h gives a significant speedup on out of order cores
        // (at the cost of a smaller slowdown on in-order cores such as A53),
        // and conserves register space (no need to clobber v8-v15).
        uzp1            v16.8h,  v24.8h,  v25.8h // Same as xtn, xtn2

        bl              L(\type\()_8tap_filter_8)
        mov             v17.16b, v23.16b
        mov             v18.16b, v24.16b
        bl              L(\type\()_8tap_filter_8)
        mov             v19.16b, v23.16b
        mov             v20.16b, v24.16b
        bl              L(\type\()_8tap_filter_8)
        mov             v21.16b, v23.16b
        mov             v22.16b, v24.16b

88:
        smull           v2.4s,   v16.4h,  v1.h[0]
        smull2          v3.4s,   v16.8h,  v1.h[0]
        bl              L(\type\()_8tap_filter_8)
        smull           v4.4s,   v17.4h,  v1.h[0]
        smull2          v5.4s,   v17.8h,  v1.h[0]
        smlal           v2.4s,   v17.4h,  v1.h[1]
        smlal2          v3.4s,   v17.8h,  v1.h[1]
        smlal           v4.4s,   v18.4h,  v1.h[1]
        smlal2          v5.4s,   v18.8h,  v1.h[1]
        smlal           v2.4s,   v18.4h,  v1.h[2]
        smlal2          v3.4s,   v18.8h,  v1.h[2]
        smlal           v4.4s,   v19.4h,  v1.h[2]
        smlal2          v5.4s,   v19.8h,  v1.h[2]
        smlal           v2.4s,   v19.4h,  v1.h[3]
        smlal2          v3.4s,   v19.8h,  v1.h[3]
        smlal           v4.4s,   v20.4h,  v1.h[3]
        smlal2          v5.4s,   v20.8h,  v1.h[3]
        smlal           v2.4s,   v20.4h,  v1.h[4]
        smlal2          v3.4s,   v20.8h,  v1.h[4]
        smlal           v4.4s,   v21.4h,  v1.h[4]
        smlal2          v5.4s,   v21.8h,  v1.h[4]
        smlal           v2.4s,   v21.4h,  v1.h[5]
        smlal2          v3.4s,   v21.8h,  v1.h[5]
        smlal           v4.4s,   v22.4h,  v1.h[5]
        smlal2          v5.4s,   v22.8h,  v1.h[5]
        smlal           v2.4s,   v22.4h,  v1.h[6]
        smlal2          v3.4s,   v22.8h,  v1.h[6]
        smlal           v4.4s,   v23.4h,  v1.h[6]
        smlal2          v5.4s,   v23.8h,  v1.h[6]
        smlal           v2.4s,   v23.4h,  v1.h[7]
        smlal2          v3.4s,   v23.8h,  v1.h[7]
        smlal           v4.4s,   v24.4h,  v1.h[7]
        smlal2          v5.4s,   v24.8h,  v1.h[7]
.ifc \type, put
        srshl           v2.4s,   v2.4s,   v29.4s // -(6+intermediate_bits)
        srshl           v3.4s,   v3.4s,   v29.4s // -(6+intermediate_bits)
        srshl           v4.4s,   v4.4s,   v29.4s // -(6+intermediate_bits)
        srshl           v5.4s,   v5.4s,   v29.4s // -(6+intermediate_bits)
        sqxtun          v2.4h,   v2.4s
        sqxtun2         v2.8h,   v3.4s
        sqxtun          v3.4h,   v4.4s
        sqxtun2         v3.8h,   v5.4s
        umin            v2.8h,   v2.8h,   v31.8h
        umin            v3.8h,   v3.8h,   v31.8h
.else
        rshrn           v2.4h,   v2.4s,   #6
        rshrn2          v2.8h,   v3.4s,   #6
        rshrn           v3.4h,   v4.4s,   #6
        rshrn2          v3.8h,   v5.4s,   #6
        sub             v2.8h,   v2.8h,   v29.8h // PREP_BIAS
        sub             v3.8h,   v3.8h,   v29.8h // PREP_BIAS
.endif
        subs            \h,  \h,  #2
        st1             {v2.8h}, [\dst], \d_strd
        st1             {v3.8h}, [\ds2], \d_strd
        b.le            9f
        mov             v16.16b, v18.16b
        mov             v17.16b, v19.16b
        mov             v18.16b, v20.16b
        mov             v19.16b, v21.16b
        mov             v20.16b, v22.16b
        mov             v21.16b, v23.16b
        mov             v22.16b, v24.16b
        b               88b
9:
        subs            \w,  \w,  #8
        b.le            0f
        asr             \s_strd,  \s_strd,  #1
        asr             \d_strd,  \d_strd,  #1
        msub            \src,  \s_strd,  \xmy,  \src
        msub            \dst,  \d_strd,  \xmy,  \dst
        sub             \src,  \src,  \s_strd,  lsl #3
        mov             \h,  \my
        add             \src,  \src,  #16
        add             \dst,  \dst,  #16
        b               168b
0:
        ret             x15

L(\type\()_8tap_filter_8):
        ld1             {v4.8h, v5.8h},  [\sr2], \s_strd
        ld1             {v6.8h, v7.8h},  [\src], \s_strd
        smull           v25.4s,  v4.4h,   v0.h[0]
        smull2          v26.4s,  v4.8h,   v0.h[0]
        smull           v27.4s,  v6.4h,   v0.h[0]
        smull2          v28.4s,  v6.8h,   v0.h[0]
.irpc i, 1234567
        ext             v23.16b, v4.16b,  v5.16b,  #(2*\i)
        ext             v24.16b, v6.16b,  v7.16b,  #(2*\i)
        smlal           v25.4s,  v23.4h,  v0.h[\i]
        smlal2          v26.4s,  v23.8h,  v0.h[\i]
        smlal           v27.4s,  v24.4h,  v0.h[\i]
        smlal2          v28.4s,  v24.8h,  v0.h[\i]
.endr
        srshl           v25.4s,  v25.4s,  v30.4s // -(6-intermediate_bits)
        srshl           v26.4s,  v26.4s,  v30.4s // -(6-intermediate_bits)
        srshl           v27.4s,  v27.4s,  v30.4s // -(6-intermediate_bits)
        srshl           v28.4s,  v28.4s,  v30.4s // -(6-intermediate_bits)
        uzp1            v23.8h,  v25.8h,  v26.8h // Same as xtn, xtn2
        uzp1            v24.8h,  v27.8h,  v28.8h // Ditto
        ret

L(\type\()_8tap_hv_tbl):
        .hword L(\type\()_8tap_hv_tbl) - 1280b
        .hword L(\type\()_8tap_hv_tbl) -  640b
        .hword L(\type\()_8tap_hv_tbl) -  320b
        .hword L(\type\()_8tap_hv_tbl) -  160b
        .hword L(\type\()_8tap_hv_tbl) -   80b
        .hword L(\type\()_8tap_hv_tbl) -   40b
        .hword L(\type\()_8tap_hv_tbl) -   20b
        .hword 0
endfunc


function \type\()_bilin_16bpc_neon, export=1
.ifc \bdmax, w8
        ldr             w8,  [sp]
.endif
        dup             v1.8h,   \mx
        dup             v3.8h,   \my
        mov             w10, #16
        sub             w9,  w10, \mx
        sub             w10, w10, \my
        dup             v0.8h,   w9
        dup             v2.8h,   w10
.ifc \type, prep
        uxtw            \d_strd, \w
        lsl             \d_strd, \d_strd, #1
.endif

        clz             \bdmax,   \bdmax       // bitdepth_max
        clz             w9,  \w
        sub             \bdmax,   \bdmax,  #18 // intermediate_bits = clz(bitdepth_max) - 18
        mov             w11, #4
        sub             w9,  w9,  #24
        sub             w11, w11, \bdmax  // 4 - intermediate_bits
        add             w12, \bdmax, #4   // 4 + intermediate_bits
        cbnz            \mx, L(\type\()_bilin_h)
        cbnz            \my, L(\type\()_bilin_v)
        b               \type\()_neon

L(\type\()_bilin_h):
        cbnz            \my, L(\type\()_bilin_hv)

        adr             x10, L(\type\()_bilin_h_tbl)
        dup             v31.8h,  w11      // 4 - intermediate_bits
        ldrh            w9,  [x10, x9, lsl #1]
        neg             v31.8h,  v31.8h   // -(4-intermediate_bits)
.ifc \type, put
        dup             v30.8h,  \bdmax   // intermediate_bits
.else
        movi            v29.8h,  #(PREP_BIAS >> 8), lsl #8
.endif
        sub             x10, x10, w9, uxtw
.ifc \type, put
        neg             v30.8h,  v30.8h   // -intermediate_bits
.endif
        br              x10

20:     // 2xN h
        AARCH64_VALID_JUMP_TARGET
.ifc \type, put
        add             \ds2,  \dst,  \d_strd
        add             \sr2,  \src,  \s_strd
        lsl             \d_strd,  \d_strd,  #1
        lsl             \s_strd,  \s_strd,  #1
2:
        ld1             {v4.4h},  [\src], \s_strd
        ld1             {v6.4h},  [\sr2], \s_strd
        ext             v5.8b,   v4.8b,   v4.8b,   #2
        ext             v7.8b,   v6.8b,   v6.8b,   #2
        trn1            v4.2s,   v4.2s,   v6.2s
        trn1            v5.2s,   v5.2s,   v7.2s
        subs            \h,  \h,  #2
        mul             v4.4h,   v4.4h,   v0.4h
        mla             v4.4h,   v5.4h,   v1.4h
        urshl           v4.4h,   v4.4h,   v31.4h
        urshl           v4.4h,   v4.4h,   v30.4h
        st1             {v4.s}[0], [\dst], \d_strd
        st1             {v4.s}[1], [\ds2], \d_strd
        b.gt            2b
        ret
.endif

40:     // 4xN h
        AARCH64_VALID_JUMP_TARGET
        add             \ds2,  \dst,  \d_strd
        add             \sr2,  \src,  \s_strd
        lsl             \d_strd,  \d_strd,  #1
        lsl             \s_strd,  \s_strd,  #1
4:
        ld1             {v4.8h}, [\src], \s_strd
        ld1             {v6.8h}, [\sr2], \s_strd
        ext             v5.16b,  v4.16b,  v4.16b,  #2
        ext             v7.16b,  v6.16b,  v6.16b,  #2
        trn1            v4.2d,   v4.2d,   v6.2d
        trn1            v5.2d,   v5.2d,   v7.2d
        subs            \h,  \h,  #2
        mul             v4.8h,   v4.8h,   v0.8h
        mla             v4.8h,   v5.8h,   v1.8h
        urshl           v4.8h,   v4.8h,   v31.8h
.ifc \type, put
        urshl           v4.8h,   v4.8h,   v30.8h
.else
        sub             v4.8h,   v4.8h,   v29.8h
.endif
        st1             {v4.d}[0], [\dst], \d_strd
        st1             {v4.d}[1], [\ds2], \d_strd
        b.gt            4b
        ret

80:     // 8xN h
        AARCH64_VALID_JUMP_TARGET
        add             \ds2,  \dst,  \d_strd
        add             \sr2,  \src,  \s_strd
        lsl             \d_strd,  \d_strd,  #1
        lsl             \s_strd,  \s_strd,  #1
8:
        ldr             h5,  [\src, #16]
        ldr             h7,  [\sr2, #16]
        ld1             {v4.8h}, [\src], \s_strd
        ld1             {v6.8h}, [\sr2], \s_strd
        ext             v5.16b,  v4.16b,  v5.16b,  #2
        ext             v7.16b,  v6.16b,  v7.16b,  #2
        subs            \h,  \h,  #2
        mul             v4.8h,   v4.8h,   v0.8h
        mla             v4.8h,   v5.8h,   v1.8h
        mul             v6.8h,   v6.8h,   v0.8h
        mla             v6.8h,   v7.8h,   v1.8h
        urshl           v4.8h,   v4.8h,   v31.8h
        urshl           v6.8h,   v6.8h,   v31.8h
.ifc \type, put
        urshl           v4.8h,   v4.8h,   v30.8h
        urshl           v6.8h,   v6.8h,   v30.8h
.else
        sub             v4.8h,   v4.8h,   v29.8h
        sub             v6.8h,   v6.8h,   v29.8h
.endif
        st1             {v4.8h}, [\dst], \d_strd
        st1             {v6.8h}, [\ds2], \d_strd
        b.gt            8b
        ret
160:
320:
640:
1280:   // 16xN, 32xN, ... h
        AARCH64_VALID_JUMP_TARGET
        add             \ds2,  \dst,  \d_strd
        add             \sr2,  \src,  \s_strd
        lsl             \s_strd,  \s_strd,  #1

        sub             \s_strd,  \s_strd,  \w, uxtw #1
        sub             \s_strd,  \s_strd,  #16
.ifc \type, put
        lsl             \d_strd,  \d_strd,  #1
        sub             \d_strd,  \d_strd,  \w, uxtw #1
.endif
161:
        ld1             {v16.8h},  [\src], #16
        ld1             {v21.8h},  [\sr2], #16
        mov             \mx, \w

16:
        ld1             {v17.8h, v18.8h},  [\src], #32
        ld1             {v22.8h, v23.8h},  [\sr2], #32
        ext             v19.16b, v16.16b, v17.16b, #2
        ext             v20.16b, v17.16b, v18.16b, #2
        ext             v24.16b, v21.16b, v22.16b, #2
        ext             v25.16b, v22.16b, v23.16b, #2
        mul             v16.8h,  v16.8h,  v0.8h
        mla             v16.8h,  v19.8h,  v1.8h
        mul             v17.8h,  v17.8h,  v0.8h
        mla             v17.8h,  v20.8h,  v1.8h
        mul             v21.8h,  v21.8h,  v0.8h
        mla             v21.8h,  v24.8h,  v1.8h
        mul             v22.8h,  v22.8h,  v0.8h
        mla             v22.8h,  v25.8h,  v1.8h
        urshl           v16.8h,  v16.8h,  v31.8h
        urshl           v17.8h,  v17.8h,  v31.8h
        urshl           v21.8h,  v21.8h,  v31.8h
        urshl           v22.8h,  v22.8h,  v31.8h
        subs            \mx, \mx, #16
.ifc \type, put
        urshl           v16.8h,  v16.8h,  v30.8h
        urshl           v17.8h,  v17.8h,  v30.8h
        urshl           v21.8h,  v21.8h,  v30.8h
        urshl           v22.8h,  v22.8h,  v30.8h
.else
        sub             v16.8h,  v16.8h,  v29.8h
        sub             v17.8h,  v17.8h,  v29.8h
        sub             v21.8h,  v21.8h,  v29.8h
        sub             v22.8h,  v22.8h,  v29.8h
.endif
        st1             {v16.8h, v17.8h}, [\dst], #32
        st1             {v21.8h, v22.8h}, [\ds2], #32
        b.le            9f

        mov             v16.16b, v18.16b
        mov             v21.16b, v23.16b
        b               16b

9:
        add             \dst,  \dst,  \d_strd
        add             \ds2,  \ds2,  \d_strd
        add             \src,  \src,  \s_strd
        add             \sr2,  \sr2,  \s_strd

        subs            \h,  \h,  #2
        b.gt            161b
        ret

L(\type\()_bilin_h_tbl):
        .hword L(\type\()_bilin_h_tbl) - 1280b
        .hword L(\type\()_bilin_h_tbl) -  640b
        .hword L(\type\()_bilin_h_tbl) -  320b
        .hword L(\type\()_bilin_h_tbl) -  160b
        .hword L(\type\()_bilin_h_tbl) -   80b
        .hword L(\type\()_bilin_h_tbl) -   40b
        .hword L(\type\()_bilin_h_tbl) -   20b
        .hword 0


L(\type\()_bilin_v):
        cmp             \h,  #4
        adr             x10, L(\type\()_bilin_v_tbl)
.ifc \type, prep
        dup             v31.8h,  w11      // 4 - intermediate_bits
.endif
        ldrh            w9,  [x10, x9, lsl #1]
.ifc \type, prep
        movi            v29.8h,  #(PREP_BIAS >> 8), lsl #8
        neg             v31.8h,  v31.8h   // -(4-intermediate_bits)
.endif
        sub             x10, x10, w9, uxtw
        br              x10

20:     // 2xN v
        AARCH64_VALID_JUMP_TARGET
.ifc \type, put
        cmp             \h,  #2
        add             \ds2,  \dst,  \d_strd
        add             \sr2,  \src,  \s_strd
        lsl             \s_strd,  \s_strd,  #1
        lsl             \d_strd,  \d_strd,  #1

        // 2x2 v
        ld1             {v16.s}[0], [\src], \s_strd
        b.gt            24f
22:
        ld1             {v17.s}[0], [\sr2], \s_strd
        ld1             {v18.s}[0], [\src], \s_strd
        trn1            v16.2s,  v16.2s,  v17.2s
        trn1            v17.2s,  v17.2s,  v18.2s
        mul             v4.4h,   v16.4h,  v2.4h
        mla             v4.4h,   v17.4h,  v3.4h
        urshr           v4.8h,   v4.8h,   #4
        st1             {v4.s}[0], [\dst]
        st1             {v4.s}[1], [\ds2]
        ret
24:     // 2x4, 2x6, 2x8, ... v
        ld1             {v17.s}[0], [\sr2], \s_strd
        ld1             {v18.s}[0], [\src], \s_strd
        ld1             {v19.s}[0], [\sr2], \s_strd
        ld1             {v20.s}[0], [\src], \s_strd
        sub             \h,  \h,  #4
        trn1            v16.2s,  v16.2s,  v17.2s
        trn1            v17.2s,  v17.2s,  v18.2s
        trn1            v18.2s,  v18.2s,  v19.2s
        trn1            v19.2s,  v19.2s,  v20.2s
        trn1            v16.2d,  v16.2d,  v18.2d
        trn1            v17.2d,  v17.2d,  v19.2d
        mul             v4.8h,   v16.8h,  v2.8h
        mla             v4.8h,   v17.8h,  v3.8h
        cmp             \h,  #2
        urshr           v4.8h,   v4.8h,   #4
        st1             {v4.s}[0], [\dst], \d_strd
        st1             {v4.s}[1], [\ds2], \d_strd
        st1             {v4.s}[2], [\dst], \d_strd
        st1             {v4.s}[3], [\ds2], \d_strd
        b.lt            0f
        mov             v16.8b,  v20.8b
        b.eq            22b
        b               24b
0:
        ret
.endif

40:     // 4xN v
        AARCH64_VALID_JUMP_TARGET
        add             \ds2,  \dst,  \d_strd
        add             \sr2,  \src,  \s_strd
        lsl             \s_strd,  \s_strd,  #1
        lsl             \d_strd,  \d_strd,  #1
        ld1             {v16.4h}, [\src], \s_strd
4:
        ld1             {v17.4h}, [\sr2], \s_strd
        ld1             {v18.4h}, [\src], \s_strd
        trn1            v16.2d,  v16.2d,  v17.2d
        trn1            v17.2d,  v17.2d,  v18.2d
        mul             v4.8h,   v16.8h,  v2.8h
        mla             v4.8h,   v17.8h,  v3.8h
        subs            \h,  \h,  #2
.ifc \type, put
        urshr           v4.8h,   v4.8h,   #4
.else
        urshl           v4.8h,   v4.8h,   v31.8h
        sub             v4.8h,   v4.8h,   v29.8h
.endif
        st1             {v4.d}[0], [\dst], \d_strd
        st1             {v4.d}[1], [\ds2], \d_strd
        b.le            0f
        mov             v16.8b,  v18.8b
        b               4b
0:
        ret

80:     // 8xN v
        AARCH64_VALID_JUMP_TARGET
        add             \ds2,  \dst,  \d_strd
        add             \sr2,  \src,  \s_strd
        lsl             \s_strd,  \s_strd,  #1
        lsl             \d_strd,  \d_strd,  #1
        ld1             {v16.8h}, [\src], \s_strd
8:
        ld1             {v17.8h}, [\sr2], \s_strd
        ld1             {v18.8h}, [\src], \s_strd
        mul             v4.8h,   v16.8h,  v2.8h
        mla             v4.8h,   v17.8h,  v3.8h
        mul             v5.8h,   v17.8h,  v2.8h
        mla             v5.8h,   v18.8h,  v3.8h
        subs            \h,  \h,  #2
.ifc \type, put
        urshr           v4.8h,   v4.8h,   #4
        urshr           v5.8h,   v5.8h,   #4
.else
        urshl           v4.8h,   v4.8h,   v31.8h
        urshl           v5.8h,   v5.8h,   v31.8h
        sub             v4.8h,   v4.8h,   v29.8h
        sub             v5.8h,   v5.8h,   v29.8h
.endif
        st1             {v4.8h}, [\dst], \d_strd
        st1             {v5.8h}, [\ds2], \d_strd
        b.le            0f
        mov             v16.16b, v18.16b
        b               8b
0:
        ret

160:    // 16xN, 32xN, ...
320:
640:
1280:
        AARCH64_VALID_JUMP_TARGET
        mov             \my, \h
1:
        add             \ds2, \dst, \d_strd
        add             \sr2, \src, \s_strd
        lsl             \s_strd, \s_strd, #1
        lsl             \d_strd, \d_strd, #1

        ld1             {v16.8h, v17.8h}, [\src], \s_strd
2:
        ld1             {v18.8h, v19.8h}, [\sr2], \s_strd
        ld1             {v20.8h, v21.8h}, [\src], \s_strd
        mul             v4.8h,   v16.8h,  v2.8h
        mla             v4.8h,   v18.8h,  v3.8h
        mul             v5.8h,   v17.8h,  v2.8h
        mla             v5.8h,   v19.8h,  v3.8h
        mul             v6.8h,   v18.8h,  v2.8h
        mla             v6.8h,   v20.8h,  v3.8h
        mul             v7.8h,   v19.8h,  v2.8h
        mla             v7.8h,   v21.8h,  v3.8h
        subs            \h,  \h,  #2
.ifc \type, put
        urshr           v4.8h,   v4.8h,   #4
        urshr           v5.8h,   v5.8h,   #4
        urshr           v6.8h,   v6.8h,   #4
        urshr           v7.8h,   v7.8h,   #4
.else
        urshl           v4.8h,   v4.8h,   v31.8h
        urshl           v5.8h,   v5.8h,   v31.8h
        urshl           v6.8h,   v6.8h,   v31.8h
        urshl           v7.8h,   v7.8h,   v31.8h
        sub             v4.8h,   v4.8h,   v29.8h
        sub             v5.8h,   v5.8h,   v29.8h
        sub             v6.8h,   v6.8h,   v29.8h
        sub             v7.8h,   v7.8h,   v29.8h
.endif
        st1             {v4.8h, v5.8h}, [\dst], \d_strd
        st1             {v6.8h, v7.8h}, [\ds2], \d_strd
        b.le            9f
        mov             v16.16b, v20.16b
        mov             v17.16b, v21.16b
        b               2b
9:
        subs            \w,  \w,  #16
        b.le            0f
        asr             \s_strd, \s_strd, #1
        asr             \d_strd, \d_strd, #1
        msub            \src, \s_strd, \xmy, \src
        msub            \dst, \d_strd, \xmy, \dst
        sub             \src, \src, \s_strd, lsl #1
        mov             \h,  \my
        add             \src, \src, #32
        add             \dst, \dst, #32
        b               1b
0:
        ret

L(\type\()_bilin_v_tbl):
        .hword L(\type\()_bilin_v_tbl) - 1280b
        .hword L(\type\()_bilin_v_tbl) -  640b
        .hword L(\type\()_bilin_v_tbl) -  320b
        .hword L(\type\()_bilin_v_tbl) -  160b
        .hword L(\type\()_bilin_v_tbl) -   80b
        .hword L(\type\()_bilin_v_tbl) -   40b
        .hword L(\type\()_bilin_v_tbl) -   20b
        .hword 0

L(\type\()_bilin_hv):
        adr             x10, L(\type\()_bilin_hv_tbl)
        dup             v31.8h,  w11      // 4 - intermediate_bits
        ldrh            w9,  [x10, x9, lsl #1]
        neg             v31.8h,  v31.8h   // -(4-intermediate_bits)
.ifc \type, put
        dup             v30.4s,  w12      // 4 + intermediate_bits
.else
        movi            v29.8h,  #(PREP_BIAS >> 8), lsl #8
.endif
        sub             x10, x10, w9, uxtw
.ifc \type, put
        neg             v30.4s,  v30.4s   // -(4+intermediate_bits)
.endif
        br              x10

20:     // 2xN hv
        AARCH64_VALID_JUMP_TARGET
.ifc \type, put
        add             \sr2, \src, \s_strd
        add             \ds2, \dst, \d_strd
        lsl             \s_strd, \s_strd, #1
        lsl             \d_strd, \d_strd, #1

        ld1             {v20.4h},  [\src], \s_strd
        ext             v21.8b,  v20.8b,  v20.8b,  #2
        mul             v16.4h,  v20.4h,  v0.4h
        mla             v16.4h,  v21.4h,  v1.4h
        urshl           v16.4h,  v16.4h,  v31.4h

2:
        ld1             {v22.4h},  [\sr2], \s_strd
        ld1             {v24.4h},  [\src], \s_strd
        ext             v23.8b,  v22.8b,  v22.8b,  #2
        ext             v25.8b,  v24.8b,  v24.8b,  #2
        trn1            v22.2s,  v22.2s,  v24.2s
        trn1            v23.2s,  v23.2s,  v25.2s
        mul             v17.4h,  v22.4h,  v0.4h
        mla             v17.4h,  v23.4h,  v1.4h
        urshl           v17.4h,  v17.4h,  v31.4h

        trn1            v16.2s,  v16.2s,  v17.2s

        umull           v4.4s,   v16.4h,  v2.4h
        umlal           v4.4s,   v17.4h,  v3.4h
        urshl           v4.4s,   v4.4s,   v30.4s
        xtn             v4.4h,   v4.4s
        subs            \h,  \h,  #2
        st1             {v4.s}[0], [\dst], \d_strd
        st1             {v4.s}[1], [\ds2], \d_strd
        b.le            0f
        trn2            v16.2s,  v17.2s,  v17.2s
        b               2b
0:
        ret
.endif

40:     // 4xN hv
        AARCH64_VALID_JUMP_TARGET
        add             \sr2, \src, \s_strd
        add             \ds2, \dst, \d_strd
        lsl             \s_strd, \s_strd, #1
        lsl             \d_strd, \d_strd, #1

        ld1             {v20.8h},  [\src], \s_strd
        ext             v21.16b, v20.16b, v20.16b, #2
        mul             v16.4h,  v20.4h,  v0.4h
        mla             v16.4h,  v21.4h,  v1.4h
        urshl           v16.4h,  v16.4h,  v31.4h

4:
        ld1             {v22.8h},  [\sr2], \s_strd
        ld1             {v24.8h},  [\src], \s_strd
        ext             v23.16b, v22.16b, v22.16b, #2
        ext             v25.16b, v24.16b, v24.16b, #2
        trn1            v22.2d,  v22.2d,  v24.2d
        trn1            v23.2d,  v23.2d,  v25.2d
        mul             v17.8h,  v22.8h,  v0.8h
        mla             v17.8h,  v23.8h,  v1.8h
        urshl           v17.8h,  v17.8h,  v31.8h

        trn1            v16.2d,  v16.2d,  v17.2d

        umull           v4.4s,   v16.4h,  v2.4h
        umlal           v4.4s,   v17.4h,  v3.4h
        umull2          v5.4s,   v16.8h,  v2.8h
        umlal2          v5.4s,   v17.8h,  v3.8h
.ifc \type, put
        urshl           v4.4s,   v4.4s,   v30.4s
        urshl           v5.4s,   v5.4s,   v30.4s
        uzp1            v4.8h,   v4.8h,   v5.8h  // Same as xtn, xtn2
.else
        rshrn           v4.4h,   v4.4s,   #4
        rshrn2          v4.8h,   v5.4s,   #4
        sub             v4.8h,   v4.8h,   v29.8h
.endif
        subs            \h,  \h,  #2
        st1             {v4.d}[0], [\dst], \d_strd
        st1             {v4.d}[1], [\ds2], \d_strd
        b.le            0f
        trn2            v16.2d,  v17.2d,  v17.2d
        b               4b
0:
        ret

80:     // 8xN, 16xN, ... hv
160:
320:
640:
1280:
        AARCH64_VALID_JUMP_TARGET
        mov             \my, \h

1:
        add             \sr2, \src, \s_strd
        add             \ds2, \dst, \d_strd
        lsl             \s_strd, \s_strd, #1
        lsl             \d_strd, \d_strd, #1

        ldr             h21, [\src, #16]
        ld1             {v20.8h},  [\src], \s_strd
        ext             v21.16b, v20.16b, v21.16b, #2
        mul             v16.8h,  v20.8h,  v0.8h
        mla             v16.8h,  v21.8h,  v1.8h
        urshl           v16.8h,  v16.8h,  v31.8h

2:
        ldr             h23, [\sr2, #16]
        ld1             {v22.8h},  [\sr2], \s_strd
        ldr             h25, [\src, #16]
        ld1             {v24.8h},  [\src], \s_strd
        ext             v23.16b, v22.16b, v23.16b, #2
        ext             v25.16b, v24.16b, v25.16b, #2
        mul             v17.8h,  v22.8h,  v0.8h
        mla             v17.8h,  v23.8h,  v1.8h
        mul             v18.8h,  v24.8h,  v0.8h
        mla             v18.8h,  v25.8h,  v1.8h
        urshl           v17.8h,  v17.8h,  v31.8h
        urshl           v18.8h,  v18.8h,  v31.8h

        umull           v4.4s,   v16.4h,  v2.4h
        umlal           v4.4s,   v17.4h,  v3.4h
        umull2          v5.4s,   v16.8h,  v2.8h
        umlal2          v5.4s,   v17.8h,  v3.8h
        umull           v6.4s,   v17.4h,  v2.4h
        umlal           v6.4s,   v18.4h,  v3.4h
        umull2          v7.4s,   v17.8h,  v2.8h
        umlal2          v7.4s,   v18.8h,  v3.8h
.ifc \type, put
        urshl           v4.4s,   v4.4s,   v30.4s
        urshl           v5.4s,   v5.4s,   v30.4s
        urshl           v6.4s,   v6.4s,   v30.4s
        urshl           v7.4s,   v7.4s,   v30.4s
        uzp1            v4.8h,   v4.8h,   v5.8h  // Same as xtn, xtn2
        uzp1            v5.8h,   v6.8h,   v7.8h  // Ditto
.else
        rshrn           v4.4h,   v4.4s,   #4
        rshrn2          v4.8h,   v5.4s,   #4
        rshrn           v5.4h,   v6.4s,   #4
        rshrn2          v5.8h,   v7.4s,   #4
        sub             v4.8h,   v4.8h,   v29.8h
        sub             v5.8h,   v5.8h,   v29.8h
.endif
        subs            \h,  \h,  #2
        st1             {v4.8h}, [\dst], \d_strd
        st1             {v5.8h}, [\ds2], \d_strd
        b.le            9f
        mov             v16.16b, v18.16b
        b               2b
9:
        subs            \w,  \w,  #8
        b.le            0f
        asr             \s_strd,  \s_strd,  #1
        asr             \d_strd,  \d_strd,  #1
        msub            \src,  \s_strd,  \xmy,  \src
        msub            \dst,  \d_strd,  \xmy,  \dst
        sub             \src,  \src,  \s_strd,  lsl #1
        mov             \h,  \my
        add             \src,  \src,  #16
        add             \dst,  \dst,  #16
        b               1b
0:
        ret

L(\type\()_bilin_hv_tbl):
        .hword L(\type\()_bilin_hv_tbl) - 1280b
        .hword L(\type\()_bilin_hv_tbl) -  640b
        .hword L(\type\()_bilin_hv_tbl) -  320b
        .hword L(\type\()_bilin_hv_tbl) -  160b
        .hword L(\type\()_bilin_hv_tbl) -   80b
        .hword L(\type\()_bilin_hv_tbl) -   40b
        .hword L(\type\()_bilin_hv_tbl) -   20b
        .hword 0
endfunc
.endm

filter_fn put,  x0, x1, x2, x3, w4, w5, w6, x6, w7, x7, w8, x9, x10
filter_fn prep, x0, x8, x1, x2, w3, w4, w5, x5, w6, x6, w7, x9, x10

.macro load_filter_row dst, src, inc
        asr             w13, \src, #10
        add             \src, \src, \inc
        ldr             \dst, [x11, w13, sxtw #3]
.endm

function warp_filter_horz_neon
        add             w12, w5,  #512

        ld1             {v16.8h, v17.8h}, [x2], x3

        load_filter_row d0, w12, w7
        load_filter_row d1, w12, w7
        load_filter_row d2, w12, w7
        sxtl            v0.8h,   v0.8b
        load_filter_row d3, w12, w7
        sxtl            v1.8h,   v1.8b
        load_filter_row d4, w12, w7
        sxtl            v2.8h,   v2.8b
        load_filter_row d5, w12, w7
        sxtl            v3.8h,   v3.8b
        load_filter_row d6, w12, w7
        sxtl            v4.8h,   v4.8b
        load_filter_row d7, w12, w7
        sxtl            v5.8h,   v5.8b
        ext             v18.16b, v16.16b, v17.16b, #2*1
        smull           v8.4s,   v16.4h,  v0.4h
        smull2          v9.4s,   v16.8h,  v0.8h
        sxtl            v6.8h,   v6.8b
        ext             v19.16b, v16.16b, v17.16b, #2*2
        smull           v10.4s,  v18.4h,  v1.4h
        smull2          v11.4s,  v18.8h,  v1.8h
        sxtl            v7.8h,   v7.8b
        ext             v20.16b, v16.16b, v17.16b, #2*3
        smull           v0.4s,   v19.4h,  v2.4h
        smull2          v1.4s,   v19.8h,  v2.8h
        ext             v21.16b, v16.16b, v17.16b, #2*4
        addp            v8.4s,   v8.4s,   v9.4s
        smull           v2.4s,   v20.4h,  v3.4h
        smull2          v3.4s,   v20.8h,  v3.8h
        ext             v22.16b, v16.16b, v17.16b, #2*5
        addp            v9.4s,   v10.4s,  v11.4s
        smull           v10.4s,  v21.4h,  v4.4h
        smull2          v11.4s,  v21.8h,  v4.8h
        ext             v23.16b, v16.16b, v17.16b, #2*6
        addp            v0.4s,   v0.4s,   v1.4s
        smull           v18.4s,  v22.4h,  v5.4h
        smull2          v19.4s,  v22.8h,  v5.8h
        ext             v16.16b, v16.16b, v17.16b, #2*7
        addp            v1.4s,   v2.4s,   v3.4s
        addp            v2.4s,   v10.4s,  v11.4s
        smull           v20.4s,  v23.4h,  v6.4h
        smull2          v21.4s,  v23.8h,  v6.8h
        addp            v3.4s,   v18.4s,  v19.4s
        smull           v22.4s,  v16.4h,  v7.4h
        smull2          v23.4s,  v16.8h,  v7.8h
        addp            v4.4s,   v20.4s,  v21.4s
        addp            v5.4s,   v22.4s,  v23.4s

        addp            v8.4s,   v8.4s,   v9.4s
        addp            v0.4s,   v0.4s,   v1.4s
        addp            v2.4s,   v2.4s,   v3.4s
        addp            v4.4s,   v4.4s,   v5.4s

        addp            v16.4s,  v8.4s,   v0.4s
        addp            v17.4s,  v2.4s,   v4.4s

        add             w5,  w5,  w8

        srshl           v16.4s,  v16.4s,  v14.4s // -(7 - intermediate_bits)
        srshl           v17.4s,  v17.4s,  v14.4s // -(7 - intermediate_bits)

        ret
endfunc

// void dav1d_warp_affine_8x8_16bpc_neon(
//         pixel *dst, const ptrdiff_t dst_stride,
//         const pixel *src, const ptrdiff_t src_stride,
//         const int16_t *const abcd, int mx, int my,
//         const int bitdepth_max)
.macro warp t
function warp_affine_8x8\t\()_16bpc_neon, export=1
        stp             d8,  d9,  [sp, #-0x40]!
        stp             d10, d11, [sp, #0x10]
        stp             d12, d13, [sp, #0x20]
        stp             d14, d15, [sp, #0x30]

.ifb \t
        dup             v15.8h,  w7        // bitdepth_max
.else
        movi            v15.8h,  #(PREP_BIAS >> 8), lsl #8
.endif
        clz             w7,  w7
                                           // intermediate_bits = clz(bitdepth_max) - 18
.ifb \t
        sub             w8,  w7,  #11      // 7 + intermediate_bits = clz(bitdepth_max) - 18 + 7
.endif
        sub             w7,  w7,  #25      // -(7 - intermediate_bits)
.ifb \t
        neg             w8,  w8            // -(7 + intermediate_bits)
.endif
        dup             v14.4s,  w7        // -(7 - intermediate_bits)
.ifb \t
        dup             v13.4s,  w8        // -(7 + intermediate_bits)
.endif

        ldr             x4,  [x4]
        sbfx            x7,  x4, #0,  #16
        sbfx            x8,  x4, #16, #16
        sbfx            x9,  x4, #32, #16
        sbfx            x4,  x4, #48, #16
        mov             w10, #8
        sub             x2,  x2,  x3, lsl #1
        sub             x2,  x2,  x3
        sub             x2,  x2,  #6
        movrel          x11, X(mc_warp_filter), 64*8
        mov             x15, x30
.ifnb \t
        lsl             x1,  x1,  #1
.endif

        bl              warp_filter_horz_neon
        uzp1            v24.8h,  v16.8h,  v17.8h // Same as xtn, xtn2
        bl              warp_filter_horz_neon
        uzp1            v25.8h,  v16.8h,  v17.8h // Ditto
        bl              warp_filter_horz_neon
        uzp1            v26.8h,  v16.8h,  v17.8h // Ditto
        bl              warp_filter_horz_neon
        uzp1            v27.8h,  v16.8h,  v17.8h // Ditto
        bl              warp_filter_horz_neon
        uzp1            v28.8h,  v16.8h,  v17.8h // Ditto
        bl              warp_filter_horz_neon
        uzp1            v29.8h,  v16.8h,  v17.8h // Ditto
        bl              warp_filter_horz_neon
        uzp1            v30.8h,  v16.8h,  v17.8h // Ditto

1:
        add             w14, w6,  #512
        bl              warp_filter_horz_neon
        uzp1            v31.8h,  v16.8h,  v17.8h // Same as xtn, xtn2

        load_filter_row d0, w14, w9
        load_filter_row d1, w14, w9
        load_filter_row d2, w14, w9
        load_filter_row d3, w14, w9
        load_filter_row d4, w14, w9
        load_filter_row d5, w14, w9
        load_filter_row d6, w14, w9
        load_filter_row d7, w14, w9
        transpose_8x8b_xtl v0, v1, v2, v3, v4, v5, v6, v7, sxtl

        // This ordering of smull/smlal/smull2/smlal2 is highly
        // beneficial for Cortex A53 here.
        smull           v16.4s,  v24.4h,  v0.4h
        smlal           v16.4s,  v25.4h,  v1.4h
        smlal           v16.4s,  v26.4h,  v2.4h
        smlal           v16.4s,  v27.4h,  v3.4h
        smlal           v16.4s,  v28.4h,  v4.4h
        smlal           v16.4s,  v29.4h,  v5.4h
        smlal           v16.4s,  v30.4h,  v6.4h
        smlal           v16.4s,  v31.4h,  v7.4h
        smull2          v17.4s,  v24.8h,  v0.8h
        smlal2          v17.4s,  v25.8h,  v1.8h
        smlal2          v17.4s,  v26.8h,  v2.8h
        smlal2          v17.4s,  v27.8h,  v3.8h
        smlal2          v17.4s,  v28.8h,  v4.8h
        smlal2          v17.4s,  v29.8h,  v5.8h
        smlal2          v17.4s,  v30.8h,  v6.8h
        smlal2          v17.4s,  v31.8h,  v7.8h

        mov             v24.16b, v25.16b
        mov             v25.16b, v26.16b
.ifb \t
        srshl           v16.4s,  v16.4s,  v13.4s // -(7 + intermediate_bits)
        srshl           v17.4s,  v17.4s,  v13.4s // -(7 + intermediate_bits)
.else
        rshrn           v16.4h,  v16.4s,  #7
        rshrn2          v16.8h,  v17.4s,  #7
.endif
        mov             v26.16b, v27.16b
.ifb \t
        sqxtun          v16.4h,  v16.4s
        sqxtun2         v16.8h,  v17.4s
.else
        sub             v16.8h,  v16.8h,  v15.8h // PREP_BIAS
.endif
        mov             v27.16b, v28.16b
        mov             v28.16b, v29.16b
.ifb \t
        umin            v16.8h,  v16.8h,  v15.8h // bitdepth_max
.endif
        mov             v29.16b, v30.16b
        mov             v30.16b, v31.16b
        subs            w10, w10, #1
        st1             {v16.8h}, [x0], x1

        add             w6,  w6,  w4
        b.gt            1b

        ldp             d14, d15, [sp, #0x30]
        ldp             d12, d13, [sp, #0x20]
        ldp             d10, d11, [sp, #0x10]
        ldp             d8,  d9,  [sp], 0x40

        ret             x15
endfunc
.endm

warp
warp t

// void dav1d_emu_edge_16bpc_neon(
//         const intptr_t bw, const intptr_t bh,
//         const intptr_t iw, const intptr_t ih,
//         const intptr_t x, const intptr_t y,
//         pixel *dst, const ptrdiff_t dst_stride,
//         const pixel *ref, const ptrdiff_t ref_stride)
function emu_edge_16bpc_neon, export=1
        ldp             x8,  x9,  [sp]

        // ref += iclip(y, 0, ih - 1) * PXSTRIDE(ref_stride)
        // ref += iclip(x, 0, iw - 1)
        sub             x12, x3,  #1           // ih - 1
        cmp             x5,  x3
        sub             x13, x2,  #1           // iw - 1
        csel            x12, x12, x5,  ge      // min(y, ih - 1)
        cmp             x4,  x2
        bic             x12, x12, x12, asr #63 // max(min(y, ih - 1), 0)
        csel            x13, x13, x4,  ge      // min(x, iw - 1)
        bic             x13, x13, x13, asr #63 // max(min(x, iw - 1), 0)
        madd            x8,  x12, x9,  x8      // ref += iclip() * stride
        add             x8,  x8,  x13, lsl #1  // ref += iclip()

        // bottom_ext = iclip(y + bh - ih, 0, bh - 1)
        // top_ext = iclip(-y, 0, bh - 1)
        add             x10, x5,  x1           // y + bh
        neg             x5,  x5                // -y
        sub             x10, x10, x3           // y + bh - ih
        sub             x12, x1,  #1           // bh - 1
        cmp             x10, x1
        bic             x5,  x5,  x5,  asr #63 // max(-y, 0)
        csel            x10, x10, x12, lt      // min(y + bh - ih, bh-1)
        cmp             x5,  x1
        bic             x10, x10, x10, asr #63 // max(min(y + bh - ih, bh-1), 0)
        csel            x5,  x5,  x12, lt      // min(max(-y, 0), bh-1)

        // right_ext = iclip(x + bw - iw, 0, bw - 1)
        // left_ext = iclip(-x, 0, bw - 1)
        add             x11, x4,  x0           // x + bw
        neg             x4,  x4                // -x
        sub             x11, x11, x2           // x + bw - iw
        sub             x13, x0,  #1           // bw - 1
        cmp             x11, x0
        bic             x4,  x4,  x4,  asr #63 // max(-x, 0)
        csel            x11, x11, x13, lt      // min(x + bw - iw, bw-1)
        cmp             x4,  x0
        bic             x11, x11, x11, asr #63 // max(min(x + bw - iw, bw-1), 0)
        csel            x4,  x4,  x13, lt      // min(max(-x, 0), bw - 1)

        // center_h = bh - top_ext - bottom_ext
        // dst += top_ext * PXSTRIDE(dst_stride)
        // center_w = bw - left_ext - right_ext
        sub             x1,  x1,  x5           // bh - top_ext
        madd            x6,  x5,  x7,  x6
        sub             x2,  x0,  x4           // bw - left_ext
        sub             x1,  x1,  x10          // center_h = bh - top_ext - bottom_ext
        sub             x2,  x2,  x11          // center_w = bw - left_ext - right_ext

        mov             x14, x6                // backup of dst

.macro v_loop need_left, need_right
0:
.if \need_left
        ld1r            {v0.8h}, [x8]
        mov             x12, x6                // out = dst
        mov             x3,  x4
        mov             v1.16b,  v0.16b
1:
        subs            x3,  x3,  #16
        st1             {v0.8h, v1.8h}, [x12], #32
        b.gt            1b
.endif
        mov             x13, x8
        add             x12, x6,  x4, lsl #1   // out = dst + left_ext
        mov             x3,  x2
1:
        ld1             {v0.8h, v1.8h, v2.8h, v3.8h}, [x13], #64
        subs            x3,  x3,  #32
        st1             {v0.8h, v1.8h, v2.8h, v3.8h}, [x12], #64
        b.gt            1b
.if \need_right
        add             x3,  x8,  x2, lsl #1   // in + center_w
        sub             x3,  x3,  #2           // in + center_w - 1
        add             x12, x6,  x4, lsl #1   // dst + left_ext
        ld1r            {v0.8h}, [x3]
        add             x12, x12, x2, lsl #1   // out = dst + left_ext + center_w
        mov             x3,  x11
        mov             v1.16b,  v0.16b
1:
        subs            x3,  x3,  #16
        st1             {v0.8h, v1.8h}, [x12], #32
        b.gt            1b
.endif

        subs            x1,  x1,  #1           // center_h--
        add             x6,  x6,  x7
        add             x8,  x8,  x9
        b.gt            0b
.endm

        cbz             x4,  2f
        // need_left
        cbz             x11, 3f
        // need_left + need_right
        v_loop          1,   1
        b               5f

2:
        // !need_left
        cbz             x11, 4f
        // !need_left + need_right
        v_loop          0,   1
        b               5f

3:
        // need_left + !need_right
        v_loop          1,   0
        b               5f

4:
        // !need_left + !need_right
        v_loop          0,   0

5:

        cbz             x10, 3f
        // need_bottom
        sub             x8,  x6,  x7           // ref = dst - stride
        mov             x4,  x0
1:
        ld1             {v0.8h, v1.8h, v2.8h, v3.8h}, [x8], #64
        mov             x3,  x10
2:
        subs            x3,  x3,  #1
        st1             {v0.8h, v1.8h, v2.8h, v3.8h}, [x6], x7
        b.gt            2b
        msub            x6,  x7,  x10,  x6     // dst -= bottom_ext * stride
        subs            x4,  x4,  #32          // bw -= 32
        add             x6,  x6,  #64          // dst += 32
        b.gt            1b

3:
        cbz             x5,  3f
        // need_top
        msub            x6,  x7,  x5,  x14     // dst = stored_dst - top_ext * stride
1:
        ld1             {v0.8h, v1.8h, v2.8h, v3.8h}, [x14], #64
        mov             x3,  x5
2:
        subs            x3,  x3,  #1
        st1             {v0.8h, v1.8h, v2.8h, v3.8h}, [x6], x7
        b.gt            2b
        msub            x6,  x7,  x5,  x6      // dst -= top_ext * stride
        subs            x0,  x0,  #32          // bw -= 32
        add             x6,  x6,  #64          // dst += 32
        b.gt            1b

3:
        ret
endfunc