summaryrefslogtreecommitdiffstats
path: root/third_party/dav1d/src/x86/filmgrain_sse.asm
blob: 0172f987607a7717009bdf3b58a17e2283129fac (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
; Copyright © 2019-2021, VideoLAN and dav1d authors
; Copyright © 2019, Two Orioles, LLC
; 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 "config.asm"
%include "ext/x86/x86inc.asm"
%include "x86/filmgrain_common.asm"

SECTION_RODATA

pw_1024: times 8 dw 1024
pb_27_17_17_27: db 27, 17, 17, 27
                times 6 db 0, 32
pb_23_22_h: db 23, 22
            times 7 db 0, 32
pb_27_17: times 8 db 27, 17
pb_17_27: times 8 db 17, 27
pb_23_22: times 8 db 23, 22
pb_mask: db 0, 0x80, 0x80, 0, 0x80, 0, 0, 0x80, 0x80, 0, 0, 0x80, 0, 0x80, 0x80, 0
rnd_next_upperbit_mask: dw 0x100B, 0x2016, 0x402C, 0x8058
byte_blend: db 0, 0, 0, 0xff, 0, 0, 0, 0
pw_seed_xor: times 2 dw 0xb524
             times 2 dw 0x49d8
pb_1: times 4 db 1
hmul_bits: dw 32768, 16384, 8192, 4096
round: dw 2048, 1024, 512
mul_bits: dw 256, 128, 64, 32, 16
round_vals: dw 32, 64, 128, 256, 512
max: dw 255, 240, 235
min: dw 0, 16
pw_1: dw 1

%macro JMP_TABLE 2-*
    %xdefine %1_8bpc_%2_table %%table
    %xdefine %%base %1_8bpc_%2_table
    %xdefine %%prefix mangle(private_prefix %+ _%1_8bpc_%2)
    %%table:
    %rep %0 - 2
        dd %%prefix %+ .ar%3 - %%base
        %rotate 1
    %endrep
%endmacro

JMP_TABLE generate_grain_y, ssse3, 0, 1, 2, 3
JMP_TABLE generate_grain_uv_420, ssse3, 0, 1, 2, 3
JMP_TABLE generate_grain_uv_422, ssse3, 0, 1, 2, 3
JMP_TABLE generate_grain_uv_444, ssse3, 0, 1, 2, 3

SECTION .text

%if ARCH_X86_32
%define PIC_ptr(a) base+a
%else
%define PIC_ptr(a) a
%endif

%macro SCRATCH 3
%if ARCH_X86_32
    mova [rsp+%3*mmsize], m%1
%define m%2 [rsp+%3*mmsize]
%else
    SWAP             %1, %2
%endif
%endmacro

INIT_XMM ssse3
cglobal generate_grain_y_8bpc, 2, 7 + 2 * ARCH_X86_64, 16, buf, fg_data
    LEA              r4, $$
%define base r4-$$
    movq             m1, [base+rnd_next_upperbit_mask]
    movq             m4, [base+mul_bits]
    movq             m7, [base+hmul_bits]
    mov             r2d, [fg_dataq+FGData.grain_scale_shift]
    movd             m2, [base+round+r2*2]
    movd             m0, [fg_dataq+FGData.seed]
    mova             m5, [base+pb_mask]
    pshuflw          m2, m2, q0000
    pshuflw          m0, m0, q0000
    mov              r2, -73*82
    sub            bufq, r2
    lea              r3, [base+gaussian_sequence]
.loop:
    pand             m6, m0, m1
    psrlw            m3, m6, 10
    por              m6, m3            ; bits 0xf, 0x1e, 0x3c and 0x78 are set
    pmullw           m6, m4            ; bits 0x0f00 are set
    pshufb           m3, m5, m6        ; set 15th bit for next 4 seeds
    psllq            m6, m3, 30
    por              m3, m6
    psllq            m6, m3, 15
    por              m3, m6            ; aggregate each bit into next seed's high bit
    pmulhuw          m6, m0, m7
    por              m3, m6            ; 4 next output seeds
    pshuflw          m0, m3, q3333
    psrlw            m3, 5
%if ARCH_X86_64
    movq             r6, m3
    mov              r8, r6
    movzx           r5d, r6w
    shr             r6d, 16
    shr              r8, 32
    movzx            r7, r8w
    shr              r8, 16

    movd             m6, [r3+r5*2]
    pinsrw           m6, [r3+r6*2], 1
    pinsrw           m6, [r3+r7*2], 2
    pinsrw           m6, [r3+r8*2], 3
%else
    movd             r6, m3
    pshuflw          m3, m3, q3232
    movzx            r5, r6w
    shr              r6, 16

    movd             m6, [r3+r5*2]
    pinsrw           m6, [r3+r6*2], 1

    movd             r6, m3
    movzx            r5, r6w
    shr              r6, 16

    pinsrw           m6, [r3+r5*2], 2
    pinsrw           m6, [r3+r6*2], 3
%endif
    pmulhrsw         m6, m2
    packsswb         m6, m6
    movd      [bufq+r2], m6
    add              r2, 4
    jl .loop

    ; auto-regression code
    movsxd           r2, [fg_dataq+FGData.ar_coeff_lag]
    movsxd           r2, [base+generate_grain_y_8bpc_ssse3_table+r2*4]
    lea              r2, [r2+base+generate_grain_y_8bpc_ssse3_table]
    jmp              r2

.ar1:
%if ARCH_X86_32
    DEFINE_ARGS buf, fg_data, cf3, unused, val3, min, max
%elif WIN64
    DEFINE_ARGS shift, fg_data, cf3, buf, val3, min, max, x, val0
    mov            bufq, r0
%else
    DEFINE_ARGS buf, fg_data, cf3, shift, val3, min, max, x, val0
%endif
    movsx          cf3d, byte [fg_dataq+FGData.ar_coeffs_y+3]
    movd             m4, [fg_dataq+FGData.ar_coeffs_y]
    mov             ecx, [fg_dataq+FGData.ar_coeff_shift]
%if ARCH_X86_32
    mov             r1m, cf3d
    DEFINE_ARGS buf, shift, val3, min, max, x, val0
%define hd r0mp
%define cf3d r1mp
%elif WIN64
    DEFINE_ARGS shift, h, cf3, buf, val3, min, max, x, val0
%else
    DEFINE_ARGS buf, h, cf3, shift, val3, min, max, x, val0
%endif
    pxor             m6, m6
    pcmpgtb          m7, m6, m4
    punpcklbw        m4, m7
    pinsrw           m4, [base+pw_1], 3
    pshufd           m5, m4, q1111
    pshufd           m4, m4, q0000
    movd             m3, [base+round_vals+shiftq*2-12]    ; rnd
    pshuflw          m3, m3, q0000
    sub            bufq, 82*73-(82*3+79)
    mov              hd, 70
    mov            mind, -128
    mov            maxd, 127
.y_loop_ar1:
    mov              xq, -76
    movsx         val3d, byte [bufq+xq-1]
.x_loop_ar1:
    movq             m0, [bufq+xq-82-1]     ; top/left
    pcmpgtb          m7, m6, m0
    punpcklbw        m0, m7
    psrldq           m2, m0, 2              ; top
    psrldq           m1, m0, 4              ; top/right
    punpcklwd        m0, m2
    punpcklwd        m1, m3
    pmaddwd          m0, m4
    pmaddwd          m1, m5
    paddd            m0, m1
.x_loop_ar1_inner:
    movd          val0d, m0
    psrldq           m0, 4
    imul          val3d, cf3d
    add           val3d, val0d
    sar           val3d, shiftb
    movsx         val0d, byte [bufq+xq]
    add           val3d, val0d
    cmp           val3d, maxd
    cmovns        val3d, maxd
    cmp           val3d, mind
    cmovs         val3d, mind
    mov  byte [bufq+xq], val3b
    ; keep val3d in-place as left for next x iteration
    inc              xq
    jz .x_loop_ar1_end
    test             xq, 3
    jnz .x_loop_ar1_inner
    jmp .x_loop_ar1

.x_loop_ar1_end:
    add            bufq, 82
    dec              hd
    jg .y_loop_ar1
.ar0:
    RET

.ar2:
%if ARCH_X86_32
%assign stack_offset_old stack_offset
    ALLOC_STACK -16*8
%endif
    DEFINE_ARGS buf, fg_data, shift
    mov          shiftd, [fg_dataq+FGData.ar_coeff_shift]
    movd             m6, [base+round_vals-12+shiftq*2]
    movd             m7, [base+byte_blend+1]
    SCRATCH           7, 15, 7
    movq             m0, [fg_dataq+FGData.ar_coeffs_y+0]    ; cf0-7
    movd             m1, [fg_dataq+FGData.ar_coeffs_y+8]    ; cf8-11
    pxor             m7, m7
    pshuflw          m6, m6, q0000
    punpcklwd        m6, m7
    pcmpgtb          m4, m7, m0
    pcmpgtb          m5, m7, m1
    punpcklbw        m0, m4
    punpcklbw        m1, m5
    DEFINE_ARGS buf, fg_data, h, x
    pshufd           m4, m1, q0000
    pshufd           m5, m1, q1111
    pshufd           m3, m0, q3333
    pshufd           m2, m0, q2222
    pshufd           m1, m0, q1111
    pshufd           m0, m0, q0000
    SCRATCH           0, 8,  0
    SCRATCH           1, 9,  1
    SCRATCH           2, 10, 2
    SCRATCH           3, 11, 3
    SCRATCH           4, 12, 4
    SCRATCH           5, 13, 5
    SCRATCH           6, 14, 6
    sub            bufq, 82*73-(82*3+79)
    mov              hd, 70
.y_loop_ar2:
    mov              xq, -76

.x_loop_ar2:
    movq             m0, [bufq+xq-82*2-2]   ; y=-2,x=[-2,+5]
    movhps           m0, [bufq+xq-82*1-2]   ; y=-1,x=[-2,+5]
    pcmpgtb          m2, m7, m0
    punpckhbw        m1, m0, m2
    punpcklbw        m0, m2
    psrldq           m5, m0, 2              ; y=-2,x=[-1,+5]
    psrldq           m3, m1, 2              ; y=-1,x=[-1,+5]
    psrldq           m4, m1, 4              ; y=-1,x=[+0,+5]
    punpcklwd        m2, m0, m5
    punpcklwd        m3, m4
    pmaddwd          m2, m8
    pmaddwd          m3, m11
    paddd            m2, m3

    psrldq           m4, m0, 4              ; y=-2,x=[+0,+5]
    psrldq           m5, m0, 6              ; y=-2,x=[+1,+5]
    psrldq           m6, m0, 8              ; y=-2,x=[+2,+5]
    punpcklwd        m4, m5
    punpcklwd        m6, m1
    psrldq           m5, m1, 6              ; y=-1,x=[+1,+5]
    psrldq           m1, m1, 8              ; y=-1,x=[+2,+5]
    punpcklwd        m5, m1
    pmaddwd          m4, m9
    pmaddwd          m6, m10
    pmaddwd          m5, m12
    paddd            m4, m6
    paddd            m2, m5
    paddd            m2, m4
    paddd            m2, m14

    movq             m0, [bufq+xq-2]        ; y=0,x=[-2,+5]
.x_loop_ar2_inner:
    pcmpgtb          m4, m7, m0
    punpcklbw        m1, m0, m4
    pmaddwd          m3, m1, m13
    paddd            m3, m2
    psrldq           m1, 4                  ; y=0,x=0
    psrldq           m2, 4                  ; shift top to next pixel
    psrad            m3, [fg_dataq+FGData.ar_coeff_shift]
    ; don't packssdw since we only care about one value
    paddw            m3, m1
    packsswb         m3, m3
    pslldq           m3, 2
    pand             m3, m15
    pandn            m1, m15, m0
    por              m0, m1, m3
    psrldq           m0, 1
    ; overwrite 2 pixels, but that's ok
    movd      [bufq+xq-1], m0
    inc              xq
    jz .x_loop_ar2_end
    test             xq, 3
    jnz .x_loop_ar2_inner
    jmp .x_loop_ar2

.x_loop_ar2_end:
    add            bufq, 82
    dec              hd
    jg .y_loop_ar2
    RET

.ar3:
    DEFINE_ARGS buf, fg_data, shift
%if ARCH_X86_32
%assign stack_offset stack_offset_old
    ALLOC_STACK  -16*14
%elif WIN64
    SUB             rsp, 16*6
%assign stack_size_padded (stack_size_padded+16*6)
%assign stack_size (stack_size+16*6)
%else
    ALLOC_STACK  -16*6
%endif
    mov          shiftd, [fg_dataq+FGData.ar_coeff_shift]
    movd             m6, [base+round_vals-12+shiftq*2]
    movd             m7, [base+byte_blend]
    movu             m0, [fg_dataq+FGData.ar_coeffs_y+ 0]   ; cf0-15
    movq             m2, [fg_dataq+FGData.ar_coeffs_y+16]   ; cf16-23
    pxor             m3, m3
    pcmpgtb          m4, m3, m0
    pcmpgtb          m3, m2
    pshuflw          m6, m6, q0000
    SCRATCH           6, 14, 12
    SCRATCH           7, 15, 13
    punpckhbw        m1, m0, m4
    punpcklbw        m0, m4
    punpcklbw        m2, m3
    pshufd           m3, m0, q1111
    pshufd           m4, m0, q2222
    pshufd           m5, m0, q3333
    pshufd           m0, m0, q0000
    mova    [rsp+ 0*16], m0
    mova    [rsp+ 1*16], m3
    mova    [rsp+ 2*16], m4
    mova    [rsp+ 3*16], m5
    pshufd           m6, m1, q1111
    pshufd           m7, m1, q2222
    pshufd           m5, m1, q3333
    pshufd           m1, m1, q0000
    pshufd           m3, m2, q1111
    psrldq           m0, m2, 10
    pinsrw           m2, [base+pw_1], 5
    pshufd           m4, m2, q2222
    pshufd           m2, m2, q0000
    pinsrw           m0, [base+round_vals+shiftq*2-10], 3
    mova    [rsp+ 4*16], m1
    mova    [rsp+ 5*16], m6
    SCRATCH           7, 8,  6
    SCRATCH           5, 9,  7
    SCRATCH           2, 10, 8
    SCRATCH           3, 11, 9
    SCRATCH           4, 12, 10
    SCRATCH           0, 13, 11
    DEFINE_ARGS buf, fg_data, h, x
    sub            bufq, 82*73-(82*3+79)
    mov              hd, 70
.y_loop_ar3:
    mov              xq, -76

.x_loop_ar3:
    movu             m0, [bufq+xq-82*3-3]   ; y=-3,x=[-3,+12]
    pxor             m3, m3
    pcmpgtb          m3, m0
    punpckhbw        m2, m0, m3
    punpcklbw        m0, m3

    psrldq           m5, m0, 2
    psrldq           m6, m0, 4
    psrldq           m7, m0, 6
    punpcklwd        m4, m0, m5
    punpcklwd        m6, m7
    pmaddwd          m4, [rsp+ 0*16]
    pmaddwd          m6, [rsp+ 1*16]
    paddd            m4, m6

    movu             m1, [bufq+xq-82*2-3]   ; y=-2,x=[-3,+12]
    pxor             m5, m5
    pcmpgtb          m5, m1
    punpckhbw        m3, m1, m5
    punpcklbw        m1, m5
    palignr          m6, m2, m0, 10
    palignr          m7, m2, m0, 12
    psrldq           m0, 8
    punpcklwd        m0, m6
    punpcklwd        m7, m1
    pmaddwd          m0, [rsp+ 2*16]
    pmaddwd          m7, [rsp+ 3*16]
    paddd            m0, m7
    paddd            m0, m4

    psrldq           m4, m1, 2
    psrldq           m5, m1, 4
    psrldq           m6, m1, 6
    psrldq           m7, m1, 8
    punpcklwd        m4, m5
    punpcklwd        m6, m7
    pmaddwd          m4, [rsp+ 4*16]
    pmaddwd          m6, [rsp+ 5*16]
    paddd            m4, m6
    paddd            m0, m4

    movu             m2, [bufq+xq-82*1-3]   ; y=-1,x=[-3,+12]
    pxor             m7, m7
    pcmpgtb          m7, m2
    punpckhbw        m5, m2, m7
    punpcklbw        m2, m7
    palignr          m7, m3, m1, 10
    palignr          m3, m1, 12
    psrldq           m1, m2, 2
    punpcklwd        m7, m3
    punpcklwd        m3, m2, m1
    pmaddwd          m7, m8
    pmaddwd          m3, m9
    paddd            m7, m3
    paddd            m0, m7

    psrldq           m6, m2, 4
    psrldq           m1, m2, 6
    psrldq           m3, m2, 8
    palignr          m4, m5, m2, 10
    palignr          m5, m5, m2, 12

    punpcklwd        m6, m1
    punpcklwd        m3, m4
    punpcklwd        m5, m14
    pmaddwd          m6, m10
    pmaddwd          m3, m11
    pmaddwd          m5, m12
    paddd            m0, m6
    paddd            m3, m5
    paddd            m0, m3

    movq             m1, [bufq+xq-3]        ; y=0,x=[-3,+4]
.x_loop_ar3_inner:
    pxor             m5, m5
    pcmpgtb          m5, m1
    punpcklbw        m2, m1, m5
    pmaddwd          m2, m13
    pshufd           m3, m2, q1111
    paddd            m2, m3                 ; left+cur
    paddd            m2, m0                 ; add top
    psrldq           m0, 4
    psrad            m2, [fg_dataq+FGData.ar_coeff_shift]
    ; don't packssdw since we only care about one value
    packsswb         m2, m2
    pslldq           m2, 3
    pand             m2, m15
    pandn            m3, m15, m1
    por              m1, m2, m3
    movd    [bufq+xq-3], m1
    psrldq           m1, 1
    inc              xq
    jz .x_loop_ar3_end
    test             xq, 3
    jnz .x_loop_ar3_inner
    jmp .x_loop_ar3

.x_loop_ar3_end:
    add            bufq, 82
    dec              hd
    jg .y_loop_ar3
    RET

%macro generate_grain_uv_fn 3 ; ss_name, ss_x, ss_y
INIT_XMM ssse3
cglobal generate_grain_uv_%1_8bpc, 1, 7 + 3 * ARCH_X86_64, 16, buf, bufy, fg_data, uv
    movifnidn        r2, r2mp
    movifnidn        r3, r3mp
    LEA              r4, $$
%define base r4-$$
    movq             m1, [base+rnd_next_upperbit_mask]
    movq             m4, [base+mul_bits]
    movq             m7, [base+hmul_bits]
    mov             r5d, [fg_dataq+FGData.grain_scale_shift]
    movd             m6, [base+round+r5*2]
    mova             m5, [base+pb_mask]
    movd             m0, [fg_dataq+FGData.seed]
    movd             m2, [base+pw_seed_xor+uvq*4]
    pxor             m0, m2
    pshuflw          m6, m6, q0000
    pshuflw          m0, m0, q0000
    lea              r6, [base+gaussian_sequence]
%if %2
%if ARCH_X86_64
    mov             r7d, 73-35*%3
%else
    mov            r3mp, 73-35*%3
%endif
    add            bufq, 44
.loop_y:
    mov              r5, -44
.loop_x:
%else
    mov              r5, -82*73
    sub            bufq, r5
.loop:
%endif
    pand             m2, m0, m1
    psrlw            m3, m2, 10
    por              m2, m3             ; bits 0xf, 0x1e, 0x3c and 0x78 are set
    pmullw           m2, m4             ; bits 0x0f00 are set
    pshufb           m3, m5, m2         ; set 15th bit for next 4 seeds
    psllq            m2, m3, 30
    por              m3, m2
    psllq            m2, m3, 15
    por              m3, m2             ; aggregate each bit into next seed's high bit
    pmulhuw          m2, m0, m7
    por              m2, m3             ; 4 next output seeds
    pshuflw          m0, m2, q3333
    psrlw            m2, 5
%if ARCH_X86_64
    movd            r9d, m2
    pshuflw          m2, m2, q3232
    movzx            r8, r9w
    shr              r9, 16

    movd             m3, [r6+r8*2]
    pinsrw           m3, [r6+r9*2], 1

    movd            r9d, m2
    movzx            r8, r9w
    shr              r9, 16

    pinsrw           m3, [r6+r8*2], 2
    pinsrw           m3, [r6+r9*2], 3
%else
    movd             r2, m2
    pshuflw          m2, m2, q3232
    movzx            r1, r2w
    shr              r2, 16

    movd             m3, [r6+r1*2]
    pinsrw           m3, [r6+r2*2], 1

    movd             r2, m2
    movzx            r1, r2w
    shr              r2, 16

    pinsrw           m3, [r6+r1*2], 2
    pinsrw           m3, [r6+r2*2], 3
%endif
    pmulhrsw         m3, m6
    packsswb         m3, m3
    movd      [bufq+r5], m3
    add              r5, 4
%if %2
    jl .loop_x
    add            bufq, 82
%if ARCH_X86_64
    dec             r7d
%else
    dec            r3mp
%endif
    jg .loop_y
%else
    jl .loop
%endif

%if ARCH_X86_32
    mov              r2, r2mp
%endif

    ; auto-regression code
    movsxd           r5, [fg_dataq+FGData.ar_coeff_lag]
    movsxd           r5, [base+generate_grain_uv_%1_8bpc_ssse3_table+r5*4]
    lea              r5, [r5+base+generate_grain_uv_%1_8bpc_ssse3_table]
    jmp              r5

.ar0:
    DEFINE_ARGS buf, bufy, fg_data, uv, unused, shift
    movifnidn     bufyq, bufymp
%if ARCH_X86_32
%assign stack_offset_old stack_offset
    ALLOC_STACK   -2*16
%endif
    imul            uvd, 28
    mov          shiftd, [fg_dataq+FGData.ar_coeff_shift]
    movd             m5, [fg_dataq+FGData.ar_coeffs_uv+uvq]
    movd             m4, [base+hmul_bits+shiftq*2]
    DEFINE_ARGS buf, bufy, h, x
    pxor             m0, m0
    pcmpgtb          m0, m5
    punpcklbw        m5, m0
    movd             m7, [base+pb_1]
%if %2
    movd             m6, [base+hmul_bits+2+%3*2]
%endif
    pshuflw          m5, m5, q0000
    pshuflw          m4, m4, q0000
    pshufd           m7, m7, q0000
%if %2
    pshuflw          m6, m6, q0000
%endif
    punpcklqdq       m5, m5
    punpcklqdq       m4, m4
%if %2
    punpcklqdq       m6, m6
%endif
    pcmpeqw          m1, m1
    pslldq           m1, 12>>%2
    SCRATCH           1, 8, 0
    SCRATCH           4, 9, 1
%if %2
    sub            bufq, 82*(73-35*%3)+82-(82*3+41)
%else
    sub            bufq, 82*70-3
%endif
    add           bufyq, 3+82*3
    mov              hd, 70-35*%3
.y_loop_ar0:
    xor              xd, xd
.x_loop_ar0:
    ; first 32 pixels
%if %2
    movu             m1, [bufyq+xq*2]
%if %3
    movu             m2, [bufyq+xq*2+82]
%endif
    movu             m3, [bufyq+xq*2+16]
%if %3
    movu             m4, [bufyq+xq*2+82+16]
%endif
    pmaddubsw        m0, m7, m1
%if %3
    pmaddubsw        m1, m7, m2
%endif
    pmaddubsw        m2, m7, m3
%if %3
    pmaddubsw        m3, m7, m4
    paddw            m0, m1
    paddw            m2, m3
%endif
    pmulhrsw         m0, m6
    pmulhrsw         m2, m6
%else
    movu             m0, [bufyq+xq]
    pxor             m6, m6
    pcmpgtb          m6, m0
    punpckhbw        m2, m0, m6
    punpcklbw        m0, m6
%endif
    pmullw           m0, m5
    pmullw           m2, m5
    pmulhrsw         m0, m9
    pmulhrsw         m2, m9
    movu             m1, [bufq+xq]
    pxor             m4, m4
    pcmpgtb          m4, m1
    punpckhbw        m3, m1, m4
%if %2
    punpcklbw        m1, m4
    paddw            m2, m3
    paddw            m0, m1
%else
    punpcklbw        m6, m1, m4
    paddw            m2, m3
    paddw            m0, m6
%endif
    packsswb         m0, m2
%if %2
    movu      [bufq+xq], m0
    add              xd, 16
    cmp              xd, 32
    jl .x_loop_ar0

    ; last 6/12 pixels
    movu             m1, [bufyq+xq*(1+%2)]
%if %3
    movu             m2, [bufyq+xq*2+82]
%endif
    pmaddubsw        m0, m7, m1
%if %3
    pmaddubsw        m1, m7, m2
    paddw            m0, m1
%endif
    pmulhrsw         m0, m6
    pmullw           m0, m5
    pmulhrsw         m0, m9
    movq             m1, [bufq+xq]
    pxor             m4, m4
    pcmpgtb          m4, m1
    punpcklbw        m2, m1, m4
    paddw            m0, m2
    packsswb         m0, m0
    pandn            m2, m8, m0
    pand             m1, m8
    por              m2, m1
    movq      [bufq+xq], m2
%else
    add              xd, 16
    cmp              xd, 80
    je .y_loop_final_ar0
    movu   [bufq+xq-16], m0
    jmp .x_loop_ar0
.y_loop_final_ar0:
    pandn            m2, m8, m0
    pand             m1, m8
    por              m2, m1
    movu   [bufq+xq-16], m2
%endif

    add            bufq, 82
    add           bufyq, 82<<%3
    dec              hd
    jg .y_loop_ar0
    RET

.ar1:
%if ARCH_X86_32
%assign stack_offset stack_offset_old
%assign stack_size_padded 0
%xdefine rstk rsp
%endif
    DEFINE_ARGS buf, bufy, fg_data, uv, val3, cf3, min, max, x
    imul            uvd, 28
    movsx          cf3d, byte [fg_dataq+FGData.ar_coeffs_uv+uvq+3]
    movd             m4, [fg_dataq+FGData.ar_coeffs_uv+uvq-1]
    pinsrw           m4, [fg_dataq+FGData.ar_coeffs_uv+uvq+4], 2
%if ARCH_X86_32
    mov            r3mp, cf3d
    DEFINE_ARGS buf, shift, fg_data, val3, min, max, x
%elif WIN64
    DEFINE_ARGS shift, bufy, fg_data, buf, val3, cf3, min, max, x
    mov            bufq, r0
%else
    DEFINE_ARGS buf, bufy, fg_data, shift, val3, cf3, min, max, x
%endif
    mov          shiftd, [fg_dataq+FGData.ar_coeff_shift]
    movd             m3, [base+round_vals+shiftq*2-12]    ; rnd
%if %2
    movd             m7, [base+pb_1]
    movd             m6, [base+hmul_bits+2+%3*2]
%endif
    psrldq           m4, 1
%if ARCH_X86_32
    DEFINE_ARGS buf, shift, val0, val3, min, max, x
%elif WIN64
    DEFINE_ARGS shift, bufy, h, buf, val3, cf3, min, max, x, val0
%else
    DEFINE_ARGS buf, bufy, h, shift, val3, cf3, min, max, x, val0
%endif
    pxor             m5, m5
    punpcklwd        m3, m5
%if %2
    punpcklwd        m6, m6
%endif
    pcmpgtb          m5, m4
    punpcklbw        m4, m5
    pshufd           m5, m4, q1111
    pshufd           m4, m4, q0000
    pshufd           m3, m3, q0000
%if %2
    pshufd           m7, m7, q0000
    pshufd           m6, m6, q0000
    sub            bufq, 82*(73-35*%3)+44-(82*3+41)
%else
    sub            bufq, 82*69+3
%endif
%if ARCH_X86_32
    add            r1mp, 79+82*3
    mov            r0mp, 70-35*%3
%else
    add           bufyq, 79+82*3
    mov              hd, 70-35*%3
%endif
    mov            mind, -128
    mov            maxd, 127
.y_loop_ar1:
    mov              xq, -(76>>%2)
    movsx         val3d, byte [bufq+xq-1]
.x_loop_ar1:
%if %2
%if ARCH_X86_32
    mov              r2, r1mp
    movq             m0, [r2+xq*2]
%if %3
    movq             m1, [r2+xq*2+82]
%endif
%else
    movq             m0, [bufyq+xq*2]
%if %3
    movq             m1, [bufyq+xq*2+82]
%endif
%endif
    pmaddubsw        m2, m7, m0
%if %3
    pmaddubsw        m0, m7, m1
    paddw            m2, m0
%endif
    pmulhrsw         m2, m6
%else
%if ARCH_X86_32
    mov              r2, r1mp
    movd             m2, [r2+xq]
%else
    movd             m2, [bufyq+xq]
%endif
    pxor             m0, m0
    pcmpgtb          m0, m2
    punpcklbw        m2, m0
%endif

    movq             m0, [bufq+xq-82-1]     ; top/left
    pxor             m1, m1
    pcmpgtb          m1, m0
    punpcklbw        m0, m1
    psrldq           m1, m0, 4              ; top/right
    punpcklwd        m1, m2
    psrldq           m2, m0, 2              ; top
    punpcklwd        m0, m2
    pmaddwd          m0, m4
    pmaddwd          m1, m5
    paddd            m0, m1
    paddd            m0, m3
.x_loop_ar1_inner:
    movd          val0d, m0
    psrldq           m0, 4
%if ARCH_X86_32
    imul          val3d, r3mp
%else
    imul          val3d, cf3d
%endif
    add           val3d, val0d
    sar           val3d, shiftb
    movsx         val0d, byte [bufq+xq]
    add           val3d, val0d
    cmp           val3d, maxd
    cmovns        val3d, maxd
    cmp           val3d, mind
    cmovs         val3d, mind
    mov  byte [bufq+xq], val3b
    ; keep val3d in-place as left for next x iteration
    inc              xq
    jz .x_loop_ar1_end
    test             xq, 3
    jnz .x_loop_ar1_inner
    jmp .x_loop_ar1

.x_loop_ar1_end:
    add            bufq, 82
%if ARCH_X86_32
    add            r1mp, 82<<%3
    dec            r0mp
%else
    add           bufyq, 82<<%3
    dec              hd
%endif
    jg .y_loop_ar1
    RET

.ar2:
%if ARCH_X86_32
%assign stack_offset stack_offset_old
%assign stack_size_padded 0
%xdefine rstk rsp
    ALLOC_STACK   -8*16
%endif
    DEFINE_ARGS buf, bufy, fg_data, uv, unused, shift
    movifnidn     bufyq, bufymp
    mov          shiftd, [fg_dataq+FGData.ar_coeff_shift]
    imul            uvd, 28
    movd             m7, [base+round_vals-12+shiftq*2]
    movu             m0, [fg_dataq+FGData.ar_coeffs_uv+uvq+0]   ; cf0-12
    pxor             m2, m2
    pcmpgtb          m2, m0
    punpckhbw        m1, m0, m2
    punpcklbw        m0, m2
    pinsrw           m1, [base+pw_1], 5
    punpcklwd        m7, m7
    pshufd           m7, m7, q0000
    DEFINE_ARGS buf, bufy, fg_data, h, unused, x
    pshufd           m4, m1, q0000
    pshufd           m5, m1, q1111
    pshufd           m6, m1, q2222
    pshufd           m3, m0, q3333
    pshufd           m2, m0, q2222
    pshufd           m1, m0, q1111
    pshufd           m0, m0, q0000
    SCRATCH           0, 8,  0
    SCRATCH           1, 9,  1
    SCRATCH           2, 10, 2
    SCRATCH           3, 11, 3
    SCRATCH           4, 12, 4
    SCRATCH           5, 13, 5
    SCRATCH           6, 14, 6
    SCRATCH           7, 15, 7
%if %2
    movd             m7, [base+hmul_bits+2+%3*2]
    movd             m6, [base+pb_1]
    punpcklwd        m7, m7
    pshufd           m6, m6, q0000
    pshufd           m7, m7, q0000
    sub            bufq, 82*(73-35*%3)+44-(82*3+41)
%else
    sub            bufq, 82*69+3
%endif
    add           bufyq, 79+82*3
    mov              hd, 70-35*%3
.y_loop_ar2:
    mov              xq, -(76>>%2)

.x_loop_ar2:
    pxor             m2, m2
    movq             m0, [bufq+xq-82*2-2]   ; y=-2,x=[-2,+5]
    movhps           m0, [bufq+xq-82*1-2]   ; y=-1,x=[-2,+5]
    pcmpgtb          m2, m0
    punpckhbw        m1, m0, m2
    punpcklbw        m0, m2
    psrldq           m5, m0, 2              ; y=-2,x=[-1,+5]
    psrldq           m3, m1, 2              ; y=-1,x=[-1,+5]
    psrldq           m4, m1, 4              ; y=-1,x=[+0,+5]
    punpcklwd        m2, m0, m5
    punpcklwd        m3, m4
    pmaddwd          m2, m8
    pmaddwd          m3, m11
    paddd            m2, m3

    psrldq           m4, m0, 4              ; y=-2,x=[+0,+5]
    psrldq           m5, m0, 6              ; y=-2,x=[+1,+5]
    psrldq           m0, 8                  ; y=-2,x=[+2,+5]
    punpcklwd        m4, m5
    punpcklwd        m0, m1
    psrldq           m3, m1, 6              ; y=-1,x=[+1,+5]
    psrldq           m1, m1, 8              ; y=-1,x=[+2,+5]
    punpcklwd        m3, m1
    pmaddwd          m4, m9
    pmaddwd          m0, m10
    pmaddwd          m3, m12
    paddd            m4, m0
    paddd            m2, m3
    paddd            m2, m4

%if %2
    movq             m1, [bufyq+xq*2]
%if %3
    movq             m3, [bufyq+xq*2+82]
%endif
    pmaddubsw        m0, m6, m1
%if %3
    pmaddubsw        m1, m6, m3
    paddw            m0, m1
%endif
    pmulhrsw         m0, m7
%else
    movd             m0, [bufyq+xq]
    pxor             m1, m1
    pcmpgtb          m1, m0
    punpcklbw        m0, m1
%endif
    punpcklwd        m0, m15
    pmaddwd          m0, m14
    paddd            m2, m0

    movq             m0, [bufq+xq-2]        ; y=0,x=[-2,+5]
    pxor             m4, m4
    movd             m5, [base+byte_blend+1]
    punpcklbw        m5, m5
.x_loop_ar2_inner:
    pcmpgtb          m1, m4, m0
    punpcklbw        m0, m1
    pmaddwd          m3, m0, m13
    paddd            m3, m2
    psrldq           m2, 4                  ; shift top to next pixel
    psrad            m3, [fg_dataq+FGData.ar_coeff_shift]
    pslldq           m3, 4
    pand             m3, m5
    paddw            m0, m3
    packsswb         m0, m0
    movd    [bufq+xq-2], m0
    psrldq           m0, 1
    inc              xq
    jz .x_loop_ar2_end
    test             xq, 3
    jnz .x_loop_ar2_inner
    jmp .x_loop_ar2

.x_loop_ar2_end:
    add            bufq, 82
    add           bufyq, 82<<%3
    dec              hd
    jg .y_loop_ar2
    RET

.ar3:
%if ARCH_X86_32
%assign stack_offset stack_offset_old
%assign stack_size_padded 0
%xdefine rstk rsp
%endif
    DEFINE_ARGS buf, bufy, fg_data, uv, unused, shift
    movifnidn     bufyq, bufymp
%if ARCH_X86_32
    ALLOC_STACK  -15*16
%else
    SUB             rsp, 16*7
%assign stack_size_padded (stack_size_padded+16*7)
%assign stack_size (stack_size+16*7)
%endif
    mov          shiftd, [fg_dataq+FGData.ar_coeff_shift]
    imul            uvd, 28

    movu             m0, [fg_dataq+FGData.ar_coeffs_uv+uvq+ 0]   ; cf0-15
    pxor             m3, m3
    pcmpgtb          m3, m0
    punpckhbw        m1, m0, m3
    punpcklbw        m0, m3
    pshufd           m2, m0, q1111
    pshufd           m3, m0, q2222
    pshufd           m4, m0, q3333
    pshufd           m0, m0, q0000
    pshufd           m5, m1, q1111
    pshufd           m6, m1, q2222
    pshufd           m7, m1, q3333
    pshufd           m1, m1, q0000
    mova    [rsp+ 0*16], m0
    mova    [rsp+ 1*16], m2
    mova    [rsp+ 2*16], m3
    mova    [rsp+ 3*16], m4
    mova    [rsp+ 4*16], m1
    mova    [rsp+ 5*16], m5
    mova    [rsp+ 6*16], m6
    SCRATCH           7, 8, 7

    movu             m2, [fg_dataq+FGData.ar_coeffs_uv+uvq+16]   ; cf16-24 [24=luma]
    pxor             m4, m4
    pcmpgtb          m4, m2
    punpckhbw        m5, m2, m4
    punpcklbw        m2, m4
    pshufd           m4, m2, q3232
    punpcklwd        m3, m4, m5
    pshuflw          m5, m4, q3321
    pshufd           m4, m3, q0000
    pshufd           m3, m2, q1111
    pshufd           m2, m2, q0000
    pinsrw           m5, [base+round_vals+shiftq*2-10], 3
    SCRATCH           2, 9,  8
    SCRATCH           3, 10, 9
    SCRATCH           4, 11, 10
    SCRATCH           5, 12, 11

    movd             m2, [base+round_vals-12+shiftq*2]
%if %2
    movd             m1, [base+pb_1]
    movd             m3, [base+hmul_bits+2+%3*2]
%endif
    pxor             m0, m0
    punpcklwd        m2, m0
%if %2
    punpcklwd        m3, m3
%endif
    pshufd           m2, m2, q0000
%if %2
    pshufd           m1, m1, q0000
    pshufd           m3, m3, q0000
    SCRATCH           1, 13, 12
%endif
    SCRATCH           2, 14, 13
%if %2
    SCRATCH           3, 15, 14
%endif

    DEFINE_ARGS buf, bufy, fg_data, h, unused, x
%if %2
    sub            bufq, 82*(73-35*%3)+44-(82*3+41)
%else
    sub            bufq, 82*69+3
%endif
    add           bufyq, 79+82*3
    mov              hd, 70-35*%3
.y_loop_ar3:
    mov              xq, -(76>>%2)

.x_loop_ar3:
    movu             m0, [bufq+xq-82*3-3]   ; y=-3,x=[-3,+12]
    pxor             m4, m4
    pcmpgtb          m4, m0
    punpckhbw        m3, m0, m4
    punpcklbw        m0, m4

    psrldq           m5, m0, 2
    psrldq           m6, m0, 4
    psrldq           m7, m0, 6
    punpcklwd        m4, m0, m5
    punpcklwd        m6, m7
    pmaddwd          m4, [rsp+ 0*16]
    pmaddwd          m6, [rsp+ 1*16]
    paddd            m4, m6

    palignr          m2, m3, m0, 10
    palignr          m3, m0, 12
    psrldq           m0, 8

    movu             m1, [bufq+xq-82*2-3]   ; y=-2,x=[-3,+12]
    pxor             m6, m6
    pcmpgtb          m6, m1
    punpckhbw        m5, m1, m6
    punpcklbw        m1, m6

    punpcklwd        m0, m2
    punpcklwd        m3, m1
    pmaddwd          m0, [rsp+ 2*16]
    pmaddwd          m3, [rsp+ 3*16]
    paddd            m0, m3
    paddd            m0, m4

    movu             m2, [bufq+xq-82*1-3]   ; y=-1,x=[-3,+12]
    pxor             m7, m7
    pcmpgtb          m7, m2
    punpckhbw        m6, m2, m7
    punpcklbw        m2, m7

    palignr          m3, m5, m1, 10
    palignr          m5, m1, 12
    psrldq           m4, m2, 2

    punpcklwd        m3, m5
    punpcklwd        m5, m2, m4
    pmaddwd          m3, [rsp+ 6*16]
    pmaddwd          m5, m8
    paddd            m3, m5
    paddd            m0, m3

    psrldq           m3, m1, 2
    psrldq           m4, m1, 4
    psrldq           m5, m1, 6
    psrldq           m1, 8

    punpcklwd        m3, m4
    punpcklwd        m5, m1
    pmaddwd          m3, [rsp+ 4*16]
    pmaddwd          m5, [rsp+ 5*16]
    paddd            m3, m5
    paddd            m0, m3

%if %2
    movq             m1, [bufyq+xq*2]
%if %3
    movq             m3, [bufyq+xq*2+82]
%endif
    pmaddubsw        m7, m13, m1
%if %3
    pmaddubsw        m5, m13, m3
    paddw            m7, m5
%endif
    pmulhrsw         m7, m15
%else
    movd             m7, [bufyq+xq]
    pxor             m1, m1
    pcmpgtb          m1, m7
    punpcklbw        m7, m1
%endif

    psrldq           m1, m2, 4
    psrldq           m3, m2, 6
    palignr          m4, m6, m2, 10
    palignr          m6, m2, 12
    psrldq           m2, 8

    punpcklwd        m1, m3
    punpcklwd        m2, m4
    punpcklwd        m6, m7
    pmaddwd          m1, m9
    pmaddwd          m2, m10
    pmaddwd          m6, m11
    paddd            m1, m2
    paddd            m0, m6
    paddd            m0, m1
    paddd            m0, m14

    movq             m1, [bufq+xq-3]        ; y=0,x=[-3,+4]
    pxor             m4, m4
    movd             m5, [base+byte_blend]
.x_loop_ar3_inner:
    pcmpgtb          m2, m4, m1
    punpcklbw        m3, m1, m2
    pmaddwd          m2, m3, m12
    pshufd           m3, m2, q1111
    paddd            m2, m3                 ; left+cur
    paddd            m2, m0                 ; add top
    psrldq           m0, 4
    psrad            m2, [fg_dataq+FGData.ar_coeff_shift]
    ; don't packssdw, we only care about one value
    packsswb         m2, m2
    pandn            m3, m5, m1
    pslld            m2, 24
    pand             m2, m5
    por              m1, m2, m3
    movd    [bufq+xq-3], m1
    psrldq           m1, 1
    inc              xq
    jz .x_loop_ar3_end
    test             xq, 3
    jnz .x_loop_ar3_inner
    jmp .x_loop_ar3

.x_loop_ar3_end:
    add            bufq, 82
    add           bufyq, 82<<%3
    dec              hd
    jg .y_loop_ar3
    RET
%endmacro

generate_grain_uv_fn 420, 1, 1
generate_grain_uv_fn 422, 1, 0
generate_grain_uv_fn 444, 0, 0

%macro vpgatherdw 5-6 ; dst, src, base, tmp_gpr[x2], tmp_xmm_reg
%assign %%idx 0
%define %%tmp %2
%if %0 == 6
%define %%tmp %6
%endif
%rep 4
%if %%idx == 0
    movd        %5 %+ d, %2
    pshuflw       %%tmp, %2, q3232
%else
    movd        %5 %+ d, %%tmp
%if %%idx == 2
    punpckhqdq    %%tmp, %%tmp
%elif %%idx == 4
    psrlq         %%tmp, 32
%endif
%endif
    movzx       %4 %+ d, %5 %+ w
    shr         %5 %+ d, 16

%if %%idx == 0
    movd             %1, [%3+%4]
%else
    pinsrw           %1, [%3+%4], %%idx + 0
%endif
    pinsrw           %1, [%3+%5], %%idx + 1
%assign %%idx %%idx+2
%endrep
%endmacro

INIT_XMM ssse3
; fgy_32x32xn(dst, src, stride, fg_data, w, scaling, grain_lut, h, sby)
%if ARCH_X86_32
%if STACK_ALIGNMENT < mmsize
cglobal fgy_32x32xn_8bpc, 0, 7, 16, 0 - (5 * mmsize + 16 * gprsize), \
        dst, src, scaling, unused1, fg_data, picptr, unused2
    ; copy stack arguments to new position post-alignment, so that we
    ; don't have to keep the old stack location in a separate register
    mov              r0, r0m
    mov              r1, r2m
    mov              r2, r4m
    mov              r3, r6m
    mov              r4, r7m
    mov              r5, r8m

    mov [rsp+5*mmsize+ 4*gprsize], r0
    mov [rsp+5*mmsize+ 6*gprsize], r1
    mov [rsp+5*mmsize+ 8*gprsize], r2
    mov [rsp+5*mmsize+10*gprsize], r3
    mov [rsp+5*mmsize+11*gprsize], r4
    mov [rsp+5*mmsize+12*gprsize], r5
%else
cglobal fgy_32x32xn_8bpc, 0, 7, 16, 5 * mmsize + 4 * gprsize, \
        dst, src, scaling, unused1, fg_data, picptr, unused2
%endif
    mov            srcq, srcm
    mov        fg_dataq, r3m
    mov        scalingq, r5m
%if STACK_ALIGNMENT < mmsize
%define r0m [rsp+5*mmsize+ 4*gprsize]
%define r1m [rsp+5*mmsize+ 5*gprsize]
%define r2m [rsp+5*mmsize+ 6*gprsize]
%define r3m [rsp+5*mmsize+ 7*gprsize]
%define r4m [rsp+5*mmsize+ 8*gprsize]
%define r5m [rsp+5*mmsize+ 9*gprsize]
%define r6m [rsp+5*mmsize+10*gprsize]
%define r7m [rsp+5*mmsize+11*gprsize]
%define r8m [rsp+5*mmsize+12*gprsize]
%endif
    LEA              r5, pb_mask
%define base r5-pb_mask
    mov             r5m, picptrq
%else
cglobal fgy_32x32xn_8bpc, 6, 15, 16, dst, src, stride, fg_data, w, scaling, grain_lut
    lea              r7, [pb_mask]
%define base r7-pb_mask
%endif
    mov             r6d, [fg_dataq+FGData.scaling_shift]
    movd             m3, [base+mul_bits+r6*2-14]
    mov             r6d, [fg_dataq+FGData.clip_to_restricted_range]
    movd             m4, [base+max+r6*4]
    movd             m5, [base+min+r6*2]
    punpcklwd        m3, m3
    punpcklwd        m4, m4
    punpcklwd        m5, m5
    pshufd           m3, m3, q0000
    pshufd           m4, m4, q0000
    pshufd           m5, m5, q0000
    SCRATCH           3, 11, 0
    SCRATCH           4, 12, 1
    SCRATCH           5, 13, 2

%if ARCH_X86_32
    DEFINE_ARGS dst, src, scaling, sby, fg_data, picptr, overlap
%else
    DEFINE_ARGS dst, src, stride, fg_data, w, scaling, grain_lut, unused, sby, see, overlap
%endif

    mov            sbyd, r8m
    mov        overlapd, [fg_dataq+FGData.overlap_flag] ; left_overlap: overlap & 1
    test       overlapd, overlapd
    jz .no_vertical_overlap
    mova             m6, [base+pw_1024]
    mova             m7, [base+pb_27_17_17_27]
    SCRATCH           6, 14, 3
    SCRATCH           7, 15, 4
    test           sbyd, sbyd
    jnz .vertical_overlap
    ; fall-through

.no_vertical_overlap:
    mov             r8m, overlapd
%if ARCH_X86_32
    DEFINE_ARGS dst, src, scaling, see, fg_data, picptr, unused
    imul           seed, (173 << 24) | 37
%else
    imul           seed, sbyd, (173 << 24) | 37
%endif
    add            seed, (105 << 24) | 178
    rol            seed, 8
    movzx          seed, seew
    xor            seed, [fg_dataq+FGData.seed]

%if ARCH_X86_32
    DEFINE_ARGS dst, src, scaling, see, w, picptr, src_bak

    mov             r3m, seed
    mov              wq, r4m
%else
    DEFINE_ARGS dst, src, stride, src_bak, w, scaling, grain_lut, \
                unused1, unused2, see, unused3
%endif

    lea        src_bakq, [srcq+wq]
    neg              wq
    sub           dstmp, srcq
%if ARCH_X86_32
    mov             r1m, src_bakq
    mov             r4m, wq
    DEFINE_ARGS dst, src, scaling, see, unused1, unused2, unused3
%endif

.loop_x:
%if ARCH_X86_32
    mov            seed, r3m
%endif
    mov             r6d, seed
    or             seed, 0xEFF4
    shr             r6d, 1
    test           seeb, seeh
    lea            seed, [r6+0x8000]
    cmovp          seed, r6d                ; updated seed
%if ARCH_X86_32
    mov             r3m, seed

    DEFINE_ARGS dst, src, scaling, offy, unused1, unused2, offx

    mov           offxd, offyd
%else
    DEFINE_ARGS dst, src, stride, src_bak, w, scaling, grain_lut, \
                offx, offy, see, unused

    mov           offyd, seed
    mov           offxd, seed
%endif
    ror           offyd, 8
    shr           offxd, 12
    and           offyd, 0xf
    imul          offyd, 164
    lea           offyq, [offyq+offxq*2+747] ; offy*stride+offx

%if ARCH_X86_32
    ; r0m=dst, r1m=src_bak, r2m=stride, r3m=see, r4m=w, r5m=picptr,
    ; r6m=grain_lut, r7m=h, r8m=overlap_v|h
    DEFINE_ARGS dst, src, scaling, offxy, h, picptr, grain_lut
%else
    DEFINE_ARGS dst, src, stride, src_bak, w, scaling, grain_lut, \
                h, offxy, see, unused
%endif

.loop_x_odd:
    mov              hd, r7m
    mov      grain_lutq, grain_lutmp
.loop_y:
    ; src
    mova             m0, [srcq]
    pxor             m2, m2
    punpckhbw        m1, m0, m2
    punpcklbw        m0, m2                 ; m0-1: src as word

    ; scaling[src]
%if ARCH_X86_32
    vpgatherdw       m4, m0, scalingq-1, r0, r5, m3
    vpgatherdw       m5, m1, scalingq-1, r0, r5, m3
%else
    vpgatherdw       m4, m0, scalingq-1, r12, r13, m3
    vpgatherdw       m5, m1, scalingq-1, r12, r13, m3
%endif
    REPX {psrlw x, 8}, m4, m5

    ; grain = grain_lut[offy+y][offx+x]
    movu             m3, [grain_lutq+offxyq]
    pcmpgtb          m7, m2, m3
    punpcklbw        m2, m3, m7
    punpckhbw        m3, m7

    ; noise = round2(scaling[src] * grain, scaling_shift)
    pmullw           m2, m4
    pmullw           m3, m5
    pmulhrsw         m2, m11
    pmulhrsw         m3, m11

    ; dst = clip_pixel(src, noise)
    paddw            m0, m2
    paddw            m1, m3
    pmaxsw           m0, m13
    pmaxsw           m1, m13
    pminsw           m0, m12
    pminsw           m1, m12
    packuswb         m0, m1
    movifnidn      dstq, dstmp
    mova    [dstq+srcq], m0

    add            srcq, r2mp
    add      grain_lutq, 82
    dec              hd
    jg .loop_y

%if ARCH_X86_32
    add            r4mp, 16
%else
    add              wq, 16
%endif
    jge .end
%if ARCH_X86_32
    mov            srcq, r1mp
    add            srcq, r4mp
%else
    lea            srcq, [src_bakq+wq]
%endif
    btc       dword r8m, 2
    jc .next_blk

    add          offxyd, 16
    test      dword r8m, 2              ; r8m & 2 = have_top_overlap
    jz .loop_x_odd

%if ARCH_X86_32
    add dword [rsp+5*mmsize+1*gprsize], 16
%else
    add            r11d, 16             ; top_offxyd
%endif
    jnz .loop_x_odd_v_overlap

.next_blk:
    test      dword r8m, 1
    jz .loop_x

    test      dword r8m, 2
    jnz .loop_x_hv_overlap

    ; horizontal overlap (without vertical overlap)
.loop_x_h_overlap:
%if ARCH_X86_32
    ; r0m=dst, r1m=src_bak, r2m=stride, r3m=see, r4m=w, r5m=picptr,
    ; r6m=grain_lut, r7m=h, r8m=overlap_v|h
    DEFINE_ARGS dst, src, scaling, offxy, unused1, unused2, unused3

    add          offxyd, 16                 ; left_offxyd
    mov [rsp+5*mmsize+0*gprsize], offxyd

    DEFINE_ARGS dst, src, scaling, see, unused1, unused2, unused3

    mov            seed, r3m
%else
    DEFINE_ARGS dst, src, stride, src_bak, w, scaling, grain_lut, \
                offx, offy, see, left_offxy

    lea     left_offxyd, [offyd+16]         ; previous column's offy*stride+offx
%endif

    mov             r6d, seed
    or             seed, 0xEFF4
    shr             r6d, 1
    test           seeb, seeh
    lea            seed, [r6+0x8000]
    cmovp          seed, r6d                ; updated seed

%if ARCH_X86_32
    mov             r3m, seed

    DEFINE_ARGS dst, src, scaling, offy, unused1, unused2, offx

    mov           offxd, offyd
%else
    mov           offyd, seed
    mov           offxd, seed
%endif
    ror           offyd, 8
    shr           offxd, 12
    and           offyd, 0xf
    imul          offyd, 164
    lea           offyq, [offyq+offxq*2+747] ; offy*stride+offx

%if ARCH_X86_32
    DEFINE_ARGS dst, src, scaling, offxy, h, picptr, grain_lut
%else
    DEFINE_ARGS dst, src, stride, src_bak, w, scaling, grain_lut, \
                h, offxy, see, left_offxy
%endif

    mov              hd, r7m
    mov      grain_lutq, grain_lutmp
.loop_y_h_overlap:
    ; src
    mova             m0, [srcq]
    pxor             m2, m2
    punpckhbw        m1, m0, m2
    punpcklbw        m0, m2                 ; m0-1: src as word

    ; scaling[src]
%if ARCH_X86_32
    vpgatherdw       m4, m0, scalingq-1, r0, r5, m3
    vpgatherdw       m5, m1, scalingq-1, r0, r5, m3
%else
    vpgatherdw       m4, m0, scalingq-1, r12, r13, m3
    vpgatherdw       m5, m1, scalingq-1, r12, r13, m3
%endif
    REPX {psrlw x, 8}, m4, m5

    ; grain = grain_lut[offy+y][offx+x]
    movu             m3, [grain_lutq+offxyq]
%if ARCH_X86_32
    mov              r5, [rsp+5*mmsize+0*gprsize]
    movd             m7, [grain_lutq+r5]
%else
    movd             m7, [grain_lutq+left_offxyq]
%endif
    punpcklbw        m7, m3
    pmaddubsw        m6, m15, m7
    pmulhrsw         m6, m14
    packsswb         m6, m6
    shufps           m6, m3, q3210
    pcmpgtb          m2, m6
    punpcklbw        m7, m6, m2
    punpckhbw        m6, m2

    ; noise = round2(scaling[src] * grain, scaling_shift)
    pmullw           m7, m4
    pmullw           m6, m5
    pmulhrsw         m7, m11
    pmulhrsw         m6, m11

    ; dst = clip_pixel(src, noise)
    paddw            m0, m7
    paddw            m1, m6
    pmaxsw           m0, m13
    pmaxsw           m1, m13
    pminsw           m0, m12
    pminsw           m1, m12
    packuswb         m0, m1
    movifnidn      dstq, dstmp
    mova    [dstq+srcq], m0

    add            srcq, r2mp
    add      grain_lutq, 82
    dec              hd
    jg .loop_y_h_overlap

%if ARCH_X86_32
    add            r4mp, 16
%else
    add              wq, 16
%endif
    jge .end
%if ARCH_X86_32
    mov            srcq, r1m
    add            srcq, r4m
%else
    lea            srcq, [src_bakq+wq]
%endif
    xor       dword r8m, 4
    add          offxyd, 16

    ; since this half-block had left-overlap, the next does not
    test      dword r8m, 2              ; have_top_overlap
    jz .loop_x_odd
%if ARCH_X86_32
    add dword [rsp+5*mmsize+1*gprsize], 16
%else
    add            r11d, 16             ; top_offxyd
%endif
    jmp .loop_x_odd_v_overlap

.end:
    RET

.vertical_overlap:
%if ARCH_X86_32
    DEFINE_ARGS dst, src, scaling, sby, fg_data, picptr, overlap
%else
    DEFINE_ARGS dst, src, stride, fg_data, w, scaling, grain_lut, tmp, sby, see, overlap
%endif

    or         overlapd, 2                  ; top_overlap: overlap & 2
    mov             r8m, overlapd
    movzx          sbyd, sbyb
%if ARCH_X86_32
    imul             r4, [fg_dataq+FGData.seed], 0x00010001
    DEFINE_ARGS tmp, src, scaling, sby, see, picptr, unused
%else
    imul           seed, [fg_dataq+FGData.seed], 0x00010001
%endif
    imul           tmpd, sbyd, 173 * 0x00010001
    imul           sbyd, 37 * 0x01000100
    add            tmpd, (105 << 16) | 188
    add            sbyd, (178 << 24) | (141 << 8)
    and            tmpd, 0x00ff00ff
    and            sbyd, 0xff00ff00
    xor            seed, tmpd
%if ARCH_X86_32
    xor            sbyd, seed               ; (cur_seed << 16) | top_seed

    DEFINE_ARGS dst, src, scaling, see, w, picptr, src_bak

    mov             r3m, seed
    mov              wq, r4m
%else
    xor            seed, sbyd               ; (cur_seed << 16) | top_seed

    DEFINE_ARGS dst, src, stride, src_bak, w, scaling, grain_lut, \
                tmp, unused2, see, unused3
%endif

    lea        src_bakq, [srcq+wq]
    neg              wq
    sub           dstmp, srcq
%if ARCH_X86_32
    mov             r1m, src_bakq
    mov             r4m, wq
    DEFINE_ARGS tmp, src, scaling, see, unused1, picptr, unused2
%endif

.loop_x_v_overlap:
%if ARCH_X86_32
    mov            seed, r3m
%endif
    ; we assume from the block above that bits 8-15 of tmpd are zero'ed,
    ; because of the 'and tmpd, 0x00ff00ff' above
    mov             r6d, seed
    or             seed, 0xeff4eff4
    test           seeb, seeh
    setp           tmpb                     ; parity of top_seed
    shr            seed, 16
    shl            tmpd, 16
    test           seeb, seeh
    setp           tmpb                     ; parity of cur_seed
    or              r6d, 0x00010001
    xor            tmpd, r6d
    mov            seed, tmpd
    ror            seed, 1                  ; updated (cur_seed << 16) | top_seed

%if ARCH_X86_32
    mov             r3m, seed

    DEFINE_ARGS dst, src, scaling, offy, unused1, unused2, offx

    mov           offxd, offyd
%else
    DEFINE_ARGS dst, src, stride, src_bak, w, scaling, grain_lut, \
                offx, offy, see, unused, top_offxy

    mov           offyd, seed
    mov           offxd, seed
%endif

    ror           offyd, 8
    ror           offxd, 12
    and           offyd, 0xf000f
    and           offxd, 0xf000f
    imul          offyd, 164
    ; offxy=offy*stride+offx, (cur_offxy << 16) | top_offxy
    lea           offyq, [offyq+offxq*2+0x10001*747+32*82]

%if ARCH_X86_32
    DEFINE_ARGS top_offxy, src, scaling, offxy, h, picptr, grain_lut
%else
    DEFINE_ARGS dst, src, stride, src_bak, w, scaling, grain_lut, \
                h, offxy, see, unused, top_offxy
%endif

    movzx    top_offxyd, offxyw
%if ARCH_X86_32
    mov [rsp+5*mmsize+1*gprsize], top_offxyd

    DEFINE_ARGS dst, src, scaling, offxy, h, picptr, grain_lut
%endif
    shr          offxyd, 16

.loop_x_odd_v_overlap:
%if ARCH_X86_32
    mov              r5, r5m
    lea              r5, [base+pb_27_17]
    mov [rsp+5*mmsize+12], r5
%else
    mova             m8, [pb_27_17]
%endif
    mov              hd, r7m
    mov      grain_lutq, grain_lutmp
.loop_y_v_overlap:
    ; src
    mova             m0, [srcq]
    pxor             m2, m2
    punpckhbw        m1, m0, m2
    punpcklbw        m0, m2                 ; m0-1: src as word

    ; scaling[src]
%if ARCH_X86_32
    vpgatherdw       m4, m0, scalingq-1, r0, r5, m3
    vpgatherdw       m5, m1, scalingq-1, r0, r5, m3
%else
    vpgatherdw       m4, m0, scalingq-1, r12, r13, m3
    vpgatherdw       m5, m1, scalingq-1, r12, r13, m3
%endif
    REPX {psrlw x, 8}, m4, m5

    ; grain = grain_lut[offy+y][offx+x]
    movu             m3, [grain_lutq+offxyq]
%if ARCH_X86_32
    mov              r5, [rsp+5*mmsize+1*gprsize]
    movu             m7, [grain_lutq+r5]
%else
    movu             m7, [grain_lutq+top_offxyq]
%endif
    punpckhbw        m6, m7, m3
    punpcklbw        m7, m3
%if ARCH_X86_32
    mov              r5, [rsp+5*mmsize+12]
    pmaddubsw        m3, [r5], m6
    pmaddubsw        m6, [r5], m7
%else
    pmaddubsw        m3, m8, m6
    pmaddubsw        m6, m8, m7
%endif
    pmulhrsw         m3, m14
    pmulhrsw         m6, m14
    packsswb         m6, m3
    pcmpgtb          m7, m2, m6
    punpcklbw        m2, m6, m7
    punpckhbw        m6, m7

    ; noise = round2(scaling[src] * grain, scaling_shift)
    pmullw           m2, m4
    pmullw           m6, m5
    pmulhrsw         m2, m11
    pmulhrsw         m6, m11

    ; dst = clip_pixel(src, noise)
    paddw            m0, m2
    paddw            m1, m6
    pmaxsw           m0, m13
    pmaxsw           m1, m13
    pminsw           m0, m12
    pminsw           m1, m12
    packuswb         m0, m1
    movifnidn      dstq, dstmp
    mova    [dstq+srcq], m0

%if ARCH_X86_32
    add dword [rsp+5*mmsize+12], mmsize
%else
    mova             m8, [pb_17_27]
%endif
    add            srcq, r2mp
    add      grain_lutq, 82
    dec              hw
    jz .end_y_v_overlap
    ; 2 lines get vertical overlap, then fall back to non-overlap code for
    ; remaining (up to) 30 lines
    btc              hd, 16
    jnc .loop_y_v_overlap
    jmp .loop_y

.end_y_v_overlap:
%if ARCH_X86_32
    add            r4mp, 16
%else
    add              wq, 16
%endif
    jge .end_hv
%if ARCH_X86_32
    mov            srcq, r1mp
    add            srcq, r4mp
%else
    lea            srcq, [src_bakq+wq]
%endif
    btc       dword r8m, 2
    jc .loop_x_hv_overlap
    add          offxyd, 16
%if ARCH_X86_32
    add dword [rsp+5*mmsize+1*gprsize], 16
%else
    add      top_offxyd, 16
%endif
    jmp .loop_x_odd_v_overlap

.loop_x_hv_overlap:
%if ARCH_X86_32
    mov              r5, r5m
    lea              r5, [base+pb_27_17]
    mov [rsp+5*mmsize+12], r5

    DEFINE_ARGS tmp, src, scaling, offxy, w, picptr, src_bak

    mov              r5, [rsp+5*mmsize+1*gprsize]
    mov              r4, offxyd
    add              r5, 16
    add              r4, 16
    mov [rsp+5*mmsize+2*gprsize], r5        ; topleft_offxy
    mov [rsp+5*mmsize+0*gprsize], r4        ; left_offxy

    DEFINE_ARGS tmp, src, scaling, see, w, picptr, src_bak

    xor            tmpd, tmpd
    mov            seed, r3m
%else
    mova             m8, [pb_27_17]

    DEFINE_ARGS dst, src, stride, src_bak, w, scaling, grain_lut, \
                tmp, unused2, see, unused3

    ; we assume from the block above that bits 8-15 of tmpd are zero'ed
%endif
    mov             r6d, seed
    or             seed, 0xeff4eff4
    test           seeb, seeh
    setp           tmpb                     ; parity of top_seed
    shr            seed, 16
    shl            tmpd, 16
    test           seeb, seeh
    setp           tmpb                     ; parity of cur_seed
    or              r6d, 0x00010001
    xor            tmpd, r6d
    mov            seed, tmpd
    ror            seed, 1                  ; updated (cur_seed << 16) | top_seed

%if ARCH_X86_32
    mov             r3m, seed

    DEFINE_ARGS dst, src, scaling, offy, unused1, unused2, offx

    mov           offxd, offyd
%else
    DEFINE_ARGS dst, src, stride, src_bak, w, scaling, grain_lut, \
                offx, offy, see, left_offxy, top_offxy, topleft_offxy

    lea  topleft_offxyq, [top_offxyq+16]
    lea     left_offxyq, [offyq+16]
    mov           offyd, seed
    mov           offxd, seed
%endif
    ror           offyd, 8
    ror           offxd, 12
    and           offyd, 0xf000f
    and           offxd, 0xf000f
    imul          offyd, 164
    ; offxy=offy*stride+offx, (cur_offxy << 16) | top_offxy
    lea           offyq, [offyq+offxq*2+0x10001*747+32*82]

%if ARCH_X86_32
    DEFINE_ARGS dst, src, scaling, offxy, h, picptr, grain_lut

    movzx            r5, offxyw             ; top_offxy
    mov [rsp+5*mmsize+1*gprsize], r5
%else
    DEFINE_ARGS dst, src, stride, src_bak, w, scaling, grain_lut, \
                h, offxy, see, left_offxy, top_offxy, topleft_offxy

    movzx    top_offxyd, offxyw
%endif
    shr          offxyd, 16

    mov              hd, r7m
    mov      grain_lutq, grain_lutmp
.loop_y_hv_overlap:
    ; grain = grain_lut[offy+y][offx+x]
    movu             m3, [grain_lutq+offxyq]
%if ARCH_X86_32
    mov              r5, [rsp+5*mmsize+1*gprsize]   ; top_offxy
    mov              r0, [rsp+5*mmsize+0*gprsize]   ; left_offxy
    movu             m6, [grain_lutq+r5]
    mov              r5, [rsp+5*mmsize+2*gprsize]   ; topleft_offxy
    movd             m4, [grain_lutq+r0]
    movd             m7, [grain_lutq+r5]
%else
    movu             m6, [grain_lutq+top_offxyq]
    movd             m4, [grain_lutq+left_offxyq]
    movd             m7, [grain_lutq+topleft_offxyq]
%endif
    ; do h interpolation first (so top | top/left -> top, left | cur -> cur)
    punpcklbw        m4, m3
    punpcklbw        m7, m6
    pmaddubsw        m2, m15, m4
    pmaddubsw        m4, m15, m7
    pmulhrsw         m2, m14
    pmulhrsw         m4, m14
    packsswb         m2, m2
    packsswb         m4, m4
    shufps           m2, m3, q3210
    shufps           m4, m6, q3210
    ; followed by v interpolation (top | cur -> cur)
    punpcklbw        m3, m4, m2
    punpckhbw        m4, m2
%if ARCH_X86_32
    mov              r5, [rsp+5*mmsize+12]
    pmaddubsw        m7, [r5], m4
    pmaddubsw        m4, [r5], m3
%else
    pmaddubsw        m7, m8, m4
    pmaddubsw        m4, m8, m3
%endif
    pmulhrsw         m7, m14
    pmulhrsw         m4, m14
    packsswb         m4, m7
    pxor             m2, m2
    pcmpgtb          m7, m2, m4
    punpcklbw        m3, m4, m7
    punpckhbw        m4, m7

    ; src
    mova             m0, [srcq]
    punpckhbw        m1, m0, m2
    punpcklbw        m0, m2                 ; m0-1: src as word

    ; scaling[src]
%if ARCH_X86_32
    vpgatherdw       m5, m0, scalingq-1, r0, r5, m7
    vpgatherdw       m6, m1, scalingq-1, r0, r5, m7
%else
    vpgatherdw       m5, m0, scalingq-1, r13, r14, m7
    vpgatherdw       m6, m1, scalingq-1, r13, r14, m7
%endif
    REPX {psrlw x, 8}, m5, m6

    ; noise = round2(scaling[src] * grain, scaling_shift)
    pmullw           m3, m5
    pmullw           m4, m6
    pmulhrsw         m3, m11
    pmulhrsw         m4, m11

    ; dst = clip_pixel(src, noise)
    paddw            m0, m3
    paddw            m1, m4
    pmaxsw           m0, m13
    pmaxsw           m1, m13
    pminsw           m0, m12
    pminsw           m1, m12
    packuswb         m0, m1
    movifnidn      dstq, dstmp
    mova    [dstq+srcq], m0

%if ARCH_X86_32
    add dword [rsp+5*mmsize+12], mmsize
%else
    mova             m8, [pb_17_27]
%endif
    add            srcq, r2mp
    add      grain_lutq, 82
    dec              hw
    jz .end_y_hv_overlap
    ; 2 lines get vertical overlap, then fall back to non-overlap code for
    ; remaining (up to) 30 lines
    btc              hd, 16
    jnc .loop_y_hv_overlap
    jmp .loop_y_h_overlap

.end_y_hv_overlap:
%if ARCH_X86_32
    add            r4mp, 16
%else
    add              wq, 16
%endif
    jge .end_hv
%if ARCH_X86_32
    mov            srcq, r1m
    add            srcq, r4m
%else
    lea            srcq, [src_bakq+wq]
%endif
    xor       dword r8m, 4
    add          offxyd, 16
%if ARCH_X86_32
    add dword [rsp+5*mmsize+1*gprsize], 16
%else
    add      top_offxyd, 16
%endif
    jmp .loop_x_odd_v_overlap

.end_hv:
    RET

%macro FGUV_FN 3 ; name, ss_hor, ss_ver
INIT_XMM ssse3
%if ARCH_X86_32
; fguv_32x32xn_i420_ssse3(dst, src, stride, fg_data, w, scaling, grain_lut, h,
;                         sby, luma, lstride, uv_pl, is_id)
%if STACK_ALIGNMENT < mmsize
DECLARE_ARG 0, 1, 2, 3, 4, 5, 6, 7, 8
cglobal fguv_32x32xn_i%1_8bpc, 0, 7, 8, 0 - (7 * mmsize + (13 + 3) * gprsize), \
        tmp, src, scaling, h, fg_data, picptr, unused
    mov              r0, r0m
    mov              r1, r2m
    mov              r2, r4m
    mov              r3, r6m
    mov              r4, r7m
    mov [rsp+7*mmsize+3*gprsize], r0
    mov [rsp+7*mmsize+5*gprsize], r1
    mov [rsp+7*mmsize+7*gprsize], r2
    mov [rsp+7*mmsize+9*gprsize], r3
    mov [rsp+7*mmsize+10*gprsize], r4

    mov              r0, r8m
    mov              r1, r9m
    mov              r2, r10m
    mov              r4, r11m
    mov              r3, r12m
    mov [rsp+7*mmsize+11*gprsize], r0
    mov [rsp+7*mmsize+12*gprsize], r1
    mov [rsp+7*mmsize+13*gprsize], r2
    mov [rsp+7*mmsize+14*gprsize], r4
%else
cglobal fguv_32x32xn_i%1_8bpc, 0, 7, 8, 7 * mmsize + (4) * gprsize, \
        tmp, src, scaling, h, fg_data, picptr, unused
%endif
    mov            srcq, srcm
    mov        fg_dataq, r3m
    mov        scalingq, r5m
%if STACK_ALIGNMENT < mmsize
%define r0m [rsp+7*mmsize+ 3*gprsize]
%define r1m [rsp+7*mmsize+ 4*gprsize]
%define r2m [rsp+7*mmsize+ 5*gprsize]
%define r3m [rsp+7*mmsize+ 6*gprsize]
%define r4m [rsp+7*mmsize+ 7*gprsize]
%define r5m [rsp+7*mmsize+ 8*gprsize]
%define r6m [rsp+7*mmsize+ 9*gprsize]
%define r7m [rsp+7*mmsize+10*gprsize]
%define r8m [rsp+7*mmsize+11*gprsize]
%define r9m [rsp+7*mmsize+12*gprsize]
%define r10m [rsp+7*mmsize+13*gprsize]
%define r11m [rsp+7*mmsize+14*gprsize]
%define r12m [rsp+7*mmsize+15*gprsize]
%endif
    LEA              r5, pb_mask
%define base r5-pb_mask
    mov             r5m, r5
%else
cglobal fguv_32x32xn_i%1_8bpc, 6, 15, 16, dst, src, stride, fg_data, w, scaling, \
                                     grain_lut, tmp, sby, luma, lstride, uv_pl, is_id
    lea              r8, [pb_mask]
%define base r8-pb_mask
%endif
    mov             r6d, [fg_dataq+FGData.scaling_shift]
    movd             m3, [base+mul_bits+r6*2-14]
    mov             r6d, [fg_dataq+FGData.clip_to_restricted_range]
    lea            tmpd, [r6d*2]
%if ARCH_X86_32 && STACK_ALIGNMENT < mmsize
    test             r3, r3
%else
    cmp      dword r12m, 0                      ; is_idm
%endif
    movd             m5, [base+min+r6*2]
    cmovne          r6d, tmpd
    movd             m4, [base+max+r6*2]
    punpcklwd        m3, m3
    punpcklwd        m5, m5
    punpcklwd        m4, m4
    pshufd           m3, m3, q0000
    pshufd           m5, m5, q0000
    pshufd           m4, m4, q0000
    SCRATCH           3, 11, 0
    SCRATCH           4, 12, 1
    SCRATCH           5, 13, 2

    cmp byte [fg_dataq+FGData.chroma_scaling_from_luma], 0
    jne .csfl

%macro %%FGUV_32x32xN_LOOP 3 ; not-csfl, ss_hor, ss_ver
%if ARCH_X86_32
    DEFINE_ARGS dst, src, scaling, sby, fg_data, picptr, overlap
%else
    DEFINE_ARGS dst, src, stride, fg_data, w, scaling, grain_lut, unused, sby, see, overlap
%endif

%if %1
    mov             r6d, dword r11m
    movd             m0, [fg_dataq+FGData.uv_mult+r6*4]
    movd             m1, [fg_dataq+FGData.uv_luma_mult+r6*4]
    punpcklbw        m6, m1, m0
    movd             m7, [fg_dataq+FGData.uv_offset+r6*4]
    punpcklwd        m6, m6
    punpcklwd        m7, m7
    pshufd           m6, m6, q0000
    pshufd           m7, m7, q0000
    SCRATCH           6, 14, 3
    SCRATCH           7, 15, 4
%endif

    mov            sbyd, r8m
    mov        overlapd, [fg_dataq+FGData.overlap_flag] ; left_overlap: overlap & 1
    test       overlapd, overlapd
    jz %%no_vertical_overlap
%if ARCH_X86_32
%if %2
    mova             m1, [base+pb_23_22_h]
%else
    mova             m1, [base+pb_27_17_17_27]
%endif
    mova             m0, [base+pw_1024]
%else
%if %2
    mova             m1, [pb_23_22_h]
%else
    mova             m1, [pb_27_17_17_27]
%endif
    mova             m0, [pw_1024]
%endif
    SCRATCH           0, 8, 5
    SCRATCH           1, 9, 6
    test           sbyd, sbyd
    jnz %%vertical_overlap
    ; fall-through

%%no_vertical_overlap:
    mov             r8m, overlapd
%if ARCH_X86_32
    DEFINE_ARGS dst, src, scaling, see, fg_data, picptr, overlap
    imul           seed, (173 << 24) | 37
%else
    imul           seed, sbyd, (173 << 24) | 37
%endif
    add            seed, (105 << 24) | 178
    rol            seed, 8
    movzx          seed, seew
    xor            seed, [fg_dataq+FGData.seed]

%if ARCH_X86_32
    mov             r3m, seed

    DEFINE_ARGS luma, src, scaling, see, w, picptr, src_bak
%define luma_bakq lumaq

    mov              wq, r4m
%if %3
    shl           r10mp, 1
%endif
%else
    DEFINE_ARGS dst, src, stride, luma, w, scaling, grain_lut, \
                unused2, unused3, see, overlap, unused4, src_bak, lstride, luma_bak

    mov        lstrideq, r10mp
%endif

    mov           lumaq, r9mp
    lea        src_bakq, [srcq+wq]
    lea       luma_bakq, [lumaq+wq*(1+%2)]
    neg              wq
    sub            r0mp, srcq
%if ARCH_X86_32
    mov             r1m, src_bakq
    mov            r11m, luma_bakq
    mov             r4m, wq

    DEFINE_ARGS tmp, src, scaling, see, unused1, picptr, unused2
%else
    mov           r11mp, src_bakq
    mov           r12mp, strideq
%endif

%%loop_x:
%if ARCH_X86_32
    mov            seed, r3m
%endif
    mov             r6d, seed
    or             seed, 0xEFF4
    shr             r6d, 1
    test           seeb, seeh
    lea            seed, [r6+0x8000]
    cmovp          seed, r6d               ; updated seed
%if ARCH_X86_32
    mov             r3m, seed

    DEFINE_ARGS dst, src, scaling, offy, w, picptr, offx

    mov           offxd, offyd
%else
    DEFINE_ARGS dst, src, stride, luma, w, scaling, grain_lut, \
                offx, offy, see, overlap, unused1, unused2, lstride

    mov           offyd, seed
    mov           offxd, seed
%endif
    ror           offyd, 8
    shr           offxd, 12
    and           offyd, 0xf
    imul          offyd, 164>>%3
    lea           offyq, [offyq+offxq*(2-%2)+(3+(6>>%3))*82+(3+(6>>%2))]  ; offy*stride+offx

%if ARCH_X86_32
    DEFINE_ARGS luma, src, scaling, offxy, h, picptr, grain_lut
%else
    DEFINE_ARGS dst, src, stride, luma, w, scaling, grain_lut, \
                h, offxy, see, overlap, unused1, unused2, lstride, luma_bak
%endif

%%loop_x_odd:
    mov              hd, r7m
    mov      grain_lutq, grain_lutmp
%%loop_y:
    ; src
%if ARCH_X86_32
    mov           lumaq, r9mp
%endif
%if %2
    mova             m4, [lumaq+ 0]
    mova             m6, [lumaq+16]
    mova             m0, [srcq]
%if ARCH_X86_32
    add           lumaq, r10mp
    mov            r9mp, lumaq
    mov              r5, r5m
    movd             m7, [base+pb_1]
%else
    movd             m7, [pb_1]
%endif
    pshufd           m7, m7, q0000
    pxor             m2, m2
    pmaddubsw        m4, m7
    pmaddubsw        m6, m7
    pavgw            m4, m2
    pavgw            m6, m2
%else
    mova             m4, [lumaq]
    mova             m0, [srcq]
%if ARCH_X86_32
    add           lumaq, r10mp
    mov            r9mp, lumaq
%endif
    pxor             m2, m2
%endif

%if %1
%if %2
    packuswb         m4, m6                 ; luma
%endif
    punpckhbw        m6, m4, m0
    punpcklbw        m4, m0                 ; { luma, chroma }
    pmaddubsw        m6, m14
    pmaddubsw        m4, m14
    psraw            m6, 6
    psraw            m4, 6
    paddw            m6, m15
    paddw            m4, m15
    packuswb         m4, m6                 ; pack+unpack = clip
    punpckhbw        m6, m4, m2
    punpcklbw        m4, m2
%elif %2 == 0
    punpckhbw        m6, m4, m2
    punpcklbw        m4, m2
%endif

    ; scaling[luma_src]
%if ARCH_X86_32
    vpgatherdw       m7, m4, scalingq-1, r0, r5
    vpgatherdw       m5, m6, scalingq-1, r0, r5
%else
    vpgatherdw       m7, m4, scalingq-1, r12, r2
    vpgatherdw       m5, m6, scalingq-1, r12, r2
%endif
    REPX {psrlw x, 8}, m7, m5

    ; unpack chroma_source
    punpckhbw        m1, m0, m2
    punpcklbw        m0, m2                 ; m0-1: src as word

    ; grain = grain_lut[offy+y][offx+x]
    movu             m3, [grain_lutq+offxyq+ 0]
    pcmpgtb          m6, m2, m3
    punpcklbw        m2, m3, m6
    punpckhbw        m3, m6

    ; noise = round2(scaling[luma_src] * grain, scaling_shift)
    pmullw           m2, m7
    pmullw           m3, m5
    pmulhrsw         m2, m11
    pmulhrsw         m3, m11

%if ARCH_X86_32
    DEFINE_ARGS dst, src, scaling, offxy, h, picptr, grain_lut
%endif

    ; dst = clip_pixel(src, noise)
    paddw            m0, m2
    paddw            m1, m3
    pmaxsw           m0, m13
    pmaxsw           m1, m13
    pminsw           m0, m12
    pminsw           m1, m12
    packuswb         m0, m1
    movifnidn      dstq, dstmp
    mova    [dstq+srcq], m0

%if ARCH_X86_32
    add            srcq, r2mp
    ; we already incremented lumaq above
%else
    add            srcq, r12mp
%if %3
    lea           lumaq, [lumaq+lstrideq*2]
%else
    add           lumaq, lstrideq
%endif
%endif
    add      grain_lutq, 82
    dec              hw
    jg %%loop_y

%if ARCH_X86_32
    DEFINE_ARGS luma, src, scaling, offxy, w, picptr, grain_lut

    mov              wq, r4m
%endif
    add              wq, 16
    jge %%end
%if ARCH_X86_32
    mov            srcq, r1mp
    mov           lumaq, r11mp
%else
    mov            srcq, r11mp
%endif
    lea           lumaq, [luma_bakq+wq*(1+%2)]
    add            srcq, wq
%if ARCH_X86_32
    mov             r4m, wq
    mov             r9m, lumaq
%endif
%if %2 == 0
    ; adjust top_offxy
%if ARCH_X86_32
    add dword [rsp+7*mmsize+1*gprsize], 16
%else
    add            r11d, 16
%endif
    add          offxyd, 16
    btc       dword r8m, 2
    jc %%loop_x_even
    test      dword r8m, 2
    jz %%loop_x_odd
    jmp %%loop_x_odd_v_overlap
%%loop_x_even:
%endif
    test      dword r8m, 1
    jz %%loop_x

    ; r8m = sbym
    test      dword r8m, 2
    jne %%loop_x_hv_overlap

    ; horizontal overlap (without vertical overlap)
%%loop_x_h_overlap:
%if ARCH_X86_32
%if %2
    lea              r6, [offxyd+16]
    mov [rsp+7*mmsize+0*gprsize], r6
%else
    mov [rsp+7*mmsize+0*gprsize], offxyd
%endif

    DEFINE_ARGS luma, src, scaling, see, w, picptr, grain_lut

    mov            seed, r3m
%else
    DEFINE_ARGS dst, src, stride, luma, w, scaling, grain_lut, \
                offx, offy, see, left_offxy, unused1, unused2, lstride

%if %2
    lea     left_offxyd, [offyd+16]         ; previous column's offy*stride+offx
%else
    mov     left_offxyd, offyd
%endif
%endif
    mov             r6d, seed
    or             seed, 0xEFF4
    shr             r6d, 1
    test           seeb, seeh
    lea            seed, [r6+0x8000]
    cmovp          seed, r6d                ; updated seed

%if ARCH_X86_32
    mov             r3m, seed

    DEFINE_ARGS luma, src, scaling, offy, w, picptr, offx

    mov          offxd, offyd
%else
    DEFINE_ARGS dst, src, stride, luma, w, scaling, grain_lut, \
                offx, offy, see, left_offxy, unused1, unused2, lstride

    mov           offyd, seed
    mov           offxd, seed
%endif
    ror           offyd, 8
    shr           offxd, 12
    and           offyd, 0xf
    imul          offyd, 164>>%3
    lea           offyq, [offyq+offxq*(2-%2)+(3+(6>>%3))*82+3+(6>>%2)]  ; offy*stride+offx

%if ARCH_X86_32
    DEFINE_ARGS luma, src, scaling, offxy, h, picptr, grain_lut
%else
    DEFINE_ARGS dst, src, stride, luma, w, scaling, grain_lut, \
                h, offxy, see, left_offxy, unused1, unused2, lstride, luma_bak
%endif

    mov              hd, r7m
    mov      grain_lutq, grain_lutmp
%%loop_y_h_overlap:
    ; src
%if ARCH_X86_32
    mov           lumaq, r9mp
%endif
%if %2
    mova             m4, [lumaq+ 0]
    mova             m6, [lumaq+16]
    mova             m0, [srcq]
%if ARCH_X86_32
    add           lumaq, r10mp
    mov            r9mp, lumaq
    mov              r5, r5m
    movd             m7, [base+pb_1]
%else
    movd             m7, [pb_1]
%endif
    pshufd           m7, m7, q0000
    pxor             m2, m2
    pmaddubsw        m4, m7
    pmaddubsw        m6, m7
    pavgw            m4, m2
    pavgw            m6, m2
%else
    mova             m4, [lumaq]
    mova             m0, [srcq]
%if ARCH_X86_32
    add           lumaq, r10mp
    mov            r9mp, lumaq
%endif
    pxor             m2, m2
%endif

%if %1
%if %2
    packuswb         m4, m6                 ; luma
%endif
    punpckhbw        m6, m4, m0
    punpcklbw        m4, m0                 ; { luma, chroma }
    pmaddubsw        m6, m14
    pmaddubsw        m4, m14
    psraw            m6, 6
    psraw            m4, 6
    paddw            m6, m15
    paddw            m4, m15
    packuswb         m4, m6                 ; pack+unpack = clip
    punpckhbw        m6, m4, m2
    punpcklbw        m4, m2
%elif %2 == 0
    punpckhbw        m6, m4, m2
    punpcklbw        m4, m2
%endif

    ; scaling[luma_src]
%if ARCH_X86_32
    vpgatherdw       m7, m4, scalingq-1, r0, r5
    vpgatherdw       m5, m6, scalingq-1, r0, r5
%else
    vpgatherdw       m7, m4, scalingq-1, r12, r2
    vpgatherdw       m5, m6, scalingq-1, r12, r2
%endif
    REPX {psrlw x, 8}, m7, m5

    ; unpack chroma_source
    punpckhbw        m1, m0, m2
    punpcklbw        m0, m2                 ; m0-1: src as word

    ; grain = grain_lut[offy+y][offx+x]
    movu             m4, [grain_lutq+offxyq+ 0]
%if ARCH_X86_32
    mov              r0, [rsp+7*mmsize+0*gprsize]
    movd             m2, [grain_lutq+r0+ 0]
%else
    movd             m2, [grain_lutq+left_offxyq+ 0]
%endif
    punpcklbw        m2, m4
    pmaddubsw        m3, m9, m2
    pmulhrsw         m3, m8
    packsswb         m3, m3
    shufps           m3, m4, q3210
    pxor             m4, m4
    pcmpgtb          m4, m3
    punpcklbw        m2, m3, m4
    punpckhbw        m3, m4

    ; noise = round2(scaling[luma_src] * grain, scaling_shift)
    pmullw           m2, m7
    pmullw           m3, m5
    pmulhrsw         m2, m11
    pmulhrsw         m3, m11

%if ARCH_X86_32
    DEFINE_ARGS dst, src, scaling, offxy, h, picptr, grain_lut
%endif

    ; dst = clip_pixel(src, noise)
    paddw            m0, m2
    paddw            m1, m3
    pmaxsw           m0, m13
    pmaxsw           m1, m13
    pminsw           m0, m12
    pminsw           m1, m12
    packuswb         m0, m1
    movifnidn      dstq, dstmp
    mova    [dstq+srcq], m0

%if ARCH_X86_32
    add            srcq, r2mp
    ; lumaq has already been incremented above
%else
    add            srcq, r12mp
%if %3
    lea           lumaq, [lumaq+lstrideq*2]
%else
    add           lumaq, lstrideq
%endif
%endif
    add      grain_lutq, 82
    dec              hw
    jg %%loop_y_h_overlap

%if ARCH_X86_32
    DEFINE_ARGS luma, src, scaling, offxy, w, picptr, grain_lut

    mov              wq, r4m
%endif
    add              wq, 16
    jge %%end
%if ARCH_X86_32
    mov            srcq, r1mp
    mov           lumaq, r11mp
%else
    mov            srcq, r11mp
%endif
    lea           lumaq, [luma_bakq+wq*(1+%2)]
    add            srcq, wq
%if ARCH_X86_32
    mov             r4m, wq
    mov             r9m, lumaq
%endif
%if %2 == 0
    xor       dword r8m, 4
    ; adjust top_offxyd
%if ARCH_X86_32
    add dword [rsp+7*mmsize+1*gprsize], 16
%else
    add            r11d, 16
%endif
    add          offxyd, 16
%endif

    ; r8m = sbym
    test      dword r8m, 2
%if %2
    jne %%loop_x_hv_overlap
    jmp %%loop_x_h_overlap
%else
    jne %%loop_x_odd_v_overlap
    jmp %%loop_x_odd
%endif

%%end:
    RET

%%vertical_overlap:
%if ARCH_X86_32
    DEFINE_ARGS dst, src, scaling, sby, fg_data, picptr, overlap
%else
    DEFINE_ARGS dst, src, stride, fg_data, w, scaling, grain_lut, tmp, sby, see, overlap
%endif

    or         overlapd, 2                  ; top_overlap: overlap & 2
    mov             r8m, overlapd
    movzx          sbyd, sbyb
%if ARCH_X86_32
    imul             r4, [fg_dataq+FGData.seed], 0x00010001
    DEFINE_ARGS tmp, src, scaling, sby, see, picptr, unused
%else
    imul           seed, [fg_dataq+FGData.seed], 0x00010001
%endif
    imul           tmpd, sbyd, 173 * 0x00010001
    imul           sbyd, 37 * 0x01000100
    add            tmpd, (105 << 16) | 188
    add            sbyd, (178 << 24) | (141 << 8)
    and            tmpd, 0x00ff00ff
    and            sbyd, 0xff00ff00
    xor            seed, tmpd
%if ARCH_X86_32
    xor            sbyd, seed               ; (cur_seed << 16) | top_seed

    DEFINE_ARGS luma, src, scaling, see, w, picptr, src_bak

    mov             r3m, seed
    mov              wq, r4m
%if %3
    shl           r10mp, 1
%endif
%else
    xor            seed, sbyd               ; (cur_seed << 16) | top_seed

    DEFINE_ARGS dst, src, stride, luma, w, scaling, grain_lut, \
                tmp, unused2, see, overlap, unused3, src_bak, lstride, luma_bak

    mov        lstrideq, r10mp
%endif

    mov           lumaq, r9mp
    lea        src_bakq, [srcq+wq]
    lea       luma_bakq, [lumaq+wq*(1+%2)]
    neg              wq
    sub            r0mp, srcq
%if ARCH_X86_32
    mov             r1m, src_bakq
    mov            r11m, luma_bakq
    mov             r4m, wq

    DEFINE_ARGS tmp, src, scaling, see, unused1, picptr, unused2
%else
    mov           r11mp, src_bakq
    mov           r12mp, strideq
%endif

%%loop_x_v_overlap:
%if ARCH_X86_32
    mov            seed, r3m
    xor            tmpd, tmpd
%endif
    ; we assume from the block above that bits 8-15 of tmpd are zero'ed
    mov             r6d, seed
    or             seed, 0xeff4eff4
    test           seeb, seeh
    setp           tmpb                     ; parity of top_seed
    shr            seed, 16
    shl            tmpd, 16
    test           seeb, seeh
    setp           tmpb                     ; parity of cur_seed
    or              r6d, 0x00010001
    xor            tmpd, r6d
    mov            seed, tmpd
    ror            seed, 1                  ; updated (cur_seed << 16) | top_seed

%if ARCH_X86_32
    mov             r3m, seed

    DEFINE_ARGS dst, src, scaling, offy, h, picptr, offx

    mov           offxd, offyd
%else
    DEFINE_ARGS dst, src, stride, luma, w, scaling, grain_lut, \
                offx, offy, see, overlap, top_offxy, unused, lstride

    mov           offxd, seed
    mov           offyd, seed
%endif
    ror           offyd, 8
    ror           offxd, 12
    and           offyd, 0xf000f
    and           offxd, 0xf000f
    imul          offyd, 164>>%3
    ; offxy=offy*stride+offx, (cur_offxy << 16) | top_offxy
    lea           offyq, [offyq+offxq*(2-%2)+0x10001*((3+(6>>%3))*82+3+(6>>%2))+(32>>%3)*82]

%if ARCH_X86_32
    DEFINE_ARGS tmp, src, scaling, offxy, h, picptr, top_offxy
%else
    DEFINE_ARGS dst, src, stride, luma, w, scaling, grain_lut, \
                h, offxy, see, overlap, top_offxy, unused, lstride, luma_bak
%endif

    movzx    top_offxyd, offxyw
    shr          offxyd, 16
%if ARCH_X86_32
    mov [rsp+7*mmsize+1*gprsize], top_offxyd

    DEFINE_ARGS luma, src, scaling, offxy, h, picptr, grain_lut
%endif

%%loop_x_odd_v_overlap:
    mov              hd, r7m
    mov      grain_lutq, grain_lutmp
%if ARCH_X86_32
    mov              r5, r5m
%endif
%if %3
    mova             m1, [PIC_ptr(pb_23_22)]
%else
    mova             m1, [PIC_ptr(pb_27_17)]
%endif
%%loop_y_v_overlap:
%if ARCH_X86_32
    mov           lumaq, r9mp
%endif
%if %2
    mova             m4, [lumaq+ 0]
    mova             m6, [lumaq+16]
    mova             m0, [srcq]
%if ARCH_X86_32
    add           lumaq, r10mp
    mov            r9mp, lumaq
    mov              r5, r5m
    movd             m7, [base+pb_1]
%else
    movd             m7, [pb_1]
%endif
    pshufd           m7, m7, q0000
    pxor             m2, m2
    pmaddubsw        m4, m7
    pmaddubsw        m6, m7
    pavgw            m4, m2
    pavgw            m6, m2
%else
    mova             m4, [lumaq]
    mova             m0, [srcq]
%if ARCH_X86_32
    add           lumaq, r10mp
    mov            r9mp, lumaq
%endif
    pxor             m2, m2
%endif

%if %1
%if %2
    packuswb         m4, m6                 ; luma
%endif
    punpckhbw        m6, m4, m0
    punpcklbw        m4, m0                 ; { luma, chroma }
    pmaddubsw        m6, m14
    pmaddubsw        m4, m14
    psraw            m6, 6
    psraw            m4, 6
    paddw            m6, m15
    paddw            m4, m15
    packuswb         m4, m6                 ; pack+unpack = clip
    punpckhbw        m6, m4, m2
    punpcklbw        m4, m2
%elif %2 == 0
    punpckhbw        m6, m4, m2
    punpcklbw        m4, m2
%endif

    ; scaling[luma_src]
%if ARCH_X86_32
    vpgatherdw       m7, m4, scalingq-1, r0, r5
    vpgatherdw       m5, m6, scalingq-1, r0, r5
%else
    vpgatherdw       m7, m4, scalingq-1, r12, r2
    vpgatherdw       m5, m6, scalingq-1, r12, r2
%endif
    REPX {psrlw x, 8}, m7, m5

    ; grain = grain_lut[offy+y][offx+x]
    movu             m3, [grain_lutq+offxyq]
%if ARCH_X86_32
    mov              r0, [rsp+7*mmsize+1*gprsize]
    movu             m4, [grain_lutq+r0]
%else
    movu             m4, [grain_lutq+top_offxyq]
%endif
    punpckhbw        m6, m4, m3
    punpcklbw        m4, m3
    pmaddubsw        m2, m1, m6
    pmaddubsw        m3, m1, m4
    pmulhrsw         m2, m8
    pmulhrsw         m3, m8
    packsswb         m3, m2
    pxor             m6, m6
    pcmpgtb          m6, m3
    punpcklbw        m2, m3, m6
    punpckhbw        m3, m6

    ; noise = round2(scaling[luma_src] * grain, scaling_shift)
    pmullw           m2, m7
    pmullw           m3, m5
    pmulhrsw         m2, m11
    pmulhrsw         m3, m11

    ; unpack chroma_source
    pxor             m4, m4
    punpckhbw        m6, m0, m4
    punpcklbw        m0, m4                 ; m0-1: src as word

%if ARCH_X86_32
    DEFINE_ARGS dst, src, scaling, offxy, h, picptr, grain_lut
%endif

    ; dst = clip_pixel(src, noise)
    paddw            m0, m2
    paddw            m6, m3
    pmaxsw           m0, m13
    pmaxsw           m6, m13
    pminsw           m0, m12
    pminsw           m6, m12
    packuswb         m0, m6
    movifnidn      dstq, dstmp
    mova    [dstq+srcq], m0

    dec              hw
    je %%end_y_v_overlap
%if ARCH_X86_32
    add            srcq, r2mp
    ; lumaq has already been incremented above
%else
    add            srcq, r12mp
%if %3
    lea           lumaq, [lumaq+lstrideq*2]
%else
    add           lumaq, lstrideq
%endif
%endif
    add      grain_lutq, 82
%if %3 == 0
    btc              hd, 16
%if ARCH_X86_32
    mov              r5, r5m
%endif
    mova             m1, [PIC_ptr(pb_17_27)]
    jnc %%loop_y_v_overlap
%endif
    jmp %%loop_y

%%end_y_v_overlap:
%if ARCH_X86_32
    DEFINE_ARGS luma, src, scaling, offxy, w, picptr, grain_lut

    mov              wq, r4m
%endif
    add              wq, 16
    jge %%end_hv
%if ARCH_X86_32
    mov            srcq, r1mp
    mov           lumaq, r11mp
%else
    mov            srcq, r11mp
%endif
    lea           lumaq, [luma_bakq+wq*(1+%2)]
    add            srcq, wq
%if ARCH_X86_32
    mov             r4m, wq
    mov             r9m, lumaq
%endif

%if %2
    ; since fg_dataq.overlap is guaranteed to be set, we never jump
    ; back to .loop_x_v_overlap, and instead always fall-through to
    ; h+v overlap
%else
%if ARCH_X86_32
    add dword [rsp+7*mmsize+1*gprsize], 16
%else
    add      top_offxyd, 16
%endif
    add          offxyd, 16
    btc       dword r8m, 2
    jnc %%loop_x_odd_v_overlap
%endif

%%loop_x_hv_overlap:
%if ARCH_X86_32
    DEFINE_ARGS tmp, src, scaling, offxy, w, picptr, unused

    mov              r6, [rsp+7*mmsize+1*gprsize]
%if %2
    lea              r0, [r3d+16]
    add              r6, 16
    mov [rsp+7*mmsize+0*gprsize], r0        ; left_offxy
%else
    mov [rsp+7*mmsize+0*gprsize], r3        ; left_offxy
%endif
    mov [rsp+7*mmsize+2*gprsize], r6        ; topleft_offxy

    DEFINE_ARGS tmp, src, scaling, see, w, picptr, unused

    mov            seed, r3m
    xor            tmpd, tmpd
%else
    DEFINE_ARGS dst, src, stride, luma, w, scaling, grain_lut, \
                tmp, offxy, see, left_offxy, top_offxy, topleft_offxy, lstride

%if %2
    lea  topleft_offxyq, [top_offxyq+16]
    lea     left_offxyq, [offxyq+16]
%else
    mov  topleft_offxyq, top_offxyq
    mov     left_offxyq, offxyq
%endif

    ; we assume from the block above that bits 8-15 of tmpd are zero'ed
%endif
    mov             r6d, seed
    or             seed, 0xeff4eff4
    test           seeb, seeh
    setp           tmpb                     ; parity of top_seed
    shr            seed, 16
    shl            tmpd, 16
    test           seeb, seeh
    setp           tmpb                     ; parity of cur_seed
    or              r6d, 0x00010001
    xor            tmpd, r6d
    mov            seed, tmpd
    ror            seed, 1                  ; updated (cur_seed << 16) | top_seed

%if ARCH_X86_32
    mov             r3m, seed

    DEFINE_ARGS tmp, src, scaling, offy, w, picptr, offx

    mov           offxd, offyd
%else
    DEFINE_ARGS dst, src, stride, luma, w, scaling, grain_lut, \
                offx, offy, see, left_offxy, top_offxy, topleft_offxy, lstride

    mov           offxd, seed
    mov           offyd, seed
%endif
    ror           offyd, 8
    ror           offxd, 12
    and           offyd, 0xf000f
    and           offxd, 0xf000f
    imul          offyd, 164>>%3
    ; offxy=offy*stride+offx, (cur_offxy << 16) | top_offxy
    lea           offyq, [offyq+offxq*(2-%2)+0x10001*((3+(6>>%3))*82+3+(6>>%2))+(32>>%3)*82]

%if ARCH_X86_32
    DEFINE_ARGS top_offxy, src, scaling, offxy, h, picptr, grain_lut
%else
    DEFINE_ARGS dst, src, stride, luma, w, scaling, grain_lut, \
                h, offxy, see, left_offxy, top_offxy, topleft_offxy, lstride, luma_bak
%endif

    movzx    top_offxyd, offxyw
    shr          offxyd, 16
%if ARCH_X86_32
    mov [rsp+7*mmsize+1*gprsize], top_offxyd
%endif

    mov              hd, r7m
    mov      grain_lutq, grain_lutmp
%if ARCH_X86_32
    mov              r5, r5m
%endif
%if %3
    mova             m3, [PIC_ptr(pb_23_22)]
%else
    mova             m3, [PIC_ptr(pb_27_17)]
%endif
%%loop_y_hv_overlap:
    ; grain = grain_lut[offy+y][offx+x]
%if ARCH_X86_32
    mov              r0, [rsp+7*mmsize+2*gprsize]       ; topleft_offxy
    mov              r5, [rsp+7*mmsize+1*gprsize]       ; top_offxy
    movd             m1, [grain_lutq+r0]
    mov              r0, [rsp+7*mmsize+0*gprsize]       ; left_offxy
%else
    movd             m1, [grain_lutq+topleft_offxyq]
%endif
    movu             m2, [grain_lutq+offxyq]
%if ARCH_X86_32
    movu             m6, [grain_lutq+r5]
    movd             m4, [grain_lutq+r0]
%else
    movu             m6, [grain_lutq+top_offxyq]
    movd             m4, [grain_lutq+left_offxyq]
%endif
    ; do h interpolation first (so top | top/left -> top, left | cur -> cur)
    punpcklbw        m1, m6
    punpcklbw        m4, m2
    pmaddubsw        m0, m9, m1
    pmaddubsw        m1, m9, m4
    REPX {pmulhrsw x, m8}, m0, m1
    packsswb         m0, m1
    shufps           m4, m0, m2, q3232
    shufps           m0, m6, q3210
    ; followed by v interpolation (top | cur -> cur)
    punpcklbw        m2, m0, m4
    punpckhbw        m0, m4
    pmaddubsw        m4, m3, m0
    pmaddubsw        m1, m3, m2
    pmulhrsw         m4, m8
    pmulhrsw         m1, m8
    packsswb         m1, m4

    ; src
%if ARCH_X86_32
    DEFINE_ARGS luma, src, scaling, offxy, w, picptr, grain_lut

    mov           lumaq, r9mp
%endif
%if %2
    mova             m4, [lumaq+ 0]
    mova             m6, [lumaq+16]
    mova             m0, [srcq]
%if ARCH_X86_32
    add           lumaq, r10mp
    mov            r9mp, lumaq
    mov              r5, r5m
    movd             m7, [base+pb_1]
%else
    movd             m7, [pb_1]
%endif
    pshufd           m7, m7, q0000
    pxor             m2, m2
    pmaddubsw        m4, m7
    pmaddubsw        m6, m7
    pavgw            m4, m2
    pavgw            m6, m2
%else
    mova             m4, [lumaq]
    mova             m0, [srcq]
%if ARCH_X86_32
    add           lumaq, r10mp
    mov            r9mp, lumaq
%endif
    pxor             m2, m2
%endif

%if %1
%if %2
    packuswb         m4, m6                 ; luma
%endif
    punpckhbw        m6, m4, m0
    punpcklbw        m4, m0                 ; { luma, chroma }
    pmaddubsw        m6, m14
    pmaddubsw        m4, m14
    psraw            m6, 6
    psraw            m4, 6
    paddw            m6, m15
    paddw            m4, m15
    packuswb         m4, m6                 ; pack+unpack = clip
    punpckhbw        m6, m4, m2
    punpcklbw        m4, m2
%elif %2 == 0
    punpckhbw        m6, m4, m2
    punpcklbw        m4, m2
%endif

    ; scaling[src]
%if ARCH_X86_32
    vpgatherdw       m7, m4, scalingq-1, r0, r5
    vpgatherdw       m5, m6, scalingq-1, r0, r5
%else
%if %3
    vpgatherdw       m7, m4, scalingq-1, r2, r12
    vpgatherdw       m5, m6, scalingq-1, r2, r12
%else
    vpgatherdw       m7, m4, scalingq-1, r2, r13
    vpgatherdw       m5, m6, scalingq-1, r2, r13
%endif
%endif
    REPX {psrlw x, 8}, m7, m5

    ; unpack grain
    pxor             m4, m4
    pcmpgtb          m4, m1
    punpcklbw        m2, m1, m4
    punpckhbw        m1, m4

    ; noise = round2(scaling[src] * grain, scaling_shift)
    pmullw           m2, m7
    pmullw           m1, m5
    pmulhrsw         m2, m11
    pmulhrsw         m1, m11

%if ARCH_X86_32
    DEFINE_ARGS dst, src, scaling, offxy, h, picptr, grain_lut
%endif

    ; unpack chroma source
    pxor             m4, m4
    punpckhbw        m5, m0, m4
    punpcklbw        m0, m4                 ; m0-1: src as word

    ; dst = clip_pixel(src, noise)
    paddw            m0, m2
    paddw            m5, m1
    pmaxsw           m0, m13
    pmaxsw           m5, m13
    pminsw           m0, m12
    pminsw           m5, m12
    packuswb         m0, m5
    movifnidn      dstq, dstmp
    mova    [dstq+srcq], m0

%if ARCH_X86_32
    add            srcq, r2mp
    ; lumaq has been adjusted above already
%else
    add            srcq, r12mp
%if %3
    lea           lumaq, [lumaq+lstrideq*(1+%2)]
%else
    add           lumaq, r10mp
%endif
%endif
    add      grain_lutq, 82
    dec              hw
%if %3
    jg %%loop_y_h_overlap
%else
    jle %%end_y_hv_overlap
%if ARCH_X86_32
    mov              r5, r5m
%endif
    mova             m3, [PIC_ptr(pb_17_27)]
    btc              hd, 16
    jnc %%loop_y_hv_overlap
%if ARCH_X86_64
    mov        lstrideq, r10mp
%endif
    jmp %%loop_y_h_overlap
%%end_y_hv_overlap:
%if ARCH_X86_64
    mov        lstrideq, r10mp
%endif
%endif

%if ARCH_X86_32
    DEFINE_ARGS luma, src, scaling, offxy, w, picptr, grain_lut

    mov              wq, r4m
%endif
    add              wq, 16
    jge %%end_hv
%if ARCH_X86_32
    mov            srcq, r1mp
    mov           lumaq, r11mp
%else
    mov            srcq, r11mp
%endif
    lea           lumaq, [luma_bakq+wq*(1+%2)]
    add            srcq, wq
%if ARCH_X86_32
    mov             r4m, wq
    mov             r9m, lumaq
%endif
%if %2
    jmp %%loop_x_hv_overlap
%else
%if ARCH_X86_32
    add dword [rsp+7*mmsize+1*gprsize], 16
%else
    add      top_offxyd, 16
%endif
    add          offxyd, 16
    xor       dword r8m, 4
    jmp %%loop_x_odd_v_overlap
%endif

%%end_hv:
    RET
%endmacro

    %%FGUV_32x32xN_LOOP 1, %2, %3
.csfl:
    %%FGUV_32x32xN_LOOP 0, %2, %3
%endmacro

FGUV_FN 420, 1, 1

%if STACK_ALIGNMENT < mmsize
DECLARE_ARG 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12
%endif

FGUV_FN 422, 1, 0

%if STACK_ALIGNMENT < mmsize
DECLARE_ARG 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12
%endif

FGUV_FN 444, 0, 0