summaryrefslogtreecommitdiff
path: root/libavcodec/x86/vp8dsp.asm
blob: 75c0b56f9460e2b2580c659296d0a396d273b75c (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
;******************************************************************************
;* VP8 MMXEXT optimizations
;* Copyright (c) 2010 Ronald S. Bultje <rsbultje@gmail.com>
;* Copyright (c) 2010 Jason Garrett-Glaser <darkshikari@gmail.com>
;*
;* This file is part of FFmpeg.
;*
;* FFmpeg is free software; you can redistribute it and/or
;* modify it under the terms of the GNU Lesser General Public
;* License as published by the Free Software Foundation; either
;* version 2.1 of the License, or (at your option) any later version.
;*
;* FFmpeg is distributed in the hope that it will be useful,
;* but WITHOUT ANY WARRANTY; without even the implied warranty of
;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
;* Lesser General Public License for more details.
;*
;* You should have received a copy of the GNU Lesser General Public
;* License along with FFmpeg; if not, write to the Free Software
;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
;******************************************************************************

%include "libavutil/x86/x86inc.asm"
%include "libavutil/x86/x86util.asm"

SECTION_RODATA

fourtap_filter_hw_m: times 4 dw  -6, 123
                     times 4 dw  12,  -1
                     times 4 dw  -9,  93
                     times 4 dw  50,  -6
                     times 4 dw  -6,  50
                     times 4 dw  93,  -9
                     times 4 dw  -1,  12
                     times 4 dw 123,  -6

sixtap_filter_hw_m:  times 4 dw   2, -11
                     times 4 dw 108,  36
                     times 4 dw  -8,   1
                     times 4 dw   3, -16
                     times 4 dw  77,  77
                     times 4 dw -16,   3
                     times 4 dw   1,  -8
                     times 4 dw  36, 108
                     times 4 dw -11,   2

fourtap_filter_hb_m: times 8 db  -6, 123
                     times 8 db  12,  -1
                     times 8 db  -9,  93
                     times 8 db  50,  -6
                     times 8 db  -6,  50
                     times 8 db  93,  -9
                     times 8 db  -1,  12
                     times 8 db 123,  -6

sixtap_filter_hb_m:  times 8 db   2,   1
                     times 8 db -11, 108
                     times 8 db  36,  -8
                     times 8 db   3,   3
                     times 8 db -16,  77
                     times 8 db  77, -16
                     times 8 db   1,   2
                     times 8 db  -8,  36
                     times 8 db 108, -11

fourtap_filter_v_m:  times 8 dw  -6
                     times 8 dw 123
                     times 8 dw  12
                     times 8 dw  -1
                     times 8 dw  -9
                     times 8 dw  93
                     times 8 dw  50
                     times 8 dw  -6
                     times 8 dw  -6
                     times 8 dw  50
                     times 8 dw  93
                     times 8 dw  -9
                     times 8 dw  -1
                     times 8 dw  12
                     times 8 dw 123
                     times 8 dw  -6

sixtap_filter_v_m:   times 8 dw   2
                     times 8 dw -11
                     times 8 dw 108
                     times 8 dw  36
                     times 8 dw  -8
                     times 8 dw   1
                     times 8 dw   3
                     times 8 dw -16
                     times 8 dw  77
                     times 8 dw  77
                     times 8 dw -16
                     times 8 dw   3
                     times 8 dw   1
                     times 8 dw  -8
                     times 8 dw  36
                     times 8 dw 108
                     times 8 dw -11
                     times 8 dw   2

bilinear_filter_vw_m: times 8 dw 1
                      times 8 dw 2
                      times 8 dw 3
                      times 8 dw 4
                      times 8 dw 5
                      times 8 dw 6
                      times 8 dw 7

bilinear_filter_vb_m: times 8 db 7, 1
                      times 8 db 6, 2
                      times 8 db 5, 3
                      times 8 db 4, 4
                      times 8 db 3, 5
                      times 8 db 2, 6
                      times 8 db 1, 7

%ifdef PIC
%define fourtap_filter_hw  picregq
%define sixtap_filter_hw   picregq
%define fourtap_filter_hb  picregq
%define sixtap_filter_hb   picregq
%define fourtap_filter_v   picregq
%define sixtap_filter_v    picregq
%define bilinear_filter_vw picregq
%define bilinear_filter_vb picregq
%define npicregs 1
%else
%define fourtap_filter_hw  fourtap_filter_hw_m
%define sixtap_filter_hw   sixtap_filter_hw_m
%define fourtap_filter_hb  fourtap_filter_hb_m
%define sixtap_filter_hb   sixtap_filter_hb_m
%define fourtap_filter_v   fourtap_filter_v_m
%define sixtap_filter_v    sixtap_filter_v_m
%define bilinear_filter_vw bilinear_filter_vw_m
%define bilinear_filter_vb bilinear_filter_vb_m
%define npicregs 0
%endif

filter_h2_shuf:  db 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5,  6, 6,  7,  7,  8
filter_h4_shuf:  db 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7,  8, 8,  9,  9, 10

filter_h6_shuf1: db 0, 5, 1, 6, 2, 7, 3, 8, 4, 9, 5, 10, 6, 11,  7, 12
filter_h6_shuf2: db 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6,  7, 7,  8,  8,  9
filter_h6_shuf3: db 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8,  9, 9, 10, 10, 11

pw_256:  times 8 dw 256

pw_20091: times 4 dw 20091
pw_17734: times 4 dw 17734

pb_27_63: times 8 db 27, 63
pb_18_63: times 8 db 18, 63
pb_9_63:  times 8 db  9, 63

cextern pb_1
cextern pw_3
cextern pb_3
cextern pw_4
cextern pb_4
cextern pw_9
cextern pw_18
cextern pw_27
cextern pw_63
cextern pw_64
cextern pb_80
cextern pb_F8
cextern pb_FE

SECTION .text

;-----------------------------------------------------------------------------
; subpel MC functions:
;
; void put_vp8_epel<size>_h<htap>v<vtap>_<opt>(uint8_t *dst, int deststride,
;                                              uint8_t *src, int srcstride,
;                                              int height,   int mx, int my);
;-----------------------------------------------------------------------------

%macro FILTER_SSSE3 1
cglobal put_vp8_epel%1_h6, 6, 6 + npicregs, 8, dst, dststride, src, srcstride, height, mx, picreg
    lea      mxd, [mxq*3]
    mova      m3, [filter_h6_shuf2]
    mova      m4, [filter_h6_shuf3]
%ifdef PIC
    lea  picregq, [sixtap_filter_hb_m]
%endif
    mova      m5, [sixtap_filter_hb+mxq*8-48] ; set up 6tap filter in bytes
    mova      m6, [sixtap_filter_hb+mxq*8-32]
    mova      m7, [sixtap_filter_hb+mxq*8-16]

.nextrow
    movu      m0, [srcq-2]
    mova      m1, m0
    mova      m2, m0
%if mmsize == 8
; For epel4, we need 9 bytes, but only 8 get loaded; to compensate, do the
; shuffle with a memory operand
    punpcklbw m0, [srcq+3]
%else
    pshufb    m0, [filter_h6_shuf1]
%endif
    pshufb    m1, m3
    pshufb    m2, m4
    pmaddubsw m0, m5
    pmaddubsw m1, m6
    pmaddubsw m2, m7
    paddsw    m0, m1
    paddsw    m0, m2
    pmulhrsw  m0, [pw_256]
    packuswb  m0, m0
    movh  [dstq], m0        ; store

    ; go to next line
    add     dstq, dststrideq
    add     srcq, srcstrideq
    dec  heightd            ; next row
    jg .nextrow
    REP_RET

cglobal put_vp8_epel%1_h4, 6, 6 + npicregs, 7, dst, dststride, src, srcstride, height, mx, picreg
    shl      mxd, 4
    mova      m2, [pw_256]
    mova      m3, [filter_h2_shuf]
    mova      m4, [filter_h4_shuf]
%ifdef PIC
    lea  picregq, [fourtap_filter_hb_m]
%endif
    mova      m5, [fourtap_filter_hb+mxq-16] ; set up 4tap filter in bytes
    mova      m6, [fourtap_filter_hb+mxq]

.nextrow
    movu      m0, [srcq-1]
    mova      m1, m0
    pshufb    m0, m3
    pshufb    m1, m4
    pmaddubsw m0, m5
    pmaddubsw m1, m6
    paddsw    m0, m1
    pmulhrsw  m0, m2
    packuswb  m0, m0
    movh  [dstq], m0        ; store

    ; go to next line
    add     dstq, dststrideq
    add     srcq, srcstrideq
    dec  heightd            ; next row
    jg .nextrow
    REP_RET

cglobal put_vp8_epel%1_v4, 7, 7, 8, dst, dststride, src, srcstride, height, picreg, my
    shl      myd, 4
%ifdef PIC
    lea  picregq, [fourtap_filter_hb_m]
%endif
    mova      m5, [fourtap_filter_hb+myq-16]
    mova      m6, [fourtap_filter_hb+myq]
    mova      m7, [pw_256]

    ; read 3 lines
    sub     srcq, srcstrideq
    movh      m0, [srcq]
    movh      m1, [srcq+  srcstrideq]
    movh      m2, [srcq+2*srcstrideq]
    add     srcq, srcstrideq

.nextrow
    movh      m3, [srcq+2*srcstrideq]      ; read new row
    mova      m4, m0
    mova      m0, m1
    punpcklbw m4, m1
    mova      m1, m2
    punpcklbw m2, m3
    pmaddubsw m4, m5
    pmaddubsw m2, m6
    paddsw    m4, m2
    mova      m2, m3
    pmulhrsw  m4, m7
    packuswb  m4, m4
    movh  [dstq], m4

    ; go to next line
    add      dstq, dststrideq
    add      srcq, srcstrideq
    dec   heightd                          ; next row
    jg .nextrow
    REP_RET

cglobal put_vp8_epel%1_v6, 7, 7, 8, dst, dststride, src, srcstride, height, picreg, my
    lea      myd, [myq*3]
%ifdef PIC
    lea  picregq, [sixtap_filter_hb_m]
%endif
    lea      myq, [sixtap_filter_hb+myq*8]

    ; read 5 lines
    sub     srcq, srcstrideq
    sub     srcq, srcstrideq
    movh      m0, [srcq]
    movh      m1, [srcq+srcstrideq]
    movh      m2, [srcq+srcstrideq*2]
    lea     srcq, [srcq+srcstrideq*2]
    add     srcq, srcstrideq
    movh      m3, [srcq]
    movh      m4, [srcq+srcstrideq]

.nextrow
    movh      m5, [srcq+2*srcstrideq]      ; read new row
    mova      m6, m0
    punpcklbw m6, m5
    mova      m0, m1
    punpcklbw m1, m2
    mova      m7, m3
    punpcklbw m7, m4
    pmaddubsw m6, [myq-48]
    pmaddubsw m1, [myq-32]
    pmaddubsw m7, [myq-16]
    paddsw    m6, m1
    paddsw    m6, m7
    mova      m1, m2
    mova      m2, m3
    pmulhrsw  m6, [pw_256]
    mova      m3, m4
    packuswb  m6, m6
    mova      m4, m5
    movh  [dstq], m6

    ; go to next line
    add      dstq, dststrideq
    add      srcq, srcstrideq
    dec   heightd                          ; next row
    jg .nextrow
    REP_RET
%endmacro

INIT_MMX ssse3
FILTER_SSSE3 4
INIT_XMM ssse3
FILTER_SSSE3 8

; 4x4 block, H-only 4-tap filter
INIT_MMX mmx2
cglobal put_vp8_epel4_h4, 6, 6 + npicregs, 0, dst, dststride, src, srcstride, height, mx, picreg
    shl       mxd, 4
%ifdef PIC
    lea   picregq, [fourtap_filter_hw_m]
%endif
    movq      mm4, [fourtap_filter_hw+mxq-16] ; set up 4tap filter in words
    movq      mm5, [fourtap_filter_hw+mxq]
    movq      mm7, [pw_64]
    pxor      mm6, mm6

.nextrow
    movq      mm1, [srcq-1]                ; (ABCDEFGH) load 8 horizontal pixels

    ; first set of 2 pixels
    movq      mm2, mm1                     ; byte ABCD..
    punpcklbw mm1, mm6                     ; byte->word ABCD
    pshufw    mm0, mm2, 9                  ; byte CDEF..
    punpcklbw mm0, mm6                     ; byte->word CDEF
    pshufw    mm3, mm1, 0x94               ; word ABBC
    pshufw    mm1, mm0, 0x94               ; word CDDE
    pmaddwd   mm3, mm4                     ; multiply 2px with F0/F1
    movq      mm0, mm1                     ; backup for second set of pixels
    pmaddwd   mm1, mm5                     ; multiply 2px with F2/F3
    paddd     mm3, mm1                     ; finish 1st 2px

    ; second set of 2 pixels, use backup of above
    punpckhbw mm2, mm6                     ; byte->word EFGH
    pmaddwd   mm0, mm4                     ; multiply backed up 2px with F0/F1
    pshufw    mm1, mm2, 0x94               ; word EFFG
    pmaddwd   mm1, mm5                     ; multiply 2px with F2/F3
    paddd     mm0, mm1                     ; finish 2nd 2px

    ; merge two sets of 2 pixels into one set of 4, round/clip/store
    packssdw  mm3, mm0                     ; merge dword->word (4px)
    paddsw    mm3, mm7                     ; rounding
    psraw     mm3, 7
    packuswb  mm3, mm6                     ; clip and word->bytes
    movd   [dstq], mm3                     ; store

    ; go to next line
    add      dstq, dststrideq
    add      srcq, srcstrideq
    dec   heightd                          ; next row
    jg .nextrow
    REP_RET

; 4x4 block, H-only 6-tap filter
INIT_MMX mmx2
cglobal put_vp8_epel4_h6, 6, 6 + npicregs, 0, dst, dststride, src, srcstride, height, mx, picreg
    lea       mxd, [mxq*3]
%ifdef PIC
    lea   picregq, [sixtap_filter_hw_m]
%endif
    movq      mm4, [sixtap_filter_hw+mxq*8-48] ; set up 4tap filter in words
    movq      mm5, [sixtap_filter_hw+mxq*8-32]
    movq      mm6, [sixtap_filter_hw+mxq*8-16]
    movq      mm7, [pw_64]
    pxor      mm3, mm3

.nextrow
    movq      mm1, [srcq-2]                ; (ABCDEFGH) load 8 horizontal pixels

    ; first set of 2 pixels
    movq      mm2, mm1                     ; byte ABCD..
    punpcklbw mm1, mm3                     ; byte->word ABCD
    pshufw    mm0, mm2, 0x9                ; byte CDEF..
    punpckhbw mm2, mm3                     ; byte->word EFGH
    punpcklbw mm0, mm3                     ; byte->word CDEF
    pshufw    mm1, mm1, 0x94               ; word ABBC
    pshufw    mm2, mm2, 0x94               ; word EFFG
    pmaddwd   mm1, mm4                     ; multiply 2px with F0/F1
    pshufw    mm3, mm0, 0x94               ; word CDDE
    movq      mm0, mm3                     ; backup for second set of pixels
    pmaddwd   mm3, mm5                     ; multiply 2px with F2/F3
    paddd     mm1, mm3                     ; add to 1st 2px cache
    movq      mm3, mm2                     ; backup for second set of pixels
    pmaddwd   mm2, mm6                     ; multiply 2px with F4/F5
    paddd     mm1, mm2                     ; finish 1st 2px

    ; second set of 2 pixels, use backup of above
    movd      mm2, [srcq+3]                ; byte FGHI (prevent overreads)
    pmaddwd   mm0, mm4                     ; multiply 1st backed up 2px with F0/F1
    pmaddwd   mm3, mm5                     ; multiply 2nd backed up 2px with F2/F3
    paddd     mm0, mm3                     ; add to 2nd 2px cache
    pxor      mm3, mm3
    punpcklbw mm2, mm3                     ; byte->word FGHI
    pshufw    mm2, mm2, 0xE9               ; word GHHI
    pmaddwd   mm2, mm6                     ; multiply 2px with F4/F5
    paddd     mm0, mm2                     ; finish 2nd 2px

    ; merge two sets of 2 pixels into one set of 4, round/clip/store
    packssdw  mm1, mm0                     ; merge dword->word (4px)
    paddsw    mm1, mm7                     ; rounding
    psraw     mm1, 7
    packuswb  mm1, mm3                     ; clip and word->bytes
    movd   [dstq], mm1                     ; store

    ; go to next line
    add      dstq, dststrideq
    add      srcq, srcstrideq
    dec   heightd                          ; next row
    jg .nextrow
    REP_RET

INIT_XMM sse2
cglobal put_vp8_epel8_h4, 6, 6 + npicregs, 10, dst, dststride, src, srcstride, height, mx, picreg
    shl      mxd, 5
%ifdef PIC
    lea  picregq, [fourtap_filter_v_m]
%endif
    lea      mxq, [fourtap_filter_v+mxq-32]
    pxor      m7, m7
    mova      m4, [pw_64]
    mova      m5, [mxq+ 0]
    mova      m6, [mxq+16]
%ifdef m8
    mova      m8, [mxq+32]
    mova      m9, [mxq+48]
%endif
.nextrow
    movq      m0, [srcq-1]
    movq      m1, [srcq-0]
    movq      m2, [srcq+1]
    movq      m3, [srcq+2]
    punpcklbw m0, m7
    punpcklbw m1, m7
    punpcklbw m2, m7
    punpcklbw m3, m7
    pmullw    m0, m5
    pmullw    m1, m6
%ifdef m8
    pmullw    m2, m8
    pmullw    m3, m9
%else
    pmullw    m2, [mxq+32]
    pmullw    m3, [mxq+48]
%endif
    paddsw    m0, m1
    paddsw    m2, m3
    paddsw    m0, m2
    paddsw    m0, m4
    psraw     m0, 7
    packuswb  m0, m7
    movh  [dstq], m0        ; store

    ; go to next line
    add     dstq, dststrideq
    add     srcq, srcstrideq
    dec  heightd            ; next row
    jg .nextrow
    REP_RET

INIT_XMM sse2
cglobal put_vp8_epel8_h6, 6, 6 + npicregs, 14, dst, dststride, src, srcstride, height, mx, picreg
    lea      mxd, [mxq*3]
    shl      mxd, 4
%ifdef PIC
    lea  picregq, [sixtap_filter_v_m]
%endif
    lea      mxq, [sixtap_filter_v+mxq-96]
    pxor      m7, m7
    mova      m6, [pw_64]
%ifdef m8
    mova      m8, [mxq+ 0]
    mova      m9, [mxq+16]
    mova     m10, [mxq+32]
    mova     m11, [mxq+48]
    mova     m12, [mxq+64]
    mova     m13, [mxq+80]
%endif
.nextrow
    movq      m0, [srcq-2]
    movq      m1, [srcq-1]
    movq      m2, [srcq-0]
    movq      m3, [srcq+1]
    movq      m4, [srcq+2]
    movq      m5, [srcq+3]
    punpcklbw m0, m7
    punpcklbw m1, m7
    punpcklbw m2, m7
    punpcklbw m3, m7
    punpcklbw m4, m7
    punpcklbw m5, m7
%ifdef m8
    pmullw    m0, m8
    pmullw    m1, m9
    pmullw    m2, m10
    pmullw    m3, m11
    pmullw    m4, m12
    pmullw    m5, m13
%else
    pmullw    m0, [mxq+ 0]
    pmullw    m1, [mxq+16]
    pmullw    m2, [mxq+32]
    pmullw    m3, [mxq+48]
    pmullw    m4, [mxq+64]
    pmullw    m5, [mxq+80]
%endif
    paddsw    m1, m4
    paddsw    m0, m5
    paddsw    m1, m2
    paddsw    m0, m3
    paddsw    m0, m1
    paddsw    m0, m6
    psraw     m0, 7
    packuswb  m0, m7
    movh  [dstq], m0        ; store

    ; go to next line
    add     dstq, dststrideq
    add     srcq, srcstrideq
    dec  heightd            ; next row
    jg .nextrow
    REP_RET

%macro FILTER_V 1
; 4x4 block, V-only 4-tap filter
cglobal put_vp8_epel%1_v4, 7, 7, 8, dst, dststride, src, srcstride, height, picreg, my
    shl      myd, 5
%ifdef PIC
    lea  picregq, [fourtap_filter_v_m]
%endif
    lea      myq, [fourtap_filter_v+myq-32]
    mova      m6, [pw_64]
    pxor      m7, m7
    mova      m5, [myq+48]

    ; read 3 lines
    sub     srcq, srcstrideq
    movh      m0, [srcq]
    movh      m1, [srcq+  srcstrideq]
    movh      m2, [srcq+2*srcstrideq]
    add     srcq, srcstrideq
    punpcklbw m0, m7
    punpcklbw m1, m7
    punpcklbw m2, m7

.nextrow
    ; first calculate negative taps (to prevent losing positive overflows)
    movh      m4, [srcq+2*srcstrideq]      ; read new row
    punpcklbw m4, m7
    mova      m3, m4
    pmullw    m0, [myq+0]
    pmullw    m4, m5
    paddsw    m4, m0

    ; then calculate positive taps
    mova      m0, m1
    pmullw    m1, [myq+16]
    paddsw    m4, m1
    mova      m1, m2
    pmullw    m2, [myq+32]
    paddsw    m4, m2
    mova      m2, m3

    ; round/clip/store
    paddsw    m4, m6
    psraw     m4, 7
    packuswb  m4, m7
    movh  [dstq], m4

    ; go to next line
    add     dstq, dststrideq
    add     srcq, srcstrideq
    dec  heightd                           ; next row
    jg .nextrow
    REP_RET


; 4x4 block, V-only 6-tap filter
cglobal put_vp8_epel%1_v6, 7, 7, 8, dst, dststride, src, srcstride, height, picreg, my
    shl      myd, 4
    lea      myq, [myq*3]
%ifdef PIC
    lea  picregq, [sixtap_filter_v_m]
%endif
    lea      myq, [sixtap_filter_v+myq-96]
    pxor      m7, m7

    ; read 5 lines
    sub     srcq, srcstrideq
    sub     srcq, srcstrideq
    movh      m0, [srcq]
    movh      m1, [srcq+srcstrideq]
    movh      m2, [srcq+srcstrideq*2]
    lea     srcq, [srcq+srcstrideq*2]
    add     srcq, srcstrideq
    movh      m3, [srcq]
    movh      m4, [srcq+srcstrideq]
    punpcklbw m0, m7
    punpcklbw m1, m7
    punpcklbw m2, m7
    punpcklbw m3, m7
    punpcklbw m4, m7

.nextrow
    ; first calculate negative taps (to prevent losing positive overflows)
    mova      m5, m1
    pmullw    m5, [myq+16]
    mova      m6, m4
    pmullw    m6, [myq+64]
    paddsw    m6, m5

    ; then calculate positive taps
    movh      m5, [srcq+2*srcstrideq]      ; read new row
    punpcklbw m5, m7
    pmullw    m0, [myq+0]
    paddsw    m6, m0
    mova      m0, m1
    mova      m1, m2
    pmullw    m2, [myq+32]
    paddsw    m6, m2
    mova      m2, m3
    pmullw    m3, [myq+48]
    paddsw    m6, m3
    mova      m3, m4
    mova      m4, m5
    pmullw    m5, [myq+80]
    paddsw    m6, m5

    ; round/clip/store
    paddsw    m6, [pw_64]
    psraw     m6, 7
    packuswb  m6, m7
    movh  [dstq], m6

    ; go to next line
    add     dstq, dststrideq
    add     srcq, srcstrideq
    dec  heightd                           ; next row
    jg .nextrow
    REP_RET
%endmacro

INIT_MMX mmx2
FILTER_V 4
INIT_XMM sse2
FILTER_V 8

%macro FILTER_BILINEAR 1
cglobal put_vp8_bilinear%1_v, 7, 7, 7, dst, dststride, src, srcstride, height, picreg, my
    shl      myd, 4
%ifdef PIC
    lea  picregq, [bilinear_filter_vw_m]
%endif
    pxor      m6, m6
    mova      m5, [bilinear_filter_vw+myq-1*16]
    neg      myq
    mova      m4, [bilinear_filter_vw+myq+7*16]
.nextrow
    movh      m0, [srcq+srcstrideq*0]
    movh      m1, [srcq+srcstrideq*1]
    movh      m3, [srcq+srcstrideq*2]
    punpcklbw m0, m6
    punpcklbw m1, m6
    punpcklbw m3, m6
    mova      m2, m1
    pmullw    m0, m4
    pmullw    m1, m5
    pmullw    m2, m4
    pmullw    m3, m5
    paddsw    m0, m1
    paddsw    m2, m3
    psraw     m0, 2
    psraw     m2, 2
    pavgw     m0, m6
    pavgw     m2, m6
%if mmsize == 8
    packuswb  m0, m0
    packuswb  m2, m2
    movh   [dstq+dststrideq*0], m0
    movh   [dstq+dststrideq*1], m2
%else
    packuswb  m0, m2
    movh   [dstq+dststrideq*0], m0
    movhps [dstq+dststrideq*1], m0
%endif

    lea     dstq, [dstq+dststrideq*2]
    lea     srcq, [srcq+srcstrideq*2]
    sub  heightd, 2
    jg .nextrow
    REP_RET

cglobal put_vp8_bilinear%1_h, 6, 6 + npicregs, 7, dst, dststride, src, srcstride, height, mx, picreg
    shl      mxd, 4
%ifdef PIC
    lea  picregq, [bilinear_filter_vw_m]
%endif
    pxor      m6, m6
    mova      m5, [bilinear_filter_vw+mxq-1*16]
    neg      mxq
    mova      m4, [bilinear_filter_vw+mxq+7*16]
.nextrow
    movh      m0, [srcq+srcstrideq*0+0]
    movh      m1, [srcq+srcstrideq*0+1]
    movh      m2, [srcq+srcstrideq*1+0]
    movh      m3, [srcq+srcstrideq*1+1]
    punpcklbw m0, m6
    punpcklbw m1, m6
    punpcklbw m2, m6
    punpcklbw m3, m6
    pmullw    m0, m4
    pmullw    m1, m5
    pmullw    m2, m4
    pmullw    m3, m5
    paddsw    m0, m1
    paddsw    m2, m3
    psraw     m0, 2
    psraw     m2, 2
    pavgw     m0, m6
    pavgw     m2, m6
%if mmsize == 8
    packuswb  m0, m0
    packuswb  m2, m2
    movh   [dstq+dststrideq*0], m0
    movh   [dstq+dststrideq*1], m2
%else
    packuswb  m0, m2
    movh   [dstq+dststrideq*0], m0
    movhps [dstq+dststrideq*1], m0
%endif

    lea     dstq, [dstq+dststrideq*2]
    lea     srcq, [srcq+srcstrideq*2]
    sub  heightd, 2
    jg .nextrow
    REP_RET
%endmacro

INIT_MMX mmx2
FILTER_BILINEAR 4
INIT_XMM sse2
FILTER_BILINEAR 8

%macro FILTER_BILINEAR_SSSE3 1
cglobal put_vp8_bilinear%1_v, 7, 7, 5, dst, dststride, src, srcstride, height, picreg, my
    shl      myd, 4
%ifdef PIC
    lea  picregq, [bilinear_filter_vb_m]
%endif
    pxor      m4, m4
    mova      m3, [bilinear_filter_vb+myq-16]
.nextrow
    movh      m0, [srcq+srcstrideq*0]
    movh      m1, [srcq+srcstrideq*1]
    movh      m2, [srcq+srcstrideq*2]
    punpcklbw m0, m1
    punpcklbw m1, m2
    pmaddubsw m0, m3
    pmaddubsw m1, m3
    psraw     m0, 2
    psraw     m1, 2
    pavgw     m0, m4
    pavgw     m1, m4
%if mmsize==8
    packuswb  m0, m0
    packuswb  m1, m1
    movh   [dstq+dststrideq*0], m0
    movh   [dstq+dststrideq*1], m1
%else
    packuswb  m0, m1
    movh   [dstq+dststrideq*0], m0
    movhps [dstq+dststrideq*1], m0
%endif

    lea     dstq, [dstq+dststrideq*2]
    lea     srcq, [srcq+srcstrideq*2]
    sub  heightd, 2
    jg .nextrow
    REP_RET

cglobal put_vp8_bilinear%1_h, 6, 6 + npicregs, 5, dst, dststride, src, srcstride, height, mx, picreg
    shl      mxd, 4
%ifdef PIC
    lea  picregq, [bilinear_filter_vb_m]
%endif
    pxor      m4, m4
    mova      m2, [filter_h2_shuf]
    mova      m3, [bilinear_filter_vb+mxq-16]
.nextrow
    movu      m0, [srcq+srcstrideq*0]
    movu      m1, [srcq+srcstrideq*1]
    pshufb    m0, m2
    pshufb    m1, m2
    pmaddubsw m0, m3
    pmaddubsw m1, m3
    psraw     m0, 2
    psraw     m1, 2
    pavgw     m0, m4
    pavgw     m1, m4
%if mmsize==8
    packuswb  m0, m0
    packuswb  m1, m1
    movh   [dstq+dststrideq*0], m0
    movh   [dstq+dststrideq*1], m1
%else
    packuswb  m0, m1
    movh   [dstq+dststrideq*0], m0
    movhps [dstq+dststrideq*1], m0
%endif

    lea     dstq, [dstq+dststrideq*2]
    lea     srcq, [srcq+srcstrideq*2]
    sub  heightd, 2
    jg .nextrow
    REP_RET
%endmacro

INIT_MMX ssse3
FILTER_BILINEAR_SSSE3 4
INIT_XMM ssse3
FILTER_BILINEAR_SSSE3 8

INIT_MMX mmx
cglobal put_vp8_pixels8, 5, 5, 0, dst, dststride, src, srcstride, height
.nextrow:
    movq    mm0, [srcq+srcstrideq*0]
    movq    mm1, [srcq+srcstrideq*1]
    lea    srcq, [srcq+srcstrideq*2]
    movq [dstq+dststrideq*0], mm0
    movq [dstq+dststrideq*1], mm1
    lea    dstq, [dstq+dststrideq*2]
    sub heightd, 2
    jg .nextrow
    REP_RET

%if ARCH_X86_32
INIT_MMX mmx
cglobal put_vp8_pixels16, 5, 5, 0, dst, dststride, src, srcstride, height
.nextrow:
    movq    mm0, [srcq+srcstrideq*0+0]
    movq    mm1, [srcq+srcstrideq*0+8]
    movq    mm2, [srcq+srcstrideq*1+0]
    movq    mm3, [srcq+srcstrideq*1+8]
    lea    srcq, [srcq+srcstrideq*2]
    movq [dstq+dststrideq*0+0], mm0
    movq [dstq+dststrideq*0+8], mm1
    movq [dstq+dststrideq*1+0], mm2
    movq [dstq+dststrideq*1+8], mm3
    lea    dstq, [dstq+dststrideq*2]
    sub heightd, 2
    jg .nextrow
    REP_RET
%endif

INIT_XMM sse
cglobal put_vp8_pixels16, 5, 5, 2, dst, dststride, src, srcstride, height
.nextrow:
    movups xmm0, [srcq+srcstrideq*0]
    movups xmm1, [srcq+srcstrideq*1]
    lea    srcq, [srcq+srcstrideq*2]
    movaps [dstq+dststrideq*0], xmm0
    movaps [dstq+dststrideq*1], xmm1
    lea    dstq, [dstq+dststrideq*2]
    sub heightd, 2
    jg .nextrow
    REP_RET

;-----------------------------------------------------------------------------
; void vp8_idct_dc_add_<opt>(uint8_t *dst, DCTELEM block[16], int stride);
;-----------------------------------------------------------------------------

%macro ADD_DC 4
    %4        m2, [dst1q+%3]
    %4        m3, [dst1q+strideq+%3]
    %4        m4, [dst2q+%3]
    %4        m5, [dst2q+strideq+%3]
    paddusb   m2, %1
    paddusb   m3, %1
    paddusb   m4, %1
    paddusb   m5, %1
    psubusb   m2, %2
    psubusb   m3, %2
    psubusb   m4, %2
    psubusb   m5, %2
    %4 [dst1q+%3], m2
    %4 [dst1q+strideq+%3], m3
    %4 [dst2q+%3], m4
    %4 [dst2q+strideq+%3], m5
%endmacro

INIT_MMX mmx
cglobal vp8_idct_dc_add, 3, 3, 0, dst, block, stride
    ; load data
    movd       m0, [blockq]

    ; calculate DC
    paddw      m0, [pw_4]
    pxor       m1, m1
    psraw      m0, 3
    movd [blockq], m1
    psubw      m1, m0
    packuswb   m0, m0
    packuswb   m1, m1
    punpcklbw  m0, m0
    punpcklbw  m1, m1
    punpcklwd  m0, m0
    punpcklwd  m1, m1

    ; add DC
    DEFINE_ARGS dst1, dst2, stride
    lea     dst2q, [dst1q+strideq*2]
    ADD_DC     m0, m1, 0, movh
    RET

INIT_XMM sse4
cglobal vp8_idct_dc_add, 3, 3, 6, dst, block, stride
    ; load data
    movd       m0, [blockq]
    pxor       m1, m1

    ; calculate DC
    paddw      m0, [pw_4]
    movd [blockq], m1
    DEFINE_ARGS dst1, dst2, stride
    lea     dst2q, [dst1q+strideq*2]
    movd       m2, [dst1q]
    movd       m3, [dst1q+strideq]
    movd       m4, [dst2q]
    movd       m5, [dst2q+strideq]
    psraw      m0, 3
    pshuflw    m0, m0, 0
    punpcklqdq m0, m0
    punpckldq  m2, m3
    punpckldq  m4, m5
    punpcklbw  m2, m1
    punpcklbw  m4, m1
    paddw      m2, m0
    paddw      m4, m0
    packuswb   m2, m4
    movd   [dst1q], m2
    pextrd [dst1q+strideq], m2, 1
    pextrd [dst2q], m2, 2
    pextrd [dst2q+strideq], m2, 3
    RET

;-----------------------------------------------------------------------------
; void vp8_idct_dc_add4y_<opt>(uint8_t *dst, DCTELEM block[4][16], int stride);
;-----------------------------------------------------------------------------

%if ARCH_X86_32
INIT_MMX mmx
cglobal vp8_idct_dc_add4y, 3, 3, 0, dst, block, stride
    ; load data
    movd      m0, [blockq+32*0] ; A
    movd      m1, [blockq+32*2] ; C
    punpcklwd m0, [blockq+32*1] ; A B
    punpcklwd m1, [blockq+32*3] ; C D
    punpckldq m0, m1        ; A B C D
    pxor      m6, m6

    ; calculate DC
    paddw     m0, [pw_4]
    movd [blockq+32*0], m6
    movd [blockq+32*1], m6
    movd [blockq+32*2], m6
    movd [blockq+32*3], m6
    psraw     m0, 3
    psubw     m6, m0
    packuswb  m0, m0
    packuswb  m6, m6
    punpcklbw m0, m0 ; AABBCCDD
    punpcklbw m6, m6 ; AABBCCDD
    movq      m1, m0
    movq      m7, m6
    punpcklbw m0, m0 ; AAAABBBB
    punpckhbw m1, m1 ; CCCCDDDD
    punpcklbw m6, m6 ; AAAABBBB
    punpckhbw m7, m7 ; CCCCDDDD

    ; add DC
    DEFINE_ARGS dst1, dst2, stride
    lea    dst2q, [dst1q+strideq*2]
    ADD_DC    m0, m6, 0, mova
    ADD_DC    m1, m7, 8, mova
    RET
%endif

INIT_XMM sse2
cglobal vp8_idct_dc_add4y, 3, 3, 6, dst, block, stride
    ; load data
    movd      m0, [blockq+32*0] ; A
    movd      m1, [blockq+32*2] ; C
    punpcklwd m0, [blockq+32*1] ; A B
    punpcklwd m1, [blockq+32*3] ; C D
    punpckldq m0, m1        ; A B C D
    pxor      m1, m1

    ; calculate DC
    paddw     m0, [pw_4]
    movd [blockq+32*0], m1
    movd [blockq+32*1], m1
    movd [blockq+32*2], m1
    movd [blockq+32*3], m1
    psraw     m0, 3
    psubw     m1, m0
    packuswb  m0, m0
    packuswb  m1, m1
    punpcklbw m0, m0
    punpcklbw m1, m1
    punpcklbw m0, m0
    punpcklbw m1, m1

    ; add DC
    DEFINE_ARGS dst1, dst2, stride
    lea    dst2q, [dst1q+strideq*2]
    ADD_DC    m0, m1, 0, mova
    RET

;-----------------------------------------------------------------------------
; void vp8_idct_dc_add4uv_<opt>(uint8_t *dst, DCTELEM block[4][16], int stride);
;-----------------------------------------------------------------------------

INIT_MMX mmx
cglobal vp8_idct_dc_add4uv, 3, 3, 0, dst, block, stride
    ; load data
    movd      m0, [blockq+32*0] ; A
    movd      m1, [blockq+32*2] ; C
    punpcklwd m0, [blockq+32*1] ; A B
    punpcklwd m1, [blockq+32*3] ; C D
    punpckldq m0, m1        ; A B C D
    pxor      m6, m6

    ; calculate DC
    paddw     m0, [pw_4]
    movd [blockq+32*0], m6
    movd [blockq+32*1], m6
    movd [blockq+32*2], m6
    movd [blockq+32*3], m6
    psraw     m0, 3
    psubw     m6, m0
    packuswb  m0, m0
    packuswb  m6, m6
    punpcklbw m0, m0 ; AABBCCDD
    punpcklbw m6, m6 ; AABBCCDD
    movq      m1, m0
    movq      m7, m6
    punpcklbw m0, m0 ; AAAABBBB
    punpckhbw m1, m1 ; CCCCDDDD
    punpcklbw m6, m6 ; AAAABBBB
    punpckhbw m7, m7 ; CCCCDDDD

    ; add DC
    DEFINE_ARGS dst1, dst2, stride
    lea    dst2q, [dst1q+strideq*2]
    ADD_DC    m0, m6, 0, mova
    lea    dst1q, [dst1q+strideq*4]
    lea    dst2q, [dst2q+strideq*4]
    ADD_DC    m1, m7, 0, mova
    RET

;-----------------------------------------------------------------------------
; void vp8_idct_add_<opt>(uint8_t *dst, DCTELEM block[16], int stride);
;-----------------------------------------------------------------------------

; calculate %1=mul_35468(%1)-mul_20091(%2); %2=mul_20091(%1)+mul_35468(%2)
;           this macro assumes that m6/m7 have words for 20091/17734 loaded
%macro VP8_MULTIPLY_SUMSUB 4
    mova      %3, %1
    mova      %4, %2
    pmulhw    %3, m6 ;20091(1)
    pmulhw    %4, m6 ;20091(2)
    paddw     %3, %1
    paddw     %4, %2
    paddw     %1, %1
    paddw     %2, %2
    pmulhw    %1, m7 ;35468(1)
    pmulhw    %2, m7 ;35468(2)
    psubw     %1, %4
    paddw     %2, %3
%endmacro

; calculate x0=%1+%3; x1=%1-%3
;           x2=mul_35468(%2)-mul_20091(%4); x3=mul_20091(%2)+mul_35468(%4)
;           %1=x0+x3 (tmp0); %2=x1+x2 (tmp1); %3=x1-x2 (tmp2); %4=x0-x3 (tmp3)
;           %5/%6 are temporary registers
;           we assume m6/m7 have constant words 20091/17734 loaded in them
%macro VP8_IDCT_TRANSFORM4x4_1D 6
    SUMSUB_BA         w, %3,  %1,  %5     ;t0, t1
    VP8_MULTIPLY_SUMSUB m%2, m%4, m%5,m%6 ;t2, t3
    SUMSUB_BA         w, %4,  %3,  %5     ;tmp0, tmp3
    SUMSUB_BA         w, %2,  %1,  %5     ;tmp1, tmp2
    SWAP                 %4,  %1
    SWAP                 %4,  %3
%endmacro

%macro VP8_IDCT_ADD 0
cglobal vp8_idct_add, 3, 3, 0, dst, block, stride
    ; load block data
    movq         m0, [blockq+ 0]
    movq         m1, [blockq+ 8]
    movq         m2, [blockq+16]
    movq         m3, [blockq+24]
    movq         m6, [pw_20091]
    movq         m7, [pw_17734]
%if cpuflag(sse)
    xorps      xmm0, xmm0
    movaps [blockq+ 0], xmm0
    movaps [blockq+16], xmm0
%else
    pxor         m4, m4
    movq [blockq+ 0], m4
    movq [blockq+ 8], m4
    movq [blockq+16], m4
    movq [blockq+24], m4
%endif

    ; actual IDCT
    VP8_IDCT_TRANSFORM4x4_1D 0, 1, 2, 3, 4, 5
    TRANSPOSE4x4W            0, 1, 2, 3, 4
    paddw        m0, [pw_4]
    VP8_IDCT_TRANSFORM4x4_1D 0, 1, 2, 3, 4, 5
    TRANSPOSE4x4W            0, 1, 2, 3, 4

    ; store
    pxor         m4, m4
    DEFINE_ARGS dst1, dst2, stride
    lea       dst2q, [dst1q+2*strideq]
    STORE_DIFFx2 m0, m1, m6, m7, m4, 3, dst1q, strideq
    STORE_DIFFx2 m2, m3, m6, m7, m4, 3, dst2q, strideq

    RET
%endmacro

%if ARCH_X86_32
INIT_MMX mmx
VP8_IDCT_ADD
%endif
INIT_MMX sse
VP8_IDCT_ADD

;-----------------------------------------------------------------------------
; void vp8_luma_dc_wht_mmxext(DCTELEM block[4][4][16], DCTELEM dc[16])
;-----------------------------------------------------------------------------

%macro SCATTER_WHT 3
    movd dc1d, m%1
    movd dc2d, m%2
    mov [blockq+2*16*(0+%3)], dc1w
    mov [blockq+2*16*(1+%3)], dc2w
    shr  dc1d, 16
    shr  dc2d, 16
    psrlq m%1, 32
    psrlq m%2, 32
    mov [blockq+2*16*(4+%3)], dc1w
    mov [blockq+2*16*(5+%3)], dc2w
    movd dc1d, m%1
    movd dc2d, m%2
    mov [blockq+2*16*(8+%3)], dc1w
    mov [blockq+2*16*(9+%3)], dc2w
    shr  dc1d, 16
    shr  dc2d, 16
    mov [blockq+2*16*(12+%3)], dc1w
    mov [blockq+2*16*(13+%3)], dc2w
%endmacro

%macro HADAMARD4_1D 4
    SUMSUB_BADC w, %2, %1, %4, %3
    SUMSUB_BADC w, %4, %2, %3, %1
    SWAP %1, %4, %3
%endmacro

%macro VP8_DC_WHT 0
cglobal vp8_luma_dc_wht, 2, 3, 0, block, dc1, dc2
    movq          m0, [dc1q]
    movq          m1, [dc1q+8]
    movq          m2, [dc1q+16]
    movq          m3, [dc1q+24]
%if cpuflag(sse)
    xorps      xmm0, xmm0
    movaps [dc1q+ 0], xmm0
    movaps [dc1q+16], xmm0
%else
    pxor         m4, m4
    movq  [dc1q+ 0], m4
    movq  [dc1q+ 8], m4
    movq  [dc1q+16], m4
    movq  [dc1q+24], m4
%endif
    HADAMARD4_1D  0, 1, 2, 3
    TRANSPOSE4x4W 0, 1, 2, 3, 4
    paddw         m0, [pw_3]
    HADAMARD4_1D  0, 1, 2, 3
    psraw         m0, 3
    psraw         m1, 3
    psraw         m2, 3
    psraw         m3, 3
    SCATTER_WHT   0, 1, 0
    SCATTER_WHT   2, 3, 2
    RET
%endmacro

%if ARCH_X86_32
INIT_MMX mmx
VP8_DC_WHT
%endif
INIT_MMX sse
VP8_DC_WHT

;-----------------------------------------------------------------------------
; void vp8_h/v_loop_filter_simple_<opt>(uint8_t *dst, int stride, int flim);
;-----------------------------------------------------------------------------

; macro called with 7 mm register indexes as argument, and 4 regular registers
;
; first 4 mm registers will carry the transposed pixel data
; the other three are scratchspace (one would be sufficient, but this allows
; for more spreading/pipelining and thus faster execution on OOE CPUs)
;
; first two regular registers are buf+4*stride and buf+5*stride
; third is -stride, fourth is +stride
%macro READ_8x4_INTERLEAVED 11
    ; interleave 8 (A-H) rows of 4 pixels each
    movd          m%1, [%8+%10*4]   ; A0-3
    movd          m%5, [%9+%10*4]   ; B0-3
    movd          m%2, [%8+%10*2]   ; C0-3
    movd          m%6, [%8+%10]     ; D0-3
    movd          m%3, [%8]         ; E0-3
    movd          m%7, [%9]         ; F0-3
    movd          m%4, [%9+%11]     ; G0-3
    punpcklbw     m%1, m%5          ; A/B interleaved
    movd          m%5, [%9+%11*2]   ; H0-3
    punpcklbw     m%2, m%6          ; C/D interleaved
    punpcklbw     m%3, m%7          ; E/F interleaved
    punpcklbw     m%4, m%5          ; G/H interleaved
%endmacro

; macro called with 7 mm register indexes as argument, and 5 regular registers
; first 11 mean the same as READ_8x4_TRANSPOSED above
; fifth regular register is scratchspace to reach the bottom 8 rows, it
; will be set to second regular register + 8*stride at the end
%macro READ_16x4_INTERLEAVED 12
    ; transpose 16 (A-P) rows of 4 pixels each
    lea           %12, [r0+8*r2]

    ; read (and interleave) those addressable by %8 (=r0), A/C/D/E/I/K/L/M
    movd          m%1, [%8+%10*4]   ; A0-3
    movd          m%3, [%12+%10*4]  ; I0-3
    movd          m%2, [%8+%10*2]   ; C0-3
    movd          m%4, [%12+%10*2]  ; K0-3
    movd          m%6, [%8+%10]     ; D0-3
    movd          m%5, [%12+%10]    ; L0-3
    movd          m%7, [%12]        ; M0-3
    add           %12, %11
    punpcklbw     m%1, m%3          ; A/I
    movd          m%3, [%8]         ; E0-3
    punpcklbw     m%2, m%4          ; C/K
    punpcklbw     m%6, m%5          ; D/L
    punpcklbw     m%3, m%7          ; E/M
    punpcklbw     m%2, m%6          ; C/D/K/L interleaved

    ; read (and interleave) those addressable by %9 (=r4), B/F/G/H/J/N/O/P
    movd         m%5, [%9+%10*4]   ; B0-3
    movd         m%4, [%12+%10*4]  ; J0-3
    movd         m%7, [%9]         ; F0-3
    movd         m%6, [%12]        ; N0-3
    punpcklbw    m%5, m%4          ; B/J
    punpcklbw    m%7, m%6          ; F/N
    punpcklbw    m%1, m%5          ; A/B/I/J interleaved
    punpcklbw    m%3, m%7          ; E/F/M/N interleaved
    movd         m%4, [%9+%11]     ; G0-3
    movd         m%6, [%12+%11]    ; O0-3
    movd         m%5, [%9+%11*2]   ; H0-3
    movd         m%7, [%12+%11*2]  ; P0-3
    punpcklbw    m%4, m%6          ; G/O
    punpcklbw    m%5, m%7          ; H/P
    punpcklbw    m%4, m%5          ; G/H/O/P interleaved
%endmacro

; write 4 mm registers of 2 dwords each
; first four arguments are mm register indexes containing source data
; last four are registers containing buf+4*stride, buf+5*stride,
; -stride and +stride
%macro WRITE_4x2D 8
    ; write out (2 dwords per register)
    movd    [%5+%7*4], m%1
    movd    [%5+%7*2], m%2
    movd         [%5], m%3
    movd      [%6+%8], m%4
    punpckhdq     m%1, m%1
    punpckhdq     m%2, m%2
    punpckhdq     m%3, m%3
    punpckhdq     m%4, m%4
    movd    [%6+%7*4], m%1
    movd      [%5+%7], m%2
    movd         [%6], m%3
    movd    [%6+%8*2], m%4
%endmacro

; write 4 xmm registers of 4 dwords each
; arguments same as WRITE_2x4D, but with an extra register, so that the 5 regular
; registers contain buf+4*stride, buf+5*stride, buf+12*stride, -stride and +stride
; we add 1*stride to the third regular registry in the process
; the 10th argument is 16 if it's a Y filter (i.e. all regular registers cover the
; same memory region), or 8 if they cover two separate buffers (third one points to
; a different memory region than the first two), allowing for more optimal code for
; the 16-width case
%macro WRITE_4x4D 10
    ; write out (4 dwords per register), start with dwords zero
    movd    [%5+%8*4], m%1
    movd         [%5], m%2
    movd    [%7+%8*4], m%3
    movd         [%7], m%4

    ; store dwords 1
    psrldq        m%1, 4
    psrldq        m%2, 4
    psrldq        m%3, 4
    psrldq        m%4, 4
    movd    [%6+%8*4], m%1
    movd         [%6], m%2
%if %10 == 16
    movd    [%6+%9*4], m%3
%endif
    movd      [%7+%9], m%4

    ; write dwords 2
    psrldq        m%1, 4
    psrldq        m%2, 4
%if %10 == 8
    movd    [%5+%8*2], m%1
    movd          %5d, m%3
%endif
    psrldq        m%3, 4
    psrldq        m%4, 4
%if %10 == 16
    movd    [%5+%8*2], m%1
%endif
    movd      [%6+%9], m%2
    movd    [%7+%8*2], m%3
    movd    [%7+%9*2], m%4
    add            %7, %9

    ; store dwords 3
    psrldq        m%1, 4
    psrldq        m%2, 4
    psrldq        m%3, 4
    psrldq        m%4, 4
%if %10 == 8
    mov     [%7+%8*4], %5d
    movd    [%6+%8*2], m%1
%else
    movd      [%5+%8], m%1
%endif
    movd    [%6+%9*2], m%2
    movd    [%7+%8*2], m%3
    movd    [%7+%9*2], m%4
%endmacro

; write 4 or 8 words in the mmx/xmm registers as 8 lines
; 1 and 2 are the registers to write, this can be the same (for SSE2)
; for pre-SSE4:
; 3 is a general-purpose register that we will clobber
; for SSE4:
; 3 is a pointer to the destination's 5th line
; 4 is a pointer to the destination's 4th line
; 5/6 is -stride and +stride
%macro WRITE_2x4W 6
    movd            %3d, %1
    punpckhdq        %1, %1
    mov       [%4+%5*4], %3w
    shr              %3, 16
    add              %4, %6
    mov       [%4+%5*4], %3w

    movd            %3d, %1
    add              %4, %5
    mov       [%4+%5*2], %3w
    shr              %3, 16
    mov       [%4+%5  ], %3w

    movd            %3d, %2
    punpckhdq        %2, %2
    mov       [%4     ], %3w
    shr              %3, 16
    mov       [%4+%6  ], %3w

    movd            %3d, %2
    add              %4, %6
    mov       [%4+%6  ], %3w
    shr              %3, 16
    mov       [%4+%6*2], %3w
    add              %4, %5
%endmacro

%macro WRITE_8W 5
%if cpuflag(sse4)
    pextrw    [%3+%4*4], %1, 0
    pextrw    [%2+%4*4], %1, 1
    pextrw    [%3+%4*2], %1, 2
    pextrw    [%3+%4  ], %1, 3
    pextrw    [%3     ], %1, 4
    pextrw    [%2     ], %1, 5
    pextrw    [%2+%5  ], %1, 6
    pextrw    [%2+%5*2], %1, 7
%else
    movd            %2d, %1
    psrldq           %1, 4
    mov       [%3+%4*4], %2w
    shr              %2, 16
    add              %3, %5
    mov       [%3+%4*4], %2w

    movd            %2d, %1
    psrldq           %1, 4
    add              %3, %4
    mov       [%3+%4*2], %2w
    shr              %2, 16
    mov       [%3+%4  ], %2w

    movd            %2d, %1
    psrldq           %1, 4
    mov       [%3     ], %2w
    shr              %2, 16
    mov       [%3+%5  ], %2w

    movd            %2d, %1
    add              %3, %5
    mov       [%3+%5  ], %2w
    shr              %2, 16
    mov       [%3+%5*2], %2w
%endif
%endmacro

%macro SIMPLE_LOOPFILTER 2
cglobal vp8_%1_loop_filter_simple, 3, %2, 8, dst, stride, flim, cntr
%if mmsize == 8 ; mmx/mmxext
    mov         cntrq, 2
%endif
%if cpuflag(ssse3)
    pxor           m0, m0
%endif
    SPLATB_REG     m7, flim, m0     ; splat "flim" into register

    ; set up indexes to address 4 rows
%if mmsize == 8
    DEFINE_ARGS dst1, mstride, stride, cntr, dst2
%else
    DEFINE_ARGS dst1, mstride, stride, dst3, dst2
%endif
    mov       strideq, mstrideq
    neg      mstrideq
%ifidn %1, h
    lea         dst1q, [dst1q+4*strideq-2]
%endif

%if mmsize == 8 ; mmx / mmxext
.next8px
%endif
%ifidn %1, v
    ; read 4 half/full rows of pixels
    mova           m0, [dst1q+mstrideq*2]    ; p1
    mova           m1, [dst1q+mstrideq]      ; p0
    mova           m2, [dst1q]               ; q0
    mova           m3, [dst1q+ strideq]      ; q1
%else ; h
    lea         dst2q, [dst1q+ strideq]

%if mmsize == 8 ; mmx/mmxext
    READ_8x4_INTERLEAVED  0, 1, 2, 3, 4, 5, 6, dst1q, dst2q, mstrideq, strideq
%else ; sse2
    READ_16x4_INTERLEAVED 0, 1, 2, 3, 4, 5, 6, dst1q, dst2q, mstrideq, strideq, dst3q
%endif
    TRANSPOSE4x4W         0, 1, 2, 3, 4
%endif

    ; simple_limit
    mova           m5, m2           ; m5=backup of q0
    mova           m6, m1           ; m6=backup of p0
    psubusb        m1, m2           ; p0-q0
    psubusb        m2, m6           ; q0-p0
    por            m1, m2           ; FFABS(p0-q0)
    paddusb        m1, m1           ; m1=FFABS(p0-q0)*2

    mova           m4, m3
    mova           m2, m0
    psubusb        m3, m0           ; q1-p1
    psubusb        m0, m4           ; p1-q1
    por            m3, m0           ; FFABS(p1-q1)
    mova           m0, [pb_80]
    pxor           m2, m0
    pxor           m4, m0
    psubsb         m2, m4           ; m2=p1-q1 (signed) backup for below
    pand           m3, [pb_FE]
    psrlq          m3, 1            ; m3=FFABS(p1-q1)/2, this can be used signed
    paddusb        m3, m1
    psubusb        m3, m7
    pxor           m1, m1
    pcmpeqb        m3, m1           ; abs(p0-q0)*2+abs(p1-q1)/2<=flim mask(0xff/0x0)

    ; filter_common (use m2/p1-q1, m4=q0, m6=p0, m5/q0-p0 and m3/mask)
    mova           m4, m5
    pxor           m5, m0
    pxor           m0, m6
    psubsb         m5, m0           ; q0-p0 (signed)
    paddsb         m2, m5
    paddsb         m2, m5
    paddsb         m2, m5           ; a=(p1-q1) + 3*(q0-p0)
    pand           m2, m3           ; apply filter mask (m3)

    mova           m3, [pb_F8]
    mova           m1, m2
    paddsb         m2, [pb_4]       ; f1<<3=a+4
    paddsb         m1, [pb_3]       ; f2<<3=a+3
    pand           m2, m3
    pand           m1, m3           ; cache f2<<3

    pxor           m0, m0
    pxor           m3, m3
    pcmpgtb        m0, m2           ; which values are <0?
    psubb          m3, m2           ; -f1<<3
    psrlq          m2, 3            ; +f1
    psrlq          m3, 3            ; -f1
    pand           m3, m0
    pandn          m0, m2
    psubusb        m4, m0
    paddusb        m4, m3           ; q0-f1

    pxor           m0, m0
    pxor           m3, m3
    pcmpgtb        m0, m1           ; which values are <0?
    psubb          m3, m1           ; -f2<<3
    psrlq          m1, 3            ; +f2
    psrlq          m3, 3            ; -f2
    pand           m3, m0
    pandn          m0, m1
    paddusb        m6, m0
    psubusb        m6, m3           ; p0+f2

    ; store
%ifidn %1, v
    mova      [dst1q], m4
    mova [dst1q+mstrideq], m6
%else ; h
    inc        dst1q
    SBUTTERFLY    bw, 6, 4, 0

%if mmsize == 16 ; sse2
%if cpuflag(sse4)
    inc         dst2q
%endif
    WRITE_8W       m6, dst2q, dst1q, mstrideq, strideq
    lea         dst2q, [dst3q+mstrideq+1]
%if cpuflag(sse4)
    inc         dst3q
%endif
    WRITE_8W       m4, dst3q, dst2q, mstrideq, strideq
%else ; mmx/mmxext
    WRITE_2x4W     m6, m4, dst2q, dst1q, mstrideq, strideq
%endif
%endif

%if mmsize == 8 ; mmx/mmxext
    ; next 8 pixels
%ifidn %1, v
    add         dst1q, 8            ; advance 8 cols = pixels
%else ; h
    lea         dst1q, [dst1q+strideq*8-1]  ; advance 8 rows = lines
%endif
    dec         cntrq
    jg .next8px
    REP_RET
%else ; sse2
    RET
%endif
%endmacro

%if ARCH_X86_32
INIT_MMX mmx
SIMPLE_LOOPFILTER v, 4
SIMPLE_LOOPFILTER h, 5
INIT_MMX mmx2
SIMPLE_LOOPFILTER v, 4
SIMPLE_LOOPFILTER h, 5
%endif

INIT_XMM sse2
SIMPLE_LOOPFILTER v, 3
SIMPLE_LOOPFILTER h, 5
INIT_XMM ssse3
SIMPLE_LOOPFILTER v, 3
SIMPLE_LOOPFILTER h, 5
INIT_XMM sse4
SIMPLE_LOOPFILTER h, 5

;-----------------------------------------------------------------------------
; void vp8_h/v_loop_filter<size>_inner_<opt>(uint8_t *dst, [uint8_t *v,] int stride,
;                                            int flimE, int flimI, int hev_thr);
;-----------------------------------------------------------------------------

%macro INNER_LOOPFILTER 2
%if %2 == 8 ; chroma
cglobal vp8_%1_loop_filter8uv_inner, 6, 6, 13, dst, dst8, stride, flimE, flimI, hevthr
%else ; luma
cglobal vp8_%1_loop_filter16y_inner, 5, 5, 13, dst, stride, flimE, flimI, hevthr
%endif

%if cpuflag(ssse3)
    pxor             m7, m7
%endif
%ifndef m8   ; stack layout: [0]=E, [1]=I, [2]=hev_thr
%ifidn %1, v ;               [3]=hev() result
%assign pad 16 + mmsize * 4 - gprsize - (stack_offset & 15)
%else ; h    ; extra storage space for transposes
%assign pad 16 + mmsize * 5 - gprsize - (stack_offset & 15)
%endif
    ; splat function arguments
    SPLATB_REG       m0, flimEq, m7   ; E
    SPLATB_REG       m1, flimIq, m7   ; I
    SPLATB_REG       m2, hevthrq, m7  ; hev_thresh

    SUB             rsp, pad

%define m_flimE    [rsp]
%define m_flimI    [rsp+mmsize]
%define m_hevthr   [rsp+mmsize*2]
%define m_maskres  [rsp+mmsize*3]
%define m_p0backup [rsp+mmsize*3]
%define m_q0backup [rsp+mmsize*4]

    mova        m_flimE, m0
    mova        m_flimI, m1
    mova       m_hevthr, m2
%else
%define m_flimE    m9
%define m_flimI    m10
%define m_hevthr   m11
%define m_maskres  m12
%define m_p0backup m12
%define m_q0backup m8

    ; splat function arguments
    SPLATB_REG  m_flimE, flimEq, m7   ; E
    SPLATB_REG  m_flimI, flimIq, m7   ; I
    SPLATB_REG m_hevthr, hevthrq, m7  ; hev_thresh
%endif

%if %2 == 8 ; chroma
    DEFINE_ARGS dst1, dst8, mstride, stride, dst2
%elif mmsize == 8
    DEFINE_ARGS dst1, mstride, stride, dst2, cntr
    mov           cntrq, 2
%else
    DEFINE_ARGS dst1, mstride, stride, dst2, dst8
%endif
    mov         strideq, mstrideq
    neg        mstrideq
%ifidn %1, h
    lea           dst1q, [dst1q+strideq*4-4]
%if %2 == 8 ; chroma
    lea           dst8q, [dst8q+strideq*4-4]
%endif
%endif

%if mmsize == 8
.next8px:
%endif
    ; read
    lea           dst2q, [dst1q+strideq]
%ifidn %1, v
%if %2 == 8 && mmsize == 16
%define movrow movh
%else
%define movrow mova
%endif
    movrow           m0, [dst1q+mstrideq*4] ; p3
    movrow           m1, [dst2q+mstrideq*4] ; p2
    movrow           m2, [dst1q+mstrideq*2] ; p1
    movrow           m5, [dst2q]            ; q1
    movrow           m6, [dst2q+ strideq*1] ; q2
    movrow           m7, [dst2q+ strideq*2] ; q3
%if mmsize == 16 && %2 == 8
    movhps           m0, [dst8q+mstrideq*4]
    movhps           m2, [dst8q+mstrideq*2]
    add           dst8q, strideq
    movhps           m1, [dst8q+mstrideq*4]
    movhps           m5, [dst8q]
    movhps           m6, [dst8q+ strideq  ]
    movhps           m7, [dst8q+ strideq*2]
    add           dst8q, mstrideq
%endif
%elif mmsize == 8 ; mmx/mmxext (h)
    ; read 8 rows of 8px each
    movu             m0, [dst1q+mstrideq*4]
    movu             m1, [dst2q+mstrideq*4]
    movu             m2, [dst1q+mstrideq*2]
    movu             m3, [dst1q+mstrideq  ]
    movu             m4, [dst1q]
    movu             m5, [dst2q]
    movu             m6, [dst2q+ strideq  ]

    ; 8x8 transpose
    TRANSPOSE4x4B     0, 1, 2, 3, 7
    mova     m_q0backup, m1
    movu             m7, [dst2q+ strideq*2]
    TRANSPOSE4x4B     4, 5, 6, 7, 1
    SBUTTERFLY       dq, 0, 4, 1     ; p3/p2
    SBUTTERFLY       dq, 2, 6, 1     ; q0/q1
    SBUTTERFLY       dq, 3, 7, 1     ; q2/q3
    mova             m1, m_q0backup
    mova     m_q0backup, m2          ; store q0
    SBUTTERFLY       dq, 1, 5, 2     ; p1/p0
    mova     m_p0backup, m5          ; store p0
    SWAP              1, 4
    SWAP              2, 4
    SWAP              6, 3
    SWAP              5, 3
%else ; sse2 (h)
%if %2 == 16
    lea           dst8q, [dst1q+ strideq*8]
%endif

    ; read 16 rows of 8px each, interleave
    movh             m0, [dst1q+mstrideq*4]
    movh             m1, [dst8q+mstrideq*4]
    movh             m2, [dst1q+mstrideq*2]
    movh             m5, [dst8q+mstrideq*2]
    movh             m3, [dst1q+mstrideq  ]
    movh             m6, [dst8q+mstrideq  ]
    movh             m4, [dst1q]
    movh             m7, [dst8q]
    punpcklbw        m0, m1          ; A/I
    punpcklbw        m2, m5          ; C/K
    punpcklbw        m3, m6          ; D/L
    punpcklbw        m4, m7          ; E/M

    add           dst8q, strideq
    movh             m1, [dst2q+mstrideq*4]
    movh             m6, [dst8q+mstrideq*4]
    movh             m5, [dst2q]
    movh             m7, [dst8q]
    punpcklbw        m1, m6          ; B/J
    punpcklbw        m5, m7          ; F/N
    movh             m6, [dst2q+ strideq  ]
    movh             m7, [dst8q+ strideq  ]
    punpcklbw        m6, m7          ; G/O

    ; 8x16 transpose
    TRANSPOSE4x4B     0, 1, 2, 3, 7
%ifdef m8
    SWAP              1, 8
%else
    mova     m_q0backup, m1
%endif
    movh             m7, [dst2q+ strideq*2]
    movh             m1, [dst8q+ strideq*2]
    punpcklbw        m7, m1          ; H/P
    TRANSPOSE4x4B     4, 5, 6, 7, 1
    SBUTTERFLY       dq, 0, 4, 1     ; p3/p2
    SBUTTERFLY       dq, 2, 6, 1     ; q0/q1
    SBUTTERFLY       dq, 3, 7, 1     ; q2/q3
%ifdef m8
    SWAP              1, 8
    SWAP              2, 8
%else
    mova             m1, m_q0backup
    mova     m_q0backup, m2          ; store q0
%endif
    SBUTTERFLY       dq, 1, 5, 2     ; p1/p0
%ifdef m12
    SWAP              5, 12
%else
    mova     m_p0backup, m5          ; store p0
%endif
    SWAP              1, 4
    SWAP              2, 4
    SWAP              6, 3
    SWAP              5, 3
%endif

    ; normal_limit for p3-p2, p2-p1, q3-q2 and q2-q1
    mova             m4, m1
    SWAP              4, 1
    psubusb          m4, m0          ; p2-p3
    psubusb          m0, m1          ; p3-p2
    por              m0, m4          ; abs(p3-p2)

    mova             m4, m2
    SWAP              4, 2
    psubusb          m4, m1          ; p1-p2
    psubusb          m1, m2          ; p2-p1
    por              m1, m4          ; abs(p2-p1)

    mova             m4, m6
    SWAP              4, 6
    psubusb          m4, m7          ; q2-q3
    psubusb          m7, m6          ; q3-q2
    por              m7, m4          ; abs(q3-q2)

    mova             m4, m5
    SWAP              4, 5
    psubusb          m4, m6          ; q1-q2
    psubusb          m6, m5          ; q2-q1
    por              m6, m4          ; abs(q2-q1)

%if notcpuflag(mmx2)
    mova             m4, m_flimI
    pxor             m3, m3
    psubusb          m0, m4
    psubusb          m1, m4
    psubusb          m7, m4
    psubusb          m6, m4
    pcmpeqb          m0, m3          ; abs(p3-p2) <= I
    pcmpeqb          m1, m3          ; abs(p2-p1) <= I
    pcmpeqb          m7, m3          ; abs(q3-q2) <= I
    pcmpeqb          m6, m3          ; abs(q2-q1) <= I
    pand             m0, m1
    pand             m7, m6
    pand             m0, m7
%else ; mmxext/sse2
    pmaxub           m0, m1
    pmaxub           m6, m7
    pmaxub           m0, m6
%endif

    ; normal_limit and high_edge_variance for p1-p0, q1-q0
    SWAP              7, 3           ; now m7 is zero
%ifidn %1, v
    movrow           m3, [dst1q+mstrideq  ] ; p0
%if mmsize == 16 && %2 == 8
    movhps           m3, [dst8q+mstrideq  ]
%endif
%elifdef m12
    SWAP              3, 12
%else
    mova             m3, m_p0backup
%endif

    mova             m1, m2
    SWAP              1, 2
    mova             m6, m3
    SWAP              3, 6
    psubusb          m1, m3          ; p1-p0
    psubusb          m6, m2          ; p0-p1
    por              m1, m6          ; abs(p1-p0)
%if notcpuflag(mmx2)
    mova             m6, m1
    psubusb          m1, m4
    psubusb          m6, m_hevthr
    pcmpeqb          m1, m7          ; abs(p1-p0) <= I
    pcmpeqb          m6, m7          ; abs(p1-p0) <= hev_thresh
    pand             m0, m1
    mova      m_maskres, m6
%else ; mmxext/sse2
    pmaxub           m0, m1          ; max_I
    SWAP              1, 4           ; max_hev_thresh
%endif

    SWAP              6, 4           ; now m6 is I
%ifidn %1, v
    movrow           m4, [dst1q]     ; q0
%if mmsize == 16 && %2 == 8
    movhps           m4, [dst8q]
%endif
%elifdef m8
    SWAP              4, 8
%else
    mova             m4, m_q0backup
%endif
    mova             m1, m4
    SWAP              1, 4
    mova             m7, m5
    SWAP              7, 5
    psubusb          m1, m5          ; q0-q1
    psubusb          m7, m4          ; q1-q0
    por              m1, m7          ; abs(q1-q0)
%if notcpuflag(mmx2)
    mova             m7, m1
    psubusb          m1, m6
    psubusb          m7, m_hevthr
    pxor             m6, m6
    pcmpeqb          m1, m6          ; abs(q1-q0) <= I
    pcmpeqb          m7, m6          ; abs(q1-q0) <= hev_thresh
    mova             m6, m_maskres
    pand             m0, m1          ; abs([pq][321]-[pq][210]) <= I
    pand             m6, m7
%else ; mmxext/sse2
    pxor             m7, m7
    pmaxub           m0, m1
    pmaxub           m6, m1
    psubusb          m0, m_flimI
    psubusb          m6, m_hevthr
    pcmpeqb          m0, m7          ; max(abs(..)) <= I
    pcmpeqb          m6, m7          ; !(max(abs..) > thresh)
%endif
%ifdef m12
    SWAP              6, 12
%else
    mova      m_maskres, m6          ; !(abs(p1-p0) > hev_t || abs(q1-q0) > hev_t)
%endif

    ; simple_limit
    mova             m1, m3
    SWAP              1, 3
    mova             m6, m4          ; keep copies of p0/q0 around for later use
    SWAP              6, 4
    psubusb          m1, m4          ; p0-q0
    psubusb          m6, m3          ; q0-p0
    por              m1, m6          ; abs(q0-p0)
    paddusb          m1, m1          ; m1=2*abs(q0-p0)

    mova             m7, m2
    SWAP              7, 2
    mova             m6, m5
    SWAP              6, 5
    psubusb          m7, m5          ; p1-q1
    psubusb          m6, m2          ; q1-p1
    por              m7, m6          ; abs(q1-p1)
    pxor             m6, m6
    pand             m7, [pb_FE]
    psrlq            m7, 1           ; abs(q1-p1)/2
    paddusb          m7, m1          ; abs(q0-p0)*2+abs(q1-p1)/2
    psubusb          m7, m_flimE
    pcmpeqb          m7, m6          ; abs(q0-p0)*2+abs(q1-p1)/2 <= E
    pand             m0, m7          ; normal_limit result

    ; filter_common; at this point, m2-m5=p1-q1 and m0 is filter_mask
%ifdef m8 ; x86-64 && sse2
    mova             m8, [pb_80]
%define m_pb_80 m8
%else ; x86-32 or mmx/mmxext
%define m_pb_80 [pb_80]
%endif
    mova             m1, m4
    mova             m7, m3
    pxor             m1, m_pb_80
    pxor             m7, m_pb_80
    psubsb           m1, m7          ; (signed) q0-p0
    mova             m6, m2
    mova             m7, m5
    pxor             m6, m_pb_80
    pxor             m7, m_pb_80
    psubsb           m6, m7          ; (signed) p1-q1
    mova             m7, m_maskres
    pandn            m7, m6
    paddsb           m7, m1
    paddsb           m7, m1
    paddsb           m7, m1          ; 3*(q0-p0)+is4tap?(p1-q1)

    pand             m7, m0
    mova             m1, [pb_F8]
    mova             m6, m7
    paddsb           m7, [pb_3]
    paddsb           m6, [pb_4]
    pand             m7, m1
    pand             m6, m1

    pxor             m1, m1
    pxor             m0, m0
    pcmpgtb          m1, m7
    psubb            m0, m7
    psrlq            m7, 3           ; +f2
    psrlq            m0, 3           ; -f2
    pand             m0, m1
    pandn            m1, m7
    psubusb          m3, m0
    paddusb          m3, m1          ; p0+f2

    pxor             m1, m1
    pxor             m0, m0
    pcmpgtb          m0, m6
    psubb            m1, m6
    psrlq            m6, 3           ; +f1
    psrlq            m1, 3           ; -f1
    pand             m1, m0
    pandn            m0, m6
    psubusb          m4, m0
    paddusb          m4, m1          ; q0-f1

%ifdef m12
    SWAP              6, 12
%else
    mova             m6, m_maskres
%endif
%if notcpuflag(mmx2)
    mova             m7, [pb_1]
%else ; mmxext/sse2
    pxor             m7, m7
%endif
    pand             m0, m6
    pand             m1, m6
%if notcpuflag(mmx2)
    paddusb          m0, m7
    pand             m1, [pb_FE]
    pandn            m7, m0
    psrlq            m1, 1
    psrlq            m7, 1
    SWAP              0, 7
%else ; mmxext/sse2
    psubusb          m1, [pb_1]
    pavgb            m0, m7          ; a
    pavgb            m1, m7          ; -a
%endif
    psubusb          m5, m0
    psubusb          m2, m1
    paddusb          m5, m1          ; q1-a
    paddusb          m2, m0          ; p1+a

    ; store
%ifidn %1, v
    movrow [dst1q+mstrideq*2], m2
    movrow [dst1q+mstrideq  ], m3
    movrow      [dst1q], m4
    movrow [dst1q+ strideq  ], m5
%if mmsize == 16 && %2 == 8
    movhps [dst8q+mstrideq*2], m2
    movhps [dst8q+mstrideq  ], m3
    movhps      [dst8q], m4
    movhps [dst8q+ strideq  ], m5
%endif
%else ; h
    add           dst1q, 2
    add           dst2q, 2

    ; 4x8/16 transpose
    TRANSPOSE4x4B     2, 3, 4, 5, 6

%if mmsize == 8 ; mmx/mmxext (h)
    WRITE_4x2D        2, 3, 4, 5, dst1q, dst2q, mstrideq, strideq
%else ; sse2 (h)
    lea           dst8q, [dst8q+mstrideq  +2]
    WRITE_4x4D        2, 3, 4, 5, dst1q, dst2q, dst8q, mstrideq, strideq, %2
%endif
%endif

%if mmsize == 8
%if %2 == 8 ; chroma
%ifidn %1, h
    sub           dst1q, 2
%endif
    cmp           dst1q, dst8q
    mov           dst1q, dst8q
    jnz .next8px
%else
%ifidn %1, h
    lea           dst1q, [dst1q+ strideq*8-2]
%else ; v
    add           dst1q, 8
%endif
    dec           cntrq
    jg .next8px
%endif
%endif

%ifndef m8 ; sse2 on x86-32 or mmx/mmxext
    ADD             rsp, pad
%endif
    RET
%endmacro

%if ARCH_X86_32
INIT_MMX mmx
INNER_LOOPFILTER v, 16
INNER_LOOPFILTER h, 16
INNER_LOOPFILTER v,  8
INNER_LOOPFILTER h,  8

INIT_MMX mmx2
INNER_LOOPFILTER v, 16
INNER_LOOPFILTER h, 16
INNER_LOOPFILTER v,  8
INNER_LOOPFILTER h,  8
%endif

INIT_XMM sse2
INNER_LOOPFILTER v, 16
INNER_LOOPFILTER h, 16
INNER_LOOPFILTER v,  8
INNER_LOOPFILTER h,  8

INIT_XMM ssse3
INNER_LOOPFILTER v, 16
INNER_LOOPFILTER h, 16
INNER_LOOPFILTER v,  8
INNER_LOOPFILTER h,  8

;-----------------------------------------------------------------------------
; void vp8_h/v_loop_filter<size>_mbedge_<opt>(uint8_t *dst, [uint8_t *v,] int stride,
;                                            int flimE, int flimI, int hev_thr);
;-----------------------------------------------------------------------------

%macro MBEDGE_LOOPFILTER 2
%if %2 == 8 ; chroma
cglobal vp8_%1_loop_filter8uv_mbedge, 6, 6, 15, dst1, dst8, stride, flimE, flimI, hevthr
%else ; luma
cglobal vp8_%1_loop_filter16y_mbedge, 5, 5, 15, dst1, stride, flimE, flimI, hevthr
%endif

%if cpuflag(ssse3)
    pxor             m7, m7
%endif
%ifndef m8       ; stack layout: [0]=E, [1]=I, [2]=hev_thr
%if mmsize == 16 ;               [3]=hev() result
                 ;               [4]=filter tmp result
                 ;               [5]/[6] = p2/q2 backup
                 ;               [7]=lim_res sign result
%assign pad 16 + mmsize * 7 - gprsize - (stack_offset & 15)
%else ; 8        ; extra storage space for transposes
%assign pad 16 + mmsize * 8 - gprsize - (stack_offset & 15)
%endif
    ; splat function arguments
    SPLATB_REG       m0, flimEq, m7   ; E
    SPLATB_REG       m1, flimIq, m7   ; I
    SPLATB_REG       m2, hevthrq, m7  ; hev_thresh

    SUB             rsp, pad

%define m_flimE    [rsp]
%define m_flimI    [rsp+mmsize]
%define m_hevthr   [rsp+mmsize*2]
%define m_maskres  [rsp+mmsize*3]
%define m_limres   [rsp+mmsize*4]
%define m_p0backup [rsp+mmsize*3]
%define m_q0backup [rsp+mmsize*4]
%define m_p2backup [rsp+mmsize*5]
%define m_q2backup [rsp+mmsize*6]
%if mmsize == 16
%define m_limsign  [rsp]
%else
%define m_limsign  [rsp+mmsize*7]
%endif

    mova        m_flimE, m0
    mova        m_flimI, m1
    mova       m_hevthr, m2
%else ; sse2 on x86-64
%define m_flimE    m9
%define m_flimI    m10
%define m_hevthr   m11
%define m_maskres  m12
%define m_limres   m8
%define m_p0backup m12
%define m_q0backup m8
%define m_p2backup m13
%define m_q2backup m14
%define m_limsign  m9

    ; splat function arguments
    SPLATB_REG  m_flimE, flimEq, m7   ; E
    SPLATB_REG  m_flimI, flimIq, m7   ; I
    SPLATB_REG m_hevthr, hevthrq, m7  ; hev_thresh
%endif

%if %2 == 8 ; chroma
    DEFINE_ARGS dst1, dst8, mstride, stride, dst2
%elif mmsize == 8
    DEFINE_ARGS dst1, mstride, stride, dst2, cntr
    mov           cntrq, 2
%else
    DEFINE_ARGS dst1, mstride, stride, dst2, dst8
%endif
    mov         strideq, mstrideq
    neg        mstrideq
%ifidn %1, h
    lea           dst1q, [dst1q+strideq*4-4]
%if %2 == 8 ; chroma
    lea           dst8q, [dst8q+strideq*4-4]
%endif
%endif

%if mmsize == 8
.next8px:
%endif
    ; read
    lea           dst2q, [dst1q+ strideq  ]
%ifidn %1, v
%if %2 == 8 && mmsize == 16
%define movrow movh
%else
%define movrow mova
%endif
    movrow           m0, [dst1q+mstrideq*4] ; p3
    movrow           m1, [dst2q+mstrideq*4] ; p2
    movrow           m2, [dst1q+mstrideq*2] ; p1
    movrow           m5, [dst2q]            ; q1
    movrow           m6, [dst2q+ strideq  ] ; q2
    movrow           m7, [dst2q+ strideq*2] ; q3
%if mmsize == 16 && %2 == 8
    movhps           m0, [dst8q+mstrideq*4]
    movhps           m2, [dst8q+mstrideq*2]
    add           dst8q, strideq
    movhps           m1, [dst8q+mstrideq*4]
    movhps           m5, [dst8q]
    movhps           m6, [dst8q+ strideq  ]
    movhps           m7, [dst8q+ strideq*2]
    add           dst8q, mstrideq
%endif
%elif mmsize == 8 ; mmx/mmxext (h)
    ; read 8 rows of 8px each
    movu             m0, [dst1q+mstrideq*4]
    movu             m1, [dst2q+mstrideq*4]
    movu             m2, [dst1q+mstrideq*2]
    movu             m3, [dst1q+mstrideq  ]
    movu             m4, [dst1q]
    movu             m5, [dst2q]
    movu             m6, [dst2q+ strideq  ]

    ; 8x8 transpose
    TRANSPOSE4x4B     0, 1, 2, 3, 7
    mova     m_q0backup, m1
    movu             m7, [dst2q+ strideq*2]
    TRANSPOSE4x4B     4, 5, 6, 7, 1
    SBUTTERFLY       dq, 0, 4, 1     ; p3/p2
    SBUTTERFLY       dq, 2, 6, 1     ; q0/q1
    SBUTTERFLY       dq, 3, 7, 1     ; q2/q3
    mova             m1, m_q0backup
    mova     m_q0backup, m2          ; store q0
    SBUTTERFLY       dq, 1, 5, 2     ; p1/p0
    mova     m_p0backup, m5          ; store p0
    SWAP              1, 4
    SWAP              2, 4
    SWAP              6, 3
    SWAP              5, 3
%else ; sse2 (h)
%if %2 == 16
    lea           dst8q, [dst1q+ strideq*8  ]
%endif

    ; read 16 rows of 8px each, interleave
    movh             m0, [dst1q+mstrideq*4]
    movh             m1, [dst8q+mstrideq*4]
    movh             m2, [dst1q+mstrideq*2]
    movh             m5, [dst8q+mstrideq*2]
    movh             m3, [dst1q+mstrideq  ]
    movh             m6, [dst8q+mstrideq  ]
    movh             m4, [dst1q]
    movh             m7, [dst8q]
    punpcklbw        m0, m1          ; A/I
    punpcklbw        m2, m5          ; C/K
    punpcklbw        m3, m6          ; D/L
    punpcklbw        m4, m7          ; E/M

    add           dst8q, strideq
    movh             m1, [dst2q+mstrideq*4]
    movh             m6, [dst8q+mstrideq*4]
    movh             m5, [dst2q]
    movh             m7, [dst8q]
    punpcklbw        m1, m6          ; B/J
    punpcklbw        m5, m7          ; F/N
    movh             m6, [dst2q+ strideq  ]
    movh             m7, [dst8q+ strideq  ]
    punpcklbw        m6, m7          ; G/O

    ; 8x16 transpose
    TRANSPOSE4x4B     0, 1, 2, 3, 7
%ifdef m8
    SWAP              1, 8
%else
    mova     m_q0backup, m1
%endif
    movh             m7, [dst2q+ strideq*2]
    movh             m1, [dst8q+ strideq*2]
    punpcklbw        m7, m1          ; H/P
    TRANSPOSE4x4B     4, 5, 6, 7, 1
    SBUTTERFLY       dq, 0, 4, 1     ; p3/p2
    SBUTTERFLY       dq, 2, 6, 1     ; q0/q1
    SBUTTERFLY       dq, 3, 7, 1     ; q2/q3
%ifdef m8
    SWAP              1, 8
    SWAP              2, 8
%else
    mova             m1, m_q0backup
    mova     m_q0backup, m2          ; store q0
%endif
    SBUTTERFLY       dq, 1, 5, 2     ; p1/p0
%ifdef m12
    SWAP              5, 12
%else
    mova     m_p0backup, m5          ; store p0
%endif
    SWAP              1, 4
    SWAP              2, 4
    SWAP              6, 3
    SWAP              5, 3
%endif

    ; normal_limit for p3-p2, p2-p1, q3-q2 and q2-q1
    mova             m4, m1
    SWAP              4, 1
    psubusb          m4, m0          ; p2-p3
    psubusb          m0, m1          ; p3-p2
    por              m0, m4          ; abs(p3-p2)

    mova             m4, m2
    SWAP              4, 2
    psubusb          m4, m1          ; p1-p2
    mova     m_p2backup, m1
    psubusb          m1, m2          ; p2-p1
    por              m1, m4          ; abs(p2-p1)

    mova             m4, m6
    SWAP              4, 6
    psubusb          m4, m7          ; q2-q3
    psubusb          m7, m6          ; q3-q2
    por              m7, m4          ; abs(q3-q2)

    mova             m4, m5
    SWAP              4, 5
    psubusb          m4, m6          ; q1-q2
    mova     m_q2backup, m6
    psubusb          m6, m5          ; q2-q1
    por              m6, m4          ; abs(q2-q1)

%if notcpuflag(mmx2)
    mova             m4, m_flimI
    pxor             m3, m3
    psubusb          m0, m4
    psubusb          m1, m4
    psubusb          m7, m4
    psubusb          m6, m4
    pcmpeqb          m0, m3          ; abs(p3-p2) <= I
    pcmpeqb          m1, m3          ; abs(p2-p1) <= I
    pcmpeqb          m7, m3          ; abs(q3-q2) <= I
    pcmpeqb          m6, m3          ; abs(q2-q1) <= I
    pand             m0, m1
    pand             m7, m6
    pand             m0, m7
%else ; mmxext/sse2
    pmaxub           m0, m1
    pmaxub           m6, m7
    pmaxub           m0, m6
%endif

    ; normal_limit and high_edge_variance for p1-p0, q1-q0
    SWAP              7, 3           ; now m7 is zero
%ifidn %1, v
    movrow           m3, [dst1q+mstrideq  ] ; p0
%if mmsize == 16 && %2 == 8
    movhps           m3, [dst8q+mstrideq  ]
%endif
%elifdef m12
    SWAP              3, 12
%else
    mova             m3, m_p0backup
%endif

    mova             m1, m2
    SWAP              1, 2
    mova             m6, m3
    SWAP              3, 6
    psubusb          m1, m3          ; p1-p0
    psubusb          m6, m2          ; p0-p1
    por              m1, m6          ; abs(p1-p0)
%if notcpuflag(mmx2)
    mova             m6, m1
    psubusb          m1, m4
    psubusb          m6, m_hevthr
    pcmpeqb          m1, m7          ; abs(p1-p0) <= I
    pcmpeqb          m6, m7          ; abs(p1-p0) <= hev_thresh
    pand             m0, m1
    mova      m_maskres, m6
%else ; mmxext/sse2
    pmaxub           m0, m1          ; max_I
    SWAP              1, 4           ; max_hev_thresh
%endif

    SWAP              6, 4           ; now m6 is I
%ifidn %1, v
    movrow           m4, [dst1q]     ; q0
%if mmsize == 16 && %2 == 8
    movhps           m4, [dst8q]
%endif
%elifdef m8
    SWAP              4, 8
%else
    mova             m4, m_q0backup
%endif
    mova             m1, m4
    SWAP              1, 4
    mova             m7, m5
    SWAP              7, 5
    psubusb          m1, m5          ; q0-q1
    psubusb          m7, m4          ; q1-q0
    por              m1, m7          ; abs(q1-q0)
%if notcpuflag(mmx2)
    mova             m7, m1
    psubusb          m1, m6
    psubusb          m7, m_hevthr
    pxor             m6, m6
    pcmpeqb          m1, m6          ; abs(q1-q0) <= I
    pcmpeqb          m7, m6          ; abs(q1-q0) <= hev_thresh
    mova             m6, m_maskres
    pand             m0, m1          ; abs([pq][321]-[pq][210]) <= I
    pand             m6, m7
%else ; mmxext/sse2
    pxor             m7, m7
    pmaxub           m0, m1
    pmaxub           m6, m1
    psubusb          m0, m_flimI
    psubusb          m6, m_hevthr
    pcmpeqb          m0, m7          ; max(abs(..)) <= I
    pcmpeqb          m6, m7          ; !(max(abs..) > thresh)
%endif
%ifdef m12
    SWAP              6, 12
%else
    mova      m_maskres, m6          ; !(abs(p1-p0) > hev_t || abs(q1-q0) > hev_t)
%endif

    ; simple_limit
    mova             m1, m3
    SWAP              1, 3
    mova             m6, m4          ; keep copies of p0/q0 around for later use
    SWAP              6, 4
    psubusb          m1, m4          ; p0-q0
    psubusb          m6, m3          ; q0-p0
    por              m1, m6          ; abs(q0-p0)
    paddusb          m1, m1          ; m1=2*abs(q0-p0)

    mova             m7, m2
    SWAP              7, 2
    mova             m6, m5
    SWAP              6, 5
    psubusb          m7, m5          ; p1-q1
    psubusb          m6, m2          ; q1-p1
    por              m7, m6          ; abs(q1-p1)
    pxor             m6, m6
    pand             m7, [pb_FE]
    psrlq            m7, 1           ; abs(q1-p1)/2
    paddusb          m7, m1          ; abs(q0-p0)*2+abs(q1-p1)/2
    psubusb          m7, m_flimE
    pcmpeqb          m7, m6          ; abs(q0-p0)*2+abs(q1-p1)/2 <= E
    pand             m0, m7          ; normal_limit result

    ; filter_common; at this point, m2-m5=p1-q1 and m0 is filter_mask
%ifdef m8 ; x86-64 && sse2
    mova             m8, [pb_80]
%define m_pb_80 m8
%else ; x86-32 or mmx/mmxext
%define m_pb_80 [pb_80]
%endif
    mova             m1, m4
    mova             m7, m3
    pxor             m1, m_pb_80
    pxor             m7, m_pb_80
    psubsb           m1, m7          ; (signed) q0-p0
    mova             m6, m2
    mova             m7, m5
    pxor             m6, m_pb_80
    pxor             m7, m_pb_80
    psubsb           m6, m7          ; (signed) p1-q1
    mova             m7, m_maskres
    paddsb           m6, m1
    paddsb           m6, m1
    paddsb           m6, m1
    pand             m6, m0
%ifdef m8
    mova       m_limres, m6          ; 3*(qp-p0)+(p1-q1) masked for filter_mbedge
    pand       m_limres, m7
%else
    mova             m0, m6
    pand             m0, m7
    mova       m_limres, m0
%endif
    pandn            m7, m6          ; 3*(q0-p0)+(p1-q1) masked for filter_common

    mova             m1, [pb_F8]
    mova             m6, m7
    paddsb           m7, [pb_3]
    paddsb           m6, [pb_4]
    pand             m7, m1
    pand             m6, m1

    pxor             m1, m1
    pxor             m0, m0
    pcmpgtb          m1, m7
    psubb            m0, m7
    psrlq            m7, 3           ; +f2
    psrlq            m0, 3           ; -f2
    pand             m0, m1
    pandn            m1, m7
    psubusb          m3, m0
    paddusb          m3, m1          ; p0+f2

    pxor             m1, m1
    pxor             m0, m0
    pcmpgtb          m0, m6
    psubb            m1, m6
    psrlq            m6, 3           ; +f1
    psrlq            m1, 3           ; -f1
    pand             m1, m0
    pandn            m0, m6
    psubusb          m4, m0
    paddusb          m4, m1          ; q0-f1

    ; filter_mbedge (m2-m5 = p1-q1; lim_res carries w)
%if cpuflag(ssse3)
    mova             m7, [pb_1]
%else
    mova             m7, [pw_63]
%endif
%ifdef m8
    SWAP              1, 8
%else
    mova             m1, m_limres
%endif
    pxor             m0, m0
    mova             m6, m1
    pcmpgtb          m0, m1         ; which are negative
%if cpuflag(ssse3)
    punpcklbw        m6, m7         ; interleave with "1" for rounding
    punpckhbw        m1, m7
%else
    punpcklbw        m6, m0         ; signed byte->word
    punpckhbw        m1, m0
%endif
    mova      m_limsign, m0
%if cpuflag(ssse3)
    mova             m7, [pb_27_63]
%ifndef m8
    mova       m_limres, m1
%endif
%ifdef m10
    SWAP              0, 10         ; don't lose lim_sign copy
%endif
    mova             m0, m7
    pmaddubsw        m7, m6
    SWAP              6, 7
    pmaddubsw        m0, m1
    SWAP              1, 0
%ifdef m10
    SWAP              0, 10
%else
    mova             m0, m_limsign
%endif
%else
    mova      m_maskres, m6         ; backup for later in filter
    mova       m_limres, m1
    pmullw          m6, [pw_27]
    pmullw          m1, [pw_27]
    paddw           m6, m7
    paddw           m1, m7
%endif
    psraw           m6, 7
    psraw           m1, 7
    packsswb        m6, m1          ; a0
    pxor            m1, m1
    psubb           m1, m6
    pand            m1, m0          ; -a0
    pandn           m0, m6          ; +a0
%if cpuflag(ssse3)
    mova            m6, [pb_18_63]  ; pipelining
%endif
    psubusb         m3, m1
    paddusb         m4, m1
    paddusb         m3, m0          ; p0+a0
    psubusb         m4, m0          ; q0-a0

%if cpuflag(ssse3)
    SWAP             6, 7
%ifdef m10
    SWAP             1, 10
%else
    mova            m1, m_limres
%endif
    mova            m0, m7
    pmaddubsw       m7, m6
    SWAP             6, 7
    pmaddubsw       m0, m1
    SWAP             1, 0
%ifdef m10
    SWAP             0, 10
%endif
    mova            m0, m_limsign
%else
    mova            m6, m_maskres
    mova            m1, m_limres
    pmullw          m6, [pw_18]
    pmullw          m1, [pw_18]
    paddw           m6, m7
    paddw           m1, m7
%endif
    mova            m0, m_limsign
    psraw           m6, 7
    psraw           m1, 7
    packsswb        m6, m1          ; a1
    pxor            m1, m1
    psubb           m1, m6
    pand            m1, m0          ; -a1
    pandn           m0, m6          ; +a1
%if cpuflag(ssse3)
    mova            m6, [pb_9_63]
%endif
    psubusb         m2, m1
    paddusb         m5, m1
    paddusb         m2, m0          ; p1+a1
    psubusb         m5, m0          ; q1-a1

%if cpuflag(ssse3)
    SWAP             6, 7
%ifdef m10
    SWAP             1, 10
%else
    mova            m1, m_limres
%endif
    mova            m0, m7
    pmaddubsw       m7, m6
    SWAP             6, 7
    pmaddubsw       m0, m1
    SWAP             1, 0
%else
%ifdef m8
    SWAP             6, 12
    SWAP             1, 8
%else
    mova            m6, m_maskres
    mova            m1, m_limres
%endif
    pmullw          m6, [pw_9]
    pmullw          m1, [pw_9]
    paddw           m6, m7
    paddw           m1, m7
%endif
%ifdef m9
    SWAP             7, 9
%else
    mova            m7, m_limsign
%endif
    psraw           m6, 7
    psraw           m1, 7
    packsswb        m6, m1          ; a1
    pxor            m0, m0
    psubb           m0, m6
    pand            m0, m7          ; -a1
    pandn           m7, m6          ; +a1
%ifdef m8
    SWAP             1, 13
    SWAP             6, 14
%else
    mova            m1, m_p2backup
    mova            m6, m_q2backup
%endif
    psubusb         m1, m0
    paddusb         m6, m0
    paddusb         m1, m7          ; p1+a1
    psubusb         m6, m7          ; q1-a1

    ; store
%ifidn %1, v
    movrow [dst2q+mstrideq*4], m1
    movrow [dst1q+mstrideq*2], m2
    movrow [dst1q+mstrideq  ], m3
    movrow     [dst1q], m4
    movrow     [dst2q], m5
    movrow [dst2q+ strideq  ], m6
%if mmsize == 16 && %2 == 8
    add           dst8q, mstrideq
    movhps [dst8q+mstrideq*2], m1
    movhps [dst8q+mstrideq  ], m2
    movhps     [dst8q], m3
    add          dst8q, strideq
    movhps     [dst8q], m4
    movhps [dst8q+ strideq  ], m5
    movhps [dst8q+ strideq*2], m6
%endif
%else ; h
    inc          dst1q
    inc          dst2q

    ; 4x8/16 transpose
    TRANSPOSE4x4B    1, 2, 3, 4, 0
    SBUTTERFLY      bw, 5, 6, 0

%if mmsize == 8 ; mmx/mmxext (h)
    WRITE_4x2D       1, 2, 3, 4, dst1q, dst2q, mstrideq, strideq
    add          dst1q, 4
    WRITE_2x4W      m5, m6, dst2q, dst1q, mstrideq, strideq
%else ; sse2 (h)
    lea          dst8q, [dst8q+mstrideq+1]
    WRITE_4x4D       1, 2, 3, 4, dst1q, dst2q, dst8q, mstrideq, strideq, %2
    lea          dst1q, [dst2q+mstrideq+4]
    lea          dst8q, [dst8q+mstrideq+4]
%if cpuflag(sse4)
    add          dst2q, 4
%endif
    WRITE_8W        m5, dst2q, dst1q,  mstrideq, strideq
%if cpuflag(sse4)
    lea          dst2q, [dst8q+ strideq  ]
%endif
    WRITE_8W        m6, dst2q, dst8q, mstrideq, strideq
%endif
%endif

%if mmsize == 8
%if %2 == 8 ; chroma
%ifidn %1, h
    sub          dst1q, 5
%endif
    cmp          dst1q, dst8q
    mov          dst1q, dst8q
    jnz .next8px
%else
%ifidn %1, h
    lea          dst1q, [dst1q+ strideq*8-5]
%else ; v
    add          dst1q, 8
%endif
    dec          cntrq
    jg .next8px
%endif
%endif

%ifndef m8 ; sse2 on x86-32 or mmx/mmxext
    ADD            rsp, pad
%endif
    RET
%endmacro

%if ARCH_X86_32
INIT_MMX mmx
MBEDGE_LOOPFILTER v, 16
MBEDGE_LOOPFILTER h, 16
MBEDGE_LOOPFILTER v,  8
MBEDGE_LOOPFILTER h,  8

INIT_MMX mmx2
MBEDGE_LOOPFILTER v, 16
MBEDGE_LOOPFILTER h, 16
MBEDGE_LOOPFILTER v,  8
MBEDGE_LOOPFILTER h,  8
%endif

INIT_XMM sse2
MBEDGE_LOOPFILTER v, 16
MBEDGE_LOOPFILTER h, 16
MBEDGE_LOOPFILTER v,  8
MBEDGE_LOOPFILTER h,  8

INIT_XMM ssse3
MBEDGE_LOOPFILTER v, 16
MBEDGE_LOOPFILTER h, 16
MBEDGE_LOOPFILTER v,  8
MBEDGE_LOOPFILTER h,  8

INIT_XMM sse4
MBEDGE_LOOPFILTER h, 16
MBEDGE_LOOPFILTER h,  8