aboutsummaryrefslogtreecommitdiffstats
path: root/libavcodec/arm/vp8dsp_armv6.S
blob: 4e7b78361e33f71a0d86493a241806d451a150cc (plain) (blame)
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
/**
 * VP8 ARMv6 optimisations
 *
 * Copyright (c) 2011 The WebM project authors. All Rights Reserved.
 * Copyright (c) 2010 Rob Clark <rob@ti.com>
 * Copyright (c) 2011 Mans Rullgard <mans@mansr.com>
 *
 * This file is part of Libav.
 *
 * Libav 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.
 *
 * Libav 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 Libav; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 *
 * This code was partially ported from libvpx, which uses this license:
 *
 * Use of this source code is governed by a BSD-style license
 * that can be found in the LICENSE file in the root of the source
 * tree. An additional intellectual property rights grant can be found
 * in the file PATENTS.  All contributing project authors may
 * be found in the AUTHORS file in the root of the source tree.
 *
 * (Note that the "LICENSE", "AUTHORS" and "PATENTS" files can be
 *  found in the libvpx source tree.)
 */

#include "asm.S"

@ idct

@ void vp8_luma_dc_wht(DCTELEM block[4][4][16], DCTELEM dc[16])
function ff_vp8_luma_dc_wht_armv6, export=1
        push           {r4 - r10, lr}

        @ load dc[] and zero memory
        mov             r12, #0
        ldr             r2, [r1]                @ dc0[0,1]
        ldr             r3, [r1,  #4]           @ dc0[2,3]
        ldr             r4, [r1,  #8]           @ dc1[0,1]
        ldr             r5, [r1,  #12]          @ dc1[2,3]
        ldr             r6, [r1,  #16]          @ dc2[0,1]
        ldr             r7, [r1,  #20]          @ dc2[2,3]
        ldr             r8, [r1,  #24]          @ dc3[0,1]
        ldr             r9, [r1,  #28]          @ dc3[2,3]
        str             r12,[r1]
        str             r12,[r1,  #4]
        str             r12,[r1,  #8]
        str             r12,[r1,  #12]
        str             r12,[r1,  #16]
        str             r12,[r1,  #20]
        str             r12,[r1,  #24]
        str             r12,[r1,  #28]

        @ loop1
        uadd16          r12, r2,  r8            @ t0[0,1]
        uadd16          r14, r3,  r9            @ t0[2,3]
        usub16          r2,  r2,  r8            @ t3[0,1]
        usub16          r3,  r3,  r9            @ t3[2,3]
        uadd16          r8,  r4,  r6            @ t1[0,1]
        uadd16          r9,  r5,  r7            @ t1[2,3]
        usub16          r4,  r4,  r6            @ t2[0,1]
        usub16          r5,  r5,  r7            @ t2[2,3]

        uadd16          r6,  r12, r8            @ dc0[0,1]
        uadd16          r7,  r14, r9            @ dc0[2,3]
        usub16          r12, r12, r8            @ dc2[0,1]
        usub16          r14, r14, r9            @ dc2[2,3]
        uadd16          r8,  r2,  r4            @ dc1[0,1]
        uadd16          r9,  r3,  r5            @ dc1[2,3]
        usub16          r2,  r2,  r4            @ dc3[0,1]
        usub16          r3,  r3,  r5            @ dc3[2,3]

        mov             r1,  #3
        orr             r1,  r1,  #0x30000      @ 3 | 3 (round)

        @ "transpose"
        pkhbt           r4,  r6,  r8,  lsl #16  @ dc{0,1}[0]
        pkhtb           r6,  r8,  r6,  asr #16  @ dc{0,1}[1]
        pkhbt           r5,  r12, r2,  lsl #16  @ dc{2,3}[0]
        pkhtb           r12, r2,  r12, asr #16  @ dc{2,3}[1]
        pkhbt           r8,  r7,  r9,  lsl #16  @ dc{0,1}[2]
        uadd16          r4,  r4,  r1
        uadd16          r5,  r5,  r1
        pkhtb           r7,  r9,  r7,  asr #16  @ dc{0,1}[3]
        pkhbt           r2,  r14, r3,  lsl #16  @ dc{2,3}[2]
        pkhtb           r14, r3,  r14, asr #16  @ dc{2,3}[3]

        @ loop2
        uadd16          r9,  r4,  r7            @ t0[0,1]
        uadd16          r3,  r5,  r14           @ t0[2,3]
        usub16          r4,  r4,  r7            @ t3[0,1]
        usub16          r5,  r5,  r14           @ t3[2,3]
        uadd16          r7,  r6,  r8            @ t1[0,1]
        uadd16          r14, r12, r2            @ t1[2,3]
        usub16          r6,  r6,  r8            @ t2[0,1]
        usub16          r12, r12, r2            @ t2[2,3]

        uadd16          r8,  r9,  r7            @ block[0,1][0]
        uadd16          r2,  r3,  r14           @ block[2,3][0]
        usub16          r9,  r9,  r7            @ block[0,1][2]
        usub16          r3,  r3,  r14           @ block[2,3][2]
        uadd16          r7,  r4,  r6            @ block[0,1][1]
        uadd16          r14, r5,  r12           @ block[2,3][1]
        usub16          r4,  r4,  r6            @ block[0,1][3]
        usub16          r5,  r5,  r12           @ block[2,3][3]

        @ store
        mov             r6,  r8,  asr #19       @ block[1][0]
        mov             r12, r7,  asr #19       @ block[1][1]
        mov             r1,  r9,  asr #19       @ block[1][2]
        mov             r10, r4,  asr #19       @ block[1][3]
        sxth            r8,  r8
        sxth            r7,  r7
        sxth            r9,  r9
        sxth            r4,  r4
        asr             r8,  #3                 @ block[0][0]
        asr             r7,  #3                 @ block[0][1]
        asr             r9,  #3                 @ block[0][2]
        asr             r4,  #3                 @ block[0][3]

        strh            r8, [r0], #32
        strh            r7, [r0], #32
        strh            r9, [r0], #32
        strh            r4, [r0], #32
        strh            r6, [r0], #32
        strh            r12,[r0], #32
        strh            r1, [r0], #32
        strh            r10,[r0], #32

        mov             r6,  r2,  asr #19       @ block[3][0]
        mov             r12, r14, asr #19       @ block[3][1]
        mov             r1,  r3,  asr #19       @ block[3][2]
        mov             r10, r5,  asr #19       @ block[3][3]
        sxth            r2,  r2
        sxth            r14, r14
        sxth            r3,  r3
        sxth            r5,  r5
        asr             r2,  #3                 @ block[2][0]
        asr             r14, #3                 @ block[2][1]
        asr             r3,  #3                 @ block[2][2]
        asr             r5,  #3                 @ block[2][3]

        strh            r2, [r0], #32
        strh            r14,[r0], #32
        strh            r3, [r0], #32
        strh            r5, [r0], #32
        strh            r6, [r0], #32
        strh            r12,[r0], #32
        strh            r1, [r0], #32
        strh            r10,[r0], #32

        pop            {r4 - r10, pc}
endfunc

@ void vp8_luma_dc_wht_dc(DCTELEM block[4][4][16], DCTELEM dc[16])
function ff_vp8_luma_dc_wht_dc_armv6, export=1
        ldrsh           r2, [r1]
        mov             r3,  #0
        add             r2,  r2,  #3
        strh            r3, [r1]
        asr             r2,  r2,  #3
    .rept 16
        strh            r2, [r0], #32
    .endr
        bx              lr
endfunc

@ void vp8_idct_add(uint8_t *dst, DCTELEM block[16], int stride)
function ff_vp8_idct_add_armv6, export=1
        push           {r4 - r11, lr}
        sub             sp,  sp,  #32

        mov             r3,  #0x00004E00        @ cos
        orr             r3,  r3, #0x0000007B    @ cospi8sqrt2minus1 = 20091
        mov             r4,  #0x00008A00        @ sin
        orr             r4,  r4, #0x0000008C    @ sinpi8sqrt2 = 35468
        mov             r5,  #0x2               @ i=2
1:
        ldr             r6, [r1, #8]            @  i5 | i4  = block1[1] | block1[0]
        ldr             r12,[r1, #24]           @ i13 | i12 = block3[1] | block3[0]
        ldr             r14,[r1, #16]           @  i9 | i8  = block2[1] | block2[0]

        smulwt          r9,  r3,  r6            @ (ip[5] * cospi8sqrt2minus1) >> 16
        smulwb          r7,  r3,  r6            @ (ip[4] * cospi8sqrt2minus1) >> 16
        smulwt          r10, r4,  r6            @ (ip[5] * sinpi8sqrt2) >> 16
        smulwb          r8,  r4,  r6            @ (ip[4] * sinpi8sqrt2) >> 16
        pkhbt           r7,  r7,  r9,  lsl #16  @ 5c | 4c
        smulwt          r11, r3,  r12           @ (ip[13] * cospi8sqrt2minus1) >> 16
        pkhbt           r8,  r8,  r10, lsl #16  @ 5s | 4s         = t2 first half
        uadd16          r6,  r6,  r7            @ 5c+5 | 4c+4     = t3 first half
        smulwt          r7,  r4,  r12           @ (ip[13] * sinpi8sqrt2) >> 16
        smulwb          r9,  r3,  r12           @ (ip[12] * cospi8sqrt2minus1) >> 16
        smulwb          r10, r4,  r12           @ (ip[12] * sinpi8sqrt2) >> 16

        subs            r5,  r5,  #1            @ i--
        pkhbt           r9,  r9,  r11, lsl #16  @ 13c | 12c
        ldr             r11,[r1]                @  i1 | i0
        pkhbt           r10, r10, r7,  lsl #16  @ 13s | 12s       = t3 second half
        uadd16          r7,  r12, r9            @ 13c+13 | 12c+12 = t2 second half
        usub16          r7,  r8,  r7            @ c = t2
        uadd16          r6,  r6,  r10           @ d = t3
        uadd16          r10, r11, r14           @ a = t0
        usub16          r8,  r11, r14           @ b = t1
        uadd16          r9,  r10, r6            @ a+d = tmp{0,1}[0]
        usub16          r10, r10, r6            @ a-d = tmp{0,1}[3]
        uadd16          r6,  r8,  r7            @ b+c = tmp{0,1}[1]
        usub16          r7,  r8,  r7            @ b-c = tmp{0,1}[2]
        mov             r8,  #0
        str             r6, [sp,  #8]           @  o5 | o4
        str             r7, [sp,  #16]          @  o9 | o8
        str             r10,[sp,  #24]          @ o13 | o12
        str             r9, [sp], #4            @  o1 | o0
        str             r8, [r1,  #24]
        str             r8, [r1,  #16]
        str             r8, [r1,  #8]
        str             r8, [r1], #4
        bne             1b

        mov             r5,  #0x2               @ i=2
        sub             sp,  sp, #8
2:
        ldr             r6, [sp,  #8]           @ i5 | i4 = tmp{0,1}[1]
        ldr             r14,[sp,  #4]           @ i3 | i2 = tmp{2,3}[0]
        ldr             r12,[sp,  #12]          @ i7 | i6 = tmp{2,3}[1]
        ldr             r1, [sp], #16           @ i1 | i0 = tmp{0,1}[0]
        smulwt          r9,  r3,  r6            @ (ip[5] * cospi8sqrt2minus1) >> 16
        smulwt          r7,  r3,  r1            @ (ip[1] * cospi8sqrt2minus1) >> 16
        smulwt          r10, r4,  r6            @ (ip[5] * sinpi8sqrt2) >> 16
        smulwt          r8,  r4,  r1            @ (ip[1] * sinpi8sqrt2) >> 16
        pkhbt           r11, r1,  r6,  lsl #16  @ i4 | i0 = t0/t1 first half
        pkhbt           r7,  r7,  r9,  lsl #16  @ 5c | 1c
        pkhbt           r8,  r8,  r10, lsl #16  @ 5s | 1s = temp1 = t2 first half
        pkhtb           r1,  r6,  r1,  asr #16  @ i5 | i1
        uadd16          r1,  r7,  r1            @ 5c+5 | 1c+1 = temp2 (d) = t3 first half
        pkhbt           r9,  r14, r12, lsl #16  @ i6 | i2 = t0/t1 second half
        uadd16          r10, r11, r9            @ a = t0
        usub16          r9,  r11, r9            @ b = t1
        pkhtb           r6,  r12, r14, asr #16  @ i7 | i3
        subs            r5,  r5,  #0x1          @ i--
        smulwt          r7,  r3,  r6            @ (ip[7] * cospi8sqrt2minus1) >> 16
        smulwt          r11, r4,  r6            @ (ip[7] * sinpi8sqrt2) >> 16
        smulwb          r12, r3,  r6            @ (ip[3] * cospi8sqrt2minus1) >> 16
        smulwb          r14, r4,  r6            @ (ip[3] * sinpi8sqrt2) >> 16

        pkhbt           r7,  r12, r7,  lsl #16  @ 7c | 3c
        pkhbt           r11, r14, r11, lsl #16  @ 7s | 3s = temp1 (d) = t3 second half
        mov             r14, #0x4               @ set up 4's
        orr             r14, r14, #0x40000      @ 4|4
        uadd16          r6,  r7,  r6            @ 7c+7 | 3c+3 = temp2 (c) = t2 second half
        usub16          r12, r8,  r6            @ c (o5 | o1) = t2
        uadd16          r6,  r11, r1            @ d (o7 | o3) = t3
        uadd16          r10, r10, r14           @ t0 + 4
        uadd16          r9,  r9,  r14           @ t1 + 4
        uadd16          r7,  r10, r6            @ a+d = dst{0,1}[0]
        usub16          r6,  r10, r6            @ a-d = dst{0,1}[3]
        uadd16          r10, r9,  r12           @ b+c = dst{0,1}[1]
        usub16          r1,  r9,  r12           @ b-c = dst{0,1}[2]

        mov             r9,  r6,  asr #3        @ o[1][3]
        mov             r12, r1,  asr #3        @ o[1][2]
        pkhtb           r8,  r12, r7,  asr #19  @ o[1][0,2]
        pkhtb           r11, r9,  r10, asr #19  @ o[1][1,3]
        ldr             r12,[r0]
        ldr             r9, [r0,  r2]
        sxth            r7,  r7
        sxth            r6,  r6
        sxth            r10, r10
        sxth            r1,  r1
        asr             r7,  #3                 @ o[0][0]
        asr             r10, #3                 @ o[0][1]
        pkhbt           r7,  r7,  r1,  lsl #13  @ o[0][0,2]
        pkhbt           r10, r10, r6,  lsl #13  @ o[0][1,3]

        uxtab16         r7,  r7,  r12
        uxtab16         r10, r10, r12, ror #8
        uxtab16         r8,  r8,  r9
        uxtab16         r11, r11, r9,  ror #8
        usat16          r7,  #8,  r7
        usat16          r10, #8,  r10
        usat16          r8,  #8,  r8
        usat16          r11, #8,  r11
        orr             r7,  r7,  r10, lsl #8
        orr             r8,  r8,  r11, lsl #8
        str             r8, [r0,  r2]
        str_post        r7,  r0,  r2,  lsl #1

        bne             2b

        pop            {r4 - r11, pc}
endfunc

@ void vp8_idct_dc_add(uint8_t *dst, DCTELEM block[16], int stride)
function ff_vp8_idct_dc_add_armv6, export=1
        push           {r4 - r5,  lr}
        ldrsh           r3, [r1]
        mov             r4,  #0
        add             r3,  r3,  #4
        asr             r3,  #3
        strh            r4, [r1], #32
        ldr             r4, [r0,  r2]
        ldr_post        r5,  r0,  r2,  lsl #1
        pkhbt           r3,  r3,  r3,  lsl #16

        uxtab16         lr,  r3,  r5            @ a1+2 | a1+0
        uxtab16         r5,  r3,  r5,  ror #8   @ a1+3 | a1+1
        uxtab16         r12, r3,  r4
        uxtab16         r4,  r3,  r4,  ror #8
        usat16          lr,  #8,  lr
        usat16          r5,  #8,  r5
        usat16          r12, #8,  r12
        usat16          r4,  #8,  r4
        orr             lr,  lr,  r5,  lsl #8
        orr             r12, r12, r4,  lsl #8
        ldr             r5, [r0]
        ldr             r4, [r0,  r2]
        sub             r0,  r0,  r2,  lsl #1
        str             r12,[r0,  r2]
        str_post        lr,  r0,  r2,  lsl #1

        uxtab16         lr,  r3,  r5
        uxtab16         r5,  r3,  r5,  ror #8
        uxtab16         r12, r3,  r4
        uxtab16         r4,  r3,  r4,  ror #8
        usat16          lr,  #8,  lr
        usat16          r5,  #8,  r5
        usat16          r12, #8,  r12
        usat16          r4,  #8,  r4
        orr             lr,  lr,  r5,  lsl #8
        orr             r12, r12, r4,  lsl #8

        str             r12,[r0,  r2]
        str_post        lr,  r0,  r2,  lsl #1

        pop            {r4 - r5,  pc}
endfunc

@ void vp8_idct_dc_add4uv(uint8_t *dst, DCTELEM block[4][16], int stride)
function ff_vp8_idct_dc_add4uv_armv6, export=1
        push           {lr}

        bl              ff_vp8_idct_dc_add_armv6
        sub             r0,  r0,  r2,  lsl #2
        add             r0,  r0,  #4
        bl              ff_vp8_idct_dc_add_armv6
        sub             r0,  r0,  #4
        bl              ff_vp8_idct_dc_add_armv6
        sub             r0,  r0,  r2,  lsl #2
        add             r0,  r0,  #4
        bl              ff_vp8_idct_dc_add_armv6

        pop            {pc}
endfunc

@ void vp8_idct_dc_add4y(uint8_t *dst, DCTELEM block[4][16], int stride)
function ff_vp8_idct_dc_add4y_armv6, export=1
        push           {lr}

        bl              ff_vp8_idct_dc_add_armv6
        sub             r0,  r0,  r2,  lsl #2
        add             r0,  r0,  #4
        bl              ff_vp8_idct_dc_add_armv6
        sub             r0,  r0,  r2,  lsl #2
        add             r0,  r0,  #4
        bl              ff_vp8_idct_dc_add_armv6
        sub             r0,  r0,  r2,  lsl #2
        add             r0,  r0,  #4
        bl              ff_vp8_idct_dc_add_armv6

        pop            {pc}
endfunc

@ loopfilter

@ void vp8_v_loop_filter16_simple(uint8_t *dst, int stride, int flim)
function ff_vp8_v_loop_filter16_simple_armv6, export=1
        push           {r4 - r11, lr}

        ldr_dpren       r3,  r0,  r1,  lsl #1   @ p1
        ldr_dpren       r4,  r0,  r1            @ p0
        ldr             r5, [r0]                @ q0
        ldr             r6, [r0,  r1]           @ q1
        orr             r2,  r2,  r2,  lsl #16
        mov             r9,  #4                 @ count
        mov             lr,  #0                 @ need 0 in a couple places
        orr             r12, r2,  r2,  lsl #8   @ splat int -> byte
        ldr             r2,  c0x80808080

1:
        @ vp8_simple_filter_mask()
        uqsub8          r7,  r3,  r6            @ p1 - q1
        uqsub8          r8,  r6,  r3            @ q1 - p1
        uqsub8          r10, r4,  r5            @ p0 - q0
        uqsub8          r11, r5,  r4            @ q0 - p0
        orr             r8,  r8,  r7            @ abs(p1 - q1)
        orr             r10, r10, r11           @ abs(p0 - q0)
        uqadd8          r10, r10, r10           @ abs(p0 - q0) * 2
        uhadd8          r8,  r8,  lr            @ abs(p1 - q2) >> 1
        uqadd8          r10, r10, r8            @ abs(p0 - q0)*2 + abs(p1 - q1)/2
        mvn             r8,  #0
        usub8           r10, r12, r10           @ compare to flimit. usub8 sets GE flags
        sel             r10, r8,  lr            @ filter mask: F or 0
        cmp             r10, #0
        beq             2f                      @ skip filtering if all masks are 0x00

        @ vp8_simple_filter()
        eor             r3,  r3,  r2            @ p1 offset to convert to a signed value
        eor             r6,  r6,  r2            @ q1 offset to convert to a signed value
        eor             r4,  r4,  r2            @ p0 offset to convert to a signed value
        eor             r5,  r5,  r2            @ q0 offset to convert to a signed value

        qsub8           r3,  r3,  r6            @ vp8_filter = p1 - q1
        qsub8           r6,  r5,  r4            @ q0 - p0
        qadd8           r3,  r3,  r6            @ += q0 - p0
        ldr             r7,  c0x04040404
        qadd8           r3,  r3,  r6            @ += q0 - p0
        ldr             r8,  c0x03030303
        qadd8           r3,  r3,  r6            @ vp8_filter = p1-q1 + 3*(q0-p0))
        @STALL
        and             r3,  r3,  r10           @ vp8_filter &= mask

        qadd8           r7,  r3,  r7            @ Filter1 = vp8_filter + 4
        qadd8           r8,  r3,  r8            @ Filter2 = vp8_filter + 3

        shadd8          r7,  r7,  lr
        shadd8          r8,  r8,  lr
        shadd8          r7,  r7,  lr
        shadd8          r8,  r8,  lr
        shadd8          r7,  r7,  lr            @ Filter1 >>= 3
        shadd8          r8,  r8,  lr            @ Filter2 >>= 3

        qsub8           r5,  r5,  r7            @ u = q0 - Filter1
        qadd8           r4,  r4,  r8            @ u = p0 + Filter2
        eor             r5,  r5,  r2            @ *oq0 = u^0x80
        eor             r4,  r4,  r2            @ *op0 = u^0x80
T       sub             r7,  r0,  r1
        str             r5, [r0]                @ store oq0 result
A       str             r4, [r0, -r1]           @ store op0 result
T       str             r4, [r7]

2:
        subs            r9,  r9,  #1            @ counter--
        add             r0,  r0,  #4            @ next row
T       itttt           ne
A       ldrne           r3, [r0, -r1,  lsl #1]  @ p1
T       subne           r3,  r0,  r1,  lsl #1
T       ldrne           r3, [r3]                @ p1
A       ldrne           r4, [r0, -r1]           @ p0
T       subne           r4,  r0,  r1
T       ldrne           r4, [r4]                @ p0
T       itt             ne
        ldrne           r5, [r0]                @ q0
        ldrne           r6, [r0,  r1]           @ q1

        bne             1b

        pop            {r4 - r11, pc}
endfunc

c0x01010101: .long 0x01010101
c0x03030303: .long 0x03030303
c0x04040404: .long 0x04040404
c0x7F7F7F7F: .long 0x7F7F7F7F
c0x80808080: .long 0x80808080

@ void vp8_v_loop_filter16_inner(uint8_t *dst, int stride,
@                                int fE, int fI, int hev_thresh)
@ and
@ void vp8_v_loop_filter8uv_inner(uint8_t *dstU, uint8_t *dstV, int stride,
@                                 int fE, int fI, int hev_thresh)
@ call:
@ void vp8_v_loop_filter_inner(uint8_t *dst, int stride,
@                              int fE, int fI, int hev_thresh, int count)
function ff_vp8_v_loop_filter_inner_armv6, export=1
        push           {r4 - r11, lr}

        sub             r0,  r0,  r1,  lsl #2   @ move r0 pointer down by 4 lines
        ldr             r5, [sp,  #40]          @ counter
        ldr             r6, [sp,  #36]          @ load thresh address
        sub             sp,  sp,  #16           @ create temp buffer

        ldr             r10,[r0,  r1]           @ p2
        ldr_post        r9,  r0,  r1,  lsl #1   @ p3
        ldr             r12,[r0,  r1]           @ p0
        ldr_post        r11, r0,  r1,  lsl #1   @ p1

        orr             r2,  r2,  r2,  lsl #16
        orr             r3,  r3,  r3,  lsl #16
        orr             r6,  r6,  r6,  lsl #16
        orr             r4,  r2,  r2,  lsl #8   @ flimE  splat int -> byte
        orr             r2,  r3,  r3,  lsl #8   @ flimI  splat int -> byte
        orr             r3,  r6,  r6,  lsl #8   @ thresh splat int -> byte

1:
        @ vp8_filter_mask() function
        @ calculate breakout conditions
        uqsub8          r6,  r9,  r10           @ p3 - p2
        uqsub8          r7,  r10, r9            @ p2 - p3
        uqsub8          r8,  r10, r11           @ p2 - p1
        uqsub8          r10, r11, r10           @ p1 - p2

        orr             r6,  r6,  r7            @ abs (p3-p2)
        orr             r8,  r8,  r10           @ abs (p2-p1)
        uqsub8          lr,  r6,  r2            @ compare to limit. lr: vp8_filter_mask
        uqsub8          r8,  r8,  r2            @ compare to limit
        uqsub8          r6,  r11, r12           @ p1 - p0
        orr             lr,  lr,  r8
        uqsub8          r7,  r12, r11           @ p0 - p1
        ldr             r10,[r0,  r1]           @ q1
        ldr_post        r9,  r0,  r1,  lsl #1   @ q0
        orr             r6,  r6,  r7            @ abs (p1-p0)
        uqsub8          r7,  r6,  r2            @ compare to limit
        uqsub8          r8,  r6,  r3            @ compare to thresh  -- save r8 for later
        orr             lr,  lr,  r7

        uqsub8          r6,  r11, r10           @ p1 - q1
        uqsub8          r7,  r10, r11           @ q1 - p1
        uqsub8          r11, r12, r9            @ p0 - q0
        uqsub8          r12, r9,  r12           @ q0 - p0
        orr             r6,  r6,  r7            @ abs (p1-q1)
        ldr             r7,  c0x7F7F7F7F
        orr             r12, r11, r12           @ abs (p0-q0)
        ldr_post        r11, r0,  r1            @ q2
        uqadd8          r12, r12, r12           @ abs (p0-q0) * 2
        and             r6,  r7,  r6,  lsr #1   @ abs (p1-q1) / 2
        uqsub8          r7,  r9,  r10           @ q0 - q1
        uqadd8          r12, r12, r6            @ abs (p0-q0)*2 + abs (p1-q1)/2
        uqsub8          r6,  r10, r9            @ q1 - q0
        uqsub8          r12, r12, r4            @ compare to flimit
        uqsub8          r9,  r11, r10           @ q2 - q1

        orr             lr, lr, r12

        ldr_post        r12, r0,  r1            @ q3
        uqsub8          r10, r10, r11           @ q1 - q2
        orr             r6,  r7,  r6            @ abs (q1-q0)
        orr             r10, r9,  r10           @ abs (q2-q1)
        uqsub8          r7,  r6,  r2            @ compare to limit
        uqsub8          r10, r10, r2            @ compare to limit
        uqsub8          r6,  r6,  r3            @ compare to thresh -- save r6 for later
        orr             lr,  lr,  r7
        orr             lr,  lr,  r10

        uqsub8          r10, r12, r11           @ q3 - q2
        uqsub8          r9,  r11, r12           @ q2 - q3

        mvn             r11, #0                 @ r11 == -1

        orr             r10, r10, r9            @ abs (q3-q2)
        uqsub8          r10, r10, r2            @ compare to limit

        mov             r12, #0
        orr             lr,  lr,  r10
        sub             r0,  r0,  r1,  lsl #2

        usub8           lr,  r12, lr            @ use usub8 instead of ssub8
        sel             lr,  r11, r12           @ filter mask: lr

        cmp             lr,  #0
        beq             2f                      @ skip filtering

        sub             r0,  r0,  r1,  lsl #1   @ move r0 pointer down by 6 lines

        @vp8_hevmask() function
        @calculate high edge variance
        orr             r10, r6,  r8            @ calculate vp8_hevmask

        usub8           r10, r12, r10           @ use usub8 instead of ssub8
        sel             r6,  r12, r11           @ obtain vp8_hevmask: r6

        @vp8_filter() function
        ldr             r8, [r0,  r1]           @ p0
        ldr_post        r7,  r0,  r1,  lsl #1   @ p1
        ldr             r12, c0x80808080
        ldr             r10,[r0,  r1]           @ q1
        ldr_post        r9,  r0,  r1,  lsl #1   @ q0

        eor             r7,  r7,  r12           @ p1 offset to convert to a signed value
        eor             r8,  r8,  r12           @ p0 offset to convert to a signed value
        eor             r9,  r9,  r12           @ q0 offset to convert to a signed value
        eor             r10, r10, r12           @ q1 offset to convert to a signed value

        str             r9, [sp]                @ store qs0 temporarily
        str             r8, [sp,  #4]           @ store ps0 temporarily
        str             r10,[sp,  #8]           @ store qs1 temporarily
        str             r7, [sp,  #12]          @ store ps1 temporarily

        qsub8           r7,  r7,  r10           @ vp8_signed_char_clamp(ps1-qs1)
        qsub8           r8,  r9,  r8            @ vp8_signed_char_clamp(vp8_filter + 3 * ( qs0 - ps0))

        and             r7,  r7,  r6            @ vp8_filter (r7) &= hev

        qadd8           r7,  r7,  r8
        ldr             r9,  c0x03030303        @ r9 = 3 --modified for vp8

        qadd8           r7,  r7,  r8
        ldr             r10, c0x04040404

        qadd8           r7,  r7,  r8
        and             r7,  r7,  lr            @ vp8_filter &= mask@

        qadd8           r8,  r7,  r9            @ Filter2 (r8) = vp8_signed_char_clamp(vp8_filter+3)
        qadd8           r7,  r7,  r10           @ vp8_filter = vp8_signed_char_clamp(vp8_filter+4)

        mov             r9,  #0
        shadd8          r8,  r8,  r9            @ Filter2 >>= 3
        shadd8          r7,  r7,  r9            @ vp8_filter >>= 3
        shadd8          r8,  r8,  r9
        shadd8          r7,  r7,  r9
        shadd8          lr,  r8,  r9            @ lr: Filter2
        shadd8          r7,  r7,  r9            @ r7: filter

        @calculate output

        ldr             r8, [sp]                @ load qs0
        ldr             r9, [sp,  #4]           @ load ps0

        ldr             r10, c0x01010101

        qsub8           r8,  r8,  r7            @ u = vp8_signed_char_clamp(qs0 - vp8_filter)
        qadd8           r9,  r9,  lr            @ u = vp8_signed_char_clamp(ps0 + Filter2)

        mov             lr,  #0
        sadd8           r7,  r7,  r10           @ vp8_filter += 1
        shadd8          r7,  r7,  lr            @ vp8_filter >>= 1

        ldr             r11,[sp,  #12]          @ load ps1
        ldr             r10,[sp,  #8]           @ load qs1

        bic             r7,  r7,  r6            @ vp8_filter &= ~hev
        sub             r0,  r0,  r1,  lsl #2

        qadd8           r11, r11, r7            @ u = vp8_signed_char_clamp(ps1 + vp8_filter)
        qsub8           r10, r10, r7            @ u = vp8_signed_char_clamp(qs1 - vp8_filter)

        eor             r11, r11, r12           @ *op1 = u^0x80
        eor             r9,  r9,  r12           @ *op0 = u^0x80
        eor             r8,  r8,  r12           @ *oq0 = u^0x80
        eor             r10, r10, r12           @ *oq1 = u^0x80
        str             r9, [r0,  r1]           @ store op0 result
        str_post        r11, r0,  r1,  lsl #1   @ store op1
        str             r10,[r0,  r1]           @ store oq1
        str_post        r8,  r0,  r1,  lsl #1   @ store oq0 result

        sub             r0,  r0,  r1,  lsl #1

2:
        add             r0,  r0,  #4
        sub             r0,  r0,  r1,  lsl #2

        subs            r5,  r5,  #1
T       ittt            ne
        ldrne           r10,[r0,  r1]           @ p2
A       ldrne           r9, [r0], r1,  lsl #1   @ p3
T       ldrne           r9, [r0]                @ p3
T       addne           r0,  r0,  r1,  lsl #1
T       ittt            ne
        ldrne           r12,[r0,  r1]           @ p0
A       ldrne           r11,[r0], r1,  lsl #1   @ p1
T       ldrne           r11,[r0]                @ p3
T       addne           r0,  r0,  r1,  lsl #1

        bne             1b

        add             sp,  sp,  #16
        pop            {r4 - r11, pc}
endfunc

@ void vp8_v_loop_filter16(uint8_t *dst, int stride,
@                          int fE, int fI, int hev_thresh)
@ and
@ void vp8_v_loop_filter8uv(uint8_t *dstU, uint8_t *dstV, int stride,
@                           int fE, int fI, int hev_thresh)
@ call:
@ void vp8_v_loop_filter(uint8_t *dst, int stride,
@                        int fE, int fI, int hev_thresh, int count)
function ff_vp8_v_loop_filter_armv6, export=1
        push           {r4 - r11, lr}

        sub             r0,  r0,  r1,  lsl #2   @ move r0 pointer down by 4 lines
        ldr             r5, [sp,  #40]          @ counter
        ldr             r6, [sp,  #36]          @ load thresh address
        sub             sp,  sp,  #16           @ create temp buffer

        ldr             r10,[r0,  r1]           @ p2
        ldr_post        r9,  r0,  r1,  lsl #1   @ p3
        ldr             r12,[r0,  r1]           @ p0
        ldr_post        r11, r0,  r1,  lsl #1   @ p1

        orr             r2,  r2,  r2,  lsl #16
        orr             r3,  r3,  r3,  lsl #16
        orr             r6,  r6,  r6,  lsl #16
        orr             r4,  r2,  r2,  lsl #8   @ flimE  splat int -> byte
        orr             r2,  r3,  r3,  lsl #8   @ flimI  splat int -> byte
        orr             r3,  r6,  r6,  lsl #8   @ thresh splat int -> byte

1:
        @ vp8_filter_mask() function
        @ calculate breakout conditions
        uqsub8          r6,  r9,  r10           @ p3 - p2
        uqsub8          r7,  r10, r9            @ p2 - p3
        uqsub8          r8,  r10, r11           @ p2 - p1
        uqsub8          r10, r11, r10           @ p1 - p2

        orr             r6,  r6,  r7            @ abs (p3-p2)
        orr             r8,  r8,  r10           @ abs (p2-p1)
        uqsub8          lr,  r6,  r2            @ compare to limit. lr: vp8_filter_mask
        uqsub8          r8,  r8,  r2            @ compare to limit

        uqsub8          r6,  r11, r12           @ p1 - p0
        orr             lr,  lr,  r8
        uqsub8          r7,  r12, r11           @ p0 - p1
        ldr             r10,[r0,  r1]           @ q1
        ldr_post        r9,  r0,  r1,  lsl #1   @ q0
        orr             r6,  r6,  r7            @ abs (p1-p0)
        uqsub8          r7,  r6,  r2            @ compare to limit
        uqsub8          r8,  r6,  r3            @ compare to thresh  -- save r8 for later
        orr             lr,  lr,  r7

        uqsub8          r6,  r11, r10           @ p1 - q1
        uqsub8          r7,  r10, r11           @ q1 - p1
        uqsub8          r11, r12, r9            @ p0 - q0
        uqsub8          r12, r9,  r12           @ q0 - p0
        orr             r6,  r6,  r7            @ abs (p1-q1)
        ldr             r7,  c0x7F7F7F7F
        orr             r12, r11, r12           @ abs (p0-q0)
        ldr_post        r11, r0,  r1            @ q2
        uqadd8          r12, r12, r12           @ abs (p0-q0) * 2
        and             r6,  r7,  r6,  lsr #1   @ abs (p1-q1) / 2
        uqsub8          r7,  r9,  r10           @ q0 - q1
        uqadd8          r12, r12, r6            @ abs (p0-q0)*2 + abs (p1-q1)/2
        uqsub8          r6,  r10, r9            @ q1 - q0
        uqsub8          r12, r12, r4            @ compare to flimit
        uqsub8          r9,  r11, r10           @ q2 - q1

        orr             lr,  lr,  r12

        ldr_post        r12, r0,  r1            @ q3

        uqsub8          r10, r10, r11           @ q1 - q2
        orr             r6,  r7,  r6            @ abs (q1-q0)
        orr             r10, r9,  r10           @ abs (q2-q1)
        uqsub8          r7,  r6,  r2            @ compare to limit
        uqsub8          r10, r10, r2            @ compare to limit
        uqsub8          r6,  r6,  r3            @ compare to thresh -- save r6 for later
        orr             lr,  lr,  r7
        orr             lr,  lr,  r10

        uqsub8          r10, r12, r11           @ q3 - q2
        uqsub8          r9,  r11, r12           @ q2 - q3

        mvn             r11, #0                 @ r11 == -1

        orr             r10, r10, r9            @ abs (q3-q2)
        uqsub8          r10, r10, r2            @ compare to limit

        mov             r12, #0

        orr             lr,  lr,  r10

        usub8           lr,  r12, lr            @ use usub8 instead of ssub8
        sel             lr,  r11, r12           @ filter mask: lr

        cmp             lr,  #0
        beq             2f                      @ skip filtering

        @vp8_hevmask() function
        @calculate high edge variance
        sub             r0,  r0,  r1,  lsl #2   @ move r0 pointer down by 6 lines
        sub             r0,  r0,  r1,  lsl #1

        orr             r10, r6,  r8

        usub8           r10, r12, r10
        sel             r6,  r12, r11           @ hev mask: r6

        @vp8_mbfilter() function
        @p2, q2 are only needed at the end. Do not need to load them in now.
        ldr             r8, [r0,  r1]           @ p0
        ldr_post        r7,  r0,  r1,  lsl #1   @ p1
        ldr             r12, c0x80808080
        ldr_post        r9,  r0,  r1            @ q0
        ldr             r10,[r0]                @ q1

        eor             r7,  r7,  r12           @ ps1
        eor             r8,  r8,  r12           @ ps0
        eor             r9,  r9,  r12           @ qs0
        eor             r10, r10, r12           @ qs1

        qsub8           r12, r9,  r8            @ vp8_signed_char_clamp(vp8_filter + 3 * ( qs0 - ps0))
        str             r7, [sp,  #12]          @ store ps1 temporarily
        qsub8           r7,  r7,  r10           @ vp8_signed_char_clamp(ps1-qs1)
        str             r10,[sp,  #8]           @ store qs1 temporarily
        qadd8           r7,  r7,  r12
        str             r9, [sp]                @ store qs0 temporarily
        qadd8           r7,  r7,  r12
        str             r8, [sp,  #4]           @ store ps0 temporarily
        qadd8           r7,  r7,  r12           @ vp8_filter: r7

        ldr             r10, c0x03030303        @ r10 = 3 --modified for vp8
        ldr             r9,  c0x04040404

        and             r7,  r7,  lr            @ vp8_filter &= mask (lr is free)

        mov             r12, r7                 @ Filter2: r12
        and             r12, r12, r6            @ Filter2 &= hev

        @save bottom 3 bits so that we round one side +4 and the other +3
        qadd8           r8,  r12, r9            @ Filter1 (r8) = vp8_signed_char_clamp(Filter2+4)
        qadd8           r12, r12, r10           @ Filter2 (r12) = vp8_signed_char_clamp(Filter2+3)

        mov             r10, #0
        shadd8          r8,  r8,  r10           @ Filter1 >>= 3
        shadd8          r12, r12, r10           @ Filter2 >>= 3
        shadd8          r8,  r8,  r10
        shadd8          r12, r12, r10
        shadd8          r8,  r8,  r10           @ r8: Filter1
        shadd8          r12, r12, r10           @ r12: Filter2

        ldr             r9, [sp]                @ load qs0
        ldr             r11,[sp,  #4]           @ load ps0

        qsub8           r9,  r9,  r8            @ qs0 = vp8_signed_char_clamp(qs0 - Filter1)
        qadd8           r11, r11, r12           @ ps0 = vp8_signed_char_clamp(ps0 + Filter2)

        bic             r12, r7,  r6            @ vp8_filter &= ~hev    ( r6 is free)

        @roughly 3/7th difference across boundary
        mov             lr,  #0x1b              @ 27
        mov             r7,  #0x3f              @ 63

        sxtb16          r6,  r12
        sxtb16          r10, r12, ror #8
        smlabb          r8,  r6,  lr,  r7
        smlatb          r6,  r6,  lr,  r7
        smlabb          r7,  r10, lr,  r7
        smultb          r10, r10, lr
        ssat            r8,  #8,  r8,  asr #7
        ssat            r6,  #8,  r6,  asr #7
        add             r10, r10, #63
        ssat            r7,  #8,  r7,  asr #7
        ssat            r10, #8,  r10, asr #7

        ldr             lr,  c0x80808080

        pkhbt           r6,  r8,  r6,  lsl #16
        pkhbt           r10, r7,  r10, lsl #16
        uxtb16          r6,  r6
        uxtb16          r10, r10

        sub             r0,  r0,  r1

        orr             r10, r6,  r10, lsl #8   @ u = vp8_signed_char_clamp((63 + Filter2 * 27)>>7)

        qsub8           r8,  r9,  r10           @ s = vp8_signed_char_clamp(qs0 - u)
        qadd8           r10, r11, r10           @ s = vp8_signed_char_clamp(ps0 + u)
        eor             r8,  r8,  lr            @ *oq0 = s^0x80
        str             r8, [r0]                @ store *oq0
        sub             r0,  r0,  r1
        eor             r10, r10, lr            @ *op0 = s^0x80
        str             r10,[r0]                @ store *op0

        @roughly 2/7th difference across boundary
        mov             lr,  #0x12              @ 18
        mov             r7,  #0x3f              @ 63

        sxtb16          r6,  r12
        sxtb16          r10, r12, ror #8
        smlabb          r8,  r6,  lr,  r7
        smlatb          r6,  r6,  lr,  r7
        smlabb          r9,  r10, lr,  r7
        smlatb          r10, r10, lr,  r7
        ssat            r8,  #8,  r8,  asr #7
        ssat            r6,  #8,  r6,  asr #7
        ssat            r9,  #8,  r9,  asr #7
        ssat            r10, #8,  r10, asr #7

        ldr             lr,  c0x80808080

        pkhbt           r6,  r8,  r6,  lsl #16
        pkhbt           r10, r9,  r10, lsl #16

        ldr             r9,  [sp,  #8]          @ load qs1
        ldr             r11, [sp,  #12]         @ load ps1

        uxtb16          r6,  r6
        uxtb16          r10, r10

        sub             r0,  r0,  r1

        orr             r10, r6,  r10, lsl #8   @ u = vp8_signed_char_clamp((63 + Filter2 * 18)>>7)

        qadd8           r11, r11, r10           @ s = vp8_signed_char_clamp(ps1 + u)
        qsub8           r8,  r9,  r10           @ s = vp8_signed_char_clamp(qs1 - u)
        eor             r11, r11, lr            @ *op1 = s^0x80
        str_post        r11, r0,  r1            @ store *op1
        eor             r8,  r8,  lr            @ *oq1 = s^0x80
        add             r0,  r0,  r1,  lsl #1

        mov             r7,  #0x3f              @ 63

        str_post        r8,  r0,  r1            @ store *oq1

        @roughly 1/7th difference across boundary
        mov             lr,  #0x9               @ 9
        ldr             r9, [r0]                @ load q2

        sxtb16          r6,  r12
        sxtb16          r10, r12, ror #8
        smlabb          r8,  r6,  lr,  r7
        smlatb          r6,  r6,  lr,  r7
        smlabb          r12, r10, lr,  r7
        smlatb          r10, r10, lr,  r7
        ssat            r8,  #8,  r8,  asr #7
        ssat            r6,  #8,  r6,  asr #7
        ssat            r12, #8,  r12, asr #7
        ssat            r10, #8,  r10, asr #7

        sub             r0,  r0,  r1,  lsl #2

        pkhbt           r6,  r8,  r6,  lsl #16
        pkhbt           r10, r12, r10, lsl #16

        sub             r0,  r0,  r1
        ldr             lr,  c0x80808080

        ldr             r11, [r0]               @ load p2

        uxtb16          r6,  r6
        uxtb16          r10, r10

        eor             r9,  r9,  lr
        eor             r11, r11, lr

        orr             r10, r6,  r10, lsl #8   @ u = vp8_signed_char_clamp((63 + Filter2 * 9)>>7)

        qadd8           r8,  r11, r10           @ s = vp8_signed_char_clamp(ps2 + u)
        qsub8           r10, r9,  r10           @ s = vp8_signed_char_clamp(qs2 - u)
        eor             r8,  r8,  lr            @ *op2 = s^0x80
        str_post        r8,  r0,  r1,  lsl #2   @ store *op2
        add             r0,  r0,  r1
        eor             r10, r10, lr            @ *oq2 = s^0x80
        str_post        r10, r0,  r1,  lsl #1   @ store *oq2

2:
        add             r0,  r0,  #4
        sub             r0,  r0,  r1,  lsl #3
        subs            r5,  r5,  #1

T       ittt            ne
        ldrne           r10,[r0,  r1]           @ p2
A       ldrne           r9, [r0], r1,  lsl #1   @ p3
T       ldrne           r9, [r0]                @ p3
T       addne           r0,  r0,  r1,  lsl #1
T       ittt            ne
        ldrne           r12,[r0,  r1]           @ p0
A       ldrne           r11,[r0], r1,  lsl #1   @ p1
T       ldrne           r11,[r0]                @ p3
T       addne           r0,  r0,  r1,  lsl #1

        bne             1b

        add             sp,  sp,  #16
        pop            {r4 - r11, pc}
endfunc

.macro TRANSPOSE_MATRIX i0, i1, i2, i3, o3, o2, o1, o0
        @ input:  $0, $1, $2, $3
        @ output: $4, $5, $6, $7
        @ i0: 03 02 01 00
        @ i1: 13 12 11 10
        @ i2: 23 22 21 20
        @ i3: 33 32 31 30
        @     o3 o2 o1 o0

        uxtb16          \o1, \i1                @ xx 12 xx 10
        uxtb16          \o0, \i0                @ xx 02 xx 00
        uxtb16          \o3, \i3                @ xx 32 xx 30
        uxtb16          \o2, \i2                @ xx 22 xx 20
        orr             \o1, \o0, \o1, lsl #8   @ 12 02 10 00
        orr             \o3, \o2, \o3, lsl #8   @ 32 22 30 20

        uxtb16          \i1, \i1, ror #8        @ xx 13 xx 11
        uxtb16          \i3, \i3, ror #8        @ xx 33 xx 31
        uxtb16          \i0, \i0, ror #8        @ xx 03 xx 01
        uxtb16          \i2, \i2, ror #8        @ xx 23 xx 21
        orr             \i0, \i0, \i1, lsl #8   @ 13 03 11 01
        orr             \i2, \i2, \i3, lsl #8   @ 33 23 31 21

        pkhtb           \o2, \o3, \o1, asr #16  @ 32 22 12 02   -- p1
        pkhbt           \o0, \o1, \o3, lsl #16  @ 30 20 10 00   -- p3

        pkhtb           \o3, \i2, \i0, asr #16  @ 33 23 13 03   -- p0
        pkhbt           \o1, \i0, \i2, lsl #16  @ 31 21 11 01   -- p2
.endm

@ void vp8_h_loop_filter16_simple(uint8_t *dst, int stride, int flim)
function ff_vp8_h_loop_filter16_simple_armv6, export=1
        push           {r4 - r11, lr}
        orr             r12, r2,  r2,  lsl #16
        ldr             r2,  c0x80808080
        orr             r12, r12, r12, lsl #8

        @ load soure data to r7, r8, r9, r10
        sub             r0,  r0,  #2
        ldr             r8, [r0,  r1]
        ldr_post        r7,  r0,  r1,  lsl #1
        ldr             r10,[r0,  r1]
        ldr_post        r9,  r0,  r1,  lsl #1
        add             r0,  r0,  #2

        mov             r11, #4                 @ count (r11) for 4-in-parallel
1:
        @transpose r7, r8, r9, r10 to r3, r4, r5, r6
        TRANSPOSE_MATRIX r7, r8, r9, r10, r6, r5, r4, r3

        @ vp8_simple_filter_mask() function
        uqsub8          r7,  r3,  r6            @ p1 - q1
        uqsub8          r8,  r6,  r3            @ q1 - p1
        uqsub8          r9,  r4,  r5            @ p0 - q0
        uqsub8          r10, r5,  r4            @ q0 - p0
        orr             r7,  r7,  r8            @ abs(p1 - q1)
        orr             r9,  r9,  r10           @ abs(p0 - q0)
        mov             r8,  #0
        uqadd8          r9,  r9,  r9            @ abs(p0 - q0) * 2
        uhadd8          r7,  r7,  r8            @ abs(p1 - q1) / 2
        uqadd8          r7,  r7,  r9            @ abs(p0 - q0)*2 + abs(p1 - q1)/2
        mvn             r10, #0                 @ r10 == -1

        usub8           r7,  r12, r7            @ compare to flimit
        sel             lr,  r10, r8            @ filter mask

        cmp             lr,  #0
        beq             2f                      @ skip filtering

        @vp8_simple_filter() function
        eor             r3,  r3,  r2            @ p1 offset to convert to a signed value
        eor             r6,  r6,  r2            @ q1 offset to convert to a signed value
        eor             r4,  r4,  r2            @ p0 offset to convert to a signed value
        eor             r5,  r5,  r2            @ q0 offset to convert to a signed value

        qsub8           r3,  r3,  r6            @ vp8_filter = p1 - q1
        qsub8           r6,  r5,  r4            @ q0 - p0

        qadd8           r3,  r3,  r6            @ vp8_filter += q0 - p0
        ldr             r9,  c0x03030303        @ r9 = 3

        qadd8           r3,  r3,  r6            @ vp8_filter += q0 - p0
        ldr             r7,  c0x04040404

        qadd8           r3,  r3,  r6            @ vp8_filter = p1-q1 + 3*(q0-p0))
        @STALL
        and             r3,  r3,  lr            @ vp8_filter &= mask

        qadd8           r9,  r3,  r9            @ Filter2 = vp8_filter + 3
        qadd8           r3,  r3,  r7            @ Filter1 = vp8_filter + 4

        shadd8          r9,  r9,  r8
        shadd8          r3,  r3,  r8
        shadd8          r9,  r9,  r8
        shadd8          r3,  r3,  r8
        shadd8          r9,  r9,  r8            @ Filter2 >>= 3
        shadd8          r3,  r3,  r8            @ Filter1 >>= 3

        @calculate output
        sub             r0,  r0,  r1,  lsl #2

        qadd8           r4,  r4,  r9            @ u = p0 + Filter2
        qsub8           r5,  r5,  r3            @ u = q0 - Filter1
        eor             r4,  r4,  r2            @ *op0 = u^0x80
        eor             r5,  r5,  r2            @ *oq0 = u^0x80

        strb            r4, [r0,  #-1]          @ store the result
        mov             r4,  r4,  lsr #8
        strb_post       r5,  r0,  r1
        mov             r5,  r5,  lsr #8

        strb            r4, [r0,  #-1]
        mov             r4,  r4,  lsr #8
        strb_post       r5,  r0,  r1
        mov             r5,  r5,  lsr #8

        strb            r4, [r0,  #-1]
        mov             r4,  r4,  lsr #8
        strb_post       r5,  r0,  r1
        mov             r5,  r5,  lsr #8

        strb            r4, [r0,  #-1]
        strb_post       r5,  r0,  r1

2:
        subs            r11, r11, #1

        @ load soure data to r7, r8, r9, r10
        sub             r0,  r0,  #2
T       ittt            ne
        ldrne           r8, [r0,  r1]
A       ldrne           r7, [r0], r1,  lsl #1
T       ldrne           r7, [r0]
T       addne           r0,  r0,  r1,  lsl #1
T       ittt            ne
        ldrne           r10,[r0,  r1]
A       ldrne           r9, [r0], r1,  lsl #1
T       ldrne           r9, [r0]
T       addne           r0,  r0,  r1,  lsl #1
        add             r0,  r0,  #2

        bne             1b

        pop            {r4 - r11, pc}
endfunc

@ void vp8_h_loop_filter16_inner(uint8_t *dst, int stride,
@                                int fE, int fI, int hev_thresh)
@ and
@ void vp8_h_loop_filter8uv_inner(uint8_t *dstU, uint8_t *dstV, int stride,
@                          int fE, int fI, int hev_thresh)
@ call:
@ void vp8_h_loop_filter_inner(uint8_t *dst, int stride,
@                              int fE, int fI, int hev_thresh, int count)
function ff_vp8_h_loop_filter_inner_armv6, export=1
        push           {r4 - r11, lr}

        sub             r0,  r0,  #4            @ move r0 pointer down by 4
        ldr             r5, [sp,  #40]          @ counter
        ldr             r9, [sp,  #36]          @ load thresh address
        sub             sp,  sp,  #16           @ create temp buffer

        ldr             r7, [r0,  r1]           @ transpose will make it into p3-p0
        ldr_post        r6,  r0,  r1,  lsl #1   @ load source data
        ldr             lr, [r0,  r1]
        ldr_post        r8,  r0,  r1,  lsl #1

        orr             r2,  r2,  r2,  lsl #16
        orr             r3,  r3,  r3,  lsl #16
        orr             r9,  r9,  r9,  lsl #16
        orr             r4,  r2,  r2,  lsl #8   @ flimE  splat int -> byte
        orr             r2,  r3,  r3,  lsl #8   @ flimI  splat int -> byte
        orr             r3,  r9,  r9,  lsl #8   @ thresh splat int -> byte

1:
        @ vp8_filter_mask() function
        @ calculate breakout conditions
        @ transpose the source data for 4-in-parallel operation
        TRANSPOSE_MATRIX r6, r7, r8, lr, r12, r11, r10, r9

        uqsub8          r7,  r9,  r10           @ p3 - p2
        uqsub8          r8,  r10, r9            @ p2 - p3
        uqsub8          r9,  r10, r11           @ p2 - p1
        uqsub8          r10, r11, r10           @ p1 - p2
        orr             r7,  r7,  r8            @ abs (p3-p2)
        orr             r10, r9,  r10           @ abs (p2-p1)
        uqsub8          lr,  r7,  r2            @ compare to limit. lr: vp8_filter_mask
        uqsub8          r10, r10, r2            @ compare to limit

        sub             r0,  r0,  r1,  lsl #2   @ move r0 pointer down by 4 lines

        orr             lr,  lr,  r10

        uqsub8          r6,  r11, r12           @ p1 - p0
        uqsub8          r7,  r12, r11           @ p0 - p1
        add             r0,  r0,  #4            @ move r0 pointer up by 4
        orr             r6,  r6,  r7            @ abs (p1-p0)
        str             r11,[sp,  #12]          @ save p1
        uqsub8          r10, r6,  r2            @ compare to limit
        uqsub8          r11, r6,  r3            @ compare to thresh
        orr             lr,  lr,  r10

        @ transpose uses 8 regs(r6 - r12 and lr). Need to save reg value now
        @ transpose the source data for 4-in-parallel operation
        str             r11,[sp]                @ push r11 to stack
        ldr             r7, [r0,  r1]
        ldr_post        r6,  r0,  r1,  lsl #1   @ load source data
        str             r12,[sp,  #4]           @ save current reg before load q0 - q3 data
        str             lr, [sp,  #8]
        ldr             lr, [r0,  r1]
        ldr_post        r8,  r0,  r1,  lsl #1

        TRANSPOSE_MATRIX r6, r7, r8, lr, r12, r11, r10, r9

        ldr             lr, [sp, #8]            @ load back (f)limit accumulator

        uqsub8          r6,  r12, r11           @ q3 - q2
        uqsub8          r7,  r11, r12           @ q2 - q3
        uqsub8          r12, r11, r10           @ q2 - q1
        uqsub8          r11, r10, r11           @ q1 - q2
        orr             r6,  r6,  r7            @ abs (q3-q2)
        orr             r7,  r12, r11           @ abs (q2-q1)
        uqsub8          r6,  r6,  r2            @ compare to limit
        uqsub8          r7,  r7,  r2            @ compare to limit
        ldr             r11,[sp,  #4]           @ load back p0
        ldr             r12,[sp,  #12]          @ load back p1
        orr             lr,  lr,  r6
        orr             lr,  lr,  r7

        uqsub8          r6,  r11, r9            @ p0 - q0
        uqsub8          r7,  r9,  r11           @ q0 - p0
        uqsub8          r8,  r12, r10           @ p1 - q1
        uqsub8          r11, r10, r12           @ q1 - p1
        orr             r6,  r6,  r7            @ abs (p0-q0)
        ldr             r7,  c0x7F7F7F7F
        orr             r8,  r8,  r11           @ abs (p1-q1)
        uqadd8          r6,  r6,  r6            @ abs (p0-q0) * 2
        and             r8,  r7,  r8,  lsr #1   @ abs (p1-q1) / 2
        uqsub8          r11, r10, r9            @ q1 - q0
        uqadd8          r6,  r8,  r6            @ abs (p0-q0)*2 + abs (p1-q1)/2
        uqsub8          r12, r9,  r10           @ q0 - q1
        uqsub8          r6,  r6,  r4            @ compare to flimit

        orr             r9,  r11, r12           @ abs (q1-q0)
        uqsub8          r8,  r9,  r2            @ compare to limit
        uqsub8          r10, r9,  r3            @ compare to thresh
        orr             lr,  lr,  r6
        orr             lr,  lr,  r8

        mvn             r11, #0                 @ r11 == -1
        mov             r12, #0

        usub8           lr,  r12, lr
        ldr             r9, [sp]                @ load the compared result
        sel             lr,  r11, r12           @ filter mask: lr

        cmp             lr,  #0
        beq             2f                      @ skip filtering

        @vp8_hevmask() function
        @calculate high edge variance
        sub             r0,  r0,  r1,  lsl #2   @ move r0 pointer down by 4 lines

        orr             r9,  r9,  r10

        ldrh            r7, [r0,  #-2]
        ldrh_post       r8,  r0,  r1

        usub8           r9,  r12, r9
        sel             r6,  r12, r11           @ hev mask: r6

        @vp8_filter() function
        @ load soure data to r6, r11, r12, lr
        ldrh            r9, [r0,  #-2]
        ldrh_post       r10, r0,  r1

        pkhbt           r12, r7,  r8,  lsl #16

        ldrh            r7, [r0,  #-2]
        ldrh_post       r8,  r0,  r1

        pkhbt           r11, r9,  r10, lsl #16

        ldrh            r9, [r0,  #-2]
        ldrh_post       r10, r0,  r1

        @ Transpose needs 8 regs(r6 - r12, and lr). Save r6 and lr first
        str             r6, [sp]
        str             lr, [sp,  #4]

        pkhbt           r6,  r7,  r8,  lsl #16
        pkhbt           lr,  r9,  r10, lsl #16

        @transpose r12, r11, r6, lr to r7, r8, r9, r10
        TRANSPOSE_MATRIX r12, r11, r6, lr, r10, r9, r8, r7

        @load back hev_mask r6 and filter_mask lr
        ldr             r12, c0x80808080
        ldr             r6, [sp]
        ldr             lr, [sp,  #4]

        eor             r7,  r7,  r12           @ p1 offset to convert to a signed value
        eor             r8,  r8,  r12           @ p0 offset to convert to a signed value
        eor             r9,  r9,  r12           @ q0 offset to convert to a signed value
        eor             r10, r10, r12           @ q1 offset to convert to a signed value

        str             r9, [sp]                @ store qs0 temporarily
        str             r8, [sp,  #4]           @ store ps0 temporarily
        str             r10,[sp,  #8]           @ store qs1 temporarily
        str             r7, [sp,  #12]          @ store ps1 temporarily

        qsub8           r7,  r7,  r10           @ vp8_signed_char_clamp(ps1-qs1)
        qsub8           r8,  r9,  r8            @ vp8_signed_char_clamp(vp8_filter + 3 * ( qs0 - ps0))

        and             r7,  r7,  r6            @  vp8_filter (r7) &= hev (r7 : filter)

        qadd8           r7,  r7,  r8
        ldr             r9,  c0x03030303        @ r9 = 3 --modified for vp8

        qadd8           r7,  r7,  r8
        ldr             r10, c0x04040404

        qadd8           r7,  r7,  r8

        and             r7,  r7,  lr            @ vp8_filter &= mask

        qadd8           r8,  r7,  r9            @ Filter2 (r8) = vp8_signed_char_clamp(vp8_filter+3)
        qadd8           r7,  r7,  r10           @ vp8_filter = vp8_signed_char_clamp(vp8_filter+4)

        mov             r9,  #0
        shadd8          r8,  r8,  r9            @ Filter2 >>= 3
        shadd8          r7,  r7,  r9            @ vp8_filter >>= 3
        shadd8          r8,  r8,  r9
        shadd8          r7,  r7,  r9
        shadd8          lr,  r8,  r9            @ lr: filter2
        shadd8          r7,  r7,  r9            @ r7: filter

        @calculate output
        ldr             r8, [sp]                @ load qs0
        ldr             r9, [sp,  #4]           @ load ps0

        ldr             r10, c0x01010101

        qsub8           r8,  r8,  r7            @ u = vp8_signed_char_clamp(qs0 - vp8_filter)
        qadd8           r9,  r9,  lr            @ u = vp8_signed_char_clamp(ps0 + Filter2)

        eor             r8,  r8,  r12
        eor             r9,  r9,  r12

        mov             lr,  #0

        sadd8           r7,  r7,  r10
        shadd8          r7,  r7,  lr

        ldr             r10,[sp,  #8]           @ load qs1
        ldr             r11,[sp,  #12]          @ load ps1

        bic             r7,  r7,  r6            @ r7: vp8_filter

        qsub8           r10, r10, r7            @ u = vp8_signed_char_clamp(qs1 - vp8_filter)
        qadd8           r11, r11, r7            @ u = vp8_signed_char_clamp(ps1 + vp8_filter)
        eor             r10, r10, r12
        eor             r11, r11, r12

        sub             r0,  r0,  r1,  lsl #2

        @we can use TRANSPOSE_MATRIX macro to transpose output - input: q1, q0, p0, p1
        TRANSPOSE_MATRIX r11, r9, r8, r10, lr, r12, r7, r6

        strh            r6, [r0,  #-2]          @ store the result
        mov             r6,  r6,  lsr #16
        strh_post       r6,  r0,  r1

        strh            r7, [r0,  #-2]
        mov             r7,  r7,  lsr #16
        strh_post       r7,  r0,  r1

        strh            r12, [r0,  #-2]
        mov             r12,  r12, lsr #16
        strh_post       r12,  r0,  r1

        strh            lr, [r0,  #-2]
        mov             lr,  lr,  lsr #16
        strh_post       lr,  r0,  r1

2:
        sub             r0,  r0,  #4
        subs            r5,  r5,  #1

T       ittt            ne
        ldrne           r7, [r0,  r1]
A       ldrne           r6, [r0], r1,  lsl #1   @ load source data
T       ldrne           r6, [r0]                @ load source data
T       addne           r0,  r0,  r1,  lsl #1
T       ittt            ne
        ldrne           lr, [r0,  r1]
A       ldrne           r8, [r0], r1,  lsl #1
T       ldrne           r8, [r0]
T       addne           r0,  r0,  r1,  lsl #1

        bne             1b

        add             sp, sp, #16
        pop            {r4 - r11, pc}
endfunc

@ void vp8_h_loop_filter16(uint8_t *dst, int stride,
@                          int fE, int fI, int hev_thresh)
@ and
@ void vp8_h_loop_filter8uv(uint8_t *dstU, uint8_t *dstV, int stride,
@                           int fE, int fI, int hev_thresh)
@ call:
@ void vp8_h_loop_filter(uint8_t *dst, int stride,
@                        int fE, int fI, int hev_thresh, int count)
function ff_vp8_h_loop_filter_armv6, export=1
        push           {r4 - r11, lr}

        sub             r0,  r0,  #4            @ move r0 pointer down by 4
        ldr             r5, [sp,  #40]          @ counter
        ldr             r9, [sp,  #36]          @ load thresh address
        sub             sp,  sp,  #16           @ create temp buffer

        ldr             r7, [r0,  r1]           @ transpose will make it into p3-p0
        ldr_post        r6,  r0,  r1,  lsl #1   @ load source data
        ldr             lr, [r0,  r1]
        ldr_post        r8,  r0,  r1,  lsl #1

        orr             r2,  r2,  r2,  lsl #16
        orr             r3,  r3,  r3,  lsl #16
        orr             r9,  r9,  r9,  lsl #16
        orr             r4,  r2,  r2,  lsl #8   @ flimE  splat int -> byte
        orr             r2,  r3,  r3,  lsl #8   @ flimI  splat int -> byte
        orr             r3,  r9,  r9,  lsl #8   @ thresh splat int -> byte

1:
        @ vp8_filter_mask() function
        @ calculate breakout conditions
        @ transpose the source data for 4-in-parallel operation
        TRANSPOSE_MATRIX r6, r7, r8, lr, r12, r11, r10, r9

        uqsub8          r7,  r9,  r10           @ p3 - p2
        uqsub8          r8,  r10, r9            @ p2 - p3
        uqsub8          r9,  r10, r11           @ p2 - p1
        uqsub8          r10, r11, r10           @ p1 - p2
        orr             r7,  r7,  r8            @ abs (p3-p2)
        orr             r10, r9,  r10           @ abs (p2-p1)
        uqsub8          lr,  r7,  r2            @ compare to limit. lr: vp8_filter_mask
        uqsub8          r10, r10, r2            @ compare to limit

        sub             r0,  r0,  r1,  lsl #2   @ move r0 pointer down by 4 lines

        orr             lr,  lr,  r10

        uqsub8          r6,  r11, r12           @ p1 - p0
        uqsub8          r7,  r12, r11           @ p0 - p1
        add             r0,  r0,  #4            @ move r0 pointer up by 4
        orr             r6,  r6,  r7            @ abs (p1-p0)
        str             r11,[sp,  #12]          @ save p1
        uqsub8          r10, r6,  r2            @ compare to limit
        uqsub8          r11, r6,  r3            @ compare to thresh
        orr             lr,  lr,  r10

        @ transpose uses 8 regs(r6 - r12 and lr). Need to save reg value now
        @ transpose the source data for 4-in-parallel operation
        str             r11,[sp]                @ push r11 to stack
        ldr             r7, [r0,  r1]
        ldr_post        r6,  r0,  r1,  lsl #1   @ load source data
        str             r12,[sp,  #4]           @ save current reg before load q0 - q3 data
        str             lr, [sp,  #8]
        ldr             lr, [r0,  r1]
        ldr_post        r8,  r0,  r1,  lsl #1

        TRANSPOSE_MATRIX r6, r7, r8, lr, r12, r11, r10, r9

        ldr             lr, [sp,  #8]           @ load back (f)limit accumulator

        uqsub8          r6,  r12, r11           @ q3 - q2
        uqsub8          r7,  r11, r12           @ q2 - q3
        uqsub8          r12, r11, r10           @ q2 - q1
        uqsub8          r11, r10, r11           @ q1 - q2
        orr             r6,  r6,  r7            @ abs (q3-q2)
        orr             r7,  r12, r11           @ abs (q2-q1)
        uqsub8          r6,  r6,  r2            @ compare to limit
        uqsub8          r7,  r7,  r2            @ compare to limit
        ldr             r11,[sp,  #4]           @ load back p0
        ldr             r12,[sp,  #12]          @ load back p1
        orr             lr,  lr,  r6
        orr             lr,  lr,  r7

        uqsub8          r6,  r11, r9            @ p0 - q0
        uqsub8          r7,  r9,  r11           @ q0 - p0
        uqsub8          r8,  r12, r10           @ p1 - q1
        uqsub8          r11, r10, r12           @ q1 - p1
        orr             r6,  r6,  r7            @ abs (p0-q0)
        ldr             r7,  c0x7F7F7F7F
        orr             r8,  r8,  r11           @ abs (p1-q1)
        uqadd8          r6,  r6,  r6            @ abs (p0-q0) * 2
        and             r8,  r7,  r8,  lsr #1   @ abs (p1-q1) / 2
        uqsub8          r11, r10, r9            @ q1 - q0
        uqadd8          r6,  r8,  r6            @ abs (p0-q0)*2 + abs (p1-q1)/2
        uqsub8          r12, r9,  r10           @ q0 - q1
        uqsub8          r6,  r6,  r4            @ compare to flimit

        orr             r9,  r11, r12           @ abs (q1-q0)
        uqsub8          r8,  r9,  r2            @ compare to limit
        uqsub8          r10, r9,  r3            @ compare to thresh
        orr             lr,  lr,  r6
        orr             lr,  lr,  r8

        mvn             r11,  #0                @ r11 == -1
        mov             r12,  #0

        usub8           lr,  r12, lr
        ldr             r9, [sp]                @ load the compared result
        sel             lr,  r11, r12           @ filter mask: lr

        cmp             lr,  #0
        beq             2f                      @ skip filtering


        @vp8_hevmask() function
        @calculate high edge variance
        sub             r0,  r0,  r1,  lsl #2   @ move r0 pointer down by 4 lines

        orr             r9,  r9,  r10

        ldrh            r7, [r0,  #-2]
        ldrh_post       r8,  r0,  r1

        usub8           r9,  r12, r9
        sel             r6,  r12, r11           @ hev mask: r6


        @ vp8_mbfilter() function
        @ p2, q2 are only needed at the end. do not need to load them in now.
        @ Transpose needs 8 regs(r6 - r12, and lr). Save r6 and lr first
        @ load soure data to r6, r11, r12, lr
        ldrh            r9, [r0,  #-2]
        ldrh_post       r10, r0,  r1

        pkhbt           r12, r7,  r8,  lsl #16

        ldrh            r7, [r0,  #-2]
        ldrh_post       r8,  r0,  r1

        pkhbt           r11, r9,  r10, lsl #16

        ldrh            r9, [r0,  #-2]
        ldrh_post       r10, r0,  r1

        str             r6, [sp]                @ save r6
        str             lr, [sp,  #4]           @ save lr

        pkhbt           r6,  r7,  r8,  lsl #16
        pkhbt           lr,  r9,  r10, lsl #16

        @transpose r12, r11, r6, lr to p1, p0, q0, q1
        TRANSPOSE_MATRIX r12, r11, r6, lr, r10, r9, r8, r7

        @load back hev_mask r6 and filter_mask lr
        ldr             r12, c0x80808080
        ldr             r6, [sp]
        ldr             lr, [sp,  #4]

        eor             r7,  r7,  r12           @ ps1
        eor             r8,  r8,  r12           @ ps0
        eor             r9,  r9,  r12           @ qs0
        eor             r10, r10, r12           @ qs1

        qsub8           r12, r9,  r8            @ vp8_signed_char_clamp(vp8_filter + 3 * ( qs0 - ps0))
        str             r7, [sp,  #12]          @ store ps1 temporarily
        qsub8           r7,  r7,  r10           @ vp8_signed_char_clamp(ps1-qs1)
        str             r10,[sp,  #8]           @ store qs1 temporarily
        qadd8           r7,  r7,  r12
        str             r9, [sp]                @ store qs0 temporarily
        qadd8           r7,  r7,  r12
        str             r8, [sp,  #4]           @ store ps0 temporarily
        qadd8           r7,  r7,  r12           @ vp8_filter: r7

        ldr             r10, c0x03030303        @ r10 = 3 --modified for vp8
        ldr             r9,  c0x04040404

        and             r7,  r7,  lr            @ vp8_filter &= mask (lr is free)

        mov             r12, r7                 @ Filter2: r12
        and             r12, r12, r6            @ Filter2 &= hev

        @save bottom 3 bits so that we round one side +4 and the other +3
        qadd8           r8,  r12, r9            @ Filter1 (r8) = vp8_signed_char_clamp(Filter2+4)
        qadd8           r12, r12, r10           @ Filter2 (r12) = vp8_signed_char_clamp(Filter2+3)

        mov             r10, #0
        shadd8          r8,  r8,  r10           @ Filter1 >>= 3
        shadd8          r12, r12, r10           @ Filter2 >>= 3
        shadd8          r8,  r8,  r10
        shadd8          r12, r12, r10
        shadd8          r8,  r8,  r10           @ r8: Filter1
        shadd8          r12, r12, r10           @ r12: Filter2

        ldr             r9, [sp]                @ load qs0
        ldr             r11,[sp,  #4]           @ load ps0

        qsub8           r9,  r9,  r8            @ qs0 = vp8_signed_char_clamp(qs0 - Filter1)
        qadd8           r11, r11, r12           @ ps0 = vp8_signed_char_clamp(ps0 + Filter2)

        bic             r12, r7,  r6            @vp8_filter &= ~hev    ( r6 is free)

        @roughly 3/7th difference across boundary
        mov             lr,  #0x1b              @ 27
        mov             r7,  #0x3f              @ 63

        sxtb16          r6,  r12
        sxtb16          r10, r12, ror #8
        smlabb          r8,  r6,  lr,  r7
        smlatb          r6,  r6,  lr,  r7
        smlabb          r7,  r10, lr,  r7
        smultb          r10, r10, lr
        ssat            r8,  #8,  r8,  asr #7
        ssat            r6,  #8,  r6,  asr #7
        add             r10, r10, #63
        ssat            r7,  #8,  r7,  asr #7
        ssat            r10, #8,  r10, asr #7

        ldr             lr, c0x80808080

        pkhbt           r6,  r8,  r6,  lsl #16
        pkhbt           r10, r7,  r10, lsl #16
        uxtb16          r6,  r6
        uxtb16          r10, r10

        sub             r0,  r0,  r1,  lsl #2   @ move r0 pointer down by 4 lines

        orr             r10, r6,  r10, lsl #8   @ u = vp8_signed_char_clamp((63 + Filter2 * 27)>>7)

        qsub8           r8,  r9,  r10           @ s = vp8_signed_char_clamp(qs0 - u)
        qadd8           r10, r11, r10           @ s = vp8_signed_char_clamp(ps0 + u)
        eor             r8,  r8,  lr            @ *oq0 = s^0x80
        eor             r10, r10, lr            @ *op0 = s^0x80

        strb            r10,[r0,  #-1]          @ store op0 result
        strb_post       r8,  r0,  r1            @ store oq0 result
        mov             r10, r10, lsr #8
        mov             r8,  r8,  lsr #8
        strb            r10,[r0,  #-1]
        strb_post       r8,  r0,  r1
        mov             r10, r10, lsr #8
        mov             r8,  r8,  lsr #8
        strb            r10,[r0,  #-1]
        strb_post       r8,  r0,  r1
        mov             r10, r10, lsr #8
        mov             r8,  r8,  lsr #8
        strb            r10,[r0,  #-1]
        strb_post       r8,  r0,  r1

        @roughly 2/7th difference across boundary
        mov             lr,  #0x12              @ 18
        mov             r7,  #0x3f              @ 63

        sxtb16          r6,  r12
        sxtb16          r10, r12, ror #8
        smlabb          r8,  r6,  lr,  r7
        smlatb          r6,  r6,  lr,  r7
        smlabb          r9,  r10, lr,  r7
        smlatb          r10, r10, lr,  r7
        ssat            r8,  #8,  r8,  asr #7
        ssat            r6,  #8,  r6,  asr #7
        ssat            r9,  #8,  r9,  asr #7
        ssat            r10, #8,  r10, asr #7

        sub             r0,  r0,  r1,  lsl #2   @ move r0 pointer down by 4 lines

        pkhbt           r6,  r8,  r6,  lsl #16
        pkhbt           r10, r9,  r10, lsl #16

        ldr             r9, [sp,  #8]           @ load qs1
        ldr             r11,[sp,  #12]          @ load ps1
        ldr             lr,  c0x80808080

        uxtb16          r6,  r6
        uxtb16          r10, r10

        add             r0,  r0,  #2

        orr             r10, r6,  r10, lsl #8   @ u = vp8_signed_char_clamp((63 + Filter2 * 18)>>7)

        qsub8           r8,  r9,  r10           @ s = vp8_signed_char_clamp(qs1 - u)
        qadd8           r10, r11, r10           @ s = vp8_signed_char_clamp(ps1 + u)
        eor             r8,  r8,  lr            @ *oq1 = s^0x80
        eor             r10, r10, lr            @ *op1 = s^0x80

        ldrb            r11,[r0,  #-5]          @ load p2 for 1/7th difference across boundary
        strb            r10,[r0,  #-4]          @ store op1
        strb            r8, [r0,  #-1]          @ store oq1
        ldrb_post       r9,  r0,  r1            @ load q2 for 1/7th difference across boundary

        mov             r10, r10, lsr #8
        mov             r8,  r8,  lsr #8

        ldrb            r6, [r0,  #-5]
        strb            r10,[r0,  #-4]
        strb            r8, [r0,  #-1]
        ldrb_post       r7,  r0,  r1

        mov             r10, r10, lsr #8
        mov             r8,  r8,  lsr #8
        orr             r11, r11, r6,  lsl #8
        orr             r9,  r9,  r7,  lsl #8

        ldrb            r6, [r0,  #-5]
        strb            r10,[r0,  #-4]
        strb            r8, [r0,  #-1]
        ldrb_post       r7,  r0,  r1

        mov             r10, r10, lsr #8
        mov             r8,  r8,  lsr #8
        orr             r11, r11, r6,  lsl #16
        orr             r9,  r9,  r7,  lsl #16

        ldrb            r6, [r0,  #-5]
        strb            r10,[r0,  #-4]
        strb            r8, [r0,  #-1]
        ldrb_post       r7,  r0,  r1
        orr             r11, r11, r6,  lsl #24
        orr             r9,  r9,  r7,  lsl #24

        @roughly 1/7th difference across boundary
        eor             r9,  r9,  lr
        eor             r11, r11, lr

        mov             lr,  #0x9               @ 9
        mov             r7,  #0x3f              @ 63

        sxtb16          r6,  r12
        sxtb16          r10, r12, ror #8
        smlabb          r8,  r6,  lr,  r7
        smlatb          r6,  r6,  lr,  r7
        smlabb          r12, r10, lr,  r7
        smlatb          r10, r10, lr,  r7
        ssat            r8,  #8,  r8,  asr #7
        ssat            r6,  #8,  r6,  asr #7
        ssat            r12, #8,  r12, asr #7
        ssat            r10, #8,  r10, asr #7

        sub             r0,  r0,  r1,  lsl #2

        pkhbt           r6,  r8,  r6,  lsl #16
        pkhbt           r10, r12, r10, lsl #16

        uxtb16          r6,  r6
        uxtb16          r10, r10

        ldr             lr,  c0x80808080

        orr             r10, r6,  r10, lsl #8   @ u = vp8_signed_char_clamp((63 + Filter2 * 9)>>7)

        qadd8           r8,  r11, r10           @ s = vp8_signed_char_clamp(ps2 + u)
        qsub8           r10, r9,  r10           @ s = vp8_signed_char_clamp(qs2 - u)
        eor             r8,  r8,  lr            @ *op2 = s^0x80
        eor             r10, r10, lr            @ *oq2 = s^0x80

        strb            r8, [r0,  #-5]          @ store *op2
        strb_post       r10, r0,  r1            @ store *oq2
        mov             r8,  r8,  lsr #8
        mov             r10, r10, lsr #8
        strb            r8, [r0,  #-5]
        strb_post       r10, r0,  r1
        mov             r8,  r8,  lsr #8
        mov             r10, r10, lsr #8
        strb            r8, [r0,  #-5]
        strb_post       r10, r0,  r1
        mov             r8,  r8,  lsr #8
        mov             r10, r10, lsr #8
        strb            r8, [r0,  #-5]
        strb_post       r10, r0,  r1

        @adjust r0 pointer for next loop
        sub             r0,  r0,  #2

2:
        sub             r0,  r0,  #4
        subs            r5,  r5,  #1

T       ittt            ne
        ldrne           r7, [r0,  r1]
A       ldrne           r6, [r0], r1,  lsl #1   @ load source data
T       ldrne           r6, [r0]
T       addne           r0,  r0,  r1,  lsl #1
T       ittt            ne
        ldrne           lr, [r0,  r1]
A       ldrne           r8, [r0], r1,  lsl #1
T       ldrne           r8, [r0]
T       addne           r0,  r0,  r1,  lsl #1

        bne             1b

        add             sp,  sp,  #16
        pop            {r4 - r11, pc}
endfunc

@ MC

@ void put_vp8_pixels16(uint8_t *dst, int dststride, uint8_t *src,
@                       int srcstride, int h, int mx, int my)
function ff_put_vp8_pixels16_armv6, export=1
        push           {r4 - r11}
        ldr             r12,[sp,  #32]          @ h
1:
        subs            r12, r12, #2
        ldr             r5, [r2,  #4]
        ldr             r6, [r2,  #8]
        ldr             r7, [r2,  #12]
        ldr_post        r4,  r2,  r3
        ldr             r9, [r2,  #4]
        ldr             r10,[r2,  #8]
        ldr             r11,[r2,  #12]
        ldr_post        r8,  r2,  r3
        strd            r6,  r7, [r0,  #8]
        strd_post       r4,  r5,  r0,  r1
        strd            r10, r11,[r0,  #8]
        strd_post       r8,  r9,  r0,  r1
        bgt             1b
        pop            {r4 - r11}
        bx              lr
endfunc

@ void put_vp8_pixels8(uint8_t *dst, int dststride, uint8_t *src,
@                      int srcstride, int h, int mx, int my)
function ff_put_vp8_pixels8_armv6, export=1
        push           {r4 - r11}
        ldr             r12,[sp,  #32]          @ h
1:
        subs            r12, r12, #4
        ldr             r5, [r2,  #4]
        ldr_post        r4,  r2,  r3
        ldr             r7, [r2,  #4]
        ldr_post        r6,  r2,  r3
        ldr             r9, [r2,  #4]
        ldr_post        r8,  r2,  r3
        ldr             r11,[r2,  #4]
        ldr_post        r10, r2,  r3
        strd_post       r4,  r5,  r0,  r1
        strd_post       r6,  r7,  r0,  r1
        strd_post       r8,  r9,  r0,  r1
        strd_post       r10, r11, r0,  r1
        bgt             1b
        pop            {r4 - r11}
        bx              lr
endfunc

@ void put_vp8_pixels4(uint8_t *dst, int dststride, uint8_t *src,
@                      int srcstride, int h, int mx, int my)
function ff_put_vp8_pixels4_armv6, export=1
        ldr             r12, [sp, #0]           @ h
        push           {r4 - r6, lr}
1:
        subs            r12, r12, #4
        ldr             r5, [r2,  r3]
        ldr_post        r4,  r2,  r3,  lsl #1
        ldr             lr, [r2,  r3]
        ldr_post        r6,  r2,  r3,  lsl #1
        str             r5, [r0,  r1]
        str_post        r4,  r0,  r1,  lsl #1
        str             lr, [r0,  r1]
        str_post        r6,  r0,  r1,  lsl #1
        bgt             1b
        pop            {r4 - r6,  pc}
endfunc

@ note: worst case sum of all 6-tap filter values * 255 is 0x7f80 so 16 bit
@ arithmatic can be used to apply filters
const sixtap_filters_13245600, align=4
        .short     2, 108, -11,  36,  -8, 1, 0, 0
        .short     3,  77, -16,  77, -16, 3, 0, 0
        .short     1,  36,  -8, 108, -11, 2, 0, 0
endconst
const fourtap_filters_1324, align=4
        .short     -6,  12, 123, -1
        .short     -9,  50,  93, -6
        .short     -6,  93,  50, -9
        .short     -1, 123,  12, -6
endconst

@ void put_vp8_epel_h6(uint8_t *dst, int dststride, uint8_t *src,
@                      int srcstride, int w, int h, int mx)
function ff_put_vp8_epel_h6_armv6, export=1
        push           {r4 - r11, lr}

        sub             r2,  r2,  #2
        movrel          lr,  sixtap_filters_13245600 - 16
        ldr             r12,[sp,  #44]          @ vp8_filter index
        ldr             r4, [sp,  #36]          @ width
        add             lr,  lr,  r12, lsl #3
        sub             r3,  r3,  r4            @ src_stride - block_width
        sub             r1,  r1,  r4            @ dst_stride - block_width
        lsr             r4, #2

        str             r4, [sp,  #36]          @ "4-in-parallel" loop counter @40
        str             r3, [sp,  #44]          @ src_stride - block_width @48
        push           {r1}                     @ dst_stride - block_width @0
                                                @ height @44

        ldr             r1, [lr], #4            @ coefficients
        ldr             r3, [lr], #4
        ldr             lr, [lr]
1:
        @ 3 loads, 10 shuffles and then mul/acc/add/shr
        @ o0: i0/i1/i2/i3/i4/i5 -> i0/i2 (ld1) | i1/i3 (ld1)   | i4/i5 (ld2)
        @ o1: i1/i2/i3/i4/i5/i6 -> i1/i3 (ld1) | i2/i4 (ld2)   | i5/i6 (ld2/3)
        @ o2: i2/i3/i4/i5/i6/i7 -> i2/i4 (ld2) | i3/i5 (ld2)   | i6/i7 (ld3)
        @ o3: i3/i4/i5/i6/i7/i8 -> i3/i5 (ld2) | i4/i6 (ld2/3) | i7/i8 (ld3)
        ldr             r7, [r2,  #5]           @ ld3 -> src[5-8]
        ldr             r6, [r2,  #2]           @ ld2 -> src[2-5]
        ldr             r5, [r2], #4            @ ld1 -> src[0-3]

        pkhtb           r7,  r7,  r7,  asr #8   @ src[8,7,7,6]
        uxtb16          r9,  r6,  ror #8        @ src[5] | src[3]
        uxtb16          r6,  r6                 @ src[4] | src[2]
        uxtb16          r8,  r5,  ror #8        @ src[3] | src[1]
        uxtb16          r11, r7,  ror #8        @ src[8] | src[7]
        uxtb16          r7,  r7                 @ src[7] | src[6]
        pkhtb           r10, r9,  r6,  asr #16  @ src[5] | src[4]
        uxtb16          r5,  r5                 @ src[2] | src[0]

        smuad           r11, r11, lr            @ filter[3][2] -> r11
        subs            r4,  r4,  #1
        pkhbt           r12, r10, r7,  lsl #16  @ src[6] | src[4]
        smuad           r7,  r7,  lr            @ filter[2][2] -> r7
        smuad           r5,  r5,  r1            @ filter[0][0] -> r5
        smlad           r11, r9,  r1,  r11      @ filter[3][0] -> r11
        smlad           r7,  r9,  r3,  r7       @ filter[2][1] -> r7
        smuad           r9,  r8,  r1            @ filter[1][0] -> r9
        smlad           r5,  r8,  r3,  r5       @ filter[0][1] -> r5
        pkhtb           r8,  r12, r10, asr #16  @ src[6] | src[5]
        smlad           r11, r12, r3,  r11      @ filter[3][1] -> r11
        smlad           r9,  r6,  r3,  r9       @ filter[1][1] -> r9
        smlad           r5,  r10, lr,  r5       @ filter[0][2] -> r5
        smlad           r7,  r6,  r1,  r7       @ filter[2][0] -> r7
        smlad           r9,  r8,  lr,  r9       @ filter[1][2] -> r9

        add             r5,  r5,  #0x40         @ round_shift_and_clamp[0]
        add             r9,  r9,  #0x40         @ round_shift_and_clamp[1]
        add             r7,  r7,  #0x40         @ round_shift_and_clamp[2]
        add             r11, r11, #0x40         @ round_shift_and_clamp[3]

        usat            r5,  #8,  r5,  asr #7
        usat            r9,  #8,  r9,  asr #7
        usat            r7,  #8,  r7,  asr #7
        usat            r11, #8,  r11, asr #7

        strb            r5, [r0], #1            @ store res[0]
        strb            r9, [r0], #1            @ store res[1]
        strb            r7, [r0], #1            @ store res[2]
        strb            r11,[r0], #1            @ store res[3]

        bne             1b

        ldr             r12,[sp,  #44]          @ height = outer-loop counter
        subs            r12, r12, #1
T       itttt           ne
        ldrne           r4, [sp,  #40]          @ 4-in-parallel loop counter
        ldrne           r5, [sp,  #48]
        ldrne           r6, [sp]
        strne           r12,[sp,  #44]
        add             r2,  r2,  r5            @ move to next input/output lines
        add             r0,  r0,  r6

        bne             1b

        add             sp,  sp,  #4            @ restore stack after push{r1} above
        pop            {r4 - r11, pc}
endfunc

@ void put_vp8_epel_v6(uint8_t *dst, int dststride, uint8_t *src,
@                      int srcstride, int w, int h, int my)
function ff_put_vp8_epel_v6_armv6, export=1
        push           {r4 - r11, lr}

        movrel          lr,  sixtap_filters_13245600 - 16
        ldr             r12,[sp,  #44]          @ vp8_filter index
        ldr             r4, [sp,  #36]          @ width
        add             lr,  lr,  r12, lsl #3
        sub             r1,  r1,  r4            @ dst_stride - block_width
        lsr             r4,  #2

        str             r4, [sp,  #36]          @ "4-in-parallel" loop counter @40
        str             r3, [sp,  #44]          @ src_stride - block_width @48
        push           {r1}                     @ dst_stride - block_width @0
                                                @ height @44
1:
        add             r1,  r3,  r3,  lsl #1   @ stride * 3
        ldr_dpren       r5,  r2,  r3            @ src[0,1,2,3 + stride * 1]
        ldr             r6, [r2,  r3]           @ src[0,1,2,3 + stride * 3]
        ldr             r7, [r2,  r3,  lsl #1]  @ src[0,1,2,3 + stride * 4]
        ldr             r8, [r2,  r1]           @ src[0,1,2,3 + stride * 5]

        @ byte -> word and "transpose"
        uxtb16          r9,  r5,  ror #8        @ src[3 + stride*1] | src[1 + stride*1]
        uxtb16          r10, r6,  ror #8        @ src[3 + stride*3] | src[1 + stride*3]
        uxtb16          r11, r7,  ror #8        @ src[3 + stride*4] | src[1 + stride*4]
        uxtb16          r12, r8,  ror #8        @ src[3 + stride*5] | src[1 + stride*5]
        uxtb16          r5,  r5                 @ src[2 + stride*1] | src[0 + stride*1]
        uxtb16          r6,  r6                 @ src[2 + stride*3] | src[0 + stride*3]
        uxtb16          r7,  r7                 @ src[2 + stride*4] | src[0 + stride*4]
        uxtb16          r8,  r8                 @ src[2 + stride*5] | src[0 + stride*5]
        pkhbt           r1,  r9,  r10, lsl #16  @ src[1 + stride*3] | src[1 + stride*1]
        pkhtb           r9,  r10, r9,  asr #16  @ src[3 + stride*3] | src[3 + stride*1]
        pkhbt           r10, r11, r12, lsl #16  @ src[1 + stride*5] | src[1 + stride*4]
        pkhtb           r11, r12, r11, asr #16  @ src[3 + stride*5] | src[3 + stride*4]
        pkhbt           r12, r5,  r6,  lsl #16  @ src[0 + stride*3] | src[0 + stride*1]
        pkhtb           r5,  r6,  r5,  asr #16  @ src[2 + stride*3] | src[2 + stride*1]
        pkhbt           r6,  r7,  r8,  lsl #16  @ src[0 + stride*5] | src[0 + stride*4]
        pkhtb           r7,  r8,  r7,  asr #16  @ src[2 + stride*5] | src[2 + stride*4]

        ldr             r8, [lr,  #4]           @ stall - if only I had more registers...
        smuad           r12, r12, r8            @ filter[0][1]
        smuad           r1,  r1,  r8            @ filter[1][1]
        smuad           r5,  r5,  r8            @ filter[2][1]
        smuad           r9,  r9,  r8            @ filter[3][1]
        ldr             r8, [lr,  #8]           @ stall - if only I had more registers...
        smlad           r12, r6,  r8, r12       @ filter[0][2]
        smlad           r1,  r10, r8, r1        @ filter[1][2]
        ldr_dpren       r6,  r2,  r3,  lsl #1   @ src[0,1,2,3 + stride *  0]
        ldr             r10,[r2], #4            @ src[0,1,2,3 + stride *  2]
        smlad           r5,  r7,  r8, r5        @ filter[2][2]
        smlad           r9,  r11, r8, r9        @ filter[3][2]

        uxtb16          r7,  r6,  ror #8        @ src[3 + stride*0] | src[1 + stride*0]
        uxtb16          r11, r10, ror #8        @ src[3 + stride*2] | src[1 + stride*2]
        uxtb16          r6,  r6                 @ src[2 + stride*0] | src[0 + stride*0]
        uxtb16          r10, r10                @ src[2 + stride*2] | src[0 + stride*2]

        pkhbt           r8,  r7,  r11, lsl #16  @ src[1 + stride*2] | src[1 + stride*0]
        pkhtb           r7,  r11, r7,  asr #16  @ src[3 + stride*2] | src[3 + stride*0]
        pkhbt           r11, r6,  r10, lsl #16  @ src[0 + stride*2] | src[0 + stride*0]
        pkhtb           r6,  r10, r6,  asr #16  @ src[2 + stride*2] | src[2 + stride*0]

        ldr             r10,[lr]                @ stall - if only I had more registers...
        subs            r4,  r4,  #1            @ counter--
        smlad           r12, r11, r10, r12      @ filter[0][0]
        smlad           r1,  r8,  r10, r1       @ filter[1][0]
        smlad           r5,  r6,  r10, r5       @ filter[2][0]
        smlad           r9,  r7,  r10, r9       @ filter[3][0]

        add             r12, r12, #0x40         @ round_shift_and_clamp[0]
        add             r1,  r1,  #0x40         @ round_shift_and_clamp[1]
        add             r5,  r5,  #0x40         @ round_shift_and_clamp[2]
        add             r9,  r9,  #0x40         @ round_shift_and_clamp[3]

        usat            r12, #8,  r12, asr #7
        usat            r1,  #8,  r1,  asr #7
        usat            r5,  #8,  r5,  asr #7
        usat            r9,  #8,  r9,  asr #7

        strb            r12,[r0], #1            @ store res[0]
        strb            r1, [r0], #1            @ store res[1]
        strb            r5, [r0], #1            @ store res[2]
        strb            r9, [r0], #1            @ store res[3]

        bne             1b

        ldr             r12,[sp,  #44]          @ height = outer-loop counter
        subs            r12, r12, #1
T       itttt           ne
        ldrne           r4, [sp,  #40]          @ 4-in-parallel loop counter
        ldrne           r6, [sp,  #0]
        subne           r2,  r2,  r4,  lsl #2
        strne           r12,[sp,  #44]
        add             r0,  r0,  r6
        add             r2,  r2,  r3            @ move to next input/output lines

        bne             1b

        add             sp,  sp,  #4            @ restore stack after push{r1} above
        pop            {r4 - r11, pc}
endfunc

@ void put_vp8_epel_h4(uint8_t *dst, int dststride, uint8_t *src,
@                      int srcstride, int w, int h, int mx)
function ff_put_vp8_epel_h4_armv6, export=1
        push           {r4 - r11, lr}

        subs            r2,  r2,  #1
        movrel          lr,  fourtap_filters_1324 - 4
        ldr             r4, [sp,  #36]          @ width
        ldr             r12,[sp,  #44]          @ vp8_filter index
        add             lr,  lr,  r12, lsl #2
        sub             r3,  r3,  r4            @ src_stride - block_width
        sub             r1,  r1,  r4            @ dst_stride - block_width
        ldr             r5,  [lr]
        ldr             r6,  [lr,  #4]
        asr             r4,  #2

        ldr             lr, [sp,  #40]          @ height = outer-loop counter
        str             r4, [sp,  #36]          @ "4-in-parallel" inner loop counter
1:
        @ 3 loads, 5 uxtb16s and then mul/acc/add/shr
        @ o0: i0/i1/i2/i3 -> i0/i2(ld1) + i1/i3(ld1)
        @ o1: i1/i2/i3/i4 -> i1/i3(ld1) + i2/i4(ld2)
        @ o2: i2/i3/i4/i5 -> i2/i4(ld2) + i3/i5(ld2)
        @ o3: i3/i4/i5/i6 -> i3/i5(ld2) + i4/i6(ld3)
        ldr             r9, [r2,  #3]           @ load source data
        ldr             r8, [r2,  #2]
        ldr             r7, [r2], #4

        uxtb16          r9,  r9,  ror #8        @ src[6] | src[4]
        uxtb16          r10, r8,  ror #8        @ src[5] | src[3]
        uxtb16          r8,  r8                 @ src[4] | src[2]
        uxtb16          r11, r7,  ror #8        @ src[3] | src[1]
        uxtb16          r7,  r7                 @ src[2] | src[0]

        smuad           r9,  r9,  r6            @ filter[3][1] -> r9
        smuad           r12, r10, r6            @ filter[2][1] -> r12
        smuad           r7,  r7,  r5            @ filter[0][0] -> r7
        smlad           r9,  r10, r5,  r9       @ filter[3][0] -> r9
        smuad           r10, r11, r5            @ filter[1][0] -> r10
        smlad           r12, r8,  r5,  r12      @ filter[2][0] -> r12
        smlad           r7,  r11, r6,  r7       @ filter[0][1] -> r7
        smlad           r10, r8,  r6,  r10      @ filter[1][1] -> r10

        subs            r4,  r4,  #1            @ counter--

        add             r7,  r7,  #0x40         @ round_shift_and_clamp[0]
        add             r10, r10, #0x40         @ round_shift_and_clamp[1]
        add             r12, r12, #0x40         @ round_shift_and_clamp[2]
        add             r9,  r9,  #0x40         @ round_shift_and_clamp[3]

        usat            r7,  #8,  r7,  asr #7
        usat            r10, #8,  r10, asr #7
        usat            r12, #8,  r12, asr #7
        usat            r9,  #8,  r9,  asr #7

        strb            r7, [r0], #1            @ store res[0]
        strb            r10,[r0], #1            @ store res[1]
        strb            r12,[r0], #1            @ store res[2]
        strb            r9, [r0], #1            @ store res[3]

        bne             1b

        subs            lr,  lr,  #1
T       it              ne
        ldrne           r4, [sp,  #36]          @ 4-in-parallel loop counter
        add             r2,  r2,  r3            @ move to next input/output lines
        add             r0,  r0,  r1

        bne             1b

        pop            {r4 - r11, pc}
endfunc

@ void put_vp8_epel_v4(uint8_t *dst, int dststride, uint8_t *src,
@                      int srcstride, int w, int h, int my)
function ff_put_vp8_epel_v4_armv6, export=1
        push           {r4 - r11, lr}

        movrel          lr,  fourtap_filters_1324 - 4
        ldr             r12,[sp,  #44]          @ vp8_filter index
        ldr             r4, [sp,  #36]          @ width
        add             lr,  lr,  r12, lsl #2
        sub             r1,  r1,  r4            @ dst_stride - block_width
        asr             r4,  #2
        ldr             r5, [lr]
        ldr             r6, [lr,  #4]

        str             r4, [sp,  #36]          @ "4-in-parallel" loop counter @40
        str             r3, [sp,  #44]          @ src_stride @48
        push           {r1}                     @ dst_stride - block_width @36
                                                @ height @44
1:
        ldr             lr, [r2,  r3, lsl #1]   @ load source pixels
        ldr             r12,[r2,  r3]
        ldr_dpren       r7,  r2,  r3
        ldr             r11,[r2], #4

        @ byte -> word and "transpose"
        uxtb16          r8,  lr,  ror #8        @ src[3 + stride*3] | src[1 + stride*3]
        uxtb16          r9,  r12, ror #8        @ src[3 + stride*2] | src[1 + stride*2]
        uxtb16          r3,  r7,  ror #8        @ src[3 + stride*0] | src[1 + stride*0]
        uxtb16          r1,  r11, ror #8        @ src[3 + stride*1] | src[1 + stride*1]
        uxtb16          lr,  lr                 @ src[2 + stride*3] | src[0 + stride*3]
        uxtb16          r12, r12                @ src[2 + stride*2] | src[0 + stride*2]
        uxtb16          r7,  r7                 @ src[2 + stride*0] | src[0 + stride*0]
        uxtb16          r11, r11                @ src[2 + stride*1] | src[0 + stride*1]
        pkhbt           r10, r1,  r8,  lsl #16  @ src[1 + stride*3] | src[1 + stride*1]
        pkhtb           r1,  r8,  r1,  asr #16  @ src[3 + stride*3] | src[3 + stride*1]
        pkhbt           r8,  r3,  r9,  lsl #16  @ src[1 + stride*2] | src[1 + stride*0]
        pkhtb           r3,  r9,  r3,  asr #16  @ src[3 + stride*2] | src[3 + stride*0]
        pkhbt           r9,  r11, lr,  lsl #16  @ src[0 + stride*3] | src[0 + stride*1]
        pkhtb           r11, lr,  r11, asr #16  @ src[2 + stride*3] | src[2 + stride*1]
        pkhbt           lr,  r7,  r12, lsl #16  @ src[0 + stride*2] | src[0 + stride*0]
        pkhtb           r7,  r12, r7,  asr #16  @ src[2 + stride*2] | src[2 + stride*0]

        smuad           r9,  r9,  r6            @ filter[0][1]
        smuad           r10, r10, r6            @ filter[1][1]
        smuad           r11, r11, r6            @ filter[2][1]
        smuad           r1,  r1,  r6            @ filter[3][1]
        smlad           r9,  lr,  r5, r9        @ filter[0][0]
        smlad           r10, r8,  r5, r10       @ filter[1][0]
        smlad           r11, r7,  r5, r11       @ filter[2][0]
        smlad           r1,  r3,  r5, r1        @ filter[3][0]

        subs            r4,  r4,  #1            @ counter--
        ldr             r3, [sp,  #48]          @ FIXME prevent clobber of r3 above?

        add             r9,  r9,  #0x40         @ round_shift_and_clamp[0]
        add             r10, r10, #0x40         @ round_shift_and_clamp[1]
        add             r11, r11, #0x40         @ round_shift_and_clamp[2]
        add             r1,  r1,  #0x40         @ round_shift_and_clamp[3]

        usat            r9,  #8,  r9,  asr #7
        usat            r10, #8,  r10, asr #7
        usat            r11, #8,  r11, asr #7
        usat            r1,  #8,  r1,  asr #7

        strb            r9, [r0], #1            @ store result
        strb            r10,[r0], #1
        strb            r11,[r0], #1
        strb            r1, [r0], #1

        bne             1b

        ldr             r12,[sp,  #44]          @ height = outer-loop counter
        subs            r12, r12, #1
T       ittt            ne
        ldrne           r4, [sp,  #40]          @ 4-in-parallel loop counter
        ldrne           r9, [sp,  #0]
        strne           r12,[sp,  #44]
        sub             r2,  r2,  r4,  lsl #2
        add             r0,  r0,  r9
        add             r2,  r2,  r3            @ move to next input/output lines

        bne             1b

        add             sp,  sp,  #4            @ restore stack after push{r1} above
        pop            {r4 - r11, pc}
endfunc

@ void put_vp8_bilin_h(uint8_t *dst, int dststride, uint8_t *src,
@                      int srcstride, int w, int h, int mx)
function ff_put_vp8_bilin_h_armv6, export=1
        push           {r4 - r9,  lr}

        ldr             r8, [sp,  #36]          @ vp8_filter index
        ldr             r12,[sp,  #32]          @ height = outer-loop counter
        ldr             r4, [sp,  #28]          @ width
        lsl             r5,  r8,  #16           @ mx << 16
        sub             r3,  r3,  r4            @ src_stride - block_width
        sub             r1,  r1,  r4            @ dst_stride - block_width
        asr             r4,  #2
        sub             r5,  r5,  r8            @ (mx << 16) | (-mx)
        str             r4, [sp,  #28]          @ "4-in-parallel" loop counter
        add             r5,  r5,  #8            @ (8 - mx) | (mx << 16) = filter coefficients
1:
        ldrb            r6, [r2], #1            @ load source data
        ldrb            r7, [r2], #1
        ldrb            r8, [r2], #1
        ldrb            r9, [r2], #1
        ldrb            lr, [r2]

        pkhbt           r6,  r6,  r7,  lsl #16  @ src[1] | src[0]
        pkhbt           r7,  r7,  r8,  lsl #16  @ src[2] | src[1]
        pkhbt           r8,  r8,  r9,  lsl #16  @ src[3] | src[2]
        pkhbt           r9,  r9,  lr,  lsl #16  @ src[4] | src[3]

        smuad           r6,  r6,  r5            @ apply the filter
        smuad           r7,  r7,  r5
        smuad           r8,  r8,  r5
        smuad           r9,  r9,  r5

        subs            r4,  r4,  #1            @ counter--

        add             r6,  r6,  #0x4          @ round_shift_and_clamp
        add             r7,  r7,  #0x4
        add             r8,  r8,  #0x4
        add             r9,  r9,  #0x4

        asr             r6,  #3
        asr             r7,  #3
        pkhbt           r6,  r6,  r8,  lsl #13
        pkhbt           r7,  r7,  r9,  lsl #13
        orr             r6,  r6,  r7,  lsl #8
        str             r6, [r0], #4            @ store result

        bne             1b

        ldr             r4, [sp,  #28]          @ 4-in-parallel loop counter
        subs            r12, r12, #1

        add             r2,  r2,  r3            @ move to next input/output lines
        add             r0,  r0,  r1

        bne             1b

        pop            {r4 - r9,  pc}
endfunc

@ void put_vp8_bilin_v(uint8_t *dst, int dststride, uint8_t *src,
@                      int srcstride, int w, int h, int my)
function ff_put_vp8_bilin_v_armv6, export=1
        push           {r4 - r11, lr}

        ldr             r11,[sp,  #44]          @ vp8_filter index
        ldr             r4, [sp,  #36]          @ width
        mov             r5,  r11, lsl #16       @ mx << 16
        ldr             r12,[sp,  #40]          @ height = outer-loop counter
        sub             r1,  r1,  r4
        sub             r5,  r5,  r11           @ (mx << 16) | (-mx)
        asr             r4,  #2
        add             r5,  r5,  #8            @ (8 - mx) | (mx << 16) = filter coefficients
        str             r4, [sp,  #36]          @ "4-in-parallel" loop counter
1:
        ldrb            r10,[r2,  r3]           @ load the data
        ldrb            r6, [r2], #1
        ldrb            r11,[r2,  r3]
        ldrb            r7, [r2], #1
        ldrb            lr, [r2,  r3]
        ldrb            r8, [r2], #1
        ldrb            r9, [r2,  r3]
        pkhbt           r6,  r6,  r10, lsl #16
        ldrb            r10,[r2], #1
        pkhbt           r7,  r7,  r11, lsl #16
        pkhbt           r8,  r8,  lr,  lsl #16
        pkhbt           r9,  r10, r9,  lsl #16

        smuad           r6,  r6,  r5            @ apply the filter
        smuad           r7,  r7,  r5
        smuad           r8,  r8,  r5
        smuad           r9,  r9,  r5

        subs            r4,  r4,  #1            @ counter--

        add             r6,  r6,  #0x4          @ round_shift_and_clamp
        add             r7,  r7,  #0x4
        add             r8,  r8,  #0x4
        add             r9,  r9,  #0x4

        asr             r6,  #3
        asr             r7,  #3
        pkhbt           r6,  r6,  r8,  lsl #13
        pkhbt           r7,  r7,  r9,  lsl #13
        orr             r6,  r6,  r7,  lsl #8
        str             r6, [r0], #4            @ store result

        bne             1b

        ldr             r4, [sp,  #36]          @ 4-in-parallel loop counter
        subs            r12, r12, #1

        add             r2,  r2,  r3            @ move to next input/output lines
        add             r0,  r0,  r1
        sub             r2,  r2,  r4,  lsl #2

        bne             1b
        pop            {r4 - r11, pc}
endfunc