summaryrefslogtreecommitdiffstats
path: root/third_party/aom/aom_dsp/mips/macros_msa.h
blob: 48fbcfd47f09aba8f768ce2d8e3ce795b9c5f621 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
/*
 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
 *
 * This source code is subject to the terms of the BSD 2 Clause License and
 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
 * was not distributed with this source code in the LICENSE file, you can
 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
 * Media Patent License 1.0 was not distributed with this source code in the
 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
 */

#ifndef AOM_DSP_MIPS_MACROS_MSA_H_
#define AOM_DSP_MIPS_MACROS_MSA_H_

#include <msa.h>

#include "./aom_config.h"
#include "aom/aom_integer.h"

#define LD_B(RTYPE, psrc) *((const RTYPE *)(psrc))
#define LD_UB(...) LD_B(v16u8, __VA_ARGS__)
#define LD_SB(...) LD_B(v16i8, __VA_ARGS__)

#define LD_H(RTYPE, psrc) *((const RTYPE *)(psrc))
#define LD_UH(...) LD_H(v8u16, __VA_ARGS__)
#define LD_SH(...) LD_H(v8i16, __VA_ARGS__)

#define LD_W(RTYPE, psrc) *((const RTYPE *)(psrc))
#define LD_SW(...) LD_W(v4i32, __VA_ARGS__)

#define ST_B(RTYPE, in, pdst) *((RTYPE *)(pdst)) = (in)
#define ST_UB(...) ST_B(v16u8, __VA_ARGS__)
#define ST_SB(...) ST_B(v16i8, __VA_ARGS__)

#define ST_H(RTYPE, in, pdst) *((RTYPE *)(pdst)) = (in)
#define ST_SH(...) ST_H(v8i16, __VA_ARGS__)

#define ST_W(RTYPE, in, pdst) *((RTYPE *)(pdst)) = (in)
#define ST_SW(...) ST_W(v4i32, __VA_ARGS__)

#if (__mips_isa_rev >= 6)
#define LH(psrc)                                          \
  ({                                                      \
    const uint8_t *psrc_m = (const uint8_t *)(psrc);      \
    uint16_t val_m;                                       \
                                                          \
    __asm__ __volatile__("lh  %[val_m],  %[psrc_m]  \n\t" \
                                                          \
                         : [val_m] "=r"(val_m)            \
                         : [psrc_m] "m"(*psrc_m));        \
                                                          \
    val_m;                                                \
  })

#define LW(psrc)                                          \
  ({                                                      \
    const uint8_t *psrc_m = (const uint8_t *)(psrc);      \
    uint32_t val_m;                                       \
                                                          \
    __asm__ __volatile__("lw  %[val_m],  %[psrc_m]  \n\t" \
                                                          \
                         : [val_m] "=r"(val_m)            \
                         : [psrc_m] "m"(*psrc_m));        \
                                                          \
    val_m;                                                \
  })

#if (__mips == 64)
#define LD(psrc)                                          \
  ({                                                      \
    const uint8_t *psrc_m = (const uint8_t *)(psrc);      \
    uint64_t val_m = 0;                                   \
                                                          \
    __asm__ __volatile__("ld  %[val_m],  %[psrc_m]  \n\t" \
                                                          \
                         : [val_m] "=r"(val_m)            \
                         : [psrc_m] "m"(*psrc_m));        \
                                                          \
    val_m;                                                \
  })
#else  // !(__mips == 64)
#define LD(psrc)                                            \
  ({                                                        \
    const uint8_t *psrc_m = (const uint8_t *)(psrc);        \
    uint32_t val0_m, val1_m;                                \
    uint64_t val_m = 0;                                     \
                                                            \
    val0_m = LW(psrc_m);                                    \
    val1_m = LW(psrc_m + 4);                                \
                                                            \
    val_m = (uint64_t)(val1_m);                             \
    val_m = (uint64_t)((val_m << 32) & 0xFFFFFFFF00000000); \
    val_m = (uint64_t)(val_m | (uint64_t)val0_m);           \
                                                            \
    val_m;                                                  \
  })
#endif  // (__mips == 64)

#define SH(val, pdst)                                     \
  {                                                       \
    uint8_t *pdst_m = (uint8_t *)(pdst);                  \
    const uint16_t val_m = (val);                         \
                                                          \
    __asm__ __volatile__("sh  %[val_m],  %[pdst_m]  \n\t" \
                                                          \
                         : [pdst_m] "=m"(*pdst_m)         \
                         : [val_m] "r"(val_m));           \
  }

#define SW(val, pdst)                                     \
  {                                                       \
    uint8_t *pdst_m = (uint8_t *)(pdst);                  \
    const uint32_t val_m = (val);                         \
                                                          \
    __asm__ __volatile__("sw  %[val_m],  %[pdst_m]  \n\t" \
                                                          \
                         : [pdst_m] "=m"(*pdst_m)         \
                         : [val_m] "r"(val_m));           \
  }

#define SD(val, pdst)                                     \
  {                                                       \
    uint8_t *pdst_m = (uint8_t *)(pdst);                  \
    const uint64_t val_m = (val);                         \
                                                          \
    __asm__ __volatile__("sd  %[val_m],  %[pdst_m]  \n\t" \
                                                          \
                         : [pdst_m] "=m"(*pdst_m)         \
                         : [val_m] "r"(val_m));           \
  }
#else  // !(__mips_isa_rev >= 6)
#define LH(psrc)                                           \
  ({                                                       \
    const uint8_t *psrc_m = (const uint8_t *)(psrc);       \
    uint16_t val_m;                                        \
                                                           \
    __asm__ __volatile__("ulh  %[val_m],  %[psrc_m]  \n\t" \
                                                           \
                         : [val_m] "=r"(val_m)             \
                         : [psrc_m] "m"(*psrc_m));         \
                                                           \
    val_m;                                                 \
  })

#define LW(psrc)                                           \
  ({                                                       \
    const uint8_t *psrc_m = (const uint8_t *)(psrc);       \
    uint32_t val_m;                                        \
                                                           \
    __asm__ __volatile__("ulw  %[val_m],  %[psrc_m]  \n\t" \
                                                           \
                         : [val_m] "=r"(val_m)             \
                         : [psrc_m] "m"(*psrc_m));         \
                                                           \
    val_m;                                                 \
  })

#if (__mips == 64)
#define LD(psrc)                                           \
  ({                                                       \
    const uint8_t *psrc_m = (const uint8_t *)(psrc);       \
    uint64_t val_m = 0;                                    \
                                                           \
    __asm__ __volatile__("uld  %[val_m],  %[psrc_m]  \n\t" \
                                                           \
                         : [val_m] "=r"(val_m)             \
                         : [psrc_m] "m"(*psrc_m));         \
                                                           \
    val_m;                                                 \
  })
#else  // !(__mips == 64)
#define LD(psrc)                                                              \
  ({                                                                          \
    const uint8_t *psrc_m1 = (const uint8_t *)(psrc);                         \
    uint32_t val0_m, val1_m;                                                  \
    uint64_t val_m_combined = 0;                                              \
                                                                              \
    val0_m = LW(psrc_m1);                                                     \
    val1_m = LW(psrc_m1 + 4);                                                 \
                                                                              \
    val_m_combined = (uint64_t)(val1_m);                                      \
    val_m_combined = (uint64_t)((val_m_combined << 32) & 0xFFFFFFFF00000000); \
    val_m_combined = (uint64_t)(val_m_combined | (uint64_t)val0_m);           \
                                                                              \
    val_m_combined;                                                           \
  })
#endif  // (__mips == 64)

#define SH(val, pdst)                                      \
  {                                                        \
    uint8_t *pdst_m = (uint8_t *)(pdst);                   \
    const uint16_t val_m = (val);                          \
                                                           \
    __asm__ __volatile__("ush  %[val_m],  %[pdst_m]  \n\t" \
                                                           \
                         : [pdst_m] "=m"(*pdst_m)          \
                         : [val_m] "r"(val_m));            \
  }

#define SW(val, pdst)                                      \
  {                                                        \
    uint8_t *pdst_m = (uint8_t *)(pdst);                   \
    const uint32_t val_m = (val);                          \
                                                           \
    __asm__ __volatile__("usw  %[val_m],  %[pdst_m]  \n\t" \
                                                           \
                         : [pdst_m] "=m"(*pdst_m)          \
                         : [val_m] "r"(val_m));            \
  }

#define SD(val, pdst)                                        \
  {                                                          \
    uint8_t *pdst_m1 = (uint8_t *)(pdst);                    \
    uint32_t val0_m, val1_m;                                 \
                                                             \
    val0_m = (uint32_t)((val)&0x00000000FFFFFFFF);           \
    val1_m = (uint32_t)(((val) >> 32) & 0x00000000FFFFFFFF); \
                                                             \
    SW(val0_m, pdst_m1);                                     \
    SW(val1_m, pdst_m1 + 4);                                 \
  }
#endif  // (__mips_isa_rev >= 6)

/* Description : Load 4 words with stride
   Arguments   : Inputs  - psrc, stride
                 Outputs - out0, out1, out2, out3
   Details     : Load word in 'out0' from (psrc)
                 Load word in 'out1' from (psrc + stride)
                 Load word in 'out2' from (psrc + 2 * stride)
                 Load word in 'out3' from (psrc + 3 * stride)
*/
#define LW4(psrc, stride, out0, out1, out2, out3) \
  {                                               \
    out0 = LW((psrc));                            \
    out1 = LW((psrc) + stride);                   \
    out2 = LW((psrc) + 2 * stride);               \
    out3 = LW((psrc) + 3 * stride);               \
  }

/* Description : Load double words with stride
   Arguments   : Inputs  - psrc, stride
                 Outputs - out0, out1
   Details     : Load double word in 'out0' from (psrc)
                 Load double word in 'out1' from (psrc + stride)
*/
#define LD2(psrc, stride, out0, out1) \
  {                                   \
    out0 = LD((psrc));                \
    out1 = LD((psrc) + stride);       \
  }
#define LD4(psrc, stride, out0, out1, out2, out3) \
  {                                               \
    LD2((psrc), stride, out0, out1);              \
    LD2((psrc) + 2 * stride, stride, out2, out3); \
  }

/* Description : Store 4 words with stride
   Arguments   : Inputs - in0, in1, in2, in3, pdst, stride
   Details     : Store word from 'in0' to (pdst)
                 Store word from 'in1' to (pdst + stride)
                 Store word from 'in2' to (pdst + 2 * stride)
                 Store word from 'in3' to (pdst + 3 * stride)
*/
#define SW4(in0, in1, in2, in3, pdst, stride) \
  {                                           \
    SW(in0, (pdst))                           \
    SW(in1, (pdst) + stride);                 \
    SW(in2, (pdst) + 2 * stride);             \
    SW(in3, (pdst) + 3 * stride);             \
  }

/* Description : Store 4 double words with stride
   Arguments   : Inputs - in0, in1, in2, in3, pdst, stride
   Details     : Store double word from 'in0' to (pdst)
                 Store double word from 'in1' to (pdst + stride)
                 Store double word from 'in2' to (pdst + 2 * stride)
                 Store double word from 'in3' to (pdst + 3 * stride)
*/
#define SD4(in0, in1, in2, in3, pdst, stride) \
  {                                           \
    SD(in0, (pdst))                           \
    SD(in1, (pdst) + stride);                 \
    SD(in2, (pdst) + 2 * stride);             \
    SD(in3, (pdst) + 3 * stride);             \
  }

/* Description : Load vectors with 16 byte elements with stride
   Arguments   : Inputs  - psrc, stride
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Load 16 byte elements in 'out0' from (psrc)
                 Load 16 byte elements in 'out1' from (psrc + stride)
*/
#define LD_B2(RTYPE, psrc, stride, out0, out1) \
  {                                            \
    out0 = LD_B(RTYPE, (psrc));                \
    out1 = LD_B(RTYPE, (psrc) + stride);       \
  }
#define LD_UB2(...) LD_B2(v16u8, __VA_ARGS__)
#define LD_SB2(...) LD_B2(v16i8, __VA_ARGS__)

#define LD_B3(RTYPE, psrc, stride, out0, out1, out2) \
  {                                                  \
    LD_B2(RTYPE, (psrc), stride, out0, out1);        \
    out2 = LD_B(RTYPE, (psrc) + 2 * stride);         \
  }
#define LD_UB3(...) LD_B3(v16u8, __VA_ARGS__)

#define LD_B4(RTYPE, psrc, stride, out0, out1, out2, out3) \
  {                                                        \
    LD_B2(RTYPE, (psrc), stride, out0, out1);              \
    LD_B2(RTYPE, (psrc) + 2 * stride, stride, out2, out3); \
  }
#define LD_UB4(...) LD_B4(v16u8, __VA_ARGS__)
#define LD_SB4(...) LD_B4(v16i8, __VA_ARGS__)

#define LD_B5(RTYPE, psrc, stride, out0, out1, out2, out3, out4) \
  {                                                              \
    LD_B4(RTYPE, (psrc), stride, out0, out1, out2, out3);        \
    out4 = LD_B(RTYPE, (psrc) + 4 * stride);                     \
  }
#define LD_UB5(...) LD_B5(v16u8, __VA_ARGS__)
#define LD_SB5(...) LD_B5(v16i8, __VA_ARGS__)

#define LD_B7(RTYPE, psrc, stride, out0, out1, out2, out3, out4, out5, out6) \
  {                                                                          \
    LD_B5(RTYPE, (psrc), stride, out0, out1, out2, out3, out4);              \
    LD_B2(RTYPE, (psrc) + 5 * stride, stride, out5, out6);                   \
  }
#define LD_SB7(...) LD_B7(v16i8, __VA_ARGS__)

#define LD_B8(RTYPE, psrc, stride, out0, out1, out2, out3, out4, out5, out6, \
              out7)                                                          \
  {                                                                          \
    LD_B4(RTYPE, (psrc), stride, out0, out1, out2, out3);                    \
    LD_B4(RTYPE, (psrc) + 4 * stride, stride, out4, out5, out6, out7);       \
  }
#define LD_UB8(...) LD_B8(v16u8, __VA_ARGS__)
#define LD_SB8(...) LD_B8(v16i8, __VA_ARGS__)

/* Description : Load vectors with 8 halfword elements with stride
   Arguments   : Inputs  - psrc, stride
                 Outputs - out0, out1
   Details     : Load 8 halfword elements in 'out0' from (psrc)
                 Load 8 halfword elements in 'out1' from (psrc + stride)
*/
#define LD_H2(RTYPE, psrc, stride, out0, out1) \
  {                                            \
    out0 = LD_H(RTYPE, (psrc));                \
    out1 = LD_H(RTYPE, (psrc) + (stride));     \
  }
#define LD_SH2(...) LD_H2(v8i16, __VA_ARGS__)

#define LD_H4(RTYPE, psrc, stride, out0, out1, out2, out3) \
  {                                                        \
    LD_H2(RTYPE, (psrc), stride, out0, out1);              \
    LD_H2(RTYPE, (psrc) + 2 * stride, stride, out2, out3); \
  }
#define LD_SH4(...) LD_H4(v8i16, __VA_ARGS__)

#define LD_H8(RTYPE, psrc, stride, out0, out1, out2, out3, out4, out5, out6, \
              out7)                                                          \
  {                                                                          \
    LD_H4(RTYPE, (psrc), stride, out0, out1, out2, out3);                    \
    LD_H4(RTYPE, (psrc) + 4 * stride, stride, out4, out5, out6, out7);       \
  }
#define LD_SH8(...) LD_H8(v8i16, __VA_ARGS__)

#define LD_H16(RTYPE, psrc, stride, out0, out1, out2, out3, out4, out5, out6,  \
               out7, out8, out9, out10, out11, out12, out13, out14, out15)     \
  {                                                                            \
    LD_H8(RTYPE, (psrc), stride, out0, out1, out2, out3, out4, out5, out6,     \
          out7);                                                               \
    LD_H8(RTYPE, (psrc) + 8 * stride, stride, out8, out9, out10, out11, out12, \
          out13, out14, out15);                                                \
  }
#define LD_SH16(...) LD_H16(v8i16, __VA_ARGS__)

/* Description : Load 4x4 block of signed halfword elements from 1D source
                 data into 4 vectors (Each vector with 4 signed halfwords)
   Arguments   : Input   - psrc
                 Outputs - out0, out1, out2, out3
*/
#define LD4x4_SH(psrc, out0, out1, out2, out3)            \
  {                                                       \
    out0 = LD_SH(psrc);                                   \
    out2 = LD_SH(psrc + 8);                               \
    out1 = (v8i16)__msa_ilvl_d((v2i64)out0, (v2i64)out0); \
    out3 = (v8i16)__msa_ilvl_d((v2i64)out2, (v2i64)out2); \
  }

/* Description : Load 2 vectors of signed word elements with stride
   Arguments   : Inputs  - psrc, stride
                 Outputs - out0, out1
                 Return Type - signed word
*/
#define LD_SW2(psrc, stride, out0, out1) \
  {                                      \
    out0 = LD_SW((psrc));                \
    out1 = LD_SW((psrc) + stride);       \
  }

/* Description : Store vectors of 16 byte elements with stride
   Arguments   : Inputs - in0, in1, pdst, stride
   Details     : Store 16 byte elements from 'in0' to (pdst)
                 Store 16 byte elements from 'in1' to (pdst + stride)
*/
#define ST_B2(RTYPE, in0, in1, pdst, stride) \
  {                                          \
    ST_B(RTYPE, in0, (pdst));                \
    ST_B(RTYPE, in1, (pdst) + stride);       \
  }
#define ST_UB2(...) ST_B2(v16u8, __VA_ARGS__)

#define ST_B4(RTYPE, in0, in1, in2, in3, pdst, stride)   \
  {                                                      \
    ST_B2(RTYPE, in0, in1, (pdst), stride);              \
    ST_B2(RTYPE, in2, in3, (pdst) + 2 * stride, stride); \
  }
#define ST_UB4(...) ST_B4(v16u8, __VA_ARGS__)

#define ST_B8(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, pdst, stride) \
  {                                                                        \
    ST_B4(RTYPE, in0, in1, in2, in3, pdst, stride);                        \
    ST_B4(RTYPE, in4, in5, in6, in7, (pdst) + 4 * stride, stride);         \
  }
#define ST_UB8(...) ST_B8(v16u8, __VA_ARGS__)

/* Description : Store vectors of 8 halfword elements with stride
   Arguments   : Inputs - in0, in1, pdst, stride
   Details     : Store 8 halfword elements from 'in0' to (pdst)
                 Store 8 halfword elements from 'in1' to (pdst + stride)
*/
#define ST_H2(RTYPE, in0, in1, pdst, stride) \
  {                                          \
    ST_H(RTYPE, in0, (pdst));                \
    ST_H(RTYPE, in1, (pdst) + stride);       \
  }
#define ST_SH2(...) ST_H2(v8i16, __VA_ARGS__)

#define ST_H4(RTYPE, in0, in1, in2, in3, pdst, stride)   \
  {                                                      \
    ST_H2(RTYPE, in0, in1, (pdst), stride);              \
    ST_H2(RTYPE, in2, in3, (pdst) + 2 * stride, stride); \
  }
#define ST_SH4(...) ST_H4(v8i16, __VA_ARGS__)

#define ST_H8(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, pdst, stride) \
  {                                                                        \
    ST_H4(RTYPE, in0, in1, in2, in3, (pdst), stride);                      \
    ST_H4(RTYPE, in4, in5, in6, in7, (pdst) + 4 * stride, stride);         \
  }
#define ST_SH8(...) ST_H8(v8i16, __VA_ARGS__)

/* Description : Store vectors of word elements with stride
   Arguments   : Inputs - in0, in1, pdst, stride
   Details     : Store 4 word elements from 'in0' to (pdst)
                 Store 4 word elements from 'in1' to (pdst + stride)
*/
#define ST_SW2(in0, in1, pdst, stride) \
  {                                    \
    ST_SW(in0, (pdst));                \
    ST_SW(in1, (pdst) + stride);       \
  }

/* Description : Store 2x4 byte block to destination memory from input vector
   Arguments   : Inputs - in, stidx, pdst, stride
   Details     : Index 'stidx' halfword element from 'in' vector is copied to
                 the GP register and stored to (pdst)
                 Index 'stidx+1' halfword element from 'in' vector is copied to
                 the GP register and stored to (pdst + stride)
                 Index 'stidx+2' halfword element from 'in' vector is copied to
                 the GP register and stored to (pdst + 2 * stride)
                 Index 'stidx+3' halfword element from 'in' vector is copied to
                 the GP register and stored to (pdst + 3 * stride)
*/
#define ST2x4_UB(in, stidx, pdst, stride)            \
  {                                                  \
    uint16_t out0_m, out1_m, out2_m, out3_m;         \
    uint8_t *pblk_2x4_m = (uint8_t *)(pdst);         \
                                                     \
    out0_m = __msa_copy_u_h((v8i16)in, (stidx));     \
    out1_m = __msa_copy_u_h((v8i16)in, (stidx + 1)); \
    out2_m = __msa_copy_u_h((v8i16)in, (stidx + 2)); \
    out3_m = __msa_copy_u_h((v8i16)in, (stidx + 3)); \
                                                     \
    SH(out0_m, pblk_2x4_m);                          \
    SH(out1_m, pblk_2x4_m + stride);                 \
    SH(out2_m, pblk_2x4_m + 2 * stride);             \
    SH(out3_m, pblk_2x4_m + 3 * stride);             \
  }

/* Description : Store 4x2 byte block to destination memory from input vector
   Arguments   : Inputs - in, pdst, stride
   Details     : Index 0 word element from 'in' vector is copied to the GP
                 register and stored to (pdst)
                 Index 1 word element from 'in' vector is copied to the GP
                 register and stored to (pdst + stride)
*/
#define ST4x2_UB(in, pdst, stride)           \
  {                                          \
    uint32_t out0_m, out1_m;                 \
    uint8_t *pblk_4x2_m = (uint8_t *)(pdst); \
                                             \
    out0_m = __msa_copy_u_w((v4i32)in, 0);   \
    out1_m = __msa_copy_u_w((v4i32)in, 1);   \
                                             \
    SW(out0_m, pblk_4x2_m);                  \
    SW(out1_m, pblk_4x2_m + stride);         \
  }

/* Description : Store 4x4 byte block to destination memory from input vector
   Arguments   : Inputs - in0, in1, pdst, stride
   Details     : 'Idx0' word element from input vector 'in0' is copied to the
                 GP register and stored to (pdst)
                 'Idx1' word element from input vector 'in0' is copied to the
                 GP register and stored to (pdst + stride)
                 'Idx2' word element from input vector 'in0' is copied to the
                 GP register and stored to (pdst + 2 * stride)
                 'Idx3' word element from input vector 'in0' is copied to the
                 GP register and stored to (pdst + 3 * stride)
*/
#define ST4x4_UB(in0, in1, idx0, idx1, idx2, idx3, pdst, stride) \
  {                                                              \
    uint32_t out0_m, out1_m, out2_m, out3_m;                     \
    uint8_t *pblk_4x4_m = (uint8_t *)(pdst);                     \
                                                                 \
    out0_m = __msa_copy_u_w((v4i32)in0, idx0);                   \
    out1_m = __msa_copy_u_w((v4i32)in0, idx1);                   \
    out2_m = __msa_copy_u_w((v4i32)in1, idx2);                   \
    out3_m = __msa_copy_u_w((v4i32)in1, idx3);                   \
                                                                 \
    SW4(out0_m, out1_m, out2_m, out3_m, pblk_4x4_m, stride);     \
  }
#define ST4x8_UB(in0, in1, pdst, stride)                           \
  {                                                                \
    uint8_t *pblk_4x8 = (uint8_t *)(pdst);                         \
                                                                   \
    ST4x4_UB(in0, in0, 0, 1, 2, 3, pblk_4x8, stride);              \
    ST4x4_UB(in1, in1, 0, 1, 2, 3, pblk_4x8 + 4 * stride, stride); \
  }

/* Description : Store 8x1 byte block to destination memory from input vector
   Arguments   : Inputs - in, pdst
   Details     : Index 0 double word element from 'in' vector is copied to the
                 GP register and stored to (pdst)
*/
#define ST8x1_UB(in, pdst)                 \
  {                                        \
    uint64_t out0_m;                       \
                                           \
    out0_m = __msa_copy_u_d((v2i64)in, 0); \
    SD(out0_m, pdst);                      \
  }

/* Description : Store 8x2 byte block to destination memory from input vector
   Arguments   : Inputs - in, pdst, stride
   Details     : Index 0 double word element from 'in' vector is copied to the
                 GP register and stored to (pdst)
                 Index 1 double word element from 'in' vector is copied to the
                 GP register and stored to (pdst + stride)
*/
#define ST8x2_UB(in, pdst, stride)           \
  {                                          \
    uint64_t out0_m, out1_m;                 \
    uint8_t *pblk_8x2_m = (uint8_t *)(pdst); \
                                             \
    out0_m = __msa_copy_u_d((v2i64)in, 0);   \
    out1_m = __msa_copy_u_d((v2i64)in, 1);   \
                                             \
    SD(out0_m, pblk_8x2_m);                  \
    SD(out1_m, pblk_8x2_m + stride);         \
  }

/* Description : Store 8x4 byte block to destination memory from input
                 vectors
   Arguments   : Inputs - in0, in1, pdst, stride
   Details     : Index 0 double word element from 'in0' vector is copied to the
                 GP register and stored to (pdst)
                 Index 1 double word element from 'in0' vector is copied to the
                 GP register and stored to (pdst + stride)
                 Index 0 double word element from 'in1' vector is copied to the
                 GP register and stored to (pdst + 2 * stride)
                 Index 1 double word element from 'in1' vector is copied to the
                 GP register and stored to (pdst + 3 * stride)
*/
#define ST8x4_UB(in0, in1, pdst, stride)                     \
  {                                                          \
    uint64_t out0_m, out1_m, out2_m, out3_m;                 \
    uint8_t *pblk_8x4_m = (uint8_t *)(pdst);                 \
                                                             \
    out0_m = __msa_copy_u_d((v2i64)in0, 0);                  \
    out1_m = __msa_copy_u_d((v2i64)in0, 1);                  \
    out2_m = __msa_copy_u_d((v2i64)in1, 0);                  \
    out3_m = __msa_copy_u_d((v2i64)in1, 1);                  \
                                                             \
    SD4(out0_m, out1_m, out2_m, out3_m, pblk_8x4_m, stride); \
  }

/* Description : average with rounding (in0 + in1 + 1) / 2.
   Arguments   : Inputs  - in0, in1, in2, in3,
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Each unsigned byte element from 'in0' vector is added with
                 each unsigned byte element from 'in1' vector. Then the average
                 with rounding is calculated and written to 'out0'
*/
#define AVER_UB2(RTYPE, in0, in1, in2, in3, out0, out1)   \
  {                                                       \
    out0 = (RTYPE)__msa_aver_u_b((v16u8)in0, (v16u8)in1); \
    out1 = (RTYPE)__msa_aver_u_b((v16u8)in2, (v16u8)in3); \
  }
#define AVER_UB2_UB(...) AVER_UB2(v16u8, __VA_ARGS__)

#define AVER_UB4(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, \
                 out2, out3)                                                \
  {                                                                         \
    AVER_UB2(RTYPE, in0, in1, in2, in3, out0, out1)                         \
    AVER_UB2(RTYPE, in4, in5, in6, in7, out2, out3)                         \
  }
#define AVER_UB4_UB(...) AVER_UB4(v16u8, __VA_ARGS__)

/* Description : Immediate number of elements to slide with zero
   Arguments   : Inputs  - in0, in1, slide_val
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Byte elements from 'zero_m' vector are slid into 'in0' by
                 value specified in the 'slide_val'
*/
#define SLDI_B2_0(RTYPE, in0, in1, out0, out1, slide_val)             \
  {                                                                   \
    v16i8 zero_m = { 0 };                                             \
    out0 = (RTYPE)__msa_sldi_b((v16i8)zero_m, (v16i8)in0, slide_val); \
    out1 = (RTYPE)__msa_sldi_b((v16i8)zero_m, (v16i8)in1, slide_val); \
  }
#define SLDI_B2_0_SW(...) SLDI_B2_0(v4i32, __VA_ARGS__)

#define SLDI_B4_0(RTYPE, in0, in1, in2, in3, out0, out1, out2, out3, \
                  slide_val)                                         \
  {                                                                  \
    SLDI_B2_0(RTYPE, in0, in1, out0, out1, slide_val);               \
    SLDI_B2_0(RTYPE, in2, in3, out2, out3, slide_val);               \
  }
#define SLDI_B4_0_UB(...) SLDI_B4_0(v16u8, __VA_ARGS__)

/* Description : Immediate number of elements to slide
   Arguments   : Inputs  - in0_0, in0_1, in1_0, in1_1, slide_val
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Byte elements from 'in0_0' vector are slid into 'in1_0' by
                 value specified in the 'slide_val'
*/
#define SLDI_B2(RTYPE, in0_0, in0_1, in1_0, in1_1, out0, out1, slide_val) \
  {                                                                       \
    out0 = (RTYPE)__msa_sldi_b((v16i8)in0_0, (v16i8)in1_0, slide_val);    \
    out1 = (RTYPE)__msa_sldi_b((v16i8)in0_1, (v16i8)in1_1, slide_val);    \
  }
#define SLDI_B2_UB(...) SLDI_B2(v16u8, __VA_ARGS__)
#define SLDI_B2_SH(...) SLDI_B2(v8i16, __VA_ARGS__)

#define SLDI_B3(RTYPE, in0_0, in0_1, in0_2, in1_0, in1_1, in1_2, out0, out1, \
                out2, slide_val)                                             \
  {                                                                          \
    SLDI_B2(RTYPE, in0_0, in0_1, in1_0, in1_1, out0, out1, slide_val)        \
    out2 = (RTYPE)__msa_sldi_b((v16i8)in0_2, (v16i8)in1_2, slide_val);       \
  }
#define SLDI_B3_SB(...) SLDI_B3(v16i8, __VA_ARGS__)
#define SLDI_B3_UH(...) SLDI_B3(v8u16, __VA_ARGS__)

/* Description : Shuffle byte vector elements as per mask vector
   Arguments   : Inputs  - in0, in1, in2, in3, mask0, mask1
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Byte elements from 'in0' & 'in1' are copied selectively to
                 'out0' as per control vector 'mask0'
*/
#define VSHF_B2(RTYPE, in0, in1, in2, in3, mask0, mask1, out0, out1)  \
  {                                                                   \
    out0 = (RTYPE)__msa_vshf_b((v16i8)mask0, (v16i8)in1, (v16i8)in0); \
    out1 = (RTYPE)__msa_vshf_b((v16i8)mask1, (v16i8)in3, (v16i8)in2); \
  }
#define VSHF_B2_UB(...) VSHF_B2(v16u8, __VA_ARGS__)
#define VSHF_B2_SB(...) VSHF_B2(v16i8, __VA_ARGS__)
#define VSHF_B2_UH(...) VSHF_B2(v8u16, __VA_ARGS__)

#define VSHF_B4(RTYPE, in0, in1, mask0, mask1, mask2, mask3, out0, out1, out2, \
                out3)                                                          \
  {                                                                            \
    VSHF_B2(RTYPE, in0, in1, in0, in1, mask0, mask1, out0, out1);              \
    VSHF_B2(RTYPE, in0, in1, in0, in1, mask2, mask3, out2, out3);              \
  }
#define VSHF_B4_SB(...) VSHF_B4(v16i8, __VA_ARGS__)
#define VSHF_B4_SH(...) VSHF_B4(v8i16, __VA_ARGS__)

/* Description : Dot product of byte vector elements
   Arguments   : Inputs  - mult0, mult1, cnst0, cnst1
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Unsigned byte elements from 'mult0' are multiplied with
                 unsigned byte elements from 'cnst0' producing a result
                 twice the size of input i.e. unsigned halfword.
                 The multiplication result of adjacent odd-even elements
                 are added together and written to the 'out0' vector
*/
#define DOTP_UB2(RTYPE, mult0, mult1, cnst0, cnst1, out0, out1) \
  {                                                             \
    out0 = (RTYPE)__msa_dotp_u_h((v16u8)mult0, (v16u8)cnst0);   \
    out1 = (RTYPE)__msa_dotp_u_h((v16u8)mult1, (v16u8)cnst1);   \
  }
#define DOTP_UB2_UH(...) DOTP_UB2(v8u16, __VA_ARGS__)

#define DOTP_UB4(RTYPE, mult0, mult1, mult2, mult3, cnst0, cnst1, cnst2, \
                 cnst3, out0, out1, out2, out3)                          \
  {                                                                      \
    DOTP_UB2(RTYPE, mult0, mult1, cnst0, cnst1, out0, out1);             \
    DOTP_UB2(RTYPE, mult2, mult3, cnst2, cnst3, out2, out3);             \
  }
#define DOTP_UB4_UH(...) DOTP_UB4(v8u16, __VA_ARGS__)

/* Description : Dot product of byte vector elements
   Arguments   : Inputs  - mult0, mult1, cnst0, cnst1
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Signed byte elements from 'mult0' are multiplied with
                 signed byte elements from 'cnst0' producing a result
                 twice the size of input i.e. signed halfword.
                 The multiplication result of adjacent odd-even elements
                 are added together and written to the 'out0' vector
*/
#define DOTP_SB2(RTYPE, mult0, mult1, cnst0, cnst1, out0, out1) \
  {                                                             \
    out0 = (RTYPE)__msa_dotp_s_h((v16i8)mult0, (v16i8)cnst0);   \
    out1 = (RTYPE)__msa_dotp_s_h((v16i8)mult1, (v16i8)cnst1);   \
  }
#define DOTP_SB2_SH(...) DOTP_SB2(v8i16, __VA_ARGS__)

#define DOTP_SB4(RTYPE, mult0, mult1, mult2, mult3, cnst0, cnst1, cnst2, \
                 cnst3, out0, out1, out2, out3)                          \
  {                                                                      \
    DOTP_SB2(RTYPE, mult0, mult1, cnst0, cnst1, out0, out1);             \
    DOTP_SB2(RTYPE, mult2, mult3, cnst2, cnst3, out2, out3);             \
  }
#define DOTP_SB4_SH(...) DOTP_SB4(v8i16, __VA_ARGS__)

/* Description : Dot product of halfword vector elements
   Arguments   : Inputs  - mult0, mult1, cnst0, cnst1
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Signed halfword elements from 'mult0' are multiplied with
                 signed halfword elements from 'cnst0' producing a result
                 twice the size of input i.e. signed word.
                 The multiplication result of adjacent odd-even elements
                 are added together and written to the 'out0' vector
*/
#define DOTP_SH2(RTYPE, mult0, mult1, cnst0, cnst1, out0, out1) \
  {                                                             \
    out0 = (RTYPE)__msa_dotp_s_w((v8i16)mult0, (v8i16)cnst0);   \
    out1 = (RTYPE)__msa_dotp_s_w((v8i16)mult1, (v8i16)cnst1);   \
  }
#define DOTP_SH2_SW(...) DOTP_SH2(v4i32, __VA_ARGS__)

#define DOTP_SH4(RTYPE, mult0, mult1, mult2, mult3, cnst0, cnst1, cnst2, \
                 cnst3, out0, out1, out2, out3)                          \
  {                                                                      \
    DOTP_SH2(RTYPE, mult0, mult1, cnst0, cnst1, out0, out1);             \
    DOTP_SH2(RTYPE, mult2, mult3, cnst2, cnst3, out2, out3);             \
  }
#define DOTP_SH4_SW(...) DOTP_SH4(v4i32, __VA_ARGS__)

/* Description : Dot product of word vector elements
   Arguments   : Inputs  - mult0, mult1, cnst0, cnst1
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Signed word elements from 'mult0' are multiplied with
                 signed word elements from 'cnst0' producing a result
                 twice the size of input i.e. signed double word.
                 The multiplication result of adjacent odd-even elements
                 are added together and written to the 'out0' vector
*/
#define DOTP_SW2(RTYPE, mult0, mult1, cnst0, cnst1, out0, out1) \
  {                                                             \
    out0 = (RTYPE)__msa_dotp_s_d((v4i32)mult0, (v4i32)cnst0);   \
    out1 = (RTYPE)__msa_dotp_s_d((v4i32)mult1, (v4i32)cnst1);   \
  }
#define DOTP_SW2_SD(...) DOTP_SW2(v2i64, __VA_ARGS__)

/* Description : Dot product & addition of byte vector elements
   Arguments   : Inputs  - mult0, mult1, cnst0, cnst1
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Signed byte elements from 'mult0' are multiplied with
                 signed byte elements from 'cnst0' producing a result
                 twice the size of input i.e. signed halfword.
                 The multiplication result of adjacent odd-even elements
                 are added to the 'out0' vector
*/
#define DPADD_SB2(RTYPE, mult0, mult1, cnst0, cnst1, out0, out1)            \
  {                                                                         \
    out0 = (RTYPE)__msa_dpadd_s_h((v8i16)out0, (v16i8)mult0, (v16i8)cnst0); \
    out1 = (RTYPE)__msa_dpadd_s_h((v8i16)out1, (v16i8)mult1, (v16i8)cnst1); \
  }
#define DPADD_SB2_SH(...) DPADD_SB2(v8i16, __VA_ARGS__)

#define DPADD_SB4(RTYPE, mult0, mult1, mult2, mult3, cnst0, cnst1, cnst2, \
                  cnst3, out0, out1, out2, out3)                          \
  {                                                                       \
    DPADD_SB2(RTYPE, mult0, mult1, cnst0, cnst1, out0, out1);             \
    DPADD_SB2(RTYPE, mult2, mult3, cnst2, cnst3, out2, out3);             \
  }
#define DPADD_SB4_SH(...) DPADD_SB4(v8i16, __VA_ARGS__)

/* Description : Dot product & addition of halfword vector elements
   Arguments   : Inputs  - mult0, mult1, cnst0, cnst1
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Signed halfword elements from 'mult0' are multiplied with
                 signed halfword elements from 'cnst0' producing a result
                 twice the size of input i.e. signed word.
                 The multiplication result of adjacent odd-even elements
                 are added to the 'out0' vector
*/
#define DPADD_SH2(RTYPE, mult0, mult1, cnst0, cnst1, out0, out1)            \
  {                                                                         \
    out0 = (RTYPE)__msa_dpadd_s_w((v4i32)out0, (v8i16)mult0, (v8i16)cnst0); \
    out1 = (RTYPE)__msa_dpadd_s_w((v4i32)out1, (v8i16)mult1, (v8i16)cnst1); \
  }
#define DPADD_SH2_SW(...) DPADD_SH2(v4i32, __VA_ARGS__)

/* Description : Dot product & addition of double word vector elements
   Arguments   : Inputs  - mult0, mult1
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Each signed word element from 'mult0' is multiplied with itself
                 producing an intermediate result twice the size of input
                 i.e. signed double word
                 The multiplication result of adjacent odd-even elements
                 are added to the 'out0' vector
*/
#define DPADD_SD2(RTYPE, mult0, mult1, out0, out1)                          \
  {                                                                         \
    out0 = (RTYPE)__msa_dpadd_s_d((v2i64)out0, (v4i32)mult0, (v4i32)mult0); \
    out1 = (RTYPE)__msa_dpadd_s_d((v2i64)out1, (v4i32)mult1, (v4i32)mult1); \
  }
#define DPADD_SD2_SD(...) DPADD_SD2(v2i64, __VA_ARGS__)

/* Description : Minimum values between unsigned elements of
                 either vector are copied to the output vector
   Arguments   : Inputs  - in0, in1, min_vec
                 Outputs - in place operation
                 Return Type - as per RTYPE
   Details     : Minimum of unsigned halfword element values from 'in0' and
                 'min_vec' are written to output vector 'in0'
*/
#define MIN_UH2(RTYPE, in0, in1, min_vec)            \
  {                                                  \
    in0 = (RTYPE)__msa_min_u_h((v8u16)in0, min_vec); \
    in1 = (RTYPE)__msa_min_u_h((v8u16)in1, min_vec); \
  }
#define MIN_UH2_UH(...) MIN_UH2(v8u16, __VA_ARGS__)

#define MIN_UH4(RTYPE, in0, in1, in2, in3, min_vec) \
  {                                                 \
    MIN_UH2(RTYPE, in0, in1, min_vec);              \
    MIN_UH2(RTYPE, in2, in3, min_vec);              \
  }
#define MIN_UH4_UH(...) MIN_UH4(v8u16, __VA_ARGS__)

/* Description : Clips all signed halfword elements of input vector
                 between 0 & 255
   Arguments   : Input  - in
                 Output - out_m
                 Return Type - signed halfword
*/
#define CLIP_SH_0_255(in)                              \
  ({                                                   \
    v8i16 max_m = __msa_ldi_h(255);                    \
    v8i16 out_m;                                       \
                                                       \
    out_m = __msa_maxi_s_h((v8i16)in, 0);              \
    out_m = __msa_min_s_h((v8i16)max_m, (v8i16)out_m); \
    out_m;                                             \
  })
#define CLIP_SH2_0_255(in0, in1) \
  {                              \
    in0 = CLIP_SH_0_255(in0);    \
    in1 = CLIP_SH_0_255(in1);    \
  }
#define CLIP_SH4_0_255(in0, in1, in2, in3) \
  {                                        \
    CLIP_SH2_0_255(in0, in1);              \
    CLIP_SH2_0_255(in2, in3);              \
  }

/* Description : Horizontal addition of 4 signed word elements of input vector
   Arguments   : Input  - in       (signed word vector)
                 Output - sum_m    (i32 sum)
                 Return Type - signed word (GP)
   Details     : 4 signed word elements of 'in' vector are added together and
                 the resulting integer sum is returned
*/
#define HADD_SW_S32(in)                            \
  ({                                               \
    v2i64 res0_m, res1_m;                          \
    int32_t sum_m;                                 \
                                                   \
    res0_m = __msa_hadd_s_d((v4i32)in, (v4i32)in); \
    res1_m = __msa_splati_d(res0_m, 1);            \
    res0_m = res0_m + res1_m;                      \
    sum_m = __msa_copy_s_w((v4i32)res0_m, 0);      \
    sum_m;                                         \
  })

/* Description : Horizontal addition of 8 unsigned halfword elements
   Arguments   : Inputs  - in       (unsigned halfword vector)
                 Outputs - sum_m    (u32 sum)
                 Return Type - unsigned word
   Details     : 8 unsigned halfword elements of input vector are added
                 together and the resulting integer sum is returned
*/
#define HADD_UH_U32(in)                               \
  ({                                                  \
    v4u32 res_m;                                      \
    v2u64 res0_m, res1_m;                             \
    uint32_t sum_m;                                   \
                                                      \
    res_m = __msa_hadd_u_w((v8u16)in, (v8u16)in);     \
    res0_m = __msa_hadd_u_d(res_m, res_m);            \
    res1_m = (v2u64)__msa_splati_d((v2i64)res0_m, 1); \
    res0_m = res0_m + res1_m;                         \
    sum_m = __msa_copy_u_w((v4i32)res0_m, 0);         \
    sum_m;                                            \
  })

/* Description : Horizontal addition of unsigned byte vector elements
   Arguments   : Inputs  - in0, in1
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Each unsigned odd byte element from 'in0' is added to
                 even unsigned byte element from 'in0' (pairwise) and the
                 halfword result is written to 'out0'
*/
#define HADD_UB2(RTYPE, in0, in1, out0, out1)             \
  {                                                       \
    out0 = (RTYPE)__msa_hadd_u_h((v16u8)in0, (v16u8)in0); \
    out1 = (RTYPE)__msa_hadd_u_h((v16u8)in1, (v16u8)in1); \
  }
#define HADD_UB2_UH(...) HADD_UB2(v8u16, __VA_ARGS__)

#define HADD_UB4(RTYPE, in0, in1, in2, in3, out0, out1, out2, out3) \
  {                                                                 \
    HADD_UB2(RTYPE, in0, in1, out0, out1);                          \
    HADD_UB2(RTYPE, in2, in3, out2, out3);                          \
  }
#define HADD_UB4_UH(...) HADD_UB4(v8u16, __VA_ARGS__)

/* Description : Horizontal subtraction of unsigned byte vector elements
   Arguments   : Inputs  - in0, in1
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Each unsigned odd byte element from 'in0' is subtracted from
                 even unsigned byte element from 'in0' (pairwise) and the
                 halfword result is written to 'out0'
*/
#define HSUB_UB2(RTYPE, in0, in1, out0, out1)             \
  {                                                       \
    out0 = (RTYPE)__msa_hsub_u_h((v16u8)in0, (v16u8)in0); \
    out1 = (RTYPE)__msa_hsub_u_h((v16u8)in1, (v16u8)in1); \
  }
#define HSUB_UB2_SH(...) HSUB_UB2(v8i16, __VA_ARGS__)

/* Description : SAD (Sum of Absolute Difference)
   Arguments   : Inputs  - in0, in1, ref0, ref1
                 Outputs - sad_m                 (halfword vector)
                 Return Type - unsigned halfword
   Details     : Absolute difference of all the byte elements from 'in0' with
                 'ref0' is calculated and preserved in 'diff0'. Then even-odd
                 pairs are added together to generate 8 halfword results.
*/
#define SAD_UB2_UH(in0, in1, ref0, ref1)                     \
  ({                                                         \
    v16u8 diff0_m, diff1_m;                                  \
    v8u16 sad_m = { 0 };                                     \
                                                             \
    diff0_m = __msa_asub_u_b((v16u8)in0, (v16u8)ref0);       \
    diff1_m = __msa_asub_u_b((v16u8)in1, (v16u8)ref1);       \
                                                             \
    sad_m += __msa_hadd_u_h((v16u8)diff0_m, (v16u8)diff0_m); \
    sad_m += __msa_hadd_u_h((v16u8)diff1_m, (v16u8)diff1_m); \
                                                             \
    sad_m;                                                   \
  })

/* Description : Horizontal subtraction of signed halfword vector elements
   Arguments   : Inputs  - in0, in1
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Each signed odd halfword element from 'in0' is subtracted from
                 even signed halfword element from 'in0' (pairwise) and the
                 word result is written to 'out0'
*/
#define HSUB_UH2(RTYPE, in0, in1, out0, out1)             \
  {                                                       \
    out0 = (RTYPE)__msa_hsub_s_w((v8i16)in0, (v8i16)in0); \
    out1 = (RTYPE)__msa_hsub_s_w((v8i16)in1, (v8i16)in1); \
  }
#define HSUB_UH2_SW(...) HSUB_UH2(v4i32, __VA_ARGS__)

/* Description : Set element n input vector to GPR value
   Arguments   : Inputs - in0, in1, in2, in3
                 Output - out
                 Return Type - as per RTYPE
   Details     : Set element 0 in vector 'out' to value specified in 'in0'
*/
#define INSERT_W2(RTYPE, in0, in1, out)              \
  {                                                  \
    out = (RTYPE)__msa_insert_w((v4i32)out, 0, in0); \
    out = (RTYPE)__msa_insert_w((v4i32)out, 1, in1); \
  }
#define INSERT_W2_SB(...) INSERT_W2(v16i8, __VA_ARGS__)

#define INSERT_W4(RTYPE, in0, in1, in2, in3, out)    \
  {                                                  \
    out = (RTYPE)__msa_insert_w((v4i32)out, 0, in0); \
    out = (RTYPE)__msa_insert_w((v4i32)out, 1, in1); \
    out = (RTYPE)__msa_insert_w((v4i32)out, 2, in2); \
    out = (RTYPE)__msa_insert_w((v4i32)out, 3, in3); \
  }
#define INSERT_W4_UB(...) INSERT_W4(v16u8, __VA_ARGS__)
#define INSERT_W4_SB(...) INSERT_W4(v16i8, __VA_ARGS__)

#define INSERT_D2(RTYPE, in0, in1, out)              \
  {                                                  \
    out = (RTYPE)__msa_insert_d((v2i64)out, 0, in0); \
    out = (RTYPE)__msa_insert_d((v2i64)out, 1, in1); \
  }
#define INSERT_D2_UB(...) INSERT_D2(v16u8, __VA_ARGS__)
#define INSERT_D2_SB(...) INSERT_D2(v16i8, __VA_ARGS__)

/* Description : Interleave even byte elements from vectors
   Arguments   : Inputs  - in0, in1, in2, in3
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Even byte elements of 'in0' and 'in1' are interleaved
                 and written to 'out0'
*/
#define ILVEV_B2(RTYPE, in0, in1, in2, in3, out0, out1)  \
  {                                                      \
    out0 = (RTYPE)__msa_ilvev_b((v16i8)in1, (v16i8)in0); \
    out1 = (RTYPE)__msa_ilvev_b((v16i8)in3, (v16i8)in2); \
  }
#define ILVEV_B2_UB(...) ILVEV_B2(v16u8, __VA_ARGS__)
#define ILVEV_B2_SH(...) ILVEV_B2(v8i16, __VA_ARGS__)

/* Description : Interleave even halfword elements from vectors
   Arguments   : Inputs  - in0, in1, in2, in3
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Even halfword elements of 'in0' and 'in1' are interleaved
                 and written to 'out0'
*/
#define ILVEV_H2(RTYPE, in0, in1, in2, in3, out0, out1)  \
  {                                                      \
    out0 = (RTYPE)__msa_ilvev_h((v8i16)in1, (v8i16)in0); \
    out1 = (RTYPE)__msa_ilvev_h((v8i16)in3, (v8i16)in2); \
  }
#define ILVEV_H2_UB(...) ILVEV_H2(v16u8, __VA_ARGS__)
#define ILVEV_H2_SH(...) ILVEV_H2(v8i16, __VA_ARGS__)
#define ILVEV_H2_SW(...) ILVEV_H2(v4i32, __VA_ARGS__)

/* Description : Interleave even word elements from vectors
   Arguments   : Inputs  - in0, in1, in2, in3
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Even word elements of 'in0' and 'in1' are interleaved
                 and written to 'out0'
*/
#define ILVEV_W2(RTYPE, in0, in1, in2, in3, out0, out1)  \
  {                                                      \
    out0 = (RTYPE)__msa_ilvev_w((v4i32)in1, (v4i32)in0); \
    out1 = (RTYPE)__msa_ilvev_w((v4i32)in3, (v4i32)in2); \
  }
#define ILVEV_W2_SB(...) ILVEV_W2(v16i8, __VA_ARGS__)

/* Description : Interleave even double word elements from vectors
   Arguments   : Inputs  - in0, in1, in2, in3
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Even double word elements of 'in0' and 'in1' are interleaved
                 and written to 'out0'
*/
#define ILVEV_D2(RTYPE, in0, in1, in2, in3, out0, out1)  \
  {                                                      \
    out0 = (RTYPE)__msa_ilvev_d((v2i64)in1, (v2i64)in0); \
    out1 = (RTYPE)__msa_ilvev_d((v2i64)in3, (v2i64)in2); \
  }
#define ILVEV_D2_UB(...) ILVEV_D2(v16u8, __VA_ARGS__)

/* Description : Interleave left half of byte elements from vectors
   Arguments   : Inputs  - in0, in1, in2, in3
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Left half of byte elements of 'in0' and 'in1' are interleaved
                 and written to 'out0'.
*/
#define ILVL_B2(RTYPE, in0, in1, in2, in3, out0, out1)  \
  {                                                     \
    out0 = (RTYPE)__msa_ilvl_b((v16i8)in0, (v16i8)in1); \
    out1 = (RTYPE)__msa_ilvl_b((v16i8)in2, (v16i8)in3); \
  }
#define ILVL_B2_UB(...) ILVL_B2(v16u8, __VA_ARGS__)
#define ILVL_B2_SB(...) ILVL_B2(v16i8, __VA_ARGS__)
#define ILVL_B2_UH(...) ILVL_B2(v8u16, __VA_ARGS__)
#define ILVL_B2_SH(...) ILVL_B2(v8i16, __VA_ARGS__)

#define ILVL_B4(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, \
                out2, out3)                                                \
  {                                                                        \
    ILVL_B2(RTYPE, in0, in1, in2, in3, out0, out1);                        \
    ILVL_B2(RTYPE, in4, in5, in6, in7, out2, out3);                        \
  }
#define ILVL_B4_SB(...) ILVL_B4(v16i8, __VA_ARGS__)
#define ILVL_B4_SH(...) ILVL_B4(v8i16, __VA_ARGS__)
#define ILVL_B4_UH(...) ILVL_B4(v8u16, __VA_ARGS__)

/* Description : Interleave left half of halfword elements from vectors
   Arguments   : Inputs  - in0, in1, in2, in3
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Left half of halfword elements of 'in0' and 'in1' are
                 interleaved and written to 'out0'.
*/
#define ILVL_H2(RTYPE, in0, in1, in2, in3, out0, out1)  \
  {                                                     \
    out0 = (RTYPE)__msa_ilvl_h((v8i16)in0, (v8i16)in1); \
    out1 = (RTYPE)__msa_ilvl_h((v8i16)in2, (v8i16)in3); \
  }
#define ILVL_H2_SH(...) ILVL_H2(v8i16, __VA_ARGS__)
#define ILVL_H2_SW(...) ILVL_H2(v4i32, __VA_ARGS__)

/* Description : Interleave left half of word elements from vectors
   Arguments   : Inputs  - in0, in1, in2, in3
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Left half of word elements of 'in0' and 'in1' are interleaved
                 and written to 'out0'.
*/
#define ILVL_W2(RTYPE, in0, in1, in2, in3, out0, out1)  \
  {                                                     \
    out0 = (RTYPE)__msa_ilvl_w((v4i32)in0, (v4i32)in1); \
    out1 = (RTYPE)__msa_ilvl_w((v4i32)in2, (v4i32)in3); \
  }
#define ILVL_W2_UB(...) ILVL_W2(v16u8, __VA_ARGS__)
#define ILVL_W2_SH(...) ILVL_W2(v8i16, __VA_ARGS__)

/* Description : Interleave right half of byte elements from vectors
   Arguments   : Inputs  - in0, in1, in2, in3
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Right half of byte elements of 'in0' and 'in1' are interleaved
                 and written to out0.
*/
#define ILVR_B2(RTYPE, in0, in1, in2, in3, out0, out1)  \
  {                                                     \
    out0 = (RTYPE)__msa_ilvr_b((v16i8)in0, (v16i8)in1); \
    out1 = (RTYPE)__msa_ilvr_b((v16i8)in2, (v16i8)in3); \
  }
#define ILVR_B2_UB(...) ILVR_B2(v16u8, __VA_ARGS__)
#define ILVR_B2_SB(...) ILVR_B2(v16i8, __VA_ARGS__)
#define ILVR_B2_UH(...) ILVR_B2(v8u16, __VA_ARGS__)
#define ILVR_B2_SH(...) ILVR_B2(v8i16, __VA_ARGS__)

#define ILVR_B4(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, \
                out2, out3)                                                \
  {                                                                        \
    ILVR_B2(RTYPE, in0, in1, in2, in3, out0, out1);                        \
    ILVR_B2(RTYPE, in4, in5, in6, in7, out2, out3);                        \
  }
#define ILVR_B4_UB(...) ILVR_B4(v16u8, __VA_ARGS__)
#define ILVR_B4_SB(...) ILVR_B4(v16i8, __VA_ARGS__)
#define ILVR_B4_UH(...) ILVR_B4(v8u16, __VA_ARGS__)
#define ILVR_B4_SH(...) ILVR_B4(v8i16, __VA_ARGS__)

#define ILVR_B8(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, in8, in9, in10, \
                in11, in12, in13, in14, in15, out0, out1, out2, out3, out4,    \
                out5, out6, out7)                                              \
  {                                                                            \
    ILVR_B4(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, out2,   \
            out3);                                                             \
    ILVR_B4(RTYPE, in8, in9, in10, in11, in12, in13, in14, in15, out4, out5,   \
            out6, out7);                                                       \
  }
#define ILVR_B8_UH(...) ILVR_B8(v8u16, __VA_ARGS__)

/* Description : Interleave right half of halfword elements from vectors
   Arguments   : Inputs  - in0, in1, in2, in3
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Right half of halfword elements of 'in0' and 'in1' are
                 interleaved and written to 'out0'.
*/
#define ILVR_H2(RTYPE, in0, in1, in2, in3, out0, out1)  \
  {                                                     \
    out0 = (RTYPE)__msa_ilvr_h((v8i16)in0, (v8i16)in1); \
    out1 = (RTYPE)__msa_ilvr_h((v8i16)in2, (v8i16)in3); \
  }
#define ILVR_H2_SH(...) ILVR_H2(v8i16, __VA_ARGS__)
#define ILVR_H2_SW(...) ILVR_H2(v4i32, __VA_ARGS__)

#define ILVR_H4(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, \
                out2, out3)                                                \
  {                                                                        \
    ILVR_H2(RTYPE, in0, in1, in2, in3, out0, out1);                        \
    ILVR_H2(RTYPE, in4, in5, in6, in7, out2, out3);                        \
  }
#define ILVR_H4_SH(...) ILVR_H4(v8i16, __VA_ARGS__)

#define ILVR_W2(RTYPE, in0, in1, in2, in3, out0, out1)  \
  {                                                     \
    out0 = (RTYPE)__msa_ilvr_w((v4i32)in0, (v4i32)in1); \
    out1 = (RTYPE)__msa_ilvr_w((v4i32)in2, (v4i32)in3); \
  }
#define ILVR_W2_UB(...) ILVR_W2(v16u8, __VA_ARGS__)
#define ILVR_W2_SH(...) ILVR_W2(v8i16, __VA_ARGS__)

#define ILVR_W4(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, \
                out2, out3)                                                \
  {                                                                        \
    ILVR_W2(RTYPE, in0, in1, in2, in3, out0, out1);                        \
    ILVR_W2(RTYPE, in4, in5, in6, in7, out2, out3);                        \
  }
#define ILVR_W4_UB(...) ILVR_W4(v16u8, __VA_ARGS__)

/* Description : Interleave right half of double word elements from vectors
   Arguments   : Inputs  - in0, in1, in2, in3
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Right half of double word elements of 'in0' and 'in1' are
                 interleaved and written to 'out0'.
*/
#define ILVR_D2(RTYPE, in0, in1, in2, in3, out0, out1)      \
  {                                                         \
    out0 = (RTYPE)__msa_ilvr_d((v2i64)(in0), (v2i64)(in1)); \
    out1 = (RTYPE)__msa_ilvr_d((v2i64)(in2), (v2i64)(in3)); \
  }
#define ILVR_D2_UB(...) ILVR_D2(v16u8, __VA_ARGS__)
#define ILVR_D2_SB(...) ILVR_D2(v16i8, __VA_ARGS__)
#define ILVR_D2_SH(...) ILVR_D2(v8i16, __VA_ARGS__)

#define ILVR_D3(RTYPE, in0, in1, in2, in3, in4, in5, out0, out1, out2) \
  {                                                                    \
    ILVR_D2(RTYPE, in0, in1, in2, in3, out0, out1);                    \
    out2 = (RTYPE)__msa_ilvr_d((v2i64)(in4), (v2i64)(in5));            \
  }
#define ILVR_D3_SB(...) ILVR_D3(v16i8, __VA_ARGS__)

#define ILVR_D4(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, \
                out2, out3)                                                \
  {                                                                        \
    ILVR_D2(RTYPE, in0, in1, in2, in3, out0, out1);                        \
    ILVR_D2(RTYPE, in4, in5, in6, in7, out2, out3);                        \
  }
#define ILVR_D4_SB(...) ILVR_D4(v16i8, __VA_ARGS__)
#define ILVR_D4_UB(...) ILVR_D4(v16u8, __VA_ARGS__)

/* Description : Interleave both left and right half of input vectors
   Arguments   : Inputs  - in0, in1
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Right half of byte elements from 'in0' and 'in1' are
                 interleaved and written to 'out0'
*/
#define ILVRL_B2(RTYPE, in0, in1, out0, out1)           \
  {                                                     \
    out0 = (RTYPE)__msa_ilvr_b((v16i8)in0, (v16i8)in1); \
    out1 = (RTYPE)__msa_ilvl_b((v16i8)in0, (v16i8)in1); \
  }
#define ILVRL_B2_UB(...) ILVRL_B2(v16u8, __VA_ARGS__)
#define ILVRL_B2_SB(...) ILVRL_B2(v16i8, __VA_ARGS__)
#define ILVRL_B2_UH(...) ILVRL_B2(v8u16, __VA_ARGS__)
#define ILVRL_B2_SH(...) ILVRL_B2(v8i16, __VA_ARGS__)

#define ILVRL_H2(RTYPE, in0, in1, out0, out1)           \
  {                                                     \
    out0 = (RTYPE)__msa_ilvr_h((v8i16)in0, (v8i16)in1); \
    out1 = (RTYPE)__msa_ilvl_h((v8i16)in0, (v8i16)in1); \
  }
#define ILVRL_H2_SH(...) ILVRL_H2(v8i16, __VA_ARGS__)
#define ILVRL_H2_SW(...) ILVRL_H2(v4i32, __VA_ARGS__)

#define ILVRL_W2(RTYPE, in0, in1, out0, out1)           \
  {                                                     \
    out0 = (RTYPE)__msa_ilvr_w((v4i32)in0, (v4i32)in1); \
    out1 = (RTYPE)__msa_ilvl_w((v4i32)in0, (v4i32)in1); \
  }
#define ILVRL_W2_UB(...) ILVRL_W2(v16u8, __VA_ARGS__)
#define ILVRL_W2_SH(...) ILVRL_W2(v8i16, __VA_ARGS__)
#define ILVRL_W2_SW(...) ILVRL_W2(v4i32, __VA_ARGS__)

/* Description : Saturate the halfword element values to the max
                 unsigned value of (sat_val + 1) bits
                 The element data width remains unchanged
   Arguments   : Inputs  - in0, in1, sat_val
                 Outputs - in place operation
                 Return Type - as per RTYPE
   Details     : Each unsigned halfword element from 'in0' is saturated to the
                 value generated with (sat_val + 1) bit range.
                 The results are written in place
*/
#define SAT_UH2(RTYPE, in0, in1, sat_val)            \
  {                                                  \
    in0 = (RTYPE)__msa_sat_u_h((v8u16)in0, sat_val); \
    in1 = (RTYPE)__msa_sat_u_h((v8u16)in1, sat_val); \
  }
#define SAT_UH2_UH(...) SAT_UH2(v8u16, __VA_ARGS__)

#define SAT_UH4(RTYPE, in0, in1, in2, in3, sat_val) \
  {                                                 \
    SAT_UH2(RTYPE, in0, in1, sat_val);              \
    SAT_UH2(RTYPE, in2, in3, sat_val)               \
  }
#define SAT_UH4_UH(...) SAT_UH4(v8u16, __VA_ARGS__)

/* Description : Saturate the halfword element values to the max
                 unsigned value of (sat_val + 1) bits
                 The element data width remains unchanged
   Arguments   : Inputs  - in0, in1, sat_val
                 Outputs - in place operation
                 Return Type - as per RTYPE
   Details     : Each unsigned halfword element from 'in0' is saturated to the
                 value generated with (sat_val + 1) bit range
                 The results are written in place
*/
#define SAT_SH2(RTYPE, in0, in1, sat_val)            \
  {                                                  \
    in0 = (RTYPE)__msa_sat_s_h((v8i16)in0, sat_val); \
    in1 = (RTYPE)__msa_sat_s_h((v8i16)in1, sat_val); \
  }
#define SAT_SH2_SH(...) SAT_SH2(v8i16, __VA_ARGS__)

#define SAT_SH4(RTYPE, in0, in1, in2, in3, sat_val) \
  {                                                 \
    SAT_SH2(RTYPE, in0, in1, sat_val);              \
    SAT_SH2(RTYPE, in2, in3, sat_val);              \
  }
#define SAT_SH4_SH(...) SAT_SH4(v8i16, __VA_ARGS__)

/* Description : Indexed halfword element values are replicated to all
                 elements in output vector
   Arguments   : Inputs  - in, idx0, idx1
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : 'idx0' element value from 'in' vector is replicated to all
                  elements in 'out0' vector
                  Valid index range for halfword operation is 0-7
*/
#define SPLATI_H2(RTYPE, in, idx0, idx1, out0, out1) \
  {                                                  \
    out0 = (RTYPE)__msa_splati_h((v8i16)in, idx0);   \
    out1 = (RTYPE)__msa_splati_h((v8i16)in, idx1);   \
  }
#define SPLATI_H2_SH(...) SPLATI_H2(v8i16, __VA_ARGS__)

#define SPLATI_H4(RTYPE, in, idx0, idx1, idx2, idx3, out0, out1, out2, out3) \
  {                                                                          \
    SPLATI_H2(RTYPE, in, idx0, idx1, out0, out1);                            \
    SPLATI_H2(RTYPE, in, idx2, idx3, out2, out3);                            \
  }
#define SPLATI_H4_SB(...) SPLATI_H4(v16i8, __VA_ARGS__)
#define SPLATI_H4_SH(...) SPLATI_H4(v8i16, __VA_ARGS__)

/* Description : Pack even byte elements of vector pairs
   Arguments   : Inputs  - in0, in1, in2, in3
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Even byte elements of 'in0' are copied to the left half of
                 'out0' & even byte elements of 'in1' are copied to the right
                 half of 'out0'.
*/
#define PCKEV_B2(RTYPE, in0, in1, in2, in3, out0, out1)  \
  {                                                      \
    out0 = (RTYPE)__msa_pckev_b((v16i8)in0, (v16i8)in1); \
    out1 = (RTYPE)__msa_pckev_b((v16i8)in2, (v16i8)in3); \
  }
#define PCKEV_B2_SB(...) PCKEV_B2(v16i8, __VA_ARGS__)
#define PCKEV_B2_UB(...) PCKEV_B2(v16u8, __VA_ARGS__)
#define PCKEV_B2_SH(...) PCKEV_B2(v8i16, __VA_ARGS__)

#define PCKEV_B4(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, \
                 out2, out3)                                                \
  {                                                                         \
    PCKEV_B2(RTYPE, in0, in1, in2, in3, out0, out1);                        \
    PCKEV_B2(RTYPE, in4, in5, in6, in7, out2, out3);                        \
  }
#define PCKEV_B4_SB(...) PCKEV_B4(v16i8, __VA_ARGS__)
#define PCKEV_B4_UB(...) PCKEV_B4(v16u8, __VA_ARGS__)
#define PCKEV_B4_SH(...) PCKEV_B4(v8i16, __VA_ARGS__)

/* Description : Pack even halfword elements of vector pairs
   Arguments   : Inputs  - in0, in1, in2, in3
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Even halfword elements of 'in0' are copied to the left half of
                 'out0' & even halfword elements of 'in1' are copied to the
                 right half of 'out0'.
*/
#define PCKEV_H2(RTYPE, in0, in1, in2, in3, out0, out1)  \
  {                                                      \
    out0 = (RTYPE)__msa_pckev_h((v8i16)in0, (v8i16)in1); \
    out1 = (RTYPE)__msa_pckev_h((v8i16)in2, (v8i16)in3); \
  }
#define PCKEV_H2_SH(...) PCKEV_H2(v8i16, __VA_ARGS__)
#define PCKEV_H2_SW(...) PCKEV_H2(v4i32, __VA_ARGS__)

#define PCKEV_H4(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, \
                 out2, out3)                                                \
  {                                                                         \
    PCKEV_H2(RTYPE, in0, in1, in2, in3, out0, out1);                        \
    PCKEV_H2(RTYPE, in4, in5, in6, in7, out2, out3);                        \
  }
#define PCKEV_H4_SH(...) PCKEV_H4(v8i16, __VA_ARGS__)

/* Description : Pack even double word elements of vector pairs
   Arguments   : Inputs  - in0, in1, in2, in3
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Even double elements of 'in0' are copied to the left half of
                 'out0' & even double elements of 'in1' are copied to the right
                 half of 'out0'.
*/
#define PCKEV_D2(RTYPE, in0, in1, in2, in3, out0, out1)  \
  {                                                      \
    out0 = (RTYPE)__msa_pckev_d((v2i64)in0, (v2i64)in1); \
    out1 = (RTYPE)__msa_pckev_d((v2i64)in2, (v2i64)in3); \
  }
#define PCKEV_D2_UB(...) PCKEV_D2(v16u8, __VA_ARGS__)
#define PCKEV_D2_SH(...) PCKEV_D2(v8i16, __VA_ARGS__)

#define PCKEV_D4(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, \
                 out2, out3)                                                \
  {                                                                         \
    PCKEV_D2(RTYPE, in0, in1, in2, in3, out0, out1);                        \
    PCKEV_D2(RTYPE, in4, in5, in6, in7, out2, out3);                        \
  }
#define PCKEV_D4_UB(...) PCKEV_D4(v16u8, __VA_ARGS__)

/* Description : Each byte element is logically xor'ed with immediate 128
   Arguments   : Inputs  - in0, in1
                 Outputs - in place operation
                 Return Type - as per RTYPE
   Details     : Each unsigned byte element from input vector 'in0' is
                 logically xor'ed with 128 and the result is stored in-place.
*/
#define XORI_B2_128(RTYPE, in0, in1)            \
  {                                             \
    in0 = (RTYPE)__msa_xori_b((v16u8)in0, 128); \
    in1 = (RTYPE)__msa_xori_b((v16u8)in1, 128); \
  }
#define XORI_B2_128_UB(...) XORI_B2_128(v16u8, __VA_ARGS__)
#define XORI_B2_128_SB(...) XORI_B2_128(v16i8, __VA_ARGS__)

#define XORI_B3_128(RTYPE, in0, in1, in2)       \
  {                                             \
    XORI_B2_128(RTYPE, in0, in1);               \
    in2 = (RTYPE)__msa_xori_b((v16u8)in2, 128); \
  }
#define XORI_B3_128_SB(...) XORI_B3_128(v16i8, __VA_ARGS__)

#define XORI_B4_128(RTYPE, in0, in1, in2, in3) \
  {                                            \
    XORI_B2_128(RTYPE, in0, in1);              \
    XORI_B2_128(RTYPE, in2, in3);              \
  }
#define XORI_B4_128_UB(...) XORI_B4_128(v16u8, __VA_ARGS__)
#define XORI_B4_128_SB(...) XORI_B4_128(v16i8, __VA_ARGS__)

#define XORI_B7_128(RTYPE, in0, in1, in2, in3, in4, in5, in6) \
  {                                                           \
    XORI_B4_128(RTYPE, in0, in1, in2, in3);                   \
    XORI_B3_128(RTYPE, in4, in5, in6);                        \
  }
#define XORI_B7_128_SB(...) XORI_B7_128(v16i8, __VA_ARGS__)

/* Description : Average of signed halfword elements -> (a + b) / 2
   Arguments   : Inputs  - in0, in1, in2, in3, in4, in5, in6, in7
                 Outputs - out0, out1, out2, out3
                 Return Type - as per RTYPE
   Details     : Each signed halfword element from 'in0' is added to each
                 signed halfword element of 'in1' with full precision resulting
                 in one extra bit in the result. The result is then divided by
                 2 and written to 'out0'
*/
#define AVE_SH4(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, \
                out2, out3)                                                \
  {                                                                        \
    out0 = (RTYPE)__msa_ave_s_h((v8i16)in0, (v8i16)in1);                   \
    out1 = (RTYPE)__msa_ave_s_h((v8i16)in2, (v8i16)in3);                   \
    out2 = (RTYPE)__msa_ave_s_h((v8i16)in4, (v8i16)in5);                   \
    out3 = (RTYPE)__msa_ave_s_h((v8i16)in6, (v8i16)in7);                   \
  }
#define AVE_SH4_SH(...) AVE_SH4(v8i16, __VA_ARGS__)

/* Description : Addition of signed halfword elements and signed saturation
   Arguments   : Inputs  - in0, in1, in2, in3
                 Outputs - out0, out1
                 Return Type - as per RTYPE
   Details     : Signed halfword elements from 'in0' are added to signed
                 halfword elements of 'in1'. The result is then signed saturated
                 between halfword data type range
*/
#define ADDS_SH2(RTYPE, in0, in1, in2, in3, out0, out1)   \
  {                                                       \
    out0 = (RTYPE)__msa_adds_s_h((v8i16)in0, (v8i16)in1); \
    out1 = (RTYPE)__msa_adds_s_h((v8i16)in2, (v8i16)in3); \
  }
#define ADDS_SH2_SH(...) ADDS_SH2(v8i16, __VA_ARGS__)

#define ADDS_SH4(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, \
                 out2, out3)                                                \
  {                                                                         \
    ADDS_SH2(RTYPE, in0, in1, in2, in3, out0, out1);                        \
    ADDS_SH2(RTYPE, in4, in5, in6, in7, out2, out3);                        \
  }
#define ADDS_SH4_SH(...) ADDS_SH4(v8i16, __VA_ARGS__)

/* Description : Shift left all elements of vector (generic for all data types)
   Arguments   : Inputs  - in0, in1, in2, in3, shift
                 Outputs - in place operation
                 Return Type - as per input vector RTYPE
   Details     : Each element of vector 'in0' is left shifted by 'shift' and
                 the result is written in-place.
*/
#define SLLI_4V(in0, in1, in2, in3, shift) \
  {                                        \
    in0 = in0 << shift;                    \
    in1 = in1 << shift;                    \
    in2 = in2 << shift;                    \
    in3 = in3 << shift;                    \
  }

/* Description : Arithmetic shift right all elements of vector
                 (generic for all data types)
   Arguments   : Inputs  - in0, in1, in2, in3, shift
                 Outputs - in place operation
                 Return Type - as per input vector RTYPE
   Details     : Each element of vector 'in0' is right shifted by 'shift' and
                 the result is written in-place. 'shift' is a GP variable.
*/
#define SRA_4V(in0, in1, in2, in3, shift) \
  {                                       \
    in0 = in0 >> shift;                   \
    in1 = in1 >> shift;                   \
    in2 = in2 >> shift;                   \
    in3 = in3 >> shift;                   \
  }

/* Description : Shift right arithmetic rounded words
   Arguments   : Inputs  - in0, in1, shift
                 Outputs - in place operation
                 Return Type - as per RTYPE
   Details     : Each element of vector 'in0' is shifted right arithmetically by
                 the number of bits in the corresponding element in the vector
                 'shift'. The last discarded bit is added to shifted value for
                 rounding and the result is written in-place.
                 'shift' is a vector.
*/
#define SRAR_W2(RTYPE, in0, in1, shift)                  \
  {                                                      \
    in0 = (RTYPE)__msa_srar_w((v4i32)in0, (v4i32)shift); \
    in1 = (RTYPE)__msa_srar_w((v4i32)in1, (v4i32)shift); \
  }

#define SRAR_W4(RTYPE, in0, in1, in2, in3, shift) \
  {                                               \
    SRAR_W2(RTYPE, in0, in1, shift)               \
    SRAR_W2(RTYPE, in2, in3, shift)               \
  }
#define SRAR_W4_SW(...) SRAR_W4(v4i32, __VA_ARGS__)

/* Description : Shift right arithmetic rounded (immediate)
   Arguments   : Inputs  - in0, in1, shift
                 Outputs - in place operation
                 Return Type - as per RTYPE
   Details     : Each element of vector 'in0' is shifted right arithmetically by
                 the value in 'shift'. The last discarded bit is added to the
                 shifted value for rounding and the result is written in-place.
                 'shift' is an immediate value.
*/
#define SRARI_H2(RTYPE, in0, in1, shift)           \
  {                                                \
    in0 = (RTYPE)__msa_srari_h((v8i16)in0, shift); \
    in1 = (RTYPE)__msa_srari_h((v8i16)in1, shift); \
  }
#define SRARI_H2_UH(...) SRARI_H2(v8u16, __VA_ARGS__)
#define SRARI_H2_SH(...) SRARI_H2(v8i16, __VA_ARGS__)

#define SRARI_H4(RTYPE, in0, in1, in2, in3, shift) \
  {                                                \
    SRARI_H2(RTYPE, in0, in1, shift);              \
    SRARI_H2(RTYPE, in2, in3, shift);              \
  }
#define SRARI_H4_UH(...) SRARI_H4(v8u16, __VA_ARGS__)
#define SRARI_H4_SH(...) SRARI_H4(v8i16, __VA_ARGS__)

#define SRARI_W2(RTYPE, in0, in1, shift)           \
  {                                                \
    in0 = (RTYPE)__msa_srari_w((v4i32)in0, shift); \
    in1 = (RTYPE)__msa_srari_w((v4i32)in1, shift); \
  }
#define SRARI_W2_SW(...) SRARI_W2(v4i32, __VA_ARGS__)

#define SRARI_W4(RTYPE, in0, in1, in2, in3, shift) \
  {                                                \
    SRARI_W2(RTYPE, in0, in1, shift);              \
    SRARI_W2(RTYPE, in2, in3, shift);              \
  }
#define SRARI_W4_SW(...) SRARI_W4(v4i32, __VA_ARGS__)

/* Description : Logical shift right all elements of vector (immediate)
   Arguments   : Inputs  - in0, in1, in2, in3, shift
                 Outputs - out0, out1, out2, out3
                 Return Type - as per RTYPE
   Details     : Each element of vector 'in0' is right shifted by 'shift' and
                 the result is written in-place. 'shift' is an immediate value.
*/
#define SRLI_H4(RTYPE, in0, in1, in2, in3, out0, out1, out2, out3, shift) \
  {                                                                       \
    out0 = (RTYPE)__msa_srli_h((v8i16)in0, shift);                        \
    out1 = (RTYPE)__msa_srli_h((v8i16)in1, shift);                        \
    out2 = (RTYPE)__msa_srli_h((v8i16)in2, shift);                        \
    out3 = (RTYPE)__msa_srli_h((v8i16)in3, shift);                        \
  }
#define SRLI_H4_SH(...) SRLI_H4(v8i16, __VA_ARGS__)

/* Description : Multiplication of pairs of vectors
   Arguments   : Inputs  - in0, in1, in2, in3
                 Outputs - out0, out1
   Details     : Each element from 'in0' is multiplied with elements from 'in1'
                 and the result is written to 'out0'
*/
#define MUL2(in0, in1, in2, in3, out0, out1) \
  {                                          \
    out0 = in0 * in1;                        \
    out1 = in2 * in3;                        \
  }
#define MUL4(in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, out2, out3) \
  {                                                                          \
    MUL2(in0, in1, in2, in3, out0, out1);                                    \
    MUL2(in4, in5, in6, in7, out2, out3);                                    \
  }

/* Description : Addition of 2 pairs of vectors
   Arguments   : Inputs  - in0, in1, in2, in3
                 Outputs - out0, out1
   Details     : Each element in 'in0' is added to 'in1' and result is written
                 to 'out0'.
*/
#define ADD2(in0, in1, in2, in3, out0, out1) \
  {                                          \
    out0 = in0 + in1;                        \
    out1 = in2 + in3;                        \
  }
#define ADD4(in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, out2, out3) \
  {                                                                          \
    ADD2(in0, in1, in2, in3, out0, out1);                                    \
    ADD2(in4, in5, in6, in7, out2, out3);                                    \
  }

/* Description : Subtraction of 2 pairs of vectors
   Arguments   : Inputs  - in0, in1, in2, in3
                 Outputs - out0, out1
   Details     : Each element in 'in1' is subtracted from 'in0' and result is
                 written to 'out0'.
*/
#define SUB2(in0, in1, in2, in3, out0, out1) \
  {                                          \
    out0 = in0 - in1;                        \
    out1 = in2 - in3;                        \
  }
#define SUB4(in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, out2, out3) \
  {                                                                          \
    out0 = in0 - in1;                                                        \
    out1 = in2 - in3;                                                        \
    out2 = in4 - in5;                                                        \
    out3 = in6 - in7;                                                        \
  }

/* Description : Sign extend halfword elements from right half of the vector
   Arguments   : Input  - in    (halfword vector)
                 Output - out   (sign extended word vector)
                 Return Type - signed word
   Details     : Sign bit of halfword elements from input vector 'in' is
                 extracted and interleaved with same vector 'in0' to generate
                 4 word elements keeping sign intact
*/
#define UNPCK_R_SH_SW(in, out)                    \
  {                                               \
    v8i16 sign_m;                                 \
                                                  \
    sign_m = __msa_clti_s_h((v8i16)in, 0);        \
    out = (v4i32)__msa_ilvr_h(sign_m, (v8i16)in); \
  }

/* Description : Zero extend unsigned byte elements to halfword elements
   Arguments   : Input   - in          (unsigned byte vector)
                 Outputs - out0, out1  (unsigned  halfword vectors)
                 Return Type - signed halfword
   Details     : Zero extended right half of vector is returned in 'out0'
                 Zero extended left half of vector is returned in 'out1'
*/
#define UNPCK_UB_SH(in, out0, out1)      \
  {                                      \
    v16i8 zero_m = { 0 };                \
                                         \
    ILVRL_B2_SH(zero_m, in, out0, out1); \
  }

/* Description : Sign extend halfword elements from input vector and return
                 the result in pair of vectors
   Arguments   : Input   - in            (halfword vector)
                 Outputs - out0, out1   (sign extended word vectors)
                 Return Type - signed word
   Details     : Sign bit of halfword elements from input vector 'in' is
                 extracted and interleaved right with same vector 'in0' to
                 generate 4 signed word elements in 'out0'
                 Then interleaved left with same vector 'in0' to
                 generate 4 signed word elements in 'out1'
*/
#define UNPCK_SH_SW(in, out0, out1)       \
  {                                       \
    v8i16 tmp_m;                          \
                                          \
    tmp_m = __msa_clti_s_h((v8i16)in, 0); \
    ILVRL_H2_SW(tmp_m, in, out0, out1);   \
  }

/* Description : Butterfly of 4 input vectors
   Arguments   : Inputs  - in0, in1, in2, in3
                 Outputs - out0, out1, out2, out3
   Details     : Butterfly operation
*/
#define BUTTERFLY_4(in0, in1, in2, in3, out0, out1, out2, out3) \
  {                                                             \
    out0 = in0 + in3;                                           \
    out1 = in1 + in2;                                           \
                                                                \
    out2 = in1 - in2;                                           \
    out3 = in0 - in3;                                           \
  }

/* Description : Butterfly of 8 input vectors
   Arguments   : Inputs  - in0 ...  in7
                 Outputs - out0 .. out7
   Details     : Butterfly operation
*/
#define BUTTERFLY_8(in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, out2, \
                    out3, out4, out5, out6, out7)                             \
  {                                                                           \
    out0 = in0 + in7;                                                         \
    out1 = in1 + in6;                                                         \
    out2 = in2 + in5;                                                         \
    out3 = in3 + in4;                                                         \
                                                                              \
    out4 = in3 - in4;                                                         \
    out5 = in2 - in5;                                                         \
    out6 = in1 - in6;                                                         \
    out7 = in0 - in7;                                                         \
  }

/* Description : Butterfly of 16 input vectors
   Arguments   : Inputs  - in0 ...  in15
                 Outputs - out0 .. out15
   Details     : Butterfly operation
*/
#define BUTTERFLY_16(in0, in1, in2, in3, in4, in5, in6, in7, in8, in9, in10,  \
                     in11, in12, in13, in14, in15, out0, out1, out2, out3,    \
                     out4, out5, out6, out7, out8, out9, out10, out11, out12, \
                     out13, out14, out15)                                     \
  {                                                                           \
    out0 = in0 + in15;                                                        \
    out1 = in1 + in14;                                                        \
    out2 = in2 + in13;                                                        \
    out3 = in3 + in12;                                                        \
    out4 = in4 + in11;                                                        \
    out5 = in5 + in10;                                                        \
    out6 = in6 + in9;                                                         \
    out7 = in7 + in8;                                                         \
                                                                              \
    out8 = in7 - in8;                                                         \
    out9 = in6 - in9;                                                         \
    out10 = in5 - in10;                                                       \
    out11 = in4 - in11;                                                       \
    out12 = in3 - in12;                                                       \
    out13 = in2 - in13;                                                       \
    out14 = in1 - in14;                                                       \
    out15 = in0 - in15;                                                       \
  }

/* Description : Transpose input 8x8 byte block
   Arguments   : Inputs  - in0, in1, in2, in3, in4, in5, in6, in7
                 Outputs - out0, out1, out2, out3, out4, out5, out6, out7
                 Return Type - as per RTYPE
*/
#define TRANSPOSE8x8_UB(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, out0,   \
                        out1, out2, out3, out4, out5, out6, out7)              \
  {                                                                            \
    v16i8 tmp0_m, tmp1_m, tmp2_m, tmp3_m;                                      \
    v16i8 tmp4_m, tmp5_m, tmp6_m, tmp7_m;                                      \
                                                                               \
    ILVR_B4_SB(in2, in0, in3, in1, in6, in4, in7, in5, tmp0_m, tmp1_m, tmp2_m, \
               tmp3_m);                                                        \
    ILVRL_B2_SB(tmp1_m, tmp0_m, tmp4_m, tmp5_m);                               \
    ILVRL_B2_SB(tmp3_m, tmp2_m, tmp6_m, tmp7_m);                               \
    ILVRL_W2(RTYPE, tmp6_m, tmp4_m, out0, out2);                               \
    ILVRL_W2(RTYPE, tmp7_m, tmp5_m, out4, out6);                               \
    SLDI_B2_0(RTYPE, out0, out2, out1, out3, 8);                               \
    SLDI_B2_0(RTYPE, out4, out6, out5, out7, 8);                               \
  }
#define TRANSPOSE8x8_UB_UB(...) TRANSPOSE8x8_UB(v16u8, __VA_ARGS__)

/* Description : Transpose 16x8 block into 8x16 with byte elements in vectors
   Arguments   : Inputs  - in0, in1, in2, in3, in4, in5, in6, in7,
                           in8, in9, in10, in11, in12, in13, in14, in15
                 Outputs - out0, out1, out2, out3, out4, out5, out6, out7
                 Return Type - unsigned byte
*/
#define TRANSPOSE16x8_UB_UB(in0, in1, in2, in3, in4, in5, in6, in7, in8, in9, \
                            in10, in11, in12, in13, in14, in15, out0, out1,   \
                            out2, out3, out4, out5, out6, out7)               \
  {                                                                           \
    v16u8 tmp0_m, tmp1_m, tmp2_m, tmp3_m;                                     \
    v16u8 tmp4_m, tmp5_m, tmp6_m, tmp7_m;                                     \
                                                                              \
    ILVEV_D2_UB(in0, in8, in1, in9, out7, out6);                              \
    ILVEV_D2_UB(in2, in10, in3, in11, out5, out4);                            \
    ILVEV_D2_UB(in4, in12, in5, in13, out3, out2);                            \
    ILVEV_D2_UB(in6, in14, in7, in15, out1, out0);                            \
                                                                              \
    tmp0_m = (v16u8)__msa_ilvev_b((v16i8)out6, (v16i8)out7);                  \
    tmp4_m = (v16u8)__msa_ilvod_b((v16i8)out6, (v16i8)out7);                  \
    tmp1_m = (v16u8)__msa_ilvev_b((v16i8)out4, (v16i8)out5);                  \
    tmp5_m = (v16u8)__msa_ilvod_b((v16i8)out4, (v16i8)out5);                  \
    out5 = (v16u8)__msa_ilvev_b((v16i8)out2, (v16i8)out3);                    \
    tmp6_m = (v16u8)__msa_ilvod_b((v16i8)out2, (v16i8)out3);                  \
    out7 = (v16u8)__msa_ilvev_b((v16i8)out0, (v16i8)out1);                    \
    tmp7_m = (v16u8)__msa_ilvod_b((v16i8)out0, (v16i8)out1);                  \
                                                                              \
    ILVEV_H2_UB(tmp0_m, tmp1_m, out5, out7, tmp2_m, tmp3_m);                  \
    out0 = (v16u8)__msa_ilvev_w((v4i32)tmp3_m, (v4i32)tmp2_m);                \
    out4 = (v16u8)__msa_ilvod_w((v4i32)tmp3_m, (v4i32)tmp2_m);                \
                                                                              \
    tmp2_m = (v16u8)__msa_ilvod_h((v8i16)tmp1_m, (v8i16)tmp0_m);              \
    tmp3_m = (v16u8)__msa_ilvod_h((v8i16)out7, (v8i16)out5);                  \
    out2 = (v16u8)__msa_ilvev_w((v4i32)tmp3_m, (v4i32)tmp2_m);                \
    out6 = (v16u8)__msa_ilvod_w((v4i32)tmp3_m, (v4i32)tmp2_m);                \
                                                                              \
    ILVEV_H2_UB(tmp4_m, tmp5_m, tmp6_m, tmp7_m, tmp2_m, tmp3_m);              \
    out1 = (v16u8)__msa_ilvev_w((v4i32)tmp3_m, (v4i32)tmp2_m);                \
    out5 = (v16u8)__msa_ilvod_w((v4i32)tmp3_m, (v4i32)tmp2_m);                \
                                                                              \
    tmp2_m = (v16u8)__msa_ilvod_h((v8i16)tmp5_m, (v8i16)tmp4_m);              \
    tmp2_m = (v16u8)__msa_ilvod_h((v8i16)tmp5_m, (v8i16)tmp4_m);              \
    tmp3_m = (v16u8)__msa_ilvod_h((v8i16)tmp7_m, (v8i16)tmp6_m);              \
    tmp3_m = (v16u8)__msa_ilvod_h((v8i16)tmp7_m, (v8i16)tmp6_m);              \
    out3 = (v16u8)__msa_ilvev_w((v4i32)tmp3_m, (v4i32)tmp2_m);                \
    out7 = (v16u8)__msa_ilvod_w((v4i32)tmp3_m, (v4i32)tmp2_m);                \
  }

/* Description : Transpose 4x4 block with half word elements in vectors
   Arguments   : Inputs  - in0, in1, in2, in3
                 Outputs - out0, out1, out2, out3
                 Return Type - signed halfword
*/
#define TRANSPOSE4x4_SH_SH(in0, in1, in2, in3, out0, out1, out2, out3) \
  {                                                                    \
    v8i16 s0_m, s1_m;                                                  \
                                                                       \
    ILVR_H2_SH(in1, in0, in3, in2, s0_m, s1_m);                        \
    ILVRL_W2_SH(s1_m, s0_m, out0, out2);                               \
    out1 = (v8i16)__msa_ilvl_d((v2i64)out0, (v2i64)out0);              \
    out3 = (v8i16)__msa_ilvl_d((v2i64)out0, (v2i64)out2);              \
  }

/* Description : Transpose 4x8 block with half word elements in vectors
   Arguments   : Inputs  - in0, in1, in2, in3, in4, in5, in6, in7
                 Outputs - out0, out1, out2, out3, out4, out5, out6, out7
                 Return Type - signed halfword
*/
#define TRANSPOSE4X8_SH_SH(in0, in1, in2, in3, in4, in5, in6, in7, out0, out1, \
                           out2, out3, out4, out5, out6, out7)                 \
  {                                                                            \
    v8i16 tmp0_m, tmp1_m, tmp2_m, tmp3_m;                                      \
    v8i16 tmp0_n, tmp1_n, tmp2_n, tmp3_n;                                      \
    v8i16 zero_m = { 0 };                                                      \
                                                                               \
    ILVR_H4_SH(in1, in0, in3, in2, in5, in4, in7, in6, tmp0_n, tmp1_n, tmp2_n, \
               tmp3_n);                                                        \
    ILVRL_W2_SH(tmp1_n, tmp0_n, tmp0_m, tmp2_m);                               \
    ILVRL_W2_SH(tmp3_n, tmp2_n, tmp1_m, tmp3_m);                               \
                                                                               \
    out0 = (v8i16)__msa_ilvr_d((v2i64)tmp1_m, (v2i64)tmp0_m);                  \
    out1 = (v8i16)__msa_ilvl_d((v2i64)tmp1_m, (v2i64)tmp0_m);                  \
    out2 = (v8i16)__msa_ilvr_d((v2i64)tmp3_m, (v2i64)tmp2_m);                  \
    out3 = (v8i16)__msa_ilvl_d((v2i64)tmp3_m, (v2i64)tmp2_m);                  \
                                                                               \
    out4 = zero_m;                                                             \
    out5 = zero_m;                                                             \
    out6 = zero_m;                                                             \
    out7 = zero_m;                                                             \
  }

/* Description : Transpose 8x4 block with half word elements in vectors
   Arguments   : Inputs  - in0, in1, in2, in3, in4, in5, in6, in7
                 Outputs - out0, out1, out2, out3, out4, out5, out6, out7
                 Return Type - signed halfword
*/
#define TRANSPOSE8X4_SH_SH(in0, in1, in2, in3, out0, out1, out2, out3) \
  {                                                                    \
    v8i16 tmp0_m, tmp1_m, tmp2_m, tmp3_m;                              \
                                                                       \
    ILVR_H2_SH(in1, in0, in3, in2, tmp0_m, tmp1_m);                    \
    ILVL_H2_SH(in1, in0, in3, in2, tmp2_m, tmp3_m);                    \
    ILVR_W2_SH(tmp1_m, tmp0_m, tmp3_m, tmp2_m, out0, out2);            \
    ILVL_W2_SH(tmp1_m, tmp0_m, tmp3_m, tmp2_m, out1, out3);            \
  }

/* Description : Transpose 8x8 block with half word elements in vectors
   Arguments   : Inputs  - in0, in1, in2, in3, in4, in5, in6, in7
                 Outputs - out0, out1, out2, out3, out4, out5, out6, out7
                 Return Type - as per RTYPE
*/
#define TRANSPOSE8x8_H(RTYPE, in0, in1, in2, in3, in4, in5, in6, in7, out0, \
                       out1, out2, out3, out4, out5, out6, out7)            \
  {                                                                         \
    v8i16 s0_m, s1_m;                                                       \
    v8i16 tmp0_m, tmp1_m, tmp2_m, tmp3_m;                                   \
    v8i16 tmp4_m, tmp5_m, tmp6_m, tmp7_m;                                   \
                                                                            \
    ILVR_H2_SH(in6, in4, in7, in5, s0_m, s1_m);                             \
    ILVRL_H2_SH(s1_m, s0_m, tmp0_m, tmp1_m);                                \
    ILVL_H2_SH(in6, in4, in7, in5, s0_m, s1_m);                             \
    ILVRL_H2_SH(s1_m, s0_m, tmp2_m, tmp3_m);                                \
    ILVR_H2_SH(in2, in0, in3, in1, s0_m, s1_m);                             \
    ILVRL_H2_SH(s1_m, s0_m, tmp4_m, tmp5_m);                                \
    ILVL_H2_SH(in2, in0, in3, in1, s0_m, s1_m);                             \
    ILVRL_H2_SH(s1_m, s0_m, tmp6_m, tmp7_m);                                \
    PCKEV_D4(RTYPE, tmp0_m, tmp4_m, tmp1_m, tmp5_m, tmp2_m, tmp6_m, tmp3_m, \
             tmp7_m, out0, out2, out4, out6);                               \
    out1 = (RTYPE)__msa_pckod_d((v2i64)tmp0_m, (v2i64)tmp4_m);              \
    out3 = (RTYPE)__msa_pckod_d((v2i64)tmp1_m, (v2i64)tmp5_m);              \
    out5 = (RTYPE)__msa_pckod_d((v2i64)tmp2_m, (v2i64)tmp6_m);              \
    out7 = (RTYPE)__msa_pckod_d((v2i64)tmp3_m, (v2i64)tmp7_m);              \
  }
#define TRANSPOSE8x8_SH_SH(...) TRANSPOSE8x8_H(v8i16, __VA_ARGS__)

/* Description : Transpose 4x4 block with word elements in vectors
   Arguments   : Inputs  - in0, in1, in2, in3
                 Outputs - out0, out1, out2, out3
                 Return Type - signed word
*/
#define TRANSPOSE4x4_SW_SW(in0, in1, in2, in3, out0, out1, out2, out3) \
  {                                                                    \
    v4i32 s0_m, s1_m, s2_m, s3_m;                                      \
                                                                       \
    ILVRL_W2_SW(in1, in0, s0_m, s1_m);                                 \
    ILVRL_W2_SW(in3, in2, s2_m, s3_m);                                 \
                                                                       \
    out0 = (v4i32)__msa_ilvr_d((v2i64)s2_m, (v2i64)s0_m);              \
    out1 = (v4i32)__msa_ilvl_d((v2i64)s2_m, (v2i64)s0_m);              \
    out2 = (v4i32)__msa_ilvr_d((v2i64)s3_m, (v2i64)s1_m);              \
    out3 = (v4i32)__msa_ilvl_d((v2i64)s3_m, (v2i64)s1_m);              \
  }

/* Description : Add block 4x4
   Arguments   : Inputs - in0, in1, in2, in3, pdst, stride
   Details     : Least significant 4 bytes from each input vector are added to
                 the destination bytes, clipped between 0-255 and stored.
*/
#define ADDBLK_ST4x4_UB(in0, in1, in2, in3, pdst, stride)        \
  {                                                              \
    uint32_t src0_m, src1_m, src2_m, src3_m;                     \
    v8i16 inp0_m, inp1_m, res0_m, res1_m;                        \
    v16i8 dst0_m = { 0 };                                        \
    v16i8 dst1_m = { 0 };                                        \
    v16i8 zero_m = { 0 };                                        \
                                                                 \
    ILVR_D2_SH(in1, in0, in3, in2, inp0_m, inp1_m)               \
    LW4(pdst, stride, src0_m, src1_m, src2_m, src3_m);           \
    INSERT_W2_SB(src0_m, src1_m, dst0_m);                        \
    INSERT_W2_SB(src2_m, src3_m, dst1_m);                        \
    ILVR_B2_SH(zero_m, dst0_m, zero_m, dst1_m, res0_m, res1_m);  \
    ADD2(res0_m, inp0_m, res1_m, inp1_m, res0_m, res1_m);        \
    CLIP_SH2_0_255(res0_m, res1_m);                              \
    PCKEV_B2_SB(res0_m, res0_m, res1_m, res1_m, dst0_m, dst1_m); \
    ST4x4_UB(dst0_m, dst1_m, 0, 1, 0, 1, pdst, stride);          \
  }

/* Description : Pack even elements of input vectors & xor with 128
   Arguments   : Inputs - in0, in1
                 Output - out_m
                 Return Type - unsigned byte
   Details     : Signed byte even elements from 'in0' and 'in1' are packed
                 together in one vector and the resulting vector is xor'ed with
                 128 to shift the range from signed to unsigned byte
*/
#define PCKEV_XORI128_UB(in0, in1)                        \
  ({                                                      \
    v16u8 out_m;                                          \
                                                          \
    out_m = (v16u8)__msa_pckev_b((v16i8)in1, (v16i8)in0); \
    out_m = (v16u8)__msa_xori_b((v16u8)out_m, 128);       \
    out_m;                                                \
  })

/* Description : Converts inputs to unsigned bytes, interleave, average & store
                 as 8x4 unsigned byte block
   Arguments   : Inputs - in0, in1, in2, in3, dst0, dst1, dst2, dst3,
                          pdst, stride
*/
#define CONVERT_UB_AVG_ST8x4_UB(in0, in1, in2, in3, dst0, dst1, dst2, dst3, \
                                pdst, stride)                               \
  {                                                                         \
    v16u8 tmp0_m, tmp1_m, tmp2_m, tmp3_m;                                   \
                                                                            \
    tmp0_m = PCKEV_XORI128_UB(in0, in1);                                    \
    tmp1_m = PCKEV_XORI128_UB(in2, in3);                                    \
    ILVR_D2_UB(dst1, dst0, dst3, dst2, tmp2_m, tmp3_m);                     \
    AVER_UB2_UB(tmp0_m, tmp2_m, tmp1_m, tmp3_m, tmp0_m, tmp1_m);            \
    ST8x4_UB(tmp0_m, tmp1_m, pdst, stride);                                 \
  }

/* Description : Pack even byte elements and store byte vector in destination
                 memory
   Arguments   : Inputs - in0, in1, pdst
*/
#define PCKEV_ST_SB(in0, in1, pdst)                \
  {                                                \
    v16i8 tmp_m;                                   \
                                                   \
    tmp_m = __msa_pckev_b((v16i8)in1, (v16i8)in0); \
    ST_SB(tmp_m, (pdst));                          \
  }

/* Description : Horizontal 2 tap filter kernel code
   Arguments   : Inputs - in0, in1, mask, coeff, shift
*/
#define HORIZ_2TAP_FILT_UH(in0, in1, mask, coeff, shift)        \
  ({                                                            \
    v16i8 tmp0_m;                                               \
    v8u16 tmp1_m;                                               \
                                                                \
    tmp0_m = __msa_vshf_b((v16i8)mask, (v16i8)in1, (v16i8)in0); \
    tmp1_m = __msa_dotp_u_h((v16u8)tmp0_m, (v16u8)coeff);       \
    tmp1_m = (v8u16)__msa_srari_h((v8i16)tmp1_m, shift);        \
                                                                \
    tmp1_m;                                                     \
  })
#endif /* AOM_DSP_MIPS_MACROS_MSA_H_ */