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
| /*===---- xmmintrin.h - Implementation of SSE intrinsics on PowerPC --------===
*
* Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
* See https://llvm.org/LICENSE.txt for license information.
* SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
*
*===-----------------------------------------------------------------------===
*/
/* Implemented from the specification included in the Intel C++ Compiler
User Guide and Reference, version 9.0. */
#ifndef NO_WARN_X86_INTRINSICS
/* This header file is to help porting code using Intel intrinsics
explicitly from x86_64 to powerpc64/powerpc64le.
Since X86 SSE intrinsics mainly handles __m128 type, PowerPC
VMX/VSX ISA is a good match for vector float SIMD operations.
However scalar float operations in vector (XMM) registers require
the POWER8 VSX ISA (2.07) level. There are differences for data
format and placement of float scalars in the vector register, which
require extra steps to match SSE scalar float semantics on POWER.
It should be noted that there's much difference between X86_64's
MXSCR and PowerISA's FPSCR/VSCR registers. It's recommended to use
portable <fenv.h> instead of access MXSCR directly.
Most SSE scalar float intrinsic operations can be performed more
efficiently as C language float scalar operations or optimized to
use vector SIMD operations. We recommend this for new applications. */
#error "Please read comment above. Use -DNO_WARN_X86_INTRINSICS to disable this error."
#endif
#ifndef _XMMINTRIN_H_INCLUDED
#define _XMMINTRIN_H_INCLUDED
#if defined(__linux__) && defined(__ppc64__)
/* Define four value permute mask */
#define _MM_SHUFFLE(w,x,y,z) (((w) << 6) | ((x) << 4) | ((y) << 2) | (z))
#include <altivec.h>
/* Avoid collisions between altivec.h and strict adherence to C++ and
C11 standards. This should eventually be done inside altivec.h itself,
but only after testing a full distro build. */
#if defined(__STRICT_ANSI__) && (defined(__cplusplus) || \
(defined(__STDC_VERSION__) && \
__STDC_VERSION__ >= 201112L))
#undef vector
#undef pixel
#undef bool
#endif
/* We need type definitions from the MMX header file. */
#include <mmintrin.h>
/* Get _mm_malloc () and _mm_free (). */
#if __STDC_HOSTED__
#include <mm_malloc.h>
#endif
/* The Intel API is flexible enough that we must allow aliasing with other
vector types, and their scalar components. */
typedef float __m128 __attribute__ ((__vector_size__ (16), __may_alias__));
/* Unaligned version of the same type. */
typedef float __m128_u __attribute__ ((__vector_size__ (16), __may_alias__,
__aligned__ (1)));
/* Internal data types for implementing the intrinsics. */
typedef float __v4sf __attribute__ ((__vector_size__ (16)));
/* Create an undefined vector. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_undefined_ps (void)
{
__m128 __Y = __Y;
return __Y;
}
/* Create a vector of zeros. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_setzero_ps (void)
{
return __extension__ (__m128){ 0.0f, 0.0f, 0.0f, 0.0f };
}
/* Load four SPFP values from P. The address must be 16-byte aligned. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_load_ps (float const *__P)
{
return ((__m128)vec_ld(0, (__v4sf*)__P));
}
/* Load four SPFP values from P. The address need not be 16-byte aligned. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_loadu_ps (float const *__P)
{
return (vec_vsx_ld(0, __P));
}
/* Load four SPFP values in reverse order. The address must be aligned. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_loadr_ps (float const *__P)
{
__v4sf __tmp;
__m128 result;
static const __vector unsigned char permute_vector =
{ 0x1C, 0x1D, 0x1E, 0x1F, 0x18, 0x19, 0x1A, 0x1B, 0x14, 0x15, 0x16,
0x17, 0x10, 0x11, 0x12, 0x13 };
__tmp = vec_ld (0, (__v4sf *) __P);
result = (__m128) vec_perm (__tmp, __tmp, permute_vector);
return result;
}
/* Create a vector with all four elements equal to F. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_set1_ps (float __F)
{
return __extension__ (__m128)(__v4sf){ __F, __F, __F, __F };
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_set_ps1 (float __F)
{
return _mm_set1_ps (__F);
}
/* Create the vector [Z Y X W]. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_set_ps (const float __Z, const float __Y, const float __X, const float __W)
{
return __extension__ (__m128)(__v4sf){ __W, __X, __Y, __Z };
}
/* Create the vector [W X Y Z]. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_setr_ps (float __Z, float __Y, float __X, float __W)
{
return __extension__ (__m128)(__v4sf){ __Z, __Y, __X, __W };
}
/* Store four SPFP values. The address must be 16-byte aligned. */
extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_store_ps (float *__P, __m128 __A)
{
vec_st((__v4sf)__A, 0, (__v4sf*)__P);
}
/* Store four SPFP values. The address need not be 16-byte aligned. */
extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_storeu_ps (float *__P, __m128 __A)
{
*(__m128_u *)__P = __A;
}
/* Store four SPFP values in reverse order. The address must be aligned. */
extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_storer_ps (float *__P, __m128 __A)
{
__v4sf __tmp;
static const __vector unsigned char permute_vector =
{ 0x1C, 0x1D, 0x1E, 0x1F, 0x18, 0x19, 0x1A, 0x1B, 0x14, 0x15, 0x16,
0x17, 0x10, 0x11, 0x12, 0x13 };
__tmp = (__m128) vec_perm (__A, __A, permute_vector);
_mm_store_ps (__P, __tmp);
}
/* Store the lower SPFP value across four words. */
extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_store1_ps (float *__P, __m128 __A)
{
__v4sf __va = vec_splat((__v4sf)__A, 0);
_mm_store_ps (__P, __va);
}
extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_store_ps1 (float *__P, __m128 __A)
{
_mm_store1_ps (__P, __A);
}
/* Create a vector with element 0 as F and the rest zero. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_set_ss (float __F)
{
return __extension__ (__m128)(__v4sf){ __F, 0.0f, 0.0f, 0.0f };
}
/* Sets the low SPFP value of A from the low value of B. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_move_ss (__m128 __A, __m128 __B)
{
static const __vector unsigned int mask = {0xffffffff, 0, 0, 0};
return (vec_sel ((__v4sf)__A, (__v4sf)__B, mask));
}
/* Create a vector with element 0 as *P and the rest zero. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_load_ss (float const *__P)
{
return _mm_set_ss (*__P);
}
/* Stores the lower SPFP value. */
extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_store_ss (float *__P, __m128 __A)
{
*__P = ((__v4sf)__A)[0];
}
/* Perform the respective operation on the lower SPFP (single-precision
floating-point) values of A and B; the upper three SPFP values are
passed through from A. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_add_ss (__m128 __A, __m128 __B)
{
#ifdef _ARCH_PWR7
__m128 a, b, c;
static const __vector unsigned int mask = {0xffffffff, 0, 0, 0};
/* PowerISA VSX does not allow partial (for just lower double)
results. So to insure we don't generate spurious exceptions
(from the upper double values) we splat the lower double
before we to the operation. */
a = vec_splat (__A, 0);
b = vec_splat (__B, 0);
c = a + b;
/* Then we merge the lower float result with the original upper
float elements from __A. */
return (vec_sel (__A, c, mask));
#else
__A[0] = __A[0] + __B[0];
return (__A);
#endif
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_sub_ss (__m128 __A, __m128 __B)
{
#ifdef _ARCH_PWR7
__m128 a, b, c;
static const __vector unsigned int mask = {0xffffffff, 0, 0, 0};
/* PowerISA VSX does not allow partial (for just lower double)
results. So to insure we don't generate spurious exceptions
(from the upper double values) we splat the lower double
before we to the operation. */
a = vec_splat (__A, 0);
b = vec_splat (__B, 0);
c = a - b;
/* Then we merge the lower float result with the original upper
float elements from __A. */
return (vec_sel (__A, c, mask));
#else
__A[0] = __A[0] - __B[0];
return (__A);
#endif
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_mul_ss (__m128 __A, __m128 __B)
{
#ifdef _ARCH_PWR7
__m128 a, b, c;
static const __vector unsigned int mask = {0xffffffff, 0, 0, 0};
/* PowerISA VSX does not allow partial (for just lower double)
results. So to insure we don't generate spurious exceptions
(from the upper double values) we splat the lower double
before we to the operation. */
a = vec_splat (__A, 0);
b = vec_splat (__B, 0);
c = a * b;
/* Then we merge the lower float result with the original upper
float elements from __A. */
return (vec_sel (__A, c, mask));
#else
__A[0] = __A[0] * __B[0];
return (__A);
#endif
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_div_ss (__m128 __A, __m128 __B)
{
#ifdef _ARCH_PWR7
__m128 a, b, c;
static const __vector unsigned int mask = {0xffffffff, 0, 0, 0};
/* PowerISA VSX does not allow partial (for just lower double)
results. So to insure we don't generate spurious exceptions
(from the upper double values) we splat the lower double
before we to the operation. */
a = vec_splat (__A, 0);
b = vec_splat (__B, 0);
c = a / b;
/* Then we merge the lower float result with the original upper
float elements from __A. */
return (vec_sel (__A, c, mask));
#else
__A[0] = __A[0] / __B[0];
return (__A);
#endif
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_sqrt_ss (__m128 __A)
{
__m128 a, c;
static const __vector unsigned int mask = {0xffffffff, 0, 0, 0};
/* PowerISA VSX does not allow partial (for just lower double)
* results. So to insure we don't generate spurious exceptions
* (from the upper double values) we splat the lower double
* before we to the operation. */
a = vec_splat (__A, 0);
c = vec_sqrt (a);
/* Then we merge the lower float result with the original upper
* float elements from __A. */
return (vec_sel (__A, c, mask));
}
/* Perform the respective operation on the four SPFP values in A and B. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_add_ps (__m128 __A, __m128 __B)
{
return (__m128) ((__v4sf)__A + (__v4sf)__B);
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_sub_ps (__m128 __A, __m128 __B)
{
return (__m128) ((__v4sf)__A - (__v4sf)__B);
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_mul_ps (__m128 __A, __m128 __B)
{
return (__m128) ((__v4sf)__A * (__v4sf)__B);
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_div_ps (__m128 __A, __m128 __B)
{
return (__m128) ((__v4sf)__A / (__v4sf)__B);
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_sqrt_ps (__m128 __A)
{
return (vec_sqrt ((__v4sf)__A));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_rcp_ps (__m128 __A)
{
return (vec_re ((__v4sf)__A));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_rsqrt_ps (__m128 __A)
{
return (vec_rsqrte (__A));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_rcp_ss (__m128 __A)
{
__m128 a, c;
static const __vector unsigned int mask = {0xffffffff, 0, 0, 0};
/* PowerISA VSX does not allow partial (for just lower double)
* results. So to insure we don't generate spurious exceptions
* (from the upper double values) we splat the lower double
* before we to the operation. */
a = vec_splat (__A, 0);
c = _mm_rcp_ps (a);
/* Then we merge the lower float result with the original upper
* float elements from __A. */
return (vec_sel (__A, c, mask));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_rsqrt_ss (__m128 __A)
{
__m128 a, c;
static const __vector unsigned int mask = {0xffffffff, 0, 0, 0};
/* PowerISA VSX does not allow partial (for just lower double)
* results. So to insure we don't generate spurious exceptions
* (from the upper double values) we splat the lower double
* before we to the operation. */
a = vec_splat (__A, 0);
c = vec_rsqrte (a);
/* Then we merge the lower float result with the original upper
* float elements from __A. */
return (vec_sel (__A, c, mask));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_min_ss (__m128 __A, __m128 __B)
{
__v4sf a, b, c;
static const __vector unsigned int mask = {0xffffffff, 0, 0, 0};
/* PowerISA VSX does not allow partial (for just lower float)
* results. So to insure we don't generate spurious exceptions
* (from the upper float values) we splat the lower float
* before we to the operation. */
a = vec_splat ((__v4sf)__A, 0);
b = vec_splat ((__v4sf)__B, 0);
c = vec_min (a, b);
/* Then we merge the lower float result with the original upper
* float elements from __A. */
return (vec_sel ((__v4sf)__A, c, mask));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_max_ss (__m128 __A, __m128 __B)
{
__v4sf a, b, c;
static const __vector unsigned int mask = {0xffffffff, 0, 0, 0};
/* PowerISA VSX does not allow partial (for just lower float)
* results. So to insure we don't generate spurious exceptions
* (from the upper float values) we splat the lower float
* before we to the operation. */
a = vec_splat (__A, 0);
b = vec_splat (__B, 0);
c = vec_max (a, b);
/* Then we merge the lower float result with the original upper
* float elements from __A. */
return (vec_sel ((__v4sf)__A, c, mask));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_min_ps (__m128 __A, __m128 __B)
{
__vector __bool int m = vec_cmpgt ((__v4sf) __B, (__v4sf) __A);
return vec_sel (__B, __A, m);
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_max_ps (__m128 __A, __m128 __B)
{
__vector __bool int m = vec_cmpgt ((__v4sf) __A, (__v4sf) __B);
return vec_sel (__B, __A, m);
}
/* Perform logical bit-wise operations on 128-bit values. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_and_ps (__m128 __A, __m128 __B)
{
return ((__m128)vec_and ((__v4sf)__A, (__v4sf)__B));
// return __builtin_ia32_andps (__A, __B);
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_andnot_ps (__m128 __A, __m128 __B)
{
return ((__m128)vec_andc ((__v4sf)__B, (__v4sf)__A));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_or_ps (__m128 __A, __m128 __B)
{
return ((__m128)vec_or ((__v4sf)__A, (__v4sf)__B));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_xor_ps (__m128 __A, __m128 __B)
{
return ((__m128)vec_xor ((__v4sf)__A, (__v4sf)__B));
}
/* Perform a comparison on the four SPFP values of A and B. For each
element, if the comparison is true, place a mask of all ones in the
result, otherwise a mask of zeros. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmpeq_ps (__m128 __A, __m128 __B)
{
return ((__m128)vec_cmpeq ((__v4sf)__A,(__v4sf) __B));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmplt_ps (__m128 __A, __m128 __B)
{
return ((__m128)vec_cmplt ((__v4sf)__A, (__v4sf)__B));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmple_ps (__m128 __A, __m128 __B)
{
return ((__m128)vec_cmple ((__v4sf)__A, (__v4sf)__B));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmpgt_ps (__m128 __A, __m128 __B)
{
return ((__m128)vec_cmpgt ((__v4sf)__A, (__v4sf)__B));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmpge_ps (__m128 __A, __m128 __B)
{
return ((__m128)vec_cmpge ((__v4sf)__A, (__v4sf)__B));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmpneq_ps (__m128 __A, __m128 __B)
{
__v4sf temp = (__v4sf ) vec_cmpeq ((__v4sf) __A, (__v4sf)__B);
return ((__m128)vec_nor (temp, temp));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmpnlt_ps (__m128 __A, __m128 __B)
{
return ((__m128)vec_cmpge ((__v4sf)__A, (__v4sf)__B));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmpnle_ps (__m128 __A, __m128 __B)
{
return ((__m128)vec_cmpgt ((__v4sf)__A, (__v4sf)__B));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmpngt_ps (__m128 __A, __m128 __B)
{
return ((__m128)vec_cmple ((__v4sf)__A, (__v4sf)__B));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmpnge_ps (__m128 __A, __m128 __B)
{
return ((__m128)vec_cmplt ((__v4sf)__A, (__v4sf)__B));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmpord_ps (__m128 __A, __m128 __B)
{
__vector unsigned int a, b;
__vector unsigned int c, d;
static const __vector unsigned int float_exp_mask =
{ 0x7f800000, 0x7f800000, 0x7f800000, 0x7f800000 };
a = (__vector unsigned int) vec_abs ((__v4sf)__A);
b = (__vector unsigned int) vec_abs ((__v4sf)__B);
c = (__vector unsigned int) vec_cmpgt (float_exp_mask, a);
d = (__vector unsigned int) vec_cmpgt (float_exp_mask, b);
return ((__m128 ) vec_and (c, d));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmpunord_ps (__m128 __A, __m128 __B)
{
__vector unsigned int a, b;
__vector unsigned int c, d;
static const __vector unsigned int float_exp_mask =
{ 0x7f800000, 0x7f800000, 0x7f800000, 0x7f800000 };
a = (__vector unsigned int) vec_abs ((__v4sf)__A);
b = (__vector unsigned int) vec_abs ((__v4sf)__B);
c = (__vector unsigned int) vec_cmpgt (a, float_exp_mask);
d = (__vector unsigned int) vec_cmpgt (b, float_exp_mask);
return ((__m128 ) vec_or (c, d));
}
/* Perform a comparison on the lower SPFP values of A and B. If the
comparison is true, place a mask of all ones in the result, otherwise a
mask of zeros. The upper three SPFP values are passed through from A. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmpeq_ss (__m128 __A, __m128 __B)
{
static const __vector unsigned int mask =
{ 0xffffffff, 0, 0, 0 };
__v4sf a, b, c;
/* PowerISA VMX does not allow partial (for just element 0)
* results. So to insure we don't generate spurious exceptions
* (from the upper elements) we splat the lower float
* before we to the operation. */
a = vec_splat ((__v4sf) __A, 0);
b = vec_splat ((__v4sf) __B, 0);
c = (__v4sf) vec_cmpeq(a, b);
/* Then we merge the lower float result with the original upper
* float elements from __A. */
return ((__m128)vec_sel ((__v4sf)__A, c, mask));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmplt_ss (__m128 __A, __m128 __B)
{
static const __vector unsigned int mask =
{ 0xffffffff, 0, 0, 0 };
__v4sf a, b, c;
/* PowerISA VMX does not allow partial (for just element 0)
* results. So to insure we don't generate spurious exceptions
* (from the upper elements) we splat the lower float
* before we to the operation. */
a = vec_splat ((__v4sf) __A, 0);
b = vec_splat ((__v4sf) __B, 0);
c = (__v4sf) vec_cmplt(a, b);
/* Then we merge the lower float result with the original upper
* float elements from __A. */
return ((__m128)vec_sel ((__v4sf)__A, c, mask));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmple_ss (__m128 __A, __m128 __B)
{
static const __vector unsigned int mask =
{ 0xffffffff, 0, 0, 0 };
__v4sf a, b, c;
/* PowerISA VMX does not allow partial (for just element 0)
* results. So to insure we don't generate spurious exceptions
* (from the upper elements) we splat the lower float
* before we to the operation. */
a = vec_splat ((__v4sf) __A, 0);
b = vec_splat ((__v4sf) __B, 0);
c = (__v4sf) vec_cmple(a, b);
/* Then we merge the lower float result with the original upper
* float elements from __A. */
return ((__m128)vec_sel ((__v4sf)__A, c, mask));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmpgt_ss (__m128 __A, __m128 __B)
{
static const __vector unsigned int mask =
{ 0xffffffff, 0, 0, 0 };
__v4sf a, b, c;
/* PowerISA VMX does not allow partial (for just element 0)
* results. So to insure we don't generate spurious exceptions
* (from the upper elements) we splat the lower float
* before we to the operation. */
a = vec_splat ((__v4sf) __A, 0);
b = vec_splat ((__v4sf) __B, 0);
c = (__v4sf) vec_cmpgt(a, b);
/* Then we merge the lower float result with the original upper
* float elements from __A. */
return ((__m128)vec_sel ((__v4sf)__A, c, mask));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmpge_ss (__m128 __A, __m128 __B)
{
static const __vector unsigned int mask =
{ 0xffffffff, 0, 0, 0 };
__v4sf a, b, c;
/* PowerISA VMX does not allow partial (for just element 0)
* results. So to insure we don't generate spurious exceptions
* (from the upper elements) we splat the lower float
* before we to the operation. */
a = vec_splat ((__v4sf) __A, 0);
b = vec_splat ((__v4sf) __B, 0);
c = (__v4sf) vec_cmpge(a, b);
/* Then we merge the lower float result with the original upper
* float elements from __A. */
return ((__m128)vec_sel ((__v4sf)__A, c, mask));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmpneq_ss (__m128 __A, __m128 __B)
{
static const __vector unsigned int mask =
{ 0xffffffff, 0, 0, 0 };
__v4sf a, b, c;
/* PowerISA VMX does not allow partial (for just element 0)
* results. So to insure we don't generate spurious exceptions
* (from the upper elements) we splat the lower float
* before we to the operation. */
a = vec_splat ((__v4sf) __A, 0);
b = vec_splat ((__v4sf) __B, 0);
c = (__v4sf) vec_cmpeq(a, b);
c = vec_nor (c, c);
/* Then we merge the lower float result with the original upper
* float elements from __A. */
return ((__m128)vec_sel ((__v4sf)__A, c, mask));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmpnlt_ss (__m128 __A, __m128 __B)
{
static const __vector unsigned int mask =
{ 0xffffffff, 0, 0, 0 };
__v4sf a, b, c;
/* PowerISA VMX does not allow partial (for just element 0)
* results. So to insure we don't generate spurious exceptions
* (from the upper elements) we splat the lower float
* before we to the operation. */
a = vec_splat ((__v4sf) __A, 0);
b = vec_splat ((__v4sf) __B, 0);
c = (__v4sf) vec_cmpge(a, b);
/* Then we merge the lower float result with the original upper
* float elements from __A. */
return ((__m128)vec_sel ((__v4sf)__A, c, mask));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmpnle_ss (__m128 __A, __m128 __B)
{
static const __vector unsigned int mask =
{ 0xffffffff, 0, 0, 0 };
__v4sf a, b, c;
/* PowerISA VMX does not allow partial (for just element 0)
* results. So to insure we don't generate spurious exceptions
* (from the upper elements) we splat the lower float
* before we to the operation. */
a = vec_splat ((__v4sf) __A, 0);
b = vec_splat ((__v4sf) __B, 0);
c = (__v4sf) vec_cmpgt(a, b);
/* Then we merge the lower float result with the original upper
* float elements from __A. */
return ((__m128)vec_sel ((__v4sf)__A, c, mask));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmpngt_ss (__m128 __A, __m128 __B)
{
static const __vector unsigned int mask =
{ 0xffffffff, 0, 0, 0 };
__v4sf a, b, c;
/* PowerISA VMX does not allow partial (for just element 0)
* results. So to insure we don't generate spurious exceptions
* (from the upper elements) we splat the lower float
* before we to the operation. */
a = vec_splat ((__v4sf) __A, 0);
b = vec_splat ((__v4sf) __B, 0);
c = (__v4sf) vec_cmple(a, b);
/* Then we merge the lower float result with the original upper
* float elements from __A. */
return ((__m128)vec_sel ((__v4sf)__A, c, mask));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmpnge_ss (__m128 __A, __m128 __B)
{
static const __vector unsigned int mask =
{ 0xffffffff, 0, 0, 0 };
__v4sf a, b, c;
/* PowerISA VMX does not allow partial (for just element 0)
* results. So to insure we don't generate spurious exceptions
* (from the upper elements) we splat the lower float
* before we do the operation. */
a = vec_splat ((__v4sf) __A, 0);
b = vec_splat ((__v4sf) __B, 0);
c = (__v4sf) vec_cmplt(a, b);
/* Then we merge the lower float result with the original upper
* float elements from __A. */
return ((__m128)vec_sel ((__v4sf)__A, c, mask));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmpord_ss (__m128 __A, __m128 __B)
{
__vector unsigned int a, b;
__vector unsigned int c, d;
static const __vector unsigned int float_exp_mask =
{ 0x7f800000, 0x7f800000, 0x7f800000, 0x7f800000 };
static const __vector unsigned int mask =
{ 0xffffffff, 0, 0, 0 };
a = (__vector unsigned int) vec_abs ((__v4sf)__A);
b = (__vector unsigned int) vec_abs ((__v4sf)__B);
c = (__vector unsigned int) vec_cmpgt (float_exp_mask, a);
d = (__vector unsigned int) vec_cmpgt (float_exp_mask, b);
c = vec_and (c, d);
/* Then we merge the lower float result with the original upper
* float elements from __A. */
return ((__m128)vec_sel ((__v4sf)__A, (__v4sf)c, mask));
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cmpunord_ss (__m128 __A, __m128 __B)
{
__vector unsigned int a, b;
__vector unsigned int c, d;
static const __vector unsigned int float_exp_mask =
{ 0x7f800000, 0x7f800000, 0x7f800000, 0x7f800000 };
static const __vector unsigned int mask =
{ 0xffffffff, 0, 0, 0 };
a = (__vector unsigned int) vec_abs ((__v4sf)__A);
b = (__vector unsigned int) vec_abs ((__v4sf)__B);
c = (__vector unsigned int) vec_cmpgt (a, float_exp_mask);
d = (__vector unsigned int) vec_cmpgt (b, float_exp_mask);
c = vec_or (c, d);
/* Then we merge the lower float result with the original upper
* float elements from __A. */
return ((__m128)vec_sel ((__v4sf)__A, (__v4sf)c, mask));
}
/* Compare the lower SPFP values of A and B and return 1 if true
and 0 if false. */
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_comieq_ss (__m128 __A, __m128 __B)
{
return (__A[0] == __B[0]);
}
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_comilt_ss (__m128 __A, __m128 __B)
{
return (__A[0] < __B[0]);
}
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_comile_ss (__m128 __A, __m128 __B)
{
return (__A[0] <= __B[0]);
}
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_comigt_ss (__m128 __A, __m128 __B)
{
return (__A[0] > __B[0]);
}
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_comige_ss (__m128 __A, __m128 __B)
{
return (__A[0] >= __B[0]);
}
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_comineq_ss (__m128 __A, __m128 __B)
{
return (__A[0] != __B[0]);
}
/* FIXME
* The __mm_ucomi??_ss implementations below are exactly the same as
* __mm_comi??_ss because GCC for PowerPC only generates unordered
* compares (scalar and vector).
* Technically __mm_comieq_ss et al should be using the ordered
* compare and signal for QNaNs.
* The __mm_ucomieq_sd et all should be OK, as is.
*/
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_ucomieq_ss (__m128 __A, __m128 __B)
{
return (__A[0] == __B[0]);
}
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_ucomilt_ss (__m128 __A, __m128 __B)
{
return (__A[0] < __B[0]);
}
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_ucomile_ss (__m128 __A, __m128 __B)
{
return (__A[0] <= __B[0]);
}
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_ucomigt_ss (__m128 __A, __m128 __B)
{
return (__A[0] > __B[0]);
}
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_ucomige_ss (__m128 __A, __m128 __B)
{
return (__A[0] >= __B[0]);
}
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_ucomineq_ss (__m128 __A, __m128 __B)
{
return (__A[0] != __B[0]);
}
extern __inline float __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvtss_f32 (__m128 __A)
{
return ((__v4sf)__A)[0];
}
/* Convert the lower SPFP value to a 32-bit integer according to the current
rounding mode. */
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvtss_si32 (__m128 __A)
{
__m64 res = 0;
#ifdef _ARCH_PWR8
double dtmp;
__asm__(
#ifdef __LITTLE_ENDIAN__
"xxsldwi %x0,%x0,%x0,3;\n"
#endif
"xscvspdp %x2,%x0;\n"
"fctiw %2,%2;\n"
"mfvsrd %1,%x2;\n"
: "+wa" (__A),
"=r" (res),
"=f" (dtmp)
: );
#else
res = __builtin_rint(__A[0]);
#endif
return (res);
}
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvt_ss2si (__m128 __A)
{
return _mm_cvtss_si32 (__A);
}
/* Convert the lower SPFP value to a 32-bit integer according to the
current rounding mode. */
/* Intel intrinsic. */
extern __inline long long __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvtss_si64 (__m128 __A)
{
__m64 res = 0;
#ifdef _ARCH_PWR8
double dtmp;
__asm__(
#ifdef __LITTLE_ENDIAN__
"xxsldwi %x0,%x0,%x0,3;\n"
#endif
"xscvspdp %x2,%x0;\n"
"fctid %2,%2;\n"
"mfvsrd %1,%x2;\n"
: "+wa" (__A),
"=r" (res),
"=f" (dtmp)
: );
#else
res = __builtin_llrint(__A[0]);
#endif
return (res);
}
/* Microsoft intrinsic. */
extern __inline long long __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvtss_si64x (__m128 __A)
{
return _mm_cvtss_si64 ((__v4sf) __A);
}
/* Constants for use with _mm_prefetch. */
enum _mm_hint
{
/* _MM_HINT_ET is _MM_HINT_T with set 3rd bit. */
_MM_HINT_ET0 = 7,
_MM_HINT_ET1 = 6,
_MM_HINT_T0 = 3,
_MM_HINT_T1 = 2,
_MM_HINT_T2 = 1,
_MM_HINT_NTA = 0
};
/* Loads one cache line from address P to a location "closer" to the
processor. The selector I specifies the type of prefetch operation. */
extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_prefetch (const void *__P, enum _mm_hint __I)
{
/* Current PowerPC will ignores the hint parameters. */
__builtin_prefetch (__P);
}
/* Convert the two lower SPFP values to 32-bit integers according to the
current rounding mode. Return the integers in packed form. */
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvtps_pi32 (__m128 __A)
{
/* Splat two lower SPFP values to both halves. */
__v4sf temp, rounded;
__vector unsigned long long result;
/* Splat two lower SPFP values to both halves. */
temp = (__v4sf) vec_splat ((__vector long long)__A, 0);
rounded = vec_rint(temp);
result = (__vector unsigned long long) vec_cts (rounded, 0);
return (__m64) ((__vector long long) result)[0];
}
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvt_ps2pi (__m128 __A)
{
return _mm_cvtps_pi32 (__A);
}
/* Truncate the lower SPFP value to a 32-bit integer. */
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvttss_si32 (__m128 __A)
{
/* Extract the lower float element. */
float temp = __A[0];
/* truncate to 32-bit integer and return. */
return temp;
}
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvtt_ss2si (__m128 __A)
{
return _mm_cvttss_si32 (__A);
}
/* Intel intrinsic. */
extern __inline long long __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvttss_si64 (__m128 __A)
{
/* Extract the lower float element. */
float temp = __A[0];
/* truncate to 32-bit integer and return. */
return temp;
}
/* Microsoft intrinsic. */
extern __inline long long __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvttss_si64x (__m128 __A)
{
/* Extract the lower float element. */
float temp = __A[0];
/* truncate to 32-bit integer and return. */
return temp;
}
/* Truncate the two lower SPFP values to 32-bit integers. Return the
integers in packed form. */
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvttps_pi32 (__m128 __A)
{
__v4sf temp;
__vector unsigned long long result;
/* Splat two lower SPFP values to both halves. */
temp = (__v4sf) vec_splat ((__vector long long)__A, 0);
result = (__vector unsigned long long) vec_cts (temp, 0);
return (__m64) ((__vector long long) result)[0];
}
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvtt_ps2pi (__m128 __A)
{
return _mm_cvttps_pi32 (__A);
}
/* Convert B to a SPFP value and insert it as element zero in A. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvtsi32_ss (__m128 __A, int __B)
{
float temp = __B;
__A[0] = temp;
return __A;
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvt_si2ss (__m128 __A, int __B)
{
return _mm_cvtsi32_ss (__A, __B);
}
/* Convert B to a SPFP value and insert it as element zero in A. */
/* Intel intrinsic. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvtsi64_ss (__m128 __A, long long __B)
{
float temp = __B;
__A[0] = temp;
return __A;
}
/* Microsoft intrinsic. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvtsi64x_ss (__m128 __A, long long __B)
{
return _mm_cvtsi64_ss (__A, __B);
}
/* Convert the two 32-bit values in B to SPFP form and insert them
as the two lower elements in A. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvtpi32_ps (__m128 __A, __m64 __B)
{
__vector signed int vm1;
__vector float vf1;
vm1 = (__vector signed int) (__vector unsigned long long) {__B, __B};
vf1 = (__vector float) vec_ctf (vm1, 0);
return ((__m128) (__vector unsigned long long)
{ ((__vector unsigned long long)vf1) [0],
((__vector unsigned long long)__A) [1]});
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvt_pi2ps (__m128 __A, __m64 __B)
{
return _mm_cvtpi32_ps (__A, __B);
}
/* Convert the four signed 16-bit values in A to SPFP form. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvtpi16_ps (__m64 __A)
{
__vector signed short vs8;
__vector signed int vi4;
__vector float vf1;
vs8 = (__vector signed short) (__vector unsigned long long) { __A, __A };
vi4 = vec_vupklsh (vs8);
vf1 = (__vector float) vec_ctf (vi4, 0);
return (__m128) vf1;
}
/* Convert the four unsigned 16-bit values in A to SPFP form. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvtpu16_ps (__m64 __A)
{
const __vector unsigned short zero =
{ 0, 0, 0, 0, 0, 0, 0, 0 };
__vector unsigned short vs8;
__vector unsigned int vi4;
__vector float vf1;
vs8 = (__vector unsigned short) (__vector unsigned long long) { __A, __A };
vi4 = (__vector unsigned int) vec_mergel
#ifdef __LITTLE_ENDIAN__
(vs8, zero);
#else
(zero, vs8);
#endif
vf1 = (__vector float) vec_ctf (vi4, 0);
return (__m128) vf1;
}
/* Convert the low four signed 8-bit values in A to SPFP form. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvtpi8_ps (__m64 __A)
{
__vector signed char vc16;
__vector signed short vs8;
__vector signed int vi4;
__vector float vf1;
vc16 = (__vector signed char) (__vector unsigned long long) { __A, __A };
vs8 = vec_vupkhsb (vc16);
vi4 = vec_vupkhsh (vs8);
vf1 = (__vector float) vec_ctf (vi4, 0);
return (__m128) vf1;
}
/* Convert the low four unsigned 8-bit values in A to SPFP form. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvtpu8_ps (__m64 __A)
{
const __vector unsigned char zero =
{ 0, 0, 0, 0, 0, 0, 0, 0 };
__vector unsigned char vc16;
__vector unsigned short vs8;
__vector unsigned int vi4;
__vector float vf1;
vc16 = (__vector unsigned char) (__vector unsigned long long) { __A, __A };
#ifdef __LITTLE_ENDIAN__
vs8 = (__vector unsigned short) vec_mergel (vc16, zero);
vi4 = (__vector unsigned int) vec_mergeh (vs8,
(__vector unsigned short) zero);
#else
vs8 = (__vector unsigned short) vec_mergel (zero, vc16);
vi4 = (__vector unsigned int) vec_mergeh ((__vector unsigned short) zero,
vs8);
#endif
vf1 = (__vector float) vec_ctf (vi4, 0);
return (__m128) vf1;
}
/* Convert the four signed 32-bit values in A and B to SPFP form. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvtpi32x2_ps (__m64 __A, __m64 __B)
{
__vector signed int vi4;
__vector float vf4;
vi4 = (__vector signed int) (__vector unsigned long long) { __A, __B };
vf4 = (__vector float) vec_ctf (vi4, 0);
return (__m128) vf4;
}
/* Convert the four SPFP values in A to four signed 16-bit integers. */
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvtps_pi16 (__m128 __A)
{
__v4sf rounded;
__vector signed int temp;
__vector unsigned long long result;
rounded = vec_rint(__A);
temp = vec_cts (rounded, 0);
result = (__vector unsigned long long) vec_pack (temp, temp);
return (__m64) ((__vector long long) result)[0];
}
/* Convert the four SPFP values in A to four signed 8-bit integers. */
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_cvtps_pi8 (__m128 __A)
{
__v4sf rounded;
__vector signed int tmp_i;
static const __vector signed int zero = {0, 0, 0, 0};
__vector signed short tmp_s;
__vector signed char res_v;
rounded = vec_rint(__A);
tmp_i = vec_cts (rounded, 0);
tmp_s = vec_pack (tmp_i, zero);
res_v = vec_pack (tmp_s, tmp_s);
return (__m64) ((__vector long long) res_v)[0];
}
/* Selects four specific SPFP values from A and B based on MASK. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_shuffle_ps (__m128 __A, __m128 __B, int const __mask)
{
unsigned long element_selector_10 = __mask & 0x03;
unsigned long element_selector_32 = (__mask >> 2) & 0x03;
unsigned long element_selector_54 = (__mask >> 4) & 0x03;
unsigned long element_selector_76 = (__mask >> 6) & 0x03;
static const unsigned int permute_selectors[4] =
{
#ifdef __LITTLE_ENDIAN__
0x03020100, 0x07060504, 0x0B0A0908, 0x0F0E0D0C
#else
0x00010203, 0x04050607, 0x08090A0B, 0x0C0D0E0F
#endif
};
__vector unsigned int t;
t[0] = permute_selectors[element_selector_10];
t[1] = permute_selectors[element_selector_32];
t[2] = permute_selectors[element_selector_54] + 0x10101010;
t[3] = permute_selectors[element_selector_76] + 0x10101010;
return vec_perm ((__v4sf) __A, (__v4sf)__B, (__vector unsigned char)t);
}
/* Selects and interleaves the upper two SPFP values from A and B. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_unpackhi_ps (__m128 __A, __m128 __B)
{
return (__m128) vec_vmrglw ((__v4sf) __A, (__v4sf)__B);
}
/* Selects and interleaves the lower two SPFP values from A and B. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_unpacklo_ps (__m128 __A, __m128 __B)
{
return (__m128) vec_vmrghw ((__v4sf) __A, (__v4sf)__B);
}
/* Sets the upper two SPFP values with 64-bits of data loaded from P;
the lower two values are passed through from A. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_loadh_pi (__m128 __A, __m64 const *__P)
{
__vector unsigned long long __a = (__vector unsigned long long)__A;
__vector unsigned long long __p = vec_splats(*__P);
__a [1] = __p [1];
return (__m128)__a;
}
/* Stores the upper two SPFP values of A into P. */
extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_storeh_pi (__m64 *__P, __m128 __A)
{
__vector unsigned long long __a = (__vector unsigned long long) __A;
*__P = __a[1];
}
/* Moves the upper two values of B into the lower two values of A. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_movehl_ps (__m128 __A, __m128 __B)
{
return (__m128) vec_mergel ((__vector unsigned long long)__B,
(__vector unsigned long long)__A);
}
/* Moves the lower two values of B into the upper two values of A. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_movelh_ps (__m128 __A, __m128 __B)
{
return (__m128) vec_mergeh ((__vector unsigned long long)__A,
(__vector unsigned long long)__B);
}
/* Sets the lower two SPFP values with 64-bits of data loaded from P;
the upper two values are passed through from A. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_loadl_pi (__m128 __A, __m64 const *__P)
{
__vector unsigned long long __a = (__vector unsigned long long)__A;
__vector unsigned long long __p = vec_splats(*__P);
__a [0] = __p [0];
return (__m128)__a;
}
/* Stores the lower two SPFP values of A into P. */
extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_storel_pi (__m64 *__P, __m128 __A)
{
__vector unsigned long long __a = (__vector unsigned long long) __A;
*__P = __a[0];
}
#ifdef _ARCH_PWR8
/* Intrinsic functions that require PowerISA 2.07 minimum. */
/* Creates a 4-bit mask from the most significant bits of the SPFP values. */
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_movemask_ps (__m128 __A)
{
__vector unsigned long long result;
static const __vector unsigned int perm_mask =
{
#ifdef __LITTLE_ENDIAN__
0x00204060, 0x80808080, 0x80808080, 0x80808080
#else
0x80808080, 0x80808080, 0x80808080, 0x00204060
#endif
};
result = ((__vector unsigned long long)
vec_vbpermq ((__vector unsigned char) __A,
(__vector unsigned char) perm_mask));
#ifdef __LITTLE_ENDIAN__
return result[1];
#else
return result[0];
#endif
}
#endif /* _ARCH_PWR8 */
/* Create a vector with all four elements equal to *P. */
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_load1_ps (float const *__P)
{
return _mm_set1_ps (*__P);
}
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_load_ps1 (float const *__P)
{
return _mm_load1_ps (__P);
}
/* Extracts one of the four words of A. The selector N must be immediate. */
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_extract_pi16 (__m64 const __A, int const __N)
{
unsigned int shiftr = __N & 3;
#ifdef __BIG_ENDIAN__
shiftr = 3 - shiftr;
#endif
return ((__A >> (shiftr * 16)) & 0xffff);
}
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_m_pextrw (__m64 const __A, int const __N)
{
return _mm_extract_pi16 (__A, __N);
}
/* Inserts word D into one of four words of A. The selector N must be
immediate. */
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_insert_pi16 (__m64 const __A, int const __D, int const __N)
{
const int shiftl = (__N & 3) * 16;
const __m64 shiftD = (const __m64) __D << shiftl;
const __m64 mask = 0xffffUL << shiftl;
__m64 result = (__A & (~mask)) | (shiftD & mask);
return (result);
}
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_m_pinsrw (__m64 const __A, int const __D, int const __N)
{
return _mm_insert_pi16 (__A, __D, __N);
}
/* Compute the element-wise maximum of signed 16-bit values. */
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_max_pi16 (__m64 __A, __m64 __B)
{
#if _ARCH_PWR8
__vector signed short a, b, r;
__vector __bool short c;
a = (__vector signed short)vec_splats (__A);
b = (__vector signed short)vec_splats (__B);
c = (__vector __bool short)vec_cmpgt (a, b);
r = vec_sel (b, a, c);
return (__m64) ((__vector long long) r)[0];
#else
__m64_union m1, m2, res;
m1.as_m64 = __A;
m2.as_m64 = __B;
res.as_short[0] =
(m1.as_short[0] > m2.as_short[0]) ? m1.as_short[0] : m2.as_short[0];
res.as_short[1] =
(m1.as_short[1] > m2.as_short[1]) ? m1.as_short[1] : m2.as_short[1];
res.as_short[2] =
(m1.as_short[2] > m2.as_short[2]) ? m1.as_short[2] : m2.as_short[2];
res.as_short[3] =
(m1.as_short[3] > m2.as_short[3]) ? m1.as_short[3] : m2.as_short[3];
return (__m64) res.as_m64;
#endif
}
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_m_pmaxsw (__m64 __A, __m64 __B)
{
return _mm_max_pi16 (__A, __B);
}
/* Compute the element-wise maximum of unsigned 8-bit values. */
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_max_pu8 (__m64 __A, __m64 __B)
{
#if _ARCH_PWR8
__vector unsigned char a, b, r;
__vector __bool char c;
a = (__vector unsigned char)vec_splats (__A);
b = (__vector unsigned char)vec_splats (__B);
c = (__vector __bool char)vec_cmpgt (a, b);
r = vec_sel (b, a, c);
return (__m64) ((__vector long long) r)[0];
#else
__m64_union m1, m2, res;
long i;
m1.as_m64 = __A;
m2.as_m64 = __B;
for (i = 0; i < 8; i++)
res.as_char[i] =
((unsigned char) m1.as_char[i] > (unsigned char) m2.as_char[i]) ?
m1.as_char[i] : m2.as_char[i];
return (__m64) res.as_m64;
#endif
}
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_m_pmaxub (__m64 __A, __m64 __B)
{
return _mm_max_pu8 (__A, __B);
}
/* Compute the element-wise minimum of signed 16-bit values. */
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_min_pi16 (__m64 __A, __m64 __B)
{
#if _ARCH_PWR8
__vector signed short a, b, r;
__vector __bool short c;
a = (__vector signed short)vec_splats (__A);
b = (__vector signed short)vec_splats (__B);
c = (__vector __bool short)vec_cmplt (a, b);
r = vec_sel (b, a, c);
return (__m64) ((__vector long long) r)[0];
#else
__m64_union m1, m2, res;
m1.as_m64 = __A;
m2.as_m64 = __B;
res.as_short[0] =
(m1.as_short[0] < m2.as_short[0]) ? m1.as_short[0] : m2.as_short[0];
res.as_short[1] =
(m1.as_short[1] < m2.as_short[1]) ? m1.as_short[1] : m2.as_short[1];
res.as_short[2] =
(m1.as_short[2] < m2.as_short[2]) ? m1.as_short[2] : m2.as_short[2];
res.as_short[3] =
(m1.as_short[3] < m2.as_short[3]) ? m1.as_short[3] : m2.as_short[3];
return (__m64) res.as_m64;
#endif
}
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_m_pminsw (__m64 __A, __m64 __B)
{
return _mm_min_pi16 (__A, __B);
}
/* Compute the element-wise minimum of unsigned 8-bit values. */
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_min_pu8 (__m64 __A, __m64 __B)
{
#if _ARCH_PWR8
__vector unsigned char a, b, r;
__vector __bool char c;
a = (__vector unsigned char)vec_splats (__A);
b = (__vector unsigned char)vec_splats (__B);
c = (__vector __bool char)vec_cmplt (a, b);
r = vec_sel (b, a, c);
return (__m64) ((__vector long long) r)[0];
#else
__m64_union m1, m2, res;
long i;
m1.as_m64 = __A;
m2.as_m64 = __B;
for (i = 0; i < 8; i++)
res.as_char[i] =
((unsigned char) m1.as_char[i] < (unsigned char) m2.as_char[i]) ?
m1.as_char[i] : m2.as_char[i];
return (__m64) res.as_m64;
#endif
}
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_m_pminub (__m64 __A, __m64 __B)
{
return _mm_min_pu8 (__A, __B);
}
/* Create an 8-bit mask of the signs of 8-bit values. */
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_movemask_pi8 (__m64 __A)
{
unsigned long long p =
#ifdef __LITTLE_ENDIAN__
0x0008101820283038UL; // permute control for sign bits
#else
0x3830282018100800UL; // permute control for sign bits
#endif
return __builtin_bpermd (p, __A);
}
extern __inline int __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_m_pmovmskb (__m64 __A)
{
return _mm_movemask_pi8 (__A);
}
/* Multiply four unsigned 16-bit values in A by four unsigned 16-bit values
in B and produce the high 16 bits of the 32-bit results. */
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_mulhi_pu16 (__m64 __A, __m64 __B)
{
__vector unsigned short a, b;
__vector unsigned short c;
__vector unsigned int w0, w1;
__vector unsigned char xform1 = {
#ifdef __LITTLE_ENDIAN__
0x02, 0x03, 0x12, 0x13, 0x06, 0x07, 0x16, 0x17,
0x0A, 0x0B, 0x1A, 0x1B, 0x0E, 0x0F, 0x1E, 0x1F
#else
0x00, 0x01, 0x10, 0x11, 0x04, 0x05, 0x14, 0x15,
0x00, 0x01, 0x10, 0x11, 0x04, 0x05, 0x14, 0x15
#endif
};
a = (__vector unsigned short)vec_splats (__A);
b = (__vector unsigned short)vec_splats (__B);
w0 = vec_vmuleuh (a, b);
w1 = vec_vmulouh (a, b);
c = (__vector unsigned short)vec_perm (w0, w1, xform1);
return (__m64) ((__vector long long) c)[0];
}
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_m_pmulhuw (__m64 __A, __m64 __B)
{
return _mm_mulhi_pu16 (__A, __B);
}
/* Return a combination of the four 16-bit values in A. The selector
must be an immediate. */
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_shuffle_pi16 (__m64 __A, int const __N)
{
unsigned long element_selector_10 = __N & 0x03;
unsigned long element_selector_32 = (__N >> 2) & 0x03;
unsigned long element_selector_54 = (__N >> 4) & 0x03;
unsigned long element_selector_76 = (__N >> 6) & 0x03;
static const unsigned short permute_selectors[4] =
{
#ifdef __LITTLE_ENDIAN__
0x0908, 0x0B0A, 0x0D0C, 0x0F0E
#else
0x0607, 0x0405, 0x0203, 0x0001
#endif
};
__m64_union t;
__vector unsigned long long a, p, r;
#ifdef __LITTLE_ENDIAN__
t.as_short[0] = permute_selectors[element_selector_10];
t.as_short[1] = permute_selectors[element_selector_32];
t.as_short[2] = permute_selectors[element_selector_54];
t.as_short[3] = permute_selectors[element_selector_76];
#else
t.as_short[3] = permute_selectors[element_selector_10];
t.as_short[2] = permute_selectors[element_selector_32];
t.as_short[1] = permute_selectors[element_selector_54];
t.as_short[0] = permute_selectors[element_selector_76];
#endif
p = vec_splats (t.as_m64);
a = vec_splats (__A);
r = vec_perm (a, a, (__vector unsigned char)p);
return (__m64) ((__vector long long) r)[0];
}
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_m_pshufw (__m64 __A, int const __N)
{
return _mm_shuffle_pi16 (__A, __N);
}
/* Conditionally store byte elements of A into P. The high bit of each
byte in the selector N determines whether the corresponding byte from
A is stored. */
extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_maskmove_si64 (__m64 __A, __m64 __N, char *__P)
{
__m64 hibit = 0x8080808080808080UL;
__m64 mask, tmp;
__m64 *p = (__m64*)__P;
tmp = *p;
mask = _mm_cmpeq_pi8 ((__N & hibit), hibit);
tmp = (tmp & (~mask)) | (__A & mask);
*p = tmp;
}
extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_m_maskmovq (__m64 __A, __m64 __N, char *__P)
{
_mm_maskmove_si64 (__A, __N, __P);
}
/* Compute the rounded averages of the unsigned 8-bit values in A and B. */
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_avg_pu8 (__m64 __A, __m64 __B)
{
__vector unsigned char a, b, c;
a = (__vector unsigned char)vec_splats (__A);
b = (__vector unsigned char)vec_splats (__B);
c = vec_avg (a, b);
return (__m64) ((__vector long long) c)[0];
}
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_m_pavgb (__m64 __A, __m64 __B)
{
return _mm_avg_pu8 (__A, __B);
}
/* Compute the rounded averages of the unsigned 16-bit values in A and B. */
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_avg_pu16 (__m64 __A, __m64 __B)
{
__vector unsigned short a, b, c;
a = (__vector unsigned short)vec_splats (__A);
b = (__vector unsigned short)vec_splats (__B);
c = vec_avg (a, b);
return (__m64) ((__vector long long) c)[0];
}
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_m_pavgw (__m64 __A, __m64 __B)
{
return _mm_avg_pu16 (__A, __B);
}
/* Compute the sum of the absolute differences of the unsigned 8-bit
values in A and B. Return the value in the lower 16-bit word; the
upper words are cleared. */
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_sad_pu8 (__m64 __A, __m64 __B)
{
__vector unsigned char a, b;
__vector unsigned char vmin, vmax, vabsdiff;
__vector signed int vsum;
const __vector unsigned int zero =
{ 0, 0, 0, 0 };
__m64_union result = {0};
a = (__vector unsigned char) (__vector unsigned long long) { 0UL, __A };
b = (__vector unsigned char) (__vector unsigned long long) { 0UL, __B };
vmin = vec_min (a, b);
vmax = vec_max (a, b);
vabsdiff = vec_sub (vmax, vmin);
/* Sum four groups of bytes into integers. */
vsum = (__vector signed int) vec_sum4s (vabsdiff, zero);
/* Sum across four integers with integer result. */
vsum = vec_sums (vsum, (__vector signed int) zero);
/* The sum is in the right most 32-bits of the vector result.
Transfer to a GPR and truncate to 16 bits. */
result.as_short[0] = vsum[3];
return result.as_m64;
}
extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_m_psadbw (__m64 __A, __m64 __B)
{
return _mm_sad_pu8 (__A, __B);
}
/* Stores the data in A to the address P without polluting the caches. */
extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_stream_pi (__m64 *__P, __m64 __A)
{
/* Use the data cache block touch for store transient. */
__asm__ (
" dcbtstt 0,%0"
:
: "b" (__P)
: "memory"
);
*__P = __A;
}
/* Likewise. The address must be 16-byte aligned. */
extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_stream_ps (float *__P, __m128 __A)
{
/* Use the data cache block touch for store transient. */
__asm__ (
" dcbtstt 0,%0"
:
: "b" (__P)
: "memory"
);
_mm_store_ps (__P, __A);
}
/* Guarantees that every preceding store is globally visible before
any subsequent store. */
extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_sfence (void)
{
/* Generate a light weight sync. */
__atomic_thread_fence (__ATOMIC_RELEASE);
}
/* The execution of the next instruction is delayed by an implementation
specific amount of time. The instruction does not modify the
architectural state. This is after the pop_options pragma because
it does not require SSE support in the processor--the encoding is a
nop on processors that do not support it. */
extern __inline void __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm_pause (void)
{
/* There is no exact match with this construct, but the following is
close to the desired effect. */
#if _ARCH_PWR8
/* On power8 and later processors we can depend on Program Priority
(PRI) and associated "very low" PPI setting. Since we don't know
what PPI this thread is running at we: 1) save the current PRI
from the PPR SPR into a local GRP, 2) set the PRI to "very low*
via the special or 31,31,31 encoding. 3) issue an "isync" to
insure the PRI change takes effect before we execute any more
instructions.
Now we can execute a lwsync (release barrier) while we execute
this thread at "very low" PRI. Finally we restore the original
PRI and continue execution. */
unsigned long __PPR;
__asm__ volatile (
" mfppr %0;"
" or 31,31,31;"
" isync;"
" lwsync;"
" isync;"
" mtppr %0;"
: "=r" (__PPR)
:
: "memory"
);
#else
/* For older processor where we may not even have Program Priority
controls we can only depend on Heavy Weight Sync. */
__atomic_thread_fence (__ATOMIC_SEQ_CST);
#endif
}
/* Transpose the 4x4 matrix composed of row[0-3]. */
#define _MM_TRANSPOSE4_PS(row0, row1, row2, row3) \
do { \
__v4sf __r0 = (row0), __r1 = (row1), __r2 = (row2), __r3 = (row3); \
__v4sf __t0 = vec_vmrghw (__r0, __r1); \
__v4sf __t1 = vec_vmrghw (__r2, __r3); \
__v4sf __t2 = vec_vmrglw (__r0, __r1); \
__v4sf __t3 = vec_vmrglw (__r2, __r3); \
(row0) = (__v4sf)vec_mergeh ((__vector long long)__t0, \
(__vector long long)__t1); \
(row1) = (__v4sf)vec_mergel ((__vector long long)__t0, \
(__vector long long)__t1); \
(row2) = (__v4sf)vec_mergeh ((__vector long long)__t2, \
(__vector long long)__t3); \
(row3) = (__v4sf)vec_mergel ((__vector long long)__t2, \
(__vector long long)__t3); \
} while (0)
/* For backward source compatibility. */
//# include <emmintrin.h>
#else
#include_next <xmmintrin.h>
#endif /* defined(__linux__) && defined(__ppc64__) */
#endif /* _XMMINTRIN_H_INCLUDED */
|