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
use crate::abi::{Abi, FnAbi, FnAbiLlvmExt, LlvmType, PassMode};
use crate::builder::Builder;
use crate::context::CodegenCx;
use crate::llvm;
use crate::type_::Type;
use crate::type_of::LayoutLlvmExt;
use crate::va_arg::emit_va_arg;
use crate::value::Value;

use rustc_codegen_ssa::base::{compare_simd_types, wants_msvc_seh};
use rustc_codegen_ssa::common::span_invalid_monomorphization_error;
use rustc_codegen_ssa::common::{IntPredicate, TypeKind};
use rustc_codegen_ssa::mir::operand::OperandRef;
use rustc_codegen_ssa::mir::place::PlaceRef;
use rustc_codegen_ssa::traits::*;
use rustc_hir as hir;
use rustc_middle::ty::layout::{FnAbiOf, HasTyCtxt, LayoutOf};
use rustc_middle::ty::{self, Ty};
use rustc_middle::{bug, span_bug};
use rustc_span::{sym, symbol::kw, Span, Symbol};
use rustc_target::abi::{self, Align, HasDataLayout, Primitive};
use rustc_target::spec::{HasTargetSpec, PanicStrategy};

use std::cmp::Ordering;
use std::iter;

fn get_simple_intrinsic<'ll>(
    cx: &CodegenCx<'ll, '_>,
    name: Symbol,
) -> Option<(&'ll Type, &'ll Value)> {
    let llvm_name = match name {
        sym::sqrtf32 => "llvm.sqrt.f32",
        sym::sqrtf64 => "llvm.sqrt.f64",
        sym::powif32 => "llvm.powi.f32",
        sym::powif64 => "llvm.powi.f64",
        sym::sinf32 => "llvm.sin.f32",
        sym::sinf64 => "llvm.sin.f64",
        sym::cosf32 => "llvm.cos.f32",
        sym::cosf64 => "llvm.cos.f64",
        sym::powf32 => "llvm.pow.f32",
        sym::powf64 => "llvm.pow.f64",
        sym::expf32 => "llvm.exp.f32",
        sym::expf64 => "llvm.exp.f64",
        sym::exp2f32 => "llvm.exp2.f32",
        sym::exp2f64 => "llvm.exp2.f64",
        sym::logf32 => "llvm.log.f32",
        sym::logf64 => "llvm.log.f64",
        sym::log10f32 => "llvm.log10.f32",
        sym::log10f64 => "llvm.log10.f64",
        sym::log2f32 => "llvm.log2.f32",
        sym::log2f64 => "llvm.log2.f64",
        sym::fmaf32 => "llvm.fma.f32",
        sym::fmaf64 => "llvm.fma.f64",
        sym::fabsf32 => "llvm.fabs.f32",
        sym::fabsf64 => "llvm.fabs.f64",
        sym::minnumf32 => "llvm.minnum.f32",
        sym::minnumf64 => "llvm.minnum.f64",
        sym::maxnumf32 => "llvm.maxnum.f32",
        sym::maxnumf64 => "llvm.maxnum.f64",
        sym::copysignf32 => "llvm.copysign.f32",
        sym::copysignf64 => "llvm.copysign.f64",
        sym::floorf32 => "llvm.floor.f32",
        sym::floorf64 => "llvm.floor.f64",
        sym::ceilf32 => "llvm.ceil.f32",
        sym::ceilf64 => "llvm.ceil.f64",
        sym::truncf32 => "llvm.trunc.f32",
        sym::truncf64 => "llvm.trunc.f64",
        sym::rintf32 => "llvm.rint.f32",
        sym::rintf64 => "llvm.rint.f64",
        sym::nearbyintf32 => "llvm.nearbyint.f32",
        sym::nearbyintf64 => "llvm.nearbyint.f64",
        sym::roundf32 => "llvm.round.f32",
        sym::roundf64 => "llvm.round.f64",
        _ => return None,
    };
    Some(cx.get_intrinsic(llvm_name))
}

impl<'ll, 'tcx> IntrinsicCallMethods<'tcx> for Builder<'_, 'll, 'tcx> {
    fn codegen_intrinsic_call(
        &mut self,
        instance: ty::Instance<'tcx>,
        fn_abi: &FnAbi<'tcx, Ty<'tcx>>,
        args: &[OperandRef<'tcx, &'ll Value>],
        llresult: &'ll Value,
        span: Span,
    ) {
        let tcx = self.tcx;
        let callee_ty = instance.ty(tcx, ty::ParamEnv::reveal_all());

        let ty::FnDef(def_id, substs) = *callee_ty.kind() else {
            bug!("expected fn item type, found {}", callee_ty);
        };

        let sig = callee_ty.fn_sig(tcx);
        let sig = tcx.normalize_erasing_late_bound_regions(ty::ParamEnv::reveal_all(), sig);
        let arg_tys = sig.inputs();
        let ret_ty = sig.output();
        let name = tcx.item_name(def_id);

        let llret_ty = self.layout_of(ret_ty).llvm_type(self);
        let result = PlaceRef::new_sized(llresult, fn_abi.ret.layout);

        let simple = get_simple_intrinsic(self, name);
        let llval = match name {
            _ if simple.is_some() => {
                let (simple_ty, simple_fn) = simple.unwrap();
                self.call(
                    simple_ty,
                    simple_fn,
                    &args.iter().map(|arg| arg.immediate()).collect::<Vec<_>>(),
                    None,
                )
            }
            sym::likely => {
                self.call_intrinsic("llvm.expect.i1", &[args[0].immediate(), self.const_bool(true)])
            }
            sym::unlikely => self
                .call_intrinsic("llvm.expect.i1", &[args[0].immediate(), self.const_bool(false)]),
            kw::Try => {
                try_intrinsic(
                    self,
                    args[0].immediate(),
                    args[1].immediate(),
                    args[2].immediate(),
                    llresult,
                );
                return;
            }
            sym::breakpoint => self.call_intrinsic("llvm.debugtrap", &[]),
            sym::va_copy => {
                self.call_intrinsic("llvm.va_copy", &[args[0].immediate(), args[1].immediate()])
            }
            sym::va_arg => {
                match fn_abi.ret.layout.abi {
                    abi::Abi::Scalar(scalar) => {
                        match scalar.primitive() {
                            Primitive::Int(..) => {
                                if self.cx().size_of(ret_ty).bytes() < 4 {
                                    // `va_arg` should not be called on an integer type
                                    // less than 4 bytes in length. If it is, promote
                                    // the integer to an `i32` and truncate the result
                                    // back to the smaller type.
                                    let promoted_result = emit_va_arg(self, args[0], tcx.types.i32);
                                    self.trunc(promoted_result, llret_ty)
                                } else {
                                    emit_va_arg(self, args[0], ret_ty)
                                }
                            }
                            Primitive::F64 | Primitive::Pointer => {
                                emit_va_arg(self, args[0], ret_ty)
                            }
                            // `va_arg` should never be used with the return type f32.
                            Primitive::F32 => bug!("the va_arg intrinsic does not work with `f32`"),
                        }
                    }
                    _ => bug!("the va_arg intrinsic does not work with non-scalar types"),
                }
            }

            sym::volatile_load | sym::unaligned_volatile_load => {
                let tp_ty = substs.type_at(0);
                let ptr = args[0].immediate();
                let load = if let PassMode::Cast(ty) = fn_abi.ret.mode {
                    let llty = ty.llvm_type(self);
                    let ptr = self.pointercast(ptr, self.type_ptr_to(llty));
                    self.volatile_load(llty, ptr)
                } else {
                    self.volatile_load(self.layout_of(tp_ty).llvm_type(self), ptr)
                };
                let align = if name == sym::unaligned_volatile_load {
                    1
                } else {
                    self.align_of(tp_ty).bytes() as u32
                };
                unsafe {
                    llvm::LLVMSetAlignment(load, align);
                }
                self.to_immediate(load, self.layout_of(tp_ty))
            }
            sym::volatile_store => {
                let dst = args[0].deref(self.cx());
                args[1].val.volatile_store(self, dst);
                return;
            }
            sym::unaligned_volatile_store => {
                let dst = args[0].deref(self.cx());
                args[1].val.unaligned_volatile_store(self, dst);
                return;
            }
            sym::prefetch_read_data
            | sym::prefetch_write_data
            | sym::prefetch_read_instruction
            | sym::prefetch_write_instruction => {
                let (rw, cache_type) = match name {
                    sym::prefetch_read_data => (0, 1),
                    sym::prefetch_write_data => (1, 1),
                    sym::prefetch_read_instruction => (0, 0),
                    sym::prefetch_write_instruction => (1, 0),
                    _ => bug!(),
                };
                self.call_intrinsic(
                    "llvm.prefetch",
                    &[
                        args[0].immediate(),
                        self.const_i32(rw),
                        args[1].immediate(),
                        self.const_i32(cache_type),
                    ],
                )
            }
            sym::ctlz
            | sym::ctlz_nonzero
            | sym::cttz
            | sym::cttz_nonzero
            | sym::ctpop
            | sym::bswap
            | sym::bitreverse
            | sym::rotate_left
            | sym::rotate_right
            | sym::saturating_add
            | sym::saturating_sub => {
                let ty = arg_tys[0];
                match int_type_width_signed(ty, self) {
                    Some((width, signed)) => match name {
                        sym::ctlz | sym::cttz => {
                            let y = self.const_bool(false);
                            self.call_intrinsic(
                                &format!("llvm.{}.i{}", name, width),
                                &[args[0].immediate(), y],
                            )
                        }
                        sym::ctlz_nonzero => {
                            let y = self.const_bool(true);
                            let llvm_name = &format!("llvm.ctlz.i{}", width);
                            self.call_intrinsic(llvm_name, &[args[0].immediate(), y])
                        }
                        sym::cttz_nonzero => {
                            let y = self.const_bool(true);
                            let llvm_name = &format!("llvm.cttz.i{}", width);
                            self.call_intrinsic(llvm_name, &[args[0].immediate(), y])
                        }
                        sym::ctpop => self.call_intrinsic(
                            &format!("llvm.ctpop.i{}", width),
                            &[args[0].immediate()],
                        ),
                        sym::bswap => {
                            if width == 8 {
                                args[0].immediate() // byte swap a u8/i8 is just a no-op
                            } else {
                                self.call_intrinsic(
                                    &format!("llvm.bswap.i{}", width),
                                    &[args[0].immediate()],
                                )
                            }
                        }
                        sym::bitreverse => self.call_intrinsic(
                            &format!("llvm.bitreverse.i{}", width),
                            &[args[0].immediate()],
                        ),
                        sym::rotate_left | sym::rotate_right => {
                            let is_left = name == sym::rotate_left;
                            let val = args[0].immediate();
                            let raw_shift = args[1].immediate();
                            // rotate = funnel shift with first two args the same
                            let llvm_name =
                                &format!("llvm.fsh{}.i{}", if is_left { 'l' } else { 'r' }, width);
                            self.call_intrinsic(llvm_name, &[val, val, raw_shift])
                        }
                        sym::saturating_add | sym::saturating_sub => {
                            let is_add = name == sym::saturating_add;
                            let lhs = args[0].immediate();
                            let rhs = args[1].immediate();
                            let llvm_name = &format!(
                                "llvm.{}{}.sat.i{}",
                                if signed { 's' } else { 'u' },
                                if is_add { "add" } else { "sub" },
                                width
                            );
                            self.call_intrinsic(llvm_name, &[lhs, rhs])
                        }
                        _ => bug!(),
                    },
                    None => {
                        span_invalid_monomorphization_error(
                            tcx.sess,
                            span,
                            &format!(
                                "invalid monomorphization of `{}` intrinsic: \
                                      expected basic integer type, found `{}`",
                                name, ty
                            ),
                        );
                        return;
                    }
                }
            }

            sym::raw_eq => {
                use abi::Abi::*;
                let tp_ty = substs.type_at(0);
                let layout = self.layout_of(tp_ty).layout;
                let use_integer_compare = match layout.abi() {
                    Scalar(_) | ScalarPair(_, _) => true,
                    Uninhabited | Vector { .. } => false,
                    Aggregate { .. } => {
                        // For rusty ABIs, small aggregates are actually passed
                        // as `RegKind::Integer` (see `FnAbi::adjust_for_abi`),
                        // so we re-use that same threshold here.
                        layout.size() <= self.data_layout().pointer_size * 2
                    }
                };

                let a = args[0].immediate();
                let b = args[1].immediate();
                if layout.size().bytes() == 0 {
                    self.const_bool(true)
                } else if use_integer_compare {
                    let integer_ty = self.type_ix(layout.size().bits());
                    let ptr_ty = self.type_ptr_to(integer_ty);
                    let a_ptr = self.bitcast(a, ptr_ty);
                    let a_val = self.load(integer_ty, a_ptr, layout.align().abi);
                    let b_ptr = self.bitcast(b, ptr_ty);
                    let b_val = self.load(integer_ty, b_ptr, layout.align().abi);
                    self.icmp(IntPredicate::IntEQ, a_val, b_val)
                } else {
                    let i8p_ty = self.type_i8p();
                    let a_ptr = self.bitcast(a, i8p_ty);
                    let b_ptr = self.bitcast(b, i8p_ty);
                    let n = self.const_usize(layout.size().bytes());
                    let cmp = self.call_intrinsic("memcmp", &[a_ptr, b_ptr, n]);
                    match self.cx.sess().target.arch.as_ref() {
                        "avr" | "msp430" => self.icmp(IntPredicate::IntEQ, cmp, self.const_i16(0)),
                        _ => self.icmp(IntPredicate::IntEQ, cmp, self.const_i32(0)),
                    }
                }
            }

            sym::black_box => {
                args[0].val.store(self, result);

                // We need to "use" the argument in some way LLVM can't introspect, and on
                // targets that support it we can typically leverage inline assembly to do
                // this. LLVM's interpretation of inline assembly is that it's, well, a black
                // box. This isn't the greatest implementation since it probably deoptimizes
                // more than we want, but it's so far good enough.
                crate::asm::inline_asm_call(
                    self,
                    "",
                    "r,~{memory}",
                    &[result.llval],
                    self.type_void(),
                    true,
                    false,
                    llvm::AsmDialect::Att,
                    &[span],
                    false,
                    None,
                )
                .unwrap_or_else(|| bug!("failed to generate inline asm call for `black_box`"));

                // We have copied the value to `result` already.
                return;
            }

            _ if name.as_str().starts_with("simd_") => {
                match generic_simd_intrinsic(self, name, callee_ty, args, ret_ty, llret_ty, span) {
                    Ok(llval) => llval,
                    Err(()) => return,
                }
            }

            _ => bug!("unknown intrinsic '{}'", name),
        };

        if !fn_abi.ret.is_ignore() {
            if let PassMode::Cast(ty) = fn_abi.ret.mode {
                let ptr_llty = self.type_ptr_to(ty.llvm_type(self));
                let ptr = self.pointercast(result.llval, ptr_llty);
                self.store(llval, ptr, result.align);
            } else {
                OperandRef::from_immediate_or_packed_pair(self, llval, result.layout)
                    .val
                    .store(self, result);
            }
        }
    }

    fn abort(&mut self) {
        self.call_intrinsic("llvm.trap", &[]);
    }

    fn assume(&mut self, val: Self::Value) {
        self.call_intrinsic("llvm.assume", &[val]);
    }

    fn expect(&mut self, cond: Self::Value, expected: bool) -> Self::Value {
        self.call_intrinsic("llvm.expect.i1", &[cond, self.const_bool(expected)])
    }

    fn type_test(&mut self, pointer: Self::Value, typeid: Self::Value) -> Self::Value {
        // Test the called operand using llvm.type.test intrinsic. The LowerTypeTests link-time
        // optimization pass replaces calls to this intrinsic with code to test type membership.
        let i8p_ty = self.type_i8p();
        let bitcast = self.bitcast(pointer, i8p_ty);
        self.call_intrinsic("llvm.type.test", &[bitcast, typeid])
    }

    fn type_checked_load(
        &mut self,
        llvtable: &'ll Value,
        vtable_byte_offset: u64,
        typeid: &'ll Value,
    ) -> Self::Value {
        let vtable_byte_offset = self.const_i32(vtable_byte_offset as i32);
        self.call_intrinsic("llvm.type.checked.load", &[llvtable, vtable_byte_offset, typeid])
    }

    fn va_start(&mut self, va_list: &'ll Value) -> &'ll Value {
        self.call_intrinsic("llvm.va_start", &[va_list])
    }

    fn va_end(&mut self, va_list: &'ll Value) -> &'ll Value {
        self.call_intrinsic("llvm.va_end", &[va_list])
    }
}

fn try_intrinsic<'ll>(
    bx: &mut Builder<'_, 'll, '_>,
    try_func: &'ll Value,
    data: &'ll Value,
    catch_func: &'ll Value,
    dest: &'ll Value,
) {
    if bx.sess().panic_strategy() == PanicStrategy::Abort {
        let try_func_ty = bx.type_func(&[bx.type_i8p()], bx.type_void());
        bx.call(try_func_ty, try_func, &[data], None);
        // Return 0 unconditionally from the intrinsic call;
        // we can never unwind.
        let ret_align = bx.tcx().data_layout.i32_align.abi;
        bx.store(bx.const_i32(0), dest, ret_align);
    } else if wants_msvc_seh(bx.sess()) {
        codegen_msvc_try(bx, try_func, data, catch_func, dest);
    } else if bx.sess().target.os == "emscripten" {
        codegen_emcc_try(bx, try_func, data, catch_func, dest);
    } else {
        codegen_gnu_try(bx, try_func, data, catch_func, dest);
    }
}

// MSVC's definition of the `rust_try` function.
//
// This implementation uses the new exception handling instructions in LLVM
// which have support in LLVM for SEH on MSVC targets. Although these
// instructions are meant to work for all targets, as of the time of this
// writing, however, LLVM does not recommend the usage of these new instructions
// as the old ones are still more optimized.
fn codegen_msvc_try<'ll>(
    bx: &mut Builder<'_, 'll, '_>,
    try_func: &'ll Value,
    data: &'ll Value,
    catch_func: &'ll Value,
    dest: &'ll Value,
) {
    let (llty, llfn) = get_rust_try_fn(bx, &mut |mut bx| {
        bx.set_personality_fn(bx.eh_personality());

        let normal = bx.append_sibling_block("normal");
        let catchswitch = bx.append_sibling_block("catchswitch");
        let catchpad_rust = bx.append_sibling_block("catchpad_rust");
        let catchpad_foreign = bx.append_sibling_block("catchpad_foreign");
        let caught = bx.append_sibling_block("caught");

        let try_func = llvm::get_param(bx.llfn(), 0);
        let data = llvm::get_param(bx.llfn(), 1);
        let catch_func = llvm::get_param(bx.llfn(), 2);

        // We're generating an IR snippet that looks like:
        //
        //   declare i32 @rust_try(%try_func, %data, %catch_func) {
        //      %slot = alloca i8*
        //      invoke %try_func(%data) to label %normal unwind label %catchswitch
        //
        //   normal:
        //      ret i32 0
        //
        //   catchswitch:
        //      %cs = catchswitch within none [%catchpad_rust, %catchpad_foreign] unwind to caller
        //
        //   catchpad_rust:
        //      %tok = catchpad within %cs [%type_descriptor, 8, %slot]
        //      %ptr = load %slot
        //      call %catch_func(%data, %ptr)
        //      catchret from %tok to label %caught
        //
        //   catchpad_foreign:
        //      %tok = catchpad within %cs [null, 64, null]
        //      call %catch_func(%data, null)
        //      catchret from %tok to label %caught
        //
        //   caught:
        //      ret i32 1
        //   }
        //
        // This structure follows the basic usage of throw/try/catch in LLVM.
        // For example, compile this C++ snippet to see what LLVM generates:
        //
        //      struct rust_panic {
        //          rust_panic(const rust_panic&);
        //          ~rust_panic();
        //
        //          void* x[2];
        //      };
        //
        //      int __rust_try(
        //          void (*try_func)(void*),
        //          void *data,
        //          void (*catch_func)(void*, void*) noexcept
        //      ) {
        //          try {
        //              try_func(data);
        //              return 0;
        //          } catch(rust_panic& a) {
        //              catch_func(data, &a);
        //              return 1;
        //          } catch(...) {
        //              catch_func(data, NULL);
        //              return 1;
        //          }
        //      }
        //
        // More information can be found in libstd's seh.rs implementation.
        let ptr_align = bx.tcx().data_layout.pointer_align.abi;
        let slot = bx.alloca(bx.type_i8p(), ptr_align);
        let try_func_ty = bx.type_func(&[bx.type_i8p()], bx.type_void());
        bx.invoke(try_func_ty, try_func, &[data], normal, catchswitch, None);

        bx.switch_to_block(normal);
        bx.ret(bx.const_i32(0));

        bx.switch_to_block(catchswitch);
        let cs = bx.catch_switch(None, None, &[catchpad_rust, catchpad_foreign]);

        // We can't use the TypeDescriptor defined in libpanic_unwind because it
        // might be in another DLL and the SEH encoding only supports specifying
        // a TypeDescriptor from the current module.
        //
        // However this isn't an issue since the MSVC runtime uses string
        // comparison on the type name to match TypeDescriptors rather than
        // pointer equality.
        //
        // So instead we generate a new TypeDescriptor in each module that uses
        // `try` and let the linker merge duplicate definitions in the same
        // module.
        //
        // When modifying, make sure that the type_name string exactly matches
        // the one used in src/libpanic_unwind/seh.rs.
        let type_info_vtable = bx.declare_global("??_7type_info@@6B@", bx.type_i8p());
        let type_name = bx.const_bytes(b"rust_panic\0");
        let type_info =
            bx.const_struct(&[type_info_vtable, bx.const_null(bx.type_i8p()), type_name], false);
        let tydesc = bx.declare_global("__rust_panic_type_info", bx.val_ty(type_info));
        unsafe {
            llvm::LLVMRustSetLinkage(tydesc, llvm::Linkage::LinkOnceODRLinkage);
            llvm::SetUniqueComdat(bx.llmod, tydesc);
            llvm::LLVMSetInitializer(tydesc, type_info);
        }

        // The flag value of 8 indicates that we are catching the exception by
        // reference instead of by value. We can't use catch by value because
        // that requires copying the exception object, which we don't support
        // since our exception object effectively contains a Box.
        //
        // Source: MicrosoftCXXABI::getAddrOfCXXCatchHandlerType in clang
        bx.switch_to_block(catchpad_rust);
        let flags = bx.const_i32(8);
        let funclet = bx.catch_pad(cs, &[tydesc, flags, slot]);
        let ptr = bx.load(bx.type_i8p(), slot, ptr_align);
        let catch_ty = bx.type_func(&[bx.type_i8p(), bx.type_i8p()], bx.type_void());
        bx.call(catch_ty, catch_func, &[data, ptr], Some(&funclet));
        bx.catch_ret(&funclet, caught);

        // The flag value of 64 indicates a "catch-all".
        bx.switch_to_block(catchpad_foreign);
        let flags = bx.const_i32(64);
        let null = bx.const_null(bx.type_i8p());
        let funclet = bx.catch_pad(cs, &[null, flags, null]);
        bx.call(catch_ty, catch_func, &[data, null], Some(&funclet));
        bx.catch_ret(&funclet, caught);

        bx.switch_to_block(caught);
        bx.ret(bx.const_i32(1));
    });

    // Note that no invoke is used here because by definition this function
    // can't panic (that's what it's catching).
    let ret = bx.call(llty, llfn, &[try_func, data, catch_func], None);
    let i32_align = bx.tcx().data_layout.i32_align.abi;
    bx.store(ret, dest, i32_align);
}

// Definition of the standard `try` function for Rust using the GNU-like model
// of exceptions (e.g., the normal semantics of LLVM's `landingpad` and `invoke`
// instructions).
//
// This codegen is a little surprising because we always call a shim
// function instead of inlining the call to `invoke` manually here. This is done
// because in LLVM we're only allowed to have one personality per function
// definition. The call to the `try` intrinsic is being inlined into the
// function calling it, and that function may already have other personality
// functions in play. By calling a shim we're guaranteed that our shim will have
// the right personality function.
fn codegen_gnu_try<'ll>(
    bx: &mut Builder<'_, 'll, '_>,
    try_func: &'ll Value,
    data: &'ll Value,
    catch_func: &'ll Value,
    dest: &'ll Value,
) {
    let (llty, llfn) = get_rust_try_fn(bx, &mut |mut bx| {
        // Codegens the shims described above:
        //
        //   bx:
        //      invoke %try_func(%data) normal %normal unwind %catch
        //
        //   normal:
        //      ret 0
        //
        //   catch:
        //      (%ptr, _) = landingpad
        //      call %catch_func(%data, %ptr)
        //      ret 1
        let then = bx.append_sibling_block("then");
        let catch = bx.append_sibling_block("catch");

        let try_func = llvm::get_param(bx.llfn(), 0);
        let data = llvm::get_param(bx.llfn(), 1);
        let catch_func = llvm::get_param(bx.llfn(), 2);
        let try_func_ty = bx.type_func(&[bx.type_i8p()], bx.type_void());
        bx.invoke(try_func_ty, try_func, &[data], then, catch, None);

        bx.switch_to_block(then);
        bx.ret(bx.const_i32(0));

        // Type indicator for the exception being thrown.
        //
        // The first value in this tuple is a pointer to the exception object
        // being thrown.  The second value is a "selector" indicating which of
        // the landing pad clauses the exception's type had been matched to.
        // rust_try ignores the selector.
        bx.switch_to_block(catch);
        let lpad_ty = bx.type_struct(&[bx.type_i8p(), bx.type_i32()], false);
        let vals = bx.landing_pad(lpad_ty, bx.eh_personality(), 1);
        let tydesc = bx.const_null(bx.type_i8p());
        bx.add_clause(vals, tydesc);
        let ptr = bx.extract_value(vals, 0);
        let catch_ty = bx.type_func(&[bx.type_i8p(), bx.type_i8p()], bx.type_void());
        bx.call(catch_ty, catch_func, &[data, ptr], None);
        bx.ret(bx.const_i32(1));
    });

    // Note that no invoke is used here because by definition this function
    // can't panic (that's what it's catching).
    let ret = bx.call(llty, llfn, &[try_func, data, catch_func], None);
    let i32_align = bx.tcx().data_layout.i32_align.abi;
    bx.store(ret, dest, i32_align);
}

// Variant of codegen_gnu_try used for emscripten where Rust panics are
// implemented using C++ exceptions. Here we use exceptions of a specific type
// (`struct rust_panic`) to represent Rust panics.
fn codegen_emcc_try<'ll>(
    bx: &mut Builder<'_, 'll, '_>,
    try_func: &'ll Value,
    data: &'ll Value,
    catch_func: &'ll Value,
    dest: &'ll Value,
) {
    let (llty, llfn) = get_rust_try_fn(bx, &mut |mut bx| {
        // Codegens the shims described above:
        //
        //   bx:
        //      invoke %try_func(%data) normal %normal unwind %catch
        //
        //   normal:
        //      ret 0
        //
        //   catch:
        //      (%ptr, %selector) = landingpad
        //      %rust_typeid = @llvm.eh.typeid.for(@_ZTI10rust_panic)
        //      %is_rust_panic = %selector == %rust_typeid
        //      %catch_data = alloca { i8*, i8 }
        //      %catch_data[0] = %ptr
        //      %catch_data[1] = %is_rust_panic
        //      call %catch_func(%data, %catch_data)
        //      ret 1
        let then = bx.append_sibling_block("then");
        let catch = bx.append_sibling_block("catch");

        let try_func = llvm::get_param(bx.llfn(), 0);
        let data = llvm::get_param(bx.llfn(), 1);
        let catch_func = llvm::get_param(bx.llfn(), 2);
        let try_func_ty = bx.type_func(&[bx.type_i8p()], bx.type_void());
        bx.invoke(try_func_ty, try_func, &[data], then, catch, None);

        bx.switch_to_block(then);
        bx.ret(bx.const_i32(0));

        // Type indicator for the exception being thrown.
        //
        // The first value in this tuple is a pointer to the exception object
        // being thrown.  The second value is a "selector" indicating which of
        // the landing pad clauses the exception's type had been matched to.
        bx.switch_to_block(catch);
        let tydesc = bx.eh_catch_typeinfo();
        let lpad_ty = bx.type_struct(&[bx.type_i8p(), bx.type_i32()], false);
        let vals = bx.landing_pad(lpad_ty, bx.eh_personality(), 2);
        bx.add_clause(vals, tydesc);
        bx.add_clause(vals, bx.const_null(bx.type_i8p()));
        let ptr = bx.extract_value(vals, 0);
        let selector = bx.extract_value(vals, 1);

        // Check if the typeid we got is the one for a Rust panic.
        let rust_typeid = bx.call_intrinsic("llvm.eh.typeid.for", &[tydesc]);
        let is_rust_panic = bx.icmp(IntPredicate::IntEQ, selector, rust_typeid);
        let is_rust_panic = bx.zext(is_rust_panic, bx.type_bool());

        // We need to pass two values to catch_func (ptr and is_rust_panic), so
        // create an alloca and pass a pointer to that.
        let ptr_align = bx.tcx().data_layout.pointer_align.abi;
        let i8_align = bx.tcx().data_layout.i8_align.abi;
        let catch_data_type = bx.type_struct(&[bx.type_i8p(), bx.type_bool()], false);
        let catch_data = bx.alloca(catch_data_type, ptr_align);
        let catch_data_0 =
            bx.inbounds_gep(catch_data_type, catch_data, &[bx.const_usize(0), bx.const_usize(0)]);
        bx.store(ptr, catch_data_0, ptr_align);
        let catch_data_1 =
            bx.inbounds_gep(catch_data_type, catch_data, &[bx.const_usize(0), bx.const_usize(1)]);
        bx.store(is_rust_panic, catch_data_1, i8_align);
        let catch_data = bx.bitcast(catch_data, bx.type_i8p());

        let catch_ty = bx.type_func(&[bx.type_i8p(), bx.type_i8p()], bx.type_void());
        bx.call(catch_ty, catch_func, &[data, catch_data], None);
        bx.ret(bx.const_i32(1));
    });

    // Note that no invoke is used here because by definition this function
    // can't panic (that's what it's catching).
    let ret = bx.call(llty, llfn, &[try_func, data, catch_func], None);
    let i32_align = bx.tcx().data_layout.i32_align.abi;
    bx.store(ret, dest, i32_align);
}

// Helper function to give a Block to a closure to codegen a shim function.
// This is currently primarily used for the `try` intrinsic functions above.
fn gen_fn<'ll, 'tcx>(
    cx: &CodegenCx<'ll, 'tcx>,
    name: &str,
    rust_fn_sig: ty::PolyFnSig<'tcx>,
    codegen: &mut dyn FnMut(Builder<'_, 'll, 'tcx>),
) -> (&'ll Type, &'ll Value) {
    let fn_abi = cx.fn_abi_of_fn_ptr(rust_fn_sig, ty::List::empty());
    let llty = fn_abi.llvm_type(cx);
    let llfn = cx.declare_fn(name, fn_abi);
    cx.set_frame_pointer_type(llfn);
    cx.apply_target_cpu_attr(llfn);
    // FIXME(eddyb) find a nicer way to do this.
    unsafe { llvm::LLVMRustSetLinkage(llfn, llvm::Linkage::InternalLinkage) };
    let llbb = Builder::append_block(cx, llfn, "entry-block");
    let bx = Builder::build(cx, llbb);
    codegen(bx);
    (llty, llfn)
}

// Helper function used to get a handle to the `__rust_try` function used to
// catch exceptions.
//
// This function is only generated once and is then cached.
fn get_rust_try_fn<'ll, 'tcx>(
    cx: &CodegenCx<'ll, 'tcx>,
    codegen: &mut dyn FnMut(Builder<'_, 'll, 'tcx>),
) -> (&'ll Type, &'ll Value) {
    if let Some(llfn) = cx.rust_try_fn.get() {
        return llfn;
    }

    // Define the type up front for the signature of the rust_try function.
    let tcx = cx.tcx;
    let i8p = tcx.mk_mut_ptr(tcx.types.i8);
    // `unsafe fn(*mut i8) -> ()`
    let try_fn_ty = tcx.mk_fn_ptr(ty::Binder::dummy(tcx.mk_fn_sig(
        iter::once(i8p),
        tcx.mk_unit(),
        false,
        hir::Unsafety::Unsafe,
        Abi::Rust,
    )));
    // `unsafe fn(*mut i8, *mut i8) -> ()`
    let catch_fn_ty = tcx.mk_fn_ptr(ty::Binder::dummy(tcx.mk_fn_sig(
        [i8p, i8p].iter().cloned(),
        tcx.mk_unit(),
        false,
        hir::Unsafety::Unsafe,
        Abi::Rust,
    )));
    // `unsafe fn(unsafe fn(*mut i8) -> (), *mut i8, unsafe fn(*mut i8, *mut i8) -> ()) -> i32`
    let rust_fn_sig = ty::Binder::dummy(cx.tcx.mk_fn_sig(
        [try_fn_ty, i8p, catch_fn_ty].into_iter(),
        tcx.types.i32,
        false,
        hir::Unsafety::Unsafe,
        Abi::Rust,
    ));
    let rust_try = gen_fn(cx, "__rust_try", rust_fn_sig, codegen);
    cx.rust_try_fn.set(Some(rust_try));
    rust_try
}

fn generic_simd_intrinsic<'ll, 'tcx>(
    bx: &mut Builder<'_, 'll, 'tcx>,
    name: Symbol,
    callee_ty: Ty<'tcx>,
    args: &[OperandRef<'tcx, &'ll Value>],
    ret_ty: Ty<'tcx>,
    llret_ty: &'ll Type,
    span: Span,
) -> Result<&'ll Value, ()> {
    // macros for error handling:
    #[allow(unused_macro_rules)]
    macro_rules! emit_error {
        ($msg: tt) => {
            emit_error!($msg, )
        };
        ($msg: tt, $($fmt: tt)*) => {
            span_invalid_monomorphization_error(
                bx.sess(), span,
                &format!(concat!("invalid monomorphization of `{}` intrinsic: ", $msg),
                         name, $($fmt)*));
        }
    }

    macro_rules! return_error {
        ($($fmt: tt)*) => {
            {
                emit_error!($($fmt)*);
                return Err(());
            }
        }
    }

    macro_rules! require {
        ($cond: expr, $($fmt: tt)*) => {
            if !$cond {
                return_error!($($fmt)*);
            }
        };
    }

    macro_rules! require_simd {
        ($ty: expr, $position: expr) => {
            require!($ty.is_simd(), "expected SIMD {} type, found non-SIMD `{}`", $position, $ty)
        };
    }

    let tcx = bx.tcx();
    let sig =
        tcx.normalize_erasing_late_bound_regions(ty::ParamEnv::reveal_all(), callee_ty.fn_sig(tcx));
    let arg_tys = sig.inputs();

    if name == sym::simd_select_bitmask {
        require_simd!(arg_tys[1], "argument");
        let (len, _) = arg_tys[1].simd_size_and_type(bx.tcx());

        let expected_int_bits = (len.max(8) - 1).next_power_of_two();
        let expected_bytes = len / 8 + ((len % 8 > 0) as u64);

        let mask_ty = arg_tys[0];
        let mask = match mask_ty.kind() {
            ty::Int(i) if i.bit_width() == Some(expected_int_bits) => args[0].immediate(),
            ty::Uint(i) if i.bit_width() == Some(expected_int_bits) => args[0].immediate(),
            ty::Array(elem, len)
                if matches!(elem.kind(), ty::Uint(ty::UintTy::U8))
                    && len.try_eval_usize(bx.tcx, ty::ParamEnv::reveal_all())
                        == Some(expected_bytes) =>
            {
                let place = PlaceRef::alloca(bx, args[0].layout);
                args[0].val.store(bx, place);
                let int_ty = bx.type_ix(expected_bytes * 8);
                let ptr = bx.pointercast(place.llval, bx.cx.type_ptr_to(int_ty));
                bx.load(int_ty, ptr, Align::ONE)
            }
            _ => return_error!(
                "invalid bitmask `{}`, expected `u{}` or `[u8; {}]`",
                mask_ty,
                expected_int_bits,
                expected_bytes
            ),
        };

        let i1 = bx.type_i1();
        let im = bx.type_ix(len);
        let i1xn = bx.type_vector(i1, len);
        let m_im = bx.trunc(mask, im);
        let m_i1s = bx.bitcast(m_im, i1xn);
        return Ok(bx.select(m_i1s, args[1].immediate(), args[2].immediate()));
    }

    // every intrinsic below takes a SIMD vector as its first argument
    require_simd!(arg_tys[0], "input");
    let in_ty = arg_tys[0];

    let comparison = match name {
        sym::simd_eq => Some(hir::BinOpKind::Eq),
        sym::simd_ne => Some(hir::BinOpKind::Ne),
        sym::simd_lt => Some(hir::BinOpKind::Lt),
        sym::simd_le => Some(hir::BinOpKind::Le),
        sym::simd_gt => Some(hir::BinOpKind::Gt),
        sym::simd_ge => Some(hir::BinOpKind::Ge),
        _ => None,
    };

    let (in_len, in_elem) = arg_tys[0].simd_size_and_type(bx.tcx());
    if let Some(cmp_op) = comparison {
        require_simd!(ret_ty, "return");

        let (out_len, out_ty) = ret_ty.simd_size_and_type(bx.tcx());
        require!(
            in_len == out_len,
            "expected return type with length {} (same as input type `{}`), \
             found `{}` with length {}",
            in_len,
            in_ty,
            ret_ty,
            out_len
        );
        require!(
            bx.type_kind(bx.element_type(llret_ty)) == TypeKind::Integer,
            "expected return type with integer elements, found `{}` with non-integer `{}`",
            ret_ty,
            out_ty
        );

        return Ok(compare_simd_types(
            bx,
            args[0].immediate(),
            args[1].immediate(),
            in_elem,
            llret_ty,
            cmp_op,
        ));
    }

    if let Some(stripped) = name.as_str().strip_prefix("simd_shuffle") {
        // If this intrinsic is the older "simd_shuffleN" form, simply parse the integer.
        // If there is no suffix, use the index array length.
        let n: u64 = if stripped.is_empty() {
            // Make sure this is actually an array, since typeck only checks the length-suffixed
            // version of this intrinsic.
            match args[2].layout.ty.kind() {
                ty::Array(ty, len) if matches!(ty.kind(), ty::Uint(ty::UintTy::U32)) => {
                    len.try_eval_usize(bx.cx.tcx, ty::ParamEnv::reveal_all()).unwrap_or_else(|| {
                        span_bug!(span, "could not evaluate shuffle index array length")
                    })
                }
                _ => return_error!(
                    "simd_shuffle index must be an array of `u32`, got `{}`",
                    args[2].layout.ty
                ),
            }
        } else {
            stripped.parse().unwrap_or_else(|_| {
                span_bug!(span, "bad `simd_shuffle` instruction only caught in codegen?")
            })
        };

        require_simd!(ret_ty, "return");
        let (out_len, out_ty) = ret_ty.simd_size_and_type(bx.tcx());
        require!(
            out_len == n,
            "expected return type of length {}, found `{}` with length {}",
            n,
            ret_ty,
            out_len
        );
        require!(
            in_elem == out_ty,
            "expected return element type `{}` (element of input `{}`), \
             found `{}` with element type `{}`",
            in_elem,
            in_ty,
            ret_ty,
            out_ty
        );

        let total_len = u128::from(in_len) * 2;

        let vector = args[2].immediate();

        let indices: Option<Vec<_>> = (0..n)
            .map(|i| {
                let arg_idx = i;
                let val = bx.const_get_elt(vector, i as u64);
                match bx.const_to_opt_u128(val, true) {
                    None => {
                        emit_error!("shuffle index #{} is not a constant", arg_idx);
                        None
                    }
                    Some(idx) if idx >= total_len => {
                        emit_error!(
                            "shuffle index #{} is out of bounds (limit {})",
                            arg_idx,
                            total_len
                        );
                        None
                    }
                    Some(idx) => Some(bx.const_i32(idx as i32)),
                }
            })
            .collect();
        let Some(indices) = indices else {
            return Ok(bx.const_null(llret_ty));
        };

        return Ok(bx.shuffle_vector(
            args[0].immediate(),
            args[1].immediate(),
            bx.const_vector(&indices),
        ));
    }

    if name == sym::simd_insert {
        require!(
            in_elem == arg_tys[2],
            "expected inserted type `{}` (element of input `{}`), found `{}`",
            in_elem,
            in_ty,
            arg_tys[2]
        );
        return Ok(bx.insert_element(
            args[0].immediate(),
            args[2].immediate(),
            args[1].immediate(),
        ));
    }
    if name == sym::simd_extract {
        require!(
            ret_ty == in_elem,
            "expected return type `{}` (element of input `{}`), found `{}`",
            in_elem,
            in_ty,
            ret_ty
        );
        return Ok(bx.extract_element(args[0].immediate(), args[1].immediate()));
    }

    if name == sym::simd_select {
        let m_elem_ty = in_elem;
        let m_len = in_len;
        require_simd!(arg_tys[1], "argument");
        let (v_len, _) = arg_tys[1].simd_size_and_type(bx.tcx());
        require!(
            m_len == v_len,
            "mismatched lengths: mask length `{}` != other vector length `{}`",
            m_len,
            v_len
        );
        match m_elem_ty.kind() {
            ty::Int(_) => {}
            _ => return_error!("mask element type is `{}`, expected `i_`", m_elem_ty),
        }
        // truncate the mask to a vector of i1s
        let i1 = bx.type_i1();
        let i1xn = bx.type_vector(i1, m_len as u64);
        let m_i1s = bx.trunc(args[0].immediate(), i1xn);
        return Ok(bx.select(m_i1s, args[1].immediate(), args[2].immediate()));
    }

    if name == sym::simd_bitmask {
        // The `fn simd_bitmask(vector) -> unsigned integer` intrinsic takes a
        // vector mask and returns the most significant bit (MSB) of each lane in the form
        // of either:
        // * an unsigned integer
        // * an array of `u8`
        // If the vector has less than 8 lanes, a u8 is returned with zeroed trailing bits.
        //
        // The bit order of the result depends on the byte endianness, LSB-first for little
        // endian and MSB-first for big endian.
        let expected_int_bits = in_len.max(8);
        let expected_bytes = expected_int_bits / 8 + ((expected_int_bits % 8 > 0) as u64);

        // Integer vector <i{in_bitwidth} x in_len>:
        let (i_xn, in_elem_bitwidth) = match in_elem.kind() {
            ty::Int(i) => (
                args[0].immediate(),
                i.bit_width().unwrap_or_else(|| bx.data_layout().pointer_size.bits()),
            ),
            ty::Uint(i) => (
                args[0].immediate(),
                i.bit_width().unwrap_or_else(|| bx.data_layout().pointer_size.bits()),
            ),
            _ => return_error!(
                "vector argument `{}`'s element type `{}`, expected integer element type",
                in_ty,
                in_elem
            ),
        };

        // Shift the MSB to the right by "in_elem_bitwidth - 1" into the first bit position.
        let shift_indices =
            vec![
                bx.cx.const_int(bx.type_ix(in_elem_bitwidth), (in_elem_bitwidth - 1) as _);
                in_len as _
            ];
        let i_xn_msb = bx.lshr(i_xn, bx.const_vector(shift_indices.as_slice()));
        // Truncate vector to an <i1 x N>
        let i1xn = bx.trunc(i_xn_msb, bx.type_vector(bx.type_i1(), in_len));
        // Bitcast <i1 x N> to iN:
        let i_ = bx.bitcast(i1xn, bx.type_ix(in_len));

        match ret_ty.kind() {
            ty::Uint(i) if i.bit_width() == Some(expected_int_bits) => {
                // Zero-extend iN to the bitmask type:
                return Ok(bx.zext(i_, bx.type_ix(expected_int_bits)));
            }
            ty::Array(elem, len)
                if matches!(elem.kind(), ty::Uint(ty::UintTy::U8))
                    && len.try_eval_usize(bx.tcx, ty::ParamEnv::reveal_all())
                        == Some(expected_bytes) =>
            {
                // Zero-extend iN to the array length:
                let ze = bx.zext(i_, bx.type_ix(expected_bytes * 8));

                // Convert the integer to a byte array
                let ptr = bx.alloca(bx.type_ix(expected_bytes * 8), Align::ONE);
                bx.store(ze, ptr, Align::ONE);
                let array_ty = bx.type_array(bx.type_i8(), expected_bytes);
                let ptr = bx.pointercast(ptr, bx.cx.type_ptr_to(array_ty));
                return Ok(bx.load(array_ty, ptr, Align::ONE));
            }
            _ => return_error!(
                "cannot return `{}`, expected `u{}` or `[u8; {}]`",
                ret_ty,
                expected_int_bits,
                expected_bytes
            ),
        }
    }

    fn simd_simple_float_intrinsic<'ll, 'tcx>(
        name: Symbol,
        in_elem: Ty<'_>,
        in_ty: Ty<'_>,
        in_len: u64,
        bx: &mut Builder<'_, 'll, 'tcx>,
        span: Span,
        args: &[OperandRef<'tcx, &'ll Value>],
    ) -> Result<&'ll Value, ()> {
        #[allow(unused_macro_rules)]
        macro_rules! emit_error {
            ($msg: tt) => {
                emit_error!($msg, )
            };
            ($msg: tt, $($fmt: tt)*) => {
                span_invalid_monomorphization_error(
                    bx.sess(), span,
                    &format!(concat!("invalid monomorphization of `{}` intrinsic: ", $msg),
                             name, $($fmt)*));
            }
        }
        macro_rules! return_error {
            ($($fmt: tt)*) => {
                {
                    emit_error!($($fmt)*);
                    return Err(());
                }
            }
        }

        let (elem_ty_str, elem_ty) = if let ty::Float(f) = in_elem.kind() {
            let elem_ty = bx.cx.type_float_from_ty(*f);
            match f.bit_width() {
                32 => ("f32", elem_ty),
                64 => ("f64", elem_ty),
                _ => {
                    return_error!(
                        "unsupported element type `{}` of floating-point vector `{}`",
                        f.name_str(),
                        in_ty
                    );
                }
            }
        } else {
            return_error!("`{}` is not a floating-point type", in_ty);
        };

        let vec_ty = bx.type_vector(elem_ty, in_len);

        let (intr_name, fn_ty) = match name {
            sym::simd_ceil => ("ceil", bx.type_func(&[vec_ty], vec_ty)),
            sym::simd_fabs => ("fabs", bx.type_func(&[vec_ty], vec_ty)),
            sym::simd_fcos => ("cos", bx.type_func(&[vec_ty], vec_ty)),
            sym::simd_fexp2 => ("exp2", bx.type_func(&[vec_ty], vec_ty)),
            sym::simd_fexp => ("exp", bx.type_func(&[vec_ty], vec_ty)),
            sym::simd_flog10 => ("log10", bx.type_func(&[vec_ty], vec_ty)),
            sym::simd_flog2 => ("log2", bx.type_func(&[vec_ty], vec_ty)),
            sym::simd_flog => ("log", bx.type_func(&[vec_ty], vec_ty)),
            sym::simd_floor => ("floor", bx.type_func(&[vec_ty], vec_ty)),
            sym::simd_fma => ("fma", bx.type_func(&[vec_ty, vec_ty, vec_ty], vec_ty)),
            sym::simd_fpowi => ("powi", bx.type_func(&[vec_ty, bx.type_i32()], vec_ty)),
            sym::simd_fpow => ("pow", bx.type_func(&[vec_ty, vec_ty], vec_ty)),
            sym::simd_fsin => ("sin", bx.type_func(&[vec_ty], vec_ty)),
            sym::simd_fsqrt => ("sqrt", bx.type_func(&[vec_ty], vec_ty)),
            sym::simd_round => ("round", bx.type_func(&[vec_ty], vec_ty)),
            sym::simd_trunc => ("trunc", bx.type_func(&[vec_ty], vec_ty)),
            _ => return_error!("unrecognized intrinsic `{}`", name),
        };
        let llvm_name = &format!("llvm.{0}.v{1}{2}", intr_name, in_len, elem_ty_str);
        let f = bx.declare_cfn(llvm_name, llvm::UnnamedAddr::No, fn_ty);
        let c =
            bx.call(fn_ty, f, &args.iter().map(|arg| arg.immediate()).collect::<Vec<_>>(), None);
        Ok(c)
    }

    if std::matches!(
        name,
        sym::simd_ceil
            | sym::simd_fabs
            | sym::simd_fcos
            | sym::simd_fexp2
            | sym::simd_fexp
            | sym::simd_flog10
            | sym::simd_flog2
            | sym::simd_flog
            | sym::simd_floor
            | sym::simd_fma
            | sym::simd_fpow
            | sym::simd_fpowi
            | sym::simd_fsin
            | sym::simd_fsqrt
            | sym::simd_round
            | sym::simd_trunc
    ) {
        return simd_simple_float_intrinsic(name, in_elem, in_ty, in_len, bx, span, args);
    }

    // FIXME: use:
    //  https://github.com/llvm-mirror/llvm/blob/master/include/llvm/IR/Function.h#L182
    //  https://github.com/llvm-mirror/llvm/blob/master/include/llvm/IR/Intrinsics.h#L81
    fn llvm_vector_str(
        elem_ty: Ty<'_>,
        vec_len: u64,
        no_pointers: usize,
        bx: &Builder<'_, '_, '_>,
    ) -> String {
        let p0s: String = "p0".repeat(no_pointers);
        match *elem_ty.kind() {
            ty::Int(v) => format!(
                "v{}{}i{}",
                vec_len,
                p0s,
                // Normalize to prevent crash if v: IntTy::Isize
                v.normalize(bx.target_spec().pointer_width).bit_width().unwrap()
            ),
            ty::Uint(v) => format!(
                "v{}{}i{}",
                vec_len,
                p0s,
                // Normalize to prevent crash if v: UIntTy::Usize
                v.normalize(bx.target_spec().pointer_width).bit_width().unwrap()
            ),
            ty::Float(v) => format!("v{}{}f{}", vec_len, p0s, v.bit_width()),
            _ => unreachable!(),
        }
    }

    fn llvm_vector_ty<'ll>(
        cx: &CodegenCx<'ll, '_>,
        elem_ty: Ty<'_>,
        vec_len: u64,
        mut no_pointers: usize,
    ) -> &'ll Type {
        // FIXME: use cx.layout_of(ty).llvm_type() ?
        let mut elem_ty = match *elem_ty.kind() {
            ty::Int(v) => cx.type_int_from_ty(v),
            ty::Uint(v) => cx.type_uint_from_ty(v),
            ty::Float(v) => cx.type_float_from_ty(v),
            _ => unreachable!(),
        };
        while no_pointers > 0 {
            elem_ty = cx.type_ptr_to(elem_ty);
            no_pointers -= 1;
        }
        cx.type_vector(elem_ty, vec_len)
    }

    if name == sym::simd_gather {
        // simd_gather(values: <N x T>, pointers: <N x *_ T>,
        //             mask: <N x i{M}>) -> <N x T>
        // * N: number of elements in the input vectors
        // * T: type of the element to load
        // * M: any integer width is supported, will be truncated to i1

        // All types must be simd vector types
        require_simd!(in_ty, "first");
        require_simd!(arg_tys[1], "second");
        require_simd!(arg_tys[2], "third");
        require_simd!(ret_ty, "return");

        // Of the same length:
        let (out_len, _) = arg_tys[1].simd_size_and_type(bx.tcx());
        let (out_len2, _) = arg_tys[2].simd_size_and_type(bx.tcx());
        require!(
            in_len == out_len,
            "expected {} argument with length {} (same as input type `{}`), \
             found `{}` with length {}",
            "second",
            in_len,
            in_ty,
            arg_tys[1],
            out_len
        );
        require!(
            in_len == out_len2,
            "expected {} argument with length {} (same as input type `{}`), \
             found `{}` with length {}",
            "third",
            in_len,
            in_ty,
            arg_tys[2],
            out_len2
        );

        // The return type must match the first argument type
        require!(ret_ty == in_ty, "expected return type `{}`, found `{}`", in_ty, ret_ty);

        // This counts how many pointers
        fn ptr_count(t: Ty<'_>) -> usize {
            match t.kind() {
                ty::RawPtr(p) => 1 + ptr_count(p.ty),
                _ => 0,
            }
        }

        // Non-ptr type
        fn non_ptr(t: Ty<'_>) -> Ty<'_> {
            match t.kind() {
                ty::RawPtr(p) => non_ptr(p.ty),
                _ => t,
            }
        }

        // The second argument must be a simd vector with an element type that's a pointer
        // to the element type of the first argument
        let (_, element_ty0) = arg_tys[0].simd_size_and_type(bx.tcx());
        let (_, element_ty1) = arg_tys[1].simd_size_and_type(bx.tcx());
        let (pointer_count, underlying_ty) = match element_ty1.kind() {
            ty::RawPtr(p) if p.ty == in_elem => (ptr_count(element_ty1), non_ptr(element_ty1)),
            _ => {
                require!(
                    false,
                    "expected element type `{}` of second argument `{}` \
                        to be a pointer to the element type `{}` of the first \
                        argument `{}`, found `{}` != `*_ {}`",
                    element_ty1,
                    arg_tys[1],
                    in_elem,
                    in_ty,
                    element_ty1,
                    in_elem
                );
                unreachable!();
            }
        };
        assert!(pointer_count > 0);
        assert_eq!(pointer_count - 1, ptr_count(element_ty0));
        assert_eq!(underlying_ty, non_ptr(element_ty0));

        // The element type of the third argument must be a signed integer type of any width:
        let (_, element_ty2) = arg_tys[2].simd_size_and_type(bx.tcx());
        match element_ty2.kind() {
            ty::Int(_) => (),
            _ => {
                require!(
                    false,
                    "expected element type `{}` of third argument `{}` \
                                 to be a signed integer type",
                    element_ty2,
                    arg_tys[2]
                );
            }
        }

        // Alignment of T, must be a constant integer value:
        let alignment_ty = bx.type_i32();
        let alignment = bx.const_i32(bx.align_of(in_elem).bytes() as i32);

        // Truncate the mask vector to a vector of i1s:
        let (mask, mask_ty) = {
            let i1 = bx.type_i1();
            let i1xn = bx.type_vector(i1, in_len);
            (bx.trunc(args[2].immediate(), i1xn), i1xn)
        };

        // Type of the vector of pointers:
        let llvm_pointer_vec_ty = llvm_vector_ty(bx, underlying_ty, in_len, pointer_count);
        let llvm_pointer_vec_str = llvm_vector_str(underlying_ty, in_len, pointer_count, bx);

        // Type of the vector of elements:
        let llvm_elem_vec_ty = llvm_vector_ty(bx, underlying_ty, in_len, pointer_count - 1);
        let llvm_elem_vec_str = llvm_vector_str(underlying_ty, in_len, pointer_count - 1, bx);

        let llvm_intrinsic =
            format!("llvm.masked.gather.{}.{}", llvm_elem_vec_str, llvm_pointer_vec_str);
        let fn_ty = bx.type_func(
            &[llvm_pointer_vec_ty, alignment_ty, mask_ty, llvm_elem_vec_ty],
            llvm_elem_vec_ty,
        );
        let f = bx.declare_cfn(&llvm_intrinsic, llvm::UnnamedAddr::No, fn_ty);
        let v =
            bx.call(fn_ty, f, &[args[1].immediate(), alignment, mask, args[0].immediate()], None);
        return Ok(v);
    }

    if name == sym::simd_scatter {
        // simd_scatter(values: <N x T>, pointers: <N x *mut T>,
        //             mask: <N x i{M}>) -> ()
        // * N: number of elements in the input vectors
        // * T: type of the element to load
        // * M: any integer width is supported, will be truncated to i1

        // All types must be simd vector types
        require_simd!(in_ty, "first");
        require_simd!(arg_tys[1], "second");
        require_simd!(arg_tys[2], "third");

        // Of the same length:
        let (element_len1, _) = arg_tys[1].simd_size_and_type(bx.tcx());
        let (element_len2, _) = arg_tys[2].simd_size_and_type(bx.tcx());
        require!(
            in_len == element_len1,
            "expected {} argument with length {} (same as input type `{}`), \
            found `{}` with length {}",
            "second",
            in_len,
            in_ty,
            arg_tys[1],
            element_len1
        );
        require!(
            in_len == element_len2,
            "expected {} argument with length {} (same as input type `{}`), \
            found `{}` with length {}",
            "third",
            in_len,
            in_ty,
            arg_tys[2],
            element_len2
        );

        // This counts how many pointers
        fn ptr_count(t: Ty<'_>) -> usize {
            match t.kind() {
                ty::RawPtr(p) => 1 + ptr_count(p.ty),
                _ => 0,
            }
        }

        // Non-ptr type
        fn non_ptr(t: Ty<'_>) -> Ty<'_> {
            match t.kind() {
                ty::RawPtr(p) => non_ptr(p.ty),
                _ => t,
            }
        }

        // The second argument must be a simd vector with an element type that's a pointer
        // to the element type of the first argument
        let (_, element_ty0) = arg_tys[0].simd_size_and_type(bx.tcx());
        let (_, element_ty1) = arg_tys[1].simd_size_and_type(bx.tcx());
        let (_, element_ty2) = arg_tys[2].simd_size_and_type(bx.tcx());
        let (pointer_count, underlying_ty) = match element_ty1.kind() {
            ty::RawPtr(p) if p.ty == in_elem && p.mutbl == hir::Mutability::Mut => {
                (ptr_count(element_ty1), non_ptr(element_ty1))
            }
            _ => {
                require!(
                    false,
                    "expected element type `{}` of second argument `{}` \
                        to be a pointer to the element type `{}` of the first \
                        argument `{}`, found `{}` != `*mut {}`",
                    element_ty1,
                    arg_tys[1],
                    in_elem,
                    in_ty,
                    element_ty1,
                    in_elem
                );
                unreachable!();
            }
        };
        assert!(pointer_count > 0);
        assert_eq!(pointer_count - 1, ptr_count(element_ty0));
        assert_eq!(underlying_ty, non_ptr(element_ty0));

        // The element type of the third argument must be a signed integer type of any width:
        match element_ty2.kind() {
            ty::Int(_) => (),
            _ => {
                require!(
                    false,
                    "expected element type `{}` of third argument `{}` \
                         be a signed integer type",
                    element_ty2,
                    arg_tys[2]
                );
            }
        }

        // Alignment of T, must be a constant integer value:
        let alignment_ty = bx.type_i32();
        let alignment = bx.const_i32(bx.align_of(in_elem).bytes() as i32);

        // Truncate the mask vector to a vector of i1s:
        let (mask, mask_ty) = {
            let i1 = bx.type_i1();
            let i1xn = bx.type_vector(i1, in_len);
            (bx.trunc(args[2].immediate(), i1xn), i1xn)
        };

        let ret_t = bx.type_void();

        // Type of the vector of pointers:
        let llvm_pointer_vec_ty = llvm_vector_ty(bx, underlying_ty, in_len, pointer_count);
        let llvm_pointer_vec_str = llvm_vector_str(underlying_ty, in_len, pointer_count, bx);

        // Type of the vector of elements:
        let llvm_elem_vec_ty = llvm_vector_ty(bx, underlying_ty, in_len, pointer_count - 1);
        let llvm_elem_vec_str = llvm_vector_str(underlying_ty, in_len, pointer_count - 1, bx);

        let llvm_intrinsic =
            format!("llvm.masked.scatter.{}.{}", llvm_elem_vec_str, llvm_pointer_vec_str);
        let fn_ty =
            bx.type_func(&[llvm_elem_vec_ty, llvm_pointer_vec_ty, alignment_ty, mask_ty], ret_t);
        let f = bx.declare_cfn(&llvm_intrinsic, llvm::UnnamedAddr::No, fn_ty);
        let v =
            bx.call(fn_ty, f, &[args[0].immediate(), args[1].immediate(), alignment, mask], None);
        return Ok(v);
    }

    macro_rules! arith_red {
        ($name:ident : $integer_reduce:ident, $float_reduce:ident, $ordered:expr, $op:ident,
         $identity:expr) => {
            if name == sym::$name {
                require!(
                    ret_ty == in_elem,
                    "expected return type `{}` (element of input `{}`), found `{}`",
                    in_elem,
                    in_ty,
                    ret_ty
                );
                return match in_elem.kind() {
                    ty::Int(_) | ty::Uint(_) => {
                        let r = bx.$integer_reduce(args[0].immediate());
                        if $ordered {
                            // if overflow occurs, the result is the
                            // mathematical result modulo 2^n:
                            Ok(bx.$op(args[1].immediate(), r))
                        } else {
                            Ok(bx.$integer_reduce(args[0].immediate()))
                        }
                    }
                    ty::Float(f) => {
                        let acc = if $ordered {
                            // ordered arithmetic reductions take an accumulator
                            args[1].immediate()
                        } else {
                            // unordered arithmetic reductions use the identity accumulator
                            match f.bit_width() {
                                32 => bx.const_real(bx.type_f32(), $identity),
                                64 => bx.const_real(bx.type_f64(), $identity),
                                v => return_error!(
                                    r#"
unsupported {} from `{}` with element `{}` of size `{}` to `{}`"#,
                                    sym::$name,
                                    in_ty,
                                    in_elem,
                                    v,
                                    ret_ty
                                ),
                            }
                        };
                        Ok(bx.$float_reduce(acc, args[0].immediate()))
                    }
                    _ => return_error!(
                        "unsupported {} from `{}` with element `{}` to `{}`",
                        sym::$name,
                        in_ty,
                        in_elem,
                        ret_ty
                    ),
                };
            }
        };
    }

    arith_red!(simd_reduce_add_ordered: vector_reduce_add, vector_reduce_fadd, true, add, 0.0);
    arith_red!(simd_reduce_mul_ordered: vector_reduce_mul, vector_reduce_fmul, true, mul, 1.0);
    arith_red!(
        simd_reduce_add_unordered: vector_reduce_add,
        vector_reduce_fadd_fast,
        false,
        add,
        0.0
    );
    arith_red!(
        simd_reduce_mul_unordered: vector_reduce_mul,
        vector_reduce_fmul_fast,
        false,
        mul,
        1.0
    );

    macro_rules! minmax_red {
        ($name:ident: $int_red:ident, $float_red:ident) => {
            if name == sym::$name {
                require!(
                    ret_ty == in_elem,
                    "expected return type `{}` (element of input `{}`), found `{}`",
                    in_elem,
                    in_ty,
                    ret_ty
                );
                return match in_elem.kind() {
                    ty::Int(_i) => Ok(bx.$int_red(args[0].immediate(), true)),
                    ty::Uint(_u) => Ok(bx.$int_red(args[0].immediate(), false)),
                    ty::Float(_f) => Ok(bx.$float_red(args[0].immediate())),
                    _ => return_error!(
                        "unsupported {} from `{}` with element `{}` to `{}`",
                        sym::$name,
                        in_ty,
                        in_elem,
                        ret_ty
                    ),
                };
            }
        };
    }

    minmax_red!(simd_reduce_min: vector_reduce_min, vector_reduce_fmin);
    minmax_red!(simd_reduce_max: vector_reduce_max, vector_reduce_fmax);

    minmax_red!(simd_reduce_min_nanless: vector_reduce_min, vector_reduce_fmin_fast);
    minmax_red!(simd_reduce_max_nanless: vector_reduce_max, vector_reduce_fmax_fast);

    macro_rules! bitwise_red {
        ($name:ident : $red:ident, $boolean:expr) => {
            if name == sym::$name {
                let input = if !$boolean {
                    require!(
                        ret_ty == in_elem,
                        "expected return type `{}` (element of input `{}`), found `{}`",
                        in_elem,
                        in_ty,
                        ret_ty
                    );
                    args[0].immediate()
                } else {
                    match in_elem.kind() {
                        ty::Int(_) | ty::Uint(_) => {}
                        _ => return_error!(
                            "unsupported {} from `{}` with element `{}` to `{}`",
                            sym::$name,
                            in_ty,
                            in_elem,
                            ret_ty
                        ),
                    }

                    // boolean reductions operate on vectors of i1s:
                    let i1 = bx.type_i1();
                    let i1xn = bx.type_vector(i1, in_len as u64);
                    bx.trunc(args[0].immediate(), i1xn)
                };
                return match in_elem.kind() {
                    ty::Int(_) | ty::Uint(_) => {
                        let r = bx.$red(input);
                        Ok(if !$boolean { r } else { bx.zext(r, bx.type_bool()) })
                    }
                    _ => return_error!(
                        "unsupported {} from `{}` with element `{}` to `{}`",
                        sym::$name,
                        in_ty,
                        in_elem,
                        ret_ty
                    ),
                };
            }
        };
    }

    bitwise_red!(simd_reduce_and: vector_reduce_and, false);
    bitwise_red!(simd_reduce_or: vector_reduce_or, false);
    bitwise_red!(simd_reduce_xor: vector_reduce_xor, false);
    bitwise_red!(simd_reduce_all: vector_reduce_and, true);
    bitwise_red!(simd_reduce_any: vector_reduce_or, true);

    if name == sym::simd_cast || name == sym::simd_as {
        require_simd!(ret_ty, "return");
        let (out_len, out_elem) = ret_ty.simd_size_and_type(bx.tcx());
        require!(
            in_len == out_len,
            "expected return type with length {} (same as input type `{}`), \
                  found `{}` with length {}",
            in_len,
            in_ty,
            ret_ty,
            out_len
        );
        // casting cares about nominal type, not just structural type
        if in_elem == out_elem {
            return Ok(args[0].immediate());
        }

        enum Style {
            Float,
            Int(/* is signed? */ bool),
            Unsupported,
        }

        let (in_style, in_width) = match in_elem.kind() {
            // vectors of pointer-sized integers should've been
            // disallowed before here, so this unwrap is safe.
            ty::Int(i) => (
                Style::Int(true),
                i.normalize(bx.tcx().sess.target.pointer_width).bit_width().unwrap(),
            ),
            ty::Uint(u) => (
                Style::Int(false),
                u.normalize(bx.tcx().sess.target.pointer_width).bit_width().unwrap(),
            ),
            ty::Float(f) => (Style::Float, f.bit_width()),
            _ => (Style::Unsupported, 0),
        };
        let (out_style, out_width) = match out_elem.kind() {
            ty::Int(i) => (
                Style::Int(true),
                i.normalize(bx.tcx().sess.target.pointer_width).bit_width().unwrap(),
            ),
            ty::Uint(u) => (
                Style::Int(false),
                u.normalize(bx.tcx().sess.target.pointer_width).bit_width().unwrap(),
            ),
            ty::Float(f) => (Style::Float, f.bit_width()),
            _ => (Style::Unsupported, 0),
        };

        match (in_style, out_style) {
            (Style::Int(in_is_signed), Style::Int(_)) => {
                return Ok(match in_width.cmp(&out_width) {
                    Ordering::Greater => bx.trunc(args[0].immediate(), llret_ty),
                    Ordering::Equal => args[0].immediate(),
                    Ordering::Less => {
                        if in_is_signed {
                            bx.sext(args[0].immediate(), llret_ty)
                        } else {
                            bx.zext(args[0].immediate(), llret_ty)
                        }
                    }
                });
            }
            (Style::Int(in_is_signed), Style::Float) => {
                return Ok(if in_is_signed {
                    bx.sitofp(args[0].immediate(), llret_ty)
                } else {
                    bx.uitofp(args[0].immediate(), llret_ty)
                });
            }
            (Style::Float, Style::Int(out_is_signed)) => {
                return Ok(match (out_is_signed, name == sym::simd_as) {
                    (false, false) => bx.fptoui(args[0].immediate(), llret_ty),
                    (true, false) => bx.fptosi(args[0].immediate(), llret_ty),
                    (_, true) => bx.cast_float_to_int(out_is_signed, args[0].immediate(), llret_ty),
                });
            }
            (Style::Float, Style::Float) => {
                return Ok(match in_width.cmp(&out_width) {
                    Ordering::Greater => bx.fptrunc(args[0].immediate(), llret_ty),
                    Ordering::Equal => args[0].immediate(),
                    Ordering::Less => bx.fpext(args[0].immediate(), llret_ty),
                });
            }
            _ => { /* Unsupported. Fallthrough. */ }
        }
        require!(
            false,
            "unsupported cast from `{}` with element `{}` to `{}` with element `{}`",
            in_ty,
            in_elem,
            ret_ty,
            out_elem
        );
    }
    macro_rules! arith_binary {
        ($($name: ident: $($($p: ident),* => $call: ident),*;)*) => {
            $(if name == sym::$name {
                match in_elem.kind() {
                    $($(ty::$p(_))|* => {
                        return Ok(bx.$call(args[0].immediate(), args[1].immediate()))
                    })*
                    _ => {},
                }
                require!(false,
                         "unsupported operation on `{}` with element `{}`",
                         in_ty,
                         in_elem)
            })*
        }
    }
    arith_binary! {
        simd_add: Uint, Int => add, Float => fadd;
        simd_sub: Uint, Int => sub, Float => fsub;
        simd_mul: Uint, Int => mul, Float => fmul;
        simd_div: Uint => udiv, Int => sdiv, Float => fdiv;
        simd_rem: Uint => urem, Int => srem, Float => frem;
        simd_shl: Uint, Int => shl;
        simd_shr: Uint => lshr, Int => ashr;
        simd_and: Uint, Int => and;
        simd_or: Uint, Int => or;
        simd_xor: Uint, Int => xor;
        simd_fmax: Float => maxnum;
        simd_fmin: Float => minnum;

    }
    macro_rules! arith_unary {
        ($($name: ident: $($($p: ident),* => $call: ident),*;)*) => {
            $(if name == sym::$name {
                match in_elem.kind() {
                    $($(ty::$p(_))|* => {
                        return Ok(bx.$call(args[0].immediate()))
                    })*
                    _ => {},
                }
                require!(false,
                         "unsupported operation on `{}` with element `{}`",
                         in_ty,
                         in_elem)
            })*
        }
    }
    arith_unary! {
        simd_neg: Int => neg, Float => fneg;
    }

    if name == sym::simd_arith_offset {
        // This also checks that the first operand is a ptr type.
        let pointee = in_elem.builtin_deref(true).unwrap_or_else(|| {
            span_bug!(span, "must be called with a vector of pointer types as first argument")
        });
        let layout = bx.layout_of(pointee.ty);
        let ptrs = args[0].immediate();
        // The second argument must be a ptr-sized integer.
        // (We don't care about the signedness, this is wrapping anyway.)
        let (_offsets_len, offsets_elem) = arg_tys[1].simd_size_and_type(bx.tcx());
        if !matches!(offsets_elem.kind(), ty::Int(ty::IntTy::Isize) | ty::Uint(ty::UintTy::Usize)) {
            span_bug!(
                span,
                "must be called with a vector of pointer-sized integers as second argument"
            );
        }
        let offsets = args[1].immediate();

        return Ok(bx.gep(bx.backend_type(layout), ptrs, &[offsets]));
    }

    if name == sym::simd_saturating_add || name == sym::simd_saturating_sub {
        let lhs = args[0].immediate();
        let rhs = args[1].immediate();
        let is_add = name == sym::simd_saturating_add;
        let ptr_bits = bx.tcx().data_layout.pointer_size.bits() as _;
        let (signed, elem_width, elem_ty) = match *in_elem.kind() {
            ty::Int(i) => (true, i.bit_width().unwrap_or(ptr_bits), bx.cx.type_int_from_ty(i)),
            ty::Uint(i) => (false, i.bit_width().unwrap_or(ptr_bits), bx.cx.type_uint_from_ty(i)),
            _ => {
                return_error!(
                    "expected element type `{}` of vector type `{}` \
                     to be a signed or unsigned integer type",
                    arg_tys[0].simd_size_and_type(bx.tcx()).1,
                    arg_tys[0]
                );
            }
        };
        let llvm_intrinsic = &format!(
            "llvm.{}{}.sat.v{}i{}",
            if signed { 's' } else { 'u' },
            if is_add { "add" } else { "sub" },
            in_len,
            elem_width
        );
        let vec_ty = bx.cx.type_vector(elem_ty, in_len as u64);

        let fn_ty = bx.type_func(&[vec_ty, vec_ty], vec_ty);
        let f = bx.declare_cfn(llvm_intrinsic, llvm::UnnamedAddr::No, fn_ty);
        let v = bx.call(fn_ty, f, &[lhs, rhs], None);
        return Ok(v);
    }

    span_bug!(span, "unknown SIMD intrinsic");
}

// Returns the width of an int Ty, and if it's signed or not
// Returns None if the type is not an integer
// FIXME: there’s multiple of this functions, investigate using some of the already existing
// stuffs.
fn int_type_width_signed(ty: Ty<'_>, cx: &CodegenCx<'_, '_>) -> Option<(u64, bool)> {
    match ty.kind() {
        ty::Int(t) => {
            Some((t.bit_width().unwrap_or(u64::from(cx.tcx.sess.target.pointer_width)), true))
        }
        ty::Uint(t) => {
            Some((t.bit_width().unwrap_or(u64::from(cx.tcx.sess.target.pointer_width)), false))
        }
        _ => None,
    }
}