aboutsummaryrefslogtreecommitdiffstats
path: root/lib/CodeGen/CGStmtOpenMP.cpp
blob: 68555128ea015022975ceca573d451b90d8c10c1 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
//===--- CGStmtOpenMP.cpp - Emit LLVM Code from Statements ----------------===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This contains code to emit OpenMP nodes as LLVM code.
//
//===----------------------------------------------------------------------===//

#include "CGOpenMPRuntime.h"
#include "CodeGenFunction.h"
#include "CodeGenModule.h"
#include "TargetInfo.h"
#include "clang/AST/Stmt.h"
#include "clang/AST/StmtOpenMP.h"
using namespace clang;
using namespace CodeGen;

void CodeGenFunction::GenerateOpenMPCapturedVars(
    const CapturedStmt &S, SmallVectorImpl<llvm::Value *> &CapturedVars) {
  const RecordDecl *RD = S.getCapturedRecordDecl();
  auto CurField = RD->field_begin();
  auto CurCap = S.captures().begin();
  for (CapturedStmt::const_capture_init_iterator I = S.capture_init_begin(),
                                                 E = S.capture_init_end();
       I != E; ++I, ++CurField, ++CurCap) {
    if (CurField->hasCapturedVLAType()) {
      auto VAT = CurField->getCapturedVLAType();
      auto *Val = VLASizeMap[VAT->getSizeExpr()];
      CapturedVars.push_back(Val);
    } else if (CurCap->capturesThis())
      CapturedVars.push_back(CXXThisValue);
    else if (CurCap->capturesVariableByCopy())
      CapturedVars.push_back(
          EmitLoadOfLValue(EmitLValue(*I), SourceLocation()).getScalarVal());
    else {
      assert(CurCap->capturesVariable() && "Expected capture by reference.");
      CapturedVars.push_back(EmitLValue(*I).getAddress().getPointer());
    }
  }
}

static Address castValueFromUintptr(CodeGenFunction &CGF, QualType DstType,
                                    StringRef Name, LValue AddrLV,
                                    bool isReferenceType = false) {
  ASTContext &Ctx = CGF.getContext();

  auto *CastedPtr = CGF.EmitScalarConversion(
      AddrLV.getAddress().getPointer(), Ctx.getUIntPtrType(),
      Ctx.getPointerType(DstType), SourceLocation());
  auto TmpAddr =
      CGF.MakeNaturalAlignAddrLValue(CastedPtr, Ctx.getPointerType(DstType))
          .getAddress();

  // If we are dealing with references we need to return the address of the
  // reference instead of the reference of the value.
  if (isReferenceType) {
    QualType RefType = Ctx.getLValueReferenceType(DstType);
    auto *RefVal = TmpAddr.getPointer();
    TmpAddr = CGF.CreateMemTemp(RefType, Twine(Name) + ".ref");
    auto TmpLVal = CGF.MakeAddrLValue(TmpAddr, RefType);
    CGF.EmitScalarInit(RefVal, TmpLVal);
  }

  return TmpAddr;
}

llvm::Function *
CodeGenFunction::GenerateOpenMPCapturedStmtFunction(const CapturedStmt &S) {
  assert(
      CapturedStmtInfo &&
      "CapturedStmtInfo should be set when generating the captured function");
  const CapturedDecl *CD = S.getCapturedDecl();
  const RecordDecl *RD = S.getCapturedRecordDecl();
  assert(CD->hasBody() && "missing CapturedDecl body");

  // Build the argument list.
  ASTContext &Ctx = CGM.getContext();
  FunctionArgList Args;
  Args.append(CD->param_begin(),
              std::next(CD->param_begin(), CD->getContextParamPosition()));
  auto I = S.captures().begin();
  for (auto *FD : RD->fields()) {
    QualType ArgType = FD->getType();
    IdentifierInfo *II = nullptr;
    VarDecl *CapVar = nullptr;

    // If this is a capture by copy and the type is not a pointer, the outlined
    // function argument type should be uintptr and the value properly casted to
    // uintptr. This is necessary given that the runtime library is only able to
    // deal with pointers. We can pass in the same way the VLA type sizes to the
    // outlined function.
    if ((I->capturesVariableByCopy() && !ArgType->isAnyPointerType()) ||
        I->capturesVariableArrayType())
      ArgType = Ctx.getUIntPtrType();

    if (I->capturesVariable() || I->capturesVariableByCopy()) {
      CapVar = I->getCapturedVar();
      II = CapVar->getIdentifier();
    } else if (I->capturesThis())
      II = &getContext().Idents.get("this");
    else {
      assert(I->capturesVariableArrayType());
      II = &getContext().Idents.get("vla");
    }
    if (ArgType->isVariablyModifiedType())
      ArgType = getContext().getVariableArrayDecayedType(ArgType);
    Args.push_back(ImplicitParamDecl::Create(getContext(), nullptr,
                                             FD->getLocation(), II, ArgType));
    ++I;
  }
  Args.append(
      std::next(CD->param_begin(), CD->getContextParamPosition() + 1),
      CD->param_end());

  // Create the function declaration.
  FunctionType::ExtInfo ExtInfo;
  const CGFunctionInfo &FuncInfo =
      CGM.getTypes().arrangeFreeFunctionDeclaration(Ctx.VoidTy, Args, ExtInfo,
                                                    /*IsVariadic=*/false);
  llvm::FunctionType *FuncLLVMTy = CGM.getTypes().GetFunctionType(FuncInfo);

  llvm::Function *F = llvm::Function::Create(
      FuncLLVMTy, llvm::GlobalValue::InternalLinkage,
      CapturedStmtInfo->getHelperName(), &CGM.getModule());
  CGM.SetInternalFunctionAttributes(CD, F, FuncInfo);
  if (CD->isNothrow())
    F->addFnAttr(llvm::Attribute::NoUnwind);

  // Generate the function.
  StartFunction(CD, Ctx.VoidTy, F, FuncInfo, Args, CD->getLocation(),
                CD->getBody()->getLocStart());
  unsigned Cnt = CD->getContextParamPosition();
  I = S.captures().begin();
  for (auto *FD : RD->fields()) {
    // If we are capturing a pointer by copy we don't need to do anything, just
    // use the value that we get from the arguments.
    if (I->capturesVariableByCopy() && FD->getType()->isAnyPointerType()) {
      setAddrOfLocalVar(I->getCapturedVar(), GetAddrOfLocalVar(Args[Cnt]));
      ++Cnt, ++I;
      continue;
    }

    LValue ArgLVal =
        MakeAddrLValue(GetAddrOfLocalVar(Args[Cnt]), Args[Cnt]->getType(),
                       AlignmentSource::Decl);
    if (FD->hasCapturedVLAType()) {
      LValue CastedArgLVal =
          MakeAddrLValue(castValueFromUintptr(*this, FD->getType(),
                                              Args[Cnt]->getName(), ArgLVal),
                         FD->getType(), AlignmentSource::Decl);
      auto *ExprArg =
          EmitLoadOfLValue(CastedArgLVal, SourceLocation()).getScalarVal();
      auto VAT = FD->getCapturedVLAType();
      VLASizeMap[VAT->getSizeExpr()] = ExprArg;
    } else if (I->capturesVariable()) {
      auto *Var = I->getCapturedVar();
      QualType VarTy = Var->getType();
      Address ArgAddr = ArgLVal.getAddress();
      if (!VarTy->isReferenceType()) {
        ArgAddr = EmitLoadOfReference(
            ArgAddr, ArgLVal.getType()->castAs<ReferenceType>());
      }
      setAddrOfLocalVar(
          Var, Address(ArgAddr.getPointer(), getContext().getDeclAlign(Var)));
    } else if (I->capturesVariableByCopy()) {
      assert(!FD->getType()->isAnyPointerType() &&
             "Not expecting a captured pointer.");
      auto *Var = I->getCapturedVar();
      QualType VarTy = Var->getType();
      setAddrOfLocalVar(I->getCapturedVar(),
                        castValueFromUintptr(*this, FD->getType(),
                                             Args[Cnt]->getName(), ArgLVal,
                                             VarTy->isReferenceType()));
    } else {
      // If 'this' is captured, load it into CXXThisValue.
      assert(I->capturesThis());
      CXXThisValue =
          EmitLoadOfLValue(ArgLVal, Args[Cnt]->getLocation()).getScalarVal();
    }
    ++Cnt, ++I;
  }

  PGO.assignRegionCounters(GlobalDecl(CD), F);
  CapturedStmtInfo->EmitBody(*this, CD->getBody());
  FinishFunction(CD->getBodyRBrace());

  return F;
}

//===----------------------------------------------------------------------===//
//                              OpenMP Directive Emission
//===----------------------------------------------------------------------===//
void CodeGenFunction::EmitOMPAggregateAssign(
    Address DestAddr, Address SrcAddr, QualType OriginalType,
    const llvm::function_ref<void(Address, Address)> &CopyGen) {
  // Perform element-by-element initialization.
  QualType ElementTy;

  // Drill down to the base element type on both arrays.
  auto ArrayTy = OriginalType->getAsArrayTypeUnsafe();
  auto NumElements = emitArrayLength(ArrayTy, ElementTy, DestAddr);
  SrcAddr = Builder.CreateElementBitCast(SrcAddr, DestAddr.getElementType());

  auto SrcBegin = SrcAddr.getPointer();
  auto DestBegin = DestAddr.getPointer();
  // Cast from pointer to array type to pointer to single element.
  auto DestEnd = Builder.CreateGEP(DestBegin, NumElements);
  // The basic structure here is a while-do loop.
  auto BodyBB = createBasicBlock("omp.arraycpy.body");
  auto DoneBB = createBasicBlock("omp.arraycpy.done");
  auto IsEmpty =
      Builder.CreateICmpEQ(DestBegin, DestEnd, "omp.arraycpy.isempty");
  Builder.CreateCondBr(IsEmpty, DoneBB, BodyBB);

  // Enter the loop body, making that address the current address.
  auto EntryBB = Builder.GetInsertBlock();
  EmitBlock(BodyBB);

  CharUnits ElementSize = getContext().getTypeSizeInChars(ElementTy);

  llvm::PHINode *SrcElementPHI =
    Builder.CreatePHI(SrcBegin->getType(), 2, "omp.arraycpy.srcElementPast");
  SrcElementPHI->addIncoming(SrcBegin, EntryBB);
  Address SrcElementCurrent =
      Address(SrcElementPHI,
              SrcAddr.getAlignment().alignmentOfArrayElement(ElementSize));

  llvm::PHINode *DestElementPHI =
    Builder.CreatePHI(DestBegin->getType(), 2, "omp.arraycpy.destElementPast");
  DestElementPHI->addIncoming(DestBegin, EntryBB);
  Address DestElementCurrent =
    Address(DestElementPHI,
            DestAddr.getAlignment().alignmentOfArrayElement(ElementSize));

  // Emit copy.
  CopyGen(DestElementCurrent, SrcElementCurrent);

  // Shift the address forward by one element.
  auto DestElementNext = Builder.CreateConstGEP1_32(
      DestElementPHI, /*Idx0=*/1, "omp.arraycpy.dest.element");
  auto SrcElementNext = Builder.CreateConstGEP1_32(
      SrcElementPHI, /*Idx0=*/1, "omp.arraycpy.src.element");
  // Check whether we've reached the end.
  auto Done =
      Builder.CreateICmpEQ(DestElementNext, DestEnd, "omp.arraycpy.done");
  Builder.CreateCondBr(Done, DoneBB, BodyBB);
  DestElementPHI->addIncoming(DestElementNext, Builder.GetInsertBlock());
  SrcElementPHI->addIncoming(SrcElementNext, Builder.GetInsertBlock());

  // Done.
  EmitBlock(DoneBB, /*IsFinished=*/true);
}

/// \brief Emit initialization of arrays of complex types.
/// \param DestAddr Address of the array.
/// \param Type Type of array.
/// \param Init Initial expression of array.
static void EmitOMPAggregateInit(CodeGenFunction &CGF, Address DestAddr,
                                 QualType Type, const Expr *Init) {
  // Perform element-by-element initialization.
  QualType ElementTy;

  // Drill down to the base element type on both arrays.
  auto ArrayTy = Type->getAsArrayTypeUnsafe();
  auto NumElements = CGF.emitArrayLength(ArrayTy, ElementTy, DestAddr);
  DestAddr =
      CGF.Builder.CreateElementBitCast(DestAddr, DestAddr.getElementType());

  auto DestBegin = DestAddr.getPointer();
  // Cast from pointer to array type to pointer to single element.
  auto DestEnd = CGF.Builder.CreateGEP(DestBegin, NumElements);
  // The basic structure here is a while-do loop.
  auto BodyBB = CGF.createBasicBlock("omp.arrayinit.body");
  auto DoneBB = CGF.createBasicBlock("omp.arrayinit.done");
  auto IsEmpty =
      CGF.Builder.CreateICmpEQ(DestBegin, DestEnd, "omp.arrayinit.isempty");
  CGF.Builder.CreateCondBr(IsEmpty, DoneBB, BodyBB);

  // Enter the loop body, making that address the current address.
  auto EntryBB = CGF.Builder.GetInsertBlock();
  CGF.EmitBlock(BodyBB);

  CharUnits ElementSize = CGF.getContext().getTypeSizeInChars(ElementTy);

  llvm::PHINode *DestElementPHI = CGF.Builder.CreatePHI(
      DestBegin->getType(), 2, "omp.arraycpy.destElementPast");
  DestElementPHI->addIncoming(DestBegin, EntryBB);
  Address DestElementCurrent =
      Address(DestElementPHI,
              DestAddr.getAlignment().alignmentOfArrayElement(ElementSize));

  // Emit copy.
  {
    CodeGenFunction::RunCleanupsScope InitScope(CGF);
    CGF.EmitAnyExprToMem(Init, DestElementCurrent, ElementTy.getQualifiers(),
                         /*IsInitializer=*/false);
  }

  // Shift the address forward by one element.
  auto DestElementNext = CGF.Builder.CreateConstGEP1_32(
      DestElementPHI, /*Idx0=*/1, "omp.arraycpy.dest.element");
  // Check whether we've reached the end.
  auto Done =
      CGF.Builder.CreateICmpEQ(DestElementNext, DestEnd, "omp.arraycpy.done");
  CGF.Builder.CreateCondBr(Done, DoneBB, BodyBB);
  DestElementPHI->addIncoming(DestElementNext, CGF.Builder.GetInsertBlock());

  // Done.
  CGF.EmitBlock(DoneBB, /*IsFinished=*/true);
}

void CodeGenFunction::EmitOMPCopy(QualType OriginalType, Address DestAddr,
                                  Address SrcAddr, const VarDecl *DestVD,
                                  const VarDecl *SrcVD, const Expr *Copy) {
  if (OriginalType->isArrayType()) {
    auto *BO = dyn_cast<BinaryOperator>(Copy);
    if (BO && BO->getOpcode() == BO_Assign) {
      // Perform simple memcpy for simple copying.
      EmitAggregateAssign(DestAddr, SrcAddr, OriginalType);
    } else {
      // For arrays with complex element types perform element by element
      // copying.
      EmitOMPAggregateAssign(
          DestAddr, SrcAddr, OriginalType,
          [this, Copy, SrcVD, DestVD](Address DestElement, Address SrcElement) {
            // Working with the single array element, so have to remap
            // destination and source variables to corresponding array
            // elements.
            CodeGenFunction::OMPPrivateScope Remap(*this);
            Remap.addPrivate(DestVD, [DestElement]() -> Address {
              return DestElement;
            });
            Remap.addPrivate(
                SrcVD, [SrcElement]() -> Address { return SrcElement; });
            (void)Remap.Privatize();
            EmitIgnoredExpr(Copy);
          });
    }
  } else {
    // Remap pseudo source variable to private copy.
    CodeGenFunction::OMPPrivateScope Remap(*this);
    Remap.addPrivate(SrcVD, [SrcAddr]() -> Address { return SrcAddr; });
    Remap.addPrivate(DestVD, [DestAddr]() -> Address { return DestAddr; });
    (void)Remap.Privatize();
    // Emit copying of the whole variable.
    EmitIgnoredExpr(Copy);
  }
}

bool CodeGenFunction::EmitOMPFirstprivateClause(const OMPExecutableDirective &D,
                                                OMPPrivateScope &PrivateScope) {
  if (!HaveInsertPoint())
    return false;
  llvm::DenseSet<const VarDecl *> EmittedAsFirstprivate;
  for (const auto *C : D.getClausesOfKind<OMPFirstprivateClause>()) {
    auto IRef = C->varlist_begin();
    auto InitsRef = C->inits().begin();
    for (auto IInit : C->private_copies()) {
      auto *OrigVD = cast<VarDecl>(cast<DeclRefExpr>(*IRef)->getDecl());
      if (EmittedAsFirstprivate.count(OrigVD) == 0) {
        EmittedAsFirstprivate.insert(OrigVD);
        auto *VD = cast<VarDecl>(cast<DeclRefExpr>(IInit)->getDecl());
        auto *VDInit = cast<VarDecl>(cast<DeclRefExpr>(*InitsRef)->getDecl());
        bool IsRegistered;
        DeclRefExpr DRE(
            const_cast<VarDecl *>(OrigVD),
            /*RefersToEnclosingVariableOrCapture=*/CapturedStmtInfo->lookup(
                OrigVD) != nullptr,
            (*IRef)->getType(), VK_LValue, (*IRef)->getExprLoc());
        Address OriginalAddr = EmitLValue(&DRE).getAddress();
        QualType Type = OrigVD->getType();
        if (Type->isArrayType()) {
          // Emit VarDecl with copy init for arrays.
          // Get the address of the original variable captured in current
          // captured region.
          IsRegistered = PrivateScope.addPrivate(OrigVD, [&]() -> Address {
            auto Emission = EmitAutoVarAlloca(*VD);
            auto *Init = VD->getInit();
            if (!isa<CXXConstructExpr>(Init) || isTrivialInitializer(Init)) {
              // Perform simple memcpy.
              EmitAggregateAssign(Emission.getAllocatedAddress(), OriginalAddr,
                                  Type);
            } else {
              EmitOMPAggregateAssign(
                  Emission.getAllocatedAddress(), OriginalAddr, Type,
                  [this, VDInit, Init](Address DestElement,
                                       Address SrcElement) {
                    // Clean up any temporaries needed by the initialization.
                    RunCleanupsScope InitScope(*this);
                    // Emit initialization for single element.
                    setAddrOfLocalVar(VDInit, SrcElement);
                    EmitAnyExprToMem(Init, DestElement,
                                     Init->getType().getQualifiers(),
                                     /*IsInitializer*/ false);
                    LocalDeclMap.erase(VDInit);
                  });
            }
            EmitAutoVarCleanups(Emission);
            return Emission.getAllocatedAddress();
          });
        } else {
          IsRegistered = PrivateScope.addPrivate(OrigVD, [&]() -> Address {
            // Emit private VarDecl with copy init.
            // Remap temp VDInit variable to the address of the original
            // variable
            // (for proper handling of captured global variables).
            setAddrOfLocalVar(VDInit, OriginalAddr);
            EmitDecl(*VD);
            LocalDeclMap.erase(VDInit);
            return GetAddrOfLocalVar(VD);
          });
        }
        assert(IsRegistered &&
               "firstprivate var already registered as private");
        // Silence the warning about unused variable.
        (void)IsRegistered;
      }
      ++IRef, ++InitsRef;
    }
  }
  return !EmittedAsFirstprivate.empty();
}

void CodeGenFunction::EmitOMPPrivateClause(
    const OMPExecutableDirective &D,
    CodeGenFunction::OMPPrivateScope &PrivateScope) {
  if (!HaveInsertPoint())
    return;
  llvm::DenseSet<const VarDecl *> EmittedAsPrivate;
  for (const auto *C : D.getClausesOfKind<OMPPrivateClause>()) {
    auto IRef = C->varlist_begin();
    for (auto IInit : C->private_copies()) {
      auto *OrigVD = cast<VarDecl>(cast<DeclRefExpr>(*IRef)->getDecl());
      if (EmittedAsPrivate.insert(OrigVD->getCanonicalDecl()).second) {
        auto VD = cast<VarDecl>(cast<DeclRefExpr>(IInit)->getDecl());
        bool IsRegistered =
            PrivateScope.addPrivate(OrigVD, [&]() -> Address {
              // Emit private VarDecl with copy init.
              EmitDecl(*VD);
              return GetAddrOfLocalVar(VD);
            });
        assert(IsRegistered && "private var already registered as private");
        // Silence the warning about unused variable.
        (void)IsRegistered;
      }
      ++IRef;
    }
  }
}

bool CodeGenFunction::EmitOMPCopyinClause(const OMPExecutableDirective &D) {
  if (!HaveInsertPoint())
    return false;
  // threadprivate_var1 = master_threadprivate_var1;
  // operator=(threadprivate_var2, master_threadprivate_var2);
  // ...
  // __kmpc_barrier(&loc, global_tid);
  llvm::DenseSet<const VarDecl *> CopiedVars;
  llvm::BasicBlock *CopyBegin = nullptr, *CopyEnd = nullptr;
  for (const auto *C : D.getClausesOfKind<OMPCopyinClause>()) {
    auto IRef = C->varlist_begin();
    auto ISrcRef = C->source_exprs().begin();
    auto IDestRef = C->destination_exprs().begin();
    for (auto *AssignOp : C->assignment_ops()) {
      auto *VD = cast<VarDecl>(cast<DeclRefExpr>(*IRef)->getDecl());
      QualType Type = VD->getType();
      if (CopiedVars.insert(VD->getCanonicalDecl()).second) {

        // Get the address of the master variable. If we are emitting code with
        // TLS support, the address is passed from the master as field in the
        // captured declaration.
        Address MasterAddr = Address::invalid();
        if (getLangOpts().OpenMPUseTLS &&
            getContext().getTargetInfo().isTLSSupported()) {
          assert(CapturedStmtInfo->lookup(VD) &&
                 "Copyin threadprivates should have been captured!");
          DeclRefExpr DRE(const_cast<VarDecl *>(VD), true, (*IRef)->getType(),
                          VK_LValue, (*IRef)->getExprLoc());
          MasterAddr = EmitLValue(&DRE).getAddress();
          LocalDeclMap.erase(VD);
        } else {
          MasterAddr =
            Address(VD->isStaticLocal() ? CGM.getStaticLocalDeclAddress(VD)
                                        : CGM.GetAddrOfGlobal(VD),
                    getContext().getDeclAlign(VD));
        }
        // Get the address of the threadprivate variable.
        Address PrivateAddr = EmitLValue(*IRef).getAddress();
        if (CopiedVars.size() == 1) {
          // At first check if current thread is a master thread. If it is, no
          // need to copy data.
          CopyBegin = createBasicBlock("copyin.not.master");
          CopyEnd = createBasicBlock("copyin.not.master.end");
          Builder.CreateCondBr(
              Builder.CreateICmpNE(
                  Builder.CreatePtrToInt(MasterAddr.getPointer(), CGM.IntPtrTy),
                  Builder.CreatePtrToInt(PrivateAddr.getPointer(), CGM.IntPtrTy)),
              CopyBegin, CopyEnd);
          EmitBlock(CopyBegin);
        }
        auto *SrcVD = cast<VarDecl>(cast<DeclRefExpr>(*ISrcRef)->getDecl());
        auto *DestVD = cast<VarDecl>(cast<DeclRefExpr>(*IDestRef)->getDecl());
        EmitOMPCopy(Type, PrivateAddr, MasterAddr, DestVD, SrcVD, AssignOp);
      }
      ++IRef;
      ++ISrcRef;
      ++IDestRef;
    }
  }
  if (CopyEnd) {
    // Exit out of copying procedure for non-master thread.
    EmitBlock(CopyEnd, /*IsFinished=*/true);
    return true;
  }
  return false;
}

bool CodeGenFunction::EmitOMPLastprivateClauseInit(
    const OMPExecutableDirective &D, OMPPrivateScope &PrivateScope) {
  if (!HaveInsertPoint())
    return false;
  bool HasAtLeastOneLastprivate = false;
  llvm::DenseSet<const VarDecl *> AlreadyEmittedVars;
  for (const auto *C : D.getClausesOfKind<OMPLastprivateClause>()) {
    HasAtLeastOneLastprivate = true;
    auto IRef = C->varlist_begin();
    auto IDestRef = C->destination_exprs().begin();
    for (auto *IInit : C->private_copies()) {
      // Keep the address of the original variable for future update at the end
      // of the loop.
      auto *OrigVD = cast<VarDecl>(cast<DeclRefExpr>(*IRef)->getDecl());
      if (AlreadyEmittedVars.insert(OrigVD->getCanonicalDecl()).second) {
        auto *DestVD = cast<VarDecl>(cast<DeclRefExpr>(*IDestRef)->getDecl());
        PrivateScope.addPrivate(DestVD, [this, OrigVD, IRef]() -> Address {
          DeclRefExpr DRE(
              const_cast<VarDecl *>(OrigVD),
              /*RefersToEnclosingVariableOrCapture=*/CapturedStmtInfo->lookup(
                  OrigVD) != nullptr,
              (*IRef)->getType(), VK_LValue, (*IRef)->getExprLoc());
          return EmitLValue(&DRE).getAddress();
        });
        // Check if the variable is also a firstprivate: in this case IInit is
        // not generated. Initialization of this variable will happen in codegen
        // for 'firstprivate' clause.
        if (IInit) {
          auto *VD = cast<VarDecl>(cast<DeclRefExpr>(IInit)->getDecl());
          bool IsRegistered =
              PrivateScope.addPrivate(OrigVD, [&]() -> Address {
                // Emit private VarDecl with copy init.
                EmitDecl(*VD);
                return GetAddrOfLocalVar(VD);
              });
          assert(IsRegistered &&
                 "lastprivate var already registered as private");
          (void)IsRegistered;
        }
      }
      ++IRef, ++IDestRef;
    }
  }
  return HasAtLeastOneLastprivate;
}

void CodeGenFunction::EmitOMPLastprivateClauseFinal(
    const OMPExecutableDirective &D, llvm::Value *IsLastIterCond) {
  if (!HaveInsertPoint())
    return;
  // Emit following code:
  // if (<IsLastIterCond>) {
  //   orig_var1 = private_orig_var1;
  //   ...
  //   orig_varn = private_orig_varn;
  // }
  llvm::BasicBlock *ThenBB = nullptr;
  llvm::BasicBlock *DoneBB = nullptr;
  if (IsLastIterCond) {
    ThenBB = createBasicBlock(".omp.lastprivate.then");
    DoneBB = createBasicBlock(".omp.lastprivate.done");
    Builder.CreateCondBr(IsLastIterCond, ThenBB, DoneBB);
    EmitBlock(ThenBB);
  }
  llvm::DenseMap<const Decl *, const Expr *> LoopCountersAndUpdates;
  const Expr *LastIterVal = nullptr;
  const Expr *IVExpr = nullptr;
  const Expr *IncExpr = nullptr;
  if (auto *LoopDirective = dyn_cast<OMPLoopDirective>(&D)) {
    if (isOpenMPWorksharingDirective(D.getDirectiveKind())) {
      LastIterVal = cast<VarDecl>(cast<DeclRefExpr>(
                                      LoopDirective->getUpperBoundVariable())
                                      ->getDecl())
                        ->getAnyInitializer();
      IVExpr = LoopDirective->getIterationVariable();
      IncExpr = LoopDirective->getInc();
      auto IUpdate = LoopDirective->updates().begin();
      for (auto *E : LoopDirective->counters()) {
        auto *D = cast<DeclRefExpr>(E)->getDecl()->getCanonicalDecl();
        LoopCountersAndUpdates[D] = *IUpdate;
        ++IUpdate;
      }
    }
  }
  {
    llvm::DenseSet<const VarDecl *> AlreadyEmittedVars;
    bool FirstLCV = true;
    for (const auto *C : D.getClausesOfKind<OMPLastprivateClause>()) {
      auto IRef = C->varlist_begin();
      auto ISrcRef = C->source_exprs().begin();
      auto IDestRef = C->destination_exprs().begin();
      for (auto *AssignOp : C->assignment_ops()) {
        auto *PrivateVD = cast<VarDecl>(cast<DeclRefExpr>(*IRef)->getDecl());
        QualType Type = PrivateVD->getType();
        auto *CanonicalVD = PrivateVD->getCanonicalDecl();
        if (AlreadyEmittedVars.insert(CanonicalVD).second) {
          // If lastprivate variable is a loop control variable for loop-based
          // directive, update its value before copyin back to original
          // variable.
          if (auto *UpExpr = LoopCountersAndUpdates.lookup(CanonicalVD)) {
            if (FirstLCV && LastIterVal) {
              EmitAnyExprToMem(LastIterVal, EmitLValue(IVExpr).getAddress(),
                               IVExpr->getType().getQualifiers(),
                               /*IsInitializer=*/false);
              EmitIgnoredExpr(IncExpr);
              FirstLCV = false;
            }
            EmitIgnoredExpr(UpExpr);
          }
          auto *SrcVD = cast<VarDecl>(cast<DeclRefExpr>(*ISrcRef)->getDecl());
          auto *DestVD = cast<VarDecl>(cast<DeclRefExpr>(*IDestRef)->getDecl());
          // Get the address of the original variable.
          Address OriginalAddr = GetAddrOfLocalVar(DestVD);
          // Get the address of the private variable.
          Address PrivateAddr = GetAddrOfLocalVar(PrivateVD);
          if (auto RefTy = PrivateVD->getType()->getAs<ReferenceType>())
            PrivateAddr =
              Address(Builder.CreateLoad(PrivateAddr),
                      getNaturalTypeAlignment(RefTy->getPointeeType()));
          EmitOMPCopy(Type, OriginalAddr, PrivateAddr, DestVD, SrcVD, AssignOp);
        }
        ++IRef;
        ++ISrcRef;
        ++IDestRef;
      }
    }
  }
  if (IsLastIterCond) {
    EmitBlock(DoneBB, /*IsFinished=*/true);
  }
}

void CodeGenFunction::EmitOMPReductionClauseInit(
    const OMPExecutableDirective &D,
    CodeGenFunction::OMPPrivateScope &PrivateScope) {
  if (!HaveInsertPoint())
    return;
  for (const auto *C : D.getClausesOfKind<OMPReductionClause>()) {
    auto ILHS = C->lhs_exprs().begin();
    auto IRHS = C->rhs_exprs().begin();
    auto IPriv = C->privates().begin();
    for (auto IRef : C->varlists()) {
      auto *LHSVD = cast<VarDecl>(cast<DeclRefExpr>(*ILHS)->getDecl());
      auto *RHSVD = cast<VarDecl>(cast<DeclRefExpr>(*IRHS)->getDecl());
      auto *PrivateVD = cast<VarDecl>(cast<DeclRefExpr>(*IPriv)->getDecl());
      if (auto *OASE = dyn_cast<OMPArraySectionExpr>(IRef)) {
        auto *Base = OASE->getBase()->IgnoreParenImpCasts();
        while (auto *TempOASE = dyn_cast<OMPArraySectionExpr>(Base))
          Base = TempOASE->getBase()->IgnoreParenImpCasts();
        while (auto *TempASE = dyn_cast<ArraySubscriptExpr>(Base))
          Base = TempASE->getBase()->IgnoreParenImpCasts();
        auto *DE = cast<DeclRefExpr>(Base);
        auto *OrigVD = cast<VarDecl>(DE->getDecl());
        auto OASELValueLB = EmitOMPArraySectionExpr(OASE);
        auto OASELValueUB =
            EmitOMPArraySectionExpr(OASE, /*IsLowerBound=*/false);
        auto OriginalBaseLValue = EmitLValue(DE);
        auto BaseLValue = OriginalBaseLValue;
        auto *Zero = Builder.getInt64(/*C=*/0);
        llvm::SmallVector<llvm::Value *, 4> Indexes;
        Indexes.push_back(Zero);
        auto *ItemTy =
            OASELValueLB.getPointer()->getType()->getPointerElementType();
        auto *Ty = BaseLValue.getPointer()->getType()->getPointerElementType();
        while (Ty != ItemTy) {
          Indexes.push_back(Zero);
          Ty = Ty->getPointerElementType();
        }
        BaseLValue = MakeAddrLValue(
            Address(Builder.CreateInBoundsGEP(BaseLValue.getPointer(), Indexes),
                    OASELValueLB.getAlignment()),
            OASELValueLB.getType(), OASELValueLB.getAlignmentSource());
        // Store the address of the original variable associated with the LHS
        // implicit variable.
        PrivateScope.addPrivate(LHSVD, [this, OASELValueLB]() -> Address {
          return OASELValueLB.getAddress();
        });
        // Emit reduction copy.
        bool IsRegistered = PrivateScope.addPrivate(
            OrigVD, [this, PrivateVD, BaseLValue, OASELValueLB, OASELValueUB,
                     OriginalBaseLValue]() -> Address {
              // Emit VarDecl with copy init for arrays.
              // Get the address of the original variable captured in current
              // captured region.
              auto *Size = Builder.CreatePtrDiff(OASELValueUB.getPointer(),
                                                 OASELValueLB.getPointer());
              Size = Builder.CreateNUWAdd(
                  Size, llvm::ConstantInt::get(Size->getType(), /*V=*/1));
              CodeGenFunction::OpaqueValueMapping OpaqueMap(
                  *this, cast<OpaqueValueExpr>(
                             getContext()
                                 .getAsVariableArrayType(PrivateVD->getType())
                                 ->getSizeExpr()),
                  RValue::get(Size));
              EmitVariablyModifiedType(PrivateVD->getType());
              auto Emission = EmitAutoVarAlloca(*PrivateVD);
              auto Addr = Emission.getAllocatedAddress();
              auto *Init = PrivateVD->getInit();
              EmitOMPAggregateInit(*this, Addr, PrivateVD->getType(), Init);
              EmitAutoVarCleanups(Emission);
              // Emit private VarDecl with reduction init.
              auto *Offset = Builder.CreatePtrDiff(BaseLValue.getPointer(),
                                                   OASELValueLB.getPointer());
              auto *Ptr = Builder.CreateGEP(Addr.getPointer(), Offset);
              Ptr = Builder.CreatePointerBitCastOrAddrSpaceCast(
                  Ptr, OriginalBaseLValue.getPointer()->getType());
              return Address(Ptr, OriginalBaseLValue.getAlignment());
            });
        assert(IsRegistered && "private var already registered as private");
        // Silence the warning about unused variable.
        (void)IsRegistered;
        PrivateScope.addPrivate(RHSVD, [this, PrivateVD]() -> Address {
          return GetAddrOfLocalVar(PrivateVD);
        });
      } else if (auto *ASE = dyn_cast<ArraySubscriptExpr>(IRef)) {
        auto *Base = ASE->getBase()->IgnoreParenImpCasts();
        while (auto *TempASE = dyn_cast<ArraySubscriptExpr>(Base))
          Base = TempASE->getBase()->IgnoreParenImpCasts();
        auto *DE = cast<DeclRefExpr>(Base);
        auto *OrigVD = cast<VarDecl>(DE->getDecl());
        auto ASELValue = EmitLValue(ASE);
        auto OriginalBaseLValue = EmitLValue(DE);
        auto BaseLValue = OriginalBaseLValue;
        auto *Zero = Builder.getInt64(/*C=*/0);
        llvm::SmallVector<llvm::Value *, 4> Indexes;
        Indexes.push_back(Zero);
        auto *ItemTy =
            ASELValue.getPointer()->getType()->getPointerElementType();
        auto *Ty = BaseLValue.getPointer()->getType()->getPointerElementType();
        while (Ty != ItemTy) {
          Indexes.push_back(Zero);
          Ty = Ty->getPointerElementType();
        }
        BaseLValue = MakeAddrLValue(
            Address(Builder.CreateInBoundsGEP(BaseLValue.getPointer(), Indexes),
                    ASELValue.getAlignment()),
            ASELValue.getType(), ASELValue.getAlignmentSource());
        // Store the address of the original variable associated with the LHS
        // implicit variable.
        PrivateScope.addPrivate(LHSVD, [this, ASELValue]() -> Address {
          return ASELValue.getAddress();
        });
        // Emit reduction copy.
        bool IsRegistered = PrivateScope.addPrivate(
            OrigVD, [this, PrivateVD, BaseLValue, ASELValue,
                     OriginalBaseLValue]() -> Address {
              // Emit private VarDecl with reduction init.
              EmitDecl(*PrivateVD);
              auto Addr = GetAddrOfLocalVar(PrivateVD);
              auto *Offset = Builder.CreatePtrDiff(BaseLValue.getPointer(),
                                                   ASELValue.getPointer());
              auto *Ptr = Builder.CreateGEP(Addr.getPointer(), Offset);
              Ptr = Builder.CreatePointerBitCastOrAddrSpaceCast(
                  Ptr, OriginalBaseLValue.getPointer()->getType());
              return Address(Ptr, OriginalBaseLValue.getAlignment());
            });
        assert(IsRegistered && "private var already registered as private");
        // Silence the warning about unused variable.
        (void)IsRegistered;
        PrivateScope.addPrivate(RHSVD, [this, PrivateVD]() -> Address {
          return GetAddrOfLocalVar(PrivateVD);
        });
      } else {
        auto *OrigVD = cast<VarDecl>(cast<DeclRefExpr>(IRef)->getDecl());
        // Store the address of the original variable associated with the LHS
        // implicit variable.
        PrivateScope.addPrivate(LHSVD, [this, OrigVD, IRef]() -> Address {
          DeclRefExpr DRE(const_cast<VarDecl *>(OrigVD),
                          CapturedStmtInfo->lookup(OrigVD) != nullptr,
                          IRef->getType(), VK_LValue, IRef->getExprLoc());
          return EmitLValue(&DRE).getAddress();
        });
        // Emit reduction copy.
        bool IsRegistered =
            PrivateScope.addPrivate(OrigVD, [this, PrivateVD]() -> Address {
              // Emit private VarDecl with reduction init.
              EmitDecl(*PrivateVD);
              return GetAddrOfLocalVar(PrivateVD);
            });
        assert(IsRegistered && "private var already registered as private");
        // Silence the warning about unused variable.
        (void)IsRegistered;
        PrivateScope.addPrivate(RHSVD, [this, PrivateVD]() -> Address {
          return GetAddrOfLocalVar(PrivateVD);
        });
      }
      ++ILHS, ++IRHS, ++IPriv;
    }
  }
}

void CodeGenFunction::EmitOMPReductionClauseFinal(
    const OMPExecutableDirective &D) {
  if (!HaveInsertPoint())
    return;
  llvm::SmallVector<const Expr *, 8> Privates;
  llvm::SmallVector<const Expr *, 8> LHSExprs;
  llvm::SmallVector<const Expr *, 8> RHSExprs;
  llvm::SmallVector<const Expr *, 8> ReductionOps;
  bool HasAtLeastOneReduction = false;
  for (const auto *C : D.getClausesOfKind<OMPReductionClause>()) {
    HasAtLeastOneReduction = true;
    Privates.append(C->privates().begin(), C->privates().end());
    LHSExprs.append(C->lhs_exprs().begin(), C->lhs_exprs().end());
    RHSExprs.append(C->rhs_exprs().begin(), C->rhs_exprs().end());
    ReductionOps.append(C->reduction_ops().begin(), C->reduction_ops().end());
  }
  if (HasAtLeastOneReduction) {
    // Emit nowait reduction if nowait clause is present or directive is a
    // parallel directive (it always has implicit barrier).
    CGM.getOpenMPRuntime().emitReduction(
        *this, D.getLocEnd(), Privates, LHSExprs, RHSExprs, ReductionOps,
        D.getSingleClause<OMPNowaitClause>() ||
            isOpenMPParallelDirective(D.getDirectiveKind()) ||
            D.getDirectiveKind() == OMPD_simd,
        D.getDirectiveKind() == OMPD_simd);
  }
}

static void emitCommonOMPParallelDirective(CodeGenFunction &CGF,
                                           const OMPExecutableDirective &S,
                                           OpenMPDirectiveKind InnermostKind,
                                           const RegionCodeGenTy &CodeGen) {
  auto CS = cast<CapturedStmt>(S.getAssociatedStmt());
  llvm::SmallVector<llvm::Value *, 16> CapturedVars;
  CGF.GenerateOpenMPCapturedVars(*CS, CapturedVars);
  auto OutlinedFn = CGF.CGM.getOpenMPRuntime().emitParallelOutlinedFunction(
      S, *CS->getCapturedDecl()->param_begin(), InnermostKind, CodeGen);
  if (const auto *NumThreadsClause = S.getSingleClause<OMPNumThreadsClause>()) {
    CodeGenFunction::RunCleanupsScope NumThreadsScope(CGF);
    auto NumThreads = CGF.EmitScalarExpr(NumThreadsClause->getNumThreads(),
                                         /*IgnoreResultAssign*/ true);
    CGF.CGM.getOpenMPRuntime().emitNumThreadsClause(
        CGF, NumThreads, NumThreadsClause->getLocStart());
  }
  if (const auto *ProcBindClause = S.getSingleClause<OMPProcBindClause>()) {
    CodeGenFunction::RunCleanupsScope NumThreadsScope(CGF);
    CGF.CGM.getOpenMPRuntime().emitProcBindClause(
        CGF, ProcBindClause->getProcBindKind(), ProcBindClause->getLocStart());
  }
  const Expr *IfCond = nullptr;
  for (const auto *C : S.getClausesOfKind<OMPIfClause>()) {
    if (C->getNameModifier() == OMPD_unknown ||
        C->getNameModifier() == OMPD_parallel) {
      IfCond = C->getCondition();
      break;
    }
  }
  CGF.CGM.getOpenMPRuntime().emitParallelCall(CGF, S.getLocStart(), OutlinedFn,
                                              CapturedVars, IfCond);
}

void CodeGenFunction::EmitOMPParallelDirective(const OMPParallelDirective &S) {
  LexicalScope Scope(*this, S.getSourceRange());
  // Emit parallel region as a standalone region.
  auto &&CodeGen = [&S](CodeGenFunction &CGF) {
    OMPPrivateScope PrivateScope(CGF);
    bool Copyins = CGF.EmitOMPCopyinClause(S);
    bool Firstprivates = CGF.EmitOMPFirstprivateClause(S, PrivateScope);
    if (Copyins || Firstprivates) {
      // Emit implicit barrier to synchronize threads and avoid data races on
      // initialization of firstprivate variables or propagation master's thread
      // values of threadprivate variables to local instances of that variables
      // of all other implicit threads.
      CGF.CGM.getOpenMPRuntime().emitBarrierCall(
          CGF, S.getLocStart(), OMPD_unknown, /*EmitChecks=*/false,
          /*ForceSimpleCall=*/true);
    }
    CGF.EmitOMPPrivateClause(S, PrivateScope);
    CGF.EmitOMPReductionClauseInit(S, PrivateScope);
    (void)PrivateScope.Privatize();
    CGF.EmitStmt(cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
    CGF.EmitOMPReductionClauseFinal(S);
  };
  emitCommonOMPParallelDirective(*this, S, OMPD_parallel, CodeGen);
}

void CodeGenFunction::EmitOMPLoopBody(const OMPLoopDirective &D,
                                      JumpDest LoopExit) {
  RunCleanupsScope BodyScope(*this);
  // Update counters values on current iteration.
  for (auto I : D.updates()) {
    EmitIgnoredExpr(I);
  }
  // Update the linear variables.
  for (const auto *C : D.getClausesOfKind<OMPLinearClause>()) {
    for (auto U : C->updates()) {
      EmitIgnoredExpr(U);
    }
  }

  // On a continue in the body, jump to the end.
  auto Continue = getJumpDestInCurrentScope("omp.body.continue");
  BreakContinueStack.push_back(BreakContinue(LoopExit, Continue));
  // Emit loop body.
  EmitStmt(D.getBody());
  // The end (updates/cleanups).
  EmitBlock(Continue.getBlock());
  BreakContinueStack.pop_back();
    // TODO: Update lastprivates if the SeparateIter flag is true.
    // This will be implemented in a follow-up OMPLastprivateClause patch, but
    // result should be still correct without it, as we do not make these
    // variables private yet.
}

void CodeGenFunction::EmitOMPInnerLoop(
    const Stmt &S, bool RequiresCleanup, const Expr *LoopCond,
    const Expr *IncExpr,
    const llvm::function_ref<void(CodeGenFunction &)> &BodyGen,
    const llvm::function_ref<void(CodeGenFunction &)> &PostIncGen) {
  auto LoopExit = getJumpDestInCurrentScope("omp.inner.for.end");

  // Start the loop with a block that tests the condition.
  auto CondBlock = createBasicBlock("omp.inner.for.cond");
  EmitBlock(CondBlock);
  LoopStack.push(CondBlock);

  // If there are any cleanups between here and the loop-exit scope,
  // create a block to stage a loop exit along.
  auto ExitBlock = LoopExit.getBlock();
  if (RequiresCleanup)
    ExitBlock = createBasicBlock("omp.inner.for.cond.cleanup");

  auto LoopBody = createBasicBlock("omp.inner.for.body");

  // Emit condition.
  EmitBranchOnBoolExpr(LoopCond, LoopBody, ExitBlock, getProfileCount(&S));
  if (ExitBlock != LoopExit.getBlock()) {
    EmitBlock(ExitBlock);
    EmitBranchThroughCleanup(LoopExit);
  }

  EmitBlock(LoopBody);
  incrementProfileCounter(&S);

  // Create a block for the increment.
  auto Continue = getJumpDestInCurrentScope("omp.inner.for.inc");
  BreakContinueStack.push_back(BreakContinue(LoopExit, Continue));

  BodyGen(*this);

  // Emit "IV = IV + 1" and a back-edge to the condition block.
  EmitBlock(Continue.getBlock());
  EmitIgnoredExpr(IncExpr);
  PostIncGen(*this);
  BreakContinueStack.pop_back();
  EmitBranch(CondBlock);
  LoopStack.pop();
  // Emit the fall-through block.
  EmitBlock(LoopExit.getBlock());
}

void CodeGenFunction::EmitOMPLinearClauseInit(const OMPLoopDirective &D) {
  if (!HaveInsertPoint())
    return;
  // Emit inits for the linear variables.
  for (const auto *C : D.getClausesOfKind<OMPLinearClause>()) {
    for (auto Init : C->inits()) {
      auto *VD = cast<VarDecl>(cast<DeclRefExpr>(Init)->getDecl());
      auto *OrigVD = cast<VarDecl>(
          cast<DeclRefExpr>(VD->getInit()->IgnoreImpCasts())->getDecl());
      DeclRefExpr DRE(const_cast<VarDecl *>(OrigVD),
                      CapturedStmtInfo->lookup(OrigVD) != nullptr,
                      VD->getInit()->getType(), VK_LValue,
                      VD->getInit()->getExprLoc());
      AutoVarEmission Emission = EmitAutoVarAlloca(*VD);
      EmitExprAsInit(&DRE, VD,
               MakeAddrLValue(Emission.getAllocatedAddress(), VD->getType()),
                     /*capturedByInit=*/false);
      EmitAutoVarCleanups(Emission);
    }
    // Emit the linear steps for the linear clauses.
    // If a step is not constant, it is pre-calculated before the loop.
    if (auto CS = cast_or_null<BinaryOperator>(C->getCalcStep()))
      if (auto SaveRef = cast<DeclRefExpr>(CS->getLHS())) {
        EmitVarDecl(*cast<VarDecl>(SaveRef->getDecl()));
        // Emit calculation of the linear step.
        EmitIgnoredExpr(CS);
      }
  }
}

static void emitLinearClauseFinal(CodeGenFunction &CGF,
                                  const OMPLoopDirective &D) {
  if (!CGF.HaveInsertPoint())
    return;
  // Emit the final values of the linear variables.
  for (const auto *C : D.getClausesOfKind<OMPLinearClause>()) {
    auto IC = C->varlist_begin();
    for (auto F : C->finals()) {
      auto *OrigVD = cast<VarDecl>(cast<DeclRefExpr>(*IC)->getDecl());
      DeclRefExpr DRE(const_cast<VarDecl *>(OrigVD),
                      CGF.CapturedStmtInfo->lookup(OrigVD) != nullptr,
                      (*IC)->getType(), VK_LValue, (*IC)->getExprLoc());
      Address OrigAddr = CGF.EmitLValue(&DRE).getAddress();
      CodeGenFunction::OMPPrivateScope VarScope(CGF);
      VarScope.addPrivate(OrigVD,
                          [OrigAddr]() -> Address { return OrigAddr; });
      (void)VarScope.Privatize();
      CGF.EmitIgnoredExpr(F);
      ++IC;
    }
  }
}

static void emitAlignedClause(CodeGenFunction &CGF,
                              const OMPExecutableDirective &D) {
  if (!CGF.HaveInsertPoint())
    return;
  for (const auto *Clause : D.getClausesOfKind<OMPAlignedClause>()) {
    unsigned ClauseAlignment = 0;
    if (auto AlignmentExpr = Clause->getAlignment()) {
      auto AlignmentCI =
          cast<llvm::ConstantInt>(CGF.EmitScalarExpr(AlignmentExpr));
      ClauseAlignment = static_cast<unsigned>(AlignmentCI->getZExtValue());
    }
    for (auto E : Clause->varlists()) {
      unsigned Alignment = ClauseAlignment;
      if (Alignment == 0) {
        // OpenMP [2.8.1, Description]
        // If no optional parameter is specified, implementation-defined default
        // alignments for SIMD instructions on the target platforms are assumed.
        Alignment =
            CGF.getContext()
                .toCharUnitsFromBits(CGF.getContext().getOpenMPDefaultSimdAlign(
                    E->getType()->getPointeeType()))
                .getQuantity();
      }
      assert((Alignment == 0 || llvm::isPowerOf2_32(Alignment)) &&
             "alignment is not power of 2");
      if (Alignment != 0) {
        llvm::Value *PtrValue = CGF.EmitScalarExpr(E);
        CGF.EmitAlignmentAssumption(PtrValue, Alignment);
      }
    }
  }
}

static void emitPrivateLoopCounters(CodeGenFunction &CGF,
                                    CodeGenFunction::OMPPrivateScope &LoopScope,
                                    ArrayRef<Expr *> Counters,
                                    ArrayRef<Expr *> PrivateCounters) {
  if (!CGF.HaveInsertPoint())
    return;
  auto I = PrivateCounters.begin();
  for (auto *E : Counters) {
    auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
    auto *PrivateVD = cast<VarDecl>(cast<DeclRefExpr>(*I)->getDecl());
    Address Addr = Address::invalid();
    (void)LoopScope.addPrivate(PrivateVD, [&]() -> Address {
      // Emit var without initialization.
      auto VarEmission = CGF.EmitAutoVarAlloca(*PrivateVD);
      CGF.EmitAutoVarCleanups(VarEmission);
      Addr = VarEmission.getAllocatedAddress();
      return Addr;
    });
    (void)LoopScope.addPrivate(VD, [&]() -> Address { return Addr; });
    ++I;
  }
}

static void emitPreCond(CodeGenFunction &CGF, const OMPLoopDirective &S,
                        const Expr *Cond, llvm::BasicBlock *TrueBlock,
                        llvm::BasicBlock *FalseBlock, uint64_t TrueCount) {
  if (!CGF.HaveInsertPoint())
    return;
  {
    CodeGenFunction::OMPPrivateScope PreCondScope(CGF);
    emitPrivateLoopCounters(CGF, PreCondScope, S.counters(),
                            S.private_counters());
    (void)PreCondScope.Privatize();
    // Get initial values of real counters.
    for (auto I : S.inits()) {
      CGF.EmitIgnoredExpr(I);
    }
  }
  // Check that loop is executed at least one time.
  CGF.EmitBranchOnBoolExpr(Cond, TrueBlock, FalseBlock, TrueCount);
}

static void
emitPrivateLinearVars(CodeGenFunction &CGF, const OMPExecutableDirective &D,
                      CodeGenFunction::OMPPrivateScope &PrivateScope) {
  if (!CGF.HaveInsertPoint())
    return;
  for (const auto *C : D.getClausesOfKind<OMPLinearClause>()) {
    auto CurPrivate = C->privates().begin();
    for (auto *E : C->varlists()) {
      auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
      auto *PrivateVD =
          cast<VarDecl>(cast<DeclRefExpr>(*CurPrivate)->getDecl());
      bool IsRegistered = PrivateScope.addPrivate(VD, [&]() -> Address {
        // Emit private VarDecl with copy init.
        CGF.EmitVarDecl(*PrivateVD);
        return CGF.GetAddrOfLocalVar(PrivateVD);
      });
      assert(IsRegistered && "linear var already registered as private");
      // Silence the warning about unused variable.
      (void)IsRegistered;
      ++CurPrivate;
    }
  }
}

static void emitSimdlenSafelenClause(CodeGenFunction &CGF,
                                     const OMPExecutableDirective &D,
                                     bool IsMonotonic) {
  if (!CGF.HaveInsertPoint())
    return;
  if (const auto *C = D.getSingleClause<OMPSimdlenClause>()) {
    RValue Len = CGF.EmitAnyExpr(C->getSimdlen(), AggValueSlot::ignored(),
                                 /*ignoreResult=*/true);
    llvm::ConstantInt *Val = cast<llvm::ConstantInt>(Len.getScalarVal());
    CGF.LoopStack.setVectorizeWidth(Val->getZExtValue());
    // In presence of finite 'safelen', it may be unsafe to mark all
    // the memory instructions parallel, because loop-carried
    // dependences of 'safelen' iterations are possible.
    if (!IsMonotonic)
      CGF.LoopStack.setParallel(!D.getSingleClause<OMPSafelenClause>());
  } else if (const auto *C = D.getSingleClause<OMPSafelenClause>()) {
    RValue Len = CGF.EmitAnyExpr(C->getSafelen(), AggValueSlot::ignored(),
                                 /*ignoreResult=*/true);
    llvm::ConstantInt *Val = cast<llvm::ConstantInt>(Len.getScalarVal());
    CGF.LoopStack.setVectorizeWidth(Val->getZExtValue());
    // In presence of finite 'safelen', it may be unsafe to mark all
    // the memory instructions parallel, because loop-carried
    // dependences of 'safelen' iterations are possible.
    CGF.LoopStack.setParallel(false);
  }
}

void CodeGenFunction::EmitOMPSimdInit(const OMPLoopDirective &D,
                                      bool IsMonotonic) {
  // Walk clauses and process safelen/lastprivate.
  LoopStack.setParallel(!IsMonotonic);
  LoopStack.setVectorizeEnable(true);
  emitSimdlenSafelenClause(*this, D, IsMonotonic);
}

void CodeGenFunction::EmitOMPSimdFinal(const OMPLoopDirective &D) {
  if (!HaveInsertPoint())
    return;
  auto IC = D.counters().begin();
  for (auto F : D.finals()) {
    auto *OrigVD = cast<VarDecl>(cast<DeclRefExpr>((*IC))->getDecl());
    if (LocalDeclMap.count(OrigVD) || CapturedStmtInfo->lookup(OrigVD)) {
      DeclRefExpr DRE(const_cast<VarDecl *>(OrigVD),
                      CapturedStmtInfo->lookup(OrigVD) != nullptr,
                      (*IC)->getType(), VK_LValue, (*IC)->getExprLoc());
      Address OrigAddr = EmitLValue(&DRE).getAddress();
      OMPPrivateScope VarScope(*this);
      VarScope.addPrivate(OrigVD,
                          [OrigAddr]() -> Address { return OrigAddr; });
      (void)VarScope.Privatize();
      EmitIgnoredExpr(F);
    }
    ++IC;
  }
  emitLinearClauseFinal(*this, D);
}

void CodeGenFunction::EmitOMPSimdDirective(const OMPSimdDirective &S) {
  auto &&CodeGen = [&S](CodeGenFunction &CGF) {
    // if (PreCond) {
    //   for (IV in 0..LastIteration) BODY;
    //   <Final counter/linear vars updates>;
    // }
    //

    // Emit: if (PreCond) - begin.
    // If the condition constant folds and can be elided, avoid emitting the
    // whole loop.
    bool CondConstant;
    llvm::BasicBlock *ContBlock = nullptr;
    if (CGF.ConstantFoldsToSimpleInteger(S.getPreCond(), CondConstant)) {
      if (!CondConstant)
        return;
    } else {
      auto *ThenBlock = CGF.createBasicBlock("simd.if.then");
      ContBlock = CGF.createBasicBlock("simd.if.end");
      emitPreCond(CGF, S, S.getPreCond(), ThenBlock, ContBlock,
                  CGF.getProfileCount(&S));
      CGF.EmitBlock(ThenBlock);
      CGF.incrementProfileCounter(&S);
    }

    // Emit the loop iteration variable.
    const Expr *IVExpr = S.getIterationVariable();
    const VarDecl *IVDecl = cast<VarDecl>(cast<DeclRefExpr>(IVExpr)->getDecl());
    CGF.EmitVarDecl(*IVDecl);
    CGF.EmitIgnoredExpr(S.getInit());

    // Emit the iterations count variable.
    // If it is not a variable, Sema decided to calculate iterations count on
    // each iteration (e.g., it is foldable into a constant).
    if (auto LIExpr = dyn_cast<DeclRefExpr>(S.getLastIteration())) {
      CGF.EmitVarDecl(*cast<VarDecl>(LIExpr->getDecl()));
      // Emit calculation of the iterations count.
      CGF.EmitIgnoredExpr(S.getCalcLastIteration());
    }

    CGF.EmitOMPSimdInit(S);

    emitAlignedClause(CGF, S);
    CGF.EmitOMPLinearClauseInit(S);
    bool HasLastprivateClause;
    {
      OMPPrivateScope LoopScope(CGF);
      emitPrivateLoopCounters(CGF, LoopScope, S.counters(),
                              S.private_counters());
      emitPrivateLinearVars(CGF, S, LoopScope);
      CGF.EmitOMPPrivateClause(S, LoopScope);
      CGF.EmitOMPReductionClauseInit(S, LoopScope);
      HasLastprivateClause = CGF.EmitOMPLastprivateClauseInit(S, LoopScope);
      (void)LoopScope.Privatize();
      CGF.EmitOMPInnerLoop(S, LoopScope.requiresCleanups(), S.getCond(),
                           S.getInc(),
                           [&S](CodeGenFunction &CGF) {
                             CGF.EmitOMPLoopBody(S, JumpDest());
                             CGF.EmitStopPoint(&S);
                           },
                           [](CodeGenFunction &) {});
      // Emit final copy of the lastprivate variables at the end of loops.
      if (HasLastprivateClause) {
        CGF.EmitOMPLastprivateClauseFinal(S);
      }
      CGF.EmitOMPReductionClauseFinal(S);
    }
    CGF.EmitOMPSimdFinal(S);
    // Emit: if (PreCond) - end.
    if (ContBlock) {
      CGF.EmitBranch(ContBlock);
      CGF.EmitBlock(ContBlock, true);
    }
  };
  CGM.getOpenMPRuntime().emitInlinedDirective(*this, OMPD_simd, CodeGen);
}

void CodeGenFunction::EmitOMPForOuterLoop(
    OpenMPScheduleClauseKind ScheduleKind, bool IsMonotonic,
    const OMPLoopDirective &S, OMPPrivateScope &LoopScope, bool Ordered,
    Address LB, Address UB, Address ST, Address IL, llvm::Value *Chunk) {
  auto &RT = CGM.getOpenMPRuntime();

  // Dynamic scheduling of the outer loop (dynamic, guided, auto, runtime).
  const bool DynamicOrOrdered = Ordered || RT.isDynamic(ScheduleKind);

  assert((Ordered ||
          !RT.isStaticNonchunked(ScheduleKind, /*Chunked=*/Chunk != nullptr)) &&
         "static non-chunked schedule does not need outer loop");

  // Emit outer loop.
  //
  // OpenMP [2.7.1, Loop Construct, Description, table 2-1]
  // When schedule(dynamic,chunk_size) is specified, the iterations are
  // distributed to threads in the team in chunks as the threads request them.
  // Each thread executes a chunk of iterations, then requests another chunk,
  // until no chunks remain to be distributed. Each chunk contains chunk_size
  // iterations, except for the last chunk to be distributed, which may have
  // fewer iterations. When no chunk_size is specified, it defaults to 1.
  //
  // When schedule(guided,chunk_size) is specified, the iterations are assigned
  // to threads in the team in chunks as the executing threads request them.
  // Each thread executes a chunk of iterations, then requests another chunk,
  // until no chunks remain to be assigned. For a chunk_size of 1, the size of
  // each chunk is proportional to the number of unassigned iterations divided
  // by the number of threads in the team, decreasing to 1. For a chunk_size
  // with value k (greater than 1), the size of each chunk is determined in the
  // same way, with the restriction that the chunks do not contain fewer than k
  // iterations (except for the last chunk to be assigned, which may have fewer
  // than k iterations).
  //
  // When schedule(auto) is specified, the decision regarding scheduling is
  // delegated to the compiler and/or runtime system. The programmer gives the
  // implementation the freedom to choose any possible mapping of iterations to
  // threads in the team.
  //
  // When schedule(runtime) is specified, the decision regarding scheduling is
  // deferred until run time, and the schedule and chunk size are taken from the
  // run-sched-var ICV. If the ICV is set to auto, the schedule is
  // implementation defined
  //
  // while(__kmpc_dispatch_next(&LB, &UB)) {
  //   idx = LB;
  //   while (idx <= UB) { BODY; ++idx;
  //   __kmpc_dispatch_fini_(4|8)[u](); // For ordered loops only.
  //   } // inner loop
  // }
  //
  // OpenMP [2.7.1, Loop Construct, Description, table 2-1]
  // When schedule(static, chunk_size) is specified, iterations are divided into
  // chunks of size chunk_size, and the chunks are assigned to the threads in
  // the team in a round-robin fashion in the order of the thread number.
  //
  // while(UB = min(UB, GlobalUB), idx = LB, idx < UB) {
  //   while (idx <= UB) { BODY; ++idx; } // inner loop
  //   LB = LB + ST;
  //   UB = UB + ST;
  // }
  //

  const Expr *IVExpr = S.getIterationVariable();
  const unsigned IVSize = getContext().getTypeSize(IVExpr->getType());
  const bool IVSigned = IVExpr->getType()->hasSignedIntegerRepresentation();

  if (DynamicOrOrdered) {
    llvm::Value *UBVal = EmitScalarExpr(S.getLastIteration());
    RT.emitForDispatchInit(*this, S.getLocStart(), ScheduleKind,
                           IVSize, IVSigned, Ordered, UBVal, Chunk);
  } else {
    RT.emitForStaticInit(*this, S.getLocStart(), ScheduleKind,
                         IVSize, IVSigned, Ordered, IL, LB, UB, ST, Chunk);
  }

  auto LoopExit = getJumpDestInCurrentScope("omp.dispatch.end");

  // Start the loop with a block that tests the condition.
  auto CondBlock = createBasicBlock("omp.dispatch.cond");
  EmitBlock(CondBlock);
  LoopStack.push(CondBlock);

  llvm::Value *BoolCondVal = nullptr;
  if (!DynamicOrOrdered) {
    // UB = min(UB, GlobalUB)
    EmitIgnoredExpr(S.getEnsureUpperBound());
    // IV = LB
    EmitIgnoredExpr(S.getInit());
    // IV < UB
    BoolCondVal = EvaluateExprAsBool(S.getCond());
  } else {
    BoolCondVal = RT.emitForNext(*this, S.getLocStart(), IVSize, IVSigned,
                                    IL, LB, UB, ST);
  }

  // If there are any cleanups between here and the loop-exit scope,
  // create a block to stage a loop exit along.
  auto ExitBlock = LoopExit.getBlock();
  if (LoopScope.requiresCleanups())
    ExitBlock = createBasicBlock("omp.dispatch.cleanup");

  auto LoopBody = createBasicBlock("omp.dispatch.body");
  Builder.CreateCondBr(BoolCondVal, LoopBody, ExitBlock);
  if (ExitBlock != LoopExit.getBlock()) {
    EmitBlock(ExitBlock);
    EmitBranchThroughCleanup(LoopExit);
  }
  EmitBlock(LoopBody);

  // Emit "IV = LB" (in case of static schedule, we have already calculated new
  // LB for loop condition and emitted it above).
  if (DynamicOrOrdered)
    EmitIgnoredExpr(S.getInit());

  // Create a block for the increment.
  auto Continue = getJumpDestInCurrentScope("omp.dispatch.inc");
  BreakContinueStack.push_back(BreakContinue(LoopExit, Continue));

  // Generate !llvm.loop.parallel metadata for loads and stores for loops
  // with dynamic/guided scheduling and without ordered clause.
  if (!isOpenMPSimdDirective(S.getDirectiveKind()))
    LoopStack.setParallel(!IsMonotonic);
  else
    EmitOMPSimdInit(S, IsMonotonic);

  SourceLocation Loc = S.getLocStart();
  EmitOMPInnerLoop(S, LoopScope.requiresCleanups(), S.getCond(), S.getInc(),
                   [&S, LoopExit](CodeGenFunction &CGF) {
                     CGF.EmitOMPLoopBody(S, LoopExit);
                     CGF.EmitStopPoint(&S);
                   },
                   [Ordered, IVSize, IVSigned, Loc](CodeGenFunction &CGF) {
                     if (Ordered) {
                       CGF.CGM.getOpenMPRuntime().emitForOrderedIterationEnd(
                           CGF, Loc, IVSize, IVSigned);
                     }
                   });

  EmitBlock(Continue.getBlock());
  BreakContinueStack.pop_back();
  if (!DynamicOrOrdered) {
    // Emit "LB = LB + Stride", "UB = UB + Stride".
    EmitIgnoredExpr(S.getNextLowerBound());
    EmitIgnoredExpr(S.getNextUpperBound());
  }

  EmitBranch(CondBlock);
  LoopStack.pop();
  // Emit the fall-through block.
  EmitBlock(LoopExit.getBlock());

  // Tell the runtime we are done.
  if (!DynamicOrOrdered)
    RT.emitForStaticFinish(*this, S.getLocEnd());
}

/// \brief Emit a helper variable and return corresponding lvalue.
static LValue EmitOMPHelperVar(CodeGenFunction &CGF,
                               const DeclRefExpr *Helper) {
  auto VDecl = cast<VarDecl>(Helper->getDecl());
  CGF.EmitVarDecl(*VDecl);
  return CGF.EmitLValue(Helper);
}

namespace {
  struct ScheduleKindModifiersTy {
    OpenMPScheduleClauseKind Kind;
    OpenMPScheduleClauseModifier M1;
    OpenMPScheduleClauseModifier M2;
    ScheduleKindModifiersTy(OpenMPScheduleClauseKind Kind,
                            OpenMPScheduleClauseModifier M1,
                            OpenMPScheduleClauseModifier M2)
        : Kind(Kind), M1(M1), M2(M2) {}
  };
} // namespace

static std::pair<llvm::Value * /*Chunk*/, ScheduleKindModifiersTy>
emitScheduleClause(CodeGenFunction &CGF, const OMPLoopDirective &S,
                   bool OuterRegion) {
  // Detect the loop schedule kind and chunk.
  auto ScheduleKind = OMPC_SCHEDULE_unknown;
  OpenMPScheduleClauseModifier M1 = OMPC_SCHEDULE_MODIFIER_unknown;
  OpenMPScheduleClauseModifier M2 = OMPC_SCHEDULE_MODIFIER_unknown;
  llvm::Value *Chunk = nullptr;
  if (const auto *C = S.getSingleClause<OMPScheduleClause>()) {
    ScheduleKind = C->getScheduleKind();
    M1 = C->getFirstScheduleModifier();
    M2 = C->getSecondScheduleModifier();
    if (const auto *Ch = C->getChunkSize()) {
      if (auto *ImpRef = cast_or_null<DeclRefExpr>(C->getHelperChunkSize())) {
        if (OuterRegion) {
          const VarDecl *ImpVar = cast<VarDecl>(ImpRef->getDecl());
          CGF.EmitVarDecl(*ImpVar);
          CGF.EmitStoreThroughLValue(
              CGF.EmitAnyExpr(Ch),
              CGF.MakeAddrLValue(CGF.GetAddrOfLocalVar(ImpVar),
                                 ImpVar->getType()));
        } else {
          Ch = ImpRef;
        }
      }
      if (!C->getHelperChunkSize() || !OuterRegion) {
        Chunk = CGF.EmitScalarExpr(Ch);
        Chunk = CGF.EmitScalarConversion(Chunk, Ch->getType(),
                                         S.getIterationVariable()->getType(),
                                         S.getLocStart());
      }
    }
  }
  return std::make_pair(Chunk, ScheduleKindModifiersTy(ScheduleKind, M1, M2));
}

bool CodeGenFunction::EmitOMPWorksharingLoop(const OMPLoopDirective &S) {
  // Emit the loop iteration variable.
  auto IVExpr = cast<DeclRefExpr>(S.getIterationVariable());
  auto IVDecl = cast<VarDecl>(IVExpr->getDecl());
  EmitVarDecl(*IVDecl);

  // Emit the iterations count variable.
  // If it is not a variable, Sema decided to calculate iterations count on each
  // iteration (e.g., it is foldable into a constant).
  if (auto LIExpr = dyn_cast<DeclRefExpr>(S.getLastIteration())) {
    EmitVarDecl(*cast<VarDecl>(LIExpr->getDecl()));
    // Emit calculation of the iterations count.
    EmitIgnoredExpr(S.getCalcLastIteration());
  }

  auto &RT = CGM.getOpenMPRuntime();

  bool HasLastprivateClause;
  // Check pre-condition.
  {
    // Skip the entire loop if we don't meet the precondition.
    // If the condition constant folds and can be elided, avoid emitting the
    // whole loop.
    bool CondConstant;
    llvm::BasicBlock *ContBlock = nullptr;
    if (ConstantFoldsToSimpleInteger(S.getPreCond(), CondConstant)) {
      if (!CondConstant)
        return false;
    } else {
      auto *ThenBlock = createBasicBlock("omp.precond.then");
      ContBlock = createBasicBlock("omp.precond.end");
      emitPreCond(*this, S, S.getPreCond(), ThenBlock, ContBlock,
                  getProfileCount(&S));
      EmitBlock(ThenBlock);
      incrementProfileCounter(&S);
    }

    emitAlignedClause(*this, S);
    EmitOMPLinearClauseInit(S);
    // Emit 'then' code.
    {
      // Emit helper vars inits.
      LValue LB =
          EmitOMPHelperVar(*this, cast<DeclRefExpr>(S.getLowerBoundVariable()));
      LValue UB =
          EmitOMPHelperVar(*this, cast<DeclRefExpr>(S.getUpperBoundVariable()));
      LValue ST =
          EmitOMPHelperVar(*this, cast<DeclRefExpr>(S.getStrideVariable()));
      LValue IL =
          EmitOMPHelperVar(*this, cast<DeclRefExpr>(S.getIsLastIterVariable()));

      OMPPrivateScope LoopScope(*this);
      if (EmitOMPFirstprivateClause(S, LoopScope)) {
        // Emit implicit barrier to synchronize threads and avoid data races on
        // initialization of firstprivate variables.
        CGM.getOpenMPRuntime().emitBarrierCall(
            *this, S.getLocStart(), OMPD_unknown, /*EmitChecks=*/false,
            /*ForceSimpleCall=*/true);
      }
      EmitOMPPrivateClause(S, LoopScope);
      HasLastprivateClause = EmitOMPLastprivateClauseInit(S, LoopScope);
      EmitOMPReductionClauseInit(S, LoopScope);
      emitPrivateLoopCounters(*this, LoopScope, S.counters(),
                              S.private_counters());
      emitPrivateLinearVars(*this, S, LoopScope);
      (void)LoopScope.Privatize();

      // Detect the loop schedule kind and chunk.
      llvm::Value *Chunk;
      OpenMPScheduleClauseKind ScheduleKind;
      auto ScheduleInfo =
          emitScheduleClause(*this, S, /*OuterRegion=*/false);
      Chunk = ScheduleInfo.first;
      ScheduleKind = ScheduleInfo.second.Kind;
      const OpenMPScheduleClauseModifier M1 = ScheduleInfo.second.M1;
      const OpenMPScheduleClauseModifier M2 = ScheduleInfo.second.M2;
      const unsigned IVSize = getContext().getTypeSize(IVExpr->getType());
      const bool IVSigned = IVExpr->getType()->hasSignedIntegerRepresentation();
      const bool Ordered = S.getSingleClause<OMPOrderedClause>() != nullptr;
      // OpenMP 4.5, 2.7.1 Loop Construct, Description.
      // If the static schedule kind is specified or if the ordered clause is
      // specified, and if no monotonic modifier is specified, the effect will
      // be as if the monotonic modifier was specified.
      if (RT.isStaticNonchunked(ScheduleKind,
                                /* Chunked */ Chunk != nullptr) &&
          !Ordered) {
        if (isOpenMPSimdDirective(S.getDirectiveKind()))
          EmitOMPSimdInit(S, /*IsMonotonic=*/true);
        // OpenMP [2.7.1, Loop Construct, Description, table 2-1]
        // When no chunk_size is specified, the iteration space is divided into
        // chunks that are approximately equal in size, and at most one chunk is
        // distributed to each thread. Note that the size of the chunks is
        // unspecified in this case.
        RT.emitForStaticInit(*this, S.getLocStart(), ScheduleKind,
                             IVSize, IVSigned, Ordered,
                             IL.getAddress(), LB.getAddress(),
                             UB.getAddress(), ST.getAddress());
        auto LoopExit =
            getJumpDestInCurrentScope(createBasicBlock("omp.loop.exit"));
        // UB = min(UB, GlobalUB);
        EmitIgnoredExpr(S.getEnsureUpperBound());
        // IV = LB;
        EmitIgnoredExpr(S.getInit());
        // while (idx <= UB) { BODY; ++idx; }
        EmitOMPInnerLoop(S, LoopScope.requiresCleanups(), S.getCond(),
                         S.getInc(),
                         [&S, LoopExit](CodeGenFunction &CGF) {
                           CGF.EmitOMPLoopBody(S, LoopExit);
                           CGF.EmitStopPoint(&S);
                         },
                         [](CodeGenFunction &) {});
        EmitBlock(LoopExit.getBlock());
        // Tell the runtime we are done.
        RT.emitForStaticFinish(*this, S.getLocStart());
      } else {
        const bool IsMonotonic = Ordered ||
                                 ScheduleKind == OMPC_SCHEDULE_static ||
                                 ScheduleKind == OMPC_SCHEDULE_unknown ||
                                 M1 == OMPC_SCHEDULE_MODIFIER_monotonic ||
                                 M2 == OMPC_SCHEDULE_MODIFIER_monotonic;
        // Emit the outer loop, which requests its work chunk [LB..UB] from
        // runtime and runs the inner loop to process it.
        EmitOMPForOuterLoop(ScheduleKind, IsMonotonic, S, LoopScope, Ordered,
                            LB.getAddress(), UB.getAddress(), ST.getAddress(),
                            IL.getAddress(), Chunk);
      }
      EmitOMPReductionClauseFinal(S);
      // Emit final copy of the lastprivate variables if IsLastIter != 0.
      if (HasLastprivateClause)
        EmitOMPLastprivateClauseFinal(
            S, Builder.CreateIsNotNull(EmitLoadOfScalar(IL, S.getLocStart())));
    }
    if (isOpenMPSimdDirective(S.getDirectiveKind())) {
      EmitOMPSimdFinal(S);
    }
    // We're now done with the loop, so jump to the continuation block.
    if (ContBlock) {
      EmitBranch(ContBlock);
      EmitBlock(ContBlock, true);
    }
  }
  return HasLastprivateClause;
}

void CodeGenFunction::EmitOMPForDirective(const OMPForDirective &S) {
  LexicalScope Scope(*this, S.getSourceRange());
  bool HasLastprivates = false;
  auto &&CodeGen = [&S, &HasLastprivates](CodeGenFunction &CGF) {
    HasLastprivates = CGF.EmitOMPWorksharingLoop(S);
  };
  CGM.getOpenMPRuntime().emitInlinedDirective(*this, OMPD_for, CodeGen,
                                              S.hasCancel());

  // Emit an implicit barrier at the end.
  if (!S.getSingleClause<OMPNowaitClause>() || HasLastprivates) {
    CGM.getOpenMPRuntime().emitBarrierCall(*this, S.getLocStart(), OMPD_for);
  }
}

void CodeGenFunction::EmitOMPForSimdDirective(const OMPForSimdDirective &S) {
  LexicalScope Scope(*this, S.getSourceRange());
  bool HasLastprivates = false;
  auto &&CodeGen = [&S, &HasLastprivates](CodeGenFunction &CGF) {
    HasLastprivates = CGF.EmitOMPWorksharingLoop(S);
  };
  CGM.getOpenMPRuntime().emitInlinedDirective(*this, OMPD_simd, CodeGen);

  // Emit an implicit barrier at the end.
  if (!S.getSingleClause<OMPNowaitClause>() || HasLastprivates) {
    CGM.getOpenMPRuntime().emitBarrierCall(*this, S.getLocStart(), OMPD_for);
  }
}

static LValue createSectionLVal(CodeGenFunction &CGF, QualType Ty,
                                const Twine &Name,
                                llvm::Value *Init = nullptr) {
  auto LVal = CGF.MakeAddrLValue(CGF.CreateMemTemp(Ty, Name), Ty);
  if (Init)
    CGF.EmitScalarInit(Init, LVal);
  return LVal;
}

OpenMPDirectiveKind
CodeGenFunction::EmitSections(const OMPExecutableDirective &S) {
  auto *Stmt = cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt();
  auto *CS = dyn_cast<CompoundStmt>(Stmt);
  if (CS && CS->size() > 1) {
    bool HasLastprivates = false;
    auto &&CodeGen = [&S, CS, &HasLastprivates](CodeGenFunction &CGF) {
      auto &C = CGF.CGM.getContext();
      auto KmpInt32Ty = C.getIntTypeForBitwidth(/*DestWidth=*/32, /*Signed=*/1);
      // Emit helper vars inits.
      LValue LB = createSectionLVal(CGF, KmpInt32Ty, ".omp.sections.lb.",
                                    CGF.Builder.getInt32(0));
      auto *GlobalUBVal = CGF.Builder.getInt32(CS->size() - 1);
      LValue UB =
          createSectionLVal(CGF, KmpInt32Ty, ".omp.sections.ub.", GlobalUBVal);
      LValue ST = createSectionLVal(CGF, KmpInt32Ty, ".omp.sections.st.",
                                    CGF.Builder.getInt32(1));
      LValue IL = createSectionLVal(CGF, KmpInt32Ty, ".omp.sections.il.",
                                    CGF.Builder.getInt32(0));
      // Loop counter.
      LValue IV = createSectionLVal(CGF, KmpInt32Ty, ".omp.sections.iv.");
      OpaqueValueExpr IVRefExpr(S.getLocStart(), KmpInt32Ty, VK_LValue);
      CodeGenFunction::OpaqueValueMapping OpaqueIV(CGF, &IVRefExpr, IV);
      OpaqueValueExpr UBRefExpr(S.getLocStart(), KmpInt32Ty, VK_LValue);
      CodeGenFunction::OpaqueValueMapping OpaqueUB(CGF, &UBRefExpr, UB);
      // Generate condition for loop.
      BinaryOperator Cond(&IVRefExpr, &UBRefExpr, BO_LE, C.BoolTy, VK_RValue,
                          OK_Ordinary, S.getLocStart(),
                          /*fpContractable=*/false);
      // Increment for loop counter.
      UnaryOperator Inc(&IVRefExpr, UO_PreInc, KmpInt32Ty, VK_RValue,
                        OK_Ordinary, S.getLocStart());
      auto BodyGen = [CS, &S, &IV](CodeGenFunction &CGF) {
        // Iterate through all sections and emit a switch construct:
        // switch (IV) {
        //   case 0:
        //     <SectionStmt[0]>;
        //     break;
        // ...
        //   case <NumSection> - 1:
        //     <SectionStmt[<NumSection> - 1]>;
        //     break;
        // }
        // .omp.sections.exit:
        auto *ExitBB = CGF.createBasicBlock(".omp.sections.exit");
        auto *SwitchStmt = CGF.Builder.CreateSwitch(
            CGF.EmitLoadOfLValue(IV, S.getLocStart()).getScalarVal(), ExitBB,
            CS->size());
        unsigned CaseNumber = 0;
        for (auto *SubStmt : CS->children()) {
          auto CaseBB = CGF.createBasicBlock(".omp.sections.case");
          CGF.EmitBlock(CaseBB);
          SwitchStmt->addCase(CGF.Builder.getInt32(CaseNumber), CaseBB);
          CGF.EmitStmt(SubStmt);
          CGF.EmitBranch(ExitBB);
          ++CaseNumber;
        }
        CGF.EmitBlock(ExitBB, /*IsFinished=*/true);
      };

      CodeGenFunction::OMPPrivateScope LoopScope(CGF);
      if (CGF.EmitOMPFirstprivateClause(S, LoopScope)) {
        // Emit implicit barrier to synchronize threads and avoid data races on
        // initialization of firstprivate variables.
        CGF.CGM.getOpenMPRuntime().emitBarrierCall(
            CGF, S.getLocStart(), OMPD_unknown, /*EmitChecks=*/false,
            /*ForceSimpleCall=*/true);
      }
      CGF.EmitOMPPrivateClause(S, LoopScope);
      HasLastprivates = CGF.EmitOMPLastprivateClauseInit(S, LoopScope);
      CGF.EmitOMPReductionClauseInit(S, LoopScope);
      (void)LoopScope.Privatize();

      // Emit static non-chunked loop.
      CGF.CGM.getOpenMPRuntime().emitForStaticInit(
          CGF, S.getLocStart(), OMPC_SCHEDULE_static, /*IVSize=*/32,
          /*IVSigned=*/true, /*Ordered=*/false, IL.getAddress(),
          LB.getAddress(), UB.getAddress(), ST.getAddress());
      // UB = min(UB, GlobalUB);
      auto *UBVal = CGF.EmitLoadOfScalar(UB, S.getLocStart());
      auto *MinUBGlobalUB = CGF.Builder.CreateSelect(
          CGF.Builder.CreateICmpSLT(UBVal, GlobalUBVal), UBVal, GlobalUBVal);
      CGF.EmitStoreOfScalar(MinUBGlobalUB, UB);
      // IV = LB;
      CGF.EmitStoreOfScalar(CGF.EmitLoadOfScalar(LB, S.getLocStart()), IV);
      // while (idx <= UB) { BODY; ++idx; }
      CGF.EmitOMPInnerLoop(S, /*RequiresCleanup=*/false, &Cond, &Inc, BodyGen,
                           [](CodeGenFunction &) {});
      // Tell the runtime we are done.
      CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocStart());
      CGF.EmitOMPReductionClauseFinal(S);

      // Emit final copy of the lastprivate variables if IsLastIter != 0.
      if (HasLastprivates)
        CGF.EmitOMPLastprivateClauseFinal(
            S, CGF.Builder.CreateIsNotNull(
                   CGF.EmitLoadOfScalar(IL, S.getLocStart())));
    };

    bool HasCancel = false;
    if (auto *OSD = dyn_cast<OMPSectionsDirective>(&S))
      HasCancel = OSD->hasCancel();
    else if (auto *OPSD = dyn_cast<OMPParallelSectionsDirective>(&S))
      HasCancel = OPSD->hasCancel();
    CGM.getOpenMPRuntime().emitInlinedDirective(*this, OMPD_sections, CodeGen,
                                                HasCancel);
    // Emit barrier for lastprivates only if 'sections' directive has 'nowait'
    // clause. Otherwise the barrier will be generated by the codegen for the
    // directive.
    if (HasLastprivates && S.getSingleClause<OMPNowaitClause>()) {
      // Emit implicit barrier to synchronize threads and avoid data races on
      // initialization of firstprivate variables.
      CGM.getOpenMPRuntime().emitBarrierCall(*this, S.getLocStart(),
                                             OMPD_unknown);
    }
    return OMPD_sections;
  }
  // If only one section is found - no need to generate loop, emit as a single
  // region.
  bool HasFirstprivates;
  // No need to generate reductions for sections with single section region, we
  // can use original shared variables for all operations.
  bool HasReductions = S.hasClausesOfKind<OMPReductionClause>();
  // No need to generate lastprivates for sections with single section region,
  // we can use original shared variable for all calculations with barrier at
  // the end of the sections.
  bool HasLastprivates = S.hasClausesOfKind<OMPLastprivateClause>();
  auto &&CodeGen = [Stmt, &S, &HasFirstprivates](CodeGenFunction &CGF) {
    CodeGenFunction::OMPPrivateScope SingleScope(CGF);
    HasFirstprivates = CGF.EmitOMPFirstprivateClause(S, SingleScope);
    CGF.EmitOMPPrivateClause(S, SingleScope);
    (void)SingleScope.Privatize();

    CGF.EmitStmt(Stmt);
  };
  CGM.getOpenMPRuntime().emitSingleRegion(*this, CodeGen, S.getLocStart(),
                                          llvm::None, llvm::None, llvm::None,
                                          llvm::None);
  // Emit barrier for firstprivates, lastprivates or reductions only if
  // 'sections' directive has 'nowait' clause. Otherwise the barrier will be
  // generated by the codegen for the directive.
  if ((HasFirstprivates || HasLastprivates || HasReductions) &&
      S.getSingleClause<OMPNowaitClause>()) {
    // Emit implicit barrier to synchronize threads and avoid data races on
    // initialization of firstprivate variables.
    CGM.getOpenMPRuntime().emitBarrierCall(*this, S.getLocStart(), OMPD_unknown,
                                           /*EmitChecks=*/false,
                                           /*ForceSimpleCall=*/true);
  }
  return OMPD_single;
}

void CodeGenFunction::EmitOMPSectionsDirective(const OMPSectionsDirective &S) {
  LexicalScope Scope(*this, S.getSourceRange());
  OpenMPDirectiveKind EmittedAs = EmitSections(S);
  // Emit an implicit barrier at the end.
  if (!S.getSingleClause<OMPNowaitClause>()) {
    CGM.getOpenMPRuntime().emitBarrierCall(*this, S.getLocStart(), EmittedAs);
  }
}

void CodeGenFunction::EmitOMPSectionDirective(const OMPSectionDirective &S) {
  LexicalScope Scope(*this, S.getSourceRange());
  auto &&CodeGen = [&S](CodeGenFunction &CGF) {
    CGF.EmitStmt(cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
  };
  CGM.getOpenMPRuntime().emitInlinedDirective(*this, OMPD_section, CodeGen,
                                              S.hasCancel());
}

void CodeGenFunction::EmitOMPSingleDirective(const OMPSingleDirective &S) {
  llvm::SmallVector<const Expr *, 8> CopyprivateVars;
  llvm::SmallVector<const Expr *, 8> DestExprs;
  llvm::SmallVector<const Expr *, 8> SrcExprs;
  llvm::SmallVector<const Expr *, 8> AssignmentOps;
  // Check if there are any 'copyprivate' clauses associated with this
  // 'single'
  // construct.
  // Build a list of copyprivate variables along with helper expressions
  // (<source>, <destination>, <destination>=<source> expressions)
  for (const auto *C : S.getClausesOfKind<OMPCopyprivateClause>()) {
    CopyprivateVars.append(C->varlists().begin(), C->varlists().end());
    DestExprs.append(C->destination_exprs().begin(),
                     C->destination_exprs().end());
    SrcExprs.append(C->source_exprs().begin(), C->source_exprs().end());
    AssignmentOps.append(C->assignment_ops().begin(),
                         C->assignment_ops().end());
  }
  LexicalScope Scope(*this, S.getSourceRange());
  // Emit code for 'single' region along with 'copyprivate' clauses
  bool HasFirstprivates;
  auto &&CodeGen = [&S, &HasFirstprivates](CodeGenFunction &CGF) {
    CodeGenFunction::OMPPrivateScope SingleScope(CGF);
    HasFirstprivates = CGF.EmitOMPFirstprivateClause(S, SingleScope);
    CGF.EmitOMPPrivateClause(S, SingleScope);
    (void)SingleScope.Privatize();

    CGF.EmitStmt(cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
  };
  CGM.getOpenMPRuntime().emitSingleRegion(*this, CodeGen, S.getLocStart(),
                                          CopyprivateVars, DestExprs, SrcExprs,
                                          AssignmentOps);
  // Emit an implicit barrier at the end (to avoid data race on firstprivate
  // init or if no 'nowait' clause was specified and no 'copyprivate' clause).
  if ((!S.getSingleClause<OMPNowaitClause>() || HasFirstprivates) &&
      CopyprivateVars.empty()) {
    CGM.getOpenMPRuntime().emitBarrierCall(
        *this, S.getLocStart(),
        S.getSingleClause<OMPNowaitClause>() ? OMPD_unknown : OMPD_single);
  }
}

void CodeGenFunction::EmitOMPMasterDirective(const OMPMasterDirective &S) {
  LexicalScope Scope(*this, S.getSourceRange());
  auto &&CodeGen = [&S](CodeGenFunction &CGF) {
    CGF.EmitStmt(cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
  };
  CGM.getOpenMPRuntime().emitMasterRegion(*this, CodeGen, S.getLocStart());
}

void CodeGenFunction::EmitOMPCriticalDirective(const OMPCriticalDirective &S) {
  LexicalScope Scope(*this, S.getSourceRange());
  auto &&CodeGen = [&S](CodeGenFunction &CGF) {
    CGF.EmitStmt(cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
  };
  Expr *Hint = nullptr;
  if (auto *HintClause = S.getSingleClause<OMPHintClause>())
    Hint = HintClause->getHint();
  CGM.getOpenMPRuntime().emitCriticalRegion(*this,
                                            S.getDirectiveName().getAsString(),
                                            CodeGen, S.getLocStart(), Hint);
}

void CodeGenFunction::EmitOMPParallelForDirective(
    const OMPParallelForDirective &S) {
  // Emit directive as a combined directive that consists of two implicit
  // directives: 'parallel' with 'for' directive.
  LexicalScope Scope(*this, S.getSourceRange());
  (void)emitScheduleClause(*this, S, /*OuterRegion=*/true);
  auto &&CodeGen = [&S](CodeGenFunction &CGF) {
    CGF.EmitOMPWorksharingLoop(S);
  };
  emitCommonOMPParallelDirective(*this, S, OMPD_for, CodeGen);
}

void CodeGenFunction::EmitOMPParallelForSimdDirective(
    const OMPParallelForSimdDirective &S) {
  // Emit directive as a combined directive that consists of two implicit
  // directives: 'parallel' with 'for' directive.
  LexicalScope Scope(*this, S.getSourceRange());
  (void)emitScheduleClause(*this, S, /*OuterRegion=*/true);
  auto &&CodeGen = [&S](CodeGenFunction &CGF) {
    CGF.EmitOMPWorksharingLoop(S);
  };
  emitCommonOMPParallelDirective(*this, S, OMPD_simd, CodeGen);
}

void CodeGenFunction::EmitOMPParallelSectionsDirective(
    const OMPParallelSectionsDirective &S) {
  // Emit directive as a combined directive that consists of two implicit
  // directives: 'parallel' with 'sections' directive.
  LexicalScope Scope(*this, S.getSourceRange());
  auto &&CodeGen = [&S](CodeGenFunction &CGF) {
    (void)CGF.EmitSections(S);
  };
  emitCommonOMPParallelDirective(*this, S, OMPD_sections, CodeGen);
}

void CodeGenFunction::EmitOMPTaskDirective(const OMPTaskDirective &S) {
  // Emit outlined function for task construct.
  LexicalScope Scope(*this, S.getSourceRange());
  auto CS = cast<CapturedStmt>(S.getAssociatedStmt());
  auto CapturedStruct = GenerateCapturedStmtArgument(*CS);
  auto *I = CS->getCapturedDecl()->param_begin();
  auto *PartId = std::next(I);
  // The first function argument for tasks is a thread id, the second one is a
  // part id (0 for tied tasks, >=0 for untied task).
  llvm::DenseSet<const VarDecl *> EmittedAsPrivate;
  // Get list of private variables.
  llvm::SmallVector<const Expr *, 8> PrivateVars;
  llvm::SmallVector<const Expr *, 8> PrivateCopies;
  for (const auto *C : S.getClausesOfKind<OMPPrivateClause>()) {
    auto IRef = C->varlist_begin();
    for (auto *IInit : C->private_copies()) {
      auto *OrigVD = cast<VarDecl>(cast<DeclRefExpr>(*IRef)->getDecl());
      if (EmittedAsPrivate.insert(OrigVD->getCanonicalDecl()).second) {
        PrivateVars.push_back(*IRef);
        PrivateCopies.push_back(IInit);
      }
      ++IRef;
    }
  }
  EmittedAsPrivate.clear();
  // Get list of firstprivate variables.
  llvm::SmallVector<const Expr *, 8> FirstprivateVars;
  llvm::SmallVector<const Expr *, 8> FirstprivateCopies;
  llvm::SmallVector<const Expr *, 8> FirstprivateInits;
  for (const auto *C : S.getClausesOfKind<OMPFirstprivateClause>()) {
    auto IRef = C->varlist_begin();
    auto IElemInitRef = C->inits().begin();
    for (auto *IInit : C->private_copies()) {
      auto *OrigVD = cast<VarDecl>(cast<DeclRefExpr>(*IRef)->getDecl());
      if (EmittedAsPrivate.insert(OrigVD->getCanonicalDecl()).second) {
        FirstprivateVars.push_back(*IRef);
        FirstprivateCopies.push_back(IInit);
        FirstprivateInits.push_back(*IElemInitRef);
      }
      ++IRef, ++IElemInitRef;
    }
  }
  // Build list of dependences.
  llvm::SmallVector<std::pair<OpenMPDependClauseKind, const Expr *>, 8>
      Dependences;
  for (const auto *C : S.getClausesOfKind<OMPDependClause>()) {
    for (auto *IRef : C->varlists()) {
      Dependences.push_back(std::make_pair(C->getDependencyKind(), IRef));
    }
  }
  auto &&CodeGen = [PartId, &S, &PrivateVars, &FirstprivateVars](
      CodeGenFunction &CGF) {
    // Set proper addresses for generated private copies.
    auto *CS = cast<CapturedStmt>(S.getAssociatedStmt());
    OMPPrivateScope Scope(CGF);
    if (!PrivateVars.empty() || !FirstprivateVars.empty()) {
      auto *CopyFn = CGF.Builder.CreateLoad(
          CGF.GetAddrOfLocalVar(CS->getCapturedDecl()->getParam(3)));
      auto *PrivatesPtr = CGF.Builder.CreateLoad(
          CGF.GetAddrOfLocalVar(CS->getCapturedDecl()->getParam(2)));
      // Map privates.
      llvm::SmallVector<std::pair<const VarDecl *, Address>, 16>
          PrivatePtrs;
      llvm::SmallVector<llvm::Value *, 16> CallArgs;
      CallArgs.push_back(PrivatesPtr);
      for (auto *E : PrivateVars) {
        auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
        Address PrivatePtr =
            CGF.CreateMemTemp(CGF.getContext().getPointerType(E->getType()));
        PrivatePtrs.push_back(std::make_pair(VD, PrivatePtr));
        CallArgs.push_back(PrivatePtr.getPointer());
      }
      for (auto *E : FirstprivateVars) {
        auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
        Address PrivatePtr =
            CGF.CreateMemTemp(CGF.getContext().getPointerType(E->getType()));
        PrivatePtrs.push_back(std::make_pair(VD, PrivatePtr));
        CallArgs.push_back(PrivatePtr.getPointer());
      }
      CGF.EmitRuntimeCall(CopyFn, CallArgs);
      for (auto &&Pair : PrivatePtrs) {
        Address Replacement(CGF.Builder.CreateLoad(Pair.second),
                            CGF.getContext().getDeclAlign(Pair.first));
        Scope.addPrivate(Pair.first, [Replacement]() { return Replacement; });
      }
    }
    (void)Scope.Privatize();
    if (*PartId) {
      // TODO: emit code for untied tasks.
    }
    CGF.EmitStmt(CS->getCapturedStmt());
  };
  auto OutlinedFn = CGM.getOpenMPRuntime().emitTaskOutlinedFunction(
      S, *I, OMPD_task, CodeGen);
  // Check if we should emit tied or untied task.
  bool Tied = !S.getSingleClause<OMPUntiedClause>();
  // Check if the task is final
  llvm::PointerIntPair<llvm::Value *, 1, bool> Final;
  if (const auto *Clause = S.getSingleClause<OMPFinalClause>()) {
    // If the condition constant folds and can be elided, try to avoid emitting
    // the condition and the dead arm of the if/else.
    auto *Cond = Clause->getCondition();
    bool CondConstant;
    if (ConstantFoldsToSimpleInteger(Cond, CondConstant))
      Final.setInt(CondConstant);
    else
      Final.setPointer(EvaluateExprAsBool(Cond));
  } else {
    // By default the task is not final.
    Final.setInt(/*IntVal=*/false);
  }
  auto SharedsTy = getContext().getRecordType(CS->getCapturedRecordDecl());
  const Expr *IfCond = nullptr;
  for (const auto *C : S.getClausesOfKind<OMPIfClause>()) {
    if (C->getNameModifier() == OMPD_unknown ||
        C->getNameModifier() == OMPD_task) {
      IfCond = C->getCondition();
      break;
    }
  }
  CGM.getOpenMPRuntime().emitTaskCall(
      *this, S.getLocStart(), S, Tied, Final, OutlinedFn, SharedsTy,
      CapturedStruct, IfCond, PrivateVars, PrivateCopies, FirstprivateVars,
      FirstprivateCopies, FirstprivateInits, Dependences);
}

void CodeGenFunction::EmitOMPTaskyieldDirective(
    const OMPTaskyieldDirective &S) {
  CGM.getOpenMPRuntime().emitTaskyieldCall(*this, S.getLocStart());
}

void CodeGenFunction::EmitOMPBarrierDirective(const OMPBarrierDirective &S) {
  CGM.getOpenMPRuntime().emitBarrierCall(*this, S.getLocStart(), OMPD_barrier);
}

void CodeGenFunction::EmitOMPTaskwaitDirective(const OMPTaskwaitDirective &S) {
  CGM.getOpenMPRuntime().emitTaskwaitCall(*this, S.getLocStart());
}

void CodeGenFunction::EmitOMPTaskgroupDirective(
    const OMPTaskgroupDirective &S) {
  LexicalScope Scope(*this, S.getSourceRange());
  auto &&CodeGen = [&S](CodeGenFunction &CGF) {
    CGF.EmitStmt(cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
  };
  CGM.getOpenMPRuntime().emitTaskgroupRegion(*this, CodeGen, S.getLocStart());
}

void CodeGenFunction::EmitOMPFlushDirective(const OMPFlushDirective &S) {
  CGM.getOpenMPRuntime().emitFlush(*this, [&]() -> ArrayRef<const Expr *> {
    if (const auto *FlushClause = S.getSingleClause<OMPFlushClause>()) {
      return llvm::makeArrayRef(FlushClause->varlist_begin(),
                                FlushClause->varlist_end());
    }
    return llvm::None;
  }(), S.getLocStart());
}

void CodeGenFunction::EmitOMPDistributeDirective(
    const OMPDistributeDirective &S) {
  llvm_unreachable("CodeGen for 'omp distribute' is not supported yet.");
}

static llvm::Function *emitOutlinedOrderedFunction(CodeGenModule &CGM,
                                                   const CapturedStmt *S) {
  CodeGenFunction CGF(CGM, /*suppressNewContext=*/true);
  CodeGenFunction::CGCapturedStmtInfo CapStmtInfo;
  CGF.CapturedStmtInfo = &CapStmtInfo;
  auto *Fn = CGF.GenerateOpenMPCapturedStmtFunction(*S);
  Fn->addFnAttr(llvm::Attribute::NoInline);
  return Fn;
}

void CodeGenFunction::EmitOMPOrderedDirective(const OMPOrderedDirective &S) {
  if (!S.getAssociatedStmt())
    return;
  LexicalScope Scope(*this, S.getSourceRange());
  auto *C = S.getSingleClause<OMPSIMDClause>();
  auto &&CodeGen = [&S, C, this](CodeGenFunction &CGF) {
    if (C) {
      auto CS = cast<CapturedStmt>(S.getAssociatedStmt());
      llvm::SmallVector<llvm::Value *, 16> CapturedVars;
      CGF.GenerateOpenMPCapturedVars(*CS, CapturedVars);
      auto *OutlinedFn = emitOutlinedOrderedFunction(CGM, CS);
      CGF.EmitNounwindRuntimeCall(OutlinedFn, CapturedVars);
    } else {
      CGF.EmitStmt(
          cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
    }
  };
  CGM.getOpenMPRuntime().emitOrderedRegion(*this, CodeGen, S.getLocStart(), !C);
}

static llvm::Value *convertToScalarValue(CodeGenFunction &CGF, RValue Val,
                                         QualType SrcType, QualType DestType,
                                         SourceLocation Loc) {
  assert(CGF.hasScalarEvaluationKind(DestType) &&
         "DestType must have scalar evaluation kind.");
  assert(!Val.isAggregate() && "Must be a scalar or complex.");
  return Val.isScalar()
             ? CGF.EmitScalarConversion(Val.getScalarVal(), SrcType, DestType,
                                        Loc)
             : CGF.EmitComplexToScalarConversion(Val.getComplexVal(), SrcType,
                                                 DestType, Loc);
}

static CodeGenFunction::ComplexPairTy
convertToComplexValue(CodeGenFunction &CGF, RValue Val, QualType SrcType,
                      QualType DestType, SourceLocation Loc) {
  assert(CGF.getEvaluationKind(DestType) == TEK_Complex &&
         "DestType must have complex evaluation kind.");
  CodeGenFunction::ComplexPairTy ComplexVal;
  if (Val.isScalar()) {
    // Convert the input element to the element type of the complex.
    auto DestElementType = DestType->castAs<ComplexType>()->getElementType();
    auto ScalarVal = CGF.EmitScalarConversion(Val.getScalarVal(), SrcType,
                                              DestElementType, Loc);
    ComplexVal = CodeGenFunction::ComplexPairTy(
        ScalarVal, llvm::Constant::getNullValue(ScalarVal->getType()));
  } else {
    assert(Val.isComplex() && "Must be a scalar or complex.");
    auto SrcElementType = SrcType->castAs<ComplexType>()->getElementType();
    auto DestElementType = DestType->castAs<ComplexType>()->getElementType();
    ComplexVal.first = CGF.EmitScalarConversion(
        Val.getComplexVal().first, SrcElementType, DestElementType, Loc);
    ComplexVal.second = CGF.EmitScalarConversion(
        Val.getComplexVal().second, SrcElementType, DestElementType, Loc);
  }
  return ComplexVal;
}

static void emitSimpleAtomicStore(CodeGenFunction &CGF, bool IsSeqCst,
                                  LValue LVal, RValue RVal) {
  if (LVal.isGlobalReg()) {
    CGF.EmitStoreThroughGlobalRegLValue(RVal, LVal);
  } else {
    CGF.EmitAtomicStore(RVal, LVal, IsSeqCst ? llvm::SequentiallyConsistent
                                             : llvm::Monotonic,
                        LVal.isVolatile(), /*IsInit=*/false);
  }
}

void CodeGenFunction::emitOMPSimpleStore(LValue LVal, RValue RVal,
                                         QualType RValTy, SourceLocation Loc) {
  switch (getEvaluationKind(LVal.getType())) {
  case TEK_Scalar:
    EmitStoreThroughLValue(RValue::get(convertToScalarValue(
                               *this, RVal, RValTy, LVal.getType(), Loc)),
                           LVal);
    break;
  case TEK_Complex:
    EmitStoreOfComplex(
        convertToComplexValue(*this, RVal, RValTy, LVal.getType(), Loc), LVal,
        /*isInit=*/false);
    break;
  case TEK_Aggregate:
    llvm_unreachable("Must be a scalar or complex.");
  }
}

static void EmitOMPAtomicReadExpr(CodeGenFunction &CGF, bool IsSeqCst,
                                  const Expr *X, const Expr *V,
                                  SourceLocation Loc) {
  // v = x;
  assert(V->isLValue() && "V of 'omp atomic read' is not lvalue");
  assert(X->isLValue() && "X of 'omp atomic read' is not lvalue");
  LValue XLValue = CGF.EmitLValue(X);
  LValue VLValue = CGF.EmitLValue(V);
  RValue Res = XLValue.isGlobalReg()
                   ? CGF.EmitLoadOfLValue(XLValue, Loc)
                   : CGF.EmitAtomicLoad(XLValue, Loc,
                                        IsSeqCst ? llvm::SequentiallyConsistent
                                                 : llvm::Monotonic,
                                        XLValue.isVolatile());
  // OpenMP, 2.12.6, atomic Construct
  // Any atomic construct with a seq_cst clause forces the atomically
  // performed operation to include an implicit flush operation without a
  // list.
  if (IsSeqCst)
    CGF.CGM.getOpenMPRuntime().emitFlush(CGF, llvm::None, Loc);
  CGF.emitOMPSimpleStore(VLValue, Res, X->getType().getNonReferenceType(), Loc);
}

static void EmitOMPAtomicWriteExpr(CodeGenFunction &CGF, bool IsSeqCst,
                                   const Expr *X, const Expr *E,
                                   SourceLocation Loc) {
  // x = expr;
  assert(X->isLValue() && "X of 'omp atomic write' is not lvalue");
  emitSimpleAtomicStore(CGF, IsSeqCst, CGF.EmitLValue(X), CGF.EmitAnyExpr(E));
  // OpenMP, 2.12.6, atomic Construct
  // Any atomic construct with a seq_cst clause forces the atomically
  // performed operation to include an implicit flush operation without a
  // list.
  if (IsSeqCst)
    CGF.CGM.getOpenMPRuntime().emitFlush(CGF, llvm::None, Loc);
}

static std::pair<bool, RValue> emitOMPAtomicRMW(CodeGenFunction &CGF, LValue X,
                                                RValue Update,
                                                BinaryOperatorKind BO,
                                                llvm::AtomicOrdering AO,
                                                bool IsXLHSInRHSPart) {
  auto &Context = CGF.CGM.getContext();
  // Allow atomicrmw only if 'x' and 'update' are integer values, lvalue for 'x'
  // expression is simple and atomic is allowed for the given type for the
  // target platform.
  if (BO == BO_Comma || !Update.isScalar() ||
      !Update.getScalarVal()->getType()->isIntegerTy() ||
      !X.isSimple() || (!isa<llvm::ConstantInt>(Update.getScalarVal()) &&
                        (Update.getScalarVal()->getType() !=
                         X.getAddress().getElementType())) ||
      !X.getAddress().getElementType()->isIntegerTy() ||
      !Context.getTargetInfo().hasBuiltinAtomic(
          Context.getTypeSize(X.getType()), Context.toBits(X.getAlignment())))
    return std::make_pair(false, RValue::get(nullptr));

  llvm::AtomicRMWInst::BinOp RMWOp;
  switch (BO) {
  case BO_Add:
    RMWOp = llvm::AtomicRMWInst::Add;
    break;
  case BO_Sub:
    if (!IsXLHSInRHSPart)
      return std::make_pair(false, RValue::get(nullptr));
    RMWOp = llvm::AtomicRMWInst::Sub;
    break;
  case BO_And:
    RMWOp = llvm::AtomicRMWInst::And;
    break;
  case BO_Or:
    RMWOp = llvm::AtomicRMWInst::Or;
    break;
  case BO_Xor:
    RMWOp = llvm::AtomicRMWInst::Xor;
    break;
  case BO_LT:
    RMWOp = X.getType()->hasSignedIntegerRepresentation()
                ? (IsXLHSInRHSPart ? llvm::AtomicRMWInst::Min
                                   : llvm::AtomicRMWInst::Max)
                : (IsXLHSInRHSPart ? llvm::AtomicRMWInst::UMin
                                   : llvm::AtomicRMWInst::UMax);
    break;
  case BO_GT:
    RMWOp = X.getType()->hasSignedIntegerRepresentation()
                ? (IsXLHSInRHSPart ? llvm::AtomicRMWInst::Max
                                   : llvm::AtomicRMWInst::Min)
                : (IsXLHSInRHSPart ? llvm::AtomicRMWInst::UMax
                                   : llvm::AtomicRMWInst::UMin);
    break;
  case BO_Assign:
    RMWOp = llvm::AtomicRMWInst::Xchg;
    break;
  case BO_Mul:
  case BO_Div:
  case BO_Rem:
  case BO_Shl:
  case BO_Shr:
  case BO_LAnd:
  case BO_LOr:
    return std::make_pair(false, RValue::get(nullptr));
  case BO_PtrMemD:
  case BO_PtrMemI:
  case BO_LE:
  case BO_GE:
  case BO_EQ:
  case BO_NE:
  case BO_AddAssign:
  case BO_SubAssign:
  case BO_AndAssign:
  case BO_OrAssign:
  case BO_XorAssign:
  case BO_MulAssign:
  case BO_DivAssign:
  case BO_RemAssign:
  case BO_ShlAssign:
  case BO_ShrAssign:
  case BO_Comma:
    llvm_unreachable("Unsupported atomic update operation");
  }
  auto *UpdateVal = Update.getScalarVal();
  if (auto *IC = dyn_cast<llvm::ConstantInt>(UpdateVal)) {
    UpdateVal = CGF.Builder.CreateIntCast(
        IC, X.getAddress().getElementType(),
        X.getType()->hasSignedIntegerRepresentation());
  }
  auto *Res = CGF.Builder.CreateAtomicRMW(RMWOp, X.getPointer(), UpdateVal, AO);
  return std::make_pair(true, RValue::get(Res));
}

std::pair<bool, RValue> CodeGenFunction::EmitOMPAtomicSimpleUpdateExpr(
    LValue X, RValue E, BinaryOperatorKind BO, bool IsXLHSInRHSPart,
    llvm::AtomicOrdering AO, SourceLocation Loc,
    const llvm::function_ref<RValue(RValue)> &CommonGen) {
  // Update expressions are allowed to have the following forms:
  // x binop= expr; -> xrval + expr;
  // x++, ++x -> xrval + 1;
  // x--, --x -> xrval - 1;
  // x = x binop expr; -> xrval binop expr
  // x = expr Op x; - > expr binop xrval;
  auto Res = emitOMPAtomicRMW(*this, X, E, BO, AO, IsXLHSInRHSPart);
  if (!Res.first) {
    if (X.isGlobalReg()) {
      // Emit an update expression: 'xrval' binop 'expr' or 'expr' binop
      // 'xrval'.
      EmitStoreThroughLValue(CommonGen(EmitLoadOfLValue(X, Loc)), X);
    } else {
      // Perform compare-and-swap procedure.
      EmitAtomicUpdate(X, AO, CommonGen, X.getType().isVolatileQualified());
    }
  }
  return Res;
}

static void EmitOMPAtomicUpdateExpr(CodeGenFunction &CGF, bool IsSeqCst,
                                    const Expr *X, const Expr *E,
                                    const Expr *UE, bool IsXLHSInRHSPart,
                                    SourceLocation Loc) {
  assert(isa<BinaryOperator>(UE->IgnoreImpCasts()) &&
         "Update expr in 'atomic update' must be a binary operator.");
  auto *BOUE = cast<BinaryOperator>(UE->IgnoreImpCasts());
  // Update expressions are allowed to have the following forms:
  // x binop= expr; -> xrval + expr;
  // x++, ++x -> xrval + 1;
  // x--, --x -> xrval - 1;
  // x = x binop expr; -> xrval binop expr
  // x = expr Op x; - > expr binop xrval;
  assert(X->isLValue() && "X of 'omp atomic update' is not lvalue");
  LValue XLValue = CGF.EmitLValue(X);
  RValue ExprRValue = CGF.EmitAnyExpr(E);
  auto AO = IsSeqCst ? llvm::SequentiallyConsistent : llvm::Monotonic;
  auto *LHS = cast<OpaqueValueExpr>(BOUE->getLHS()->IgnoreImpCasts());
  auto *RHS = cast<OpaqueValueExpr>(BOUE->getRHS()->IgnoreImpCasts());
  auto *XRValExpr = IsXLHSInRHSPart ? LHS : RHS;
  auto *ERValExpr = IsXLHSInRHSPart ? RHS : LHS;
  auto Gen =
      [&CGF, UE, ExprRValue, XRValExpr, ERValExpr](RValue XRValue) -> RValue {
        CodeGenFunction::OpaqueValueMapping MapExpr(CGF, ERValExpr, ExprRValue);
        CodeGenFunction::OpaqueValueMapping MapX(CGF, XRValExpr, XRValue);
        return CGF.EmitAnyExpr(UE);
      };
  (void)CGF.EmitOMPAtomicSimpleUpdateExpr(
      XLValue, ExprRValue, BOUE->getOpcode(), IsXLHSInRHSPart, AO, Loc, Gen);
  // OpenMP, 2.12.6, atomic Construct
  // Any atomic construct with a seq_cst clause forces the atomically
  // performed operation to include an implicit flush operation without a
  // list.
  if (IsSeqCst)
    CGF.CGM.getOpenMPRuntime().emitFlush(CGF, llvm::None, Loc);
}

static RValue convertToType(CodeGenFunction &CGF, RValue Value,
                            QualType SourceType, QualType ResType,
                            SourceLocation Loc) {
  switch (CGF.getEvaluationKind(ResType)) {
  case TEK_Scalar:
    return RValue::get(
        convertToScalarValue(CGF, Value, SourceType, ResType, Loc));
  case TEK_Complex: {
    auto Res = convertToComplexValue(CGF, Value, SourceType, ResType, Loc);
    return RValue::getComplex(Res.first, Res.second);
  }
  case TEK_Aggregate:
    break;
  }
  llvm_unreachable("Must be a scalar or complex.");
}

static void EmitOMPAtomicCaptureExpr(CodeGenFunction &CGF, bool IsSeqCst,
                                     bool IsPostfixUpdate, const Expr *V,
                                     const Expr *X, const Expr *E,
                                     const Expr *UE, bool IsXLHSInRHSPart,
                                     SourceLocation Loc) {
  assert(X->isLValue() && "X of 'omp atomic capture' is not lvalue");
  assert(V->isLValue() && "V of 'omp atomic capture' is not lvalue");
  RValue NewVVal;
  LValue VLValue = CGF.EmitLValue(V);
  LValue XLValue = CGF.EmitLValue(X);
  RValue ExprRValue = CGF.EmitAnyExpr(E);
  auto AO = IsSeqCst ? llvm::SequentiallyConsistent : llvm::Monotonic;
  QualType NewVValType;
  if (UE) {
    // 'x' is updated with some additional value.
    assert(isa<BinaryOperator>(UE->IgnoreImpCasts()) &&
           "Update expr in 'atomic capture' must be a binary operator.");
    auto *BOUE = cast<BinaryOperator>(UE->IgnoreImpCasts());
    // Update expressions are allowed to have the following forms:
    // x binop= expr; -> xrval + expr;
    // x++, ++x -> xrval + 1;
    // x--, --x -> xrval - 1;
    // x = x binop expr; -> xrval binop expr
    // x = expr Op x; - > expr binop xrval;
    auto *LHS = cast<OpaqueValueExpr>(BOUE->getLHS()->IgnoreImpCasts());
    auto *RHS = cast<OpaqueValueExpr>(BOUE->getRHS()->IgnoreImpCasts());
    auto *XRValExpr = IsXLHSInRHSPart ? LHS : RHS;
    NewVValType = XRValExpr->getType();
    auto *ERValExpr = IsXLHSInRHSPart ? RHS : LHS;
    auto &&Gen = [&CGF, &NewVVal, UE, ExprRValue, XRValExpr, ERValExpr,
                  IsSeqCst, IsPostfixUpdate](RValue XRValue) -> RValue {
      CodeGenFunction::OpaqueValueMapping MapExpr(CGF, ERValExpr, ExprRValue);
      CodeGenFunction::OpaqueValueMapping MapX(CGF, XRValExpr, XRValue);
      RValue Res = CGF.EmitAnyExpr(UE);
      NewVVal = IsPostfixUpdate ? XRValue : Res;
      return Res;
    };
    auto Res = CGF.EmitOMPAtomicSimpleUpdateExpr(
        XLValue, ExprRValue, BOUE->getOpcode(), IsXLHSInRHSPart, AO, Loc, Gen);
    if (Res.first) {
      // 'atomicrmw' instruction was generated.
      if (IsPostfixUpdate) {
        // Use old value from 'atomicrmw'.
        NewVVal = Res.second;
      } else {
        // 'atomicrmw' does not provide new value, so evaluate it using old
        // value of 'x'.
        CodeGenFunction::OpaqueValueMapping MapExpr(CGF, ERValExpr, ExprRValue);
        CodeGenFunction::OpaqueValueMapping MapX(CGF, XRValExpr, Res.second);
        NewVVal = CGF.EmitAnyExpr(UE);
      }
    }
  } else {
    // 'x' is simply rewritten with some 'expr'.
    NewVValType = X->getType().getNonReferenceType();
    ExprRValue = convertToType(CGF, ExprRValue, E->getType(),
                               X->getType().getNonReferenceType(), Loc);
    auto &&Gen = [&CGF, &NewVVal, ExprRValue](RValue XRValue) -> RValue {
      NewVVal = XRValue;
      return ExprRValue;
    };
    // Try to perform atomicrmw xchg, otherwise simple exchange.
    auto Res = CGF.EmitOMPAtomicSimpleUpdateExpr(
        XLValue, ExprRValue, /*BO=*/BO_Assign, /*IsXLHSInRHSPart=*/false, AO,
        Loc, Gen);
    if (Res.first) {
      // 'atomicrmw' instruction was generated.
      NewVVal = IsPostfixUpdate ? Res.second : ExprRValue;
    }
  }
  // Emit post-update store to 'v' of old/new 'x' value.
  CGF.emitOMPSimpleStore(VLValue, NewVVal, NewVValType, Loc);
  // OpenMP, 2.12.6, atomic Construct
  // Any atomic construct with a seq_cst clause forces the atomically
  // performed operation to include an implicit flush operation without a
  // list.
  if (IsSeqCst)
    CGF.CGM.getOpenMPRuntime().emitFlush(CGF, llvm::None, Loc);
}

static void EmitOMPAtomicExpr(CodeGenFunction &CGF, OpenMPClauseKind Kind,
                              bool IsSeqCst, bool IsPostfixUpdate,
                              const Expr *X, const Expr *V, const Expr *E,
                              const Expr *UE, bool IsXLHSInRHSPart,
                              SourceLocation Loc) {
  switch (Kind) {
  case OMPC_read:
    EmitOMPAtomicReadExpr(CGF, IsSeqCst, X, V, Loc);
    break;
  case OMPC_write:
    EmitOMPAtomicWriteExpr(CGF, IsSeqCst, X, E, Loc);
    break;
  case OMPC_unknown:
  case OMPC_update:
    EmitOMPAtomicUpdateExpr(CGF, IsSeqCst, X, E, UE, IsXLHSInRHSPart, Loc);
    break;
  case OMPC_capture:
    EmitOMPAtomicCaptureExpr(CGF, IsSeqCst, IsPostfixUpdate, V, X, E, UE,
                             IsXLHSInRHSPart, Loc);
    break;
  case OMPC_if:
  case OMPC_final:
  case OMPC_num_threads:
  case OMPC_private:
  case OMPC_firstprivate:
  case OMPC_lastprivate:
  case OMPC_reduction:
  case OMPC_safelen:
  case OMPC_simdlen:
  case OMPC_collapse:
  case OMPC_default:
  case OMPC_seq_cst:
  case OMPC_shared:
  case OMPC_linear:
  case OMPC_aligned:
  case OMPC_copyin:
  case OMPC_copyprivate:
  case OMPC_flush:
  case OMPC_proc_bind:
  case OMPC_schedule:
  case OMPC_ordered:
  case OMPC_nowait:
  case OMPC_untied:
  case OMPC_threadprivate:
  case OMPC_depend:
  case OMPC_mergeable:
  case OMPC_device:
  case OMPC_threads:
  case OMPC_simd:
  case OMPC_map:
  case OMPC_num_teams:
  case OMPC_thread_limit:
  case OMPC_priority:
  case OMPC_grainsize:
  case OMPC_nogroup:
  case OMPC_num_tasks:
  case OMPC_hint:
    llvm_unreachable("Clause is not allowed in 'omp atomic'.");
  }
}

void CodeGenFunction::EmitOMPAtomicDirective(const OMPAtomicDirective &S) {
  bool IsSeqCst = S.getSingleClause<OMPSeqCstClause>();
  OpenMPClauseKind Kind = OMPC_unknown;
  for (auto *C : S.clauses()) {
    // Find first clause (skip seq_cst clause, if it is first).
    if (C->getClauseKind() != OMPC_seq_cst) {
      Kind = C->getClauseKind();
      break;
    }
  }

  const auto *CS =
      S.getAssociatedStmt()->IgnoreContainers(/*IgnoreCaptured=*/true);
  if (const auto *EWC = dyn_cast<ExprWithCleanups>(CS)) {
    enterFullExpression(EWC);
  }
  // Processing for statements under 'atomic capture'.
  if (const auto *Compound = dyn_cast<CompoundStmt>(CS)) {
    for (const auto *C : Compound->body()) {
      if (const auto *EWC = dyn_cast<ExprWithCleanups>(C)) {
        enterFullExpression(EWC);
      }
    }
  }

  LexicalScope Scope(*this, S.getSourceRange());
  auto &&CodeGen = [&S, Kind, IsSeqCst, CS](CodeGenFunction &CGF) {
    CGF.EmitStopPoint(CS);
    EmitOMPAtomicExpr(CGF, Kind, IsSeqCst, S.isPostfixUpdate(), S.getX(),
                      S.getV(), S.getExpr(), S.getUpdateExpr(),
                      S.isXLHSInRHSPart(), S.getLocStart());
  };
  CGM.getOpenMPRuntime().emitInlinedDirective(*this, OMPD_atomic, CodeGen);
}

void CodeGenFunction::EmitOMPTargetDirective(const OMPTargetDirective &S) {
  LexicalScope Scope(*this, S.getSourceRange());
  const CapturedStmt &CS = *cast<CapturedStmt>(S.getAssociatedStmt());

  llvm::SmallVector<llvm::Value *, 16> CapturedVars;
  GenerateOpenMPCapturedVars(CS, CapturedVars);

  llvm::Function *Fn = nullptr;
  llvm::Constant *FnID = nullptr;

  // Check if we have any if clause associated with the directive.
  const Expr *IfCond = nullptr;

  if (auto *C = S.getSingleClause<OMPIfClause>()) {
    IfCond = C->getCondition();
  }

  // Check if we have any device clause associated with the directive.
  const Expr *Device = nullptr;
  if (auto *C = S.getSingleClause<OMPDeviceClause>()) {
    Device = C->getDevice();
  }

  // Check if we have an if clause whose conditional always evaluates to false
  // or if we do not have any targets specified. If so the target region is not
  // an offload entry point.
  bool IsOffloadEntry = true;
  if (IfCond) {
    bool Val;
    if (ConstantFoldsToSimpleInteger(IfCond, Val) && !Val)
      IsOffloadEntry = false;
  }
  if (CGM.getLangOpts().OMPTargetTriples.empty())
    IsOffloadEntry = false;

  assert(CurFuncDecl && "No parent declaration for target region!");
  StringRef ParentName;
  // In case we have Ctors/Dtors we use the complete type variant to produce
  // the mangling of the device outlined kernel.
  if (auto *D = dyn_cast<CXXConstructorDecl>(CurFuncDecl))
    ParentName = CGM.getMangledName(GlobalDecl(D, Ctor_Complete));
  else if (auto *D = dyn_cast<CXXDestructorDecl>(CurFuncDecl))
    ParentName = CGM.getMangledName(GlobalDecl(D, Dtor_Complete));
  else
    ParentName =
        CGM.getMangledName(GlobalDecl(cast<FunctionDecl>(CurFuncDecl)));

  CGM.getOpenMPRuntime().emitTargetOutlinedFunction(S, ParentName, Fn, FnID,
                                                    IsOffloadEntry);

  CGM.getOpenMPRuntime().emitTargetCall(*this, S, Fn, FnID, IfCond, Device,
                                        CapturedVars);
}

void CodeGenFunction::EmitOMPTeamsDirective(const OMPTeamsDirective &) {
  llvm_unreachable("CodeGen for 'omp teams' is not supported yet.");
}

void CodeGenFunction::EmitOMPCancellationPointDirective(
    const OMPCancellationPointDirective &S) {
  CGM.getOpenMPRuntime().emitCancellationPointCall(*this, S.getLocStart(),
                                                   S.getCancelRegion());
}

void CodeGenFunction::EmitOMPCancelDirective(const OMPCancelDirective &S) {
  const Expr *IfCond = nullptr;
  for (const auto *C : S.getClausesOfKind<OMPIfClause>()) {
    if (C->getNameModifier() == OMPD_unknown ||
        C->getNameModifier() == OMPD_cancel) {
      IfCond = C->getCondition();
      break;
    }
  }
  CGM.getOpenMPRuntime().emitCancelCall(*this, S.getLocStart(), IfCond,
                                        S.getCancelRegion());
}

CodeGenFunction::JumpDest
CodeGenFunction::getOMPCancelDestination(OpenMPDirectiveKind Kind) {
  if (Kind == OMPD_parallel || Kind == OMPD_task)
    return ReturnBlock;
  assert(Kind == OMPD_for || Kind == OMPD_section || Kind == OMPD_sections ||
         Kind == OMPD_parallel_sections || Kind == OMPD_parallel_for);
  return BreakContinueStack.back().BreakBlock;
}

// Generate the instructions for '#pragma omp target data' directive.
void CodeGenFunction::EmitOMPTargetDataDirective(
    const OMPTargetDataDirective &S) {
  // emit the code inside the construct for now
  auto CS = cast<CapturedStmt>(S.getAssociatedStmt());
  CGM.getOpenMPRuntime().emitInlinedDirective(
      *this, OMPD_target_data,
      [&CS](CodeGenFunction &CGF) { CGF.EmitStmt(CS->getCapturedStmt()); });
}

void CodeGenFunction::EmitOMPTaskLoopDirective(const OMPTaskLoopDirective &S) {
  // emit the code inside the construct for now
  auto CS = cast<CapturedStmt>(S.getAssociatedStmt());
  CGM.getOpenMPRuntime().emitInlinedDirective(
      *this, OMPD_taskloop,
      [&CS](CodeGenFunction &CGF) { CGF.EmitStmt(CS->getCapturedStmt()); });
}

void CodeGenFunction::EmitOMPTaskLoopSimdDirective(
    const OMPTaskLoopSimdDirective &S) {
  // emit the code inside the construct for now
  auto CS = cast<CapturedStmt>(S.getAssociatedStmt());
  CGM.getOpenMPRuntime().emitInlinedDirective(
      *this, OMPD_taskloop_simd,
      [&CS](CodeGenFunction &CGF) { CGF.EmitStmt(CS->getCapturedStmt()); });
}