aboutsummaryrefslogtreecommitdiffstats
path: root/runtime/src/kmp_csupport.cpp
blob: 3afefed560fd669a950ea59eea290dc3d6306b75 (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
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
4071
4072
4073
4074
4075
4076
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
/*
 * kmp_csupport.cpp -- kfront linkage support for OpenMP.
 */

//===----------------------------------------------------------------------===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is dual licensed under the MIT and the University of Illinois Open
// Source Licenses. See LICENSE.txt for details.
//
//===----------------------------------------------------------------------===//

#define __KMP_IMP
#include "omp.h" /* extern "C" declarations of user-visible routines */
#include "kmp.h"
#include "kmp_error.h"
#include "kmp_i18n.h"
#include "kmp_itt.h"
#include "kmp_lock.h"
#include "kmp_stats.h"

#if OMPT_SUPPORT
#include "ompt-specific.h"
#endif

#define MAX_MESSAGE 512

// flags will be used in future, e.g. to implement openmp_strict library
// restrictions

/*!
 * @ingroup STARTUP_SHUTDOWN
 * @param loc   in   source location information
 * @param flags in   for future use (currently ignored)
 *
 * Initialize the runtime library. This call is optional; if it is not made then
 * it will be implicitly called by attempts to use other library functions.
 */
void __kmpc_begin(ident_t *loc, kmp_int32 flags) {
  // By default __kmpc_begin() is no-op.
  char *env;
  if ((env = getenv("KMP_INITIAL_THREAD_BIND")) != NULL &&
      __kmp_str_match_true(env)) {
    __kmp_middle_initialize();
    KC_TRACE(10, ("__kmpc_begin: middle initialization called\n"));
  } else if (__kmp_ignore_mppbeg() == FALSE) {
    // By default __kmp_ignore_mppbeg() returns TRUE.
    __kmp_internal_begin();
    KC_TRACE(10, ("__kmpc_begin: called\n"));
  }
}

/*!
 * @ingroup STARTUP_SHUTDOWN
 * @param loc source location information
 *
 * Shutdown the runtime library. This is also optional, and even if called will
 * not do anything unless the `KMP_IGNORE_MPPEND` environment variable is set to
 * zero.
 */
void __kmpc_end(ident_t *loc) {
  // By default, __kmp_ignore_mppend() returns TRUE which makes __kmpc_end()
  // call no-op. However, this can be overridden with KMP_IGNORE_MPPEND
  // environment variable. If KMP_IGNORE_MPPEND is 0, __kmp_ignore_mppend()
  // returns FALSE and __kmpc_end() will unregister this root (it can cause
  // library shut down).
  if (__kmp_ignore_mppend() == FALSE) {
    KC_TRACE(10, ("__kmpc_end: called\n"));
    KA_TRACE(30, ("__kmpc_end\n"));

    __kmp_internal_end_thread(-1);
  }
#if KMP_OS_WINDOWS && OMPT_SUPPORT
  // Normal exit process on Windows does not allow worker threads of the final
  // parallel region to finish reporting their events, so shutting down the
  // library here fixes the issue at least for the cases where __kmpc_end() is
  // placed properly.
  if (ompt_enabled.enabled)
    __kmp_internal_end_library(__kmp_gtid_get_specific());
#endif
}

/*!
@ingroup THREAD_STATES
@param loc Source location information.
@return The global thread index of the active thread.

This function can be called in any context.

If the runtime has ony been entered at the outermost level from a
single (necessarily non-OpenMP<sup>*</sup>) thread, then the thread number is
that which would be returned by omp_get_thread_num() in the outermost
active parallel construct. (Or zero if there is no active parallel
construct, since the master thread is necessarily thread zero).

If multiple non-OpenMP threads all enter an OpenMP construct then this
will be a unique thread identifier among all the threads created by
the OpenMP runtime (but the value cannote be defined in terms of
OpenMP thread ids returned by omp_get_thread_num()).
*/
kmp_int32 __kmpc_global_thread_num(ident_t *loc) {
  kmp_int32 gtid = __kmp_entry_gtid();

  KC_TRACE(10, ("__kmpc_global_thread_num: T#%d\n", gtid));

  return gtid;
}

/*!
@ingroup THREAD_STATES
@param loc Source location information.
@return The number of threads under control of the OpenMP<sup>*</sup> runtime

This function can be called in any context.
It returns the total number of threads under the control of the OpenMP runtime.
That is not a number that can be determined by any OpenMP standard calls, since
the library may be called from more than one non-OpenMP thread, and this
reflects the total over all such calls. Similarly the runtime maintains
underlying threads even when they are not active (since the cost of creating
and destroying OS threads is high), this call counts all such threads even if
they are not waiting for work.
*/
kmp_int32 __kmpc_global_num_threads(ident_t *loc) {
  KC_TRACE(10,
           ("__kmpc_global_num_threads: num_threads = %d\n", __kmp_all_nth));

  return TCR_4(__kmp_all_nth);
}

/*!
@ingroup THREAD_STATES
@param loc Source location information.
@return The thread number of the calling thread in the innermost active parallel
construct.
*/
kmp_int32 __kmpc_bound_thread_num(ident_t *loc) {
  KC_TRACE(10, ("__kmpc_bound_thread_num: called\n"));
  return __kmp_tid_from_gtid(__kmp_entry_gtid());
}

/*!
@ingroup THREAD_STATES
@param loc Source location information.
@return The number of threads in the innermost active parallel construct.
*/
kmp_int32 __kmpc_bound_num_threads(ident_t *loc) {
  KC_TRACE(10, ("__kmpc_bound_num_threads: called\n"));

  return __kmp_entry_thread()->th.th_team->t.t_nproc;
}

/*!
 * @ingroup DEPRECATED
 * @param loc location description
 *
 * This function need not be called. It always returns TRUE.
 */
kmp_int32 __kmpc_ok_to_fork(ident_t *loc) {
#ifndef KMP_DEBUG

  return TRUE;

#else

  const char *semi2;
  const char *semi3;
  int line_no;

  if (__kmp_par_range == 0) {
    return TRUE;
  }
  semi2 = loc->psource;
  if (semi2 == NULL) {
    return TRUE;
  }
  semi2 = strchr(semi2, ';');
  if (semi2 == NULL) {
    return TRUE;
  }
  semi2 = strchr(semi2 + 1, ';');
  if (semi2 == NULL) {
    return TRUE;
  }
  if (__kmp_par_range_filename[0]) {
    const char *name = semi2 - 1;
    while ((name > loc->psource) && (*name != '/') && (*name != ';')) {
      name--;
    }
    if ((*name == '/') || (*name == ';')) {
      name++;
    }
    if (strncmp(__kmp_par_range_filename, name, semi2 - name)) {
      return __kmp_par_range < 0;
    }
  }
  semi3 = strchr(semi2 + 1, ';');
  if (__kmp_par_range_routine[0]) {
    if ((semi3 != NULL) && (semi3 > semi2) &&
        (strncmp(__kmp_par_range_routine, semi2 + 1, semi3 - semi2 - 1))) {
      return __kmp_par_range < 0;
    }
  }
  if (KMP_SSCANF(semi3 + 1, "%d", &line_no) == 1) {
    if ((line_no >= __kmp_par_range_lb) && (line_no <= __kmp_par_range_ub)) {
      return __kmp_par_range > 0;
    }
    return __kmp_par_range < 0;
  }
  return TRUE;

#endif /* KMP_DEBUG */
}

/*!
@ingroup THREAD_STATES
@param loc Source location information.
@return 1 if this thread is executing inside an active parallel region, zero if
not.
*/
kmp_int32 __kmpc_in_parallel(ident_t *loc) {
  return __kmp_entry_thread()->th.th_root->r.r_active;
}

/*!
@ingroup PARALLEL
@param loc source location information
@param global_tid global thread number
@param num_threads number of threads requested for this parallel construct

Set the number of threads to be used by the next fork spawned by this thread.
This call is only required if the parallel construct has a `num_threads` clause.
*/
void __kmpc_push_num_threads(ident_t *loc, kmp_int32 global_tid,
                             kmp_int32 num_threads) {
  KA_TRACE(20, ("__kmpc_push_num_threads: enter T#%d num_threads=%d\n",
                global_tid, num_threads));

  __kmp_push_num_threads(loc, global_tid, num_threads);
}

void __kmpc_pop_num_threads(ident_t *loc, kmp_int32 global_tid) {
  KA_TRACE(20, ("__kmpc_pop_num_threads: enter\n"));

  /* the num_threads are automatically popped */
}

#if OMP_40_ENABLED

void __kmpc_push_proc_bind(ident_t *loc, kmp_int32 global_tid,
                           kmp_int32 proc_bind) {
  KA_TRACE(20, ("__kmpc_push_proc_bind: enter T#%d proc_bind=%d\n", global_tid,
                proc_bind));

  __kmp_push_proc_bind(loc, global_tid, (kmp_proc_bind_t)proc_bind);
}

#endif /* OMP_40_ENABLED */

/*!
@ingroup PARALLEL
@param loc  source location information
@param argc  total number of arguments in the ellipsis
@param microtask  pointer to callback routine consisting of outlined parallel
construct
@param ...  pointers to shared variables that aren't global

Do the actual fork and call the microtask in the relevant number of threads.
*/
void __kmpc_fork_call(ident_t *loc, kmp_int32 argc, kmpc_micro microtask, ...) {
  int gtid = __kmp_entry_gtid();

#if (KMP_STATS_ENABLED)
  // If we were in a serial region, then stop the serial timer, record
  // the event, and start parallel region timer
  stats_state_e previous_state = KMP_GET_THREAD_STATE();
  if (previous_state == stats_state_e::SERIAL_REGION) {
    KMP_EXCHANGE_PARTITIONED_TIMER(OMP_parallel_overhead);
  } else {
    KMP_PUSH_PARTITIONED_TIMER(OMP_parallel_overhead);
  }
  int inParallel = __kmpc_in_parallel(loc);
  if (inParallel) {
    KMP_COUNT_BLOCK(OMP_NESTED_PARALLEL);
  } else {
    KMP_COUNT_BLOCK(OMP_PARALLEL);
  }
#endif

  // maybe to save thr_state is enough here
  {
    va_list ap;
    va_start(ap, microtask);

#if OMPT_SUPPORT
    ompt_frame_t *ompt_frame;
    if (ompt_enabled.enabled) {
      kmp_info_t *master_th = __kmp_threads[gtid];
      kmp_team_t *parent_team = master_th->th.th_team;
      ompt_lw_taskteam_t *lwt = parent_team->t.ompt_serialized_team_info;
      if (lwt)
        ompt_frame = &(lwt->ompt_task_info.frame);
      else {
        int tid = __kmp_tid_from_gtid(gtid);
        ompt_frame = &(
            parent_team->t.t_implicit_task_taskdata[tid].ompt_task_info.frame);
      }
      ompt_frame->enter_frame.ptr = OMPT_GET_FRAME_ADDRESS(0);
      OMPT_STORE_RETURN_ADDRESS(gtid);
    }
#endif

#if INCLUDE_SSC_MARKS
    SSC_MARK_FORKING();
#endif
    __kmp_fork_call(loc, gtid, fork_context_intel, argc,
                    VOLATILE_CAST(microtask_t) microtask, // "wrapped" task
                    VOLATILE_CAST(launch_t) __kmp_invoke_task_func,
/* TODO: revert workaround for Intel(R) 64 tracker #96 */
#if (KMP_ARCH_X86_64 || KMP_ARCH_ARM || KMP_ARCH_AARCH64) && KMP_OS_LINUX
                    &ap
#else
                    ap
#endif
                    );
#if INCLUDE_SSC_MARKS
    SSC_MARK_JOINING();
#endif
    __kmp_join_call(loc, gtid
#if OMPT_SUPPORT
                    ,
                    fork_context_intel
#endif
                    );

    va_end(ap);
  }

#if KMP_STATS_ENABLED
  if (previous_state == stats_state_e::SERIAL_REGION) {
    KMP_EXCHANGE_PARTITIONED_TIMER(OMP_serial);
  } else {
    KMP_POP_PARTITIONED_TIMER();
  }
#endif // KMP_STATS_ENABLED
}

#if OMP_40_ENABLED
/*!
@ingroup PARALLEL
@param loc source location information
@param global_tid global thread number
@param num_teams number of teams requested for the teams construct
@param num_threads number of threads per team requested for the teams construct

Set the number of teams to be used by the teams construct.
This call is only required if the teams construct has a `num_teams` clause
or a `thread_limit` clause (or both).
*/
void __kmpc_push_num_teams(ident_t *loc, kmp_int32 global_tid,
                           kmp_int32 num_teams, kmp_int32 num_threads) {
  KA_TRACE(20,
           ("__kmpc_push_num_teams: enter T#%d num_teams=%d num_threads=%d\n",
            global_tid, num_teams, num_threads));

  __kmp_push_num_teams(loc, global_tid, num_teams, num_threads);
}

/*!
@ingroup PARALLEL
@param loc  source location information
@param argc  total number of arguments in the ellipsis
@param microtask  pointer to callback routine consisting of outlined teams
construct
@param ...  pointers to shared variables that aren't global

Do the actual fork and call the microtask in the relevant number of threads.
*/
void __kmpc_fork_teams(ident_t *loc, kmp_int32 argc, kmpc_micro microtask,
                       ...) {
  int gtid = __kmp_entry_gtid();
  kmp_info_t *this_thr = __kmp_threads[gtid];
  va_list ap;
  va_start(ap, microtask);

  KMP_COUNT_BLOCK(OMP_TEAMS);

  // remember teams entry point and nesting level
  this_thr->th.th_teams_microtask = microtask;
  this_thr->th.th_teams_level =
      this_thr->th.th_team->t.t_level; // AC: can be >0 on host

#if OMPT_SUPPORT
  kmp_team_t *parent_team = this_thr->th.th_team;
  int tid = __kmp_tid_from_gtid(gtid);
  if (ompt_enabled.enabled) {
    parent_team->t.t_implicit_task_taskdata[tid]
        .ompt_task_info.frame.enter_frame.ptr = OMPT_GET_FRAME_ADDRESS(0);
  }
  OMPT_STORE_RETURN_ADDRESS(gtid);
#endif

  // check if __kmpc_push_num_teams called, set default number of teams
  // otherwise
  if (this_thr->th.th_teams_size.nteams == 0) {
    __kmp_push_num_teams(loc, gtid, 0, 0);
  }
  KMP_DEBUG_ASSERT(this_thr->th.th_set_nproc >= 1);
  KMP_DEBUG_ASSERT(this_thr->th.th_teams_size.nteams >= 1);
  KMP_DEBUG_ASSERT(this_thr->th.th_teams_size.nth >= 1);

  __kmp_fork_call(loc, gtid, fork_context_intel, argc,
                  VOLATILE_CAST(microtask_t)
                      __kmp_teams_master, // "wrapped" task
                  VOLATILE_CAST(launch_t) __kmp_invoke_teams_master,
#if (KMP_ARCH_X86_64 || KMP_ARCH_ARM || KMP_ARCH_AARCH64) && KMP_OS_LINUX
                  &ap
#else
                  ap
#endif
                  );
  __kmp_join_call(loc, gtid
#if OMPT_SUPPORT
                  ,
                  fork_context_intel
#endif
                  );

  this_thr->th.th_teams_microtask = NULL;
  this_thr->th.th_teams_level = 0;
  *(kmp_int64 *)(&this_thr->th.th_teams_size) = 0L;
  va_end(ap);
}
#endif /* OMP_40_ENABLED */

// I don't think this function should ever have been exported.
// The __kmpc_ prefix was misapplied.  I'm fairly certain that no generated
// openmp code ever called it, but it's been exported from the RTL for so
// long that I'm afraid to remove the definition.
int __kmpc_invoke_task_func(int gtid) { return __kmp_invoke_task_func(gtid); }

/*!
@ingroup PARALLEL
@param loc  source location information
@param global_tid  global thread number

Enter a serialized parallel construct. This interface is used to handle a
conditional parallel region, like this,
@code
#pragma omp parallel if (condition)
@endcode
when the condition is false.
*/
void __kmpc_serialized_parallel(ident_t *loc, kmp_int32 global_tid) {
// The implementation is now in kmp_runtime.cpp so that it can share static
// functions with kmp_fork_call since the tasks to be done are similar in
// each case.
#if OMPT_SUPPORT
  OMPT_STORE_RETURN_ADDRESS(global_tid);
#endif
  __kmp_serialized_parallel(loc, global_tid);
}

/*!
@ingroup PARALLEL
@param loc  source location information
@param global_tid  global thread number

Leave a serialized parallel construct.
*/
void __kmpc_end_serialized_parallel(ident_t *loc, kmp_int32 global_tid) {
  kmp_internal_control_t *top;
  kmp_info_t *this_thr;
  kmp_team_t *serial_team;

  KC_TRACE(10,
           ("__kmpc_end_serialized_parallel: called by T#%d\n", global_tid));

  /* skip all this code for autopar serialized loops since it results in
     unacceptable overhead */
  if (loc != NULL && (loc->flags & KMP_IDENT_AUTOPAR))
    return;

  // Not autopar code
  if (!TCR_4(__kmp_init_parallel))
    __kmp_parallel_initialize();

  this_thr = __kmp_threads[global_tid];
  serial_team = this_thr->th.th_serial_team;

#if OMP_45_ENABLED
  kmp_task_team_t *task_team = this_thr->th.th_task_team;

  // we need to wait for the proxy tasks before finishing the thread
  if (task_team != NULL && task_team->tt.tt_found_proxy_tasks)
    __kmp_task_team_wait(this_thr, serial_team USE_ITT_BUILD_ARG(NULL));
#endif

  KMP_MB();
  KMP_DEBUG_ASSERT(serial_team);
  KMP_ASSERT(serial_team->t.t_serialized);
  KMP_DEBUG_ASSERT(this_thr->th.th_team == serial_team);
  KMP_DEBUG_ASSERT(serial_team != this_thr->th.th_root->r.r_root_team);
  KMP_DEBUG_ASSERT(serial_team->t.t_threads);
  KMP_DEBUG_ASSERT(serial_team->t.t_threads[0] == this_thr);

#if OMPT_SUPPORT
  if (ompt_enabled.enabled &&
      this_thr->th.ompt_thread_info.state != ompt_state_overhead) {
    OMPT_CUR_TASK_INFO(this_thr)->frame.exit_frame = ompt_data_none;
    if (ompt_enabled.ompt_callback_implicit_task) {
      ompt_callbacks.ompt_callback(ompt_callback_implicit_task)(
          ompt_scope_end, NULL, OMPT_CUR_TASK_DATA(this_thr), 1,
          OMPT_CUR_TASK_INFO(this_thr)->thread_num, ompt_task_implicit);
    }

    // reset clear the task id only after unlinking the task
    ompt_data_t *parent_task_data;
    __ompt_get_task_info_internal(1, NULL, &parent_task_data, NULL, NULL, NULL);

    if (ompt_enabled.ompt_callback_parallel_end) {
      ompt_callbacks.ompt_callback(ompt_callback_parallel_end)(
          &(serial_team->t.ompt_team_info.parallel_data), parent_task_data,
          ompt_parallel_invoker_program, OMPT_LOAD_RETURN_ADDRESS(global_tid));
    }
    __ompt_lw_taskteam_unlink(this_thr);
    this_thr->th.ompt_thread_info.state = ompt_state_overhead;
  }
#endif

  /* If necessary, pop the internal control stack values and replace the team
   * values */
  top = serial_team->t.t_control_stack_top;
  if (top && top->serial_nesting_level == serial_team->t.t_serialized) {
    copy_icvs(&serial_team->t.t_threads[0]->th.th_current_task->td_icvs, top);
    serial_team->t.t_control_stack_top = top->next;
    __kmp_free(top);
  }

  // if( serial_team -> t.t_serialized > 1 )
  serial_team->t.t_level--;

  /* pop dispatch buffers stack */
  KMP_DEBUG_ASSERT(serial_team->t.t_dispatch->th_disp_buffer);
  {
    dispatch_private_info_t *disp_buffer =
        serial_team->t.t_dispatch->th_disp_buffer;
    serial_team->t.t_dispatch->th_disp_buffer =
        serial_team->t.t_dispatch->th_disp_buffer->next;
    __kmp_free(disp_buffer);
  }
#if OMP_50_ENABLED
  this_thr->th.th_def_allocator = serial_team->t.t_def_allocator; // restore
#endif

  --serial_team->t.t_serialized;
  if (serial_team->t.t_serialized == 0) {

/* return to the parallel section */

#if KMP_ARCH_X86 || KMP_ARCH_X86_64
    if (__kmp_inherit_fp_control && serial_team->t.t_fp_control_saved) {
      __kmp_clear_x87_fpu_status_word();
      __kmp_load_x87_fpu_control_word(&serial_team->t.t_x87_fpu_control_word);
      __kmp_load_mxcsr(&serial_team->t.t_mxcsr);
    }
#endif /* KMP_ARCH_X86 || KMP_ARCH_X86_64 */

    this_thr->th.th_team = serial_team->t.t_parent;
    this_thr->th.th_info.ds.ds_tid = serial_team->t.t_master_tid;

    /* restore values cached in the thread */
    this_thr->th.th_team_nproc = serial_team->t.t_parent->t.t_nproc; /*  JPH */
    this_thr->th.th_team_master =
        serial_team->t.t_parent->t.t_threads[0]; /* JPH */
    this_thr->th.th_team_serialized = this_thr->th.th_team->t.t_serialized;

    /* TODO the below shouldn't need to be adjusted for serialized teams */
    this_thr->th.th_dispatch =
        &this_thr->th.th_team->t.t_dispatch[serial_team->t.t_master_tid];

    __kmp_pop_current_task_from_thread(this_thr);

    KMP_ASSERT(this_thr->th.th_current_task->td_flags.executing == 0);
    this_thr->th.th_current_task->td_flags.executing = 1;

    if (__kmp_tasking_mode != tskm_immediate_exec) {
      // Copy the task team from the new child / old parent team to the thread.
      this_thr->th.th_task_team =
          this_thr->th.th_team->t.t_task_team[this_thr->th.th_task_state];
      KA_TRACE(20,
               ("__kmpc_end_serialized_parallel: T#%d restoring task_team %p / "
                "team %p\n",
                global_tid, this_thr->th.th_task_team, this_thr->th.th_team));
    }
  } else {
    if (__kmp_tasking_mode != tskm_immediate_exec) {
      KA_TRACE(20, ("__kmpc_end_serialized_parallel: T#%d decreasing nesting "
                    "depth of serial team %p to %d\n",
                    global_tid, serial_team, serial_team->t.t_serialized));
    }
  }

  if (__kmp_env_consistency_check)
    __kmp_pop_parallel(global_tid, NULL);
#if OMPT_SUPPORT
  if (ompt_enabled.enabled)
    this_thr->th.ompt_thread_info.state =
        ((this_thr->th.th_team_serialized) ? ompt_state_work_serial
                                           : ompt_state_work_parallel);
#endif
}

/*!
@ingroup SYNCHRONIZATION
@param loc  source location information.

Execute <tt>flush</tt>. This is implemented as a full memory fence. (Though
depending on the memory ordering convention obeyed by the compiler
even that may not be necessary).
*/
void __kmpc_flush(ident_t *loc) {
  KC_TRACE(10, ("__kmpc_flush: called\n"));

  /* need explicit __mf() here since use volatile instead in library */
  KMP_MB(); /* Flush all pending memory write invalidates.  */

#if (KMP_ARCH_X86 || KMP_ARCH_X86_64)
#if KMP_MIC
// fence-style instructions do not exist, but lock; xaddl $0,(%rsp) can be used.
// We shouldn't need it, though, since the ABI rules require that
// * If the compiler generates NGO stores it also generates the fence
// * If users hand-code NGO stores they should insert the fence
// therefore no incomplete unordered stores should be visible.
#else
  // C74404
  // This is to address non-temporal store instructions (sfence needed).
  // The clflush instruction is addressed either (mfence needed).
  // Probably the non-temporal load monvtdqa instruction should also be
  // addressed.
  // mfence is a SSE2 instruction. Do not execute it if CPU is not SSE2.
  if (!__kmp_cpuinfo.initialized) {
    __kmp_query_cpuid(&__kmp_cpuinfo);
  }
  if (!__kmp_cpuinfo.sse2) {
    // CPU cannot execute SSE2 instructions.
  } else {
#if KMP_COMPILER_ICC
    _mm_mfence();
#elif KMP_COMPILER_MSVC
    MemoryBarrier();
#else
    __sync_synchronize();
#endif // KMP_COMPILER_ICC
  }
#endif // KMP_MIC
#elif (KMP_ARCH_ARM || KMP_ARCH_AARCH64 || KMP_ARCH_MIPS || KMP_ARCH_MIPS64)
// Nothing to see here move along
#elif KMP_ARCH_PPC64
// Nothing needed here (we have a real MB above).
#if KMP_OS_CNK
  // The flushing thread needs to yield here; this prevents a
  // busy-waiting thread from saturating the pipeline. flush is
  // often used in loops like this:
  // while (!flag) {
  //   #pragma omp flush(flag)
  // }
  // and adding the yield here is good for at least a 10x speedup
  // when running >2 threads per core (on the NAS LU benchmark).
  __kmp_yield(TRUE);
#endif
#else
#error Unknown or unsupported architecture
#endif

#if OMPT_SUPPORT && OMPT_OPTIONAL
  if (ompt_enabled.ompt_callback_flush) {
    ompt_callbacks.ompt_callback(ompt_callback_flush)(
        __ompt_get_thread_data_internal(), OMPT_GET_RETURN_ADDRESS(0));
  }
#endif
}

/* -------------------------------------------------------------------------- */
/*!
@ingroup SYNCHRONIZATION
@param loc source location information
@param global_tid thread id.

Execute a barrier.
*/
void __kmpc_barrier(ident_t *loc, kmp_int32 global_tid) {
  KMP_COUNT_BLOCK(OMP_BARRIER);
  KC_TRACE(10, ("__kmpc_barrier: called T#%d\n", global_tid));

  if (!TCR_4(__kmp_init_parallel))
    __kmp_parallel_initialize();

  if (__kmp_env_consistency_check) {
    if (loc == 0) {
      KMP_WARNING(ConstructIdentInvalid); // ??? What does it mean for the user?
    }

    __kmp_check_barrier(global_tid, ct_barrier, loc);
  }

#if OMPT_SUPPORT
  ompt_frame_t *ompt_frame;
  if (ompt_enabled.enabled) {
    __ompt_get_task_info_internal(0, NULL, NULL, &ompt_frame, NULL, NULL);
    if (ompt_frame->enter_frame.ptr == NULL)
      ompt_frame->enter_frame.ptr = OMPT_GET_FRAME_ADDRESS(0);
    OMPT_STORE_RETURN_ADDRESS(global_tid);
  }
#endif
  __kmp_threads[global_tid]->th.th_ident = loc;
  // TODO: explicit barrier_wait_id:
  //   this function is called when 'barrier' directive is present or
  //   implicit barrier at the end of a worksharing construct.
  // 1) better to add a per-thread barrier counter to a thread data structure
  // 2) set to 0 when a new team is created
  // 4) no sync is required

  __kmp_barrier(bs_plain_barrier, global_tid, FALSE, 0, NULL, NULL);
#if OMPT_SUPPORT && OMPT_OPTIONAL
  if (ompt_enabled.enabled) {
    ompt_frame->enter_frame = ompt_data_none;
  }
#endif
}

/* The BARRIER for a MASTER section is always explicit   */
/*!
@ingroup WORK_SHARING
@param loc  source location information.
@param global_tid  global thread number .
@return 1 if this thread should execute the <tt>master</tt> block, 0 otherwise.
*/
kmp_int32 __kmpc_master(ident_t *loc, kmp_int32 global_tid) {
  int status = 0;

  KC_TRACE(10, ("__kmpc_master: called T#%d\n", global_tid));

  if (!TCR_4(__kmp_init_parallel))
    __kmp_parallel_initialize();

  if (KMP_MASTER_GTID(global_tid)) {
    KMP_COUNT_BLOCK(OMP_MASTER);
    KMP_PUSH_PARTITIONED_TIMER(OMP_master);
    status = 1;
  }

#if OMPT_SUPPORT && OMPT_OPTIONAL
  if (status) {
    if (ompt_enabled.ompt_callback_master) {
      kmp_info_t *this_thr = __kmp_threads[global_tid];
      kmp_team_t *team = this_thr->th.th_team;

      int tid = __kmp_tid_from_gtid(global_tid);
      ompt_callbacks.ompt_callback(ompt_callback_master)(
          ompt_scope_begin, &(team->t.ompt_team_info.parallel_data),
          &(team->t.t_implicit_task_taskdata[tid].ompt_task_info.task_data),
          OMPT_GET_RETURN_ADDRESS(0));
    }
  }
#endif

  if (__kmp_env_consistency_check) {
#if KMP_USE_DYNAMIC_LOCK
    if (status)
      __kmp_push_sync(global_tid, ct_master, loc, NULL, 0);
    else
      __kmp_check_sync(global_tid, ct_master, loc, NULL, 0);
#else
    if (status)
      __kmp_push_sync(global_tid, ct_master, loc, NULL);
    else
      __kmp_check_sync(global_tid, ct_master, loc, NULL);
#endif
  }

  return status;
}

/*!
@ingroup WORK_SHARING
@param loc  source location information.
@param global_tid  global thread number .

Mark the end of a <tt>master</tt> region. This should only be called by the
thread that executes the <tt>master</tt> region.
*/
void __kmpc_end_master(ident_t *loc, kmp_int32 global_tid) {
  KC_TRACE(10, ("__kmpc_end_master: called T#%d\n", global_tid));

  KMP_DEBUG_ASSERT(KMP_MASTER_GTID(global_tid));
  KMP_POP_PARTITIONED_TIMER();

#if OMPT_SUPPORT && OMPT_OPTIONAL
  kmp_info_t *this_thr = __kmp_threads[global_tid];
  kmp_team_t *team = this_thr->th.th_team;
  if (ompt_enabled.ompt_callback_master) {
    int tid = __kmp_tid_from_gtid(global_tid);
    ompt_callbacks.ompt_callback(ompt_callback_master)(
        ompt_scope_end, &(team->t.ompt_team_info.parallel_data),
        &(team->t.t_implicit_task_taskdata[tid].ompt_task_info.task_data),
        OMPT_GET_RETURN_ADDRESS(0));
  }
#endif

  if (__kmp_env_consistency_check) {
    if (global_tid < 0)
      KMP_WARNING(ThreadIdentInvalid);

    if (KMP_MASTER_GTID(global_tid))
      __kmp_pop_sync(global_tid, ct_master, loc);
  }
}

/*!
@ingroup WORK_SHARING
@param loc  source location information.
@param gtid  global thread number.

Start execution of an <tt>ordered</tt> construct.
*/
void __kmpc_ordered(ident_t *loc, kmp_int32 gtid) {
  int cid = 0;
  kmp_info_t *th;
  KMP_DEBUG_ASSERT(__kmp_init_serial);

  KC_TRACE(10, ("__kmpc_ordered: called T#%d\n", gtid));

  if (!TCR_4(__kmp_init_parallel))
    __kmp_parallel_initialize();

#if USE_ITT_BUILD
  __kmp_itt_ordered_prep(gtid);
// TODO: ordered_wait_id
#endif /* USE_ITT_BUILD */

  th = __kmp_threads[gtid];

#if OMPT_SUPPORT && OMPT_OPTIONAL
  kmp_team_t *team;
  ompt_wait_id_t lck;
  void *codeptr_ra;
  if (ompt_enabled.enabled) {
    OMPT_STORE_RETURN_ADDRESS(gtid);
    team = __kmp_team_from_gtid(gtid);
    lck = (ompt_wait_id_t)(uintptr_t)&team->t.t_ordered.dt.t_value;
    /* OMPT state update */
    th->th.ompt_thread_info.wait_id = lck;
    th->th.ompt_thread_info.state = ompt_state_wait_ordered;

    /* OMPT event callback */
    codeptr_ra = OMPT_LOAD_RETURN_ADDRESS(gtid);
    if (ompt_enabled.ompt_callback_mutex_acquire) {
      ompt_callbacks.ompt_callback(ompt_callback_mutex_acquire)(
          ompt_mutex_ordered, omp_lock_hint_none, kmp_mutex_impl_spin, lck,
          codeptr_ra);
    }
  }
#endif

  if (th->th.th_dispatch->th_deo_fcn != 0)
    (*th->th.th_dispatch->th_deo_fcn)(&gtid, &cid, loc);
  else
    __kmp_parallel_deo(&gtid, &cid, loc);

#if OMPT_SUPPORT && OMPT_OPTIONAL
  if (ompt_enabled.enabled) {
    /* OMPT state update */
    th->th.ompt_thread_info.state = ompt_state_work_parallel;
    th->th.ompt_thread_info.wait_id = 0;

    /* OMPT event callback */
    if (ompt_enabled.ompt_callback_mutex_acquired) {
      ompt_callbacks.ompt_callback(ompt_callback_mutex_acquired)(
          ompt_mutex_ordered, (ompt_wait_id_t)(uintptr_t)lck, codeptr_ra);
    }
  }
#endif

#if USE_ITT_BUILD
  __kmp_itt_ordered_start(gtid);
#endif /* USE_ITT_BUILD */
}

/*!
@ingroup WORK_SHARING
@param loc  source location information.
@param gtid  global thread number.

End execution of an <tt>ordered</tt> construct.
*/
void __kmpc_end_ordered(ident_t *loc, kmp_int32 gtid) {
  int cid = 0;
  kmp_info_t *th;

  KC_TRACE(10, ("__kmpc_end_ordered: called T#%d\n", gtid));

#if USE_ITT_BUILD
  __kmp_itt_ordered_end(gtid);
// TODO: ordered_wait_id
#endif /* USE_ITT_BUILD */

  th = __kmp_threads[gtid];

  if (th->th.th_dispatch->th_dxo_fcn != 0)
    (*th->th.th_dispatch->th_dxo_fcn)(&gtid, &cid, loc);
  else
    __kmp_parallel_dxo(&gtid, &cid, loc);

#if OMPT_SUPPORT && OMPT_OPTIONAL
  OMPT_STORE_RETURN_ADDRESS(gtid);
  if (ompt_enabled.ompt_callback_mutex_released) {
    ompt_callbacks.ompt_callback(ompt_callback_mutex_released)(
        ompt_mutex_ordered,
        (ompt_wait_id_t)(uintptr_t)&__kmp_team_from_gtid(gtid)
            ->t.t_ordered.dt.t_value,
        OMPT_LOAD_RETURN_ADDRESS(gtid));
  }
#endif
}

#if KMP_USE_DYNAMIC_LOCK

static __forceinline void
__kmp_init_indirect_csptr(kmp_critical_name *crit, ident_t const *loc,
                          kmp_int32 gtid, kmp_indirect_locktag_t tag) {
  // Pointer to the allocated indirect lock is written to crit, while indexing
  // is ignored.
  void *idx;
  kmp_indirect_lock_t **lck;
  lck = (kmp_indirect_lock_t **)crit;
  kmp_indirect_lock_t *ilk = __kmp_allocate_indirect_lock(&idx, gtid, tag);
  KMP_I_LOCK_FUNC(ilk, init)(ilk->lock);
  KMP_SET_I_LOCK_LOCATION(ilk, loc);
  KMP_SET_I_LOCK_FLAGS(ilk, kmp_lf_critical_section);
  KA_TRACE(20,
           ("__kmp_init_indirect_csptr: initialized indirect lock #%d\n", tag));
#if USE_ITT_BUILD
  __kmp_itt_critical_creating(ilk->lock, loc);
#endif
  int status = KMP_COMPARE_AND_STORE_PTR(lck, nullptr, ilk);
  if (status == 0) {
#if USE_ITT_BUILD
    __kmp_itt_critical_destroyed(ilk->lock);
#endif
    // We don't really need to destroy the unclaimed lock here since it will be
    // cleaned up at program exit.
    // KMP_D_LOCK_FUNC(&idx, destroy)((kmp_dyna_lock_t *)&idx);
  }
  KMP_DEBUG_ASSERT(*lck != NULL);
}

// Fast-path acquire tas lock
#define KMP_ACQUIRE_TAS_LOCK(lock, gtid)                                       \
  {                                                                            \
    kmp_tas_lock_t *l = (kmp_tas_lock_t *)lock;                                \
    kmp_int32 tas_free = KMP_LOCK_FREE(tas);                                   \
    kmp_int32 tas_busy = KMP_LOCK_BUSY(gtid + 1, tas);                         \
    if (KMP_ATOMIC_LD_RLX(&l->lk.poll) != tas_free ||                          \
        !__kmp_atomic_compare_store_acq(&l->lk.poll, tas_free, tas_busy)) {    \
      kmp_uint32 spins;                                                        \
      KMP_FSYNC_PREPARE(l);                                                    \
      KMP_INIT_YIELD(spins);                                                   \
      if (TCR_4(__kmp_nth) >                                                   \
          (__kmp_avail_proc ? __kmp_avail_proc : __kmp_xproc)) {               \
        KMP_YIELD(TRUE);                                                       \
      } else {                                                                 \
        KMP_YIELD_SPIN(spins);                                                 \
      }                                                                        \
      kmp_backoff_t backoff = __kmp_spin_backoff_params;                       \
      while (                                                                  \
          KMP_ATOMIC_LD_RLX(&l->lk.poll) != tas_free ||                        \
          !__kmp_atomic_compare_store_acq(&l->lk.poll, tas_free, tas_busy)) {  \
        __kmp_spin_backoff(&backoff);                                          \
        if (TCR_4(__kmp_nth) >                                                 \
            (__kmp_avail_proc ? __kmp_avail_proc : __kmp_xproc)) {             \
          KMP_YIELD(TRUE);                                                     \
        } else {                                                               \
          KMP_YIELD_SPIN(spins);                                               \
        }                                                                      \
      }                                                                        \
    }                                                                          \
    KMP_FSYNC_ACQUIRED(l);                                                     \
  }

// Fast-path test tas lock
#define KMP_TEST_TAS_LOCK(lock, gtid, rc)                                      \
  {                                                                            \
    kmp_tas_lock_t *l = (kmp_tas_lock_t *)lock;                                \
    kmp_int32 tas_free = KMP_LOCK_FREE(tas);                                   \
    kmp_int32 tas_busy = KMP_LOCK_BUSY(gtid + 1, tas);                         \
    rc = KMP_ATOMIC_LD_RLX(&l->lk.poll) == tas_free &&                         \
         __kmp_atomic_compare_store_acq(&l->lk.poll, tas_free, tas_busy);      \
  }

// Fast-path release tas lock
#define KMP_RELEASE_TAS_LOCK(lock, gtid)                                       \
  { KMP_ATOMIC_ST_REL(&((kmp_tas_lock_t *)lock)->lk.poll, KMP_LOCK_FREE(tas)); }

#if KMP_USE_FUTEX

#include <sys/syscall.h>
#include <unistd.h>
#ifndef FUTEX_WAIT
#define FUTEX_WAIT 0
#endif
#ifndef FUTEX_WAKE
#define FUTEX_WAKE 1
#endif

// Fast-path acquire futex lock
#define KMP_ACQUIRE_FUTEX_LOCK(lock, gtid)                                     \
  {                                                                            \
    kmp_futex_lock_t *ftx = (kmp_futex_lock_t *)lock;                          \
    kmp_int32 gtid_code = (gtid + 1) << 1;                                     \
    KMP_MB();                                                                  \
    KMP_FSYNC_PREPARE(ftx);                                                    \
    kmp_int32 poll_val;                                                        \
    while ((poll_val = KMP_COMPARE_AND_STORE_RET32(                            \
                &(ftx->lk.poll), KMP_LOCK_FREE(futex),                         \
                KMP_LOCK_BUSY(gtid_code, futex))) != KMP_LOCK_FREE(futex)) {   \
      kmp_int32 cond = KMP_LOCK_STRIP(poll_val) & 1;                           \
      if (!cond) {                                                             \
        if (!KMP_COMPARE_AND_STORE_RET32(&(ftx->lk.poll), poll_val,            \
                                         poll_val |                            \
                                             KMP_LOCK_BUSY(1, futex))) {       \
          continue;                                                            \
        }                                                                      \
        poll_val |= KMP_LOCK_BUSY(1, futex);                                   \
      }                                                                        \
      kmp_int32 rc;                                                            \
      if ((rc = syscall(__NR_futex, &(ftx->lk.poll), FUTEX_WAIT, poll_val,     \
                        NULL, NULL, 0)) != 0) {                                \
        continue;                                                              \
      }                                                                        \
      gtid_code |= 1;                                                          \
    }                                                                          \
    KMP_FSYNC_ACQUIRED(ftx);                                                   \
  }

// Fast-path test futex lock
#define KMP_TEST_FUTEX_LOCK(lock, gtid, rc)                                    \
  {                                                                            \
    kmp_futex_lock_t *ftx = (kmp_futex_lock_t *)lock;                          \
    if (KMP_COMPARE_AND_STORE_ACQ32(&(ftx->lk.poll), KMP_LOCK_FREE(futex),     \
                                    KMP_LOCK_BUSY(gtid + 1 << 1, futex))) {    \
      KMP_FSYNC_ACQUIRED(ftx);                                                 \
      rc = TRUE;                                                               \
    } else {                                                                   \
      rc = FALSE;                                                              \
    }                                                                          \
  }

// Fast-path release futex lock
#define KMP_RELEASE_FUTEX_LOCK(lock, gtid)                                     \
  {                                                                            \
    kmp_futex_lock_t *ftx = (kmp_futex_lock_t *)lock;                          \
    KMP_MB();                                                                  \
    KMP_FSYNC_RELEASING(ftx);                                                  \
    kmp_int32 poll_val =                                                       \
        KMP_XCHG_FIXED32(&(ftx->lk.poll), KMP_LOCK_FREE(futex));               \
    if (KMP_LOCK_STRIP(poll_val) & 1) {                                        \
      syscall(__NR_futex, &(ftx->lk.poll), FUTEX_WAKE,                         \
              KMP_LOCK_BUSY(1, futex), NULL, NULL, 0);                         \
    }                                                                          \
    KMP_MB();                                                                  \
    KMP_YIELD(TCR_4(__kmp_nth) >                                               \
              (__kmp_avail_proc ? __kmp_avail_proc : __kmp_xproc));            \
  }

#endif // KMP_USE_FUTEX

#else // KMP_USE_DYNAMIC_LOCK

static kmp_user_lock_p __kmp_get_critical_section_ptr(kmp_critical_name *crit,
                                                      ident_t const *loc,
                                                      kmp_int32 gtid) {
  kmp_user_lock_p *lck_pp = (kmp_user_lock_p *)crit;

  // Because of the double-check, the following load doesn't need to be volatile
  kmp_user_lock_p lck = (kmp_user_lock_p)TCR_PTR(*lck_pp);

  if (lck == NULL) {
    void *idx;

    // Allocate & initialize the lock.
    // Remember alloc'ed locks in table in order to free them in __kmp_cleanup()
    lck = __kmp_user_lock_allocate(&idx, gtid, kmp_lf_critical_section);
    __kmp_init_user_lock_with_checks(lck);
    __kmp_set_user_lock_location(lck, loc);
#if USE_ITT_BUILD
    __kmp_itt_critical_creating(lck);
// __kmp_itt_critical_creating() should be called *before* the first usage
// of underlying lock. It is the only place where we can guarantee it. There
// are chances the lock will destroyed with no usage, but it is not a
// problem, because this is not real event seen by user but rather setting
// name for object (lock). See more details in kmp_itt.h.
#endif /* USE_ITT_BUILD */

    // Use a cmpxchg instruction to slam the start of the critical section with
    // the lock pointer.  If another thread beat us to it, deallocate the lock,
    // and use the lock that the other thread allocated.
    int status = KMP_COMPARE_AND_STORE_PTR(lck_pp, 0, lck);

    if (status == 0) {
// Deallocate the lock and reload the value.
#if USE_ITT_BUILD
      __kmp_itt_critical_destroyed(lck);
// Let ITT know the lock is destroyed and the same memory location may be reused
// for another purpose.
#endif /* USE_ITT_BUILD */
      __kmp_destroy_user_lock_with_checks(lck);
      __kmp_user_lock_free(&idx, gtid, lck);
      lck = (kmp_user_lock_p)TCR_PTR(*lck_pp);
      KMP_DEBUG_ASSERT(lck != NULL);
    }
  }
  return lck;
}

#endif // KMP_USE_DYNAMIC_LOCK

/*!
@ingroup WORK_SHARING
@param loc  source location information.
@param global_tid  global thread number .
@param crit identity of the critical section. This could be a pointer to a lock
associated with the critical section, or some other suitably unique value.

Enter code protected by a `critical` construct.
This function blocks until the executing thread can enter the critical section.
*/
void __kmpc_critical(ident_t *loc, kmp_int32 global_tid,
                     kmp_critical_name *crit) {
#if KMP_USE_DYNAMIC_LOCK
#if OMPT_SUPPORT && OMPT_OPTIONAL
  OMPT_STORE_RETURN_ADDRESS(global_tid);
#endif // OMPT_SUPPORT
  __kmpc_critical_with_hint(loc, global_tid, crit, omp_lock_hint_none);
#else
  KMP_COUNT_BLOCK(OMP_CRITICAL);
#if OMPT_SUPPORT && OMPT_OPTIONAL
  ompt_state_t prev_state = ompt_state_undefined;
  ompt_thread_info_t ti;
#endif
  kmp_user_lock_p lck;

  KC_TRACE(10, ("__kmpc_critical: called T#%d\n", global_tid));

  // TODO: add THR_OVHD_STATE

  KMP_PUSH_PARTITIONED_TIMER(OMP_critical_wait);
  KMP_CHECK_USER_LOCK_INIT();

  if ((__kmp_user_lock_kind == lk_tas) &&
      (sizeof(lck->tas.lk.poll) <= OMP_CRITICAL_SIZE)) {
    lck = (kmp_user_lock_p)crit;
  }
#if KMP_USE_FUTEX
  else if ((__kmp_user_lock_kind == lk_futex) &&
           (sizeof(lck->futex.lk.poll) <= OMP_CRITICAL_SIZE)) {
    lck = (kmp_user_lock_p)crit;
  }
#endif
  else { // ticket, queuing or drdpa
    lck = __kmp_get_critical_section_ptr(crit, loc, global_tid);
  }

  if (__kmp_env_consistency_check)
    __kmp_push_sync(global_tid, ct_critical, loc, lck);

// since the critical directive binds to all threads, not just the current
// team we have to check this even if we are in a serialized team.
// also, even if we are the uber thread, we still have to conduct the lock,
// as we have to contend with sibling threads.

#if USE_ITT_BUILD
  __kmp_itt_critical_acquiring(lck);
#endif /* USE_ITT_BUILD */
#if OMPT_SUPPORT && OMPT_OPTIONAL
  OMPT_STORE_RETURN_ADDRESS(gtid);
  void *codeptr_ra = NULL;
  if (ompt_enabled.enabled) {
    ti = __kmp_threads[global_tid]->th.ompt_thread_info;
    /* OMPT state update */
    prev_state = ti.state;
    ti.wait_id = (ompt_wait_id_t)(uintptr_t)lck;
    ti.state = ompt_state_wait_critical;

    /* OMPT event callback */
    codeptr_ra = OMPT_LOAD_RETURN_ADDRESS(gtid);
    if (ompt_enabled.ompt_callback_mutex_acquire) {
      ompt_callbacks.ompt_callback(ompt_callback_mutex_acquire)(
          ompt_mutex_critical, omp_lock_hint_none, __ompt_get_mutex_impl_type(),
          (ompt_wait_id_t)(uintptr_t)lck, codeptr_ra);
    }
  }
#endif
  // Value of 'crit' should be good for using as a critical_id of the critical
  // section directive.
  __kmp_acquire_user_lock_with_checks(lck, global_tid);

#if USE_ITT_BUILD
  __kmp_itt_critical_acquired(lck);
#endif /* USE_ITT_BUILD */
#if OMPT_SUPPORT && OMPT_OPTIONAL
  if (ompt_enabled.enabled) {
    /* OMPT state update */
    ti.state = prev_state;
    ti.wait_id = 0;

    /* OMPT event callback */
    if (ompt_enabled.ompt_callback_mutex_acquired) {
      ompt_callbacks.ompt_callback(ompt_callback_mutex_acquired)(
          ompt_mutex_critical, (ompt_wait_id_t)(uintptr_t)lck, codeptr_ra);
    }
  }
#endif
  KMP_POP_PARTITIONED_TIMER();

  KMP_PUSH_PARTITIONED_TIMER(OMP_critical);
  KA_TRACE(15, ("__kmpc_critical: done T#%d\n", global_tid));
#endif // KMP_USE_DYNAMIC_LOCK
}

#if KMP_USE_DYNAMIC_LOCK

// Converts the given hint to an internal lock implementation
static __forceinline kmp_dyna_lockseq_t __kmp_map_hint_to_lock(uintptr_t hint) {
#if KMP_USE_TSX
#define KMP_TSX_LOCK(seq) lockseq_##seq
#else
#define KMP_TSX_LOCK(seq) __kmp_user_lock_seq
#endif

#if KMP_ARCH_X86 || KMP_ARCH_X86_64
#define KMP_CPUINFO_RTM (__kmp_cpuinfo.rtm)
#else
#define KMP_CPUINFO_RTM 0
#endif

  // Hints that do not require further logic
  if (hint & kmp_lock_hint_hle)
    return KMP_TSX_LOCK(hle);
  if (hint & kmp_lock_hint_rtm)
    return KMP_CPUINFO_RTM ? KMP_TSX_LOCK(rtm) : __kmp_user_lock_seq;
  if (hint & kmp_lock_hint_adaptive)
    return KMP_CPUINFO_RTM ? KMP_TSX_LOCK(adaptive) : __kmp_user_lock_seq;

  // Rule out conflicting hints first by returning the default lock
  if ((hint & omp_lock_hint_contended) && (hint & omp_lock_hint_uncontended))
    return __kmp_user_lock_seq;
  if ((hint & omp_lock_hint_speculative) &&
      (hint & omp_lock_hint_nonspeculative))
    return __kmp_user_lock_seq;

  // Do not even consider speculation when it appears to be contended
  if (hint & omp_lock_hint_contended)
    return lockseq_queuing;

  // Uncontended lock without speculation
  if ((hint & omp_lock_hint_uncontended) && !(hint & omp_lock_hint_speculative))
    return lockseq_tas;

  // HLE lock for speculation
  if (hint & omp_lock_hint_speculative)
    return KMP_TSX_LOCK(hle);

  return __kmp_user_lock_seq;
}

#if OMPT_SUPPORT && OMPT_OPTIONAL
#if KMP_USE_DYNAMIC_LOCK
static kmp_mutex_impl_t
__ompt_get_mutex_impl_type(void *user_lock, kmp_indirect_lock_t *ilock = 0) {
  if (user_lock) {
    switch (KMP_EXTRACT_D_TAG(user_lock)) {
    case 0:
      break;
#if KMP_USE_FUTEX
    case locktag_futex:
      return kmp_mutex_impl_queuing;
#endif
    case locktag_tas:
      return kmp_mutex_impl_spin;
#if KMP_USE_TSX
    case locktag_hle:
      return kmp_mutex_impl_speculative;
#endif
    default:
      return kmp_mutex_impl_none;
    }
    ilock = KMP_LOOKUP_I_LOCK(user_lock);
  }
  KMP_ASSERT(ilock);
  switch (ilock->type) {
#if KMP_USE_TSX
  case locktag_adaptive:
  case locktag_rtm:
    return kmp_mutex_impl_speculative;
#endif
  case locktag_nested_tas:
    return kmp_mutex_impl_spin;
#if KMP_USE_FUTEX
  case locktag_nested_futex:
#endif
  case locktag_ticket:
  case locktag_queuing:
  case locktag_drdpa:
  case locktag_nested_ticket:
  case locktag_nested_queuing:
  case locktag_nested_drdpa:
    return kmp_mutex_impl_queuing;
  default:
    return kmp_mutex_impl_none;
  }
}
#else
// For locks without dynamic binding
static kmp_mutex_impl_t __ompt_get_mutex_impl_type() {
  switch (__kmp_user_lock_kind) {
  case lk_tas:
    return kmp_mutex_impl_spin;
#if KMP_USE_FUTEX
  case lk_futex:
#endif
  case lk_ticket:
  case lk_queuing:
  case lk_drdpa:
    return kmp_mutex_impl_queuing;
#if KMP_USE_TSX
  case lk_hle:
  case lk_rtm:
  case lk_adaptive:
    return kmp_mutex_impl_speculative;
#endif
  default:
    return kmp_mutex_impl_none;
  }
}
#endif // KMP_USE_DYNAMIC_LOCK
#endif // OMPT_SUPPORT && OMPT_OPTIONAL

/*!
@ingroup WORK_SHARING
@param loc  source location information.
@param global_tid  global thread number.
@param crit identity of the critical section. This could be a pointer to a lock
associated with the critical section, or some other suitably unique value.
@param hint the lock hint.

Enter code protected by a `critical` construct with a hint. The hint value is
used to suggest a lock implementation. This function blocks until the executing
thread can enter the critical section unless the hint suggests use of
speculative execution and the hardware supports it.
*/
void __kmpc_critical_with_hint(ident_t *loc, kmp_int32 global_tid,
                               kmp_critical_name *crit, uint32_t hint) {
  KMP_COUNT_BLOCK(OMP_CRITICAL);
  kmp_user_lock_p lck;
#if OMPT_SUPPORT && OMPT_OPTIONAL
  ompt_state_t prev_state = ompt_state_undefined;
  ompt_thread_info_t ti;
  // This is the case, if called from __kmpc_critical:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(global_tid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
#endif

  KC_TRACE(10, ("__kmpc_critical: called T#%d\n", global_tid));

  kmp_dyna_lock_t *lk = (kmp_dyna_lock_t *)crit;
  // Check if it is initialized.
  KMP_PUSH_PARTITIONED_TIMER(OMP_critical_wait);
  if (*lk == 0) {
    kmp_dyna_lockseq_t lckseq = __kmp_map_hint_to_lock(hint);
    if (KMP_IS_D_LOCK(lckseq)) {
      KMP_COMPARE_AND_STORE_ACQ32((volatile kmp_int32 *)crit, 0,
                                  KMP_GET_D_TAG(lckseq));
    } else {
      __kmp_init_indirect_csptr(crit, loc, global_tid, KMP_GET_I_TAG(lckseq));
    }
  }
  // Branch for accessing the actual lock object and set operation. This
  // branching is inevitable since this lock initialization does not follow the
  // normal dispatch path (lock table is not used).
  if (KMP_EXTRACT_D_TAG(lk) != 0) {
    lck = (kmp_user_lock_p)lk;
    if (__kmp_env_consistency_check) {
      __kmp_push_sync(global_tid, ct_critical, loc, lck,
                      __kmp_map_hint_to_lock(hint));
    }
#if USE_ITT_BUILD
    __kmp_itt_critical_acquiring(lck);
#endif
#if OMPT_SUPPORT && OMPT_OPTIONAL
    if (ompt_enabled.enabled) {
      ti = __kmp_threads[global_tid]->th.ompt_thread_info;
      /* OMPT state update */
      prev_state = ti.state;
      ti.wait_id = (ompt_wait_id_t)(uintptr_t)lck;
      ti.state = ompt_state_wait_critical;

      /* OMPT event callback */
      if (ompt_enabled.ompt_callback_mutex_acquire) {
        ompt_callbacks.ompt_callback(ompt_callback_mutex_acquire)(
            ompt_mutex_critical, (unsigned int)hint,
            __ompt_get_mutex_impl_type(crit), (ompt_wait_id_t)(uintptr_t)lck,
            codeptr);
      }
    }
#endif
#if KMP_USE_INLINED_TAS
    if (__kmp_user_lock_seq == lockseq_tas && !__kmp_env_consistency_check) {
      KMP_ACQUIRE_TAS_LOCK(lck, global_tid);
    } else
#elif KMP_USE_INLINED_FUTEX
    if (__kmp_user_lock_seq == lockseq_futex && !__kmp_env_consistency_check) {
      KMP_ACQUIRE_FUTEX_LOCK(lck, global_tid);
    } else
#endif
    {
      KMP_D_LOCK_FUNC(lk, set)(lk, global_tid);
    }
  } else {
    kmp_indirect_lock_t *ilk = *((kmp_indirect_lock_t **)lk);
    lck = ilk->lock;
    if (__kmp_env_consistency_check) {
      __kmp_push_sync(global_tid, ct_critical, loc, lck,
                      __kmp_map_hint_to_lock(hint));
    }
#if USE_ITT_BUILD
    __kmp_itt_critical_acquiring(lck);
#endif
#if OMPT_SUPPORT && OMPT_OPTIONAL
    if (ompt_enabled.enabled) {
      ti = __kmp_threads[global_tid]->th.ompt_thread_info;
      /* OMPT state update */
      prev_state = ti.state;
      ti.wait_id = (ompt_wait_id_t)(uintptr_t)lck;
      ti.state = ompt_state_wait_critical;

      /* OMPT event callback */
      if (ompt_enabled.ompt_callback_mutex_acquire) {
        ompt_callbacks.ompt_callback(ompt_callback_mutex_acquire)(
            ompt_mutex_critical, (unsigned int)hint,
            __ompt_get_mutex_impl_type(0, ilk), (ompt_wait_id_t)(uintptr_t)lck,
            codeptr);
      }
    }
#endif
    KMP_I_LOCK_FUNC(ilk, set)(lck, global_tid);
  }
  KMP_POP_PARTITIONED_TIMER();

#if USE_ITT_BUILD
  __kmp_itt_critical_acquired(lck);
#endif /* USE_ITT_BUILD */
#if OMPT_SUPPORT && OMPT_OPTIONAL
  if (ompt_enabled.enabled) {
    /* OMPT state update */
    ti.state = prev_state;
    ti.wait_id = 0;

    /* OMPT event callback */
    if (ompt_enabled.ompt_callback_mutex_acquired) {
      ompt_callbacks.ompt_callback(ompt_callback_mutex_acquired)(
          ompt_mutex_critical, (ompt_wait_id_t)(uintptr_t)lck, codeptr);
    }
  }
#endif

  KMP_PUSH_PARTITIONED_TIMER(OMP_critical);
  KA_TRACE(15, ("__kmpc_critical: done T#%d\n", global_tid));
} // __kmpc_critical_with_hint

#endif // KMP_USE_DYNAMIC_LOCK

/*!
@ingroup WORK_SHARING
@param loc  source location information.
@param global_tid  global thread number .
@param crit identity of the critical section. This could be a pointer to a lock
associated with the critical section, or some other suitably unique value.

Leave a critical section, releasing any lock that was held during its execution.
*/
void __kmpc_end_critical(ident_t *loc, kmp_int32 global_tid,
                         kmp_critical_name *crit) {
  kmp_user_lock_p lck;

  KC_TRACE(10, ("__kmpc_end_critical: called T#%d\n", global_tid));

#if KMP_USE_DYNAMIC_LOCK
  if (KMP_IS_D_LOCK(__kmp_user_lock_seq)) {
    lck = (kmp_user_lock_p)crit;
    KMP_ASSERT(lck != NULL);
    if (__kmp_env_consistency_check) {
      __kmp_pop_sync(global_tid, ct_critical, loc);
    }
#if USE_ITT_BUILD
    __kmp_itt_critical_releasing(lck);
#endif
#if KMP_USE_INLINED_TAS
    if (__kmp_user_lock_seq == lockseq_tas && !__kmp_env_consistency_check) {
      KMP_RELEASE_TAS_LOCK(lck, global_tid);
    } else
#elif KMP_USE_INLINED_FUTEX
    if (__kmp_user_lock_seq == lockseq_futex && !__kmp_env_consistency_check) {
      KMP_RELEASE_FUTEX_LOCK(lck, global_tid);
    } else
#endif
    {
      KMP_D_LOCK_FUNC(lck, unset)((kmp_dyna_lock_t *)lck, global_tid);
    }
  } else {
    kmp_indirect_lock_t *ilk =
        (kmp_indirect_lock_t *)TCR_PTR(*((kmp_indirect_lock_t **)crit));
    KMP_ASSERT(ilk != NULL);
    lck = ilk->lock;
    if (__kmp_env_consistency_check) {
      __kmp_pop_sync(global_tid, ct_critical, loc);
    }
#if USE_ITT_BUILD
    __kmp_itt_critical_releasing(lck);
#endif
    KMP_I_LOCK_FUNC(ilk, unset)(lck, global_tid);
  }

#else // KMP_USE_DYNAMIC_LOCK

  if ((__kmp_user_lock_kind == lk_tas) &&
      (sizeof(lck->tas.lk.poll) <= OMP_CRITICAL_SIZE)) {
    lck = (kmp_user_lock_p)crit;
  }
#if KMP_USE_FUTEX
  else if ((__kmp_user_lock_kind == lk_futex) &&
           (sizeof(lck->futex.lk.poll) <= OMP_CRITICAL_SIZE)) {
    lck = (kmp_user_lock_p)crit;
  }
#endif
  else { // ticket, queuing or drdpa
    lck = (kmp_user_lock_p)TCR_PTR(*((kmp_user_lock_p *)crit));
  }

  KMP_ASSERT(lck != NULL);

  if (__kmp_env_consistency_check)
    __kmp_pop_sync(global_tid, ct_critical, loc);

#if USE_ITT_BUILD
  __kmp_itt_critical_releasing(lck);
#endif /* USE_ITT_BUILD */
  // Value of 'crit' should be good for using as a critical_id of the critical
  // section directive.
  __kmp_release_user_lock_with_checks(lck, global_tid);

#endif // KMP_USE_DYNAMIC_LOCK

#if OMPT_SUPPORT && OMPT_OPTIONAL
  /* OMPT release event triggers after lock is released; place here to trigger
   * for all #if branches */
  OMPT_STORE_RETURN_ADDRESS(global_tid);
  if (ompt_enabled.ompt_callback_mutex_released) {
    ompt_callbacks.ompt_callback(ompt_callback_mutex_released)(
        ompt_mutex_critical, (ompt_wait_id_t)(uintptr_t)lck,
        OMPT_LOAD_RETURN_ADDRESS(0));
  }
#endif

  KMP_POP_PARTITIONED_TIMER();
  KA_TRACE(15, ("__kmpc_end_critical: done T#%d\n", global_tid));
}

/*!
@ingroup SYNCHRONIZATION
@param loc source location information
@param global_tid thread id.
@return one if the thread should execute the master block, zero otherwise

Start execution of a combined barrier and master. The barrier is executed inside
this function.
*/
kmp_int32 __kmpc_barrier_master(ident_t *loc, kmp_int32 global_tid) {
  int status;

  KC_TRACE(10, ("__kmpc_barrier_master: called T#%d\n", global_tid));

  if (!TCR_4(__kmp_init_parallel))
    __kmp_parallel_initialize();

  if (__kmp_env_consistency_check)
    __kmp_check_barrier(global_tid, ct_barrier, loc);

#if OMPT_SUPPORT
  ompt_frame_t *ompt_frame;
  if (ompt_enabled.enabled) {
    __ompt_get_task_info_internal(0, NULL, NULL, &ompt_frame, NULL, NULL);
    if (ompt_frame->enter_frame.ptr == NULL)
      ompt_frame->enter_frame.ptr = OMPT_GET_FRAME_ADDRESS(0);
    OMPT_STORE_RETURN_ADDRESS(global_tid);
  }
#endif
#if USE_ITT_NOTIFY
  __kmp_threads[global_tid]->th.th_ident = loc;
#endif
  status = __kmp_barrier(bs_plain_barrier, global_tid, TRUE, 0, NULL, NULL);
#if OMPT_SUPPORT && OMPT_OPTIONAL
  if (ompt_enabled.enabled) {
    ompt_frame->enter_frame = ompt_data_none;
  }
#endif

  return (status != 0) ? 0 : 1;
}

/*!
@ingroup SYNCHRONIZATION
@param loc source location information
@param global_tid thread id.

Complete the execution of a combined barrier and master. This function should
only be called at the completion of the <tt>master</tt> code. Other threads will
still be waiting at the barrier and this call releases them.
*/
void __kmpc_end_barrier_master(ident_t *loc, kmp_int32 global_tid) {
  KC_TRACE(10, ("__kmpc_end_barrier_master: called T#%d\n", global_tid));

  __kmp_end_split_barrier(bs_plain_barrier, global_tid);
}

/*!
@ingroup SYNCHRONIZATION
@param loc source location information
@param global_tid thread id.
@return one if the thread should execute the master block, zero otherwise

Start execution of a combined barrier and master(nowait) construct.
The barrier is executed inside this function.
There is no equivalent "end" function, since the
*/
kmp_int32 __kmpc_barrier_master_nowait(ident_t *loc, kmp_int32 global_tid) {
  kmp_int32 ret;

  KC_TRACE(10, ("__kmpc_barrier_master_nowait: called T#%d\n", global_tid));

  if (!TCR_4(__kmp_init_parallel))
    __kmp_parallel_initialize();

  if (__kmp_env_consistency_check) {
    if (loc == 0) {
      KMP_WARNING(ConstructIdentInvalid); // ??? What does it mean for the user?
    }
    __kmp_check_barrier(global_tid, ct_barrier, loc);
  }

#if OMPT_SUPPORT
  ompt_frame_t *ompt_frame;
  if (ompt_enabled.enabled) {
    __ompt_get_task_info_internal(0, NULL, NULL, &ompt_frame, NULL, NULL);
    if (ompt_frame->enter_frame.ptr == NULL)
      ompt_frame->enter_frame.ptr = OMPT_GET_FRAME_ADDRESS(0);
    OMPT_STORE_RETURN_ADDRESS(global_tid);
  }
#endif
#if USE_ITT_NOTIFY
  __kmp_threads[global_tid]->th.th_ident = loc;
#endif
  __kmp_barrier(bs_plain_barrier, global_tid, FALSE, 0, NULL, NULL);
#if OMPT_SUPPORT && OMPT_OPTIONAL
  if (ompt_enabled.enabled) {
    ompt_frame->enter_frame = ompt_data_none;
  }
#endif

  ret = __kmpc_master(loc, global_tid);

  if (__kmp_env_consistency_check) {
    /*  there's no __kmpc_end_master called; so the (stats) */
    /*  actions of __kmpc_end_master are done here          */

    if (global_tid < 0) {
      KMP_WARNING(ThreadIdentInvalid);
    }
    if (ret) {
      /* only one thread should do the pop since only */
      /* one did the push (see __kmpc_master())       */

      __kmp_pop_sync(global_tid, ct_master, loc);
    }
  }

  return (ret);
}

/* The BARRIER for a SINGLE process section is always explicit   */
/*!
@ingroup WORK_SHARING
@param loc  source location information
@param global_tid  global thread number
@return One if this thread should execute the single construct, zero otherwise.

Test whether to execute a <tt>single</tt> construct.
There are no implicit barriers in the two "single" calls, rather the compiler
should introduce an explicit barrier if it is required.
*/

kmp_int32 __kmpc_single(ident_t *loc, kmp_int32 global_tid) {
  kmp_int32 rc = __kmp_enter_single(global_tid, loc, TRUE);

  if (rc) {
    // We are going to execute the single statement, so we should count it.
    KMP_COUNT_BLOCK(OMP_SINGLE);
    KMP_PUSH_PARTITIONED_TIMER(OMP_single);
  }

#if OMPT_SUPPORT && OMPT_OPTIONAL
  kmp_info_t *this_thr = __kmp_threads[global_tid];
  kmp_team_t *team = this_thr->th.th_team;
  int tid = __kmp_tid_from_gtid(global_tid);

  if (ompt_enabled.enabled) {
    if (rc) {
      if (ompt_enabled.ompt_callback_work) {
        ompt_callbacks.ompt_callback(ompt_callback_work)(
            ompt_work_single_executor, ompt_scope_begin,
            &(team->t.ompt_team_info.parallel_data),
            &(team->t.t_implicit_task_taskdata[tid].ompt_task_info.task_data),
            1, OMPT_GET_RETURN_ADDRESS(0));
      }
    } else {
      if (ompt_enabled.ompt_callback_work) {
        ompt_callbacks.ompt_callback(ompt_callback_work)(
            ompt_work_single_other, ompt_scope_begin,
            &(team->t.ompt_team_info.parallel_data),
            &(team->t.t_implicit_task_taskdata[tid].ompt_task_info.task_data),
            1, OMPT_GET_RETURN_ADDRESS(0));
        ompt_callbacks.ompt_callback(ompt_callback_work)(
            ompt_work_single_other, ompt_scope_end,
            &(team->t.ompt_team_info.parallel_data),
            &(team->t.t_implicit_task_taskdata[tid].ompt_task_info.task_data),
            1, OMPT_GET_RETURN_ADDRESS(0));
      }
    }
  }
#endif

  return rc;
}

/*!
@ingroup WORK_SHARING
@param loc  source location information
@param global_tid  global thread number

Mark the end of a <tt>single</tt> construct.  This function should
only be called by the thread that executed the block of code protected
by the `single` construct.
*/
void __kmpc_end_single(ident_t *loc, kmp_int32 global_tid) {
  __kmp_exit_single(global_tid);
  KMP_POP_PARTITIONED_TIMER();

#if OMPT_SUPPORT && OMPT_OPTIONAL
  kmp_info_t *this_thr = __kmp_threads[global_tid];
  kmp_team_t *team = this_thr->th.th_team;
  int tid = __kmp_tid_from_gtid(global_tid);

  if (ompt_enabled.ompt_callback_work) {
    ompt_callbacks.ompt_callback(ompt_callback_work)(
        ompt_work_single_executor, ompt_scope_end,
        &(team->t.ompt_team_info.parallel_data),
        &(team->t.t_implicit_task_taskdata[tid].ompt_task_info.task_data), 1,
        OMPT_GET_RETURN_ADDRESS(0));
  }
#endif
}

/*!
@ingroup WORK_SHARING
@param loc Source location
@param global_tid Global thread id

Mark the end of a statically scheduled loop.
*/
void __kmpc_for_static_fini(ident_t *loc, kmp_int32 global_tid) {
  KMP_POP_PARTITIONED_TIMER();
  KE_TRACE(10, ("__kmpc_for_static_fini called T#%d\n", global_tid));

#if OMPT_SUPPORT && OMPT_OPTIONAL
  if (ompt_enabled.ompt_callback_work) {
    ompt_work_t ompt_work_type = ompt_work_loop;
    ompt_team_info_t *team_info = __ompt_get_teaminfo(0, NULL);
    ompt_task_info_t *task_info = __ompt_get_task_info_object(0);
    // Determine workshare type
    if (loc != NULL) {
      if ((loc->flags & KMP_IDENT_WORK_LOOP) != 0) {
        ompt_work_type = ompt_work_loop;
      } else if ((loc->flags & KMP_IDENT_WORK_SECTIONS) != 0) {
        ompt_work_type = ompt_work_sections;
      } else if ((loc->flags & KMP_IDENT_WORK_DISTRIBUTE) != 0) {
        ompt_work_type = ompt_work_distribute;
      } else {
        // use default set above.
        // a warning about this case is provided in __kmpc_for_static_init
      }
      KMP_DEBUG_ASSERT(ompt_work_type);
    }
    ompt_callbacks.ompt_callback(ompt_callback_work)(
        ompt_work_type, ompt_scope_end, &(team_info->parallel_data),
        &(task_info->task_data), 0, OMPT_GET_RETURN_ADDRESS(0));
  }
#endif
  if (__kmp_env_consistency_check)
    __kmp_pop_workshare(global_tid, ct_pdo, loc);
}

// User routines which take C-style arguments (call by value)
// different from the Fortran equivalent routines

void ompc_set_num_threads(int arg) {
  // !!!!! TODO: check the per-task binding
  __kmp_set_num_threads(arg, __kmp_entry_gtid());
}

void ompc_set_dynamic(int flag) {
  kmp_info_t *thread;

  /* For the thread-private implementation of the internal controls */
  thread = __kmp_entry_thread();

  __kmp_save_internal_controls(thread);

  set__dynamic(thread, flag ? TRUE : FALSE);
}

void ompc_set_nested(int flag) {
  kmp_info_t *thread;

  /* For the thread-private internal controls implementation */
  thread = __kmp_entry_thread();

  __kmp_save_internal_controls(thread);

  set__nested(thread, flag ? TRUE : FALSE);
}

void ompc_set_max_active_levels(int max_active_levels) {
  /* TO DO */
  /* we want per-task implementation of this internal control */

  /* For the per-thread internal controls implementation */
  __kmp_set_max_active_levels(__kmp_entry_gtid(), max_active_levels);
}

void ompc_set_schedule(omp_sched_t kind, int modifier) {
  // !!!!! TODO: check the per-task binding
  __kmp_set_schedule(__kmp_entry_gtid(), (kmp_sched_t)kind, modifier);
}

int ompc_get_ancestor_thread_num(int level) {
  return __kmp_get_ancestor_thread_num(__kmp_entry_gtid(), level);
}

int ompc_get_team_size(int level) {
  return __kmp_get_team_size(__kmp_entry_gtid(), level);
}

#if OMP_50_ENABLED
/* OpenMP 5.0 Affinity Format API */

void ompc_set_affinity_format(char const *format) {
  if (!__kmp_init_serial) {
    __kmp_serial_initialize();
  }
  __kmp_strncpy_truncate(__kmp_affinity_format, KMP_AFFINITY_FORMAT_SIZE,
                         format, KMP_STRLEN(format) + 1);
}

size_t ompc_get_affinity_format(char *buffer, size_t size) {
  size_t format_size;
  if (!__kmp_init_serial) {
    __kmp_serial_initialize();
  }
  format_size = KMP_STRLEN(__kmp_affinity_format);
  if (buffer && size) {
    __kmp_strncpy_truncate(buffer, size, __kmp_affinity_format,
                           format_size + 1);
  }
  return format_size;
}

void ompc_display_affinity(char const *format) {
  int gtid;
  if (!TCR_4(__kmp_init_middle)) {
    __kmp_middle_initialize();
  }
  gtid = __kmp_get_gtid();
  __kmp_aux_display_affinity(gtid, format);
}

size_t ompc_capture_affinity(char *buffer, size_t buf_size,
                             char const *format) {
  int gtid;
  size_t num_required;
  kmp_str_buf_t capture_buf;
  if (!TCR_4(__kmp_init_middle)) {
    __kmp_middle_initialize();
  }
  gtid = __kmp_get_gtid();
  __kmp_str_buf_init(&capture_buf);
  num_required = __kmp_aux_capture_affinity(gtid, format, &capture_buf);
  if (buffer && buf_size) {
    __kmp_strncpy_truncate(buffer, buf_size, capture_buf.str,
                           capture_buf.used + 1);
  }
  __kmp_str_buf_free(&capture_buf);
  return num_required;
}
#endif /* OMP_50_ENABLED */

void kmpc_set_stacksize(int arg) {
  // __kmp_aux_set_stacksize initializes the library if needed
  __kmp_aux_set_stacksize(arg);
}

void kmpc_set_stacksize_s(size_t arg) {
  // __kmp_aux_set_stacksize initializes the library if needed
  __kmp_aux_set_stacksize(arg);
}

void kmpc_set_blocktime(int arg) {
  int gtid, tid;
  kmp_info_t *thread;

  gtid = __kmp_entry_gtid();
  tid = __kmp_tid_from_gtid(gtid);
  thread = __kmp_thread_from_gtid(gtid);

  __kmp_aux_set_blocktime(arg, thread, tid);
}

void kmpc_set_library(int arg) {
  // __kmp_user_set_library initializes the library if needed
  __kmp_user_set_library((enum library_type)arg);
}

void kmpc_set_defaults(char const *str) {
  // __kmp_aux_set_defaults initializes the library if needed
  __kmp_aux_set_defaults(str, KMP_STRLEN(str));
}

void kmpc_set_disp_num_buffers(int arg) {
  // ignore after initialization because some teams have already
  // allocated dispatch buffers
  if (__kmp_init_serial == 0 && arg > 0)
    __kmp_dispatch_num_buffers = arg;
}

int kmpc_set_affinity_mask_proc(int proc, void **mask) {
#if defined(KMP_STUB) || !KMP_AFFINITY_SUPPORTED
  return -1;
#else
  if (!TCR_4(__kmp_init_middle)) {
    __kmp_middle_initialize();
  }
  return __kmp_aux_set_affinity_mask_proc(proc, mask);
#endif
}

int kmpc_unset_affinity_mask_proc(int proc, void **mask) {
#if defined(KMP_STUB) || !KMP_AFFINITY_SUPPORTED
  return -1;
#else
  if (!TCR_4(__kmp_init_middle)) {
    __kmp_middle_initialize();
  }
  return __kmp_aux_unset_affinity_mask_proc(proc, mask);
#endif
}

int kmpc_get_affinity_mask_proc(int proc, void **mask) {
#if defined(KMP_STUB) || !KMP_AFFINITY_SUPPORTED
  return -1;
#else
  if (!TCR_4(__kmp_init_middle)) {
    __kmp_middle_initialize();
  }
  return __kmp_aux_get_affinity_mask_proc(proc, mask);
#endif
}

/* -------------------------------------------------------------------------- */
/*!
@ingroup THREADPRIVATE
@param loc       source location information
@param gtid      global thread number
@param cpy_size  size of the cpy_data buffer
@param cpy_data  pointer to data to be copied
@param cpy_func  helper function to call for copying data
@param didit     flag variable: 1=single thread; 0=not single thread

__kmpc_copyprivate implements the interface for the private data broadcast
needed for the copyprivate clause associated with a single region in an
OpenMP<sup>*</sup> program (both C and Fortran).
All threads participating in the parallel region call this routine.
One of the threads (called the single thread) should have the <tt>didit</tt>
variable set to 1 and all other threads should have that variable set to 0.
All threads pass a pointer to a data buffer (cpy_data) that they have built.

The OpenMP specification forbids the use of nowait on the single region when a
copyprivate clause is present. However, @ref __kmpc_copyprivate implements a
barrier internally to avoid race conditions, so the code generation for the
single region should avoid generating a barrier after the call to @ref
__kmpc_copyprivate.

The <tt>gtid</tt> parameter is the global thread id for the current thread.
The <tt>loc</tt> parameter is a pointer to source location information.

Internal implementation: The single thread will first copy its descriptor
address (cpy_data) to a team-private location, then the other threads will each
call the function pointed to by the parameter cpy_func, which carries out the
copy by copying the data using the cpy_data buffer.

The cpy_func routine used for the copy and the contents of the data area defined
by cpy_data and cpy_size may be built in any fashion that will allow the copy
to be done. For instance, the cpy_data buffer can hold the actual data to be
copied or it may hold a list of pointers to the data. The cpy_func routine must
interpret the cpy_data buffer appropriately.

The interface to cpy_func is as follows:
@code
void cpy_func( void *destination, void *source )
@endcode
where void *destination is the cpy_data pointer for the thread being copied to
and void *source is the cpy_data pointer for the thread being copied from.
*/
void __kmpc_copyprivate(ident_t *loc, kmp_int32 gtid, size_t cpy_size,
                        void *cpy_data, void (*cpy_func)(void *, void *),
                        kmp_int32 didit) {
  void **data_ptr;

  KC_TRACE(10, ("__kmpc_copyprivate: called T#%d\n", gtid));

  KMP_MB();

  data_ptr = &__kmp_team_from_gtid(gtid)->t.t_copypriv_data;

  if (__kmp_env_consistency_check) {
    if (loc == 0) {
      KMP_WARNING(ConstructIdentInvalid);
    }
  }

  // ToDo: Optimize the following two barriers into some kind of split barrier

  if (didit)
    *data_ptr = cpy_data;

#if OMPT_SUPPORT
  ompt_frame_t *ompt_frame;
  if (ompt_enabled.enabled) {
    __ompt_get_task_info_internal(0, NULL, NULL, &ompt_frame, NULL, NULL);
    if (ompt_frame->enter_frame.ptr == NULL)
      ompt_frame->enter_frame.ptr = OMPT_GET_FRAME_ADDRESS(0);
    OMPT_STORE_RETURN_ADDRESS(gtid);
  }
#endif
/* This barrier is not a barrier region boundary */
#if USE_ITT_NOTIFY
  __kmp_threads[gtid]->th.th_ident = loc;
#endif
  __kmp_barrier(bs_plain_barrier, gtid, FALSE, 0, NULL, NULL);

  if (!didit)
    (*cpy_func)(cpy_data, *data_ptr);

// Consider next barrier a user-visible barrier for barrier region boundaries
// Nesting checks are already handled by the single construct checks

#if OMPT_SUPPORT
  if (ompt_enabled.enabled) {
    OMPT_STORE_RETURN_ADDRESS(gtid);
  }
#endif
#if USE_ITT_NOTIFY
  __kmp_threads[gtid]->th.th_ident = loc; // TODO: check if it is needed (e.g.
// tasks can overwrite the location)
#endif
  __kmp_barrier(bs_plain_barrier, gtid, FALSE, 0, NULL, NULL);
#if OMPT_SUPPORT && OMPT_OPTIONAL
  if (ompt_enabled.enabled) {
    ompt_frame->enter_frame = ompt_data_none;
  }
#endif
}

/* -------------------------------------------------------------------------- */

#define INIT_LOCK __kmp_init_user_lock_with_checks
#define INIT_NESTED_LOCK __kmp_init_nested_user_lock_with_checks
#define ACQUIRE_LOCK __kmp_acquire_user_lock_with_checks
#define ACQUIRE_LOCK_TIMED __kmp_acquire_user_lock_with_checks_timed
#define ACQUIRE_NESTED_LOCK __kmp_acquire_nested_user_lock_with_checks
#define ACQUIRE_NESTED_LOCK_TIMED                                              \
  __kmp_acquire_nested_user_lock_with_checks_timed
#define RELEASE_LOCK __kmp_release_user_lock_with_checks
#define RELEASE_NESTED_LOCK __kmp_release_nested_user_lock_with_checks
#define TEST_LOCK __kmp_test_user_lock_with_checks
#define TEST_NESTED_LOCK __kmp_test_nested_user_lock_with_checks
#define DESTROY_LOCK __kmp_destroy_user_lock_with_checks
#define DESTROY_NESTED_LOCK __kmp_destroy_nested_user_lock_with_checks

// TODO: Make check abort messages use location info & pass it into
// with_checks routines

#if KMP_USE_DYNAMIC_LOCK

// internal lock initializer
static __forceinline void __kmp_init_lock_with_hint(ident_t *loc, void **lock,
                                                    kmp_dyna_lockseq_t seq) {
  if (KMP_IS_D_LOCK(seq)) {
    KMP_INIT_D_LOCK(lock, seq);
#if USE_ITT_BUILD
    __kmp_itt_lock_creating((kmp_user_lock_p)lock, NULL);
#endif
  } else {
    KMP_INIT_I_LOCK(lock, seq);
#if USE_ITT_BUILD
    kmp_indirect_lock_t *ilk = KMP_LOOKUP_I_LOCK(lock);
    __kmp_itt_lock_creating(ilk->lock, loc);
#endif
  }
}

// internal nest lock initializer
static __forceinline void
__kmp_init_nest_lock_with_hint(ident_t *loc, void **lock,
                               kmp_dyna_lockseq_t seq) {
#if KMP_USE_TSX
  // Don't have nested lock implementation for speculative locks
  if (seq == lockseq_hle || seq == lockseq_rtm || seq == lockseq_adaptive)
    seq = __kmp_user_lock_seq;
#endif
  switch (seq) {
  case lockseq_tas:
    seq = lockseq_nested_tas;
    break;
#if KMP_USE_FUTEX
  case lockseq_futex:
    seq = lockseq_nested_futex;
    break;
#endif
  case lockseq_ticket:
    seq = lockseq_nested_ticket;
    break;
  case lockseq_queuing:
    seq = lockseq_nested_queuing;
    break;
  case lockseq_drdpa:
    seq = lockseq_nested_drdpa;
    break;
  default:
    seq = lockseq_nested_queuing;
  }
  KMP_INIT_I_LOCK(lock, seq);
#if USE_ITT_BUILD
  kmp_indirect_lock_t *ilk = KMP_LOOKUP_I_LOCK(lock);
  __kmp_itt_lock_creating(ilk->lock, loc);
#endif
}

/* initialize the lock with a hint */
void __kmpc_init_lock_with_hint(ident_t *loc, kmp_int32 gtid, void **user_lock,
                                uintptr_t hint) {
  KMP_DEBUG_ASSERT(__kmp_init_serial);
  if (__kmp_env_consistency_check && user_lock == NULL) {
    KMP_FATAL(LockIsUninitialized, "omp_init_lock_with_hint");
  }

  __kmp_init_lock_with_hint(loc, user_lock, __kmp_map_hint_to_lock(hint));

#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.ompt_callback_lock_init) {
    ompt_callbacks.ompt_callback(ompt_callback_lock_init)(
        ompt_mutex_lock, (omp_lock_hint_t)hint,
        __ompt_get_mutex_impl_type(user_lock),
        (ompt_wait_id_t)(uintptr_t)user_lock, codeptr);
  }
#endif
}

/* initialize the lock with a hint */
void __kmpc_init_nest_lock_with_hint(ident_t *loc, kmp_int32 gtid,
                                     void **user_lock, uintptr_t hint) {
  KMP_DEBUG_ASSERT(__kmp_init_serial);
  if (__kmp_env_consistency_check && user_lock == NULL) {
    KMP_FATAL(LockIsUninitialized, "omp_init_nest_lock_with_hint");
  }

  __kmp_init_nest_lock_with_hint(loc, user_lock, __kmp_map_hint_to_lock(hint));

#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.ompt_callback_lock_init) {
    ompt_callbacks.ompt_callback(ompt_callback_lock_init)(
        ompt_mutex_nest_lock, (omp_lock_hint_t)hint,
        __ompt_get_mutex_impl_type(user_lock),
        (ompt_wait_id_t)(uintptr_t)user_lock, codeptr);
  }
#endif
}

#endif // KMP_USE_DYNAMIC_LOCK

/* initialize the lock */
void __kmpc_init_lock(ident_t *loc, kmp_int32 gtid, void **user_lock) {
#if KMP_USE_DYNAMIC_LOCK

  KMP_DEBUG_ASSERT(__kmp_init_serial);
  if (__kmp_env_consistency_check && user_lock == NULL) {
    KMP_FATAL(LockIsUninitialized, "omp_init_lock");
  }
  __kmp_init_lock_with_hint(loc, user_lock, __kmp_user_lock_seq);

#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.ompt_callback_lock_init) {
    ompt_callbacks.ompt_callback(ompt_callback_lock_init)(
        ompt_mutex_lock, omp_lock_hint_none,
        __ompt_get_mutex_impl_type(user_lock),
        (ompt_wait_id_t)(uintptr_t)user_lock, codeptr);
  }
#endif

#else // KMP_USE_DYNAMIC_LOCK

  static char const *const func = "omp_init_lock";
  kmp_user_lock_p lck;
  KMP_DEBUG_ASSERT(__kmp_init_serial);

  if (__kmp_env_consistency_check) {
    if (user_lock == NULL) {
      KMP_FATAL(LockIsUninitialized, func);
    }
  }

  KMP_CHECK_USER_LOCK_INIT();

  if ((__kmp_user_lock_kind == lk_tas) &&
      (sizeof(lck->tas.lk.poll) <= OMP_LOCK_T_SIZE)) {
    lck = (kmp_user_lock_p)user_lock;
  }
#if KMP_USE_FUTEX
  else if ((__kmp_user_lock_kind == lk_futex) &&
           (sizeof(lck->futex.lk.poll) <= OMP_LOCK_T_SIZE)) {
    lck = (kmp_user_lock_p)user_lock;
  }
#endif
  else {
    lck = __kmp_user_lock_allocate(user_lock, gtid, 0);
  }
  INIT_LOCK(lck);
  __kmp_set_user_lock_location(lck, loc);

#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.ompt_callback_lock_init) {
    ompt_callbacks.ompt_callback(ompt_callback_lock_init)(
        ompt_mutex_lock, omp_lock_hint_none, __ompt_get_mutex_impl_type(),
        (ompt_wait_id_t)(uintptr_t)user_lock, codeptr);
  }
#endif

#if USE_ITT_BUILD
  __kmp_itt_lock_creating(lck);
#endif /* USE_ITT_BUILD */

#endif // KMP_USE_DYNAMIC_LOCK
} // __kmpc_init_lock

/* initialize the lock */
void __kmpc_init_nest_lock(ident_t *loc, kmp_int32 gtid, void **user_lock) {
#if KMP_USE_DYNAMIC_LOCK

  KMP_DEBUG_ASSERT(__kmp_init_serial);
  if (__kmp_env_consistency_check && user_lock == NULL) {
    KMP_FATAL(LockIsUninitialized, "omp_init_nest_lock");
  }
  __kmp_init_nest_lock_with_hint(loc, user_lock, __kmp_user_lock_seq);

#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.ompt_callback_lock_init) {
    ompt_callbacks.ompt_callback(ompt_callback_lock_init)(
        ompt_mutex_nest_lock, omp_lock_hint_none,
        __ompt_get_mutex_impl_type(user_lock),
        (ompt_wait_id_t)(uintptr_t)user_lock, codeptr);
  }
#endif

#else // KMP_USE_DYNAMIC_LOCK

  static char const *const func = "omp_init_nest_lock";
  kmp_user_lock_p lck;
  KMP_DEBUG_ASSERT(__kmp_init_serial);

  if (__kmp_env_consistency_check) {
    if (user_lock == NULL) {
      KMP_FATAL(LockIsUninitialized, func);
    }
  }

  KMP_CHECK_USER_LOCK_INIT();

  if ((__kmp_user_lock_kind == lk_tas) &&
      (sizeof(lck->tas.lk.poll) + sizeof(lck->tas.lk.depth_locked) <=
       OMP_NEST_LOCK_T_SIZE)) {
    lck = (kmp_user_lock_p)user_lock;
  }
#if KMP_USE_FUTEX
  else if ((__kmp_user_lock_kind == lk_futex) &&
           (sizeof(lck->futex.lk.poll) + sizeof(lck->futex.lk.depth_locked) <=
            OMP_NEST_LOCK_T_SIZE)) {
    lck = (kmp_user_lock_p)user_lock;
  }
#endif
  else {
    lck = __kmp_user_lock_allocate(user_lock, gtid, 0);
  }

  INIT_NESTED_LOCK(lck);
  __kmp_set_user_lock_location(lck, loc);

#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.ompt_callback_lock_init) {
    ompt_callbacks.ompt_callback(ompt_callback_lock_init)(
        ompt_mutex_nest_lock, omp_lock_hint_none, __ompt_get_mutex_impl_type(),
        (ompt_wait_id_t)(uintptr_t)user_lock, codeptr);
  }
#endif

#if USE_ITT_BUILD
  __kmp_itt_lock_creating(lck);
#endif /* USE_ITT_BUILD */

#endif // KMP_USE_DYNAMIC_LOCK
} // __kmpc_init_nest_lock

void __kmpc_destroy_lock(ident_t *loc, kmp_int32 gtid, void **user_lock) {
#if KMP_USE_DYNAMIC_LOCK

#if USE_ITT_BUILD
  kmp_user_lock_p lck;
  if (KMP_EXTRACT_D_TAG(user_lock) == 0) {
    lck = ((kmp_indirect_lock_t *)KMP_LOOKUP_I_LOCK(user_lock))->lock;
  } else {
    lck = (kmp_user_lock_p)user_lock;
  }
  __kmp_itt_lock_destroyed(lck);
#endif
#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.ompt_callback_lock_destroy) {
    kmp_user_lock_p lck;
    if (KMP_EXTRACT_D_TAG(user_lock) == 0) {
      lck = ((kmp_indirect_lock_t *)KMP_LOOKUP_I_LOCK(user_lock))->lock;
    } else {
      lck = (kmp_user_lock_p)user_lock;
    }
    ompt_callbacks.ompt_callback(ompt_callback_lock_destroy)(
        ompt_mutex_lock, (ompt_wait_id_t)(uintptr_t)user_lock, codeptr);
  }
#endif
  KMP_D_LOCK_FUNC(user_lock, destroy)((kmp_dyna_lock_t *)user_lock);
#else
  kmp_user_lock_p lck;

  if ((__kmp_user_lock_kind == lk_tas) &&
      (sizeof(lck->tas.lk.poll) <= OMP_LOCK_T_SIZE)) {
    lck = (kmp_user_lock_p)user_lock;
  }
#if KMP_USE_FUTEX
  else if ((__kmp_user_lock_kind == lk_futex) &&
           (sizeof(lck->futex.lk.poll) <= OMP_LOCK_T_SIZE)) {
    lck = (kmp_user_lock_p)user_lock;
  }
#endif
  else {
    lck = __kmp_lookup_user_lock(user_lock, "omp_destroy_lock");
  }

#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.ompt_callback_lock_destroy) {
    ompt_callbacks.ompt_callback(ompt_callback_lock_destroy)(
        ompt_mutex_lock, (ompt_wait_id_t)(uintptr_t)user_lock, codeptr);
  }
#endif

#if USE_ITT_BUILD
  __kmp_itt_lock_destroyed(lck);
#endif /* USE_ITT_BUILD */
  DESTROY_LOCK(lck);

  if ((__kmp_user_lock_kind == lk_tas) &&
      (sizeof(lck->tas.lk.poll) <= OMP_LOCK_T_SIZE)) {
    ;
  }
#if KMP_USE_FUTEX
  else if ((__kmp_user_lock_kind == lk_futex) &&
           (sizeof(lck->futex.lk.poll) <= OMP_LOCK_T_SIZE)) {
    ;
  }
#endif
  else {
    __kmp_user_lock_free(user_lock, gtid, lck);
  }
#endif // KMP_USE_DYNAMIC_LOCK
} // __kmpc_destroy_lock

/* destroy the lock */
void __kmpc_destroy_nest_lock(ident_t *loc, kmp_int32 gtid, void **user_lock) {
#if KMP_USE_DYNAMIC_LOCK

#if USE_ITT_BUILD
  kmp_indirect_lock_t *ilk = KMP_LOOKUP_I_LOCK(user_lock);
  __kmp_itt_lock_destroyed(ilk->lock);
#endif
#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.ompt_callback_lock_destroy) {
    ompt_callbacks.ompt_callback(ompt_callback_lock_destroy)(
        ompt_mutex_nest_lock, (ompt_wait_id_t)(uintptr_t)user_lock, codeptr);
  }
#endif
  KMP_D_LOCK_FUNC(user_lock, destroy)((kmp_dyna_lock_t *)user_lock);

#else // KMP_USE_DYNAMIC_LOCK

  kmp_user_lock_p lck;

  if ((__kmp_user_lock_kind == lk_tas) &&
      (sizeof(lck->tas.lk.poll) + sizeof(lck->tas.lk.depth_locked) <=
       OMP_NEST_LOCK_T_SIZE)) {
    lck = (kmp_user_lock_p)user_lock;
  }
#if KMP_USE_FUTEX
  else if ((__kmp_user_lock_kind == lk_futex) &&
           (sizeof(lck->futex.lk.poll) + sizeof(lck->futex.lk.depth_locked) <=
            OMP_NEST_LOCK_T_SIZE)) {
    lck = (kmp_user_lock_p)user_lock;
  }
#endif
  else {
    lck = __kmp_lookup_user_lock(user_lock, "omp_destroy_nest_lock");
  }

#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.ompt_callback_lock_destroy) {
    ompt_callbacks.ompt_callback(ompt_callback_lock_destroy)(
        ompt_mutex_nest_lock, (ompt_wait_id_t)(uintptr_t)user_lock, codeptr);
  }
#endif

#if USE_ITT_BUILD
  __kmp_itt_lock_destroyed(lck);
#endif /* USE_ITT_BUILD */

  DESTROY_NESTED_LOCK(lck);

  if ((__kmp_user_lock_kind == lk_tas) &&
      (sizeof(lck->tas.lk.poll) + sizeof(lck->tas.lk.depth_locked) <=
       OMP_NEST_LOCK_T_SIZE)) {
    ;
  }
#if KMP_USE_FUTEX
  else if ((__kmp_user_lock_kind == lk_futex) &&
           (sizeof(lck->futex.lk.poll) + sizeof(lck->futex.lk.depth_locked) <=
            OMP_NEST_LOCK_T_SIZE)) {
    ;
  }
#endif
  else {
    __kmp_user_lock_free(user_lock, gtid, lck);
  }
#endif // KMP_USE_DYNAMIC_LOCK
} // __kmpc_destroy_nest_lock

void __kmpc_set_lock(ident_t *loc, kmp_int32 gtid, void **user_lock) {
  KMP_COUNT_BLOCK(OMP_set_lock);
#if KMP_USE_DYNAMIC_LOCK
  int tag = KMP_EXTRACT_D_TAG(user_lock);
#if USE_ITT_BUILD
  __kmp_itt_lock_acquiring(
      (kmp_user_lock_p)
          user_lock); // itt function will get to the right lock object.
#endif
#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.ompt_callback_mutex_acquire) {
    ompt_callbacks.ompt_callback(ompt_callback_mutex_acquire)(
        ompt_mutex_lock, omp_lock_hint_none,
        __ompt_get_mutex_impl_type(user_lock),
        (ompt_wait_id_t)(uintptr_t)user_lock, codeptr);
  }
#endif
#if KMP_USE_INLINED_TAS
  if (tag == locktag_tas && !__kmp_env_consistency_check) {
    KMP_ACQUIRE_TAS_LOCK(user_lock, gtid);
  } else
#elif KMP_USE_INLINED_FUTEX
  if (tag == locktag_futex && !__kmp_env_consistency_check) {
    KMP_ACQUIRE_FUTEX_LOCK(user_lock, gtid);
  } else
#endif
  {
    __kmp_direct_set[tag]((kmp_dyna_lock_t *)user_lock, gtid);
  }
#if USE_ITT_BUILD
  __kmp_itt_lock_acquired((kmp_user_lock_p)user_lock);
#endif
#if OMPT_SUPPORT && OMPT_OPTIONAL
  if (ompt_enabled.ompt_callback_mutex_acquired) {
    ompt_callbacks.ompt_callback(ompt_callback_mutex_acquired)(
        ompt_mutex_lock, (ompt_wait_id_t)(uintptr_t)user_lock, codeptr);
  }
#endif

#else // KMP_USE_DYNAMIC_LOCK

  kmp_user_lock_p lck;

  if ((__kmp_user_lock_kind == lk_tas) &&
      (sizeof(lck->tas.lk.poll) <= OMP_LOCK_T_SIZE)) {
    lck = (kmp_user_lock_p)user_lock;
  }
#if KMP_USE_FUTEX
  else if ((__kmp_user_lock_kind == lk_futex) &&
           (sizeof(lck->futex.lk.poll) <= OMP_LOCK_T_SIZE)) {
    lck = (kmp_user_lock_p)user_lock;
  }
#endif
  else {
    lck = __kmp_lookup_user_lock(user_lock, "omp_set_lock");
  }

#if USE_ITT_BUILD
  __kmp_itt_lock_acquiring(lck);
#endif /* USE_ITT_BUILD */
#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.ompt_callback_mutex_acquire) {
    ompt_callbacks.ompt_callback(ompt_callback_mutex_acquire)(
        ompt_mutex_lock, omp_lock_hint_none, __ompt_get_mutex_impl_type(),
        (ompt_wait_id_t)(uintptr_t)lck, codeptr);
  }
#endif

  ACQUIRE_LOCK(lck, gtid);

#if USE_ITT_BUILD
  __kmp_itt_lock_acquired(lck);
#endif /* USE_ITT_BUILD */

#if OMPT_SUPPORT && OMPT_OPTIONAL
  if (ompt_enabled.ompt_callback_mutex_acquired) {
    ompt_callbacks.ompt_callback(ompt_callback_mutex_acquired)(
        ompt_mutex_lock, (ompt_wait_id_t)(uintptr_t)lck, codeptr);
  }
#endif

#endif // KMP_USE_DYNAMIC_LOCK
}

void __kmpc_set_nest_lock(ident_t *loc, kmp_int32 gtid, void **user_lock) {
#if KMP_USE_DYNAMIC_LOCK

#if USE_ITT_BUILD
  __kmp_itt_lock_acquiring((kmp_user_lock_p)user_lock);
#endif
#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.enabled) {
    if (ompt_enabled.ompt_callback_mutex_acquire) {
      ompt_callbacks.ompt_callback(ompt_callback_mutex_acquire)(
          ompt_mutex_nest_lock, omp_lock_hint_none,
          __ompt_get_mutex_impl_type(user_lock),
          (ompt_wait_id_t)(uintptr_t)user_lock, codeptr);
    }
  }
#endif
  int acquire_status =
      KMP_D_LOCK_FUNC(user_lock, set)((kmp_dyna_lock_t *)user_lock, gtid);
  (void) acquire_status;
#if USE_ITT_BUILD
  __kmp_itt_lock_acquired((kmp_user_lock_p)user_lock);
#endif

#if OMPT_SUPPORT && OMPT_OPTIONAL
  if (ompt_enabled.enabled) {
    if (acquire_status == KMP_LOCK_ACQUIRED_FIRST) {
      if (ompt_enabled.ompt_callback_mutex_acquired) {
        // lock_first
        ompt_callbacks.ompt_callback(ompt_callback_mutex_acquired)(
            ompt_mutex_nest_lock, (ompt_wait_id_t)(uintptr_t)user_lock,
            codeptr);
      }
    } else {
      if (ompt_enabled.ompt_callback_nest_lock) {
        // lock_next
        ompt_callbacks.ompt_callback(ompt_callback_nest_lock)(
            ompt_scope_begin, (ompt_wait_id_t)(uintptr_t)user_lock, codeptr);
      }
    }
  }
#endif

#else // KMP_USE_DYNAMIC_LOCK
  int acquire_status;
  kmp_user_lock_p lck;

  if ((__kmp_user_lock_kind == lk_tas) &&
      (sizeof(lck->tas.lk.poll) + sizeof(lck->tas.lk.depth_locked) <=
       OMP_NEST_LOCK_T_SIZE)) {
    lck = (kmp_user_lock_p)user_lock;
  }
#if KMP_USE_FUTEX
  else if ((__kmp_user_lock_kind == lk_futex) &&
           (sizeof(lck->futex.lk.poll) + sizeof(lck->futex.lk.depth_locked) <=
            OMP_NEST_LOCK_T_SIZE)) {
    lck = (kmp_user_lock_p)user_lock;
  }
#endif
  else {
    lck = __kmp_lookup_user_lock(user_lock, "omp_set_nest_lock");
  }

#if USE_ITT_BUILD
  __kmp_itt_lock_acquiring(lck);
#endif /* USE_ITT_BUILD */
#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.enabled) {
    if (ompt_enabled.ompt_callback_mutex_acquire) {
      ompt_callbacks.ompt_callback(ompt_callback_mutex_acquire)(
          ompt_mutex_nest_lock, omp_lock_hint_none,
          __ompt_get_mutex_impl_type(), (ompt_wait_id_t)(uintptr_t)lck,
          codeptr);
    }
  }
#endif

  ACQUIRE_NESTED_LOCK(lck, gtid, &acquire_status);

#if USE_ITT_BUILD
  __kmp_itt_lock_acquired(lck);
#endif /* USE_ITT_BUILD */

#if OMPT_SUPPORT && OMPT_OPTIONAL
  if (ompt_enabled.enabled) {
    if (acquire_status == KMP_LOCK_ACQUIRED_FIRST) {
      if (ompt_enabled.ompt_callback_mutex_acquired) {
        // lock_first
        ompt_callbacks.ompt_callback(ompt_callback_mutex_acquired)(
            ompt_mutex_nest_lock, (ompt_wait_id_t)(uintptr_t)lck, codeptr);
      }
    } else {
      if (ompt_enabled.ompt_callback_nest_lock) {
        // lock_next
        ompt_callbacks.ompt_callback(ompt_callback_nest_lock)(
            ompt_scope_begin, (ompt_wait_id_t)(uintptr_t)lck, codeptr);
      }
    }
  }
#endif

#endif // KMP_USE_DYNAMIC_LOCK
}

void __kmpc_unset_lock(ident_t *loc, kmp_int32 gtid, void **user_lock) {
#if KMP_USE_DYNAMIC_LOCK

  int tag = KMP_EXTRACT_D_TAG(user_lock);
#if USE_ITT_BUILD
  __kmp_itt_lock_releasing((kmp_user_lock_p)user_lock);
#endif
#if KMP_USE_INLINED_TAS
  if (tag == locktag_tas && !__kmp_env_consistency_check) {
    KMP_RELEASE_TAS_LOCK(user_lock, gtid);
  } else
#elif KMP_USE_INLINED_FUTEX
  if (tag == locktag_futex && !__kmp_env_consistency_check) {
    KMP_RELEASE_FUTEX_LOCK(user_lock, gtid);
  } else
#endif
  {
    __kmp_direct_unset[tag]((kmp_dyna_lock_t *)user_lock, gtid);
  }

#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.ompt_callback_mutex_released) {
    ompt_callbacks.ompt_callback(ompt_callback_mutex_released)(
        ompt_mutex_lock, (ompt_wait_id_t)(uintptr_t)user_lock, codeptr);
  }
#endif

#else // KMP_USE_DYNAMIC_LOCK

  kmp_user_lock_p lck;

  /* Can't use serial interval since not block structured */
  /* release the lock */

  if ((__kmp_user_lock_kind == lk_tas) &&
      (sizeof(lck->tas.lk.poll) <= OMP_LOCK_T_SIZE)) {
#if KMP_OS_LINUX &&                                                            \
    (KMP_ARCH_X86 || KMP_ARCH_X86_64 || KMP_ARCH_ARM || KMP_ARCH_AARCH64)
// "fast" path implemented to fix customer performance issue
#if USE_ITT_BUILD
    __kmp_itt_lock_releasing((kmp_user_lock_p)user_lock);
#endif /* USE_ITT_BUILD */
    TCW_4(((kmp_user_lock_p)user_lock)->tas.lk.poll, 0);
    KMP_MB();

#if OMPT_SUPPORT && OMPT_OPTIONAL
    // This is the case, if called from omp_init_lock_with_hint:
    void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
    if (!codeptr)
      codeptr = OMPT_GET_RETURN_ADDRESS(0);
    if (ompt_enabled.ompt_callback_mutex_released) {
      ompt_callbacks.ompt_callback(ompt_callback_mutex_released)(
          ompt_mutex_lock, (ompt_wait_id_t)(uintptr_t)lck, codeptr);
    }
#endif

    return;
#else
    lck = (kmp_user_lock_p)user_lock;
#endif
  }
#if KMP_USE_FUTEX
  else if ((__kmp_user_lock_kind == lk_futex) &&
           (sizeof(lck->futex.lk.poll) <= OMP_LOCK_T_SIZE)) {
    lck = (kmp_user_lock_p)user_lock;
  }
#endif
  else {
    lck = __kmp_lookup_user_lock(user_lock, "omp_unset_lock");
  }

#if USE_ITT_BUILD
  __kmp_itt_lock_releasing(lck);
#endif /* USE_ITT_BUILD */

  RELEASE_LOCK(lck, gtid);

#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.ompt_callback_mutex_released) {
    ompt_callbacks.ompt_callback(ompt_callback_mutex_released)(
        ompt_mutex_lock, (ompt_wait_id_t)(uintptr_t)lck, codeptr);
  }
#endif

#endif // KMP_USE_DYNAMIC_LOCK
}

/* release the lock */
void __kmpc_unset_nest_lock(ident_t *loc, kmp_int32 gtid, void **user_lock) {
#if KMP_USE_DYNAMIC_LOCK

#if USE_ITT_BUILD
  __kmp_itt_lock_releasing((kmp_user_lock_p)user_lock);
#endif
  int release_status =
      KMP_D_LOCK_FUNC(user_lock, unset)((kmp_dyna_lock_t *)user_lock, gtid);
  (void) release_status;

#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.enabled) {
    if (release_status == KMP_LOCK_RELEASED) {
      if (ompt_enabled.ompt_callback_mutex_released) {
        // release_lock_last
        ompt_callbacks.ompt_callback(ompt_callback_mutex_released)(
            ompt_mutex_nest_lock, (ompt_wait_id_t)(uintptr_t)user_lock,
            codeptr);
      }
    } else if (ompt_enabled.ompt_callback_nest_lock) {
      // release_lock_prev
      ompt_callbacks.ompt_callback(ompt_callback_nest_lock)(
          ompt_scope_end, (ompt_wait_id_t)(uintptr_t)user_lock, codeptr);
    }
  }
#endif

#else // KMP_USE_DYNAMIC_LOCK

  kmp_user_lock_p lck;

  /* Can't use serial interval since not block structured */

  if ((__kmp_user_lock_kind == lk_tas) &&
      (sizeof(lck->tas.lk.poll) + sizeof(lck->tas.lk.depth_locked) <=
       OMP_NEST_LOCK_T_SIZE)) {
#if KMP_OS_LINUX &&                                                            \
    (KMP_ARCH_X86 || KMP_ARCH_X86_64 || KMP_ARCH_ARM || KMP_ARCH_AARCH64)
    // "fast" path implemented to fix customer performance issue
    kmp_tas_lock_t *tl = (kmp_tas_lock_t *)user_lock;
#if USE_ITT_BUILD
    __kmp_itt_lock_releasing((kmp_user_lock_p)user_lock);
#endif /* USE_ITT_BUILD */

#if OMPT_SUPPORT && OMPT_OPTIONAL
    int release_status = KMP_LOCK_STILL_HELD;
#endif

    if (--(tl->lk.depth_locked) == 0) {
      TCW_4(tl->lk.poll, 0);
#if OMPT_SUPPORT && OMPT_OPTIONAL
      release_status = KMP_LOCK_RELEASED;
#endif
    }
    KMP_MB();

#if OMPT_SUPPORT && OMPT_OPTIONAL
    // This is the case, if called from omp_init_lock_with_hint:
    void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
    if (!codeptr)
      codeptr = OMPT_GET_RETURN_ADDRESS(0);
    if (ompt_enabled.enabled) {
      if (release_status == KMP_LOCK_RELEASED) {
        if (ompt_enabled.ompt_callback_mutex_released) {
          // release_lock_last
          ompt_callbacks.ompt_callback(ompt_callback_mutex_released)(
              ompt_mutex_nest_lock, (ompt_wait_id_t)(uintptr_t)lck, codeptr);
        }
      } else if (ompt_enabled.ompt_callback_nest_lock) {
        // release_lock_previous
        ompt_callbacks.ompt_callback(ompt_callback_nest_lock)(
            ompt_mutex_scope_end, (ompt_wait_id_t)(uintptr_t)lck, codeptr);
      }
    }
#endif

    return;
#else
    lck = (kmp_user_lock_p)user_lock;
#endif
  }
#if KMP_USE_FUTEX
  else if ((__kmp_user_lock_kind == lk_futex) &&
           (sizeof(lck->futex.lk.poll) + sizeof(lck->futex.lk.depth_locked) <=
            OMP_NEST_LOCK_T_SIZE)) {
    lck = (kmp_user_lock_p)user_lock;
  }
#endif
  else {
    lck = __kmp_lookup_user_lock(user_lock, "omp_unset_nest_lock");
  }

#if USE_ITT_BUILD
  __kmp_itt_lock_releasing(lck);
#endif /* USE_ITT_BUILD */

  int release_status;
  release_status = RELEASE_NESTED_LOCK(lck, gtid);
#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.enabled) {
    if (release_status == KMP_LOCK_RELEASED) {
      if (ompt_enabled.ompt_callback_mutex_released) {
        // release_lock_last
        ompt_callbacks.ompt_callback(ompt_callback_mutex_released)(
            ompt_mutex_nest_lock, (ompt_wait_id_t)(uintptr_t)lck, codeptr);
      }
    } else if (ompt_enabled.ompt_callback_nest_lock) {
      // release_lock_previous
      ompt_callbacks.ompt_callback(ompt_callback_nest_lock)(
          ompt_mutex_scope_end, (ompt_wait_id_t)(uintptr_t)lck, codeptr);
    }
  }
#endif

#endif // KMP_USE_DYNAMIC_LOCK
}

/* try to acquire the lock */
int __kmpc_test_lock(ident_t *loc, kmp_int32 gtid, void **user_lock) {
  KMP_COUNT_BLOCK(OMP_test_lock);

#if KMP_USE_DYNAMIC_LOCK
  int rc;
  int tag = KMP_EXTRACT_D_TAG(user_lock);
#if USE_ITT_BUILD
  __kmp_itt_lock_acquiring((kmp_user_lock_p)user_lock);
#endif
#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.ompt_callback_mutex_acquire) {
    ompt_callbacks.ompt_callback(ompt_callback_mutex_acquire)(
        ompt_mutex_lock, omp_lock_hint_none,
        __ompt_get_mutex_impl_type(user_lock),
        (ompt_wait_id_t)(uintptr_t)user_lock, codeptr);
  }
#endif
#if KMP_USE_INLINED_TAS
  if (tag == locktag_tas && !__kmp_env_consistency_check) {
    KMP_TEST_TAS_LOCK(user_lock, gtid, rc);
  } else
#elif KMP_USE_INLINED_FUTEX
  if (tag == locktag_futex && !__kmp_env_consistency_check) {
    KMP_TEST_FUTEX_LOCK(user_lock, gtid, rc);
  } else
#endif
  {
    rc = __kmp_direct_test[tag]((kmp_dyna_lock_t *)user_lock, gtid);
  }
  if (rc) {
#if USE_ITT_BUILD
    __kmp_itt_lock_acquired((kmp_user_lock_p)user_lock);
#endif
#if OMPT_SUPPORT && OMPT_OPTIONAL
    if (ompt_enabled.ompt_callback_mutex_acquired) {
      ompt_callbacks.ompt_callback(ompt_callback_mutex_acquired)(
          ompt_mutex_lock, (ompt_wait_id_t)(uintptr_t)user_lock, codeptr);
    }
#endif
    return FTN_TRUE;
  } else {
#if USE_ITT_BUILD
    __kmp_itt_lock_cancelled((kmp_user_lock_p)user_lock);
#endif
    return FTN_FALSE;
  }

#else // KMP_USE_DYNAMIC_LOCK

  kmp_user_lock_p lck;
  int rc;

  if ((__kmp_user_lock_kind == lk_tas) &&
      (sizeof(lck->tas.lk.poll) <= OMP_LOCK_T_SIZE)) {
    lck = (kmp_user_lock_p)user_lock;
  }
#if KMP_USE_FUTEX
  else if ((__kmp_user_lock_kind == lk_futex) &&
           (sizeof(lck->futex.lk.poll) <= OMP_LOCK_T_SIZE)) {
    lck = (kmp_user_lock_p)user_lock;
  }
#endif
  else {
    lck = __kmp_lookup_user_lock(user_lock, "omp_test_lock");
  }

#if USE_ITT_BUILD
  __kmp_itt_lock_acquiring(lck);
#endif /* USE_ITT_BUILD */
#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.ompt_callback_mutex_acquire) {
    ompt_callbacks.ompt_callback(ompt_callback_mutex_acquire)(
        ompt_mutex_lock, omp_lock_hint_none, __ompt_get_mutex_impl_type(),
        (ompt_wait_id_t)(uintptr_t)lck, codeptr);
  }
#endif

  rc = TEST_LOCK(lck, gtid);
#if USE_ITT_BUILD
  if (rc) {
    __kmp_itt_lock_acquired(lck);
  } else {
    __kmp_itt_lock_cancelled(lck);
  }
#endif /* USE_ITT_BUILD */
#if OMPT_SUPPORT && OMPT_OPTIONAL
  if (rc && ompt_enabled.ompt_callback_mutex_acquired) {
    ompt_callbacks.ompt_callback(ompt_callback_mutex_acquired)(
        ompt_mutex_lock, (ompt_wait_id_t)(uintptr_t)lck, codeptr);
  }
#endif

  return (rc ? FTN_TRUE : FTN_FALSE);

/* Can't use serial interval since not block structured */

#endif // KMP_USE_DYNAMIC_LOCK
}

/* try to acquire the lock */
int __kmpc_test_nest_lock(ident_t *loc, kmp_int32 gtid, void **user_lock) {
#if KMP_USE_DYNAMIC_LOCK
  int rc;
#if USE_ITT_BUILD
  __kmp_itt_lock_acquiring((kmp_user_lock_p)user_lock);
#endif
#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.ompt_callback_mutex_acquire) {
    ompt_callbacks.ompt_callback(ompt_callback_mutex_acquire)(
        ompt_mutex_nest_lock, omp_lock_hint_none,
        __ompt_get_mutex_impl_type(user_lock),
        (ompt_wait_id_t)(uintptr_t)user_lock, codeptr);
  }
#endif
  rc = KMP_D_LOCK_FUNC(user_lock, test)((kmp_dyna_lock_t *)user_lock, gtid);
#if USE_ITT_BUILD
  if (rc) {
    __kmp_itt_lock_acquired((kmp_user_lock_p)user_lock);
  } else {
    __kmp_itt_lock_cancelled((kmp_user_lock_p)user_lock);
  }
#endif
#if OMPT_SUPPORT && OMPT_OPTIONAL
  if (ompt_enabled.enabled && rc) {
    if (rc == 1) {
      if (ompt_enabled.ompt_callback_mutex_acquired) {
        // lock_first
        ompt_callbacks.ompt_callback(ompt_callback_mutex_acquired)(
            ompt_mutex_nest_lock, (ompt_wait_id_t)(uintptr_t)user_lock,
            codeptr);
      }
    } else {
      if (ompt_enabled.ompt_callback_nest_lock) {
        // lock_next
        ompt_callbacks.ompt_callback(ompt_callback_nest_lock)(
            ompt_scope_begin, (ompt_wait_id_t)(uintptr_t)user_lock, codeptr);
      }
    }
  }
#endif
  return rc;

#else // KMP_USE_DYNAMIC_LOCK

  kmp_user_lock_p lck;
  int rc;

  if ((__kmp_user_lock_kind == lk_tas) &&
      (sizeof(lck->tas.lk.poll) + sizeof(lck->tas.lk.depth_locked) <=
       OMP_NEST_LOCK_T_SIZE)) {
    lck = (kmp_user_lock_p)user_lock;
  }
#if KMP_USE_FUTEX
  else if ((__kmp_user_lock_kind == lk_futex) &&
           (sizeof(lck->futex.lk.poll) + sizeof(lck->futex.lk.depth_locked) <=
            OMP_NEST_LOCK_T_SIZE)) {
    lck = (kmp_user_lock_p)user_lock;
  }
#endif
  else {
    lck = __kmp_lookup_user_lock(user_lock, "omp_test_nest_lock");
  }

#if USE_ITT_BUILD
  __kmp_itt_lock_acquiring(lck);
#endif /* USE_ITT_BUILD */

#if OMPT_SUPPORT && OMPT_OPTIONAL
  // This is the case, if called from omp_init_lock_with_hint:
  void *codeptr = OMPT_LOAD_RETURN_ADDRESS(gtid);
  if (!codeptr)
    codeptr = OMPT_GET_RETURN_ADDRESS(0);
  if (ompt_enabled.enabled) &&
        ompt_enabled.ompt_callback_mutex_acquire) {
      ompt_callbacks.ompt_callback(ompt_callback_mutex_acquire)(
          ompt_mutex_nest_lock, omp_lock_hint_none,
          __ompt_get_mutex_impl_type(), (ompt_wait_id_t)(uintptr_t)lck,
          codeptr);
    }
#endif

  rc = TEST_NESTED_LOCK(lck, gtid);
#if USE_ITT_BUILD
  if (rc) {
    __kmp_itt_lock_acquired(lck);
  } else {
    __kmp_itt_lock_cancelled(lck);
  }
#endif /* USE_ITT_BUILD */
#if OMPT_SUPPORT && OMPT_OPTIONAL
  if (ompt_enabled.enabled && rc) {
    if (rc == 1) {
      if (ompt_enabled.ompt_callback_mutex_acquired) {
        // lock_first
        ompt_callbacks.ompt_callback(ompt_callback_mutex_acquired)(
            ompt_mutex_nest_lock, (ompt_wait_id_t)(uintptr_t)lck, codeptr);
      }
    } else {
      if (ompt_enabled.ompt_callback_nest_lock) {
        // lock_next
        ompt_callbacks.ompt_callback(ompt_callback_nest_lock)(
            ompt_mutex_scope_begin, (ompt_wait_id_t)(uintptr_t)lck, codeptr);
      }
    }
  }
#endif
  return rc;

/* Can't use serial interval since not block structured */

#endif // KMP_USE_DYNAMIC_LOCK
}

// Interface to fast scalable reduce methods routines

// keep the selected method in a thread local structure for cross-function
// usage: will be used in __kmpc_end_reduce* functions;
// another solution: to re-determine the method one more time in
// __kmpc_end_reduce* functions (new prototype required then)
// AT: which solution is better?
#define __KMP_SET_REDUCTION_METHOD(gtid, rmethod)                              \
  ((__kmp_threads[(gtid)]->th.th_local.packed_reduction_method) = (rmethod))

#define __KMP_GET_REDUCTION_METHOD(gtid)                                       \
  (__kmp_threads[(gtid)]->th.th_local.packed_reduction_method)

// description of the packed_reduction_method variable: look at the macros in
// kmp.h

// used in a critical section reduce block
static __forceinline void
__kmp_enter_critical_section_reduce_block(ident_t *loc, kmp_int32 global_tid,
                                          kmp_critical_name *crit) {

  // this lock was visible to a customer and to the threading profile tool as a
  // serial overhead span (although it's used for an internal purpose only)
  //            why was it visible in previous implementation?
  //            should we keep it visible in new reduce block?
  kmp_user_lock_p lck;

#if KMP_USE_DYNAMIC_LOCK

  kmp_dyna_lock_t *lk = (kmp_dyna_lock_t *)crit;
  // Check if it is initialized.
  if (*lk == 0) {
    if (KMP_IS_D_LOCK(__kmp_user_lock_seq)) {
      KMP_COMPARE_AND_STORE_ACQ32((volatile kmp_int32 *)crit, 0,
                                  KMP_GET_D_TAG(__kmp_user_lock_seq));
    } else {
      __kmp_init_indirect_csptr(crit, loc, global_tid,
                                KMP_GET_I_TAG(__kmp_user_lock_seq));
    }
  }
  // Branch for accessing the actual lock object and set operation. This
  // branching is inevitable since this lock initialization does not follow the
  // normal dispatch path (lock table is not used).
  if (KMP_EXTRACT_D_TAG(lk) != 0) {
    lck = (kmp_user_lock_p)lk;
    KMP_DEBUG_ASSERT(lck != NULL);
    if (__kmp_env_consistency_check) {
      __kmp_push_sync(global_tid, ct_critical, loc, lck, __kmp_user_lock_seq);
    }
    KMP_D_LOCK_FUNC(lk, set)(lk, global_tid);
  } else {
    kmp_indirect_lock_t *ilk = *((kmp_indirect_lock_t **)lk);
    lck = ilk->lock;
    KMP_DEBUG_ASSERT(lck != NULL);
    if (__kmp_env_consistency_check) {
      __kmp_push_sync(global_tid, ct_critical, loc, lck, __kmp_user_lock_seq);
    }
    KMP_I_LOCK_FUNC(ilk, set)(lck, global_tid);
  }

#else // KMP_USE_DYNAMIC_LOCK

  // We know that the fast reduction code is only emitted by Intel compilers
  // with 32 byte critical sections. If there isn't enough space, then we
  // have to use a pointer.
  if (__kmp_base_user_lock_size <= INTEL_CRITICAL_SIZE) {
    lck = (kmp_user_lock_p)crit;
  } else {
    lck = __kmp_get_critical_section_ptr(crit, loc, global_tid);
  }
  KMP_DEBUG_ASSERT(lck != NULL);

  if (__kmp_env_consistency_check)
    __kmp_push_sync(global_tid, ct_critical, loc, lck);

  __kmp_acquire_user_lock_with_checks(lck, global_tid);

#endif // KMP_USE_DYNAMIC_LOCK
}

// used in a critical section reduce block
static __forceinline void
__kmp_end_critical_section_reduce_block(ident_t *loc, kmp_int32 global_tid,
                                        kmp_critical_name *crit) {

  kmp_user_lock_p lck;

#if KMP_USE_DYNAMIC_LOCK

  if (KMP_IS_D_LOCK(__kmp_user_lock_seq)) {
    lck = (kmp_user_lock_p)crit;
    if (__kmp_env_consistency_check)
      __kmp_pop_sync(global_tid, ct_critical, loc);
    KMP_D_LOCK_FUNC(lck, unset)((kmp_dyna_lock_t *)lck, global_tid);
  } else {
    kmp_indirect_lock_t *ilk =
        (kmp_indirect_lock_t *)TCR_PTR(*((kmp_indirect_lock_t **)crit));
    if (__kmp_env_consistency_check)
      __kmp_pop_sync(global_tid, ct_critical, loc);
    KMP_I_LOCK_FUNC(ilk, unset)(ilk->lock, global_tid);
  }

#else // KMP_USE_DYNAMIC_LOCK

  // We know that the fast reduction code is only emitted by Intel compilers
  // with 32 byte critical sections. If there isn't enough space, then we have
  // to use a pointer.
  if (__kmp_base_user_lock_size > 32) {
    lck = *((kmp_user_lock_p *)crit);
    KMP_ASSERT(lck != NULL);
  } else {
    lck = (kmp_user_lock_p)crit;
  }

  if (__kmp_env_consistency_check)
    __kmp_pop_sync(global_tid, ct_critical, loc);

  __kmp_release_user_lock_with_checks(lck, global_tid);

#endif // KMP_USE_DYNAMIC_LOCK
} // __kmp_end_critical_section_reduce_block

#if OMP_40_ENABLED
static __forceinline int
__kmp_swap_teams_for_teams_reduction(kmp_info_t *th, kmp_team_t **team_p,
                                     int *task_state) {
  kmp_team_t *team;

  // Check if we are inside the teams construct?
  if (th->th.th_teams_microtask) {
    *team_p = team = th->th.th_team;
    if (team->t.t_level == th->th.th_teams_level) {
      // This is reduction at teams construct.
      KMP_DEBUG_ASSERT(!th->th.th_info.ds.ds_tid); // AC: check that tid == 0
      // Let's swap teams temporarily for the reduction.
      th->th.th_info.ds.ds_tid = team->t.t_master_tid;
      th->th.th_team = team->t.t_parent;
      th->th.th_team_nproc = th->th.th_team->t.t_nproc;
      th->th.th_task_team = th->th.th_team->t.t_task_team[0];
      *task_state = th->th.th_task_state;
      th->th.th_task_state = 0;

      return 1;
    }
  }
  return 0;
}

static __forceinline void
__kmp_restore_swapped_teams(kmp_info_t *th, kmp_team_t *team, int task_state) {
  // Restore thread structure swapped in __kmp_swap_teams_for_teams_reduction.
  th->th.th_info.ds.ds_tid = 0;
  th->th.th_team = team;
  th->th.th_team_nproc = team->t.t_nproc;
  th->th.th_task_team = team->t.t_task_team[task_state];
  th->th.th_task_state = task_state;
}
#endif

/* 2.a.i. Reduce Block without a terminating barrier */
/*!
@ingroup SYNCHRONIZATION
@param loc source location information
@param global_tid global thread number
@param num_vars number of items (variables) to be reduced
@param reduce_size size of data in bytes to be reduced
@param reduce_data pointer to data to be reduced
@param reduce_func callback function providing reduction operation on two
operands and returning result of reduction in lhs_data
@param lck pointer to the unique lock data structure
@result 1 for the master thread, 0 for all other team threads, 2 for all team
threads if atomic reduction needed

The nowait version is used for a reduce clause with the nowait argument.
*/
kmp_int32
__kmpc_reduce_nowait(ident_t *loc, kmp_int32 global_tid, kmp_int32 num_vars,
                     size_t reduce_size, void *reduce_data,
                     void (*reduce_func)(void *lhs_data, void *rhs_data),
                     kmp_critical_name *lck) {

  KMP_COUNT_BLOCK(REDUCE_nowait);
  int retval = 0;
  PACKED_REDUCTION_METHOD_T packed_reduction_method;
#if OMP_40_ENABLED
  kmp_info_t *th;
  kmp_team_t *team;
  int teams_swapped = 0, task_state;
#endif
  KA_TRACE(10, ("__kmpc_reduce_nowait() enter: called T#%d\n", global_tid));

  // why do we need this initialization here at all?
  // Reduction clause can not be used as a stand-alone directive.

  // do not call __kmp_serial_initialize(), it will be called by
  // __kmp_parallel_initialize() if needed
  // possible detection of false-positive race by the threadchecker ???
  if (!TCR_4(__kmp_init_parallel))
    __kmp_parallel_initialize();

// check correctness of reduce block nesting
#if KMP_USE_DYNAMIC_LOCK
  if (__kmp_env_consistency_check)
    __kmp_push_sync(global_tid, ct_reduce, loc, NULL, 0);
#else
  if (__kmp_env_consistency_check)
    __kmp_push_sync(global_tid, ct_reduce, loc, NULL);
#endif

#if OMP_40_ENABLED
  th = __kmp_thread_from_gtid(global_tid);
  teams_swapped = __kmp_swap_teams_for_teams_reduction(th, &team, &task_state);
#endif // OMP_40_ENABLED

  // packed_reduction_method value will be reused by __kmp_end_reduce* function,
  // the value should be kept in a variable
  // the variable should be either a construct-specific or thread-specific
  // property, not a team specific property
  //     (a thread can reach the next reduce block on the next construct, reduce
  //     method may differ on the next construct)
  // an ident_t "loc" parameter could be used as a construct-specific property
  // (what if loc == 0?)
  //     (if both construct-specific and team-specific variables were shared,
  //     then unness extra syncs should be needed)
  // a thread-specific variable is better regarding two issues above (next
  // construct and extra syncs)
  // a thread-specific "th_local.reduction_method" variable is used currently
  // each thread executes 'determine' and 'set' lines (no need to execute by one
  // thread, to avoid unness extra syncs)

  packed_reduction_method = __kmp_determine_reduction_method(
      loc, global_tid, num_vars, reduce_size, reduce_data, reduce_func, lck);
  __KMP_SET_REDUCTION_METHOD(global_tid, packed_reduction_method);

  if (packed_reduction_method == critical_reduce_block) {

    __kmp_enter_critical_section_reduce_block(loc, global_tid, lck);
    retval = 1;

  } else if (packed_reduction_method == empty_reduce_block) {

    // usage: if team size == 1, no synchronization is required ( Intel
    // platforms only )
    retval = 1;

  } else if (packed_reduction_method == atomic_reduce_block) {

    retval = 2;

    // all threads should do this pop here (because __kmpc_end_reduce_nowait()
    // won't be called by the code gen)
    //     (it's not quite good, because the checking block has been closed by
    //     this 'pop',
    //      but atomic operation has not been executed yet, will be executed
    //      slightly later, literally on next instruction)
    if (__kmp_env_consistency_check)
      __kmp_pop_sync(global_tid, ct_reduce, loc);

  } else if (TEST_REDUCTION_METHOD(packed_reduction_method,
                                   tree_reduce_block)) {

// AT: performance issue: a real barrier here
// AT:     (if master goes slow, other threads are blocked here waiting for the
// master to come and release them)
// AT:     (it's not what a customer might expect specifying NOWAIT clause)
// AT:     (specifying NOWAIT won't result in improvement of performance, it'll
// be confusing to a customer)
// AT: another implementation of *barrier_gather*nowait() (or some other design)
// might go faster and be more in line with sense of NOWAIT
// AT: TO DO: do epcc test and compare times

// this barrier should be invisible to a customer and to the threading profile
// tool (it's neither a terminating barrier nor customer's code, it's
// used for an internal purpose)
#if OMPT_SUPPORT
    // JP: can this barrier potentially leed to task scheduling?
    // JP: as long as there is a barrier in the implementation, OMPT should and
    // will provide the barrier events
    //         so we set-up the necessary frame/return addresses.
    ompt_frame_t *ompt_frame;
    if (ompt_enabled.enabled) {
      __ompt_get_task_info_internal(0, NULL, NULL, &ompt_frame, NULL, NULL);
      if (ompt_frame->enter_frame.ptr == NULL)
        ompt_frame->enter_frame.ptr = OMPT_GET_FRAME_ADDRESS(0);
      OMPT_STORE_RETURN_ADDRESS(global_tid);
    }
#endif
#if USE_ITT_NOTIFY
    __kmp_threads[global_tid]->th.th_ident = loc;
#endif
    retval =
        __kmp_barrier(UNPACK_REDUCTION_BARRIER(packed_reduction_method),
                      global_tid, FALSE, reduce_size, reduce_data, reduce_func);
    retval = (retval != 0) ? (0) : (1);
#if OMPT_SUPPORT && OMPT_OPTIONAL
    if (ompt_enabled.enabled) {
      ompt_frame->enter_frame = ompt_data_none;
    }
#endif

    // all other workers except master should do this pop here
    //     ( none of other workers will get to __kmpc_end_reduce_nowait() )
    if (__kmp_env_consistency_check) {
      if (retval == 0) {
        __kmp_pop_sync(global_tid, ct_reduce, loc);
      }
    }

  } else {

    // should never reach this block
    KMP_ASSERT(0); // "unexpected method"
  }
#if OMP_40_ENABLED
  if (teams_swapped) {
    __kmp_restore_swapped_teams(th, team, task_state);
  }
#endif
  KA_TRACE(
      10,
      ("__kmpc_reduce_nowait() exit: called T#%d: method %08x, returns %08x\n",
       global_tid, packed_reduction_method, retval));

  return retval;
}

/*!
@ingroup SYNCHRONIZATION
@param loc source location information
@param global_tid global thread id.
@param lck pointer to the unique lock data structure

Finish the execution of a reduce nowait.
*/
void __kmpc_end_reduce_nowait(ident_t *loc, kmp_int32 global_tid,
                              kmp_critical_name *lck) {

  PACKED_REDUCTION_METHOD_T packed_reduction_method;

  KA_TRACE(10, ("__kmpc_end_reduce_nowait() enter: called T#%d\n", global_tid));

  packed_reduction_method = __KMP_GET_REDUCTION_METHOD(global_tid);

  if (packed_reduction_method == critical_reduce_block) {

    __kmp_end_critical_section_reduce_block(loc, global_tid, lck);

  } else if (packed_reduction_method == empty_reduce_block) {

    // usage: if team size == 1, no synchronization is required ( on Intel
    // platforms only )

  } else if (packed_reduction_method == atomic_reduce_block) {

    // neither master nor other workers should get here
    //     (code gen does not generate this call in case 2: atomic reduce block)
    // actually it's better to remove this elseif at all;
    // after removal this value will checked by the 'else' and will assert

  } else if (TEST_REDUCTION_METHOD(packed_reduction_method,
                                   tree_reduce_block)) {

    // only master gets here

  } else {

    // should never reach this block
    KMP_ASSERT(0); // "unexpected method"
  }

  if (__kmp_env_consistency_check)
    __kmp_pop_sync(global_tid, ct_reduce, loc);

  KA_TRACE(10, ("__kmpc_end_reduce_nowait() exit: called T#%d: method %08x\n",
                global_tid, packed_reduction_method));

  return;
}

/* 2.a.ii. Reduce Block with a terminating barrier */

/*!
@ingroup SYNCHRONIZATION
@param loc source location information
@param global_tid global thread number
@param num_vars number of items (variables) to be reduced
@param reduce_size size of data in bytes to be reduced
@param reduce_data pointer to data to be reduced
@param reduce_func callback function providing reduction operation on two
operands and returning result of reduction in lhs_data
@param lck pointer to the unique lock data structure
@result 1 for the master thread, 0 for all other team threads, 2 for all team
threads if atomic reduction needed

A blocking reduce that includes an implicit barrier.
*/
kmp_int32 __kmpc_reduce(ident_t *loc, kmp_int32 global_tid, kmp_int32 num_vars,
                        size_t reduce_size, void *reduce_data,
                        void (*reduce_func)(void *lhs_data, void *rhs_data),
                        kmp_critical_name *lck) {
  KMP_COUNT_BLOCK(REDUCE_wait);
  int retval = 0;
  PACKED_REDUCTION_METHOD_T packed_reduction_method;
#if OMP_40_ENABLED
  kmp_info_t *th;
  kmp_team_t *team;
  int teams_swapped = 0, task_state;
#endif

  KA_TRACE(10, ("__kmpc_reduce() enter: called T#%d\n", global_tid));

  // why do we need this initialization here at all?
  // Reduction clause can not be a stand-alone directive.

  // do not call __kmp_serial_initialize(), it will be called by
  // __kmp_parallel_initialize() if needed
  // possible detection of false-positive race by the threadchecker ???
  if (!TCR_4(__kmp_init_parallel))
    __kmp_parallel_initialize();

// check correctness of reduce block nesting
#if KMP_USE_DYNAMIC_LOCK
  if (__kmp_env_consistency_check)
    __kmp_push_sync(global_tid, ct_reduce, loc, NULL, 0);
#else
  if (__kmp_env_consistency_check)
    __kmp_push_sync(global_tid, ct_reduce, loc, NULL);
#endif

#if OMP_40_ENABLED
  th = __kmp_thread_from_gtid(global_tid);
  teams_swapped = __kmp_swap_teams_for_teams_reduction(th, &team, &task_state);
#endif // OMP_40_ENABLED

  packed_reduction_method = __kmp_determine_reduction_method(
      loc, global_tid, num_vars, reduce_size, reduce_data, reduce_func, lck);
  __KMP_SET_REDUCTION_METHOD(global_tid, packed_reduction_method);

  if (packed_reduction_method == critical_reduce_block) {

    __kmp_enter_critical_section_reduce_block(loc, global_tid, lck);
    retval = 1;

  } else if (packed_reduction_method == empty_reduce_block) {

    // usage: if team size == 1, no synchronization is required ( Intel
    // platforms only )
    retval = 1;

  } else if (packed_reduction_method == atomic_reduce_block) {

    retval = 2;

  } else if (TEST_REDUCTION_METHOD(packed_reduction_method,
                                   tree_reduce_block)) {

// case tree_reduce_block:
// this barrier should be visible to a customer and to the threading profile
// tool (it's a terminating barrier on constructs if NOWAIT not specified)
#if OMPT_SUPPORT
    ompt_frame_t *ompt_frame;
    if (ompt_enabled.enabled) {
      __ompt_get_task_info_internal(0, NULL, NULL, &ompt_frame, NULL, NULL);
      if (ompt_frame->enter_frame.ptr == NULL)
        ompt_frame->enter_frame.ptr = OMPT_GET_FRAME_ADDRESS(0);
      OMPT_STORE_RETURN_ADDRESS(global_tid);
    }
#endif
#if USE_ITT_NOTIFY
    __kmp_threads[global_tid]->th.th_ident =
        loc; // needed for correct notification of frames
#endif
    retval =
        __kmp_barrier(UNPACK_REDUCTION_BARRIER(packed_reduction_method),
                      global_tid, TRUE, reduce_size, reduce_data, reduce_func);
    retval = (retval != 0) ? (0) : (1);
#if OMPT_SUPPORT && OMPT_OPTIONAL
    if (ompt_enabled.enabled) {
      ompt_frame->enter_frame = ompt_data_none;
    }
#endif

    // all other workers except master should do this pop here
    // ( none of other workers except master will enter __kmpc_end_reduce() )
    if (__kmp_env_consistency_check) {
      if (retval == 0) { // 0: all other workers; 1: master
        __kmp_pop_sync(global_tid, ct_reduce, loc);
      }
    }

  } else {

    // should never reach this block
    KMP_ASSERT(0); // "unexpected method"
  }
#if OMP_40_ENABLED
  if (teams_swapped) {
    __kmp_restore_swapped_teams(th, team, task_state);
  }
#endif

  KA_TRACE(10,
           ("__kmpc_reduce() exit: called T#%d: method %08x, returns %08x\n",
            global_tid, packed_reduction_method, retval));

  return retval;
}

/*!
@ingroup SYNCHRONIZATION
@param loc source location information
@param global_tid global thread id.
@param lck pointer to the unique lock data structure

Finish the execution of a blocking reduce.
The <tt>lck</tt> pointer must be the same as that used in the corresponding
start function.
*/
void __kmpc_end_reduce(ident_t *loc, kmp_int32 global_tid,
                       kmp_critical_name *lck) {

  PACKED_REDUCTION_METHOD_T packed_reduction_method;
#if OMP_40_ENABLED
  kmp_info_t *th;
  kmp_team_t *team;
  int teams_swapped = 0, task_state;
#endif

  KA_TRACE(10, ("__kmpc_end_reduce() enter: called T#%d\n", global_tid));

#if OMP_40_ENABLED
  th = __kmp_thread_from_gtid(global_tid);
  teams_swapped = __kmp_swap_teams_for_teams_reduction(th, &team, &task_state);
#endif // OMP_40_ENABLED

  packed_reduction_method = __KMP_GET_REDUCTION_METHOD(global_tid);

  // this barrier should be visible to a customer and to the threading profile
  // tool (it's a terminating barrier on constructs if NOWAIT not specified)

  if (packed_reduction_method == critical_reduce_block) {

    __kmp_end_critical_section_reduce_block(loc, global_tid, lck);

// TODO: implicit barrier: should be exposed
#if OMPT_SUPPORT
    ompt_frame_t *ompt_frame;
    if (ompt_enabled.enabled) {
      __ompt_get_task_info_internal(0, NULL, NULL, &ompt_frame, NULL, NULL);
      if (ompt_frame->enter_frame.ptr == NULL)
        ompt_frame->enter_frame.ptr = OMPT_GET_FRAME_ADDRESS(0);
      OMPT_STORE_RETURN_ADDRESS(global_tid);
    }
#endif
#if USE_ITT_NOTIFY
    __kmp_threads[global_tid]->th.th_ident = loc;
#endif
    __kmp_barrier(bs_plain_barrier, global_tid, FALSE, 0, NULL, NULL);
#if OMPT_SUPPORT && OMPT_OPTIONAL
    if (ompt_enabled.enabled) {
      ompt_frame->enter_frame = ompt_data_none;
    }
#endif

  } else if (packed_reduction_method == empty_reduce_block) {

// usage: if team size==1, no synchronization is required (Intel platforms only)

// TODO: implicit barrier: should be exposed
#if OMPT_SUPPORT
    ompt_frame_t *ompt_frame;
    if (ompt_enabled.enabled) {
      __ompt_get_task_info_internal(0, NULL, NULL, &ompt_frame, NULL, NULL);
      if (ompt_frame->enter_frame.ptr == NULL)
        ompt_frame->enter_frame.ptr = OMPT_GET_FRAME_ADDRESS(0);
      OMPT_STORE_RETURN_ADDRESS(global_tid);
    }
#endif
#if USE_ITT_NOTIFY
    __kmp_threads[global_tid]->th.th_ident = loc;
#endif
    __kmp_barrier(bs_plain_barrier, global_tid, FALSE, 0, NULL, NULL);
#if OMPT_SUPPORT && OMPT_OPTIONAL
    if (ompt_enabled.enabled) {
      ompt_frame->enter_frame = ompt_data_none;
    }
#endif

  } else if (packed_reduction_method == atomic_reduce_block) {

#if OMPT_SUPPORT
    ompt_frame_t *ompt_frame;
    if (ompt_enabled.enabled) {
      __ompt_get_task_info_internal(0, NULL, NULL, &ompt_frame, NULL, NULL);
      if (ompt_frame->enter_frame.ptr == NULL)
        ompt_frame->enter_frame.ptr = OMPT_GET_FRAME_ADDRESS(0);
      OMPT_STORE_RETURN_ADDRESS(global_tid);
    }
#endif
// TODO: implicit barrier: should be exposed
#if USE_ITT_NOTIFY
    __kmp_threads[global_tid]->th.th_ident = loc;
#endif
    __kmp_barrier(bs_plain_barrier, global_tid, FALSE, 0, NULL, NULL);
#if OMPT_SUPPORT && OMPT_OPTIONAL
    if (ompt_enabled.enabled) {
      ompt_frame->enter_frame = ompt_data_none;
    }
#endif

  } else if (TEST_REDUCTION_METHOD(packed_reduction_method,
                                   tree_reduce_block)) {

    // only master executes here (master releases all other workers)
    __kmp_end_split_barrier(UNPACK_REDUCTION_BARRIER(packed_reduction_method),
                            global_tid);

  } else {

    // should never reach this block
    KMP_ASSERT(0); // "unexpected method"
  }
#if OMP_40_ENABLED
  if (teams_swapped) {
    __kmp_restore_swapped_teams(th, team, task_state);
  }
#endif

  if (__kmp_env_consistency_check)
    __kmp_pop_sync(global_tid, ct_reduce, loc);

  KA_TRACE(10, ("__kmpc_end_reduce() exit: called T#%d: method %08x\n",
                global_tid, packed_reduction_method));

  return;
}

#undef __KMP_GET_REDUCTION_METHOD
#undef __KMP_SET_REDUCTION_METHOD

/* end of interface to fast scalable reduce routines */

kmp_uint64 __kmpc_get_taskid() {

  kmp_int32 gtid;
  kmp_info_t *thread;

  gtid = __kmp_get_gtid();
  if (gtid < 0) {
    return 0;
  }
  thread = __kmp_thread_from_gtid(gtid);
  return thread->th.th_current_task->td_task_id;

} // __kmpc_get_taskid

kmp_uint64 __kmpc_get_parent_taskid() {

  kmp_int32 gtid;
  kmp_info_t *thread;
  kmp_taskdata_t *parent_task;

  gtid = __kmp_get_gtid();
  if (gtid < 0) {
    return 0;
  }
  thread = __kmp_thread_from_gtid(gtid);
  parent_task = thread->th.th_current_task->td_parent;
  return (parent_task == NULL ? 0 : parent_task->td_task_id);

} // __kmpc_get_parent_taskid

#if OMP_45_ENABLED
/*!
@ingroup WORK_SHARING
@param loc  source location information.
@param gtid  global thread number.
@param num_dims  number of associated doacross loops.
@param dims  info on loops bounds.

Initialize doacross loop information.
Expect compiler send us inclusive bounds,
e.g. for(i=2;i<9;i+=2) lo=2, up=8, st=2.
*/
void __kmpc_doacross_init(ident_t *loc, int gtid, int num_dims,
                          const struct kmp_dim *dims) {
  int j, idx;
  kmp_int64 last, trace_count;
  kmp_info_t *th = __kmp_threads[gtid];
  kmp_team_t *team = th->th.th_team;
  kmp_uint32 *flags;
  kmp_disp_t *pr_buf = th->th.th_dispatch;
  dispatch_shared_info_t *sh_buf;

  KA_TRACE(
      20,
      ("__kmpc_doacross_init() enter: called T#%d, num dims %d, active %d\n",
       gtid, num_dims, !team->t.t_serialized));
  KMP_DEBUG_ASSERT(dims != NULL);
  KMP_DEBUG_ASSERT(num_dims > 0);

  if (team->t.t_serialized) {
    KA_TRACE(20, ("__kmpc_doacross_init() exit: serialized team\n"));
    return; // no dependencies if team is serialized
  }
  KMP_DEBUG_ASSERT(team->t.t_nproc > 1);
  idx = pr_buf->th_doacross_buf_idx++; // Increment index of shared buffer for
  // the next loop
  sh_buf = &team->t.t_disp_buffer[idx % __kmp_dispatch_num_buffers];

  // Save bounds info into allocated private buffer
  KMP_DEBUG_ASSERT(pr_buf->th_doacross_info == NULL);
  pr_buf->th_doacross_info = (kmp_int64 *)__kmp_thread_malloc(
      th, sizeof(kmp_int64) * (4 * num_dims + 1));
  KMP_DEBUG_ASSERT(pr_buf->th_doacross_info != NULL);
  pr_buf->th_doacross_info[0] =
      (kmp_int64)num_dims; // first element is number of dimensions
  // Save also address of num_done in order to access it later without knowing
  // the buffer index
  pr_buf->th_doacross_info[1] = (kmp_int64)&sh_buf->doacross_num_done;
  pr_buf->th_doacross_info[2] = dims[0].lo;
  pr_buf->th_doacross_info[3] = dims[0].up;
  pr_buf->th_doacross_info[4] = dims[0].st;
  last = 5;
  for (j = 1; j < num_dims; ++j) {
    kmp_int64
        range_length; // To keep ranges of all dimensions but the first dims[0]
    if (dims[j].st == 1) { // most common case
      // AC: should we care of ranges bigger than LLONG_MAX? (not for now)
      range_length = dims[j].up - dims[j].lo + 1;
    } else {
      if (dims[j].st > 0) {
        KMP_DEBUG_ASSERT(dims[j].up > dims[j].lo);
        range_length = (kmp_uint64)(dims[j].up - dims[j].lo) / dims[j].st + 1;
      } else { // negative increment
        KMP_DEBUG_ASSERT(dims[j].lo > dims[j].up);
        range_length =
            (kmp_uint64)(dims[j].lo - dims[j].up) / (-dims[j].st) + 1;
      }
    }
    pr_buf->th_doacross_info[last++] = range_length;
    pr_buf->th_doacross_info[last++] = dims[j].lo;
    pr_buf->th_doacross_info[last++] = dims[j].up;
    pr_buf->th_doacross_info[last++] = dims[j].st;
  }

  // Compute total trip count.
  // Start with range of dims[0] which we don't need to keep in the buffer.
  if (dims[0].st == 1) { // most common case
    trace_count = dims[0].up - dims[0].lo + 1;
  } else if (dims[0].st > 0) {
    KMP_DEBUG_ASSERT(dims[0].up > dims[0].lo);
    trace_count = (kmp_uint64)(dims[0].up - dims[0].lo) / dims[0].st + 1;
  } else { // negative increment
    KMP_DEBUG_ASSERT(dims[0].lo > dims[0].up);
    trace_count = (kmp_uint64)(dims[0].lo - dims[0].up) / (-dims[0].st) + 1;
  }
  for (j = 1; j < num_dims; ++j) {
    trace_count *= pr_buf->th_doacross_info[4 * j + 1]; // use kept ranges
  }
  KMP_DEBUG_ASSERT(trace_count > 0);

  // Check if shared buffer is not occupied by other loop (idx -
  // __kmp_dispatch_num_buffers)
  if (idx != sh_buf->doacross_buf_idx) {
    // Shared buffer is occupied, wait for it to be free
    __kmp_wait_yield_4((volatile kmp_uint32 *)&sh_buf->doacross_buf_idx, idx,
                       __kmp_eq_4, NULL);
  }
#if KMP_32_BIT_ARCH
  // Check if we are the first thread. After the CAS the first thread gets 0,
  // others get 1 if initialization is in progress, allocated pointer otherwise.
  // Treat pointer as volatile integer (value 0 or 1) until memory is allocated.
  flags = (kmp_uint32 *)KMP_COMPARE_AND_STORE_RET32(
      (volatile kmp_int32 *)&sh_buf->doacross_flags, NULL, 1);
#else
  flags = (kmp_uint32 *)KMP_COMPARE_AND_STORE_RET64(
      (volatile kmp_int64 *)&sh_buf->doacross_flags, NULL, 1LL);
#endif
  if (flags == NULL) {
    // we are the first thread, allocate the array of flags
    size_t size = trace_count / 8 + 8; // in bytes, use single bit per iteration
    flags = (kmp_uint32 *)__kmp_thread_calloc(th, size, 1);
    KMP_MB();
    sh_buf->doacross_flags = flags;
  } else if (flags == (kmp_uint32 *)1) {
#if KMP_32_BIT_ARCH
    // initialization is still in progress, need to wait
    while (*(volatile kmp_int32 *)&sh_buf->doacross_flags == 1)
#else
    while (*(volatile kmp_int64 *)&sh_buf->doacross_flags == 1LL)
#endif
      KMP_YIELD(TRUE);
    KMP_MB();
  } else {
    KMP_MB();
  }
  KMP_DEBUG_ASSERT(sh_buf->doacross_flags > (kmp_uint32 *)1); // check ptr value
  pr_buf->th_doacross_flags =
      sh_buf->doacross_flags; // save private copy in order to not
  // touch shared buffer on each iteration
  KA_TRACE(20, ("__kmpc_doacross_init() exit: T#%d\n", gtid));
}

void __kmpc_doacross_wait(ident_t *loc, int gtid, const kmp_int64 *vec) {
  kmp_int32 shft, num_dims, i;
  kmp_uint32 flag;
  kmp_int64 iter_number; // iteration number of "collapsed" loop nest
  kmp_info_t *th = __kmp_threads[gtid];
  kmp_team_t *team = th->th.th_team;
  kmp_disp_t *pr_buf;
  kmp_int64 lo, up, st;

  KA_TRACE(20, ("__kmpc_doacross_wait() enter: called T#%d\n", gtid));
  if (team->t.t_serialized) {
    KA_TRACE(20, ("__kmpc_doacross_wait() exit: serialized team\n"));
    return; // no dependencies if team is serialized
  }

  // calculate sequential iteration number and check out-of-bounds condition
  pr_buf = th->th.th_dispatch;
  KMP_DEBUG_ASSERT(pr_buf->th_doacross_info != NULL);
  num_dims = pr_buf->th_doacross_info[0];
  lo = pr_buf->th_doacross_info[2];
  up = pr_buf->th_doacross_info[3];
  st = pr_buf->th_doacross_info[4];
  if (st == 1) { // most common case
    if (vec[0] < lo || vec[0] > up) {
      KA_TRACE(20, ("__kmpc_doacross_wait() exit: T#%d iter %lld is out of "
                    "bounds [%lld,%lld]\n",
                    gtid, vec[0], lo, up));
      return;
    }
    iter_number = vec[0] - lo;
  } else if (st > 0) {
    if (vec[0] < lo || vec[0] > up) {
      KA_TRACE(20, ("__kmpc_doacross_wait() exit: T#%d iter %lld is out of "
                    "bounds [%lld,%lld]\n",
                    gtid, vec[0], lo, up));
      return;
    }
    iter_number = (kmp_uint64)(vec[0] - lo) / st;
  } else { // negative increment
    if (vec[0] > lo || vec[0] < up) {
      KA_TRACE(20, ("__kmpc_doacross_wait() exit: T#%d iter %lld is out of "
                    "bounds [%lld,%lld]\n",
                    gtid, vec[0], lo, up));
      return;
    }
    iter_number = (kmp_uint64)(lo - vec[0]) / (-st);
  }
  for (i = 1; i < num_dims; ++i) {
    kmp_int64 iter, ln;
    kmp_int32 j = i * 4;
    ln = pr_buf->th_doacross_info[j + 1];
    lo = pr_buf->th_doacross_info[j + 2];
    up = pr_buf->th_doacross_info[j + 3];
    st = pr_buf->th_doacross_info[j + 4];
    if (st == 1) {
      if (vec[i] < lo || vec[i] > up) {
        KA_TRACE(20, ("__kmpc_doacross_wait() exit: T#%d iter %lld is out of "
                      "bounds [%lld,%lld]\n",
                      gtid, vec[i], lo, up));
        return;
      }
      iter = vec[i] - lo;
    } else if (st > 0) {
      if (vec[i] < lo || vec[i] > up) {
        KA_TRACE(20, ("__kmpc_doacross_wait() exit: T#%d iter %lld is out of "
                      "bounds [%lld,%lld]\n",
                      gtid, vec[i], lo, up));
        return;
      }
      iter = (kmp_uint64)(vec[i] - lo) / st;
    } else { // st < 0
      if (vec[i] > lo || vec[i] < up) {
        KA_TRACE(20, ("__kmpc_doacross_wait() exit: T#%d iter %lld is out of "
                      "bounds [%lld,%lld]\n",
                      gtid, vec[i], lo, up));
        return;
      }
      iter = (kmp_uint64)(lo - vec[i]) / (-st);
    }
    iter_number = iter + ln * iter_number;
  }
  shft = iter_number % 32; // use 32-bit granularity
  iter_number >>= 5; // divided by 32
  flag = 1 << shft;
  while ((flag & pr_buf->th_doacross_flags[iter_number]) == 0) {
    KMP_YIELD(TRUE);
  }
  KMP_MB();
  KA_TRACE(20,
           ("__kmpc_doacross_wait() exit: T#%d wait for iter %lld completed\n",
            gtid, (iter_number << 5) + shft));
}

void __kmpc_doacross_post(ident_t *loc, int gtid, const kmp_int64 *vec) {
  kmp_int32 shft, num_dims, i;
  kmp_uint32 flag;
  kmp_int64 iter_number; // iteration number of "collapsed" loop nest
  kmp_info_t *th = __kmp_threads[gtid];
  kmp_team_t *team = th->th.th_team;
  kmp_disp_t *pr_buf;
  kmp_int64 lo, st;

  KA_TRACE(20, ("__kmpc_doacross_post() enter: called T#%d\n", gtid));
  if (team->t.t_serialized) {
    KA_TRACE(20, ("__kmpc_doacross_post() exit: serialized team\n"));
    return; // no dependencies if team is serialized
  }

  // calculate sequential iteration number (same as in "wait" but no
  // out-of-bounds checks)
  pr_buf = th->th.th_dispatch;
  KMP_DEBUG_ASSERT(pr_buf->th_doacross_info != NULL);
  num_dims = pr_buf->th_doacross_info[0];
  lo = pr_buf->th_doacross_info[2];
  st = pr_buf->th_doacross_info[4];
  if (st == 1) { // most common case
    iter_number = vec[0] - lo;
  } else if (st > 0) {
    iter_number = (kmp_uint64)(vec[0] - lo) / st;
  } else { // negative increment
    iter_number = (kmp_uint64)(lo - vec[0]) / (-st);
  }
  for (i = 1; i < num_dims; ++i) {
    kmp_int64 iter, ln;
    kmp_int32 j = i * 4;
    ln = pr_buf->th_doacross_info[j + 1];
    lo = pr_buf->th_doacross_info[j + 2];
    st = pr_buf->th_doacross_info[j + 4];
    if (st == 1) {
      iter = vec[i] - lo;
    } else if (st > 0) {
      iter = (kmp_uint64)(vec[i] - lo) / st;
    } else { // st < 0
      iter = (kmp_uint64)(lo - vec[i]) / (-st);
    }
    iter_number = iter + ln * iter_number;
  }
  shft = iter_number % 32; // use 32-bit granularity
  iter_number >>= 5; // divided by 32
  flag = 1 << shft;
  KMP_MB();
  if ((flag & pr_buf->th_doacross_flags[iter_number]) == 0)
    KMP_TEST_THEN_OR32(&pr_buf->th_doacross_flags[iter_number], flag);
  KA_TRACE(20, ("__kmpc_doacross_post() exit: T#%d iter %lld posted\n", gtid,
                (iter_number << 5) + shft));
}

void __kmpc_doacross_fini(ident_t *loc, int gtid) {
  kmp_int32 num_done;
  kmp_info_t *th = __kmp_threads[gtid];
  kmp_team_t *team = th->th.th_team;
  kmp_disp_t *pr_buf = th->th.th_dispatch;

  KA_TRACE(20, ("__kmpc_doacross_fini() enter: called T#%d\n", gtid));
  if (team->t.t_serialized) {
    KA_TRACE(20, ("__kmpc_doacross_fini() exit: serialized team %p\n", team));
    return; // nothing to do
  }
  num_done = KMP_TEST_THEN_INC32((kmp_int32 *)pr_buf->th_doacross_info[1]) + 1;
  if (num_done == th->th.th_team_nproc) {
    // we are the last thread, need to free shared resources
    int idx = pr_buf->th_doacross_buf_idx - 1;
    dispatch_shared_info_t *sh_buf =
        &team->t.t_disp_buffer[idx % __kmp_dispatch_num_buffers];
    KMP_DEBUG_ASSERT(pr_buf->th_doacross_info[1] ==
                     (kmp_int64)&sh_buf->doacross_num_done);
    KMP_DEBUG_ASSERT(num_done == sh_buf->doacross_num_done);
    KMP_DEBUG_ASSERT(idx == sh_buf->doacross_buf_idx);
    __kmp_thread_free(th, CCAST(kmp_uint32 *, sh_buf->doacross_flags));
    sh_buf->doacross_flags = NULL;
    sh_buf->doacross_num_done = 0;
    sh_buf->doacross_buf_idx +=
        __kmp_dispatch_num_buffers; // free buffer for future re-use
  }
  // free private resources (need to keep buffer index forever)
  pr_buf->th_doacross_flags = NULL;
  __kmp_thread_free(th, (void *)pr_buf->th_doacross_info);
  pr_buf->th_doacross_info = NULL;
  KA_TRACE(20, ("__kmpc_doacross_fini() exit: T#%d\n", gtid));
}
#endif

#if OMP_50_ENABLED
int __kmpc_get_target_offload(void) {
  if (!__kmp_init_serial) {
    __kmp_serial_initialize();
  }
  return __kmp_target_offload;
}
#endif // OMP_50_ENABLED

// end of file //