aboutsummaryrefslogtreecommitdiff
path: root/runtime/src/kmp_atomic.cpp
blob: f870e308e13d9bf586641686a368cea62fee1590 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
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
/*
 * kmp_atomic.cpp -- ATOMIC implementation routines
 */


//===----------------------------------------------------------------------===//
//
//                     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.
//
//===----------------------------------------------------------------------===//


#include "kmp_atomic.h"
#include "kmp.h" // TRUE, asm routines prototypes

typedef unsigned char uchar;
typedef unsigned short ushort;

/*!
@defgroup ATOMIC_OPS Atomic Operations
These functions are used for implementing the many different varieties of atomic
operations.

The compiler is at liberty to inline atomic operations that are naturally
supported by the target architecture. For instance on IA-32 architecture an
atomic like this can be inlined
@code
static int s = 0;
#pragma omp atomic
    s++;
@endcode
using the single instruction: `lock; incl s`

However the runtime does provide entrypoints for these operations to support
compilers that choose not to inline them. (For instance,
`__kmpc_atomic_fixed4_add` could be used to perform the increment above.)

The names of the functions are encoded by using the data type name and the
operation name, as in these tables.

Data Type  | Data type encoding
-----------|---------------
int8_t     | `fixed1`
uint8_t    | `fixed1u`
int16_t    | `fixed2`
uint16_t   | `fixed2u`
int32_t    | `fixed4`
uint32_t   | `fixed4u`
int32_t    | `fixed8`
uint32_t   | `fixed8u`
float      | `float4`
double     | `float8`
float 10 (8087 eighty bit float)  | `float10`
complex<float>   |  `cmplx4`
complex<double>  | `cmplx8`
complex<float10> | `cmplx10`
<br>

Operation | Operation encoding
----------|-------------------
+ | add
- | sub
\* | mul
/ | div
& | andb
<< | shl
\>\> | shr
\| | orb
^  | xor
&& | andl
\|\| | orl
maximum | max
minimum | min
.eqv.   | eqv
.neqv.  | neqv

<br>
For non-commutative operations, `_rev` can also be added for the reversed
operation. For the functions that capture the result, the suffix `_cpt` is
added.

Update Functions
================
The general form of an atomic function that just performs an update (without a
`capture`)
@code
void __kmpc_atomic_<datatype>_<operation>( ident_t *id_ref, int gtid, TYPE *
lhs, TYPE rhs );
@endcode
@param ident_t  a pointer to source location
@param gtid  the global thread id
@param lhs   a pointer to the left operand
@param rhs   the right operand

`capture` functions
===================
The capture functions perform an atomic update and return a result, which is
either the value before the capture, or that after. They take an additional
argument to determine which result is returned.
Their general form is therefore
@code
TYPE __kmpc_atomic_<datatype>_<operation>_cpt( ident_t *id_ref, int gtid, TYPE *
lhs, TYPE rhs, int flag );
@endcode
@param ident_t  a pointer to source location
@param gtid  the global thread id
@param lhs   a pointer to the left operand
@param rhs   the right operand
@param flag  one if the result is to be captured *after* the operation, zero if
captured *before*.

The one set of exceptions to this is the `complex<float>` type where the value
is not returned, rather an extra argument pointer is passed.

They look like
@code
void __kmpc_atomic_cmplx4_<op>_cpt(  ident_t *id_ref, int gtid, kmp_cmplx32 *
lhs, kmp_cmplx32 rhs, kmp_cmplx32 * out, int flag );
@endcode

Read and Write Operations
=========================
The OpenMP<sup>*</sup> standard now supports atomic operations that simply
ensure that the value is read or written atomically, with no modification
performed. In many cases on IA-32 architecture these operations can be inlined
since the architecture guarantees that no tearing occurs on aligned objects
accessed with a single memory operation of up to 64 bits in size.

The general form of the read operations is
@code
TYPE __kmpc_atomic_<type>_rd ( ident_t *id_ref, int gtid, TYPE * loc );
@endcode

For the write operations the form is
@code
void __kmpc_atomic_<type>_wr ( ident_t *id_ref, int gtid, TYPE * lhs, TYPE rhs
);
@endcode

Full list of functions
======================
This leads to the generation of 376 atomic functions, as follows.

Functons for integers
---------------------
There are versions here for integers of size 1,2,4 and 8 bytes both signed and
unsigned (where that matters).
@code
    __kmpc_atomic_fixed1_add
    __kmpc_atomic_fixed1_add_cpt
    __kmpc_atomic_fixed1_add_fp
    __kmpc_atomic_fixed1_andb
    __kmpc_atomic_fixed1_andb_cpt
    __kmpc_atomic_fixed1_andl
    __kmpc_atomic_fixed1_andl_cpt
    __kmpc_atomic_fixed1_div
    __kmpc_atomic_fixed1_div_cpt
    __kmpc_atomic_fixed1_div_cpt_rev
    __kmpc_atomic_fixed1_div_float8
    __kmpc_atomic_fixed1_div_fp
    __kmpc_atomic_fixed1_div_rev
    __kmpc_atomic_fixed1_eqv
    __kmpc_atomic_fixed1_eqv_cpt
    __kmpc_atomic_fixed1_max
    __kmpc_atomic_fixed1_max_cpt
    __kmpc_atomic_fixed1_min
    __kmpc_atomic_fixed1_min_cpt
    __kmpc_atomic_fixed1_mul
    __kmpc_atomic_fixed1_mul_cpt
    __kmpc_atomic_fixed1_mul_float8
    __kmpc_atomic_fixed1_mul_fp
    __kmpc_atomic_fixed1_neqv
    __kmpc_atomic_fixed1_neqv_cpt
    __kmpc_atomic_fixed1_orb
    __kmpc_atomic_fixed1_orb_cpt
    __kmpc_atomic_fixed1_orl
    __kmpc_atomic_fixed1_orl_cpt
    __kmpc_atomic_fixed1_rd
    __kmpc_atomic_fixed1_shl
    __kmpc_atomic_fixed1_shl_cpt
    __kmpc_atomic_fixed1_shl_cpt_rev
    __kmpc_atomic_fixed1_shl_rev
    __kmpc_atomic_fixed1_shr
    __kmpc_atomic_fixed1_shr_cpt
    __kmpc_atomic_fixed1_shr_cpt_rev
    __kmpc_atomic_fixed1_shr_rev
    __kmpc_atomic_fixed1_sub
    __kmpc_atomic_fixed1_sub_cpt
    __kmpc_atomic_fixed1_sub_cpt_rev
    __kmpc_atomic_fixed1_sub_fp
    __kmpc_atomic_fixed1_sub_rev
    __kmpc_atomic_fixed1_swp
    __kmpc_atomic_fixed1_wr
    __kmpc_atomic_fixed1_xor
    __kmpc_atomic_fixed1_xor_cpt
    __kmpc_atomic_fixed1u_add_fp
    __kmpc_atomic_fixed1u_sub_fp
    __kmpc_atomic_fixed1u_mul_fp
    __kmpc_atomic_fixed1u_div
    __kmpc_atomic_fixed1u_div_cpt
    __kmpc_atomic_fixed1u_div_cpt_rev
    __kmpc_atomic_fixed1u_div_fp
    __kmpc_atomic_fixed1u_div_rev
    __kmpc_atomic_fixed1u_shr
    __kmpc_atomic_fixed1u_shr_cpt
    __kmpc_atomic_fixed1u_shr_cpt_rev
    __kmpc_atomic_fixed1u_shr_rev
    __kmpc_atomic_fixed2_add
    __kmpc_atomic_fixed2_add_cpt
    __kmpc_atomic_fixed2_add_fp
    __kmpc_atomic_fixed2_andb
    __kmpc_atomic_fixed2_andb_cpt
    __kmpc_atomic_fixed2_andl
    __kmpc_atomic_fixed2_andl_cpt
    __kmpc_atomic_fixed2_div
    __kmpc_atomic_fixed2_div_cpt
    __kmpc_atomic_fixed2_div_cpt_rev
    __kmpc_atomic_fixed2_div_float8
    __kmpc_atomic_fixed2_div_fp
    __kmpc_atomic_fixed2_div_rev
    __kmpc_atomic_fixed2_eqv
    __kmpc_atomic_fixed2_eqv_cpt
    __kmpc_atomic_fixed2_max
    __kmpc_atomic_fixed2_max_cpt
    __kmpc_atomic_fixed2_min
    __kmpc_atomic_fixed2_min_cpt
    __kmpc_atomic_fixed2_mul
    __kmpc_atomic_fixed2_mul_cpt
    __kmpc_atomic_fixed2_mul_float8
    __kmpc_atomic_fixed2_mul_fp
    __kmpc_atomic_fixed2_neqv
    __kmpc_atomic_fixed2_neqv_cpt
    __kmpc_atomic_fixed2_orb
    __kmpc_atomic_fixed2_orb_cpt
    __kmpc_atomic_fixed2_orl
    __kmpc_atomic_fixed2_orl_cpt
    __kmpc_atomic_fixed2_rd
    __kmpc_atomic_fixed2_shl
    __kmpc_atomic_fixed2_shl_cpt
    __kmpc_atomic_fixed2_shl_cpt_rev
    __kmpc_atomic_fixed2_shl_rev
    __kmpc_atomic_fixed2_shr
    __kmpc_atomic_fixed2_shr_cpt
    __kmpc_atomic_fixed2_shr_cpt_rev
    __kmpc_atomic_fixed2_shr_rev
    __kmpc_atomic_fixed2_sub
    __kmpc_atomic_fixed2_sub_cpt
    __kmpc_atomic_fixed2_sub_cpt_rev
    __kmpc_atomic_fixed2_sub_fp
    __kmpc_atomic_fixed2_sub_rev
    __kmpc_atomic_fixed2_swp
    __kmpc_atomic_fixed2_wr
    __kmpc_atomic_fixed2_xor
    __kmpc_atomic_fixed2_xor_cpt
    __kmpc_atomic_fixed2u_add_fp
    __kmpc_atomic_fixed2u_sub_fp
    __kmpc_atomic_fixed2u_mul_fp
    __kmpc_atomic_fixed2u_div
    __kmpc_atomic_fixed2u_div_cpt
    __kmpc_atomic_fixed2u_div_cpt_rev
    __kmpc_atomic_fixed2u_div_fp
    __kmpc_atomic_fixed2u_div_rev
    __kmpc_atomic_fixed2u_shr
    __kmpc_atomic_fixed2u_shr_cpt
    __kmpc_atomic_fixed2u_shr_cpt_rev
    __kmpc_atomic_fixed2u_shr_rev
    __kmpc_atomic_fixed4_add
    __kmpc_atomic_fixed4_add_cpt
    __kmpc_atomic_fixed4_add_fp
    __kmpc_atomic_fixed4_andb
    __kmpc_atomic_fixed4_andb_cpt
    __kmpc_atomic_fixed4_andl
    __kmpc_atomic_fixed4_andl_cpt
    __kmpc_atomic_fixed4_div
    __kmpc_atomic_fixed4_div_cpt
    __kmpc_atomic_fixed4_div_cpt_rev
    __kmpc_atomic_fixed4_div_float8
    __kmpc_atomic_fixed4_div_fp
    __kmpc_atomic_fixed4_div_rev
    __kmpc_atomic_fixed4_eqv
    __kmpc_atomic_fixed4_eqv_cpt
    __kmpc_atomic_fixed4_max
    __kmpc_atomic_fixed4_max_cpt
    __kmpc_atomic_fixed4_min
    __kmpc_atomic_fixed4_min_cpt
    __kmpc_atomic_fixed4_mul
    __kmpc_atomic_fixed4_mul_cpt
    __kmpc_atomic_fixed4_mul_float8
    __kmpc_atomic_fixed4_mul_fp
    __kmpc_atomic_fixed4_neqv
    __kmpc_atomic_fixed4_neqv_cpt
    __kmpc_atomic_fixed4_orb
    __kmpc_atomic_fixed4_orb_cpt
    __kmpc_atomic_fixed4_orl
    __kmpc_atomic_fixed4_orl_cpt
    __kmpc_atomic_fixed4_rd
    __kmpc_atomic_fixed4_shl
    __kmpc_atomic_fixed4_shl_cpt
    __kmpc_atomic_fixed4_shl_cpt_rev
    __kmpc_atomic_fixed4_shl_rev
    __kmpc_atomic_fixed4_shr
    __kmpc_atomic_fixed4_shr_cpt
    __kmpc_atomic_fixed4_shr_cpt_rev
    __kmpc_atomic_fixed4_shr_rev
    __kmpc_atomic_fixed4_sub
    __kmpc_atomic_fixed4_sub_cpt
    __kmpc_atomic_fixed4_sub_cpt_rev
    __kmpc_atomic_fixed4_sub_fp
    __kmpc_atomic_fixed4_sub_rev
    __kmpc_atomic_fixed4_swp
    __kmpc_atomic_fixed4_wr
    __kmpc_atomic_fixed4_xor
    __kmpc_atomic_fixed4_xor_cpt
    __kmpc_atomic_fixed4u_add_fp
    __kmpc_atomic_fixed4u_sub_fp
    __kmpc_atomic_fixed4u_mul_fp
    __kmpc_atomic_fixed4u_div
    __kmpc_atomic_fixed4u_div_cpt
    __kmpc_atomic_fixed4u_div_cpt_rev
    __kmpc_atomic_fixed4u_div_fp
    __kmpc_atomic_fixed4u_div_rev
    __kmpc_atomic_fixed4u_shr
    __kmpc_atomic_fixed4u_shr_cpt
    __kmpc_atomic_fixed4u_shr_cpt_rev
    __kmpc_atomic_fixed4u_shr_rev
    __kmpc_atomic_fixed8_add
    __kmpc_atomic_fixed8_add_cpt
    __kmpc_atomic_fixed8_add_fp
    __kmpc_atomic_fixed8_andb
    __kmpc_atomic_fixed8_andb_cpt
    __kmpc_atomic_fixed8_andl
    __kmpc_atomic_fixed8_andl_cpt
    __kmpc_atomic_fixed8_div
    __kmpc_atomic_fixed8_div_cpt
    __kmpc_atomic_fixed8_div_cpt_rev
    __kmpc_atomic_fixed8_div_float8
    __kmpc_atomic_fixed8_div_fp
    __kmpc_atomic_fixed8_div_rev
    __kmpc_atomic_fixed8_eqv
    __kmpc_atomic_fixed8_eqv_cpt
    __kmpc_atomic_fixed8_max
    __kmpc_atomic_fixed8_max_cpt
    __kmpc_atomic_fixed8_min
    __kmpc_atomic_fixed8_min_cpt
    __kmpc_atomic_fixed8_mul
    __kmpc_atomic_fixed8_mul_cpt
    __kmpc_atomic_fixed8_mul_float8
    __kmpc_atomic_fixed8_mul_fp
    __kmpc_atomic_fixed8_neqv
    __kmpc_atomic_fixed8_neqv_cpt
    __kmpc_atomic_fixed8_orb
    __kmpc_atomic_fixed8_orb_cpt
    __kmpc_atomic_fixed8_orl
    __kmpc_atomic_fixed8_orl_cpt
    __kmpc_atomic_fixed8_rd
    __kmpc_atomic_fixed8_shl
    __kmpc_atomic_fixed8_shl_cpt
    __kmpc_atomic_fixed8_shl_cpt_rev
    __kmpc_atomic_fixed8_shl_rev
    __kmpc_atomic_fixed8_shr
    __kmpc_atomic_fixed8_shr_cpt
    __kmpc_atomic_fixed8_shr_cpt_rev
    __kmpc_atomic_fixed8_shr_rev
    __kmpc_atomic_fixed8_sub
    __kmpc_atomic_fixed8_sub_cpt
    __kmpc_atomic_fixed8_sub_cpt_rev
    __kmpc_atomic_fixed8_sub_fp
    __kmpc_atomic_fixed8_sub_rev
    __kmpc_atomic_fixed8_swp
    __kmpc_atomic_fixed8_wr
    __kmpc_atomic_fixed8_xor
    __kmpc_atomic_fixed8_xor_cpt
    __kmpc_atomic_fixed8u_add_fp
    __kmpc_atomic_fixed8u_sub_fp
    __kmpc_atomic_fixed8u_mul_fp
    __kmpc_atomic_fixed8u_div
    __kmpc_atomic_fixed8u_div_cpt
    __kmpc_atomic_fixed8u_div_cpt_rev
    __kmpc_atomic_fixed8u_div_fp
    __kmpc_atomic_fixed8u_div_rev
    __kmpc_atomic_fixed8u_shr
    __kmpc_atomic_fixed8u_shr_cpt
    __kmpc_atomic_fixed8u_shr_cpt_rev
    __kmpc_atomic_fixed8u_shr_rev
@endcode

Functions for floating point
----------------------------
There are versions here for floating point numbers of size 4, 8, 10 and 16
bytes. (Ten byte floats are used by X87, but are now rare).
@code
    __kmpc_atomic_float4_add
    __kmpc_atomic_float4_add_cpt
    __kmpc_atomic_float4_add_float8
    __kmpc_atomic_float4_add_fp
    __kmpc_atomic_float4_div
    __kmpc_atomic_float4_div_cpt
    __kmpc_atomic_float4_div_cpt_rev
    __kmpc_atomic_float4_div_float8
    __kmpc_atomic_float4_div_fp
    __kmpc_atomic_float4_div_rev
    __kmpc_atomic_float4_max
    __kmpc_atomic_float4_max_cpt
    __kmpc_atomic_float4_min
    __kmpc_atomic_float4_min_cpt
    __kmpc_atomic_float4_mul
    __kmpc_atomic_float4_mul_cpt
    __kmpc_atomic_float4_mul_float8
    __kmpc_atomic_float4_mul_fp
    __kmpc_atomic_float4_rd
    __kmpc_atomic_float4_sub
    __kmpc_atomic_float4_sub_cpt
    __kmpc_atomic_float4_sub_cpt_rev
    __kmpc_atomic_float4_sub_float8
    __kmpc_atomic_float4_sub_fp
    __kmpc_atomic_float4_sub_rev
    __kmpc_atomic_float4_swp
    __kmpc_atomic_float4_wr
    __kmpc_atomic_float8_add
    __kmpc_atomic_float8_add_cpt
    __kmpc_atomic_float8_add_fp
    __kmpc_atomic_float8_div
    __kmpc_atomic_float8_div_cpt
    __kmpc_atomic_float8_div_cpt_rev
    __kmpc_atomic_float8_div_fp
    __kmpc_atomic_float8_div_rev
    __kmpc_atomic_float8_max
    __kmpc_atomic_float8_max_cpt
    __kmpc_atomic_float8_min
    __kmpc_atomic_float8_min_cpt
    __kmpc_atomic_float8_mul
    __kmpc_atomic_float8_mul_cpt
    __kmpc_atomic_float8_mul_fp
    __kmpc_atomic_float8_rd
    __kmpc_atomic_float8_sub
    __kmpc_atomic_float8_sub_cpt
    __kmpc_atomic_float8_sub_cpt_rev
    __kmpc_atomic_float8_sub_fp
    __kmpc_atomic_float8_sub_rev
    __kmpc_atomic_float8_swp
    __kmpc_atomic_float8_wr
    __kmpc_atomic_float10_add
    __kmpc_atomic_float10_add_cpt
    __kmpc_atomic_float10_add_fp
    __kmpc_atomic_float10_div
    __kmpc_atomic_float10_div_cpt
    __kmpc_atomic_float10_div_cpt_rev
    __kmpc_atomic_float10_div_fp
    __kmpc_atomic_float10_div_rev
    __kmpc_atomic_float10_mul
    __kmpc_atomic_float10_mul_cpt
    __kmpc_atomic_float10_mul_fp
    __kmpc_atomic_float10_rd
    __kmpc_atomic_float10_sub
    __kmpc_atomic_float10_sub_cpt
    __kmpc_atomic_float10_sub_cpt_rev
    __kmpc_atomic_float10_sub_fp
    __kmpc_atomic_float10_sub_rev
    __kmpc_atomic_float10_swp
    __kmpc_atomic_float10_wr
    __kmpc_atomic_float16_add
    __kmpc_atomic_float16_add_cpt
    __kmpc_atomic_float16_div
    __kmpc_atomic_float16_div_cpt
    __kmpc_atomic_float16_div_cpt_rev
    __kmpc_atomic_float16_div_rev
    __kmpc_atomic_float16_max
    __kmpc_atomic_float16_max_cpt
    __kmpc_atomic_float16_min
    __kmpc_atomic_float16_min_cpt
    __kmpc_atomic_float16_mul
    __kmpc_atomic_float16_mul_cpt
    __kmpc_atomic_float16_rd
    __kmpc_atomic_float16_sub
    __kmpc_atomic_float16_sub_cpt
    __kmpc_atomic_float16_sub_cpt_rev
    __kmpc_atomic_float16_sub_rev
    __kmpc_atomic_float16_swp
    __kmpc_atomic_float16_wr
@endcode

Functions for Complex types
---------------------------
Functions for complex types whose component floating point variables are of size
4,8,10 or 16 bytes. The names here are based on the size of the component float,
*not* the size of the complex type. So `__kmpc_atomc_cmplx8_add` is an operation
on a `complex<double>` or `complex(kind=8)`, *not* `complex<float>`.

@code
    __kmpc_atomic_cmplx4_add
    __kmpc_atomic_cmplx4_add_cmplx8
    __kmpc_atomic_cmplx4_add_cpt
    __kmpc_atomic_cmplx4_div
    __kmpc_atomic_cmplx4_div_cmplx8
    __kmpc_atomic_cmplx4_div_cpt
    __kmpc_atomic_cmplx4_div_cpt_rev
    __kmpc_atomic_cmplx4_div_rev
    __kmpc_atomic_cmplx4_mul
    __kmpc_atomic_cmplx4_mul_cmplx8
    __kmpc_atomic_cmplx4_mul_cpt
    __kmpc_atomic_cmplx4_rd
    __kmpc_atomic_cmplx4_sub
    __kmpc_atomic_cmplx4_sub_cmplx8
    __kmpc_atomic_cmplx4_sub_cpt
    __kmpc_atomic_cmplx4_sub_cpt_rev
    __kmpc_atomic_cmplx4_sub_rev
    __kmpc_atomic_cmplx4_swp
    __kmpc_atomic_cmplx4_wr
    __kmpc_atomic_cmplx8_add
    __kmpc_atomic_cmplx8_add_cpt
    __kmpc_atomic_cmplx8_div
    __kmpc_atomic_cmplx8_div_cpt
    __kmpc_atomic_cmplx8_div_cpt_rev
    __kmpc_atomic_cmplx8_div_rev
    __kmpc_atomic_cmplx8_mul
    __kmpc_atomic_cmplx8_mul_cpt
    __kmpc_atomic_cmplx8_rd
    __kmpc_atomic_cmplx8_sub
    __kmpc_atomic_cmplx8_sub_cpt
    __kmpc_atomic_cmplx8_sub_cpt_rev
    __kmpc_atomic_cmplx8_sub_rev
    __kmpc_atomic_cmplx8_swp
    __kmpc_atomic_cmplx8_wr
    __kmpc_atomic_cmplx10_add
    __kmpc_atomic_cmplx10_add_cpt
    __kmpc_atomic_cmplx10_div
    __kmpc_atomic_cmplx10_div_cpt
    __kmpc_atomic_cmplx10_div_cpt_rev
    __kmpc_atomic_cmplx10_div_rev
    __kmpc_atomic_cmplx10_mul
    __kmpc_atomic_cmplx10_mul_cpt
    __kmpc_atomic_cmplx10_rd
    __kmpc_atomic_cmplx10_sub
    __kmpc_atomic_cmplx10_sub_cpt
    __kmpc_atomic_cmplx10_sub_cpt_rev
    __kmpc_atomic_cmplx10_sub_rev
    __kmpc_atomic_cmplx10_swp
    __kmpc_atomic_cmplx10_wr
    __kmpc_atomic_cmplx16_add
    __kmpc_atomic_cmplx16_add_cpt
    __kmpc_atomic_cmplx16_div
    __kmpc_atomic_cmplx16_div_cpt
    __kmpc_atomic_cmplx16_div_cpt_rev
    __kmpc_atomic_cmplx16_div_rev
    __kmpc_atomic_cmplx16_mul
    __kmpc_atomic_cmplx16_mul_cpt
    __kmpc_atomic_cmplx16_rd
    __kmpc_atomic_cmplx16_sub
    __kmpc_atomic_cmplx16_sub_cpt
    __kmpc_atomic_cmplx16_sub_cpt_rev
    __kmpc_atomic_cmplx16_swp
    __kmpc_atomic_cmplx16_wr
@endcode
*/

/*!
@ingroup ATOMIC_OPS
@{
*/

/*
 * Global vars
 */

#ifndef KMP_GOMP_COMPAT
int __kmp_atomic_mode = 1; // Intel perf
#else
int __kmp_atomic_mode = 2; // GOMP compatibility
#endif /* KMP_GOMP_COMPAT */

KMP_ALIGN(128)

// Control access to all user coded atomics in Gnu compat mode
kmp_atomic_lock_t __kmp_atomic_lock;
// Control access to all user coded atomics for 1-byte fixed data types
kmp_atomic_lock_t __kmp_atomic_lock_1i;
// Control access to all user coded atomics for 2-byte fixed data types
kmp_atomic_lock_t __kmp_atomic_lock_2i;
// Control access to all user coded atomics for 4-byte fixed data types
kmp_atomic_lock_t __kmp_atomic_lock_4i;
// Control access to all user coded atomics for kmp_real32 data type
kmp_atomic_lock_t __kmp_atomic_lock_4r;
// Control access to all user coded atomics for 8-byte fixed data types
kmp_atomic_lock_t __kmp_atomic_lock_8i;
// Control access to all user coded atomics for kmp_real64 data type
kmp_atomic_lock_t __kmp_atomic_lock_8r;
// Control access to all user coded atomics for complex byte data type
kmp_atomic_lock_t __kmp_atomic_lock_8c;
// Control access to all user coded atomics for long double data type
kmp_atomic_lock_t __kmp_atomic_lock_10r;
// Control access to all user coded atomics for _Quad data type
kmp_atomic_lock_t __kmp_atomic_lock_16r;
// Control access to all user coded atomics for double complex data type
kmp_atomic_lock_t __kmp_atomic_lock_16c;
// Control access to all user coded atomics for long double complex type
kmp_atomic_lock_t __kmp_atomic_lock_20c;
// Control access to all user coded atomics for _Quad complex data type
kmp_atomic_lock_t __kmp_atomic_lock_32c;

/* 2007-03-02:
   Without "volatile" specifier in OP_CMPXCHG and MIN_MAX_CMPXCHG we have a bug
   on *_32 and *_32e. This is just a temporary workaround for the problem. It
   seems the right solution is writing OP_CMPXCHG and MIN_MAX_CMPXCHG routines
   in assembler language. */
#define KMP_ATOMIC_VOLATILE volatile

#if (KMP_ARCH_X86) && KMP_HAVE_QUAD

static inline void operator+=(Quad_a4_t &lhs, Quad_a4_t &rhs) {
  lhs.q += rhs.q;
}
static inline void operator-=(Quad_a4_t &lhs, Quad_a4_t &rhs) {
  lhs.q -= rhs.q;
}
static inline void operator*=(Quad_a4_t &lhs, Quad_a4_t &rhs) {
  lhs.q *= rhs.q;
}
static inline void operator/=(Quad_a4_t &lhs, Quad_a4_t &rhs) {
  lhs.q /= rhs.q;
}
static inline bool operator<(Quad_a4_t &lhs, Quad_a4_t &rhs) {
  return lhs.q < rhs.q;
}
static inline bool operator>(Quad_a4_t &lhs, Quad_a4_t &rhs) {
  return lhs.q > rhs.q;
}

static inline void operator+=(Quad_a16_t &lhs, Quad_a16_t &rhs) {
  lhs.q += rhs.q;
}
static inline void operator-=(Quad_a16_t &lhs, Quad_a16_t &rhs) {
  lhs.q -= rhs.q;
}
static inline void operator*=(Quad_a16_t &lhs, Quad_a16_t &rhs) {
  lhs.q *= rhs.q;
}
static inline void operator/=(Quad_a16_t &lhs, Quad_a16_t &rhs) {
  lhs.q /= rhs.q;
}
static inline bool operator<(Quad_a16_t &lhs, Quad_a16_t &rhs) {
  return lhs.q < rhs.q;
}
static inline bool operator>(Quad_a16_t &lhs, Quad_a16_t &rhs) {
  return lhs.q > rhs.q;
}

static inline void operator+=(kmp_cmplx128_a4_t &lhs, kmp_cmplx128_a4_t &rhs) {
  lhs.q += rhs.q;
}
static inline void operator-=(kmp_cmplx128_a4_t &lhs, kmp_cmplx128_a4_t &rhs) {
  lhs.q -= rhs.q;
}
static inline void operator*=(kmp_cmplx128_a4_t &lhs, kmp_cmplx128_a4_t &rhs) {
  lhs.q *= rhs.q;
}
static inline void operator/=(kmp_cmplx128_a4_t &lhs, kmp_cmplx128_a4_t &rhs) {
  lhs.q /= rhs.q;
}

static inline void operator+=(kmp_cmplx128_a16_t &lhs,
                              kmp_cmplx128_a16_t &rhs) {
  lhs.q += rhs.q;
}
static inline void operator-=(kmp_cmplx128_a16_t &lhs,
                              kmp_cmplx128_a16_t &rhs) {
  lhs.q -= rhs.q;
}
static inline void operator*=(kmp_cmplx128_a16_t &lhs,
                              kmp_cmplx128_a16_t &rhs) {
  lhs.q *= rhs.q;
}
static inline void operator/=(kmp_cmplx128_a16_t &lhs,
                              kmp_cmplx128_a16_t &rhs) {
  lhs.q /= rhs.q;
}

#endif

// ATOMIC implementation routines -----------------------------------------
// One routine for each operation and operand type.
// All routines declarations looks like
// void __kmpc_atomic_RTYPE_OP( ident_t*, int, TYPE *lhs, TYPE rhs );

#define KMP_CHECK_GTID                                                         \
  if (gtid == KMP_GTID_UNKNOWN) {                                              \
    gtid = __kmp_entry_gtid();                                                 \
  } // check and get gtid when needed

// Beginning of a definition (provides name, parameters, gebug trace)
//     TYPE_ID - operands type and size (fixed*, fixed*u for signed, unsigned
//     fixed)
//     OP_ID   - operation identifier (add, sub, mul, ...)
//     TYPE    - operands' type
#define ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, RET_TYPE)                           \
  RET_TYPE __kmpc_atomic_##TYPE_ID##_##OP_ID(ident_t *id_ref, int gtid,        \
                                             TYPE *lhs, TYPE rhs) {            \
    KMP_DEBUG_ASSERT(__kmp_init_serial);                                       \
    KA_TRACE(100, ("__kmpc_atomic_" #TYPE_ID "_" #OP_ID ": T#%d\n", gtid));

// ------------------------------------------------------------------------
// Lock variables used for critical sections for various size operands
#define ATOMIC_LOCK0 __kmp_atomic_lock // all types, for Gnu compat
#define ATOMIC_LOCK1i __kmp_atomic_lock_1i // char
#define ATOMIC_LOCK2i __kmp_atomic_lock_2i // short
#define ATOMIC_LOCK4i __kmp_atomic_lock_4i // long int
#define ATOMIC_LOCK4r __kmp_atomic_lock_4r // float
#define ATOMIC_LOCK8i __kmp_atomic_lock_8i // long long int
#define ATOMIC_LOCK8r __kmp_atomic_lock_8r // double
#define ATOMIC_LOCK8c __kmp_atomic_lock_8c // float complex
#define ATOMIC_LOCK10r __kmp_atomic_lock_10r // long double
#define ATOMIC_LOCK16r __kmp_atomic_lock_16r // _Quad
#define ATOMIC_LOCK16c __kmp_atomic_lock_16c // double complex
#define ATOMIC_LOCK20c __kmp_atomic_lock_20c // long double complex
#define ATOMIC_LOCK32c __kmp_atomic_lock_32c // _Quad complex

// ------------------------------------------------------------------------
// Operation on *lhs, rhs bound by critical section
//     OP     - operator (it's supposed to contain an assignment)
//     LCK_ID - lock identifier
// Note: don't check gtid as it should always be valid
// 1, 2-byte - expect valid parameter, other - check before this macro
#define OP_CRITICAL(OP, LCK_ID)                                                \
  __kmp_acquire_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);                       \
                                                                               \
  (*lhs) OP(rhs);                                                              \
                                                                               \
  __kmp_release_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);

// ------------------------------------------------------------------------
// For GNU compatibility, we may need to use a critical section,
// even though it is not required by the ISA.
//
// On IA-32 architecture, all atomic operations except for fixed 4 byte add,
// sub, and bitwise logical ops, and 1 & 2 byte logical ops use a common
// critical section.  On Intel(R) 64, all atomic operations are done with fetch
// and add or compare and exchange.  Therefore, the FLAG parameter to this
// macro is either KMP_ARCH_X86 or 0 (or 1, for Intel-specific extension which
// require a critical section, where we predict that they will be implemented
// in the Gnu codegen by calling GOMP_atomic_start() / GOMP_atomic_end()).
//
// When the OP_GOMP_CRITICAL macro is used in a *CRITICAL* macro construct,
// the FLAG parameter should always be 1.  If we know that we will be using
// a critical section, then we want to make certain that we use the generic
// lock __kmp_atomic_lock to protect the atomic update, and not of of the
// locks that are specialized based upon the size or type of the data.
//
// If FLAG is 0, then we are relying on dead code elimination by the build
// compiler to get rid of the useless block of code, and save a needless
// branch at runtime.

#ifdef KMP_GOMP_COMPAT
#define OP_GOMP_CRITICAL(OP, FLAG)                                             \
  if ((FLAG) && (__kmp_atomic_mode == 2)) {                                    \
    KMP_CHECK_GTID;                                                            \
    OP_CRITICAL(OP, 0);                                                        \
    return;                                                                    \
  }
#else
#define OP_GOMP_CRITICAL(OP, FLAG)
#endif /* KMP_GOMP_COMPAT */

#if KMP_MIC
#define KMP_DO_PAUSE _mm_delay_32(1)
#else
#define KMP_DO_PAUSE KMP_CPU_PAUSE()
#endif /* KMP_MIC */

// ------------------------------------------------------------------------
// Operation on *lhs, rhs using "compare_and_store" routine
//     TYPE    - operands' type
//     BITS    - size in bits, used to distinguish low level calls
//     OP      - operator
#define OP_CMPXCHG(TYPE, BITS, OP)                                             \
  {                                                                            \
    TYPE old_value, new_value;                                                 \
    old_value = *(TYPE volatile *)lhs;                                         \
    new_value = old_value OP rhs;                                              \
    while (!KMP_COMPARE_AND_STORE_ACQ##BITS(                                   \
        (kmp_int##BITS *)lhs, *VOLATILE_CAST(kmp_int##BITS *) & old_value,     \
        *VOLATILE_CAST(kmp_int##BITS *) & new_value)) {                        \
      KMP_DO_PAUSE;                                                            \
                                                                               \
      old_value = *(TYPE volatile *)lhs;                                       \
      new_value = old_value OP rhs;                                            \
    }                                                                          \
  }

#if USE_CMPXCHG_FIX
// 2007-06-25:
// workaround for C78287 (complex(kind=4) data type). lin_32, lin_32e, win_32
// and win_32e are affected (I verified the asm). Compiler ignores the volatile
// qualifier of the temp_val in the OP_CMPXCHG macro. This is a problem of the
// compiler. Related tracker is C76005, targeted to 11.0. I verified the asm of
// the workaround.
#define OP_CMPXCHG_WORKAROUND(TYPE, BITS, OP)                                  \
  {                                                                            \
    struct _sss {                                                              \
      TYPE cmp;                                                                \
      kmp_int##BITS *vvv;                                                      \
    };                                                                         \
    struct _sss old_value, new_value;                                          \
    old_value.vvv = (kmp_int##BITS *)&old_value.cmp;                           \
    new_value.vvv = (kmp_int##BITS *)&new_value.cmp;                           \
    *old_value.vvv = *(volatile kmp_int##BITS *)lhs;                           \
    new_value.cmp = old_value.cmp OP rhs;                                      \
    while (!KMP_COMPARE_AND_STORE_ACQ##BITS(                                   \
        (kmp_int##BITS *)lhs, *VOLATILE_CAST(kmp_int##BITS *) old_value.vvv,   \
        *VOLATILE_CAST(kmp_int##BITS *) new_value.vvv)) {                      \
      KMP_DO_PAUSE;                                                            \
                                                                               \
      *old_value.vvv = *(volatile kmp_int##BITS *)lhs;                         \
      new_value.cmp = old_value.cmp OP rhs;                                    \
    }                                                                          \
  }
// end of the first part of the workaround for C78287
#endif // USE_CMPXCHG_FIX

#if KMP_ARCH_X86 || KMP_ARCH_X86_64

// ------------------------------------------------------------------------
// X86 or X86_64: no alignment problems ====================================
#define ATOMIC_FIXED_ADD(TYPE_ID, OP_ID, TYPE, BITS, OP, LCK_ID, MASK,         \
                         GOMP_FLAG)                                            \
  ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, void)                                     \
  OP_GOMP_CRITICAL(OP## =, GOMP_FLAG)                                          \
  /* OP used as a sign for subtraction: (lhs-rhs) --> (lhs+-rhs) */            \
  KMP_TEST_THEN_ADD##BITS(lhs, OP rhs);                                        \
  }
// -------------------------------------------------------------------------
#define ATOMIC_CMPXCHG(TYPE_ID, OP_ID, TYPE, BITS, OP, LCK_ID, MASK,           \
                       GOMP_FLAG)                                              \
  ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, void)                                     \
  OP_GOMP_CRITICAL(OP## =, GOMP_FLAG)                                          \
  OP_CMPXCHG(TYPE, BITS, OP)                                                   \
  }
#if USE_CMPXCHG_FIX
// -------------------------------------------------------------------------
// workaround for C78287 (complex(kind=4) data type)
#define ATOMIC_CMPXCHG_WORKAROUND(TYPE_ID, OP_ID, TYPE, BITS, OP, LCK_ID,      \
                                  MASK, GOMP_FLAG)                             \
  ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, void)                                     \
  OP_GOMP_CRITICAL(OP## =, GOMP_FLAG)                                          \
  OP_CMPXCHG_WORKAROUND(TYPE, BITS, OP)                                        \
  }
// end of the second part of the workaround for C78287
#endif

#else
// -------------------------------------------------------------------------
// Code for other architectures that don't handle unaligned accesses.
#define ATOMIC_FIXED_ADD(TYPE_ID, OP_ID, TYPE, BITS, OP, LCK_ID, MASK,         \
                         GOMP_FLAG)                                            \
  ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, void)                                     \
  OP_GOMP_CRITICAL(OP## =, GOMP_FLAG)                                          \
  if (!((kmp_uintptr_t)lhs & 0x##MASK)) {                                      \
    /* OP used as a sign for subtraction: (lhs-rhs) --> (lhs+-rhs) */          \
    KMP_TEST_THEN_ADD##BITS(lhs, OP rhs);                                      \
  } else {                                                                     \
    KMP_CHECK_GTID;                                                            \
    OP_CRITICAL(OP## =, LCK_ID) /* unaligned address - use critical */         \
  }                                                                            \
  }
// -------------------------------------------------------------------------
#define ATOMIC_CMPXCHG(TYPE_ID, OP_ID, TYPE, BITS, OP, LCK_ID, MASK,           \
                       GOMP_FLAG)                                              \
  ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, void)                                     \
  OP_GOMP_CRITICAL(OP## =, GOMP_FLAG)                                          \
  if (!((kmp_uintptr_t)lhs & 0x##MASK)) {                                      \
    OP_CMPXCHG(TYPE, BITS, OP) /* aligned address */                           \
  } else {                                                                     \
    KMP_CHECK_GTID;                                                            \
    OP_CRITICAL(OP## =, LCK_ID) /* unaligned address - use critical */         \
  }                                                                            \
  }
#if USE_CMPXCHG_FIX
// -------------------------------------------------------------------------
// workaround for C78287 (complex(kind=4) data type)
#define ATOMIC_CMPXCHG_WORKAROUND(TYPE_ID, OP_ID, TYPE, BITS, OP, LCK_ID,      \
                                  MASK, GOMP_FLAG)                             \
  ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, void)                                     \
  OP_GOMP_CRITICAL(OP## =, GOMP_FLAG)                                          \
  if (!((kmp_uintptr_t)lhs & 0x##MASK)) {                                      \
    OP_CMPXCHG(TYPE, BITS, OP) /* aligned address */                           \
  } else {                                                                     \
    KMP_CHECK_GTID;                                                            \
    OP_CRITICAL(OP## =, LCK_ID) /* unaligned address - use critical */         \
  }                                                                            \
  }
// end of the second part of the workaround for C78287
#endif // USE_CMPXCHG_FIX
#endif /* KMP_ARCH_X86 || KMP_ARCH_X86_64 */

// Routines for ATOMIC 4-byte operands addition and subtraction
ATOMIC_FIXED_ADD(fixed4, add, kmp_int32, 32, +, 4i, 3,
                 0) // __kmpc_atomic_fixed4_add
ATOMIC_FIXED_ADD(fixed4, sub, kmp_int32, 32, -, 4i, 3,
                 0) // __kmpc_atomic_fixed4_sub

ATOMIC_CMPXCHG(float4, add, kmp_real32, 32, +, 4r, 3,
               KMP_ARCH_X86) // __kmpc_atomic_float4_add
ATOMIC_CMPXCHG(float4, sub, kmp_real32, 32, -, 4r, 3,
               KMP_ARCH_X86) // __kmpc_atomic_float4_sub

// Routines for ATOMIC 8-byte operands addition and subtraction
ATOMIC_FIXED_ADD(fixed8, add, kmp_int64, 64, +, 8i, 7,
                 KMP_ARCH_X86) // __kmpc_atomic_fixed8_add
ATOMIC_FIXED_ADD(fixed8, sub, kmp_int64, 64, -, 8i, 7,
                 KMP_ARCH_X86) // __kmpc_atomic_fixed8_sub

ATOMIC_CMPXCHG(float8, add, kmp_real64, 64, +, 8r, 7,
               KMP_ARCH_X86) // __kmpc_atomic_float8_add
ATOMIC_CMPXCHG(float8, sub, kmp_real64, 64, -, 8r, 7,
               KMP_ARCH_X86) // __kmpc_atomic_float8_sub

// ------------------------------------------------------------------------
// Entries definition for integer operands
//     TYPE_ID - operands type and size (fixed4, float4)
//     OP_ID   - operation identifier (add, sub, mul, ...)
//     TYPE    - operand type
//     BITS    - size in bits, used to distinguish low level calls
//     OP      - operator (used in critical section)
//     LCK_ID  - lock identifier, used to possibly distinguish lock variable
//     MASK    - used for alignment check

//               TYPE_ID,OP_ID,  TYPE,   BITS,OP,LCK_ID,MASK,GOMP_FLAG
// ------------------------------------------------------------------------
// Routines for ATOMIC integer operands, other operators
// ------------------------------------------------------------------------
//              TYPE_ID,OP_ID, TYPE,          OP, LCK_ID, GOMP_FLAG
ATOMIC_CMPXCHG(fixed1, add, kmp_int8, 8, +, 1i, 0,
               KMP_ARCH_X86) // __kmpc_atomic_fixed1_add
ATOMIC_CMPXCHG(fixed1, andb, kmp_int8, 8, &, 1i, 0,
               0) // __kmpc_atomic_fixed1_andb
ATOMIC_CMPXCHG(fixed1, div, kmp_int8, 8, /, 1i, 0,
               KMP_ARCH_X86) // __kmpc_atomic_fixed1_div
ATOMIC_CMPXCHG(fixed1u, div, kmp_uint8, 8, /, 1i, 0,
               KMP_ARCH_X86) // __kmpc_atomic_fixed1u_div
ATOMIC_CMPXCHG(fixed1, mul, kmp_int8, 8, *, 1i, 0,
               KMP_ARCH_X86) // __kmpc_atomic_fixed1_mul
ATOMIC_CMPXCHG(fixed1, orb, kmp_int8, 8, |, 1i, 0,
               0) // __kmpc_atomic_fixed1_orb
ATOMIC_CMPXCHG(fixed1, shl, kmp_int8, 8, <<, 1i, 0,
               KMP_ARCH_X86) // __kmpc_atomic_fixed1_shl
ATOMIC_CMPXCHG(fixed1, shr, kmp_int8, 8, >>, 1i, 0,
               KMP_ARCH_X86) // __kmpc_atomic_fixed1_shr
ATOMIC_CMPXCHG(fixed1u, shr, kmp_uint8, 8, >>, 1i, 0,
               KMP_ARCH_X86) // __kmpc_atomic_fixed1u_shr
ATOMIC_CMPXCHG(fixed1, sub, kmp_int8, 8, -, 1i, 0,
               KMP_ARCH_X86) // __kmpc_atomic_fixed1_sub
ATOMIC_CMPXCHG(fixed1, xor, kmp_int8, 8, ^, 1i, 0,
               0) // __kmpc_atomic_fixed1_xor
ATOMIC_CMPXCHG(fixed2, add, kmp_int16, 16, +, 2i, 1,
               KMP_ARCH_X86) // __kmpc_atomic_fixed2_add
ATOMIC_CMPXCHG(fixed2, andb, kmp_int16, 16, &, 2i, 1,
               0) // __kmpc_atomic_fixed2_andb
ATOMIC_CMPXCHG(fixed2, div, kmp_int16, 16, /, 2i, 1,
               KMP_ARCH_X86) // __kmpc_atomic_fixed2_div
ATOMIC_CMPXCHG(fixed2u, div, kmp_uint16, 16, /, 2i, 1,
               KMP_ARCH_X86) // __kmpc_atomic_fixed2u_div
ATOMIC_CMPXCHG(fixed2, mul, kmp_int16, 16, *, 2i, 1,
               KMP_ARCH_X86) // __kmpc_atomic_fixed2_mul
ATOMIC_CMPXCHG(fixed2, orb, kmp_int16, 16, |, 2i, 1,
               0) // __kmpc_atomic_fixed2_orb
ATOMIC_CMPXCHG(fixed2, shl, kmp_int16, 16, <<, 2i, 1,
               KMP_ARCH_X86) // __kmpc_atomic_fixed2_shl
ATOMIC_CMPXCHG(fixed2, shr, kmp_int16, 16, >>, 2i, 1,
               KMP_ARCH_X86) // __kmpc_atomic_fixed2_shr
ATOMIC_CMPXCHG(fixed2u, shr, kmp_uint16, 16, >>, 2i, 1,
               KMP_ARCH_X86) // __kmpc_atomic_fixed2u_shr
ATOMIC_CMPXCHG(fixed2, sub, kmp_int16, 16, -, 2i, 1,
               KMP_ARCH_X86) // __kmpc_atomic_fixed2_sub
ATOMIC_CMPXCHG(fixed2, xor, kmp_int16, 16, ^, 2i, 1,
               0) // __kmpc_atomic_fixed2_xor
ATOMIC_CMPXCHG(fixed4, andb, kmp_int32, 32, &, 4i, 3,
               0) // __kmpc_atomic_fixed4_andb
ATOMIC_CMPXCHG(fixed4, div, kmp_int32, 32, /, 4i, 3,
               KMP_ARCH_X86) // __kmpc_atomic_fixed4_div
ATOMIC_CMPXCHG(fixed4u, div, kmp_uint32, 32, /, 4i, 3,
               KMP_ARCH_X86) // __kmpc_atomic_fixed4u_div
ATOMIC_CMPXCHG(fixed4, mul, kmp_int32, 32, *, 4i, 3,
               KMP_ARCH_X86) // __kmpc_atomic_fixed4_mul
ATOMIC_CMPXCHG(fixed4, orb, kmp_int32, 32, |, 4i, 3,
               0) // __kmpc_atomic_fixed4_orb
ATOMIC_CMPXCHG(fixed4, shl, kmp_int32, 32, <<, 4i, 3,
               KMP_ARCH_X86) // __kmpc_atomic_fixed4_shl
ATOMIC_CMPXCHG(fixed4, shr, kmp_int32, 32, >>, 4i, 3,
               KMP_ARCH_X86) // __kmpc_atomic_fixed4_shr
ATOMIC_CMPXCHG(fixed4u, shr, kmp_uint32, 32, >>, 4i, 3,
               KMP_ARCH_X86) // __kmpc_atomic_fixed4u_shr
ATOMIC_CMPXCHG(fixed4, xor, kmp_int32, 32, ^, 4i, 3,
               0) // __kmpc_atomic_fixed4_xor
ATOMIC_CMPXCHG(fixed8, andb, kmp_int64, 64, &, 8i, 7,
               KMP_ARCH_X86) // __kmpc_atomic_fixed8_andb
ATOMIC_CMPXCHG(fixed8, div, kmp_int64, 64, /, 8i, 7,
               KMP_ARCH_X86) // __kmpc_atomic_fixed8_div
ATOMIC_CMPXCHG(fixed8u, div, kmp_uint64, 64, /, 8i, 7,
               KMP_ARCH_X86) // __kmpc_atomic_fixed8u_div
ATOMIC_CMPXCHG(fixed8, mul, kmp_int64, 64, *, 8i, 7,
               KMP_ARCH_X86) // __kmpc_atomic_fixed8_mul
ATOMIC_CMPXCHG(fixed8, orb, kmp_int64, 64, |, 8i, 7,
               KMP_ARCH_X86) // __kmpc_atomic_fixed8_orb
ATOMIC_CMPXCHG(fixed8, shl, kmp_int64, 64, <<, 8i, 7,
               KMP_ARCH_X86) // __kmpc_atomic_fixed8_shl
ATOMIC_CMPXCHG(fixed8, shr, kmp_int64, 64, >>, 8i, 7,
               KMP_ARCH_X86) // __kmpc_atomic_fixed8_shr
ATOMIC_CMPXCHG(fixed8u, shr, kmp_uint64, 64, >>, 8i, 7,
               KMP_ARCH_X86) // __kmpc_atomic_fixed8u_shr
ATOMIC_CMPXCHG(fixed8, xor, kmp_int64, 64, ^, 8i, 7,
               KMP_ARCH_X86) // __kmpc_atomic_fixed8_xor
ATOMIC_CMPXCHG(float4, div, kmp_real32, 32, /, 4r, 3,
               KMP_ARCH_X86) // __kmpc_atomic_float4_div
ATOMIC_CMPXCHG(float4, mul, kmp_real32, 32, *, 4r, 3,
               KMP_ARCH_X86) // __kmpc_atomic_float4_mul
ATOMIC_CMPXCHG(float8, div, kmp_real64, 64, /, 8r, 7,
               KMP_ARCH_X86) // __kmpc_atomic_float8_div
ATOMIC_CMPXCHG(float8, mul, kmp_real64, 64, *, 8r, 7,
               KMP_ARCH_X86) // __kmpc_atomic_float8_mul
//              TYPE_ID,OP_ID, TYPE,          OP, LCK_ID, GOMP_FLAG

/* ------------------------------------------------------------------------ */
/* Routines for C/C++ Reduction operators && and ||                         */

// ------------------------------------------------------------------------
// Need separate macros for &&, || because there is no combined assignment
//   TODO: eliminate ATOMIC_CRIT_{L,EQV} macros as not used
#define ATOMIC_CRIT_L(TYPE_ID, OP_ID, TYPE, OP, LCK_ID, GOMP_FLAG)             \
  ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, void)                                     \
  OP_GOMP_CRITICAL(= *lhs OP, GOMP_FLAG)                                       \
  OP_CRITICAL(= *lhs OP, LCK_ID)                                               \
  }

#if KMP_ARCH_X86 || KMP_ARCH_X86_64

// ------------------------------------------------------------------------
// X86 or X86_64: no alignment problems ===================================
#define ATOMIC_CMPX_L(TYPE_ID, OP_ID, TYPE, BITS, OP, LCK_ID, MASK, GOMP_FLAG) \
  ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, void)                                     \
  OP_GOMP_CRITICAL(= *lhs OP, GOMP_FLAG)                                       \
  OP_CMPXCHG(TYPE, BITS, OP)                                                   \
  }

#else
// ------------------------------------------------------------------------
// Code for other architectures that don't handle unaligned accesses.
#define ATOMIC_CMPX_L(TYPE_ID, OP_ID, TYPE, BITS, OP, LCK_ID, MASK, GOMP_FLAG) \
  ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, void)                                     \
  OP_GOMP_CRITICAL(= *lhs OP, GOMP_FLAG)                                       \
  if (!((kmp_uintptr_t)lhs & 0x##MASK)) {                                      \
    OP_CMPXCHG(TYPE, BITS, OP) /* aligned address */                           \
  } else {                                                                     \
    KMP_CHECK_GTID;                                                            \
    OP_CRITICAL(= *lhs OP, LCK_ID) /* unaligned - use critical */              \
  }                                                                            \
  }
#endif /* KMP_ARCH_X86 || KMP_ARCH_X86_64 */

ATOMIC_CMPX_L(fixed1, andl, char, 8, &&, 1i, 0,
              KMP_ARCH_X86) // __kmpc_atomic_fixed1_andl
ATOMIC_CMPX_L(fixed1, orl, char, 8, ||, 1i, 0,
              KMP_ARCH_X86) // __kmpc_atomic_fixed1_orl
ATOMIC_CMPX_L(fixed2, andl, short, 16, &&, 2i, 1,
              KMP_ARCH_X86) // __kmpc_atomic_fixed2_andl
ATOMIC_CMPX_L(fixed2, orl, short, 16, ||, 2i, 1,
              KMP_ARCH_X86) // __kmpc_atomic_fixed2_orl
ATOMIC_CMPX_L(fixed4, andl, kmp_int32, 32, &&, 4i, 3,
              0) // __kmpc_atomic_fixed4_andl
ATOMIC_CMPX_L(fixed4, orl, kmp_int32, 32, ||, 4i, 3,
              0) // __kmpc_atomic_fixed4_orl
ATOMIC_CMPX_L(fixed8, andl, kmp_int64, 64, &&, 8i, 7,
              KMP_ARCH_X86) // __kmpc_atomic_fixed8_andl
ATOMIC_CMPX_L(fixed8, orl, kmp_int64, 64, ||, 8i, 7,
              KMP_ARCH_X86) // __kmpc_atomic_fixed8_orl

/* ------------------------------------------------------------------------- */
/* Routines for Fortran operators that matched no one in C:                  */
/* MAX, MIN, .EQV., .NEQV.                                                   */
/* Operators .AND., .OR. are covered by __kmpc_atomic_*_{andl,orl}           */
/* Intrinsics IAND, IOR, IEOR are covered by __kmpc_atomic_*_{andb,orb,xor}  */

// -------------------------------------------------------------------------
// MIN and MAX need separate macros
// OP - operator to check if we need any actions?
#define MIN_MAX_CRITSECT(OP, LCK_ID)                                           \
  __kmp_acquire_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);                       \
                                                                               \
  if (*lhs OP rhs) { /* still need actions? */                                 \
    *lhs = rhs;                                                                \
  }                                                                            \
  __kmp_release_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);

// -------------------------------------------------------------------------
#ifdef KMP_GOMP_COMPAT
#define GOMP_MIN_MAX_CRITSECT(OP, FLAG)                                        \
  if ((FLAG) && (__kmp_atomic_mode == 2)) {                                    \
    KMP_CHECK_GTID;                                                            \
    MIN_MAX_CRITSECT(OP, 0);                                                   \
    return;                                                                    \
  }
#else
#define GOMP_MIN_MAX_CRITSECT(OP, FLAG)
#endif /* KMP_GOMP_COMPAT */

// -------------------------------------------------------------------------
#define MIN_MAX_CMPXCHG(TYPE, BITS, OP)                                        \
  {                                                                            \
    TYPE KMP_ATOMIC_VOLATILE temp_val;                                         \
    TYPE old_value;                                                            \
    temp_val = *lhs;                                                           \
    old_value = temp_val;                                                      \
    while (old_value OP rhs && /* still need actions? */                       \
           !KMP_COMPARE_AND_STORE_ACQ##BITS(                                   \
               (kmp_int##BITS *)lhs,                                           \
               *VOLATILE_CAST(kmp_int##BITS *) & old_value,                    \
               *VOLATILE_CAST(kmp_int##BITS *) & rhs)) {                       \
      KMP_CPU_PAUSE();                                                         \
      temp_val = *lhs;                                                         \
      old_value = temp_val;                                                    \
    }                                                                          \
  }

// -------------------------------------------------------------------------
// 1-byte, 2-byte operands - use critical section
#define MIN_MAX_CRITICAL(TYPE_ID, OP_ID, TYPE, OP, LCK_ID, GOMP_FLAG)          \
  ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, void)                                     \
  if (*lhs OP rhs) { /* need actions? */                                       \
    GOMP_MIN_MAX_CRITSECT(OP, GOMP_FLAG)                                       \
    MIN_MAX_CRITSECT(OP, LCK_ID)                                               \
  }                                                                            \
  }

#if KMP_ARCH_X86 || KMP_ARCH_X86_64

// -------------------------------------------------------------------------
// X86 or X86_64: no alignment problems ====================================
#define MIN_MAX_COMPXCHG(TYPE_ID, OP_ID, TYPE, BITS, OP, LCK_ID, MASK,         \
                         GOMP_FLAG)                                            \
  ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, void)                                     \
  if (*lhs OP rhs) {                                                           \
    GOMP_MIN_MAX_CRITSECT(OP, GOMP_FLAG)                                       \
    MIN_MAX_CMPXCHG(TYPE, BITS, OP)                                            \
  }                                                                            \
  }

#else
// -------------------------------------------------------------------------
// Code for other architectures that don't handle unaligned accesses.
#define MIN_MAX_COMPXCHG(TYPE_ID, OP_ID, TYPE, BITS, OP, LCK_ID, MASK,         \
                         GOMP_FLAG)                                            \
  ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, void)                                     \
  if (*lhs OP rhs) {                                                           \
    GOMP_MIN_MAX_CRITSECT(OP, GOMP_FLAG)                                       \
    if (!((kmp_uintptr_t)lhs & 0x##MASK)) {                                    \
      MIN_MAX_CMPXCHG(TYPE, BITS, OP) /* aligned address */                    \
    } else {                                                                   \
      KMP_CHECK_GTID;                                                          \
      MIN_MAX_CRITSECT(OP, LCK_ID) /* unaligned address */                     \
    }                                                                          \
  }                                                                            \
  }
#endif /* KMP_ARCH_X86 || KMP_ARCH_X86_64 */

MIN_MAX_COMPXCHG(fixed1, max, char, 8, <, 1i, 0,
                 KMP_ARCH_X86) // __kmpc_atomic_fixed1_max
MIN_MAX_COMPXCHG(fixed1, min, char, 8, >, 1i, 0,
                 KMP_ARCH_X86) // __kmpc_atomic_fixed1_min
MIN_MAX_COMPXCHG(fixed2, max, short, 16, <, 2i, 1,
                 KMP_ARCH_X86) // __kmpc_atomic_fixed2_max
MIN_MAX_COMPXCHG(fixed2, min, short, 16, >, 2i, 1,
                 KMP_ARCH_X86) // __kmpc_atomic_fixed2_min
MIN_MAX_COMPXCHG(fixed4, max, kmp_int32, 32, <, 4i, 3,
                 0) // __kmpc_atomic_fixed4_max
MIN_MAX_COMPXCHG(fixed4, min, kmp_int32, 32, >, 4i, 3,
                 0) // __kmpc_atomic_fixed4_min
MIN_MAX_COMPXCHG(fixed8, max, kmp_int64, 64, <, 8i, 7,
                 KMP_ARCH_X86) // __kmpc_atomic_fixed8_max
MIN_MAX_COMPXCHG(fixed8, min, kmp_int64, 64, >, 8i, 7,
                 KMP_ARCH_X86) // __kmpc_atomic_fixed8_min
MIN_MAX_COMPXCHG(float4, max, kmp_real32, 32, <, 4r, 3,
                 KMP_ARCH_X86) // __kmpc_atomic_float4_max
MIN_MAX_COMPXCHG(float4, min, kmp_real32, 32, >, 4r, 3,
                 KMP_ARCH_X86) // __kmpc_atomic_float4_min
MIN_MAX_COMPXCHG(float8, max, kmp_real64, 64, <, 8r, 7,
                 KMP_ARCH_X86) // __kmpc_atomic_float8_max
MIN_MAX_COMPXCHG(float8, min, kmp_real64, 64, >, 8r, 7,
                 KMP_ARCH_X86) // __kmpc_atomic_float8_min
#if KMP_HAVE_QUAD
MIN_MAX_CRITICAL(float16, max, QUAD_LEGACY, <, 16r,
                 1) // __kmpc_atomic_float16_max
MIN_MAX_CRITICAL(float16, min, QUAD_LEGACY, >, 16r,
                 1) // __kmpc_atomic_float16_min
#if (KMP_ARCH_X86)
MIN_MAX_CRITICAL(float16, max_a16, Quad_a16_t, <, 16r,
                 1) // __kmpc_atomic_float16_max_a16
MIN_MAX_CRITICAL(float16, min_a16, Quad_a16_t, >, 16r,
                 1) // __kmpc_atomic_float16_min_a16
#endif
#endif
// ------------------------------------------------------------------------
// Need separate macros for .EQV. because of the need of complement (~)
// OP ignored for critical sections, ^=~ used instead
#define ATOMIC_CRIT_EQV(TYPE_ID, OP_ID, TYPE, OP, LCK_ID, GOMP_FLAG)           \
  ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, void)                                     \
  OP_GOMP_CRITICAL(^= ~, GOMP_FLAG) /* send assignment */                      \
  OP_CRITICAL(^= ~, LCK_ID) /* send assignment and complement */               \
  }

// ------------------------------------------------------------------------
#if KMP_ARCH_X86 || KMP_ARCH_X86_64
// ------------------------------------------------------------------------
// X86 or X86_64: no alignment problems ===================================
#define ATOMIC_CMPX_EQV(TYPE_ID, OP_ID, TYPE, BITS, OP, LCK_ID, MASK,          \
                        GOMP_FLAG)                                             \
  ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, void)                                     \
  OP_GOMP_CRITICAL(^= ~, GOMP_FLAG) /* send assignment */                      \
  OP_CMPXCHG(TYPE, BITS, OP)                                                   \
  }
// ------------------------------------------------------------------------
#else
// ------------------------------------------------------------------------
// Code for other architectures that don't handle unaligned accesses.
#define ATOMIC_CMPX_EQV(TYPE_ID, OP_ID, TYPE, BITS, OP, LCK_ID, MASK,          \
                        GOMP_FLAG)                                             \
  ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, void)                                     \
  OP_GOMP_CRITICAL(^= ~, GOMP_FLAG)                                            \
  if (!((kmp_uintptr_t)lhs & 0x##MASK)) {                                      \
    OP_CMPXCHG(TYPE, BITS, OP) /* aligned address */                           \
  } else {                                                                     \
    KMP_CHECK_GTID;                                                            \
    OP_CRITICAL(^= ~, LCK_ID) /* unaligned address - use critical */           \
  }                                                                            \
  }
#endif /* KMP_ARCH_X86 || KMP_ARCH_X86_64 */

ATOMIC_CMPXCHG(fixed1, neqv, kmp_int8, 8, ^, 1i, 0,
               KMP_ARCH_X86) // __kmpc_atomic_fixed1_neqv
ATOMIC_CMPXCHG(fixed2, neqv, kmp_int16, 16, ^, 2i, 1,
               KMP_ARCH_X86) // __kmpc_atomic_fixed2_neqv
ATOMIC_CMPXCHG(fixed4, neqv, kmp_int32, 32, ^, 4i, 3,
               KMP_ARCH_X86) // __kmpc_atomic_fixed4_neqv
ATOMIC_CMPXCHG(fixed8, neqv, kmp_int64, 64, ^, 8i, 7,
               KMP_ARCH_X86) // __kmpc_atomic_fixed8_neqv
ATOMIC_CMPX_EQV(fixed1, eqv, kmp_int8, 8, ^~, 1i, 0,
                KMP_ARCH_X86) // __kmpc_atomic_fixed1_eqv
ATOMIC_CMPX_EQV(fixed2, eqv, kmp_int16, 16, ^~, 2i, 1,
                KMP_ARCH_X86) // __kmpc_atomic_fixed2_eqv
ATOMIC_CMPX_EQV(fixed4, eqv, kmp_int32, 32, ^~, 4i, 3,
                KMP_ARCH_X86) // __kmpc_atomic_fixed4_eqv
ATOMIC_CMPX_EQV(fixed8, eqv, kmp_int64, 64, ^~, 8i, 7,
                KMP_ARCH_X86) // __kmpc_atomic_fixed8_eqv

// ------------------------------------------------------------------------
// Routines for Extended types: long double, _Quad, complex flavours (use
// critical section)
//     TYPE_ID, OP_ID, TYPE - detailed above
//     OP      - operator
//     LCK_ID  - lock identifier, used to possibly distinguish lock variable
#define ATOMIC_CRITICAL(TYPE_ID, OP_ID, TYPE, OP, LCK_ID, GOMP_FLAG)           \
  ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, void)                                     \
  OP_GOMP_CRITICAL(OP## =, GOMP_FLAG) /* send assignment */                    \
  OP_CRITICAL(OP## =, LCK_ID) /* send assignment */                            \
  }

/* ------------------------------------------------------------------------- */
// routines for long double type
ATOMIC_CRITICAL(float10, add, long double, +, 10r,
                1) // __kmpc_atomic_float10_add
ATOMIC_CRITICAL(float10, sub, long double, -, 10r,
                1) // __kmpc_atomic_float10_sub
ATOMIC_CRITICAL(float10, mul, long double, *, 10r,
                1) // __kmpc_atomic_float10_mul
ATOMIC_CRITICAL(float10, div, long double, /, 10r,
                1) // __kmpc_atomic_float10_div
#if KMP_HAVE_QUAD
// routines for _Quad type
ATOMIC_CRITICAL(float16, add, QUAD_LEGACY, +, 16r,
                1) // __kmpc_atomic_float16_add
ATOMIC_CRITICAL(float16, sub, QUAD_LEGACY, -, 16r,
                1) // __kmpc_atomic_float16_sub
ATOMIC_CRITICAL(float16, mul, QUAD_LEGACY, *, 16r,
                1) // __kmpc_atomic_float16_mul
ATOMIC_CRITICAL(float16, div, QUAD_LEGACY, /, 16r,
                1) // __kmpc_atomic_float16_div
#if (KMP_ARCH_X86)
ATOMIC_CRITICAL(float16, add_a16, Quad_a16_t, +, 16r,
                1) // __kmpc_atomic_float16_add_a16
ATOMIC_CRITICAL(float16, sub_a16, Quad_a16_t, -, 16r,
                1) // __kmpc_atomic_float16_sub_a16
ATOMIC_CRITICAL(float16, mul_a16, Quad_a16_t, *, 16r,
                1) // __kmpc_atomic_float16_mul_a16
ATOMIC_CRITICAL(float16, div_a16, Quad_a16_t, /, 16r,
                1) // __kmpc_atomic_float16_div_a16
#endif
#endif
// routines for complex types

#if USE_CMPXCHG_FIX
// workaround for C78287 (complex(kind=4) data type)
ATOMIC_CMPXCHG_WORKAROUND(cmplx4, add, kmp_cmplx32, 64, +, 8c, 7,
                          1) // __kmpc_atomic_cmplx4_add
ATOMIC_CMPXCHG_WORKAROUND(cmplx4, sub, kmp_cmplx32, 64, -, 8c, 7,
                          1) // __kmpc_atomic_cmplx4_sub
ATOMIC_CMPXCHG_WORKAROUND(cmplx4, mul, kmp_cmplx32, 64, *, 8c, 7,
                          1) // __kmpc_atomic_cmplx4_mul
ATOMIC_CMPXCHG_WORKAROUND(cmplx4, div, kmp_cmplx32, 64, /, 8c, 7,
                          1) // __kmpc_atomic_cmplx4_div
// end of the workaround for C78287
#else
ATOMIC_CRITICAL(cmplx4, add, kmp_cmplx32, +, 8c, 1) // __kmpc_atomic_cmplx4_add
ATOMIC_CRITICAL(cmplx4, sub, kmp_cmplx32, -, 8c, 1) // __kmpc_atomic_cmplx4_sub
ATOMIC_CRITICAL(cmplx4, mul, kmp_cmplx32, *, 8c, 1) // __kmpc_atomic_cmplx4_mul
ATOMIC_CRITICAL(cmplx4, div, kmp_cmplx32, /, 8c, 1) // __kmpc_atomic_cmplx4_div
#endif // USE_CMPXCHG_FIX

ATOMIC_CRITICAL(cmplx8, add, kmp_cmplx64, +, 16c, 1) // __kmpc_atomic_cmplx8_add
ATOMIC_CRITICAL(cmplx8, sub, kmp_cmplx64, -, 16c, 1) // __kmpc_atomic_cmplx8_sub
ATOMIC_CRITICAL(cmplx8, mul, kmp_cmplx64, *, 16c, 1) // __kmpc_atomic_cmplx8_mul
ATOMIC_CRITICAL(cmplx8, div, kmp_cmplx64, /, 16c, 1) // __kmpc_atomic_cmplx8_div
ATOMIC_CRITICAL(cmplx10, add, kmp_cmplx80, +, 20c,
                1) // __kmpc_atomic_cmplx10_add
ATOMIC_CRITICAL(cmplx10, sub, kmp_cmplx80, -, 20c,
                1) // __kmpc_atomic_cmplx10_sub
ATOMIC_CRITICAL(cmplx10, mul, kmp_cmplx80, *, 20c,
                1) // __kmpc_atomic_cmplx10_mul
ATOMIC_CRITICAL(cmplx10, div, kmp_cmplx80, /, 20c,
                1) // __kmpc_atomic_cmplx10_div
#if KMP_HAVE_QUAD
ATOMIC_CRITICAL(cmplx16, add, CPLX128_LEG, +, 32c,
                1) // __kmpc_atomic_cmplx16_add
ATOMIC_CRITICAL(cmplx16, sub, CPLX128_LEG, -, 32c,
                1) // __kmpc_atomic_cmplx16_sub
ATOMIC_CRITICAL(cmplx16, mul, CPLX128_LEG, *, 32c,
                1) // __kmpc_atomic_cmplx16_mul
ATOMIC_CRITICAL(cmplx16, div, CPLX128_LEG, /, 32c,
                1) // __kmpc_atomic_cmplx16_div
#if (KMP_ARCH_X86)
ATOMIC_CRITICAL(cmplx16, add_a16, kmp_cmplx128_a16_t, +, 32c,
                1) // __kmpc_atomic_cmplx16_add_a16
ATOMIC_CRITICAL(cmplx16, sub_a16, kmp_cmplx128_a16_t, -, 32c,
                1) // __kmpc_atomic_cmplx16_sub_a16
ATOMIC_CRITICAL(cmplx16, mul_a16, kmp_cmplx128_a16_t, *, 32c,
                1) // __kmpc_atomic_cmplx16_mul_a16
ATOMIC_CRITICAL(cmplx16, div_a16, kmp_cmplx128_a16_t, /, 32c,
                1) // __kmpc_atomic_cmplx16_div_a16
#endif
#endif

#if OMP_40_ENABLED

// OpenMP 4.0: x = expr binop x for non-commutative operations.
// Supported only on IA-32 architecture and Intel(R) 64
#if KMP_ARCH_X86 || KMP_ARCH_X86_64

// ------------------------------------------------------------------------
// Operation on *lhs, rhs bound by critical section
//     OP     - operator (it's supposed to contain an assignment)
//     LCK_ID - lock identifier
// Note: don't check gtid as it should always be valid
// 1, 2-byte - expect valid parameter, other - check before this macro
#define OP_CRITICAL_REV(OP, LCK_ID)                                            \
  __kmp_acquire_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);                       \
                                                                               \
  (*lhs) = (rhs)OP(*lhs);                                                      \
                                                                               \
  __kmp_release_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);

#ifdef KMP_GOMP_COMPAT
#define OP_GOMP_CRITICAL_REV(OP, FLAG)                                         \
  if ((FLAG) && (__kmp_atomic_mode == 2)) {                                    \
    KMP_CHECK_GTID;                                                            \
    OP_CRITICAL_REV(OP, 0);                                                    \
    return;                                                                    \
  }
#else
#define OP_GOMP_CRITICAL_REV(OP, FLAG)
#endif /* KMP_GOMP_COMPAT */

// Beginning of a definition (provides name, parameters, gebug trace)
//     TYPE_ID - operands type and size (fixed*, fixed*u for signed, unsigned
//     fixed)
//     OP_ID   - operation identifier (add, sub, mul, ...)
//     TYPE    - operands' type
#define ATOMIC_BEGIN_REV(TYPE_ID, OP_ID, TYPE, RET_TYPE)                       \
  RET_TYPE __kmpc_atomic_##TYPE_ID##_##OP_ID##_rev(ident_t *id_ref, int gtid,  \
                                                   TYPE *lhs, TYPE rhs) {      \
    KMP_DEBUG_ASSERT(__kmp_init_serial);                                       \
    KA_TRACE(100, ("__kmpc_atomic_" #TYPE_ID "_" #OP_ID "_rev: T#%d\n", gtid));

// ------------------------------------------------------------------------
// Operation on *lhs, rhs using "compare_and_store" routine
//     TYPE    - operands' type
//     BITS    - size in bits, used to distinguish low level calls
//     OP      - operator
// Note: temp_val introduced in order to force the compiler to read
//       *lhs only once (w/o it the compiler reads *lhs twice)
#define OP_CMPXCHG_REV(TYPE, BITS, OP)                                         \
  {                                                                            \
    TYPE KMP_ATOMIC_VOLATILE temp_val;                                         \
    TYPE old_value, new_value;                                                 \
    temp_val = *lhs;                                                           \
    old_value = temp_val;                                                      \
    new_value = rhs OP old_value;                                              \
    while (!KMP_COMPARE_AND_STORE_ACQ##BITS(                                   \
        (kmp_int##BITS *)lhs, *VOLATILE_CAST(kmp_int##BITS *) & old_value,     \
        *VOLATILE_CAST(kmp_int##BITS *) & new_value)) {                        \
      KMP_DO_PAUSE;                                                            \
                                                                               \
      temp_val = *lhs;                                                         \
      old_value = temp_val;                                                    \
      new_value = rhs OP old_value;                                            \
    }                                                                          \
  }

// -------------------------------------------------------------------------
#define ATOMIC_CMPXCHG_REV(TYPE_ID, OP_ID, TYPE, BITS, OP, LCK_ID, GOMP_FLAG)  \
  ATOMIC_BEGIN_REV(TYPE_ID, OP_ID, TYPE, void)                                 \
  OP_GOMP_CRITICAL_REV(OP, GOMP_FLAG)                                          \
  OP_CMPXCHG_REV(TYPE, BITS, OP)                                               \
  }

// ------------------------------------------------------------------------
// Entries definition for integer operands
//     TYPE_ID - operands type and size (fixed4, float4)
//     OP_ID   - operation identifier (add, sub, mul, ...)
//     TYPE    - operand type
//     BITS    - size in bits, used to distinguish low level calls
//     OP      - operator (used in critical section)
//     LCK_ID  - lock identifier, used to possibly distinguish lock variable

//               TYPE_ID,OP_ID,  TYPE,   BITS,OP,LCK_ID,GOMP_FLAG
// ------------------------------------------------------------------------
// Routines for ATOMIC integer operands, other operators
// ------------------------------------------------------------------------
//                  TYPE_ID,OP_ID, TYPE,    BITS, OP, LCK_ID, GOMP_FLAG
ATOMIC_CMPXCHG_REV(fixed1, div, kmp_int8, 8, /, 1i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1_div_rev
ATOMIC_CMPXCHG_REV(fixed1u, div, kmp_uint8, 8, /, 1i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1u_div_rev
ATOMIC_CMPXCHG_REV(fixed1, shl, kmp_int8, 8, <<, 1i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1_shl_rev
ATOMIC_CMPXCHG_REV(fixed1, shr, kmp_int8, 8, >>, 1i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1_shr_rev
ATOMIC_CMPXCHG_REV(fixed1u, shr, kmp_uint8, 8, >>, 1i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1u_shr_rev
ATOMIC_CMPXCHG_REV(fixed1, sub, kmp_int8, 8, -, 1i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1_sub_rev

ATOMIC_CMPXCHG_REV(fixed2, div, kmp_int16, 16, /, 2i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2_div_rev
ATOMIC_CMPXCHG_REV(fixed2u, div, kmp_uint16, 16, /, 2i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2u_div_rev
ATOMIC_CMPXCHG_REV(fixed2, shl, kmp_int16, 16, <<, 2i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2_shl_rev
ATOMIC_CMPXCHG_REV(fixed2, shr, kmp_int16, 16, >>, 2i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2_shr_rev
ATOMIC_CMPXCHG_REV(fixed2u, shr, kmp_uint16, 16, >>, 2i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2u_shr_rev
ATOMIC_CMPXCHG_REV(fixed2, sub, kmp_int16, 16, -, 2i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2_sub_rev

ATOMIC_CMPXCHG_REV(fixed4, div, kmp_int32, 32, /, 4i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed4_div_rev
ATOMIC_CMPXCHG_REV(fixed4u, div, kmp_uint32, 32, /, 4i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed4u_div_rev
ATOMIC_CMPXCHG_REV(fixed4, shl, kmp_int32, 32, <<, 4i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed4_shl_rev
ATOMIC_CMPXCHG_REV(fixed4, shr, kmp_int32, 32, >>, 4i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed4_shr_rev
ATOMIC_CMPXCHG_REV(fixed4u, shr, kmp_uint32, 32, >>, 4i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed4u_shr_rev
ATOMIC_CMPXCHG_REV(fixed4, sub, kmp_int32, 32, -, 4i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed4_sub_rev

ATOMIC_CMPXCHG_REV(fixed8, div, kmp_int64, 64, /, 8i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8_div_rev
ATOMIC_CMPXCHG_REV(fixed8u, div, kmp_uint64, 64, /, 8i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8u_div_rev
ATOMIC_CMPXCHG_REV(fixed8, shl, kmp_int64, 64, <<, 8i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8_shl_rev
ATOMIC_CMPXCHG_REV(fixed8, shr, kmp_int64, 64, >>, 8i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8_shr_rev
ATOMIC_CMPXCHG_REV(fixed8u, shr, kmp_uint64, 64, >>, 8i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8u_shr_rev
ATOMIC_CMPXCHG_REV(fixed8, sub, kmp_int64, 64, -, 8i,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8_sub_rev

ATOMIC_CMPXCHG_REV(float4, div, kmp_real32, 32, /, 4r,
                   KMP_ARCH_X86) // __kmpc_atomic_float4_div_rev
ATOMIC_CMPXCHG_REV(float4, sub, kmp_real32, 32, -, 4r,
                   KMP_ARCH_X86) // __kmpc_atomic_float4_sub_rev

ATOMIC_CMPXCHG_REV(float8, div, kmp_real64, 64, /, 8r,
                   KMP_ARCH_X86) // __kmpc_atomic_float8_div_rev
ATOMIC_CMPXCHG_REV(float8, sub, kmp_real64, 64, -, 8r,
                   KMP_ARCH_X86) // __kmpc_atomic_float8_sub_rev
//                  TYPE_ID,OP_ID, TYPE,     BITS,OP,LCK_ID, GOMP_FLAG

// ------------------------------------------------------------------------
// Routines for Extended types: long double, _Quad, complex flavours (use
// critical section)
//     TYPE_ID, OP_ID, TYPE - detailed above
//     OP      - operator
//     LCK_ID  - lock identifier, used to possibly distinguish lock variable
#define ATOMIC_CRITICAL_REV(TYPE_ID, OP_ID, TYPE, OP, LCK_ID, GOMP_FLAG)       \
  ATOMIC_BEGIN_REV(TYPE_ID, OP_ID, TYPE, void)                                 \
  OP_GOMP_CRITICAL_REV(OP, GOMP_FLAG)                                          \
  OP_CRITICAL_REV(OP, LCK_ID)                                                  \
  }

/* ------------------------------------------------------------------------- */
// routines for long double type
ATOMIC_CRITICAL_REV(float10, sub, long double, -, 10r,
                    1) // __kmpc_atomic_float10_sub_rev
ATOMIC_CRITICAL_REV(float10, div, long double, /, 10r,
                    1) // __kmpc_atomic_float10_div_rev
#if KMP_HAVE_QUAD
// routines for _Quad type
ATOMIC_CRITICAL_REV(float16, sub, QUAD_LEGACY, -, 16r,
                    1) // __kmpc_atomic_float16_sub_rev
ATOMIC_CRITICAL_REV(float16, div, QUAD_LEGACY, /, 16r,
                    1) // __kmpc_atomic_float16_div_rev
#if (KMP_ARCH_X86)
ATOMIC_CRITICAL_REV(float16, sub_a16, Quad_a16_t, -, 16r,
                    1) // __kmpc_atomic_float16_sub_a16_rev
ATOMIC_CRITICAL_REV(float16, div_a16, Quad_a16_t, /, 16r,
                    1) // __kmpc_atomic_float16_div_a16_rev
#endif
#endif

// routines for complex types
ATOMIC_CRITICAL_REV(cmplx4, sub, kmp_cmplx32, -, 8c,
                    1) // __kmpc_atomic_cmplx4_sub_rev
ATOMIC_CRITICAL_REV(cmplx4, div, kmp_cmplx32, /, 8c,
                    1) // __kmpc_atomic_cmplx4_div_rev
ATOMIC_CRITICAL_REV(cmplx8, sub, kmp_cmplx64, -, 16c,
                    1) // __kmpc_atomic_cmplx8_sub_rev
ATOMIC_CRITICAL_REV(cmplx8, div, kmp_cmplx64, /, 16c,
                    1) // __kmpc_atomic_cmplx8_div_rev
ATOMIC_CRITICAL_REV(cmplx10, sub, kmp_cmplx80, -, 20c,
                    1) // __kmpc_atomic_cmplx10_sub_rev
ATOMIC_CRITICAL_REV(cmplx10, div, kmp_cmplx80, /, 20c,
                    1) // __kmpc_atomic_cmplx10_div_rev
#if KMP_HAVE_QUAD
ATOMIC_CRITICAL_REV(cmplx16, sub, CPLX128_LEG, -, 32c,
                    1) // __kmpc_atomic_cmplx16_sub_rev
ATOMIC_CRITICAL_REV(cmplx16, div, CPLX128_LEG, /, 32c,
                    1) // __kmpc_atomic_cmplx16_div_rev
#if (KMP_ARCH_X86)
ATOMIC_CRITICAL_REV(cmplx16, sub_a16, kmp_cmplx128_a16_t, -, 32c,
                    1) // __kmpc_atomic_cmplx16_sub_a16_rev
ATOMIC_CRITICAL_REV(cmplx16, div_a16, kmp_cmplx128_a16_t, /, 32c,
                    1) // __kmpc_atomic_cmplx16_div_a16_rev
#endif
#endif

#endif // KMP_ARCH_X86 || KMP_ARCH_X86_64
// End of OpenMP 4.0: x = expr binop x for non-commutative operations.

#endif // OMP_40_ENABLED

/* ------------------------------------------------------------------------ */
/* Routines for mixed types of LHS and RHS, when RHS is "larger"            */
/* Note: in order to reduce the total number of types combinations          */
/*       it is supposed that compiler converts RHS to longest floating type,*/
/*       that is _Quad, before call to any of these routines                */
/* Conversion to _Quad will be done by the compiler during calculation,     */
/*    conversion back to TYPE - before the assignment, like:                */
/*    *lhs = (TYPE)( (_Quad)(*lhs) OP rhs )                                 */
/* Performance penalty expected because of SW emulation use                 */
/* ------------------------------------------------------------------------ */

#define ATOMIC_BEGIN_MIX(TYPE_ID, TYPE, OP_ID, RTYPE_ID, RTYPE)                \
  void __kmpc_atomic_##TYPE_ID##_##OP_ID##_##RTYPE_ID(                         \
      ident_t *id_ref, int gtid, TYPE *lhs, RTYPE rhs) {                       \
    KMP_DEBUG_ASSERT(__kmp_init_serial);                                       \
    KA_TRACE(100,                                                              \
             ("__kmpc_atomic_" #TYPE_ID "_" #OP_ID "_" #RTYPE_ID ": T#%d\n",   \
              gtid));

// -------------------------------------------------------------------------
#define ATOMIC_CRITICAL_FP(TYPE_ID, TYPE, OP_ID, OP, RTYPE_ID, RTYPE, LCK_ID,  \
                           GOMP_FLAG)                                          \
  ATOMIC_BEGIN_MIX(TYPE_ID, TYPE, OP_ID, RTYPE_ID, RTYPE)                      \
  OP_GOMP_CRITICAL(OP## =, GOMP_FLAG) /* send assignment */                    \
  OP_CRITICAL(OP## =, LCK_ID) /* send assignment */                            \
  }

// -------------------------------------------------------------------------
#if KMP_ARCH_X86 || KMP_ARCH_X86_64
// -------------------------------------------------------------------------
// X86 or X86_64: no alignment problems ====================================
#define ATOMIC_CMPXCHG_MIX(TYPE_ID, TYPE, OP_ID, BITS, OP, RTYPE_ID, RTYPE,    \
                           LCK_ID, MASK, GOMP_FLAG)                            \
  ATOMIC_BEGIN_MIX(TYPE_ID, TYPE, OP_ID, RTYPE_ID, RTYPE)                      \
  OP_GOMP_CRITICAL(OP## =, GOMP_FLAG)                                          \
  OP_CMPXCHG(TYPE, BITS, OP)                                                   \
  }
// -------------------------------------------------------------------------
#else
// ------------------------------------------------------------------------
// Code for other architectures that don't handle unaligned accesses.
#define ATOMIC_CMPXCHG_MIX(TYPE_ID, TYPE, OP_ID, BITS, OP, RTYPE_ID, RTYPE,    \
                           LCK_ID, MASK, GOMP_FLAG)                            \
  ATOMIC_BEGIN_MIX(TYPE_ID, TYPE, OP_ID, RTYPE_ID, RTYPE)                      \
  OP_GOMP_CRITICAL(OP## =, GOMP_FLAG)                                          \
  if (!((kmp_uintptr_t)lhs & 0x##MASK)) {                                      \
    OP_CMPXCHG(TYPE, BITS, OP) /* aligned address */                           \
  } else {                                                                     \
    KMP_CHECK_GTID;                                                            \
    OP_CRITICAL(OP## =, LCK_ID) /* unaligned address - use critical */         \
  }                                                                            \
  }
#endif /* KMP_ARCH_X86 || KMP_ARCH_X86_64 */

// -------------------------------------------------------------------------
#if KMP_ARCH_X86 || KMP_ARCH_X86_64
// -------------------------------------------------------------------------
#define ATOMIC_CMPXCHG_REV_MIX(TYPE_ID, TYPE, OP_ID, BITS, OP, RTYPE_ID,       \
                               RTYPE, LCK_ID, MASK, GOMP_FLAG)                 \
  ATOMIC_BEGIN_MIX(TYPE_ID, TYPE, OP_ID, RTYPE_ID, RTYPE)                      \
  OP_GOMP_CRITICAL_REV(OP, GOMP_FLAG)                                          \
  OP_CMPXCHG_REV(TYPE, BITS, OP)                                               \
  }
#define ATOMIC_CRITICAL_REV_FP(TYPE_ID, TYPE, OP_ID, OP, RTYPE_ID, RTYPE,      \
                               LCK_ID, GOMP_FLAG)                              \
  ATOMIC_BEGIN_MIX(TYPE_ID, TYPE, OP_ID, RTYPE_ID, RTYPE)                      \
  OP_GOMP_CRITICAL_REV(OP, GOMP_FLAG)                                          \
  OP_CRITICAL_REV(OP, LCK_ID)                                                  \
  }
#endif /* KMP_ARCH_X86 || KMP_ARCH_X86_64 */

// RHS=float8
ATOMIC_CMPXCHG_MIX(fixed1, char, mul, 8, *, float8, kmp_real64, 1i, 0,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1_mul_float8
ATOMIC_CMPXCHG_MIX(fixed1, char, div, 8, /, float8, kmp_real64, 1i, 0,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1_div_float8
ATOMIC_CMPXCHG_MIX(fixed2, short, mul, 16, *, float8, kmp_real64, 2i, 1,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2_mul_float8
ATOMIC_CMPXCHG_MIX(fixed2, short, div, 16, /, float8, kmp_real64, 2i, 1,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2_div_float8
ATOMIC_CMPXCHG_MIX(fixed4, kmp_int32, mul, 32, *, float8, kmp_real64, 4i, 3,
                   0) // __kmpc_atomic_fixed4_mul_float8
ATOMIC_CMPXCHG_MIX(fixed4, kmp_int32, div, 32, /, float8, kmp_real64, 4i, 3,
                   0) // __kmpc_atomic_fixed4_div_float8
ATOMIC_CMPXCHG_MIX(fixed8, kmp_int64, mul, 64, *, float8, kmp_real64, 8i, 7,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8_mul_float8
ATOMIC_CMPXCHG_MIX(fixed8, kmp_int64, div, 64, /, float8, kmp_real64, 8i, 7,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8_div_float8
ATOMIC_CMPXCHG_MIX(float4, kmp_real32, add, 32, +, float8, kmp_real64, 4r, 3,
                   KMP_ARCH_X86) // __kmpc_atomic_float4_add_float8
ATOMIC_CMPXCHG_MIX(float4, kmp_real32, sub, 32, -, float8, kmp_real64, 4r, 3,
                   KMP_ARCH_X86) // __kmpc_atomic_float4_sub_float8
ATOMIC_CMPXCHG_MIX(float4, kmp_real32, mul, 32, *, float8, kmp_real64, 4r, 3,
                   KMP_ARCH_X86) // __kmpc_atomic_float4_mul_float8
ATOMIC_CMPXCHG_MIX(float4, kmp_real32, div, 32, /, float8, kmp_real64, 4r, 3,
                   KMP_ARCH_X86) // __kmpc_atomic_float4_div_float8

// RHS=float16 (deprecated, to be removed when we are sure the compiler does not
// use them)
#if KMP_HAVE_QUAD
ATOMIC_CMPXCHG_MIX(fixed1, char, add, 8, +, fp, _Quad, 1i, 0,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1_add_fp
ATOMIC_CMPXCHG_MIX(fixed1u, uchar, add, 8, +, fp, _Quad, 1i, 0,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1u_add_fp
ATOMIC_CMPXCHG_MIX(fixed1, char, sub, 8, -, fp, _Quad, 1i, 0,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1_sub_fp
ATOMIC_CMPXCHG_MIX(fixed1u, uchar, sub, 8, -, fp, _Quad, 1i, 0,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1u_sub_fp
ATOMIC_CMPXCHG_MIX(fixed1, char, mul, 8, *, fp, _Quad, 1i, 0,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1_mul_fp
ATOMIC_CMPXCHG_MIX(fixed1u, uchar, mul, 8, *, fp, _Quad, 1i, 0,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1u_mul_fp
ATOMIC_CMPXCHG_MIX(fixed1, char, div, 8, /, fp, _Quad, 1i, 0,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1_div_fp
ATOMIC_CMPXCHG_MIX(fixed1u, uchar, div, 8, /, fp, _Quad, 1i, 0,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1u_div_fp

ATOMIC_CMPXCHG_MIX(fixed2, short, add, 16, +, fp, _Quad, 2i, 1,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2_add_fp
ATOMIC_CMPXCHG_MIX(fixed2u, ushort, add, 16, +, fp, _Quad, 2i, 1,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2u_add_fp
ATOMIC_CMPXCHG_MIX(fixed2, short, sub, 16, -, fp, _Quad, 2i, 1,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2_sub_fp
ATOMIC_CMPXCHG_MIX(fixed2u, ushort, sub, 16, -, fp, _Quad, 2i, 1,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2u_sub_fp
ATOMIC_CMPXCHG_MIX(fixed2, short, mul, 16, *, fp, _Quad, 2i, 1,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2_mul_fp
ATOMIC_CMPXCHG_MIX(fixed2u, ushort, mul, 16, *, fp, _Quad, 2i, 1,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2u_mul_fp
ATOMIC_CMPXCHG_MIX(fixed2, short, div, 16, /, fp, _Quad, 2i, 1,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2_div_fp
ATOMIC_CMPXCHG_MIX(fixed2u, ushort, div, 16, /, fp, _Quad, 2i, 1,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2u_div_fp

ATOMIC_CMPXCHG_MIX(fixed4, kmp_int32, add, 32, +, fp, _Quad, 4i, 3,
                   0) // __kmpc_atomic_fixed4_add_fp
ATOMIC_CMPXCHG_MIX(fixed4u, kmp_uint32, add, 32, +, fp, _Quad, 4i, 3,
                   0) // __kmpc_atomic_fixed4u_add_fp
ATOMIC_CMPXCHG_MIX(fixed4, kmp_int32, sub, 32, -, fp, _Quad, 4i, 3,
                   0) // __kmpc_atomic_fixed4_sub_fp
ATOMIC_CMPXCHG_MIX(fixed4u, kmp_uint32, sub, 32, -, fp, _Quad, 4i, 3,
                   0) // __kmpc_atomic_fixed4u_sub_fp
ATOMIC_CMPXCHG_MIX(fixed4, kmp_int32, mul, 32, *, fp, _Quad, 4i, 3,
                   0) // __kmpc_atomic_fixed4_mul_fp
ATOMIC_CMPXCHG_MIX(fixed4u, kmp_uint32, mul, 32, *, fp, _Quad, 4i, 3,
                   0) // __kmpc_atomic_fixed4u_mul_fp
ATOMIC_CMPXCHG_MIX(fixed4, kmp_int32, div, 32, /, fp, _Quad, 4i, 3,
                   0) // __kmpc_atomic_fixed4_div_fp
ATOMIC_CMPXCHG_MIX(fixed4u, kmp_uint32, div, 32, /, fp, _Quad, 4i, 3,
                   0) // __kmpc_atomic_fixed4u_div_fp

ATOMIC_CMPXCHG_MIX(fixed8, kmp_int64, add, 64, +, fp, _Quad, 8i, 7,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8_add_fp
ATOMIC_CMPXCHG_MIX(fixed8u, kmp_uint64, add, 64, +, fp, _Quad, 8i, 7,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8u_add_fp
ATOMIC_CMPXCHG_MIX(fixed8, kmp_int64, sub, 64, -, fp, _Quad, 8i, 7,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8_sub_fp
ATOMIC_CMPXCHG_MIX(fixed8u, kmp_uint64, sub, 64, -, fp, _Quad, 8i, 7,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8u_sub_fp
ATOMIC_CMPXCHG_MIX(fixed8, kmp_int64, mul, 64, *, fp, _Quad, 8i, 7,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8_mul_fp
ATOMIC_CMPXCHG_MIX(fixed8u, kmp_uint64, mul, 64, *, fp, _Quad, 8i, 7,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8u_mul_fp
ATOMIC_CMPXCHG_MIX(fixed8, kmp_int64, div, 64, /, fp, _Quad, 8i, 7,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8_div_fp
ATOMIC_CMPXCHG_MIX(fixed8u, kmp_uint64, div, 64, /, fp, _Quad, 8i, 7,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8u_div_fp

ATOMIC_CMPXCHG_MIX(float4, kmp_real32, add, 32, +, fp, _Quad, 4r, 3,
                   KMP_ARCH_X86) // __kmpc_atomic_float4_add_fp
ATOMIC_CMPXCHG_MIX(float4, kmp_real32, sub, 32, -, fp, _Quad, 4r, 3,
                   KMP_ARCH_X86) // __kmpc_atomic_float4_sub_fp
ATOMIC_CMPXCHG_MIX(float4, kmp_real32, mul, 32, *, fp, _Quad, 4r, 3,
                   KMP_ARCH_X86) // __kmpc_atomic_float4_mul_fp
ATOMIC_CMPXCHG_MIX(float4, kmp_real32, div, 32, /, fp, _Quad, 4r, 3,
                   KMP_ARCH_X86) // __kmpc_atomic_float4_div_fp

ATOMIC_CMPXCHG_MIX(float8, kmp_real64, add, 64, +, fp, _Quad, 8r, 7,
                   KMP_ARCH_X86) // __kmpc_atomic_float8_add_fp
ATOMIC_CMPXCHG_MIX(float8, kmp_real64, sub, 64, -, fp, _Quad, 8r, 7,
                   KMP_ARCH_X86) // __kmpc_atomic_float8_sub_fp
ATOMIC_CMPXCHG_MIX(float8, kmp_real64, mul, 64, *, fp, _Quad, 8r, 7,
                   KMP_ARCH_X86) // __kmpc_atomic_float8_mul_fp
ATOMIC_CMPXCHG_MIX(float8, kmp_real64, div, 64, /, fp, _Quad, 8r, 7,
                   KMP_ARCH_X86) // __kmpc_atomic_float8_div_fp

ATOMIC_CRITICAL_FP(float10, long double, add, +, fp, _Quad, 10r,
                   1) // __kmpc_atomic_float10_add_fp
ATOMIC_CRITICAL_FP(float10, long double, sub, -, fp, _Quad, 10r,
                   1) // __kmpc_atomic_float10_sub_fp
ATOMIC_CRITICAL_FP(float10, long double, mul, *, fp, _Quad, 10r,
                   1) // __kmpc_atomic_float10_mul_fp
ATOMIC_CRITICAL_FP(float10, long double, div, /, fp, _Quad, 10r,
                   1) // __kmpc_atomic_float10_div_fp

#if KMP_ARCH_X86 || KMP_ARCH_X86_64
// Reverse operations
ATOMIC_CMPXCHG_REV_MIX(fixed1, char, sub_rev, 8, -, fp, _Quad, 1i, 0,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed1_sub_rev_fp
ATOMIC_CMPXCHG_REV_MIX(fixed1u, uchar, sub_rev, 8, -, fp, _Quad, 1i, 0,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed1u_sub_rev_fp
ATOMIC_CMPXCHG_REV_MIX(fixed1, char, div_rev, 8, /, fp, _Quad, 1i, 0,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed1_div_rev_fp
ATOMIC_CMPXCHG_REV_MIX(fixed1u, uchar, div_rev, 8, /, fp, _Quad, 1i, 0,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed1u_div_rev_fp

ATOMIC_CMPXCHG_REV_MIX(fixed2, short, sub_rev, 16, -, fp, _Quad, 2i, 1,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed2_sub_rev_fp
ATOMIC_CMPXCHG_REV_MIX(fixed2u, ushort, sub_rev, 16, -, fp, _Quad, 2i, 1,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed2u_sub_rev_fp
ATOMIC_CMPXCHG_REV_MIX(fixed2, short, div_rev, 16, /, fp, _Quad, 2i, 1,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed2_div_rev_fp
ATOMIC_CMPXCHG_REV_MIX(fixed2u, ushort, div_rev, 16, /, fp, _Quad, 2i, 1,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed2u_div_rev_fp

ATOMIC_CMPXCHG_REV_MIX(fixed4, kmp_int32, sub_rev, 32, -, fp, _Quad, 4i, 3,
                       0) // __kmpc_atomic_fixed4_sub_rev_fp
ATOMIC_CMPXCHG_REV_MIX(fixed4u, kmp_uint32, sub_rev, 32, -, fp, _Quad, 4i, 3,
                       0) // __kmpc_atomic_fixed4u_sub_rev_fp
ATOMIC_CMPXCHG_REV_MIX(fixed4, kmp_int32, div_rev, 32, /, fp, _Quad, 4i, 3,
                       0) // __kmpc_atomic_fixed4_div_rev_fp
ATOMIC_CMPXCHG_REV_MIX(fixed4u, kmp_uint32, div_rev, 32, /, fp, _Quad, 4i, 3,
                       0) // __kmpc_atomic_fixed4u_div_rev_fp

ATOMIC_CMPXCHG_REV_MIX(fixed8, kmp_int64, sub_rev, 64, -, fp, _Quad, 8i, 7,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed8_sub_rev_fp
ATOMIC_CMPXCHG_REV_MIX(fixed8u, kmp_uint64, sub_rev, 64, -, fp, _Quad, 8i, 7,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed8u_sub_rev_fp
ATOMIC_CMPXCHG_REV_MIX(fixed8, kmp_int64, div_rev, 64, /, fp, _Quad, 8i, 7,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed8_div_rev_fp
ATOMIC_CMPXCHG_REV_MIX(fixed8u, kmp_uint64, div_rev, 64, /, fp, _Quad, 8i, 7,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed8u_div_rev_fp

ATOMIC_CMPXCHG_REV_MIX(float4, kmp_real32, sub_rev, 32, -, fp, _Quad, 4r, 3,
                       KMP_ARCH_X86) // __kmpc_atomic_float4_sub_rev_fp
ATOMIC_CMPXCHG_REV_MIX(float4, kmp_real32, div_rev, 32, /, fp, _Quad, 4r, 3,
                       KMP_ARCH_X86) // __kmpc_atomic_float4_div_rev_fp

ATOMIC_CMPXCHG_REV_MIX(float8, kmp_real64, sub_rev, 64, -, fp, _Quad, 8r, 7,
                       KMP_ARCH_X86) // __kmpc_atomic_float8_sub_rev_fp
ATOMIC_CMPXCHG_REV_MIX(float8, kmp_real64, div_rev, 64, /, fp, _Quad, 8r, 7,
                       KMP_ARCH_X86) // __kmpc_atomic_float8_div_rev_fp

ATOMIC_CRITICAL_REV_FP(float10, long double, sub_rev, -, fp, _Quad, 10r,
                       1) // __kmpc_atomic_float10_sub_rev_fp
ATOMIC_CRITICAL_REV_FP(float10, long double, div_rev, /, fp, _Quad, 10r,
                       1) // __kmpc_atomic_float10_div_rev_fp
#endif /* KMP_ARCH_X86 || KMP_ARCH_X86_64 */

#endif

#if KMP_ARCH_X86 || KMP_ARCH_X86_64
// ------------------------------------------------------------------------
// X86 or X86_64: no alignment problems ====================================
#if USE_CMPXCHG_FIX
// workaround for C78287 (complex(kind=4) data type)
#define ATOMIC_CMPXCHG_CMPLX(TYPE_ID, TYPE, OP_ID, BITS, OP, RTYPE_ID, RTYPE,  \
                             LCK_ID, MASK, GOMP_FLAG)                          \
  ATOMIC_BEGIN_MIX(TYPE_ID, TYPE, OP_ID, RTYPE_ID, RTYPE)                      \
  OP_GOMP_CRITICAL(OP## =, GOMP_FLAG)                                          \
  OP_CMPXCHG_WORKAROUND(TYPE, BITS, OP)                                        \
  }
// end of the second part of the workaround for C78287
#else
#define ATOMIC_CMPXCHG_CMPLX(TYPE_ID, TYPE, OP_ID, BITS, OP, RTYPE_ID, RTYPE,  \
                             LCK_ID, MASK, GOMP_FLAG)                          \
  ATOMIC_BEGIN_MIX(TYPE_ID, TYPE, OP_ID, RTYPE_ID, RTYPE)                      \
  OP_GOMP_CRITICAL(OP## =, GOMP_FLAG)                                          \
  OP_CMPXCHG(TYPE, BITS, OP)                                                   \
  }
#endif // USE_CMPXCHG_FIX
#else
// ------------------------------------------------------------------------
// Code for other architectures that don't handle unaligned accesses.
#define ATOMIC_CMPXCHG_CMPLX(TYPE_ID, TYPE, OP_ID, BITS, OP, RTYPE_ID, RTYPE,  \
                             LCK_ID, MASK, GOMP_FLAG)                          \
  ATOMIC_BEGIN_MIX(TYPE_ID, TYPE, OP_ID, RTYPE_ID, RTYPE)                      \
  OP_GOMP_CRITICAL(OP## =, GOMP_FLAG)                                          \
  if (!((kmp_uintptr_t)lhs & 0x##MASK)) {                                      \
    OP_CMPXCHG(TYPE, BITS, OP) /* aligned address */                           \
  } else {                                                                     \
    KMP_CHECK_GTID;                                                            \
    OP_CRITICAL(OP## =, LCK_ID) /* unaligned address - use critical */         \
  }                                                                            \
  }
#endif /* KMP_ARCH_X86 || KMP_ARCH_X86_64 */

ATOMIC_CMPXCHG_CMPLX(cmplx4, kmp_cmplx32, add, 64, +, cmplx8, kmp_cmplx64, 8c,
                     7, KMP_ARCH_X86) // __kmpc_atomic_cmplx4_add_cmplx8
ATOMIC_CMPXCHG_CMPLX(cmplx4, kmp_cmplx32, sub, 64, -, cmplx8, kmp_cmplx64, 8c,
                     7, KMP_ARCH_X86) // __kmpc_atomic_cmplx4_sub_cmplx8
ATOMIC_CMPXCHG_CMPLX(cmplx4, kmp_cmplx32, mul, 64, *, cmplx8, kmp_cmplx64, 8c,
                     7, KMP_ARCH_X86) // __kmpc_atomic_cmplx4_mul_cmplx8
ATOMIC_CMPXCHG_CMPLX(cmplx4, kmp_cmplx32, div, 64, /, cmplx8, kmp_cmplx64, 8c,
                     7, KMP_ARCH_X86) // __kmpc_atomic_cmplx4_div_cmplx8

// READ, WRITE, CAPTURE are supported only on IA-32 architecture and Intel(R) 64
#if KMP_ARCH_X86 || KMP_ARCH_X86_64

// ------------------------------------------------------------------------
// Atomic READ routines

// ------------------------------------------------------------------------
// Beginning of a definition (provides name, parameters, gebug trace)
//     TYPE_ID - operands type and size (fixed*, fixed*u for signed, unsigned
//     fixed)
//     OP_ID   - operation identifier (add, sub, mul, ...)
//     TYPE    - operands' type
#define ATOMIC_BEGIN_READ(TYPE_ID, OP_ID, TYPE, RET_TYPE)                      \
  RET_TYPE __kmpc_atomic_##TYPE_ID##_##OP_ID(ident_t *id_ref, int gtid,        \
                                             TYPE *loc) {                      \
    KMP_DEBUG_ASSERT(__kmp_init_serial);                                       \
    KA_TRACE(100, ("__kmpc_atomic_" #TYPE_ID "_" #OP_ID ": T#%d\n", gtid));

// ------------------------------------------------------------------------
// Operation on *lhs, rhs using "compare_and_store_ret" routine
//     TYPE    - operands' type
//     BITS    - size in bits, used to distinguish low level calls
//     OP      - operator
// Note: temp_val introduced in order to force the compiler to read
//       *lhs only once (w/o it the compiler reads *lhs twice)
// TODO: check if it is still necessary
// Return old value regardless of the result of "compare & swap# operation
#define OP_CMPXCHG_READ(TYPE, BITS, OP)                                        \
  {                                                                            \
    TYPE KMP_ATOMIC_VOLATILE temp_val;                                         \
    union f_i_union {                                                          \
      TYPE f_val;                                                              \
      kmp_int##BITS i_val;                                                     \
    };                                                                         \
    union f_i_union old_value;                                                 \
    temp_val = *loc;                                                           \
    old_value.f_val = temp_val;                                                \
    old_value.i_val = KMP_COMPARE_AND_STORE_RET##BITS(                         \
        (kmp_int##BITS *)loc,                                                  \
        *VOLATILE_CAST(kmp_int##BITS *) & old_value.i_val,                     \
        *VOLATILE_CAST(kmp_int##BITS *) & old_value.i_val);                    \
    new_value = old_value.f_val;                                               \
    return new_value;                                                          \
  }

// -------------------------------------------------------------------------
// Operation on *lhs, rhs bound by critical section
//     OP     - operator (it's supposed to contain an assignment)
//     LCK_ID - lock identifier
// Note: don't check gtid as it should always be valid
// 1, 2-byte - expect valid parameter, other - check before this macro
#define OP_CRITICAL_READ(OP, LCK_ID)                                           \
  __kmp_acquire_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);                       \
                                                                               \
  new_value = (*loc);                                                          \
                                                                               \
  __kmp_release_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);

// -------------------------------------------------------------------------
#ifdef KMP_GOMP_COMPAT
#define OP_GOMP_CRITICAL_READ(OP, FLAG)                                        \
  if ((FLAG) && (__kmp_atomic_mode == 2)) {                                    \
    KMP_CHECK_GTID;                                                            \
    OP_CRITICAL_READ(OP, 0);                                                   \
    return new_value;                                                          \
  }
#else
#define OP_GOMP_CRITICAL_READ(OP, FLAG)
#endif /* KMP_GOMP_COMPAT */

// -------------------------------------------------------------------------
#define ATOMIC_FIXED_READ(TYPE_ID, OP_ID, TYPE, BITS, OP, GOMP_FLAG)           \
  ATOMIC_BEGIN_READ(TYPE_ID, OP_ID, TYPE, TYPE)                                \
  TYPE new_value;                                                              \
  OP_GOMP_CRITICAL_READ(OP## =, GOMP_FLAG)                                     \
  new_value = KMP_TEST_THEN_ADD##BITS(loc, OP 0);                              \
  return new_value;                                                            \
  }
// -------------------------------------------------------------------------
#define ATOMIC_CMPXCHG_READ(TYPE_ID, OP_ID, TYPE, BITS, OP, GOMP_FLAG)         \
  ATOMIC_BEGIN_READ(TYPE_ID, OP_ID, TYPE, TYPE)                                \
  TYPE new_value;                                                              \
  OP_GOMP_CRITICAL_READ(OP## =, GOMP_FLAG)                                     \
  OP_CMPXCHG_READ(TYPE, BITS, OP)                                              \
  }
// ------------------------------------------------------------------------
// Routines for Extended types: long double, _Quad, complex flavours (use
// critical section)
//     TYPE_ID, OP_ID, TYPE - detailed above
//     OP      - operator
//     LCK_ID  - lock identifier, used to possibly distinguish lock variable
#define ATOMIC_CRITICAL_READ(TYPE_ID, OP_ID, TYPE, OP, LCK_ID, GOMP_FLAG)      \
  ATOMIC_BEGIN_READ(TYPE_ID, OP_ID, TYPE, TYPE)                                \
  TYPE new_value;                                                              \
  OP_GOMP_CRITICAL_READ(OP## =, GOMP_FLAG) /* send assignment */               \
  OP_CRITICAL_READ(OP, LCK_ID) /* send assignment */                           \
  return new_value;                                                            \
  }

// ------------------------------------------------------------------------
// Fix for cmplx4 read (CQ220361) on Windows* OS. Regular routine with return
// value doesn't work.
// Let's return the read value through the additional parameter.
#if (KMP_OS_WINDOWS)

#define OP_CRITICAL_READ_WRK(OP, LCK_ID)                                       \
  __kmp_acquire_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);                       \
                                                                               \
  (*out) = (*loc);                                                             \
                                                                               \
  __kmp_release_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);
// ------------------------------------------------------------------------
#ifdef KMP_GOMP_COMPAT
#define OP_GOMP_CRITICAL_READ_WRK(OP, FLAG)                                    \
  if ((FLAG) && (__kmp_atomic_mode == 2)) {                                    \
    KMP_CHECK_GTID;                                                            \
    OP_CRITICAL_READ_WRK(OP, 0);                                               \
  }
#else
#define OP_GOMP_CRITICAL_READ_WRK(OP, FLAG)
#endif /* KMP_GOMP_COMPAT */
// ------------------------------------------------------------------------
#define ATOMIC_BEGIN_READ_WRK(TYPE_ID, OP_ID, TYPE)                            \
  void __kmpc_atomic_##TYPE_ID##_##OP_ID(TYPE *out, ident_t *id_ref, int gtid, \
                                         TYPE *loc) {                          \
    KMP_DEBUG_ASSERT(__kmp_init_serial);                                       \
    KA_TRACE(100, ("__kmpc_atomic_" #TYPE_ID "_" #OP_ID ": T#%d\n", gtid));

// ------------------------------------------------------------------------
#define ATOMIC_CRITICAL_READ_WRK(TYPE_ID, OP_ID, TYPE, OP, LCK_ID, GOMP_FLAG)  \
  ATOMIC_BEGIN_READ_WRK(TYPE_ID, OP_ID, TYPE)                                  \
  OP_GOMP_CRITICAL_READ_WRK(OP## =, GOMP_FLAG) /* send assignment */           \
  OP_CRITICAL_READ_WRK(OP, LCK_ID) /* send assignment */                       \
  }

#endif // KMP_OS_WINDOWS

// ------------------------------------------------------------------------
//                  TYPE_ID,OP_ID, TYPE,      OP, GOMP_FLAG
ATOMIC_FIXED_READ(fixed4, rd, kmp_int32, 32, +, 0) // __kmpc_atomic_fixed4_rd
ATOMIC_FIXED_READ(fixed8, rd, kmp_int64, 64, +,
                  KMP_ARCH_X86) // __kmpc_atomic_fixed8_rd
ATOMIC_CMPXCHG_READ(float4, rd, kmp_real32, 32, +,
                    KMP_ARCH_X86) // __kmpc_atomic_float4_rd
ATOMIC_CMPXCHG_READ(float8, rd, kmp_real64, 64, +,
                    KMP_ARCH_X86) // __kmpc_atomic_float8_rd

// !!! TODO: Remove lock operations for "char" since it can't be non-atomic
ATOMIC_CMPXCHG_READ(fixed1, rd, kmp_int8, 8, +,
                    KMP_ARCH_X86) // __kmpc_atomic_fixed1_rd
ATOMIC_CMPXCHG_READ(fixed2, rd, kmp_int16, 16, +,
                    KMP_ARCH_X86) // __kmpc_atomic_fixed2_rd

ATOMIC_CRITICAL_READ(float10, rd, long double, +, 10r,
                     1) // __kmpc_atomic_float10_rd
#if KMP_HAVE_QUAD
ATOMIC_CRITICAL_READ(float16, rd, QUAD_LEGACY, +, 16r,
                     1) // __kmpc_atomic_float16_rd
#endif // KMP_HAVE_QUAD

// Fix for CQ220361 on Windows* OS
#if (KMP_OS_WINDOWS)
ATOMIC_CRITICAL_READ_WRK(cmplx4, rd, kmp_cmplx32, +, 8c,
                         1) // __kmpc_atomic_cmplx4_rd
#else
ATOMIC_CRITICAL_READ(cmplx4, rd, kmp_cmplx32, +, 8c,
                     1) // __kmpc_atomic_cmplx4_rd
#endif
ATOMIC_CRITICAL_READ(cmplx8, rd, kmp_cmplx64, +, 16c,
                     1) // __kmpc_atomic_cmplx8_rd
ATOMIC_CRITICAL_READ(cmplx10, rd, kmp_cmplx80, +, 20c,
                     1) // __kmpc_atomic_cmplx10_rd
#if KMP_HAVE_QUAD
ATOMIC_CRITICAL_READ(cmplx16, rd, CPLX128_LEG, +, 32c,
                     1) // __kmpc_atomic_cmplx16_rd
#if (KMP_ARCH_X86)
ATOMIC_CRITICAL_READ(float16, a16_rd, Quad_a16_t, +, 16r,
                     1) // __kmpc_atomic_float16_a16_rd
ATOMIC_CRITICAL_READ(cmplx16, a16_rd, kmp_cmplx128_a16_t, +, 32c,
                     1) // __kmpc_atomic_cmplx16_a16_rd
#endif
#endif

// ------------------------------------------------------------------------
// Atomic WRITE routines

#define ATOMIC_XCHG_WR(TYPE_ID, OP_ID, TYPE, BITS, OP, GOMP_FLAG)              \
  ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, void)                                     \
  OP_GOMP_CRITICAL(OP, GOMP_FLAG)                                              \
  KMP_XCHG_FIXED##BITS(lhs, rhs);                                              \
  }
// ------------------------------------------------------------------------
#define ATOMIC_XCHG_FLOAT_WR(TYPE_ID, OP_ID, TYPE, BITS, OP, GOMP_FLAG)        \
  ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, void)                                     \
  OP_GOMP_CRITICAL(OP, GOMP_FLAG)                                              \
  KMP_XCHG_REAL##BITS(lhs, rhs);                                               \
  }

// ------------------------------------------------------------------------
// Operation on *lhs, rhs using "compare_and_store" routine
//     TYPE    - operands' type
//     BITS    - size in bits, used to distinguish low level calls
//     OP      - operator
// Note: temp_val introduced in order to force the compiler to read
//       *lhs only once (w/o it the compiler reads *lhs twice)
#define OP_CMPXCHG_WR(TYPE, BITS, OP)                                          \
  {                                                                            \
    TYPE KMP_ATOMIC_VOLATILE temp_val;                                         \
    TYPE old_value, new_value;                                                 \
    temp_val = *lhs;                                                           \
    old_value = temp_val;                                                      \
    new_value = rhs;                                                           \
    while (!KMP_COMPARE_AND_STORE_ACQ##BITS(                                   \
        (kmp_int##BITS *)lhs, *VOLATILE_CAST(kmp_int##BITS *) & old_value,     \
        *VOLATILE_CAST(kmp_int##BITS *) & new_value)) {                        \
      KMP_CPU_PAUSE();                                                         \
                                                                               \
      temp_val = *lhs;                                                         \
      old_value = temp_val;                                                    \
      new_value = rhs;                                                         \
    }                                                                          \
  }

// -------------------------------------------------------------------------
#define ATOMIC_CMPXCHG_WR(TYPE_ID, OP_ID, TYPE, BITS, OP, GOMP_FLAG)           \
  ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, void)                                     \
  OP_GOMP_CRITICAL(OP, GOMP_FLAG)                                              \
  OP_CMPXCHG_WR(TYPE, BITS, OP)                                                \
  }

// ------------------------------------------------------------------------
// Routines for Extended types: long double, _Quad, complex flavours (use
// critical section)
//     TYPE_ID, OP_ID, TYPE - detailed above
//     OP      - operator
//     LCK_ID  - lock identifier, used to possibly distinguish lock variable
#define ATOMIC_CRITICAL_WR(TYPE_ID, OP_ID, TYPE, OP, LCK_ID, GOMP_FLAG)        \
  ATOMIC_BEGIN(TYPE_ID, OP_ID, TYPE, void)                                     \
  OP_GOMP_CRITICAL(OP, GOMP_FLAG) /* send assignment */                        \
  OP_CRITICAL(OP, LCK_ID) /* send assignment */                                \
  }
// -------------------------------------------------------------------------

ATOMIC_XCHG_WR(fixed1, wr, kmp_int8, 8, =,
               KMP_ARCH_X86) // __kmpc_atomic_fixed1_wr
ATOMIC_XCHG_WR(fixed2, wr, kmp_int16, 16, =,
               KMP_ARCH_X86) // __kmpc_atomic_fixed2_wr
ATOMIC_XCHG_WR(fixed4, wr, kmp_int32, 32, =,
               KMP_ARCH_X86) // __kmpc_atomic_fixed4_wr
#if (KMP_ARCH_X86)
ATOMIC_CMPXCHG_WR(fixed8, wr, kmp_int64, 64, =,
                  KMP_ARCH_X86) // __kmpc_atomic_fixed8_wr
#else
ATOMIC_XCHG_WR(fixed8, wr, kmp_int64, 64, =,
               KMP_ARCH_X86) // __kmpc_atomic_fixed8_wr
#endif

ATOMIC_XCHG_FLOAT_WR(float4, wr, kmp_real32, 32, =,
                     KMP_ARCH_X86) // __kmpc_atomic_float4_wr
#if (KMP_ARCH_X86)
ATOMIC_CMPXCHG_WR(float8, wr, kmp_real64, 64, =,
                  KMP_ARCH_X86) // __kmpc_atomic_float8_wr
#else
ATOMIC_XCHG_FLOAT_WR(float8, wr, kmp_real64, 64, =,
                     KMP_ARCH_X86) // __kmpc_atomic_float8_wr
#endif

ATOMIC_CRITICAL_WR(float10, wr, long double, =, 10r,
                   1) // __kmpc_atomic_float10_wr
#if KMP_HAVE_QUAD
ATOMIC_CRITICAL_WR(float16, wr, QUAD_LEGACY, =, 16r,
                   1) // __kmpc_atomic_float16_wr
#endif
ATOMIC_CRITICAL_WR(cmplx4, wr, kmp_cmplx32, =, 8c, 1) // __kmpc_atomic_cmplx4_wr
ATOMIC_CRITICAL_WR(cmplx8, wr, kmp_cmplx64, =, 16c,
                   1) // __kmpc_atomic_cmplx8_wr
ATOMIC_CRITICAL_WR(cmplx10, wr, kmp_cmplx80, =, 20c,
                   1) // __kmpc_atomic_cmplx10_wr
#if KMP_HAVE_QUAD
ATOMIC_CRITICAL_WR(cmplx16, wr, CPLX128_LEG, =, 32c,
                   1) // __kmpc_atomic_cmplx16_wr
#if (KMP_ARCH_X86)
ATOMIC_CRITICAL_WR(float16, a16_wr, Quad_a16_t, =, 16r,
                   1) // __kmpc_atomic_float16_a16_wr
ATOMIC_CRITICAL_WR(cmplx16, a16_wr, kmp_cmplx128_a16_t, =, 32c,
                   1) // __kmpc_atomic_cmplx16_a16_wr
#endif
#endif

// ------------------------------------------------------------------------
// Atomic CAPTURE routines

// Beginning of a definition (provides name, parameters, gebug trace)
//     TYPE_ID - operands type and size (fixed*, fixed*u for signed, unsigned
//     fixed)
//     OP_ID   - operation identifier (add, sub, mul, ...)
//     TYPE    - operands' type
#define ATOMIC_BEGIN_CPT(TYPE_ID, OP_ID, TYPE, RET_TYPE)                       \
  RET_TYPE __kmpc_atomic_##TYPE_ID##_##OP_ID(ident_t *id_ref, int gtid,        \
                                             TYPE *lhs, TYPE rhs, int flag) {  \
    KMP_DEBUG_ASSERT(__kmp_init_serial);                                       \
    KA_TRACE(100, ("__kmpc_atomic_" #TYPE_ID "_" #OP_ID ": T#%d\n", gtid));

// -------------------------------------------------------------------------
// Operation on *lhs, rhs bound by critical section
//     OP     - operator (it's supposed to contain an assignment)
//     LCK_ID - lock identifier
// Note: don't check gtid as it should always be valid
// 1, 2-byte - expect valid parameter, other - check before this macro
#define OP_CRITICAL_CPT(OP, LCK_ID)                                            \
  __kmp_acquire_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);                       \
                                                                               \
  if (flag) {                                                                  \
    (*lhs) OP rhs;                                                             \
    new_value = (*lhs);                                                        \
  } else {                                                                     \
    new_value = (*lhs);                                                        \
    (*lhs) OP rhs;                                                             \
  }                                                                            \
                                                                               \
  __kmp_release_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);                       \
  return new_value;

// ------------------------------------------------------------------------
#ifdef KMP_GOMP_COMPAT
#define OP_GOMP_CRITICAL_CPT(OP, FLAG)                                         \
  if ((FLAG) && (__kmp_atomic_mode == 2)) {                                    \
    KMP_CHECK_GTID;                                                            \
    OP_CRITICAL_CPT(OP## =, 0);                                                \
  }
#else
#define OP_GOMP_CRITICAL_CPT(OP, FLAG)
#endif /* KMP_GOMP_COMPAT */

// ------------------------------------------------------------------------
// Operation on *lhs, rhs using "compare_and_store" routine
//     TYPE    - operands' type
//     BITS    - size in bits, used to distinguish low level calls
//     OP      - operator
// Note: temp_val introduced in order to force the compiler to read
//       *lhs only once (w/o it the compiler reads *lhs twice)
#define OP_CMPXCHG_CPT(TYPE, BITS, OP)                                         \
  {                                                                            \
    TYPE KMP_ATOMIC_VOLATILE temp_val;                                         \
    TYPE old_value, new_value;                                                 \
    temp_val = *lhs;                                                           \
    old_value = temp_val;                                                      \
    new_value = old_value OP rhs;                                              \
    while (!KMP_COMPARE_AND_STORE_ACQ##BITS(                                   \
        (kmp_int##BITS *)lhs, *VOLATILE_CAST(kmp_int##BITS *) & old_value,     \
        *VOLATILE_CAST(kmp_int##BITS *) & new_value)) {                        \
      KMP_CPU_PAUSE();                                                         \
                                                                               \
      temp_val = *lhs;                                                         \
      old_value = temp_val;                                                    \
      new_value = old_value OP rhs;                                            \
    }                                                                          \
    if (flag) {                                                                \
      return new_value;                                                        \
    } else                                                                     \
      return old_value;                                                        \
  }

// -------------------------------------------------------------------------
#define ATOMIC_CMPXCHG_CPT(TYPE_ID, OP_ID, TYPE, BITS, OP, GOMP_FLAG)          \
  ATOMIC_BEGIN_CPT(TYPE_ID, OP_ID, TYPE, TYPE)                                 \
  TYPE new_value;                                                              \
  OP_GOMP_CRITICAL_CPT(OP, GOMP_FLAG)                                          \
  OP_CMPXCHG_CPT(TYPE, BITS, OP)                                               \
  }

// -------------------------------------------------------------------------
#define ATOMIC_FIXED_ADD_CPT(TYPE_ID, OP_ID, TYPE, BITS, OP, GOMP_FLAG)        \
  ATOMIC_BEGIN_CPT(TYPE_ID, OP_ID, TYPE, TYPE)                                 \
  TYPE old_value, new_value;                                                   \
  OP_GOMP_CRITICAL_CPT(OP, GOMP_FLAG)                                          \
  /* OP used as a sign for subtraction: (lhs-rhs) --> (lhs+-rhs) */            \
  old_value = KMP_TEST_THEN_ADD##BITS(lhs, OP rhs);                            \
  if (flag) {                                                                  \
    return old_value OP rhs;                                                   \
  } else                                                                       \
    return old_value;                                                          \
  }
// -------------------------------------------------------------------------

ATOMIC_FIXED_ADD_CPT(fixed4, add_cpt, kmp_int32, 32, +,
                     0) // __kmpc_atomic_fixed4_add_cpt
ATOMIC_FIXED_ADD_CPT(fixed4, sub_cpt, kmp_int32, 32, -,
                     0) // __kmpc_atomic_fixed4_sub_cpt
ATOMIC_FIXED_ADD_CPT(fixed8, add_cpt, kmp_int64, 64, +,
                     KMP_ARCH_X86) // __kmpc_atomic_fixed8_add_cpt
ATOMIC_FIXED_ADD_CPT(fixed8, sub_cpt, kmp_int64, 64, -,
                     KMP_ARCH_X86) // __kmpc_atomic_fixed8_sub_cpt

ATOMIC_CMPXCHG_CPT(float4, add_cpt, kmp_real32, 32, +,
                   KMP_ARCH_X86) // __kmpc_atomic_float4_add_cpt
ATOMIC_CMPXCHG_CPT(float4, sub_cpt, kmp_real32, 32, -,
                   KMP_ARCH_X86) // __kmpc_atomic_float4_sub_cpt
ATOMIC_CMPXCHG_CPT(float8, add_cpt, kmp_real64, 64, +,
                   KMP_ARCH_X86) // __kmpc_atomic_float8_add_cpt
ATOMIC_CMPXCHG_CPT(float8, sub_cpt, kmp_real64, 64, -,
                   KMP_ARCH_X86) // __kmpc_atomic_float8_sub_cpt

// ------------------------------------------------------------------------
// Entries definition for integer operands
//     TYPE_ID - operands type and size (fixed4, float4)
//     OP_ID   - operation identifier (add, sub, mul, ...)
//     TYPE    - operand type
//     BITS    - size in bits, used to distinguish low level calls
//     OP      - operator (used in critical section)
//               TYPE_ID,OP_ID,  TYPE,   BITS,OP,GOMP_FLAG
// ------------------------------------------------------------------------
// Routines for ATOMIC integer operands, other operators
// ------------------------------------------------------------------------
//              TYPE_ID,OP_ID, TYPE,          OP,  GOMP_FLAG
ATOMIC_CMPXCHG_CPT(fixed1, add_cpt, kmp_int8, 8, +,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1_add_cpt
ATOMIC_CMPXCHG_CPT(fixed1, andb_cpt, kmp_int8, 8, &,
                   0) // __kmpc_atomic_fixed1_andb_cpt
ATOMIC_CMPXCHG_CPT(fixed1, div_cpt, kmp_int8, 8, /,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1_div_cpt
ATOMIC_CMPXCHG_CPT(fixed1u, div_cpt, kmp_uint8, 8, /,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1u_div_cpt
ATOMIC_CMPXCHG_CPT(fixed1, mul_cpt, kmp_int8, 8, *,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1_mul_cpt
ATOMIC_CMPXCHG_CPT(fixed1, orb_cpt, kmp_int8, 8, |,
                   0) // __kmpc_atomic_fixed1_orb_cpt
ATOMIC_CMPXCHG_CPT(fixed1, shl_cpt, kmp_int8, 8, <<,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1_shl_cpt
ATOMIC_CMPXCHG_CPT(fixed1, shr_cpt, kmp_int8, 8, >>,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1_shr_cpt
ATOMIC_CMPXCHG_CPT(fixed1u, shr_cpt, kmp_uint8, 8, >>,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1u_shr_cpt
ATOMIC_CMPXCHG_CPT(fixed1, sub_cpt, kmp_int8, 8, -,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1_sub_cpt
ATOMIC_CMPXCHG_CPT(fixed1, xor_cpt, kmp_int8, 8, ^,
                   0) // __kmpc_atomic_fixed1_xor_cpt
ATOMIC_CMPXCHG_CPT(fixed2, add_cpt, kmp_int16, 16, +,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2_add_cpt
ATOMIC_CMPXCHG_CPT(fixed2, andb_cpt, kmp_int16, 16, &,
                   0) // __kmpc_atomic_fixed2_andb_cpt
ATOMIC_CMPXCHG_CPT(fixed2, div_cpt, kmp_int16, 16, /,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2_div_cpt
ATOMIC_CMPXCHG_CPT(fixed2u, div_cpt, kmp_uint16, 16, /,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2u_div_cpt
ATOMIC_CMPXCHG_CPT(fixed2, mul_cpt, kmp_int16, 16, *,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2_mul_cpt
ATOMIC_CMPXCHG_CPT(fixed2, orb_cpt, kmp_int16, 16, |,
                   0) // __kmpc_atomic_fixed2_orb_cpt
ATOMIC_CMPXCHG_CPT(fixed2, shl_cpt, kmp_int16, 16, <<,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2_shl_cpt
ATOMIC_CMPXCHG_CPT(fixed2, shr_cpt, kmp_int16, 16, >>,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2_shr_cpt
ATOMIC_CMPXCHG_CPT(fixed2u, shr_cpt, kmp_uint16, 16, >>,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2u_shr_cpt
ATOMIC_CMPXCHG_CPT(fixed2, sub_cpt, kmp_int16, 16, -,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2_sub_cpt
ATOMIC_CMPXCHG_CPT(fixed2, xor_cpt, kmp_int16, 16, ^,
                   0) // __kmpc_atomic_fixed2_xor_cpt
ATOMIC_CMPXCHG_CPT(fixed4, andb_cpt, kmp_int32, 32, &,
                   0) // __kmpc_atomic_fixed4_andb_cpt
ATOMIC_CMPXCHG_CPT(fixed4, div_cpt, kmp_int32, 32, /,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed4_div_cpt
ATOMIC_CMPXCHG_CPT(fixed4u, div_cpt, kmp_uint32, 32, /,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed4u_div_cpt
ATOMIC_CMPXCHG_CPT(fixed4, mul_cpt, kmp_int32, 32, *,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed4_mul_cpt
ATOMIC_CMPXCHG_CPT(fixed4, orb_cpt, kmp_int32, 32, |,
                   0) // __kmpc_atomic_fixed4_orb_cpt
ATOMIC_CMPXCHG_CPT(fixed4, shl_cpt, kmp_int32, 32, <<,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed4_shl_cpt
ATOMIC_CMPXCHG_CPT(fixed4, shr_cpt, kmp_int32, 32, >>,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed4_shr_cpt
ATOMIC_CMPXCHG_CPT(fixed4u, shr_cpt, kmp_uint32, 32, >>,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed4u_shr_cpt
ATOMIC_CMPXCHG_CPT(fixed4, xor_cpt, kmp_int32, 32, ^,
                   0) // __kmpc_atomic_fixed4_xor_cpt
ATOMIC_CMPXCHG_CPT(fixed8, andb_cpt, kmp_int64, 64, &,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8_andb_cpt
ATOMIC_CMPXCHG_CPT(fixed8, div_cpt, kmp_int64, 64, /,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8_div_cpt
ATOMIC_CMPXCHG_CPT(fixed8u, div_cpt, kmp_uint64, 64, /,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8u_div_cpt
ATOMIC_CMPXCHG_CPT(fixed8, mul_cpt, kmp_int64, 64, *,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8_mul_cpt
ATOMIC_CMPXCHG_CPT(fixed8, orb_cpt, kmp_int64, 64, |,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8_orb_cpt
ATOMIC_CMPXCHG_CPT(fixed8, shl_cpt, kmp_int64, 64, <<,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8_shl_cpt
ATOMIC_CMPXCHG_CPT(fixed8, shr_cpt, kmp_int64, 64, >>,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8_shr_cpt
ATOMIC_CMPXCHG_CPT(fixed8u, shr_cpt, kmp_uint64, 64, >>,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8u_shr_cpt
ATOMIC_CMPXCHG_CPT(fixed8, xor_cpt, kmp_int64, 64, ^,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8_xor_cpt
ATOMIC_CMPXCHG_CPT(float4, div_cpt, kmp_real32, 32, /,
                   KMP_ARCH_X86) // __kmpc_atomic_float4_div_cpt
ATOMIC_CMPXCHG_CPT(float4, mul_cpt, kmp_real32, 32, *,
                   KMP_ARCH_X86) // __kmpc_atomic_float4_mul_cpt
ATOMIC_CMPXCHG_CPT(float8, div_cpt, kmp_real64, 64, /,
                   KMP_ARCH_X86) // __kmpc_atomic_float8_div_cpt
ATOMIC_CMPXCHG_CPT(float8, mul_cpt, kmp_real64, 64, *,
                   KMP_ARCH_X86) // __kmpc_atomic_float8_mul_cpt
//              TYPE_ID,OP_ID, TYPE,          OP,  GOMP_FLAG

// CAPTURE routines for mixed types RHS=float16
#if KMP_HAVE_QUAD

// Beginning of a definition (provides name, parameters, gebug trace)
//     TYPE_ID - operands type and size (fixed*, fixed*u for signed, unsigned
//     fixed)
//     OP_ID   - operation identifier (add, sub, mul, ...)
//     TYPE    - operands' type
#define ATOMIC_BEGIN_CPT_MIX(TYPE_ID, OP_ID, TYPE, RTYPE_ID, RTYPE)            \
  TYPE __kmpc_atomic_##TYPE_ID##_##OP_ID##_##RTYPE_ID(                         \
      ident_t *id_ref, int gtid, TYPE *lhs, RTYPE rhs, int flag) {             \
    KMP_DEBUG_ASSERT(__kmp_init_serial);                                       \
    KA_TRACE(100,                                                              \
             ("__kmpc_atomic_" #TYPE_ID "_" #OP_ID "_" #RTYPE_ID ": T#%d\n",   \
              gtid));

// -------------------------------------------------------------------------
#define ATOMIC_CMPXCHG_CPT_MIX(TYPE_ID, TYPE, OP_ID, BITS, OP, RTYPE_ID,       \
                               RTYPE, LCK_ID, MASK, GOMP_FLAG)                 \
  ATOMIC_BEGIN_CPT_MIX(TYPE_ID, OP_ID, TYPE, RTYPE_ID, RTYPE)                  \
  TYPE new_value;                                                              \
  OP_GOMP_CRITICAL_CPT(OP, GOMP_FLAG)                                          \
  OP_CMPXCHG_CPT(TYPE, BITS, OP)                                               \
  }

// -------------------------------------------------------------------------
#define ATOMIC_CRITICAL_CPT_MIX(TYPE_ID, TYPE, OP_ID, OP, RTYPE_ID, RTYPE,     \
                                LCK_ID, GOMP_FLAG)                             \
  ATOMIC_BEGIN_CPT_MIX(TYPE_ID, OP_ID, TYPE, RTYPE_ID, RTYPE)                  \
  TYPE new_value;                                                              \
  OP_GOMP_CRITICAL_CPT(OP, GOMP_FLAG) /* send assignment */                    \
  OP_CRITICAL_CPT(OP## =, LCK_ID) /* send assignment */                        \
  }

ATOMIC_CMPXCHG_CPT_MIX(fixed1, char, add_cpt, 8, +, fp, _Quad, 1i, 0,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed1_add_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed1u, uchar, add_cpt, 8, +, fp, _Quad, 1i, 0,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed1u_add_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed1, char, sub_cpt, 8, -, fp, _Quad, 1i, 0,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed1_sub_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed1u, uchar, sub_cpt, 8, -, fp, _Quad, 1i, 0,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed1u_sub_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed1, char, mul_cpt, 8, *, fp, _Quad, 1i, 0,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed1_mul_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed1u, uchar, mul_cpt, 8, *, fp, _Quad, 1i, 0,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed1u_mul_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed1, char, div_cpt, 8, /, fp, _Quad, 1i, 0,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed1_div_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed1u, uchar, div_cpt, 8, /, fp, _Quad, 1i, 0,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed1u_div_cpt_fp

ATOMIC_CMPXCHG_CPT_MIX(fixed2, short, add_cpt, 16, +, fp, _Quad, 2i, 1,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed2_add_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed2u, ushort, add_cpt, 16, +, fp, _Quad, 2i, 1,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed2u_add_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed2, short, sub_cpt, 16, -, fp, _Quad, 2i, 1,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed2_sub_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed2u, ushort, sub_cpt, 16, -, fp, _Quad, 2i, 1,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed2u_sub_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed2, short, mul_cpt, 16, *, fp, _Quad, 2i, 1,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed2_mul_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed2u, ushort, mul_cpt, 16, *, fp, _Quad, 2i, 1,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed2u_mul_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed2, short, div_cpt, 16, /, fp, _Quad, 2i, 1,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed2_div_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed2u, ushort, div_cpt, 16, /, fp, _Quad, 2i, 1,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed2u_div_cpt_fp

ATOMIC_CMPXCHG_CPT_MIX(fixed4, kmp_int32, add_cpt, 32, +, fp, _Quad, 4i, 3,
                       0) // __kmpc_atomic_fixed4_add_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed4u, kmp_uint32, add_cpt, 32, +, fp, _Quad, 4i, 3,
                       0) // __kmpc_atomic_fixed4u_add_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed4, kmp_int32, sub_cpt, 32, -, fp, _Quad, 4i, 3,
                       0) // __kmpc_atomic_fixed4_sub_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed4u, kmp_uint32, sub_cpt, 32, -, fp, _Quad, 4i, 3,
                       0) // __kmpc_atomic_fixed4u_sub_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed4, kmp_int32, mul_cpt, 32, *, fp, _Quad, 4i, 3,
                       0) // __kmpc_atomic_fixed4_mul_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed4u, kmp_uint32, mul_cpt, 32, *, fp, _Quad, 4i, 3,
                       0) // __kmpc_atomic_fixed4u_mul_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed4, kmp_int32, div_cpt, 32, /, fp, _Quad, 4i, 3,
                       0) // __kmpc_atomic_fixed4_div_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed4u, kmp_uint32, div_cpt, 32, /, fp, _Quad, 4i, 3,
                       0) // __kmpc_atomic_fixed4u_div_cpt_fp

ATOMIC_CMPXCHG_CPT_MIX(fixed8, kmp_int64, add_cpt, 64, +, fp, _Quad, 8i, 7,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed8_add_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed8u, kmp_uint64, add_cpt, 64, +, fp, _Quad, 8i, 7,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed8u_add_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed8, kmp_int64, sub_cpt, 64, -, fp, _Quad, 8i, 7,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed8_sub_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed8u, kmp_uint64, sub_cpt, 64, -, fp, _Quad, 8i, 7,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed8u_sub_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed8, kmp_int64, mul_cpt, 64, *, fp, _Quad, 8i, 7,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed8_mul_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed8u, kmp_uint64, mul_cpt, 64, *, fp, _Quad, 8i, 7,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed8u_mul_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed8, kmp_int64, div_cpt, 64, /, fp, _Quad, 8i, 7,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed8_div_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(fixed8u, kmp_uint64, div_cpt, 64, /, fp, _Quad, 8i, 7,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed8u_div_cpt_fp

ATOMIC_CMPXCHG_CPT_MIX(float4, kmp_real32, add_cpt, 32, +, fp, _Quad, 4r, 3,
                       KMP_ARCH_X86) // __kmpc_atomic_float4_add_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(float4, kmp_real32, sub_cpt, 32, -, fp, _Quad, 4r, 3,
                       KMP_ARCH_X86) // __kmpc_atomic_float4_sub_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(float4, kmp_real32, mul_cpt, 32, *, fp, _Quad, 4r, 3,
                       KMP_ARCH_X86) // __kmpc_atomic_float4_mul_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(float4, kmp_real32, div_cpt, 32, /, fp, _Quad, 4r, 3,
                       KMP_ARCH_X86) // __kmpc_atomic_float4_div_cpt_fp

ATOMIC_CMPXCHG_CPT_MIX(float8, kmp_real64, add_cpt, 64, +, fp, _Quad, 8r, 7,
                       KMP_ARCH_X86) // __kmpc_atomic_float8_add_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(float8, kmp_real64, sub_cpt, 64, -, fp, _Quad, 8r, 7,
                       KMP_ARCH_X86) // __kmpc_atomic_float8_sub_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(float8, kmp_real64, mul_cpt, 64, *, fp, _Quad, 8r, 7,
                       KMP_ARCH_X86) // __kmpc_atomic_float8_mul_cpt_fp
ATOMIC_CMPXCHG_CPT_MIX(float8, kmp_real64, div_cpt, 64, /, fp, _Quad, 8r, 7,
                       KMP_ARCH_X86) // __kmpc_atomic_float8_div_cpt_fp

ATOMIC_CRITICAL_CPT_MIX(float10, long double, add_cpt, +, fp, _Quad, 10r,
                        1) // __kmpc_atomic_float10_add_cpt_fp
ATOMIC_CRITICAL_CPT_MIX(float10, long double, sub_cpt, -, fp, _Quad, 10r,
                        1) // __kmpc_atomic_float10_sub_cpt_fp
ATOMIC_CRITICAL_CPT_MIX(float10, long double, mul_cpt, *, fp, _Quad, 10r,
                        1) // __kmpc_atomic_float10_mul_cpt_fp
ATOMIC_CRITICAL_CPT_MIX(float10, long double, div_cpt, /, fp, _Quad, 10r,
                        1) // __kmpc_atomic_float10_div_cpt_fp

#endif // KMP_HAVE_QUAD

// ------------------------------------------------------------------------
// Routines for C/C++ Reduction operators && and ||

// -------------------------------------------------------------------------
// Operation on *lhs, rhs bound by critical section
//     OP     - operator (it's supposed to contain an assignment)
//     LCK_ID - lock identifier
// Note: don't check gtid as it should always be valid
// 1, 2-byte - expect valid parameter, other - check before this macro
#define OP_CRITICAL_L_CPT(OP, LCK_ID)                                          \
  __kmp_acquire_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);                       \
                                                                               \
  if (flag) {                                                                  \
    new_value OP rhs;                                                          \
  } else                                                                       \
    new_value = (*lhs);                                                        \
                                                                               \
  __kmp_release_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);

// ------------------------------------------------------------------------
#ifdef KMP_GOMP_COMPAT
#define OP_GOMP_CRITICAL_L_CPT(OP, FLAG)                                       \
  if ((FLAG) && (__kmp_atomic_mode == 2)) {                                    \
    KMP_CHECK_GTID;                                                            \
    OP_CRITICAL_L_CPT(OP, 0);                                                  \
    return new_value;                                                          \
  }
#else
#define OP_GOMP_CRITICAL_L_CPT(OP, FLAG)
#endif /* KMP_GOMP_COMPAT */

// ------------------------------------------------------------------------
// Need separate macros for &&, || because there is no combined assignment
#define ATOMIC_CMPX_L_CPT(TYPE_ID, OP_ID, TYPE, BITS, OP, GOMP_FLAG)           \
  ATOMIC_BEGIN_CPT(TYPE_ID, OP_ID, TYPE, TYPE)                                 \
  TYPE new_value;                                                              \
  OP_GOMP_CRITICAL_L_CPT(= *lhs OP, GOMP_FLAG)                                 \
  OP_CMPXCHG_CPT(TYPE, BITS, OP)                                               \
  }

ATOMIC_CMPX_L_CPT(fixed1, andl_cpt, char, 8, &&,
                  KMP_ARCH_X86) // __kmpc_atomic_fixed1_andl_cpt
ATOMIC_CMPX_L_CPT(fixed1, orl_cpt, char, 8, ||,
                  KMP_ARCH_X86) // __kmpc_atomic_fixed1_orl_cpt
ATOMIC_CMPX_L_CPT(fixed2, andl_cpt, short, 16, &&,
                  KMP_ARCH_X86) // __kmpc_atomic_fixed2_andl_cpt
ATOMIC_CMPX_L_CPT(fixed2, orl_cpt, short, 16, ||,
                  KMP_ARCH_X86) // __kmpc_atomic_fixed2_orl_cpt
ATOMIC_CMPX_L_CPT(fixed4, andl_cpt, kmp_int32, 32, &&,
                  0) // __kmpc_atomic_fixed4_andl_cpt
ATOMIC_CMPX_L_CPT(fixed4, orl_cpt, kmp_int32, 32, ||,
                  0) // __kmpc_atomic_fixed4_orl_cpt
ATOMIC_CMPX_L_CPT(fixed8, andl_cpt, kmp_int64, 64, &&,
                  KMP_ARCH_X86) // __kmpc_atomic_fixed8_andl_cpt
ATOMIC_CMPX_L_CPT(fixed8, orl_cpt, kmp_int64, 64, ||,
                  KMP_ARCH_X86) // __kmpc_atomic_fixed8_orl_cpt

// -------------------------------------------------------------------------
// Routines for Fortran operators that matched no one in C:
// MAX, MIN, .EQV., .NEQV.
// Operators .AND., .OR. are covered by __kmpc_atomic_*_{andl,orl}_cpt
// Intrinsics IAND, IOR, IEOR are covered by __kmpc_atomic_*_{andb,orb,xor}_cpt

// -------------------------------------------------------------------------
// MIN and MAX need separate macros
// OP - operator to check if we need any actions?
#define MIN_MAX_CRITSECT_CPT(OP, LCK_ID)                                       \
  __kmp_acquire_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);                       \
                                                                               \
  if (*lhs OP rhs) { /* still need actions? */                                 \
    old_value = *lhs;                                                          \
    *lhs = rhs;                                                                \
    if (flag)                                                                  \
      new_value = rhs;                                                         \
    else                                                                       \
      new_value = old_value;                                                   \
  }                                                                            \
  __kmp_release_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);                       \
  return new_value;

// -------------------------------------------------------------------------
#ifdef KMP_GOMP_COMPAT
#define GOMP_MIN_MAX_CRITSECT_CPT(OP, FLAG)                                    \
  if ((FLAG) && (__kmp_atomic_mode == 2)) {                                    \
    KMP_CHECK_GTID;                                                            \
    MIN_MAX_CRITSECT_CPT(OP, 0);                                               \
  }
#else
#define GOMP_MIN_MAX_CRITSECT_CPT(OP, FLAG)
#endif /* KMP_GOMP_COMPAT */

// -------------------------------------------------------------------------
#define MIN_MAX_CMPXCHG_CPT(TYPE, BITS, OP)                                    \
  {                                                                            \
    TYPE KMP_ATOMIC_VOLATILE temp_val;                                         \
    /*TYPE old_value; */                                                       \
    temp_val = *lhs;                                                           \
    old_value = temp_val;                                                      \
    while (old_value OP rhs && /* still need actions? */                       \
           !KMP_COMPARE_AND_STORE_ACQ##BITS(                                   \
               (kmp_int##BITS *)lhs,                                           \
               *VOLATILE_CAST(kmp_int##BITS *) & old_value,                    \
               *VOLATILE_CAST(kmp_int##BITS *) & rhs)) {                       \
      KMP_CPU_PAUSE();                                                         \
      temp_val = *lhs;                                                         \
      old_value = temp_val;                                                    \
    }                                                                          \
    if (flag)                                                                  \
      return rhs;                                                              \
    else                                                                       \
      return old_value;                                                        \
  }

// -------------------------------------------------------------------------
// 1-byte, 2-byte operands - use critical section
#define MIN_MAX_CRITICAL_CPT(TYPE_ID, OP_ID, TYPE, OP, LCK_ID, GOMP_FLAG)      \
  ATOMIC_BEGIN_CPT(TYPE_ID, OP_ID, TYPE, TYPE)                                 \
  TYPE new_value, old_value;                                                   \
  if (*lhs OP rhs) { /* need actions? */                                       \
    GOMP_MIN_MAX_CRITSECT_CPT(OP, GOMP_FLAG)                                   \
    MIN_MAX_CRITSECT_CPT(OP, LCK_ID)                                           \
  }                                                                            \
  return *lhs;                                                                 \
  }

#define MIN_MAX_COMPXCHG_CPT(TYPE_ID, OP_ID, TYPE, BITS, OP, GOMP_FLAG)        \
  ATOMIC_BEGIN_CPT(TYPE_ID, OP_ID, TYPE, TYPE)                                 \
  TYPE new_value, old_value;                                                   \
  if (*lhs OP rhs) {                                                           \
    GOMP_MIN_MAX_CRITSECT_CPT(OP, GOMP_FLAG)                                   \
    MIN_MAX_CMPXCHG_CPT(TYPE, BITS, OP)                                        \
  }                                                                            \
  return *lhs;                                                                 \
  }

MIN_MAX_COMPXCHG_CPT(fixed1, max_cpt, char, 8, <,
                     KMP_ARCH_X86) // __kmpc_atomic_fixed1_max_cpt
MIN_MAX_COMPXCHG_CPT(fixed1, min_cpt, char, 8, >,
                     KMP_ARCH_X86) // __kmpc_atomic_fixed1_min_cpt
MIN_MAX_COMPXCHG_CPT(fixed2, max_cpt, short, 16, <,
                     KMP_ARCH_X86) // __kmpc_atomic_fixed2_max_cpt
MIN_MAX_COMPXCHG_CPT(fixed2, min_cpt, short, 16, >,
                     KMP_ARCH_X86) // __kmpc_atomic_fixed2_min_cpt
MIN_MAX_COMPXCHG_CPT(fixed4, max_cpt, kmp_int32, 32, <,
                     0) // __kmpc_atomic_fixed4_max_cpt
MIN_MAX_COMPXCHG_CPT(fixed4, min_cpt, kmp_int32, 32, >,
                     0) // __kmpc_atomic_fixed4_min_cpt
MIN_MAX_COMPXCHG_CPT(fixed8, max_cpt, kmp_int64, 64, <,
                     KMP_ARCH_X86) // __kmpc_atomic_fixed8_max_cpt
MIN_MAX_COMPXCHG_CPT(fixed8, min_cpt, kmp_int64, 64, >,
                     KMP_ARCH_X86) // __kmpc_atomic_fixed8_min_cpt
MIN_MAX_COMPXCHG_CPT(float4, max_cpt, kmp_real32, 32, <,
                     KMP_ARCH_X86) // __kmpc_atomic_float4_max_cpt
MIN_MAX_COMPXCHG_CPT(float4, min_cpt, kmp_real32, 32, >,
                     KMP_ARCH_X86) // __kmpc_atomic_float4_min_cpt
MIN_MAX_COMPXCHG_CPT(float8, max_cpt, kmp_real64, 64, <,
                     KMP_ARCH_X86) // __kmpc_atomic_float8_max_cpt
MIN_MAX_COMPXCHG_CPT(float8, min_cpt, kmp_real64, 64, >,
                     KMP_ARCH_X86) // __kmpc_atomic_float8_min_cpt
#if KMP_HAVE_QUAD
MIN_MAX_CRITICAL_CPT(float16, max_cpt, QUAD_LEGACY, <, 16r,
                     1) // __kmpc_atomic_float16_max_cpt
MIN_MAX_CRITICAL_CPT(float16, min_cpt, QUAD_LEGACY, >, 16r,
                     1) // __kmpc_atomic_float16_min_cpt
#if (KMP_ARCH_X86)
MIN_MAX_CRITICAL_CPT(float16, max_a16_cpt, Quad_a16_t, <, 16r,
                     1) // __kmpc_atomic_float16_max_a16_cpt
MIN_MAX_CRITICAL_CPT(float16, min_a16_cpt, Quad_a16_t, >, 16r,
                     1) // __kmpc_atomic_float16_mix_a16_cpt
#endif
#endif

// ------------------------------------------------------------------------
#ifdef KMP_GOMP_COMPAT
#define OP_GOMP_CRITICAL_EQV_CPT(OP, FLAG)                                     \
  if ((FLAG) && (__kmp_atomic_mode == 2)) {                                    \
    KMP_CHECK_GTID;                                                            \
    OP_CRITICAL_CPT(OP, 0);                                                    \
  }
#else
#define OP_GOMP_CRITICAL_EQV_CPT(OP, FLAG)
#endif /* KMP_GOMP_COMPAT */
// ------------------------------------------------------------------------
#define ATOMIC_CMPX_EQV_CPT(TYPE_ID, OP_ID, TYPE, BITS, OP, GOMP_FLAG)         \
  ATOMIC_BEGIN_CPT(TYPE_ID, OP_ID, TYPE, TYPE)                                 \
  TYPE new_value;                                                              \
  OP_GOMP_CRITICAL_EQV_CPT(^= ~, GOMP_FLAG) /* send assignment */              \
  OP_CMPXCHG_CPT(TYPE, BITS, OP)                                               \
  }

// ------------------------------------------------------------------------

ATOMIC_CMPXCHG_CPT(fixed1, neqv_cpt, kmp_int8, 8, ^,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed1_neqv_cpt
ATOMIC_CMPXCHG_CPT(fixed2, neqv_cpt, kmp_int16, 16, ^,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed2_neqv_cpt
ATOMIC_CMPXCHG_CPT(fixed4, neqv_cpt, kmp_int32, 32, ^,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed4_neqv_cpt
ATOMIC_CMPXCHG_CPT(fixed8, neqv_cpt, kmp_int64, 64, ^,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8_neqv_cpt
ATOMIC_CMPX_EQV_CPT(fixed1, eqv_cpt, kmp_int8, 8, ^~,
                    KMP_ARCH_X86) // __kmpc_atomic_fixed1_eqv_cpt
ATOMIC_CMPX_EQV_CPT(fixed2, eqv_cpt, kmp_int16, 16, ^~,
                    KMP_ARCH_X86) // __kmpc_atomic_fixed2_eqv_cpt
ATOMIC_CMPX_EQV_CPT(fixed4, eqv_cpt, kmp_int32, 32, ^~,
                    KMP_ARCH_X86) // __kmpc_atomic_fixed4_eqv_cpt
ATOMIC_CMPX_EQV_CPT(fixed8, eqv_cpt, kmp_int64, 64, ^~,
                    KMP_ARCH_X86) // __kmpc_atomic_fixed8_eqv_cpt

// ------------------------------------------------------------------------
// Routines for Extended types: long double, _Quad, complex flavours (use
// critical section)
//     TYPE_ID, OP_ID, TYPE - detailed above
//     OP      - operator
//     LCK_ID  - lock identifier, used to possibly distinguish lock variable
#define ATOMIC_CRITICAL_CPT(TYPE_ID, OP_ID, TYPE, OP, LCK_ID, GOMP_FLAG)       \
  ATOMIC_BEGIN_CPT(TYPE_ID, OP_ID, TYPE, TYPE)                                 \
  TYPE new_value;                                                              \
  OP_GOMP_CRITICAL_CPT(OP, GOMP_FLAG) /* send assignment */                    \
  OP_CRITICAL_CPT(OP## =, LCK_ID) /* send assignment */                        \
  }

// ------------------------------------------------------------------------
// Workaround for cmplx4. Regular routines with return value don't work
// on Win_32e. Let's return captured values through the additional parameter.
#define OP_CRITICAL_CPT_WRK(OP, LCK_ID)                                        \
  __kmp_acquire_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);                       \
                                                                               \
  if (flag) {                                                                  \
    (*lhs) OP rhs;                                                             \
    (*out) = (*lhs);                                                           \
  } else {                                                                     \
    (*out) = (*lhs);                                                           \
    (*lhs) OP rhs;                                                             \
  }                                                                            \
                                                                               \
  __kmp_release_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);                       \
  return;
// ------------------------------------------------------------------------

#ifdef KMP_GOMP_COMPAT
#define OP_GOMP_CRITICAL_CPT_WRK(OP, FLAG)                                     \
  if ((FLAG) && (__kmp_atomic_mode == 2)) {                                    \
    KMP_CHECK_GTID;                                                            \
    OP_CRITICAL_CPT_WRK(OP## =, 0);                                            \
  }
#else
#define OP_GOMP_CRITICAL_CPT_WRK(OP, FLAG)
#endif /* KMP_GOMP_COMPAT */
// ------------------------------------------------------------------------

#define ATOMIC_BEGIN_WRK(TYPE_ID, OP_ID, TYPE)                                 \
  void __kmpc_atomic_##TYPE_ID##_##OP_ID(ident_t *id_ref, int gtid, TYPE *lhs, \
                                         TYPE rhs, TYPE *out, int flag) {      \
    KMP_DEBUG_ASSERT(__kmp_init_serial);                                       \
    KA_TRACE(100, ("__kmpc_atomic_" #TYPE_ID "_" #OP_ID ": T#%d\n", gtid));
// ------------------------------------------------------------------------

#define ATOMIC_CRITICAL_CPT_WRK(TYPE_ID, OP_ID, TYPE, OP, LCK_ID, GOMP_FLAG)   \
  ATOMIC_BEGIN_WRK(TYPE_ID, OP_ID, TYPE)                                       \
  OP_GOMP_CRITICAL_CPT_WRK(OP, GOMP_FLAG)                                      \
  OP_CRITICAL_CPT_WRK(OP## =, LCK_ID)                                          \
  }
// The end of workaround for cmplx4

/* ------------------------------------------------------------------------- */
// routines for long double type
ATOMIC_CRITICAL_CPT(float10, add_cpt, long double, +, 10r,
                    1) // __kmpc_atomic_float10_add_cpt
ATOMIC_CRITICAL_CPT(float10, sub_cpt, long double, -, 10r,
                    1) // __kmpc_atomic_float10_sub_cpt
ATOMIC_CRITICAL_CPT(float10, mul_cpt, long double, *, 10r,
                    1) // __kmpc_atomic_float10_mul_cpt
ATOMIC_CRITICAL_CPT(float10, div_cpt, long double, /, 10r,
                    1) // __kmpc_atomic_float10_div_cpt
#if KMP_HAVE_QUAD
// routines for _Quad type
ATOMIC_CRITICAL_CPT(float16, add_cpt, QUAD_LEGACY, +, 16r,
                    1) // __kmpc_atomic_float16_add_cpt
ATOMIC_CRITICAL_CPT(float16, sub_cpt, QUAD_LEGACY, -, 16r,
                    1) // __kmpc_atomic_float16_sub_cpt
ATOMIC_CRITICAL_CPT(float16, mul_cpt, QUAD_LEGACY, *, 16r,
                    1) // __kmpc_atomic_float16_mul_cpt
ATOMIC_CRITICAL_CPT(float16, div_cpt, QUAD_LEGACY, /, 16r,
                    1) // __kmpc_atomic_float16_div_cpt
#if (KMP_ARCH_X86)
ATOMIC_CRITICAL_CPT(float16, add_a16_cpt, Quad_a16_t, +, 16r,
                    1) // __kmpc_atomic_float16_add_a16_cpt
ATOMIC_CRITICAL_CPT(float16, sub_a16_cpt, Quad_a16_t, -, 16r,
                    1) // __kmpc_atomic_float16_sub_a16_cpt
ATOMIC_CRITICAL_CPT(float16, mul_a16_cpt, Quad_a16_t, *, 16r,
                    1) // __kmpc_atomic_float16_mul_a16_cpt
ATOMIC_CRITICAL_CPT(float16, div_a16_cpt, Quad_a16_t, /, 16r,
                    1) // __kmpc_atomic_float16_div_a16_cpt
#endif
#endif

// routines for complex types

// cmplx4 routines to return void
ATOMIC_CRITICAL_CPT_WRK(cmplx4, add_cpt, kmp_cmplx32, +, 8c,
                        1) // __kmpc_atomic_cmplx4_add_cpt
ATOMIC_CRITICAL_CPT_WRK(cmplx4, sub_cpt, kmp_cmplx32, -, 8c,
                        1) // __kmpc_atomic_cmplx4_sub_cpt
ATOMIC_CRITICAL_CPT_WRK(cmplx4, mul_cpt, kmp_cmplx32, *, 8c,
                        1) // __kmpc_atomic_cmplx4_mul_cpt
ATOMIC_CRITICAL_CPT_WRK(cmplx4, div_cpt, kmp_cmplx32, /, 8c,
                        1) // __kmpc_atomic_cmplx4_div_cpt

ATOMIC_CRITICAL_CPT(cmplx8, add_cpt, kmp_cmplx64, +, 16c,
                    1) // __kmpc_atomic_cmplx8_add_cpt
ATOMIC_CRITICAL_CPT(cmplx8, sub_cpt, kmp_cmplx64, -, 16c,
                    1) // __kmpc_atomic_cmplx8_sub_cpt
ATOMIC_CRITICAL_CPT(cmplx8, mul_cpt, kmp_cmplx64, *, 16c,
                    1) // __kmpc_atomic_cmplx8_mul_cpt
ATOMIC_CRITICAL_CPT(cmplx8, div_cpt, kmp_cmplx64, /, 16c,
                    1) // __kmpc_atomic_cmplx8_div_cpt
ATOMIC_CRITICAL_CPT(cmplx10, add_cpt, kmp_cmplx80, +, 20c,
                    1) // __kmpc_atomic_cmplx10_add_cpt
ATOMIC_CRITICAL_CPT(cmplx10, sub_cpt, kmp_cmplx80, -, 20c,
                    1) // __kmpc_atomic_cmplx10_sub_cpt
ATOMIC_CRITICAL_CPT(cmplx10, mul_cpt, kmp_cmplx80, *, 20c,
                    1) // __kmpc_atomic_cmplx10_mul_cpt
ATOMIC_CRITICAL_CPT(cmplx10, div_cpt, kmp_cmplx80, /, 20c,
                    1) // __kmpc_atomic_cmplx10_div_cpt
#if KMP_HAVE_QUAD
ATOMIC_CRITICAL_CPT(cmplx16, add_cpt, CPLX128_LEG, +, 32c,
                    1) // __kmpc_atomic_cmplx16_add_cpt
ATOMIC_CRITICAL_CPT(cmplx16, sub_cpt, CPLX128_LEG, -, 32c,
                    1) // __kmpc_atomic_cmplx16_sub_cpt
ATOMIC_CRITICAL_CPT(cmplx16, mul_cpt, CPLX128_LEG, *, 32c,
                    1) // __kmpc_atomic_cmplx16_mul_cpt
ATOMIC_CRITICAL_CPT(cmplx16, div_cpt, CPLX128_LEG, /, 32c,
                    1) // __kmpc_atomic_cmplx16_div_cpt
#if (KMP_ARCH_X86)
ATOMIC_CRITICAL_CPT(cmplx16, add_a16_cpt, kmp_cmplx128_a16_t, +, 32c,
                    1) // __kmpc_atomic_cmplx16_add_a16_cpt
ATOMIC_CRITICAL_CPT(cmplx16, sub_a16_cpt, kmp_cmplx128_a16_t, -, 32c,
                    1) // __kmpc_atomic_cmplx16_sub_a16_cpt
ATOMIC_CRITICAL_CPT(cmplx16, mul_a16_cpt, kmp_cmplx128_a16_t, *, 32c,
                    1) // __kmpc_atomic_cmplx16_mul_a16_cpt
ATOMIC_CRITICAL_CPT(cmplx16, div_a16_cpt, kmp_cmplx128_a16_t, /, 32c,
                    1) // __kmpc_atomic_cmplx16_div_a16_cpt
#endif
#endif

#if OMP_40_ENABLED

// OpenMP 4.0: v = x = expr binop x; { v = x; x = expr binop x; } { x = expr
// binop x; v = x; }  for non-commutative operations.
// Supported only on IA-32 architecture and Intel(R) 64

// -------------------------------------------------------------------------
// Operation on *lhs, rhs bound by critical section
//     OP     - operator (it's supposed to contain an assignment)
//     LCK_ID - lock identifier
// Note: don't check gtid as it should always be valid
// 1, 2-byte - expect valid parameter, other - check before this macro
#define OP_CRITICAL_CPT_REV(OP, LCK_ID)                                        \
  __kmp_acquire_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);                       \
                                                                               \
  if (flag) {                                                                  \
    /*temp_val = (*lhs);*/                                                     \
    (*lhs) = (rhs)OP(*lhs);                                                    \
    new_value = (*lhs);                                                        \
  } else {                                                                     \
    new_value = (*lhs);                                                        \
    (*lhs) = (rhs)OP(*lhs);                                                    \
  }                                                                            \
  __kmp_release_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);                       \
  return new_value;

// ------------------------------------------------------------------------
#ifdef KMP_GOMP_COMPAT
#define OP_GOMP_CRITICAL_CPT_REV(OP, FLAG)                                     \
  if ((FLAG) && (__kmp_atomic_mode == 2)) {                                    \
    KMP_CHECK_GTID;                                                            \
    OP_CRITICAL_CPT_REV(OP, 0);                                                \
  }
#else
#define OP_GOMP_CRITICAL_CPT_REV(OP, FLAG)
#endif /* KMP_GOMP_COMPAT */

// ------------------------------------------------------------------------
// Operation on *lhs, rhs using "compare_and_store" routine
//     TYPE    - operands' type
//     BITS    - size in bits, used to distinguish low level calls
//     OP      - operator
// Note: temp_val introduced in order to force the compiler to read
//       *lhs only once (w/o it the compiler reads *lhs twice)
#define OP_CMPXCHG_CPT_REV(TYPE, BITS, OP)                                     \
  {                                                                            \
    TYPE KMP_ATOMIC_VOLATILE temp_val;                                         \
    TYPE old_value, new_value;                                                 \
    temp_val = *lhs;                                                           \
    old_value = temp_val;                                                      \
    new_value = rhs OP old_value;                                              \
    while (!KMP_COMPARE_AND_STORE_ACQ##BITS(                                   \
        (kmp_int##BITS *)lhs, *VOLATILE_CAST(kmp_int##BITS *) & old_value,     \
        *VOLATILE_CAST(kmp_int##BITS *) & new_value)) {                        \
      KMP_CPU_PAUSE();                                                         \
                                                                               \
      temp_val = *lhs;                                                         \
      old_value = temp_val;                                                    \
      new_value = rhs OP old_value;                                            \
    }                                                                          \
    if (flag) {                                                                \
      return new_value;                                                        \
    } else                                                                     \
      return old_value;                                                        \
  }

// -------------------------------------------------------------------------
#define ATOMIC_CMPXCHG_CPT_REV(TYPE_ID, OP_ID, TYPE, BITS, OP, GOMP_FLAG)      \
  ATOMIC_BEGIN_CPT(TYPE_ID, OP_ID, TYPE, TYPE)                                 \
  TYPE new_value;                                                              \
  TYPE KMP_ATOMIC_VOLATILE temp_val;                                           \
  OP_GOMP_CRITICAL_CPT_REV(OP, GOMP_FLAG)                                      \
  OP_CMPXCHG_CPT_REV(TYPE, BITS, OP)                                           \
  }

ATOMIC_CMPXCHG_CPT_REV(fixed1, div_cpt_rev, kmp_int8, 8, /,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed1_div_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed1u, div_cpt_rev, kmp_uint8, 8, /,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed1u_div_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed1, shl_cpt_rev, kmp_int8, 8, <<,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed1_shl_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed1, shr_cpt_rev, kmp_int8, 8, >>,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed1_shr_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed1u, shr_cpt_rev, kmp_uint8, 8, >>,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed1u_shr_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed1, sub_cpt_rev, kmp_int8, 8, -,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed1_sub_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed2, div_cpt_rev, kmp_int16, 16, /,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed2_div_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed2u, div_cpt_rev, kmp_uint16, 16, /,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed2u_div_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed2, shl_cpt_rev, kmp_int16, 16, <<,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed2_shl_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed2, shr_cpt_rev, kmp_int16, 16, >>,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed2_shr_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed2u, shr_cpt_rev, kmp_uint16, 16, >>,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed2u_shr_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed2, sub_cpt_rev, kmp_int16, 16, -,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed2_sub_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed4, div_cpt_rev, kmp_int32, 32, /,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed4_div_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed4u, div_cpt_rev, kmp_uint32, 32, /,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed4u_div_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed4, shl_cpt_rev, kmp_int32, 32, <<,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed4_shl_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed4, shr_cpt_rev, kmp_int32, 32, >>,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed4_shr_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed4u, shr_cpt_rev, kmp_uint32, 32, >>,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed4u_shr_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed4, sub_cpt_rev, kmp_int32, 32, -,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed4_sub_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed8, div_cpt_rev, kmp_int64, 64, /,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed8_div_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed8u, div_cpt_rev, kmp_uint64, 64, /,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed8u_div_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed8, shl_cpt_rev, kmp_int64, 64, <<,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed8_shl_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed8, shr_cpt_rev, kmp_int64, 64, >>,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed8_shr_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed8u, shr_cpt_rev, kmp_uint64, 64, >>,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed8u_shr_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(fixed8, sub_cpt_rev, kmp_int64, 64, -,
                       KMP_ARCH_X86) // __kmpc_atomic_fixed8_sub_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(float4, div_cpt_rev, kmp_real32, 32, /,
                       KMP_ARCH_X86) // __kmpc_atomic_float4_div_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(float4, sub_cpt_rev, kmp_real32, 32, -,
                       KMP_ARCH_X86) // __kmpc_atomic_float4_sub_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(float8, div_cpt_rev, kmp_real64, 64, /,
                       KMP_ARCH_X86) // __kmpc_atomic_float8_div_cpt_rev
ATOMIC_CMPXCHG_CPT_REV(float8, sub_cpt_rev, kmp_real64, 64, -,
                       KMP_ARCH_X86) // __kmpc_atomic_float8_sub_cpt_rev
//              TYPE_ID,OP_ID, TYPE,          OP,  GOMP_FLAG

// ------------------------------------------------------------------------
// Routines for Extended types: long double, _Quad, complex flavours (use
// critical section)
//     TYPE_ID, OP_ID, TYPE - detailed above
//     OP      - operator
//     LCK_ID  - lock identifier, used to possibly distinguish lock variable
#define ATOMIC_CRITICAL_CPT_REV(TYPE_ID, OP_ID, TYPE, OP, LCK_ID, GOMP_FLAG)   \
  ATOMIC_BEGIN_CPT(TYPE_ID, OP_ID, TYPE, TYPE)                                 \
  TYPE new_value;                                                              \
  TYPE KMP_ATOMIC_VOLATILE temp_val;                                           \
  /*printf("__kmp_atomic_mode = %d\n", __kmp_atomic_mode);*/                   \
  OP_GOMP_CRITICAL_CPT_REV(OP, GOMP_FLAG)                                      \
  OP_CRITICAL_CPT_REV(OP, LCK_ID)                                              \
  }

/* ------------------------------------------------------------------------- */
// routines for long double type
ATOMIC_CRITICAL_CPT_REV(float10, sub_cpt_rev, long double, -, 10r,
                        1) // __kmpc_atomic_float10_sub_cpt_rev
ATOMIC_CRITICAL_CPT_REV(float10, div_cpt_rev, long double, /, 10r,
                        1) // __kmpc_atomic_float10_div_cpt_rev
#if KMP_HAVE_QUAD
// routines for _Quad type
ATOMIC_CRITICAL_CPT_REV(float16, sub_cpt_rev, QUAD_LEGACY, -, 16r,
                        1) // __kmpc_atomic_float16_sub_cpt_rev
ATOMIC_CRITICAL_CPT_REV(float16, div_cpt_rev, QUAD_LEGACY, /, 16r,
                        1) // __kmpc_atomic_float16_div_cpt_rev
#if (KMP_ARCH_X86)
ATOMIC_CRITICAL_CPT_REV(float16, sub_a16_cpt_rev, Quad_a16_t, -, 16r,
                        1) // __kmpc_atomic_float16_sub_a16_cpt_rev
ATOMIC_CRITICAL_CPT_REV(float16, div_a16_cpt_rev, Quad_a16_t, /, 16r,
                        1) // __kmpc_atomic_float16_div_a16_cpt_rev
#endif
#endif

// routines for complex types

// ------------------------------------------------------------------------
// Workaround for cmplx4. Regular routines with return value don't work
// on Win_32e. Let's return captured values through the additional parameter.
#define OP_CRITICAL_CPT_REV_WRK(OP, LCK_ID)                                    \
  __kmp_acquire_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);                       \
                                                                               \
  if (flag) {                                                                  \
    (*lhs) = (rhs)OP(*lhs);                                                    \
    (*out) = (*lhs);                                                           \
  } else {                                                                     \
    (*out) = (*lhs);                                                           \
    (*lhs) = (rhs)OP(*lhs);                                                    \
  }                                                                            \
                                                                               \
  __kmp_release_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);                       \
  return;
// ------------------------------------------------------------------------

#ifdef KMP_GOMP_COMPAT
#define OP_GOMP_CRITICAL_CPT_REV_WRK(OP, FLAG)                                 \
  if ((FLAG) && (__kmp_atomic_mode == 2)) {                                    \
    KMP_CHECK_GTID;                                                            \
    OP_CRITICAL_CPT_REV_WRK(OP, 0);                                            \
  }
#else
#define OP_GOMP_CRITICAL_CPT_REV_WRK(OP, FLAG)
#endif /* KMP_GOMP_COMPAT */
// ------------------------------------------------------------------------

#define ATOMIC_CRITICAL_CPT_REV_WRK(TYPE_ID, OP_ID, TYPE, OP, LCK_ID,          \
                                    GOMP_FLAG)                                 \
  ATOMIC_BEGIN_WRK(TYPE_ID, OP_ID, TYPE)                                       \
  OP_GOMP_CRITICAL_CPT_REV_WRK(OP, GOMP_FLAG)                                  \
  OP_CRITICAL_CPT_REV_WRK(OP, LCK_ID)                                          \
  }
// The end of workaround for cmplx4

// !!! TODO: check if we need to return void for cmplx4 routines
// cmplx4 routines to return void
ATOMIC_CRITICAL_CPT_REV_WRK(cmplx4, sub_cpt_rev, kmp_cmplx32, -, 8c,
                            1) // __kmpc_atomic_cmplx4_sub_cpt_rev
ATOMIC_CRITICAL_CPT_REV_WRK(cmplx4, div_cpt_rev, kmp_cmplx32, /, 8c,
                            1) // __kmpc_atomic_cmplx4_div_cpt_rev

ATOMIC_CRITICAL_CPT_REV(cmplx8, sub_cpt_rev, kmp_cmplx64, -, 16c,
                        1) // __kmpc_atomic_cmplx8_sub_cpt_rev
ATOMIC_CRITICAL_CPT_REV(cmplx8, div_cpt_rev, kmp_cmplx64, /, 16c,
                        1) // __kmpc_atomic_cmplx8_div_cpt_rev
ATOMIC_CRITICAL_CPT_REV(cmplx10, sub_cpt_rev, kmp_cmplx80, -, 20c,
                        1) // __kmpc_atomic_cmplx10_sub_cpt_rev
ATOMIC_CRITICAL_CPT_REV(cmplx10, div_cpt_rev, kmp_cmplx80, /, 20c,
                        1) // __kmpc_atomic_cmplx10_div_cpt_rev
#if KMP_HAVE_QUAD
ATOMIC_CRITICAL_CPT_REV(cmplx16, sub_cpt_rev, CPLX128_LEG, -, 32c,
                        1) // __kmpc_atomic_cmplx16_sub_cpt_rev
ATOMIC_CRITICAL_CPT_REV(cmplx16, div_cpt_rev, CPLX128_LEG, /, 32c,
                        1) // __kmpc_atomic_cmplx16_div_cpt_rev
#if (KMP_ARCH_X86)
ATOMIC_CRITICAL_CPT_REV(cmplx16, sub_a16_cpt_rev, kmp_cmplx128_a16_t, -, 32c,
                        1) // __kmpc_atomic_cmplx16_sub_a16_cpt_rev
ATOMIC_CRITICAL_CPT_REV(cmplx16, div_a16_cpt_rev, kmp_cmplx128_a16_t, /, 32c,
                        1) // __kmpc_atomic_cmplx16_div_a16_cpt_rev
#endif
#endif

// Capture reverse for mixed type: RHS=float16
#if KMP_HAVE_QUAD

// Beginning of a definition (provides name, parameters, gebug trace)
//     TYPE_ID - operands type and size (fixed*, fixed*u for signed, unsigned
//     fixed)
//     OP_ID   - operation identifier (add, sub, mul, ...)
//     TYPE    - operands' type
// -------------------------------------------------------------------------
#define ATOMIC_CMPXCHG_CPT_REV_MIX(TYPE_ID, TYPE, OP_ID, BITS, OP, RTYPE_ID,   \
                                   RTYPE, LCK_ID, MASK, GOMP_FLAG)             \
  ATOMIC_BEGIN_CPT_MIX(TYPE_ID, OP_ID, TYPE, RTYPE_ID, RTYPE)                  \
  TYPE new_value;                                                              \
  OP_GOMP_CRITICAL_CPT_REV(OP, GOMP_FLAG)                                      \
  OP_CMPXCHG_CPT_REV(TYPE, BITS, OP)                                           \
  }

// -------------------------------------------------------------------------
#define ATOMIC_CRITICAL_CPT_REV_MIX(TYPE_ID, TYPE, OP_ID, OP, RTYPE_ID, RTYPE, \
                                    LCK_ID, GOMP_FLAG)                         \
  ATOMIC_BEGIN_CPT_MIX(TYPE_ID, OP_ID, TYPE, RTYPE_ID, RTYPE)                  \
  TYPE new_value;                                                              \
  OP_GOMP_CRITICAL_CPT_REV(OP, GOMP_FLAG) /* send assignment */                \
  OP_CRITICAL_CPT_REV(OP, LCK_ID) /* send assignment */                        \
  }

ATOMIC_CMPXCHG_CPT_REV_MIX(fixed1, char, sub_cpt_rev, 8, -, fp, _Quad, 1i, 0,
                           KMP_ARCH_X86) // __kmpc_atomic_fixed1_sub_cpt_rev_fp
ATOMIC_CMPXCHG_CPT_REV_MIX(fixed1u, uchar, sub_cpt_rev, 8, -, fp, _Quad, 1i, 0,
                           KMP_ARCH_X86) // __kmpc_atomic_fixed1u_sub_cpt_rev_fp
ATOMIC_CMPXCHG_CPT_REV_MIX(fixed1, char, div_cpt_rev, 8, /, fp, _Quad, 1i, 0,
                           KMP_ARCH_X86) // __kmpc_atomic_fixed1_div_cpt_rev_fp
ATOMIC_CMPXCHG_CPT_REV_MIX(fixed1u, uchar, div_cpt_rev, 8, /, fp, _Quad, 1i, 0,
                           KMP_ARCH_X86) // __kmpc_atomic_fixed1u_div_cpt_rev_fp

ATOMIC_CMPXCHG_CPT_REV_MIX(fixed2, short, sub_cpt_rev, 16, -, fp, _Quad, 2i, 1,
                           KMP_ARCH_X86) // __kmpc_atomic_fixed2_sub_cpt_rev_fp
ATOMIC_CMPXCHG_CPT_REV_MIX(fixed2u, ushort, sub_cpt_rev, 16, -, fp, _Quad, 2i,
                           1,
                           KMP_ARCH_X86) // __kmpc_atomic_fixed2u_sub_cpt_rev_fp
ATOMIC_CMPXCHG_CPT_REV_MIX(fixed2, short, div_cpt_rev, 16, /, fp, _Quad, 2i, 1,
                           KMP_ARCH_X86) // __kmpc_atomic_fixed2_div_cpt_rev_fp
ATOMIC_CMPXCHG_CPT_REV_MIX(fixed2u, ushort, div_cpt_rev, 16, /, fp, _Quad, 2i,
                           1,
                           KMP_ARCH_X86) // __kmpc_atomic_fixed2u_div_cpt_rev_fp

ATOMIC_CMPXCHG_CPT_REV_MIX(fixed4, kmp_int32, sub_cpt_rev, 32, -, fp, _Quad, 4i,
                           3, 0) // __kmpc_atomic_fixed4_sub_cpt_rev_fp
ATOMIC_CMPXCHG_CPT_REV_MIX(fixed4u, kmp_uint32, sub_cpt_rev, 32, -, fp, _Quad,
                           4i, 3, 0) // __kmpc_atomic_fixed4u_sub_cpt_rev_fp
ATOMIC_CMPXCHG_CPT_REV_MIX(fixed4, kmp_int32, div_cpt_rev, 32, /, fp, _Quad, 4i,
                           3, 0) // __kmpc_atomic_fixed4_div_cpt_rev_fp
ATOMIC_CMPXCHG_CPT_REV_MIX(fixed4u, kmp_uint32, div_cpt_rev, 32, /, fp, _Quad,
                           4i, 3, 0) // __kmpc_atomic_fixed4u_div_cpt_rev_fp

ATOMIC_CMPXCHG_CPT_REV_MIX(fixed8, kmp_int64, sub_cpt_rev, 64, -, fp, _Quad, 8i,
                           7,
                           KMP_ARCH_X86) // __kmpc_atomic_fixed8_sub_cpt_rev_fp
ATOMIC_CMPXCHG_CPT_REV_MIX(fixed8u, kmp_uint64, sub_cpt_rev, 64, -, fp, _Quad,
                           8i, 7,
                           KMP_ARCH_X86) // __kmpc_atomic_fixed8u_sub_cpt_rev_fp
ATOMIC_CMPXCHG_CPT_REV_MIX(fixed8, kmp_int64, div_cpt_rev, 64, /, fp, _Quad, 8i,
                           7,
                           KMP_ARCH_X86) // __kmpc_atomic_fixed8_div_cpt_rev_fp
ATOMIC_CMPXCHG_CPT_REV_MIX(fixed8u, kmp_uint64, div_cpt_rev, 64, /, fp, _Quad,
                           8i, 7,
                           KMP_ARCH_X86) // __kmpc_atomic_fixed8u_div_cpt_rev_fp

ATOMIC_CMPXCHG_CPT_REV_MIX(float4, kmp_real32, sub_cpt_rev, 32, -, fp, _Quad,
                           4r, 3,
                           KMP_ARCH_X86) // __kmpc_atomic_float4_sub_cpt_rev_fp
ATOMIC_CMPXCHG_CPT_REV_MIX(float4, kmp_real32, div_cpt_rev, 32, /, fp, _Quad,
                           4r, 3,
                           KMP_ARCH_X86) // __kmpc_atomic_float4_div_cpt_rev_fp

ATOMIC_CMPXCHG_CPT_REV_MIX(float8, kmp_real64, sub_cpt_rev, 64, -, fp, _Quad,
                           8r, 7,
                           KMP_ARCH_X86) // __kmpc_atomic_float8_sub_cpt_rev_fp
ATOMIC_CMPXCHG_CPT_REV_MIX(float8, kmp_real64, div_cpt_rev, 64, /, fp, _Quad,
                           8r, 7,
                           KMP_ARCH_X86) // __kmpc_atomic_float8_div_cpt_rev_fp

ATOMIC_CRITICAL_CPT_REV_MIX(float10, long double, sub_cpt_rev, -, fp, _Quad,
                            10r, 1) // __kmpc_atomic_float10_sub_cpt_rev_fp
ATOMIC_CRITICAL_CPT_REV_MIX(float10, long double, div_cpt_rev, /, fp, _Quad,
                            10r, 1) // __kmpc_atomic_float10_div_cpt_rev_fp

#endif // KMP_HAVE_QUAD

//   OpenMP 4.0 Capture-write (swap): {v = x; x = expr;}

#define ATOMIC_BEGIN_SWP(TYPE_ID, TYPE)                                        \
  TYPE __kmpc_atomic_##TYPE_ID##_swp(ident_t *id_ref, int gtid, TYPE *lhs,     \
                                     TYPE rhs) {                               \
    KMP_DEBUG_ASSERT(__kmp_init_serial);                                       \
    KA_TRACE(100, ("__kmpc_atomic_" #TYPE_ID "_swp: T#%d\n", gtid));

#define CRITICAL_SWP(LCK_ID)                                                   \
  __kmp_acquire_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);                       \
                                                                               \
  old_value = (*lhs);                                                          \
  (*lhs) = rhs;                                                                \
                                                                               \
  __kmp_release_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);                       \
  return old_value;

// ------------------------------------------------------------------------
#ifdef KMP_GOMP_COMPAT
#define GOMP_CRITICAL_SWP(FLAG)                                                \
  if ((FLAG) && (__kmp_atomic_mode == 2)) {                                    \
    KMP_CHECK_GTID;                                                            \
    CRITICAL_SWP(0);                                                           \
  }
#else
#define GOMP_CRITICAL_SWP(FLAG)
#endif /* KMP_GOMP_COMPAT */

#define ATOMIC_XCHG_SWP(TYPE_ID, TYPE, BITS, GOMP_FLAG)                        \
  ATOMIC_BEGIN_SWP(TYPE_ID, TYPE)                                              \
  TYPE old_value;                                                              \
  GOMP_CRITICAL_SWP(GOMP_FLAG)                                                 \
  old_value = KMP_XCHG_FIXED##BITS(lhs, rhs);                                  \
  return old_value;                                                            \
  }
// ------------------------------------------------------------------------
#define ATOMIC_XCHG_FLOAT_SWP(TYPE_ID, TYPE, BITS, GOMP_FLAG)                  \
  ATOMIC_BEGIN_SWP(TYPE_ID, TYPE)                                              \
  TYPE old_value;                                                              \
  GOMP_CRITICAL_SWP(GOMP_FLAG)                                                 \
  old_value = KMP_XCHG_REAL##BITS(lhs, rhs);                                   \
  return old_value;                                                            \
  }

// ------------------------------------------------------------------------
#define CMPXCHG_SWP(TYPE, BITS)                                                \
  {                                                                            \
    TYPE KMP_ATOMIC_VOLATILE temp_val;                                         \
    TYPE old_value, new_value;                                                 \
    temp_val = *lhs;                                                           \
    old_value = temp_val;                                                      \
    new_value = rhs;                                                           \
    while (!KMP_COMPARE_AND_STORE_ACQ##BITS(                                   \
        (kmp_int##BITS *)lhs, *VOLATILE_CAST(kmp_int##BITS *) & old_value,     \
        *VOLATILE_CAST(kmp_int##BITS *) & new_value)) {                        \
      KMP_CPU_PAUSE();                                                         \
                                                                               \
      temp_val = *lhs;                                                         \
      old_value = temp_val;                                                    \
      new_value = rhs;                                                         \
    }                                                                          \
    return old_value;                                                          \
  }

// -------------------------------------------------------------------------
#define ATOMIC_CMPXCHG_SWP(TYPE_ID, TYPE, BITS, GOMP_FLAG)                     \
  ATOMIC_BEGIN_SWP(TYPE_ID, TYPE)                                              \
  TYPE old_value;                                                              \
  GOMP_CRITICAL_SWP(GOMP_FLAG)                                                 \
  CMPXCHG_SWP(TYPE, BITS)                                                      \
  }

ATOMIC_XCHG_SWP(fixed1, kmp_int8, 8, KMP_ARCH_X86) // __kmpc_atomic_fixed1_swp
ATOMIC_XCHG_SWP(fixed2, kmp_int16, 16, KMP_ARCH_X86) // __kmpc_atomic_fixed2_swp
ATOMIC_XCHG_SWP(fixed4, kmp_int32, 32, KMP_ARCH_X86) // __kmpc_atomic_fixed4_swp

ATOMIC_XCHG_FLOAT_SWP(float4, kmp_real32, 32,
                      KMP_ARCH_X86) // __kmpc_atomic_float4_swp

#if (KMP_ARCH_X86)
ATOMIC_CMPXCHG_SWP(fixed8, kmp_int64, 64,
                   KMP_ARCH_X86) // __kmpc_atomic_fixed8_swp
ATOMIC_CMPXCHG_SWP(float8, kmp_real64, 64,
                   KMP_ARCH_X86) // __kmpc_atomic_float8_swp
#else
ATOMIC_XCHG_SWP(fixed8, kmp_int64, 64, KMP_ARCH_X86) // __kmpc_atomic_fixed8_swp
ATOMIC_XCHG_FLOAT_SWP(float8, kmp_real64, 64,
                      KMP_ARCH_X86) // __kmpc_atomic_float8_swp
#endif

// ------------------------------------------------------------------------
// Routines for Extended types: long double, _Quad, complex flavours (use
// critical section)
#define ATOMIC_CRITICAL_SWP(TYPE_ID, TYPE, LCK_ID, GOMP_FLAG)                  \
  ATOMIC_BEGIN_SWP(TYPE_ID, TYPE)                                              \
  TYPE old_value;                                                              \
  GOMP_CRITICAL_SWP(GOMP_FLAG)                                                 \
  CRITICAL_SWP(LCK_ID)                                                         \
  }

// ------------------------------------------------------------------------
// !!! TODO: check if we need to return void for cmplx4 routines
// Workaround for cmplx4. Regular routines with return value don't work
// on Win_32e. Let's return captured values through the additional parameter.

#define ATOMIC_BEGIN_SWP_WRK(TYPE_ID, TYPE)                                    \
  void __kmpc_atomic_##TYPE_ID##_swp(ident_t *id_ref, int gtid, TYPE *lhs,     \
                                     TYPE rhs, TYPE *out) {                    \
    KMP_DEBUG_ASSERT(__kmp_init_serial);                                       \
    KA_TRACE(100, ("__kmpc_atomic_" #TYPE_ID "_swp: T#%d\n", gtid));

#define CRITICAL_SWP_WRK(LCK_ID)                                               \
  __kmp_acquire_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);                       \
                                                                               \
  tmp = (*lhs);                                                                \
  (*lhs) = (rhs);                                                              \
  (*out) = tmp;                                                                \
  __kmp_release_atomic_lock(&ATOMIC_LOCK##LCK_ID, gtid);                       \
  return;
// ------------------------------------------------------------------------

#ifdef KMP_GOMP_COMPAT
#define GOMP_CRITICAL_SWP_WRK(FLAG)                                            \
  if ((FLAG) && (__kmp_atomic_mode == 2)) {                                    \
    KMP_CHECK_GTID;                                                            \
    CRITICAL_SWP_WRK(0);                                                       \
  }
#else
#define GOMP_CRITICAL_SWP_WRK(FLAG)
#endif /* KMP_GOMP_COMPAT */
// ------------------------------------------------------------------------

#define ATOMIC_CRITICAL_SWP_WRK(TYPE_ID, TYPE, LCK_ID, GOMP_FLAG)              \
  ATOMIC_BEGIN_SWP_WRK(TYPE_ID, TYPE)                                          \
  TYPE tmp;                                                                    \
  GOMP_CRITICAL_SWP_WRK(GOMP_FLAG)                                             \
  CRITICAL_SWP_WRK(LCK_ID)                                                     \
  }
// The end of workaround for cmplx4

ATOMIC_CRITICAL_SWP(float10, long double, 10r, 1) // __kmpc_atomic_float10_swp
#if KMP_HAVE_QUAD
ATOMIC_CRITICAL_SWP(float16, QUAD_LEGACY, 16r, 1) // __kmpc_atomic_float16_swp
#endif
// cmplx4 routine to return void
ATOMIC_CRITICAL_SWP_WRK(cmplx4, kmp_cmplx32, 8c, 1) // __kmpc_atomic_cmplx4_swp

// ATOMIC_CRITICAL_SWP( cmplx4, kmp_cmplx32,  8c,   1 )           //
// __kmpc_atomic_cmplx4_swp

ATOMIC_CRITICAL_SWP(cmplx8, kmp_cmplx64, 16c, 1) // __kmpc_atomic_cmplx8_swp
ATOMIC_CRITICAL_SWP(cmplx10, kmp_cmplx80, 20c, 1) // __kmpc_atomic_cmplx10_swp
#if KMP_HAVE_QUAD
ATOMIC_CRITICAL_SWP(cmplx16, CPLX128_LEG, 32c, 1) // __kmpc_atomic_cmplx16_swp
#if (KMP_ARCH_X86)
ATOMIC_CRITICAL_SWP(float16_a16, Quad_a16_t, 16r,
                    1) // __kmpc_atomic_float16_a16_swp
ATOMIC_CRITICAL_SWP(cmplx16_a16, kmp_cmplx128_a16_t, 32c,
                    1) // __kmpc_atomic_cmplx16_a16_swp
#endif
#endif

// End of OpenMP 4.0 Capture

#endif // OMP_40_ENABLED

#endif // KMP_ARCH_X86 || KMP_ARCH_X86_64

#undef OP_CRITICAL

/* ------------------------------------------------------------------------ */
/* Generic atomic routines                                                  */

void __kmpc_atomic_1(ident_t *id_ref, int gtid, void *lhs, void *rhs,
                     void (*f)(void *, void *, void *)) {
  KMP_DEBUG_ASSERT(__kmp_init_serial);

  if (
#if KMP_ARCH_X86 && defined(KMP_GOMP_COMPAT)
      FALSE /* must use lock */
#else
      TRUE
#endif
      ) {
    kmp_int8 old_value, new_value;

    old_value = *(kmp_int8 *)lhs;
    (*f)(&new_value, &old_value, rhs);

    /* TODO: Should this be acquire or release? */
    while (!KMP_COMPARE_AND_STORE_ACQ8((kmp_int8 *)lhs, *(kmp_int8 *)&old_value,
                                       *(kmp_int8 *)&new_value)) {
      KMP_CPU_PAUSE();

      old_value = *(kmp_int8 *)lhs;
      (*f)(&new_value, &old_value, rhs);
    }

    return;
  } else {
// All 1-byte data is of integer data type.

#ifdef KMP_GOMP_COMPAT
    if (__kmp_atomic_mode == 2) {
      __kmp_acquire_atomic_lock(&__kmp_atomic_lock, gtid);
    } else
#endif /* KMP_GOMP_COMPAT */
      __kmp_acquire_atomic_lock(&__kmp_atomic_lock_1i, gtid);

    (*f)(lhs, lhs, rhs);

#ifdef KMP_GOMP_COMPAT
    if (__kmp_atomic_mode == 2) {
      __kmp_release_atomic_lock(&__kmp_atomic_lock, gtid);
    } else
#endif /* KMP_GOMP_COMPAT */
      __kmp_release_atomic_lock(&__kmp_atomic_lock_1i, gtid);
  }
}

void __kmpc_atomic_2(ident_t *id_ref, int gtid, void *lhs, void *rhs,
                     void (*f)(void *, void *, void *)) {
  if (
#if KMP_ARCH_X86 && defined(KMP_GOMP_COMPAT)
      FALSE /* must use lock */
#elif KMP_ARCH_X86 || KMP_ARCH_X86_64
      TRUE /* no alignment problems */
#else
      !((kmp_uintptr_t)lhs & 0x1) /* make sure address is 2-byte aligned */
#endif
      ) {
    kmp_int16 old_value, new_value;

    old_value = *(kmp_int16 *)lhs;
    (*f)(&new_value, &old_value, rhs);

    /* TODO: Should this be acquire or release? */
    while (!KMP_COMPARE_AND_STORE_ACQ16(
        (kmp_int16 *)lhs, *(kmp_int16 *)&old_value, *(kmp_int16 *)&new_value)) {
      KMP_CPU_PAUSE();

      old_value = *(kmp_int16 *)lhs;
      (*f)(&new_value, &old_value, rhs);
    }

    return;
  } else {
// All 2-byte data is of integer data type.

#ifdef KMP_GOMP_COMPAT
    if (__kmp_atomic_mode == 2) {
      __kmp_acquire_atomic_lock(&__kmp_atomic_lock, gtid);
    } else
#endif /* KMP_GOMP_COMPAT */
      __kmp_acquire_atomic_lock(&__kmp_atomic_lock_2i, gtid);

    (*f)(lhs, lhs, rhs);

#ifdef KMP_GOMP_COMPAT
    if (__kmp_atomic_mode == 2) {
      __kmp_release_atomic_lock(&__kmp_atomic_lock, gtid);
    } else
#endif /* KMP_GOMP_COMPAT */
      __kmp_release_atomic_lock(&__kmp_atomic_lock_2i, gtid);
  }
}

void __kmpc_atomic_4(ident_t *id_ref, int gtid, void *lhs, void *rhs,
                     void (*f)(void *, void *, void *)) {
  KMP_DEBUG_ASSERT(__kmp_init_serial);

  if (
// FIXME: On IA-32 architecture, gcc uses cmpxchg only for 4-byte ints.
// Gomp compatibility is broken if this routine is called for floats.
#if KMP_ARCH_X86 || KMP_ARCH_X86_64
      TRUE /* no alignment problems */
#else
      !((kmp_uintptr_t)lhs & 0x3) /* make sure address is 4-byte aligned */
#endif
      ) {
    kmp_int32 old_value, new_value;

    old_value = *(kmp_int32 *)lhs;
    (*f)(&new_value, &old_value, rhs);

    /* TODO: Should this be acquire or release? */
    while (!KMP_COMPARE_AND_STORE_ACQ32(
        (kmp_int32 *)lhs, *(kmp_int32 *)&old_value, *(kmp_int32 *)&new_value)) {
      KMP_CPU_PAUSE();

      old_value = *(kmp_int32 *)lhs;
      (*f)(&new_value, &old_value, rhs);
    }

    return;
  } else {
// Use __kmp_atomic_lock_4i for all 4-byte data,
// even if it isn't of integer data type.

#ifdef KMP_GOMP_COMPAT
    if (__kmp_atomic_mode == 2) {
      __kmp_acquire_atomic_lock(&__kmp_atomic_lock, gtid);
    } else
#endif /* KMP_GOMP_COMPAT */
      __kmp_acquire_atomic_lock(&__kmp_atomic_lock_4i, gtid);

    (*f)(lhs, lhs, rhs);

#ifdef KMP_GOMP_COMPAT
    if (__kmp_atomic_mode == 2) {
      __kmp_release_atomic_lock(&__kmp_atomic_lock, gtid);
    } else
#endif /* KMP_GOMP_COMPAT */
      __kmp_release_atomic_lock(&__kmp_atomic_lock_4i, gtid);
  }
}

void __kmpc_atomic_8(ident_t *id_ref, int gtid, void *lhs, void *rhs,
                     void (*f)(void *, void *, void *)) {
  KMP_DEBUG_ASSERT(__kmp_init_serial);
  if (

#if KMP_ARCH_X86 && defined(KMP_GOMP_COMPAT)
      FALSE /* must use lock */
#elif KMP_ARCH_X86 || KMP_ARCH_X86_64
      TRUE /* no alignment problems */
#else
      !((kmp_uintptr_t)lhs & 0x7) /* make sure address is 8-byte aligned */
#endif
      ) {
    kmp_int64 old_value, new_value;

    old_value = *(kmp_int64 *)lhs;
    (*f)(&new_value, &old_value, rhs);
    /* TODO: Should this be acquire or release? */
    while (!KMP_COMPARE_AND_STORE_ACQ64(
        (kmp_int64 *)lhs, *(kmp_int64 *)&old_value, *(kmp_int64 *)&new_value)) {
      KMP_CPU_PAUSE();

      old_value = *(kmp_int64 *)lhs;
      (*f)(&new_value, &old_value, rhs);
    }

    return;
  } else {
// Use __kmp_atomic_lock_8i for all 8-byte data,
// even if it isn't of integer data type.

#ifdef KMP_GOMP_COMPAT
    if (__kmp_atomic_mode == 2) {
      __kmp_acquire_atomic_lock(&__kmp_atomic_lock, gtid);
    } else
#endif /* KMP_GOMP_COMPAT */
      __kmp_acquire_atomic_lock(&__kmp_atomic_lock_8i, gtid);

    (*f)(lhs, lhs, rhs);

#ifdef KMP_GOMP_COMPAT
    if (__kmp_atomic_mode == 2) {
      __kmp_release_atomic_lock(&__kmp_atomic_lock, gtid);
    } else
#endif /* KMP_GOMP_COMPAT */
      __kmp_release_atomic_lock(&__kmp_atomic_lock_8i, gtid);
  }
}

void __kmpc_atomic_10(ident_t *id_ref, int gtid, void *lhs, void *rhs,
                      void (*f)(void *, void *, void *)) {
  KMP_DEBUG_ASSERT(__kmp_init_serial);

#ifdef KMP_GOMP_COMPAT
  if (__kmp_atomic_mode == 2) {
    __kmp_acquire_atomic_lock(&__kmp_atomic_lock, gtid);
  } else
#endif /* KMP_GOMP_COMPAT */
    __kmp_acquire_atomic_lock(&__kmp_atomic_lock_10r, gtid);

  (*f)(lhs, lhs, rhs);

#ifdef KMP_GOMP_COMPAT
  if (__kmp_atomic_mode == 2) {
    __kmp_release_atomic_lock(&__kmp_atomic_lock, gtid);
  } else
#endif /* KMP_GOMP_COMPAT */
    __kmp_release_atomic_lock(&__kmp_atomic_lock_10r, gtid);
}

void __kmpc_atomic_16(ident_t *id_ref, int gtid, void *lhs, void *rhs,
                      void (*f)(void *, void *, void *)) {
  KMP_DEBUG_ASSERT(__kmp_init_serial);

#ifdef KMP_GOMP_COMPAT
  if (__kmp_atomic_mode == 2) {
    __kmp_acquire_atomic_lock(&__kmp_atomic_lock, gtid);
  } else
#endif /* KMP_GOMP_COMPAT */
    __kmp_acquire_atomic_lock(&__kmp_atomic_lock_16c, gtid);

  (*f)(lhs, lhs, rhs);

#ifdef KMP_GOMP_COMPAT
  if (__kmp_atomic_mode == 2) {
    __kmp_release_atomic_lock(&__kmp_atomic_lock, gtid);
  } else
#endif /* KMP_GOMP_COMPAT */
    __kmp_release_atomic_lock(&__kmp_atomic_lock_16c, gtid);
}

void __kmpc_atomic_20(ident_t *id_ref, int gtid, void *lhs, void *rhs,
                      void (*f)(void *, void *, void *)) {
  KMP_DEBUG_ASSERT(__kmp_init_serial);

#ifdef KMP_GOMP_COMPAT
  if (__kmp_atomic_mode == 2) {
    __kmp_acquire_atomic_lock(&__kmp_atomic_lock, gtid);
  } else
#endif /* KMP_GOMP_COMPAT */
    __kmp_acquire_atomic_lock(&__kmp_atomic_lock_20c, gtid);

  (*f)(lhs, lhs, rhs);

#ifdef KMP_GOMP_COMPAT
  if (__kmp_atomic_mode == 2) {
    __kmp_release_atomic_lock(&__kmp_atomic_lock, gtid);
  } else
#endif /* KMP_GOMP_COMPAT */
    __kmp_release_atomic_lock(&__kmp_atomic_lock_20c, gtid);
}

void __kmpc_atomic_32(ident_t *id_ref, int gtid, void *lhs, void *rhs,
                      void (*f)(void *, void *, void *)) {
  KMP_DEBUG_ASSERT(__kmp_init_serial);

#ifdef KMP_GOMP_COMPAT
  if (__kmp_atomic_mode == 2) {
    __kmp_acquire_atomic_lock(&__kmp_atomic_lock, gtid);
  } else
#endif /* KMP_GOMP_COMPAT */
    __kmp_acquire_atomic_lock(&__kmp_atomic_lock_32c, gtid);

  (*f)(lhs, lhs, rhs);

#ifdef KMP_GOMP_COMPAT
  if (__kmp_atomic_mode == 2) {
    __kmp_release_atomic_lock(&__kmp_atomic_lock, gtid);
  } else
#endif /* KMP_GOMP_COMPAT */
    __kmp_release_atomic_lock(&__kmp_atomic_lock_32c, gtid);
}

// AC: same two routines as GOMP_atomic_start/end, but will be called by our
// compiler; duplicated in order to not use 3-party names in pure Intel code
// TODO: consider adding GTID parameter after consultation with Ernesto/Xinmin.
void __kmpc_atomic_start(void) {
  int gtid = __kmp_entry_gtid();
  KA_TRACE(20, ("__kmpc_atomic_start: T#%d\n", gtid));
  __kmp_acquire_atomic_lock(&__kmp_atomic_lock, gtid);
}

void __kmpc_atomic_end(void) {
  int gtid = __kmp_get_gtid();
  KA_TRACE(20, ("__kmpc_atomic_end: T#%d\n", gtid));
  __kmp_release_atomic_lock(&__kmp_atomic_lock, gtid);
}

/*!
@}
*/

// end of file