aboutsummaryrefslogtreecommitdiffstats
path: root/erts/emulator/beam/erl_thr_progress.c
blob: 9678d7e08b5869ed39ee063b1edf15fae8eb2875 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
/*
 * %CopyrightBegin%
 *
 * Copyright Ericsson AB 2011-2012. All Rights Reserved.
 *
 * The contents of this file are subject to the Erlang Public License,
 * Version 1.1, (the "License"); you may not use this file except in
 * compliance with the License. You should have received a copy of the
 * Erlang Public License along with this software. If not, it can be
 * retrieved online at http://www.erlang.org/.
 *
 * Software distributed under the License is distributed on an "AS IS"
 * basis, WITHOUT WARRANTY OF ANY KIND, either express or implied. See
 * the License for the specific language governing rights and limitations
 * under the License.
 *
 * %CopyrightEnd%
 */

/*
 * Description: Thread progress information. Used by lock free algorithms
 *              to determine when all involved threads are guaranteed to
 *              have passed a specific point of execution.
 *
 *              Usage instructions below.
 *
 * Author: 	Rickard Green
 */

/*
 * ------ Usage instructions -----------------------------------------------
 *
 * This module keeps track of the progress of a set of managed threads. Only
 * threads that behave well can be allowed to be managed. A managed thread
 * should update its thread progress frequently. Currently only scheduler
 * threads, the system-message-dispatcher threads, and the aux-thread are
 * managed threads. We typically do not want any async threads as managed
 * threads since they cannot guarantee a frequent update of thread progress,
 * since they execute user implemented driver code that is assumed to be
 * time consuming.
 *
 * erts_thr_progress_current() returns the global current thread progress
 * value of managed threads. I.e., the latest progress value that all
 * managed threads have reached. Thread progress values are opaque.
 *
 * erts_thr_progress_has_reached(VAL) returns a value != 0 if current
 * global thread progress has reached or passed VAL.
 *
 * erts_thr_progress_later() returns a thread progress value in the future
 * which no managed thread have yet reached.
 *
 * All threads issue a full memory barrier when reaching a new thread
 * progress value. They only reach new thread progress values in specific
 * controlled states when calling erts_thr_progress_update(). Schedulers
 * call erts_thr_progress_update() in between execution of processes,
 * when going to sleep and when waking up.
 *
 * Sleeping managed threads are considered to have reached next thread
 * progress value immediately. They are not woken and do therefore not
 * issue any memory barriers when reaching a new thread progress value.
 * A sleeping thread do however immediately issue a memory barrier upon
 * wakeup.
 *
 * Both managed and registered unmanaged threads may request wakeup when
 * the global thread progress reach a certain value using
 * erts_thr_progress_wakeup().
 *
 * Note that thread progress values are opaque, and that you are only
 * allowed to use thread progress values retrieved from this API!
 *
 * -------------------------------------------------------------------------
 */

#ifdef HAVE_CONFIG_H
#  include "config.h"
#endif

#include <stddef.h> /* offsetof() */
#include "erl_thr_progress.h"
#include "global.h"

#ifdef ERTS_SMP

#define ERTS_THR_PRGR_DBG_CHK_WAKEUP_REQUEST_VALUE 0

#ifdef DEBUG
#undef ERTS_THR_PRGR_DBG_CHK_WAKEUP_REQUEST_VALUE
#define ERTS_THR_PRGR_DBG_CHK_WAKEUP_REQUEST_VALUE 1
#endif

#define ERTS_THR_PRGR_PRINT_LEADER 0
#define ERTS_THR_PRGR_PRINT_VAL 0
#define ERTS_THR_PRGR_PRINT_BLOCKERS 0

#define ERTS_THR_PRGR_FTL_ERR_BLCK_POLL_INTERVAL 100

#define ERTS_THR_PRGR_LFLG_BLOCK	(((erts_aint32_t) 1) << 31)
#define ERTS_THR_PRGR_LFLG_NO_LEADER	(((erts_aint32_t) 1) << 30)
#define ERTS_THR_PRGR_LFLG_WAITING_UM	(((erts_aint32_t) 1) << 29)
#define ERTS_THR_PRGR_LFLG_ACTIVE_MASK	(~(ERTS_THR_PRGR_LFLG_NO_LEADER	\
					   | ERTS_THR_PRGR_LFLG_BLOCK	\
					   | ERTS_THR_PRGR_LFLG_WAITING_UM))

#define ERTS_THR_PRGR_LFLGS_ACTIVE(LFLGS)				\
    ((LFLGS) & ERTS_THR_PRGR_LFLG_ACTIVE_MASK)

/*
 * We use a 64-bit value for thread progress. By this wrapping of
 * the thread progress will more or less never occur.
 *
 * On 32-bit systems we therefore need a double word atomic.
 */
#undef read_acqb
#define read_acqb erts_thr_prgr_read_acqb__
#undef read_nob
#define read_nob erts_thr_prgr_read_nob__

#ifdef ARCH_64

static ERTS_INLINE void
set_mb(ERTS_THR_PRGR_ATOMIC *atmc, ErtsThrPrgrVal val)
{
    erts_atomic_set_mb(atmc, val);
}

static ERTS_INLINE void
set_nob(ERTS_THR_PRGR_ATOMIC *atmc, ErtsThrPrgrVal val)
{
    erts_atomic_set_nob(atmc, val);
}

static ERTS_INLINE void
init_nob(ERTS_THR_PRGR_ATOMIC *atmc, ErtsThrPrgrVal val)
{
    erts_atomic_init_nob(atmc, val);
}

#else

#undef dw_aint_to_val
#define dw_aint_to_val erts_thr_prgr_dw_aint_to_val__

static void
val_to_dw_aint(erts_dw_aint_t *dw_aint, ErtsThrPrgrVal val)
{
#ifdef ETHR_SU_DW_NAINT_T__
    dw_aint->dw_sint = (ETHR_SU_DW_NAINT_T__) val;
#else
    dw_aint->sint[ERTS_DW_AINT_LOW_WORD]
	= (erts_aint_t) (val & 0xffffffff);
    dw_aint->sint[ERTS_DW_AINT_HIGH_WORD]
	= (erts_aint_t) ((val >> 32) & 0xffffffff);
#endif
}

static ERTS_INLINE void
set_mb(ERTS_THR_PRGR_ATOMIC *atmc, ErtsThrPrgrVal val)
{
    erts_dw_aint_t dw_aint;
    val_to_dw_aint(&dw_aint, val);
    erts_dw_atomic_set_mb(atmc, &dw_aint);
}

static ERTS_INLINE void
set_nob(ERTS_THR_PRGR_ATOMIC *atmc, ErtsThrPrgrVal val)
{
    erts_dw_aint_t dw_aint;
    val_to_dw_aint(&dw_aint, val);
    erts_dw_atomic_set_nob(atmc, &dw_aint);
}

static ERTS_INLINE void
init_nob(ERTS_THR_PRGR_ATOMIC *atmc, ErtsThrPrgrVal val)
{
    erts_dw_aint_t dw_aint;
    val_to_dw_aint(&dw_aint, val);
    erts_dw_atomic_init_nob(atmc, &dw_aint);
}

#endif

/* #define ERTS_THR_PROGRESS_STATE_DEBUG */

#ifdef ERTS_THR_PROGRESS_STATE_DEBUG

#ifdef __GNUC__
#warning "Thread progress state debug is on"
#endif

#define ERTS_THR_PROGRESS_STATE_DEBUG_LEADER	(((erts_aint32_t) 1) << 0)
#define ERTS_THR_PROGRESS_STATE_DEBUG_ACTIVE	(((erts_aint32_t) 1) << 1)

#define ERTS_THR_PROGRESS_STATE_DEBUG_INIT(ID)						\
    erts_atomic32_init_nob(&intrnl->thr[(ID)].data.state_debug,				\
			   ERTS_THR_PROGRESS_STATE_DEBUG_ACTIVE)

#define ERTS_THR_PROGRESS_STATE_DEBUG_SET_ACTIVE(ID, ON)				\
do {											\
    erts_aint32_t state_debug__;							\
    state_debug__ = erts_atomic32_read_nob(&intrnl->thr[(ID)].data.state_debug);	\
    if ((ON))										\
	state_debug__ |= ERTS_THR_PROGRESS_STATE_DEBUG_ACTIVE;				\
    else										\
	state_debug__ &= ~ERTS_THR_PROGRESS_STATE_DEBUG_ACTIVE;				\
    erts_atomic32_set_nob(&intrnl->thr[(ID)].data.state_debug, state_debug__);		\
} while (0)

#define ERTS_THR_PROGRESS_STATE_DEBUG_SET_LEADER(ID, ON)				\
do {											\
    erts_aint32_t state_debug__;							\
    state_debug__ = erts_atomic32_read_nob(&intrnl->thr[(ID)].data.state_debug);	\
    if ((ON))										\
	state_debug__ |= ERTS_THR_PROGRESS_STATE_DEBUG_LEADER;				\
    else										\
	state_debug__ &= ~ERTS_THR_PROGRESS_STATE_DEBUG_LEADER;				\
    erts_atomic32_set_nob(&intrnl->thr[(ID)].data.state_debug, state_debug__);		\
} while (0)

#else

#define ERTS_THR_PROGRESS_STATE_DEBUG_INIT(ID)
#define ERTS_THR_PROGRESS_STATE_DEBUG_SET_ACTIVE(ID, ON)
#define ERTS_THR_PROGRESS_STATE_DEBUG_SET_LEADER(ID, ON)

#endif /* ERTS_THR_PROGRESS_STATE_DEBUG */

#define ERTS_THR_PRGR_BLCKR_INVALID (~((erts_aint32_t) 0))
#define ERTS_THR_PRGR_BLCKR_UNMANAGED (((erts_aint32_t) 1) << 31)

#define ERTS_THR_PRGR_BC_FLG_NOT_BLOCKING (((erts_aint32_t) 1) << 31)

#define ERTS_THR_PRGR_BM_BITS 32
#define ERTS_THR_PRGR_BM_SHIFT 5
#define ERTS_THR_PRGR_BM_MASK 0x1f

#define ERTS_THR_PRGR_WAKEUP_DATA_MASK (ERTS_THR_PRGR_WAKEUP_DATA_SIZE - 1)

#define ERTS_THR_PRGR_WAKEUP_IX(V) \
    ((int) ((V) & ERTS_THR_PRGR_WAKEUP_DATA_MASK))

typedef struct {
    erts_atomic32_t len;
    int id[1];
} ErtsThrPrgrManagedWakeupData;

typedef struct {
    erts_atomic32_t len;
    int high_sz;
    int low_sz;
    erts_atomic32_t *high;
    erts_atomic32_t *low;
} ErtsThrPrgrUnmanagedWakeupData;

typedef struct {
    erts_atomic32_t lflgs;
    erts_atomic32_t block_count;
    erts_atomic_t blocker_event;
    erts_atomic32_t pref_wakeup_used;
    erts_atomic32_t managed_count;
    erts_atomic32_t managed_id;
    erts_atomic32_t unmanaged_id;    
    int chk_next_ix;
    struct {
	int waiting;
	erts_atomic32_t current;
    } umrefc_ix;
} ErtsThrPrgrMiscData;

typedef struct {
    ERTS_THR_PRGR_ATOMIC current;
#ifdef ERTS_THR_PROGRESS_STATE_DEBUG
    erts_atomic32_t state_debug;
#endif
} ErtsThrPrgrElement;

typedef union {
    ErtsThrPrgrElement data;
    char align__[ERTS_ALC_CACHE_LINE_ALIGN_SIZE(sizeof(ErtsThrPrgrElement))];
} ErtsThrPrgrArray;

typedef union {
    erts_atomic_t refc;
    char align__[ERTS_ALC_CACHE_LINE_ALIGN_SIZE(sizeof(erts_atomic_t))];
} ErtsThrPrgrUnmanagedRefc;

typedef struct {
    union {
	ErtsThrPrgrMiscData data;
	char align__[ERTS_ALC_CACHE_LINE_ALIGN_SIZE(
		sizeof(ErtsThrPrgrMiscData))];
    } misc;
    ErtsThrPrgrUnmanagedRefc umrefc[2];
    ErtsThrPrgrArray *thr;
    struct {
	int no;
	ErtsThrPrgrCallbacks *callbacks;
	ErtsThrPrgrManagedWakeupData *data[ERTS_THR_PRGR_WAKEUP_DATA_SIZE];
    } managed;
    struct {
	int no;
	ErtsThrPrgrCallbacks *callbacks;
	ErtsThrPrgrUnmanagedWakeupData *data[ERTS_THR_PRGR_WAKEUP_DATA_SIZE];
    } unmanaged;
} ErtsThrPrgrInternalData;

static ErtsThrPrgrInternalData *intrnl;

ErtsThrPrgr erts_thr_prgr__;

erts_tsd_key_t erts_thr_prgr_data_key__;

static void handle_wakeup_requests(ErtsThrPrgrVal current);
static int got_sched_wakeups(void);
static erts_aint32_t block_thread(ErtsThrPrgrData *tpd);

static ERTS_INLINE void
wakeup_managed(int id)
{
    ErtsThrPrgrCallbacks *cbp = &intrnl->managed.callbacks[id];
    ASSERT(0 <= id && id < intrnl->managed.no);
    cbp->wakeup(cbp->arg);
}


static ERTS_INLINE void
wakeup_unmanaged(int id)
{
    ErtsThrPrgrCallbacks *cbp = &intrnl->unmanaged.callbacks[id];
    ASSERT(0 <= id && id < intrnl->unmanaged.no);
    cbp->wakeup(cbp->arg);
}

static ERTS_INLINE ErtsThrPrgrData *
perhaps_thr_prgr_data(ErtsSchedulerData *esdp)
{
    if (esdp)
	return &esdp->thr_progress_data;
    else
	return erts_tsd_get(erts_thr_prgr_data_key__);
}

static ERTS_INLINE ErtsThrPrgrData *
thr_prgr_data(ErtsSchedulerData *esdp)
{
    ErtsThrPrgrData *tpd = perhaps_thr_prgr_data(esdp);
    ASSERT(tpd);
    return tpd;
}

static void
init_tmp_thr_prgr_data(ErtsThrPrgrData *tpd)
{
    tpd->id = -1;
    tpd->is_managed = 0;
    tpd->is_blocking = 0;
    tpd->is_temporary = 1;
#ifdef ERTS_ENABLE_LOCK_CHECK
    tpd->is_delaying = 0;
#endif
    erts_tsd_set(erts_thr_prgr_data_key__, (void *) tpd);
}

static ERTS_INLINE ErtsThrPrgrData *
tmp_thr_prgr_data(ErtsSchedulerData *esdp)
{
    ErtsThrPrgrData *tpd = perhaps_thr_prgr_data(esdp);

    if (!tpd) {
	/*
	 * We only allocate the part up to the wakeup_request field
	 * which is the first field only used by registered threads
	 */
	tpd = erts_alloc(ERTS_ALC_T_T_THR_PRGR_DATA,
			 offsetof(ErtsThrPrgrData, wakeup_request));
	init_tmp_thr_prgr_data(tpd);
    }

    return tpd;
}

static ERTS_INLINE void
return_tmp_thr_prgr_data(ErtsThrPrgrData *tpd)
{
    if (tpd->is_temporary) {
	erts_tsd_set(erts_thr_prgr_data_key__, NULL);
	erts_free(ERTS_ALC_T_T_THR_PRGR_DATA, tpd);
    }
}

static ERTS_INLINE int
block_count_dec(void)
{
    erts_aint32_t block_count;
    block_count = erts_atomic32_dec_read_mb(&intrnl->misc.data.block_count);
    if (block_count == 0) {
	erts_tse_t *event;
	event = ((erts_tse_t*)
		 erts_atomic_read_nob(&intrnl->misc.data.blocker_event));
	if (event)
	    erts_tse_set(event);
	return 1;
    }

    return (block_count & ERTS_THR_PRGR_BC_FLG_NOT_BLOCKING) == 0;
}

static ERTS_INLINE int
block_count_inc(void)
{
    erts_aint32_t block_count;
    block_count = erts_atomic32_inc_read_mb(&intrnl->misc.data.block_count);
    return (block_count & ERTS_THR_PRGR_BC_FLG_NOT_BLOCKING) == 0;
}


void
erts_thr_progress_pre_init(void)
{
    intrnl = NULL;
    erts_tsd_key_create(&erts_thr_prgr_data_key__);
    init_nob(&erts_thr_prgr__.current, 0);
}

void
erts_thr_progress_init(int no_schedulers, int managed, int unmanaged)
{
    int i, j, um_low, um_high;
    char *ptr;
    size_t cb_sz, intrnl_sz, thr_arr_sz, m_wakeup_size, um_wakeup_size,
	tot_size;

    intrnl_sz = sizeof(ErtsThrPrgrInternalData);
    intrnl_sz = ERTS_ALC_CACHE_LINE_ALIGN_SIZE(intrnl_sz);

    cb_sz = sizeof(ErtsThrPrgrCallbacks)*(managed+unmanaged);
    cb_sz = ERTS_ALC_CACHE_LINE_ALIGN_SIZE(cb_sz);

    thr_arr_sz = sizeof(ErtsThrPrgrArray)*managed;
    ASSERT(thr_arr_sz == ERTS_ALC_CACHE_LINE_ALIGN_SIZE(thr_arr_sz));

    m_wakeup_size = sizeof(ErtsThrPrgrManagedWakeupData);
    m_wakeup_size += (managed - 1)*sizeof(int);
    m_wakeup_size = ERTS_ALC_CACHE_LINE_ALIGN_SIZE(m_wakeup_size);

    um_low = (unmanaged - 1)/ERTS_THR_PRGR_BM_BITS + 1;
    um_high = (um_low - 1)/ERTS_THR_PRGR_BM_BITS + 1;

    um_wakeup_size = sizeof(ErtsThrPrgrUnmanagedWakeupData);
    um_wakeup_size += (um_high + um_low)*sizeof(erts_atomic32_t);
    um_wakeup_size = ERTS_ALC_CACHE_LINE_ALIGN_SIZE(um_wakeup_size);

    tot_size = intrnl_sz;
    tot_size += cb_sz;
    tot_size += thr_arr_sz;
    tot_size += m_wakeup_size*ERTS_THR_PRGR_WAKEUP_DATA_SIZE;
    tot_size += um_wakeup_size*ERTS_THR_PRGR_WAKEUP_DATA_SIZE;

    ptr = erts_alloc_permanent_cache_aligned(ERTS_ALC_T_THR_PRGR_IDATA,
					     tot_size);

    intrnl = (ErtsThrPrgrInternalData *) ptr;
    ptr += intrnl_sz;

    erts_atomic32_init_nob(&intrnl->misc.data.lflgs,
			   ERTS_THR_PRGR_LFLG_NO_LEADER);
    erts_atomic32_init_nob(&intrnl->misc.data.block_count,
			   (ERTS_THR_PRGR_BC_FLG_NOT_BLOCKING
			    | (erts_aint32_t) managed));
    erts_atomic_init_nob(&intrnl->misc.data.blocker_event, ERTS_AINT_NULL);
    erts_atomic32_init_nob(&intrnl->misc.data.pref_wakeup_used, 0);
    erts_atomic32_init_nob(&intrnl->misc.data.managed_count, 0);
    erts_atomic32_init_nob(&intrnl->misc.data.managed_id, no_schedulers);
    erts_atomic32_init_nob(&intrnl->misc.data.unmanaged_id, -1);
    intrnl->misc.data.chk_next_ix = 0;
    intrnl->misc.data.umrefc_ix.waiting = -1;
    erts_atomic32_init_nob(&intrnl->misc.data.umrefc_ix.current, 0);

    erts_atomic_init_nob(&intrnl->umrefc[0].refc, (erts_aint_t) 0);
    erts_atomic_init_nob(&intrnl->umrefc[1].refc, (erts_aint_t) 0);

    intrnl->thr = (ErtsThrPrgrArray *) ptr;
    ptr += thr_arr_sz;
    for (i = 0; i < managed; i++)
	init_nob(&intrnl->thr[i].data.current, 0);

    intrnl->managed.callbacks = (ErtsThrPrgrCallbacks *) ptr;
    intrnl->unmanaged.callbacks = &intrnl->managed.callbacks[managed];
    ptr += cb_sz;

    intrnl->managed.no = managed;
    for (i = 0; i < managed; i++) {
	intrnl->managed.callbacks[i].arg = NULL;
	intrnl->managed.callbacks[i].wakeup = NULL;
    }

    intrnl->unmanaged.no = unmanaged;
    for (i = 0; i < unmanaged; i++) {
	intrnl->unmanaged.callbacks[i].arg = NULL;
	intrnl->unmanaged.callbacks[i].wakeup = NULL;
    }

    for (i = 0; i < ERTS_THR_PRGR_WAKEUP_DATA_SIZE; i++) {
	intrnl->managed.data[i] = (ErtsThrPrgrManagedWakeupData *) ptr;
	erts_atomic32_init_nob(&intrnl->managed.data[i]->len, 0);
	ptr += m_wakeup_size;
    }

    for (i = 0; i < ERTS_THR_PRGR_WAKEUP_DATA_SIZE; i++) {
	erts_atomic32_t *bm;
	intrnl->unmanaged.data[i] = (ErtsThrPrgrUnmanagedWakeupData *) ptr;
	erts_atomic32_init_nob(&intrnl->unmanaged.data[i]->len, 0);
	bm = (erts_atomic32_t *) (ptr + sizeof(ErtsThrPrgrUnmanagedWakeupData));
	intrnl->unmanaged.data[i]->high = bm;
	intrnl->unmanaged.data[i]->high_sz = um_high;
	for (j = 0; j < um_high; j++)
	    erts_atomic32_init_nob(&intrnl->unmanaged.data[i]->high[j], 0);
	intrnl->unmanaged.data[i]->low
	    = &intrnl->unmanaged.data[i]->high[um_high];	
	intrnl->unmanaged.data[i]->low_sz = um_low;
	for (j = 0; j < um_low; j++)
	    erts_atomic32_init_nob(&intrnl->unmanaged.data[i]->low[j], 0);
	ptr += um_wakeup_size;
    }
    ERTS_THR_MEMORY_BARRIER;
}

static void
init_wakeup_request_array(ErtsThrPrgrVal *w)
{
    int i;
    ErtsThrPrgrVal current;

    current = read_acqb(&erts_thr_prgr__.current);
    for (i = 0; i < ERTS_THR_PRGR_WAKEUP_DATA_SIZE; i++) {
	w[i] = current - ((ErtsThrPrgrVal) (ERTS_THR_PRGR_WAKEUP_DATA_SIZE + i));
	if (w[i] > current)
	    w[i]--;
    }
}

void
erts_thr_progress_register_unmanaged_thread(ErtsThrPrgrCallbacks *callbacks)
{
    ErtsThrPrgrData *tpd = perhaps_thr_prgr_data(NULL);
    int is_blocking = 0;

    if (tpd) {
	if (!tpd->is_temporary)
	    erl_exit(ERTS_ABORT_EXIT,
		     "%s:%d:%s(): Double register of thread\n",
		     __FILE__, __LINE__, __func__);
	is_blocking = tpd->is_blocking;
	return_tmp_thr_prgr_data(tpd);
    }

    /*
     * We only allocate the part up to the leader field
     * which is the first field only used by managed threads
     */
    tpd = erts_alloc(ERTS_ALC_T_THR_PRGR_DATA,
		     offsetof(ErtsThrPrgrData, leader));
    tpd->id = (int) erts_atomic32_inc_read_nob(&intrnl->misc.data.unmanaged_id);
    tpd->is_managed = 0;
    tpd->is_blocking = is_blocking;
    tpd->is_temporary = 0;
#ifdef ERTS_ENABLE_LOCK_CHECK
    tpd->is_delaying = 0;
#endif
    ASSERT(tpd->id >= 0);
    if (tpd->id >= intrnl->unmanaged.no)
	erl_exit(ERTS_ABORT_EXIT,
		 "%s:%d:%s(): Too many unmanaged registered threads\n",
		 __FILE__, __LINE__, __func__);

    init_wakeup_request_array(&tpd->wakeup_request[0]);
    erts_tsd_set(erts_thr_prgr_data_key__, (void *) tpd);

    ASSERT(callbacks->wakeup);

    intrnl->unmanaged.callbacks[tpd->id] = *callbacks;
}


void
erts_thr_progress_register_managed_thread(ErtsSchedulerData *esdp,
					  ErtsThrPrgrCallbacks *callbacks,
					  int pref_wakeup)
{
    ErtsThrPrgrData *tpd = perhaps_thr_prgr_data(NULL);
    int is_blocking = 0, managed;

    if (tpd) {
	if (!tpd->is_temporary)
	    erl_exit(ERTS_ABORT_EXIT,
		     "%s:%d:%s(): Double register of thread\n",
		     __FILE__, __LINE__, __func__);
	is_blocking = tpd->is_blocking;
	return_tmp_thr_prgr_data(tpd);
    }

    if (esdp)
	tpd = &esdp->thr_progress_data;
    else
	tpd = erts_alloc(ERTS_ALC_T_THR_PRGR_DATA, sizeof(ErtsThrPrgrData));

    if (pref_wakeup
	&& !erts_atomic32_xchg_nob(&intrnl->misc.data.pref_wakeup_used, 1))
	tpd->id = 0;
    else if (esdp)
	tpd->id = (int) esdp->no;
    else
	tpd->id = erts_atomic32_inc_read_nob(&intrnl->misc.data.managed_id);
    ASSERT(tpd->id >= 0);
    if (tpd->id >= intrnl->managed.no)
	erl_exit(ERTS_ABORT_EXIT,
		 "%s:%d:%s(): Too many managed registered threads\n",
		 __FILE__, __LINE__, __func__);

    tpd->is_managed = 1;
    tpd->is_blocking = is_blocking;
    tpd->is_temporary = 0;
#ifdef ERTS_ENABLE_LOCK_CHECK
    tpd->is_delaying = 1;
#endif

    init_wakeup_request_array(&tpd->wakeup_request[0]);

    ERTS_THR_PROGRESS_STATE_DEBUG_INIT(tpd->id);

    tpd->leader = 0;
    tpd->active = 1;
    tpd->confirmed = 0;
    tpd->leader_state.current = ERTS_THR_PRGR_VAL_WAITING;
    erts_tsd_set(erts_thr_prgr_data_key__, (void *) tpd);

    erts_atomic32_inc_nob(&intrnl->misc.data.lflgs);

    ASSERT(callbacks->wakeup);
    ASSERT(callbacks->prepare_wait);
    ASSERT(callbacks->wait);
    ASSERT(callbacks->finalize_wait);

    intrnl->managed.callbacks[tpd->id] = *callbacks;

    callbacks->prepare_wait(callbacks->arg);
    managed = erts_atomic32_inc_read_relb(&intrnl->misc.data.managed_count);
    if (managed != intrnl->managed.no) {
	/* Wait until all managed threads have registered... */
	do {
	    callbacks->wait(callbacks->arg);
	    callbacks->prepare_wait(callbacks->arg);
	    managed = erts_atomic32_read_acqb(&intrnl->misc.data.managed_count);
	} while (managed != intrnl->managed.no);
    }
    else {
	int id;
	/* All managed threads have registered; lets go... */
	for (id = 0; id < managed; id++)
	    if (id != tpd->id)
		wakeup_managed(id);
    }
    callbacks->finalize_wait(callbacks->arg);
}

static ERTS_INLINE int
leader_update(ErtsThrPrgrData *tpd)
{
#ifdef ERTS_ENABLE_LOCK_CHECK
    erts_lc_check_exact(NULL, 0);
#endif
    if (!tpd->leader) {
	/* Probably need to block... */
	block_thread(tpd);
    }
    else {
	ErtsThrPrgrVal current;
	int ix, chk_next_ix, umrefc_ix, my_ix, no_managed, waiting_unmanaged;
	erts_aint32_t lflgs;
	ErtsThrPrgrVal next;
	erts_aint_t refc;

	my_ix = tpd->id;

	if (tpd->leader_state.current == ERTS_THR_PRGR_VAL_WAITING) {
	    /* Took over as leader from another thread */
	    tpd->leader_state.current = read_nob(&erts_thr_prgr__.current);
	    tpd->leader_state.next = tpd->leader_state.current;
	    tpd->leader_state.next++;
	    if (tpd->leader_state.next == ERTS_THR_PRGR_VAL_WAITING)
		tpd->leader_state.next = 0;
	    tpd->leader_state.chk_next_ix = intrnl->misc.data.chk_next_ix;
	    tpd->leader_state.umrefc_ix.waiting = intrnl->misc.data.umrefc_ix.waiting;
	    tpd->leader_state.umrefc_ix.current =
		(int) erts_atomic32_read_nob(&intrnl->misc.data.umrefc_ix.current);

	    if (tpd->confirmed == tpd->leader_state.current) {
		ErtsThrPrgrVal val = tpd->leader_state.current + 1;
		if (val == ERTS_THR_PRGR_VAL_WAITING)
		    val = 0;
		tpd->confirmed = val;
		set_mb(&intrnl->thr[my_ix].data.current, val);
	    }
	}


	next = tpd->leader_state.next;

	waiting_unmanaged = 0;
	umrefc_ix = -1; /* Shut up annoying warning */

	chk_next_ix = tpd->leader_state.chk_next_ix;
	no_managed = intrnl->managed.no;
	ASSERT(0 <= chk_next_ix && chk_next_ix <= no_managed);
	/* Check manged threads */
	if (chk_next_ix < no_managed) {
	    for (ix = chk_next_ix; ix < no_managed; ix++) {
		ErtsThrPrgrVal tmp;
		if (ix == my_ix)
		    continue;
		tmp = read_nob(&intrnl->thr[ix].data.current);
		if (tmp != next && tmp != ERTS_THR_PRGR_VAL_WAITING) {
		    tpd->leader_state.chk_next_ix = ix;
		    ASSERT(erts_thr_progress_has_passed__(next, tmp));
		    goto done;
		}
	    }
	}

	/* Check unmanged threads */
	waiting_unmanaged = tpd->leader_state.umrefc_ix.waiting != -1;
	umrefc_ix = (waiting_unmanaged
		     ? tpd->leader_state.umrefc_ix.waiting
		     : tpd->leader_state.umrefc_ix.current);
	refc = erts_atomic_read_nob(&intrnl->umrefc[umrefc_ix].refc);
	ASSERT(refc >= 0);
	if (refc != 0) {
	    int new_umrefc_ix;

	    if (waiting_unmanaged)
		goto done;

	    new_umrefc_ix = (umrefc_ix + 1) & 0x1;
	    tpd->leader_state.umrefc_ix.waiting = umrefc_ix;
	    tpd->leader_state.chk_next_ix = no_managed;
	    erts_atomic32_set_nob(&intrnl->misc.data.umrefc_ix.current,
				  (erts_aint32_t) new_umrefc_ix);
	    ETHR_MEMBAR(ETHR_StoreLoad);
	    refc = erts_atomic_read_nob(&intrnl->umrefc[umrefc_ix].refc);
	    ASSERT(refc >= 0);
	    waiting_unmanaged = 1;
	    if (refc != 0)
		goto done;
	}

	/* Make progress */
	current = next;

	next++;
	if (next == ERTS_THR_PRGR_VAL_WAITING)
	    next = 0;

	set_nob(&intrnl->thr[my_ix].data.current, next);
	set_mb(&erts_thr_prgr__.current, current);
	tpd->confirmed = next;
	tpd->leader_state.next = next;
	tpd->leader_state.current = current;

#if ERTS_THR_PRGR_PRINT_VAL
	if (current % 1000 == 0)
	    erts_fprintf(stderr, "%b64u\n", current);
#endif
	handle_wakeup_requests(current);

	if (waiting_unmanaged) {
	    waiting_unmanaged = 0;
	    tpd->leader_state.umrefc_ix.waiting = -1;
	    erts_atomic32_read_band_nob(&intrnl->misc.data.lflgs,
					~ERTS_THR_PRGR_LFLG_WAITING_UM);
	}
	tpd->leader_state.chk_next_ix = 0;

    done:

	if (tpd->active) {
	    lflgs = erts_atomic32_read_nob(&intrnl->misc.data.lflgs);
	    if (lflgs & ERTS_THR_PRGR_LFLG_BLOCK)
		(void) block_thread(tpd);
	}
	else {
	    int force_wakeup_check = 0;
	    erts_aint32_t set_flags = ERTS_THR_PRGR_LFLG_NO_LEADER;
	    tpd->leader = 0;
	    tpd->leader_state.current = ERTS_THR_PRGR_VAL_WAITING;
#if ERTS_THR_PRGR_PRINT_LEADER
	    erts_fprintf(stderr, "L <- %d\n", tpd->id);
#endif

	    ERTS_THR_PROGRESS_STATE_DEBUG_SET_LEADER(tpd->id, 0);

	    if (waiting_unmanaged)
		set_flags |= ERTS_THR_PRGR_LFLG_WAITING_UM;

	    lflgs = erts_atomic32_read_bor_relb(&intrnl->misc.data.lflgs,
						set_flags);
	    lflgs |= set_flags;
	    if (lflgs & ERTS_THR_PRGR_LFLG_BLOCK)
		lflgs = block_thread(tpd);

	    if (waiting_unmanaged) {
		/* Need to check umrefc again */
		ETHR_MEMBAR(ETHR_StoreLoad);
		refc = erts_atomic_read_nob(&intrnl->umrefc[umrefc_ix].refc);
		if (refc == 0) {
		    /* Need to force wakeup check */
		    force_wakeup_check = 1;
		}
	    }

	    if ((force_wakeup_check
		 || ((lflgs & (ERTS_THR_PRGR_LFLG_NO_LEADER
			       | ERTS_THR_PRGR_LFLG_WAITING_UM
			       | ERTS_THR_PRGR_LFLG_ACTIVE_MASK))
		     == ERTS_THR_PRGR_LFLG_NO_LEADER))
		&& got_sched_wakeups()) {
		/* Someone need to make progress */
		wakeup_managed(0);
	    }
	}
    }

    return tpd->leader;
}

static int
update(ErtsThrPrgrData *tpd)
{
    int res;
    ErtsThrPrgrVal val;

    if (tpd->leader)
	res = 1;
    else {
	erts_aint32_t lflgs;
	res = 0;
	val = read_acqb(&erts_thr_prgr__.current);
	if (tpd->confirmed == val) {
	    val++;
	    if (val == ERTS_THR_PRGR_VAL_WAITING)
		val = 0;
	    tpd->confirmed = val;
	    set_mb(&intrnl->thr[tpd->id].data.current, val);
	}

	lflgs = erts_atomic32_read_nob(&intrnl->misc.data.lflgs);
	if (lflgs & ERTS_THR_PRGR_LFLG_BLOCK)
	    res = 1; /* Need to block in leader_update() */

	if ((lflgs & ERTS_THR_PRGR_LFLG_NO_LEADER)
	    && (tpd->active || ERTS_THR_PRGR_LFLGS_ACTIVE(lflgs) == 0)) {
	    /* Try to take over leadership... */
	    erts_aint32_t olflgs;
	    olflgs = erts_atomic32_read_band_acqb(
		&intrnl->misc.data.lflgs,
		~ERTS_THR_PRGR_LFLG_NO_LEADER);
	    if (olflgs & ERTS_THR_PRGR_LFLG_NO_LEADER) {
		tpd->leader = 1;
#if ERTS_THR_PRGR_PRINT_LEADER
		erts_fprintf(stderr, "L -> %d\n", tpd->id);
#endif
		ERTS_THR_PROGRESS_STATE_DEBUG_SET_LEADER(tpd->id, 1);
	    }
	}
	res |= tpd->leader;
    }
    return res;
}

int
erts_thr_progress_update(ErtsSchedulerData *esdp)
{
    return update(thr_prgr_data(esdp));
}


int
erts_thr_progress_leader_update(ErtsSchedulerData *esdp)
{
    return leader_update(thr_prgr_data(esdp));
}

void
erts_thr_progress_prepare_wait(ErtsSchedulerData *esdp)
{
    erts_aint32_t lflgs;
    ErtsThrPrgrData *tpd = thr_prgr_data(esdp);

#ifdef ERTS_ENABLE_LOCK_CHECK
    erts_lc_check_exact(NULL, 0);
#endif

    block_count_dec();

    tpd->confirmed = ERTS_THR_PRGR_VAL_WAITING;
    set_mb(&intrnl->thr[tpd->id].data.current, ERTS_THR_PRGR_VAL_WAITING);

    lflgs = erts_atomic32_read_nob(&intrnl->misc.data.lflgs);

    if ((lflgs & (ERTS_THR_PRGR_LFLG_NO_LEADER
		  | ERTS_THR_PRGR_LFLG_WAITING_UM
		  | ERTS_THR_PRGR_LFLG_ACTIVE_MASK))
	== ERTS_THR_PRGR_LFLG_NO_LEADER 
	&& got_sched_wakeups()) {
	/* Someone need to make progress */
	wakeup_managed(0);
    }
}

void
erts_thr_progress_finalize_wait(ErtsSchedulerData *esdp)
{
    ErtsThrPrgrData *tpd = thr_prgr_data(esdp);
    ErtsThrPrgrVal current, val;

#ifdef ERTS_ENABLE_LOCK_CHECK
    erts_lc_check_exact(NULL, 0);
#endif

    /*
     * We aren't allowed to continue until our thread
     * progress is past global current.
     */
    val = current = read_acqb(&erts_thr_prgr__.current);
    while (1) {
	val++;
	if (val == ERTS_THR_PRGR_VAL_WAITING)
	    val = 0;
	tpd->confirmed = val;
	set_mb(&intrnl->thr[tpd->id].data.current, val);
	val = read_acqb(&erts_thr_prgr__.current);
	if (current == val)
	    break;
	current = val;
    }
    if (block_count_inc())
	block_thread(tpd);
    if (update(tpd))
	leader_update(tpd);
}

void
erts_thr_progress_active(ErtsSchedulerData *esdp, int on)
{
    ErtsThrPrgrData *tpd = thr_prgr_data(esdp);

#ifdef ERTS_ENABLE_LOCK_CHECK
    erts_lc_check_exact(NULL, 0);
#endif

    ERTS_THR_PROGRESS_STATE_DEBUG_SET_ACTIVE(tpd->id, on);

    if (on) {
	ASSERT(!tpd->active);
	tpd->active = 1;
	erts_atomic32_inc_nob(&intrnl->misc.data.lflgs);
    }
    else {
	ASSERT(tpd->active);
	tpd->active = 0;
	erts_atomic32_dec_nob(&intrnl->misc.data.lflgs);
	if (update(tpd))
	    leader_update(tpd);
    }

#ifdef DEBUG
    {
	erts_aint32_t n = erts_atomic32_read_nob(&intrnl->misc.data.lflgs);
	n &= ERTS_THR_PRGR_LFLG_ACTIVE_MASK;
	ASSERT(tpd->active <= n && n <= intrnl->managed.no);
    }
#endif

}

static ERTS_INLINE void
unmanaged_continue(ErtsThrPrgrDelayHandle handle)
{
    int umrefc_ix = (int) handle;
    erts_aint_t refc;

    ASSERT(umrefc_ix == 0 || umrefc_ix == 1);
    refc = erts_atomic_dec_read_relb(&intrnl->umrefc[umrefc_ix].refc);
    ASSERT(refc >= 0);
    if (refc == 0) {
	erts_aint_t lflgs;
	ERTS_THR_READ_MEMORY_BARRIER;
	lflgs = erts_atomic32_read_nob(&intrnl->misc.data.lflgs);
	if ((lflgs & (ERTS_THR_PRGR_LFLG_NO_LEADER
		      | ERTS_THR_PRGR_LFLG_WAITING_UM
		      | ERTS_THR_PRGR_LFLG_ACTIVE_MASK))
	    == (ERTS_THR_PRGR_LFLG_NO_LEADER|ERTS_THR_PRGR_LFLG_WAITING_UM)
	    && got_sched_wakeups()) {
	    /* Others waiting for us... */
	    wakeup_managed(0);
	}
    }
}

void
erts_thr_progress_unmanaged_continue__(ErtsThrPrgrDelayHandle handle)
{
#ifdef ERTS_ENABLE_LOCK_CHECK
    ErtsThrPrgrData *tpd = perhaps_thr_prgr_data(NULL);
    ERTS_LC_ASSERT(tpd && tpd->is_delaying);
    tpd->is_delaying = 0;
    return_tmp_thr_prgr_data(tpd);
#endif
    ASSERT(!erts_thr_progress_is_managed_thread());

    unmanaged_continue(handle);
}

ErtsThrPrgrDelayHandle
erts_thr_progress_unmanaged_delay__(void)
{
    int umrefc_ix;
    ASSERT(!erts_thr_progress_is_managed_thread());
    umrefc_ix = (int) erts_atomic32_read_acqb(&intrnl->misc.data.umrefc_ix.current);
    while (1) {
	int tmp_ix;
	erts_atomic_inc_acqb(&intrnl->umrefc[umrefc_ix].refc);
	tmp_ix = (int) erts_atomic32_read_acqb(&intrnl->misc.data.umrefc_ix.current);
	if (tmp_ix == umrefc_ix)
	    break;
	unmanaged_continue(umrefc_ix);
	umrefc_ix = tmp_ix;
    }
#ifdef ERTS_ENABLE_LOCK_CHECK
    {
	ErtsThrPrgrData *tpd = tmp_thr_prgr_data(NULL);
	tpd->is_delaying = 1;
    }
#endif
    return (ErtsThrPrgrDelayHandle) umrefc_ix;
}

static ERTS_INLINE int
has_reached_wakeup(ErtsThrPrgrVal wakeup)
{
    /*
     * Exactly the same as erts_thr_progress_has_reached(), but
     * also verify valid wakeup requests in debug mode.
     */
    ErtsThrPrgrVal current;

    current = read_acqb(&erts_thr_prgr__.current);

#if ERTS_THR_PRGR_DBG_CHK_WAKEUP_REQUEST_VALUE
    {
	ErtsThrPrgrVal limit;
	/*
	 * erts_thr_progress_later() returns values which are
	 * equal to 'current + 2', or 'current + 3'. That is, users
	 * should never get a hold of values larger than that.
	 *
	 * That is, valid values are values less than 'current + 4'.
	 *
	 * Values larger than this won't work with the wakeup
	 * algorithm.
	 */

	limit = current + 4;
	if (limit == ERTS_THR_PRGR_VAL_WAITING)
	    limit = 0;
	else if (limit < current) /* Wrapped */
	    limit += 1;

	if (!erts_thr_progress_has_passed__(limit, wakeup))
	    erl_exit(ERTS_ABORT_EXIT,
		     "Invalid wakeup request value found:"
		     " current=%b64u, wakeup=%b64u, limit=%b64u",
		     current, wakeup, limit);
    }
#endif

    if (current == wakeup)
	return 1;
    return erts_thr_progress_has_passed__(current, wakeup);
}

static void
request_wakeup_managed(ErtsThrPrgrData *tpd, ErtsThrPrgrVal value)
{
    ErtsThrPrgrManagedWakeupData *mwd;
    int ix, wix;

    /*
     * Only managed threads that aren't in waiting state
     * are allowed to call this function.
     */

    ASSERT(tpd->is_managed);
    ASSERT(tpd->confirmed != ERTS_THR_PRGR_VAL_WAITING);

    if (has_reached_wakeup(value)) {
	wakeup_managed(tpd->id);
	return;
    }

    wix = ERTS_THR_PRGR_WAKEUP_IX(value);
    if (tpd->wakeup_request[wix] == value)
	return; /* Already got a request registered */

    ASSERT(erts_thr_progress_has_passed__(value,
					  tpd->wakeup_request[wix]));


    if (tpd->confirmed == value) {
	/*
	 * We have already confirmed this value. We need to request
	 * wakeup for a value later than our latest confirmed value in
	 * order to prevent progress from reaching the requested value
	 * while we are writing the request.
	 *
	 * It is ok to move the wakeup request forward since the only
	 * guarantee we make (and can make) is that the thread will be
	 * woken some time *after* the requested value has been reached.
	 */
	value++;
	if (value == ERTS_THR_PRGR_VAL_WAITING)
	    value = 0;

	wix = ERTS_THR_PRGR_WAKEUP_IX(value);
	if (tpd->wakeup_request[wix] == value)
	    return; /* Already got a request registered */

	ASSERT(erts_thr_progress_has_passed__(value,
					      tpd->wakeup_request[wix]));
    }

    tpd->wakeup_request[wix] = value;

    mwd = intrnl->managed.data[wix];

    ix = erts_atomic32_inc_read_nob(&mwd->len) - 1;
#if ERTS_THR_PRGR_DBG_CHK_WAKEUP_REQUEST_VALUE
    if (ix >= intrnl->managed.no)
	erl_exit(ERTS_ABORT_EXIT, "Internal error: Too many wakeup requests\n");
#endif
    mwd->id[ix] = tpd->id;

    ASSERT(!erts_thr_progress_has_reached(value));

    /*
     * This thread is guarranteed to issue a full memory barrier:
     * - after the request has been written, but
     * - before the global thread progress reach the (possibly
     *   increased) requested wakeup value.
     */
}

static void
request_wakeup_unmanaged(ErtsThrPrgrData *tpd, ErtsThrPrgrVal value)
{
    int wix, ix, id, bit;
    ErtsThrPrgrUnmanagedWakeupData *umwd;

    ASSERT(!tpd->is_managed);

    /*
     * Thread progress *can* reach and pass our requested value while
     * we are writing the request.
     */

    if (has_reached_wakeup(value)) {
	wakeup_unmanaged(tpd->id);
	return;
    }

    wix = ERTS_THR_PRGR_WAKEUP_IX(value);

    if (tpd->wakeup_request[wix] == value)
	return; /* Already got a request registered */

    ASSERT(erts_thr_progress_has_passed__(value,
					  tpd->wakeup_request[wix]));

    umwd = intrnl->unmanaged.data[wix];

    id = tpd->id;

    bit = id & ERTS_THR_PRGR_BM_MASK;
    ix = id >> ERTS_THR_PRGR_BM_SHIFT;
    ASSERT(0 <= ix && ix < umwd->low_sz);
    erts_atomic32_read_bor_nob(&umwd->low[ix], 1 << bit);

    bit = ix & ERTS_THR_PRGR_BM_MASK;
    ix >>= ERTS_THR_PRGR_BM_SHIFT;
    ASSERT(0 <= ix && ix < umwd->high_sz);
    erts_atomic32_read_bor_nob(&umwd->high[ix], 1 << bit);

    erts_atomic32_inc_mb(&umwd->len);

    if (erts_thr_progress_has_reached(value))
	wakeup_unmanaged(tpd->id);
    else
	tpd->wakeup_request[wix] = value;
}

void
erts_thr_progress_wakeup(ErtsSchedulerData *esdp,
			 ErtsThrPrgrVal value)
{
    ErtsThrPrgrData *tpd = thr_prgr_data(esdp);
    ASSERT(!tpd->is_temporary);
    if (tpd->is_managed)
	request_wakeup_managed(tpd, value);
    else
	request_wakeup_unmanaged(tpd, value);
}

static void
wakeup_unmanaged_threads(ErtsThrPrgrUnmanagedWakeupData *umwd)
{
    int hix;
    for (hix = 0; hix < umwd->high_sz; hix++) {
	erts_aint32_t hmask = erts_atomic32_read_nob(&umwd->high[hix]);
	if (hmask) {
	    int hbase = hix << ERTS_THR_PRGR_BM_SHIFT;
	    int hbit;
	    for (hbit = 0; hbit < ERTS_THR_PRGR_BM_BITS; hbit++) {
		if (hmask & (1 << hbit)) {
		    erts_aint_t lmask;
		    int lix = hbase + hbit;
		    ASSERT(0 <= lix && lix < umwd->low_sz);
		    lmask = erts_atomic32_read_nob(&umwd->low[lix]);
		    if (lmask) {
			int lbase = lix << ERTS_THR_PRGR_BM_SHIFT;
			int lbit;
			for (lbit = 0; lbit < ERTS_THR_PRGR_BM_BITS; lbit++) {
			    if (lmask & (1 << lbit)) {
				int id = lbase + lbit;
				wakeup_unmanaged(id);
			    }
			}
			erts_atomic32_set_nob(&umwd->low[lix], 0);
		    }
		}
	    }
	    erts_atomic32_set_nob(&umwd->high[hix], 0);
	}
    }
}


static void
handle_wakeup_requests(ErtsThrPrgrVal current)
{
    ErtsThrPrgrManagedWakeupData *mwd;
    ErtsThrPrgrUnmanagedWakeupData *umwd;
    int wix, len, i;

    wix = ERTS_THR_PRGR_WAKEUP_IX(current);

    mwd = intrnl->managed.data[wix];
    len = erts_atomic32_read_nob(&mwd->len);
    ASSERT(len >= 0);
    if (len) {
	for (i = 0; i < len; i++)
	    wakeup_managed(mwd->id[i]);
	erts_atomic32_set_nob(&mwd->len, 0);
    }

    umwd = intrnl->unmanaged.data[wix];
    len = erts_atomic32_read_nob(&umwd->len);
    ASSERT(len >= 0);
    if (len) {
	wakeup_unmanaged_threads(umwd);
	erts_atomic32_set_nob(&umwd->len, 0);
    }

}

static int
got_sched_wakeups(void)
{
    int wix;

    ERTS_THR_MEMORY_BARRIER;

    for (wix = 0; wix < ERTS_THR_PRGR_WAKEUP_DATA_SIZE; wix++) {
 	ErtsThrPrgrManagedWakeupData **mwd = intrnl->managed.data;
	if (erts_atomic32_read_nob(&mwd[wix]->len))
	    return 1;
    }
    for (wix = 0; wix < ERTS_THR_PRGR_WAKEUP_DATA_SIZE; wix++) {
 	ErtsThrPrgrUnmanagedWakeupData **umwd = intrnl->unmanaged.data;
	if (erts_atomic32_read_nob(&umwd[wix]->len))
	    return 1;
    }
    return 0;
}

static erts_aint32_t
block_thread(ErtsThrPrgrData *tpd)
{
    erts_aint32_t lflgs;
    ErtsThrPrgrCallbacks *cbp = &intrnl->managed.callbacks[tpd->id];

    do {
	block_count_dec();

	while (1) {
	    cbp->prepare_wait(cbp->arg);
	    lflgs = erts_atomic32_read_nob(&intrnl->misc.data.lflgs);
	    if (lflgs & ERTS_THR_PRGR_LFLG_BLOCK)
		cbp->wait(cbp->arg);
	    else
		break;
	}

    } while (block_count_inc());

    cbp->finalize_wait(cbp->arg);

    return lflgs;
}

static erts_aint32_t
thr_progress_block(ErtsThrPrgrData *tpd, int wait)
{
    erts_tse_t *event = NULL; /* Remove erroneous warning... sigh... */
    erts_aint32_t lflgs, bc;

    if (tpd->is_blocking++)
	return (erts_aint32_t) 0;

    while (1) {
	lflgs = erts_atomic32_read_bor_nob(&intrnl->misc.data.lflgs,
					   ERTS_THR_PRGR_LFLG_BLOCK);
	if (lflgs & ERTS_THR_PRGR_LFLG_BLOCK)
	    block_thread(tpd);
	else
	    break;
    }

#if ERTS_THR_PRGR_PRINT_BLOCKERS
    erts_fprintf(stderr, "block(%d)\n", tpd->id);
#endif

    ASSERT(ERTS_AINT_NULL
	   == erts_atomic_read_nob(&intrnl->misc.data.blocker_event));

    if (wait) {
	event = erts_tse_fetch();
	erts_tse_reset(event);
	erts_atomic_set_nob(&intrnl->misc.data.blocker_event,
			    (erts_aint_t) event);
    }
    if (tpd->is_managed)
	erts_atomic32_dec_nob(&intrnl->misc.data.block_count);
    bc = erts_atomic32_read_band_mb(&intrnl->misc.data.block_count,
				    ~ERTS_THR_PRGR_BC_FLG_NOT_BLOCKING);
    bc &= ~ERTS_THR_PRGR_BC_FLG_NOT_BLOCKING;
    if (wait) {
	while (bc != 0) {
	    erts_tse_wait(event);
	    erts_tse_reset(event);
	    bc = erts_atomic32_read_acqb(&intrnl->misc.data.block_count);
	}
    }
    return bc;

}

void
erts_thr_progress_block(void)
{
    thr_progress_block(tmp_thr_prgr_data(NULL), 1);
}

void
erts_thr_progress_fatal_error_block(SWord timeout,
				    ErtsThrPrgrData *tmp_tpd_bufp)
{
    ErtsThrPrgrData *tpd = perhaps_thr_prgr_data(NULL);
    erts_aint32_t bc;
    SWord time_left = timeout;
    SysTimeval to;

    /*
     * Counting poll intervals may give us a too long timeout
     * if cpu is busy. If we got tolerant time of day we use it
     * to prevent this.
     */
    if (!erts_disable_tolerant_timeofday) {
	erts_get_timeval(&to);
	to.tv_sec += timeout / 1000;
	to.tv_sec += timeout % 1000;
    }

    if (!tpd) {
	/*
	 * We stack allocate since failure to allocate memory may
	 * have caused the problem in the first place. This is ok
	 * since we never complete an unblock after a fatal error
	 * block.
	 */
	tpd = tmp_tpd_bufp;
	init_tmp_thr_prgr_data(tpd);
    }

    bc = thr_progress_block(tpd, 0);
    if (bc == 0)
	return; /* Succefully blocked all managed threads */

    while (1) {
	if (erts_milli_sleep(ERTS_THR_PRGR_FTL_ERR_BLCK_POLL_INTERVAL) == 0)
	    time_left -= ERTS_THR_PRGR_FTL_ERR_BLCK_POLL_INTERVAL;
	bc = erts_atomic32_read_acqb(&intrnl->misc.data.block_count);
	if (bc == 0)
	    break; /* Succefully blocked all managed threads */
	if (time_left <= 0)
	    break; /* Timeout */
	if (!erts_disable_tolerant_timeofday) {
	    SysTimeval now;
	    erts_get_timeval(&now);
	    if (now.tv_sec > to.tv_sec)
		break; /* Timeout */
	    if (now.tv_sec == to.tv_sec && now.tv_usec >= to.tv_usec)
		break; /* Timeout */
	}
    }
}

void
erts_thr_progress_unblock(void)
{
    erts_tse_t *event;
    int id, break_id, sz, wakeup;
    ErtsThrPrgrData *tpd = thr_prgr_data(NULL);

    ASSERT(tpd->is_blocking);
    if (--tpd->is_blocking)
	return;

    sz = intrnl->managed.no;

    wakeup = 1;
    if (!tpd->is_managed)
	id = break_id = tpd->id < 0 ? 0 : tpd->id % sz;
    else {
	break_id = tpd->id;
	id = break_id + 1;
	if (id >= sz)
	    id = 0;
	if (id == break_id)
	    wakeup = 0;
	erts_atomic32_inc_nob(&intrnl->misc.data.block_count);
    }

    event = ((erts_tse_t *)
	     erts_atomic_read_nob(&intrnl->misc.data.blocker_event));
    ASSERT(event);
    erts_atomic_set_nob(&intrnl->misc.data.blocker_event, ERTS_AINT_NULL);

    erts_atomic32_read_bor_relb(&intrnl->misc.data.block_count,
				ERTS_THR_PRGR_BC_FLG_NOT_BLOCKING);
#if ERTS_THR_PRGR_PRINT_BLOCKERS
    erts_fprintf(stderr, "unblock(%d)\n", tpd->id);
#endif
    erts_atomic32_read_band_mb(&intrnl->misc.data.lflgs,
			       ~ERTS_THR_PRGR_LFLG_BLOCK);

    if (wakeup) {
	do {
	    ErtsThrPrgrVal tmp;
	    tmp = read_nob(&intrnl->thr[id].data.current);
	    if (tmp != ERTS_THR_PRGR_VAL_WAITING)
		wakeup_managed(id);
	    if (++id >= sz)
		id = 0;
	} while (id != break_id);
    }

    return_tmp_thr_prgr_data(tpd);
    erts_tse_return(event);
}

int
erts_thr_progress_is_blocking(void)
{
    ErtsThrPrgrData *tpd = perhaps_thr_prgr_data(NULL);
    return tpd && tpd->is_blocking;
}

void erts_thr_progress_dbg_print_state(void)
{
    int id;
    int sz = intrnl->managed.no;

    erts_fprintf(stderr, "--- thread progress ---\n");
    erts_fprintf(stderr,"current=%b64u\n", erts_thr_progress_current());
    for (id = 0; id < sz; id++) {
	ErtsThrPrgrVal current = read_nob(&intrnl->thr[id].data.current);
#ifdef ERTS_THR_PROGRESS_STATE_DEBUG
	erts_aint32_t state_debug;
	char *active, *leader;

	state_debug = erts_atomic32_read_nob(&intrnl->thr[id].data.state_debug);
	active = (state_debug & ERTS_THR_PROGRESS_STATE_DEBUG_ACTIVE
		  ? "true"
		  : "false");
	leader = (state_debug & ERTS_THR_PROGRESS_STATE_DEBUG_LEADER
		  ? "true"
		  : "false");
#endif
	if (current == ERTS_THR_PRGR_VAL_WAITING)
	    erts_fprintf(stderr,
			 "  id=%d, current=WAITING"
#ifdef ERTS_THR_PROGRESS_STATE_DEBUG
			 ", active=%s, leader=%s"
#endif
			 "\n", id
#ifdef ERTS_THR_PROGRESS_STATE_DEBUG
			 , active, leader
#endif
		);
	else
	    erts_fprintf(stderr,
			 "  id=%d, current=%b64u"
#ifdef ERTS_THR_PROGRESS_STATE_DEBUG
			 ", active=%s, leader=%s"
#endif
			 "\n", id, current
#ifdef ERTS_THR_PROGRESS_STATE_DEBUG
			 , active, leader
#endif
		);
    }
    erts_fprintf(stderr, "-----------------------\n");
    

}

#endif