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
|
/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
* Copyright by The HDF Group. *
* All rights reserved. *
* *
* This file is part of HDF5. The full HDF5 copyright notice, including *
* terms governing use, modification, and redistribution, is contained in *
* the COPYING file, which can be found at the root of the source code *
* distribution tree, or in https://www.hdfgroup.org/licenses. *
* If you do not have access to either file, you may request a copy from *
* help@hdfgroup.org. *
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
/*
* Collective file open optimization tests
*
*/
#include "testpar.h"
#include "H5Dprivate.h"
/* The collection of files is included below to aid
* an external "cleanup" process if required.
*
* Note that the code below relies on the ordering of this array
* since each set of three is used by the tests either to construct
* or to read and validate.
*/
#define NFILENAME 3
const char *FILENAMES[NFILENAME + 1] = {"reloc_t_pread_data_file", "reloc_t_pread_group_0_file",
"reloc_t_pread_group_1_file", NULL};
#define FILENAME_BUF_SIZE 1024
#define COUNT 1000
#define LIMIT_NPROC 6
bool pass = true;
static const char *random_hdf5_text = "Now is the time for all first-time-users of HDF5 to read their \
manual or go thru the tutorials!\n\
While you\'re at it, now is also the time to read up on MPI-IO.";
static const char *hitchhiker_quote = "A common mistake that people make when trying to design something\n\
completely foolproof is to underestimate the ingenuity of complete\n\
fools.\n";
static int generate_test_file(MPI_Comm comm, int mpi_rank, int group);
static int test_parallel_read(MPI_Comm comm, int mpi_rank, int mpi_size, int group);
static char *test_argv0 = NULL;
/*-------------------------------------------------------------------------
* Function: generate_test_file
*
* Purpose: This function is called to produce an HDF5 data file
* whose superblock is relocated to a power-of-2 boundary.
*
* Since data will be read back and validated, we generate
* data in a predictable manner rather than randomly.
* For now, we simply use the global mpi_rank of the writing
* process as a starting component for the data generation.
* Subsequent writes are increments from the initial start
* value.
*
* In the overall scheme of running the test, we'll call
* this function twice: first as a collection of all MPI
* processes and then a second time with the processes split
* more or less in half. Each sub group will operate
* collectively on their assigned file. This split into
* subgroups validates that parallel groups can successfully
* open and read data independently from the other parallel
* operations taking place.
*
* Return: Success: 0
*
* Failure: 1
*-------------------------------------------------------------------------
*/
static int
generate_test_file(MPI_Comm comm, int mpi_rank, int group_id)
{
int header = -1;
const char *fcn_name = "generate_test_file()";
const char *failure_mssg = NULL;
const char *group_filename = NULL;
char data_filename[FILENAME_BUF_SIZE];
int file_index = 0;
int group_size;
int group_rank;
int local_failure = 0;
int global_failures = 0;
hsize_t count = COUNT;
hsize_t i;
hsize_t offset;
hsize_t dims[1] = {0};
hid_t file_id = H5I_INVALID_HID;
hid_t memspace = H5I_INVALID_HID;
hid_t filespace = H5I_INVALID_HID;
hid_t fctmpl = H5I_INVALID_HID;
hid_t fapl_id = H5I_INVALID_HID;
hid_t dxpl_id = H5I_INVALID_HID;
hid_t dset_id = H5I_INVALID_HID;
hid_t dset_id_ch = H5I_INVALID_HID;
hid_t dcpl_id = H5P_DEFAULT;
hsize_t chunk[1];
float nextValue;
float *data_slice = NULL;
pass = true;
assert(comm != MPI_COMM_NULL);
if ((MPI_Comm_rank(comm, &group_rank)) != MPI_SUCCESS) {
pass = false;
failure_mssg = "generate_test_file: MPI_Comm_rank failed.\n";
}
if ((MPI_Comm_size(comm, &group_size)) != MPI_SUCCESS) {
pass = false;
failure_mssg = "generate_test_file: MPI_Comm_size failed.\n";
}
if (mpi_rank == 0) {
fprintf(stdout, "Constructing test files...");
}
/* Setup the file names
* The test specific filenames are stored as consecutive
* array entries in the global 'FILENAMES' array above.
* Here, we simply decide on the starting index for
* file construction. The reading portion of the test
* will have a similar setup process...
*/
if (pass) {
if (comm == MPI_COMM_WORLD) { /* Test 1 */
file_index = 0;
}
else if (group_id == 0) { /* Test 2 group 0 */
file_index = 1;
}
else { /* Test 2 group 1 */
file_index = 2;
}
/* The 'group_filename' is just a temp variable and
* is used to call into the h5_fixname function. No
* need to worry that we reassign it for each file!
*/
group_filename = FILENAMES[file_index];
assert(group_filename);
/* Assign the 'data_filename' */
if (h5_fixname(group_filename, H5P_DEFAULT, data_filename, sizeof(data_filename)) == NULL) {
pass = false;
failure_mssg = "h5_fixname(0) failed.\n";
}
}
/* setup data to write */
if (pass) {
if ((data_slice = (float *)malloc(COUNT * sizeof(float))) == NULL) {
pass = false;
failure_mssg = "malloc of data_slice failed.\n";
}
}
if (pass) {
nextValue = (float)(mpi_rank * COUNT);
for (i = 0; i < COUNT; i++) {
data_slice[i] = nextValue;
nextValue += 1;
}
}
/* Initialize a file creation template */
if (pass) {
if ((fctmpl = H5Pcreate(H5P_FILE_CREATE)) < 0) {
pass = false;
failure_mssg = "H5Pcreate(H5P_FILE_CREATE) failed.\n";
}
else if (H5Pset_userblock(fctmpl, 512) != SUCCEED) {
pass = false;
failure_mssg = "H5Pset_userblock(,size) failed.\n";
}
}
/* setup FAPL */
if (pass) {
if ((fapl_id = H5Pcreate(H5P_FILE_ACCESS)) < 0) {
pass = false;
failure_mssg = "H5Pcreate(H5P_FILE_ACCESS) failed.\n";
}
}
if (pass) {
if ((H5Pset_fapl_mpio(fapl_id, comm, MPI_INFO_NULL)) < 0) {
pass = false;
failure_mssg = "H5Pset_fapl_mpio() failed\n";
}
}
/* create the data file */
if (pass) {
if ((file_id = H5Fcreate(data_filename, H5F_ACC_TRUNC, fctmpl, fapl_id)) < 0) {
pass = false;
failure_mssg = "H5Fcreate() failed.\n";
}
}
/* create and write the dataset */
if (pass) {
if ((dxpl_id = H5Pcreate(H5P_DATASET_XFER)) < 0) {
pass = false;
failure_mssg = "H5Pcreate(H5P_DATASET_XFER) failed.\n";
}
}
if (pass) {
if ((H5Pset_dxpl_mpio(dxpl_id, H5FD_MPIO_COLLECTIVE)) < 0) {
pass = false;
failure_mssg = "H5Pset_dxpl_mpio() failed.\n";
}
}
if (pass) {
dims[0] = COUNT;
if ((memspace = H5Screate_simple(1, dims, NULL)) < 0) {
pass = false;
failure_mssg = "H5Screate_simple(1, dims, NULL) failed (1).\n";
}
}
if (pass) {
dims[0] *= (hsize_t)group_size;
if ((filespace = H5Screate_simple(1, dims, NULL)) < 0) {
pass = false;
failure_mssg = "H5Screate_simple(1, dims, NULL) failed (2).\n";
}
}
if (pass) {
offset = (hsize_t)group_rank * (hsize_t)COUNT;
if ((H5Sselect_hyperslab(filespace, H5S_SELECT_SET, &offset, NULL, &count, NULL)) < 0) {
pass = false;
failure_mssg = "H5Sselect_hyperslab() failed.\n";
}
}
if (pass) {
if ((dset_id = H5Dcreate2(file_id, "dataset0", H5T_NATIVE_FLOAT, filespace, H5P_DEFAULT, H5P_DEFAULT,
H5P_DEFAULT)) < 0) {
pass = false;
failure_mssg = "H5Dcreate2() failed.\n";
}
}
if (pass) {
if ((H5Dwrite(dset_id, H5T_NATIVE_FLOAT, memspace, filespace, dxpl_id, data_slice)) < 0) {
pass = false;
failure_mssg = "H5Dwrite() failed.\n";
}
}
/* create a chunked dataset */
chunk[0] = COUNT / 8;
if (pass) {
if ((dcpl_id = H5Pcreate(H5P_DATASET_CREATE)) < 0) {
pass = false;
failure_mssg = "H5Pcreate() failed.\n";
}
}
if (pass) {
if ((H5Pset_chunk(dcpl_id, 1, chunk)) < 0) {
pass = false;
failure_mssg = "H5Pset_chunk() failed.\n";
}
}
if (pass) {
if ((dset_id_ch = H5Dcreate2(file_id, "dataset0_chunked", H5T_NATIVE_FLOAT, filespace, H5P_DEFAULT,
dcpl_id, H5P_DEFAULT)) < 0) {
pass = false;
failure_mssg = "H5Dcreate2() failed.\n";
}
}
if (pass) {
if ((H5Dwrite(dset_id_ch, H5T_NATIVE_FLOAT, memspace, filespace, dxpl_id, data_slice)) < 0) {
pass = false;
failure_mssg = "H5Dwrite() failed.\n";
}
}
if (pass || (dcpl_id != -1)) {
if (H5Pclose(dcpl_id) < 0) {
pass = false;
failure_mssg = "H5Pclose(dcpl_id) failed.\n";
}
}
if (pass || (dset_id_ch != -1)) {
if (H5Dclose(dset_id_ch) < 0) {
pass = false;
failure_mssg = "H5Dclose(dset_id_ch) failed.\n";
}
}
/* close file, etc. */
if (pass || (dset_id != -1)) {
if (H5Dclose(dset_id) < 0) {
pass = false;
failure_mssg = "H5Dclose(dset_id) failed.\n";
}
}
if (pass || (memspace != -1)) {
if (H5Sclose(memspace) < 0) {
pass = false;
failure_mssg = "H5Sclose(memspace) failed.\n";
}
}
if (pass || (filespace != -1)) {
if (H5Sclose(filespace) < 0) {
pass = false;
failure_mssg = "H5Sclose(filespace) failed.\n";
}
}
if (pass || (file_id != -1)) {
if (H5Fclose(file_id) < 0) {
pass = false;
failure_mssg = "H5Fclose(file_id) failed.\n";
}
}
if (pass || (dxpl_id != -1)) {
if (H5Pclose(dxpl_id) < 0) {
pass = false;
failure_mssg = "H5Pclose(dxpl_id) failed.\n";
}
}
if (pass || (fapl_id != -1)) {
if (H5Pclose(fapl_id) < 0) {
pass = false;
failure_mssg = "H5Pclose(fapl_id) failed.\n";
}
}
if (pass || (fctmpl != -1)) {
if (H5Pclose(fctmpl) < 0) {
pass = false;
failure_mssg = "H5Pclose(fctmpl) failed.\n";
}
}
/* Add a userblock to the head of the datafile.
* We will use this to for a functional test of the
* file open optimization. This is superblock
* relocation is done by the rank 0 process associated
* with the communicator being used. For test 1, we
* utilize MPI_COMM_WORLD, so group_rank 0 is the
* same as mpi_rank 0. For test 2 which utilizes
* two groups resulting from an MPI_Comm_split, we
* will have parallel groups and hence two
* group_rank(0) processes. Each parallel group
* will create a unique file with different text
* headers and different data.
*/
if (group_rank == 0) {
const char *text_to_write;
size_t bytes_to_write;
if (group_id == 0)
text_to_write = random_hdf5_text;
else
text_to_write = hitchhiker_quote;
bytes_to_write = HDstrlen(text_to_write);
if (pass) {
if ((header = HDopen(data_filename, O_WRONLY)) < 0) {
pass = false;
failure_mssg = "HDopen(data_filename, O_WRONLY) failed.\n";
}
}
if (pass) {
HDlseek(header, 0, SEEK_SET);
if (HDwrite(header, text_to_write, bytes_to_write) < 0) {
pass = false;
failure_mssg = "Unable to write user text into file.\n";
}
}
if (pass || (header > 0)) {
if (HDclose(header) < 0) {
pass = false;
failure_mssg = "HDclose() failed.\n";
}
}
}
/* collect results from other processes.
* Only overwrite the failure message if no previous error
* has been detected
*/
local_failure = (pass ? 0 : 1);
/* This is a global all reduce (NOT group specific) */
if (MPI_Allreduce(&local_failure, &global_failures, 1, MPI_INT, MPI_SUM, MPI_COMM_WORLD) != MPI_SUCCESS) {
if (pass) {
pass = false;
failure_mssg = "MPI_Allreduce() failed.\n";
}
}
else if ((pass) && (global_failures > 0)) {
pass = false;
failure_mssg = "One or more processes report failure.\n";
}
/* report results */
if (mpi_rank == 0) {
if (pass) {
fprintf(stdout, "Done.\n");
}
else {
fprintf(stdout, "FAILED.\n");
fprintf(stdout, "%s: failure_mssg = \"%s\"\n", fcn_name, failure_mssg);
}
}
/* free data_slice if it has been allocated */
if (data_slice != NULL) {
free(data_slice);
data_slice = NULL;
}
return (!pass);
} /* generate_test_file() */
/*-------------------------------------------------------------------------
* Function: test_parallel_read
*
* Purpose: This actually tests the superblock optimization
* and covers the three primary cases we're interested in.
* 1). That HDF5 files can be opened in parallel by
* the rank 0 process and that the superblock
* offset is correctly broadcast to the other
* parallel file readers.
* 2). That a parallel application can correctly
* handle reading multiple files by using
* subgroups of MPI_COMM_WORLD and that each
* subgroup operates as described in (1) to
* collectively read the data.
* 3). Testing proc0-read-and-MPI_Bcast using
* sub-communicators, and reading into
* a memory space that is different from the
* file space, and chunked datasets.
*
* The global MPI rank is used for reading and
* writing data for process specific data in the
* dataset. We do this rather simplisticly, i.e.
* rank 0: writes/reads 0-9999
* rank 1: writes/reads 1000-1999
* rank 2: writes/reads 2000-2999
* ...
*
* Return: Success: 0
*
* Failure: 1
*-------------------------------------------------------------------------
*/
static int
test_parallel_read(MPI_Comm comm, int mpi_rank, int mpi_size, int group_id)
{
const char *failure_mssg;
const char *fcn_name = "test_parallel_read()";
const char *group_filename = NULL;
char reloc_data_filename[FILENAME_BUF_SIZE];
int local_failure = 0;
int global_failures = 0;
int group_size;
int group_rank;
hid_t fapl_id = H5I_INVALID_HID;
hid_t file_id = H5I_INVALID_HID;
hid_t dset_id = H5I_INVALID_HID;
hid_t dset_id_ch = H5I_INVALID_HID;
hid_t dxpl_id = H5P_DEFAULT;
hid_t memspace = H5I_INVALID_HID;
hid_t filespace = H5I_INVALID_HID;
hid_t filetype = H5I_INVALID_HID;
size_t filetype_size;
hssize_t dset_size;
hsize_t i;
hsize_t offset;
hsize_t count = COUNT;
hsize_t dims[1] = {0};
float nextValue;
float *data_slice = NULL;
pass = true;
assert(comm != MPI_COMM_NULL);
if ((MPI_Comm_rank(comm, &group_rank)) != MPI_SUCCESS) {
pass = false;
failure_mssg = "test_parallel_read: MPI_Comm_rank failed.\n";
}
if ((MPI_Comm_size(comm, &group_size)) != MPI_SUCCESS) {
pass = false;
failure_mssg = "test_parallel_read: MPI_Comm_size failed.\n";
}
if (mpi_rank == 0) {
if (comm == MPI_COMM_WORLD) {
TESTING("parallel file open test 1");
}
else {
TESTING("parallel file open test 2");
}
}
/* allocate space for the data_slice array */
if (pass) {
if ((data_slice = (float *)malloc(COUNT * sizeof(float))) == NULL) {
pass = false;
failure_mssg = "malloc of data_slice failed.\n";
}
}
/* Select the file file name to read
* Please see the comments in the 'generate_test_file' function
* for more details...
*/
if (pass) {
if (comm == MPI_COMM_WORLD) /* test 1 */
group_filename = FILENAMES[0];
else if (group_id == 0) /* test 2 group 0 */
group_filename = FILENAMES[1];
else /* test 2 group 1 */
group_filename = FILENAMES[2];
assert(group_filename);
if (h5_fixname(group_filename, H5P_DEFAULT, reloc_data_filename, sizeof(reloc_data_filename)) ==
NULL) {
pass = false;
failure_mssg = "h5_fixname(1) failed.\n";
}
}
/* setup FAPL */
if (pass) {
if ((fapl_id = H5Pcreate(H5P_FILE_ACCESS)) < 0) {
pass = false;
failure_mssg = "H5Pcreate(H5P_FILE_ACCESS) failed.\n";
}
}
if (pass) {
if ((H5Pset_fapl_mpio(fapl_id, comm, MPI_INFO_NULL)) < 0) {
pass = false;
failure_mssg = "H5Pset_fapl_mpio() failed\n";
}
}
/* open the file -- should have user block, exercising the optimization */
if (pass) {
if ((file_id = H5Fopen(reloc_data_filename, H5F_ACC_RDONLY, fapl_id)) < 0) {
pass = false;
failure_mssg = "H5Fopen() failed\n";
}
}
/* open the data set */
if (pass) {
if ((dset_id = H5Dopen2(file_id, "dataset0", H5P_DEFAULT)) < 0) {
pass = false;
failure_mssg = "H5Dopen2() failed\n";
}
}
/* open the chunked data set */
if (pass) {
if ((dset_id_ch = H5Dopen2(file_id, "dataset0_chunked", H5P_DEFAULT)) < 0) {
pass = false;
failure_mssg = "H5Dopen2() failed\n";
}
}
/* setup memspace */
if (pass) {
dims[0] = count;
if ((memspace = H5Screate_simple(1, dims, NULL)) < 0) {
pass = false;
failure_mssg = "H5Screate_simple(1, dims, NULL) failed\n";
}
}
/* setup filespace */
if (pass) {
if ((filespace = H5Dget_space(dset_id)) < 0) {
pass = false;
failure_mssg = "H5Dget_space(dataset) failed\n";
}
}
if (pass) {
offset = (hsize_t)group_rank * count;
if ((H5Sselect_hyperslab(filespace, H5S_SELECT_SET, &offset, NULL, &count, NULL)) < 0) {
pass = false;
failure_mssg = "H5Sselect_hyperslab() failed\n";
}
}
/* read this processes section of the data */
if (pass) {
if ((H5Dread(dset_id, H5T_NATIVE_FLOAT, memspace, filespace, H5P_DEFAULT, data_slice)) < 0) {
pass = false;
failure_mssg = "H5Dread() failed\n";
}
}
/* verify the data */
if (pass) {
nextValue = (float)((hsize_t)mpi_rank * count);
i = 0;
while ((pass) && (i < count)) {
/* what we really want is data_slice[i] != nextValue --
* the following is a circumlocution to shut up the
* the compiler.
*/
if ((data_slice[i] > nextValue) || (data_slice[i] < nextValue)) {
pass = false;
failure_mssg = "Unexpected dset contents.\n";
}
nextValue += 1;
i++;
}
}
if (pass || (memspace != -1)) {
if (H5Sclose(memspace) < 0) {
pass = false;
failure_mssg = "H5Sclose(memspace) failed.\n";
}
}
if (pass || (filespace != -1)) {
if (H5Sclose(filespace) < 0) {
pass = false;
failure_mssg = "H5Sclose(filespace) failed.\n";
}
}
/* free data_slice if it has been allocated */
if (data_slice != NULL) {
free(data_slice);
data_slice = NULL;
}
/*
* Test reading proc0-read-and-bcast with sub-communicators
*/
/* Don't test with more than LIMIT_NPROC processes to avoid memory issues */
if (group_size <= LIMIT_NPROC) {
#ifdef H5_HAVE_INSTRUMENTED_LIBRARY
bool prop_value;
#endif /* H5_HAVE_INSTRUMENTED_LIBRARY */
if ((filespace = H5Dget_space(dset_id)) < 0) {
pass = false;
failure_mssg = "H5Dget_space failed.\n";
}
if ((dset_size = H5Sget_simple_extent_npoints(filespace)) < 0) {
pass = false;
failure_mssg = "H5Sget_simple_extent_npoints failed.\n";
}
if ((filetype = H5Dget_type(dset_id)) < 0) {
pass = false;
failure_mssg = "H5Dget_type failed.\n";
}
if ((filetype_size = H5Tget_size(filetype)) == 0) {
pass = false;
failure_mssg = "H5Tget_size failed.\n";
}
if (H5Tclose(filetype) < 0) {
pass = false;
failure_mssg = "H5Tclose failed.\n";
};
if ((data_slice = (float *)malloc((size_t)dset_size * filetype_size)) == NULL) {
pass = false;
failure_mssg = "malloc of data_slice failed.\n";
}
if (pass) {
if ((dxpl_id = H5Pcreate(H5P_DATASET_XFER)) < 0) {
pass = false;
failure_mssg = "H5Pcreate(H5P_DATASET_XFER) failed.\n";
}
}
if (pass) {
if ((H5Pset_dxpl_mpio(dxpl_id, H5FD_MPIO_COLLECTIVE)) < 0) {
pass = false;
failure_mssg = "H5Pset_dxpl_mpio() failed.\n";
}
}
#ifdef H5_HAVE_INSTRUMENTED_LIBRARY
if (pass) {
prop_value = H5D_XFER_COLL_RANK0_BCAST_DEF;
if (H5Pinsert2(dxpl_id, H5D_XFER_COLL_RANK0_BCAST_NAME, H5D_XFER_COLL_RANK0_BCAST_SIZE,
&prop_value, NULL, NULL, NULL, NULL, NULL, NULL) < 0) {
pass = false;
failure_mssg = "H5Pinsert2() failed\n";
}
}
#endif /* H5_HAVE_INSTRUMENTED_LIBRARY */
/* read H5S_ALL section */
if (pass) {
if ((H5Dread(dset_id, H5T_NATIVE_FLOAT, H5S_ALL, H5S_ALL, dxpl_id, data_slice)) < 0) {
pass = false;
failure_mssg = "H5Dread() failed\n";
}
}
#ifdef H5_HAVE_INSTRUMENTED_LIBRARY
if (pass) {
prop_value = false;
if (H5Pget(dxpl_id, H5D_XFER_COLL_RANK0_BCAST_NAME, &prop_value) < 0) {
pass = false;
failure_mssg = "H5Pget() failed\n";
}
if (pass) {
if (prop_value != true) {
pass = false;
failure_mssg = "rank 0 Bcast optimization was mistakenly not performed\n";
}
}
}
#endif /* H5_HAVE_INSTRUMENTED_LIBRARY */
/* verify the data */
if (pass) {
if (comm == MPI_COMM_WORLD) /* test 1 */
nextValue = 0;
else if (group_id == 0) /* test 2 group 0 */
nextValue = 0;
else /* test 2 group 1 */
nextValue = (float)((hsize_t)(mpi_size / 2) * count);
i = 0;
while ((pass) && (i < (hsize_t)dset_size)) {
/* what we really want is data_slice[i] != nextValue --
* the following is a circumlocution to shut up the
* the compiler.
*/
if ((data_slice[i] > nextValue) || (data_slice[i] < nextValue)) {
pass = false;
failure_mssg = "Unexpected dset contents.\n";
}
nextValue += 1;
i++;
}
}
/* read H5S_ALL section for the chunked dataset */
#ifdef H5_HAVE_INSTRUMENTED_LIBRARY
if (pass) {
prop_value = H5D_XFER_COLL_RANK0_BCAST_DEF;
if (H5Pset(dxpl_id, H5D_XFER_COLL_RANK0_BCAST_NAME, &prop_value) < 0) {
pass = false;
failure_mssg = "H5Pset() failed\n";
}
}
#endif /* H5_HAVE_INSTRUMENTED_LIBRARY */
for (i = 0; i < (hsize_t)dset_size; i++) {
data_slice[i] = 0;
}
if (pass) {
if ((H5Dread(dset_id_ch, H5T_NATIVE_FLOAT, H5S_ALL, H5S_ALL, dxpl_id, data_slice)) < 0) {
pass = false;
failure_mssg = "H5Dread() failed\n";
}
}
#ifdef H5_HAVE_INSTRUMENTED_LIBRARY
if (pass) {
prop_value = false;
if (H5Pget(dxpl_id, H5D_XFER_COLL_RANK0_BCAST_NAME, &prop_value) < 0) {
pass = false;
failure_mssg = "H5Pget() failed\n";
}
if (pass) {
if (prop_value == true) {
pass = false;
failure_mssg = "rank 0 Bcast optimization was mistakenly performed for chunked dataset\n";
}
}
}
#endif /* H5_HAVE_INSTRUMENTED_LIBRARY */
/* verify the data */
if (pass) {
if (comm == MPI_COMM_WORLD) /* test 1 */
nextValue = 0;
else if (group_id == 0) /* test 2 group 0 */
nextValue = 0;
else /* test 2 group 1 */
nextValue = (float)((hsize_t)(mpi_size / 2) * count);
i = 0;
while ((pass) && (i < (hsize_t)dset_size)) {
/* what we really want is data_slice[i] != nextValue --
* the following is a circumlocution to shut up the
* the compiler.
*/
if ((data_slice[i] > nextValue) || (data_slice[i] < nextValue)) {
pass = false;
failure_mssg = "Unexpected chunked dset contents.\n";
}
nextValue += 1;
i++;
}
}
if (pass || (filespace != -1)) {
if (H5Sclose(filespace) < 0) {
pass = false;
failure_mssg = "H5Sclose(filespace) failed.\n";
}
}
/* free data_slice if it has been allocated */
if (data_slice != NULL) {
free(data_slice);
data_slice = NULL;
}
/*
* Read an H5S_ALL filespace into a hyperslab defined memory space
*/
if ((data_slice = (float *)malloc((size_t)(dset_size * 2) * filetype_size)) == NULL) {
pass = false;
failure_mssg = "malloc of data_slice failed.\n";
}
/* setup memspace */
if (pass) {
dims[0] = (hsize_t)dset_size * 2;
if ((memspace = H5Screate_simple(1, dims, NULL)) < 0) {
pass = false;
failure_mssg = "H5Screate_simple(1, dims, NULL) failed\n";
}
}
if (pass) {
offset = (hsize_t)dset_size;
if ((H5Sselect_hyperslab(memspace, H5S_SELECT_SET, &offset, NULL, &offset, NULL)) < 0) {
pass = false;
failure_mssg = "H5Sselect_hyperslab() failed\n";
}
}
#ifdef H5_HAVE_INSTRUMENTED_LIBRARY
if (pass) {
prop_value = H5D_XFER_COLL_RANK0_BCAST_DEF;
if (H5Pset(dxpl_id, H5D_XFER_COLL_RANK0_BCAST_NAME, &prop_value) < 0) {
pass = false;
failure_mssg = "H5Pset() failed\n";
}
}
#endif /* H5_HAVE_INSTRUMENTED_LIBRARY */
/* read this processes section of the data */
if (pass) {
if ((H5Dread(dset_id, H5T_NATIVE_FLOAT, memspace, H5S_ALL, dxpl_id, data_slice)) < 0) {
pass = false;
failure_mssg = "H5Dread() failed\n";
}
}
#ifdef H5_HAVE_INSTRUMENTED_LIBRARY
if (pass) {
prop_value = false;
if (H5Pget(dxpl_id, H5D_XFER_COLL_RANK0_BCAST_NAME, &prop_value) < 0) {
pass = false;
failure_mssg = "H5Pget() failed\n";
}
if (pass) {
if (prop_value != true) {
pass = false;
failure_mssg = "rank 0 Bcast optimization was mistakenly not performed\n";
}
}
}
#endif /* H5_HAVE_INSTRUMENTED_LIBRARY */
/* verify the data */
if (pass) {
if (comm == MPI_COMM_WORLD) /* test 1 */
nextValue = 0;
else if (group_id == 0) /* test 2 group 0 */
nextValue = 0;
else /* test 2 group 1 */
nextValue = (float)((hsize_t)(mpi_size / 2) * count);
i = (hsize_t)dset_size;
while ((pass) && (i < (hsize_t)dset_size)) {
/* what we really want is data_slice[i] != nextValue --
* the following is a circumlocution to shut up the
* the compiler.
*/
if ((data_slice[i] > nextValue) || (data_slice[i] < nextValue)) {
pass = false;
failure_mssg = "Unexpected dset contents.\n";
}
nextValue += 1;
i++;
}
}
if (pass || (memspace != -1)) {
if (H5Sclose(memspace) < 0) {
pass = false;
failure_mssg = "H5Sclose(memspace) failed.\n";
}
}
/* free data_slice if it has been allocated */
if (data_slice != NULL) {
free(data_slice);
data_slice = NULL;
}
if (pass || (dxpl_id != -1)) {
if (H5Pclose(dxpl_id) < 0) {
pass = false;
failure_mssg = "H5Pclose(dxpl_id) failed.\n";
}
}
}
/* close file, etc. */
if (pass || (dset_id != -1)) {
if (H5Dclose(dset_id) < 0) {
pass = false;
failure_mssg = "H5Dclose(dset_id) failed.\n";
}
}
if (pass || (dset_id_ch != -1)) {
if (H5Dclose(dset_id_ch) < 0) {
pass = false;
failure_mssg = "H5Dclose(dset_id_ch) failed.\n";
}
}
if (pass || (file_id != -1)) {
if (H5Fclose(file_id) < 0) {
pass = false;
failure_mssg = "H5Fclose(file_id) failed.\n";
}
}
if (pass || (fapl_id != -1)) {
if (H5Pclose(fapl_id) < 0) {
pass = false;
failure_mssg = "H5Pclose(fapl_id) failed.\n";
}
}
/* collect results from other processes.
* Only overwrite the failure message if no previous error
* has been detected
*/
local_failure = (pass ? 0 : 1);
if (MPI_Allreduce(&local_failure, &global_failures, 1, MPI_INT, MPI_SUM, MPI_COMM_WORLD) != MPI_SUCCESS) {
if (pass) {
pass = false;
failure_mssg = "MPI_Allreduce() failed.\n";
}
}
else if ((pass) && (global_failures > 0)) {
pass = false;
failure_mssg = "One or more processes report failure.\n";
}
/* report results and finish cleanup */
if (group_rank == 0) {
if (pass) {
PASSED();
}
else {
H5_FAILED();
fprintf(stdout, "%s: failure_mssg = \"%s\"\n", fcn_name, failure_mssg);
}
HDremove(reloc_data_filename);
}
return (!pass);
} /* test_parallel_read() */
/*-------------------------------------------------------------------------
* Function: main
*
* Purpose: To implement a parallel test which validates whether the
* new superblock lookup functionality is working correctly.
*
* The test consists of creating two separate HDF datasets
* in which random text is inserted at the start of each
* file using the 'j5jam' application. This forces the
* HDF5 file superblock to a non-zero offset.
* Having created the two independent files, we create two
* non-overlapping MPI groups, each of which is then tasked
* with the opening and validation of the data contained
* therein.
*
* Return: Success: 0
* Failure: 1
*-------------------------------------------------------------------------
*/
int
main(int argc, char **argv)
{
int nerrs = 0;
int which_group = 0;
int mpi_rank;
int mpi_size;
int split_size;
MPI_Comm group_comm = MPI_COMM_NULL;
/* I don't believe that argv[0] can ever be NULL.
* It should thus be safe to dup and save as a check
* for cmake testing. Note that in our Cmake builds,
* all executables are located in the same directory.
* We assume (but we'll check) that the h5jam utility
* is in the directory as this executable. If that
* isn't true, then we can use a relative path that
* should be valid for the autotools environment.
*/
test_argv0 = HDstrdup(argv[0]);
if ((MPI_Init(&argc, &argv)) != MPI_SUCCESS) {
fprintf(stderr, "FATAL: Unable to initialize MPI\n");
exit(EXIT_FAILURE);
}
if ((MPI_Comm_rank(MPI_COMM_WORLD, &mpi_rank)) != MPI_SUCCESS) {
fprintf(stderr, "FATAL: MPI_Comm_rank returned an error\n");
exit(EXIT_FAILURE);
}
if ((MPI_Comm_size(MPI_COMM_WORLD, &mpi_size)) != MPI_SUCCESS) {
fprintf(stderr, "FATAL: MPI_Comm_size returned an error\n");
exit(EXIT_FAILURE);
}
H5open();
if (mpi_rank == 0) {
fprintf(stdout, "========================================\n");
fprintf(stdout, "Collective file open optimization tests\n");
fprintf(stdout, " mpi_size = %d\n", mpi_size);
fprintf(stdout, "========================================\n");
}
if (mpi_size < 3) {
if (mpi_rank == 0) {
printf(" Need at least 3 processes. Exiting.\n");
}
goto finish;
}
/* ------ Create two (2) MPI groups ------
*
* We split MPI_COMM_WORLD into 2 more or less equal sized
* groups. The resulting communicators will be used to generate
* two HDF files which in turn will be opened in parallel and the
* contents verified in the second read test below.
*/
split_size = mpi_size / 2;
which_group = (mpi_rank < split_size ? 0 : 1);
if ((MPI_Comm_split(MPI_COMM_WORLD, which_group, 0, &group_comm)) != MPI_SUCCESS) {
fprintf(stderr, "FATAL: MPI_Comm_split returned an error\n");
exit(EXIT_FAILURE);
}
/* ------ Generate all files ------ */
/* We generate the file used for test 1 */
nerrs += generate_test_file(MPI_COMM_WORLD, mpi_rank, which_group);
if (nerrs > 0) {
if (mpi_rank == 0) {
printf(" Test(1) file construction failed -- skipping tests.\n");
}
goto finish;
}
/* We generate the file used for test 2 */
nerrs += generate_test_file(group_comm, mpi_rank, which_group);
if (nerrs > 0) {
if (mpi_rank == 0) {
printf(" Test(2) file construction failed -- skipping tests.\n");
}
goto finish;
}
/* Now read the generated test file (still using MPI_COMM_WORLD) */
nerrs += test_parallel_read(MPI_COMM_WORLD, mpi_rank, mpi_size, which_group);
if (nerrs > 0) {
if (mpi_rank == 0) {
printf(" Parallel read test(1) failed -- skipping tests.\n");
}
goto finish;
}
/* Update the user on our progress so far. */
if (mpi_rank == 0) {
printf(" Test 1 of 2 succeeded\n");
printf(" -- Starting multi-group parallel read test.\n");
}
/* run the 2nd set of tests */
nerrs += test_parallel_read(group_comm, mpi_rank, mpi_size, which_group);
if (nerrs > 0) {
if (mpi_rank == 0) {
printf(" Multi-group read test(2) failed\n");
}
goto finish;
}
if (mpi_rank == 0) {
printf(" Test 2 of 2 succeeded\n");
}
finish:
if ((group_comm != MPI_COMM_NULL) && (MPI_Comm_free(&group_comm)) != MPI_SUCCESS) {
fprintf(stderr, "MPI_Comm_free failed!\n");
}
/* make sure all processes are finished before final report, cleanup
* and exit.
*/
MPI_Barrier(MPI_COMM_WORLD);
if (mpi_rank == 0) { /* only process 0 reports */
const char *header = "Collective file open optimization tests";
fprintf(stdout, "===================================\n");
if (nerrs > 0) {
fprintf(stdout, "***%s detected %d failures***\n", header, nerrs);
}
else {
fprintf(stdout, "%s finished with no failures\n", header);
}
fprintf(stdout, "===================================\n");
}
/* close HDF5 library */
if (H5close() != SUCCEED) {
fprintf(stdout, "H5close() failed. (Ignoring)\n");
}
/* MPI_Finalize must be called AFTER H5close which may use MPI calls */
MPI_Finalize();
/* cannot just return (nerrs) because exit code is limited to 1byte */
return ((nerrs > 0) ? EXIT_FAILURE : EXIT_SUCCESS);
} /* main() */
|