mpi3-fortran / ompi / runtime / ompi_mpi_init.c

  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
/*
 * Copyright (c) 2004-2010 The Trustees of Indiana University and Indiana
 *                         University Research and Technology
 *                         Corporation.  All rights reserved.
 * Copyright (c) 2004-2011 The University of Tennessee and The University
 *                         of Tennessee Research Foundation.  All rights
 *                         reserved.
 * Copyright (c) 2004-2005 High Performance Computing Center Stuttgart, 
 *                         University of Stuttgart.  All rights reserved.
 * Copyright (c) 2004-2005 The Regents of the University of California.
 *                         All rights reserved.
 * Copyright (c) 2006-2012 Cisco Systems, Inc.  All rights reserved.
 * Copyright (c) 2006-2012 Los Alamos National Security, LLC.  All rights
 *                         reserved. 
 * Copyright (c) 2006-2009 University of Houston. All rights reserved.
 * Copyright (c) 2008-2009 Sun Microsystems, Inc.  All rights reserved.
 * Copyright (c) 2011      Sandia National Laboratories. All rights reserved.
 *
 * $COPYRIGHT$
 * 
 * Additional copyrights may follow
 * 
 * $HEADER$
 */

#include "ompi_config.h"

#ifdef HAVE_SYS_TIME_H
#include <sys/time.h>
#endif  /* HAVE_SYS_TIME_H */
#ifdef HAVE_PTHREAD_H
#include <pthread.h>
#endif
#ifdef HAVE_UNISTD_H
#include <unistd.h>
#endif

#include "mpi.h"
#include "opal/class/opal_list.h"
#include "opal/mca/base/base.h"
#include "opal/mca/hwloc/base/base.h"
#include "opal/runtime/opal_progress.h"
#include "opal/threads/threads.h"
#include "opal/util/arch.h"
#include "opal/util/output.h"
#include "opal/util/error.h"
#include "opal/util/stacktrace.h"
#include "opal/util/show_help.h"
#include "opal/runtime/opal.h"
#include "opal/mca/event/event.h"

#include "ompi/constants.h"
#include "ompi/mpi/fortran/base/constants.h"
#include "ompi/runtime/mpiruntime.h"
#include "ompi/runtime/params.h"
#include "ompi/runtime/ompi_module_exchange.h"
#include "ompi/communicator/communicator.h"
#include "ompi/info/info.h"
#include "ompi/errhandler/errcode.h"
#include "ompi/errhandler/errhandler.h"
#include "ompi/request/request.h"
#include "ompi/message/message.h"
#include "ompi/op/op.h"
#include "ompi/mca/op/op.h"
#include "ompi/mca/op/base/base.h"
#include "ompi/file/file.h"
#include "ompi/attribute/attribute.h"
#include "ompi/mca/allocator/base/base.h"
#include "ompi/mca/rcache/base/base.h"
#include "ompi/mca/rcache/rcache.h"
#include "ompi/mca/mpool/base/base.h"
#include "ompi/mca/pml/pml.h"
#include "ompi/mca/pml/base/base.h"
#include "ompi/mca/osc/base/base.h"
#include "ompi/mca/coll/base/base.h"
#include "ompi/mca/io/io.h"
#include "ompi/mca/io/base/base.h"
#include "ompi/mca/rte/rte.h"
#include "ompi/debuggers/debuggers.h"
#include "ompi/proc/proc.h"
#include "ompi/mca/pml/base/pml_base_bsend.h"
#include "ompi/mca/dpm/base/base.h"
#include "ompi/mca/pubsub/base/base.h"
#include "ompi/mpiext/mpiext.h"

#if OPAL_ENABLE_FT_CR == 1
#include "ompi/mca/crcp/crcp.h"
#include "ompi/mca/crcp/base/base.h"
#endif
#include "ompi/runtime/ompi_cr.h"

/* This is required for the boundaries of the hash tables used to store
 * the F90 types returned by the MPI_Type_create_f90_XXX functions.
 */
#include <float.h>

#if OPAL_CC_USE_PRAGMA_IDENT
#pragma ident OMPI_IDENT_STRING
#elif OPAL_CC_USE_IDENT
#ident OMPI_IDENT_STRING
#endif
const char ompi_version_string[] = OMPI_IDENT_STRING;

/*
 * Global variables and symbols for the MPI layer
 */

bool ompi_mpi_init_started = false;
bool ompi_mpi_initialized = false;
bool ompi_mpi_finalized = false;

bool ompi_mpi_thread_multiple = false;
int ompi_mpi_thread_requested = MPI_THREAD_SINGLE;
int ompi_mpi_thread_provided = MPI_THREAD_SINGLE;

opal_thread_t *ompi_mpi_main_thread = NULL;

bool ompi_warn_on_fork;

/*
 * These variables are for the MPI F08 bindings (F08 must bind Fortran
 * varaiables to symbols; it cannot bind Fortran variables to the
 * address of a C variable).
 */

ompi_predefined_datatype_t *ompi_mpi_character_addr = &ompi_mpi_character;
ompi_predefined_datatype_t *ompi_mpi_logical_addr   = &ompi_mpi_logical;
ompi_predefined_datatype_t *ompi_mpi_logical1_addr  = &ompi_mpi_logical1;
ompi_predefined_datatype_t *ompi_mpi_logical2_addr  = &ompi_mpi_logical2;
ompi_predefined_datatype_t *ompi_mpi_logical4_addr  = &ompi_mpi_logical4;
ompi_predefined_datatype_t *ompi_mpi_logical8_addr  = &ompi_mpi_logical8;
ompi_predefined_datatype_t *ompi_mpi_integer_addr   = &ompi_mpi_integer;
ompi_predefined_datatype_t *ompi_mpi_integer1_addr  = &ompi_mpi_integer1;
ompi_predefined_datatype_t *ompi_mpi_integer2_addr  = &ompi_mpi_integer2;
ompi_predefined_datatype_t *ompi_mpi_integer4_addr  = &ompi_mpi_integer4;
ompi_predefined_datatype_t *ompi_mpi_integer8_addr  = &ompi_mpi_integer8;
ompi_predefined_datatype_t *ompi_mpi_integer16_addr = &ompi_mpi_integer16;
ompi_predefined_datatype_t *ompi_mpi_real_addr      = &ompi_mpi_real;
ompi_predefined_datatype_t *ompi_mpi_real4_addr     = &ompi_mpi_real4;
ompi_predefined_datatype_t *ompi_mpi_real8_addr     = &ompi_mpi_real8;
ompi_predefined_datatype_t *ompi_mpi_real16_addr    = &ompi_mpi_real16;
ompi_predefined_datatype_t *ompi_mpi_dblprec_addr   = &ompi_mpi_dblprec;
ompi_predefined_datatype_t *ompi_mpi_cplex_addr     = &ompi_mpi_cplex;
ompi_predefined_datatype_t *ompi_mpi_complex8_addr  = &ompi_mpi_complex8;
ompi_predefined_datatype_t *ompi_mpi_complex16_addr = &ompi_mpi_complex16;
ompi_predefined_datatype_t *ompi_mpi_complex32_addr = &ompi_mpi_complex32;
ompi_predefined_datatype_t *ompi_mpi_dblcplex_addr  = &ompi_mpi_dblcplex;
ompi_predefined_datatype_t *ompi_mpi_2real_addr     = &ompi_mpi_2real;
ompi_predefined_datatype_t *ompi_mpi_2dblprec_addr  = &ompi_mpi_2dblprec;
ompi_predefined_datatype_t *ompi_mpi_2integer_addr  = &ompi_mpi_2integer;

struct ompi_status_public_t *ompi_mpi_status_ignore_addr =
    (ompi_status_public_t *) 0;
struct ompi_status_public_t *ompi_mpi_statuses_ignore_addr =
    (ompi_status_public_t *) 0;

#if OPAL_HAVE_POSIX_THREADS
static bool fork_warning_issued = false;
static bool atfork_called = false;

static void warn_fork_cb(void)
{
    if (ompi_mpi_initialized && !ompi_mpi_finalized && !fork_warning_issued) {
        ompi_show_help("help-mpi-runtime.txt", "mpi_init:warn-fork", true,
                       ompi_process_info.nodename, getpid(),
                       ompi_mpi_comm_world.comm.c_my_rank);
        fork_warning_issued = true;
    }
}
#endif

void ompi_warn_fork(void)
{
#if OPAL_HAVE_POSIX_THREADS
    if (ompi_warn_on_fork && !atfork_called) {
        pthread_atfork(warn_fork_cb, NULL, NULL);
        atfork_called = true;
    }
#endif
}

/*
 * These variables are here, rather than under ompi/mpi/c/foo.c
 * because it is not sufficient to have a .c file that only contains
 * variables -- you must have a function that is invoked from
 * elsewhere in the code to guarantee that all linkers will pull in
 * the .o file from the library.  Hence, although these are MPI
 * constants, we might as well just define them here (i.e., in a file
 * that already has a function that is guaranteed to be linked in,
 * rather than make a new .c file with the constants and a
 * corresponding dummy function that is invoked from this function).
 *
 * Additionally, there can be/are strange linking paths such that
 * ompi_info needs symbols such as ompi_fortran_status_ignore,
 * which, if they weren't here with a collection of other global
 * symbols that are initialized (which seems to force this .o file to
 * be pulled into the resolution process, because ompi_info certainly
 * does not call ompi_mpi_init()), would not be able to be found by
 * the OSX linker.
 *
 * NOTE: See the big comment in ompi/mpi/fortran/base/constants.h
 * about why we have four symbols for each of the common blocks (e.g.,
 * the Fortran equivalent(s) of MPI_STATUS_IGNORE).  Here, we can only
 * have *one* value (not four).  So the only thing we can do is make
 * it equal to the fortran compiler convention that was selected at
 * configure time.  Note that this is also true for the value of
 * .TRUE. from the Fortran compiler, so even though Open MPI supports
 * all four Fortran symbol conventions, it can only support one
 * convention for the two C constants (MPI_FORTRAN_STATUS[ES]_IGNORE)
 * and only support one compiler for the value of .TRUE.  Ugh!!
 *
 * Note that the casts here are ok -- we're *only* comparing pointer
 * values (i.e., they'll never be de-referenced).  The global symbols
 * are actually of type (ompi_fortran_common_t) (for alignment
 * issues), but MPI says that MPI_F_STATUS[ES]_IGNORE must be of type
 * (MPI_Fint*).  Hence, we have to cast to make compilers not
 * complain.
 */
#if OMPI_BUILD_FORTRAN_BINDINGS
#  if OMPI_FORTRAN_CAPS
MPI_Fint *MPI_F_STATUS_IGNORE = (MPI_Fint*) &MPI_FORTRAN_STATUS_IGNORE;
MPI_Fint *MPI_F_STATUSES_IGNORE = (MPI_Fint*) &MPI_FORTRAN_STATUSES_IGNORE;
#  elif OMPI_FORTRAN_PLAIN
MPI_Fint *MPI_F_STATUS_IGNORE = (MPI_Fint*) &mpi_fortran_status_ignore;
MPI_Fint *MPI_F_STATUSES_IGNORE = (MPI_Fint*) &mpi_fortran_statuses_ignore;
#  elif OMPI_FORTRAN_SINGLE_UNDERSCORE
MPI_Fint *MPI_F_STATUS_IGNORE = (MPI_Fint*) &mpi_fortran_status_ignore_;
MPI_Fint *MPI_F_STATUSES_IGNORE = (MPI_Fint*) &mpi_fortran_statuses_ignore_;
#  elif OMPI_FORTRAN_DOUBLE_UNDERSCORE
MPI_Fint *MPI_F_STATUS_IGNORE = (MPI_Fint*) &mpi_fortran_status_ignore__;
MPI_Fint *MPI_F_STATUSES_IGNORE = (MPI_Fint*) &mpi_fortran_statuses_ignore__;
#  else
#    error Unrecognized Fortran name mangling scheme
#  endif
#else
MPI_Fint *MPI_F_STATUS_IGNORE = NULL;
MPI_Fint *MPI_F_STATUSES_IGNORE = NULL;
#endif  /* OMPI_BUILD_FORTRAN_BINDINGS */


/* Constants for the Fortran layer.  These values are referred to via
   common blocks in the Fortran equivalents.  See
   ompi/mpi/fortran/base/constants.h for a more detailed explanation.

   The values are *NOT* initialized.  We do not use the values of
   these constants; only their addresses (because they're always
   passed by reference by Fortran).  

   Initializing upon instantiation these can reveal size and/or
   alignment differences between Fortran and C (!) which can cause
   warnings or errors upon linking (e.g., making static libraries with
   the intel 9.0 compilers on 64 bit platforms shows alignment
   differences between libmpi.a and the user's application, resulting
   in a linker warning).  FWIW, if you initialize these variables in
   functions (i.e., not at the instantiation in the global scope), the
   linker somehow "figures it all out" (w.r.t. different alignments
   between fortan common blocks and the corresponding C variables) and
   no linker warnings occur.

   Note that the rationale for the types of each of these variables is
   discussed in ompi/include/mpif-common.h.  Do not change the types
   without also modifying ompi/mpi/fortran/base/constants.h and
   ompi/include/mpif-common.h.
 */

#define INST(type, upper_case, lower_case, single_u, double_u)   \
type lower_case; \
type upper_case; \
type single_u;  \
type double_u

INST(int, MPI_FORTRAN_BOTTOM, mpi_fortran_bottom,
     mpi_fortran_bottom_, mpi_fortran_bottom__);
INST(int, MPI_FORTRAN_IN_PLACE, mpi_fortran_in_place,
     mpi_fortran_in_place_, mpi_fortran_in_place__);
INST(char *, MPI_FORTRAN_ARGV_NULL, mpi_fortran_argv_null,
     mpi_fortran_argv_null_, mpi_fortran_argv_null__);
INST(char *, MPI_FORTRAN_ARGVS_NULL, mpi_fortran_argvs_null,
     mpi_fortran_argvs_null_, mpi_fortran_argvs_null__);
INST(int *, MPI_FORTRAN_ERRCODES_IGNORE, mpi_fortran_errcodes_ignore,
     mpi_fortran_errcodes_ignore_, mpi_fortran_errcodes_ignore__);
INST(int *, MPI_FORTRAN_STATUS_IGNORE, mpi_fortran_status_ignore,
     mpi_fortran_status_ignore_, mpi_fortran_status_ignore__);
INST(int *, MPI_FORTRAN_STATUSES_IGNORE, mpi_fortran_statuses_ignore,
      mpi_fortran_statuses_ignore_, mpi_fortran_statuses_ignore__);

/*
 * Hash tables for MPI_Type_create_f90* functions
 */
opal_hash_table_t ompi_mpi_f90_integer_hashtable;
opal_hash_table_t ompi_mpi_f90_real_hashtable;
opal_hash_table_t ompi_mpi_f90_complex_hashtable;

/*
 * Per MPI-2:9.5.3, MPI_REGISTER_DATAREP is a memory leak.  There is
 * no way to *de*register datareps once they've been registered.  So
 * we have to track all registrations here so that they can be
 * de-registered during MPI_FINALIZE so that memory-tracking debuggers
 * don't show Open MPI as leaking memory.
 */
opal_list_t ompi_registered_datareps;


int ompi_mpi_init(int argc, char **argv, int requested, int *provided)
{
    int ret;
    ompi_proc_t** procs;
    size_t nprocs;
    char *error = NULL;
    bool timing = false;
    int param, value;
    struct timeval ompistart, ompistop;
    char *event_val = NULL;
    bool rte_setup = false;
    ompi_rte_collective_t *coll;
    char *cmd=NULL, *av=NULL;

    /* bitflag of the thread level support provided. To be used
     * for the modex in order to work in heterogeneous environments. */
    uint8_t threadlevel_bf; 

    /* Indicate that we have *started* MPI_INIT*.  MPI_FINALIZE has
       something sorta similar in a static local variable in
       ompi_mpi_finalize(). */
    ompi_mpi_init_started = true;

    /* Setup enough to check get/set MCA params */

    if (OPAL_SUCCESS != (ret = opal_init_util(&argc, &argv))) {
        error = "ompi_mpi_init: opal_init_util failed";
        goto error;
    }

    if (OPAL_SUCCESS != (ret = opal_arch_set_fortran_logical_size(sizeof(ompi_fortran_logical_t)))) {
        error = "ompi_mpi_init: opal_arch_set_fortran_logical_size failed";
        goto error;
    }

    /* _After_ opal_init_util() but _before_ orte_init(), we need to
       set an MCA param that tells libevent that it's ok to use any
       mechanism in libevent that is available on this platform (e.g.,
       epoll and friends).  Per opal/event/event.s, we default to
       select/poll -- but we know that MPI processes won't be using
       pty's with the event engine, so it's ok to relax this
       constraint and let any fd-monitoring mechanism be used. */
    ret = mca_base_param_reg_string_name("opal", "event_include",
                                         "Internal opal MCA param: tell opal_init() to use a specific mechanism in libevent",
                                         false, false, "all", &event_val);
    if (ret >= 0) {
        /* We have to explicitly "set" the MCA param value here
           because libevent initialization will re-register the MCA
           param and therefore override the default. Setting the value
           here puts the desired value ("all") in different storage
           that is not overwritten if/when the MCA param is
           re-registered. This is unless the user has specified a different
           value for this MCA parameter. Make sure we check to see if the
           default is specified before forcing "all" in case that is not what
           the user desires. Note that we do *NOT* set this value as an
           environment variable, just so that it won't be inherited by
           any spawned processes and potentially cause unintented
           side-effects with launching RTE tools... */
        if (0 == strcmp("all", event_val)) {
            mca_base_param_set_string(ret, "all");
        }
    }

    if( NULL != event_val ) {
        free(event_val);
        event_val = NULL;
    }

    /* check to see if we want timing information */
    param = mca_base_param_reg_int_name("ompi", "timing",
                                        "Request that critical timing loops be measured",
                                        false, false, 0, &value);
    if (value != 0) {
        timing = true;
        gettimeofday(&ompistart, NULL);
    }
    
    /* if we were not externally started, then we need to setup
     * some envars so the MPI_INFO_ENV can get the cmd name
     * and argv (but only if the user supplied a non-NULL argv!), and
     * the requested thread level
     */
    if (NULL == getenv("OMPI_COMMAND") && NULL != argv && NULL != argv[0]) {
        asprintf(&cmd, "OMPI_COMMAND=%s", argv[0]);
        putenv(cmd);
    }
    if (NULL == getenv("OMPI_ARGV") && 1 < argc) {
        char *tmp;
        tmp = opal_argv_join(&argv[1], ' ');
        asprintf(&av, "OMPI_ARGV=%s", tmp);
        free(tmp);
        putenv(av);
    }

    /* Setup RTE - note that we are an MPI process  */
    if (OMPI_SUCCESS != (ret = ompi_rte_init(NULL, NULL))) {
        error = "ompi_mpi_init: ompi_rte_init failed";
        goto error;
    }
    rte_setup = true;
    
    /* check for timing request - get stop time and report elapsed time if so */
    if (timing && 0 == OMPI_PROC_MY_NAME->vpid) {
        gettimeofday(&ompistop, NULL);
        opal_output(0, "ompi_mpi_init [%ld]: time from start to completion of rte_init %ld usec",
                    (long)OMPI_PROC_MY_NAME->vpid,
                    (long int)((ompistop.tv_sec - ompistart.tv_sec)*1000000 +
                               (ompistop.tv_usec - ompistart.tv_usec)));
        gettimeofday(&ompistart, NULL);
    }

#if OPAL_HAVE_HWLOC
    /* if hwloc is available but didn't get setup for some
     * reason, do so now
     */
    if (NULL == opal_hwloc_topology) {
        if (OPAL_SUCCESS != (ret = opal_hwloc_base_get_topology())) {
            error = "Topology init";
            goto error;
        }
    }
#endif

    /* Register errhandler callback - RTE will ignore if it
     * doesn't support this capability
     */
    ompi_rte_set_fault_callback(ompi_errhandler_runtime_callback);

    /* Figure out the final MPI thread levels.  If we were not
       compiled for support for MPI threads, then don't allow
       MPI_THREAD_MULTIPLE.  Set this stuff up here early in the
       process so that other components can make decisions based on
       this value. */
    /**
     * These values are monotonic; MPI_THREAD_SINGLE < MPI_THREAD_FUNNELED
     *                             < MPI_THREAD_SERIALIZED < MPI_THREAD_MULTIPLE.
     * If possible, the call will return provided = required. Failing this,
     * the call will return the least supported level such that
     * provided > required. Finally, if the user requirement cannot be
     * satisfied, then the call will return in provided the highest
     * supported level.
     */
    ompi_mpi_thread_requested = requested;

    if (OMPI_ENABLE_THREAD_MULTIPLE == 1) {
        ompi_mpi_thread_provided = *provided = requested;
        ompi_mpi_main_thread = opal_thread_get_self();
    } else {
        if (MPI_THREAD_MULTIPLE == requested) {
            ompi_mpi_thread_provided = *provided = MPI_THREAD_SERIALIZED;
        } else {
            ompi_mpi_thread_provided = *provided = requested;
        }
        ompi_mpi_main_thread = (OPAL_ENABLE_MULTI_THREADS ? opal_thread_get_self() : NULL);
    }

    ompi_mpi_thread_multiple = (ompi_mpi_thread_provided == 
                                MPI_THREAD_MULTIPLE);

    /* determine the bitflag belonging to the threadlevel_support provided */
    memset ( &threadlevel_bf, 0, sizeof(uint8_t));
    OMPI_THREADLEVEL_SET_BITFLAG ( ompi_mpi_thread_provided, threadlevel_bf );

    /* add this bitflag to the modex */
    if ( OMPI_SUCCESS != (ret = ompi_modex_send_string("MPI_THREAD_LEVEL", &threadlevel_bf, sizeof(uint8_t)))) {
        error = "ompi_mpi_init: modex send thread level";
        goto error;
    }

    /* Once we've joined the RTE, see if any MCA parameters were
       passed to the MPI level */

    if (OMPI_SUCCESS != (ret = ompi_mpi_register_params())) {
        error = "mca_mpi_register_params() failed";
        goto error;
    }

    /* initialize datatypes. This step should be done early as it will
     * create the local convertor and local arch used in the proc
     * init.
     */
    if (OMPI_SUCCESS != (ret = ompi_datatype_init())) {
        error = "ompi_datatype_init() failed";
        goto error;
    }

    /* Initialize OMPI procs */
    if (OMPI_SUCCESS != (ret = ompi_proc_init())) {
        error = "mca_proc_init() failed";
        goto error;
    }

    /* Initialize the op framework. This has to be done *after*
       ddt_init, but befor mca_coll_base_open, since some collective
       modules (e.g., the hierarchical coll component) may need ops in
       their query function. */
    if (OMPI_SUCCESS != (ret = ompi_op_base_open())) {
        error = "ompi_op_base_open() failed";
        goto error;
    }
    if (OMPI_SUCCESS != 
        (ret = ompi_op_base_find_available(OMPI_ENABLE_PROGRESS_THREADS,
                                           OMPI_ENABLE_THREAD_MULTIPLE))) {
        error = "ompi_op_base_find_available() failed";
        goto error;
    }
    if (OMPI_SUCCESS != (ret = ompi_op_init())) {
        error = "ompi_op_init() failed";
        goto error;
    }

    /* Open up MPI-related MCA components */

    if (OMPI_SUCCESS != (ret = mca_allocator_base_open())) {
        error = "mca_allocator_base_open() failed";
        goto error;
    }
    if (OMPI_SUCCESS != (ret = mca_rcache_base_open())) {
        error = "mca_rcache_base_open() failed";
        goto error;
    }
    if (OMPI_SUCCESS != (ret = mca_mpool_base_open())) {
        error = "mca_mpool_base_open() failed";
        goto error;
    }
    if (OMPI_SUCCESS != (ret = mca_pml_base_open())) {
        error = "mca_pml_base_open() failed";
        goto error;
    }
    if (OMPI_SUCCESS != (ret = mca_coll_base_open())) {
        error = "mca_coll_base_open() failed";
        goto error;
    }

    if (OMPI_SUCCESS != (ret = ompi_osc_base_open())) {
        error = "ompi_osc_base_open() failed";
        goto error;
    }

#if OPAL_ENABLE_FT_CR == 1
    if (OMPI_SUCCESS != (ret = ompi_crcp_base_open())) {
        error = "ompi_crcp_base_open() failed";
        goto error;
    }
#endif

    /* In order to reduce the common case for MPI apps (where they
       don't use MPI-2 IO or MPI-1 topology functions), the io and
       topo frameworks are initialized lazily, at the first use of
       relevant functions (e.g., MPI_FILE_*, MPI_CART_*, MPI_GRAPH_*),
       so they are not opened here. */

    /* Select which MPI components to use */

    if (OMPI_SUCCESS != 
        (ret = mca_mpool_base_init(OMPI_ENABLE_PROGRESS_THREADS,
                                   OMPI_ENABLE_THREAD_MULTIPLE))) {
        error = "mca_mpool_base_init() failed";
        goto error;
    }

    if (OMPI_SUCCESS != 
        (ret = mca_pml_base_select(OMPI_ENABLE_PROGRESS_THREADS,
                                   OMPI_ENABLE_THREAD_MULTIPLE))) {
        error = "mca_pml_base_select() failed";
        goto error;
    }

    /* check for timing request - get stop time and report elapsed time if so */
    if (timing && 0 == OMPI_PROC_MY_NAME->vpid) {
        gettimeofday(&ompistop, NULL);
        opal_output(0, "ompi_mpi_init[%ld]: time from completion of rte_init to modex %ld usec",
                    (long)OMPI_PROC_MY_NAME->vpid,
                    (long int)((ompistop.tv_sec - ompistart.tv_sec)*1000000 +
                               (ompistop.tv_usec - ompistart.tv_usec)));
        gettimeofday(&ompistart, NULL);
    }
    
    /* exchange connection info - this function also acts as a barrier
     * as it will not return until the exchange is complete
     */
    coll = OBJ_NEW(ompi_rte_collective_t);
    coll->id = ompi_process_info.peer_modex;
    if (OMPI_SUCCESS != (ret = ompi_rte_modex(coll))) {
        error = "rte_modex failed";
        goto error;
    }
    /* wait for modex to complete - this may be moved anywhere in mpi_init
     * so long as it occurs prior to calling a function that needs
     * the modex info!
     */
    while (coll->active) {
        opal_progress();  /* block in progress pending events */
    }
    OBJ_RELEASE(coll);

    if (timing && 0 == OMPI_PROC_MY_NAME->vpid) {
        gettimeofday(&ompistop, NULL);
        opal_output(0, "ompi_mpi_init[%ld]: time to execute modex %ld usec",
                    (long)OMPI_PROC_MY_NAME->vpid,
                    (long int)((ompistop.tv_sec - ompistart.tv_sec)*1000000 +
                               (ompistop.tv_usec - ompistart.tv_usec)));
        gettimeofday(&ompistart, NULL);
    }

    /* select buffered send allocator component to be used */
    ret=mca_pml_base_bsend_init(OMPI_ENABLE_THREAD_MULTIPLE);
    if( OMPI_SUCCESS != ret ) {
        error = "mca_pml_base_bsend_init() failed";
        goto error;
    }

    if (OMPI_SUCCESS != 
        (ret = mca_coll_base_find_available(OMPI_ENABLE_PROGRESS_THREADS,
                                            OMPI_ENABLE_THREAD_MULTIPLE))) {
        error = "mca_coll_base_find_available() failed";
        goto error;
    }

    if (OMPI_SUCCESS != 
        (ret = ompi_osc_base_find_available(OMPI_ENABLE_PROGRESS_THREADS,
                                            OMPI_ENABLE_THREAD_MULTIPLE))) {
        error = "ompi_osc_base_find_available() failed";
        goto error;
    }

#if OPAL_ENABLE_FT_CR == 1
    if (OMPI_SUCCESS != (ret = ompi_crcp_base_select() ) ) {
        error = "ompi_crcp_base_select() failed";
        goto error;
    }
#endif

    /* io and topo components are not selected here -- see comment
       above about the io and topo frameworks being loaded lazily */

    /* Initialize each MPI handle subsystem */
    /* initialize requests */
    if (OMPI_SUCCESS != (ret = ompi_request_init())) {
        error = "ompi_request_init() failed";
        goto error;
    }

    if (OMPI_SUCCESS != (ret = ompi_message_init())) {
        error = "ompi_message_init() failed";
        goto error;
    }

    /* initialize info */
    if (OMPI_SUCCESS != (ret = ompi_info_init())) {
        error = "ompi_info_init() failed";
        goto error;
    }

    /* initialize error handlers */
    if (OMPI_SUCCESS != (ret = ompi_errhandler_init())) {
        error = "ompi_errhandler_init() failed";
        goto error;
    }

    /* initialize error codes */
    if (OMPI_SUCCESS != (ret = ompi_mpi_errcode_init())) {
        error = "ompi_mpi_errcode_init() failed";
        goto error;
    }
    
    /* initialize internal error codes */
    if (OMPI_SUCCESS != (ret = ompi_errcode_intern_init())) {
        error = "ompi_errcode_intern_init() failed";
        goto error;
    }
     
    /* initialize groups  */
    if (OMPI_SUCCESS != (ret = ompi_group_init())) {
        error = "ompi_group_init() failed";
        goto error;
    }

    /* initialize communicators */
    if (OMPI_SUCCESS != (ret = ompi_comm_init())) {
        error = "ompi_comm_init() failed";
        goto error;
    }

    /* initialize file handles */
    if (OMPI_SUCCESS != (ret = ompi_file_init())) {
        error = "ompi_file_init() failed";
        goto error;
    }

    /* initialize windows */
    if (OMPI_SUCCESS != (ret = ompi_win_init())) {
        error = "ompi_win_init() failed";
        goto error;
    }

    /* initialize attribute meta-data structure for comm/win/dtype */
    if (OMPI_SUCCESS != (ret = ompi_attr_init())) {
        error = "ompi_attr_init() failed";
        goto error;
    }

    /* identify the architectures of remote procs and setup
     * their datatype convertors, if required
     */
    if (OMPI_SUCCESS != (ret = ompi_proc_complete_init())) {
        error = "ompi_proc_complete_init failed";
        goto error;
    }

    /* If thread support was enabled, then setup OPAL to allow for
       them. */
    if ((OMPI_ENABLE_PROGRESS_THREADS == 1) ||
        (*provided != MPI_THREAD_SINGLE)) {
        opal_set_using_threads(true);
    }

    /* start PML/BTL's */
    ret = MCA_PML_CALL(enable(true));
    if( OMPI_SUCCESS != ret ) {
        error = "PML control failed";
        goto error;
    }

    /* add all ompi_proc_t's to PML */
    if (NULL == (procs = ompi_proc_world(&nprocs))) {
        error = "ompi_proc_world() failed";
        goto error;
    }
    ret = MCA_PML_CALL(add_procs(procs, nprocs));
    free(procs);
    /* If we got "unreachable", then print a specific error message.
       Otherwise, if we got some other failure, fall through to print
       a generic message. */
    if (OMPI_ERR_UNREACH == ret) {
        ompi_show_help("help-mpi-runtime",
                       "mpi_init:startup:pml-add-procs-fail", true);
        error = NULL;
        goto error;
    } else if (OMPI_SUCCESS != ret) {
        error = "PML add procs failed";
        goto error;
    }

    MCA_PML_CALL(add_comm(&ompi_mpi_comm_world.comm));
    MCA_PML_CALL(add_comm(&ompi_mpi_comm_self.comm));

    /*
     * Dump all MCA parameters if requested
     */
    if (ompi_mpi_show_mca_params) {
        ompi_show_all_mca_params(ompi_mpi_comm_world.comm.c_my_rank, 
                                 nprocs, 
                                 ompi_process_info.nodename);
    }

    /* Do we need to wait for a debugger? */
    ompi_rte_wait_for_debugger();

    /* check for timing request - get stop time and report elapsed
       time if so, then start the clock again */
    if (timing && 0 == OMPI_PROC_MY_NAME->vpid) {
        gettimeofday(&ompistop, NULL);
        opal_output(0, "ompi_mpi_init[%ld]: time from modex to first barrier %ld usec",
                    (long)OMPI_PROC_MY_NAME->vpid,
                    (long int)((ompistop.tv_sec - ompistart.tv_sec)*1000000 +
                               (ompistop.tv_usec - ompistart.tv_usec)));
        gettimeofday(&ompistart, NULL);
    }

    /* wait for everyone to reach this point */
    coll = OBJ_NEW(ompi_rte_collective_t);
    coll->id = ompi_process_info.peer_init_barrier;
    if (OMPI_SUCCESS != (ret = ompi_rte_barrier(coll))) {
        error = "rte_barrier failed";
        goto error;
    }
    /* wait for barrier to complete */
    while (coll->active) {
        opal_progress();  /* block in progress pending events */
    }
    OBJ_RELEASE(coll);

    /* check for timing request - get stop time and report elapsed
       time if so, then start the clock again */
    if (timing && 0 == OMPI_PROC_MY_NAME->vpid) {
        gettimeofday(&ompistop, NULL);
        opal_output(0, "ompi_mpi_init[%ld]: time to execute barrier %ld usec",
                    (long)OMPI_PROC_MY_NAME->vpid,
                    (long int)((ompistop.tv_sec - ompistart.tv_sec)*1000000 +
                               (ompistop.tv_usec - ompistart.tv_usec)));
        gettimeofday(&ompistart, NULL);
    }

#if OMPI_ENABLE_PROGRESS_THREADS == 0
    /* Start setting up the event engine for MPI operations.  Don't
       block in the event library, so that communications don't take
       forever between procs in the dynamic code.  This will increase
       CPU utilization for the remainder of MPI_INIT when we are
       blocking on RTE-level events, but may greatly reduce non-TCP
       latency. */
    opal_progress_set_event_flag(OPAL_EVLOOP_NONBLOCK);
#endif
    
    /* wire up the mpi interface, if requested.  Do this after the
       non-block switch for non-TCP performance.  Do before the
       polling change as anyone with a complex wire-up is going to be
       using the oob. */
    if (OMPI_SUCCESS != (ret = ompi_init_preconnect_mpi())) {
        error = "ompi_mpi_do_preconnect_all() failed";
        goto error;
    }

    /* Setup the publish/subscribe (PUBSUB) framework */
    if (OMPI_SUCCESS != (ret = ompi_pubsub_base_open())) {
        error = "ompi_pubsub_base_open() failed";
        goto error;
    }
    if (OMPI_SUCCESS != (ret = ompi_pubsub_base_select())) {
        error = "ompi_pubsub_base_select() failed";
        goto error;
    }
    
    /* Setup the dynamic process management (DPM) framework */
    if (OMPI_SUCCESS != (ret = ompi_dpm_base_open())) {
        error = "ompi_dpm_base_open() failed";
        goto error;
    }
    if (OMPI_SUCCESS != (ret = ompi_dpm_base_select())) {
        error = "ompi_dpm_base_select() failed";
        goto error;
    }

    /* Determine the overall threadlevel support of all processes 
       in MPI_COMM_WORLD. This has to be done before calling 
       coll_base_comm_select, since some of the collective components
       e.g. hierarch, might create subcommunicators. The threadlevel
       requested by all processes is required in order to know
       which cid allocation algorithm can be used. */
    if ( OMPI_SUCCESS != 
	 ( ret = ompi_comm_cid_init ())) {
	error = "ompi_mpi_init: ompi_comm_cid_init failed";
	goto error;
    }

    /* Init coll for the comms. This has to be after dpm_base_select, 
       (since dpm.mark_dyncomm is not set in the communicator creation
       function else), but before dpm.dyncom_init, since this function
       might require collective for the CID allocation. */
    if (OMPI_SUCCESS !=
        (ret = mca_coll_base_comm_select(MPI_COMM_WORLD))) {
        error = "mca_coll_base_comm_select(MPI_COMM_WORLD) failed";
        goto error;
    }

    if (OMPI_SUCCESS != 
        (ret = mca_coll_base_comm_select(MPI_COMM_SELF))) {
        error = "mca_coll_base_comm_select(MPI_COMM_SELF) failed";
        goto error;
    }


    
    /* Check whether we have been spawned or not.  We introduce that
       at the very end, since we need collectives, datatypes, ptls
       etc. up and running here.... */
    if (OMPI_SUCCESS != (ret = ompi_dpm.dyn_init())) {
        error = "ompi_comm_dyn_init() failed";
        goto error;
    }

    /*
     * Startup the Checkpoint/Restart Mech.
     * Note: Always do this so tools don't hang when
     * in a non-checkpointable build
     */
    if (OMPI_SUCCESS != (ret = ompi_cr_init())) {
        error = "ompi_cr_init";
        goto error;
    }

    /* Undo OPAL calling opal_progress_event_users_increment() during 
       opal_init, to get better latency when not using TCP.  Do 
       this *after* dyn_init, as dyn init uses lots of RTE 
       communication and we don't want to hinder the performance of 
       that code. */ 
    opal_progress_event_users_decrement(); 

    /* see if yield_when_idle was specified - if so, use it */
    param = mca_base_param_find("mpi", NULL, "yield_when_idle");
    mca_base_param_lookup_int(param, &value);
    if (value < 0) {
        /* if no info is provided, just default to conservative */
        opal_progress_set_yield_when_idle(true);
    } else {
        /* info was provided, so set idle accordingly */
        opal_progress_set_yield_when_idle(value == 0 ? false : true);
    }
    
    param = mca_base_param_find("mpi", NULL, "event_tick_rate");
    mca_base_param_lookup_int(param, &value);
    /* negative value means use default - just don't do anything */
    if (value >= 0) {
        opal_progress_set_event_poll_rate(value);
    }

    /* At this point, we are fully configured and in MPI mode.  Any
       communication calls here will work exactly like they would in
       the user's code.  Setup the connections between procs and warm
       them up with simple sends, if requested */

    if (OMPI_SUCCESS != (ret = ompi_mpiext_init())) {
        error = "ompi_mpiext_init";
        goto error;
    }

    /* Fall through */
 error:
    if (ret != OMPI_SUCCESS) {
        /* Only print a message if one was not already printed */
        if (NULL != error) {
            const char *err_msg = opal_strerror(ret);
            /* If RTE was not setup yet, don't use ompi_show_help */
            if (rte_setup) {
                ompi_show_help("help-mpi-runtime",
                               "mpi_init:startup:internal-failure", true,
                               "MPI_INIT", "MPI_INIT", error, err_msg, ret);
            } else {
                opal_show_help("help-mpi-runtime",
                               "mpi_init:startup:internal-failure", true,
                               "MPI_INIT", "MPI_INIT", error, err_msg, ret);
            }
        }
        return ret;
    }

    /* Initialize the registered datarep list to be empty */
    OBJ_CONSTRUCT(&ompi_registered_datareps, opal_list_t);

    /* Initialize the arrays used to store the F90 types returned by the
     *  MPI_Type_create_f90_XXX functions.
     */
    OBJ_CONSTRUCT( &ompi_mpi_f90_integer_hashtable, opal_hash_table_t);
    opal_hash_table_init(&ompi_mpi_f90_integer_hashtable, 16 /* why not? */);

    OBJ_CONSTRUCT( &ompi_mpi_f90_real_hashtable, opal_hash_table_t);
    opal_hash_table_init(&ompi_mpi_f90_real_hashtable, FLT_MAX_10_EXP);

    OBJ_CONSTRUCT( &ompi_mpi_f90_complex_hashtable, opal_hash_table_t);
    opal_hash_table_init(&ompi_mpi_f90_complex_hashtable, FLT_MAX_10_EXP);

    /* All done.  Wasn't that simple? */

    ompi_mpi_initialized = true;

    /* check for timing request - get stop time and report elapsed time if so */
    if (timing && 0 == OMPI_PROC_MY_NAME->vpid) {
        gettimeofday(&ompistop, NULL);
        opal_output(0, "ompi_mpi_init[%ld]: time from barrier to complete mpi_init %ld usec",
                    (long)OMPI_PROC_MY_NAME->vpid,
                    (long int)((ompistop.tv_sec - ompistart.tv_sec)*1000000 +
                               (ompistop.tv_usec - ompistart.tv_usec)));
    }

    return MPI_SUCCESS;
}
Tip: Filter by directory path e.g. /media app.js to search for public/media/app.js.
Tip: Use camelCasing e.g. ProjME to search for ProjectModifiedEvent.java.
Tip: Filter by extension type e.g. /repo .js to search for all .js files in the /repo directory.
Tip: Separate your search with spaces e.g. /ssh pom.xml to search for src/ssh/pom.xml.
Tip: Use ↑ and ↓ arrow keys to navigate and return to view the file.
Tip: You can also navigate files with Ctrl+j (next) and Ctrl+k (previous) and view the file with Ctrl+o.
Tip: You can also navigate files with Alt+j (next) and Alt+k (previous) and view the file with Alt+o.