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
|
<!DOCTYPE html>
<html class="writer-html5" lang="en">
<head>
<meta charset="utf-8" />
<meta name="viewport" content="width=device-width, initial-scale=1.0" />
<title>11.10. Tuning Collectives — Open MPI 5.0.8 documentation</title>
<link rel="stylesheet" type="text/css" href="../_static/pygments.css" />
<link rel="stylesheet" type="text/css" href="../_static/css/theme.css" />
<!--[if lt IE 9]>
<script src="../_static/js/html5shiv.min.js"></script>
<![endif]-->
<script data-url_root="../" id="documentation_options" src="../_static/documentation_options.js"></script>
<script src="../_static/jquery.js"></script>
<script src="../_static/underscore.js"></script>
<script src="../_static/_sphinx_javascript_frameworks_compat.js"></script>
<script src="../_static/doctools.js"></script>
<script src="../_static/sphinx_highlight.js"></script>
<script src="../_static/js/theme.js"></script>
<link rel="index" title="Index" href="../genindex.html" />
<link rel="search" title="Search" href="../search.html" />
<link rel="next" title="11.11. Benchmarking Open MPI applications" href="benchmarking.html" />
<link rel="prev" title="11.9.2. ROMIO" href="mpi-io/romio.html" />
</head>
<body class="wy-body-for-nav">
<div class="wy-grid-for-nav">
<nav data-toggle="wy-nav-shift" class="wy-nav-side">
<div class="wy-side-scroll">
<div class="wy-side-nav-search" >
<a href="../index.html" class="icon icon-home">
Open MPI
</a>
<div role="search">
<form id="rtd-search-form" class="wy-form" action="../search.html" method="get">
<input type="text" name="q" placeholder="Search docs" aria-label="Search docs" />
<input type="hidden" name="check_keywords" value="yes" />
<input type="hidden" name="area" value="default" />
</form>
</div>
</div><div class="wy-menu wy-menu-vertical" data-spy="affix" role="navigation" aria-label="Navigation menu">
<ul class="current">
<li class="toctree-l1"><a class="reference internal" href="../quickstart.html">1. Quick start</a></li>
<li class="toctree-l1"><a class="reference internal" href="../getting-help.html">2. Getting help</a></li>
<li class="toctree-l1"><a class="reference internal" href="../release-notes/index.html">3. Release notes</a></li>
<li class="toctree-l1"><a class="reference internal" href="../installing-open-mpi/index.html">4. Building and installing Open MPI</a></li>
<li class="toctree-l1"><a class="reference internal" href="../features/index.html">5. Open MPI-specific features</a></li>
<li class="toctree-l1"><a class="reference internal" href="../validate.html">6. Validating your installation</a></li>
<li class="toctree-l1"><a class="reference internal" href="../version-numbering.html">7. Version numbers and compatibility</a></li>
<li class="toctree-l1"><a class="reference internal" href="../mca.html">8. The Modular Component Architecture (MCA)</a></li>
<li class="toctree-l1"><a class="reference internal" href="../building-apps/index.html">9. Building MPI applications</a></li>
<li class="toctree-l1"><a class="reference internal" href="../launching-apps/index.html">10. Launching MPI applications</a></li>
<li class="toctree-l1 current"><a class="reference internal" href="index.html">11. Run-time operation and tuning MPI applications</a><ul class="current">
<li class="toctree-l2"><a class="reference internal" href="environment-var.html">11.1. Environment variables set for MPI applications</a></li>
<li class="toctree-l2"><a class="reference internal" href="networking/index.html">11.2. Networking support</a></li>
<li class="toctree-l2"><a class="reference internal" href="multithreaded.html">11.3. Running multi-threaded MPI applications</a></li>
<li class="toctree-l2"><a class="reference internal" href="dynamic-loading.html">11.4. Dynamically loading <code class="docutils literal notranslate"><span class="pre">libmpi</span></code> at runtime</a></li>
<li class="toctree-l2"><a class="reference internal" href="fork-system-popen.html">11.5. Calling fork(), system(), or popen() in MPI processes</a></li>
<li class="toctree-l2"><a class="reference internal" href="fault-tolerance/index.html">11.6. Fault tolerance</a></li>
<li class="toctree-l2"><a class="reference internal" href="large-clusters/index.html">11.7. Large Clusters</a></li>
<li class="toctree-l2"><a class="reference internal" href="affinity.html">11.8. Processor and memory affinity</a></li>
<li class="toctree-l2"><a class="reference internal" href="mpi-io/index.html">11.9. MPI-IO tuning options</a></li>
<li class="toctree-l2 current"><a class="current reference internal" href="#">11.10. Tuning Collectives</a><ul>
<li class="toctree-l3"><a class="reference internal" href="#fixed-forced-and-dynamic-decisions">11.10.1. Fixed, Forced, and Dynamic Decisions</a></li>
<li class="toctree-l3"><a class="reference internal" href="#forcing-an-algorithm">11.10.2. Forcing an Algorithm</a></li>
<li class="toctree-l3"><a class="reference internal" href="#dynamic-decisions-and-the-rules-file">11.10.3. Dynamic Decisions and the Rules File</a></li>
<li class="toctree-l3"><a class="reference internal" href="#collectives-and-their-algorithms">11.10.4. Collectives and their Algorithms</a><ul>
<li class="toctree-l4"><a class="reference internal" href="#allgather-id-0">11.10.4.1. Allgather (Id=0)</a></li>
<li class="toctree-l4"><a class="reference internal" href="#allgatherv-id-1">11.10.4.2. Allgatherv (Id=1)</a></li>
<li class="toctree-l4"><a class="reference internal" href="#allreduce-id-2">11.10.4.3. Allreduce (Id=2)</a></li>
<li class="toctree-l4"><a class="reference internal" href="#alltoall-id-3">11.10.4.4. Alltoall (Id=3)</a></li>
<li class="toctree-l4"><a class="reference internal" href="#alltoallv-id-4">11.10.4.5. Alltoallv (Id=4)</a></li>
<li class="toctree-l4"><a class="reference internal" href="#barrier-id-6">11.10.4.6. Barrier (Id=6)</a></li>
<li class="toctree-l4"><a class="reference internal" href="#bcast-id-7">11.10.4.7. Bcast (Id=7)</a></li>
<li class="toctree-l4"><a class="reference internal" href="#exscan-id-8">11.10.4.8. Exscan (Id=8)</a></li>
<li class="toctree-l4"><a class="reference internal" href="#gather-id-9">11.10.4.9. Gather (Id=9)</a></li>
<li class="toctree-l4"><a class="reference internal" href="#reduce-id-11">11.10.4.10. Reduce (Id=11)</a></li>
<li class="toctree-l4"><a class="reference internal" href="#reduce-scatter-id-12">11.10.4.11. Reduce_scatter (Id=12)</a></li>
<li class="toctree-l4"><a class="reference internal" href="#reduce-scatter-block-id-13">11.10.4.12. Reduce_scatter_block (Id=13)</a></li>
<li class="toctree-l4"><a class="reference internal" href="#scan-id-14">11.10.4.13. Scan (Id=14)</a></li>
<li class="toctree-l4"><a class="reference internal" href="#scatter-id-15">11.10.4.14. Scatter (Id=15)</a></li>
</ul>
</li>
</ul>
</li>
<li class="toctree-l2"><a class="reference internal" href="benchmarking.html">11.11. Benchmarking Open MPI applications</a></li>
<li class="toctree-l2"><a class="reference internal" href="heterogeneity.html">11.12. Building heterogeneous MPI applications</a></li>
</ul>
</li>
<li class="toctree-l1"><a class="reference internal" href="../app-debug/index.html">12. Debugging Open MPI Parallel Applications</a></li>
<li class="toctree-l1"><a class="reference internal" href="../developers/index.html">13. Developer’s guide</a></li>
<li class="toctree-l1"><a class="reference internal" href="../contributing.html">14. Contributing to Open MPI</a></li>
<li class="toctree-l1"><a class="reference internal" href="../license/index.html">15. License</a></li>
<li class="toctree-l1"><a class="reference internal" href="../history.html">16. History of Open MPI</a></li>
<li class="toctree-l1"><a class="reference internal" href="../man-openmpi/index.html">17. Open MPI manual pages</a></li>
<li class="toctree-l1"><a class="reference internal" href="../man-openshmem/index.html">18. OpenSHMEM manual pages</a></li>
</ul>
</div>
</div>
</nav>
<section data-toggle="wy-nav-shift" class="wy-nav-content-wrap"><nav class="wy-nav-top" aria-label="Mobile navigation menu" >
<i data-toggle="wy-nav-top" class="fa fa-bars"></i>
<a href="../index.html">Open MPI</a>
</nav>
<div class="wy-nav-content">
<div class="rst-content">
<div role="navigation" aria-label="Page navigation">
<ul class="wy-breadcrumbs">
<li><a href="../index.html" class="icon icon-home" aria-label="Home"></a></li>
<li class="breadcrumb-item"><a href="index.html"><span class="section-number">11. </span>Run-time operation and tuning MPI applications</a></li>
<li class="breadcrumb-item active"><span class="section-number">11.10. </span>Tuning Collectives</li>
<li class="wy-breadcrumbs-aside">
<a href="../_sources/tuning-apps/coll-tuned.rst.txt" rel="nofollow"> View page source</a>
</li>
</ul>
<hr/>
</div>
<div role="main" class="document" itemscope="itemscope" itemtype="http://schema.org/Article">
<div itemprop="articleBody">
<style>
.wy-table-responsive table td,.wy-table-responsive table th{white-space:normal}
</style><div class="section" id="tuning-collectives">
<h1><span class="section-number">11.10. </span>Tuning Collectives<a class="headerlink" href="#tuning-collectives" title="Permalink to this heading"></a></h1>
<p>Open MPI’s <code class="docutils literal notranslate"><span class="pre">coll</span></code> framework provides a number of components implementing
collective communication, including: <code class="docutils literal notranslate"><span class="pre">han</span></code>, <code class="docutils literal notranslate"><span class="pre">libnbc</span></code>, <code class="docutils literal notranslate"><span class="pre">self</span></code>, <code class="docutils literal notranslate"><span class="pre">ucc</span></code> <code class="docutils literal notranslate"><span class="pre">base</span></code>,
<code class="docutils literal notranslate"><span class="pre">hcoll</span></code>, <code class="docutils literal notranslate"><span class="pre">sync</span></code>, <code class="docutils literal notranslate"><span class="pre">xhc</span></code>, <code class="docutils literal notranslate"><span class="pre">accelerator</span></code>, <code class="docutils literal notranslate"><span class="pre">basic</span></code>, <code class="docutils literal notranslate"><span class="pre">ftagree</span></code>, <code class="docutils literal notranslate"><span class="pre">inter</span></code>, <code class="docutils literal notranslate"><span class="pre">portals4</span></code>,
and <code class="docutils literal notranslate"><span class="pre">tuned</span></code>. Some of these components may not be available depending on how
Open MPI was compiled and what hardware is available on the system. A run-time
decision based on each component’s self reported priority, selects which
component will be used. These priorities may be adjusted on the command line
or with any of the other usual ways of setting MCA variables, giving us a way
to influence or override component selection. In the end, which of the
available components is selected depends on a number of factors such as the
underlying hardware and the whether or not a specific collective is provided by
the component as not all components implement all collectives. However, there
is always a fallback <code class="docutils literal notranslate"><span class="pre">base</span></code> component that steps in and takes over when another
component fails to provide an implementation.</p>
<p>The remainder of this section describes the tuning options available in the
<code class="docutils literal notranslate"><span class="pre">tuned</span></code> collective component.</p>
<div class="section" id="fixed-forced-and-dynamic-decisions">
<h2><span class="section-number">11.10.1. </span>Fixed, Forced, and Dynamic Decisions<a class="headerlink" href="#fixed-forced-and-dynamic-decisions" title="Permalink to this heading"></a></h2>
<p>Open MPI’s <code class="docutils literal notranslate"><span class="pre">tuned</span></code> collective component has three modes of operation, <em>fixed
decision</em>, <em>forced algorithm</em>, and <em>dynamic decision</em>. Since different
collective algorithms perform better in different situations, the purpose of
these modes is, when a collective is called, to select an appropriate
algorithm.</p>
<p>In the <em>fixed decision</em> mode a decision tree, essentially a large set of nested
if-then-else-if blocks with baked in comm and message size thresholds derived
by measuring performance on existing clusters, selects one of a number of
available algorithms for the specific collective being called. <em>Fixed
decision</em> is the default. The <em>fixed decision</em> mode can work well if your
cluster hardware is similar to the systems profiled when constructing the
decision tree. However, in some cases the <em>fixed decision</em> rules can yield
less than ideal performance. For instance, when your hardware differs
substantially from that which was used to derive the fixed rules decision
tree.</p>
<p>In the <em>dynamic decision</em> mode a user can provide a set of rules encoded in an
ASCII file that tell the <code class="docutils literal notranslate"><span class="pre">tuned</span></code> component which algorithm should be for one or
more collectives as a function of communicator and message size. For the
collectives which rules are not provided the <em>fixed decision</em> mode is used.
More detail about the dynamic decision mode and rules file can be found in
section <a class="reference internal" href="#rulesfile"><span class="std std-ref">Rules File</span></a>.</p>
<p>In the <em>forced algorithm</em> mode one can specify an algorithm for a specific
collective on the command line. This is discussed in detail in section
<a class="reference internal" href="#forcealg"><span class="std std-ref">Forcing an Algorithm</span></a>.</p>
</div>
<div class="section" id="forcing-an-algorithm">
<span id="forcealg"></span><h2><span class="section-number">11.10.2. </span>Forcing an Algorithm<a class="headerlink" href="#forcing-an-algorithm" title="Permalink to this heading"></a></h2>
<p>The simplest method of tuning is to force the <code class="docutils literal notranslate"><span class="pre">tuned</span></code> collective component to
use a specific algorithm for all calls to a specific collective. This can be
done on the command line, or by any of the other usual ways of setting MCA
variables. For example, the following command line sets the algorithm to
linear for all MPI_Alltoall calls in the entire run.</p>
<div class="highlight-sh notranslate"><div class="highlight"><pre><span></span>shell$<span class="w"> </span>mpirun<span class="w"> </span>...<span class="w"> </span>--mca<span class="w"> </span>coll_tuned_use_dynamic_rules<span class="w"> </span><span class="m">1</span><span class="w"> </span><span class="se">\</span>
<span class="w"> </span>--mca<span class="w"> </span>coll_tuned_alltoall_algorithm<span class="w"> </span><span class="m">1</span><span class="w"> </span>...
</pre></div>
</div>
<p>When an algorithm is selected this way, the <em>fixed decision</em> rules associated
with that collective are short circuited. This is an easy way to select a
collective algorithm and is most effective when there is a single communicator
and message sizes are static with in the run. On the other hand, when the
message size for a given collective varies during the run, or there are
multiple communicators of different sizes, the effectiveness of this approach
may be limited. The <a class="reference internal" href="#rulesfile"><span class="std std-ref">rules file</span></a> provides a more comprehensive
and flexible approach to tuning. For collectives where an algorithm is not
forced on the command line, the <em>fixed decision</em> rules are used.</p>
</div>
<div class="section" id="dynamic-decisions-and-the-rules-file">
<span id="rulesfile"></span><h2><span class="section-number">11.10.3. </span>Dynamic Decisions and the Rules File<a class="headerlink" href="#dynamic-decisions-and-the-rules-file" title="Permalink to this heading"></a></h2>
<p>Given that the best choice of algorithm for a given collective depends on a
number of factors only known at run time, and that some of these factors may
vary within a run, setting an algorithm on the command line often is an
ineffective means of tuning. The rules file provides a means of choosing
an algorithm at run time based on communicator and message size. The rules
file can be specified on the command line, or the other usual ways to set MCA
variables. The file is parsed during initialization and takes affect there
after.</p>
<div class="highlight-sh notranslate"><div class="highlight"><pre><span></span>shell$<span class="w"> </span>mpirun<span class="w"> </span>...<span class="w"> </span>--mca<span class="w"> </span>coll_tuned_use_dynamic_rules<span class="w"> </span><span class="m">1</span><span class="w"> </span><span class="se">\</span>
<span class="w"> </span>--mca<span class="w"> </span>coll_tuned_dynamic_rules_filename<span class="w"> </span>/path/to/my_rules.conf<span class="w"> </span>...
</pre></div>
</div>
<p>The loaded set of rules then are used to select the algorithm
to use based on the collective, the communicator size, and the message size.
Collectives for which rules have not be specified in the file will make use of
the <em>fixed decision</em> rules as usual.</p>
<p>Dynamic tuning files are organized in this format:</p>
<div class="highlight-sh notranslate"><div class="highlight"><pre><span></span><span class="linenos"> 1</span>rule-file-version-2
<span class="linenos"> 2</span><span class="m">1</span><span class="w"> </span><span class="c1"># num of collectives</span>
<span class="linenos"> 3</span><span class="m">3</span><span class="w"> </span><span class="c1"># collective ID</span>
<span class="linenos"> 4</span><span class="m">1</span><span class="w"> </span><span class="c1"># number of comm sizes</span>
<span class="linenos"> 5</span><span class="c1">#=====================</span>
<span class="linenos"> 6</span><span class="m">64</span><span class="w"> </span><span class="c1"># comm size</span>
<span class="linenos"> 7</span><span class="m">14</span><span class="w"> </span><span class="c1"># number of rules</span>
<span class="linenos"> 8</span><span class="c1"># Bytes alg topo segs reqs</span>
<span class="linenos"> 9</span><span class="c1">#----------------------</span>
<span class="linenos">10</span><span class="m">0</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">0</span>
<span class="linenos">11</span><span class="m">512000</span><span class="w"> </span><span class="m">4</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">64</span>
<span class="linenos">12</span><span class="m">1536000</span><span class="w"> </span><span class="m">4</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">64</span>
<span class="linenos">13</span><span class="m">3072000</span><span class="w"> </span><span class="m">4</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">64</span>
<span class="linenos">14</span><span class="m">6144000</span><span class="w"> </span><span class="m">4</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">64</span>
<span class="linenos">15</span><span class="m">12288000</span><span class="w"> </span><span class="m">4</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">16</span>
<span class="linenos">16</span><span class="m">24576000</span><span class="w"> </span><span class="m">4</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">16</span>
<span class="linenos">17</span><span class="m">49152000</span><span class="w"> </span><span class="m">4</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">16</span>
<span class="linenos">18</span><span class="m">98304000</span><span class="w"> </span><span class="m">4</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">16</span>
<span class="linenos">19</span><span class="m">196608000</span><span class="w"> </span><span class="m">4</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">8</span>
<span class="linenos">20</span><span class="m">393216000</span><span class="w"> </span><span class="m">4</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">8</span>
<span class="linenos">21</span><span class="m">786432000</span><span class="w"> </span><span class="m">4</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">1</span>
<span class="linenos">22</span><span class="m">1572864000</span><span class="w"> </span><span class="m">4</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">1</span>
<span class="linenos">23</span><span class="m">2621440000</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">0</span><span class="w"> </span><span class="m">0</span>
</pre></div>
</div>
<p>The rules file effectively defines, for one or more collectives, a function of
two variables, which given communicator and message size, returns an algorithm
id to use for the call. This mechanism allows one to specify for each
collective, an algorithm for any number of ranges of message and communicator
sizes. As communicators are constructed, a search of the rules table is made
using the communicator size to select a set of message size rules to be used
with that communicator. Later as the collective is invoked, a search of the
message size rules associated with the communicator is made. The rule with the
nearest (less than) matching message size specifies the algorithm that is used.
The actual definition of <em>message size</em> is dependent on the collective in
question, see the section on <a class="reference internal" href="#collectivesandalgorithms"><span class="std std-ref">collectives and algorithms</span></a>
for details.</p>
<p>One may provide rules for as many collectives, communicator sizes, and message
sizes as desired. Simply repeat the sections as needed and adjust the relevant
count parameters. One must always provide a rule for message size of zero.
Message size rules are expected in ascending order. The last parameters in the
message size rule may or may not be used and have different meaning depending
on the collective and algorithm. The first two parameters in the rule following
the algorithm ID, <cite>topo</cite> and <cite>segment size</cite>, are always required. In version 2
of the file format a third parameter, <cite>max requests</cite>, may also be provided. A
release of Open MPI at least v5.0.7 is required for version 2 features.</p>
<p>The file format version specifier, <cite>rule-file-version-N</cite> where N is an integer
greater or equal to 1, should appear on the first line of the file. If the
version specifier is not present, the file format is assumed to be version 1.
Version 2 or greater is required to use the <cite>max requests</cite> parameter. Open MPI
releases older than v5.0.7 do not support the file format version
identifier. When using older releases of Open MPI do not include a version
specifier and do not use the <cite>max requests</cite> parameter in message size rules.</p>
</div>
<div class="section" id="collectives-and-their-algorithms">
<span id="collectivesandalgorithms"></span><h2><span class="section-number">11.10.4. </span>Collectives and their Algorithms<a class="headerlink" href="#collectives-and-their-algorithms" title="Permalink to this heading"></a></h2>
<p>The following table lists the collectives
implemented by the <code class="docutils literal notranslate"><span class="pre">tuned</span></code> collective component along with the enumeration value identifying it.
It is this value that must be used in the rules file when specifying a set of rules.
The definition of <em>message size</em> is dependent on the collective and is given in the table.
Tables describing the algorithms available for each collective, and there identifiers, are linked.</p>
<table class="colwidths-given docutils align-default" id="id1">
<caption><span class="caption-text">Collectives</span><a class="headerlink" href="#id1" title="Permalink to this table"></a></caption>
<colgroup>
<col style="width: 21%" />
<col style="width: 11%" />
<col style="width: 68%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>Collective</p></th>
<th class="head"><p>Id</p></th>
<th class="head"><p>Message Size</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p><a class="reference internal" href="#allgather"><span class="std std-ref">Allgather</span></a></p></td>
<td><p>0</p></td>
<td><p>datatype size * comm size * number of elements in send buffer</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="#allgatherv"><span class="std std-ref">Allgatherv</span></a></p></td>
<td><p>1</p></td>
<td><p>datatype size * sum of number of elements that are to be received from each process (sum of recvcounts)</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="#allreduce"><span class="std std-ref">Allreduce</span></a></p></td>
<td><p>2</p></td>
<td><p>datatype size * number of elements in send buffer</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="#alltoall"><span class="std std-ref">Alltoall</span></a></p></td>
<td><p>3</p></td>
<td><p>datatype size * comm size * number of elements to send to each process</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="#alltoallv"><span class="std std-ref">Alltoallv</span></a></p></td>
<td><p>4</p></td>
<td><p>not used</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="#barrier"><span class="std std-ref">Barrier</span></a></p></td>
<td><p>6</p></td>
<td><p>not used</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="#bcast"><span class="std std-ref">Bcast</span></a></p></td>
<td><p>7</p></td>
<td><p>datatype size * number of entries in buffer</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="#exscan"><span class="std std-ref">Exscan</span></a></p></td>
<td><p>8</p></td>
<td><p>datatype size * comm size</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="#gather"><span class="std std-ref">Gather</span></a></p></td>
<td><p>9</p></td>
<td><p>datatype size * comm size * number of elements in send buffer</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="#reduce"><span class="std std-ref">Reduce</span></a></p></td>
<td><p>11</p></td>
<td><p>datatype size * number of elements in send buffer</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="#reduce-scatter"><span class="std std-ref">Reduce_scatter</span></a></p></td>
<td><p>12</p></td>
<td><p>datatype size * sum of number of elements in result distributed to each process (sum of recvcounts)</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="#reduce-scatter-block"><span class="std std-ref">Reduce_scatter_block</span></a></p></td>
<td><p>13</p></td>
<td><p>datatype size * comm size * element count per block</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="#scan"><span class="std std-ref">Scan</span></a></p></td>
<td><p>14</p></td>
<td><p>datatype size * comm size</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="#scatter"><span class="std std-ref">Scatter</span></a></p></td>
<td><p>15</p></td>
<td><p>datatype size * number of elements in send buffer</p></td>
</tr>
</tbody>
</table>
<div class="section" id="allgather-id-0">
<span id="allgather"></span><h3><span class="section-number">11.10.4.1. </span>Allgather (Id=0)<a class="headerlink" href="#allgather-id-0" title="Permalink to this heading"></a></h3>
<table class="colwidths-given docutils align-default" id="id2">
<caption><span class="caption-text">Allgather Algorithms</span><a class="headerlink" href="#id2" title="Permalink to this table"></a></caption>
<colgroup>
<col style="width: 11%" />
<col style="width: 26%" />
<col style="width: 63%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>Id</p></th>
<th class="head"><p>Name</p></th>
<th class="head"><p>Description</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p>0</p></td>
<td><p>ignore</p></td>
<td><p>Use fixed rules</p></td>
</tr>
<tr class="row-odd"><td><p>1</p></td>
<td><p>linear</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>2</p></td>
<td><p>bruck-k-fanout</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>3</p></td>
<td><p>recursive_doubling</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>4</p></td>
<td><p>ring</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>5</p></td>
<td><p>neighbor</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>6</p></td>
<td><p>two_proc</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>7</p></td>
<td><p>sparbit</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>8</p></td>
<td><p>direct-messaging</p></td>
<td><p>…</p></td>
</tr>
</tbody>
</table>
</div>
<div class="section" id="allgatherv-id-1">
<span id="allgatherv"></span><h3><span class="section-number">11.10.4.2. </span>Allgatherv (Id=1)<a class="headerlink" href="#allgatherv-id-1" title="Permalink to this heading"></a></h3>
<table class="colwidths-given docutils align-default" id="id3">
<caption><span class="caption-text">Allgatherv Algorithms</span><a class="headerlink" href="#id3" title="Permalink to this table"></a></caption>
<colgroup>
<col style="width: 11%" />
<col style="width: 26%" />
<col style="width: 63%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>Id</p></th>
<th class="head"><p>Name</p></th>
<th class="head"><p>Description</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p>0</p></td>
<td><p>ignore</p></td>
<td><p>Use fixed rules</p></td>
</tr>
<tr class="row-odd"><td><p>1</p></td>
<td><p>default</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>2</p></td>
<td><p>bruck</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>3</p></td>
<td><p>ring</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>4</p></td>
<td><p>neighbor</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>5</p></td>
<td><p>two_proc</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>6</p></td>
<td><p>sparbit</p></td>
<td><p>…</p></td>
</tr>
</tbody>
</table>
</div>
<div class="section" id="allreduce-id-2">
<span id="allreduce"></span><h3><span class="section-number">11.10.4.3. </span>Allreduce (Id=2)<a class="headerlink" href="#allreduce-id-2" title="Permalink to this heading"></a></h3>
<table class="colwidths-given docutils align-default" id="id4">
<caption><span class="caption-text">Allreduce Algorithms</span><a class="headerlink" href="#id4" title="Permalink to this table"></a></caption>
<colgroup>
<col style="width: 11%" />
<col style="width: 26%" />
<col style="width: 63%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>Id</p></th>
<th class="head"><p>Name</p></th>
<th class="head"><p>Description</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p>0</p></td>
<td><p>ignore</p></td>
<td><p>Use fixed rules</p></td>
</tr>
<tr class="row-odd"><td><p>1</p></td>
<td><p>basic_linear</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>2</p></td>
<td><p>nonoverlapping</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>3</p></td>
<td><p>recursive_doubling</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>4</p></td>
<td><p>ring</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>5</p></td>
<td><p>segmented_ring</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>6</p></td>
<td><p>rabenseifner</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>7</p></td>
<td><p>allgather_reduce</p></td>
<td><p>…</p></td>
</tr>
</tbody>
</table>
</div>
<div class="section" id="alltoall-id-3">
<span id="alltoall"></span><h3><span class="section-number">11.10.4.4. </span>Alltoall (Id=3)<a class="headerlink" href="#alltoall-id-3" title="Permalink to this heading"></a></h3>
<table class="colwidths-given docutils align-default" id="id5">
<caption><span class="caption-text">Alltoall Algorithms</span><a class="headerlink" href="#id5" title="Permalink to this table"></a></caption>
<colgroup>
<col style="width: 11%" />
<col style="width: 26%" />
<col style="width: 63%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>Id</p></th>
<th class="head"><p>Name</p></th>
<th class="head"><p>Description</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p>0</p></td>
<td><p>ignore</p></td>
<td><p>Use fixed rules</p></td>
</tr>
<tr class="row-odd"><td><p>1</p></td>
<td><p>linear</p></td>
<td><p>Launches all non-blocking send/recv pairs and wait for them to complete.</p></td>
</tr>
<tr class="row-even"><td><p>2</p></td>
<td><p>pairwise</p></td>
<td><p>For comm size P, implemented as P rounds of blocking MPI_Sendrecv</p></td>
</tr>
<tr class="row-odd"><td><p>3</p></td>
<td><p>modified_bruck</p></td>
<td><p>An algorithm exploiting network packet quantization to achieve O(log) time complexity. Typically best for very small message sizes.</p></td>
</tr>
<tr class="row-even"><td><p>4</p></td>
<td><p>linear_sync</p></td>
<td><p>Keep N non-blocking MPI_Isend/Irecv pairs in flight at all times. N is set by the coll_tuned_alltoall_max_requests MCA variable.</p></td>
</tr>
<tr class="row-odd"><td><p>5</p></td>
<td><p>two_proc</p></td>
<td><p>An implementation tailored for alltoall between 2 ranks, otherwise it is not used.</p></td>
</tr>
</tbody>
</table>
</div>
<div class="section" id="alltoallv-id-4">
<span id="alltoallv"></span><h3><span class="section-number">11.10.4.5. </span>Alltoallv (Id=4)<a class="headerlink" href="#alltoallv-id-4" title="Permalink to this heading"></a></h3>
<table class="colwidths-given docutils align-default" id="id6">
<caption><span class="caption-text">Alltoallv Algorithms</span><a class="headerlink" href="#id6" title="Permalink to this table"></a></caption>
<colgroup>
<col style="width: 11%" />
<col style="width: 26%" />
<col style="width: 63%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>Id</p></th>
<th class="head"><p>Name</p></th>
<th class="head"><p>Description</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p>0</p></td>
<td><p>ignore</p></td>
<td><p>Use fixed rules</p></td>
</tr>
<tr class="row-odd"><td><p>1</p></td>
<td><p>basic_linear</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>2</p></td>
<td><p>pairwise</p></td>
<td><p>…</p></td>
</tr>
</tbody>
</table>
</div>
<div class="section" id="barrier-id-6">
<span id="barrier"></span><h3><span class="section-number">11.10.4.6. </span>Barrier (Id=6)<a class="headerlink" href="#barrier-id-6" title="Permalink to this heading"></a></h3>
<table class="colwidths-given docutils align-default" id="id7">
<caption><span class="caption-text">Barrier Algorithms</span><a class="headerlink" href="#id7" title="Permalink to this table"></a></caption>
<colgroup>
<col style="width: 11%" />
<col style="width: 26%" />
<col style="width: 63%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>Id</p></th>
<th class="head"><p>Name</p></th>
<th class="head"><p>Description</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p>0</p></td>
<td><p>ignore</p></td>
<td><p>Use fixed rules</p></td>
</tr>
<tr class="row-odd"><td><p>1</p></td>
<td><p>linear</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>2</p></td>
<td><p>double_ring</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>3</p></td>
<td><p>recursive_doubling</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>4</p></td>
<td><p>bruck</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>5</p></td>
<td><p>two_proc</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>6</p></td>
<td><p>tree</p></td>
<td><p>…</p></td>
</tr>
</tbody>
</table>
</div>
<div class="section" id="bcast-id-7">
<span id="bcast"></span><h3><span class="section-number">11.10.4.7. </span>Bcast (Id=7)<a class="headerlink" href="#bcast-id-7" title="Permalink to this heading"></a></h3>
<table class="colwidths-given docutils align-default" id="id8">
<caption><span class="caption-text">Bcast Algorithms</span><a class="headerlink" href="#id8" title="Permalink to this table"></a></caption>
<colgroup>
<col style="width: 11%" />
<col style="width: 26%" />
<col style="width: 63%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>Id</p></th>
<th class="head"><p>Name</p></th>
<th class="head"><p>Description</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p>0</p></td>
<td><p>ignore</p></td>
<td><p>Use fixed rules</p></td>
</tr>
<tr class="row-odd"><td><p>1</p></td>
<td><p>basic_linear</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>2</p></td>
<td><p>chain</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>3</p></td>
<td><p>pipeline</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>4</p></td>
<td><p>split_binary_tree</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>5</p></td>
<td><p>binary_tree</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>6</p></td>
<td><p>binomial</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>7</p></td>
<td><p>knomial</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>8</p></td>
<td><p>scatter_allgather</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>9</p></td>
<td><p>scatter_allgather_ring</p></td>
<td><p>…</p></td>
</tr>
</tbody>
</table>
</div>
<div class="section" id="exscan-id-8">
<span id="exscan"></span><h3><span class="section-number">11.10.4.8. </span>Exscan (Id=8)<a class="headerlink" href="#exscan-id-8" title="Permalink to this heading"></a></h3>
<table class="colwidths-given docutils align-default" id="id9">
<caption><span class="caption-text">Exscan Algorithms</span><a class="headerlink" href="#id9" title="Permalink to this table"></a></caption>
<colgroup>
<col style="width: 11%" />
<col style="width: 26%" />
<col style="width: 63%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>Id</p></th>
<th class="head"><p>Name</p></th>
<th class="head"><p>Description</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p>0</p></td>
<td><p>ignore</p></td>
<td><p>Use fixed rules</p></td>
</tr>
<tr class="row-odd"><td><p>1</p></td>
<td><p>linear</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>2</p></td>
<td><p>recursive_doubling</p></td>
<td><p>…</p></td>
</tr>
</tbody>
</table>
</div>
<div class="section" id="gather-id-9">
<span id="gather"></span><h3><span class="section-number">11.10.4.9. </span>Gather (Id=9)<a class="headerlink" href="#gather-id-9" title="Permalink to this heading"></a></h3>
<table class="colwidths-given docutils align-default" id="id10">
<caption><span class="caption-text">Gather Algorithms</span><a class="headerlink" href="#id10" title="Permalink to this table"></a></caption>
<colgroup>
<col style="width: 11%" />
<col style="width: 26%" />
<col style="width: 63%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>Id</p></th>
<th class="head"><p>Name</p></th>
<th class="head"><p>Description</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p>0</p></td>
<td><p>ignore</p></td>
<td><p>Use fixed rules</p></td>
</tr>
<tr class="row-odd"><td><p>1</p></td>
<td><p>basic_linear</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>2</p></td>
<td><p>binomial</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>3</p></td>
<td><p>linear_sync</p></td>
<td><p>…</p></td>
</tr>
</tbody>
</table>
</div>
<div class="section" id="reduce-id-11">
<span id="reduce"></span><h3><span class="section-number">11.10.4.10. </span>Reduce (Id=11)<a class="headerlink" href="#reduce-id-11" title="Permalink to this heading"></a></h3>
<table class="colwidths-given docutils align-default" id="id11">
<caption><span class="caption-text">Reduce Algorithms</span><a class="headerlink" href="#id11" title="Permalink to this table"></a></caption>
<colgroup>
<col style="width: 11%" />
<col style="width: 26%" />
<col style="width: 63%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>Id</p></th>
<th class="head"><p>Name</p></th>
<th class="head"><p>Description</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p>0</p></td>
<td><p>ignore</p></td>
<td><p>Use fixed rules</p></td>
</tr>
<tr class="row-odd"><td><p>1</p></td>
<td><p>linear</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>2</p></td>
<td><p>chain</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>3</p></td>
<td><p>pipeline</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>4</p></td>
<td><p>binary</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>5</p></td>
<td><p>binomial</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>6</p></td>
<td><p>in-order_binary</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>7</p></td>
<td><p>rabenseifner</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>8</p></td>
<td><p>knomial</p></td>
<td><p>…</p></td>
</tr>
</tbody>
</table>
</div>
<div class="section" id="reduce-scatter-id-12">
<span id="reduce-scatter"></span><h3><span class="section-number">11.10.4.11. </span>Reduce_scatter (Id=12)<a class="headerlink" href="#reduce-scatter-id-12" title="Permalink to this heading"></a></h3>
<table class="colwidths-given docutils align-default" id="id12">
<caption><span class="caption-text">Reduce_scatter Algorithms</span><a class="headerlink" href="#id12" title="Permalink to this table"></a></caption>
<colgroup>
<col style="width: 11%" />
<col style="width: 26%" />
<col style="width: 63%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>Id</p></th>
<th class="head"><p>Name</p></th>
<th class="head"><p>Description</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p>0</p></td>
<td><p>ignore</p></td>
<td><p>Use fixed rules</p></td>
</tr>
<tr class="row-odd"><td><p>1</p></td>
<td><p>non-overlapping</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>2</p></td>
<td><p>recursive_halving</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>3</p></td>
<td><p>ring</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>4</p></td>
<td><p>butterfly</p></td>
<td><p>…</p></td>
</tr>
</tbody>
</table>
</div>
<div class="section" id="reduce-scatter-block-id-13">
<span id="reduce-scatter-block"></span><h3><span class="section-number">11.10.4.12. </span>Reduce_scatter_block (Id=13)<a class="headerlink" href="#reduce-scatter-block-id-13" title="Permalink to this heading"></a></h3>
<table class="colwidths-given docutils align-default" id="id13">
<caption><span class="caption-text">Reduce_scatter_block Algorithms</span><a class="headerlink" href="#id13" title="Permalink to this table"></a></caption>
<colgroup>
<col style="width: 11%" />
<col style="width: 26%" />
<col style="width: 63%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>Id</p></th>
<th class="head"><p>Name</p></th>
<th class="head"><p>Description</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p>0</p></td>
<td><p>ignore</p></td>
<td><p>Use fixed rules</p></td>
</tr>
<tr class="row-odd"><td><p>1</p></td>
<td><p>basic_linear</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>2</p></td>
<td><p>recursive_doubling</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>3</p></td>
<td><p>recursive_halving</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>4</p></td>
<td><p>butterfly</p></td>
<td><p>…</p></td>
</tr>
</tbody>
</table>
</div>
<div class="section" id="scan-id-14">
<span id="scan"></span><h3><span class="section-number">11.10.4.13. </span>Scan (Id=14)<a class="headerlink" href="#scan-id-14" title="Permalink to this heading"></a></h3>
<table class="colwidths-given docutils align-default" id="id14">
<caption><span class="caption-text">Scan Algorithms</span><a class="headerlink" href="#id14" title="Permalink to this table"></a></caption>
<colgroup>
<col style="width: 11%" />
<col style="width: 26%" />
<col style="width: 63%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>Id</p></th>
<th class="head"><p>Name</p></th>
<th class="head"><p>Description</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p>0</p></td>
<td><p>ignore</p></td>
<td><p>Use fixed rules</p></td>
</tr>
<tr class="row-odd"><td><p>1</p></td>
<td><p>linear</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>2</p></td>
<td><p>recursive_doubling</p></td>
<td><p>…</p></td>
</tr>
</tbody>
</table>
</div>
<div class="section" id="scatter-id-15">
<span id="scatter"></span><h3><span class="section-number">11.10.4.14. </span>Scatter (Id=15)<a class="headerlink" href="#scatter-id-15" title="Permalink to this heading"></a></h3>
<table class="colwidths-given docutils align-default" id="id15">
<caption><span class="caption-text">Scatter Algorithms</span><a class="headerlink" href="#id15" title="Permalink to this table"></a></caption>
<colgroup>
<col style="width: 11%" />
<col style="width: 26%" />
<col style="width: 63%" />
</colgroup>
<thead>
<tr class="row-odd"><th class="head"><p>Id</p></th>
<th class="head"><p>Name</p></th>
<th class="head"><p>Description</p></th>
</tr>
</thead>
<tbody>
<tr class="row-even"><td><p>0</p></td>
<td><p>ignore</p></td>
<td><p>Use fixed rules</p></td>
</tr>
<tr class="row-odd"><td><p>1</p></td>
<td><p>basic_linear</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-even"><td><p>2</p></td>
<td><p>binomial</p></td>
<td><p>…</p></td>
</tr>
<tr class="row-odd"><td><p>3</p></td>
<td><p>linear_nb</p></td>
<td><p>…</p></td>
</tr>
</tbody>
</table>
</div>
</div>
</div>
</div>
</div>
<footer><div class="rst-footer-buttons" role="navigation" aria-label="Footer">
<a href="mpi-io/romio.html" class="btn btn-neutral float-left" title="11.9.2. ROMIO" accesskey="p" rel="prev"><span class="fa fa-arrow-circle-left" aria-hidden="true"></span> Previous</a>
<a href="benchmarking.html" class="btn btn-neutral float-right" title="11.11. Benchmarking Open MPI applications" accesskey="n" rel="next">Next <span class="fa fa-arrow-circle-right" aria-hidden="true"></span></a>
</div>
<hr/>
<div role="contentinfo">
<p>© Copyright 2003-2025, The Open MPI Community.
<span class="lastupdated">Last updated on 2025-05-30 16:41:43 UTC.
</span></p>
</div>
Built with <a href="https://www.sphinx-doc.org/">Sphinx</a> using a
<a href="https://github.com/readthedocs/sphinx_rtd_theme">theme</a>
provided by <a href="https://readthedocs.org">Read the Docs</a>.
</footer>
</div>
</div>
</section>
</div>
<script>
jQuery(function () {
SphinxRtdTheme.Navigation.enable(true);
});
</script>
</body>
</html>
|
