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
|
<!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>11.2.3. Shared Memory — 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="11.2.4. InifiniBand / RoCE support" href="ib-and-roce.html" />
<link rel="prev" title="11.2.2. TCP" href="tcp.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 current"><a class="reference internal" href="index.html">11.2. Networking support</a><ul class="current">
<li class="toctree-l3"><a class="reference internal" href="ofi.html">11.2.1. OpenFabrics Interfaces (OFI) / Libfabric support</a></li>
<li class="toctree-l3"><a class="reference internal" href="tcp.html">11.2.2. TCP</a></li>
<li class="toctree-l3 current"><a class="current reference internal" href="#">11.2.3. Shared Memory</a><ul>
<li class="toctree-l4"><a class="reference internal" href="#the-sm-btl">11.2.3.1. The sm BTL</a></li>
<li class="toctree-l4"><a class="reference internal" href="#specifying-the-use-of-sm-for-mpi-messages">11.2.3.2. Specifying the Use of sm for MPI Messages</a></li>
<li class="toctree-l4"><a class="reference internal" href="#tuning-parameters-to-improve-performance">11.2.3.3. Tuning Parameters to Improve Performance</a></li>
<li class="toctree-l4"><a class="reference internal" href="#shared-memory-mechanisms">11.2.3.4. Shared Memory Mechanisms</a></li>
<li class="toctree-l4"><a class="reference internal" href="#shared-memory-mapping-on-the-filesystem">11.2.3.5. Shared Memory Mapping on the Filesystem</a></li>
</ul>
</li>
<li class="toctree-l3"><a class="reference internal" href="ib-and-roce.html">11.2.4. InifiniBand / RoCE support</a></li>
<li class="toctree-l3"><a class="reference internal" href="iwarp.html">11.2.5. iWARP Support</a></li>
<li class="toctree-l3"><a class="reference internal" href="cuda.html">11.2.6. CUDA</a></li>
<li class="toctree-l3"><a class="reference internal" href="rocm.html">11.2.7. ROCm</a></li>
</ul>
</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"><a class="reference internal" href="../coll-tuned.html">11.10. Tuning Collectives</a></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"><a href="index.html"><span class="section-number">11.2. </span>Networking support</a></li>
<li class="breadcrumb-item active"><span class="section-number">11.2.3. </span>Shared Memory</li>
<li class="wy-breadcrumbs-aside">
<a href="../../_sources/tuning-apps/networking/shared-memory.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="shared-memory">
<h1><span class="section-number">11.2.3. </span>Shared Memory<a class="headerlink" href="#shared-memory" title="Permalink to this heading"></a></h1>
<div class="section" id="the-sm-btl">
<h2><span class="section-number">11.2.3.1. </span>The sm BTL<a class="headerlink" href="#the-sm-btl" title="Permalink to this heading"></a></h2>
<p>The <code class="docutils literal notranslate"><span class="pre">sm</span></code> BTL is a low-latency, high-bandwidth mechanism for
transferring data between two processes via shared memory. This BTL
can only be used between processes executing on the same node.</p>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p>Between Open MPI version 1.8.0 and 4.1.x, the shared memory
BTL was named <code class="docutils literal notranslate"><span class="pre">vader</span></code>. As of Open MPI version 5.0.0, the
BTL has been renamed <code class="docutils literal notranslate"><span class="pre">sm</span></code>.</p>
</div>
<div class="admonition warning">
<p class="admonition-title">Warning</p>
<p>In Open MPI version 5.0.x, the name <code class="docutils literal notranslate"><span class="pre">vader</span></code> is simply
an alias for the <code class="docutils literal notranslate"><span class="pre">sm</span></code> BTL. Similarly, all
<code class="docutils literal notranslate"><span class="pre">vader_</span></code>-prefixed MCA parameters are automatically
aliased to their corresponding <code class="docutils literal notranslate"><span class="pre">sm_</span></code>-prefixed MCA
parameter.</p>
<p>This alias mechanism is a legacy transition device, and
will likely disappear in a future release of Open MPI.</p>
</div>
</div>
<hr class="docutils" />
<div class="section" id="specifying-the-use-of-sm-for-mpi-messages">
<h2><span class="section-number">11.2.3.2. </span>Specifying the Use of sm for MPI Messages<a class="headerlink" href="#specifying-the-use-of-sm-for-mpi-messages" title="Permalink to this heading"></a></h2>
<p>Typically, it is unnecessary to do so; OMPI will use the best BTL available
for each communication.</p>
<p>Nevertheless, you may use the MCA parameter <code class="docutils literal notranslate"><span class="pre">btl</span></code>. You should also
specify the <code class="docutils literal notranslate"><span class="pre">self</span></code> BTL for communications between a process and
itself. Furthermore, if not all processes in your job will run on the
same, single node, then you also need to specify a BTL for internode
communications. For example:</p>
<div class="highlight-sh notranslate"><div class="highlight"><pre><span></span>shell$<span class="w"> </span>mpirun<span class="w"> </span>--mca<span class="w"> </span>btl<span class="w"> </span>self,sm,tcp<span class="w"> </span>-n<span class="w"> </span><span class="m">16</span><span class="w"> </span>./a.out
</pre></div>
</div>
</div>
<hr class="docutils" />
<div class="section" id="tuning-parameters-to-improve-performance">
<h2><span class="section-number">11.2.3.3. </span>Tuning Parameters to Improve Performance<a class="headerlink" href="#tuning-parameters-to-improve-performance" title="Permalink to this heading"></a></h2>
<p>Mostly, the default values of the MCA parameters have already
been chosen to give good performance. To improve performance further
is a little bit of an art. Sometimes, it’s a matter of trading off
performance for memory.</p>
<ul class="simple">
<li><p><code class="docutils literal notranslate"><span class="pre">btl_sm_eager_limit</span></code>: If message data plus header information fits
within this limit, the message is sent “eagerly” — that is, a
sender attempts to write its entire message to shared buffers
without waiting for a receiver to be ready. Above this size, a
sender will only write the first part of a message, then wait for
the receiver to acknowledge its readiness before continuing. Eager
sends <em>can</em> improve performance by decoupling senders from
receivers.</p></li>
<li><p><code class="docutils literal notranslate"><span class="pre">btl_sm_max_send_size</span></code>: Large messages are sent in fragments of
this size. Larger segments <em>can</em> lead to greater efficiencies,
though they could perhaps also inhibit pipelining between sender and
receiver.</p></li>
<li><p><code class="docutils literal notranslate"><span class="pre">btl_sm_free_list_num</span></code>: This is the initial number of fragments on
each (eager and max) free list. The free lists can grow in response
to resource congestion, but you can increase this parameter to
pre-reserve space for more fragments.</p></li>
<li><p><code class="docutils literal notranslate"><span class="pre">btl_sm_backing_directory</span></code>: Directory to place backing files for
shared memory communication. This directory should be on a local
filesystem such as <code class="docutils literal notranslate"><span class="pre">/tmp</span></code> or <code class="docutils literal notranslate"><span class="pre">/dev/shm</span></code> (default: (linux) <code class="docutils literal notranslate"><span class="pre">/dev/shm</span></code>,
(others) session directory)</p></li>
</ul>
</div>
<hr class="docutils" />
<div class="section" id="shared-memory-mechanisms">
<h2><span class="section-number">11.2.3.4. </span>Shared Memory Mechanisms<a class="headerlink" href="#shared-memory-mechanisms" title="Permalink to this heading"></a></h2>
<p>The <code class="docutils literal notranslate"><span class="pre">sm</span></code> BTL supports two modes of shared memory communication:</p>
<ol class="arabic">
<li><p><strong>Two-copy:</strong> Otherwise known as “copy-in / copy-out”, this mode is
where the sender copies data into shared memory and the receiver
copies the data out.</p>
<p>This mechanism is always available.</p>
</li>
<li><p><strong>Single copy:</strong> In this mode, the sender or receiver makes a
single copy of the message data from the source buffer in one
process to the destination buffer in another process. Open MPI
supports three flavors of shared memory single-copy transfers:</p>
<ul>
<li><p><a class="reference external" href="https://knem.gitlabpages.inria.fr/">Linux KNEM</a>. This is a
standalone Linux kernel module, made specifically for HPC and MPI
libraries to enable high-performance single-copy message
transfers.</p>
<p>Open MPI must be able to find the KNEM header files in order to
build support for KNEM.</p>
</li>
<li><p><a class="reference external" href="https://github.com/hjelmn/xpmem">Linux XPMEM</a>. Similar to
KNEM, this is a standalone Linux kernel module, made specifically
for HPC and MPI libraries to enable high-performance single-copy
message transfers. It is derived from the Cray XPMEM system.</p>
<p>Open MPI must be able to find the XPMEM header files in order to
build support for XPMEM.</p>
</li>
<li><p>Linux Cross-Memory Attach (CMA). This mechanism is built-in to
modern versions of the Linux kernel. Although more performance
than the two-copy shared memory transfer mechanism, CMA is the
lowest performance of the single-copy mechanisms. However, CMA
is likely the most widely available because it is enabled by
default in several modern Linux distributions.</p>
<p>Open MPI must be built on a Linux system with a recent enough
Glibc and kernel version in order to build support for Linux CMA.</p>
</li>
</ul>
</li>
</ol>
<p>Which mechanism is used at run time depends both on how Open MPI was
built and how your system is configured. You can check to see which
single-copy mechanisms Open MPI was built with via two mechanisms:</p>
<ol class="arabic">
<li><p>At the end of running <code class="docutils literal notranslate"><span class="pre">configure</span></code>, Open MPI emits a list of
transports for which it found relevant header files and libraries
such that it will be able to build support for them. You might see
lines like this, for example:</p>
<div class="highlight-text notranslate"><div class="highlight"><pre><span></span>Shared memory/copy in+copy out: yes
Shared memory/Linux CMA: yes
Shared memory/Linux KNEM: no
Shared memory/XPMEM: no
</pre></div>
</div>
<p>The above output indicates that Open MPI will be built with 2-copy
(as mentioned above, 2-copy is <em>always</em> available) and with Linux
CMA support. KNEM and XPMEM support will <em>not</em> be built.</p>
</li>
<li><p>After Open MPI is installed, the <code class="docutils literal notranslate"><span class="pre">ompi_info</span></code> command can show
which <code class="docutils literal notranslate"><span class="pre">smsc</span></code> (shared memory single copy) components are
available:</p>
<div class="highlight-text notranslate"><div class="highlight"><pre><span></span>shell$ ompi_info | grep smsc
MCA smsc: cma (MCA v2.1.0, API v1.0.0, Component v5.1.0)
</pre></div>
</div>
<p>This Open MPI installation only supports the Linux CMA single-copy
mechanism.</p>
</li>
</ol>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p>As implied by the SMSC component names, none of them are
supported on macOS. macOS users will use the two-copy mechanism.</p>
</div>
</div>
<hr class="docutils" />
<div class="section" id="shared-memory-mapping-on-the-filesystem">
<h2><span class="section-number">11.2.3.5. </span>Shared Memory Mapping on the Filesystem<a class="headerlink" href="#shared-memory-mapping-on-the-filesystem" title="Permalink to this heading"></a></h2>
<p>The default location of the file is in the <code class="docutils literal notranslate"><span class="pre">/dev/shm</span></code> directory. If <code class="docutils literal notranslate"><span class="pre">/dev/shm</span></code>
does not exist on the system, the default location will be the OMPI session
directory. The path is typically something like:
<code class="docutils literal notranslate"><span class="pre">/dev/shm/sm_segment.nodename.user_id.job_id.my_node_rank</span></code>.
For example, the full path could be: <code class="docutils literal notranslate"><span class="pre">/dev/shm/sm_segment.x.1000.23c70000.0</span></code>.</p>
<p>You can use the MCA parameter <code class="docutils literal notranslate"><span class="pre">btl_sm_backing_directory</span></code> to place the
directory in a non-default location.</p>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p>The session directory can be customized via
PRRTE using <code class="docutils literal notranslate"><span class="pre">--prtemca</span> <span class="pre">prte_tmpdir_base</span> <span class="pre">/path/to/somewhere</span></code>.</p>
</div>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p>Even when using single-copy methods like CMA, a shared memory file is still
created for managing connection metadata.</p>
</div>
</div>
</div>
</div>
</div>
<footer><div class="rst-footer-buttons" role="navigation" aria-label="Footer">
<a href="tcp.html" class="btn btn-neutral float-left" title="11.2.2. TCP" accesskey="p" rel="prev"><span class="fa fa-arrow-circle-left" aria-hidden="true"></span> Previous</a>
<a href="ib-and-roce.html" class="btn btn-neutral float-right" title="11.2.4. InifiniBand / RoCE support" 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>
|
