8 <object id="TOC" data="graphical-toc.svg" type="image/svg+xml"></object>
10 window.onload=function() { // Wait for the SVG to be loaded before changing it
11 var elem=document.querySelector("#TOC").contentDocument.getElementById("ConfigBox")
12 elem.style="opacity:0.93999999;fill:#ff0000;fill-opacity:0.1;stroke:#000000;stroke-width:0.35277778;stroke-linecap:round;stroke-linejoin:round;stroke-miterlimit:4;stroke-dasharray:none;stroke-dashoffset:0;stroke-opacity:1";
18 A number of options can be given at runtime to change the default
19 SimGrid behavior. For a complete list of all configuration options
20 accepted by the SimGrid version used in your simulator, simply pass
21 the --help configuration flag to your program. If some of the options
22 are not documented on this page, this is a bug that you should please
23 report so that we can fix it. Note that some of the options presented
24 here may not be available in your simulators, depending on the
25 :ref:`compile-time options <install_src_config>` that you used.
27 Setting Configuration Items
28 ---------------------------
30 There is several way to pass configuration options to the simulators.
31 The most common way is to use the ``--cfg`` command line argument. For
32 example, to set the item ``Item`` to the value ``Value``, simply
33 type the following on the command-line:
35 .. code-block:: console
37 $ my_simulator --cfg=Item:Value (other arguments)
39 Several ``--cfg`` command line arguments can naturally be used. If you
40 need to include spaces in the argument, don't forget to quote the
41 argument. You can even escape the included quotes (write ``@'`` for ``'`` if
42 you have your argument between simple quotes).
44 Another solution is to use the ``<config>`` tag in the platform file. The
45 only restriction is that this tag must occur before the first
46 platform element (be it ``<zone>``, ``<cluster>``, ``<peer>`` or whatever).
47 The ``<config>`` tag takes an ``id`` attribute, but it is currently
48 ignored so you don't really need to pass it. The important part is that
49 within that tag, you can pass one or several ``<prop>`` tags to specify
50 the configuration to use. For example, setting ``Item`` to ``Value``
51 can be done by adding the following to the beginning of your platform
57 <prop id="Item" value="Value"/>
60 A last solution is to pass your configuration directly in your program
61 with :cpp:func:`simgrid::s4u::Engine::set_config` or :cpp:func:`MSG_config`.
65 #include <simgrid/s4u.hpp>
67 int main(int argc, char *argv[]) {
68 simgrid::s4u::Engine e(&argc, argv);
70 simgrid::s4u::Engine::set_config("Item:Value");
77 Existing Configuration Items
78 ----------------------------
81 The full list can be retrieved by passing ``--help`` and
82 ``--help-cfg`` to an executable that uses SimGrid. Try passing
83 ``help`` as a value to get the list of values accepted by a given
84 option. For example, ``--cfg=plugin:help`` will give you the list
85 of plugins available in your installation of SimGrid.
87 - **bmf/max-iterations:** :ref:`cfg=bmf/max-iterations`
88 - **bmf/precision:** :ref:`cfg=bmf/precision`
90 - **contexts/factory:** :ref:`cfg=contexts/factory`
91 - **contexts/guard-size:** :ref:`cfg=contexts/guard-size`
92 - **contexts/nthreads:** :ref:`cfg=contexts/nthreads`
93 - **contexts/stack-size:** :ref:`cfg=contexts/stack-size`
94 - **contexts/synchro:** :ref:`cfg=contexts/synchro`
96 - **cpu/maxmin-selective-update:** :ref:`Cpu Optimization Level <options_model_optim>`
97 - **cpu/model:** :ref:`options_model_select`
98 - **cpu/optim:** :ref:`Cpu Optimization Level <options_model_optim>`
100 - **debug/breakpoint:** :ref:`cfg=debug/breakpoint`
101 - **debug/clean-atexit:** :ref:`cfg=debug/clean-atexit`
102 - **debug/verbose-exit:** :ref:`cfg=debug/verbose-exit`
104 - **exception/cutpath:** :ref:`cfg=exception/cutpath`
106 - **host/model:** :ref:`options_model_select`
108 - **maxmin/precision:** :ref:`cfg=maxmin/precision`
109 - **maxmin/concurrency-limit:** :ref:`cfg=maxmin/concurrency-limit`
111 - **msg/debug-multiple-use:** :ref:`cfg=msg/debug-multiple-use`
113 - **model-check:** :ref:`options_modelchecking`
114 - **model-check/checkpoint:** :ref:`cfg=model-check/checkpoint`
115 - **model-check/communications-determinism:** :ref:`cfg=model-check/communications-determinism`
116 - **model-check/dot-output:** :ref:`cfg=model-check/dot-output`
117 - **model-check/max-depth:** :ref:`cfg=model-check/max-depth`
118 - **model-check/property:** :ref:`cfg=model-check/property`
119 - **model-check/reduction:** :ref:`cfg=model-check/reduction`
120 - **model-check/replay:** :ref:`cfg=model-check/replay`
121 - **model-check/send-determinism:** :ref:`cfg=model-check/send-determinism`
122 - **model-check/setenv:** :ref:`cfg=model-check/setenv`
123 - **model-check/termination:** :ref:`cfg=model-check/termination`
124 - **model-check/timeout:** :ref:`cfg=model-check/timeout`
125 - **model-check/visited:** :ref:`cfg=model-check/visited`
127 - **network/bandwidth-factor:** :ref:`cfg=network/bandwidth-factor`
128 - **network/crosstraffic:** :ref:`cfg=network/crosstraffic`
129 - **network/latency-factor:** :ref:`cfg=network/latency-factor`
130 - **network/loopback-lat:** :ref:`cfg=network/loopback`
131 - **network/loopback-bw:** :ref:`cfg=network/loopback`
132 - **network/maxmin-selective-update:** :ref:`Network Optimization Level <options_model_optim>`
133 - **network/model:** :ref:`options_model_select`
134 - **network/optim:** :ref:`Network Optimization Level <options_model_optim>`
135 - **network/TCP-gamma:** :ref:`cfg=network/TCP-gamma`
136 - **network/weight-S:** :ref:`cfg=network/weight-S`
138 - **ns3/TcpModel:** :ref:`options_pls`
139 - **ns3/seed:** :ref:`options_pls`
140 - **path:** :ref:`cfg=path`
141 - **plugin:** :ref:`cfg=plugin`
143 - **storage/max_file_descriptors:** :ref:`cfg=storage/max_file_descriptors`
145 - **surf/precision:** :ref:`cfg=surf/precision`
147 - **For collective operations of SMPI,** please refer to Section :ref:`cfg=smpi/coll-selector`
148 - **smpi/auto-shared-malloc-thresh:** :ref:`cfg=smpi/auto-shared-malloc-thresh`
149 - **smpi/async-small-thresh:** :ref:`cfg=smpi/async-small-thresh`
150 - **smpi/barrier-finalization:** :ref:`cfg=smpi/barrier-finalization`
151 - **smpi/barrier-collectives:** :ref:`cfg=smpi/barrier-collectives`
152 - **smpi/buffering:** :ref:`cfg=smpi/buffering`
153 - **smpi/coll-selector:** :ref:`cfg=smpi/coll-selector`
154 - **smpi/comp-adjustment-file:** :ref:`cfg=smpi/comp-adjustment-file`
155 - **smpi/cpu-threshold:** :ref:`cfg=smpi/cpu-threshold`
156 - **smpi/display-allocs:** :ref:`cfg=smpi/display-allocs`
157 - **smpi/display-timing:** :ref:`cfg=smpi/display-timing`
158 - **smpi/errors-are-fatal:** :ref:`cfg=smpi/errors-are-fatal`
159 - **smpi/grow-injected-times:** :ref:`cfg=smpi/grow-injected-times`
160 - **smpi/host-speed:** :ref:`cfg=smpi/host-speed`
161 - **smpi/IB-penalty-factors:** :ref:`cfg=smpi/IB-penalty-factors`
162 - **smpi/iprobe:** :ref:`cfg=smpi/iprobe`
163 - **smpi/iprobe-cpu-usage:** :ref:`cfg=smpi/iprobe-cpu-usage`
164 - **smpi/init:** :ref:`cfg=smpi/init`
165 - **smpi/keep-temps:** :ref:`cfg=smpi/keep-temps`
166 - **smpi/ois:** :ref:`cfg=smpi/ois`
167 - **smpi/or:** :ref:`cfg=smpi/or`
168 - **smpi/os:** :ref:`cfg=smpi/os`
169 - **smpi/papi-events:** :ref:`cfg=smpi/papi-events`
170 - **smpi/pedantic:** :ref:`cfg=smpi/pedantic`
171 - **smpi/privatization:** :ref:`cfg=smpi/privatization`
172 - **smpi/privatize-libs:** :ref:`cfg=smpi/privatize-libs`
173 - **smpi/send-is-detached-thresh:** :ref:`cfg=smpi/send-is-detached-thresh`
174 - **smpi/shared-malloc:** :ref:`cfg=smpi/shared-malloc`
175 - **smpi/shared-malloc-hugepage:** :ref:`cfg=smpi/shared-malloc-hugepage`
176 - **smpi/simulate-computation:** :ref:`cfg=smpi/simulate-computation`
177 - **smpi/test:** :ref:`cfg=smpi/test`
178 - **smpi/wtime:** :ref:`cfg=smpi/wtime`
179 - **smpi/list-leaks** :ref:`cfg=smpi/list-leaks`
181 - **Tracing configuration options** can be found in Section :ref:`tracing_tracing_options`
183 - **storage/model:** :ref:`options_model_select`
185 - **vm/model:** :ref:`options_model_select`
189 Configuring the Platform Models
190 -------------------------------
192 .. _options_model_select:
194 Choosing the Platform Models
195 ............................
197 SimGrid comes with several network, CPU and disk models built in,
198 and you can change the used model at runtime by changing the passed
199 configuration. The three main configuration items are given below.
200 For each of these items, passing the special ``help`` value gives you
201 a short description of all possible values (for example,
202 ``--cfg=network/model:help`` will present all provided network
203 models). Also, ``--help-models`` should provide information about all
204 models for all existing resources.
206 - ``network/model``: specify the used network model. Possible values:
208 - **LV08 (default one):** Realistic network analytic model
209 (slow-start modeled by multiplying latency by 13.01, bandwidth by
210 .97; bottleneck sharing uses a payload of S=20537 for evaluating
211 RTT). Described in `Accuracy Study and Improvement of Network
212 Simulation in the SimGrid Framework
213 <http://mescal.imag.fr/membres/arnaud.legrand/articles/simutools09.pdf>`_.
214 - **Constant:** Simplistic network model where all communication
215 take a constant time (one second). This model provides the lowest
216 realism, but is (marginally) faster.
217 - **SMPI:** Realistic network model specifically tailored for HPC
218 settings (accurate modeling of slow start with correction factors on
219 three intervals: < 1KiB, < 64 KiB, >= 64 KiB). This model can be
220 :ref:`further configured <options_model_network>`.
221 - **IB:** Realistic network model specifically tailored for HPC
222 settings with InfiniBand networks (accurate modeling contention
223 behavior, based on the model explained in `this PhD work
224 <http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf>`_.
225 This model can be :ref:`further configured <options_model_network>`.
226 - **CM02:** Legacy network analytic model. Very similar to LV08, but
227 without corrective factors. The timings of small messages are thus
228 poorly modeled. This model is described in `A Network Model for
229 Simulation of Grid Application
230 <https://hal.inria.fr/inria-00071989/document>`_.
231 - **ns-3** (only available if you compiled SimGrid accordingly):
232 Use the packet-level network
233 simulators as network models (see :ref:`model_ns3`).
234 This model can be :ref:`further configured <options_pls>`.
236 - ``cpu/model``: specify the used CPU model. We have only one model
239 - **Cas01:** Simplistic CPU model (time=size/speed)
241 - ``host/model``: The host concept is the aggregation of a CPU with a
242 network card. Three models exists, but actually, only 2 of them are
243 interesting. The "compound" one is simply due to the way our
244 internal code is organized, and can easily be ignored. So at the
245 end, you have two host models: The default one allows aggregation of
246 an existing CPU model with an existing network model, but does not
247 allow parallel tasks because these beasts need some collaboration
248 between the network and CPU model.
250 - **default:** Default host model. Currently, CPU:Cas01 and
251 network:LV08 (with cross traffic enabled)
252 - **compound:** Host model that is automatically chosen if
253 you change the network and CPU models
254 - **ptask_L07:** Host model somehow similar to Cas01+CM02 but
255 allowing "parallel tasks", that are intended to model the moldable
256 tasks of the grid scheduling literature.
258 - ``storage/model``: specify the used storage model. Only one model is
260 - ``vm/model``: specify the model for virtual machines. Only one model
263 .. todo: make 'compound' the default host model.
265 .. _options_model_solver:
270 The different models rely on a linear inequalities solver to share
271 the underlying resources. SimGrid allows you to change the solver, but
272 be cautious, **don't change it unless you are 100% sure**.
274 - items ``cpu/solver``, ``network/solver``, ``disk/solver`` and ``host/solver``
275 allow you to change the solver for each model:
277 - **maxmin:** The default solver for all models except ptask. Provides a
278 max-min fairness allocation.
279 - **fairbottleneck:** The default solver for ptasks. Extends max-min to
280 allow heterogeneous resources.
281 - **bmf:** More realistic solver for heterogeneous resource sharing.
282 Implements BMF (Bottleneck max fairness) fairness. To be used with
283 parallel tasks instead of fair-bottleneck.
285 .. _options_model_optim:
290 The network and CPU models that are based on linear inequalities solver (that
291 is, all our analytical models) accept specific optimization
294 - items ``network/optim`` and ``cpu/optim`` (both default to 'Lazy'):
296 - **Lazy:** Lazy action management (partial invalidation in lmm +
297 heap in action remaining).
298 - **TI:** Trace integration. Highly optimized mode when using
299 availability traces (only available for the Cas01 CPU model for
301 - **Full:** Full update of remaining and variables. Slow but may be
302 useful when debugging.
304 - items ``network/maxmin-selective-update`` and
305 ``cpu/maxmin-selective-update``: configure whether the underlying
306 should be lazily updated or not. It should have no impact on the
307 computed timings, but should speed up the computation. |br| It is
308 still possible to disable this feature because it can reveal
309 counter-productive in very specific scenarios where the
310 interaction level is high. In particular, if all your
311 communication share a given backbone link, you should disable it:
312 without it, a simple regular loop is used to update each
313 communication. With it, each of them is still updated (because of
314 the dependency induced by the backbone), but through a complicated
315 and slow pattern that follows the actual dependencies.
317 .. _cfg=bmf/precision:
318 .. _cfg=maxmin/precision:
319 .. _cfg=surf/precision:
324 **Option** ``maxmin/precision`` **Default:** 1e-5 (in flops or bytes) |br|
325 **Option** ``surf/precision`` **Default:** 1e-9 (in seconds) |br|
326 **Option** ``bmf/precision`` **Default:** 1e-12 (no unit)
328 The analytical models handle a lot of floating point values. It is
329 possible to change the epsilon used to update and compare them through
330 this configuration item. Changing it may speedup the simulation by
331 discarding very small actions, at the price of a reduced numerical
332 precision. You can modify separately the precision used to manipulate
333 timings (in seconds) and the one used to manipulate amounts of work
336 .. _cfg=maxmin/concurrency-limit:
341 **Option** ``maxmin/concurrency-limit`` **Default:** -1 (no limit)
343 The maximum number of variables per resource can be tuned through this
344 option. You can have as many simultaneous actions per resources as you
345 want. If your simulation presents a very high level of concurrency, it
346 may help to use e.g. 100 as a value here. It means that at most 100
347 actions can consume a resource at a given time. The extraneous actions
348 are queued and wait until the amount of concurrency of the considered
349 resource lowers under the given boundary.
351 Such limitations help both to the simulation speed and simulation accuracy
352 on highly constrained scenarios, but the simulation speed suffers of this
353 setting on regular (less constrained) scenarios so it is off by default.
355 .. _cfg=bmf/max-iterations:
360 **Option** ``bmf/max-iterations`` **Default:** 1000
362 It may happen in some settings that the BMF solver fails to converge to
363 a solution, so there is a hard limit on the amount of iteration count to
364 avoid infinite loops.
366 .. _options_model_network:
368 Configuring the Network Model
369 .............................
371 .. _cfg=network/TCP-gamma:
373 Maximal TCP Window Size
374 ^^^^^^^^^^^^^^^^^^^^^^^
376 **Option** ``network/TCP-gamma`` **Default:** 4194304
378 The analytical models need to know the maximal TCP window size to take
379 the TCP congestion mechanism into account. On Linux, this value can
380 be retrieved using the following commands. Both give a set of values,
381 and you should use the last one, which is the maximal size.
383 .. code-block:: console
385 $ cat /proc/sys/net/ipv4/tcp_rmem # gives the sender window
386 $ cat /proc/sys/net/ipv4/tcp_wmem # gives the receiver window
388 .. _cfg=network/bandwidth-factor:
389 .. _cfg=network/latency-factor:
390 .. _cfg=network/weight-S:
392 Manual calibration factors
393 ^^^^^^^^^^^^^^^^^^^^^^^^^^
395 SimGrid can take network irregularities such as a slow startup or changing behavior depending on the message size into account.
396 The values provided by default were computed through data fitting one the timings of either packet-level simulators or direct
397 experiments on real platforms. These default values should be OK for most users, but simulation realism is really important to
398 you, you probably want to recalibrate the models (i.e., devise sensible values for your specific settings). This section only
399 describes how to pass new values to the models while the calibration process involved in the computation of these values is
400 described :ref:`in the relevant chapter <models_calibration>`.
402 We found out that many networking effects can be realistically modeled with three following correction factors. They were shown
403 to be enough to capture slow-start effects, the different transmission modes of MPI systems (eager vs. rendez-vous mode), or the
404 non linear effects of wifi sharing.
406 **Option** ``network/latency-factor`` **Default:** 1.0, but overridden by most models
408 This option specifies a multiplier to apply to the platform latency. The factor can either be a constant to apply to any
409 communication, or it can depend on the message size. The ``CM02`` model does not use any correction factor, so the
410 latency-factor remains to 1. The ``LV08`` model sets it to 13.01 to model slow-start, while the ``SMPI`` model has several
411 values, each for an interval of sizes. The default SMPI setting given below specifies for example that a message smaller than
412 257 bytes will get a latency multiplier of 2.01467 while a message which size is in [15424, 65472] will get a latency multiplier
413 of 3.48845. The ``wifi`` model goes further and uses a callback in the program to compute the factor that must be non-linear in
416 This multiplier is applied to the latency computed from the platform, that is the sum of all link latencies over the :ref:`network
417 path <platform_routing>` used by the considered communication.
419 Constant factors are easy to express, but the interval-based syntax used in SMPI is somewhat complex. It expects a set of
420 factors separated by semicolons, each of the form ``boundary:factor``. For example if your specification is
421 ``0:1;1000:2;5000:3``, it means that on [0, 1000) the factor is 1. On [1000,5000), the factor is 2 while the factor is 3 for
422 5000 and beyond. If your first interval does include size=0, then the default value of 1 is used before. Changing the factor
423 callback is not possible from the command line and must be done from your code, as shown in `this example
424 <https://framagit.org/simgrid/simgrid/tree/master/examples/cpp/network-factors/s4u-network-factors.cpp>`_. Note that the chosen
425 model only provide some default setting, not more. You can pick a ``LV08`` to get some of the settings, and override the latency
426 with an interval-based value.
428 SMPI default value: 65472:11.6436; 15424:3.48845; 9376:2.59299; 5776:2.18796; 3484:1.88101; 1426:1.61075; 732:1.9503;
429 257:1.95341;0:2.01467 (interval boundaries are sorted automatically). These values were computed by data fitting on the Stampede
430 Supercomputer at TACC, with optimal deployment of processes on nodes. To accurately model your settings, you should redo the
431 :ref:`calibration <models_calibration>`.
433 **Option** ``network/bandwidth-factor`` **Default:** 1.0, but overridden by most models
435 Setting this option automatically adjusts the bandwidth used by any given communication. As with latency-factor above, the value
436 can be a constant (``CM02`` uses 1 -- no correction -- while ``LV08`` uses 0.97 to discount TCP headers while computing the
437 payload bandwidth), interval-based (as the default provided by the ``SMPI``), or using in-program callbacks (as with ``wifi``).
439 SMPI default value: 65472:0.940694;15424:0.697866;9376:0.58729;5776:1.08739;3484:0.77493;1426:0.608902;732:0.341987;257:0.338112;0:0.812084
440 This was also computed on the Stampede Supercomputer.
442 **Option** ``network/weight-S`` **Default:** depends on the model
444 Value of the bottleneck sharing, that is used to calculate RTT. Described in `Accuracy Study and Improvement of Network
445 Simulation in the SimGrid Framework <http://mescal.imag.fr/membres/arnaud.legrand/articles/simutools09.pdf>`_
447 Default values for ``CM02`` is 0. ``LV08`` sets it to 20537 while both ``SMPI`` and ``IB`` set it to 8775.
449 .. _cfg=network/loopback:
451 Configuring loopback link
452 ^^^^^^^^^^^^^^^^^^^^^^^^^
454 Several network model provide an implicit loopback link to account for local
455 communication on a host. By default it has a 10GBps bandwidth and a null latency.
456 This can be changed with ``network/loopback-lat`` and ``network/loopback-bw``
459 .. _cfg=smpi/IB-penalty-factors:
464 InfiniBand network behavior can be modeled through 3 parameters
465 ``smpi/IB-penalty-factors:"βe;βs;γs"``, as explained in `the PhD
466 thesis of Jean-Marc Vincent
467 <http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf>`_ (in French)
468 or more concisely in `this paper <https://hal.inria.fr/hal-00953618/document>`_,
469 even if that paper does only describe models for myrinet and ethernet.
470 You can see in Fig 2 some results for Infiniband, for example. This model
471 may be outdated by now for modern infiniband, anyway, so a new
472 validation would be good.
474 The three paramaters are defined as follows:
476 - βs: penalty factor for outgoing messages, computed by running a simple send to
477 two nodes and checking slowdown compared to a single send to one node,
479 - βe: penalty factor for ingoing messages, same computation method but with one
480 node receiving several messages
481 - γr: slowdown factor when communication buffer memory is saturated. It needs a
482 more complicated pattern to run in order to be computed (5.3 in the thesis,
483 page 107), and formula in the end is γr = time(c)/(3×βe×time(ref)), where
484 time(ref) is the time of a single comm with no contention).
486 Once these values are computed, a penalty is assessed for each message (this is
487 the part implemented in the simulator) as shown page 106 of the thesis. Here is
488 a simple translation of this text. First, some notations:
490 - ∆e(e) which corresponds to the incoming degree of node e, that is to say the number of communications having as destination node e.
491 - ∆s (s) which corresponds to the degree outgoing from node s, that is to say the number of communications sent by node s.
492 - Φ (e) which corresponds to the number of communications destined for the node e but coming from a different node.
493 - Ω (s, e) which corresponds to the number of messages coming from node s to node e. If node e only receives communications from different nodes then Φ (e) = ∆e (e). On the other hand if, for example, there are three messages coming from node s and going from node e then Φ (e) 6 = ∆e (e) and Ω (s, e) = 3
495 To determine the penalty for a communication, two values need to be calculated. First, the penalty caused by the conflict in transmission, noted ps.
498 - if ∆s (i) = 1 then ps = 1.
499 - if ∆s (i) ≥ 2 and ∆e (i) ≥ 3 then ps = ∆s (i) × βs × γr
500 - else, ps = ∆s (i) × βs
503 Then, the penalty caused by the conflict in reception (noted pe) should be computed as follows:
505 - if ∆e (i) = 1 then pe = 1
506 - else, pe = Φ (e) × βe × Ω (s, e)
508 Finally, the penalty associated with the communication is:
509 p = max (ps ∈ s, pe)
511 .. _cfg=network/crosstraffic:
513 Simulating Cross-Traffic
514 ^^^^^^^^^^^^^^^^^^^^^^^^
516 Since SimGrid v3.7, cross-traffic effects can be taken into account in
517 analytical simulations. It means that ongoing and incoming
518 communication flows are treated independently. In addition, the LV08
519 model adds 0.05 of usage on the opposite direction for each new
520 created flow. This can be useful to simulate some important TCP
521 phenomena such as ack compression.
523 For that to work, your platform must have two links for each
524 pair of interconnected hosts. An example of usable platform is
525 available in ``examples/platforms/crosstraffic.xml``.
527 This is activated through the ``network/crosstraffic`` item, that
528 can be set to 0 (disable this feature) or 1 (enable it).
530 Note that with the default host model this option is activated by default.
532 .. _cfg=smpi/async-small-thresh:
534 Simulating Asynchronous Send
535 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
537 (this configuration item is experimental and may change or disappear)
539 It is possible to specify that messages below a certain size (in bytes) will be
540 sent as soon as the call to MPI_Send is issued, without waiting for
541 the correspondent receive. This threshold can be configured through
542 the ``smpi/async-small-thresh`` item. The default value is 0. This
543 behavior can also be manually set for mailboxes, by setting the
544 receiving mode of the mailbox with a call to
545 :cpp:func:`MSG_mailbox_set_async`. After this, all messages sent to
546 this mailbox will have this behavior regardless of the message size.
548 This value needs to be smaller than or equals to the threshold set at
549 :ref:`cfg=smpi/send-is-detached-thresh`, because asynchronous messages
550 are meant to be detached as well.
557 **Option** ``ns3/TcpModel`` **Default:** "default" (ns-3 default)
559 When using ns-3, there is an extra item ``ns3/TcpModel``, corresponding
560 to the ``ns3::TcpL4Protocol::SocketType`` configuration item in
561 ns-3. The only valid values (enforced on the SimGrid side) are
562 'default' (no change to the ns-3 configuration), 'NewReno' or 'Reno' or
565 **Option** ``ns3/seed`` **Default:** "" (don't set the seed in ns-3)
567 This option is the random seed to provide to ns-3 with
568 ``ns3::RngSeedManager::SetSeed`` and ``ns3::RngSeedManager::SetRun``.
570 If left blank, no seed is set in ns-3. If the value 'time' is
571 provided, the current amount of seconds since epoch is used as a seed.
572 Otherwise, the provided value must be a number to use as a seed.
574 Configuring the Storage model
575 .............................
577 .. _cfg=storage/max_file_descriptors:
579 File Descriptor Count per Host
580 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
582 **Option** ``storage/max_file_descriptors`` **Default:** 1024
584 Each host maintains a fixed-size array of its file descriptors. You
585 can change its size through this item to either enlarge it if your
586 application requires it or to reduce it to save memory space.
593 SimGrid plugins allow one to extend the framework without changing its
594 source code directly. Read the source code of the existing plugins to
595 learn how to do so (in ``src/plugins``), and ask your questions to the
596 usual channels (Stack Overflow, Mailing list, IRC). The basic idea is
597 that plugins usually register callbacks to some signals of interest.
598 If they need to store some information about a given object (Link, CPU
599 or Actor), they do so through the use of a dedicated object extension.
601 Some of the existing plugins can be activated from the command line,
602 meaning that you can activate them from the command line without any
603 modification to your simulation code. For example, you can activate
604 the host energy plugin by adding ``--cfg=plugin:host_energy`` to your
607 Here is a partial list of plugins that can be activated this way. You can get
608 the full list by passing ``--cfg=plugin:help`` to your simulator.
610 - :ref:`Host Energy <plugin_host_energy>`: models the energy dissipation of the compute units.
611 - :ref:`Link Energy <plugin_link_energy>`: models the energy dissipation of the network.
612 - :ref:`Host Load <plugin_host_load>`: monitors the load of the compute units.
614 .. _options_modelchecking:
616 Configuring the Model-Checking
617 ------------------------------
619 To enable SimGrid's model-checking support, the program should
620 be executed using the simgrid-mc wrapper:
622 .. code-block:: console
624 $ simgrid-mc ./my_program
626 Safety properties are expressed as assertions using the function
627 :cpp:func:`void MC_assert(int prop)`.
629 .. _cfg=smpi/buffering:
631 Specifying the MPI buffering behavior
632 .....................................
634 **Option** ``smpi/buffering`` **Default:** infty
636 Buffering in MPI has a huge impact on the communication semantic. For example,
637 standard blocking sends are synchronous calls when the system buffers are full
638 while these calls can complete immediately without even requiring a matching
639 receive call for small messages sent when the system buffers are empty.
641 In SMPI, this depends on the message size, that is compared against two thresholds:
643 - if (size < :ref:`smpi/async-small-thresh <cfg=smpi/async-small-thresh>`) then
644 MPI_Send returns immediately, even if the corresponding receive has not be issued yet.
645 - if (:ref:`smpi/async-small-thresh <cfg=smpi/async-small-thresh>` < size < :ref:`smpi/send-is-detached-thresh <cfg=smpi/send-is-detached-thresh>`) then
646 MPI_Send returns as soon as the corresponding receive has been issued. This is known as the eager mode.
647 - if (:ref:`smpi/send-is-detached-thresh <cfg=smpi/send-is-detached-thresh>` < size) then
648 MPI_Send returns only when the message has actually been sent over the network. This is known as the rendez-vous mode.
650 The ``smpi/buffering`` (only valid with MC) option gives an easier interface to choose between these semantics. It can take two values:
652 - **zero:** means that buffering should be disabled. All communications are actually blocking.
653 - **infty:** means that buffering should be made infinite. All communications are non-blocking.
655 .. _cfg=model-check/property:
657 Specifying a liveness property
658 ..............................
660 **Option** ``model-check/property`` **Default:** unset
662 If you want to specify liveness properties, you have to pass them on
663 the command line, specifying the name of the file containing the
664 property, as formatted by the `ltl2ba <https://github.com/utwente-fmt/ltl2ba>`_ program.
665 Note that ltl2ba is not part of SimGrid and must be installed separately.
667 .. code-block:: console
669 $ simgrid-mc ./my_program --cfg=model-check/property:<filename>
671 .. _cfg=model-check/checkpoint:
673 Going for Stateful Verification
674 ...............................
676 By default, the system is backtracked to its initial state to explore
677 another path, instead of backtracking to the exact step before the fork
678 that we want to explore (this is called stateless verification). This
679 is done this way because saving intermediate states can rapidly
680 exhaust the available memory. If you want, you can change the value of
681 the ``model-check/checkpoint`` item. For example,
682 ``--cfg=model-check/checkpoint:1`` asks to take a checkpoint every
683 step. Beware, this will certainly explode your memory. Larger values
684 are probably better, make sure to experiment a bit to find the right
685 setting for your specific system.
687 .. _cfg=model-check/reduction:
689 Specifying the kind of reduction
690 ................................
692 The main issue when using the model-checking is the state space
693 explosion. You can activate some reduction technique with
694 ``--cfg=model-check/reduction:<technique>``. For now, this
695 configuration variable can take 2 values:
697 - **none:** Do not apply any kind of reduction (mandatory for
698 liveness properties, as our current DPOR algorithm breaks cycles)
699 - **dpor:** Apply Dynamic Partial Ordering Reduction. Only valid if
700 you verify local safety properties (default value for safety
703 Another way to mitigate the state space explosion is to search for
704 cycles in the exploration with the :ref:`cfg=model-check/visited`
705 configuration. Note that DPOR and state-equality reduction may not
706 play well together. You should choose between them.
708 Our current DPOR implementation could be improved in may ways. We are
709 currently improving its efficiency (both in term of reduction ability
710 and computational speed), and future work could make it compatible
711 with liveness properties.
713 .. _cfg=model-check/visited:
715 Size of Cycle Detection Set (state equality reduction)
716 ......................................................
718 Mc SimGrid can be asked to search for cycles during the exploration,
719 i.e. situations where a new explored state is in fact the same state
720 than a previous one.. This can prove useful to mitigate the state
721 space explosion with safety properties, and this is the crux when
722 searching for counter-examples to the liveness properties.
724 Note that this feature may break the current implementation of the
725 DPOR reduction technique.
727 The ``model-check/visited`` item is the maximum number of states, which
728 are stored in memory. If the maximum number of snapshotted state is
729 reached, some states will be removed from the memory and some cycles
730 might be missed. Small values can lead to incorrect verifications, but
731 large values can exhaust your memory and be CPU intensive as each new
732 state must be compared to that amount of older saved states.
734 The default settings depend on the kind of exploration. With safety
735 checking, no state is snapshotted and cycles cannot be detected. With
736 liveness checking, all states are snapshotted because missing a cycle
737 could hinder the exploration soundness.
739 .. _cfg=model-check/termination:
741 Non-Termination Detection
742 .........................
744 The ``model-check/termination`` configuration item can be used to
745 report if a non-termination execution path has been found. This is a
746 path with a cycle, which means that the program might never terminate.
748 This only works in safety mode, not in liveness mode.
750 This options is disabled by default.
752 .. _cfg=model-check/dot-output:
757 If set, the ``model-check/dot-output`` configuration item is the name
758 of a file in which to write a dot file of the path leading to the
759 property violation discovered (safety or liveness violation), as well
760 as the cycle for liveness properties. This dot file can then be fed to the
761 graphviz dot tool to generate a corresponding graphical representation.
763 .. _cfg=model-check/max-depth:
765 Exploration Depth Limit
766 .......................
768 The ``model-check/max-depth`` can set the maximum depth of the
769 exploration graph of the model checker. If this limit is reached, a
770 logging message is sent and the results might not be exact.
772 By default, the exploration is limited to the depth of 1000.
774 .. _cfg=model-check/timeout:
779 By default, the model checker does not handle timeout conditions: the `wait`
780 operations never time out. With the ``model-check/timeout`` configuration item
781 set to **yes**, the model checker will explore timeouts of `wait` operations.
783 .. _cfg=model-check/communications-determinism:
784 .. _cfg=model-check/send-determinism:
786 Communication Determinism
787 .........................
789 The ``model-check/communications-determinism`` and
790 ``model-check/send-determinism`` items can be used to select the
791 communication determinism mode of the model checker, which checks
792 determinism properties of the communications of an application.
794 .. _cfg=model-check/setenv:
796 Passing environment variables
797 .............................
799 You can specify extra environment variables to be set in the verified application
800 with ``model-check/setenv``. For example, you can preload a library as follows:
801 ``-cfg=model-check/setenv:LD_PRELOAD=toto;LD_LIBRARY_PATH=/tmp``.
805 Verification Performance Considerations
806 .......................................
808 The size of the stacks can have a huge impact on the memory
809 consumption when using model-checking. By default, each snapshot will
810 save a copy of the whole stacks and not only of the part that is
811 really meaningful: you should expect the contribution of the memory
812 consumption of the snapshots to be:
813 :math:`\text{number of processes} \times \text{stack size} \times \text{number of states}`.
815 When compiled against the model checker, the stacks are not
816 protected with guards: if the stack size is too small for your
817 application, the stack will silently overflow into other parts of the
818 memory (see :ref:`contexts/guard-size <cfg=contexts/guard-size>`).
820 .. _cfg=model-check/replay:
822 Replaying buggy execution paths from the model checker
823 ......................................................
825 Debugging the problems reported by the model checker is challenging:
826 First, the application under verification cannot be debugged with gdb
827 because the model checker already traces it. Then, the model checker may
828 explore several execution paths before encountering the issue, making it
829 very difficult to understand the output. Fortunately, SimGrid provides
830 the execution path leading to any reported issue so that you can replay
831 this path reported by the model checker, enabling the usage of classical
834 When the model checker finds an interesting path in the application
835 execution graph (where a safety or liveness property is violated), it
836 generates an identifier for this path. Here is an example of the output:
838 .. code-block:: console
840 [ 0.000000] (0:@) Check a safety property
841 [ 0.000000] (0:@) **************************
842 [ 0.000000] (0:@) *** PROPERTY NOT VALID ***
843 [ 0.000000] (0:@) **************************
844 [ 0.000000] (0:@) Counter-example execution trace:
845 [ 0.000000] (0:@) [(1)Tremblay (app)] MC_RANDOM(3)
846 [ 0.000000] (0:@) [(1)Tremblay (app)] MC_RANDOM(4)
847 [ 0.000000] (0:@) Path = 1/3;1/4
848 [ 0.000000] (0:@) Expanded states = 27
849 [ 0.000000] (0:@) Visited states = 68
850 [ 0.000000] (0:@) Executed transitions = 46
852 The interesting line is ``Path = 1/3;1/4``, which means that you should use
853 ``--cfg=model-check/replay:1/3;1/4`` to replay your application on the buggy
854 execution path. All options (but the model checker related ones) must
855 remain the same. In particular, if you ran your application with
856 ``smpirun -wrapper simgrid-mc``, then do it again. Remove all
857 MC-related options, keep non-MC-related ones and add
858 ``--cfg=model-check/replay:???``.
860 Currently, if the path is of the form ``X;Y;Z``, each number denotes
861 the actor's pid that is selected at each indecision point. If it's of
862 the form ``X/a;Y/b``, the X and Y are the selected pids while the a
863 and b are the return values of their simcalls. In the previous
864 example, ``1/3;1/4``, you can see from the full output that the actor
865 1 is doing MC_RANDOM simcalls, so the 3 and 4 simply denote the values
866 that these simcall return on the execution branch leading to the
869 Configuring the User Code Virtualization
870 ----------------------------------------
872 .. _cfg=contexts/factory:
874 Selecting the Virtualization Factory
875 ....................................
877 **Option** contexts/factory **Default:** "raw"
879 In SimGrid, the user code is virtualized in a specific mechanism that
880 allows the simulation kernel to control its execution: when a user
881 process requires a blocking action (such as sending a message), it is
882 interrupted, and only gets released when the simulated clock reaches
883 the point where the blocking operation is done. This is explained
884 graphically in the `relevant tutorial, available online
885 <https://simgrid.org/tutorials/simgrid-simix-101.pdf>`_.
887 In SimGrid, the containers in which user processes are virtualized are
888 called contexts. Several context factory are provided, and you can
889 select the one you want to use with the ``contexts/factory``
890 configuration item. Some of the following may not exist on your
891 machine because of portability issues. In any case, the default one
892 should be the most effcient one (please report bugs if the
893 auto-detection fails for you). They are approximately sorted here from
894 the slowest to the most efficient:
896 - **thread:** very slow factory using full featured threads (either
897 pthreads or windows native threads). They are slow but very
898 standard. Some debuggers or profilers only work with this factory.
899 - **java:** Java applications are virtualized onto java threads (that
900 are regular pthreads registered to the JVM)
901 - **ucontext:** fast factory using System V contexts (Linux and FreeBSD only)
902 - **boost:** This uses the `context
903 implementation <http://www.boost.org/doc/libs/1_59_0/libs/context/doc/html/index.html>`_
904 of the boost library for a performance that is comparable to our
906 |br| Install the relevant library (e.g. with the
907 libboost-contexts-dev package on Debian/Ubuntu) and recompile
909 - **raw:** amazingly fast factory using a context switching mechanism
910 of our own, directly implemented in assembly (only available for x86
911 and amd64 platforms for now) and without any unneeded system call.
913 The main reason to change this setting is when the debugging tools become
914 fooled by the optimized context factories. Threads are the most
915 debugging-friendly contexts, as they allow one to set breakpoints
916 anywhere with gdb and visualize backtraces for all processes, in order
917 to debug concurrency issues. Valgrind is also more comfortable with
918 threads, but it should be usable with all factories (Exception: the
919 callgrind tool really dislikes raw and ucontext factories).
921 .. _cfg=contexts/stack-size:
923 Adapting the Stack Size
924 .......................
926 **Option** ``contexts/stack-size`` **Default:** 8192 KiB
928 Each virtualized used process is executed using a specific system
929 stack. The size of this stack has a huge impact on the simulation
930 scalability, but its default value is rather large. This is because
931 the error messages that you get when the stack size is too small are
932 rather disturbing: this leads to stack overflow (overwriting other
933 stacks), leading to segfaults with corrupted stack traces.
935 If you want to push the scalability limits of your code, you might
936 want to reduce the ``contexts/stack-size`` item. Its default value is
937 8192 (in KiB), while our Chord simulation works with stacks as small
938 as 16 KiB, for example. You can ensure that some actors have a specific
939 size by simply changing the value of this configuration item before
940 creating these actors. The :cpp:func:`simgrid::s4u::Engine::set_config`
941 functions are handy for that.
943 This *setting is ignored* when using the thread factory (because there
944 is no way to modify the stack size with C++ system threads). Instead,
945 you should compile SimGrid and your application with
946 ``-fsplit-stack``. Note that this compilation flag is not compatible
947 with the model checker right now.
949 The operating system should only allocate memory for the pages of the
950 stack which are actually used and you might not need to use this in
951 most cases. However, this setting is very important when using the
952 model checker (see :ref:`options_mc_perf`).
954 .. _cfg=contexts/guard-size:
956 Disabling Stack Guard Pages
957 ...........................
959 **Option** ``contexts/guard-size`` **Default** 1 page in most case (0 pages on Windows or with MC)
961 Unless you use the threads context factory (see
962 :ref:`cfg=contexts/factory`), a stack guard page is usually used
963 which prevents the stack of a given actor from overflowing on another
964 stack. But the performance impact may become prohibitive when the
965 amount of actors increases. The option ``contexts/guard-size`` is the
966 number of stack guard pages used. By setting it to 0, no guard pages
967 will be used: in this case, you should avoid using small stacks (with
968 :ref:`contexts/stack-size <cfg=contexts/stack-size>`) as the stack
969 will silently overflow on other parts of the memory.
971 When no stack guard page is created, stacks may then silently overflow
972 on other parts of the memory if their size is too small for the
975 .. _cfg=contexts/nthreads:
976 .. _cfg=contexts/synchro:
978 Running User Code in Parallel
979 .............................
981 Parallel execution of the user code is only considered stable in
982 SimGrid v3.7 and higher, and mostly for MSG simulations. SMPI
983 simulations may well fail in parallel mode. It is described in
984 `INRIA RR-7653 <http://hal.inria.fr/inria-00602216/>`_.
986 If you are using the **ucontext** or **raw** context factories, you can
987 request to execute the user code in parallel. Several threads are
988 launched, each of them handling the same number of user contexts at each
989 run. To activate this, set the ``contexts/nthreads`` item to the amount
990 of cores that you have in your computer (or lower than 1 to have the
991 amount of cores auto-detected).
993 When parallel execution is activated, you can choose the
994 synchronization schema used with the ``contexts/synchro`` item,
995 which value is either:
997 - **futex:** ultra optimized synchronisation schema, based on futexes
998 (fast user-mode mutexes), and thus only available on Linux systems.
999 This is the default mode when available.
1000 - **posix:** slow but portable synchronisation using only POSIX
1002 - **busy_wait:** not really a synchronisation: the worker threads
1003 constantly request new contexts to execute. It should be the most
1004 efficient synchronisation schema, but it loads all the cores of
1005 your machine for no good reason. You probably prefer the other less
1008 Configuring the Tracing
1009 -----------------------
1011 The :ref:`tracing subsystem <outcome_vizu>` can be configured in
1012 several different ways depending on the used interface (S4U, SMPI)
1013 and the kind of traces that needs to be obtained. See the
1014 :ref:`Tracing Configuration Options subsection
1015 <tracing_tracing_options>` for a full description of each
1016 configuration option.
1018 We detail here a simple way to get the traces working for you, even if
1019 you never used the tracing API.
1022 - Any SimGrid-based simulator (MSG, SMPI, ...) and raw traces:
1024 .. code-block:: none
1026 --cfg=tracing:yes --cfg=tracing/uncategorized:yes
1028 The first parameter activates the tracing subsystem, and the second
1029 tells it to trace host and link utilization (without any
1032 - MSG-based simulator and categorized traces (you need to
1033 declare categories and classify your tasks according to them)
1035 .. code-block:: none
1037 --cfg=tracing:yes --cfg=tracing/categorized:yes
1039 The first parameter activates the tracing subsystem, and the second
1040 tells it to trace host and link categorized utilization.
1042 - SMPI simulator and traces for a space/time view:
1044 .. code-block:: console
1046 $ smpirun -trace ...
1048 The `-trace` parameter for the smpirun script runs the simulation
1049 with ``--cfg=tracing:yes --cfg=tracing/smpi:yes``. Check the
1050 smpirun's `-help` parameter for additional tracing options.
1052 Sometimes you might want to put additional information on the trace to
1053 correctly identify them later, or to provide data that can be used to
1054 reproduce an experiment. You have two ways to do that:
1056 - Add a string on top of the trace file as comment:
1058 .. code-block:: none
1060 --cfg=tracing/comment:my_simulation_identifier
1062 - Add the contents of a textual file on top of the trace file as comment:
1064 .. code-block:: none
1066 --cfg=tracing/comment-file:my_file_with_additional_information.txt
1068 Please, use these two parameters (for comments) to make reproducible
1069 simulations. For additional details about this and all tracing
1070 options, check See the :ref:`tracing_tracing_options`.
1075 .. _cfg=msg/debug-multiple-use:
1080 **Option** ``msg/debug-multiple-use`` **Default:** off
1082 Sometimes your application may try to send a task that is still being
1083 executed somewhere else, making it impossible to send this task. However,
1084 for debugging purposes, one may want to know what the other host is/was
1085 doing. This option shows a backtrace of the other process.
1090 The SMPI interface provides several specific configuration items.
1091 These are not easy to see with ``--help-cfg``, since SMPI binaries are usually launched through the ``smiprun`` script.
1093 .. _cfg=smpi/host-speed:
1094 .. _cfg=smpi/cpu-threshold:
1095 .. _cfg=smpi/simulate-computation:
1097 Automatic Benchmarking of SMPI Code
1098 ...................................
1100 In SMPI, the sequential code is automatically benchmarked, and these
1101 computations are automatically reported to the simulator. That is to
1102 say that if you have a large computation between a ``MPI_Recv()`` and
1103 a ``MPI_Send()``, SMPI will automatically benchmark the duration of
1104 this code, and create an execution task within the simulator to take
1105 this into account. For that, the actual duration is measured on the
1106 host machine and then scaled to the power of the corresponding
1107 simulated machine. The variable ``smpi/host-speed`` allows one to
1108 specify the computational speed of the host machine (in flop/s by
1109 default) to use when scaling the execution times.
1111 The default value is ``smpi/host-speed=20kf`` (= 20,000 flop/s). This
1112 is probably underestimated for most machines, leading SimGrid to
1113 overestimate the amount of flops in the execution blocks that are
1114 automatically injected in the simulator. As a result, the execution
1115 time of the whole application will probably be overestimated until you
1116 use a realistic value.
1118 When the code consists of numerous consecutive MPI calls, the
1119 previous mechanism feeds the simulation kernel with numerous tiny
1120 computations. The ``smpi/cpu-threshold`` item becomes handy when this
1121 impacts badly on the simulation performance. It specifies a threshold (in
1122 seconds) below which the execution chunks are not reported to the
1123 simulation kernel (default value: 1e-6).
1125 .. note:: The option ``smpi/cpu-threshold`` ignores any computation
1126 time spent below this threshold. SMPI does not consider the
1127 `amount of time` of these computations; there is no offset for
1128 this. Hence, a value that is too small, may lead to unreliable
1131 In some cases, however, one may wish to disable simulation of
1132 the computation of an application. This is the case when SMPI is used not to
1133 simulate an MPI application, but instead an MPI code that performs
1134 "live replay" of another MPI app (e.g., ScalaTrace's replay tool, or
1135 various on-line simulators that run an app at scale). In this case the
1136 computation of the replay/simulation logic should not be simulated by
1137 SMPI. Instead, the replay tool or on-line simulator will issue
1138 "computation events", which correspond to the actual MPI simulation
1139 being replayed/simulated. At the moment, these computation events can
1140 be simulated using SMPI by calling internal smpi_execute*() functions.
1142 To disable the benchmarking/simulation of a computation in the simulated
1143 application, the variable ``smpi/simulate-computation`` should be set
1144 to **no**. This option just ignores the timings in your simulation; it
1145 still executes the computations itself. If you want to stop SMPI from
1146 doing that, you should check the SMPI_SAMPLE macros, documented in
1147 Section :ref:`SMPI_use_faster`.
1149 +------------------------------------+-------------------------+-----------------------------+
1150 | Solution | Computations executed? | Computations simulated? |
1151 +====================================+=========================+=============================+
1152 | --cfg=smpi/simulate-computation:no | Yes | Never |
1153 +------------------------------------+-------------------------+-----------------------------+
1154 | --cfg=smpi/cpu-threshold:42 | Yes, in all cases | If it lasts over 42 seconds |
1155 +------------------------------------+-------------------------+-----------------------------+
1156 | SMPI_SAMPLE() macro | Only once per loop nest | Always |
1157 +------------------------------------+-------------------------+-----------------------------+
1159 .. _cfg=smpi/comp-adjustment-file:
1161 Slow-down or speed-up parts of your code
1162 ........................................
1164 **Option** ``smpi/comp-adjustment-file:`` **Default:** unset
1166 This option allows you to pass a file that contains two columns: The
1167 first column defines the section that will be subject to a speedup;
1168 the second column is the speedup. For instance:
1170 .. code-block:: none
1172 "start:stop","ratio"
1173 "exchange_1.f:30:exchange_1.f:130",1.18244559422142
1175 The first line is the header - you must include it. The following
1176 line means that the code between two consecutive MPI calls on line 30
1177 in exchange_1.f and line 130 in exchange_1.f should receive a speedup
1178 of 1.18244559422142. The value for the second column is therefore a
1179 speedup, if it is larger than 1 and a slowdown if it is smaller
1180 than 1. Nothing will be changed if it is equal to 1.
1182 Of course, you can set any arbitrary filenames you want (so the start
1183 and end don't have to be in the same file), but be aware that this
1184 mechanism only supports `consecutive calls!`
1186 Please note that you must pass the ``-trace-call-location`` flag to
1187 smpicc or smpiff, respectively. This flag activates some internal
1188 macro definitions that help with obtaining the call location.
1190 Bandwidth and latency factors
1191 .............................
1193 Adapting the bandwidth and latency acurately to the network conditions is of a paramount importance to get realistic results.
1194 This is done through the :ref:`network/bandwidth-factor <cfg=network/bandwidth-factor>` and :ref:`network/latency-factor
1195 <cfg=network/latency-factor>` items. You probably also want to read the following section: :ref:`models_calibration`.
1197 .. _cfg=smpi/display-timing:
1199 Reporting Simulation Time
1200 .........................
1202 **Option** ``smpi/display-timing`` **Default:** 0 (false)
1204 Most of the time, you run MPI code with SMPI to compute the time it
1205 would take to run it on a platform. But since the code is run through
1206 the ``smpirun`` script, you don't have any control on the launcher
1207 code, making it difficult to report the simulated time when the
1208 simulation ends. If you enable the ``smpi/display-timing`` item,
1209 ``smpirun`` will display this information when the simulation
1211 SMPI will also display information about the amout of real time spent
1212 in application code and in SMPI internals, to provide hints about the
1213 need to use sampling to reduce simulation time.
1215 .. _cfg=smpi/display-allocs:
1217 Reporting memory allocations
1218 ............................
1220 **Option** ``smpi/display-allocs`` **Default:** 0 (false)
1222 SMPI intercepts malloc and calloc calls performed inside the running
1223 application, if it wasn't compiled with SMPI_NO_OVERRIDE_MALLOC.
1224 With this option, SMPI will show at the end of execution the amount of
1225 memory allocated through these calls, and locate the most expensive one.
1226 This helps finding the targets for manual memory sharing, or the threshold
1227 to use for smpi/auto-shared-malloc-thresh option (see :ref:`cfg=smpi/auto-shared-malloc-thresh`).
1229 .. _cfg=smpi/keep-temps:
1231 Keeping temporary files after simulation
1232 ........................................
1234 **Option** ``smpi/keep-temps`` **default:** 0 (false)
1236 SMPI usually generates a lot of temporary files that are cleaned after
1237 use. This option requests to preserve them, for example to debug or
1238 profile your code. Indeed, the binary files are removed very early
1239 under the dlopen privatization schema, which tends to fool the
1242 .. _cfg=smpi/papi-events:
1244 Trace hardware counters with PAPI
1245 .................................
1247 **Option** ``smpi/papi-events`` **default:** unset
1249 When the PAPI support is compiled into SimGrid, this option takes the
1250 names of PAPI counters and adds their respective values to the trace
1251 files (See Section :ref:`tracing_tracing_options`).
1255 This feature currently requires superuser privileges, as registers
1256 are queried. Only use this feature with code you trust! Call
1257 smpirun for instance via ``smpirun -wrapper "sudo "
1258 <your-parameters>`` or run ``sudo sh -c "echo 0 >
1259 /proc/sys/kernel/perf_event_paranoid"`` In the later case, sudo
1260 will not be required.
1262 It is planned to make this feature available on a per-process (or per-thread?) basis.
1263 The first draft, however, just implements a "global" (i.e., for all processes) set
1264 of counters, the "default" set.
1266 .. code-block:: none
1268 --cfg=smpi/papi-events:"default:PAPI_L3_LDM:PAPI_L2_LDM"
1270 .. _cfg=smpi/privatization:
1272 Automatic Privatization of Global Variables
1273 ...........................................
1275 **Option** ``smpi/privatization`` **default:** "dlopen" (when using smpirun)
1277 MPI executables are usually meant to be executed in separate
1278 processes, but SMPI is executed in only one process. Global variables
1279 from executables will be placed in the same memory region and shared
1280 between processes, causing intricate bugs. Several options are
1281 possible to avoid this, as described in the main `SMPI publication
1282 <https://hal.inria.fr/hal-01415484>`_ and in the :ref:`SMPI
1283 documentation <SMPI_what_globals>`. SimGrid provides two ways of
1284 automatically privatizing the globals, and this option allows one to
1285 choose between them.
1287 - **no** (default when not using smpirun): Do not automatically
1288 privatize variables. Pass ``-no-privatize`` to smpirun to disable
1290 - **dlopen** or **yes** (default when using smpirun): Link multiple
1291 times against the binary.
1292 - **mmap** (slower, but maybe somewhat more stable):
1293 Runtime automatic switching of the data segments.
1296 This configuration option cannot be set in your platform file. You can only
1297 pass it as an argument to smpirun.
1299 .. _cfg=smpi/privatize-libs:
1301 Automatic privatization of global variables inside external libraries
1302 .....................................................................
1304 **Option** ``smpi/privatize-libs`` **default:** unset
1306 **Linux/BSD only:** When using dlopen (default) privatization,
1307 privatize specific shared libraries with internal global variables, if
1308 they can't be linked statically. For example libgfortran is usually
1309 used for Fortran I/O and indexes in files can be mixed up.
1311 Multiple libraries can be given, semicolon separated.
1313 This configuration option can only use either full paths to libraries,
1314 or full names. Check with ldd the name of the library you want to
1317 .. code-block:: console
1321 libgfortran.so.3 => /usr/lib/x86_64-linux-gnu/libgfortran.so.3 (0x00007fbb4d91b000)
1324 Then you can use ``--cfg=smpi/privatize-libs:libgfortran.so.3``
1325 or ``--cfg=smpi/privatize-libs:/usr/lib/x86_64-linux-gnu/libgfortran.so.3``,
1326 but not ``libgfortran`` nor ``libgfortran.so``.
1328 .. _cfg=smpi/send-is-detached-thresh:
1330 Simulating MPI detached send
1331 ............................
1333 **Option** ``smpi/send-is-detached-thresh`` **default:** 65536
1335 This threshold specifies the size in bytes under which the send will
1336 return immediately. This is different from the threshold detailed in
1337 :ref:`cfg=smpi/async-small-thresh` because the message is not
1338 really sent when the send is posted. SMPI still waits for the
1339 corresponding receive to be posted, in order to perform the communication
1342 .. _cfg=smpi/coll-selector:
1344 Simulating MPI collective algorithms
1345 ....................................
1347 **Option** ``smpi/coll-selector`` **Possible values:** naive (default), ompi, mpich
1349 SMPI implements more than 100 different algorithms for MPI collective
1350 communication, to accurately simulate the behavior of most of the
1351 existing MPI libraries. The ``smpi/coll-selector`` item can be used to
1352 select the decision logic either of the OpenMPI or the MPICH libraries. (By
1353 default SMPI uses naive version of collective operations.)
1355 Each collective operation can be manually selected with a
1356 ``smpi/collective_name:algo_name``. Available algorithms are listed in
1357 :ref:`SMPI_use_colls`.
1359 .. TODO:: All available collective algorithms will be made available
1360 via the ``smpirun --help-coll`` command.
1362 .. _cfg=smpi/barrier-collectives:
1364 Add a barrier in all collectives
1365 ................................
1367 **Option** ``smpi/barrier-collectives`` **default:** off
1369 This option adds a simple barrier in some collective operations to catch dangerous
1370 code that may or may not work depending on the MPI implementation: Bcast, Exscan,
1371 Gather, Gatherv, Scan, Scatter, Scatterv and Reduce.
1373 For example, the following code works with OpenMPI while it deadlocks in MPICH and
1374 Intel MPI. Broadcast seem to be "fire and forget" in OpenMPI while other
1375 implementations expect to receive a message.
1380 MPI_Bcast(buf1, buff_size, MPI_CHAR, 0, newcom);
1381 MPI_Send(&buf2, buff_size, MPI_CHAR, 1, tag, newcom);
1382 } else if (rank==1) {
1383 MPI_Recv(&buf2, buff_size, MPI_CHAR, 0, tag, newcom, MPI_STATUS_IGNORE);
1384 MPI_Bcast(buf1, buff_size, MPI_CHAR, 0, newcom);
1387 The barrier is only simulated and does not involve any additional message (it is a S4U barrier).
1388 This option is disabled by default, and activated by the `-analyze` flag of smpirun.
1390 .. _cfg=smpi/barrier-finalization:
1392 Add a barrier in MPI_Finalize
1393 .............................
1395 **Option** ``smpi/finalization-barrier`` **default:** off
1397 By default, SMPI processes are destroyed as soon as soon as their code ends,
1398 so after a successful MPI_Finalize call returns. In some rare cases, some data
1399 might have been attached to MPI objects still active in the remaining processes,
1400 and can be destroyed eagerly by the finished process.
1401 If your code shows issues at finalization, such as segmentation fault, triggering
1402 this option will add an explicit MPI_Barrier(MPI_COMM_WORLD) call inside the
1403 MPI_Finalize, so that all processes will terminate at almost the same point.
1404 It might affect the total timing by the cost of a barrier.
1406 .. _cfg=smpi/errors-are-fatal:
1408 Disable MPI fatal errors
1409 ........................
1411 **Option** ``smpi/errors-are-fatal`` **default:** on
1413 By default, SMPI processes will crash if a MPI error code is returned. MPI allows
1414 to explicitely set MPI_ERRORS_RETURN errhandler to avoid this behaviour. This flag
1415 will turn on this behaviour by default (for all concerned types and errhandlers).
1416 This can ease debugging by going after the first reported error.
1418 .. _cfg=smpi/pedantic:
1420 Disable pedantic MPI errors
1421 ...........................
1423 **Option** ``smpi/pedantic`` **default:** on
1425 By default, SMPI will report all errors it finds in MPI codes. Some of these errors
1426 may not be considered as errors by all developers. This flag can be turned off to
1427 avoid reporting some usually harmless mistakes.
1428 Concerned errors list (will be expanded in the future):
1430 - Calling MPI_Win_fence only once in a program, hence just opening an epoch without
1433 .. _cfg=smpi/iprobe:
1435 Inject constant times for MPI_Iprobe
1436 ....................................
1438 **Option** ``smpi/iprobe`` **default:** 0.0001
1440 The behavior and motivation for this configuration option is identical
1441 with :ref:`smpi/test <cfg=smpi/test>`, but for the function
1444 .. _cfg=smpi/iprobe-cpu-usage:
1446 Reduce speed for iprobe calls
1447 .............................
1449 **Option** ``smpi/iprobe-cpu-usage`` **default:** 1 (no change)
1451 MPI_Iprobe calls can be heavily used in applications. To account
1452 correctly for the energy that cores spend probing, it is necessary to
1453 reduce the load that these calls cause inside SimGrid.
1455 For instance, we measured a maximum power consumption of 220 W for a
1456 particular application but only 180 W while this application was
1457 probing. Hence, the correct factor that should be passed to this
1458 option would be 180/220 = 0.81.
1462 Inject constant times for MPI_Init
1463 ..................................
1465 **Option** ``smpi/init`` **default:** 0
1467 The behavior and motivation for this configuration option is identical
1468 with :ref:`smpi/test <cfg=smpi/test>`, but for the function ``MPI_Init()``.
1472 Inject constant times for MPI_Isend()
1473 .....................................
1475 **Option** ``smpi/ois``
1477 The behavior and motivation for this configuration option is identical
1478 with :ref:`smpi/os <cfg=smpi/os>`, but for the function ``MPI_Isend()``.
1482 Inject constant times for MPI_send()
1483 ....................................
1485 **Option** ``smpi/os``
1487 In several network models such as LogP, send (MPI_Send, MPI_Isend) and
1488 receive (MPI_Recv) operations incur costs (i.e., they consume CPU
1489 time). SMPI can factor these costs in as well, but the user has to
1490 configure SMPI accordingly as these values may vary by machine. This
1491 can be done by using ``smpi/os`` for MPI_Send operations; for MPI_Isend
1492 and MPI_Recv, use ``smpi/ois`` and ``smpi/or``, respectively. These work
1493 exactly as ``smpi/ois``.
1495 This item can consist of multiple sections; each section takes three
1496 values, for example ``1:3:2;10:5:1``. The sections are divided by ";"
1497 so this example contains two sections. Furthermore, each section
1498 consists of three values.
1500 1. The first value denotes the minimum size in bytes for this section to take effect;
1501 read it as "if message size is greater than this value (and other section has a larger
1502 first value that is also smaller than the message size), use this".
1503 In the first section above, this value is "1".
1505 2. The second value is the startup time; this is a constant value that will always
1506 be charged, no matter what the size of the message. In the first section above,
1509 3. The third value is the `per-byte` cost. That is, it is charged for every
1510 byte of the message (incurring cost messageSize*cost_per_byte)
1511 and hence accounts also for larger messages. In the first
1512 section of the example above, this value is "2".
1514 Now, SMPI always checks which section it should use for a given
1515 message; that is, if a message of size 11 is sent with the
1516 configuration of the example above, only the second section will be
1517 used, not the first, as the first value of the second section is
1518 closer to the message size. Hence, when ``smpi/os=1:3:2;10:5:1``, a
1519 message of size 11 incurs the following cost inside MPI_Send:
1520 ``5+11*1`` because 5 is the startup cost and 1 is the cost per byte.
1522 Note that the order of sections can be arbitrary; they will be ordered internally.
1526 Inject constant times for MPI_Recv()
1527 ....................................
1529 **Option** ``smpi/or``
1531 The behavior and motivation for this configuration option is identical
1532 with :ref:`smpi/os <cfg=smpi/os>`, but for the function ``MPI_Recv()``.
1535 .. _cfg=smpi/grow-injected-times:
1537 Inject constant times for MPI_Test
1538 ..................................
1540 **Option** ``smpi/test`` **default:** 0.0001
1542 By setting this option, you can control the amount of time a process
1543 sleeps when MPI_Test() is called; this is important, because SimGrid
1544 normally only advances the time while communication is happening and
1545 thus, MPI_Test will not add to the time, resulting in deadlock if it is
1546 used as a break-condition as in the following example:
1551 MPI_Test(request, flag, status);
1555 To speed up execution, we use a counter to keep track of how often we
1556 checked if the handle is now valid or not. Hence, we actually
1557 use counter*SLEEP_TIME, that is, the time MPI_Test() causes the
1558 process to sleep increases linearly with the number of previously
1559 failed tests. This behavior can be disabled by setting
1560 ``smpi/grow-injected-times`` to **no**. This will also disable this
1561 behavior for MPI_Iprobe.
1563 .. _cfg=smpi/shared-malloc:
1564 .. _cfg=smpi/shared-malloc-hugepage:
1569 **Option** ``smpi/shared-malloc`` **Possible values:** global (default), local
1571 If your simulation consumes too much memory, you may want to modify
1572 your code so that the working areas are shared by all MPI ranks. For
1573 example, in a block-cyclic matrix multiplication, you will only
1574 allocate one set of blocks, and all processes will share them.
1575 Naturally, this will lead to very wrong results, but this will save a
1576 lot of memory. So this is still desirable for some studies. For more on
1577 the motivation for that feature, please refer to the `relevant section
1578 <https://simgrid.github.io/SMPI_CourseWare/topic_understanding_performance/matrixmultiplication>`_
1579 of the SMPI CourseWare (see Activity #2.2 of the pointed
1580 assignment). In practice, change the calls for malloc() and free() into
1581 SMPI_SHARED_MALLOC() and SMPI_SHARED_FREE().
1583 SMPI provides two algorithms for this feature. The first one, called
1584 ``local``, allocates one block per call to SMPI_SHARED_MALLOC()
1585 (each call site gets its own block) ,and this block is shared
1586 among all MPI ranks. This is implemented with the shm_* functions
1587 to create a new POSIX shared memory object (kept in RAM, in /dev/shm)
1588 for each shared block.
1590 With the ``global`` algorithm, each call to SMPI_SHARED_MALLOC()
1591 returns a new address, but it only points to a shadow block: its memory
1592 area is mapped on a 1 MiB file on disk. If the returned block is of size
1593 N MiB, then the same file is mapped N times to cover the whole block.
1594 At the end, no matter how many times you call SMPI_SHARED_MALLOC, this will
1595 only consume 1 MiB in memory.
1597 You can disable this behavior and come back to regular mallocs (for
1598 example for debugging purposes) using ``no`` as a value.
1600 If you want to keep private some parts of the buffer, for instance if these
1601 parts are used by the application logic and should not be corrupted, you
1602 can use SMPI_PARTIAL_SHARED_MALLOC(size, offsets, offsets_count). For example:
1606 mem = SMPI_PARTIAL_SHARED_MALLOC(500, {27,42 , 100,200}, 2);
1608 This will allocate 500 bytes to mem, such that mem[27..41] and
1609 mem[100..199] are shared while other area remain private.
1611 Then, it can be deallocated by calling SMPI_SHARED_FREE(mem).
1613 When smpi/shared-malloc:global is used, the memory consumption problem
1614 is solved, but it may induce too much load on the kernel's pages table.
1615 In this case, you should use huge pages so that the kernel creates only one
1616 entry per MB of malloced data instead of one entry per 4 kB.
1617 To activate this, you must mount a hugetlbfs on your system and allocate
1618 at least one huge page:
1620 .. code-block:: console
1623 $ sudo mount none /home/huge -t hugetlbfs -o rw,mode=0777
1624 $ sudo sh -c 'echo 1 > /proc/sys/vm/nr_hugepages' # echo more if you need more
1626 Then, you can pass the option
1627 ``--cfg=smpi/shared-malloc-hugepage:/home/huge`` to smpirun to
1628 actually activate the huge page support in shared mallocs.
1630 .. _cfg=smpi/auto-shared-malloc-thresh:
1632 Automatically share allocations
1633 ...............................
1635 **Option** ``smpi/auto-shared-malloc-thresh:`` **Default:** 0 (false)
1636 This value in bytes represents the size above which all allocations
1637 will be "shared" by default (as if they were performed through
1638 SMPI_SHARED_MALLOC macros). Default = 0 = disabled feature.
1639 The value must be carefully chosen to only select data buffers which
1640 will not modify execution path or cause crash if their content is false.
1641 Option :ref:`cfg=smpi/display-allocs` can be used to locate the largest
1642 allocation detected in a run, and provide a good starting threshold.
1643 Note : malloc, calloc and free are overridden by smpicc/cxx by default.
1644 This can cause some troubles if codes are already overriding these. If this
1645 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
1646 help, but will make this feature unusable.
1650 Inject constant times for MPI_Wtime, gettimeofday and clock_gettime
1651 ...................................................................
1653 **Option** ``smpi/wtime`` **default:** 10 ns
1655 This option controls the amount of (simulated) time spent in calls to
1656 MPI_Wtime(), gettimeofday() and clock_gettime(). If you set this value
1657 to 0, the simulated clock is not advanced in these calls, which leads
1658 to issues if your application contains such a loop:
1662 while(MPI_Wtime() < some_time_bound) {
1663 /* some tests, with no communication nor computation */
1666 When the option smpi/wtime is set to 0, the time advances only on
1667 communications and computations. So the previous code results in an
1668 infinite loop: the current [simulated] time will never reach
1669 ``some_time_bound``. This infinite loop is avoided when that option
1670 is set to a small value, as it is by default since SimGrid v3.21.
1672 Note that if your application does not contain any loop depending on
1673 the current time only, then setting this option to a non-zero value
1674 will slow down your simulations by a tiny bit: the simulation loop has
1675 to be broken out of and reset each time your code asks for the current time.
1676 If the simulation speed really matters to you, you can avoid this
1677 extra delay by setting smpi/wtime to 0.
1679 .. _cfg=smpi/list-leaks:
1681 Report leaked MPI objects
1682 .........................
1684 **Option** ``smpi/list-leaks`` **default:** 0
1686 This option controls whether to report leaked MPI objects.
1687 The parameter is the number of leaks to report.
1689 Other Configurations
1690 --------------------
1692 .. _cfg=debug/clean-atexit:
1694 Cleanup at Termination
1695 ......................
1697 **Option** ``debug/clean-atexit`` **default:** on
1699 If your code is segfaulting during its finalization, it may help to
1700 disable this option to request that SimGrid not attempt any cleanups at
1701 the end of the simulation. Since the Unix process is ending anyway,
1702 the operating system will wipe it all.
1709 **Option** ``path`` **default:** . (current dir)
1711 It is possible to specify a list of directories to search in for the
1712 trace files (see :ref:`pf_trace`) by using this configuration
1713 item. To add several directory to the path, set the configuration
1714 item several times, as in ``--cfg=path:toto --cfg=path:tutu``
1716 .. _cfg=debug/breakpoint:
1721 **Option** ``debug/breakpoint`` **default:** unset
1723 This configuration option sets a breakpoint: when the simulated clock
1724 reaches the given time, a SIGTRAP is raised. This can be used to stop
1725 the execution and get a backtrace with a debugger.
1727 It is also possible to set the breakpoint from inside the debugger, by
1728 writing in global variable simgrid::kernel::cfg_breakpoint. For example,
1731 .. code-block:: none
1733 set variable simgrid::kernel::cfg_breakpoint = 3.1416
1735 .. _cfg=debug/verbose-exit:
1740 **Option** ``debug/verbose-exit`` **default:** on
1742 By default, when Ctrl-C is pressed, the status of all existing actors
1743 is displayed before exiting the simulation. This is very useful to
1744 debug your code, but it can become troublesome if you have many
1745 actors. Set this configuration item to **off** to disable this
1748 .. _cfg=exception/cutpath:
1750 Truncate local path from exception backtrace
1751 ............................................
1753 **Option** ``exception/cutpath`` **default:** off
1755 This configuration option is used to remove the path from the
1756 backtrace shown when an exception is thrown. This is mainly useful for
1757 the tests: the full file path would makes the tests non-reproducible because
1758 the paths of source files depend of the build settings. That would
1759 break most of the tests since their output is continually compared.
1763 Logging configuration
1764 ---------------------
1766 As introduced in :ref:`outcome_logs`, the SimGrid logging mechanism allows to configure at runtime the messages that should be displayed and those that should be omitted. Each
1767 message produced in the code is given a category (denoting its topic) and a priority. Then at runtime, each category is given a threshold (only messages of priority higher than
1768 that threshold are displayed), a layout (deciding how the messages in this category are formatted), and an appender (deciding what to do with the message: either print on stderr or
1771 This section explains how to configure this logging features. You can also refer to the documentation of the :ref:`programmer's interface <logging_prog>`, that allows to produce
1772 messages from your code.
1774 Most of the time, the logging mechanism is configured at runtime using the ``--log`` command-line argument, even if you can also use :c:func:`xbt_log_control_set()` to control it from
1775 your program. To pass configure more than one setting, you can either pass several ``--log`` arguments, or separate your settings with spaces, that must be quoted accordingly. In
1776 practice, the following is equivalent to the above settings: ``--log=root.thresh:error --log=s4u_host.thresh:debug``.
1778 If you want to specify more than one setting, you can either pass several ``--log`` argument to your program as above, or separate them with spaces. In this case, you want to quote
1779 your settings, as in ``--log="root.thresh:error s4u_host.thresh:debug"``. The parameters are interpreted in order, from left to right.
1782 Threshold configuration
1783 .......................
1785 The keyword ``threshold`` controls which logging event will get displayed in a given category. For example, ``--log=root.threshold:debug`` displays *every* message produced in the
1786 ``root`` category and its subcategories (i.e., every message produced -- this is *extremely* verbose), while ``--log=root.thres:critical`` turns almost everything off. As you can
1787 see, ``threshold`` can be abbreviated here.
1789 Existing thresholds:
1791 - ``trace`` some functions display a message at this level when entering or returning
1792 - ``debug`` output that is mostly useful when debugging the corresponding module.
1793 - ``verbose`` verbose output that is only mildly interesting and can easily be ignored
1794 - ``info`` usual output (this is the default threshold of all categories)
1795 - ``warning`` minor issue encountered
1796 - ``error`` issue encountered
1797 - ``critical`` major issue encountered, such as assertions failures
1801 Format configuration
1802 ....................
1804 The keyword ``fmt`` controls the layout (the format) of a logging category. For example, ``--log=root.fmt:%m`` reduces the output to the user-message only, removing any decoration such
1805 as the date, or the actor ID, everything. Existing format directives:
1808 - %n: line separator (LOG4J compatible)
1809 - %e: plain old space (SimGrid extension)
1811 - %m: user-provided message
1813 - %c: Category name (LOG4J compatible)
1814 - %p: Priority name (LOG4J compatible)
1816 - %h: Hostname (SimGrid extension)
1817 - %a: Actor name (SimGrid extension -- note that with SMPI this is the integer value of the process rank)
1818 - %i: Actor PID (SimGrid extension -- this is a 'i' as in 'i'dea)
1819 - %t: Thread "name" (LOG4J compatible -- actually the address of the thread in memory)
1821 - %F: file name where the log event was raised (LOG4J compatible)
1822 - %l: location where the log event was raised (LOG4J compatible, like '%%F:%%L' -- this is a l as in 'l'etter)
1823 - %L: line number where the log event was raised (LOG4J compatible)
1824 - %M: function name (LOG4J compatible -- called method name here of course).
1826 - %d: date (UNIX-like epoch)
1827 - %r: application age (time elapsed since the beginning of the application)
1830 ``--log=root.fmt:'[%h:%a:(%i) %r] %l: %m%n'`` gives you the default layout used for info messages while ``--log=root.fmt:'[%h:%a:(%i) %r] %l: [%c/%p] %m%n'`` gives you the default
1831 layout for the other priorities (it adds the source code location). Also, the actor identification is omitted by the default layout for the messages coming directly from the
1832 SimGrid kernel, so info messages are formatted with ``[%r] [%c/%p] %m%n`` in this case. When specifying the layout manually, such distinctions are currently impossible, and the
1833 provided layout is used for every messages.
1835 As with printf, you can specify the precision and width of the fields. For example, ``%.4r`` limits the date precision to four digits while ``%15h`` limits the host name to at most
1839 If you want to have spaces in your log format, you should protect it. Otherwise, SimGrid will consider that this is a space-separated list of several parameters. But you should
1840 also protect it from the shell that also splits command line arguments on spaces. At the end, you should use something such as ``--log="'root.fmt:%l: [%p/%c]: %m%n'"``.
1841 Another option is to use the ``%e`` directive for spaces, as in ``--log=root.fmt:%l:%e[%p/%c]:%e%m%n``.
1846 The keyword ``app`` controls the appended of a logging category. For example ``--log=root.app:file:mylogfile`` redirects every output to the file ``mylogfile``.
1848 With the ``splitfile`` appender, a new file is created when the size of the output reaches the specified size. The format is ``--log=root.app:splitfile:<size>:<file name>``. For
1849 example, ``--log=root.app:splitfile:500:mylog_%`` creates log files of at most 500 bytes, using the names ``mylog_0``, ``mylog_1``, ``mylog_2``, etc.
1851 The ``rollfile`` appender uses one file only, but the file is emptied and recreated when its size reaches the specified maximum. For example, ``--log=root.app:rollfile:500:mylog``
1852 ensures that the log file ``mylog`` will never overpass 500 bytes in size.
1854 Any appender setup this way have its own layout format, that you may change afterward. When specifying a new appender, its additivity is set to false to prevent log event displayed
1855 by this appender to "leak" to any other appender higher in the hierarchy. You can naturally change that if you want your messages to be displayed twice.
1860 The keyword ``add`` controls the additivity of a logging category. By default, the messages are only passed one appender only: the more specific, i.e. the first one found when
1861 climbing the tree from the category in which they were produced. In Log4J parlance, it is said that the default additivity of appenders is false. If you change this setting to
1862 ``on`` (or ``yes`` or ``1``), the produced messages will also be passed to the upper appender.
1864 Let's consider a more complex example: ``--log="root.app:file:all.log s4u.app:file:iface.log xbt.app:file:xbt.log xbt.add:yes``. Here, the logging of s4u will be sent to the
1865 ``iface.log`` file; the logging of the xbt toolbox will be sent to both the ``xbt.log`` file and the ``all.log`` file (because xbt additivity was enabled); and every other loggings
1866 will only be sent to ``all.log``.
1871 ``--help-logs`` displays a complete help message about logging in SimGrid.
1873 ``--help-log-categories`` displays the actual hierarchy of log categories for this binary.
1875 ``--log=no_loc`` hides the source locations (file names and line numbers) from the messages. This is useful to make tests reproducible.