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 - **contexts/factory:** :ref:`cfg=contexts/factory`
88 - **contexts/guard-size:** :ref:`cfg=contexts/guard-size`
89 - **contexts/nthreads:** :ref:`cfg=contexts/nthreads`
90 - **contexts/stack-size:** :ref:`cfg=contexts/stack-size`
91 - **contexts/synchro:** :ref:`cfg=contexts/synchro`
93 - **cpu/maxmin-selective-update:** :ref:`Cpu Optimization Level <options_model_optim>`
94 - **cpu/model:** :ref:`options_model_select`
95 - **cpu/optim:** :ref:`Cpu Optimization Level <options_model_optim>`
97 - **debug/breakpoint:** :ref:`cfg=debug/breakpoint`
98 - **debug/clean-atexit:** :ref:`cfg=debug/clean-atexit`
99 - **debug/verbose-exit:** :ref:`cfg=debug/verbose-exit`
101 - **exception/cutpath:** :ref:`cfg=exception/cutpath`
103 - **host/model:** :ref:`options_model_select`
105 - **maxmin/precision:** :ref:`cfg=maxmin/precision`
106 - **maxmin/concurrency-limit:** :ref:`cfg=maxmin/concurrency-limit`
108 - **msg/debug-multiple-use:** :ref:`cfg=msg/debug-multiple-use`
110 - **model-check:** :ref:`options_modelchecking`
111 - **model-check/checkpoint:** :ref:`cfg=model-check/checkpoint`
112 - **model-check/communications-determinism:** :ref:`cfg=model-check/communications-determinism`
113 - **model-check/dot-output:** :ref:`cfg=model-check/dot-output`
114 - **model-check/max-depth:** :ref:`cfg=model-check/max-depth`
115 - **model-check/property:** :ref:`cfg=model-check/property`
116 - **model-check/reduction:** :ref:`cfg=model-check/reduction`
117 - **model-check/replay:** :ref:`cfg=model-check/replay`
118 - **model-check/send-determinism:** :ref:`cfg=model-check/send-determinism`
119 - **model-check/termination:** :ref:`cfg=model-check/termination`
120 - **model-check/timeout:** :ref:`cfg=model-check/timeout`
121 - **model-check/visited:** :ref:`cfg=model-check/visited`
123 - **network/bandwidth-factor:** :ref:`cfg=network/bandwidth-factor`
124 - **network/crosstraffic:** :ref:`cfg=network/crosstraffic`
125 - **network/latency-factor:** :ref:`cfg=network/latency-factor`
126 - **network/loopback-lat:** :ref:`cfg=network/loopback`
127 - **network/loopback-bw:** :ref:`cfg=network/loopback`
128 - **network/maxmin-selective-update:** :ref:`Network Optimization Level <options_model_optim>`
129 - **network/model:** :ref:`options_model_select`
130 - **network/optim:** :ref:`Network Optimization Level <options_model_optim>`
131 - **network/TCP-gamma:** :ref:`cfg=network/TCP-gamma`
132 - **network/weight-S:** :ref:`cfg=network/weight-S`
134 - **ns3/TcpModel:** :ref:`options_pls`
135 - **ns3/seed:** :ref:`options_pls`
136 - **path:** :ref:`cfg=path`
137 - **plugin:** :ref:`cfg=plugin`
139 - **storage/max_file_descriptors:** :ref:`cfg=storage/max_file_descriptors`
141 - **surf/precision:** :ref:`cfg=surf/precision`
143 - **For collective operations of SMPI,** please refer to Section :ref:`cfg=smpi/coll-selector`
144 - **smpi/auto-shared-malloc-thresh:** :ref:`cfg=smpi/auto-shared-malloc-thresh`
145 - **smpi/async-small-thresh:** :ref:`cfg=smpi/async-small-thresh`
146 - **smpi/buffering:** :ref:`cfg=smpi/buffering`
147 - **smpi/bw-factor:** :ref:`cfg=smpi/bw-factor`
148 - **smpi/coll-selector:** :ref:`cfg=smpi/coll-selector`
149 - **smpi/comp-adjustment-file:** :ref:`cfg=smpi/comp-adjustment-file`
150 - **smpi/cpu-threshold:** :ref:`cfg=smpi/cpu-threshold`
151 - **smpi/display-allocs:** :ref:`cfg=smpi/display-allocs`
152 - **smpi/display-timing:** :ref:`cfg=smpi/display-timing`
153 - **smpi/errors-are-fatal:** :ref:`cfg=smpi/errors-are-fatal`
154 - **smpi/finalization-barrier:** :ref:`cfg=smpi/finalization-barrier`
155 - **smpi/grow-injected-times:** :ref:`cfg=smpi/grow-injected-times`
156 - **smpi/host-speed:** :ref:`cfg=smpi/host-speed`
157 - **smpi/IB-penalty-factors:** :ref:`cfg=smpi/IB-penalty-factors`
158 - **smpi/iprobe:** :ref:`cfg=smpi/iprobe`
159 - **smpi/iprobe-cpu-usage:** :ref:`cfg=smpi/iprobe-cpu-usage`
160 - **smpi/init:** :ref:`cfg=smpi/init`
161 - **smpi/keep-temps:** :ref:`cfg=smpi/keep-temps`
162 - **smpi/lat-factor:** :ref:`cfg=smpi/lat-factor`
163 - **smpi/ois:** :ref:`cfg=smpi/ois`
164 - **smpi/or:** :ref:`cfg=smpi/or`
165 - **smpi/os:** :ref:`cfg=smpi/os`
166 - **smpi/papi-events:** :ref:`cfg=smpi/papi-events`
167 - **smpi/pedantic:** :ref:`cfg=smpi/pedantic`
168 - **smpi/privatization:** :ref:`cfg=smpi/privatization`
169 - **smpi/privatize-libs:** :ref:`cfg=smpi/privatize-libs`
170 - **smpi/send-is-detached-thresh:** :ref:`cfg=smpi/send-is-detached-thresh`
171 - **smpi/shared-malloc:** :ref:`cfg=smpi/shared-malloc`
172 - **smpi/shared-malloc-hugepage:** :ref:`cfg=smpi/shared-malloc-hugepage`
173 - **smpi/simulate-computation:** :ref:`cfg=smpi/simulate-computation`
174 - **smpi/test:** :ref:`cfg=smpi/test`
175 - **smpi/wtime:** :ref:`cfg=smpi/wtime`
176 - **smpi/list-leaks** :ref:`cfg=smpi/list-leaks`
178 - **Tracing configuration options** can be found in Section :ref:`tracing_tracing_options`
180 - **storage/model:** :ref:`options_model_select`
182 - **vm/model:** :ref:`options_model_select`
186 Configuring the Platform Models
187 -------------------------------
189 .. _options_model_select:
191 Choosing the Platform Models
192 ............................
194 SimGrid comes with several network, CPU and disk models built in,
195 and you can change the used model at runtime by changing the passed
196 configuration. The three main configuration items are given below.
197 For each of these items, passing the special ``help`` value gives you
198 a short description of all possible values (for example,
199 ``--cfg=network/model:help`` will present all provided network
200 models). Also, ``--help-models`` should provide information about all
201 models for all existing resources.
203 - ``network/model``: specify the used network model. Possible values:
205 - **LV08 (default one):** Realistic network analytic model
206 (slow-start modeled by multiplying latency by 13.01, bandwidth by
207 .97; bottleneck sharing uses a payload of S=20537 for evaluating
208 RTT). Described in `Accuracy Study and Improvement of Network
209 Simulation in the SimGrid Framework
210 <http://mescal.imag.fr/membres/arnaud.legrand/articles/simutools09.pdf>`_.
211 - **Constant:** Simplistic network model where all communication
212 take a constant time (one second). This model provides the lowest
213 realism, but is (marginally) faster.
214 - **SMPI:** Realistic network model specifically tailored for HPC
215 settings (accurate modeling of slow start with correction factors on
216 three intervals: < 1KiB, < 64 KiB, >= 64 KiB). This model can be
217 :ref:`further configured <options_model_network>`.
218 - **IB:** Realistic network model specifically tailored for HPC
219 settings with InfiniBand networks (accurate modeling contention
220 behavior, based on the model explained in `this PhD work
221 <http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf>`_.
222 This model can be :ref:`further configured <options_model_network>`.
223 - **CM02:** Legacy network analytic model. Very similar to LV08, but
224 without corrective factors. The timings of small messages are thus
225 poorly modeled. This model is described in `A Network Model for
226 Simulation of Grid Application
227 <https://hal.inria.fr/inria-00071989/document>`_.
228 - **ns-3** (only available if you compiled SimGrid accordingly):
229 Use the packet-level network
230 simulators as network models (see :ref:`model_ns3`).
231 This model can be :ref:`further configured <options_pls>`.
233 - ``cpu/model``: specify the used CPU model. We have only one model
236 - **Cas01:** Simplistic CPU model (time=size/speed)
238 - ``host/model``: The host concept is the aggregation of a CPU with a
239 network card. Three models exists, but actually, only 2 of them are
240 interesting. The "compound" one is simply due to the way our
241 internal code is organized, and can easily be ignored. So at the
242 end, you have two host models: The default one allows aggregation of
243 an existing CPU model with an existing network model, but does not
244 allow parallel tasks because these beasts need some collaboration
245 between the network and CPU model.
247 - **default:** Default host model. Currently, CPU:Cas01 and
248 network:LV08 (with cross traffic enabled)
249 - **compound:** Host model that is automatically chosen if
250 you change the network and CPU models
251 - **ptask_L07:** Host model somehow similar to Cas01+CM02 but
252 allowing "parallel tasks", that are intended to model the moldable
253 tasks of the grid scheduling literature.
255 - ``storage/model``: specify the used storage model. Only one model is
257 - ``vm/model``: specify the model for virtual machines. Only one model
260 .. todo: make 'compound' the default host model.
262 .. _options_model_solver:
267 The different models rely on a linear inequalities solver to share
268 the underlying resources. SimGrid allows you to change the solver, but
269 be cautious, **don't change it unless you are 100% sure**.
271 - items ``cpu/solver``, ``network/solver``, ``disk/solver`` and ``host/solver``
272 allow you to change the solver for each model:
274 - **maxmin:** The default solver for all models except ptask. Provides a
275 max-min fairness allocation.
276 - **fairbottleneck:** The default solver for ptasks. Extends max-min to
277 allow heterogeneous resources.
278 - **bmf:** More realistic solver for heterogeneous resource sharing.
279 Implements BMF (Bottleneck max fairness) fairness. To be used with
280 parallel tasks instead of fair-bottleneck.
282 .. _options_model_optim:
287 The network and CPU models that are based on linear inequalities solver (that
288 is, all our analytical models) accept specific optimization
291 - items ``network/optim`` and ``cpu/optim`` (both default to 'Lazy'):
293 - **Lazy:** Lazy action management (partial invalidation in lmm +
294 heap in action remaining).
295 - **TI:** Trace integration. Highly optimized mode when using
296 availability traces (only available for the Cas01 CPU model for
298 - **Full:** Full update of remaining and variables. Slow but may be
299 useful when debugging.
301 - items ``network/maxmin-selective-update`` and
302 ``cpu/maxmin-selective-update``: configure whether the underlying
303 should be lazily updated or not. It should have no impact on the
304 computed timings, but should speed up the computation. |br| It is
305 still possible to disable this feature because it can reveal
306 counter-productive in very specific scenarios where the
307 interaction level is high. In particular, if all your
308 communication share a given backbone link, you should disable it:
309 without it, a simple regular loop is used to update each
310 communication. With it, each of them is still updated (because of
311 the dependency induced by the backbone), but through a complicated
312 and slow pattern that follows the actual dependencies.
314 .. _cfg=maxmin/precision:
315 .. _cfg=surf/precision:
320 **Option** ``maxmin/precision`` **Default:** 1e-5 (in flops or bytes) |br|
321 **Option** ``surf/precision`` **Default:** 1e-9 (in seconds) |br|
322 **Option** ``bmf/precision`` **Default:** 1e-12 (no unit)
324 The analytical models handle a lot of floating point values. It is
325 possible to change the epsilon used to update and compare them through
326 this configuration item. Changing it may speedup the simulation by
327 discarding very small actions, at the price of a reduced numerical
328 precision. You can modify separately the precision used to manipulate
329 timings (in seconds) and the one used to manipulate amounts of work
332 .. _cfg=maxmin/concurrency-limit:
337 **Option** ``maxmin/concurrency-limit`` **Default:** -1 (no limit)
339 The maximum number of variables per resource can be tuned through this
340 option. You can have as many simultaneous actions per resources as you
341 want. If your simulation presents a very high level of concurrency, it
342 may help to use e.g. 100 as a value here. It means that at most 100
343 actions can consume a resource at a given time. The extraneous actions
344 are queued and wait until the amount of concurrency of the considered
345 resource lowers under the given boundary.
347 Such limitations help both to the simulation speed and simulation accuracy
348 on highly constrained scenarios, but the simulation speed suffers of this
349 setting on regular (less constrained) scenarios so it is off by default.
351 .. _options_model_network:
353 Configuring the Network Model
354 .............................
356 .. _cfg=network/TCP-gamma:
358 Maximal TCP Window Size
359 ^^^^^^^^^^^^^^^^^^^^^^^
361 **Option** ``network/TCP-gamma`` **Default:** 4194304
363 The analytical models need to know the maximal TCP window size to take
364 the TCP congestion mechanism into account. On Linux, this value can
365 be retrieved using the following commands. Both give a set of values,
366 and you should use the last one, which is the maximal size.
368 .. code-block:: console
370 $ cat /proc/sys/net/ipv4/tcp_rmem # gives the sender window
371 $ cat /proc/sys/net/ipv4/tcp_wmem # gives the receiver window
373 .. _cfg=network/bandwidth-factor:
374 .. _cfg=network/latency-factor:
375 .. _cfg=network/weight-S:
377 Correcting Important Network Parameters
378 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
380 SimGrid can take network irregularities such as a slow startup or
381 changing behavior depending on the message size into account. You
382 should not change these values unless you really know what you're
383 doing. The corresponding values were computed through data fitting
384 one the timings of packet-level simulators, as described in `Accuracy
385 Study and Improvement of Network Simulation in the SimGrid Framework
386 <http://mescal.imag.fr/membres/arnaud.legrand/articles/simutools09.pdf>`_.
388 - **network/latency-factor**: apply a multiplier to latency.
389 Models the TCP slow-start mechanism.
390 - **network/bandwidth-factor**: actual bandwidth perceived by the
392 - **network/weight-S**: bottleneck sharing constant parameter. Used
395 These parameters are the same for all communications in your simulation,
396 independently of message size or source/destination hosts. A more flexible
397 mechanism based on callbacks was introduced in SimGrid. It provides the user
398 a callback that will be called for each communication, allowing the user
399 to set different latency and bandwidth factors, based on the message size, links used
400 or zones traversed. To more details of how to use it, please look at the
401 `examples/cpp/network-factors/s4u-network-factors.cpp <https://framagit.org/simgrid/simgrid/tree/master/examples/cpp/network-factors/s4u-network-factors.cpp>`_.
404 If you are using the SMPI model, these correction coefficients are
405 themselves corrected by constant values depending on the size of the
406 exchange. By default SMPI uses factors computed on the Stampede
407 Supercomputer at TACC, with optimal deployment of processes on
408 nodes. Again, only hardcore experts should bother about this fact.
409 For more details, see SMPI sections about :ref:`cfg=smpi/bw-factor` and :ref:`cfg=smpi/lat-factor`.
412 .. _cfg=smpi/IB-penalty-factors:
417 InfiniBand network behavior can be modeled through 3 parameters
418 ``smpi/IB-penalty-factors:"βe;βs;γs"``, as explained in `this PhD
420 <http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf>`_ (in French)
421 or more concisely in `this paper <https://hal.inria.fr/hal-00953618/document>`_,
422 even if that paper does only describe models for myrinet and ethernet.
423 You can see in Fig 2 some results for Infiniband, for example. This model
424 may be outdated by now for modern infiniband, anyway, so a new
425 validation would be good.
427 The three paramaters are defined as follows:
429 - βs: penalty factor for outgoing messages, computed by running a simple send to
430 two nodes and checking slowdown compared to a single send to one node,
432 - βe: penalty factor for ingoing messages, same computation method but with one
433 node receiving several messages
434 - γr: slowdown factor when communication buffer memory is saturated. It needs a
435 more complicated pattern to run in order to be computed (5.3 in the thesis,
436 page 107), and formula in the end is γr = time(c)/(3×βe×time(ref)), where
437 time(ref) is the time of a single comm with no contention).
439 Once these values are computed, a penalty is assessed for each message (this is
440 the part implemented in the simulator) as shown page 106 of the thesis. Here is
441 a simple translation of this text. First, some notations:
443 - ∆e(e) which corresponds to the incoming degree of node e, that is to say the number of communications having as destination node e.
444 - ∆s (s) which corresponds to the degree outgoing from node s, that is to say the number of communications sent by node s.
445 - Φ (e) which corresponds to the number of communications destined for the node e but coming from a different node.
446 - 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
448 To determine the penalty for a communication, two values need to be calculated. First, the penalty caused by the conflict in transmission, noted ps.
451 - if ∆s (i) = 1 then ps = 1.
452 - if ∆s (i) ≥ 2 and ∆e (i) ≥ 3 then ps = ∆s (i) × βs × γr
453 - else, ps = ∆s (i) × βs
456 Then, the penalty caused by the conflict in reception (noted pe) should be computed as follows:
458 - if ∆e (i) = 1 then pe = 1
459 - else, pe = Φ (e) × βe × Ω (s, e)
461 Finally, the penalty associated with the communication is:
462 p = max (ps ∈ s, pe)
464 .. _cfg=network/crosstraffic:
466 Simulating Cross-Traffic
467 ^^^^^^^^^^^^^^^^^^^^^^^^
469 Since SimGrid v3.7, cross-traffic effects can be taken into account in
470 analytical simulations. It means that ongoing and incoming
471 communication flows are treated independently. In addition, the LV08
472 model adds 0.05 of usage on the opposite direction for each new
473 created flow. This can be useful to simulate some important TCP
474 phenomena such as ack compression.
476 For that to work, your platform must have two links for each
477 pair of interconnected hosts. An example of usable platform is
478 available in ``examples/platforms/crosstraffic.xml``.
480 This is activated through the ``network/crosstraffic`` item, that
481 can be set to 0 (disable this feature) or 1 (enable it).
483 Note that with the default host model this option is activated by default.
485 .. _cfg=network/loopback:
487 Configuring loopback link
488 ^^^^^^^^^^^^^^^^^^^^^^^^^
490 Several network model provide an implicit loopback link to account for local
491 communication on a host. By default it has a 10GBps bandwidth and a null latency.
492 This can be changed with ``network/loopback-lat`` and ``network/loopback-bw``
495 .. _cfg=smpi/async-small-thresh:
497 Simulating Asynchronous Send
498 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
500 (this configuration item is experimental and may change or disappear)
502 It is possible to specify that messages below a certain size (in bytes) will be
503 sent as soon as the call to MPI_Send is issued, without waiting for
504 the correspondent receive. This threshold can be configured through
505 the ``smpi/async-small-thresh`` item. The default value is 0. This
506 behavior can also be manually set for mailboxes, by setting the
507 receiving mode of the mailbox with a call to
508 :cpp:func:`MSG_mailbox_set_async`. After this, all messages sent to
509 this mailbox will have this behavior regardless of the message size.
511 This value needs to be smaller than or equals to the threshold set at
512 :ref:`cfg=smpi/send-is-detached-thresh`, because asynchronous messages
513 are meant to be detached as well.
520 **Option** ``ns3/TcpModel`` **Default:** "default" (ns-3 default)
522 When using ns-3, there is an extra item ``ns3/TcpModel``, corresponding
523 to the ``ns3::TcpL4Protocol::SocketType`` configuration item in
524 ns-3. The only valid values (enforced on the SimGrid side) are
525 'default' (no change to the ns-3 configuration), 'NewReno' or 'Reno' or
528 **Option** ``ns3/seed`` **Default:** "" (don't set the seed in ns-3)
530 This option is the random seed to provide to ns-3 with
531 ``ns3::RngSeedManager::SetSeed`` and ``ns3::RngSeedManager::SetRun``.
533 If left blank, no seed is set in ns-3. If the value 'time' is
534 provided, the current amount of seconds since epoch is used as a seed.
535 Otherwise, the provided value must be a number to use as a seed.
537 Configuring the Storage model
538 .............................
540 .. _cfg=storage/max_file_descriptors:
542 File Descriptor Count per Host
543 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
545 **Option** ``storage/max_file_descriptors`` **Default:** 1024
547 Each host maintains a fixed-size array of its file descriptors. You
548 can change its size through this item to either enlarge it if your
549 application requires it or to reduce it to save memory space.
556 SimGrid plugins allow one to extend the framework without changing its
557 source code directly. Read the source code of the existing plugins to
558 learn how to do so (in ``src/plugins``), and ask your questions to the
559 usual channels (Stack Overflow, Mailing list, IRC). The basic idea is
560 that plugins usually register callbacks to some signals of interest.
561 If they need to store some information about a given object (Link, CPU
562 or Actor), they do so through the use of a dedicated object extension.
564 Some of the existing plugins can be activated from the command line,
565 meaning that you can activate them from the command line without any
566 modification to your simulation code. For example, you can activate
567 the host energy plugin by adding ``--cfg=plugin:host_energy`` to your
570 Here is a partial list of plugins that can be activated this way. You can get
571 the full list by passing ``--cfg=plugin:help`` to your simulator.
573 - :ref:`Host Energy <plugin_host_energy>`: models the energy dissipation of the compute units.
574 - :ref:`Link Energy <plugin_link_energy>`: models the energy dissipation of the network.
575 - :ref:`Host Load <plugin_host_load>`: monitors the load of the compute units.
577 .. _options_modelchecking:
579 Configuring the Model-Checking
580 ------------------------------
582 To enable SimGrid's model-checking support, the program should
583 be executed using the simgrid-mc wrapper:
585 .. code-block:: console
587 $ simgrid-mc ./my_program
589 Safety properties are expressed as assertions using the function
590 :cpp:func:`void MC_assert(int prop)`.
592 .. _cfg=smpi/buffering:
594 Specifying the MPI buffering behavior
595 .....................................
597 **Option** ``smpi/buffering`` **Default:** infty
599 Buffering in MPI has a huge impact on the communication semantic. For example,
600 standard blocking sends are synchronous calls when the system buffers are full
601 while these calls can complete immediately without even requiring a matching
602 receive call for small messages sent when the system buffers are empty.
604 In SMPI, this depends on the message size, that is compared against two thresholds:
606 - if (size < :ref:`smpi/async-small-thresh <cfg=smpi/async-small-thresh>`) then
607 MPI_Send returns immediately, even if the corresponding receive has not be issued yet.
608 - 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
609 MPI_Send returns as soon as the corresponding receive has been issued. This is known as the eager mode.
610 - if (:ref:`smpi/send-is-detached-thresh <cfg=smpi/send-is-detached-thresh>` < size) then
611 MPI_Send returns only when the message has actually been sent over the network. This is known as the rendez-vous mode.
613 The ``smpi/buffering`` (only valid with MC) option gives an easier interface to choose between these semantics. It can take two values:
615 - **zero:** means that buffering should be disabled. All communications are actually blocking.
616 - **infty:** means that buffering should be made infinite. All communications are non-blocking.
618 .. _cfg=model-check/property:
620 Specifying a liveness property
621 ..............................
623 **Option** ``model-check/property`` **Default:** unset
625 If you want to specify liveness properties, you have to pass them on
626 the command line, specifying the name of the file containing the
627 property, as formatted by the `ltl2ba <https://github.com/utwente-fmt/ltl2ba>`_ program.
628 Note that ltl2ba is not part of SimGrid and must be installed separately.
630 .. code-block:: console
632 $ simgrid-mc ./my_program --cfg=model-check/property:<filename>
634 .. _cfg=model-check/checkpoint:
636 Going for Stateful Verification
637 ...............................
639 By default, the system is backtracked to its initial state to explore
640 another path, instead of backtracking to the exact step before the fork
641 that we want to explore (this is called stateless verification). This
642 is done this way because saving intermediate states can rapidly
643 exhaust the available memory. If you want, you can change the value of
644 the ``model-check/checkpoint`` item. For example,
645 ``--cfg=model-check/checkpoint:1`` asks to take a checkpoint every
646 step. Beware, this will certainly explode your memory. Larger values
647 are probably better, make sure to experiment a bit to find the right
648 setting for your specific system.
650 .. _cfg=model-check/reduction:
652 Specifying the kind of reduction
653 ................................
655 The main issue when using the model-checking is the state space
656 explosion. You can activate some reduction technique with
657 ``--cfg=model-check/reduction:<technique>``. For now, this
658 configuration variable can take 2 values:
660 - **none:** Do not apply any kind of reduction (mandatory for
661 liveness properties, as our current DPOR algorithm breaks cycles)
662 - **dpor:** Apply Dynamic Partial Ordering Reduction. Only valid if
663 you verify local safety properties (default value for safety
666 Another way to mitigate the state space explosion is to search for
667 cycles in the exploration with the :ref:`cfg=model-check/visited`
668 configuration. Note that DPOR and state-equality reduction may not
669 play well together. You should choose between them.
671 Our current DPOR implementation could be improved in may ways. We are
672 currently improving its efficiency (both in term of reduction ability
673 and computational speed), and future work could make it compatible
674 with liveness properties.
676 .. _cfg=model-check/visited:
678 Size of Cycle Detection Set (state equality reduction)
679 ......................................................
681 Mc SimGrid can be asked to search for cycles during the exploration,
682 i.e. situations where a new explored state is in fact the same state
683 than a previous one.. This can prove useful to mitigate the state
684 space explosion with safety properties, and this is the crux when
685 searching for counter-examples to the liveness properties.
687 Note that this feature may break the current implementation of the
688 DPOR reduction technique.
690 The ``model-check/visited`` item is the maximum number of states, which
691 are stored in memory. If the maximum number of snapshotted state is
692 reached, some states will be removed from the memory and some cycles
693 might be missed. Small values can lead to incorrect verifications, but
694 large values can exhaust your memory and be CPU intensive as each new
695 state must be compared to that amount of older saved states.
697 The default settings depend on the kind of exploration. With safety
698 checking, no state is snapshotted and cycles cannot be detected. With
699 liveness checking, all states are snapshotted because missing a cycle
700 could hinder the exploration soundness.
702 .. _cfg=model-check/termination:
704 Non-Termination Detection
705 .........................
707 The ``model-check/termination`` configuration item can be used to
708 report if a non-termination execution path has been found. This is a
709 path with a cycle, which means that the program might never terminate.
711 This only works in safety mode, not in liveness mode.
713 This options is disabled by default.
715 .. _cfg=model-check/dot-output:
720 If set, the ``model-check/dot-output`` configuration item is the name
721 of a file in which to write a dot file of the path leading to the
722 property violation discovered (safety or liveness violation), as well
723 as the cycle for liveness properties. This dot file can then be fed to the
724 graphviz dot tool to generate a corresponding graphical representation.
726 .. _cfg=model-check/max-depth:
728 Exploration Depth Limit
729 .......................
731 The ``model-check/max-depth`` can set the maximum depth of the
732 exploration graph of the model checker. If this limit is reached, a
733 logging message is sent and the results might not be exact.
735 By default, the exploration is limited to the depth of 1000.
737 .. _cfg=model-check/timeout:
742 By default, the model checker does not handle timeout conditions: the `wait`
743 operations never time out. With the ``model-check/timeout`` configuration item
744 set to **yes**, the model checker will explore timeouts of `wait` operations.
746 .. _cfg=model-check/communications-determinism:
747 .. _cfg=model-check/send-determinism:
749 Communication Determinism
750 .........................
752 The ``model-check/communications-determinism`` and
753 ``model-check/send-determinism`` items can be used to select the
754 communication determinism mode of the model checker, which checks
755 determinism properties of the communications of an application.
759 Verification Performance Considerations
760 .......................................
762 The size of the stacks can have a huge impact on the memory
763 consumption when using model-checking. By default, each snapshot will
764 save a copy of the whole stacks and not only of the part that is
765 really meaningful: you should expect the contribution of the memory
766 consumption of the snapshots to be:
767 :math:`\text{number of processes} \times \text{stack size} \times \text{number of states}`.
769 When compiled against the model checker, the stacks are not
770 protected with guards: if the stack size is too small for your
771 application, the stack will silently overflow into other parts of the
772 memory (see :ref:`contexts/guard-size <cfg=contexts/guard-size>`).
774 .. _cfg=model-check/replay:
776 Replaying buggy execution paths from the model checker
777 ......................................................
779 Debugging the problems reported by the model checker is challenging:
780 First, the application under verification cannot be debugged with gdb
781 because the model checker already traces it. Then, the model checker may
782 explore several execution paths before encountering the issue, making it
783 very difficult to understand the output. Fortunately, SimGrid provides
784 the execution path leading to any reported issue so that you can replay
785 this path reported by the model checker, enabling the usage of classical
788 When the model checker finds an interesting path in the application
789 execution graph (where a safety or liveness property is violated), it
790 generates an identifier for this path. Here is an example of the output:
792 .. code-block:: console
794 [ 0.000000] (0:@) Check a safety property
795 [ 0.000000] (0:@) **************************
796 [ 0.000000] (0:@) *** PROPERTY NOT VALID ***
797 [ 0.000000] (0:@) **************************
798 [ 0.000000] (0:@) Counter-example execution trace:
799 [ 0.000000] (0:@) [(1)Tremblay (app)] MC_RANDOM(3)
800 [ 0.000000] (0:@) [(1)Tremblay (app)] MC_RANDOM(4)
801 [ 0.000000] (0:@) Path = 1/3;1/4
802 [ 0.000000] (0:@) Expanded states = 27
803 [ 0.000000] (0:@) Visited states = 68
804 [ 0.000000] (0:@) Executed transitions = 46
806 The interesting line is ``Path = 1/3;1/4``, which means that you should use
807 ``--cfg=model-check/replay:1/3;1/4`` to replay your application on the buggy
808 execution path. All options (but the model checker related ones) must
809 remain the same. In particular, if you ran your application with
810 ``smpirun -wrapper simgrid-mc``, then do it again. Remove all
811 MC-related options, keep non-MC-related ones and add
812 ``--cfg=model-check/replay:???``.
814 Currently, if the path is of the form ``X;Y;Z``, each number denotes
815 the actor's pid that is selected at each indecision point. If it's of
816 the form ``X/a;Y/b``, the X and Y are the selected pids while the a
817 and b are the return values of their simcalls. In the previous
818 example, ``1/3;1/4``, you can see from the full output that the actor
819 1 is doing MC_RANDOM simcalls, so the 3 and 4 simply denote the values
820 that these simcall return on the execution branch leading to the
823 Configuring the User Code Virtualization
824 ----------------------------------------
826 .. _cfg=contexts/factory:
828 Selecting the Virtualization Factory
829 ....................................
831 **Option** contexts/factory **Default:** "raw"
833 In SimGrid, the user code is virtualized in a specific mechanism that
834 allows the simulation kernel to control its execution: when a user
835 process requires a blocking action (such as sending a message), it is
836 interrupted, and only gets released when the simulated clock reaches
837 the point where the blocking operation is done. This is explained
838 graphically in the `relevant tutorial, available online
839 <https://simgrid.org/tutorials/simgrid-simix-101.pdf>`_.
841 In SimGrid, the containers in which user processes are virtualized are
842 called contexts. Several context factory are provided, and you can
843 select the one you want to use with the ``contexts/factory``
844 configuration item. Some of the following may not exist on your
845 machine because of portability issues. In any case, the default one
846 should be the most effcient one (please report bugs if the
847 auto-detection fails for you). They are approximately sorted here from
848 the slowest to the most efficient:
850 - **thread:** very slow factory using full featured threads (either
851 pthreads or windows native threads). They are slow but very
852 standard. Some debuggers or profilers only work with this factory.
853 - **java:** Java applications are virtualized onto java threads (that
854 are regular pthreads registered to the JVM)
855 - **ucontext:** fast factory using System V contexts (Linux and FreeBSD only)
856 - **boost:** This uses the `context
857 implementation <http://www.boost.org/doc/libs/1_59_0/libs/context/doc/html/index.html>`_
858 of the boost library for a performance that is comparable to our
860 |br| Install the relevant library (e.g. with the
861 libboost-contexts-dev package on Debian/Ubuntu) and recompile
863 - **raw:** amazingly fast factory using a context switching mechanism
864 of our own, directly implemented in assembly (only available for x86
865 and amd64 platforms for now) and without any unneeded system call.
867 The main reason to change this setting is when the debugging tools become
868 fooled by the optimized context factories. Threads are the most
869 debugging-friendly contexts, as they allow one to set breakpoints
870 anywhere with gdb and visualize backtraces for all processes, in order
871 to debug concurrency issues. Valgrind is also more comfortable with
872 threads, but it should be usable with all factories (Exception: the
873 callgrind tool really dislikes raw and ucontext factories).
875 .. _cfg=contexts/stack-size:
877 Adapting the Stack Size
878 .......................
880 **Option** ``contexts/stack-size`` **Default:** 8192 KiB
882 Each virtualized used process is executed using a specific system
883 stack. The size of this stack has a huge impact on the simulation
884 scalability, but its default value is rather large. This is because
885 the error messages that you get when the stack size is too small are
886 rather disturbing: this leads to stack overflow (overwriting other
887 stacks), leading to segfaults with corrupted stack traces.
889 If you want to push the scalability limits of your code, you might
890 want to reduce the ``contexts/stack-size`` item. Its default value is
891 8192 (in KiB), while our Chord simulation works with stacks as small
892 as 16 KiB, for example. You can ensure that some actors have a specific
893 size by simply changing the value of this configuration item before
894 creating these actors. The :cpp:func:`simgrid::s4u::Engine::set_config`
895 functions are handy for that.
897 This *setting is ignored* when using the thread factory (because there
898 is no way to modify the stack size with C++ system threads). Instead,
899 you should compile SimGrid and your application with
900 ``-fsplit-stack``. Note that this compilation flag is not compatible
901 with the model checker right now.
903 The operating system should only allocate memory for the pages of the
904 stack which are actually used and you might not need to use this in
905 most cases. However, this setting is very important when using the
906 model checker (see :ref:`options_mc_perf`).
908 .. _cfg=contexts/guard-size:
910 Disabling Stack Guard Pages
911 ...........................
913 **Option** ``contexts/guard-size`` **Default** 1 page in most case (0 pages on Windows or with MC)
915 Unless you use the threads context factory (see
916 :ref:`cfg=contexts/factory`), a stack guard page is usually used
917 which prevents the stack of a given actor from overflowing on another
918 stack. But the performance impact may become prohibitive when the
919 amount of actors increases. The option ``contexts/guard-size`` is the
920 number of stack guard pages used. By setting it to 0, no guard pages
921 will be used: in this case, you should avoid using small stacks (with
922 :ref:`contexts/stack-size <cfg=contexts/stack-size>`) as the stack
923 will silently overflow on other parts of the memory.
925 When no stack guard page is created, stacks may then silently overflow
926 on other parts of the memory if their size is too small for the
929 .. _cfg=contexts/nthreads:
930 .. _cfg=contexts/synchro:
932 Running User Code in Parallel
933 .............................
935 Parallel execution of the user code is only considered stable in
936 SimGrid v3.7 and higher, and mostly for MSG simulations. SMPI
937 simulations may well fail in parallel mode. It is described in
938 `INRIA RR-7653 <http://hal.inria.fr/inria-00602216/>`_.
940 If you are using the **ucontext** or **raw** context factories, you can
941 request to execute the user code in parallel. Several threads are
942 launched, each of them handling the same number of user contexts at each
943 run. To activate this, set the ``contexts/nthreads`` item to the amount
944 of cores that you have in your computer (or lower than 1 to have the
945 amount of cores auto-detected).
947 When parallel execution is activated, you can choose the
948 synchronization schema used with the ``contexts/synchro`` item,
949 which value is either:
951 - **futex:** ultra optimized synchronisation schema, based on futexes
952 (fast user-mode mutexes), and thus only available on Linux systems.
953 This is the default mode when available.
954 - **posix:** slow but portable synchronisation using only POSIX
956 - **busy_wait:** not really a synchronisation: the worker threads
957 constantly request new contexts to execute. It should be the most
958 efficient synchronisation schema, but it loads all the cores of
959 your machine for no good reason. You probably prefer the other less
962 Configuring the Tracing
963 -----------------------
965 The :ref:`tracing subsystem <outcome_vizu>` can be configured in
966 several different ways depending on the used interface (S4U, SMPI)
967 and the kind of traces that needs to be obtained. See the
968 :ref:`Tracing Configuration Options subsection
969 <tracing_tracing_options>` for a full description of each
970 configuration option.
972 We detail here a simple way to get the traces working for you, even if
973 you never used the tracing API.
976 - Any SimGrid-based simulator (MSG, SMPI, ...) and raw traces:
980 --cfg=tracing:yes --cfg=tracing/uncategorized:yes
982 The first parameter activates the tracing subsystem, and the second
983 tells it to trace host and link utilization (without any
986 - MSG-based simulator and categorized traces (you need to
987 declare categories and classify your tasks according to them)
991 --cfg=tracing:yes --cfg=tracing/categorized:yes
993 The first parameter activates the tracing subsystem, and the second
994 tells it to trace host and link categorized utilization.
996 - SMPI simulator and traces for a space/time view:
998 .. code-block:: console
1000 $ smpirun -trace ...
1002 The `-trace` parameter for the smpirun script runs the simulation
1003 with ``--cfg=tracing:yes --cfg=tracing/smpi:yes``. Check the
1004 smpirun's `-help` parameter for additional tracing options.
1006 Sometimes you might want to put additional information on the trace to
1007 correctly identify them later, or to provide data that can be used to
1008 reproduce an experiment. You have two ways to do that:
1010 - Add a string on top of the trace file as comment:
1012 .. code-block:: none
1014 --cfg=tracing/comment:my_simulation_identifier
1016 - Add the contents of a textual file on top of the trace file as comment:
1018 .. code-block:: none
1020 --cfg=tracing/comment-file:my_file_with_additional_information.txt
1022 Please, use these two parameters (for comments) to make reproducible
1023 simulations. For additional details about this and all tracing
1024 options, check See the :ref:`tracing_tracing_options`.
1029 .. _cfg=msg/debug-multiple-use:
1034 **Option** ``msg/debug-multiple-use`` **Default:** off
1036 Sometimes your application may try to send a task that is still being
1037 executed somewhere else, making it impossible to send this task. However,
1038 for debugging purposes, one may want to know what the other host is/was
1039 doing. This option shows a backtrace of the other process.
1044 The SMPI interface provides several specific configuration items.
1045 These are not easy to see, since the code is usually launched through the
1046 ``smiprun`` script directly.
1048 .. _cfg=smpi/host-speed:
1049 .. _cfg=smpi/cpu-threshold:
1050 .. _cfg=smpi/simulate-computation:
1052 Automatic Benchmarking of SMPI Code
1053 ...................................
1055 In SMPI, the sequential code is automatically benchmarked, and these
1056 computations are automatically reported to the simulator. That is to
1057 say that if you have a large computation between a ``MPI_Recv()`` and
1058 a ``MPI_Send()``, SMPI will automatically benchmark the duration of
1059 this code, and create an execution task within the simulator to take
1060 this into account. For that, the actual duration is measured on the
1061 host machine and then scaled to the power of the corresponding
1062 simulated machine. The variable ``smpi/host-speed`` allows one to
1063 specify the computational speed of the host machine (in flop/s by
1064 default) to use when scaling the execution times.
1066 The default value is ``smpi/host-speed=20kf`` (= 20,000 flop/s). This
1067 is probably underestimated for most machines, leading SimGrid to
1068 overestimate the amount of flops in the execution blocks that are
1069 automatically injected in the simulator. As a result, the execution
1070 time of the whole application will probably be overestimated until you
1071 use a realistic value.
1073 When the code consists of numerous consecutive MPI calls, the
1074 previous mechanism feeds the simulation kernel with numerous tiny
1075 computations. The ``smpi/cpu-threshold`` item becomes handy when this
1076 impacts badly on the simulation performance. It specifies a threshold (in
1077 seconds) below which the execution chunks are not reported to the
1078 simulation kernel (default value: 1e-6).
1080 .. note:: The option ``smpi/cpu-threshold`` ignores any computation
1081 time spent below this threshold. SMPI does not consider the
1082 `amount of time` of these computations; there is no offset for
1083 this. Hence, a value that is too small, may lead to unreliable
1086 In some cases, however, one may wish to disable simulation of
1087 the computation of an application. This is the case when SMPI is used not to
1088 simulate an MPI application, but instead an MPI code that performs
1089 "live replay" of another MPI app (e.g., ScalaTrace's replay tool, or
1090 various on-line simulators that run an app at scale). In this case the
1091 computation of the replay/simulation logic should not be simulated by
1092 SMPI. Instead, the replay tool or on-line simulator will issue
1093 "computation events", which correspond to the actual MPI simulation
1094 being replayed/simulated. At the moment, these computation events can
1095 be simulated using SMPI by calling internal smpi_execute*() functions.
1097 To disable the benchmarking/simulation of a computation in the simulated
1098 application, the variable ``smpi/simulate-computation`` should be set
1099 to **no**. This option just ignores the timings in your simulation; it
1100 still executes the computations itself. If you want to stop SMPI from
1101 doing that, you should check the SMPI_SAMPLE macros, documented in
1102 Section :ref:`SMPI_use_faster`.
1104 +------------------------------------+-------------------------+-----------------------------+
1105 | Solution | Computations executed? | Computations simulated? |
1106 +====================================+=========================+=============================+
1107 | --cfg=smpi/simulate-computation:no | Yes | Never |
1108 +------------------------------------+-------------------------+-----------------------------+
1109 | --cfg=smpi/cpu-threshold:42 | Yes, in all cases | If it lasts over 42 seconds |
1110 +------------------------------------+-------------------------+-----------------------------+
1111 | SMPI_SAMPLE() macro | Only once per loop nest | Always |
1112 +------------------------------------+-------------------------+-----------------------------+
1114 .. _cfg=smpi/comp-adjustment-file:
1116 Slow-down or speed-up parts of your code
1117 ........................................
1119 **Option** ``smpi/comp-adjustment-file:`` **Default:** unset
1121 This option allows you to pass a file that contains two columns: The
1122 first column defines the section that will be subject to a speedup;
1123 the second column is the speedup. For instance:
1125 .. code-block:: none
1127 "start:stop","ratio"
1128 "exchange_1.f:30:exchange_1.f:130",1.18244559422142
1130 The first line is the header - you must include it. The following
1131 line means that the code between two consecutive MPI calls on line 30
1132 in exchange_1.f and line 130 in exchange_1.f should receive a speedup
1133 of 1.18244559422142. The value for the second column is therefore a
1134 speedup, if it is larger than 1 and a slowdown if it is smaller
1135 than 1. Nothing will be changed if it is equal to 1.
1137 Of course, you can set any arbitrary filenames you want (so the start
1138 and end don't have to be in the same file), but be aware that this
1139 mechanism only supports `consecutive calls!`
1141 Please note that you must pass the ``-trace-call-location`` flag to
1142 smpicc or smpiff, respectively. This flag activates some internal
1143 macro definitions that help with obtaining the call location.
1145 .. _cfg=smpi/bw-factor:
1150 **Option** ``smpi/bw-factor``
1151 |br| **Default:** 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
1153 The possible throughput of network links is often dependent on the
1154 message sizes, as protocols may adapt to different message sizes. With
1155 this option, a series of message sizes and factors are given, helping
1156 the simulation to be more realistic. For instance, the current default
1157 value means that messages with size 65472 bytes and more will get a total of
1158 MAX_BANDWIDTH*0.940694, messages of size 15424 to 65471 will get
1159 MAX_BANDWIDTH*0.697866, and so on (where MAX_BANDWIDTH denotes the
1160 bandwidth of the link).
1162 An experimental script to compute these factors is available online. See
1163 https://framagit.org/simgrid/platform-calibration/
1164 https://simgrid.org/contrib/smpi-saturation-doc.html
1166 .. _cfg=smpi/display-timing:
1168 Reporting Simulation Time
1169 .........................
1171 **Option** ``smpi/display-timing`` **Default:** 0 (false)
1173 Most of the time, you run MPI code with SMPI to compute the time it
1174 would take to run it on a platform. But since the code is run through
1175 the ``smpirun`` script, you don't have any control on the launcher
1176 code, making it difficult to report the simulated time when the
1177 simulation ends. If you enable the ``smpi/display-timing`` item,
1178 ``smpirun`` will display this information when the simulation
1180 SMPI will also display information about the amout of real time spent
1181 in application code and in SMPI internals, to provide hints about the
1182 need to use sampling to reduce simulation time.
1184 .. _cfg=smpi/display-allocs:
1186 Reporting memory allocations
1187 ............................
1189 **Option** ``smpi/display-allocs`` **Default:** 0 (false)
1191 SMPI intercepts malloc and calloc calls performed inside the running
1192 application, if it wasn't compiled with SMPI_NO_OVERRIDE_MALLOC.
1193 With this option, SMPI will show at the end of execution the amount of
1194 memory allocated through these calls, and locate the most expensive one.
1195 This helps finding the targets for manual memory sharing, or the threshold
1196 to use for smpi/auto-shared-malloc-thresh option (see :ref:`cfg=smpi/auto-shared-malloc-thresh`).
1198 .. _cfg=smpi/keep-temps:
1200 Keeping temporary files after simulation
1201 ........................................
1203 **Option** ``smpi/keep-temps`` **default:** 0 (false)
1205 SMPI usually generates a lot of temporary files that are cleaned after
1206 use. This option requests to preserve them, for example to debug or
1207 profile your code. Indeed, the binary files are removed very early
1208 under the dlopen privatization schema, which tends to fool the
1211 .. _cfg=smpi/lat-factor:
1216 **Option** ``smpi/lat-factor`` |br|
1217 **default:** 65472:11.6436;15424:3.48845;9376:2.59299;5776:2.18796;3484:1.88101;1426:1.61075;732:1.9503;257:1.95341;0:2.01467
1219 The motivation and syntax for this option is identical to the motivation/syntax
1220 of :ref:`cfg=smpi/bw-factor`.
1222 There is an important difference, though: While smpi/bw-factor `reduces` the
1223 actual bandwidth (i.e., values between 0 and 1 are valid), latency factors
1224 increase the latency, i.e., values larger than or equal to 1 are valid here.
1226 .. _cfg=smpi/papi-events:
1228 Trace hardware counters with PAPI
1229 .................................
1231 **Option** ``smpi/papi-events`` **default:** unset
1233 When the PAPI support is compiled into SimGrid, this option takes the
1234 names of PAPI counters and adds their respective values to the trace
1235 files (See Section :ref:`tracing_tracing_options`).
1239 This feature currently requires superuser privileges, as registers
1240 are queried. Only use this feature with code you trust! Call
1241 smpirun for instance via ``smpirun -wrapper "sudo "
1242 <your-parameters>`` or run ``sudo sh -c "echo 0 >
1243 /proc/sys/kernel/perf_event_paranoid"`` In the later case, sudo
1244 will not be required.
1246 It is planned to make this feature available on a per-process (or per-thread?) basis.
1247 The first draft, however, just implements a "global" (i.e., for all processes) set
1248 of counters, the "default" set.
1250 .. code-block:: none
1252 --cfg=smpi/papi-events:"default:PAPI_L3_LDM:PAPI_L2_LDM"
1254 .. _cfg=smpi/privatization:
1256 Automatic Privatization of Global Variables
1257 ...........................................
1259 **Option** ``smpi/privatization`` **default:** "dlopen" (when using smpirun)
1261 MPI executables are usually meant to be executed in separate
1262 processes, but SMPI is executed in only one process. Global variables
1263 from executables will be placed in the same memory region and shared
1264 between processes, causing intricate bugs. Several options are
1265 possible to avoid this, as described in the main `SMPI publication
1266 <https://hal.inria.fr/hal-01415484>`_ and in the :ref:`SMPI
1267 documentation <SMPI_what_globals>`. SimGrid provides two ways of
1268 automatically privatizing the globals, and this option allows one to
1269 choose between them.
1271 - **no** (default when not using smpirun): Do not automatically
1272 privatize variables. Pass ``-no-privatize`` to smpirun to disable
1274 - **dlopen** or **yes** (default when using smpirun): Link multiple
1275 times against the binary.
1276 - **mmap** (slower, but maybe somewhat more stable):
1277 Runtime automatic switching of the data segments.
1280 This configuration option cannot be set in your platform file. You can only
1281 pass it as an argument to smpirun.
1283 .. _cfg=smpi/privatize-libs:
1285 Automatic privatization of global variables inside external libraries
1286 .....................................................................
1288 **Option** ``smpi/privatize-libs`` **default:** unset
1290 **Linux/BSD only:** When using dlopen (default) privatization,
1291 privatize specific shared libraries with internal global variables, if
1292 they can't be linked statically. For example libgfortran is usually
1293 used for Fortran I/O and indexes in files can be mixed up.
1295 Multiple libraries can be given, semicolon separated.
1297 This configuration option can only use either full paths to libraries,
1298 or full names. Check with ldd the name of the library you want to
1301 .. code-block:: console
1305 libgfortran.so.3 => /usr/lib/x86_64-linux-gnu/libgfortran.so.3 (0x00007fbb4d91b000)
1308 Then you can use ``--cfg=smpi/privatize-libs:libgfortran.so.3``
1309 or ``--cfg=smpi/privatize-libs:/usr/lib/x86_64-linux-gnu/libgfortran.so.3``,
1310 but not ``libgfortran`` nor ``libgfortran.so``.
1312 .. _cfg=smpi/send-is-detached-thresh:
1314 Simulating MPI detached send
1315 ............................
1317 **Option** ``smpi/send-is-detached-thresh`` **default:** 65536
1319 This threshold specifies the size in bytes under which the send will
1320 return immediately. This is different from the threshold detailed in
1321 :ref:`cfg=smpi/async-small-thresh` because the message is not
1322 really sent when the send is posted. SMPI still waits for the
1323 corresponding receive to be posted, in order to perform the communication
1326 .. _cfg=smpi/coll-selector:
1328 Simulating MPI collective algorithms
1329 ....................................
1331 **Option** ``smpi/coll-selector`` **Possible values:** naive (default), ompi, mpich
1333 SMPI implements more than 100 different algorithms for MPI collective
1334 communication, to accurately simulate the behavior of most of the
1335 existing MPI libraries. The ``smpi/coll-selector`` item can be used to
1336 select the decision logic either of the OpenMPI or the MPICH libraries. (By
1337 default SMPI uses naive version of collective operations.)
1339 Each collective operation can be manually selected with a
1340 ``smpi/collective_name:algo_name``. Available algorithms are listed in
1341 :ref:`SMPI_use_colls`.
1343 .. TODO:: All available collective algorithms will be made available
1344 via the ``smpirun --help-coll`` command.
1346 .. _cfg=smpi/finalization-barrier:
1348 Add a barrier in MPI_Finalize
1349 .............................
1351 **Option** ``smpi/finalization-barrier`` **default:** off
1353 By default, SMPI processes are destroyed as soon as soon as their code ends,
1354 so after a successful MPI_Finalize call returns. In some rare cases, some data
1355 might have been attached to MPI objects still active in the remaining processes,
1356 and can be destroyed eagerly by the finished process.
1357 If your code shows issues at finalization, such as segmentation fault, triggering
1358 this option will add an explicit MPI_Barrier(MPI_COMM_WORLD) call inside the
1359 MPI_Finalize, so that all processes will terminate at almost the same point.
1360 It might affect the total timing by the cost of a barrier.
1362 .. _cfg=smpi/errors-are-fatal:
1364 Disable MPI fatal errors
1365 ........................
1367 **Option** ``smpi/errors-are-fatal`` **default:** on
1369 By default, SMPI processes will crash if a MPI error code is returned. MPI allows
1370 to explicitely set MPI_ERRORS_RETURN errhandler to avoid this behaviour. This flag
1371 will turn on this behaviour by default (for all concerned types and errhandlers).
1372 This can ease debugging by going after the first reported error.
1374 .. _cfg=smpi/pedantic:
1376 Disable pedantic MPI errors
1377 ...........................
1379 **Option** ``smpi/pedantic`` **default:** on
1381 By default, SMPI will report all errors it finds in MPI codes. Some of these errors
1382 may not be considered as errors by all developers. This flag can be turned off to
1383 avoid reporting some usually harmless mistakes.
1384 Concerned errors list (will be expanded in the future):
1386 - Calling MPI_Win_fence only once in a program, hence just opening an epoch without
1389 .. _cfg=smpi/iprobe:
1391 Inject constant times for MPI_Iprobe
1392 ....................................
1394 **Option** ``smpi/iprobe`` **default:** 0.0001
1396 The behavior and motivation for this configuration option is identical
1397 with :ref:`smpi/test <cfg=smpi/test>`, but for the function
1400 .. _cfg=smpi/iprobe-cpu-usage:
1402 Reduce speed for iprobe calls
1403 .............................
1405 **Option** ``smpi/iprobe-cpu-usage`` **default:** 1 (no change)
1407 MPI_Iprobe calls can be heavily used in applications. To account
1408 correctly for the energy that cores spend probing, it is necessary to
1409 reduce the load that these calls cause inside SimGrid.
1411 For instance, we measured a maximum power consumption of 220 W for a
1412 particular application but only 180 W while this application was
1413 probing. Hence, the correct factor that should be passed to this
1414 option would be 180/220 = 0.81.
1418 Inject constant times for MPI_Init
1419 ..................................
1421 **Option** ``smpi/init`` **default:** 0
1423 The behavior and motivation for this configuration option is identical
1424 with :ref:`smpi/test <cfg=smpi/test>`, but for the function ``MPI_Init()``.
1428 Inject constant times for MPI_Isend()
1429 .....................................
1431 **Option** ``smpi/ois``
1433 The behavior and motivation for this configuration option is identical
1434 with :ref:`smpi/os <cfg=smpi/os>`, but for the function ``MPI_Isend()``.
1438 Inject constant times for MPI_send()
1439 ....................................
1441 **Option** ``smpi/os``
1443 In several network models such as LogP, send (MPI_Send, MPI_Isend) and
1444 receive (MPI_Recv) operations incur costs (i.e., they consume CPU
1445 time). SMPI can factor these costs in as well, but the user has to
1446 configure SMPI accordingly as these values may vary by machine. This
1447 can be done by using ``smpi/os`` for MPI_Send operations; for MPI_Isend
1448 and MPI_Recv, use ``smpi/ois`` and ``smpi/or``, respectively. These work
1449 exactly as ``smpi/ois``.
1451 This item can consist of multiple sections; each section takes three
1452 values, for example ``1:3:2;10:5:1``. The sections are divided by ";"
1453 so this example contains two sections. Furthermore, each section
1454 consists of three values.
1456 1. The first value denotes the minimum size in bytes for this section to take effect;
1457 read it as "if message size is greater than this value (and other section has a larger
1458 first value that is also smaller than the message size), use this".
1459 In the first section above, this value is "1".
1461 2. The second value is the startup time; this is a constant value that will always
1462 be charged, no matter what the size of the message. In the first section above,
1465 3. The third value is the `per-byte` cost. That is, it is charged for every
1466 byte of the message (incurring cost messageSize*cost_per_byte)
1467 and hence accounts also for larger messages. In the first
1468 section of the example above, this value is "2".
1470 Now, SMPI always checks which section it should use for a given
1471 message; that is, if a message of size 11 is sent with the
1472 configuration of the example above, only the second section will be
1473 used, not the first, as the first value of the second section is
1474 closer to the message size. Hence, when ``smpi/os=1:3:2;10:5:1``, a
1475 message of size 11 incurs the following cost inside MPI_Send:
1476 ``5+11*1`` because 5 is the startup cost and 1 is the cost per byte.
1478 Note that the order of sections can be arbitrary; they will be ordered internally.
1482 Inject constant times for MPI_Recv()
1483 ....................................
1485 **Option** ``smpi/or``
1487 The behavior and motivation for this configuration option is identical
1488 with :ref:`smpi/os <cfg=smpi/os>`, but for the function ``MPI_Recv()``.
1491 .. _cfg=smpi/grow-injected-times:
1493 Inject constant times for MPI_Test
1494 ..................................
1496 **Option** ``smpi/test`` **default:** 0.0001
1498 By setting this option, you can control the amount of time a process
1499 sleeps when MPI_Test() is called; this is important, because SimGrid
1500 normally only advances the time while communication is happening and
1501 thus, MPI_Test will not add to the time, resulting in deadlock if it is
1502 used as a break-condition as in the following example:
1507 MPI_Test(request, flag, status);
1511 To speed up execution, we use a counter to keep track of how often we
1512 checked if the handle is now valid or not. Hence, we actually
1513 use counter*SLEEP_TIME, that is, the time MPI_Test() causes the
1514 process to sleep increases linearly with the number of previously
1515 failed tests. This behavior can be disabled by setting
1516 ``smpi/grow-injected-times`` to **no**. This will also disable this
1517 behavior for MPI_Iprobe.
1519 .. _cfg=smpi/shared-malloc:
1520 .. _cfg=smpi/shared-malloc-hugepage:
1525 **Option** ``smpi/shared-malloc`` **Possible values:** global (default), local
1527 If your simulation consumes too much memory, you may want to modify
1528 your code so that the working areas are shared by all MPI ranks. For
1529 example, in a block-cyclic matrix multiplication, you will only
1530 allocate one set of blocks, and all processes will share them.
1531 Naturally, this will lead to very wrong results, but this will save a
1532 lot of memory. So this is still desirable for some studies. For more on
1533 the motivation for that feature, please refer to the `relevant section
1534 <https://simgrid.github.io/SMPI_CourseWare/topic_understanding_performance/matrixmultiplication>`_
1535 of the SMPI CourseWare (see Activity #2.2 of the pointed
1536 assignment). In practice, change the calls for malloc() and free() into
1537 SMPI_SHARED_MALLOC() and SMPI_SHARED_FREE().
1539 SMPI provides two algorithms for this feature. The first one, called
1540 ``local``, allocates one block per call to SMPI_SHARED_MALLOC()
1541 (each call site gets its own block) ,and this block is shared
1542 among all MPI ranks. This is implemented with the shm_* functions
1543 to create a new POSIX shared memory object (kept in RAM, in /dev/shm)
1544 for each shared block.
1546 With the ``global`` algorithm, each call to SMPI_SHARED_MALLOC()
1547 returns a new address, but it only points to a shadow block: its memory
1548 area is mapped on a 1 MiB file on disk. If the returned block is of size
1549 N MiB, then the same file is mapped N times to cover the whole block.
1550 At the end, no matter how many times you call SMPI_SHARED_MALLOC, this will
1551 only consume 1 MiB in memory.
1553 You can disable this behavior and come back to regular mallocs (for
1554 example for debugging purposes) using ``no`` as a value.
1556 If you want to keep private some parts of the buffer, for instance if these
1557 parts are used by the application logic and should not be corrupted, you
1558 can use SMPI_PARTIAL_SHARED_MALLOC(size, offsets, offsets_count). For example:
1562 mem = SMPI_PARTIAL_SHARED_MALLOC(500, {27,42 , 100,200}, 2);
1564 This will allocate 500 bytes to mem, such that mem[27..41] and
1565 mem[100..199] are shared while other area remain private.
1567 Then, it can be deallocated by calling SMPI_SHARED_FREE(mem).
1569 When smpi/shared-malloc:global is used, the memory consumption problem
1570 is solved, but it may induce too much load on the kernel's pages table.
1571 In this case, you should use huge pages so that the kernel creates only one
1572 entry per MB of malloced data instead of one entry per 4 kB.
1573 To activate this, you must mount a hugetlbfs on your system and allocate
1574 at least one huge page:
1576 .. code-block:: console
1579 $ sudo mount none /home/huge -t hugetlbfs -o rw,mode=0777
1580 $ sudo sh -c 'echo 1 > /proc/sys/vm/nr_hugepages' # echo more if you need more
1582 Then, you can pass the option
1583 ``--cfg=smpi/shared-malloc-hugepage:/home/huge`` to smpirun to
1584 actually activate the huge page support in shared mallocs.
1586 .. _cfg=smpi/auto-shared-malloc-thresh:
1588 Automatically share allocations
1589 ...............................
1591 **Option** ``smpi/auto-shared-malloc-thresh:`` **Default:** 0 (false)
1592 This value in bytes represents the size above which all allocations
1593 will be "shared" by default (as if they were performed through
1594 SMPI_SHARED_MALLOC macros). Default = 0 = disabled feature.
1595 The value must be carefully chosen to only select data buffers which
1596 will not modify execution path or cause crash if their content is false.
1597 Option :ref:`cfg=smpi/display-allocs` can be used to locate the largest
1598 allocation detected in a run, and provide a good starting threshold.
1599 Note : malloc, calloc and free are overridden by smpicc/cxx by default.
1600 This can cause some troubles if codes are already overriding these. If this
1601 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
1602 help, but will make this feature unusable.
1606 Inject constant times for MPI_Wtime, gettimeofday and clock_gettime
1607 ...................................................................
1609 **Option** ``smpi/wtime`` **default:** 10 ns
1611 This option controls the amount of (simulated) time spent in calls to
1612 MPI_Wtime(), gettimeofday() and clock_gettime(). If you set this value
1613 to 0, the simulated clock is not advanced in these calls, which leads
1614 to issues if your application contains such a loop:
1618 while(MPI_Wtime() < some_time_bound) {
1619 /* some tests, with no communication nor computation */
1622 When the option smpi/wtime is set to 0, the time advances only on
1623 communications and computations. So the previous code results in an
1624 infinite loop: the current [simulated] time will never reach
1625 ``some_time_bound``. This infinite loop is avoided when that option
1626 is set to a small value, as it is by default since SimGrid v3.21.
1628 Note that if your application does not contain any loop depending on
1629 the current time only, then setting this option to a non-zero value
1630 will slow down your simulations by a tiny bit: the simulation loop has
1631 to be broken out of and reset each time your code asks for the current time.
1632 If the simulation speed really matters to you, you can avoid this
1633 extra delay by setting smpi/wtime to 0.
1635 .. _cfg=smpi/list-leaks:
1637 Report leaked MPI objects
1638 .........................
1640 **Option** ``smpi/list-leaks`` **default:** 0
1642 This option controls whether to report leaked MPI objects.
1643 The parameter is the number of leaks to report.
1645 Other Configurations
1646 --------------------
1648 .. _cfg=debug/clean-atexit:
1650 Cleanup at Termination
1651 ......................
1653 **Option** ``debug/clean-atexit`` **default:** on
1655 If your code is segfaulting during its finalization, it may help to
1656 disable this option to request that SimGrid not attempt any cleanups at
1657 the end of the simulation. Since the Unix process is ending anyway,
1658 the operating system will wipe it all.
1665 **Option** ``path`` **default:** . (current dir)
1667 It is possible to specify a list of directories to search in for the
1668 trace files (see :ref:`pf_trace`) by using this configuration
1669 item. To add several directory to the path, set the configuration
1670 item several times, as in ``--cfg=path:toto --cfg=path:tutu``
1672 .. _cfg=debug/breakpoint:
1677 **Option** ``debug/breakpoint`` **default:** unset
1679 This configuration option sets a breakpoint: when the simulated clock
1680 reaches the given time, a SIGTRAP is raised. This can be used to stop
1681 the execution and get a backtrace with a debugger.
1683 It is also possible to set the breakpoint from inside the debugger, by
1684 writing in global variable simgrid::kernel::cfg_breakpoint. For example,
1687 .. code-block:: none
1689 set variable simgrid::kernel::cfg_breakpoint = 3.1416
1691 .. _cfg=debug/verbose-exit:
1696 **Option** ``debug/verbose-exit`` **default:** on
1698 By default, when Ctrl-C is pressed, the status of all existing actors
1699 is displayed before exiting the simulation. This is very useful to
1700 debug your code, but it can become troublesome if you have many
1701 actors. Set this configuration item to **off** to disable this
1704 .. _cfg=exception/cutpath:
1706 Truncate local path from exception backtrace
1707 ............................................
1709 **Option** ``exception/cutpath`` **default:** off
1711 This configuration option is used to remove the path from the
1712 backtrace shown when an exception is thrown. This is mainly useful for
1713 the tests: the full file path would makes the tests non-reproducible because
1714 the paths of source files depend of the build settings. That would
1715 break most of the tests since their output is continually compared.
1719 Logging configuration
1720 ---------------------
1722 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
1723 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
1724 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
1727 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
1728 messages from your code.
1730 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
1731 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
1732 practice, the following is equivalent to the above settings: ``--log=root.thresh:error --log=s4u_host.thresh:debug``.
1734 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
1735 your settings, as in ``--log="root.thresh:error s4u_host.thresh:debug"``. The parameters are interpreted in order, from left to right.
1738 Threshold configuration
1739 .......................
1741 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
1742 ``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
1743 see, ``threshold`` can be abbreviated here.
1745 Existing thresholds:
1747 - ``trace`` some functions display a message at this level when entering or returning
1748 - ``debug`` output that is mostly useful when debugging the corresponding module.
1749 - ``verbose`` verbose output that is only mildly interesting and can easily be ignored
1750 - ``info`` usual output (this is the default threshold of all categories)
1751 - ``warning`` minor issue encountered
1752 - ``error`` issue encountered
1753 - ``critical`` major issue encountered, such as assertions failures
1757 Format configuration
1758 ....................
1760 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
1761 as the date, or the actor ID, everything. Existing format directives:
1764 - %n: line separator (LOG4J compatible)
1765 - %e: plain old space (SimGrid extension)
1767 - %m: user-provided message
1769 - %c: Category name (LOG4J compatible)
1770 - %p: Priority name (LOG4J compatible)
1772 - %h: Hostname (SimGrid extension)
1773 - %a: Actor name (SimGrid extension -- note that with SMPI this is the integer value of the process rank)
1774 - %i: Actor PID (SimGrid extension -- this is a 'i' as in 'i'dea)
1775 - %t: Thread "name" (LOG4J compatible -- actually the address of the thread in memory)
1777 - %F: file name where the log event was raised (LOG4J compatible)
1778 - %l: location where the log event was raised (LOG4J compatible, like '%%F:%%L' -- this is a l as in 'l'etter)
1779 - %L: line number where the log event was raised (LOG4J compatible)
1780 - %M: function name (LOG4J compatible -- called method name here of course).
1782 - %d: date (UNIX-like epoch)
1783 - %r: application age (time elapsed since the beginning of the application)
1786 ``--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
1787 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
1788 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
1789 provided layout is used for every messages.
1791 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
1795 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
1796 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'"``.
1797 Another option is to use the ``%e`` directive for spaces, as in ``--log=root.fmt:%l:%e[%p/%c]:%e%m%n``.
1802 The keyword ``app`` controls the appended of a logging category. For example ``--log=root.app:file:mylogfile`` redirects every output to the file ``mylogfile``.
1804 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
1805 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.
1807 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``
1808 ensures that the log file ``mylog`` will never overpass 500 bytes in size.
1810 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
1811 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.
1816 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
1817 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
1818 ``on`` (or ``yes`` or ``1``), the produced messages will also be passed to the upper appender.
1820 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
1821 ``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
1822 will only be sent to ``all.log``.
1827 ``--help-logs`` displays a complete help message about logging in SimGrid.
1829 ``--help-log-categories`` displays the actual hierarchy of log categories for this binary.
1831 ``--log=no_loc`` hides the source locations (file names and line numbers) from the messages. This is useful to make tests reproducible.