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/termination:** :ref:`cfg=model-check/termination`
123 - **model-check/timeout:** :ref:`cfg=model-check/timeout`
124 - **model-check/visited:** :ref:`cfg=model-check/visited`
126 - **network/bandwidth-factor:** :ref:`cfg=network/bandwidth-factor`
127 - **network/crosstraffic:** :ref:`cfg=network/crosstraffic`
128 - **network/latency-factor:** :ref:`cfg=network/latency-factor`
129 - **network/loopback-lat:** :ref:`cfg=network/loopback`
130 - **network/loopback-bw:** :ref:`cfg=network/loopback`
131 - **network/maxmin-selective-update:** :ref:`Network Optimization Level <options_model_optim>`
132 - **network/model:** :ref:`options_model_select`
133 - **network/optim:** :ref:`Network Optimization Level <options_model_optim>`
134 - **network/TCP-gamma:** :ref:`cfg=network/TCP-gamma`
135 - **network/weight-S:** :ref:`cfg=network/weight-S`
137 - **ns3/TcpModel:** :ref:`options_pls`
138 - **ns3/seed:** :ref:`options_pls`
139 - **path:** :ref:`cfg=path`
140 - **plugin:** :ref:`cfg=plugin`
142 - **storage/max_file_descriptors:** :ref:`cfg=storage/max_file_descriptors`
144 - **surf/precision:** :ref:`cfg=surf/precision`
146 - **For collective operations of SMPI,** please refer to Section :ref:`cfg=smpi/coll-selector`
147 - **smpi/auto-shared-malloc-thresh:** :ref:`cfg=smpi/auto-shared-malloc-thresh`
148 - **smpi/async-small-thresh:** :ref:`cfg=smpi/async-small-thresh`
149 - **smpi/buffering:** :ref:`cfg=smpi/buffering`
150 - **smpi/bw-factor:** :ref:`cfg=smpi/bw-factor`
151 - **smpi/coll-selector:** :ref:`cfg=smpi/coll-selector`
152 - **smpi/comp-adjustment-file:** :ref:`cfg=smpi/comp-adjustment-file`
153 - **smpi/cpu-threshold:** :ref:`cfg=smpi/cpu-threshold`
154 - **smpi/display-allocs:** :ref:`cfg=smpi/display-allocs`
155 - **smpi/display-timing:** :ref:`cfg=smpi/display-timing`
156 - **smpi/errors-are-fatal:** :ref:`cfg=smpi/errors-are-fatal`
157 - **smpi/finalization-barrier:** :ref:`cfg=smpi/finalization-barrier`
158 - **smpi/grow-injected-times:** :ref:`cfg=smpi/grow-injected-times`
159 - **smpi/host-speed:** :ref:`cfg=smpi/host-speed`
160 - **smpi/IB-penalty-factors:** :ref:`cfg=smpi/IB-penalty-factors`
161 - **smpi/iprobe:** :ref:`cfg=smpi/iprobe`
162 - **smpi/iprobe-cpu-usage:** :ref:`cfg=smpi/iprobe-cpu-usage`
163 - **smpi/init:** :ref:`cfg=smpi/init`
164 - **smpi/keep-temps:** :ref:`cfg=smpi/keep-temps`
165 - **smpi/lat-factor:** :ref:`cfg=smpi/lat-factor`
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 Correcting Important Network Parameters
393 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
395 SimGrid can take network irregularities such as a slow startup or
396 changing behavior depending on the message size into account. You
397 should not change these values unless you really know what you're
398 doing. The corresponding values were computed through data fitting
399 one the timings of packet-level simulators, as described in `Accuracy
400 Study and Improvement of Network Simulation in the SimGrid Framework
401 <http://mescal.imag.fr/membres/arnaud.legrand/articles/simutools09.pdf>`_.
403 - **network/latency-factor**: apply a multiplier to latency.
404 Models the TCP slow-start mechanism.
405 - **network/bandwidth-factor**: actual bandwidth perceived by the
407 - **network/weight-S**: bottleneck sharing constant parameter. Used
410 These parameters are the same for all communications in your simulation,
411 independently of message size or source/destination hosts. A more flexible
412 mechanism based on callbacks was introduced in SimGrid. It provides the user
413 a callback that will be called for each communication, allowing the user
414 to set different latency and bandwidth factors, based on the message size, links used
415 or zones traversed. To more details of how to use it, please look at the
416 `examples/cpp/network-factors/s4u-network-factors.cpp <https://framagit.org/simgrid/simgrid/tree/master/examples/cpp/network-factors/s4u-network-factors.cpp>`_.
419 If you are using the SMPI model, these correction coefficients are
420 themselves corrected by constant values depending on the size of the
421 exchange. By default SMPI uses factors computed on the Stampede
422 Supercomputer at TACC, with optimal deployment of processes on
423 nodes. Again, only hardcore experts should bother about this fact.
424 For more details, see SMPI sections about :ref:`cfg=smpi/bw-factor` and :ref:`cfg=smpi/lat-factor`.
427 .. _cfg=smpi/IB-penalty-factors:
432 InfiniBand network behavior can be modeled through 3 parameters
433 ``smpi/IB-penalty-factors:"βe;βs;γs"``, as explained in `this PhD
435 <http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf>`_ (in French)
436 or more concisely in `this paper <https://hal.inria.fr/hal-00953618/document>`_,
437 even if that paper does only describe models for myrinet and ethernet.
438 You can see in Fig 2 some results for Infiniband, for example. This model
439 may be outdated by now for modern infiniband, anyway, so a new
440 validation would be good.
442 The three paramaters are defined as follows:
444 - βs: penalty factor for outgoing messages, computed by running a simple send to
445 two nodes and checking slowdown compared to a single send to one node,
447 - βe: penalty factor for ingoing messages, same computation method but with one
448 node receiving several messages
449 - γr: slowdown factor when communication buffer memory is saturated. It needs a
450 more complicated pattern to run in order to be computed (5.3 in the thesis,
451 page 107), and formula in the end is γr = time(c)/(3×βe×time(ref)), where
452 time(ref) is the time of a single comm with no contention).
454 Once these values are computed, a penalty is assessed for each message (this is
455 the part implemented in the simulator) as shown page 106 of the thesis. Here is
456 a simple translation of this text. First, some notations:
458 - ∆e(e) which corresponds to the incoming degree of node e, that is to say the number of communications having as destination node e.
459 - ∆s (s) which corresponds to the degree outgoing from node s, that is to say the number of communications sent by node s.
460 - Φ (e) which corresponds to the number of communications destined for the node e but coming from a different node.
461 - 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
463 To determine the penalty for a communication, two values need to be calculated. First, the penalty caused by the conflict in transmission, noted ps.
466 - if ∆s (i) = 1 then ps = 1.
467 - if ∆s (i) ≥ 2 and ∆e (i) ≥ 3 then ps = ∆s (i) × βs × γr
468 - else, ps = ∆s (i) × βs
471 Then, the penalty caused by the conflict in reception (noted pe) should be computed as follows:
473 - if ∆e (i) = 1 then pe = 1
474 - else, pe = Φ (e) × βe × Ω (s, e)
476 Finally, the penalty associated with the communication is:
477 p = max (ps ∈ s, pe)
479 .. _cfg=network/crosstraffic:
481 Simulating Cross-Traffic
482 ^^^^^^^^^^^^^^^^^^^^^^^^
484 Since SimGrid v3.7, cross-traffic effects can be taken into account in
485 analytical simulations. It means that ongoing and incoming
486 communication flows are treated independently. In addition, the LV08
487 model adds 0.05 of usage on the opposite direction for each new
488 created flow. This can be useful to simulate some important TCP
489 phenomena such as ack compression.
491 For that to work, your platform must have two links for each
492 pair of interconnected hosts. An example of usable platform is
493 available in ``examples/platforms/crosstraffic.xml``.
495 This is activated through the ``network/crosstraffic`` item, that
496 can be set to 0 (disable this feature) or 1 (enable it).
498 Note that with the default host model this option is activated by default.
500 .. _cfg=network/loopback:
502 Configuring loopback link
503 ^^^^^^^^^^^^^^^^^^^^^^^^^
505 Several network model provide an implicit loopback link to account for local
506 communication on a host. By default it has a 10GBps bandwidth and a null latency.
507 This can be changed with ``network/loopback-lat`` and ``network/loopback-bw``
510 .. _cfg=smpi/async-small-thresh:
512 Simulating Asynchronous Send
513 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
515 (this configuration item is experimental and may change or disappear)
517 It is possible to specify that messages below a certain size (in bytes) will be
518 sent as soon as the call to MPI_Send is issued, without waiting for
519 the correspondent receive. This threshold can be configured through
520 the ``smpi/async-small-thresh`` item. The default value is 0. This
521 behavior can also be manually set for mailboxes, by setting the
522 receiving mode of the mailbox with a call to
523 :cpp:func:`MSG_mailbox_set_async`. After this, all messages sent to
524 this mailbox will have this behavior regardless of the message size.
526 This value needs to be smaller than or equals to the threshold set at
527 :ref:`cfg=smpi/send-is-detached-thresh`, because asynchronous messages
528 are meant to be detached as well.
535 **Option** ``ns3/TcpModel`` **Default:** "default" (ns-3 default)
537 When using ns-3, there is an extra item ``ns3/TcpModel``, corresponding
538 to the ``ns3::TcpL4Protocol::SocketType`` configuration item in
539 ns-3. The only valid values (enforced on the SimGrid side) are
540 'default' (no change to the ns-3 configuration), 'NewReno' or 'Reno' or
543 **Option** ``ns3/seed`` **Default:** "" (don't set the seed in ns-3)
545 This option is the random seed to provide to ns-3 with
546 ``ns3::RngSeedManager::SetSeed`` and ``ns3::RngSeedManager::SetRun``.
548 If left blank, no seed is set in ns-3. If the value 'time' is
549 provided, the current amount of seconds since epoch is used as a seed.
550 Otherwise, the provided value must be a number to use as a seed.
552 Configuring the Storage model
553 .............................
555 .. _cfg=storage/max_file_descriptors:
557 File Descriptor Count per Host
558 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
560 **Option** ``storage/max_file_descriptors`` **Default:** 1024
562 Each host maintains a fixed-size array of its file descriptors. You
563 can change its size through this item to either enlarge it if your
564 application requires it or to reduce it to save memory space.
571 SimGrid plugins allow one to extend the framework without changing its
572 source code directly. Read the source code of the existing plugins to
573 learn how to do so (in ``src/plugins``), and ask your questions to the
574 usual channels (Stack Overflow, Mailing list, IRC). The basic idea is
575 that plugins usually register callbacks to some signals of interest.
576 If they need to store some information about a given object (Link, CPU
577 or Actor), they do so through the use of a dedicated object extension.
579 Some of the existing plugins can be activated from the command line,
580 meaning that you can activate them from the command line without any
581 modification to your simulation code. For example, you can activate
582 the host energy plugin by adding ``--cfg=plugin:host_energy`` to your
585 Here is a partial list of plugins that can be activated this way. You can get
586 the full list by passing ``--cfg=plugin:help`` to your simulator.
588 - :ref:`Host Energy <plugin_host_energy>`: models the energy dissipation of the compute units.
589 - :ref:`Link Energy <plugin_link_energy>`: models the energy dissipation of the network.
590 - :ref:`Host Load <plugin_host_load>`: monitors the load of the compute units.
592 .. _options_modelchecking:
594 Configuring the Model-Checking
595 ------------------------------
597 To enable SimGrid's model-checking support, the program should
598 be executed using the simgrid-mc wrapper:
600 .. code-block:: console
602 $ simgrid-mc ./my_program
604 Safety properties are expressed as assertions using the function
605 :cpp:func:`void MC_assert(int prop)`.
607 .. _cfg=smpi/buffering:
609 Specifying the MPI buffering behavior
610 .....................................
612 **Option** ``smpi/buffering`` **Default:** infty
614 Buffering in MPI has a huge impact on the communication semantic. For example,
615 standard blocking sends are synchronous calls when the system buffers are full
616 while these calls can complete immediately without even requiring a matching
617 receive call for small messages sent when the system buffers are empty.
619 In SMPI, this depends on the message size, that is compared against two thresholds:
621 - if (size < :ref:`smpi/async-small-thresh <cfg=smpi/async-small-thresh>`) then
622 MPI_Send returns immediately, even if the corresponding receive has not be issued yet.
623 - 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
624 MPI_Send returns as soon as the corresponding receive has been issued. This is known as the eager mode.
625 - if (:ref:`smpi/send-is-detached-thresh <cfg=smpi/send-is-detached-thresh>` < size) then
626 MPI_Send returns only when the message has actually been sent over the network. This is known as the rendez-vous mode.
628 The ``smpi/buffering`` (only valid with MC) option gives an easier interface to choose between these semantics. It can take two values:
630 - **zero:** means that buffering should be disabled. All communications are actually blocking.
631 - **infty:** means that buffering should be made infinite. All communications are non-blocking.
633 .. _cfg=model-check/property:
635 Specifying a liveness property
636 ..............................
638 **Option** ``model-check/property`` **Default:** unset
640 If you want to specify liveness properties, you have to pass them on
641 the command line, specifying the name of the file containing the
642 property, as formatted by the `ltl2ba <https://github.com/utwente-fmt/ltl2ba>`_ program.
643 Note that ltl2ba is not part of SimGrid and must be installed separately.
645 .. code-block:: console
647 $ simgrid-mc ./my_program --cfg=model-check/property:<filename>
649 .. _cfg=model-check/checkpoint:
651 Going for Stateful Verification
652 ...............................
654 By default, the system is backtracked to its initial state to explore
655 another path, instead of backtracking to the exact step before the fork
656 that we want to explore (this is called stateless verification). This
657 is done this way because saving intermediate states can rapidly
658 exhaust the available memory. If you want, you can change the value of
659 the ``model-check/checkpoint`` item. For example,
660 ``--cfg=model-check/checkpoint:1`` asks to take a checkpoint every
661 step. Beware, this will certainly explode your memory. Larger values
662 are probably better, make sure to experiment a bit to find the right
663 setting for your specific system.
665 .. _cfg=model-check/reduction:
667 Specifying the kind of reduction
668 ................................
670 The main issue when using the model-checking is the state space
671 explosion. You can activate some reduction technique with
672 ``--cfg=model-check/reduction:<technique>``. For now, this
673 configuration variable can take 2 values:
675 - **none:** Do not apply any kind of reduction (mandatory for
676 liveness properties, as our current DPOR algorithm breaks cycles)
677 - **dpor:** Apply Dynamic Partial Ordering Reduction. Only valid if
678 you verify local safety properties (default value for safety
681 Another way to mitigate the state space explosion is to search for
682 cycles in the exploration with the :ref:`cfg=model-check/visited`
683 configuration. Note that DPOR and state-equality reduction may not
684 play well together. You should choose between them.
686 Our current DPOR implementation could be improved in may ways. We are
687 currently improving its efficiency (both in term of reduction ability
688 and computational speed), and future work could make it compatible
689 with liveness properties.
691 .. _cfg=model-check/visited:
693 Size of Cycle Detection Set (state equality reduction)
694 ......................................................
696 Mc SimGrid can be asked to search for cycles during the exploration,
697 i.e. situations where a new explored state is in fact the same state
698 than a previous one.. This can prove useful to mitigate the state
699 space explosion with safety properties, and this is the crux when
700 searching for counter-examples to the liveness properties.
702 Note that this feature may break the current implementation of the
703 DPOR reduction technique.
705 The ``model-check/visited`` item is the maximum number of states, which
706 are stored in memory. If the maximum number of snapshotted state is
707 reached, some states will be removed from the memory and some cycles
708 might be missed. Small values can lead to incorrect verifications, but
709 large values can exhaust your memory and be CPU intensive as each new
710 state must be compared to that amount of older saved states.
712 The default settings depend on the kind of exploration. With safety
713 checking, no state is snapshotted and cycles cannot be detected. With
714 liveness checking, all states are snapshotted because missing a cycle
715 could hinder the exploration soundness.
717 .. _cfg=model-check/termination:
719 Non-Termination Detection
720 .........................
722 The ``model-check/termination`` configuration item can be used to
723 report if a non-termination execution path has been found. This is a
724 path with a cycle, which means that the program might never terminate.
726 This only works in safety mode, not in liveness mode.
728 This options is disabled by default.
730 .. _cfg=model-check/dot-output:
735 If set, the ``model-check/dot-output`` configuration item is the name
736 of a file in which to write a dot file of the path leading to the
737 property violation discovered (safety or liveness violation), as well
738 as the cycle for liveness properties. This dot file can then be fed to the
739 graphviz dot tool to generate a corresponding graphical representation.
741 .. _cfg=model-check/max-depth:
743 Exploration Depth Limit
744 .......................
746 The ``model-check/max-depth`` can set the maximum depth of the
747 exploration graph of the model checker. If this limit is reached, a
748 logging message is sent and the results might not be exact.
750 By default, the exploration is limited to the depth of 1000.
752 .. _cfg=model-check/timeout:
757 By default, the model checker does not handle timeout conditions: the `wait`
758 operations never time out. With the ``model-check/timeout`` configuration item
759 set to **yes**, the model checker will explore timeouts of `wait` operations.
761 .. _cfg=model-check/communications-determinism:
762 .. _cfg=model-check/send-determinism:
764 Communication Determinism
765 .........................
767 The ``model-check/communications-determinism`` and
768 ``model-check/send-determinism`` items can be used to select the
769 communication determinism mode of the model checker, which checks
770 determinism properties of the communications of an application.
774 Verification Performance Considerations
775 .......................................
777 The size of the stacks can have a huge impact on the memory
778 consumption when using model-checking. By default, each snapshot will
779 save a copy of the whole stacks and not only of the part that is
780 really meaningful: you should expect the contribution of the memory
781 consumption of the snapshots to be:
782 :math:`\text{number of processes} \times \text{stack size} \times \text{number of states}`.
784 When compiled against the model checker, the stacks are not
785 protected with guards: if the stack size is too small for your
786 application, the stack will silently overflow into other parts of the
787 memory (see :ref:`contexts/guard-size <cfg=contexts/guard-size>`).
789 .. _cfg=model-check/replay:
791 Replaying buggy execution paths from the model checker
792 ......................................................
794 Debugging the problems reported by the model checker is challenging:
795 First, the application under verification cannot be debugged with gdb
796 because the model checker already traces it. Then, the model checker may
797 explore several execution paths before encountering the issue, making it
798 very difficult to understand the output. Fortunately, SimGrid provides
799 the execution path leading to any reported issue so that you can replay
800 this path reported by the model checker, enabling the usage of classical
803 When the model checker finds an interesting path in the application
804 execution graph (where a safety or liveness property is violated), it
805 generates an identifier for this path. Here is an example of the output:
807 .. code-block:: console
809 [ 0.000000] (0:@) Check a safety property
810 [ 0.000000] (0:@) **************************
811 [ 0.000000] (0:@) *** PROPERTY NOT VALID ***
812 [ 0.000000] (0:@) **************************
813 [ 0.000000] (0:@) Counter-example execution trace:
814 [ 0.000000] (0:@) [(1)Tremblay (app)] MC_RANDOM(3)
815 [ 0.000000] (0:@) [(1)Tremblay (app)] MC_RANDOM(4)
816 [ 0.000000] (0:@) Path = 1/3;1/4
817 [ 0.000000] (0:@) Expanded states = 27
818 [ 0.000000] (0:@) Visited states = 68
819 [ 0.000000] (0:@) Executed transitions = 46
821 The interesting line is ``Path = 1/3;1/4``, which means that you should use
822 ``--cfg=model-check/replay:1/3;1/4`` to replay your application on the buggy
823 execution path. All options (but the model checker related ones) must
824 remain the same. In particular, if you ran your application with
825 ``smpirun -wrapper simgrid-mc``, then do it again. Remove all
826 MC-related options, keep non-MC-related ones and add
827 ``--cfg=model-check/replay:???``.
829 Currently, if the path is of the form ``X;Y;Z``, each number denotes
830 the actor's pid that is selected at each indecision point. If it's of
831 the form ``X/a;Y/b``, the X and Y are the selected pids while the a
832 and b are the return values of their simcalls. In the previous
833 example, ``1/3;1/4``, you can see from the full output that the actor
834 1 is doing MC_RANDOM simcalls, so the 3 and 4 simply denote the values
835 that these simcall return on the execution branch leading to the
838 Configuring the User Code Virtualization
839 ----------------------------------------
841 .. _cfg=contexts/factory:
843 Selecting the Virtualization Factory
844 ....................................
846 **Option** contexts/factory **Default:** "raw"
848 In SimGrid, the user code is virtualized in a specific mechanism that
849 allows the simulation kernel to control its execution: when a user
850 process requires a blocking action (such as sending a message), it is
851 interrupted, and only gets released when the simulated clock reaches
852 the point where the blocking operation is done. This is explained
853 graphically in the `relevant tutorial, available online
854 <https://simgrid.org/tutorials/simgrid-simix-101.pdf>`_.
856 In SimGrid, the containers in which user processes are virtualized are
857 called contexts. Several context factory are provided, and you can
858 select the one you want to use with the ``contexts/factory``
859 configuration item. Some of the following may not exist on your
860 machine because of portability issues. In any case, the default one
861 should be the most effcient one (please report bugs if the
862 auto-detection fails for you). They are approximately sorted here from
863 the slowest to the most efficient:
865 - **thread:** very slow factory using full featured threads (either
866 pthreads or windows native threads). They are slow but very
867 standard. Some debuggers or profilers only work with this factory.
868 - **java:** Java applications are virtualized onto java threads (that
869 are regular pthreads registered to the JVM)
870 - **ucontext:** fast factory using System V contexts (Linux and FreeBSD only)
871 - **boost:** This uses the `context
872 implementation <http://www.boost.org/doc/libs/1_59_0/libs/context/doc/html/index.html>`_
873 of the boost library for a performance that is comparable to our
875 |br| Install the relevant library (e.g. with the
876 libboost-contexts-dev package on Debian/Ubuntu) and recompile
878 - **raw:** amazingly fast factory using a context switching mechanism
879 of our own, directly implemented in assembly (only available for x86
880 and amd64 platforms for now) and without any unneeded system call.
882 The main reason to change this setting is when the debugging tools become
883 fooled by the optimized context factories. Threads are the most
884 debugging-friendly contexts, as they allow one to set breakpoints
885 anywhere with gdb and visualize backtraces for all processes, in order
886 to debug concurrency issues. Valgrind is also more comfortable with
887 threads, but it should be usable with all factories (Exception: the
888 callgrind tool really dislikes raw and ucontext factories).
890 .. _cfg=contexts/stack-size:
892 Adapting the Stack Size
893 .......................
895 **Option** ``contexts/stack-size`` **Default:** 8192 KiB
897 Each virtualized used process is executed using a specific system
898 stack. The size of this stack has a huge impact on the simulation
899 scalability, but its default value is rather large. This is because
900 the error messages that you get when the stack size is too small are
901 rather disturbing: this leads to stack overflow (overwriting other
902 stacks), leading to segfaults with corrupted stack traces.
904 If you want to push the scalability limits of your code, you might
905 want to reduce the ``contexts/stack-size`` item. Its default value is
906 8192 (in KiB), while our Chord simulation works with stacks as small
907 as 16 KiB, for example. You can ensure that some actors have a specific
908 size by simply changing the value of this configuration item before
909 creating these actors. The :cpp:func:`simgrid::s4u::Engine::set_config`
910 functions are handy for that.
912 This *setting is ignored* when using the thread factory (because there
913 is no way to modify the stack size with C++ system threads). Instead,
914 you should compile SimGrid and your application with
915 ``-fsplit-stack``. Note that this compilation flag is not compatible
916 with the model checker right now.
918 The operating system should only allocate memory for the pages of the
919 stack which are actually used and you might not need to use this in
920 most cases. However, this setting is very important when using the
921 model checker (see :ref:`options_mc_perf`).
923 .. _cfg=contexts/guard-size:
925 Disabling Stack Guard Pages
926 ...........................
928 **Option** ``contexts/guard-size`` **Default** 1 page in most case (0 pages on Windows or with MC)
930 Unless you use the threads context factory (see
931 :ref:`cfg=contexts/factory`), a stack guard page is usually used
932 which prevents the stack of a given actor from overflowing on another
933 stack. But the performance impact may become prohibitive when the
934 amount of actors increases. The option ``contexts/guard-size`` is the
935 number of stack guard pages used. By setting it to 0, no guard pages
936 will be used: in this case, you should avoid using small stacks (with
937 :ref:`contexts/stack-size <cfg=contexts/stack-size>`) as the stack
938 will silently overflow on other parts of the memory.
940 When no stack guard page is created, stacks may then silently overflow
941 on other parts of the memory if their size is too small for the
944 .. _cfg=contexts/nthreads:
945 .. _cfg=contexts/synchro:
947 Running User Code in Parallel
948 .............................
950 Parallel execution of the user code is only considered stable in
951 SimGrid v3.7 and higher, and mostly for MSG simulations. SMPI
952 simulations may well fail in parallel mode. It is described in
953 `INRIA RR-7653 <http://hal.inria.fr/inria-00602216/>`_.
955 If you are using the **ucontext** or **raw** context factories, you can
956 request to execute the user code in parallel. Several threads are
957 launched, each of them handling the same number of user contexts at each
958 run. To activate this, set the ``contexts/nthreads`` item to the amount
959 of cores that you have in your computer (or lower than 1 to have the
960 amount of cores auto-detected).
962 When parallel execution is activated, you can choose the
963 synchronization schema used with the ``contexts/synchro`` item,
964 which value is either:
966 - **futex:** ultra optimized synchronisation schema, based on futexes
967 (fast user-mode mutexes), and thus only available on Linux systems.
968 This is the default mode when available.
969 - **posix:** slow but portable synchronisation using only POSIX
971 - **busy_wait:** not really a synchronisation: the worker threads
972 constantly request new contexts to execute. It should be the most
973 efficient synchronisation schema, but it loads all the cores of
974 your machine for no good reason. You probably prefer the other less
977 Configuring the Tracing
978 -----------------------
980 The :ref:`tracing subsystem <outcome_vizu>` can be configured in
981 several different ways depending on the used interface (S4U, SMPI)
982 and the kind of traces that needs to be obtained. See the
983 :ref:`Tracing Configuration Options subsection
984 <tracing_tracing_options>` for a full description of each
985 configuration option.
987 We detail here a simple way to get the traces working for you, even if
988 you never used the tracing API.
991 - Any SimGrid-based simulator (MSG, SMPI, ...) and raw traces:
995 --cfg=tracing:yes --cfg=tracing/uncategorized:yes
997 The first parameter activates the tracing subsystem, and the second
998 tells it to trace host and link utilization (without any
1001 - MSG-based simulator and categorized traces (you need to
1002 declare categories and classify your tasks according to them)
1004 .. code-block:: none
1006 --cfg=tracing:yes --cfg=tracing/categorized:yes
1008 The first parameter activates the tracing subsystem, and the second
1009 tells it to trace host and link categorized utilization.
1011 - SMPI simulator and traces for a space/time view:
1013 .. code-block:: console
1015 $ smpirun -trace ...
1017 The `-trace` parameter for the smpirun script runs the simulation
1018 with ``--cfg=tracing:yes --cfg=tracing/smpi:yes``. Check the
1019 smpirun's `-help` parameter for additional tracing options.
1021 Sometimes you might want to put additional information on the trace to
1022 correctly identify them later, or to provide data that can be used to
1023 reproduce an experiment. You have two ways to do that:
1025 - Add a string on top of the trace file as comment:
1027 .. code-block:: none
1029 --cfg=tracing/comment:my_simulation_identifier
1031 - Add the contents of a textual file on top of the trace file as comment:
1033 .. code-block:: none
1035 --cfg=tracing/comment-file:my_file_with_additional_information.txt
1037 Please, use these two parameters (for comments) to make reproducible
1038 simulations. For additional details about this and all tracing
1039 options, check See the :ref:`tracing_tracing_options`.
1044 .. _cfg=msg/debug-multiple-use:
1049 **Option** ``msg/debug-multiple-use`` **Default:** off
1051 Sometimes your application may try to send a task that is still being
1052 executed somewhere else, making it impossible to send this task. However,
1053 for debugging purposes, one may want to know what the other host is/was
1054 doing. This option shows a backtrace of the other process.
1059 The SMPI interface provides several specific configuration items.
1060 These are not easy to see, since the code is usually launched through the
1061 ``smiprun`` script directly.
1063 .. _cfg=smpi/host-speed:
1064 .. _cfg=smpi/cpu-threshold:
1065 .. _cfg=smpi/simulate-computation:
1067 Automatic Benchmarking of SMPI Code
1068 ...................................
1070 In SMPI, the sequential code is automatically benchmarked, and these
1071 computations are automatically reported to the simulator. That is to
1072 say that if you have a large computation between a ``MPI_Recv()`` and
1073 a ``MPI_Send()``, SMPI will automatically benchmark the duration of
1074 this code, and create an execution task within the simulator to take
1075 this into account. For that, the actual duration is measured on the
1076 host machine and then scaled to the power of the corresponding
1077 simulated machine. The variable ``smpi/host-speed`` allows one to
1078 specify the computational speed of the host machine (in flop/s by
1079 default) to use when scaling the execution times.
1081 The default value is ``smpi/host-speed=20kf`` (= 20,000 flop/s). This
1082 is probably underestimated for most machines, leading SimGrid to
1083 overestimate the amount of flops in the execution blocks that are
1084 automatically injected in the simulator. As a result, the execution
1085 time of the whole application will probably be overestimated until you
1086 use a realistic value.
1088 When the code consists of numerous consecutive MPI calls, the
1089 previous mechanism feeds the simulation kernel with numerous tiny
1090 computations. The ``smpi/cpu-threshold`` item becomes handy when this
1091 impacts badly on the simulation performance. It specifies a threshold (in
1092 seconds) below which the execution chunks are not reported to the
1093 simulation kernel (default value: 1e-6).
1095 .. note:: The option ``smpi/cpu-threshold`` ignores any computation
1096 time spent below this threshold. SMPI does not consider the
1097 `amount of time` of these computations; there is no offset for
1098 this. Hence, a value that is too small, may lead to unreliable
1101 In some cases, however, one may wish to disable simulation of
1102 the computation of an application. This is the case when SMPI is used not to
1103 simulate an MPI application, but instead an MPI code that performs
1104 "live replay" of another MPI app (e.g., ScalaTrace's replay tool, or
1105 various on-line simulators that run an app at scale). In this case the
1106 computation of the replay/simulation logic should not be simulated by
1107 SMPI. Instead, the replay tool or on-line simulator will issue
1108 "computation events", which correspond to the actual MPI simulation
1109 being replayed/simulated. At the moment, these computation events can
1110 be simulated using SMPI by calling internal smpi_execute*() functions.
1112 To disable the benchmarking/simulation of a computation in the simulated
1113 application, the variable ``smpi/simulate-computation`` should be set
1114 to **no**. This option just ignores the timings in your simulation; it
1115 still executes the computations itself. If you want to stop SMPI from
1116 doing that, you should check the SMPI_SAMPLE macros, documented in
1117 Section :ref:`SMPI_use_faster`.
1119 +------------------------------------+-------------------------+-----------------------------+
1120 | Solution | Computations executed? | Computations simulated? |
1121 +====================================+=========================+=============================+
1122 | --cfg=smpi/simulate-computation:no | Yes | Never |
1123 +------------------------------------+-------------------------+-----------------------------+
1124 | --cfg=smpi/cpu-threshold:42 | Yes, in all cases | If it lasts over 42 seconds |
1125 +------------------------------------+-------------------------+-----------------------------+
1126 | SMPI_SAMPLE() macro | Only once per loop nest | Always |
1127 +------------------------------------+-------------------------+-----------------------------+
1129 .. _cfg=smpi/comp-adjustment-file:
1131 Slow-down or speed-up parts of your code
1132 ........................................
1134 **Option** ``smpi/comp-adjustment-file:`` **Default:** unset
1136 This option allows you to pass a file that contains two columns: The
1137 first column defines the section that will be subject to a speedup;
1138 the second column is the speedup. For instance:
1140 .. code-block:: none
1142 "start:stop","ratio"
1143 "exchange_1.f:30:exchange_1.f:130",1.18244559422142
1145 The first line is the header - you must include it. The following
1146 line means that the code between two consecutive MPI calls on line 30
1147 in exchange_1.f and line 130 in exchange_1.f should receive a speedup
1148 of 1.18244559422142. The value for the second column is therefore a
1149 speedup, if it is larger than 1 and a slowdown if it is smaller
1150 than 1. Nothing will be changed if it is equal to 1.
1152 Of course, you can set any arbitrary filenames you want (so the start
1153 and end don't have to be in the same file), but be aware that this
1154 mechanism only supports `consecutive calls!`
1156 Please note that you must pass the ``-trace-call-location`` flag to
1157 smpicc or smpiff, respectively. This flag activates some internal
1158 macro definitions that help with obtaining the call location.
1160 .. _cfg=smpi/bw-factor:
1165 **Option** ``smpi/bw-factor``
1166 |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
1168 The possible throughput of network links is often dependent on the
1169 message sizes, as protocols may adapt to different message sizes. With
1170 this option, a series of message sizes and factors are given, helping
1171 the simulation to be more realistic. For instance, the current default
1172 value means that messages with size 65472 bytes and more will get a total of
1173 MAX_BANDWIDTH*0.940694, messages of size 15424 to 65471 will get
1174 MAX_BANDWIDTH*0.697866, and so on (where MAX_BANDWIDTH denotes the
1175 bandwidth of the link).
1177 An experimental script to compute these factors is available online. See
1178 https://framagit.org/simgrid/platform-calibration/
1179 https://simgrid.org/contrib/smpi-saturation-doc.html
1181 .. _cfg=smpi/display-timing:
1183 Reporting Simulation Time
1184 .........................
1186 **Option** ``smpi/display-timing`` **Default:** 0 (false)
1188 Most of the time, you run MPI code with SMPI to compute the time it
1189 would take to run it on a platform. But since the code is run through
1190 the ``smpirun`` script, you don't have any control on the launcher
1191 code, making it difficult to report the simulated time when the
1192 simulation ends. If you enable the ``smpi/display-timing`` item,
1193 ``smpirun`` will display this information when the simulation
1195 SMPI will also display information about the amout of real time spent
1196 in application code and in SMPI internals, to provide hints about the
1197 need to use sampling to reduce simulation time.
1199 .. _cfg=smpi/display-allocs:
1201 Reporting memory allocations
1202 ............................
1204 **Option** ``smpi/display-allocs`` **Default:** 0 (false)
1206 SMPI intercepts malloc and calloc calls performed inside the running
1207 application, if it wasn't compiled with SMPI_NO_OVERRIDE_MALLOC.
1208 With this option, SMPI will show at the end of execution the amount of
1209 memory allocated through these calls, and locate the most expensive one.
1210 This helps finding the targets for manual memory sharing, or the threshold
1211 to use for smpi/auto-shared-malloc-thresh option (see :ref:`cfg=smpi/auto-shared-malloc-thresh`).
1213 .. _cfg=smpi/keep-temps:
1215 Keeping temporary files after simulation
1216 ........................................
1218 **Option** ``smpi/keep-temps`` **default:** 0 (false)
1220 SMPI usually generates a lot of temporary files that are cleaned after
1221 use. This option requests to preserve them, for example to debug or
1222 profile your code. Indeed, the binary files are removed very early
1223 under the dlopen privatization schema, which tends to fool the
1226 .. _cfg=smpi/lat-factor:
1231 **Option** ``smpi/lat-factor`` |br|
1232 **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
1234 The motivation and syntax for this option is identical to the motivation/syntax
1235 of :ref:`cfg=smpi/bw-factor`.
1237 There is an important difference, though: While smpi/bw-factor `reduces` the
1238 actual bandwidth (i.e., values between 0 and 1 are valid), latency factors
1239 increase the latency, i.e., values larger than or equal to 1 are valid here.
1241 .. _cfg=smpi/papi-events:
1243 Trace hardware counters with PAPI
1244 .................................
1246 **Option** ``smpi/papi-events`` **default:** unset
1248 When the PAPI support is compiled into SimGrid, this option takes the
1249 names of PAPI counters and adds their respective values to the trace
1250 files (See Section :ref:`tracing_tracing_options`).
1254 This feature currently requires superuser privileges, as registers
1255 are queried. Only use this feature with code you trust! Call
1256 smpirun for instance via ``smpirun -wrapper "sudo "
1257 <your-parameters>`` or run ``sudo sh -c "echo 0 >
1258 /proc/sys/kernel/perf_event_paranoid"`` In the later case, sudo
1259 will not be required.
1261 It is planned to make this feature available on a per-process (or per-thread?) basis.
1262 The first draft, however, just implements a "global" (i.e., for all processes) set
1263 of counters, the "default" set.
1265 .. code-block:: none
1267 --cfg=smpi/papi-events:"default:PAPI_L3_LDM:PAPI_L2_LDM"
1269 .. _cfg=smpi/privatization:
1271 Automatic Privatization of Global Variables
1272 ...........................................
1274 **Option** ``smpi/privatization`` **default:** "dlopen" (when using smpirun)
1276 MPI executables are usually meant to be executed in separate
1277 processes, but SMPI is executed in only one process. Global variables
1278 from executables will be placed in the same memory region and shared
1279 between processes, causing intricate bugs. Several options are
1280 possible to avoid this, as described in the main `SMPI publication
1281 <https://hal.inria.fr/hal-01415484>`_ and in the :ref:`SMPI
1282 documentation <SMPI_what_globals>`. SimGrid provides two ways of
1283 automatically privatizing the globals, and this option allows one to
1284 choose between them.
1286 - **no** (default when not using smpirun): Do not automatically
1287 privatize variables. Pass ``-no-privatize`` to smpirun to disable
1289 - **dlopen** or **yes** (default when using smpirun): Link multiple
1290 times against the binary.
1291 - **mmap** (slower, but maybe somewhat more stable):
1292 Runtime automatic switching of the data segments.
1295 This configuration option cannot be set in your platform file. You can only
1296 pass it as an argument to smpirun.
1298 .. _cfg=smpi/privatize-libs:
1300 Automatic privatization of global variables inside external libraries
1301 .....................................................................
1303 **Option** ``smpi/privatize-libs`` **default:** unset
1305 **Linux/BSD only:** When using dlopen (default) privatization,
1306 privatize specific shared libraries with internal global variables, if
1307 they can't be linked statically. For example libgfortran is usually
1308 used for Fortran I/O and indexes in files can be mixed up.
1310 Multiple libraries can be given, semicolon separated.
1312 This configuration option can only use either full paths to libraries,
1313 or full names. Check with ldd the name of the library you want to
1316 .. code-block:: console
1320 libgfortran.so.3 => /usr/lib/x86_64-linux-gnu/libgfortran.so.3 (0x00007fbb4d91b000)
1323 Then you can use ``--cfg=smpi/privatize-libs:libgfortran.so.3``
1324 or ``--cfg=smpi/privatize-libs:/usr/lib/x86_64-linux-gnu/libgfortran.so.3``,
1325 but not ``libgfortran`` nor ``libgfortran.so``.
1327 .. _cfg=smpi/send-is-detached-thresh:
1329 Simulating MPI detached send
1330 ............................
1332 **Option** ``smpi/send-is-detached-thresh`` **default:** 65536
1334 This threshold specifies the size in bytes under which the send will
1335 return immediately. This is different from the threshold detailed in
1336 :ref:`cfg=smpi/async-small-thresh` because the message is not
1337 really sent when the send is posted. SMPI still waits for the
1338 corresponding receive to be posted, in order to perform the communication
1341 .. _cfg=smpi/coll-selector:
1343 Simulating MPI collective algorithms
1344 ....................................
1346 **Option** ``smpi/coll-selector`` **Possible values:** naive (default), ompi, mpich
1348 SMPI implements more than 100 different algorithms for MPI collective
1349 communication, to accurately simulate the behavior of most of the
1350 existing MPI libraries. The ``smpi/coll-selector`` item can be used to
1351 select the decision logic either of the OpenMPI or the MPICH libraries. (By
1352 default SMPI uses naive version of collective operations.)
1354 Each collective operation can be manually selected with a
1355 ``smpi/collective_name:algo_name``. Available algorithms are listed in
1356 :ref:`SMPI_use_colls`.
1358 .. TODO:: All available collective algorithms will be made available
1359 via the ``smpirun --help-coll`` command.
1361 .. _cfg=smpi/finalization-barrier:
1363 Add a barrier in MPI_Finalize
1364 .............................
1366 **Option** ``smpi/finalization-barrier`` **default:** off
1368 By default, SMPI processes are destroyed as soon as soon as their code ends,
1369 so after a successful MPI_Finalize call returns. In some rare cases, some data
1370 might have been attached to MPI objects still active in the remaining processes,
1371 and can be destroyed eagerly by the finished process.
1372 If your code shows issues at finalization, such as segmentation fault, triggering
1373 this option will add an explicit MPI_Barrier(MPI_COMM_WORLD) call inside the
1374 MPI_Finalize, so that all processes will terminate at almost the same point.
1375 It might affect the total timing by the cost of a barrier.
1377 .. _cfg=smpi/errors-are-fatal:
1379 Disable MPI fatal errors
1380 ........................
1382 **Option** ``smpi/errors-are-fatal`` **default:** on
1384 By default, SMPI processes will crash if a MPI error code is returned. MPI allows
1385 to explicitely set MPI_ERRORS_RETURN errhandler to avoid this behaviour. This flag
1386 will turn on this behaviour by default (for all concerned types and errhandlers).
1387 This can ease debugging by going after the first reported error.
1389 .. _cfg=smpi/pedantic:
1391 Disable pedantic MPI errors
1392 ...........................
1394 **Option** ``smpi/pedantic`` **default:** on
1396 By default, SMPI will report all errors it finds in MPI codes. Some of these errors
1397 may not be considered as errors by all developers. This flag can be turned off to
1398 avoid reporting some usually harmless mistakes.
1399 Concerned errors list (will be expanded in the future):
1401 - Calling MPI_Win_fence only once in a program, hence just opening an epoch without
1404 .. _cfg=smpi/iprobe:
1406 Inject constant times for MPI_Iprobe
1407 ....................................
1409 **Option** ``smpi/iprobe`` **default:** 0.0001
1411 The behavior and motivation for this configuration option is identical
1412 with :ref:`smpi/test <cfg=smpi/test>`, but for the function
1415 .. _cfg=smpi/iprobe-cpu-usage:
1417 Reduce speed for iprobe calls
1418 .............................
1420 **Option** ``smpi/iprobe-cpu-usage`` **default:** 1 (no change)
1422 MPI_Iprobe calls can be heavily used in applications. To account
1423 correctly for the energy that cores spend probing, it is necessary to
1424 reduce the load that these calls cause inside SimGrid.
1426 For instance, we measured a maximum power consumption of 220 W for a
1427 particular application but only 180 W while this application was
1428 probing. Hence, the correct factor that should be passed to this
1429 option would be 180/220 = 0.81.
1433 Inject constant times for MPI_Init
1434 ..................................
1436 **Option** ``smpi/init`` **default:** 0
1438 The behavior and motivation for this configuration option is identical
1439 with :ref:`smpi/test <cfg=smpi/test>`, but for the function ``MPI_Init()``.
1443 Inject constant times for MPI_Isend()
1444 .....................................
1446 **Option** ``smpi/ois``
1448 The behavior and motivation for this configuration option is identical
1449 with :ref:`smpi/os <cfg=smpi/os>`, but for the function ``MPI_Isend()``.
1453 Inject constant times for MPI_send()
1454 ....................................
1456 **Option** ``smpi/os``
1458 In several network models such as LogP, send (MPI_Send, MPI_Isend) and
1459 receive (MPI_Recv) operations incur costs (i.e., they consume CPU
1460 time). SMPI can factor these costs in as well, but the user has to
1461 configure SMPI accordingly as these values may vary by machine. This
1462 can be done by using ``smpi/os`` for MPI_Send operations; for MPI_Isend
1463 and MPI_Recv, use ``smpi/ois`` and ``smpi/or``, respectively. These work
1464 exactly as ``smpi/ois``.
1466 This item can consist of multiple sections; each section takes three
1467 values, for example ``1:3:2;10:5:1``. The sections are divided by ";"
1468 so this example contains two sections. Furthermore, each section
1469 consists of three values.
1471 1. The first value denotes the minimum size in bytes for this section to take effect;
1472 read it as "if message size is greater than this value (and other section has a larger
1473 first value that is also smaller than the message size), use this".
1474 In the first section above, this value is "1".
1476 2. The second value is the startup time; this is a constant value that will always
1477 be charged, no matter what the size of the message. In the first section above,
1480 3. The third value is the `per-byte` cost. That is, it is charged for every
1481 byte of the message (incurring cost messageSize*cost_per_byte)
1482 and hence accounts also for larger messages. In the first
1483 section of the example above, this value is "2".
1485 Now, SMPI always checks which section it should use for a given
1486 message; that is, if a message of size 11 is sent with the
1487 configuration of the example above, only the second section will be
1488 used, not the first, as the first value of the second section is
1489 closer to the message size. Hence, when ``smpi/os=1:3:2;10:5:1``, a
1490 message of size 11 incurs the following cost inside MPI_Send:
1491 ``5+11*1`` because 5 is the startup cost and 1 is the cost per byte.
1493 Note that the order of sections can be arbitrary; they will be ordered internally.
1497 Inject constant times for MPI_Recv()
1498 ....................................
1500 **Option** ``smpi/or``
1502 The behavior and motivation for this configuration option is identical
1503 with :ref:`smpi/os <cfg=smpi/os>`, but for the function ``MPI_Recv()``.
1506 .. _cfg=smpi/grow-injected-times:
1508 Inject constant times for MPI_Test
1509 ..................................
1511 **Option** ``smpi/test`` **default:** 0.0001
1513 By setting this option, you can control the amount of time a process
1514 sleeps when MPI_Test() is called; this is important, because SimGrid
1515 normally only advances the time while communication is happening and
1516 thus, MPI_Test will not add to the time, resulting in deadlock if it is
1517 used as a break-condition as in the following example:
1522 MPI_Test(request, flag, status);
1526 To speed up execution, we use a counter to keep track of how often we
1527 checked if the handle is now valid or not. Hence, we actually
1528 use counter*SLEEP_TIME, that is, the time MPI_Test() causes the
1529 process to sleep increases linearly with the number of previously
1530 failed tests. This behavior can be disabled by setting
1531 ``smpi/grow-injected-times`` to **no**. This will also disable this
1532 behavior for MPI_Iprobe.
1534 .. _cfg=smpi/shared-malloc:
1535 .. _cfg=smpi/shared-malloc-hugepage:
1540 **Option** ``smpi/shared-malloc`` **Possible values:** global (default), local
1542 If your simulation consumes too much memory, you may want to modify
1543 your code so that the working areas are shared by all MPI ranks. For
1544 example, in a block-cyclic matrix multiplication, you will only
1545 allocate one set of blocks, and all processes will share them.
1546 Naturally, this will lead to very wrong results, but this will save a
1547 lot of memory. So this is still desirable for some studies. For more on
1548 the motivation for that feature, please refer to the `relevant section
1549 <https://simgrid.github.io/SMPI_CourseWare/topic_understanding_performance/matrixmultiplication>`_
1550 of the SMPI CourseWare (see Activity #2.2 of the pointed
1551 assignment). In practice, change the calls for malloc() and free() into
1552 SMPI_SHARED_MALLOC() and SMPI_SHARED_FREE().
1554 SMPI provides two algorithms for this feature. The first one, called
1555 ``local``, allocates one block per call to SMPI_SHARED_MALLOC()
1556 (each call site gets its own block) ,and this block is shared
1557 among all MPI ranks. This is implemented with the shm_* functions
1558 to create a new POSIX shared memory object (kept in RAM, in /dev/shm)
1559 for each shared block.
1561 With the ``global`` algorithm, each call to SMPI_SHARED_MALLOC()
1562 returns a new address, but it only points to a shadow block: its memory
1563 area is mapped on a 1 MiB file on disk. If the returned block is of size
1564 N MiB, then the same file is mapped N times to cover the whole block.
1565 At the end, no matter how many times you call SMPI_SHARED_MALLOC, this will
1566 only consume 1 MiB in memory.
1568 You can disable this behavior and come back to regular mallocs (for
1569 example for debugging purposes) using ``no`` as a value.
1571 If you want to keep private some parts of the buffer, for instance if these
1572 parts are used by the application logic and should not be corrupted, you
1573 can use SMPI_PARTIAL_SHARED_MALLOC(size, offsets, offsets_count). For example:
1577 mem = SMPI_PARTIAL_SHARED_MALLOC(500, {27,42 , 100,200}, 2);
1579 This will allocate 500 bytes to mem, such that mem[27..41] and
1580 mem[100..199] are shared while other area remain private.
1582 Then, it can be deallocated by calling SMPI_SHARED_FREE(mem).
1584 When smpi/shared-malloc:global is used, the memory consumption problem
1585 is solved, but it may induce too much load on the kernel's pages table.
1586 In this case, you should use huge pages so that the kernel creates only one
1587 entry per MB of malloced data instead of one entry per 4 kB.
1588 To activate this, you must mount a hugetlbfs on your system and allocate
1589 at least one huge page:
1591 .. code-block:: console
1594 $ sudo mount none /home/huge -t hugetlbfs -o rw,mode=0777
1595 $ sudo sh -c 'echo 1 > /proc/sys/vm/nr_hugepages' # echo more if you need more
1597 Then, you can pass the option
1598 ``--cfg=smpi/shared-malloc-hugepage:/home/huge`` to smpirun to
1599 actually activate the huge page support in shared mallocs.
1601 .. _cfg=smpi/auto-shared-malloc-thresh:
1603 Automatically share allocations
1604 ...............................
1606 **Option** ``smpi/auto-shared-malloc-thresh:`` **Default:** 0 (false)
1607 This value in bytes represents the size above which all allocations
1608 will be "shared" by default (as if they were performed through
1609 SMPI_SHARED_MALLOC macros). Default = 0 = disabled feature.
1610 The value must be carefully chosen to only select data buffers which
1611 will not modify execution path or cause crash if their content is false.
1612 Option :ref:`cfg=smpi/display-allocs` can be used to locate the largest
1613 allocation detected in a run, and provide a good starting threshold.
1614 Note : malloc, calloc and free are overridden by smpicc/cxx by default.
1615 This can cause some troubles if codes are already overriding these. If this
1616 is the case, defining SMPI_NO_OVERRIDE_MALLOC in the compilation flags can
1617 help, but will make this feature unusable.
1621 Inject constant times for MPI_Wtime, gettimeofday and clock_gettime
1622 ...................................................................
1624 **Option** ``smpi/wtime`` **default:** 10 ns
1626 This option controls the amount of (simulated) time spent in calls to
1627 MPI_Wtime(), gettimeofday() and clock_gettime(). If you set this value
1628 to 0, the simulated clock is not advanced in these calls, which leads
1629 to issues if your application contains such a loop:
1633 while(MPI_Wtime() < some_time_bound) {
1634 /* some tests, with no communication nor computation */
1637 When the option smpi/wtime is set to 0, the time advances only on
1638 communications and computations. So the previous code results in an
1639 infinite loop: the current [simulated] time will never reach
1640 ``some_time_bound``. This infinite loop is avoided when that option
1641 is set to a small value, as it is by default since SimGrid v3.21.
1643 Note that if your application does not contain any loop depending on
1644 the current time only, then setting this option to a non-zero value
1645 will slow down your simulations by a tiny bit: the simulation loop has
1646 to be broken out of and reset each time your code asks for the current time.
1647 If the simulation speed really matters to you, you can avoid this
1648 extra delay by setting smpi/wtime to 0.
1650 .. _cfg=smpi/list-leaks:
1652 Report leaked MPI objects
1653 .........................
1655 **Option** ``smpi/list-leaks`` **default:** 0
1657 This option controls whether to report leaked MPI objects.
1658 The parameter is the number of leaks to report.
1660 Other Configurations
1661 --------------------
1663 .. _cfg=debug/clean-atexit:
1665 Cleanup at Termination
1666 ......................
1668 **Option** ``debug/clean-atexit`` **default:** on
1670 If your code is segfaulting during its finalization, it may help to
1671 disable this option to request that SimGrid not attempt any cleanups at
1672 the end of the simulation. Since the Unix process is ending anyway,
1673 the operating system will wipe it all.
1680 **Option** ``path`` **default:** . (current dir)
1682 It is possible to specify a list of directories to search in for the
1683 trace files (see :ref:`pf_trace`) by using this configuration
1684 item. To add several directory to the path, set the configuration
1685 item several times, as in ``--cfg=path:toto --cfg=path:tutu``
1687 .. _cfg=debug/breakpoint:
1692 **Option** ``debug/breakpoint`` **default:** unset
1694 This configuration option sets a breakpoint: when the simulated clock
1695 reaches the given time, a SIGTRAP is raised. This can be used to stop
1696 the execution and get a backtrace with a debugger.
1698 It is also possible to set the breakpoint from inside the debugger, by
1699 writing in global variable simgrid::kernel::cfg_breakpoint. For example,
1702 .. code-block:: none
1704 set variable simgrid::kernel::cfg_breakpoint = 3.1416
1706 .. _cfg=debug/verbose-exit:
1711 **Option** ``debug/verbose-exit`` **default:** on
1713 By default, when Ctrl-C is pressed, the status of all existing actors
1714 is displayed before exiting the simulation. This is very useful to
1715 debug your code, but it can become troublesome if you have many
1716 actors. Set this configuration item to **off** to disable this
1719 .. _cfg=exception/cutpath:
1721 Truncate local path from exception backtrace
1722 ............................................
1724 **Option** ``exception/cutpath`` **default:** off
1726 This configuration option is used to remove the path from the
1727 backtrace shown when an exception is thrown. This is mainly useful for
1728 the tests: the full file path would makes the tests non-reproducible because
1729 the paths of source files depend of the build settings. That would
1730 break most of the tests since their output is continually compared.
1734 Logging configuration
1735 ---------------------
1737 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
1738 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
1739 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
1742 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
1743 messages from your code.
1745 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
1746 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
1747 practice, the following is equivalent to the above settings: ``--log=root.thresh:error --log=s4u_host.thresh:debug``.
1749 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
1750 your settings, as in ``--log="root.thresh:error s4u_host.thresh:debug"``. The parameters are interpreted in order, from left to right.
1753 Threshold configuration
1754 .......................
1756 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
1757 ``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
1758 see, ``threshold`` can be abbreviated here.
1760 Existing thresholds:
1762 - ``trace`` some functions display a message at this level when entering or returning
1763 - ``debug`` output that is mostly useful when debugging the corresponding module.
1764 - ``verbose`` verbose output that is only mildly interesting and can easily be ignored
1765 - ``info`` usual output (this is the default threshold of all categories)
1766 - ``warning`` minor issue encountered
1767 - ``error`` issue encountered
1768 - ``critical`` major issue encountered, such as assertions failures
1772 Format configuration
1773 ....................
1775 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
1776 as the date, or the actor ID, everything. Existing format directives:
1779 - %n: line separator (LOG4J compatible)
1780 - %e: plain old space (SimGrid extension)
1782 - %m: user-provided message
1784 - %c: Category name (LOG4J compatible)
1785 - %p: Priority name (LOG4J compatible)
1787 - %h: Hostname (SimGrid extension)
1788 - %a: Actor name (SimGrid extension -- note that with SMPI this is the integer value of the process rank)
1789 - %i: Actor PID (SimGrid extension -- this is a 'i' as in 'i'dea)
1790 - %t: Thread "name" (LOG4J compatible -- actually the address of the thread in memory)
1792 - %F: file name where the log event was raised (LOG4J compatible)
1793 - %l: location where the log event was raised (LOG4J compatible, like '%%F:%%L' -- this is a l as in 'l'etter)
1794 - %L: line number where the log event was raised (LOG4J compatible)
1795 - %M: function name (LOG4J compatible -- called method name here of course).
1797 - %d: date (UNIX-like epoch)
1798 - %r: application age (time elapsed since the beginning of the application)
1801 ``--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
1802 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
1803 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
1804 provided layout is used for every messages.
1806 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
1810 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
1811 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'"``.
1812 Another option is to use the ``%e`` directive for spaces, as in ``--log=root.fmt:%l:%e[%p/%c]:%e%m%n``.
1817 The keyword ``app`` controls the appended of a logging category. For example ``--log=root.app:file:mylogfile`` redirects every output to the file ``mylogfile``.
1819 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
1820 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.
1822 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``
1823 ensures that the log file ``mylog`` will never overpass 500 bytes in size.
1825 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
1826 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.
1831 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
1832 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
1833 ``on`` (or ``yes`` or ``1``), the produced messages will also be passed to the upper appender.
1835 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
1836 ``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
1837 will only be sent to ``all.log``.
1842 ``--help-logs`` displays a complete help message about logging in SimGrid.
1844 ``--help-log-categories`` displays the actual hierarchy of log categories for this binary.
1846 ``--log=no_loc`` hides the source locations (file names and line numbers) from the messages. This is useful to make tests reproducible.