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:
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 e.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 - **path:** :ref:`cfg=path`
136 - **plugin:** :ref:`cfg=plugin`
138 - **storage/max_file_descriptors:** :ref:`cfg=storage/max_file_descriptors`
140 - **surf/precision:** :ref:`cfg=surf/precision`
142 - **For collective operations of SMPI,** please refer to Section :ref:`cfg=smpi/coll-selector`
143 - **smpi/async-small-thresh:** :ref:`cfg=smpi/async-small-thresh`
144 - **smpi/buffering:** :ref:`cfg=smpi/buffering`
145 - **smpi/bw-factor:** :ref:`cfg=smpi/bw-factor`
146 - **smpi/coll-selector:** :ref:`cfg=smpi/coll-selector`
147 - **smpi/comp-adjustment-file:** :ref:`cfg=smpi/comp-adjustment-file`
148 - **smpi/cpu-threshold:** :ref:`cfg=smpi/cpu-threshold`
149 - **smpi/display-timing:** :ref:`cfg=smpi/display-timing`
150 - **smpi/grow-injected-times:** :ref:`cfg=smpi/grow-injected-times`
151 - **smpi/host-speed:** :ref:`cfg=smpi/host-speed`
152 - **smpi/IB-penalty-factors:** :ref:`cfg=smpi/IB-penalty-factors`
153 - **smpi/iprobe:** :ref:`cfg=smpi/iprobe`
154 - **smpi/iprobe-cpu-usage:** :ref:`cfg=smpi/iprobe-cpu-usage`
155 - **smpi/init:** :ref:`cfg=smpi/init`
156 - **smpi/keep-temps:** :ref:`cfg=smpi/keep-temps`
157 - **smpi/lat-factor:** :ref:`cfg=smpi/lat-factor`
158 - **smpi/ois:** :ref:`cfg=smpi/ois`
159 - **smpi/or:** :ref:`cfg=smpi/or`
160 - **smpi/os:** :ref:`cfg=smpi/os`
161 - **smpi/papi-events:** :ref:`cfg=smpi/papi-events`
162 - **smpi/privatization:** :ref:`cfg=smpi/privatization`
163 - **smpi/privatize-libs:** :ref:`cfg=smpi/privatize-libs`
164 - **smpi/send-is-detached-thresh:** :ref:`cfg=smpi/send-is-detached-thresh`
165 - **smpi/shared-malloc:** :ref:`cfg=smpi/shared-malloc`
166 - **smpi/shared-malloc-hugepage:** :ref:`cfg=smpi/shared-malloc-hugepage`
167 - **smpi/simulate-computation:** :ref:`cfg=smpi/simulate-computation`
168 - **smpi/test:** :ref:`cfg=smpi/test`
169 - **smpi/wtime:** :ref:`cfg=smpi/wtime`
171 - **Tracing configuration options** can be found in Section :ref:`tracing_tracing_options`
173 - **storage/model:** :ref:`options_model_select`
175 - **vm/model:** :ref:`options_model_select`
179 Configuring the Platform Models
180 -------------------------------
182 .. _options_model_select:
184 Choosing the Platform Models
185 ............................
187 SimGrid comes with several network, CPU and disk models built in,
188 and you can change the used model at runtime by changing the passed
189 configuration. The three main configuration items are given below.
190 For each of these items, passing the special ``help`` value gives you
191 a short description of all possible values (for example,
192 ``--cfg=network/model:help`` will present all provided network
193 models). Also, ``--help-models`` should provide information about all
194 models for all existing resources.
196 - ``network/model``: specify the used network model. Possible values:
198 - **LV08 (default one):** Realistic network analytic model
199 (slow-start modeled by multiplying latency by 13.01, bandwidth by
200 .97; bottleneck sharing uses a payload of S=20537 for evaluating
201 RTT). Described in `Accuracy Study and Improvement of Network
202 Simulation in the SimGrid Framework
203 <http://mescal.imag.fr/membres/arnaud.legrand/articles/simutools09.pdf>`_.
204 - **Constant:** Simplistic network model where all communication
205 take a constant time (one second). This model provides the lowest
206 realism, but is (marginally) faster.
207 - **SMPI:** Realistic network model specifically tailored for HPC
208 settings (accurate modeling of slow start with correction factors on
209 three intervals: < 1KiB, < 64 KiB, >= 64 KiB). This model can be
210 :ref:`further configured <options_model_network>`.
211 - **IB:** Realistic network model specifically tailored for HPC
212 settings with InfiniBand networks (accurate modeling contention
213 behavior, based on the model explained in `this PhD work
214 <http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf>`_.
215 This model can be :ref:`further configured <options_model_network>`.
216 - **CM02:** Legacy network analytic model. Very similar to LV08, but
217 without corrective factors. The timings of small messages are thus
218 poorly modeled. This model is described in `A Network Model for
219 Simulation of Grid Application
220 <https://hal.inria.fr/inria-00071989/document>`_.
221 - **ns-3** (only available if you compiled SimGrid accordingly):
222 Use the packet-level network
223 simulators as network models (see :ref:`model_ns3`).
224 This model can be :ref:`further configured <options_pls>`.
226 - ``cpu/model``: specify the used CPU model. We have only one model
229 - **Cas01:** Simplistic CPU model (time=size/speed)
231 - ``host/model``: The host concept is the aggregation of a CPU with a
232 network card. Three models exists, but actually, only 2 of them are
233 interesting. The "compound" one is simply due to the way our
234 internal code is organized, and can easily be ignored. So at the
235 end, you have two host models: The default one allows aggregation of
236 an existing CPU model with an existing network model, but does not
237 allow parallel tasks because these beasts need some collaboration
238 between the network and CPU model. That is why, ptask_07 is used by
239 default when using SimDag.
241 - **default:** Default host model. Currently, CPU:Cas01 and
242 network:LV08 (with cross traffic enabled)
243 - **compound:** Host model that is automatically chosen if
244 you change the network and CPU models
245 - **ptask_L07:** Host model somehow similar to Cas01+CM02 but
246 allowing "parallel tasks", that are intended to model the moldable
247 tasks of the grid scheduling literature.
249 - ``storage/model``: specify the used storage model. Only one model is
251 - ``vm/model``: specify the model for virtual machines. Only one model
254 .. todo: make 'compound' the default host model.
256 .. _options_model_optim:
261 The network and CPU models that are based on lmm_solve (that
262 is, all our analytical models) accept specific optimization
265 - items ``network/optim`` and ``cpu/optim`` (both default to 'Lazy'):
267 - **Lazy:** Lazy action management (partial invalidation in lmm +
268 heap in action remaining).
269 - **TI:** Trace integration. Highly optimized mode when using
270 availability traces (only available for the Cas01 CPU model for
272 - **Full:** Full update of remaining and variables. Slow but may be
273 useful when debugging.
275 - items ``network/maxmin-selective-update`` and
276 ``cpu/maxmin-selective-update``: configure whether the underlying
277 should be lazily updated or not. It should have no impact on the
278 computed timings, but should speed up the computation. |br| It is
279 still possible to disable this feature because it can reveal
280 counter-productive in very specific scenarios where the
281 interaction level is high. In particular, if all your
282 communication share a given backbone link, you should disable it:
283 without it, a simple regular loop is used to update each
284 communication. With it, each of them is still updated (because of
285 the dependency induced by the backbone), but through a complicated
286 and slow pattern that follows the actual dependencies.
288 .. _cfg=maxmin/precision:
289 .. _cfg=surf/precision:
294 **Option** ``maxmin/precision`` **Default:** 0.00001 (in flops or bytes) |br|
295 **Option** ``surf/precision`` **Default:** 0.00001 (in seconds)
297 The analytical models handle a lot of floating point values. It is
298 possible to change the epsilon used to update and compare them through
299 this configuration item. Changing it may speedup the simulation by
300 discarding very small actions, at the price of a reduced numerical
301 precision. You can modify separately the precision used to manipulate
302 timings (in seconds) and the one used to manipulate amounts of work
305 .. _cfg=maxmin/concurrency-limit:
310 **Option** ``maxmin/concurrency-limit`` **Default:** -1 (no limit)
312 The maximum number of variables per resource can be tuned through this
313 option. You can have as many simultaneous actions per resources as you
314 want. If your simulation presents a very high level of concurrency, it
315 may help to use e.g. 100 as a value here. It means that at most 100
316 actions can consume a resource at a given time. The extraneous actions
317 are queued and wait until the amount of concurrency of the considered
318 resource lowers under the given boundary.
320 Such limitations help both to the simulation speed and simulation accuracy
321 on highly constrained scenarios, but the simulation speed suffers of this
322 setting on regular (less constrained) scenarios so it is off by default.
324 .. _options_model_network:
326 Configuring the Network Model
327 .............................
329 .. _cfg=network/TCP-gamma:
331 Maximal TCP Window Size
332 ^^^^^^^^^^^^^^^^^^^^^^^
334 **Option** ``network/TCP-gamma`` **Default:** 4194304
336 The analytical models need to know the maximal TCP window size to take
337 the TCP congestion mechanism into account. On Linux, this value can
338 be retrieved using the following commands. Both give a set of values,
339 and you should use the last one, which is the maximal size.
341 .. code-block:: shell
343 cat /proc/sys/net/ipv4/tcp_rmem # gives the sender window
344 cat /proc/sys/net/ipv4/tcp_wmem # gives the receiver window
346 .. _cfg=smpi/IB-penalty-factors:
347 .. _cfg=network/bandwidth-factor:
348 .. _cfg=network/latency-factor:
349 .. _cfg=network/weight-S:
351 Correcting Important Network Parameters
352 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
354 SimGrid can take network irregularities such as a slow startup or
355 changing behavior depending on the message size into account. You
356 should not change these values unless you really know what you're
357 doing. The corresponding values were computed through data fitting
358 one the timings of packet-level simulators, as described in `Accuracy
359 Study and Improvement of Network Simulation in the SimGrid Framework
360 <http://mescal.imag.fr/membres/arnaud.legrand/articles/simutools09.pdf>`_.
363 If you are using the SMPI model, these correction coefficients are
364 themselves corrected by constant values depending on the size of the
365 exchange. By default SMPI uses factors computed on the Stampede
366 Supercomputer at TACC, with optimal deployment of processes on
367 nodes. Again, only hardcore experts should bother about this fact.
369 InfiniBand network behavior can be modeled through 3 parameters
370 ``smpi/IB-penalty-factors:"βe;βs;γs"``, as explained in `this PhD
372 <http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf>`_.
374 .. todo:: This section should be rewritten, and actually explain the
375 options network/bandwidth-factor, network/latency-factor,
378 .. _cfg=network/crosstraffic:
380 Simulating Cross-Traffic
381 ^^^^^^^^^^^^^^^^^^^^^^^^
383 Since SimGrid v3.7, cross-traffic effects can be taken into account in
384 analytical simulations. It means that ongoing and incoming
385 communication flows are treated independently. In addition, the LV08
386 model adds 0.05 of usage on the opposite direction for each new
387 created flow. This can be useful to simulate some important TCP
388 phenomena such as ack compression.
390 For that to work, your platform must have two links for each
391 pair of interconnected hosts. An example of usable platform is
392 available in ``examples/platforms/crosstraffic.xml``.
394 This is activated through the ``network/crosstraffic`` item, that
395 can be set to 0 (disable this feature) or 1 (enable it).
397 Note that with the default host model this option is activated by default.
399 .. _cfg=network/loopback:
401 Configuring loopback link
402 ^^^^^^^^^^^^^^^^^^^^^^^^^
404 Several network model provide an implicit loopback link to account for local
405 communication on a host. By default it has a 10GBps bandwidth and a null latency.
406 This can be changed with ``network/loopback-lat`` and ``network/loopback-bw``
409 .. _cfg=smpi/async-small-thresh:
411 Simulating Asynchronous Send
412 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
414 (this configuration item is experimental and may change or disappear)
416 It is possible to specify that messages below a certain size (in bytes) will be
417 sent as soon as the call to MPI_Send is issued, without waiting for
418 the correspondant receive. This threshold can be configured through
419 the ``smpi/async-small-thresh`` item. The default value is 0. This
420 behavior can also be manually set for mailboxes, by setting the
421 receiving mode of the mailbox with a call to
422 :cpp:func:`MSG_mailbox_set_async`. After this, all messages sent to
423 this mailbox will have this behavior regardless of the message size.
425 This value needs to be smaller than or equals to the threshold set at
426 :ref:`cfg=smpi/send-is-detached-thresh`, because asynchronous messages
427 are meant to be detached as well.
434 **Option** ``ns3/TcpModel`` **Default:** "default" (ns-3 default)
436 When using ns-3, there is an extra item ``ns3/TcpModel``, corresponding
437 to the ``ns3::TcpL4Protocol::SocketType`` configuration item in
438 ns-3. The only valid values (enforced on the SimGrid side) are
439 'default' (no change to the ns-3 configuration), 'NewReno' or 'Reno' or
442 Configuring the Storage model
443 .............................
445 .. _cfg=storage/max_file_descriptors:
447 File Descriptor Cound per Host
448 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
450 **Option** ``storage/max_file_descriptors`` **Default:** 1024
452 Each host maintains a fixed-size array of its file descriptors. You
453 can change its size through this item to either enlarge it if your
454 application requires it or to reduce it to save memory space.
461 SimGrid plugins allow one to extend the framework without changing its
462 source code directly. Read the source code of the existing plugins to
463 learn how to do so (in ``src/plugins``), and ask your questions to the
464 usual channels (Stack Overflow, Mailing list, IRC). The basic idea is
465 that plugins usually register callbacks to some signals of interest.
466 If they need to store some information about a given object (Link, CPU
467 or Actor), they do so through the use of a dedicated object extension.
469 Some of the existing plugins can be activated from the command line,
470 meaning that you can activate them from the command line without any
471 modification to your simulation code. For example, you can activate
472 the host energy plugin by adding ``--cfg=plugin:host_energy`` to your
475 Here is a partial list of plugins that can be activated this way. You can get
476 the full list by passing ``--cfg=plugin:help`` to your simulator.
478 - :ref:`Host Energy <plugin_host_energy>`: models the energy dissipation of the compute units.
479 - :ref:`Link Energy <plugin_link_energy>`: models the energy dissipation of the network.
480 - :ref:`Host Load <plugin_host_load>`: monitors the load of the compute units.
482 .. _options_modelchecking:
484 Configuring the Model-Checking
485 ------------------------------
487 To enable SimGrid's model-checking support, the program should
488 be executed using the simgrid-mc wrapper:
490 .. code-block:: shell
492 simgrid-mc ./my_program
494 Safety properties are expressed as assertions using the function
495 :cpp:func:`void MC_assert(int prop)`.
497 .. _cfg=smpi/buffering:
499 Specifying the MPI buffering behavior
500 .....................................
502 **Option** ``smpi/buffering`` **Default:** infty
504 Buffering in MPI has a huge impact on the communication semantic. For example,
505 standard blocking sends are synchronous calls when the system buffers are full
506 while these calls can complete immediately without even requiring a matching
507 receive call for small messages sent when the system buffers are empty.
509 In SMPI, this depends on the message size, that is compared against two thresholds:
511 - if (size < :ref:`smpi/async-small-thresh <cfg=smpi/async-small-thresh>`) then
512 MPI_Send returns immediately, even if the corresponding receive has not be issued yet.
513 - 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
514 MPI_Send returns as soon as the corresponding receive has been issued. This is known as the eager mode.
515 - if (:ref:`smpi/send-is-detached-thresh <cfg=smpi/send-is-detached-thresh>` < size) then
516 MPI_Send returns only when the message has actually been sent over the network. This is known as the rendez-vous mode.
518 The ``smpi/buffering`` (only valid with MC) option gives an easier interface to choose between these semantics. It can take two values:
520 - **zero:** means that buffering should be disabled. All communications are actually blocking.
521 - **infty:** means that buffering should be made infinite. All communications are non-blocking.
523 .. _cfg=model-check/property:
525 Specifying a liveness property
526 ..............................
528 **Option** ``model-check/property`` **Default:** unset
530 If you want to specify liveness properties, you have to pass them on
531 the command line, specifying the name of the file containing the
532 property, as formatted by the `ltl2ba <https://github.com/utwente-fmt/ltl2ba>`_ program.
533 Note that ltl2ba is not part of SimGrid and must be installed separatly.
535 .. code-block:: shell
537 simgrid-mc ./my_program --cfg=model-check/property:<filename>
539 .. _cfg=model-check/checkpoint:
541 Going for Stateful Verification
542 ...............................
544 By default, the system is backtracked to its initial state to explore
545 another path, instead of backtracking to the exact step before the fork
546 that we want to explore (this is called stateless verification). This
547 is done this way because saving intermediate states can rapidly
548 exhaust the available memory. If you want, you can change the value of
549 the ``model-check/checkpoint`` item. For example,
550 ``--cfg=model-check/checkpoint:1`` asks to take a checkpoint every
551 step. Beware, this will certainly explode your memory. Larger values
552 are probably better, make sure to experiment a bit to find the right
553 setting for your specific system.
555 .. _cfg=model-check/reduction:
557 Specifying the kind of reduction
558 ................................
560 The main issue when using the model-checking is the state space
561 explosion. You can activate some reduction technique with
562 ``--cfg=model-check/reduction:<technique>``. For now, this
563 configuration variable can take 2 values:
565 - **none:** Do not apply any kind of reduction (mandatory for
566 liveness properties, as our current DPOR algorithm breaks cycles)
567 - **dpor:** Apply Dynamic Partial Ordering Reduction. Only valid if
568 you verify local safety properties (default value for safety
571 Another way to mitigate the state space explosion is to search for
572 cycles in the exploration with the :ref:`cfg=model-check/visited`
573 configuration. Note that DPOR and state-equality reduction may not
574 play well together. You should choose between them.
576 Our current DPOR implementation could be improved in may ways. We are
577 currently improving its efficiency (both in term of reduction ability
578 and computational speed), and future work could make it compatible
579 with liveness properties.
581 .. _cfg=model-check/visited:
583 Size of Cycle Detection Set (state equality reduction)
584 ......................................................
586 Mc SimGrid can be asked to search for cycles during the exploration,
587 i.e. situations where a new explored state is in fact the same state
588 than a previous one.. This can prove useful to mitigate the state
589 space explosion with safety properties, and this is the crux when
590 searching for counter-examples to the liveness properties.
592 Note that this feature may break the current implementation of the
593 DPOR reduction technique.
595 The ``model-check/visited`` item is the maximum number of states, which
596 are stored in memory. If the maximum number of snapshotted state is
597 reached, some states will be removed from the memory and some cycles
598 might be missed. Small values can lead to incorrect verifications, but
599 large values can exhaust your memory and be CPU intensive as each new
600 state must be compared to that amount of older saved states.
602 The default settings depend on the kind of exploration. With safety
603 checking, no state is snapshotted and cycles cannot be detected. With
604 liveness checking, all states are snapshotted because missing a cycle
605 could hinder the exploration soundness.
607 .. _cfg=model-check/termination:
609 Non-Termination Detection
610 .........................
612 The ``model-check/termination`` configuration item can be used to
613 report if a non-termination execution path has been found. This is a
614 path with a cycle, which means that the program might never terminate.
616 This only works in safety mode, not in liveness mode.
618 This options is disabled by default.
620 .. _cfg=model-check/dot-output:
625 If set, the ``model-check/dot-output`` configuration item is the name
626 of a file in which to write a dot file of the path leading to the
627 property violation discovered (safety or liveness violation), as well
628 as the cycle for liveness properties. This dot file can then be fed to the
629 graphviz dot tool to generate an corresponding graphical representation.
631 .. _cfg=model-check/max-depth:
633 Exploration Depth Limit
634 .......................
636 The ``model-checker/max-depth`` can set the maximum depth of the
637 exploration graph of the model checker. If this limit is reached, a
638 logging message is sent and the results might not be exact.
640 By default, there is no depth limit.
642 .. _cfg=model-check/timeout:
647 By default, the model checker does not handle timeout conditions: the `wait`
648 operations never time out. With the ``model-check/timeout`` configuration item
649 set to **yes**, the model checker will explore timeouts of `wait` operations.
651 .. _cfg=model-check/communications-determinism:
652 .. _cfg=model-check/send-determinism:
654 Communication Determinism
655 .........................
657 The ``model-check/communications-determinism`` and
658 ``model-check/send-determinism`` items can be used to select the
659 communication determinism mode of the model checker, which checks
660 determinism properties of the communications of an application.
664 Verification Performance Considerations
665 .......................................
667 The size of the stacks can have a huge impact on the memory
668 consumption when using model-checking. By default, each snapshot will
669 save a copy of the whole stacks and not only of the part that is
670 really meaningful: you should expect the contribution of the memory
671 consumption of the snapshots to be:
672 :math:`\text{number of processes} \times \text{stack size} \times \text{number of states}`.
674 When compiled against the model checker, the stacks are not
675 protected with guards: if the stack size is too small for your
676 application, the stack will silently overflow into other parts of the
677 memory (see :ref:`contexts/guard-size <cfg=contexts/guard-size>`).
679 .. _cfg=model-check/replay:
681 Replaying buggy execution paths from the model checker
682 ......................................................
684 Debugging the problems reported by the model checker is challenging:
685 First, the application under verification cannot be debugged with gdb
686 because the model checker already traces it. Then, the model checker may
687 explore several execution paths before encountering the issue, making it
688 very difficult to understand the output. Fortunately, SimGrid provides
689 the execution path leading to any reported issue so that you can replay
690 this path reported by the model checker, enabling the usage of classical
693 When the model checker finds an interesting path in the application
694 execution graph (where a safety or liveness property is violated), it
695 generates an identifier for this path. Here is an example of the output:
697 .. code-block:: shell
699 [ 0.000000] (0:@) Check a safety property
700 [ 0.000000] (0:@) **************************
701 [ 0.000000] (0:@) *** PROPERTY NOT VALID ***
702 [ 0.000000] (0:@) **************************
703 [ 0.000000] (0:@) Counter-example execution trace:
704 [ 0.000000] (0:@) [(1)Tremblay (app)] MC_RANDOM(3)
705 [ 0.000000] (0:@) [(1)Tremblay (app)] MC_RANDOM(4)
706 [ 0.000000] (0:@) Path = 1/3;1/4
707 [ 0.000000] (0:@) Expanded states = 27
708 [ 0.000000] (0:@) Visited states = 68
709 [ 0.000000] (0:@) Executed transitions = 46
711 The interesting line is ``Path = 1/3;1/4``, which means that you should use
712 ``--cfg=model-check/replay:1/3;1/4`` to replay your application on the buggy
713 execution path. All options (but the model checker related ones) must
714 remain the same. In particular, if you ran your application with
715 ``smpirun -wrapper simgrid-mc``, then do it again. Remove all
716 MC-related options, keep non-MC-related ones and add
717 ``--cfg=model-check/replay:???``.
719 Currently, if the path is of the form ``X;Y;Z``, each number denotes
720 the actor's pid that is selected at each indecision point. If it's of
721 the form ``X/a;Y/b``, the X and Y are the selected pids while the a
722 and b are the return values of their simcalls. In the previouse
723 example, ``1/3;1/4``, you can see from the full output that the actor
724 1 is doing MC_RANDOM simcalls, so the 3 and 4 simply denote the values
725 that these simcall return.
727 Configuring the User Code Virtualization
728 ----------------------------------------
730 .. _cfg=contexts/factory:
732 Selecting the Virtualization Factory
733 ....................................
735 **Option** contexts/factory **Default:** "raw"
737 In SimGrid, the user code is virtualized in a specific mechanism that
738 allows the simulation kernel to control its execution: when a user
739 process requires a blocking action (such as sending a message), it is
740 interrupted, and only gets released when the simulated clock reaches
741 the point where the blocking operation is done. This is explained
742 graphically in the `relevant tutorial, available online
743 <https://simgrid.org/tutorials/simgrid-simix-101.pdf>`_.
745 In SimGrid, the containers in which user processes are virtualized are
746 called contexts. Several context factory are provided, and you can
747 select the one you want to use with the ``contexts/factory``
748 configuration item. Some of the following may not exist on your
749 machine because of portability issues. In any case, the default one
750 should be the most effcient one (please report bugs if the
751 auto-detection fails for you). They are approximately sorted here from
752 the slowest to the most efficient:
754 - **thread:** very slow factory using full featured threads (either
755 pthreads or windows native threads). They are slow but very
756 standard. Some debuggers or profilers only work with this factory.
757 - **java:** Java applications are virtualized onto java threads (that
758 are regular pthreads registered to the JVM)
759 - **ucontext:** fast factory using System V contexts (Linux and FreeBSD only)
760 - **boost:** This uses the `context
761 implementation <http://www.boost.org/doc/libs/1_59_0/libs/context/doc/html/index.html>`_
762 of the boost library for a performance that is comparable to our
764 |br| Install the relevant library (e.g. with the
765 libboost-contexts-dev package on Debian/Ubuntu) and recompile
767 - **raw:** amazingly fast factory using a context switching mechanism
768 of our own, directly implemented in assembly (only available for x86
769 and amd64 platforms for now) and without any unneeded system call.
771 The main reason to change this setting is when the debugging tools become
772 fooled by the optimized context factories. Threads are the most
773 debugging-friendly contexts, as they allow one to set breakpoints
774 anywhere with gdb and visualize backtraces for all processes, in order
775 to debug concurrency issues. Valgrind is also more comfortable with
776 threads, but it should be usable with all factories (Exception: the
777 callgrind tool really dislikes raw and ucontext factories).
779 .. _cfg=contexts/stack-size:
781 Adapting the Stack Size
782 .......................
784 **Option** ``contexts/stack-size`` **Default:** 8192 KiB
786 Each virtualized used process is executed using a specific system
787 stack. The size of this stack has a huge impact on the simulation
788 scalability, but its default value is rather large. This is because
789 the error messages that you get when the stack size is too small are
790 rather disturbing: this leads to stack overflow (overwriting other
791 stacks), leading to segfaults with corrupted stack traces.
793 If you want to push the scalability limits of your code, you might
794 want to reduce the ``contexts/stack-size`` item. Its default value is
795 8192 (in KiB), while our Chord simulation works with stacks as small
796 as 16 KiB, for example. You can ensure that some actors have a specific
797 size by simply changing the value of this configuration item before
798 creating these actors. The :cpp:func:`simgrid::s4u::Engine::set_config`
799 functions are handy for that.
801 This *setting is ignored* when using the thread factory (because there
802 is no way to modify the stack size with C++ system threads). Instead,
803 you should compile SimGrid and your application with
804 ``-fsplit-stack``. Note that this compilation flag is not compatible
805 with the model checker right now.
807 The operating system should only allocate memory for the pages of the
808 stack which are actually used and you might not need to use this in
809 most cases. However, this setting is very important when using the
810 model checker (see :ref:`options_mc_perf`).
812 .. _cfg=contexts/guard-size:
814 Disabling Stack Guard Pages
815 ...........................
817 **Option** ``contexts/guard-size`` **Default** 1 page in most case (0 pages on Windows or with MC)
819 Unless you use the threads context factory (see
820 :ref:`cfg=contexts/factory`), a stack guard page is usually used
821 which prevents the stack of a given actor from overflowing on another
822 stack. But the performance impact may become prohibitive when the
823 amount of actors increases. The option ``contexts/guard-size`` is the
824 number of stack guard pages used. By setting it to 0, no guard pages
825 will be used: in this case, you should avoid using small stacks (with
826 :ref:`contexts/stack-size <cfg=contexts/stack-size>`) as the stack
827 will silently overflow on other parts of the memory.
829 When no stack guard page is created, stacks may then silently overflow
830 on other parts of the memory if their size is too small for the
833 .. _cfg=contexts/nthreads:
834 .. _cfg=contexts/synchro:
836 Running User Code in Parallel
837 .............................
839 Parallel execution of the user code is only considered stable in
840 SimGrid v3.7 and higher, and mostly for MSG simulations. SMPI
841 simulations may well fail in parallel mode. It is described in
842 `INRIA RR-7653 <http://hal.inria.fr/inria-00602216/>`_.
844 If you are using the **ucontext** or **raw** context factories, you can
845 request to execute the user code in parallel. Several threads are
846 launched, each of them handling the same number of user contexts at each
847 run. To activate this, set the ``contexts/nthreads`` item to the amount
848 of cores that you have in your computer (or lower than 1 to have the
849 amount of cores auto-detected).
851 When parallel execution is activated, you can choose the
852 synchronization schema used with the ``contexts/synchro`` item,
853 which value is either:
855 - **futex:** ultra optimized synchronisation schema, based on futexes
856 (fast user-mode mutexes), and thus only available on Linux systems.
857 This is the default mode when available.
858 - **posix:** slow but portable synchronisation using only POSIX
860 - **busy_wait:** not really a synchronisation: the worker threads
861 constantly request new contexts to execute. It should be the most
862 efficient synchronisation schema, but it loads all the cores of
863 your machine for no good reason. You probably prefer the other less
866 Configuring the Tracing
867 -----------------------
869 The :ref:`tracing subsystem <outcomes_vizu>` can be configured in
870 several different ways depending on the nature of the simulator (MSG,
871 SimDag, SMPI) and the kind of traces that need to be obtained. See the
872 :ref:`Tracing Configuration Options subsection
873 <tracing_tracing_options>` to get a detailed description of each
874 configuration option.
876 We detail here a simple way to get the traces working for you, even if
877 you never used the tracing API.
880 - Any SimGrid-based simulator (MSG, SimDag, SMPI, ...) and raw traces:
882 .. code-block:: shell
884 --cfg=tracing:yes --cfg=tracing/uncategorized:yes
886 The first parameter activates the tracing subsystem, and the second
887 tells it to trace host and link utilization (without any
890 - MSG or SimDag-based simulator and categorized traces (you need to
891 declare categories and classify your tasks according to them)
893 .. code-block:: shell
895 --cfg=tracing:yes --cfg=tracing/categorized:yes
897 The first parameter activates the tracing subsystem, and the second
898 tells it to trace host and link categorized utilization.
900 - SMPI simulator and traces for a space/time view:
902 .. code-block:: shell
906 The `-trace` parameter for the smpirun script runs the simulation
907 with ``--cfg=tracing:yes --cfg=tracing/smpi:yes``. Check the
908 smpirun's `-help` parameter for additional tracing options.
910 Sometimes you might want to put additional information on the trace to
911 correctly identify them later, or to provide data that can be used to
912 reproduce an experiment. You have two ways to do that:
914 - Add a string on top of the trace file as comment:
916 .. code-block:: shell
918 --cfg=tracing/comment:my_simulation_identifier
920 - Add the contents of a textual file on top of the trace file as comment:
922 .. code-block:: shell
924 --cfg=tracing/comment-file:my_file_with_additional_information.txt
926 Please, use these two parameters (for comments) to make reproducible
927 simulations. For additional details about this and all tracing
928 options, check See the :ref:`tracing_tracing_options`.
933 .. _cfg=msg/debug-multiple-use:
938 **Option** ``msg/debug-multiple-use`` **Default:** off
940 Sometimes your application may try to send a task that is still being
941 executed somewhere else, making it impossible to send this task. However,
942 for debugging purposes, one may want to know what the other host is/was
943 doing. This option shows a backtrace of the other process.
948 The SMPI interface provides several specific configuration items.
949 These are not easy to see, since the code is usually launched through the
950 ``smiprun`` script directly.
952 .. _cfg=smpi/host-speed:
953 .. _cfg=smpi/cpu-threshold:
954 .. _cfg=smpi/simulate-computation:
956 Automatic Benchmarking of SMPI Code
957 ...................................
959 In SMPI, the sequential code is automatically benchmarked, and these
960 computations are automatically reported to the simulator. That is to
961 say that if you have a large computation between a ``MPI_Recv()`` and
962 a ``MPI_Send()``, SMPI will automatically benchmark the duration of
963 this code, and create an execution task within the simulator to take
964 this into account. For that, the actual duration is measured on the
965 host machine and then scaled to the power of the corresponding
966 simulated machine. The variable ``smpi/host-speed`` allows one to
967 specify the computational speed of the host machine (in flop/s by
968 default) to use when scaling the execution times. It defaults to 20kf
969 (i.e., 20000 flop/s), but you really want to adjust it to get accurate
972 When the code consists of numerous consecutive MPI calls, the
973 previous mechanism feeds the simulation kernel with numerous tiny
974 computations. The ``smpi/cpu-threshold`` item becomes handy when this
975 impacts badly on the simulation performance. It specifies a threshold (in
976 seconds) below which the execution chunks are not reported to the
977 simulation kernel (default value: 1e-6).
979 .. note:: The option ``smpi/cpu-threshold`` ignores any computation
980 time spent below this threshold. SMPI does not consider the
981 `amount of time` of these computations; there is no offset for
982 this. Hence, a value that is too small, may lead to unreliable
985 In some cases, however, one may wish to disable simulation of
986 the computation of an application. This is the case when SMPI is used not to
987 simulate an MPI application, but instead an MPI code that performs
988 "live replay" of another MPI app (e.g., ScalaTrace's replay tool, or
989 various on-line simulators that run an app at scale). In this case the
990 computation of the replay/simulation logic should not be simulated by
991 SMPI. Instead, the replay tool or on-line simulator will issue
992 "computation events", which correspond to the actual MPI simulation
993 being replayed/simulated. At the moment, these computation events can
994 be simulated using SMPI by calling internal smpi_execute*() functions.
996 To disable the benchmarking/simulation of a computation in the simulated
997 application, the variable ``smpi/simulate-computation`` should be set
998 to **no**. This option just ignores the timings in your simulation; it
999 still executes the computations itself. If you want to stop SMPI from
1000 doing that, you should check the SMPI_SAMPLE macros, documented in
1001 Section :ref:`SMPI_use_faster`.
1003 +------------------------------------+-------------------------+-----------------------------+
1004 | Solution | Computations executed? | Computations simulated? |
1005 +====================================+=========================+=============================+
1006 | --cfg=smpi/simulate-computation:no | Yes | Never |
1007 +------------------------------------+-------------------------+-----------------------------+
1008 | --cfg=smpi/cpu-threshold:42 | Yes, in all cases | If it lasts over 42 seconds |
1009 +------------------------------------+-------------------------+-----------------------------+
1010 | SMPI_SAMPLE() macro | Only once per loop nest | Always |
1011 +------------------------------------+-------------------------+-----------------------------+
1013 .. _cfg=smpi/comp-adjustment-file:
1015 Slow-down or speed-up parts of your code
1016 ........................................
1018 **Option** ``smpi/comp-adjustment-file:`` **Default:** unset
1020 This option allows you to pass a file that contains two columns: The
1021 first column defines the section that will be subject to a speedup;
1022 the second column is the speedup. For instance:
1024 .. code-block:: shell
1026 "start:stop","ratio"
1027 "exchange_1.f:30:exchange_1.f:130",1.18244559422142
1029 The first line is the header - you must include it. The following
1030 line means that the code between two consecutive MPI calls on line 30
1031 in exchange_1.f and line 130 in exchange_1.f should receive a speedup
1032 of 1.18244559422142. The value for the second column is therefore a
1033 speedup, if it is larger than 1 and a slowdown if it is smaller
1034 than 1. Nothing will be changed if it is equal to 1.
1036 Of course, you can set any arbitrary filenames you want (so the start
1037 and end don't have to be in the same file), but be aware that this
1038 mechanism only supports `consecutive calls!`
1040 Please note that you must pass the ``-trace-call-location`` flag to
1041 smpicc or smpiff, respectively. This flag activates some internal
1042 macro definitions that help with obtaining the call location.
1044 .. _cfg=smpi/bw-factor:
1049 **Option** ``smpi/bw-factor``
1050 |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
1052 The possible throughput of network links is often dependent on the
1053 message sizes, as protocols may adapt to different message sizes. With
1054 this option, a series of message sizes and factors are given, helping
1055 the simulation to be more realistic. For instance, the current default
1056 value means that messages with size 65472 bytes and more will get a total of
1057 MAX_BANDWIDTH*0.940694, messages of size 15424 to 65471 will get
1058 MAX_BANDWIDTH*0.697866, and so on (where MAX_BANDWIDTH denotes the
1059 bandwidth of the link).
1061 An experimental script to compute these factors is available online. See
1062 https://framagit.org/simgrid/platform-calibration/
1063 https://simgrid.org/contrib/smpi-saturation-doc.html
1065 .. _cfg=smpi/display-timing:
1067 Reporting Simulation Time
1068 .........................
1070 **Option** ``smpi/display-timing`` **Default:** 0 (false)
1072 Most of the time, you run MPI code with SMPI to compute the time it
1073 would take to run it on a platform. But since the code is run through
1074 the ``smpirun`` script, you don't have any control on the launcher
1075 code, making it difficult to report the simulated time when the
1076 simulation ends. If you enable the ``smpi/display-timing`` item,
1077 ``smpirun`` will display this information when the simulation
1080 .. _cfg=smpi/keep-temps:
1082 Keeping temporary files after simulation
1083 ........................................
1085 **Option** ``smpi/keep-temps`` **default:** 0 (false)
1087 SMPI usually generates a lot of temporary files that are cleaned after
1088 use. This option requests to preserve them, for example to debug or
1089 profile your code. Indeed, the binary files are removed very early
1090 under the dlopen privatization schema, which tends to fool the
1093 .. _cfg=smpi/lat-factor:
1098 **Option** ``smpi/lat-factor`` |br|
1099 **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
1101 The motivation and syntax for this option is identical to the motivation/syntax
1102 of :ref:`cfg=smpi/bw-factor`.
1104 There is an important difference, though: While smpi/bw-factor `reduces` the
1105 actual bandwidth (i.e., values between 0 and 1 are valid), latency factors
1106 increase the latency, i.e., values larger than or equal to 1 are valid here.
1108 .. _cfg=smpi/papi-events:
1110 Trace hardware counters with PAPI
1111 .................................
1113 **Option** ``smpi/papi-events`` **default:** unset
1115 When the PAPI support is compiled into SimGrid, this option takes the
1116 names of PAPI counters and adds their respective values to the trace
1117 files (See Section :ref:`tracing_tracing_options`).
1121 This feature currently requires superuser privileges, as registers
1122 are queried. Only use this feature with code you trust! Call
1123 smpirun for instance via ``smpirun -wrapper "sudo "
1124 <your-parameters>`` or run ``sudo sh -c "echo 0 >
1125 /proc/sys/kernel/perf_event_paranoid"`` In the later case, sudo
1126 will not be required.
1128 It is planned to make this feature available on a per-process (or per-thread?) basis.
1129 The first draft, however, just implements a "global" (i.e., for all processes) set
1130 of counters, the "default" set.
1132 .. code-block:: shell
1134 --cfg=smpi/papi-events:"default:PAPI_L3_LDM:PAPI_L2_LDM"
1136 .. _cfg=smpi/privatization:
1138 Automatic Privatization of Global Variables
1139 ...........................................
1141 **Option** ``smpi/privatization`` **default:** "dlopen" (when using smpirun)
1143 MPI executables are usually meant to be executed in separate
1144 processes, but SMPI is executed in only one process. Global variables
1145 from executables will be placed in the same memory region and shared
1146 between processes, causing intricate bugs. Several options are
1147 possible to avoid this, as described in the main `SMPI publication
1148 <https://hal.inria.fr/hal-01415484>`_ and in the :ref:`SMPI
1149 documentation <SMPI_what_globals>`. SimGrid provides two ways of
1150 automatically privatizing the globals, and this option allows one to
1151 choose between them.
1153 - **no** (default when not using smpirun): Do not automatically
1154 privatize variables. Pass ``-no-privatize`` to smpirun to disable
1156 - **dlopen** or **yes** (default when using smpirun): Link multiple
1157 times against the binary.
1158 - **mmap** (slower, but maybe somewhat more stable):
1159 Runtime automatic switching of the data segments.
1162 This configuration option cannot be set in your platform file. You can only
1163 pass it as an argument to smpirun.
1165 .. _cfg=smpi/privatize-libs:
1167 Automatic privatization of global variables inside external libraries
1168 .....................................................................
1170 **Option** ``smpi/privatize-libs`` **default:** unset
1172 **Linux/BSD only:** When using dlopen (default) privatization,
1173 privatize specific shared libraries with internal global variables, if
1174 they can't be linked statically. For example libgfortran is usually
1175 used for Fortran I/O and indexes in files can be mixed up.
1177 Multiple libraries can be given, semicolon separated.
1179 This configuration option can only use either full paths to libraries,
1180 or full names. Check with ldd the name of the library you want to
1183 .. code-block:: shell
1187 libgfortran.so.3 => /usr/lib/x86_64-linux-gnu/libgfortran.so.3 (0x00007fbb4d91b000)
1190 Then you can use ``--cfg=smpi/privatize-libs:libgfortran.so.3``
1191 or ``--cfg=smpi/privatize-libs:/usr/lib/x86_64-linux-gnu/libgfortran.so.3``,
1192 but not ``libgfortran`` nor ``libgfortran.so``.
1194 .. _cfg=smpi/send-is-detached-thresh:
1196 Simulating MPI detached send
1197 ............................
1199 **Option** ``smpi/send-is-detached-thresh`` **default:** 65536
1201 This threshold specifies the size in bytes under which the send will
1202 return immediately. This is different from the threshold detailed in
1203 :ref:`cfg=smpi/async-small-thresh` because the message is not
1204 really sent when the send is posted. SMPI still waits for the
1205 corresponding receive to be posted, in order to perform the communication
1208 .. _cfg=smpi/coll-selector:
1210 Simulating MPI collective algorithms
1211 ....................................
1213 **Option** ``smpi/coll-selector`` **Possible values:** naive (default), ompi, mpich
1215 SMPI implements more than 100 different algorithms for MPI collective
1216 communication, to accurately simulate the behavior of most of the
1217 existing MPI libraries. The ``smpi/coll-selector`` item can be used to
1218 select the decision logic either of the OpenMPI or the MPICH libraries. (By
1219 default SMPI uses naive version of collective operations.)
1221 Each collective operation can be manually selected with a
1222 ``smpi/collective_name:algo_name``. Available algorithms are listed in
1223 :ref:`SMPI_use_colls`.
1225 .. TODO:: All available collective algorithms will be made available
1226 via the ``smpirun --help-coll`` command.
1228 .. _cfg=smpi/iprobe:
1230 Inject constant times for MPI_Iprobe
1231 ....................................
1233 **Option** ``smpi/iprobe`` **default:** 0.0001
1235 The behavior and motivation for this configuration option is identical
1236 with :ref:`smpi/test <cfg=smpi/test>`, but for the function
1239 .. _cfg=smpi/iprobe-cpu-usage:
1241 Reduce speed for iprobe calls
1242 .............................
1244 **Option** ``smpi/iprobe-cpu-usage`` **default:** 1 (no change)
1246 MPI_Iprobe calls can be heavily used in applications. To account
1247 correctly for the energy that cores spend probing, it is necessary to
1248 reduce the load that these calls cause inside SimGrid.
1250 For instance, we measured a maximum power consumption of 220 W for a
1251 particular application but only 180 W while this application was
1252 probing. Hence, the correct factor that should be passed to this
1253 option would be 180/220 = 0.81.
1257 Inject constant times for MPI_Init
1258 ..................................
1260 **Option** ``smpi/init`` **default:** 0
1262 The behavior and motivation for this configuration option is identical
1263 with :ref:`smpi/test <cfg=smpi/test>`, but for the function ``MPI_Init()``.
1267 Inject constant times for MPI_Isend()
1268 .....................................
1270 **Option** ``smpi/ois``
1272 The behavior and motivation for this configuration option is identical
1273 with :ref:`smpi/os <cfg=smpi/os>`, but for the function ``MPI_Isend()``.
1277 Inject constant times for MPI_send()
1278 ....................................
1280 **Option** ``smpi/os``
1282 In several network models such as LogP, send (MPI_Send, MPI_Isend) and
1283 receive (MPI_Recv) operations incur costs (i.e., they consume CPU
1284 time). SMPI can factor these costs in as well, but the user has to
1285 configure SMPI accordingly as these values may vary by machine. This
1286 can be done by using ``smpi/os`` for MPI_Send operations; for MPI_Isend
1287 and MPI_Recv, use ``smpi/ois`` and ``smpi/or``, respectively. These work
1288 exactly as ``smpi/ois``.
1290 This item can consist of multiple sections; each section takes three
1291 values, for example ``1:3:2;10:5:1``. The sections are divided by ";"
1292 so this example contains two sections. Furthermore, each section
1293 consists of three values.
1295 1. The first value denotes the minimum size in bytes for this section to take effect;
1296 read it as "if message size is greater than this value (and other section has a larger
1297 first value that is also smaller than the message size), use this".
1298 In the first section above, this value is "1".
1300 2. The second value is the startup time; this is a constant value that will always
1301 be charged, no matter what the size of the message. In the first section above,
1304 3. The third value is the `per-byte` cost. That is, it is charged for every
1305 byte of the message (incurring cost messageSize*cost_per_byte)
1306 and hence accounts also for larger messages. In the first
1307 section of the example above, this value is "2".
1309 Now, SMPI always checks which section it should use for a given
1310 message; that is, if a message of size 11 is sent with the
1311 configuration of the example above, only the second section will be
1312 used, not the first, as the first value of the second section is
1313 closer to the message size. Hence, when ``smpi/os=1:3:2;10:5:1``, a
1314 message of size 11 incurs the following cost inside MPI_Send:
1315 ``5+11*1`` because 5 is the startup cost and 1 is the cost per byte.
1317 Note that the order of sections can be arbitrary; they will be ordered internally.
1321 Inject constant times for MPI_Recv()
1322 ....................................
1324 **Option** ``smpi/or``
1326 The behavior and motivation for this configuration option is identical
1327 with :ref:`smpi/os <cfg=smpi/os>`, but for the function ``MPI_Recv()``.
1330 .. _cfg=smpi/grow-injected-times:
1332 Inject constant times for MPI_Test
1333 ..................................
1335 **Option** ``smpi/test`` **default:** 0.0001
1337 By setting this option, you can control the amount of time a process
1338 sleeps when MPI_Test() is called; this is important, because SimGrid
1339 normally only advances the time while communication is happening and
1340 thus, MPI_Test will not add to the time, resulting in deadlock if it is
1341 used as a break-condition as in the following example:
1346 MPI_Test(request, flag, status);
1350 To speed up execution, we use a counter to keep track of how often we
1351 checked if the handle is now valid or not. Hence, we actually
1352 use counter*SLEEP_TIME, that is, the time MPI_Test() causes the
1353 process to sleep increases linearly with the number of previously
1354 failed tests. This behavior can be disabled by setting
1355 ``smpi/grow-injected-times`` to **no**. This will also disable this
1356 behavior for MPI_Iprobe.
1358 .. _cfg=smpi/shared-malloc:
1359 .. _cfg=smpi/shared-malloc-hugepage:
1364 **Option** ``smpi/shared-malloc`` **Possible values:** global (default), local
1366 If your simulation consumes too much memory, you may want to modify
1367 your code so that the working areas are shared by all MPI ranks. For
1368 example, in a block-cyclic matrix multiplication, you will only
1369 allocate one set of blocks, and all processes will share them.
1370 Naturally, this will lead to very wrong results, but this will save a
1371 lot of memory. So this is still desirable for some studies. For more on
1372 the motivation for that feature, please refer to the `relevant section
1373 <https://simgrid.github.io/SMPI_CourseWare/topic_understanding_performance/matrixmultiplication>`_
1374 of the SMPI CourseWare (see Activity #2.2 of the pointed
1375 assignment). In practice, change the calls for malloc() and free() into
1376 SMPI_SHARED_MALLOC() and SMPI_SHARED_FREE().
1378 SMPI provides two algorithms for this feature. The first one, called
1379 ``local``, allocates one block per call to SMPI_SHARED_MALLOC()
1380 (each call site gets its own block) ,and this block is shared
1381 among all MPI ranks. This is implemented with the shm_* functions
1382 to create a new POSIX shared memory object (kept in RAM, in /dev/shm)
1383 for each shared block.
1385 With the ``global`` algorithm, each call to SMPI_SHARED_MALLOC()
1386 returns a new address, but it only points to a shadow block: its memory
1387 area is mapped on a 1 MiB file on disk. If the returned block is of size
1388 N MiB, then the same file is mapped N times to cover the whole bloc.
1389 At the end, no matter how many times you call SMPI_SHARED_MALLOC, this will
1390 only consume 1 MiB in memory.
1392 You can disable this behavior and come back to regular mallocs (for
1393 example for debugging purposes) using ``no`` as a value.
1395 If you want to keep private some parts of the buffer, for instance if these
1396 parts are used by the application logic and should not be corrupted, you
1397 can use SMPI_PARTIAL_SHARED_MALLOC(size, offsets, offsets_count). For example:
1401 mem = SMPI_PARTIAL_SHARED_MALLOC(500, {27,42 , 100,200}, 2);
1403 This will allocate 500 bytes to mem, such that mem[27..41] and
1404 mem[100..199] are shared while other area remain private.
1406 Then, it can be deallocated by calling SMPI_SHARED_FREE(mem).
1408 When smpi/shared-malloc:global is used, the memory consumption problem
1409 is solved, but it may induce too much load on the kernel's pages table.
1410 In this case, you should use huge pages so that the kernel creates only one
1411 entry per MB of malloced data instead of one entry per 4 kB.
1412 To activate this, you must mount a hugetlbfs on your system and allocate
1413 at least one huge page:
1415 .. code-block:: shell
1418 sudo mount none /home/huge -t hugetlbfs -o rw,mode=0777
1419 sudo sh -c 'echo 1 > /proc/sys/vm/nr_hugepages' # echo more if you need more
1421 Then, you can pass the option
1422 ``--cfg=smpi/shared-malloc-hugepage:/home/huge`` to smpirun to
1423 actually activate the huge page support in shared mallocs.
1427 Inject constant times for MPI_Wtime, gettimeofday and clock_gettime
1428 ...................................................................
1430 **Option** ``smpi/wtime`` **default:** 10 ns
1432 This option controls the amount of (simulated) time spent in calls to
1433 MPI_Wtime(), gettimeofday() and clock_gettime(). If you set this value
1434 to 0, the simulated clock is not advanced in these calls, which leads
1435 to issues if your application contains such a loop:
1439 while(MPI_Wtime() < some_time_bound) {
1440 /* some tests, with no communication nor computation */
1443 When the option smpi/wtime is set to 0, the time advances only on
1444 communications and computations. So the previous code results in an
1445 infinite loop: the current [simulated] time will never reach
1446 ``some_time_bound``. This infinite loop is avoided when that option
1447 is set to a small value, as it is by default since SimGrid v3.21.
1449 Note that if your application does not contain any loop depending on
1450 the current time only, then setting this option to a non-zero value
1451 will slow down your simulations by a tiny bit: the simulation loop has
1452 to be broken out of and reset each time your code asks for the current time.
1453 If the simulation speed really matters to you, you can avoid this
1454 extra delay by setting smpi/wtime to 0.
1456 Other Configurations
1457 --------------------
1459 .. _cfg=debug/clean-atexit:
1461 Cleanup at Termination
1462 ......................
1464 **Option** ``debug/clean-atexit`` **default:** on
1466 If your code is segfaulting during its finalization, it may help to
1467 disable this option to request that SimGrid not attempt any cleanups at
1468 the end of the simulation. Since the Unix process is ending anyway,
1469 the operating system will wipe it all.
1476 **Option** ``path`` **default:** . (current dir)
1478 It is possible to specify a list of directories to search in for the
1479 trace files (see :ref:`pf_trace`) by using this configuration
1480 item. To add several directory to the path, set the configuration
1481 item several times, as in ``--cfg=path:toto --cfg=path:tutu``
1483 .. _cfg=debug/breakpoint:
1488 **Option** ``debug/breakpoint`` **default:** unset
1490 This configuration option sets a breakpoint: when the simulated clock
1491 reaches the given time, a SIGTRAP is raised. This can be used to stop
1492 the execution and get a backtrace with a debugger.
1494 It is also possible to set the breakpoint from inside the debugger, by
1495 writing in global variable simgrid::simix::breakpoint. For example,
1498 .. code-block:: shell
1500 set variable simgrid::simix::breakpoint = 3.1416
1502 .. _cfg=debug/verbose-exit:
1507 **Option** ``debug/verbose-exit`` **default:** on
1509 By default, when Ctrl-C is pressed, the status of all existing actors
1510 is displayed before exiting the simulation. This is very useful to
1511 debug your code, but it can become troublesome if you have many
1512 actors. Set this configuration item to **off** to disable this
1515 .. _cfg=exception/cutpath:
1517 Truncate local path from exception backtrace
1518 ............................................
1520 **Option** ``exception/cutpath`` **default:** off
1522 This configuration option is used to remove the path from the
1523 backtrace shown when an exception is thrown. This is mainly useful for
1524 the tests: the full file path would makes the tests non-reproducible because
1525 the paths of source files depend of the build settings. That would
1526 break most of the tests since their output is continually compared.
1528 Logging Configuration
1529 ---------------------
1531 This can be done by using XBT. Go to :ref:`XBT_log` for more details.