8 <object id="TOC" data="graphical-toc.svg" width="100%" 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 ').
44 Another solution is to use the ``<config>`` tag in the platform file. The
45 only restriction is that this tag must occure 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/parallel-threshold:** :ref:`cfg=contexts/parallel-threshold`
91 - **contexts/stack-size:** :ref:`cfg=contexts/stack-size`
92 - **contexts/synchro:** :ref:`cfg=contexts/synchro`
94 - **cpu/maxmin-selective-update:** :ref:`Cpu Optimization Level <options_model_optim>`
95 - **cpu/model:** :ref:`options_model_select`
96 - **cpu/optim:** :ref:`Cpu Optimization Level <options_model_optim>`
98 - **debug/breakpoint:** :ref:`cfg=debug/breakpoint`
99 - **debug/clean-atexit:** :ref:`cfg=debug/clean-atexit`
100 - **debug/verbose-exit:** :ref:`cfg=debug/verbose-exit`
102 - **exception/cutpath:** :ref:`cfg=exception/cutpath`
104 - **host/model:** :ref:`options_model_select`
106 - **maxmin/precision:** :ref:`cfg=maxmin/precision`
107 - **maxmin/concurrency-limit:** :ref:`cfg=maxmin/concurrency-limit`
109 - **msg/debug-multiple-use:** :ref:`cfg=msg/debug-multiple-use`
111 - **model-check:** :ref:`options_modelchecking`
112 - **model-check/checkpoint:** :ref:`cfg=model-check/checkpoint`
113 - **model-check/communications-determinism:** :ref:`cfg=model-check/communications-determinism`
114 - **model-check/dot-output:** :ref:`cfg=model-check/dot-output`
115 - **model-check/hash:** :ref:`cfg=model-checker/hash`
116 - **model-check/max-depth:** :ref:`cfg=model-check/max-depth`
117 - **model-check/property:** :ref:`cfg=model-check/property`
118 - **model-check/reduction:** :ref:`cfg=model-check/reduction`
119 - **model-check/replay:** :ref:`cfg=model-check/replay`
120 - **model-check/send-determinism:** :ref:`cfg=model-check/send-determinism`
121 - **model-check/termination:** :ref:`cfg=model-check/termination`
122 - **model-check/timeout:** :ref:`cfg=model-check/timeout`
123 - **model-check/visited:** :ref:`cfg=model-check/visited`
125 - **network/bandwidth-factor:** :ref:`cfg=network/bandwidth-factor`
126 - **network/crosstraffic:** :ref:`cfg=network/crosstraffic`
127 - **network/latency-factor:** :ref:`cfg=network/latency-factor`
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/bw-factor:** :ref:`cfg=smpi/bw-factor`
145 - **smpi/coll-selector:** :ref:`cfg=smpi/coll-selector`
146 - **smpi/comp-adjustment-file:** :ref:`cfg=smpi/comp-adjustment-file`
147 - **smpi/cpu-threshold:** :ref:`cfg=smpi/cpu-threshold`
148 - **smpi/display-timing:** :ref:`cfg=smpi/display-timing`
149 - **smpi/grow-injected-times:** :ref:`cfg=smpi/grow-injected-times`
150 - **smpi/host-speed:** :ref:`cfg=smpi/host-speed`
151 - **smpi/IB-penalty-factors:** :ref:`cfg=smpi/IB-penalty-factors`
152 - **smpi/iprobe:** :ref:`cfg=smpi/iprobe`
153 - **smpi/iprobe-cpu-usage:** :ref:`cfg=smpi/iprobe-cpu-usage`
154 - **smpi/init:** :ref:`cfg=smpi/init`
155 - **smpi/keep-temps:** :ref:`cfg=smpi/keep-temps`
156 - **smpi/lat-factor:** :ref:`cfg=smpi/lat-factor`
157 - **smpi/ois:** :ref:`cfg=smpi/ois`
158 - **smpi/or:** :ref:`cfg=smpi/or`
159 - **smpi/os:** :ref:`cfg=smpi/os`
160 - **smpi/papi-events:** :ref:`cfg=smpi/papi-events`
161 - **smpi/privatization:** :ref:`cfg=smpi/privatization`
162 - **smpi/privatize-libs:** :ref:`cfg=smpi/privatize-libs`
163 - **smpi/send-is-detached-thresh:** :ref:`cfg=smpi/send-is-detached-thresh`
164 - **smpi/shared-malloc:** :ref:`cfg=smpi/shared-malloc`
165 - **smpi/shared-malloc-hugepage:** :ref:`cfg=smpi/shared-malloc-hugepage`
166 - **smpi/simulate-computation:** :ref:`cfg=smpi/simulate-computation`
167 - **smpi/test:** :ref:`cfg=smpi/test`
168 - **smpi/wtime:** :ref:`cfg=smpi/wtime`
170 - **Tracing configuration options** can be found in Section :ref:`tracing_tracing_options`
172 - **storage/model:** :ref:`options_model_select`
174 - **vm/model:** :ref:`options_model_select`
178 Configuring the Platform Models
179 -------------------------------
181 .. _options_model_select:
183 Choosing the Platform Models
184 ............................
186 SimGrid comes with several network, CPU and storage models built in,
187 and you can change the used model at runtime by changing the passed
188 configuration. The three main configuration items are given below.
189 For each of these items, passing the special ``help`` value gives you
190 a short description of all possible values (for example,
191 ``--cfg=network/model:help`` will present all provided network
192 models). Also, ``--help-models`` should provide information about all
193 models for all existing resources.
195 - ``network/model``: specify the used network model. Possible values:
197 - **LV08 (default one):** Realistic network analytic model
198 (slow-start modeled by multiplying latency by 13.01, bandwidth by
199 .97; bottleneck sharing uses a payload of S=20537 for evaluating
200 RTT). Described in `Accuracy Study and Improvement of Network
201 Simulation in the SimGrid Framework
202 <http://mescal.imag.fr/membres/arnaud.legrand/articles/simutools09.pdf>`_.
203 - **Constant:** Simplistic network model where all communication
204 take a constant time (one second). This model provides the lowest
205 realism, but is (marginally) faster.
206 - **SMPI:** Realistic network model specifically tailored for HPC
207 settings (accurate modeling of slow start with correction factors on
208 three intervals: < 1KiB, < 64 KiB, >= 64 KiB). This model can be
209 :ref:`further configured <options_model_network>`.
210 - **IB:** Realistic network model specifically tailored for HPC
211 settings with InfiniBand networks (accurate modeling contention
212 behavior, based on the model explained in `this PhD work
213 <http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf>`_.
214 This model can be :ref:`further configured <options_model_network>`.
215 - **CM02:** Legacy network analytic model. Very similar to LV08, but
216 without corrective factors. The timings of small messages are thus
217 poorly modeled. This model is described in `A Network Model for
218 Simulation of Grid Application
219 <https://hal.inria.fr/inria-00071989/document>`_.
220 - **ns-3** (only available if you compiled SimGrid accordingly):
221 Use the packet-level network
222 simulators as network models (see :ref:`model_ns3`).
223 This model can be :ref:`further configured <options_pls>`.
225 - ``cpu/model``: specify the used CPU model. We have only one model
228 - **Cas01:** Simplistic CPU model (time=size/power)
230 - ``host/model``: The host concept is the aggregation of a CPU with a
231 network card. Three models exists, but actually, only 2 of them are
232 interesting. The "compound" one is simply due to the way our
233 internal code is organized, and can easily be ignored. So at the
234 end, you have two host models: The default one allows to aggregate
235 an existing CPU model with an existing network model, but does not
236 allow parallel tasks because these beasts need some collaboration
237 between the network and CPU model. That is why, ptask_07 is used by
238 default when using SimDag.
240 - **default:** Default host model. Currently, CPU:Cas01 and
241 network:LV08 (with cross traffic enabled)
242 - **compound:** Host model that is automatically chosen if
243 you change the network and CPU models
244 - **ptask_L07:** Host model somehow similar to Cas01+CM02 but
245 allowing "parallel tasks", that are intended to model the moldable
246 tasks of the grid scheduling literature.
248 - ``storage/model``: specify the used storage model. Only one model is
250 - ``vm/model``: specify the model for virtual machines. Only one model
253 .. todo: make 'compound' the default host model.
255 .. _options_model_optim:
260 The network and CPU models that are based on lmm_solve (that
261 is, all our analytical models) accept specific optimization
264 - items ``network/optim`` and ``cpu/optim`` (both default to 'Lazy'):
266 - **Lazy:** Lazy action management (partial invalidation in lmm +
267 heap in action remaining).
268 - **TI:** Trace integration. Highly optimized mode when using
269 availability traces (only available for the Cas01 CPU model for
271 - **Full:** Full update of remaining and variables. Slow but may be
272 useful when debugging.
274 - items ``network/maxmin-selective-update`` and
275 ``cpu/maxmin-selective-update``: configure whether the underlying
276 should be lazily updated or not. It should have no impact on the
277 computed timings, but should speed up the computation. |br| It is
278 still possible to disable this feature because it can reveal
279 counter-productive in very specific scenarios where the
280 interaction level is high. In particular, if all your
281 communication share a given backbone link, you should disable it:
282 without it, a simple regular loop is used to update each
283 communication. With it, each of them is still updated (because of
284 the dependency induced by the backbone), but through a complicated
285 and slow pattern that follows the actual dependencies.
287 .. _cfg=maxmin/precision:
288 .. _cfg=surf/precision:
293 **Option** ``maxmin/precision`` **Default:** 0.00001 (in flops or bytes) |br|
294 **Option** ``surf/precision`` **Default:** 0.00001 (in seconds)
296 The analytical models handle a lot of floating point values. It is
297 possible to change the epsilon used to update and compare them through
298 this configuration item. Changing it may speedup the simulation by
299 discarding very small actions, at the price of a reduced numerical
300 precision. You can modify separately the precision used to manipulate
301 timings (in seconds) and the one used to manipulate amounts of work
304 .. _cfg=maxmin/concurrency-limit:
309 **Option** ``maxmin/concurrency-limit`` **Default:** -1 (no limit)
311 The maximum number of variables per resource can be tuned through this
312 option. You can have as many simultaneous actions per resources as you
313 want. If your simulation presents a very high level of concurrency, it
314 may help to use e.g. 100 as a value here. It means that at most 100
315 actions can consume a resource at a given time. The extraneous actions
316 are queued and wait until the amount of concurrency of the considered
317 resource lowers under the given boundary.
319 Such limitations help both to the simulation speed and simulation accuracy
320 on highly constrained scenarios, but the simulation speed suffers of this
321 setting on regular (less constrained) scenarios so it is off by default.
323 .. _options_model_network:
325 Configuring the Network Model
326 .............................
328 .. _cfg=network/TCP-gamma:
330 Maximal TCP Window Size
331 ^^^^^^^^^^^^^^^^^^^^^^^
333 **Option** ``network/TCP-gamma`` **Default:** 4194304
335 The analytical models need to know the maximal TCP window size to take
336 the TCP congestion mechanism into account. On Linux, this value can
337 be retrieved using the following commands. Both give a set of values,
338 and you should use the last one, which is the maximal size.
340 .. code-block:: shell
342 cat /proc/sys/net/ipv4/tcp_rmem # gives the sender window
343 cat /proc/sys/net/ipv4/tcp_wmem # gives the receiver window
345 .. _cfg=smpi/IB-penalty-factors:
346 .. _cfg=network/bandwidth-factor:
347 .. _cfg=network/latency-factor:
348 .. _cfg=network/weight-S:
350 Correcting Important Network Parameters
351 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
353 SimGrid can take network irregularities such as a slow startup or
354 changing behavior depending on the message size into account. You
355 should not change these values unless you really know what you're
356 doing. The corresponding values were computed through data fitting
357 one the timings of packet-level simulators, as described in `Accuracy
358 Study and Improvement of Network Simulation in the SimGrid Framework
359 <http://mescal.imag.fr/membres/arnaud.legrand/articles/simutools09.pdf>`_.
362 If you are using the SMPI model, these correction coefficients are
363 themselves corrected by constant values depending on the size of the
364 exchange. By default SMPI uses factors computed on the Stampede
365 Supercomputer at TACC, with optimal deployment of processes on
366 nodes. Again, only hardcore experts should bother about this fact.
368 InfiniBand network behavior can be modeled through 3 parameters
369 ``smpi/IB-penalty-factors:"βe;βs;γs"``, as explained in `this PhD
371 <http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf>`_.
373 .. todo:: This section should be rewritten, and actually explain the
374 options network/bandwidth-factor, network/latency-factor,
377 .. _cfg=network/crosstraffic:
379 Simulating Cross-Traffic
380 ^^^^^^^^^^^^^^^^^^^^^^^^
382 Since SimGrid v3.7, cross-traffic effects can be taken into account in
383 analytical simulations. It means that ongoing and incoming
384 communication flows are treated independently. In addition, the LV08
385 model adds 0.05 of usage on the opposite direction for each new
386 created flow. This can be useful to simulate some important TCP
387 phenomena such as ack compression.
389 For that to work, your platform must have two links for each
390 pair of interconnected hosts. An example of usable platform is
391 available in ``examples/platforms/crosstraffic.xml``.
393 This is activated through the ``network/crosstraffic`` item, that
394 can be set to 0 (disable this feature) or 1 (enable it).
396 Note that with the default host model this option is activated by default.
398 .. _cfg=smpi/async-small-thresh:
400 Simulating Asyncronous Send
401 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
403 (this configuration item is experimental and may change or disapear)
405 It is possible to specify that messages below a certain size will be
406 sent as soon as the call to MPI_Send is issued, without waiting for
407 the correspondant receive. This threshold can be configured through
408 the ``smpi/async-small-thresh`` item. The default value is 0. This
409 behavior can also be manually set for mailboxes, by setting the
410 receiving mode of the mailbox with a call to
411 :cpp:func:`MSG_mailbox_set_async`. After this, all messages sent to
412 this mailbox will have this behavior regardless of the message size.
414 This value needs to be smaller than or equals to the threshold set at
415 @ref options_model_smpi_detached , because asynchronous messages are
416 meant to be detached as well.
423 **Option** ``ns3/TcpModel`` **Default:** "default" (ns-3 default)
425 When using ns-3, there is an extra item ``ns3/TcpModel``, corresponding
426 to the ``ns3::TcpL4Protocol::SocketType`` configuration item in
427 ns-3. The only valid values (enforced on the SimGrid side) are
428 'default' (no change to the ns-3 configuration), 'NewReno' or 'Reno' or
431 Configuring the Storage model
432 .............................
434 .. _cfg=storage/max_file_descriptors:
436 File Descriptor Cound per Host
437 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
439 **Option** ``storage/max_file_descriptors`` **Default:** 1024
441 Each host maintains a fixed-size array of its file descriptors. You
442 can change its size through this item to either enlarge it if your
443 application requires it or to reduce it to save memory space.
450 SimGrid plugins allow to extend the framework without changing its
451 source code directly. Read the source code of the existing plugins to
452 learn how to do so (in ``src/plugins``), and ask your questions to the
453 usual channels (Stack Overflow, Mailing list, IRC). The basic idea is
454 that plugins usually register callbacks to some signals of interest.
455 If they need to store some information about a given object (Link, CPU
456 or Actor), they do so through the use of a dedicated object extension.
458 Some of the existing plugins can be activated from the command line,
459 meaning that you can activate them from the command line without any
460 modification to your simulation code. For example, you can activate
461 the host energy plugin by adding ``--cfg=plugin:host_energy`` to your
464 Here is the full list of plugins that can be activated this way:
466 - **host_energy:** keeps track of the energy dissipated by
467 computations. More details in @ref plugin_energy.
468 - **link_energy:** keeps track of the energy dissipated by
469 communications. More details in @ref SURF_plugin_energy.
470 - **host_load:** keeps track of the computational load.
471 More details in @ref plugin_load.
473 .. _options_modelchecking:
475 Configuring the Model-Checking
476 ------------------------------
478 To enable the SimGrid model-checking support the program should
479 be executed using the simgrid-mc wrapper:
481 .. code-block:: shell
483 simgrid-mc ./my_program
485 Safety properties are expressed as assertions using the function
486 :cpp:func:`void MC_assert(int prop)`.
488 .. _cfg=model-check/property:
490 Specifying a liveness property
491 ..............................
493 **Option** ``model-check/property`` **Default:** unset
495 If you want to specify liveness properties, you have to pass them on
496 the command line, specifying the name of the file containing the
497 property, as formatted by the ltl2ba program.
500 .. code-block:: shell
502 simgrid-mc ./my_program --cfg=model-check/property:<filename>
504 .. _cfg=model-check/checkpoint:
506 Going for Stateful Verification
507 ...............................
509 By default, the system is backtracked to its initial state to explore
510 another path instead of backtracking to the exact step before the fork
511 that we want to explore (this is called stateless verification). This
512 is done this way because saving intermediate states can rapidly
513 exhaust the available memory. If you want, you can change the value of
514 the ``model-check/checkpoint`` item. For example,
515 ``--cfg=model-check/checkpoint:1`` asks to take a checkpoint every
516 step. Beware, this will certainly explode your memory. Larger values
517 are probably better, make sure to experiment a bit to find the right
518 setting for your specific system.
520 .. _cfg=model-check/reduction:
522 Specifying the kind of reduction
523 ................................
525 The main issue when using the model-checking is the state space
526 explosion. To counter that problem, you can chose a exploration
527 reduction techniques with
528 ``--cfg=model-check/reduction:<technique>``. For now, this
529 configuration variable can take 2 values:
531 - **none:** Do not apply any kind of reduction (mandatory for now for
533 - **dpor:** Apply Dynamic Partial Ordering Reduction. Only valid if
534 you verify local safety properties (default value for safety
537 There is unfortunately no silver bullet here, and the most efficient
538 reduction techniques cannot be applied to any properties. In
539 particular, the DPOR method cannot be applied on liveness properties
540 since our implementation of DPOR may break some cycles, while cycles
541 are very important to the soundness of the exploration for liveness
544 .. _cfg=model-check/visited:
546 Size of Cycle Detection Set
547 ...........................
549 In order to detect cycles, the model-checker needs to check if a new
550 explored state is in fact the same state than a previous one. For
551 that, the model-checker can take a snapshot of each visited state:
552 this snapshot is then used to compare it with subsequent states in the
555 The ``model-check/visited`` item is the maximum number of states which
556 are stored in memory. If the maximum number of snapshotted state is
557 reached, some states will be removed from the memory and some cycles
558 might be missed. Small values can lead to incorrect verifications, but
559 large value can exhaust your memory, so choose carefully.
561 By default, no state is snapshotted and cycles cannot be detected.
563 .. _cfg=model-check/termination:
565 Non-Termination Detection
566 .........................
568 The ``model-check/termination`` configuration item can be used to
569 report if a non-termination execution path has been found. This is a
570 path with a cycle which means that the program might never terminate.
572 This only works in safety mode, not in liveness mode.
574 This options is disabled by default.
576 .. _cfg=model-check/dot-output:
581 If set, the ``model-check/dot-output`` configuration item is the name
582 of a file in which to write a dot file of the path leading the found
583 property (safety or liveness violation) as well as the cycle for
584 liveness properties. This dot file can then fed to the graphviz dot
585 tool to generate an corresponding graphical representation.
587 .. _cfg=model-check/max-depth:
589 Exploration Depth Limit
590 .......................
592 The ``model-checker/max-depth`` can set the maximum depth of the
593 exploration graph of the model-checker. If this limit is reached, a
594 logging message is sent and the results might not be exact.
596 By default, there is not depth limit.
598 .. _cfg=model-check/timeout:
603 By default, the model-checker does not handle timeout conditions: the `wait`
604 operations never time out. With the ``model-check/timeout`` configuration item
605 set to **yes**, the model-checker will explore timeouts of `wait` operations.
607 .. _cfg=model-check/communications-determinism:
608 .. _cfg=model-check/send-determinism:
610 Communication Determinism
611 .........................
613 The ``model-check/communications-determinism`` and
614 ``model-check/send-determinism`` items can be used to select the
615 communication determinism mode of the model-checker which checks
616 determinism properties of the communications of an application.
618 Verification Performance Considerations
619 .......................................
621 The size of the stacks can have a huge impact on the memory
622 consumption when using model-checking. By default, each snapshot will
623 save a copy of the whole stacks and not only of the part which is
624 really meaningful: you should expect the contribution of the memory
625 consumption of the snapshots to be @f$ @mbox{number of processes}
626 @times @mbox{stack size} @times @mbox{number of states} @f$.
628 When compiled against the model checker, the stacks are not
629 protected with guards: if the stack size is too small for your
630 application, the stack will silently overflow on other parts of the
631 memory (see :ref:`contexts/guard-size <cfg=contexts/guard-size>`).
633 .. _cfg=model-checker/hash:
638 Usually most of the time of the model-checker is spent comparing states. This
639 process is complicated and consumes a lot of bandwidth and cache.
640 In order to speedup the state comparison, the experimental ``model-checker/hash``
641 configuration item enables the computation of a hash summarizing as much
642 information of the state as possible into a single value. This hash can be used
643 to avoid most of the comparisons: the costly comparison is then only used when
644 the hashes are identical.
646 Currently most of the state is not included in the hash because the
647 implementation was found to be buggy and this options is not as useful as
648 it could be. For this reason, it is currently disabled by default.
650 .. _cfg=model-check/replay:
652 Replaying buggy execution paths out of the model-checker
653 ........................................................
655 Debugging the problems reported by the model-checker is challenging: First, the
656 application under verification cannot be debugged with gdb because the
657 model-checker already traces it. Then, the model-checker may explore several
658 execution paths before encountering the issue, making it very difficult to
659 understand the outputs. Fortunately, SimGrid provides the execution path leading
660 to any reported issue so that you can replay this path out of the model checker,
661 enabling the usage of classical debugging tools.
663 When the model-checker finds an interesting path in the application
664 execution graph (where a safety or liveness property is violated), it
665 generates an identifier for this path. Here is an example of output:
667 .. code-block:: shell
669 [ 0.000000] (0:@) Check a safety property
670 [ 0.000000] (0:@) **************************
671 [ 0.000000] (0:@) *** PROPERTY NOT VALID ***
672 [ 0.000000] (0:@) **************************
673 [ 0.000000] (0:@) Counter-example execution trace:
674 [ 0.000000] (0:@) [(1)Tremblay (app)] MC_RANDOM(3)
675 [ 0.000000] (0:@) [(1)Tremblay (app)] MC_RANDOM(4)
676 [ 0.000000] (0:@) Path = 1/3;1/4
677 [ 0.000000] (0:@) Expanded states = 27
678 [ 0.000000] (0:@) Visited states = 68
679 [ 0.000000] (0:@) Executed transitions = 46
681 The interesting line is ``Path = 1/3;1/4``, which means that you should use
682 `--cfg=model-check/replay:1/3;1/4`` to replay your application on the buggy
683 execution path. The other options should be the same (but the model-checker
684 should be disabled). Note that format and meaning of the path may change between
687 Configuring the User Code Virtualization
688 ----------------------------------------
690 .. _cfg=contexts/factory:
692 Selecting the Virtualization Factory
693 ....................................
695 **Option** contexts/factory **Default:** "raw"
697 In SimGrid, the user code is virtualized in a specific mechanism that
698 allows the simulation kernel to control its execution: when a user
699 process requires a blocking action (such as sending a message), it is
700 interrupted, and only gets released when the simulated clock reaches
701 the point where the blocking operation is done. This is explained
702 graphically in the `relevant tutorial, available online
703 <https://simgrid.org/tutorials/simgrid-simix-101.pdf>`_.
705 In SimGrid, the containers in which user processes are virtualized are
706 called contexts. Several context factory are provided, and you can
707 select the one you want to use with the ``contexts/factory``
708 configuration item. Some of the following may not exist on your
709 machine because of portability issues. In any case, the default one
710 should be the most effcient one (please report bugs if the
711 auto-detection fails for you). They are approximately sorted here from
712 the slowest to the most efficient:
714 - **thread:** very slow factory using full featured threads (either
715 pthreads or windows native threads). They are slow but very
716 standard. Some debuggers or profilers only work with this factory.
717 - **java:** Java applications are virtualized onto java threads (that
718 are regular pthreads registered to the JVM)
719 - **ucontext:** fast factory using System V contexts (Linux and FreeBSD only)
720 - **boost:** This uses the `context
721 implementation <http://www.boost.org/doc/libs/1_59_0/libs/context/doc/html/index.html>`_
722 of the boost library for a performance that is comparable to our
724 |br| Install the relevant library (e.g. with the
725 libboost-contexts-dev package on Debian/Ubuntu) and recompile
727 - **raw:** amazingly fast factory using a context switching mechanism
728 of our own, directly implemented in assembly (only available for x86
729 and amd64 platforms for now) and without any unneeded system call.
731 The main reason to change this setting is when the debugging tools get
732 fooled by the optimized context factories. Threads are the most
733 debugging-friendly contextes, as they allow to set breakpoints
734 anywhere with gdb and visualize backtraces for all processes, in order
735 to debug concurrency issues. Valgrind is also more comfortable with
736 threads, but it should be usable with all factories (Exception: the
737 callgrind tool really dislikes raw and ucontext factories).
739 .. _cfg=contexts/stack-size:
741 Adapting the Stack Size
742 .......................
744 **Option** ``contexts/stack-size`` **Default:** 8192 KiB
746 Each virtualized used process is executed using a specific system
747 stack. The size of this stack has a huge impact on the simulation
748 scalability, but its default value is rather large. This is because
749 the error messages that you get when the stack size is too small are
750 rather disturbing: this leads to stack overflow (overwriting other
751 stacks), leading to segfaults with corrupted stack traces.
753 If you want to push the scalability limits of your code, you might
754 want to reduce the ``contexts/stack-size`` item. Its default value is
755 8192 (in KiB), while our Chord simulation works with stacks as small
756 as 16 KiB, for example. This *setting is ignored* when using the
757 thread factory. Instead, you should compile SimGrid and your
758 application with ``-fsplit-stack``. Note that this compilation flag is
759 not compatible with the model-checker right now.
761 The operating system should only allocate memory for the pages of the
762 stack which are actually used and you might not need to use this in
763 most cases. However, this setting is very important when using the
764 model checker (see :ref:`options_mc_perf`).
766 .. _cfg=contexts/guard-size:
768 Disabling Stack Guard Pages
769 ...........................
771 **Option** ``contexts/guard-size`` **Default** 1 page in most case (0 pages on Windows or with MC)
773 Unless you use the threads context factory (see
774 :ref:`cfg=contexts/factory`), a stack guard page is usually used
775 which prevents the stack of a given actor from overflowing on another
776 stack. But the performance impact may become prohibitive when the
777 amount of actors increases. The option ``contexts/guard-size`` is the
778 number of stack guard pages used. By setting it to 0, no guard pages
779 will be used: in this case, you should avoid using small stacks (with
780 :ref:`contexts/stack-size <cfg=contexts/stack-size>`) as the stack
781 will silently overflow on other parts of the memory.
783 When no stack guard page is created, stacks may then silently overflow
784 on other parts of the memory if their size is too small for the
787 .. _cfg=contexts/nthreads:
788 .. _cfg=contexts/parallel-threshold:
789 .. _cfg=contexts/synchro:
791 Running User Code in Parallel
792 .............................
794 Parallel execution of the user code is only considered stable in
795 SimGrid v3.7 and higher, and mostly for MSG simulations. SMPI
796 simulations may well fail in parallel mode. It is described in
797 `INRIA RR-7653 <http://hal.inria.fr/inria-00602216/>`_.
799 If you are using the **ucontext** or **raw** context factories, you can
800 request to execute the user code in parallel. Several threads are
801 launched, each of them handling as much user contexts at each run. To
802 actiave this, set the ``contexts/nthreads`` item to the amount of
803 cores that you have in your computer (or lower than 1 to have
804 the amount of cores auto-detected).
806 Even if you asked several worker threads using the previous option,
807 you can request to start the parallel execution (and pay the
808 associated synchronization costs) only if the potential parallelism is
809 large enough. For that, set the ``contexts/parallel-threshold``
810 item to the minimal amount of user contexts needed to start the
811 parallel execution. In any given simulation round, if that amount is
812 not reached, the contexts will be run sequentially directly by the
813 main thread (thus saving the synchronization costs). Note that this
814 option is mainly useful when the grain of the user code is very fine,
815 because our synchronization is now very efficient.
817 When parallel execution is activated, you can choose the
818 synchronization schema used with the ``contexts/synchro`` item,
819 which value is either:
821 - **futex:** ultra optimized synchronisation schema, based on futexes
822 (fast user-mode mutexes), and thus only available on Linux systems.
823 This is the default mode when available.
824 - **posix:** slow but portable synchronisation using only POSIX
826 - **busy_wait:** not really a synchronisation: the worker threads
827 constantly request new contexts to execute. It should be the most
828 efficient synchronisation schema, but it loads all the cores of
829 your machine for no good reason. You probably prefer the other less
832 Configuring the Tracing
833 -----------------------
835 The :ref:`tracing subsystem <outcomes_vizu>` can be configured in
836 several different ways depending on the nature of the simulator (MSG,
837 SimDag, SMPI) and the kind of traces that need to be obtained. See the
838 :ref:`Tracing Configuration Options subsection
839 <tracing_tracing_options>` to get a detailed description of each
840 configuration option.
842 We detail here a simple way to get the traces working for you, even if
843 you never used the tracing API.
846 - Any SimGrid-based simulator (MSG, SimDag, SMPI, ...) and raw traces:
848 .. code-block:: shell
850 --cfg=tracing:yes --cfg=tracing/uncategorized:yes --cfg=triva/uncategorized:uncat.plist
852 The first parameter activates the tracing subsystem, the second
853 tells it to trace host and link utilization (without any
854 categorization) and the third creates a graph configuration file to
855 configure Triva when analysing the resulting trace file.
857 - MSG or SimDag-based simulator and categorized traces (you need to
858 declare categories and classify your tasks according to them)
860 .. code-block:: shell
862 --cfg=tracing:yes --cfg=tracing/categorized:yes --cfg=triva/categorized:cat.plist
864 The first parameter activates the tracing subsystem, the second
865 tells it to trace host and link categorized utilization and the
866 third creates a graph configuration file to configure Triva when
867 analysing the resulting trace file.
869 - SMPI simulator and traces for a space/time view:
871 .. code-block:: shell
875 The `-trace` parameter for the smpirun script runs the simulation
876 with ``--cfg=tracing:yes --cfg=tracing/smpi:yes``. Check the
877 smpirun's `-help` parameter for additional tracing options.
879 Sometimes you might want to put additional information on the trace to
880 correctly identify them later, or to provide data that can be used to
881 reproduce an experiment. You have two ways to do that:
883 - Add a string on top of the trace file as comment:
885 .. code-block:: shell
887 --cfg=tracing/comment:my_simulation_identifier
889 - Add the contents of a textual file on top of the trace file as comment:
891 .. code-block:: shell
893 --cfg=tracing/comment-file:my_file_with_additional_information.txt
895 Please, use these two parameters (for comments) to make reproducible
896 simulations. For additional details about this and all tracing
897 options, check See the :ref:`tracing_tracing_options`.
902 .. _cfg=msg/debug-multiple-use:
907 **Option** ``msg/debug-multiple-use`` **Default:** off
909 Sometimes your application may try to send a task that is still being
910 executed somewhere else, making it impossible to send this task. However,
911 for debugging purposes, one may want to know what the other host is/was
912 doing. This option shows a backtrace of the other process.
917 The SMPI interface provides several specific configuration items.
918 These are uneasy to see since the code is usually launched through the
919 ``smiprun`` script directly.
921 .. _cfg=smpi/host-speed:
922 .. _cfg=smpi/cpu-threshold:
923 .. _cfg=smpi/simulate-computation:
925 Automatic Benchmarking of SMPI Code
926 ...................................
928 In SMPI, the sequential code is automatically benchmarked, and these
929 computations are automatically reported to the simulator. That is to
930 say that if you have a large computation between a ``MPI_Recv()`` and
931 a ``MPI_Send()``, SMPI will automatically benchmark the duration of
932 this code, and create an execution task within the simulator to take
933 this into account. For that, the actual duration is measured on the
934 host machine and then scaled to the power of the corresponding
935 simulated machine. The variable ``smpi/host-speed`` allows to specify
936 the computational speed of the host machine (in flop/s) to use when
937 scaling the execution times. It defaults to 20000, but you really want
938 to update it to get accurate simulation results.
940 When the code is constituted of numerous consecutive MPI calls, the
941 previous mechanism feeds the simulation kernel with numerous tiny
942 computations. The ``smpi/cpu-threshold`` item becomes handy when this
943 impacts badly the simulation performance. It specifies a threshold (in
944 seconds) below which the execution chunks are not reported to the
945 simulation kernel (default value: 1e-6).
947 .. note:: The option ``smpi/cpu-threshold`` ignores any computation
948 time spent below this threshold. SMPI does not consider the
949 `amount` of these computations; there is no offset for this. Hence,
950 a value that is too small, may lead to unreliable simulation
953 In some cases, however, one may wish to disable simulation of
954 application computation. This is the case when SMPI is used not to
955 simulate an MPI applications, but instead an MPI code that performs
956 "live replay" of another MPI app (e.g., ScalaTrace's replay tool,
957 various on-line simulators that run an app at scale). In this case the
958 computation of the replay/simulation logic should not be simulated by
959 SMPI. Instead, the replay tool or on-line simulator will issue
960 "computation events", which correspond to the actual MPI simulation
961 being replayed/simulated. At the moment, these computation events can
962 be simulated using SMPI by calling internal smpi_execute*() functions.
964 To disable the benchmarking/simulation of computation in the simulated
965 application, the variable ``smpi/simulate-computation`` should be set
966 to no. This option just ignores the timings in your simulation; it
967 still executes the computations itself. If you want to stop SMPI from
968 doing that, you should check the SMPI_SAMPLE macros, documented in
969 Section :ref:`SMPI_adapting_speed`.
971 +------------------------------------+-------------------------+-----------------------------+
972 | Solution | Computations executed? | Computations simulated? |
973 +====================================+=========================+=============================+
974 | --cfg=smpi/simulate-computation:no | Yes | Never |
975 +------------------------------------+-------------------------+-----------------------------+
976 | --cfg=smpi/cpu-threshold:42 | Yes, in all cases | If it lasts over 42 seconds |
977 +------------------------------------+-------------------------+-----------------------------+
978 | SMPI_SAMPLE() macro | Only once per loop nest | Always |
979 +------------------------------------+-------------------------+-----------------------------+
981 .. _cfg=smpi/comp-adjustment-file:
983 Slow-down or speed-up parts of your code
984 ........................................
986 **Option** ``smpi/comp-adjustment-file:`` **Default:** unset
988 This option allows you to pass a file that contains two columns: The
989 first column defines the section that will be subject to a speedup;
990 the second column is the speedup. For instance:
992 .. code-block:: shell
995 "exchange_1.f:30:exchange_1.f:130",1.18244559422142
997 The first line is the header - you must include it. The following
998 line means that the code between two consecutive MPI calls on line 30
999 in exchange_1.f and line 130 in exchange_1.f should receive a speedup
1000 of 1.18244559422142. The value for the second column is therefore a
1001 speedup, if it is larger than 1 and a slow-down if it is smaller
1002 than 1. Nothing will be changed if it is equal to 1.
1004 Of course, you can set any arbitrary filenames you want (so the start
1005 and end don't have to be in the same file), but be aware that this
1006 mechanism only supports `consecutive calls!`
1008 Please note that you must pass the ``-trace-call-location`` flag to
1009 smpicc or smpiff, respectively. This flag activates some internal
1010 macro definitions that help with obtaining the call location.
1012 .. _cfg=smpi/bw-factor:
1017 **Option** ``smpi/bw-factor``
1018 |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
1020 The possible throughput of network links is often dependent on the
1021 message sizes, as protocols may adapt to different message sizes. With
1022 this option, a series of message sizes and factors are given, helping
1023 the simulation to be more realistic. For instance, the current default
1024 value means that messages with size 65472 and more will get a total of
1025 MAX_BANDWIDTH*0.940694, messages of size 15424 to 65471 will get
1026 MAX_BANDWIDTH*0.697866 and so on (where MAX_BANDWIDTH denotes the
1027 bandwidth of the link).
1029 An experimental script to compute these factors is available online. See
1030 https://framagit.org/simgrid/platform-calibration/
1031 https://simgrid.org/contrib/smpi-saturation-doc.html
1033 .. _cfg=smpi/display-timing:
1035 Reporting Simulation Time
1036 .........................
1038 **Option** ``smpi/display-timing`` **Default:** 0 (false)
1040 Most of the time, you run MPI code with SMPI to compute the time it
1041 would take to run it on a platform. But since the code is run through
1042 the ``smpirun`` script, you don't have any control on the launcher
1043 code, making it difficult to report the simulated time when the
1044 simulation ends. If you enable the ``smpi/display-timing`` item,
1045 ``smpirun`` will display this information when the simulation
1048 .. _cfg=smpi/keep-temps:
1050 Keeping temporary files after simulation
1051 ........................................
1053 **Option** ``smpi/keep-temps`` **default:** 0 (false)
1055 SMPI usually generates a lot of temporary files that are cleaned after
1056 use. This option request to preserve them, for example to debug or
1057 profile your code. Indeed, the binary files are removed very early
1058 under the dlopen privatization schema, which tend to fool the
1061 .. _cfg=smpi/lat-factor:
1066 **Option** ``smpi/lat-factor`` |br|
1067 **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
1069 The motivation and syntax for this option is identical to the motivation/syntax
1070 of :ref:`cfg=smpi/bw-factor`.
1072 There is an important difference, though: While smpi/bw-factor `reduces` the
1073 actual bandwidth (i.e., values between 0 and 1 are valid), latency factors
1074 increase the latency, i.e., values larger than or equal to 1 are valid here.
1076 .. _cfg=smpi/papi-events:
1078 Trace hardware counters with PAPI
1079 .................................
1081 **Option** ``smpi/papi-events`` **default:** unset
1083 When the PAPI support was compiled in SimGrid, this option takes the
1084 names of PAPI counters and adds their respective values to the trace
1085 files (See Section :ref:`tracing_tracing_options`).
1089 This feature currently requires superuser privileges, as registers
1090 are queried. Only use this feature with code you trust! Call
1091 smpirun for instance via ``smpirun -wrapper "sudo "
1092 <your-parameters>`` or run ``sudo sh -c "echo 0 >
1093 /proc/sys/kernel/perf_event_paranoid"`` In the later case, sudo
1094 will not be required.
1096 It is planned to make this feature available on a per-process (or per-thread?) basis.
1097 The first draft, however, just implements a "global" (i.e., for all processes) set
1098 of counters, the "default" set.
1100 .. code-block:: shell
1102 --cfg=smpi/papi-events:"default:PAPI_L3_LDM:PAPI_L2_LDM"
1104 .. _cfg=smpi/privatization:
1106 Automatic Privatization of Global Variables
1107 ...........................................
1109 **Option** ``smpi/privatization`` **default:** "dlopen" (when using smpirun)
1111 MPI executables are usually meant to be executed in separated
1112 processes, but SMPI is executed in only one process. Global variables
1113 from executables will be placed in the same memory zone and shared
1114 between processes, causing intricate bugs. Several options are
1115 possible to avoid this, as described in the main `SMPI publication
1116 <https://hal.inria.fr/hal-01415484>`_ and in the :ref:`SMPI
1117 documentation <SMPI_what_globals>`. SimGrid provides two ways of
1118 automatically privatizing the globals, and this option allows to
1119 choose between them.
1121 - **no** (default when not using smpirun): Do not automatically
1122 privatize variables. Pass ``-no-privatize`` to smpirun to disable
1124 - **dlopen** or **yes** (default when using smpirun): Link multiple
1125 times against the binary.
1126 - **mmap** (slower, but maybe somewhat more stable):
1127 Runtime automatic switching of the data segments.
1130 This configuration option cannot be set in your platform file. You can only
1131 pass it as an argument to smpirun.
1133 .. _cfg=smpi/privatize-libs:
1135 Automatic privatization of global variables inside external libraries
1136 .....................................................................
1138 **Option** ``smpi/privatize-libs`` **default:** unset
1140 **Linux/BSD only:** When using dlopen (default) privatization,
1141 privatize specific shared libraries with internal global variables, if
1142 they can't be linked statically. For example libgfortran is usually
1143 used for Fortran I/O and indexes in files can be mixed up.
1145 Multiple libraries can be given, semicolon separated.
1147 This configuration option can only use either full paths to libraries,
1148 or full names. Check with ldd the name of the library you want to
1151 .. code-block:: shell
1155 libgfortran.so.3 => /usr/lib/x86_64-linux-gnu/libgfortran.so.3 (0x00007fbb4d91b000)
1158 Then you can use ``--cfg=smpi/privatize-libs:libgfortran.so.3``
1159 or ``--cfg=smpi/privatize-libs:/usr/lib/x86_64-linux-gnu/libgfortran.so.3``,
1160 but not ``libgfortran`` nor ``libgfortran.so``.
1162 .. _cfg=smpi/send-is-detached-thresh:
1164 Simulating MPI detached send
1165 ............................
1167 **Option** ``smpi/send-is-detached-thresh`` **default:** 65536
1169 This threshold specifies the size in bytes under which the send will
1170 return immediately. This is different from the threshold detailed in
1171 :ref:`options_model_network_asyncsend` because the message is not
1172 effectively sent when the send is posted. SMPI still waits for the
1173 correspondant receive to be posted to perform the communication
1176 .. _cfg=smpi/coll-selector:
1178 Simulating MPI collective algorithms
1179 ....................................
1181 **Option** ``smpi/coll-selector`` **Possible values:** naive (default), ompi, mpich
1183 SMPI implements more than 100 different algorithms for MPI collective
1184 communication, to accurately simulate the behavior of most of the
1185 existing MPI libraries. The ``smpi/coll-selector`` item can be used to
1186 use the decision logic of either OpenMPI or MPICH libraries (by
1187 default SMPI uses naive version of collective operations).
1189 Each collective operation can be manually selected with a
1190 ``smpi/collective_name:algo_name``. Available algorithms are listed in
1191 :ref:`SMPI_use_colls`.
1193 .. TODO:: All available collective algorithms will be made available
1194 via the ``smpirun --help-coll`` command.
1196 .. _cfg=smpi/iprobe:
1198 Inject constant times for MPI_Iprobe
1199 ....................................
1201 **Option** ``smpi/iprobe`` **default:** 0.0001
1203 The behavior and motivation for this configuration option is identical
1204 with :ref:`smpi/test <cfg=smpi/test>`, but for the function
1207 .. _cfg=smpi/iprobe-cpu-usage:
1209 Reduce speed for iprobe calls
1210 .............................
1212 **Option** ``smpi/iprobe-cpu-usage`` **default:** 1 (no change)
1214 MPI_Iprobe calls can be heavily used in applications. To account
1215 correctly for the energy cores spend probing, it is necessary to
1216 reduce the load that these calls cause inside SimGrid.
1218 For instance, we measured a max power consumption of 220 W for a
1219 particular application but only 180 W while this application was
1220 probing. Hence, the correct factor that should be passed to this
1221 option would be 180/220 = 0.81.
1225 Inject constant times for MPI_Init
1226 ..................................
1228 **Option** ``smpi/init`` **default:** 0
1230 The behavior and motivation for this configuration option is identical
1231 with :ref:`smpi/test <cfg=smpi/test>`, but for the function ``MPI_Init()``.
1235 Inject constant times for MPI_Isend()
1236 .....................................
1238 **Option** ``smpi/ois``
1240 The behavior and motivation for this configuration option is identical
1241 with :ref:`smpi/os <cfg=smpi/os>`, but for the function ``MPI_Isend()``.
1245 Inject constant times for MPI_send()
1246 ....................................
1248 **Option** ``smpi/os``
1250 In several network models such as LogP, send (MPI_Send, MPI_Isend) and
1251 receive (MPI_Recv) operations incur costs (i.e., they consume CPU
1252 time). SMPI can factor these costs in as well, but the user has to
1253 configure SMPI accordingly as these values may vary by machine. This
1254 can be done by using ``smpi/os`` for MPI_Send operations; for MPI_Isend
1255 and MPI_Recv, use ``smpi/ois`` and ``smpi/or``, respectively. These work
1256 exactly as ``smpi/ois``.
1258 This item can consist of multiple sections; each section takes three
1259 values, for example ``1:3:2;10:5:1``. The sections are divided by ";"
1260 so this example contains two sections. Furthermore, each section
1261 consists of three values.
1263 1. The first value denotes the minimum size for this section to take effect;
1264 read it as "if message size is greater than this value (and other section has a larger
1265 first value that is also smaller than the message size), use this".
1266 In the first section above, this value is "1".
1268 2. The second value is the startup time; this is a constant value that will always
1269 be charged, no matter what the size of the message. In the first section above,
1272 3. The third value is the `per-byte` cost. That is, it is charged for every
1273 byte of the message (incurring cost messageSize*cost_per_byte)
1274 and hence accounts also for larger messages. In the first
1275 section of the example above, this value is "2".
1277 Now, SMPI always checks which section it should take for a given
1278 message; that is, if a message of size 11 is sent with the
1279 configuration of the example above, only the second section will be
1280 used, not the first, as the first value of the second section is
1281 closer to the message size. Hence, when ``smpi/os=1:3:2;10:5:1``, a
1282 message of size 11 incurs the following cost inside MPI_Send:
1283 ``5+11*1`` because 5 is the startup cost and 1 is the cost per byte.
1285 Note that the order of sections can be arbitrary; they will be ordered internally.
1289 Inject constant times for MPI_Recv()
1290 ....................................
1292 **Option** ``smpi/or``
1294 The behavior and motivation for this configuration option is identical
1295 with :ref:`smpi/os <cfg=smpi/os>`, but for the function ``MPI_Recv()``.
1298 .. _cfg=smpi/grow-injected-times:
1300 Inject constant times for MPI_Test
1301 ..................................
1303 **Option** ``smpi/test`` **default:** 0.0001
1305 By setting this option, you can control the amount of time a process
1306 sleeps when MPI_Test() is called; this is important, because SimGrid
1307 normally only advances the time while communication is happening and
1308 thus, MPI_Test will not add to the time, resulting in a deadlock if
1309 used as a break-condition as in the following example:
1314 MPI_Test(request, flag, status);
1318 To speed up execution, we use a counter to keep track on how often we
1319 already checked if the handle is now valid or not. Hence, we actually
1320 use counter*SLEEP_TIME, that is, the time MPI_Test() causes the
1321 process to sleep increases linearly with the number of previously
1322 failed tests. This behavior can be disabled by setting
1323 ``smpi/grow-injected-times`` to **no**. This will also disable this
1324 behavior for MPI_Iprobe.
1326 .. _cfg=smpi/shared-malloc:
1327 .. _cfg=smpi/shared-malloc-hugepage:
1332 **Option** ``smpi/shared-malloc`` **Possible values:** global (default), local
1334 If your simulation consumes too much memory, you may want to modify
1335 your code so that the working areas are shared by all MPI ranks. For
1336 example, in a bloc-cyclic matrix multiplication, you will only
1337 allocate one set of blocs, and every processes will share them.
1338 Naturally, this will lead to very wrong results, but this will save a
1339 lot of memory so this is still desirable for some studies. For more on
1340 the motivation for that feature, please refer to the `relevant section
1341 <https://simgrid.github.io/SMPI_CourseWare/topic_understanding_performance/matrixmultiplication>`_
1342 of the SMPI CourseWare (see Activity #2.2 of the pointed
1343 assignment). In practice, change the call to malloc() and free() into
1344 SMPI_SHARED_MALLOC() and SMPI_SHARED_FREE().
1346 SMPI provides two algorithms for this feature. The first one, called
1347 ``local``, allocates one bloc per call to SMPI_SHARED_MALLOC() in your
1348 code (each call location gets its own bloc) and this bloc is shared
1349 amongst all MPI ranks. This is implemented with the shm_* functions
1350 to create a new POSIX shared memory object (kept in RAM, in /dev/shm)
1351 for each shared bloc.
1353 With the ``global`` algorithm, each call to SMPI_SHARED_MALLOC()
1354 returns a new adress, but it only points to a shadow bloc: its memory
1355 area is mapped on a 1MiB file on disk. If the returned bloc is of size
1356 N MiB, then the same file is mapped N times to cover the whole bloc.
1357 At the end, no matter how many SMPI_SHARED_MALLOC you do, this will
1358 only consume 1 MiB in memory.
1360 You can disable this behavior and come back to regular mallocs (for
1361 example for debugging purposes) using @c "no" as a value.
1363 If you want to keep private some parts of the buffer, for instance if these
1364 parts are used by the application logic and should not be corrupted, you
1365 can use SMPI_PARTIAL_SHARED_MALLOC(size, offsets, offsets_count). Example:
1369 mem = SMPI_PARTIAL_SHARED_MALLOC(500, {27,42 , 100,200}, 2);
1371 This will allocate 500 bytes to mem, such that mem[27..41] and
1372 mem[100..199] are shared while other area remain private.
1374 Then, it can be deallocated by calling SMPI_SHARED_FREE(mem).
1376 When smpi/shared-malloc:global is used, the memory consumption problem
1377 is solved, but it may induce too much load on the kernel's pages table.
1378 In this case, you should use huge pages so that we create only one
1379 entry per Mb of malloced data instead of one entry per 4k.
1380 To activate this, you must mount a hugetlbfs on your system and allocate
1381 at least one huge page:
1383 .. code-block:: shell
1386 sudo mount none /home/huge -t hugetlbfs -o rw,mode=0777
1387 sudo sh -c 'echo 1 > /proc/sys/vm/nr_hugepages' # echo more if you need more
1389 Then, you can pass the option
1390 ``--cfg=smpi/shared-malloc-hugepage:/home/huge`` to smpirun to
1391 actually activate the huge page support in shared mallocs.
1395 Inject constant times for MPI_Wtime, gettimeofday and clock_gettime
1396 ...................................................................
1398 **Option** ``smpi/wtime`` **default:** 10 ns
1400 This option controls the amount of (simulated) time spent in calls to
1401 MPI_Wtime(), gettimeofday() and clock_gettime(). If you set this value
1402 to 0, the simulated clock is not advanced in these calls, which leads
1403 to issue if your application contains such a loop:
1407 while(MPI_Wtime() < some_time_bound) {
1408 /* some tests, with no communication nor computation */
1411 When the option smpi/wtime is set to 0, the time advances only on
1412 communications and computations, so the previous code results in an
1413 infinite loop: the current [simulated] time will never reach
1414 ``some_time_bound``. This infinite loop is avoided when that option
1415 is set to a small amount, as it is by default since SimGrid v3.21.
1417 Note that if your application does not contain any loop depending on
1418 the current time only, then setting this option to a non-zero value
1419 will slow down your simulations by a tiny bit: the simulation loop has
1420 to be broken and reset each time your code ask for the current time.
1421 If the simulation speed really matters to you, you can avoid this
1422 extra delay by setting smpi/wtime to 0.
1424 Other Configurations
1425 --------------------
1427 .. _cfg=debug/clean-atexit:
1429 Cleanup at Termination
1430 ......................
1432 **Option** ``debug/clean-atexit`` **default:** on
1434 If your code is segfaulting during its finalization, it may help to
1435 disable this option to request SimGrid to not attempt any cleanups at
1436 the end of the simulation. Since the Unix process is ending anyway,
1437 the operating system will wipe it all.
1444 **Option** ``path`` **default:** . (current dir)
1446 It is possible to specify a list of directories to search into for the
1447 trace files (see :ref:`pf_trace`) by using this configuration
1448 item. To add several directory to the path, set the configuration
1449 item several times, as in ``--cfg=path:toto --cfg=path:tutu``
1451 .. _cfg=debug/breakpoint:
1456 **Option** ``debug/breakpoint`` **default:** unset
1458 This configuration option sets a breakpoint: when the simulated clock
1459 reaches the given time, a SIGTRAP is raised. This can be used to stop
1460 the execution and get a backtrace with a debugger.
1462 It is also possible to set the breakpoint from inside the debugger, by
1463 writing in global variable simgrid::simix::breakpoint. For example,
1466 .. code-block:: shell
1468 set variable simgrid::simix::breakpoint = 3.1416
1470 .. _cfg=debug/verbose-exit:
1475 **Option** ``debug/verbose-exit`` **default:** on
1477 By default, when Ctrl-C is pressed, the status of all existing actors
1478 is displayed before exiting the simulation. This is very useful to
1479 debug your code, but it can reveal troublesome if you have many
1480 actors. Set this configuration item to **off** to disable this
1483 .. _cfg=exception/cutpath:
1485 Truncate local path from exception backtrace
1486 ............................................
1488 **Option** ``exception/cutpath`` **default:** off
1490 This configuration option is used to remove the path from the
1491 backtrace shown when an exception is thrown. This is mainly useful for
1492 the tests: the full file path makes the tests not reproducible because
1493 the path of source files depend of the build settings. That would
1494 break most of our tests as we keep comparing output.
1496 Logging Configuration
1497 ---------------------
1499 It can be done by using XBT. Go to :ref:`XBT_log` for more details.