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");
75 Existing Configuration Items
76 ----------------------------
79 The full list can be retrieved by passing ``--help`` and
80 ``--help-cfg`` to an executable that uses SimGrid.
82 - **clean-atexit:** :ref:`cfg=clean-atexit`
84 - **contexts/factory:** :ref:`cfg=contexts/factory`
85 - **contexts/guard-size:** :ref:`cfg=contexts/guard-size`
86 - **contexts/nthreads:** :ref:`cfg=contexts/nthreads`
87 - **contexts/parallel-threshold:** :ref:`cfg=contexts/parallel-threshold`
88 - **contexts/stack-size:** :ref:`cfg=contexts/stack-size`
89 - **contexts/synchro:** :ref:`cfg=contexts/synchro`
91 - **cpu/maxmin-selective-update:** :ref:`Cpu Optimization Level <options_model_optim>`
92 - **cpu/model:** :ref:`options_model_select`
93 - **cpu/optim:** :ref:`Cpu Optimization Level <options_model_optim>`
95 - **exception/cutpath:** :ref:`cfg=exception/cutpath`
97 - **host/model:** :ref:`options_model_select`
99 - **maxmin/precision:** :ref:`cfg=maxmin/precision`
100 - **maxmin/concurrency-limit:** :ref:`cfg=maxmin/concurrency-limit`
102 - **msg/debug-multiple-use:** :ref:`cfg=msg/debug-multiple-use`
104 - **model-check:** :ref:`options_modelchecking`
105 - **model-check/checkpoint:** :ref:`cfg=model-check/checkpoint`
106 - **model-check/communications-determinism:** :ref:`cfg=model-check/communications-determinism`
107 - **model-check/dot-output:** :ref:`cfg=model-check/dot-output`
108 - **model-check/hash:** :ref:`cfg=model-checker/hash`
109 - **model-check/max-depth:** :ref:`cfg=model-check/max-depth`
110 - **model-check/property:** :ref:`cfg=model-check/property`
111 - **model-check/record:** :ref:`cfg=model-check/record`
112 - **model-check/reduction:** :ref:`cfg=model-check/reduction`
113 - **model-check/replay:** :ref:`cfg=model-check/replay`
114 - **model-check/send-determinism:** :ref:`cfg=model-check/send-determinism`
115 - **model-check/sparse-checkpoint:** :ref:`cfg=model-check/sparse-checkpoint`
116 - **model-check/termination:** :ref:`cfg=model-check/termination`
117 - **model-check/timeout:** :ref:`cfg=model-check/timeout`
118 - **model-check/visited:** :ref:`cfg=model-check/visited`
120 - **network/bandwidth-factor:** :ref:`cfg=network/bandwidth-factor`
121 - **network/crosstraffic:** :ref:`cfg=network/crosstraffic`
122 - **network/latency-factor:** :ref:`cfg=network/latency-factor`
123 - **network/maxmin-selective-update:** :ref:`Network Optimization Level <options_model_optim>`
124 - **network/model:** :ref:`options_model_select`
125 - **network/optim:** :ref:`Network Optimization Level <options_model_optim>`
126 - **network/TCP-gamma:** :ref:`cfg=network/TCP-gamma`
127 - **network/weight-S:** :ref:`cfg=network/weight-S`
129 - **ns3/TcpModel:** :ref:`options_pls`
130 - **path:** :ref:`cfg=path`
131 - **plugin:** :ref:`cfg=plugin`
133 - **simix/breakpoint:** :ref:`cfg=simix/breakpoint`
135 - **storage/max_file_descriptors:** :ref:`cfg=storage/max_file_descriptors`
137 - **surf/precision:** :ref:`cfg=surf/precision`
139 - **For collective operations of SMPI,** please refer to Section :ref:`cfg=smpi/coll-selector`
140 - **smpi/async-small-thresh:** :ref:`cfg=smpi/async-small-thresh`
141 - **smpi/bw-factor:** :ref:`cfg=smpi/bw-factor`
142 - **smpi/coll-selector:** :ref:`cfg=smpi/coll-selector`
143 - **smpi/comp-adjustment-file:** :ref:`cfg=smpi/comp-adjustment-file`
144 - **smpi/cpu-threshold:** :ref:`cfg=smpi/cpu-threshold`
145 - **smpi/display-timing:** :ref:`cfg=smpi/display-timing`
146 - **smpi/grow-injected-times:** :ref:`cfg=smpi/grow-injected-times`
147 - **smpi/host-speed:** :ref:`cfg=smpi/host-speed`
148 - **smpi/IB-penalty-factors:** :ref:`cfg=smpi/IB-penalty-factors`
149 - **smpi/iprobe:** :ref:`cfg=smpi/iprobe`
150 - **smpi/iprobe-cpu-usage:** :ref:`cfg=smpi/iprobe-cpu-usage`
151 - **smpi/init:** :ref:`cfg=smpi/init`
152 - **smpi/keep-temps:** :ref:`cfg=smpi/keep-temps`
153 - **smpi/lat-factor:** :ref:`cfg=smpi/lat-factor`
154 - **smpi/ois:** :ref:`cfg=smpi/ois`
155 - **smpi/or:** :ref:`cfg=smpi/or`
156 - **smpi/os:** :ref:`cfg=smpi/os`
157 - **smpi/papi-events:** :ref:`cfg=smpi/papi-events`
158 - **smpi/privatization:** :ref:`cfg=smpi/privatization`
159 - **smpi/privatize-libs:** :ref:`cfg=smpi/privatize-libs`
160 - **smpi/send-is-detached-thresh:** :ref:`cfg=smpi/send-is-detached-thresh`
161 - **smpi/shared-malloc:** :ref:`cfg=smpi/shared-malloc`
162 - **smpi/shared-malloc-hugepage:** :ref:`cfg=smpi/shared-malloc-hugepage`
163 - **smpi/simulate-computation:** :ref:`cfg=smpi/simulate-computation`
164 - **smpi/test:** :ref:`cfg=smpi/test`
165 - **smpi/wtime:** :ref:`cfg=smpi/wtime`
167 - **Tracing configuration options** can be found in Section :ref:`tracing_tracing_options`
169 - **storage/model:** :ref:`options_model_select`
170 - **verbose-exit:** :ref:`cfg=verbose-exit`
172 - **vm/model:** :ref:`options_model_select`
176 Configuring the Platform Models
177 -------------------------------
179 .. _options_model_select:
181 Choosing the Platform Models
182 ............................
184 SimGrid comes with several network, CPU and storage models built in,
185 and you can change the used model at runtime by changing the passed
186 configuration. The three main configuration items are given below.
187 For each of these items, passing the special ``help`` value gives you
188 a short description of all possible values (for example,
189 ``--cfg=network/model:help`` will present all provided network
190 models). Also, ``--help-models`` should provide information about all
191 models for all existing resources.
193 - ``network/model``: specify the used network model. Possible values:
195 - **LV08 (default one):** Realistic network analytic model
196 (slow-start modeled by multiplying latency by 13.01, bandwidth by
197 .97; bottleneck sharing uses a payload of S=20537 for evaluating
198 RTT). Described in `Accuracy Study and Improvement of Network
199 Simulation in the SimGrid Framework
200 <http://mescal.imag.fr/membres/arnaud.legrand/articles/simutools09.pdf>`_.
201 - **Constant:** Simplistic network model where all communication
202 take a constant time (one second). This model provides the lowest
203 realism, but is (marginally) faster.
204 - **SMPI:** Realistic network model specifically tailored for HPC
205 settings (accurate modeling of slow start with correction factors on
206 three intervals: < 1KiB, < 64 KiB, >= 64 KiB). This model can be
207 :ref:`further configured <options_model_network>`.
208 - **IB:** Realistic network model specifically tailored for HPC
209 settings with InfiniBand networks (accurate modeling contention
210 behavior, based on the model explained in `this PhD work
211 <http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf>`_.
212 This model can be :ref:`further configured <options_model_network>`.
213 - **CM02:** Legacy network analytic model. Very similar to LV08, but
214 without corrective factors. The timings of small messages are thus
215 poorly modeled. This model is described in `A Network Model for
216 Simulation of Grid Application
217 <ftp://ftp.ens-lyon.fr/pub/LIP/Rapports/RR/RR2002/RR2002-40.ps.gz>`_.
218 - **Reno/Reno2/Vegas:** Models from Steven H. Low using lagrange_solve instead of
219 lmm_solve (experts only; check the code for more info).
220 - **NS3** (only available if you compiled SimGrid accordingly):
221 Use the packet-level network
222 simulators as network models (see :ref:`pls_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" (NS3 default)
425 When using NS3, there is an extra item ``ns3/TcpModel``, corresponding
426 to the ``ns3::TcpL4Protocol::SocketType`` configuration item in
427 NS3. The only valid values (enforced on the SimGrid side) are
428 'default' (no change to the NS3 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 .. _cfg=model-check/sparse-checkpoint:
620 Incremental Checkpoints
621 .......................
623 When the model-checker is configured to take a snapshot of each
624 explored state (with the ``model-checker/visited`` item), the memory
625 consumption can rapidly reach GiB ou Tib of memory. However, for many
626 workloads, the memory does not change much between different snapshots
627 and taking a complete copy of each snapshot is a waste of memory.
629 The ``model-check/sparse-checkpoint`` option item can be set to
630 **yes** to avoid making a complete copy of each snapshot. Instead,
631 each snapshot will be decomposed in blocks which will be stored
632 separately. If multiple snapshots share the same block (or if the
633 same block is used in the same snapshot), the same copy of the block
634 will be shared leading to a reduction of the memory footprint.
636 For many applications, this option considerably reduces the memory
637 consumption. In somes cases, the model-checker might be slightly
638 slower because of the time taken to manage the metadata about the
639 blocks. In other cases however, this snapshotting strategy will be
640 much faster by reducing the cache consumption. When the memory
641 consumption is important, by avoiding to hit the swap or reducing the
642 swap usage, this option might be much faster than the basic
643 snapshotting strategy.
645 This option is currently disabled by default.
647 Verification Performance Considerations
648 .......................................
650 The size of the stacks can have a huge impact on the memory
651 consumption when using model-checking. By default, each snapshot will
652 save a copy of the whole stacks and not only of the part which is
653 really meaningful: you should expect the contribution of the memory
654 consumption of the snapshots to be @f$ @mbox{number of processes}
655 @times @mbox{stack size} @times @mbox{number of states} @f$.
657 The ``model-check/sparse-checkpoint`` can be used to reduce the memory
658 consumption by trying to share memory between the different snapshots.
660 When compiled against the model checker, the stacks are not
661 protected with guards: if the stack size is too small for your
662 application, the stack will silently overflow on other parts of the
663 memory (see :ref:`contexts/guard-size <cfg=contexts/guard-size>`).
665 .. _cfg=model-checker/hash:
670 Usually most of the time of the model-checker is spent comparing states. This
671 process is complicated and consumes a lot of bandwidth and cache.
672 In order to speedup the state comparison, the experimental ``model-checker/hash``
673 configuration item enables the computation of a hash summarizing as much
674 information of the state as possible into a single value. This hash can be used
675 to avoid most of the comparisons: the costly comparison is then only used when
676 the hashes are identical.
678 Currently most of the state is not included in the hash because the
679 implementation was found to be buggy and this options is not as useful as
680 it could be. For this reason, it is currently disabled by default.
682 .. _cfg=model-check/record:
683 .. _cfg=model-check/replay:
685 Record/Replay of Verification
686 .............................
688 As the model-checker keeps jumping at different places in the execution graph,
689 it is difficult to understand what happens when trying to debug an application
690 under the model-checker. Event the output of the program is difficult to
691 interpret. Moreover, the model-checker does not behave nicely with advanced
692 debugging tools such as valgrind. For those reason, to identify a trajectory
693 in the execution graph with the model-checker and replay this trajcetory and
694 without the model-checker black-magic but with more standard tools
695 (such as a debugger, valgrind, etc.). For this reason, Simgrid implements an
696 experimental record/replay functionnality in order to record a trajectory with
697 the model-checker and replay it without the model-checker.
699 When the model-checker finds an interesting path in the application
700 execution graph (where a safety or liveness property is violated), it
701 can generate an identifier for this path. To enable this behavious the
702 ``model-check/record`` must be set to **yes**, which is not the case
705 Here is an example of output:
707 .. code-block:: shell
709 [ 0.000000] (0:@) Check a safety property
710 [ 0.000000] (0:@) **************************
711 [ 0.000000] (0:@) *** PROPERTY NOT VALID ***
712 [ 0.000000] (0:@) **************************
713 [ 0.000000] (0:@) Counter-example execution trace:
714 [ 0.000000] (0:@) Path = 1/3;1/4
715 [ 0.000000] (0:@) [(1)Tremblay (app)] MC_RANDOM(3)
716 [ 0.000000] (0:@) [(1)Tremblay (app)] MC_RANDOM(4)
717 [ 0.000000] (0:@) Expanded states = 27
718 [ 0.000000] (0:@) Visited states = 68
719 [ 0.000000] (0:@) Executed transitions = 46
721 This path can then be replayed outside of the model-checker (and even
722 in non-MC build of simgrid) by setting the ``model-check/replay`` item
723 to the given path. The other options should be the same (but the
724 model-checker should be disabled).
726 The format and meaning of the path may change between different
727 releases so the same release of Simgrid should be used for the record
728 phase and the replay phase.
730 Configuring the User Code Virtualization
731 ----------------------------------------
733 .. _cfg=contexts/factory:
735 Selecting the Virtualization Factory
736 ....................................
738 **Option** contexts/factory **Default:** "raw"
740 In SimGrid, the user code is virtualized in a specific mechanism that
741 allows the simulation kernel to control its execution: when a user
742 process requires a blocking action (such as sending a message), it is
743 interrupted, and only gets released when the simulated clock reaches
744 the point where the blocking operation is done. This is explained
745 graphically in the `relevant tutorial, available online
746 <http://simgrid.gforge.inria.fr/tutorials/simgrid-simix-101.pdf>`_.
748 In SimGrid, the containers in which user processes are virtualized are
749 called contexts. Several context factory are provided, and you can
750 select the one you want to use with the ``contexts/factory``
751 configuration item. Some of the following may not exist on your
752 machine because of portability issues. In any case, the default one
753 should be the most effcient one (please report bugs if the
754 auto-detection fails for you). They are approximately sorted here from
755 the slowest to the most efficient:
757 - **thread:** very slow factory using full featured threads (either
758 pthreads or windows native threads). They are slow but very
759 standard. Some debuggers or profilers only work with this factory.
760 - **java:** Java applications are virtualized onto java threads (that
761 are regular pthreads registered to the JVM)
762 - **ucontext:** fast factory using System V contexts (Linux and FreeBSD only)
763 - **boost:** This uses the `context
764 implementation <http://www.boost.org/doc/libs/1_59_0/libs/context/doc/html/index.html>`_
765 of the boost library for a performance that is comparable to our
767 |br| Install the relevant library (e.g. with the
768 libboost-contexts-dev package on Debian/Ubuntu) and recompile
770 - **raw:** amazingly fast factory using a context switching mechanism
771 of our own, directly implemented in assembly (only available for x86
772 and amd64 platforms for now) and without any unneeded system call.
774 The main reason to change this setting is when the debugging tools get
775 fooled by the optimized context factories. Threads are the most
776 debugging-friendly contextes, as they allow to set breakpoints
777 anywhere with gdb and visualize backtraces for all processes, in order
778 to debug concurrency issues. Valgrind is also more comfortable with
779 threads, but it should be usable with all factories (Exception: the
780 callgrind tool really dislikes raw and ucontext factories).
782 .. _cfg=contexts/stack-size:
784 Adapting the Stack Size
785 .......................
787 **Option** ``contexts/stack-size`` **Default:** 8192 KiB
789 Each virtualized used process is executed using a specific system
790 stack. The size of this stack has a huge impact on the simulation
791 scalability, but its default value is rather large. This is because
792 the error messages that you get when the stack size is too small are
793 rather disturbing: this leads to stack overflow (overwriting other
794 stacks), leading to segfaults with corrupted stack traces.
796 If you want to push the scalability limits of your code, you might
797 want to reduce the ``contexts/stack-size`` item. Its default value is
798 8192 (in KiB), while our Chord simulation works with stacks as small
799 as 16 KiB, for example. For the thread factory, the default value is
800 the one of the system but you can still change it with this parameter.
802 The operating system should only allocate memory for the pages of the
803 stack which are actually used and you might not need to use this in
804 most cases. However, this setting is very important when using the
805 model checker (see :ref:`options_mc_perf`).
807 .. _cfg=contexts/guard-size:
809 Disabling Stack Guard Pages
810 ...........................
812 **Option** ``contexts/guard-size`` **Default** 1 page in most case (0 pages on Windows or with MC)
814 A stack guard page is usually used which prevents the stack of a given
815 actor from overflowing on another stack. But the performance impact
816 may become prohibitive when the amount of actors increases. The
817 option ``contexts/guard-size`` is the number of stack guard pages
818 used. By setting it to 0, no guard pages will be used: in this case,
819 you should avoid using small stacks (with :ref:`contexts/stack-size
820 <cfg=contexts/stack-size>`) as the stack will silently overflow on
821 other parts of the memory.
823 When no stack guard page is created, stacks may then silently overflow
824 on other parts of the memory if their size is too small for the
827 .. _cfg=contexts/nthreads:
828 .. _cfg=contexts/parallel-threshold:
829 .. _cfg=contexts/synchro:
831 Running User Code in Parallel
832 .............................
834 Parallel execution of the user code is only considered stable in
835 SimGrid v3.7 and higher, and mostly for MSG simulations. SMPI
836 simulations may well fail in parallel mode. It is described in
837 `INRIA RR-7653 <http://hal.inria.fr/inria-00602216/>`_.
839 If you are using the **ucontext** or **raw** context factories, you can
840 request to execute the user code in parallel. Several threads are
841 launched, each of them handling as much user contexts at each run. To
842 actiave this, set the ``contexts/nthreads`` item to the amount of
843 cores that you have in your computer (or lower than 1 to have
844 the amount of cores auto-detected).
846 Even if you asked several worker threads using the previous option,
847 you can request to start the parallel execution (and pay the
848 associated synchronization costs) only if the potential parallelism is
849 large enough. For that, set the ``contexts/parallel-threshold``
850 item to the minimal amount of user contexts needed to start the
851 parallel execution. In any given simulation round, if that amount is
852 not reached, the contexts will be run sequentially directly by the
853 main thread (thus saving the synchronization costs). Note that this
854 option is mainly useful when the grain of the user code is very fine,
855 because our synchronization is now very efficient.
857 When parallel execution is activated, you can choose the
858 synchronization schema used with the ``contexts/synchro`` item,
859 which value is either:
861 - **futex:** ultra optimized synchronisation schema, based on futexes
862 (fast user-mode mutexes), and thus only available on Linux systems.
863 This is the default mode when available.
864 - **posix:** slow but portable synchronisation using only POSIX
866 - **busy_wait:** not really a synchronisation: the worker threads
867 constantly request new contexts to execute. It should be the most
868 efficient synchronisation schema, but it loads all the cores of
869 your machine for no good reason. You probably prefer the other less
873 Configuring the Tracing
874 -----------------------
876 The :ref:`tracing subsystem <outcomes_vizu>` can be configured in
877 several different ways depending on the nature of the simulator (MSG,
878 SimDag, SMPI) and the kind of traces that need to be obtained. See the
879 :ref:`Tracing Configuration Options subsection
880 <tracing_tracing_options>` to get a detailed description of each
881 configuration option.
883 We detail here a simple way to get the traces working for you, even if
884 you never used the tracing API.
887 - Any SimGrid-based simulator (MSG, SimDag, SMPI, ...) and raw traces:
889 .. code-block:: shell
891 --cfg=tracing:yes --cfg=tracing/uncategorized:yes --cfg=triva/uncategorized:uncat.plist
893 The first parameter activates the tracing subsystem, the second
894 tells it to trace host and link utilization (without any
895 categorization) and the third creates a graph configuration file to
896 configure Triva when analysing the resulting trace file.
898 - MSG or SimDag-based simulator and categorized traces (you need to
899 declare categories and classify your tasks according to them)
901 .. code-block:: shell
903 --cfg=tracing:yes --cfg=tracing/categorized:yes --cfg=triva/categorized:cat.plist
905 The first parameter activates the tracing subsystem, the second
906 tells it to trace host and link categorized utilization and the
907 third creates a graph configuration file to configure Triva when
908 analysing the resulting trace file.
910 - SMPI simulator and traces for a space/time view:
912 .. code-block:: shell
916 The `-trace` parameter for the smpirun script runs the simulation
917 with ``--cfg=tracing:yes --cfg=tracing/smpi:yes``. Check the
918 smpirun's `-help` parameter for additional tracing options.
920 Sometimes you might want to put additional information on the trace to
921 correctly identify them later, or to provide data that can be used to
922 reproduce an experiment. You have two ways to do that:
924 - Add a string on top of the trace file as comment:
926 .. code-block:: shell
928 --cfg=tracing/comment:my_simulation_identifier
930 - Add the contents of a textual file on top of the trace file as comment:
932 .. code-block:: shell
934 --cfg=tracing/comment-file:my_file_with_additional_information.txt
936 Please, use these two parameters (for comments) to make reproducible
937 simulations. For additional details about this and all tracing
938 options, check See the :ref:`tracing_tracing_options`.
943 .. _cfg=msg/debug-multiple-use:
948 **Option** ``msg/debug-multiple-use`` **Default:** off
950 Sometimes your application may try to send a task that is still being
951 executed somewhere else, making it impossible to send this task. However,
952 for debugging purposes, one may want to know what the other host is/was
953 doing. This option shows a backtrace of the other process.
958 The SMPI interface provides several specific configuration items.
959 These are uneasy to see since the code is usually launched through the
960 ``smiprun`` script directly.
962 .. _cfg=smpi/host-speed:
963 .. _cfg=smpi/cpu-threshold:
964 .. _cfg=smpi/simulate-computation:
966 Automatic Benchmarking of SMPI Code
967 ...................................
969 In SMPI, the sequential code is automatically benchmarked, and these
970 computations are automatically reported to the simulator. That is to
971 say that if you have a large computation between a ``MPI_Recv()`` and
972 a ``MPI_Send()``, SMPI will automatically benchmark the duration of
973 this code, and create an execution task within the simulator to take
974 this into account. For that, the actual duration is measured on the
975 host machine and then scaled to the power of the corresponding
976 simulated machine. The variable ``smpi/host-speed`` allows to specify
977 the computational speed of the host machine (in flop/s) to use when
978 scaling the execution times. It defaults to 20000, but you really want
979 to update it to get accurate simulation results.
981 When the code is constituted of numerous consecutive MPI calls, the
982 previous mechanism feeds the simulation kernel with numerous tiny
983 computations. The ``smpi/cpu-threshold`` item becomes handy when this
984 impacts badly the simulation performance. It specifies a threshold (in
985 seconds) below which the execution chunks are not reported to the
986 simulation kernel (default value: 1e-6).
988 .. note:: The option ``smpi/cpu-threshold`` ignores any computation
989 time spent below this threshold. SMPI does not consider the
990 `amount` of these computations; there is no offset for this. Hence,
991 a value that is too small, may lead to unreliable simulation
994 In some cases, however, one may wish to disable simulation of
995 application computation. This is the case when SMPI is used not to
996 simulate an MPI applications, but instead an MPI code that performs
997 "live replay" of another MPI app (e.g., ScalaTrace's replay tool,
998 various on-line simulators that run an app at scale). In this case the
999 computation of the replay/simulation logic should not be simulated by
1000 SMPI. Instead, the replay tool or on-line simulator will issue
1001 "computation events", which correspond to the actual MPI simulation
1002 being replayed/simulated. At the moment, these computation events can
1003 be simulated using SMPI by calling internal smpi_execute*() functions.
1005 To disable the benchmarking/simulation of computation in the simulated
1006 application, the variable ``smpi/simulate-computation`` should be set
1007 to no. This option just ignores the timings in your simulation; it
1008 still executes the computations itself. If you want to stop SMPI from
1009 doing that, you should check the SMPI_SAMPLE macros, documented in
1010 Section :ref:`SMPI_adapting_speed`.
1012 +------------------------------------+-------------------------+-----------------------------+
1013 | Solution | Computations executed? | Computations simulated? |
1014 +====================================+=========================+=============================+
1015 | --cfg=smpi/simulate-computation:no | Yes | Never |
1016 +------------------------------------+-------------------------+-----------------------------+
1017 | --cfg=smpi/cpu-threshold:42 | Yes, in all cases | If it lasts over 42 seconds |
1018 +------------------------------------+-------------------------+-----------------------------+
1019 | SMPI_SAMPLE() macro | Only once per loop nest | Always |
1020 +------------------------------------+-------------------------+-----------------------------+
1022 .. _cfg=smpi/comp-adjustment-file:
1024 Slow-down or speed-up parts of your code
1025 ........................................
1027 **Option** ``smpi/comp-adjustment-file:`` **Default:** unset
1029 This option allows you to pass a file that contains two columns: The
1030 first column defines the section that will be subject to a speedup;
1031 the second column is the speedup. For instance:
1033 .. code-block:: shell
1035 "start:stop","ratio"
1036 "exchange_1.f:30:exchange_1.f:130",1.18244559422142
1038 The first line is the header - you must include it. The following
1039 line means that the code between two consecutive MPI calls on line 30
1040 in exchange_1.f and line 130 in exchange_1.f should receive a speedup
1041 of 1.18244559422142. The value for the second column is therefore a
1042 speedup, if it is larger than 1 and a slow-down if it is smaller
1043 than 1. Nothing will be changed if it is equal to 1.
1045 Of course, you can set any arbitrary filenames you want (so the start
1046 and end don't have to be in the same file), but be aware that this
1047 mechanism only supports `consecutive calls!`
1049 Please note that you must pass the ``-trace-call-location`` flag to
1050 smpicc or smpiff, respectively. This flag activates some internal
1051 macro definitions that help with obtaining the call location.
1053 .. _cfg=smpi/bw-factor:
1058 **Option** ``smpi/bw-factor``
1059 |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
1061 The possible throughput of network links is often dependent on the
1062 message sizes, as protocols may adapt to different message sizes. With
1063 this option, a series of message sizes and factors are given, helping
1064 the simulation to be more realistic. For instance, the current default
1065 value means that messages with size 65472 and more will get a total of
1066 MAX_BANDWIDTH*0.940694, messages of size 15424 to 65471 will get
1067 MAX_BANDWIDTH*0.697866 and so on (where MAX_BANDWIDTH denotes the
1068 bandwidth of the link).
1070 An experimental script to compute these factors is available online. See
1071 http://simgrid.gforge.inria.fr/contrib/smpi-calibration-doc.html
1072 http://simgrid.gforge.inria.fr/contrib/smpi-saturation-doc.html
1074 .. _cfg=smpi/display-timing:
1076 Reporting Simulation Time
1077 .........................
1079 **Option** ``smpi/display-timing`` **Default:** 0 (false)
1081 Most of the time, you run MPI code with SMPI to compute the time it
1082 would take to run it on a platform. But since the code is run through
1083 the ``smpirun`` script, you don't have any control on the launcher
1084 code, making it difficult to report the simulated time when the
1085 simulation ends. If you enable the ``smpi/display-timing`` item,
1086 ``smpirun`` will display this information when the simulation
1089 .. _cfg=smpi/keep-temps:
1091 Keeping temporary files after simulation
1092 ........................................
1094 **Option** ``smpi/keep-temps`` **default:** 0 (false)
1096 SMPI usually generates a lot of temporary files that are cleaned after
1097 use. This option request to preserve them, for example to debug or
1098 profile your code. Indeed, the binary files are removed very early
1099 under the dlopen privatization schema, which tend to fool the
1102 .. _cfg=smpi/lat-factor:
1107 **Option** ``smpi/lat-factor`` |br|
1108 **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
1110 The motivation and syntax for this option is identical to the motivation/syntax
1111 of :ref:`cfg=smpi/bw-factor`.
1113 There is an important difference, though: While smpi/bw-factor `reduces` the
1114 actual bandwidth (i.e., values between 0 and 1 are valid), latency factors
1115 increase the latency, i.e., values larger than or equal to 1 are valid here.
1117 .. _cfg=smpi/papi-events:
1119 Trace hardware counters with PAPI
1120 .................................
1122 **Option** ``smpi/papi-events`` **default:** unset
1124 When the PAPI support was compiled in SimGrid, this option takes the
1125 names of PAPI counters and adds their respective values to the trace
1126 files (See Section :ref:`tracing_tracing_options`).
1130 This feature currently requires superuser privileges, as registers
1131 are queried. Only use this feature with code you trust! Call
1132 smpirun for instance via ``smpirun -wrapper "sudo "
1133 <your-parameters>`` or run ``sudo sh -c "echo 0 >
1134 /proc/sys/kernel/perf_event_paranoid"`` In the later case, sudo
1135 will not be required.
1137 It is planned to make this feature available on a per-process (or per-thread?) basis.
1138 The first draft, however, just implements a "global" (i.e., for all processes) set
1139 of counters, the "default" set.
1141 .. code-block:: shell
1143 --cfg=smpi/papi-events:"default:PAPI_L3_LDM:PAPI_L2_LDM"
1145 .. _cfg=smpi/privatization:
1147 Automatic Privatization of Global Variables
1148 ...........................................
1150 **Option** ``smpi/privatization`` **default:** "dlopen" (when using smpirun)
1152 MPI executables are usually meant to be executed in separated
1153 processes, but SMPI is executed in only one process. Global variables
1154 from executables will be placed in the same memory zone and shared
1155 between processes, causing intricate bugs. Several options are
1156 possible to avoid this, as described in the main `SMPI publication
1157 <https://hal.inria.fr/hal-01415484>`_ and in the :ref:`SMPI
1158 documentation <SMPI_what_globals>`. SimGrid provides two ways of
1159 automatically privatizing the globals, and this option allows to
1160 choose between them.
1162 - **no** (default when not using smpirun): Do not automatically
1163 privatize variables. Pass ``-no-privatize`` to smpirun to disable
1165 - **dlopen** or **yes** (default when using smpirun): Link multiple
1166 times against the binary.
1167 - **mmap** (slower, but maybe somewhat more stable):
1168 Runtime automatic switching of the data segments.
1171 This configuration option cannot be set in your platform file. You can only
1172 pass it as an argument to smpirun.
1174 .. _cfg=smpi/privatize-libs:
1176 Automatic privatization of global variables inside external libraries
1177 .....................................................................
1179 **Option** ``smpi/privatize-libs`` **default:** unset
1181 **Linux/BSD only:** When using dlopen (default) privatization,
1182 privatize specific shared libraries with internal global variables, if
1183 they can't be linked statically. For example libgfortran is usually
1184 used for Fortran I/O and indexes in files can be mixed up.
1186 Multiple libraries can be given, semicolon separated.
1188 This configuration option can only use either full paths to libraries,
1189 or full names. Check with ldd the name of the library you want to
1192 .. code-block:: shell
1196 libgfortran.so.3 => /usr/lib/x86_64-linux-gnu/libgfortran.so.3 (0x00007fbb4d91b000)
1199 Then you can use ``--cfg=smpi/privatize-libs:libgfortran.so.3``
1200 or ``--cfg=smpi/privatize-libs:/usr/lib/x86_64-linux-gnu/libgfortran.so.3``,
1201 but not ``libgfortran`` nor ``libgfortran.so``.
1203 .. _cfg=smpi/send-is-detached-thresh:
1205 Simulating MPI detached send
1206 ............................
1208 **Option** ``smpi/send-is-detached-thresh`` **default:** 65536
1210 This threshold specifies the size in bytes under which the send will
1211 return immediately. This is different from the threshold detailed in
1212 :ref:`options_model_network_asyncsend` because the message is not
1213 effectively sent when the send is posted. SMPI still waits for the
1214 correspondant receive to be posted to perform the communication
1217 .. _cfg=smpi/coll-selector:
1219 Simulating MPI collective algorithms
1220 ....................................
1222 **Option** ``smpi/coll-selector`` **Possible values:** naive (default), ompi, mpich
1224 SMPI implements more than 100 different algorithms for MPI collective
1225 communication, to accurately simulate the behavior of most of the
1226 existing MPI libraries. The ``smpi/coll-selector`` item can be used to
1227 use the decision logic of either OpenMPI or MPICH libraries (by
1228 default SMPI uses naive version of collective operations).
1230 Each collective operation can be manually selected with a
1231 ``smpi/collective_name:algo_name``. Available algorithms are listed in
1232 :ref:`SMPI_use_colls`.
1234 .. TODO:: All available collective algorithms will be made available
1235 via the ``smpirun --help-coll`` command.
1237 .. _cfg=smpi/iprobe:
1239 Inject constant times for MPI_Iprobe
1240 ....................................
1242 **Option** ``smpi/iprobe`` **default:** 0.0001
1244 The behavior and motivation for this configuration option is identical
1245 with :ref:`smpi/test <cfg=smpi/test>`, but for the function
1248 .. _cfg=smpi/iprobe-cpu-usage:
1250 Reduce speed for iprobe calls
1251 .............................
1253 **Option** ``smpi/iprobe-cpu-usage`` **default:** 1 (no change)
1255 MPI_Iprobe calls can be heavily used in applications. To account
1256 correctly for the energy cores spend probing, it is necessary to
1257 reduce the load that these calls cause inside SimGrid.
1259 For instance, we measured a max power consumption of 220 W for a
1260 particular application but only 180 W while this application was
1261 probing. Hence, the correct factor that should be passed to this
1262 option would be 180/220 = 0.81.
1266 Inject constant times for MPI_Init
1267 ..................................
1269 **Option** ``smpi/init`` **default:** 0
1271 The behavior and motivation for this configuration option is identical
1272 with :ref:`smpi/test <cfg=smpi/test>`, but for the function ``MPI_Init()``.
1276 Inject constant times for MPI_Isend()
1277 .....................................
1279 **Option** ``smpi/ois``
1281 The behavior and motivation for this configuration option is identical
1282 with :ref:`smpi/os <cfg=smpi/os>`, but for the function ``MPI_Isend()``.
1286 Inject constant times for MPI_send()
1287 ....................................
1289 **Option** ``smpi/os``
1291 In several network models such as LogP, send (MPI_Send, MPI_Isend) and
1292 receive (MPI_Recv) operations incur costs (i.e., they consume CPU
1293 time). SMPI can factor these costs in as well, but the user has to
1294 configure SMPI accordingly as these values may vary by machine. This
1295 can be done by using ``smpi/os`` for MPI_Send operations; for MPI_Isend
1296 and MPI_Recv, use ``smpi/ois`` and ``smpi/or``, respectively. These work
1297 exactly as ``smpi/ois``.
1299 This item can consist of multiple sections; each section takes three
1300 values, for example ``1:3:2;10:5:1``. The sections are divided by ";"
1301 so this example contains two sections. Furthermore, each section
1302 consists of three values.
1304 1. The first value denotes the minimum size for this section to take effect;
1305 read it as "if message size is greater than this value (and other section has a larger
1306 first value that is also smaller than the message size), use this".
1307 In the first section above, this value is "1".
1309 2. The second value is the startup time; this is a constant value that will always
1310 be charged, no matter what the size of the message. In the first section above,
1313 3. The third value is the `per-byte` cost. That is, it is charged for every
1314 byte of the message (incurring cost messageSize*cost_per_byte)
1315 and hence accounts also for larger messages. In the first
1316 section of the example above, this value is "2".
1318 Now, SMPI always checks which section it should take for a given
1319 message; that is, if a message of size 11 is sent with the
1320 configuration of the example above, only the second section will be
1321 used, not the first, as the first value of the second section is
1322 closer to the message size. Hence, when ``smpi/os=1:3:2;10:5:1``, a
1323 message of size 11 incurs the following cost inside MPI_Send:
1324 ``5+11*1`` because 5 is the startup cost and 1 is the cost per byte.
1326 Note that the order of sections can be arbitrary; they will be ordered internally.
1330 Inject constant times for MPI_Recv()
1331 ....................................
1333 **Option** ``smpi/or``
1335 The behavior and motivation for this configuration option is identical
1336 with :ref:`smpi/os <cfg=smpi/os>`, but for the function ``MPI_Recv()``.
1339 .. _cfg=smpi/grow-injected-times:
1341 Inject constant times for MPI_Test
1342 ..................................
1344 **Option** ``smpi/test`` **default:** 0.0001
1346 By setting this option, you can control the amount of time a process
1347 sleeps when MPI_Test() is called; this is important, because SimGrid
1348 normally only advances the time while communication is happening and
1349 thus, MPI_Test will not add to the time, resulting in a deadlock if
1350 used as a break-condition as in the following example:
1355 MPI_Test(request, flag, status);
1359 To speed up execution, we use a counter to keep track on how often we
1360 already checked if the handle is now valid or not. Hence, we actually
1361 use counter*SLEEP_TIME, that is, the time MPI_Test() causes the
1362 process to sleep increases linearly with the number of previously
1363 failed tests. This behavior can be disabled by setting
1364 ``smpi/grow-injected-times`` to **no**. This will also disable this
1365 behavior for MPI_Iprobe.
1367 .. _cfg=smpi/shared-malloc:
1368 .. _cfg=smpi/shared-malloc-hugepage:
1373 **Option** ``smpi/shared-malloc`` **Possible values:** global (default), local
1375 If your simulation consumes too much memory, you may want to modify
1376 your code so that the working areas are shared by all MPI ranks. For
1377 example, in a bloc-cyclic matrix multiplication, you will only
1378 allocate one set of blocs, and every processes will share them.
1379 Naturally, this will lead to very wrong results, but this will save a
1380 lot of memory so this is still desirable for some studies. For more on
1381 the motivation for that feature, please refer to the `relevant section
1382 <https://simgrid.github.io/SMPI_CourseWare/topic_understanding_performance/matrixmultiplication>`_
1383 of the SMPI CourseWare (see Activity #2.2 of the pointed
1384 assignment). In practice, change the call to malloc() and free() into
1385 SMPI_SHARED_MALLOC() and SMPI_SHARED_FREE().
1387 SMPI provides two algorithms for this feature. The first one, called
1388 ``local``, allocates one bloc per call to SMPI_SHARED_MALLOC() in your
1389 code (each call location gets its own bloc) and this bloc is shared
1390 amongst all MPI ranks. This is implemented with the shm_* functions
1391 to create a new POSIX shared memory object (kept in RAM, in /dev/shm)
1392 for each shared bloc.
1394 With the ``global`` algorithm, each call to SMPI_SHARED_MALLOC()
1395 returns a new adress, but it only points to a shadow bloc: its memory
1396 area is mapped on a 1MiB file on disk. If the returned bloc is of size
1397 N MiB, then the same file is mapped N times to cover the whole bloc.
1398 At the end, no matter how many SMPI_SHARED_MALLOC you do, this will
1399 only consume 1 MiB in memory.
1401 You can disable this behavior and come back to regular mallocs (for
1402 example for debugging purposes) using @c "no" as a value.
1404 If you want to keep private some parts of the buffer, for instance if these
1405 parts are used by the application logic and should not be corrupted, you
1406 can use SMPI_PARTIAL_SHARED_MALLOC(size, offsets, offsets_count). Example:
1410 mem = SMPI_PARTIAL_SHARED_MALLOC(500, {27,42 , 100,200}, 2);
1412 This will allocate 500 bytes to mem, such that mem[27..41] and
1413 mem[100..199] are shared while other area remain private.
1415 Then, it can be deallocated by calling SMPI_SHARED_FREE(mem).
1417 When smpi/shared-malloc:global is used, the memory consumption problem
1418 is solved, but it may induce too much load on the kernel's pages table.
1419 In this case, you should use huge pages so that we create only one
1420 entry per Mb of malloced data instead of one entry per 4k.
1421 To activate this, you must mount a hugetlbfs on your system and allocate
1422 at least one huge page:
1424 .. code-block:: shell
1427 sudo mount none /home/huge -t hugetlbfs -o rw,mode=0777
1428 sudo sh -c 'echo 1 > /proc/sys/vm/nr_hugepages' # echo more if you need more
1430 Then, you can pass the option
1431 ``--cfg=smpi/shared-malloc-hugepage:/home/huge`` to smpirun to
1432 actually activate the huge page support in shared mallocs.
1436 Inject constant times for MPI_Wtime, gettimeofday and clock_gettime
1437 ...................................................................
1439 **Option** ``smpi/wtime`` **default:** 10 ns
1441 This option controls the amount of (simulated) time spent in calls to
1442 MPI_Wtime(), gettimeofday() and clock_gettime(). If you set this value
1443 to 0, the simulated clock is not advanced in these calls, which leads
1444 to issue if your application contains such a loop:
1448 while(MPI_Wtime() < some_time_bound) {
1449 /* some tests, with no communication nor computation */
1452 When the option smpi/wtime is set to 0, the time advances only on
1453 communications and computations, so the previous code results in an
1454 infinite loop: the current [simulated] time will never reach
1455 ``some_time_bound``. This infinite loop is avoided when that option
1456 is set to a small amount, as it is by default since SimGrid v3.21.
1458 Note that if your application does not contain any loop depending on
1459 the current time only, then setting this option to a non-zero value
1460 will slow down your simulations by a tiny bit: the simulation loop has
1461 to be broken and reset each time your code ask for the current time.
1462 If the simulation speed really matters to you, you can avoid this
1463 extra delay by setting smpi/wtime to 0.
1465 Other Configurations
1466 --------------------
1468 .. _cfg=clean-atexit:
1470 Cleanup at Termination
1471 ......................
1473 **Option** ``clean-atexit`` **default:** on
1475 If your code is segfaulting during its finalization, it may help to
1476 disable this option to request SimGrid to not attempt any cleanups at
1477 the end of the simulation. Since the Unix process is ending anyway,
1478 the operating system will wipe it all.
1485 **Option** ``path`` **default:** . (current dir)
1487 It is possible to specify a list of directories to search into for the
1488 trace files (see :ref:`pf_trace`) by using this configuration
1489 item. To add several directory to the path, set the configuration
1490 item several times, as in ``--cfg=path:toto --cfg=path:tutu``
1492 .. _cfg=simix/breakpoint:
1497 **Option** ``simix/breakpoint`` **default:** unset
1499 This configuration option sets a breakpoint: when the simulated clock
1500 reaches the given time, a SIGTRAP is raised. This can be used to stop
1501 the execution and get a backtrace with a debugger.
1503 It is also possible to set the breakpoint from inside the debugger, by
1504 writing in global variable simgrid::simix::breakpoint. For example,
1507 .. code-block:: shell
1509 set variable simgrid::simix::breakpoint = 3.1416
1511 .. _cfg=verbose-exit:
1516 **Option** ``verbose-exit`` **default:** on
1518 By default, when Ctrl-C is pressed, the status of all existing actors
1519 is displayed before exiting the simulation. This is very useful to
1520 debug your code, but it can reveal troublesome if you have many
1521 actors. Set this configuration item to **off** to disable this
1524 .. _cfg=exception/cutpath:
1526 Truncate local path from exception backtrace
1527 ............................................
1529 **Option** ``exception/cutpath`` **default:** off
1531 This configuration option is used to remove the path from the
1532 backtrace shown when an exception is thrown. This is mainly useful for
1533 the tests: the full file path makes the tests not reproducible because
1534 the path of source files depend of the build settings. That would
1535 break most of our tests as we keep comparing output.
1537 Logging Configuration
1538 ---------------------
1540 It can be done by using XBT. Go to :ref:`XBT_log` for more details.