1 /*! \page options Configure SimGrid
3 A number of options can be given at runtime to change the default
4 SimGrid behavior. For a complete list of all configuration options
5 accepted by the SimGrid version used in your simulator, simply pass
6 the --help configuration flag to your program. If some of the options
7 are not documented on this page, this is a bug that you should please
8 report so that we can fix it. Note that some of the options presented
9 here may not be available in your simulators, depending on the
10 @ref install_src_config "compile-time options" that you used.
14 \section options_using Passing configuration options to the simulators
16 There is several way to pass configuration options to the simulators.
17 The most common way is to use the \c --cfg command line argument. For
18 example, to set the item \c Item to the value \c Value, simply
19 type the following: \verbatim
20 my_simulator --cfg=Item:Value (other arguments)
23 Several \c `--cfg` command line arguments can naturally be used. If you
24 need to include spaces in the argument, don't forget to quote the
25 argument. You can even escape the included quotes (write \' for ' if
26 you have your argument between ').
28 Another solution is to use the \c \<config\> tag in the platform file. The
29 only restriction is that this tag must occure before the first
30 platform element (be it \c \<AS\>, \c \<cluster\>, \c \<peer\> or whatever).
31 The \c \<config\> tag takes an \c id attribute, but it is currently
32 ignored so you don't really need to pass it. The important par is that
33 within that tag, you can pass one or several \c \<prop\> tags to specify
34 the configuration to use. For example, setting \c Item to \c Value
35 can be done by adding the following to the beginning of your platform
39 <prop id="Item" value="Value"/>
43 A last solution is to pass your configuration directly using the C
44 interface. If you happen to use the MSG interface, this is very easy
45 with the MSG_config() function. If you do not use MSG, that's a bit
46 more complex, as you have to mess with the internal configuration set
47 directly as follows. Check the \ref XBT_config "relevant page" for
48 details on all the functions you can use in this context, \c
49 _sg_cfg_set being the only configuration set currently used in
53 #include <xbt/config.h>
55 extern xbt_cfg_t _sg_cfg_set;
57 int main(int argc, char *argv[]) {
60 /* Prefer MSG_config() if you use MSG!! */
61 xbt_cfg_set_parse(_sg_cfg_set,"Item:Value");
67 \section options_model Configuring the platform models
69 \anchor options_storage_model
70 \anchor options_vm_model
71 \subsection options_model_select Selecting the platform models
73 SimGrid comes with several network, CPU and storage models built in, and you
74 can change the used model at runtime by changing the passed
75 configuration. The three main configuration items are given below.
76 For each of these items, passing the special \c help value gives
77 you a short description of all possible values. Also, \c --help-models
78 should provide information about all models for all existing resources.
79 - \b network/model: specify the used network model
80 - \b cpu/model: specify the used CPU model
81 - \b host/model: specify the used host model
82 - \b storage/model: specify the used storage model (there is currently only one such model - this option is hence only useful for future releases)
83 - \b vm/model: specify the model for virtual machines (there is currently only one such model - this option is hence only useful for future releases)
85 As of writing, the following network models are accepted. Over
86 the time new models can be added, and some experimental models can be
87 removed; check the values on your simulators for an uptodate
88 information. Note that the CM02 model is described in the research report
89 <a href="ftp://ftp.ens-lyon.fr/pub/LIP/Rapports/RR/RR2002/RR2002-40.ps.gz">A
90 Network Model for Simulation of Grid Application</a> while LV08 is
92 <a href="http://mescal.imag.fr/membres/arnaud.legrand/articles/simutools09.pdf">Accuracy Study and Improvement of Network Simulation in the SimGrid Framework</a>.
94 - \b LV08 (default one): Realistic network analytic model
95 (slow-start modeled by multiplying latency by 10.4, bandwidth by
96 .92; bottleneck sharing uses a payload of S=8775 for evaluating RTT)
97 - \anchor options_model_select_network_constant \b Constant: Simplistic network model where all communication
98 take a constant time (one second). This model provides the lowest
99 realism, but is (marginally) faster.
100 - \b SMPI: Realistic network model specifically tailored for HPC
101 settings (accurate modeling of slow start with correction factors on
102 three intervals: < 1KiB, < 64 KiB, >= 64 KiB). See also \ref
103 options_model_network_coefs "this section" for more info.
104 - \b IB: Realistic network model specifically tailored for HPC
105 settings with InfiniBand networks (accurate modeling contention
106 behavior, based on the model explained in
107 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf).
108 See also \ref options_model_network_coefs "this section" for more info.
109 - \b CM02: Legacy network analytic model (Very similar to LV08, but
110 without corrective factors. The timings of small messages are thus
112 - \b Reno: Model from Steven H. Low using lagrange_solve instead of
113 lmm_solve (experts only; check the code for more info).
114 - \b Reno2: Model from Steven H. Low using lagrange_solve instead of
115 lmm_solve (experts only; check the code for more info).
116 - \b Vegas: Model from Steven H. Low using lagrange_solve instead of
117 lmm_solve (experts only; check the code for more info).
119 If you compiled SimGrid accordingly, you can use packet-level network
120 simulators as network models (see \ref pls_ns3). In that case, you have
121 two extra models, described below, and some \ref options_pls "specific
122 additional configuration flags".
123 - \b NS3: Network pseudo-model using the NS3 tcp model
125 Concerning the CPU, we have only one model for now:
126 - \b Cas01: Simplistic CPU model (time=size/power)
128 The host concept is the aggregation of a CPU with a network
129 card. Three models exists, but actually, only 2 of them are
130 interesting. The "compound" one is simply due to the way our internal
131 code is organized, and can easily be ignored. So at the end, you have
132 two host models: The default one allows to aggregate an
133 existing CPU model with an existing network model, but does not allow
134 parallel tasks because these beasts need some collaboration between
135 the network and CPU model. That is why, ptask_07 is used by default
137 - \b default: Default host model. Currently, CPU:Cas01 and
138 network:LV08 (with cross traffic enabled)
139 - \b compound: Host model that is automatically chosen if
140 you change the network and CPU models
141 - \b ptask_L07: Host model somehow similar to Cas01+CM02 but
142 allowing "parallel tasks", that are intended to model the moldable
143 tasks of the grid scheduling literature.
145 \subsection options_generic_plugin Plugins
147 SimGrid supports the use of plugins; currently, no known plugins
148 can be activated but there are use-cases where you may want to write
149 your own plugin (for instance, for logging).
151 Plugins can for instance define own classes that inherit from
152 existing classes (for instance, a class "CpuEnergy" inherits from
153 "Cpu" to assess energy consumption).
155 The plugin connects to the code by registering callbacks using
156 ``signal.connect(callback)`` (see file ``src/surf/plugins/energy.cpp`` for
164 This option is case-sensitive: Energy and energy are not the same!
166 \subsection options_model_optim Optimization level of the platform models
168 The network and CPU models that are based on lmm_solve (that
169 is, all our analytical models) accept specific optimization
171 - items \b network/optim and \b cpu/optim (both default to 'Lazy'):
172 - \b Lazy: Lazy action management (partial invalidation in lmm +
173 heap in action remaining).
174 - \b TI: Trace integration. Highly optimized mode when using
175 availability traces (only available for the Cas01 CPU model for
177 - \b Full: Full update of remaining and variables. Slow but may be
178 useful when debugging.
179 - items \b network/maxmin-selective-update and
180 \b cpu/maxmin-selective-update: configure whether the underlying
181 should be lazily updated or not. It should have no impact on the
182 computed timings, but should speed up the computation.
184 It is still possible to disable the \c maxmin-selective-update feature
185 because it can reveal counter-productive in very specific scenarios
186 where the interaction level is high. In particular, if all your
187 communication share a given backbone link, you should disable it:
188 without \c maxmin-selective-update, every communications are updated
189 at each step through a simple loop over them. With that feature
190 enabled, every communications will still get updated in this case
191 (because of the dependency induced by the backbone), but through a
192 complicated pattern aiming at following the actual dependencies.
194 \subsection options_model_precision Numerical precision of the platform models
196 The analytical models handle a lot of floating point values. It is
197 possible to change the epsilon used to update and compare them through
198 the \b maxmin/precision item (default value: 0.00001). Changing it
199 may speedup the simulation by discarding very small actions, at the
200 price of a reduced numerical precision.
202 \subsection options_concurrency_limit Concurrency limit
204 The maximum number of variables per resource can be tuned through
205 the \b maxmin/concurrency-limit item. The default value is -1, meaning that
206 there is no such limitation. You can have as many simultaneous actions per
207 resources as you want. If your simulation presents a very high level of
208 concurrency, it may help to use e.g. 100 as a value here. It means that at
209 most 100 actions can consume a resource at a given time. The extraneous actions
210 are queued and wait until the amount of concurrency of the considered resource
211 lowers under the given boundary.
213 Such limitations help both to the simulation speed and simulation accuracy
214 on highly constrained scenarios, but the simulation speed suffers of this
215 setting on regular (less constrained) scenarios so it is off by default.
217 \subsection options_model_network Configuring the Network model
219 \subsubsection options_model_network_gamma Maximal TCP window size
221 The analytical models need to know the maximal TCP window size to take
222 the TCP congestion mechanism into account. This is set to 20000 by
223 default, but can be changed using the \b network/TCP-gamma item.
225 On linux, this value can be retrieved using the following
226 commands. Both give a set of values, and you should use the last one,
227 which is the maximal size.\verbatim
228 cat /proc/sys/net/ipv4/tcp_rmem # gives the sender window
229 cat /proc/sys/net/ipv4/tcp_wmem # gives the receiver window
232 \subsubsection options_model_network_coefs Correcting important network parameters
234 SimGrid can take network irregularities such as a slow startup or
235 changing behavior depending on the message size into account.
236 You should not change these values unless you really know what you're doing.
238 The corresponding values were computed through data fitting one the
239 timings of packet-level simulators.
242 <a href="http://mescal.imag.fr/membres/arnaud.legrand/articles/simutools09.pdf">Accuracy Study and Improvement of Network Simulation in the SimGrid Framework</a>
243 for more information about these parameters.
245 If you are using the SMPI model, these correction coefficients are
246 themselves corrected by constant values depending on the size of the
247 exchange. Again, only hardcore experts should bother about this fact.
249 InfiniBand network behavior can be modeled through 3 parameters, as explained in
250 <a href="http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf">this PhD thesis</a>.
251 These factors can be changed through the following option:
254 smpi/IB-penalty-factors:"βe;βs;γs"
257 By default SMPI uses factors computed on the Stampede Supercomputer at TACC, with optimal
258 deployment of processes on nodes.
260 \subsubsection options_model_network_crosstraffic Simulating cross-traffic
262 As of SimGrid v3.7, cross-traffic effects can be taken into account in
263 analytical simulations. It means that ongoing and incoming
264 communication flows are treated independently. In addition, the LV08
265 model adds 0.05 of usage on the opposite direction for each new
266 created flow. This can be useful to simulate some important TCP
267 phenomena such as ack compression.
269 For that to work, your platform must have two links for each
270 pair of interconnected hosts. An example of usable platform is
271 available in <tt>examples/platforms/crosstraffic.xml</tt>.
273 This is activated through the \b network/crosstraffic item, that
274 can be set to 0 (disable this feature) or 1 (enable it).
276 Note that with the default host model this option is activated by default.
278 \subsubsection options_model_network_sendergap Simulating sender gap
280 (this configuration item is experimental and may change or disapear)
282 It is possible to specify a timing gap between consecutive emission on
283 the same network card through the \b network/sender-gap item. This
284 is still under investigation as of writting, and the default value is
285 to wait 10 microseconds (1e-5 seconds) between emissions.
287 \subsubsection options_model_network_asyncsend Simulating asyncronous send
289 (this configuration item is experimental and may change or disapear)
291 It is possible to specify that messages below a certain size will be sent
292 as soon as the call to MPI_Send is issued, without waiting for the
293 correspondant receive. This threshold can be configured through the
294 \b smpi/async-small-thresh item. The default value is 0. This behavior can also be
295 manually set for MSG mailboxes, by setting the receiving mode of the mailbox
296 with a call to \ref MSG_mailbox_set_async . For MSG, all messages sent to this
297 mailbox will have this behavior, so consider using two mailboxes if needed.
299 This value needs to be smaller than or equals to the threshold set at
300 \ref options_model_smpi_detached , because asynchronous messages are
301 meant to be detached as well.
303 \subsubsection options_pls Configuring packet-level pseudo-models
305 When using the packet-level pseudo-models, several specific
306 configuration flags are provided to configure the associated tools.
307 There is by far not enough such SimGrid flags to cover every aspects
308 of the associated tools, since we only added the items that we
309 needed ourselves. Feel free to request more items (or even better:
310 provide patches adding more items).
312 When using NS3, the only existing item is \b ns3/TcpModel,
313 corresponding to the ns3::TcpL4Protocol::SocketType configuration item
314 in NS3. The only valid values (enforced on the SimGrid side) are
315 'NewReno' or 'Reno' or 'Tahoe'.
317 \subsection options_model_storage Configuring the Storage model
319 \subsubsection option_model_storage_maxfd Maximum amount of file descriptors per host
321 Each host maintains a fixed-size array of its file descriptors. You
322 can change its size (1024 by default) through the \b
323 storage/max_file_descriptors item to either enlarge it if your
324 application requires it or to reduce it to save memory space.
326 \section options_modelchecking Configuring the Model-Checking
328 To enable the SimGrid model-checking support the program should
329 be executed using the simgrid-mc wrapper:
331 simgrid-mc ./my_program
334 Safety properties are expressed as assertions using the function
336 void MC_assert(int prop);
339 \subsection options_modelchecking_liveness Specifying a liveness property
341 If you want to specify liveness properties (beware, that's
342 experimental), you have to pass them on the command line, specifying
343 the name of the file containing the property, as formatted by the
347 --cfg=model-check/property:<filename>
350 \subsection options_modelchecking_steps Going for stateful verification
352 By default, the system is backtracked to its initial state to explore
353 another path instead of backtracking to the exact step before the fork
354 that we want to explore (this is called stateless verification). This
355 is done this way because saving intermediate states can rapidly
356 exhaust the available memory. If you want, you can change the value of
357 the <tt>model-check/checkpoint</tt> variable. For example, the
358 following configuration will ask to take a checkpoint every step.
359 Beware, this will certainly explode your memory. Larger values are
360 probably better, make sure to experiment a bit to find the right
361 setting for your specific system.
364 --cfg=model-check/checkpoint:1
367 \subsection options_modelchecking_reduction Specifying the kind of reduction
369 The main issue when using the model-checking is the state space
370 explosion. To counter that problem, several exploration reduction
371 techniques can be used. There is unfortunately no silver bullet here,
372 and the most efficient reduction techniques cannot be applied to any
373 properties. In particular, the DPOR method cannot be applied on
374 liveness properties since it may break some cycles in the exploration
375 that are important to the property validity.
378 --cfg=model-check/reduction:<technique>
381 For now, this configuration variable can take 2 values:
382 * none: Do not apply any kind of reduction (mandatory for now for
384 * dpor: Apply Dynamic Partial Ordering Reduction. Only valid if you
385 verify local safety properties (default value for safety checks).
387 \subsection options_modelchecking_visited model-check/visited, Cycle detection
389 In order to detect cycles, the model-checker needs to check if a new explored
390 state is in fact the same state than a previous one. In order to do this,
391 the model-checker can take a snapshot of each visited state: this snapshot is
392 then used to compare it with subsequent states in the exploration graph.
394 The \b model-check/visited is the maximum number of states which are stored in
395 memory. If the maximum number of snapshotted state is reached some states will
396 be removed from the memory and some cycles might be missed.
398 By default, no state is snapshotted and cycles cannot be detected.
400 \subsection options_modelchecking_termination model-check/termination, Non termination detection
402 The \b model-check/termination configuration item can be used to report if a
403 non-termination execution path has been found. This is a path with a cycle
404 which means that the program might never terminate.
406 This only works in safety mode.
408 This options is disabled by default.
410 \subsection options_modelchecking_dot_output model-check/dot-output, Dot output
412 If set, the \b model-check/dot-output configuration item is the name of a file
413 in which to write a dot file of the path leading the found property (safety or
414 liveness violation) as well as the cycle for liveness properties. This dot file
415 can then fed to the graphviz dot tool to generate an corresponding graphical
418 \subsection options_modelchecking_max_depth model-check/max_depth, Depth limit
420 The \b model-checker/max-depth can set the maximum depth of the exploration
421 graph of the model-checker. If this limit is reached, a logging message is
422 sent and the results might not be exact.
424 By default, there is not depth limit.
426 \subsection options_modelchecking_timeout Handling of timeout
428 By default, the model-checker does not handle timeout conditions: the `wait`
429 operations never time out. With the \b model-check/timeout configuration item
430 set to \b yes, the model-checker will explore timeouts of `wait` operations.
432 \subsection options_modelchecking_comm_determinism Communication determinism
434 The \b model-check/communications-determinism and
435 \b model-check/send-determinism items can be used to select the communication
436 determinism mode of the model-checker which checks determinism properties of
437 the communications of an application.
439 \subsection options_modelchecking_sparse_checkpoint Per page checkpoints
441 When the model-checker is configured to take a snapshot of each explored state
442 (with the \b model-checker/visited item), the memory consumption can rapidly
443 reach GiB ou Tib of memory. However, for many workloads, the memory does not
444 change much between different snapshots and taking a complete copy of each
445 snapshot is a waste of memory.
447 The \b model-check/sparse-checkpoint option item can be set to \b yes in order
448 to avoid making a complete copy of each snapshot: instead, each snapshot will be
449 decomposed in blocks which will be stored separately.
450 If multiple snapshots share the same block (or if the same block
451 is used in the same snapshot), the same copy of the block will be shared leading
452 to a reduction of the memory footprint.
454 For many applications, this option considerably reduces the memory consumption.
455 In somes cases, the model-checker might be slightly slower because of the time
456 taken to manage the metadata about the blocks. In other cases however, this
457 snapshotting strategy will be much faster by reducing the cache consumption.
458 When the memory consumption is important, by avoiding to hit the swap or
459 reducing the swap usage, this option might be much faster than the basic
460 snapshotting strategy.
462 This option is currently disabled by default.
464 \subsection options_mc_perf Performance considerations for the model checker
466 The size of the stacks can have a huge impact on the memory
467 consumption when using model-checking. By default, each snapshot will
468 save a copy of the whole stacks and not only of the part which is
469 really meaningful: you should expect the contribution of the memory
470 consumption of the snapshots to be \f$ \mbox{number of processes}
471 \times \mbox{stack size} \times \mbox{number of states} \f$.
473 The \b model-check/sparse-checkpoint can be used to reduce the memory
474 consumption by trying to share memory between the different snapshots.
476 When compiled against the model checker, the stacks are not
477 protected with guards: if the stack size is too small for your
478 application, the stack will silently overflow on other parts of the
479 memory (see \ref options_virt_guard_size).
481 \subsection options_modelchecking_hash Hashing of the state (experimental)
483 Usually most of the time of the model-checker is spent comparing states. This
484 process is complicated and consumes a lot of bandwidth and cache.
485 In order to speedup the state comparison, the experimental \b model-checker/hash
486 configuration item enables the computation of a hash summarizing as much
487 information of the state as possible into a single value. This hash can be used
488 to avoid most of the comparisons: the costly comparison is then only used when
489 the hashes are identical.
491 Currently most of the state is not included in the hash because the
492 implementation was found to be buggy and this options is not as useful as
493 it could be. For this reason, it is currently disabled by default.
495 \subsection options_modelchecking_recordreplay Record/replay (experimental)
497 As the model-checker keeps jumping at different places in the execution graph,
498 it is difficult to understand what happens when trying to debug an application
499 under the model-checker. Event the output of the program is difficult to
500 interpret. Moreover, the model-checker does not behave nicely with advanced
501 debugging tools such as valgrind. For those reason, to identify a trajectory
502 in the execution graph with the model-checker and replay this trajcetory and
503 without the model-checker black-magic but with more standard tools
504 (such as a debugger, valgrind, etc.). For this reason, Simgrid implements an
505 experimental record/replay functionnality in order to record a trajectory with
506 the model-checker and replay it without the model-checker.
508 When the model-checker finds an interesting path in the application execution
509 graph (where a safety or liveness property is violated), it can generate an
510 identifier for this path. In order to enable this behavious the
511 \b model-check/record must be set to \b yes. By default, this behaviour is not
514 This is an example of output:
517 [ 0.000000] (0:@) Check a safety property
518 [ 0.000000] (0:@) **************************
519 [ 0.000000] (0:@) *** PROPERTY NOT VALID ***
520 [ 0.000000] (0:@) **************************
521 [ 0.000000] (0:@) Counter-example execution trace:
522 [ 0.000000] (0:@) Path = 1/3;1/4
523 [ 0.000000] (0:@) [(1)Tremblay (app)] MC_RANDOM(3)
524 [ 0.000000] (0:@) [(1)Tremblay (app)] MC_RANDOM(4)
525 [ 0.000000] (0:@) Expanded states = 27
526 [ 0.000000] (0:@) Visited states = 68
527 [ 0.000000] (0:@) Executed transitions = 46
530 This path can then be replayed outside of the model-checker (and even in
531 non-MC build of simgrid) by setting the \b model-check/replay item to the given
532 path. The other options should be the same (but the model-checker should
535 The format and meaning of the path may change between different releases so
536 the same release of Simgrid should be used for the record phase and the replay
539 \section options_virt Configuring the User Process Virtualization
541 \subsection options_virt_factory Selecting the virtualization factory
543 In SimGrid, the user code is virtualized in a specific mechanism
544 that allows the simulation kernel to control its execution: when a user
545 process requires a blocking action (such as sending a message), it is
546 interrupted, and only gets released when the simulated clock reaches
547 the point where the blocking operation is done.
549 In SimGrid, the containers in which user processes are virtualized are
550 called contexts. Several context factory are provided, and you can
551 select the one you want to use with the \b contexts/factory
552 configuration item. Some of the following may not exist on your
553 machine because of portability issues. In any case, the default one
554 should be the most effcient one (please report bugs if the
555 auto-detection fails for you). They are approximately sorted here from
556 the slowest to the most effient:
558 - \b thread: very slow factory using full featured threads (either
559 pthreads or windows native threads). They are slow but very
560 standard. Some debuggers or profilers only work with this factory.
561 - \b java: Java applications are virtualized onto java threads (that
562 are regular pthreads registered to the JVM)
563 - \b ucontext: fast factory using System V contexts (Linux and FreeBSD only)
564 - \b boost: This uses the [context implementation](http://www.boost.org/doc/libs/1_59_0/libs/context/doc/html/index.html)
565 of the boost library for a performance that is comparable to our
566 raw implementation.\nInstall the relevant library (e.g. with the
567 libboost-contexts-dev package on Debian/Ubuntu) and recompile
568 SimGrid. Note that our implementation is not compatible with recent
569 implementations of the library, and it will be hard to fix this since
570 the library's author decided to hide an API that we were using.
571 - \b raw: amazingly fast factory using a context switching mechanism
572 of our own, directly implemented in assembly (only available for x86
573 and amd64 platforms for now) and without any unneeded system call.
575 The main reason to change this setting is when the debugging tools get
576 fooled by the optimized context factories. Threads are the most
577 debugging-friendly contextes, as they allow to set breakpoints
578 anywhere with gdb and visualize backtraces for all processes, in order
579 to debug concurrency issues. Valgrind is also more comfortable with
580 threads, but it should be usable with all factories (but the callgrind
581 tool that really don't like raw and ucontext factories).
583 \subsection options_virt_stacksize Adapting the used stack size
585 Each virtualized used process is executed using a specific system
586 stack. The size of this stack has a huge impact on the simulation
587 scalability, but its default value is rather large. This is because
588 the error messages that you get when the stack size is too small are
589 rather disturbing: this leads to stack overflow (overwriting other
590 stacks), leading to segfaults with corrupted stack traces.
592 If you want to push the scalability limits of your code, you might
593 want to reduce the \b contexts/stack-size item. Its default value
594 is 8192 (in KiB), while our Chord simulation works with stacks as small
595 as 16 KiB, for example. For the thread factory, the default value
596 is the one of the system but you can still change it with this parameter.
598 The operating system should only allocate memory for the pages of the
599 stack which are actually used and you might not need to use this in
600 most cases. However, this setting is very important when using the
601 model checker (see \ref options_mc_perf).
603 \subsection options_virt_guard_size Disabling stack guard pages
605 A stack guard page is usually used which prevents the stack of a given
606 actor from overflowing on another stack. But the performance impact
607 may become prohibitive when the amount of actors increases. The
608 option \b contexts:guard-size is the number of stack guard pages used.
609 By setting it to 0, no guard pages will be used: in this case, you
610 should avoid using small stacks (\b stack-size) as the stack will
611 silently overflow on other parts of the memory.
613 When no stack guard page is created, stacks may then silently overflow
614 on other parts of the memory if their size is too small for the
615 application. This happens:
617 - on Windows systems;
618 - when the model checker is enabled;
619 - and of course when guard pages are explicitely disabled (with \b contexts:guard-size=0).
621 \subsection options_virt_parallel Running user code in parallel
623 Parallel execution of the user code is only considered stable in
624 SimGrid v3.7 and higher, and mostly for MSG simulations. SMPI
625 simulations may well fail in parallel mode. It is described in
626 <a href="http://hal.inria.fr/inria-00602216/">INRIA RR-7653</a>.
628 If you are using the \c ucontext or \c raw context factories, you can
629 request to execute the user code in parallel. Several threads are
630 launched, each of them handling as much user contexts at each run. To
631 actiave this, set the \b contexts/nthreads item to the amount of
632 cores that you have in your computer (or lower than 1 to have
633 the amount of cores auto-detected).
635 Even if you asked several worker threads using the previous option,
636 you can request to start the parallel execution (and pay the
637 associated synchronization costs) only if the potential parallelism is
638 large enough. For that, set the \b contexts/parallel-threshold
639 item to the minimal amount of user contexts needed to start the
640 parallel execution. In any given simulation round, if that amount is
641 not reached, the contexts will be run sequentially directly by the
642 main thread (thus saving the synchronization costs). Note that this
643 option is mainly useful when the grain of the user code is very fine,
644 because our synchronization is now very efficient.
646 When parallel execution is activated, you can choose the
647 synchronization schema used with the \b contexts/synchro item,
648 which value is either:
649 - \b futex: ultra optimized synchronisation schema, based on futexes
650 (fast user-mode mutexes), and thus only available on Linux systems.
651 This is the default mode when available.
652 - \b posix: slow but portable synchronisation using only POSIX
654 - \b busy_wait: not really a synchronisation: the worker threads
655 constantly request new contexts to execute. It should be the most
656 efficient synchronisation schema, but it loads all the cores of your
657 machine for no good reason. You probably prefer the other less
660 \section options_tracing Configuring the tracing subsystem
662 The \ref outcomes_vizu "tracing subsystem" can be configured in several
663 different ways depending on the nature of the simulator (MSG, SimDag,
664 SMPI) and the kind of traces that need to be obtained. See the \ref
665 tracing_tracing_options "Tracing Configuration Options subsection" to
666 get a detailed description of each configuration option.
668 We detail here a simple way to get the traces working for you, even if
669 you never used the tracing API.
672 - Any SimGrid-based simulator (MSG, SimDag, SMPI, ...) and raw traces:
674 --cfg=tracing:yes --cfg=tracing/uncategorized:yes --cfg=triva/uncategorized:uncat.plist
676 The first parameter activates the tracing subsystem, the second
677 tells it to trace host and link utilization (without any
678 categorization) and the third creates a graph configuration file
679 to configure Triva when analysing the resulting trace file.
681 - MSG or SimDag-based simulator and categorized traces (you need to declare categories and classify your tasks according to them)
683 --cfg=tracing:yes --cfg=tracing/categorized:yes --cfg=triva/categorized:cat.plist
685 The first parameter activates the tracing subsystem, the second
686 tells it to trace host and link categorized utilization and the
687 third creates a graph configuration file to configure Triva when
688 analysing the resulting trace file.
690 - SMPI simulator and traces for a space/time view:
694 The <i>-trace</i> parameter for the smpirun script runs the
695 simulation with --cfg=tracing:yes and --cfg=tracing/smpi:yes. Check the
696 smpirun's <i>-help</i> parameter for additional tracing options.
698 Sometimes you might want to put additional information on the trace to
699 correctly identify them later, or to provide data that can be used to
700 reproduce an experiment. You have two ways to do that:
702 - Add a string on top of the trace file as comment:
704 --cfg=tracing/comment:my_simulation_identifier
707 - Add the contents of a textual file on top of the trace file as comment:
709 --cfg=tracing/comment-file:my_file_with_additional_information.txt
712 Please, use these two parameters (for comments) to make reproducible
713 simulations. For additional details about this and all tracing
714 options, check See the \ref tracing_tracing_options.
716 \section options_msg Configuring MSG
718 \subsection options_msg_debug_multiple_use Debugging MSG
720 Sometimes your application may try to send a task that is still being
721 executed somewhere else, making it impossible to send this task. However,
722 for debugging purposes, one may want to know what the other host is/was
723 doing. This option shows a backtrace of the other process.
725 Enable this option by adding
728 --cfg=msg/debug-multiple-use:on
731 \section options_smpi Configuring SMPI
733 The SMPI interface provides several specific configuration items.
734 These are uneasy to see since the code is usually launched through the
735 \c smiprun script directly.
737 \subsection options_smpi_bench smpi/bench: Automatic benchmarking of SMPI code
739 In SMPI, the sequential code is automatically benchmarked, and these
740 computations are automatically reported to the simulator. That is to
741 say that if you have a large computation between a \c MPI_Recv() and a
742 \c MPI_Send(), SMPI will automatically benchmark the duration of this
743 code, and create an execution task within the simulator to take this
744 into account. For that, the actual duration is measured on the host
745 machine and then scaled to the power of the corresponding simulated
746 machine. The variable \b smpi/host-speed allows to specify the
747 computational speed of the host machine (in flop/s) to use when
748 scaling the execution times. It defaults to 20000, but you really want
749 to update it to get accurate simulation results.
751 When the code is constituted of numerous consecutive MPI calls, the
752 previous mechanism feeds the simulation kernel with numerous tiny
753 computations. The \b smpi/cpu-threshold item becomes handy when this
754 impacts badly the simulation performance. It specifies a threshold (in
755 seconds) below which the execution chunks are not reported to the
756 simulation kernel (default value: 1e-6).
759 The option smpi/cpu-threshold ignores any computation time spent
760 below this threshold. SMPI does not consider the \a amount of these
761 computations; there is no offset for this. Hence, by using a
762 value that is too low, you may end up with unreliable simulation
765 In some cases, however, one may wish to disable simulation of
766 application computation. This is the case when SMPI is used not to
767 simulate an MPI applications, but instead an MPI code that performs
768 "live replay" of another MPI app (e.g., ScalaTrace's replay tool,
769 various on-line simulators that run an app at scale). In this case the
770 computation of the replay/simulation logic should not be simulated by
771 SMPI. Instead, the replay tool or on-line simulator will issue
772 "computation events", which correspond to the actual MPI simulation
773 being replayed/simulated. At the moment, these computation events can
774 be simulated using SMPI by calling internal smpi_execute*() functions.
776 To disable the benchmarking/simulation of computation in the simulated
777 application, the variable \b smpi/simulate-computation should be set to no.
780 This option just ignores the timings in your simulation; it still executes
781 the computations itself. If you want to stop SMPI from doing that,
782 you should check the SMPI_SAMPLE macros, documented in the section
783 \ref SMPI_adapting_speed.
785 Solution | Computations actually executed? | Computations simulated ?
786 ---------------------------------- | ------------------------------- | ------------------------
787 --cfg=smpi/simulate-computation:no | Yes | No, never
788 --cfg=smpi/cpu-threshold:42 | Yes, in all cases | Only if it lasts more than 42 seconds
789 SMPI_SAMPLE() macro | Only once per loop nest (see @ref SMPI_adapting_speed "documentation") | Always
791 \subsection options_model_smpi_adj_file smpi/comp-adjustment-file: Slow-down or speed-up parts of your code.
793 This option allows you to pass a file that contains two columns: The first column
794 defines the section that will be subject to a speedup; the second column is the speedup.
800 "exchange_1.f:30:exchange_1.f:130",1.18244559422142
803 The first line is the header - you must include it.
804 The following line means that the code between two consecutive MPI calls on
805 line 30 in exchange_1.f and line 130 in exchange_1.f should receive a speedup
806 of 1.18244559422142. The value for the second column is therefore a speedup, if it is
807 larger than 1 and a slow-down if it is smaller than 1. Nothing will be changed if it is
810 Of course, you can set any arbitrary filenames you want (so the start and end don't have to be
811 in the same file), but be aware that this mechanism only supports @em consecutive calls!
814 Please note that you must pass the \b -trace-call-location flag to smpicc
815 or smpiff, respectively! This flag activates some macro definitions in our
816 mpi.h / mpi.f files that help with obtaining the call location.
818 \subsection options_model_smpi_bw_factor smpi/bw-factor: Bandwidth factors
820 The possible throughput of network links is often dependent on the
821 message sizes, as protocols may adapt to different message sizes. With
822 this option, a series of message sizes and factors are given, helping
823 the simulation to be more realistic. For instance, the current
827 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
830 So, messages with size 65472 and more will get a total of MAX_BANDWIDTH*0.940694,
831 messages of size 15424 to 65471 will get MAX_BANDWIDTH*0.697866 and so on.
832 Here, MAX_BANDWIDTH denotes the bandwidth of the link.
835 The SimGrid-Team has developed a script to help you determine these
836 values. You can find more information and the download here:
837 1. http://simgrid.gforge.inria.fr/contrib/smpi-calibration-doc.html
838 2. http://simgrid.gforge.inria.fr/contrib/smpi-saturation-doc.html
840 \subsection options_smpi_timing smpi/display-timing: Reporting simulation time
842 \b Default: 0 (false)
844 Most of the time, you run MPI code with SMPI to compute the time it
845 would take to run it on a platform. But since the
846 code is run through the \c smpirun script, you don't have any control
847 on the launcher code, making it difficult to report the simulated time
848 when the simulation ends. If you set the \b smpi/display-timing item
849 to 1, \c smpirun will display this information when the simulation ends. \verbatim
850 Simulation time: 1e3 seconds.
853 \subsection options_model_smpi_lat_factor smpi/lat-factor: Latency factors
855 The motivation and syntax for this option is identical to the motivation/syntax
856 of smpi/bw-factor, see \ref options_model_smpi_bw_factor for details.
858 There is an important difference, though: While smpi/bw-factor \a reduces the
859 actual bandwidth (i.e., values between 0 and 1 are valid), latency factors
860 increase the latency, i.e., values larger than or equal to 1 are valid here.
862 This is the default value:
865 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
869 The SimGrid-Team has developed a script to help you determine these
870 values. You can find more information and the download here:
871 1. http://simgrid.gforge.inria.fr/contrib/smpi-calibration-doc.html
872 2. http://simgrid.gforge.inria.fr/contrib/smpi-saturation-doc.html
874 \subsection options_smpi_papi_events smpi/papi-events: Trace hardware counters with PAPI
877 This option is experimental and will be subject to change.
878 This feature currently requires superuser privileges, as registers are queried.
879 Only use this feature with code you trust! Call smpirun for instance via
880 smpirun -wrapper "sudo " <your-parameters>
881 or run sudo sh -c "echo 0 > /proc/sys/kernel/perf_event_paranoid"
882 In the later case, sudo will not be required.
885 This option is only available when SimGrid was compiled with PAPI support.
887 This option takes the names of PAPI counters and adds their respective values
888 to the trace files. (See Section \ref tracing_tracing_options.)
890 It is planned to make this feature available on a per-process (or per-thread?) basis.
891 The first draft, however, just implements a "global" (i.e., for all processes) set
892 of counters, the "default" set.
895 --cfg=smpi/papi-events:"default:PAPI_L3_LDM:PAPI_L2_LDM"
898 \subsection options_smpi_privatization smpi/privatization: Automatic privatization of global variables
900 MPI executables are usually meant to be executed in separated processes, but SMPI is
901 executed in only one process. Global variables from executables will be placed
902 in the same memory zone and shared between processes, causing intricate bugs.
903 Several options are possible to avoid this, as described in the main
904 <a href="https://hal.inria.fr/hal-01415484">SMPI publication</a>.
905 SimGrid provides two ways of automatically privatizing the globals,
906 and this option allows to choose between them.
908 - <b>no</b> (default): Do not automatically privatize variables.
909 - <b>mmap</b> or <b>yes</b>: Runtime automatic switching of the data segments.\n
910 SMPI stores a copy of each global data segment for each process,
911 and at each context switch replaces the actual data with its copy
912 from the right process. No copy actually occures as this mechanism
913 uses mmap for efficiency. As such, it is for now limited to
914 systems supporting this functionnality (all Linux and most BSD).\n
915 Another limitation is that SMPI only accounts for global variables
916 defined in the executable. If the processes use external global
917 variables from dynamic libraries, they won't be switched
918 correctly. The easiest way to solve this is to statically link
919 against the library with these globals (but you should never
920 statically link against the simgrid library itself).
921 - <b>dlopen</b>: Link multiple times against the binary.\n
922 SMPI loads several copy of the same binary in memory, resulting in
923 the natural duplication global variables. Since the dynamic linker
924 refuses to link the same file several times, the binary is copied
925 in a temporary file before being dl-loaded (it is erased right
927 Note that this feature is somewhat experimental at time of writing
928 (v3.16) but seems to work.\n
929 This approach greatly speeds up the context switching, down to
930 about 40 CPU cycles with our raw contextes, instead of requesting
931 several syscalls with the \c mmap approach. Another advantage is
932 that it permits to run the SMPI contexts in parallel, which is
933 obviously not possible with the \c mmap approach.\n
934 Further work may be possible to alleviate the memory and disk
935 overconsumption. It seems that we could
936 <a href="https://lwn.net/Articles/415889/">punch holes</a>
937 in the files before dl-loading them to remove the code and
938 constants, and mmap these area onto a unique copy. This require
939 to understand the ELF layout of the file, but would
940 reduce the disk- and memory- usage to the bare minimum. In
941 addition, this would reduce the pressure on the CPU caches (in
942 particular on instruction one).
945 This configuration option cannot be set in your platform file. You can only
946 pass it as an argument to smpirun.
948 \subsection options_model_smpi_detached Simulating MPI detached send
950 This threshold specifies the size in bytes under which the send will return
951 immediately. This is different from the threshold detailed in \ref options_model_network_asyncsend
952 because the message is not effectively sent when the send is posted. SMPI still waits for the
953 correspondant receive to be posted to perform the communication operation. This threshold can be set
954 by changing the \b smpi/send-is-detached-thresh item. The default value is 65536.
956 \subsection options_model_smpi_collectives Simulating MPI collective algorithms
958 SMPI implements more than 100 different algorithms for MPI collective communication, to accurately
959 simulate the behavior of most of the existing MPI libraries. The \b smpi/coll-selector item can be used
960 to use the decision logic of either OpenMPI or MPICH libraries (values: ompi or mpich, by default SMPI
961 uses naive version of collective operations). Each collective operation can be manually selected with a
962 \b smpi/collective_name:algo_name. Available algorithms are listed in \ref SMPI_use_colls .
964 \subsection options_model_smpi_iprobe smpi/iprobe: Inject constant times for calls to MPI_Iprobe
966 \b Default value: 0.0001
968 The behavior and motivation for this configuration option is identical with \a smpi/test, see
969 Section \ref options_model_smpi_test for details.
971 \subsection options_model_smpi_iprobe_cpu_usage smpi/iprobe-cpu-usage: Reduce speed for iprobe calls
973 \b Default value: 1 (no change from default behavior)
975 MPI_Iprobe calls can be heavily used in applications. To account correctly for the energy
976 cores spend probing, it is necessary to reduce the load that these calls cause inside
979 For instance, we measured a max power consumption of 220 W for a particular application but
980 only 180 W while this application was probing. Hence, the correct factor that should
981 be passed to this option would be 180/220 = 0.81.
983 \subsection options_model_smpi_init smpi/init: Inject constant times for calls to MPI_Init
987 The behavior for this configuration option is identical with \a smpi/test, see
988 Section \ref options_model_smpi_test for details.
990 \subsection options_model_smpi_ois smpi/ois: Inject constant times for asynchronous send operations
992 This configuration option works exactly as \a smpi/os, see Section \ref options_model_smpi_os.
993 Of course, \a smpi/ois is used to account for MPI_Isend instead of MPI_Send.
995 \subsection options_model_smpi_os smpi/os: Inject constant times for send operations
997 In several network models such as LogP, send (MPI_Send, MPI_Isend) and receive (MPI_Recv)
998 operations incur costs (i.e., they consume CPU time). SMPI can factor these costs in as well, but the
999 user has to configure SMPI accordingly as these values may vary by machine.
1000 This can be done by using smpi/os for MPI_Send operations; for MPI_Isend and
1001 MPI_Recv, use \a smpi/ois and \a smpi/or, respectively. These work exactly as
1004 \a smpi/os can consist of multiple sections; each section takes three values, for example:
1010 Here, the sections are divided by ";" (that is, this example contains two sections).
1011 Furthermore, each section consists of three values.
1013 1. The first value denotes the minimum size for this section to take effect;
1014 read it as "if message size is greater than this value (and other section has a larger
1015 first value that is also smaller than the message size), use this".
1016 In the first section above, this value is "1".
1018 2. The second value is the startup time; this is a constant value that will always
1019 be charged, no matter what the size of the message. In the first section above,
1022 3. The third value is the \a per-byte cost. That is, it is charged for every
1023 byte of the message (incurring cost messageSize*cost_per_byte)
1024 and hence accounts also for larger messages. In the first
1025 section of the example above, this value is "2".
1027 Now, SMPI always checks which section it should take for a given message; that is,
1028 if a message of size 11 is sent with the configuration of the example above, only
1029 the second section will be used, not the first, as the first value of the second
1030 section is closer to the message size. Hence, a message of size 11 incurs the
1031 following cost inside MPI_Send:
1037 As 5 is the startup cost and 1 is the cost per byte.
1040 The order of sections can be arbitrary; they will be ordered internally.
1042 \subsection options_model_smpi_or smpi/or: Inject constant times for receive operations
1044 This configuration option works exactly as \a smpi/os, see Section \ref options_model_smpi_os.
1045 Of course, \a smpi/or is used to account for MPI_Recv instead of MPI_Send.
1047 \subsection options_model_smpi_test smpi/test: Inject constant times for calls to MPI_Test
1049 \b Default value: 0.0001
1051 By setting this option, you can control the amount of time a process sleeps
1052 when MPI_Test() is called; this is important, because SimGrid normally only
1053 advances the time while communication is happening and thus,
1054 MPI_Test will not add to the time, resulting in a deadlock if used as a
1061 MPI_Test(request, flag, status);
1067 Internally, in order to speed up execution, we use a counter to keep track
1068 on how often we already checked if the handle is now valid or not. Hence, we
1069 actually use counter*SLEEP_TIME, that is, the time MPI_Test() causes the process
1070 to sleep increases linearly with the number of previously failed tests. This
1071 behavior can be disabled by setting smpi/grow-injected-times to no. This will
1072 also disable this behavior for MPI_Iprobe.
1075 \subsection options_model_smpi_shared_malloc smpi/shared-malloc: Factorize malloc()s
1079 If your simulation consumes too much memory, you may want to modify
1080 your code so that the working areas are shared by all MPI ranks. For
1081 example, in a bloc-cyclic matrix multiplication, you will only
1082 allocate one set of blocs, and every processes will share them.
1083 Naturally, this will lead to very wrong results, but this will save a
1084 lot of memory so this is still desirable for some studies. For more on
1085 the motivation for that feature, please refer to the
1086 <a href="https://simgrid.github.io/SMPI_CourseWare/topic_understanding_performance/matrixmultiplication/">relevant
1087 section</a> of the SMPI CourseWare (see Activity #2.2 of the pointed
1088 assignment). In practice, change the call to malloc() and free() into
1089 SMPI_SHARED_MALLOC() and SMPI_SHARED_FREE().
1091 SMPI provides 2 algorithms for this feature. The first one, called \c
1092 local, allocates one bloc per call to SMPI_SHARED_MALLOC() in your
1093 code (each call location gets its own bloc) and this bloc is shared
1094 amongst all MPI ranks. This is implemented with the shm_* functions
1095 to create a new POSIX shared memory object (kept in RAM, in /dev/shm)
1096 for each shared bloc.
1098 With the \c global algorithm, each call to SMPI_SHARED_MALLOC()
1099 returns a new adress, but it only points to a shadow bloc: its memory
1100 area is mapped on a 1MiB file on disk. If the returned bloc is of size
1101 N MiB, then the same file is mapped N times to cover the whole bloc.
1102 At the end, no matter how many SMPI_SHARED_MALLOC you do, this will
1103 only consume 1 MiB in memory.
1105 You can disable this behavior and come back to regular mallocs (for
1106 example for debugging purposes) using \c "no" as a value.
1108 \subsection options_model_smpi_wtime smpi/wtime: Inject constant times for calls to MPI_Wtime
1112 By setting this option, you can control the amount of time a process sleeps
1113 when MPI_Wtime() is called; this is important, because SimGrid normally only
1114 advances the time while communication is happening and thus,
1115 MPI_Wtime will not add to the time, resulting in a deadlock if used as a
1121 while(MPI_Wtime() < some_time_bound) {
1126 If the time is never advanced, this loop will clearly never end as MPI_Wtime()
1127 always returns the same value. Hence, pass a (small) value to the smpi/wtime
1128 option to force a call to MPI_Wtime to advance the time as well.
1131 \section options_generic Configuring other aspects of SimGrid
1133 \subsection options_generic_clean_atexit Cleanup before termination
1135 The C / C++ standard contains a function called \b [atexit](http://www.cplusplus.com/reference/cstdlib/atexit/).
1136 atexit registers callbacks, which are called just before the program terminates.
1138 By setting the configuration option clean-atexit to 1 (true), a callback
1139 is registered and will clean up some variables and terminate/cleanup the tracing.
1141 TODO: Add when this should be used.
1143 \subsection options_generic_path XML file inclusion path
1145 It is possible to specify a list of directories to search into for the
1146 \<include\> tag in XML files by using the \b path configuration
1147 item. To add several directory to the path, set the configuration
1148 item several times, as in \verbatim
1149 --cfg=path:toto --cfg=path:tutu
1152 \subsection options_generic_exit Behavior on Ctrl-C
1154 By default, when Ctrl-C is pressed, the status of all existing
1155 simulated processes is displayed before exiting the simulation. This is very useful to debug your
1156 code, but it can reveal troublesome in some cases (such as when the
1157 amount of processes becomes really big). This behavior is disabled
1158 when \b verbose-exit is set to 0 (it is to 1 by default).
1160 \subsection options_exception_cutpath Truncate local path from exception backtrace
1163 --cfg=exceptions/cutpath:1
1166 This configuration option is used to remove the path from the
1167 backtrace shown when an exception is thrown. This is mainly useful for
1168 the tests: the full file path makes the tests not reproducible, and
1169 thus failing as we are currently comparing output. Clearly, the path
1170 used on different machines are almost guaranteed to be different and
1171 hence, the output would mismatch, causing the test to fail.
1173 \section options_log Logging Configuration
1175 It can be done by using XBT. Go to \ref XBT_log for more details.
1177 \section options_index Index of all existing configuration options
1180 Almost all options are defined in <i>src/simgrid/sg_config.c</i>. You may
1181 want to check this file, too, but this index should be somewhat complete
1182 for the moment (May 2015).
1185 \b Please \b note: You can also pass the command-line option "--help" and
1186 "--help-cfg" to an executable that uses simgrid.
1188 - \c clean-atexit: \ref options_generic_clean_atexit
1190 - \c contexts/factory: \ref options_virt_factory
1191 - \c contexts/guard-size: \ref options_virt_guard_size
1192 - \c contexts/nthreads: \ref options_virt_parallel
1193 - \c contexts/parallel_threshold: \ref options_virt_parallel
1194 - \c contexts/stack-size: \ref options_virt_stacksize
1195 - \c contexts/synchro: \ref options_virt_parallel
1197 - \c cpu/maxmin-selective-update: \ref options_model_optim
1198 - \c cpu/model: \ref options_model_select
1199 - \c cpu/optim: \ref options_model_optim
1201 - \c exception/cutpath: \ref options_exception_cutpath
1203 - \c host/model: \ref options_model_select
1205 - \c maxmin/precision: \ref options_model_precision
1206 - \c maxmin/concurrency-limit: \ref options_concurrency_limit
1208 - \c msg/debug-multiple-use: \ref options_msg_debug_multiple_use
1210 - \c model-check: \ref options_modelchecking
1211 - \c model-check/checkpoint: \ref options_modelchecking_steps
1212 - \c model-check/communications-determinism: \ref options_modelchecking_comm_determinism
1213 - \c model-check/dot-output: \ref options_modelchecking_dot_output
1214 - \c model-check/hash: \ref options_modelchecking_hash
1215 - \c model-check/property: \ref options_modelchecking_liveness
1216 - \c model-check/max-depth: \ref options_modelchecking_max_depth
1217 - \c model-check/record: \ref options_modelchecking_recordreplay
1218 - \c model-check/reduction: \ref options_modelchecking_reduction
1219 - \c model-check/replay: \ref options_modelchecking_recordreplay
1220 - \c model-check/send-determinism: \ref options_modelchecking_comm_determinism
1221 - \c model-check/sparse-checkpoint: \ref options_modelchecking_sparse_checkpoint
1222 - \c model-check/termination: \ref options_modelchecking_termination
1223 - \c model-check/timeout: \ref options_modelchecking_timeout
1224 - \c model-check/visited: \ref options_modelchecking_visited
1226 - \c network/bandwidth-factor: \ref options_model_network_coefs
1227 - \c network/crosstraffic: \ref options_model_network_crosstraffic
1228 - \c network/latency-factor: \ref options_model_network_coefs
1229 - \c network/maxmin-selective-update: \ref options_model_optim
1230 - \c network/model: \ref options_model_select
1231 - \c network/optim: \ref options_model_optim
1232 - \c network/sender_gap: \ref options_model_network_sendergap
1233 - \c network/TCP-gamma: \ref options_model_network_gamma
1234 - \c network/weight-S: \ref options_model_network_coefs
1236 - \c ns3/TcpModel: \ref options_pls
1237 - \c path: \ref options_generic_path
1238 - \c plugin: \ref options_generic_plugin
1240 - \c storage/max_file_descriptors: \ref option_model_storage_maxfd
1242 - \c surf/precision: \ref options_model_precision
1244 - \c <b>For collective operations of SMPI, please refer to Section \ref options_index_smpi_coll</b>
1245 - \c smpi/async-small-thresh: \ref options_model_network_asyncsend
1246 - \c smpi/bw-factor: \ref options_model_smpi_bw_factor
1247 - \c smpi/coll-selector: \ref options_model_smpi_collectives
1248 - \c smpi/comp-adjustment-file: \ref options_model_smpi_adj_file
1249 - \c smpi/cpu-threshold: \ref options_smpi_bench
1250 - \c smpi/display-timing: \ref options_smpi_timing
1251 - \c smpi/grow-injected-times: \ref options_model_smpi_test
1252 - \c smpi/host-speed: \ref options_smpi_bench
1253 - \c smpi/IB-penalty-factors: \ref options_model_network_coefs
1254 - \c smpi/iprobe: \ref options_model_smpi_iprobe
1255 - \c smpi/iprobe-cpu-usage: \ref options_model_smpi_iprobe_cpu_usage
1256 - \c smpi/init: \ref options_model_smpi_init
1257 - \c smpi/lat-factor: \ref options_model_smpi_lat_factor
1258 - \c smpi/ois: \ref options_model_smpi_ois
1259 - \c smpi/or: \ref options_model_smpi_or
1260 - \c smpi/os: \ref options_model_smpi_os
1261 - \c smpi/papi-events: \ref options_smpi_papi_events
1262 - \c smpi/privatization: \ref options_smpi_privatization
1263 - \c smpi/send-is-detached-thresh: \ref options_model_smpi_detached
1264 - \c smpi/shared-malloc: \ref options_model_smpi_shared_malloc
1265 - \c smpi/simulate-computation: \ref options_smpi_bench
1266 - \c smpi/test: \ref options_model_smpi_test
1267 - \c smpi/wtime: \ref options_model_smpi_wtime
1269 - \c <b>Tracing configuration options can be found in Section \ref tracing_tracing_options</b>.
1271 - \c storage/model: \ref options_storage_model
1272 - \c verbose-exit: \ref options_generic_exit
1274 - \c vm/model: \ref options_vm_model
1276 \subsection options_index_smpi_coll Index of SMPI collective algorithms options
1278 TODO: All available collective algorithms will be made available via the ``smpirun --help-coll`` command.