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 in a system can be tuned through
205 the \b maxmin/concurrency_limit item (default value: 100). Setting a higher value can lift some limitations, such as the number of concurrent processes running on a single host.
207 \subsection options_model_network Configuring the Network model
209 \subsubsection options_model_network_gamma Maximal TCP window size
211 The analytical models need to know the maximal TCP window size to take
212 the TCP congestion mechanism into account. This is set to 20000 by
213 default, but can be changed using the \b network/TCP-gamma item.
215 On linux, this value can be retrieved using the following
216 commands. Both give a set of values, and you should use the last one,
217 which is the maximal size.\verbatim
218 cat /proc/sys/net/ipv4/tcp_rmem # gives the sender window
219 cat /proc/sys/net/ipv4/tcp_wmem # gives the receiver window
222 \subsubsection options_model_network_coefs Correcting important network parameters
224 SimGrid can take network irregularities such as a slow startup or
225 changing behavior depending on the message size into account.
226 You should not change these values unless you really know what you're doing.
228 The corresponding values were computed through data fitting one the
229 timings of packet-level simulators.
232 <a href="http://mescal.imag.fr/membres/arnaud.legrand/articles/simutools09.pdf">Accuracy Study and Improvement of Network Simulation in the SimGrid Framework</a>
233 for more information about these parameters.
235 If you are using the SMPI model, these correction coefficients are
236 themselves corrected by constant values depending on the size of the
237 exchange. Again, only hardcore experts should bother about this fact.
239 InfiniBand network behavior can be modeled through 3 parameters, as explained in
240 <a href="http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf">this PhD thesis</a>.
241 These factors can be changed through the following option:
244 smpi/IB-penalty-factors:"βe;βs;γs"
247 By default SMPI uses factors computed on the Stampede Supercomputer at TACC, with optimal
248 deployment of processes on nodes.
250 \subsubsection options_model_network_crosstraffic Simulating cross-traffic
252 As of SimGrid v3.7, cross-traffic effects can be taken into account in
253 analytical simulations. It means that ongoing and incoming
254 communication flows are treated independently. In addition, the LV08
255 model adds 0.05 of usage on the opposite direction for each new
256 created flow. This can be useful to simulate some important TCP
257 phenomena such as ack compression.
259 For that to work, your platform must have two links for each
260 pair of interconnected hosts. An example of usable platform is
261 available in <tt>examples/platforms/crosstraffic.xml</tt>.
263 This is activated through the \b network/crosstraffic item, that
264 can be set to 0 (disable this feature) or 1 (enable it).
266 Note that with the default host model this option is activated by default.
268 \subsubsection options_model_network_sendergap Simulating sender gap
270 (this configuration item is experimental and may change or disapear)
272 It is possible to specify a timing gap between consecutive emission on
273 the same network card through the \b network/sender-gap item. This
274 is still under investigation as of writting, and the default value is
275 to wait 10 microseconds (1e-5 seconds) between emissions.
277 \subsubsection options_model_network_asyncsend Simulating asyncronous send
279 (this configuration item is experimental and may change or disapear)
281 It is possible to specify that messages below a certain size will be sent
282 as soon as the call to MPI_Send is issued, without waiting for the
283 correspondant receive. This threshold can be configured through the
284 \b smpi/async-small-thresh item. The default value is 0. This behavior can also be
285 manually set for MSG mailboxes, by setting the receiving mode of the mailbox
286 with a call to \ref MSG_mailbox_set_async . For MSG, all messages sent to this
287 mailbox will have this behavior, so consider using two mailboxes if needed.
289 This value needs to be smaller than or equals to the threshold set at
290 \ref options_model_smpi_detached , because asynchronous messages are
291 meant to be detached as well.
293 \subsubsection options_pls Configuring packet-level pseudo-models
295 When using the packet-level pseudo-models, several specific
296 configuration flags are provided to configure the associated tools.
297 There is by far not enough such SimGrid flags to cover every aspects
298 of the associated tools, since we only added the items that we
299 needed ourselves. Feel free to request more items (or even better:
300 provide patches adding more items).
302 When using NS3, the only existing item is \b ns3/TcpModel,
303 corresponding to the ns3::TcpL4Protocol::SocketType configuration item
304 in NS3. The only valid values (enforced on the SimGrid side) are
305 'NewReno' or 'Reno' or 'Tahoe'.
307 \subsection options_model_storage Configuring the Storage model
309 \subsubsection option_model_storage_maxfd Maximum amount of file descriptors per host
311 Each host maintains a fixed-size array of its file descriptors. You
312 can change its size (1024 by default) through the \b
313 storage/max_file_descriptors item to either enlarge it if your
314 application requires it or to reduce it to save memory space.
316 \section options_modelchecking Configuring the Model-Checking
318 To enable the SimGrid model-checking support the program should
319 be executed using the simgrid-mc wrapper:
321 simgrid-mc ./my_program
324 Safety properties are expressed as assertions using the function
326 void MC_assert(int prop);
329 \subsection options_modelchecking_liveness Specifying a liveness property
331 If you want to specify liveness properties (beware, that's
332 experimental), you have to pass them on the command line, specifying
333 the name of the file containing the property, as formatted by the
337 --cfg=model-check/property:<filename>
340 \subsection options_modelchecking_steps Going for stateful verification
342 By default, the system is backtracked to its initial state to explore
343 another path instead of backtracking to the exact step before the fork
344 that we want to explore (this is called stateless verification). This
345 is done this way because saving intermediate states can rapidly
346 exhaust the available memory. If you want, you can change the value of
347 the <tt>model-check/checkpoint</tt> variable. For example, the
348 following configuration will ask to take a checkpoint every step.
349 Beware, this will certainly explode your memory. Larger values are
350 probably better, make sure to experiment a bit to find the right
351 setting for your specific system.
354 --cfg=model-check/checkpoint:1
357 \subsection options_modelchecking_reduction Specifying the kind of reduction
359 The main issue when using the model-checking is the state space
360 explosion. To counter that problem, several exploration reduction
361 techniques can be used. There is unfortunately no silver bullet here,
362 and the most efficient reduction techniques cannot be applied to any
363 properties. In particular, the DPOR method cannot be applied on
364 liveness properties since it may break some cycles in the exploration
365 that are important to the property validity.
368 --cfg=model-check/reduction:<technique>
371 For now, this configuration variable can take 2 values:
372 * none: Do not apply any kind of reduction (mandatory for now for
374 * dpor: Apply Dynamic Partial Ordering Reduction. Only valid if you
375 verify local safety properties.
377 \subsection options_modelchecking_visited model-check/visited, Cycle detection
379 In order to detect cycles, the model-checker needs to check if a new explored
380 state is in fact the same state than a previous one. In order to do this,
381 the model-checker can take a snapshot of each visited state: this snapshot is
382 then used to compare it with subsequent states in the exploration graph.
384 The \b model-check/visited is the maximum number of states which are stored in
385 memory. If the maximum number of snapshotted state is reached some states will
386 be removed from the memory and some cycles might be missed.
388 By default, no state is snapshotted and cycles cannot be detected.
390 \subsection options_modelchecking_termination model-check/termination, Non termination detection
392 The \b model-check/termination configuration item can be used to report if a
393 non-termination execution path has been found. This is a path with a cycle
394 which means that the program might never terminate.
396 This only works in safety mode.
398 This options is disabled by default.
400 \subsection options_modelchecking_dot_output model-check/dot-output, Dot output
402 If set, the \b model-check/dot-output configuration item is the name of a file
403 in which to write a dot file of the path leading the found property (safety or
404 liveness violation) as well as the cycle for liveness properties. This dot file
405 can then fed to the graphviz dot tool to generate an corresponding graphical
408 \subsection options_modelchecking_max_depth model-check/max_depth, Depth limit
410 The \b model-checker/max-depth can set the maximum depth of the exploration
411 graph of the model-checker. If this limit is reached, a logging message is
412 sent and the results might not be exact.
414 By default, there is not depth limit.
416 \subsection options_modelchecking_timeout Handling of timeout
418 By default, the model-checker does not handle timeout conditions: the `wait`
419 operations never time out. With the \b model-check/timeout configuration item
420 set to \b yes, the model-checker will explore timeouts of `wait` operations.
422 \subsection options_modelchecking_comm_determinism Communication determinism
424 The \b model-check/communications-determinism and
425 \b model-check/send-determinism items can be used to select the communication
426 determinism mode of the model-checker which checks determinism properties of
427 the communications of an application.
429 \subsection options_modelchecking_sparse_checkpoint Per page checkpoints
431 When the model-checker is configured to take a snapshot of each explored state
432 (with the \b model-checker/visited item), the memory consumption can rapidly
433 reach GiB ou Tib of memory. However, for many workloads, the memory does not
434 change much between different snapshots and taking a complete copy of each
435 snapshot is a waste of memory.
437 The \b model-check/sparse-checkpoint option item can be set to \b yes in order
438 to avoid making a complete copy of each snapshot: instead, each snapshot will be
439 decomposed in blocks which will be stored separately.
440 If multiple snapshots share the same block (or if the same block
441 is used in the same snapshot), the same copy of the block will be shared leading
442 to a reduction of the memory footprint.
444 For many applications, this option considerably reduces the memory consumption.
445 In somes cases, the model-checker might be slightly slower because of the time
446 taken to manage the metadata about the blocks. In other cases however, this
447 snapshotting strategy will be much faster by reducing the cache consumption.
448 When the memory consumption is important, by avoiding to hit the swap or
449 reducing the swap usage, this option might be much faster than the basic
450 snapshotting strategy.
452 This option is currently disabled by default.
454 \subsection options_mc_perf Performance considerations for the model checker
456 The size of the stacks can have a huge impact on the memory
457 consumption when using model-checking. By default, each snapshot will
458 save a copy of the whole stacks and not only of the part which is
459 really meaningful: you should expect the contribution of the memory
460 consumption of the snapshots to be \f$ \mbox{number of processes}
461 \times \mbox{stack size} \times \mbox{number of states} \f$.
463 The \b model-check/sparse-checkpoint can be used to reduce the memory
464 consumption by trying to share memory between the different snapshots.
466 When compiled against the model checker, the stacks are not
467 protected with guards: if the stack size is too small for your
468 application, the stack will silently overflow on other parts of the
469 memory (see \ref options_virt_guard_size).
471 \subsection options_modelchecking_hash Hashing of the state (experimental)
473 Usually most of the time of the model-checker is spent comparing states. This
474 process is complicated and consumes a lot of bandwidth and cache.
475 In order to speedup the state comparison, the experimental \b model-checker/hash
476 configuration item enables the computation of a hash summarizing as much
477 information of the state as possible into a single value. This hash can be used
478 to avoid most of the comparisons: the costly comparison is then only used when
479 the hashes are identical.
481 Currently most of the state is not included in the hash because the
482 implementation was found to be buggy and this options is not as useful as
483 it could be. For this reason, it is currently disabled by default.
485 \subsection options_modelchecking_recordreplay Record/replay (experimental)
487 As the model-checker keeps jumping at different places in the execution graph,
488 it is difficult to understand what happens when trying to debug an application
489 under the model-checker. Event the output of the program is difficult to
490 interpret. Moreover, the model-checker does not behave nicely with advanced
491 debugging tools such as valgrind. For those reason, to identify a trajectory
492 in the execution graph with the model-checker and replay this trajcetory and
493 without the model-checker black-magic but with more standard tools
494 (such as a debugger, valgrind, etc.). For this reason, Simgrid implements an
495 experimental record/replay functionnality in order to record a trajectory with
496 the model-checker and replay it without the model-checker.
498 When the model-checker finds an interesting path in the application execution
499 graph (where a safety or liveness property is violated), it can generate an
500 identifier for this path. In order to enable this behavious the
501 \b model-check/record must be set to \b yes. By default, this behaviour is not
504 This is an example of output:
507 [ 0.000000] (0:@) Check a safety property
508 [ 0.000000] (0:@) **************************
509 [ 0.000000] (0:@) *** PROPERTY NOT VALID ***
510 [ 0.000000] (0:@) **************************
511 [ 0.000000] (0:@) Counter-example execution trace:
512 [ 0.000000] (0:@) Path = 1/3;1/4
513 [ 0.000000] (0:@) [(1)Tremblay (app)] MC_RANDOM(3)
514 [ 0.000000] (0:@) [(1)Tremblay (app)] MC_RANDOM(4)
515 [ 0.000000] (0:@) Expanded states = 27
516 [ 0.000000] (0:@) Visited states = 68
517 [ 0.000000] (0:@) Executed transitions = 46
520 This path can then be replayed outside of the model-checker (and even in
521 non-MC build of simgrid) by setting the \b model-check/replay item to the given
522 path. The other options should be the same (but the model-checker should
525 The format and meaning of the path may change between different releases so
526 the same release of Simgrid should be used for the record phase and the replay
529 \section options_virt Configuring the User Process Virtualization
531 \subsection options_virt_factory Selecting the virtualization factory
533 In SimGrid, the user code is virtualized in a specific mechanism
534 that allows the simulation kernel to control its execution: when a user
535 process requires a blocking action (such as sending a message), it is
536 interrupted, and only gets released when the simulated clock reaches
537 the point where the blocking operation is done.
539 In SimGrid, the containers in which user processes are virtualized are
540 called contexts. Several context factory are provided, and you can
541 select the one you want to use with the \b contexts/factory
542 configuration item. Some of the following may not exist on your
543 machine because of portability issues. In any case, the default one
544 should be the most effcient one (please report bugs if the
545 auto-detection fails for you). They are approximately sorted here from
546 the slowest to the most effient:
548 - \b thread: very slow factory using full featured threads (either
549 pthreads or windows native threads). They are slow but very
550 standard. Some debuggers or profilers only work with this factory.
551 - \b java: Java applications are virtualized onto java threads (that
552 are regular pthreads registered to the JVM)
553 - \b ucontext: fast factory using System V contexts (Linux and FreeBSD only)
554 - \b boost: This uses the [context implementation](http://www.boost.org/doc/libs/1_59_0/libs/context/doc/html/index.html)
555 of the boost library for a performance that is comparable to our
556 raw implementation.\nInstall the relevant library (e.g. with the
557 libboost-contexts-dev package on Debian/Ubuntu) and recompile
558 SimGrid. Note that our implementation is not compatible with recent
559 implementations of the library, and it will be hard to fix this since
560 the library's author decided to hide an API that we were using.
561 - \b raw: amazingly fast factory using a context switching mechanism
562 of our own, directly implemented in assembly (only available for x86
563 and amd64 platforms for now) and without any unneeded system call.
565 The main reason to change this setting is when the debugging tools get
566 fooled by the optimized context factories. Threads are the most
567 debugging-friendly contextes, as they allow to set breakpoints
568 anywhere with gdb and visualize backtraces for all processes, in order
569 to debug concurrency issues. Valgrind is also more comfortable with
570 threads, but it should be usable with all factories (but the callgrind
571 tool that really don't like raw and ucontext factories).
573 \subsection options_virt_stacksize Adapting the used stack size
575 Each virtualized used process is executed using a specific system
576 stack. The size of this stack has a huge impact on the simulation
577 scalability, but its default value is rather large. This is because
578 the error messages that you get when the stack size is too small are
579 rather disturbing: this leads to stack overflow (overwriting other
580 stacks), leading to segfaults with corrupted stack traces.
582 If you want to push the scalability limits of your code, you might
583 want to reduce the \b contexts/stack-size item. Its default value
584 is 8192 (in KiB), while our Chord simulation works with stacks as small
585 as 16 KiB, for example. For the thread factory, the default value
586 is the one of the system but you can still change it with this parameter.
588 The operating system should only allocate memory for the pages of the
589 stack which are actually used and you might not need to use this in
590 most cases. However, this setting is very important when using the
591 model checker (see \ref options_mc_perf).
593 \subsection options_virt_guard_size Disabling stack guard pages
595 A stack guard page is usually used which prevents the stack of a given
596 actor from overflowing on another stack. But the performance impact
597 may become prohibitive when the amount of actors increases. The
598 option \b contexts:guard-size is the number of stack guard pages used.
599 By setting it to 0, no guard pages will be used: in this case, you
600 should avoid using small stacks (\b stack-size) as the stack will
601 silently overflow on other parts of the memory.
603 When no stack guard page is created, stacks may then silently overflow
604 on other parts of the memory if their size is too small for the
605 application. This happens:
607 - on Windows systems;
608 - when the model checker is enabled;
609 - and of course when guard pages are explicitely disabled (with \b contexts:guard-size=0).
611 \subsection options_virt_parallel Running user code in parallel
613 Parallel execution of the user code is only considered stable in
614 SimGrid v3.7 and higher, and mostly for MSG simulations. SMPI
615 simulations may well fail in parallel mode. It is described in
616 <a href="http://hal.inria.fr/inria-00602216/">INRIA RR-7653</a>.
618 If you are using the \c ucontext or \c raw context factories, you can
619 request to execute the user code in parallel. Several threads are
620 launched, each of them handling as much user contexts at each run. To
621 actiave this, set the \b contexts/nthreads item to the amount of
622 cores that you have in your computer (or lower than 1 to have
623 the amount of cores auto-detected).
625 Even if you asked several worker threads using the previous option,
626 you can request to start the parallel execution (and pay the
627 associated synchronization costs) only if the potential parallelism is
628 large enough. For that, set the \b contexts/parallel-threshold
629 item to the minimal amount of user contexts needed to start the
630 parallel execution. In any given simulation round, if that amount is
631 not reached, the contexts will be run sequentially directly by the
632 main thread (thus saving the synchronization costs). Note that this
633 option is mainly useful when the grain of the user code is very fine,
634 because our synchronization is now very efficient.
636 When parallel execution is activated, you can choose the
637 synchronization schema used with the \b contexts/synchro item,
638 which value is either:
639 - \b futex: ultra optimized synchronisation schema, based on futexes
640 (fast user-mode mutexes), and thus only available on Linux systems.
641 This is the default mode when available.
642 - \b posix: slow but portable synchronisation using only POSIX
644 - \b busy_wait: not really a synchronisation: the worker threads
645 constantly request new contexts to execute. It should be the most
646 efficient synchronisation schema, but it loads all the cores of your
647 machine for no good reason. You probably prefer the other less
650 \section options_tracing Configuring the tracing subsystem
652 The \ref outcomes_vizu "tracing subsystem" can be configured in several
653 different ways depending on the nature of the simulator (MSG, SimDag,
654 SMPI) and the kind of traces that need to be obtained. See the \ref
655 tracing_tracing_options "Tracing Configuration Options subsection" to
656 get a detailed description of each configuration option.
658 We detail here a simple way to get the traces working for you, even if
659 you never used the tracing API.
662 - Any SimGrid-based simulator (MSG, SimDag, SMPI, ...) and raw traces:
664 --cfg=tracing:yes --cfg=tracing/uncategorized:yes --cfg=triva/uncategorized:uncat.plist
666 The first parameter activates the tracing subsystem, the second
667 tells it to trace host and link utilization (without any
668 categorization) and the third creates a graph configuration file
669 to configure Triva when analysing the resulting trace file.
671 - MSG or SimDag-based simulator and categorized traces (you need to declare categories and classify your tasks according to them)
673 --cfg=tracing:yes --cfg=tracing/categorized:yes --cfg=triva/categorized:cat.plist
675 The first parameter activates the tracing subsystem, the second
676 tells it to trace host and link categorized utilization and the
677 third creates a graph configuration file to configure Triva when
678 analysing the resulting trace file.
680 - SMPI simulator and traces for a space/time view:
684 The <i>-trace</i> parameter for the smpirun script runs the
685 simulation with --cfg=tracing:yes and --cfg=tracing/smpi:yes. Check the
686 smpirun's <i>-help</i> parameter for additional tracing options.
688 Sometimes you might want to put additional information on the trace to
689 correctly identify them later, or to provide data that can be used to
690 reproduce an experiment. You have two ways to do that:
692 - Add a string on top of the trace file as comment:
694 --cfg=tracing/comment:my_simulation_identifier
697 - Add the contents of a textual file on top of the trace file as comment:
699 --cfg=tracing/comment-file:my_file_with_additional_information.txt
702 Please, use these two parameters (for comments) to make reproducible
703 simulations. For additional details about this and all tracing
704 options, check See the \ref tracing_tracing_options.
706 \section options_msg Configuring MSG
708 \subsection options_msg_debug_multiple_use Debugging MSG
710 Sometimes your application may try to send a task that is still being
711 executed somewhere else, making it impossible to send this task. However,
712 for debugging purposes, one may want to know what the other host is/was
713 doing. This option shows a backtrace of the other process.
715 Enable this option by adding
718 --cfg=msg/debug-multiple-use:on
721 \section options_smpi Configuring SMPI
723 The SMPI interface provides several specific configuration items.
724 These are uneasy to see since the code is usually launched through the
725 \c smiprun script directly.
727 \subsection options_smpi_bench smpi/bench: Automatic benchmarking of SMPI code
729 In SMPI, the sequential code is automatically benchmarked, and these
730 computations are automatically reported to the simulator. That is to
731 say that if you have a large computation between a \c MPI_Recv() and a
732 \c MPI_Send(), SMPI will automatically benchmark the duration of this
733 code, and create an execution task within the simulator to take this
734 into account. For that, the actual duration is measured on the host
735 machine and then scaled to the power of the corresponding simulated
736 machine. The variable \b smpi/host-speed allows to specify the
737 computational speed of the host machine (in flop/s) to use when
738 scaling the execution times. It defaults to 20000, but you really want
739 to update it to get accurate simulation results.
741 When the code is constituted of numerous consecutive MPI calls, the
742 previous mechanism feeds the simulation kernel with numerous tiny
743 computations. The \b smpi/cpu-threshold item becomes handy when this
744 impacts badly the simulation performance. It specifies a threshold (in
745 seconds) below which the execution chunks are not reported to the
746 simulation kernel (default value: 1e-6).
749 The option smpi/cpu-threshold ignores any computation time spent
750 below this threshold. SMPI does not consider the \a amount of these
751 computations; there is no offset for this. Hence, by using a
752 value that is too low, you may end up with unreliable simulation
755 In some cases, however, one may wish to disable simulation of
756 application computation. This is the case when SMPI is used not to
757 simulate an MPI applications, but instead an MPI code that performs
758 "live replay" of another MPI app (e.g., ScalaTrace's replay tool,
759 various on-line simulators that run an app at scale). In this case the
760 computation of the replay/simulation logic should not be simulated by
761 SMPI. Instead, the replay tool or on-line simulator will issue
762 "computation events", which correspond to the actual MPI simulation
763 being replayed/simulated. At the moment, these computation events can
764 be simulated using SMPI by calling internal smpi_execute*() functions.
766 To disable the benchmarking/simulation of computation in the simulated
767 application, the variable \b smpi/simulate-computation should be set to no.
770 This option just ignores the timings in your simulation; it still executes
771 the computations itself. If you want to stop SMPI from doing that,
772 you should check the SMPI_SAMPLE macros, documented in the section
773 \ref SMPI_adapting_speed.
775 Solution | Computations actually executed? | Computations simulated ?
776 ---------------------------------- | ------------------------------- | ------------------------
777 --cfg=smpi/simulate-computation:no | Yes | No, never
778 --cfg=smpi/cpu-threshold:42 | Yes, in all cases | Only if it lasts more than 42 seconds
779 SMPI_SAMPLE() macro | Only once per loop nest (see @ref SMPI_adapting_speed "documentation") | Always
781 \subsection options_model_smpi_adj_file smpi/comp-adjustment-file: Slow-down or speed-up parts of your code.
783 This option allows you to pass a file that contains two columns: The first column
784 defines the section that will be subject to a speedup; the second column is the speedup.
790 "exchange_1.f:30:exchange_1.f:130",1.18244559422142
793 The first line is the header - you must include it.
794 The following line means that the code between two consecutive MPI calls on
795 line 30 in exchange_1.f and line 130 in exchange_1.f should receive a speedup
796 of 1.18244559422142. The value for the second column is therefore a speedup, if it is
797 larger than 1 and a slow-down if it is smaller than 1. Nothing will be changed if it is
800 Of course, you can set any arbitrary filenames you want (so the start and end don't have to be
801 in the same file), but be aware that this mechanism only supports @em consecutive calls!
804 Please note that you must pass the \b -trace-call-location flag to smpicc
805 or smpiff, respectively! This flag activates some macro definitions in our
806 mpi.h / mpi.f files that help with obtaining the call location.
808 \subsection options_model_smpi_bw_factor smpi/bw-factor: Bandwidth factors
810 The possible throughput of network links is often dependent on the
811 message sizes, as protocols may adapt to different message sizes. With
812 this option, a series of message sizes and factors are given, helping
813 the simulation to be more realistic. For instance, the current
817 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
820 So, messages with size 65472 and more will get a total of MAX_BANDWIDTH*0.940694,
821 messages of size 15424 to 65471 will get MAX_BANDWIDTH*0.697866 and so on.
822 Here, MAX_BANDWIDTH denotes the bandwidth of the link.
825 The SimGrid-Team has developed a script to help you determine these
826 values. You can find more information and the download here:
827 1. http://simgrid.gforge.inria.fr/contrib/smpi-calibration-doc.html
828 2. http://simgrid.gforge.inria.fr/contrib/smpi-saturation-doc.html
830 \subsection options_smpi_timing smpi/display-timing: Reporting simulation time
832 \b Default: 0 (false)
834 Most of the time, you run MPI code with SMPI to compute the time it
835 would take to run it on a platform. But since the
836 code is run through the \c smpirun script, you don't have any control
837 on the launcher code, making it difficult to report the simulated time
838 when the simulation ends. If you set the \b smpi/display-timing item
839 to 1, \c smpirun will display this information when the simulation ends. \verbatim
840 Simulation time: 1e3 seconds.
843 \subsection options_model_smpi_lat_factor smpi/lat-factor: Latency factors
845 The motivation and syntax for this option is identical to the motivation/syntax
846 of smpi/bw-factor, see \ref options_model_smpi_bw_factor for details.
848 There is an important difference, though: While smpi/bw-factor \a reduces the
849 actual bandwidth (i.e., values between 0 and 1 are valid), latency factors
850 increase the latency, i.e., values larger than or equal to 1 are valid here.
852 This is the default value:
855 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
859 The SimGrid-Team has developed a script to help you determine these
860 values. You can find more information and the download here:
861 1. http://simgrid.gforge.inria.fr/contrib/smpi-calibration-doc.html
862 2. http://simgrid.gforge.inria.fr/contrib/smpi-saturation-doc.html
864 \subsection options_smpi_papi_events smpi/papi-events: Trace hardware counters with PAPI
867 This option is experimental and will be subject to change.
868 This feature currently requires superuser privileges, as registers are queried.
869 Only use this feature with code you trust! Call smpirun for instance via
870 smpirun -wrapper "sudo " <your-parameters>
871 or run sudo sh -c "echo 0 > /proc/sys/kernel/perf_event_paranoid"
872 In the later case, sudo will not be required.
875 This option is only available when SimGrid was compiled with PAPI support.
877 This option takes the names of PAPI counters and adds their respective values
878 to the trace files. (See Section \ref tracing_tracing_options.)
880 It is planned to make this feature available on a per-process (or per-thread?) basis.
881 The first draft, however, just implements a "global" (i.e., for all processes) set
882 of counters, the "default" set.
885 --cfg=smpi/papi-events:"default:PAPI_L3_LDM:PAPI_L2_LDM"
888 \subsection options_smpi_global smpi/privatize-global-variables: Automatic privatization of global variables
890 MPI executables are meant to be executed in separated processes, but SMPI is
891 executed in only one process. Global variables from executables will be placed
892 in the same memory zone and shared between processes, causing hard to find bugs.
893 To avoid this, several options are possible :
894 - Manual edition of the code, for example to add __thread keyword before data
895 declaration, which allows the resulting code to work with SMPI, but only
896 if the thread factory (see \ref options_virt_factory) is used, as global
897 variables are then placed in the TLS (thread local storage) segment.
898 - Source-to-source transformation, to add a level of indirection
899 to the global variables. SMPI does this for F77 codes compiled with smpiff,
900 and used to provide coccinelle scripts for C codes, which are not functional anymore.
901 - Compilation pass, to have the compiler automatically put the data in
903 - Runtime automatic switching of the data segments. SMPI stores a copy of
904 each global data segment for each process, and at each context switch replaces
905 the actual data with its copy from the right process. This mechanism uses mmap,
906 and is for now limited to systems supporting this functionnality (all Linux
907 and some BSD should be compatible).
908 Another limitation is that SMPI only accounts for global variables defined in
909 the executable. If the processes use external global variables from dynamic
910 libraries, they won't be switched correctly. To avoid this, using static
911 linking is advised (but not with the simgrid library, to avoid replicating
912 its own global variables).
914 To use this runtime automatic switching, the variable \b smpi/privatize-global-variables
918 This configuration option cannot be set in your platform file. You can only
919 pass it as an argument to smpirun.
922 \subsection options_model_smpi_detached Simulating MPI detached send
924 This threshold specifies the size in bytes under which the send will return
925 immediately. This is different from the threshold detailed in \ref options_model_network_asyncsend
926 because the message is not effectively sent when the send is posted. SMPI still waits for the
927 correspondant receive to be posted to perform the communication operation. This threshold can be set
928 by changing the \b smpi/send-is-detached-thresh item. The default value is 65536.
930 \subsection options_model_smpi_collectives Simulating MPI collective algorithms
932 SMPI implements more than 100 different algorithms for MPI collective communication, to accurately
933 simulate the behavior of most of the existing MPI libraries. The \b smpi/coll-selector item can be used
934 to use the decision logic of either OpenMPI or MPICH libraries (values: ompi or mpich, by default SMPI
935 uses naive version of collective operations). Each collective operation can be manually selected with a
936 \b smpi/collective_name:algo_name. Available algorithms are listed in \ref SMPI_use_colls .
938 \subsection options_model_smpi_iprobe smpi/iprobe: Inject constant times for calls to MPI_Iprobe
940 \b Default value: 0.0001
942 The behavior and motivation for this configuration option is identical with \a smpi/test, see
943 Section \ref options_model_smpi_test for details.
945 \subsection options_model_smpi_init smpi/init: Inject constant times for calls to MPI_Init
949 The behavior for this configuration option is identical with \a smpi/test, see
950 Section \ref options_model_smpi_test for details.
952 \subsection options_model_smpi_ois smpi/ois: Inject constant times for asynchronous send operations
954 This configuration option works exactly as \a smpi/os, see Section \ref options_model_smpi_os.
955 Of course, \a smpi/ois is used to account for MPI_Isend instead of MPI_Send.
957 \subsection options_model_smpi_os smpi/os: Inject constant times for send operations
959 In several network models such as LogP, send (MPI_Send, MPI_Isend) and receive (MPI_Recv)
960 operations incur costs (i.e., they consume CPU time). SMPI can factor these costs in as well, but the
961 user has to configure SMPI accordingly as these values may vary by machine.
962 This can be done by using smpi/os for MPI_Send operations; for MPI_Isend and
963 MPI_Recv, use \a smpi/ois and \a smpi/or, respectively. These work exactly as
966 \a smpi/os can consist of multiple sections; each section takes three values, for example:
972 Here, the sections are divided by ";" (that is, this example contains two sections).
973 Furthermore, each section consists of three values.
975 1. The first value denotes the minimum size for this section to take effect;
976 read it as "if message size is greater than this value (and other section has a larger
977 first value that is also smaller than the message size), use this".
978 In the first section above, this value is "1".
980 2. The second value is the startup time; this is a constant value that will always
981 be charged, no matter what the size of the message. In the first section above,
984 3. The third value is the \a per-byte cost. That is, it is charged for every
985 byte of the message (incurring cost messageSize*cost_per_byte)
986 and hence accounts also for larger messages. In the first
987 section of the example above, this value is "2".
989 Now, SMPI always checks which section it should take for a given message; that is,
990 if a message of size 11 is sent with the configuration of the example above, only
991 the second section will be used, not the first, as the first value of the second
992 section is closer to the message size. Hence, a message of size 11 incurs the
993 following cost inside MPI_Send:
999 As 5 is the startup cost and 1 is the cost per byte.
1002 The order of sections can be arbitrary; they will be ordered internally.
1004 \subsection options_model_smpi_or smpi/or: Inject constant times for receive operations
1006 This configuration option works exactly as \a smpi/os, see Section \ref options_model_smpi_os.
1007 Of course, \a smpi/or is used to account for MPI_Recv instead of MPI_Send.
1009 \subsection options_model_smpi_test smpi/test: Inject constant times for calls to MPI_Test
1011 \b Default value: 0.0001
1013 By setting this option, you can control the amount of time a process sleeps
1014 when MPI_Test() is called; this is important, because SimGrid normally only
1015 advances the time while communication is happening and thus,
1016 MPI_Test will not add to the time, resulting in a deadlock if used as a
1023 MPI_Test(request, flag, status);
1029 Internally, in order to speed up execution, we use a counter to keep track
1030 on how often we already checked if the handle is now valid or not. Hence, we
1031 actually use counter*SLEEP_TIME, that is, the time MPI_Test() causes the process
1032 to sleep increases linearly with the number of previously failed tests. This
1033 behavior can be disabled by setting smpi/grow-injected-times to no. This will
1034 also disable this behavior for MPI_Iprobe.
1037 \subsection options_model_smpi_use_shared_malloc smpi/use-shared-malloc: Factorize malloc()s
1041 SMPI can use shared memory by calling shm_* functions; this might speed up the simulation.
1042 This opens or creates a new POSIX shared memory object, kept in RAM, in /dev/shm.
1044 If you want to disable this behavior, set the value to 0.
1046 \subsection options_model_smpi_wtime smpi/wtime: Inject constant times for calls to MPI_Wtime
1050 By setting this option, you can control the amount of time a process sleeps
1051 when MPI_Wtime() is called; this is important, because SimGrid normally only
1052 advances the time while communication is happening and thus,
1053 MPI_Wtime will not add to the time, resulting in a deadlock if used as a
1059 while(MPI_Wtime() < some_time_bound) {
1064 If the time is never advanced, this loop will clearly never end as MPI_Wtime()
1065 always returns the same value. Hence, pass a (small) value to the smpi/wtime
1066 option to force a call to MPI_Wtime to advance the time as well.
1069 \section options_generic Configuring other aspects of SimGrid
1071 \subsection options_generic_clean_atexit Cleanup before termination
1073 The C / C++ standard contains a function called \b [atexit](http://www.cplusplus.com/reference/cstdlib/atexit/).
1074 atexit registers callbacks, which are called just before the program terminates.
1076 By setting the configuration option clean-atexit to 1 (true), a callback
1077 is registered and will clean up some variables and terminate/cleanup the tracing.
1079 TODO: Add when this should be used.
1081 \subsection options_generic_path XML file inclusion path
1083 It is possible to specify a list of directories to search into for the
1084 \<include\> tag in XML files by using the \b path configuration
1085 item. To add several directory to the path, set the configuration
1086 item several times, as in \verbatim
1087 --cfg=path:toto --cfg=path:tutu
1090 \subsection options_generic_exit Behavior on Ctrl-C
1092 By default, when Ctrl-C is pressed, the status of all existing
1093 simulated processes is displayed before exiting the simulation. This is very useful to debug your
1094 code, but it can reveal troublesome in some cases (such as when the
1095 amount of processes becomes really big). This behavior is disabled
1096 when \b verbose-exit is set to 0 (it is to 1 by default).
1098 \subsection options_exception_cutpath Truncate local path from exception backtrace
1101 --cfg=exceptions/cutpath:1
1104 This configuration option is used to remove the path from the
1105 backtrace shown when an exception is thrown. This is mainly useful for
1106 the tests: the full file path makes the tests not reproducible, and
1107 thus failing as we are currently comparing output. Clearly, the path
1108 used on different machines are almost guaranteed to be different and
1109 hence, the output would mismatch, causing the test to fail.
1111 \section options_log Logging Configuration
1113 It can be done by using XBT. Go to \ref XBT_log for more details.
1115 \section options_index Index of all existing configuration options
1118 Almost all options are defined in <i>src/simgrid/sg_config.c</i>. You may
1119 want to check this file, too, but this index should be somewhat complete
1120 for the moment (May 2015).
1123 \b Please \b note: You can also pass the command-line option "--help" and
1124 "--help-cfg" to an executable that uses simgrid.
1126 - \c clean-atexit: \ref options_generic_clean_atexit
1128 - \c contexts/factory: \ref options_virt_factory
1129 - \c contexts/guard-size: \ref options_virt_guard_size
1130 - \c contexts/nthreads: \ref options_virt_parallel
1131 - \c contexts/parallel_threshold: \ref options_virt_parallel
1132 - \c contexts/stack-size: \ref options_virt_stacksize
1133 - \c contexts/synchro: \ref options_virt_parallel
1135 - \c cpu/maxmin-selective-update: \ref options_model_optim
1136 - \c cpu/model: \ref options_model_select
1137 - \c cpu/optim: \ref options_model_optim
1139 - \c exception/cutpath: \ref options_exception_cutpath
1141 - \c host/model: \ref options_model_select
1143 - \c maxmin/precision: \ref options_model_precision
1145 - \c msg/debug-multiple-use: \ref options_msg_debug_multiple_use
1147 - \c model-check: \ref options_modelchecking
1148 - \c model-check/checkpoint: \ref options_modelchecking_steps
1149 - \c model-check/communications-determinism: \ref options_modelchecking_comm_determinism
1150 - \c model-check/dot-output: \ref options_modelchecking_dot_output
1151 - \c model-check/hash: \ref options_modelchecking_hash
1152 - \c model-check/property: \ref options_modelchecking_liveness
1153 - \c model-check/max-depth: \ref options_modelchecking_max_depth
1154 - \c model-check/record: \ref options_modelchecking_recordreplay
1155 - \c model-check/reduction: \ref options_modelchecking_reduction
1156 - \c model-check/replay: \ref options_modelchecking_recordreplay
1157 - \c model-check/send-determinism: \ref options_modelchecking_comm_determinism
1158 - \c model-check/sparse-checkpoint: \ref options_modelchecking_sparse_checkpoint
1159 - \c model-check/termination: \ref options_modelchecking_termination
1160 - \c model-check/timeout: \ref options_modelchecking_timeout
1161 - \c model-check/visited: \ref options_modelchecking_visited
1163 - \c network/bandwidth-factor: \ref options_model_network_coefs
1164 - \c network/crosstraffic: \ref options_model_network_crosstraffic
1165 - \c network/latency-factor: \ref options_model_network_coefs
1166 - \c network/maxmin-selective-update: \ref options_model_optim
1167 - \c network/model: \ref options_model_select
1168 - \c network/optim: \ref options_model_optim
1169 - \c network/sender_gap: \ref options_model_network_sendergap
1170 - \c network/TCP-gamma: \ref options_model_network_gamma
1171 - \c network/weight-S: \ref options_model_network_coefs
1173 - \c ns3/TcpModel: \ref options_pls
1174 - \c path: \ref options_generic_path
1175 - \c plugin: \ref options_generic_plugin
1177 - \c storage/max_file_descriptors: \ref option_model_storage_maxfd
1179 - \c surf/precision: \ref options_model_precision
1181 - \c <b>For collective operations of SMPI, please refer to Section \ref options_index_smpi_coll</b>
1182 - \c smpi/async-small-thresh: \ref options_model_network_asyncsend
1183 - \c smpi/bw-factor: \ref options_model_smpi_bw_factor
1184 - \c smpi/coll-selector: \ref options_model_smpi_collectives
1185 - \c smpi/comp-adjustment-file: \ref options_model_smpi_adj_file
1186 - \c smpi/cpu-threshold: \ref options_smpi_bench
1187 - \c smpi/display-timing: \ref options_smpi_timing
1188 - \c smpi/grow-injected-times: \ref options_model_smpi_test
1189 - \c smpi/host-speed: \ref options_smpi_bench
1190 - \c smpi/IB-penalty-factors: \ref options_model_network_coefs
1191 - \c smpi/iprobe: \ref options_model_smpi_iprobe
1192 - \c smpi/init: \ref options_model_smpi_init
1193 - \c smpi/lat-factor: \ref options_model_smpi_lat_factor
1194 - \c smpi/ois: \ref options_model_smpi_ois
1195 - \c smpi/or: \ref options_model_smpi_or
1196 - \c smpi/os: \ref options_model_smpi_os
1197 - \c smpi/papi-events: \ref options_smpi_papi_events
1198 - \c smpi/privatize-global-variables: \ref options_smpi_global
1199 - \c smpi/send-is-detached-thresh: \ref options_model_smpi_detached
1200 - \c smpi/simulate-computation: \ref options_smpi_bench
1201 - \c smpi/test: \ref options_model_smpi_test
1202 - \c smpi/use-shared-malloc: \ref options_model_smpi_use_shared_malloc
1203 - \c smpi/wtime: \ref options_model_smpi_wtime
1205 - \c <b>Tracing configuration options can be found in Section \ref tracing_tracing_options</b>.
1207 - \c storage/model: \ref options_storage_model
1208 - \c verbose-exit: \ref options_generic_exit
1210 - \c vm/model: \ref options_vm_model
1212 \subsection options_index_smpi_coll Index of SMPI collective algorithms options
1214 TODO: All available collective algorithms will be made available via the ``smpirun --help-coll`` command.