1 /*! \page FAQ Frequently Asked Questions
3 \htmlinclude .FAQ.doc.toc
5 \section faq_simgrid I'm new to SimGrid. I have some questions. Where should I start?
7 You are at the right place... Having a look to these
8 <a href="http://www.loria.fr/~quinson/blog/2010/06/28/Tutorial_at_HPCS/">the slides of the HPCS'10 tutorial</a>
9 (or to these <a href="http://graal.ens-lyon.fr/~alegrand/articles/slides_g5k_simul.pdf">ancient
10 slides</a>, or to these
11 <a href="http://graal.ens-lyon.fr/~alegrand/articles/Simgrid-Introduction.pdf">"obsolete" slides</a>)
12 may give you some insights on what SimGrid can help you to do and what
13 are its limitations. Then you definitely should read the \ref
14 MSG_examples. The \ref GRAS_tut can also help you.
16 If you are stuck at any point and if this FAQ cannot help you, please drop us a
17 mail to the user mailing list: <simgrid-user@lists.gforge.inria.fr>.
19 \subsection faq_interfaces What is the difference between MSG, SimDag, and GRAS? Do they serve the same purpose?
21 It depend on how you define "purpose", I guess ;)
23 They all allow you to build a prototype of application which you can run
24 within the simulator afterward. They all share the same simulation kernel,
25 which is the core of the SimGrid project. They differ by the way you express
28 With SimDag, you express your code as a collection of interdependent
29 parallel tasks. So, in this model, applications can be seen as a DAG of
30 tasks. This is the interface of choice for people wanting to port old
31 code designed for SimGrid v1 or v2 to the framework current version.
33 With both GRAS and MSG, your application is seen as a set of communicating
34 processes, exchanging data by the way of messages and performing computation
37 The difference between both is that MSG is somehow easier to use, but GRAS
38 is not limited to the simulator. Once you're done writing your GRAS code,
39 you can run your code both in the simulator and on a real platform. For this,
40 there are two implementations of the GRAS interface, one for simulation, and one
41 for real execution. So, you just have to relink your code to choose one of
44 \subsection faq_visualization Visualizing and analyzing the results
46 It is sometime convenient to "see" how the agents are behaving. If you
47 like colors, you can use <tt>tools/MSG_visualization/colorize.pl </tt>
48 as a filter to your MSG outputs. It works directly with INFO. Beware,
49 INFO() prints on stderr. Do not forget to redirect if you want to
50 filter (e.g. with bash):
52 ./msg_test small_platform.xml small_deployment.xml 2>&1 | ../../tools/MSG_visualization/colorize.pl
55 We also have a more graphical output. Have a look at section \ref options_tracing.
57 \subsection faq_C Argh! Do I really have to code in C?
59 Currently bindings on top of MSG are supported for Java, Ruby and Lua. You can find a few
60 documentation about them on the doc page. Note that bindings are released separately from the main dist
61 and so have their own version numbers.
63 Moreover If you use C++,
64 you should be able to use the SimGrid library as a standard C library
65 and everything should work fine (simply <i>link</i> against this
66 library; recompiling SimGrid with a C++ compiler won't work and it
67 wouldn't help if you could).
70 we do not feel a real demand for any other language. But if you think there is one,
73 \section faq_howto Feature related questions
75 \subsection faq_MIA "Could you please add (your favorite feature here) to SimGrid?"
77 Here is the deal. The whole SimGrid project (MSG, SURF, GRAS, ...) is
78 meant to be kept as simple and generic as possible. We cannot add
79 functions for everybody's needs when these functions can easily be
80 built from the ones already in the API. Most of the time, it is
81 possible and when it was not possible we always have upgraded the API
82 accordingly. When somebody asks us a question like "How to do that?
83 Is there a function in the API to simply do this?", we're always glad
84 to answer and help. However if we don't need this code for our own
85 need, there is no chance we're going to write it... it's your job! :)
86 The counterpart to our answers is that once you come up with a neat
87 implementation of this feature (task duplication, RPC, thread
88 synchronization, ...), you should send it to us and we will be glad to
89 add it to the distribution. Thus, other people will take advantage of
90 it (and we don't have to answer this question again and again ;).
92 You'll find in this section a few "Missing In Action" features. Many
93 people have asked about it and we have given hints on how to simply do
94 it with MSG. Feel free to contribute...
96 \subsection faq_MIA_MSG MSG features
98 \subsubsection faq_MIA_examples I want some more complex MSG examples!
100 Many people have come to ask me a more complex example and each time,
101 they have realized afterward that the basics were in the previous three
104 Of course they have often been needing more complex functions like
105 MSG_process_suspend(), MSG_process_resume() and
106 MSG_process_isSuspended() (to perform synchronization), or
107 MSG_task_Iprobe() and MSG_process_sleep() (to avoid blocking
108 receptions), or even MSG_process_create() (to design asynchronous
109 communications or computations). But the examples are sufficient to
112 We know. We should add some more examples, but not really some more
113 complex ones... We should add some examples that illustrate some other
114 functionalists (like how to simply encode asynchronous
115 communications, RPC, process migrations, thread synchronization, ...)
116 and we will do it when we will have a little bit more time. We have
117 tried to document the examples so that they are understandable. Tell
118 us if something is not clear and once again feel free to participate!
121 \subsubsection faq_MIA_taskdup Missing in action: MSG Task duplication/replication
123 There is no task duplication in MSG. When you create a task, you can
124 process it or send it somewhere else. As soon as a process has sent
125 this task, he doesn't have this task anymore. It's gone. The receiver
126 process has got the task. However, you could decide upon receiving to
127 create a "copy" of a task but you have to handle by yourself the
128 semantic associated to this "duplication".
130 As we already told, we prefer keeping the API as simple as
131 possible. This kind of feature is rather easy to implement by users
132 and the semantic you associate really depends on people. Having a
133 *generic* task duplication mechanism is not that trivial (in
134 particular because of the data field). That is why I would recommend
135 that you write it by yourself even if I can give you advice on how to
138 You have the following functions to get information about a task:
139 MSG_task_get_name(), MSG_task_get_compute_duration(),
140 MSG_task_get_remaining_computation(), MSG_task_get_data_size(),
141 and MSG_task_get_data().
143 You could use a dictionary (#xbt_dict_t) of dynars (#xbt_dynar_t). If
144 you still don't see how to do it, please come back to us...
146 \subsubsection faq_MIA_asynchronous I want to do asynchronous communications in MSG
148 In the past (version <= 3.4), there was no function to perform asynchronous communications.
149 It could easily be implemented by creating new process when needed though. Since version 3.5,
150 we have introduced the following functions:
159 We refer you to the description of these functions for more details on their usage as well
160 as to the example section on \ref MSG_ex_asynchronous_communications.
162 \subsubsection faq_MIA_thread_synchronization I need to synchronize my MSG processes
164 You obviously cannot use pthread_mutexes of pthread_conds since we handle every
165 scheduling related decision within SimGrid.
167 In the past (version <=3.3.4) you could do it by playing with
168 MSG_process_suspend() and MSG_process_resume() or with fake communications (using MSG_task_get(),
169 MSG_task_put() and MSG_task_Iprobe()).
171 Since version 3.4, you can use classical synchronization structures. See page \ref XBT_synchro or simply check in
172 include/xbt/synchro_core.h.
174 \subsubsection faq_MIA_host_load Where is the get_host_load function hidden in MSG?
176 There is no such thing because its semantic wouldn't be really
177 clear. Of course, it is something about the amount of host throughput,
178 but there is as many definition of "host load" as people asking for
179 this function. First, you have to remember that resource availability
180 may vary over time, which make any load notion harder to define.
182 It may be instantaneous value or an average one. Moreover it may be only the
183 power of the computer, or may take the background load into account, or may
184 even take the currently running tasks into account. In some SURF models,
185 communications have an influence on computational power. Should it be taken
188 First of all, it's near to impossible to predict the load beforehand in the
189 simulator since it depends on too much parameters (background load
190 variation, bandwidth sharing algorithmic complexity) some of them even being
191 not known beforehand (other task starting at the same time). So, getting
192 this information is really hard (just like in real life). It's not just that
193 we want MSG to be as painful as real life. But as it is in some way
194 realistic, we face some of the same problems as we would face in real life.
196 How would you do it for real? The most common option is to use something
197 like NWS that performs active probes. The best solution is probably to do
198 the same within MSG, as in next code snippet. It is very close from what you
199 would have to do out of the simulator, and thus gives you information that
200 you could also get in real settings to not hinder the realism of your
204 double get_host_load() {
205 m_task_t task = MSG_task_create("test", 0.001, 0, NULL);
206 double date = MSG_get_clock();
208 MSG_task_execute(task);
209 date = MSG_get_clock() - date;
210 MSG_task_destroy(task);
215 Of course, it may not match your personal definition of "host load". In this
216 case, please detail what you mean on the mailing list, and we will extend
217 this FAQ section to fit your taste if possible.
219 \subsubsection faq_MIA_communication_time How can I get the *real* communication time?
221 Communications are synchronous and thus if you simply get the time
222 before and after a communication, you'll only get the transmission
223 time and the time spent to really communicate (it will also take into
224 account the time spent waiting for the other party to be
225 ready). However, getting the *real* communication time is not really
226 hard either. The following solution is a good starting point.
231 m_task_t task = MSG_task_create("Task", task_comp_size, task_comm_size,
232 calloc(1,sizeof(double)));
233 *((double*) task->data) = MSG_get_clock();
234 MSG_task_put(task, slaves[i % slaves_count], PORT_22);
235 XBT_INFO("Send completed");
240 m_task_t task = NULL;
243 time1 = MSG_get_clock();
244 a = MSG_task_get(&(task), PORT_22);
245 time2 = MSG_get_clock();
246 if(time1<*((double *)task->data))
247 time1 = *((double *) task->data);
248 XBT_INFO("Communication time : \"%f\" ", time2-time1);
250 MSG_task_destroy(task);
255 \subsection faq_MIA_SimDag SimDag related questions
257 \subsubsection faq_SG_comm Implementing communication delays between tasks.
259 A classic question of SimDag newcomers is about how to express a
260 communication delay between tasks. The thing is that in SimDag, both
261 computation and communication are seen as tasks. So, if you want to
262 model a data dependency between two DAG tasks t1 and t2, you have to
263 create 3 SD_tasks: t1, t2 and c and add dependencies in the following
267 SD_task_dependency_add(NULL, NULL, t1, c);
268 SD_task_dependency_add(NULL, NULL, c, t2);
271 This way task t2 cannot start before the termination of communication c
272 which in turn cannot start before t1 ends.
274 When creating task c, you have to associate an amount of data (in bytes)
275 corresponding to what has to be sent by t1 to t2.
277 Finally to schedule the communication task c, you have to build a list
278 comprising the workstations on which t1 and t2 are scheduled (w1 and w2
279 for example) and build a communication matrix that should look like
282 \subsubsection faq_SG_DAG How to implement a distributed dynamic scheduler of DAGs.
284 Distributed is somehow "contagious". If you start making distributed
285 decisions, there is no way to handle DAGs directly anymore (unless I
286 am missing something). You have to encode your DAGs in term of
287 communicating process to make the whole scheduling process
288 distributed. Here is an example of how you could do that. Assume T1
289 has to be done before T2.
292 int your_agent(int argc, char *argv[] {
294 T1 = MSG_task_create(...);
295 T2 = MSG_task_create(...);
299 if(cond) MSG_task_execute(T1);
301 if((MSG_task_get_remaining_computation(T1)=0.0) && (you_re_in_a_good_mood))
304 /* do something else */
310 If you decide that the distributed part is not that much important and that
311 DAG is really the level of abstraction you want to work with, then you should
312 give a try to \ref SD_API.
314 \subsection faq_MIA_generic Generic features
316 \subsubsection faq_more_processes Increasing the amount of simulated processes
318 Here are a few tricks you can apply if you want to increase the amount
319 of processes in your simulations.
321 - <b>A few thousands of simulated processes</b> (soft tricks)\n
322 SimGrid can use either pthreads library or the UNIX98 contexts. On
323 most systems, the number of pthreads is limited and then your
324 simulation may be limited for a stupid reason. This is especially
325 true with the current linux pthreads, and I cannot get more than
326 2000 simulated processes with pthreads on my box. The UNIX98
327 contexts allow me to raise the limit to 25,000 simulated processes
329 The <tt>--with-context</tt> option of the <tt>./configure</tt>
330 script allows you to choose between UNIX98 contexts
331 (<tt>--with-context=ucontext</tt>) and the pthread version
332 (<tt>--with-context=pthread</tt>). The default value is ucontext
333 when the script detect a working UNIX98 context implementation. On
334 Windows boxes, the provided value is discarded and an adapted
335 version is picked up.\n\n
336 We experienced some issues with contexts on some rare systems
337 (solaris 8 and lower or old alpha linuxes comes to mind). The main
338 problem is that the configure script detect the contexts as being
339 functional when it's not true. If you happen to use such a system,
340 switch manually to the pthread version, and provide us with a good
341 patch for the configure script so that it is done automatically ;)
343 - <b>Hundred thousands of simulated processes</b> (hard-core tricks)\n
344 As explained above, SimGrid can use UNIX98 contexts to represent
345 and handle the simulated processes. Thanks to this, the main
346 limitation to the number of simulated processes becomes the
347 available memory.\n\n
348 Here are some tricks I had to use in order to run a token ring
349 between 25,000 processes on my laptop (1Gb memory, 1.5Gb swap).\n
350 - First of all, make sure your code runs for a few hundreds
351 processes before trying to push the limit. Make sure it's
352 valgrind-clean, i.e. that valgrind does not report neither memory
353 error nor memory leaks. Indeed, numerous simulated processes
354 result in *fat* simulation hindering debugging.
355 - It was really boring to write 25,000 entries in the deployment
356 file, so I wrote a little script
357 <tt>examples/gras/mutual_exclusion/simple_token/make_deployment.pl</tt>, which you may
358 want to adapt to your case. You could also think about hijacking
359 the SURFXML parser (have look at \ref faq_flexml_bypassing).
360 - The deployment file became quite big, so I had to do what is in
361 the FAQ entry \ref faq_flexml_limit
362 - Each UNIX98 context has its own stack entry. As debugging this is
363 quite hairy, the default value is a bit overestimated so that
364 user doesn't get into trouble about this. You want to tune this
365 size to increase the number of processes. This is the
366 <tt>STACK_SIZE</tt> define in
367 <tt>src/xbt/xbt_context_sysv.c</tt>, which is 128kb by default.
368 Reduce this as much as you can, but be warned that if this value
369 is too low, you'll get a segfault. The token ring example, which
370 is quite simple, runs with 40kb stacks.
371 - You may tweak the logs to reduce the stack size further. When
372 logging something, we try to build the string to display in a
373 char array on the stack. The size of this array is constant (and
374 equal to XBT_LOG_BUFF_SIZE, defined in include/xbt/log/h). If the
375 string is too large to fit this buffer, we move to a dynamically
376 sized buffer. In which case, we have to traverse one time the log
377 event arguments to compute the size we need for the buffer,
378 malloc it, and traverse the argument list again to do the actual
380 The idea here is to move XBT_LOG_BUFF_SIZE to 1, forcing the logs
381 to use a dynamic array each time. This allows us to lower further
382 the stack size at the price of some performance loss...\n
383 This allowed me to run the reduce the stack size to ... 4k. Ie,
384 on my 1Gb laptop, I can run more than 250,000 processes!
386 \subsubsection faq_MIA_batch_scheduler Is there a native support for batch schedulers in SimGrid?
388 No, there is no native support for batch schedulers and none is
389 planned because this is a very specific need (and doing it in a
390 generic way is thus very hard). However some people have implemented
391 their own batch schedulers. Vincent Garonne wrote one during his PhD
392 and put his code in the contrib directory of our SVN so that other can
393 keep working on it. You may find inspiring ideas in it.
395 \subsubsection faq_MIA_checkpointing I need a checkpointing thing
397 Actually, it depends on whether you want to checkpoint the simulation, or to
398 simulate checkpoints.
400 The first one could help if your simulation is a long standing process you
401 want to keep running even on hardware issues. It could also help to
402 <i>rewind</i> the simulation by jumping sometimes on an old checkpoint to
403 cancel recent calculations.\n
404 Unfortunately, such thing will probably never exist in SG. One would have to
405 duplicate all data structures because doing a rewind at the simulator level
406 is very very hard (not talking about the malloc free operations that might
407 have been done in between). Instead, you may be interested in the Libckpt
408 library (http://www.cs.utk.edu/~plank/plank/www/libckpt.html). This is the
409 checkpointing solution used in the condor project, for example. It makes it
410 easy to create checkpoints (at the OS level, creating something like core
411 files), and rerunning them on need.
413 If you want to simulate checkpoints instead, it means that you want the
414 state of an executing task (in particular, the progress made towards
415 completion) to be saved somewhere. So if a host (and the task executing on
416 it) fails (cf. #MSG_HOST_FAILURE), then the task can be restarted
417 from the last checkpoint.\n
419 Actually, such a thing does not exist in SimGrid either, but it's just
420 because we don't think it is fundamental and it may be done in the user code
421 at relatively low cost. You could for example use a watcher that
422 periodically get the remaining amount of things to do (using
423 MSG_task_get_remaining_computation()), or fragment the task in smaller
426 \subsection faq_platform Platform building and Dynamic resources
428 \subsubsection faq_platform_example Where can I find SimGrid platform files?
430 There are several little examples in the archive, in the examples/msg
431 directory. From time to time, we are asked for other files, but we
432 don't have much at hand right now.
434 You should refer to the Platform Description Archive
435 (http://pda.gforge.inria.fr) project to see the other platform file we
436 have available, as well as the Simulacrum simulator, meant to generate
437 SimGrid platforms using all classical generation algorithms.
439 \subsubsection faq_platform_alnem How can I automatically map an existing platform?
441 We are working on a project called ALNeM (Application-Level Network
442 Mapper) which goal is to automatically discover the topology of an
443 existing network. Its output will be a platform description file
444 following the SimGrid syntax, so everybody will get the ability to map
445 their own lab network (and contribute them to the catalog project).
446 This tool is not ready yet, but it move quite fast forward. Just stay
449 \subsubsection faq_platform_synthetic Generating synthetic but realistic platforms
451 The third possibility to get a platform file (after manual or
452 automatic mapping of real platforms) is to generate synthetic
453 platforms. Getting a realistic result is not a trivial task, and
454 moreover, nobody is really able to define what "realistic" means when
455 speaking of topology files. You can find some more thoughts on this
457 <a href="http://graal.ens-lyon.fr/~alegrand/articles/Simgrid-Introduction.pdf">slides</a>.
459 If you are looking for an actual tool, there we have a little tool to
460 annotate Tiers-generated topologies. This perl-script is in
461 <tt>tools/platform_generation/</tt> directory of the SVN. Dinda et Al.
462 released a very comparable tool, and called it GridG.
464 \subsubsection faq_SURF_multicore Modeling multi-core resources
466 Since version 3.6 of simgrid we can specify the core number of a resource.
467 To use this feature use tag 'host' with 'core' attribute.
469 <?xml version='1.0'?>
470 <!DOCTYPE platform SYSTEM "http://simgrid.gforge.inria.fr/simgrid.dtd">
471 <platform version="3">
472 <AS id="AS0" routing="Full">
473 <host id="Tremblay" power="98095000" core="6"/>
478 The specified computing power will be available to up to 6 sequential
479 tasks without sharing. If more tasks are placed on this host, the
480 resource will be shared accordingly. For example, if you schedule 12
481 tasks on the host, each will get half of the computing power. Please
482 note that although sound, this model were never scientifically
483 assessed. Please keep this fact in mind when using it.
485 \subsubsection faq_SURF_dynamic Modeling dynamic resource availability
487 A nice feature of SimGrid is that it enables you to seamlessly have
488 resources whose availability change over time. When you build a
489 platform, you generally declare hosts like that:
492 <host id="host A" power="100.00"/>
495 If you want the availability of "host A" to change over time, the only
496 thing you have to do is change this definition like that:
499 <host id="host A" power="100.00" availability_file="trace_A.txt" state_file="trace_A_failure.txt"/>
502 For hosts, availability files are expressed in fraction of available
503 power. Let's have a look at what "trace_A.txt" may look like:
512 At time 0, our host will deliver 100 flop/s. At time 11.0, it will
513 deliver only 50 flop/s until time 20.0 where it will start
514 delivering 90 flop/s. Last at time 21.0 (20.0 plus the periodicity
515 1.0), we'll be back to the beginning and it will deliver 100 flop/s.
517 Now let's look at the state file:
524 A negative value means "off" while a positive one means "on". At time
525 1.0, the host is on. At time 1.0, it is turned off and at time 2.0, it
526 is turned on again until time 12 (2.0 plus the periodicity 10.0). It
527 will be turned on again at time 13.0 until time 23.0, and so on.
529 Now, let's look how the same kind of thing can be done for network
530 links. A usual declaration looks like:
533 <link id="LinkA" bandwidth="10.0" latency="0.2"/>
536 You have at your disposal the following options: bandwidth_file,
537 latency_file and state_file. The only difference with hosts is that
538 bandwidth_file and latency_file do not express fraction of available
539 power but are expressed directly in bytes per seconds and seconds.
541 \subsubsection faq_platform_multipath How to express multipath routing in platform files?
543 It is unfortunately impossible to express the fact that there is more
544 than one routing path between two given hosts. Let's consider the
545 following platform file:
548 <route src="A" dst="B">
551 <route src="B" dst="C">
554 <route src="A" dst="C">
559 Although it is perfectly valid, it does not mean that data traveling
560 from A to C can either go directly (using link 3) or through B (using
561 links 1 and 2). It simply means that the routing on the graph is not
562 trivial, and that data do not following the shortest path in number of
563 hops on this graph. Another way to say it is that there is no implicit
564 in these routing descriptions. The system will only use the routes you
565 declare (such as <route src="A" dst="C"><link_ctn
566 id="3"/></route>), without trying to build new routes by aggregating
569 You are also free to declare platform where the routing is not
570 symmetric. For example, add the following to the previous file:
573 <route src="C" dst="A">
579 This makes sure that data from C to A go through B where data from A
580 to C go directly. Don't worry about realism of such settings since
581 we've seen ways more weird situation in real settings (in fact, that's
582 the realism of very regular platforms which is questionable, but
583 that's another story).
585 \subsubsection faq_flexml_bypassing Bypassing the XML parser with your own C functions
587 So you want to bypass the XML files parser, uh? Maybe doing some parameter
588 sweep experiments on your simulations or so? This is possible, and
589 it's not even really difficult (well. Such a brutal idea could be
590 harder to implement). Here is how it goes.
592 For this, you have to first remember that the XML parsing in SimGrid is done
593 using a tool called FleXML. Given a DTD, this gives a flex-based parser. If
594 you want to bypass the parser, you need to provide some code mimicking what
595 it does and replacing it in its interactions with the SURF code. So, let's
596 have a look at these interactions.
598 FleXML parser are close to classical SAX parsers. It means that a
599 well-formed SimGrid platform XML file might result in the following
602 - start "platform_description" with attribute version="2"
603 - start "host" with attributes id="host1" power="1.0"
605 - start "host" with attributes id="host2" power="2.0"
607 - start "link" with ...
609 - start "route" with ...
610 - start "link_ctn" with ...
613 - end "platform_description"
615 The communication from the parser to the SURF code uses two means:
616 Attributes get copied into some global variables, and a surf-provided
617 function gets called by the parser for each event. For example, the event
618 - start "host" with attributes id="host1" power="1.0"
620 let the parser do something roughly equivalent to:
622 strcpy(A_host_id,"host1");
627 In SURF, we attach callbacks to the different events by initializing the
628 pointer functions to some the right surf functions. Since there can be
629 more than one callback attached to the same event (if more than one
630 model is in use, for example), they are stored in a dynar. Example in
631 workstation_ptask_L07.c:
633 /* Adding callback functions */
634 surf_parse_reset_parser();
635 surfxml_add_callback(STag_surfxml_host_cb_list, &parse_cpu_init);
636 surfxml_add_callback(STag_surfxml_prop_cb_list, &parse_properties);
637 surfxml_add_callback(STag_surfxml_link_cb_list, &parse_link_init);
638 surfxml_add_callback(STag_surfxml_route_cb_list, &parse_route_set_endpoints);
639 surfxml_add_callback(ETag_surfxml_link_c_ctn_cb_list, &parse_route_elem);
640 surfxml_add_callback(ETag_surfxml_route_cb_list, &parse_route_set_route);
643 surf_parse_open(file);
644 xbt_assert(!surf_parse(), "Parse error in %s", file);
648 So, to bypass the FleXML parser, you need to write your own version of the
649 surf_parse function, which should do the following:
650 - Fill the A_<tag>_<attribute> variables with the wanted values
651 - Call the corresponding STag_<tag>_fun function to simulate tag start
652 - Call the corresponding ETag_<tag>_fun function to simulate tag end
653 - (do the same for the next set of values, and loop)
655 Then, tell SimGrid that you want to use your own "parser" instead of the stock one:
657 surf_parse = surf_parse_bypass_environment;
658 MSG_create_environment(NULL);
659 surf_parse = surf_parse_bypass_application;
660 MSG_launch_application(NULL);
663 A set of macros are provided at the end of
664 include/surf/surfxml_parse.h to ease the writing of the bypass
665 functions. An example of this trick is distributed in the file
666 examples/msg/masterslave/masterslave_bypass.c
668 \section faq_troubleshooting Troubleshooting
670 \subsection faq_trouble_lib_compil SimGrid compilation and installation problems
672 \subsubsection faq_trouble_lib_config cmake fails!
674 We know only one reason for the configure to fail:
676 - <b>You are using a broken build environment</b>\n
677 If symptom is that the configury magic complains about gcc not being able to build
678 executables, you are probably missing the libc6-dev package. Damn Ubuntu.
680 If you experience other kind of issue, please get in touch with us. We are
681 always interested in improving our portability to new systems.
683 \subsubsection faq_trouble_distcheck Dude! "ctest" fails on my machine!
685 Don't assume we never run this target, because we do. Check
686 http://cdash.inria.fr/CDash/index.php?project=Simgrid (click on
687 previous if there is no result for today: results are produced only by
688 11am, French time) and
689 https://buildd.debian.org/status/logs.php?pkg=simgrid if you don't believe us.
691 If it's failing on your machine in a way not experienced by the
692 autobuilders above, please drop us a mail on the mailing list so that
693 we can check it out. Make sure to read \ref faq_bugrepport before you
696 \subsection faq_trouble_compil User code compilation problems
698 \subsubsection faq_trouble_err_logcat "gcc: _simgrid_this_log_category_does_not_exist__??? undeclared (first use in this function)"
700 This is because you are using the log mecanism, but you didn't created
701 any default category in this file. You should refer to \ref XBT_log
702 for all the details, but you simply forgot to call one of
703 XBT_LOG_NEW_DEFAULT_CATEGORY() or XBT_LOG_NEW_DEFAULT_SUBCATEGORY().
705 \subsubsection faq_trouble_pthreadstatic "gcc: undefined reference to pthread_key_create"
707 This indicates that one of the library SimGrid depends on (libpthread
708 here) was missing on the linking command line. Dependencies of
709 libsimgrid are expressed directly in the dynamic library, so it's
710 quite impossible that you see this message when doing dynamic linking.
712 If you compile your code statically (and if you use a pthread version
713 of SimGrid -- see \ref faq_more_processes), you must absolutely
714 specify <tt>-lpthread</tt> on the linker command line. As usual, this should
715 come after <tt>-lsimgrid</tt> on this command line.
717 \subsubsection faq_trouble_lib_msg_deprecated "gcc: undefined reference to MSG_*"
719 Since version 3.7 all the #m_channel_t mecanism is deprecated. So functions
720 about this mecanism may get removed in future releases.
724 \li XBT_PUBLIC(int) MSG_get_host_number(void);
726 \li XBT_PUBLIC(m_host_t *) MSG_get_host_table(void);
728 \li XBT_PUBLIC(MSG_error_t) MSG_get_errno(void);
730 \li XBT_PUBLIC(MSG_error_t) MSG_task_get(m_task_t * task, m_channel_t channel);
732 \li XBT_PUBLIC(MSG_error_t) MSG_task_get_with_timeout(m_task_t * task, m_channel_t channel, double max_duration);
734 \li XBT_PUBLIC(MSG_error_t) MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
736 \li XBT_PUBLIC(MSG_error_t) MSG_task_get_ext(m_task_t * task, int channel, double max_duration, m_host_t host);
738 \li XBT_PUBLIC(MSG_error_t) MSG_task_put(m_task_t task, m_host_t dest, m_channel_t channel);
740 \li XBT_PUBLIC(MSG_error_t) MSG_task_put_bounded(m_task_t task, m_host_t dest, m_channel_t channel, double max_rate);
742 \li XBT_PUBLIC(MSG_error_t) MSG_task_put_with_timeout(m_task_t task, m_host_t dest, m_channel_t channel, double max_duration);
744 \li XBT_PUBLIC(int) MSG_task_Iprobe(m_channel_t channel);
746 \li XBT_PUBLIC(int) MSG_task_probe_from(m_channel_t channel);
748 \li XBT_PUBLIC(int) MSG_task_probe_from_host(int channel, m_host_t host);
750 \li XBT_PUBLIC(MSG_error_t) MSG_set_channel_number(int number);
752 \li XBT_PUBLIC(int) MSG_get_channel_number(void);
754 If you want them you have to compile Simgrid v3.7 with option "-Denable_msg_deprecated=ON".
755 Using them should print warning to inform what new function you have to use.
757 \subsection faq_trouble_errors Runtime error messages
759 \subsubsection faq_flexml_limit "surf_parse_lex: Assertion `next limit' failed."
761 This is because your platform file is too big for the parser.
763 Actually, the message comes directly from FleXML, the technology on top of
764 which the parser is built. FleXML has the bad idea of fetching the whole
765 document in memory before parsing it. And moreover, the memory buffer size
766 must be determined at compilation time.
768 We use a value which seems big enough for our need without bloating the
769 simulators footprints. But of course your mileage may vary. In this case,
770 just edit src/surf/surfxml.l modify the definition of
771 FLEXML_BUFFERSTACKSIZE. E.g.
774 #define FLEXML_BUFFERSTACKSIZE 1000000000
777 Then recompile and everything should be fine, provided that your version of
778 Flex is recent enough (>= 2.5.31). If not the compilation process should
781 A while ago, we worked on FleXML to reduce a bit its memory consumption, but
782 these issues remain. There is two things we should do:
784 - use a dynamic buffer instead of a static one so that the only limit
785 becomes your memory, not a stupid constant fixed at compilation time
786 (maybe not so difficult).
787 - change the parser so that it does not need to get the whole file in
788 memory before parsing
789 (seems quite difficult, but I'm a complete newbe wrt flex stuff).
791 These are changes to FleXML itself, not SimGrid. But since we kinda hijacked
792 the development of FleXML, I can grant you that any patches would be really
793 welcome and quickly integrated.
795 <b>Update:</b> A new version of FleXML (1.7) was released. Most of the work
796 was done by William Dowling, who use it in his own work. The good point is
797 that it now use a dynamic buffer, and that the memory usage was greatly
798 improved. The downside is that William also changed some things internally,
799 and it breaks the hack we devised to bypass the parser, as explained in
800 \ref faq_flexml_bypassing. Indeed, this is not a classical usage of the
801 parser, and Will didn't imagine that we may have used (and even documented)
802 such a crude usage of FleXML. So, we now have to repair the bypassing
803 functionality to use the lastest FleXML version and fix the memory usage in
806 \subsubsection faq_trouble_gras_transport GRAS spits networking error messages
808 Gras, on real platforms, naturally use regular sockets to communicate. They
809 are deeply hidden in the gras abstraction, but when things go wrong, you may
810 get some weird error messages. Here are some example, with the probable
813 - <b>Transport endpoint is not connected</b>: several processes try to open
814 a server socket on the same port number of the same machine. This is
815 naturally bad and each process should pick its own port number for this.\n
816 Maybe, you just have some processes remaining from a previous experiment
818 Killing them may help, but again if you kill -KILL them, you'll have to
819 wait for a while: they didn't close there sockets properly and the system
820 needs a while to notice that this port is free again.
822 - <b>Socket closed by remote side</b>: if the remote process is not
823 supposed to close the socket at this point, it may be dead.
825 - <b>Connection reset by peer</b>: I found this on Internet about this
826 error. I think it's what's happening here, too:\n
827 <i>This basically means that a network error occurred while the client was
828 receiving data from the server. But what is really happening is that the
829 server actually accepts the connection, processes the request, and sends
830 a reply to the client. However, when the server closes the socket, the
831 client believes that the connection has been terminated abnormally
832 because the socket implementation sends a TCP reset segment telling the
833 client to throw away the data and report an error.\n
834 Sometimes, this problem is caused by not properly closing the
835 input/output streams and the socket connection. Make sure you close the
836 input/output streams and socket connection properly. If everything is
837 closed properly, however, and the problem persists, you can work around
838 it by adding a one-second sleep before closing the streams and the
839 socket. This technique, however, is not reliable and may not work on all
841 Since GRAS sockets are closed properly (repeat after me: there is no bug
842 in GRAS), it is either that you are closing your sockets on server side
843 before the client get a chance to read them (use gras_os_sleep() to delay
844 the server), or the server died awfully before the client got the data.
846 \subsubsection faq_trouble_errors_big_fat_warning I'm told that my XML files are too old.
848 The format of the XML platform description files is sometimes
849 improved. For example, we decided to change the units used in SimGrid
850 from MBytes, MFlops and seconds to Bytes, Flops and seconds to ease
851 people exchanging small messages. We also reworked the route
852 descriptions to allow more compact descriptions.
854 That is why the XML files are versionned using the 'version' attribute
855 of the root tag. Currently, it should read:
857 <platform version="2">
860 If your files are too old, you can use the simgrid_update_xml.pl
861 script which can be found in the tools directory of the archive.
863 \subsection faq_trouble_valgrind Valgrind-related and other debugger issues
865 If you don't, you really should use valgrind to debug your code, it's
868 \subsubsection faq_trouble_vg_longjmp longjmp madness in valgrind
870 This is when valgrind starts complaining about longjmp things, just like:
872 \verbatim ==21434== Conditional jump or move depends on uninitialised value(s)
873 ==21434== at 0x420DBE5: longjmp (longjmp.c:33)
875 ==21434== Use of uninitialised value of size 4
876 ==21434== at 0x420DC3A: __longjmp (__longjmp.S:48)
879 This is the sign that you didn't used the exception mecanism well. Most
880 probably, you have a <tt>return;</tt> somewhere within a <tt>TRY{}</tt>
881 block. This is <b>evil</b>, and you must not do this. Did you read the section
884 \subsubsection faq_trouble_vg_libc Valgrind spits tons of errors about backtraces!
886 It may happen that valgrind, the memory debugger beloved by any decent C
887 programmer, spits tons of warnings like the following :
888 \verbatim ==8414== Conditional jump or move depends on uninitialised value(s)
889 ==8414== at 0x400882D: (within /lib/ld-2.3.6.so)
890 ==8414== by 0x414EDE9: (within /lib/tls/i686/cmov/libc-2.3.6.so)
891 ==8414== by 0x400B105: (within /lib/ld-2.3.6.so)
892 ==8414== by 0x414F937: _dl_open (in /lib/tls/i686/cmov/libc-2.3.6.so)
893 ==8414== by 0x4150F4C: (within /lib/tls/i686/cmov/libc-2.3.6.so)
894 ==8414== by 0x400B105: (within /lib/ld-2.3.6.so)
895 ==8414== by 0x415102D: __libc_dlopen_mode (in /lib/tls/i686/cmov/libc-2.3.6.so)
896 ==8414== by 0x412D6B9: backtrace (in /lib/tls/i686/cmov/libc-2.3.6.so)
897 ==8414== by 0x8076446: xbt_dictelm_get_ext (dict_elm.c:714)
898 ==8414== by 0x80764C1: xbt_dictelm_get (dict_elm.c:732)
899 ==8414== by 0x8079010: xbt_cfg_register (config.c:208)
900 ==8414== by 0x806821B: MSG_config (msg_config.c:42)
903 This problem is somewhere in the libc when using the backtraces and there is
904 very few things we can do ourselves to fix it. Instead, here is how to tell
905 valgrind to ignore the error. Add the following to your ~/.valgrind.supp (or
906 create this file on need). Make sure to change the obj line according to
907 your personnal mileage (change 2.3.6 to the actual version you are using,
908 which you can retrieve with a simple "ls /lib/ld*.so").
911 name: Backtrace madness
918 fun:__libc_dlopen_mode
921 Then, you have to specify valgrind to use this suppression file by passing
922 the <tt>--suppressions=$HOME/.valgrind.supp</tt> option on the command line.
923 You can also add the following to your ~/.bashrc so that it gets passed
924 automatically. Actually, it passes a bit more options to valgrind, and this
925 happen to be my personnal settings. Check the valgrind documentation for
928 \verbatim export VALGRIND_OPTS="--leak-check=yes --leak-resolution=high --num-callers=40 --tool=memcheck --suppressions=$HOME/.valgrind.supp" \endverbatim
930 \subsubsection faq_trouble_backtraces Truncated backtraces
932 When debugging SimGrid, it's easier to pass the
933 --disable-compiler-optimization flag to the configure if valgrind or
934 gdb get fooled by the optimization done by the compiler. But you
935 should remove these flag when everything works before going in
936 production (before launching your 1252135 experiments), or everything
937 will run only one half of the true SimGrid potential.
939 \subsection faq_deadlock There is a deadlock in my code!!!
941 Unfortunately, we cannot debug every code written in SimGrid. We
942 furthermore believe that the framework provides ways enough
943 information to debug such informations yourself. If the textual output
944 is not enough, Make sure to check the \ref faq_visualization FAQ entry to see
945 how to get a graphical one.
947 Now, if you come up with a really simple example that deadlocks and
948 you're absolutely convinced that it should not, you can ask on the
949 list. Just be aware that you'll be severely punished if the mistake is
950 on your side... We have plenty of FAQ entries to redact and new
951 features to implement for the impenitents! ;)
953 \subsection faq_surf_network_latency I get weird timings when I play with the latencies.
955 OK, first of all, remember that units should be Bytes, Flops and
956 Seconds. If you don't use such units, some SimGrid constants (e.g. the
957 SG_TCP_CTE_GAMMA constant used in most network models) won't have the
958 right unit and you'll end up with weird results.
960 Here is what happens with a single transfer of size L on a link
961 (bw,lat) when nothing else happens.
964 0-----lat--------------------------------------------------t
965 |-----|**** real_bw =min(bw,SG_TCP_CTE_GAMMA/(2*lat)) *****|
968 In more complex situations, this min is the solution of a complex
969 max-min linear system. Have a look
970 <a href="http://lists.gforge.inria.fr/pipermail/simgrid-devel/2006-April/thread.html">here</a>
971 and read the two threads "Bug in SURF?" and "Surf bug not
972 fixed?". You'll have a few other examples of such computations. You
973 can also read "A Network Model for Simulation of Grid Application" by
974 Henri Casanova and Loris Marchal to have all the details. The fact
975 that the real_bw is smaller than bw is easy to understand. The fact
976 that real_bw is smaller than SG_TCP_CTE_GAMMA/(2*lat) is due to the
977 window-based congestion mechanism of TCP. With TCP, you can't exploit
978 your huge network capacity if you don't have a good round-trip-time
979 because of the acks...
981 Anyway, what you get is t=lat + L/min(bw,SG_TCP_CTE_GAMMA/(2*lat)).
983 * if I you set (bw,lat)=(100 000 000, 0.00001), you get t = 1.00001 (you fully
985 * if I you set (bw,lat)=(100 000 000, 0.0001), you get t = 1.0001 (you're on the
987 * if I you set (bw,lat)=(100 000 000, 0.001), you get t = 10.001 (ouch!)
989 This bound on the effective bandwidth of a flow is not the only thing
990 that may make your result be unexpected. For example, two flows
991 competing on a saturated link receive an amount of bandwidth inversely
992 proportional to their round trip time.
994 \subsection faq_bugrepport So I've found a bug in SimGrid. How to report it?
996 We do our best to make sure to hammer away any bugs of SimGrid, but this is
997 still an academic project so please be patient if/when you find bugs in it.
998 If you do, the best solution is to drop an email either on the simgrid-user
999 or the simgrid-devel mailing list and explain us about the issue. You can
1000 also decide to open a formal bug report using the
1001 <a href="https://gforge.inria.fr/tracker/?atid=165&group_id=12&func=browse">relevant
1002 interface</a>. You need to login on the server to get the ability to submit
1005 We will do our best to solve any problem repported, but you need to help us
1006 finding the issue. Just telling "it segfault" isn't enough. Telling "It
1007 segfaults when running the attached simulator" doesn't really help either.
1008 You may find the following article interesting to see how to repport
1009 informative bug repports:
1010 http://www.chiark.greenend.org.uk/~sgtatham/bugs.html (it is not SimGrid
1011 specific at all, but it's full of good advices).
1013 \author Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1017 ******************************************************************
1018 * OLD CRUFT NOT USED ANYMORE *
1019 ******************************************************************
1022 subsection faq_crosscompile Cross-compiling a Windows DLL of SimGrid from linux
1024 At the moment, we do not distribute Windows pre-compiled version of SimGrid
1025 because the support for this platform is still experimental. We know that
1026 some parts of the GRAS environment do not work, and we think that the others
1027 environments (MSG and SD) have good chances to work, but we didn't test
1028 ourselves. This section explains how we generate the SimGrid DLL so that you
1029 can build it for yourself. First of all, you need to have a version more
1030 recent than 3.1 (ie, a SVN version as time of writting).
1032 In order to cross-compile the package to windows from linux, you need to
1033 install mingw32 (minimalist gnu win32). On Debian, you can do so by
1034 installing the packages mingw32 (compiler), mingw32-binutils (linker and
1035 so), mingw32-runtime.
1037 You can use the VPATH support of configure to compile at the same time for
1038 linux and windows without dupplicating the source nor cleaning the tree
1039 between each. Just run bootstrap (if you use the SVN) to run the autotools.
1040 Then, create a linux and a win directories. Then, type:
1041 \verbatim cd linux; ../configure --srcdir=.. <usual configure flags>; make; cd ..
1042 cd win; ../configure --srcdir=.. --host=i586-mingw32msvc <flags>; make; cd ..
1044 The trick to VPATH builds is to call configure from another directory,
1045 passing it an extra --srcdir argument to tell it where all the sources are.
1046 It will understand you want to use VPATH. Then, the trick to cross-compile
1047 is simply to add a --host argument specifying the target you want to build
1048 for. The i586-mingw32msvc string is what you have to pass to use the mingw32
1049 environment as distributed in Debian.
1051 After that, you can run all make targets from both directories, and test
1052 easily that what you change for one arch does not break the other one.
1054 It is possible that this VPATH build thing breaks from time to time in the
1055 SVN since it's quite fragile, but it's granted to work in any released
1056 version. If you experience problems, drop us a mail.
1058 Another possible source of issue is that at the moment, building the
1059 examples request to use the gras_stub_generator tool, which is a compiled
1060 program, not a script. In cross-compilation, you need to cross-execute with
1061 wine for example, which is not really pleasant. We are working on this, but
1062 in the meanwhile, simply don't build the examples in cross-compilation
1063 (<tt>cd src</tt> before running make).
1065 Program (cross-)compiled with mingw32 do request an extra DLL at run-time to be
1066 usable. For example, if you want to test your build with wine, you should do
1067 the following to put this library where wine looks for DLLs.
1069 cp /usr/share/doc/mingw32-runtime/mingwm10.dll.gz ~/.wine/c/windows/system/
1070 gunzip ~/.wine/c/windows/system/mingwm10.dll.gz
1073 The DLL is built in src/.libs, and installed in the <i>prefix</i>/bin directory
1074 when you run make install.
1076 If you want to use it in a native project on windows, you need to use
1077 simgrid.dll and mingwm10.dll. For each DLL, you need to build .def file
1078 under linux (listing the defined symbols), and convert it into a .lib file
1079 under windows (specifying this in a way that windows compilers like). To
1080 generate the def files, run (under linux):
1081 \verbatim echo "LIBRARY libsimgrid-0.dll" > simgrid.def
1082 echo EXPORTS >> simgrid.def
1083 nm libsimgrid-0.dll | grep ' T _' | sed 's/.* T _//' >> simgrid.def
1084 nm libsimgrid-0.dll | grep ' D _' | sed 's/.* D _//' | sed 's/$/ DATA/' >> simgrid.def
1086 echo "LIBRARY mingwm10.dll" > mingwm10.def
1087 echo EXPORTS >> mingwm10.def
1088 nm mingwm10.dll | grep ' T _' | sed 's/.* T _//' >> mingwm10.def
1089 nm mingwm10.dll | grep ' D _' | sed 's/.* D _//' | sed 's/$/ DATA/' >> mingwm10.def
1092 To create the import .lib files, use the <tt>lib</tt> windows tool (from
1093 MSVC) the following way to produce simgrid.lib and mingwm10.lib
1094 \verbatim lib /def:simgrid.def
1095 lib /def:mingwm10.def
1098 If you happen to use Borland C Builder, the right command line is the
1099 following (note that you don't need any file.def to get this working).
1100 \verbatim implib simgrid.lib libsimgrid-0.dll
1101 implib mingwm10.lib mingwm10.dll
1104 Then, set the following parameters in Visual C++ 2005:
1105 Linker -> Input -> Additional dependencies = simgrid.lib mingwm10.lib
1107 Just in case you wonder how to generate a DLL from libtool in another
1108 project, we added -no-undefined to any lib*_la_LDFLAGS variables so that
1109 libtool accepts to generate a dynamic library under windows. Then, to make
1110 it true, we pass any dependencies (such as -lws2 under windows or -lpthread
1111 on need) on the linking line. Passing such deps is a good idea anyway so
1112 that they get noted in the library itself, avoiding the users to know about
1113 our dependencies and put them manually on their compilation line. Then we
1114 added the AC_LIBTOOL_WIN32_DLL macro just before AC_PROG_LIBTOOL in the
1115 configure.ac. It means that we exported any symbols which need to be.
1116 Nowadays, functions get automatically exported, so we don't need to load our
1117 header files with tons of __declspec(dllexport) cruft. We only need to do so
1118 for data, but there is no public data in SimGrid so we are good.