1 /*! \page FAQ Frequently Asked Questions
4 \section faq_simgrid I'm new to SimGrid. I have some questions. Where should I start?
6 You are at the right place... Having a look to these
7 <a href="http://www.loria.fr/~quinson/blog/2010/06/28/Tutorial_at_HPCS/">the slides of the HPCS'10 tutorial</a>
8 (or to these <a href="http://graal.ens-lyon.fr/~alegrand/articles/slides_g5k_simul.pdf">ancient
9 slides</a>, or to these
10 <a href="http://graal.ens-lyon.fr/~alegrand/articles/Simgrid-Introduction.pdf">"obsolete" slides</a>)
11 may give you some insights on what SimGrid can help you to do and what
12 are its limitations. Then you definitely should read the \ref
15 If you are stuck at any point and if this FAQ cannot help you, please drop us a
16 mail to the user mailing list: <simgrid-user@lists.gforge.inria.fr>.
18 \subsection faq_interfaces What is the difference between MSG and SimDag? Do they serve the same purpose?
20 It depend on how you define "purpose", I guess ;)
22 They all allow you to build a prototype of application which you can run
23 within the simulator afterward. They all share the same simulation kernel,
24 which is the core of the SimGrid project. They differ by the way you express
27 With SimDag, you express your code as a collection of interdependent
28 parallel tasks. So, in this model, applications can be seen as a DAG of
29 tasks. This is the interface of choice for people wanting to port old
30 code designed for SimGrid v1 or v2 to the framework current version.
32 With MSG, your application is seen as a set of communicating
33 processes, exchanging data by the way of messages and performing
34 computation on their own.
36 \subsection faq_visualization Visualizing and analyzing the results
38 It is sometime convenient to "see" how the agents are behaving. If you
39 like colors, you can use <tt>tools/MSG_visualization/colorize.pl </tt>
40 as a filter to your MSG outputs. It works directly with INFO. Beware,
41 INFO() prints on stderr. Do not forget to redirect if you want to
42 filter (e.g. with bash):
44 ./msg_test small_platform.xml small_deployment.xml 2>&1 | ../../tools/MSG_visualization/colorize.pl
47 We also have a more graphical output. Have a look at section \ref options_tracing.
49 \subsection faq_C Argh! Do I really have to code in C?
51 Currently bindings on top of MSG are supported for Java, Ruby and Lua. You can find a few
52 documentation about them on the doc page. Note that bindings are released separately from the main dist
53 and so have their own version numbers.
55 Moreover If you use C++,
56 you should be able to use the SimGrid library as a standard C library
57 and everything should work fine (simply <i>link</i> against this
58 library; recompiling SimGrid with a C++ compiler won't work and it
59 wouldn't help if you could).
62 we do not feel a real demand for any other language. But if you think there is one,
65 \section faq_howto Feature related questions
67 \subsection faq_MIA "Could you please add (your favorite feature here) to SimGrid?"
69 Here is the deal. The whole SimGrid project (MSG, SURF, ...) is
70 meant to be kept as simple and generic as possible. We cannot add
71 functions for everybody's needs when these functions can easily be
72 built from the ones already in the API. Most of the time, it is
73 possible and when it was not possible we always have upgraded the API
74 accordingly. When somebody asks us a question like "How to do that?
75 Is there a function in the API to simply do this?", we're always glad
76 to answer and help. However if we don't need this code for our own
77 need, there is no chance we're going to write it... it's your job! :)
78 The counterpart to our answers is that once you come up with a neat
79 implementation of this feature (task duplication, RPC, thread
80 synchronization, ...), you should send it to us and we will be glad to
81 add it to the distribution. Thus, other people will take advantage of
82 it (and we don't have to answer this question again and again ;).
84 You'll find in this section a few "Missing In Action" features. Many
85 people have asked about it and we have given hints on how to simply do
86 it with MSG. Feel free to contribute...
88 \subsection faq_MIA_MSG MSG features
90 \subsubsection faq_MIA_examples I want some more complex MSG examples!
92 Many people have come to ask me a more complex example and each time,
93 they have realized afterward that the basics were in the previous three
96 Of course they have often been needing more complex functions like
97 MSG_process_suspend(), MSG_process_resume() and
98 MSG_process_isSuspended() (to perform synchronization), or
99 MSG_task_Iprobe() and MSG_process_sleep() (to avoid blocking
100 receptions), or even MSG_process_create() (to design asynchronous
101 communications or computations). But the examples are sufficient to
104 We know. We should add some more examples, but not really some more
105 complex ones... We should add some examples that illustrate some other
106 functionalists (like how to simply encode asynchronous
107 communications, RPC, process migrations, thread synchronization, ...)
108 and we will do it when we will have a little bit more time. We have
109 tried to document the examples so that they are understandable. Tell
110 us if something is not clear and once again feel free to participate!
113 \subsubsection faq_MIA_taskdup Missing in action: MSG Task duplication/replication
115 There is no task duplication in MSG. When you create a task, you can
116 process it or send it somewhere else. As soon as a process has sent
117 this task, he doesn't have this task anymore. It's gone. The receiver
118 process has got the task. However, you could decide upon receiving to
119 create a "copy" of a task but you have to handle by yourself the
120 semantic associated to this "duplication".
122 As we already told, we prefer keeping the API as simple as
123 possible. This kind of feature is rather easy to implement by users
124 and the semantic you associate really depends on people. Having a
125 *generic* task duplication mechanism is not that trivial (in
126 particular because of the data field). That is why I would recommend
127 that you write it by yourself even if I can give you advice on how to
130 You have the following functions to get information about a task:
131 MSG_task_get_name(), MSG_task_get_compute_duration(),
132 MSG_task_get_remaining_computation(), MSG_task_get_data_size(),
133 and MSG_task_get_data().
135 You could use a dictionary (#xbt_dict_t) of dynars (#xbt_dynar_t). If
136 you still don't see how to do it, please come back to us...
138 \subsubsection faq_MIA_asynchronous I want to do asynchronous communications in MSG
140 In the past (version <= 3.4), there was no function to perform asynchronous communications.
141 It could easily be implemented by creating new process when needed though. Since version 3.5,
142 we have introduced the following functions:
151 We refer you to the description of these functions for more details on their usage as well
152 as to the example section on \ref MSG_ex_asynchronous_communications.
154 \subsubsection faq_MIA_thread_synchronization I need to synchronize my MSG processes
156 You obviously cannot use pthread_mutexes of pthread_conds since we handle every
157 scheduling related decision within SimGrid.
159 In the past (version <=3.3.4) you could do it by playing with
160 MSG_process_suspend() and MSG_process_resume() or with fake communications (using MSG_task_get(),
161 MSG_task_put() and MSG_task_Iprobe()).
163 Since version 3.4, you can use classical synchronization structures. See page \ref XBT_synchro or simply check in
164 include/xbt/synchro_core.h.
166 \subsubsection faq_MIA_host_load Where is the get_host_load function hidden in MSG?
168 There is no such thing because its semantic wouldn't be really
169 clear. Of course, it is something about the amount of host throughput,
170 but there is as many definition of "host load" as people asking for
171 this function. First, you have to remember that resource availability
172 may vary over time, which make any load notion harder to define.
174 It may be instantaneous value or an average one. Moreover it may be only the
175 power of the computer, or may take the background load into account, or may
176 even take the currently running tasks into account. In some SURF models,
177 communications have an influence on computational power. Should it be taken
180 First of all, it's near to impossible to predict the load beforehand in the
181 simulator since it depends on too much parameters (background load
182 variation, bandwidth sharing algorithmic complexity) some of them even being
183 not known beforehand (other task starting at the same time). So, getting
184 this information is really hard (just like in real life). It's not just that
185 we want MSG to be as painful as real life. But as it is in some way
186 realistic, we face some of the same problems as we would face in real life.
188 How would you do it for real? The most common option is to use something
189 like NWS that performs active probes. The best solution is probably to do
190 the same within MSG, as in next code snippet. It is very close from what you
191 would have to do out of the simulator, and thus gives you information that
192 you could also get in real settings to not hinder the realism of your
196 double get_host_load() {
197 m_task_t task = MSG_task_create("test", 0.001, 0, NULL);
198 double date = MSG_get_clock();
200 MSG_task_execute(task);
201 date = MSG_get_clock() - date;
202 MSG_task_destroy(task);
207 Of course, it may not match your personal definition of "host load". In this
208 case, please detail what you mean on the mailing list, and we will extend
209 this FAQ section to fit your taste if possible.
211 \subsubsection faq_MIA_communication_time How can I get the *real* communication time?
213 Communications are synchronous and thus if you simply get the time
214 before and after a communication, you'll only get the transmission
215 time and the time spent to really communicate (it will also take into
216 account the time spent waiting for the other party to be
217 ready). However, getting the *real* communication time is not really
218 hard either. The following solution is a good starting point.
223 m_task_t task = MSG_task_create("Task", task_comp_size, task_comm_size,
224 calloc(1,sizeof(double)));
225 *((double*) task->data) = MSG_get_clock();
226 MSG_task_put(task, slaves[i % slaves_count], PORT_22);
227 XBT_INFO("Send completed");
232 m_task_t task = NULL;
235 time1 = MSG_get_clock();
236 a = MSG_task_get(&(task), PORT_22);
237 time2 = MSG_get_clock();
238 if(time1<*((double *)task->data))
239 time1 = *((double *) task->data);
240 XBT_INFO("Communication time : \"%f\" ", time2-time1);
242 MSG_task_destroy(task);
247 \subsection faq_MIA_SimDag SimDag related questions
249 \subsubsection faq_SG_comm Implementing communication delays between tasks.
251 A classic question of SimDag newcomers is about how to express a
252 communication delay between tasks. The thing is that in SimDag, both
253 computation and communication are seen as tasks. So, if you want to
254 model a data dependency between two DAG tasks t1 and t2, you have to
255 create 3 SD_tasks: t1, t2 and c and add dependencies in the following
259 SD_task_dependency_add(NULL, NULL, t1, c);
260 SD_task_dependency_add(NULL, NULL, c, t2);
263 This way task t2 cannot start before the termination of communication c
264 which in turn cannot start before t1 ends.
266 When creating task c, you have to associate an amount of data (in bytes)
267 corresponding to what has to be sent by t1 to t2.
269 Finally to schedule the communication task c, you have to build a list
270 comprising the workstations on which t1 and t2 are scheduled (w1 and w2
271 for example) and build a communication matrix that should look like
274 \subsubsection faq_SG_DAG How to implement a distributed dynamic scheduler of DAGs.
276 Distributed is somehow "contagious". If you start making distributed
277 decisions, there is no way to handle DAGs directly anymore (unless I
278 am missing something). You have to encode your DAGs in term of
279 communicating process to make the whole scheduling process
280 distributed. Here is an example of how you could do that. Assume T1
281 has to be done before T2.
284 int your_agent(int argc, char *argv[] {
286 T1 = MSG_task_create(...);
287 T2 = MSG_task_create(...);
291 if(cond) MSG_task_execute(T1);
293 if((MSG_task_get_remaining_computation(T1)=0.0) && (you_re_in_a_good_mood))
296 /* do something else */
302 If you decide that the distributed part is not that much important and that
303 DAG is really the level of abstraction you want to work with, then you should
304 give a try to \ref SD_API.
306 \subsection faq_MIA_generic Generic features
308 \subsubsection faq_more_processes Increasing the amount of simulated processes
310 Here are a few tricks you can apply if you want to increase the amount
311 of processes in your simulations.
313 - <b>A few thousands of simulated processes</b> (soft tricks)\n
314 SimGrid can use either pthreads library or the UNIX98 contexts. On
315 most systems, the number of pthreads is limited and then your
316 simulation may be limited for a stupid reason. This is especially
317 true with the current linux pthreads, and I cannot get more than
318 2000 simulated processes with pthreads on my box. The UNIX98
319 contexts allow me to raise the limit to 25,000 simulated processes
321 The <tt>--with-context</tt> option of the <tt>./configure</tt>
322 script allows you to choose between UNIX98 contexts
323 (<tt>--with-context=ucontext</tt>) and the pthread version
324 (<tt>--with-context=pthread</tt>). The default value is ucontext
325 when the script detect a working UNIX98 context implementation. On
326 Windows boxes, the provided value is discarded and an adapted
327 version is picked up.\n\n
328 We experienced some issues with contexts on some rare systems
329 (solaris 8 and lower or old alpha linuxes comes to mind). The main
330 problem is that the configure script detect the contexts as being
331 functional when it's not true. If you happen to use such a system,
332 switch manually to the pthread version, and provide us with a good
333 patch for the configure script so that it is done automatically ;)
335 - <b>Hundred thousands of simulated processes</b> (hard-core tricks)\n
336 As explained above, SimGrid can use UNIX98 contexts to represent
337 and handle the simulated processes. Thanks to this, the main
338 limitation to the number of simulated processes becomes the
339 available memory.\n\n
340 Here are some tricks I had to use in order to run a token ring
341 between 25,000 processes on my laptop (1Gb memory, 1.5Gb swap).\n
342 - First of all, make sure your code runs for a few hundreds
343 processes before trying to push the limit. Make sure it's
344 valgrind-clean, i.e. that valgrind does not report neither memory
345 error nor memory leaks. Indeed, numerous simulated processes
346 result in *fat* simulation hindering debugging.
347 - It was really boring to write 25,000 entries in the deployment
348 file, so I wrote a little script
349 <tt>examples/gras/mutual_exclusion/simple_token/make_deployment.pl</tt>, which you may
350 want to adapt to your case. You could also think about hijacking
351 the SURFXML parser (have look at \ref pf_flexml_bypassing).
352 - The deployment file became quite big, so I had to do what is in
353 the FAQ entry \ref faq_flexml_limit
354 - Each UNIX98 context has its own stack entry. As debugging this is
355 quite hairy, the default value is a bit overestimated so that
356 user doesn't get into trouble about this. You want to tune this
357 size to increase the number of processes. This is the
358 <tt>STACK_SIZE</tt> define in
359 <tt>src/xbt/xbt_context_sysv.c</tt>, which is 128kb by default.
360 Reduce this as much as you can, but be warned that if this value
361 is too low, you'll get a segfault. The token ring example, which
362 is quite simple, runs with 40kb stacks.
363 - You may tweak the logs to reduce the stack size further. When
364 logging something, we try to build the string to display in a
365 char array on the stack. The size of this array is constant (and
366 equal to XBT_LOG_BUFF_SIZE, defined in include/xbt/log/h). If the
367 string is too large to fit this buffer, we move to a dynamically
368 sized buffer. In which case, we have to traverse one time the log
369 event arguments to compute the size we need for the buffer,
370 malloc it, and traverse the argument list again to do the actual
372 The idea here is to move XBT_LOG_BUFF_SIZE to 1, forcing the logs
373 to use a dynamic array each time. This allows us to lower further
374 the stack size at the price of some performance loss...\n
375 This allowed me to run the reduce the stack size to ... 4k. Ie,
376 on my 1Gb laptop, I can run more than 250,000 processes!
378 \subsubsection faq_MIA_batch_scheduler Is there a native support for batch schedulers in SimGrid?
380 No, there is no native support for batch schedulers and none is
381 planned because this is a very specific need (and doing it in a
382 generic way is thus very hard). However some people have implemented
383 their own batch schedulers. Vincent Garonne wrote one during his PhD
384 and put his code in the contrib directory of our SVN so that other can
385 keep working on it. You may find inspiring ideas in it.
387 \subsubsection faq_MIA_checkpointing I need a checkpointing thing
389 Actually, it depends on whether you want to checkpoint the simulation, or to
390 simulate checkpoints.
392 The first one could help if your simulation is a long standing process you
393 want to keep running even on hardware issues. It could also help to
394 <i>rewind</i> the simulation by jumping sometimes on an old checkpoint to
395 cancel recent calculations.\n
396 Unfortunately, such thing will probably never exist in SG. One would have to
397 duplicate all data structures because doing a rewind at the simulator level
398 is very very hard (not talking about the malloc free operations that might
399 have been done in between). Instead, you may be interested in the Libckpt
400 library (http://www.cs.utk.edu/~plank/plank/www/libckpt.html). This is the
401 checkpointing solution used in the condor project, for example. It makes it
402 easy to create checkpoints (at the OS level, creating something like core
403 files), and rerunning them on need.
405 If you want to simulate checkpoints instead, it means that you want the
406 state of an executing task (in particular, the progress made towards
407 completion) to be saved somewhere. So if a host (and the task executing on
408 it) fails (cf. #MSG_HOST_FAILURE), then the task can be restarted
409 from the last checkpoint.\n
411 Actually, such a thing does not exist in SimGrid either, but it's just
412 because we don't think it is fundamental and it may be done in the user code
413 at relatively low cost. You could for example use a watcher that
414 periodically get the remaining amount of things to do (using
415 MSG_task_get_remaining_computation()), or fragment the task in smaller
418 \subsection faq_platform Platform building and Dynamic resources
420 \subsubsection faq_platform_example Where can I find SimGrid platform files?
422 There are several little examples in the archive, in the examples/msg
423 directory. From time to time, we are asked for other files, but we
424 don't have much at hand right now.
426 You should refer to the Platform Description Archive
427 (http://pda.gforge.inria.fr) project to see the other platform file we
428 have available, as well as the Simulacrum simulator, meant to generate
429 SimGrid platforms using all classical generation algorithms.
431 \subsubsection faq_platform_alnem How can I automatically map an existing platform?
433 We are working on a project called ALNeM (Application-Level Network
434 Mapper) which goal is to automatically discover the topology of an
435 existing network. Its output will be a platform description file
436 following the SimGrid syntax, so everybody will get the ability to map
437 their own lab network (and contribute them to the catalog project).
438 This tool is not ready yet, but it move quite fast forward. Just stay
441 \subsubsection faq_platform_synthetic Generating synthetic but realistic platforms
443 The third possibility to get a platform file (after manual or
444 automatic mapping of real platforms) is to generate synthetic
445 platforms. Getting a realistic result is not a trivial task, and
446 moreover, nobody is really able to define what "realistic" means when
447 speaking of topology files. You can find some more thoughts on this
449 <a href="http://graal.ens-lyon.fr/~alegrand/articles/Simgrid-Introduction.pdf">slides</a>.
451 If you are looking for an actual tool, there we have a little tool to
452 annotate Tiers-generated topologies. This perl-script is in
453 <tt>tools/platform_generation/</tt> directory of the SVN. Dinda et Al.
454 released a very comparable tool, and called it GridG.
457 The specified computing power will be available to up to 6 sequential
458 tasks without sharing. If more tasks are placed on this host, the
459 resource will be shared accordingly. For example, if you schedule 12
460 tasks on the host, each will get half of the computing power. Please
461 note that although sound, this model were never scientifically
462 assessed. Please keep this fact in mind when using it.
467 \section faq_troubleshooting Troubleshooting
469 \subsection faq_trouble_lib_compil SimGrid compilation and installation problems
471 \subsubsection faq_trouble_lib_config cmake fails!
473 We know only one reason for the configure to fail:
475 - <b>You are using a broken build environment</b>\n
476 If symptom is that the configury magic complains about gcc not being able to build
477 executables, you are probably missing the libc6-dev package. Damn Ubuntu.
479 If you experience other kind of issue, please get in touch with us. We are
480 always interested in improving our portability to new systems.
482 \subsubsection faq_trouble_distcheck Dude! "ctest" fails on my machine!
484 Don't assume we never run this target, because we do. Check
485 http://cdash.inria.fr/CDash/index.php?project=Simgrid (click on
486 previous if there is no result for today: results are produced only by
487 11am, French time) and
488 https://buildd.debian.org/status/logs.php?pkg=simgrid if you don't believe us.
490 If it's failing on your machine in a way not experienced by the
491 autobuilders above, please drop us a mail on the mailing list so that
492 we can check it out. Make sure to read \ref faq_bugrepport before you
495 \subsection faq_trouble_compil User code compilation problems
497 \subsubsection faq_trouble_err_logcat "gcc: _simgrid_this_log_category_does_not_exist__??? undeclared (first use in this function)"
499 This is because you are using the log mecanism, but you didn't created
500 any default category in this file. You should refer to \ref XBT_log
501 for all the details, but you simply forgot to call one of
502 XBT_LOG_NEW_DEFAULT_CATEGORY() or XBT_LOG_NEW_DEFAULT_SUBCATEGORY().
504 \subsubsection faq_trouble_pthreadstatic "gcc: undefined reference to pthread_key_create"
506 This indicates that one of the library SimGrid depends on (libpthread
507 here) was missing on the linking command line. Dependencies of
508 libsimgrid are expressed directly in the dynamic library, so it's
509 quite impossible that you see this message when doing dynamic linking.
511 If you compile your code statically (and if you use a pthread version
512 of SimGrid -- see \ref faq_more_processes), you must absolutely
513 specify <tt>-lpthread</tt> on the linker command line. As usual, this should
514 come after <tt>-lsimgrid</tt> on this command line.
516 \subsubsection faq_trouble_lib_msg_deprecated "gcc: undefined reference to MSG_*"
518 Since version 3.7 all the m_channel_t mecanism is deprecated. So functions
519 about this mecanism may get removed in future releases.
523 \li XBT_PUBLIC(int) MSG_get_host_number(void);
525 \li XBT_PUBLIC(m_host_t *) MSG_get_host_table(void);
527 \li XBT_PUBLIC(MSG_error_t) MSG_get_errno(void);
529 \li XBT_PUBLIC(MSG_error_t) MSG_task_get(m_task_t * task, m_channel_t channel);
531 \li XBT_PUBLIC(MSG_error_t) MSG_task_get_with_timeout(m_task_t * task, m_channel_t channel, double max_duration);
533 \li XBT_PUBLIC(MSG_error_t) MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
535 \li XBT_PUBLIC(MSG_error_t) MSG_task_get_ext(m_task_t * task, int channel, double max_duration, m_host_t host);
537 \li XBT_PUBLIC(MSG_error_t) MSG_task_put(m_task_t task, m_host_t dest, m_channel_t channel);
539 \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);
541 \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);
543 \li XBT_PUBLIC(int) MSG_task_Iprobe(m_channel_t channel);
545 \li XBT_PUBLIC(int) MSG_task_probe_from(m_channel_t channel);
547 \li XBT_PUBLIC(int) MSG_task_probe_from_host(int channel, m_host_t host);
549 \li XBT_PUBLIC(MSG_error_t) MSG_set_channel_number(int number);
551 \li XBT_PUBLIC(int) MSG_get_channel_number(void);
553 If you want them you have to compile Simgrid v3.7 with option "-Denable_msg_deprecated=ON".
554 Using them should print warning to inform what new function you have to use.
556 \subsection faq_trouble_errors Runtime error messages
558 \subsubsection faq_flexml_limit "surf_parse_lex: Assertion `next limit' failed."
560 This is because your platform file is too big for the parser.
562 Actually, the message comes directly from FleXML, the technology on top of
563 which the parser is built. FleXML has the bad idea of fetching the whole
564 document in memory before parsing it. And moreover, the memory buffer size
565 must be determined at compilation time.
567 We use a value which seems big enough for our need without bloating the
568 simulators footprints. But of course your mileage may vary. In this case,
569 just edit src/surf/surfxml.l modify the definition of
570 FLEXML_BUFFERSTACKSIZE. E.g.
573 #define FLEXML_BUFFERSTACKSIZE 1000000000
576 Then recompile and everything should be fine, provided that your version of
577 Flex is recent enough (>= 2.5.31). If not the compilation process should
580 A while ago, we worked on FleXML to reduce a bit its memory consumption, but
581 these issues remain. There is two things we should do:
583 - use a dynamic buffer instead of a static one so that the only limit
584 becomes your memory, not a stupid constant fixed at compilation time
585 (maybe not so difficult).
586 - change the parser so that it does not need to get the whole file in
587 memory before parsing
588 (seems quite difficult, but I'm a complete newbe wrt flex stuff).
590 These are changes to FleXML itself, not SimGrid. But since we kinda hijacked
591 the development of FleXML, I can grant you that any patches would be really
592 welcome and quickly integrated.
594 <b>Update:</b> A new version of FleXML (1.7) was released. Most of the work
595 was done by William Dowling, who use it in his own work. The good point is
596 that it now use a dynamic buffer, and that the memory usage was greatly
597 improved. The downside is that William also changed some things internally,
598 and it breaks the hack we devised to bypass the parser, as explained in
599 \ref pf_flexml_bypassing. Indeed, this is not a classical usage of the
600 parser, and Will didn't imagine that we may have used (and even documented)
601 such a crude usage of FleXML. So, we now have to repair the bypassing
602 functionality to use the latest FleXML version and fix the memory usage in
605 \subsubsection faq_trouble_errors_big_fat_warning I'm told that my XML files are too old.
607 The format of the XML platform description files is sometimes
608 improved. For example, we decided to change the units used in SimGrid
609 from MBytes, MFlops and seconds to Bytes, Flops and seconds to ease
610 people exchanging small messages. We also reworked the route
611 descriptions to allow more compact descriptions.
613 That is why the XML files are versionned using the 'version' attribute
614 of the root tag. Currently, it should read:
616 <platform version="2">
619 If your files are too old, you can use the simgrid_update_xml.pl
620 script which can be found in the tools directory of the archive.
622 \subsection faq_trouble_valgrind Valgrind-related and other debugger issues
624 If you don't, you really should use valgrind to debug your code, it's
627 \subsubsection faq_trouble_vg_longjmp longjmp madness in valgrind
629 This is when valgrind starts complaining about longjmp things, just like:
631 \verbatim ==21434== Conditional jump or move depends on uninitialised value(s)
632 ==21434== at 0x420DBE5: longjmp (longjmp.c:33)
634 ==21434== Use of uninitialised value of size 4
635 ==21434== at 0x420DC3A: __longjmp (__longjmp.S:48)
638 This is the sign that you didn't used the exception mecanism well. Most
639 probably, you have a <tt>return;</tt> somewhere within a <tt>TRY{}</tt>
640 block. This is <b>evil</b>, and you must not do this. Did you read the section
643 \subsubsection faq_trouble_vg_libc Valgrind spits tons of errors about backtraces!
645 It may happen that valgrind, the memory debugger beloved by any decent C
646 programmer, spits tons of warnings like the following :
647 \verbatim ==8414== Conditional jump or move depends on uninitialised value(s)
648 ==8414== at 0x400882D: (within /lib/ld-2.3.6.so)
649 ==8414== by 0x414EDE9: (within /lib/tls/i686/cmov/libc-2.3.6.so)
650 ==8414== by 0x400B105: (within /lib/ld-2.3.6.so)
651 ==8414== by 0x414F937: _dl_open (in /lib/tls/i686/cmov/libc-2.3.6.so)
652 ==8414== by 0x4150F4C: (within /lib/tls/i686/cmov/libc-2.3.6.so)
653 ==8414== by 0x400B105: (within /lib/ld-2.3.6.so)
654 ==8414== by 0x415102D: __libc_dlopen_mode (in /lib/tls/i686/cmov/libc-2.3.6.so)
655 ==8414== by 0x412D6B9: backtrace (in /lib/tls/i686/cmov/libc-2.3.6.so)
656 ==8414== by 0x8076446: xbt_dictelm_get_ext (dict_elm.c:714)
657 ==8414== by 0x80764C1: xbt_dictelm_get (dict_elm.c:732)
658 ==8414== by 0x8079010: xbt_cfg_register (config.c:208)
659 ==8414== by 0x806821B: MSG_config (msg_config.c:42)
662 This problem is somewhere in the libc when using the backtraces and there is
663 very few things we can do ourselves to fix it. Instead, here is how to tell
664 valgrind to ignore the error. Add the following to your ~/.valgrind.supp (or
665 create this file on need). Make sure to change the obj line according to
666 your personnal mileage (change 2.3.6 to the actual version you are using,
667 which you can retrieve with a simple "ls /lib/ld*.so").
670 name: Backtrace madness
677 fun:__libc_dlopen_mode
680 Then, you have to specify valgrind to use this suppression file by passing
681 the <tt>--suppressions=$HOME/.valgrind.supp</tt> option on the command line.
682 You can also add the following to your ~/.bashrc so that it gets passed
683 automatically. Actually, it passes a bit more options to valgrind, and this
684 happen to be my personnal settings. Check the valgrind documentation for
687 \verbatim export VALGRIND_OPTS="--leak-check=yes --leak-resolution=high --num-callers=40 --tool=memcheck --suppressions=$HOME/.valgrind.supp" \endverbatim
689 \subsubsection faq_trouble_backtraces Truncated backtraces
691 When debugging SimGrid, it's easier to pass the
692 --disable-compiler-optimization flag to the configure if valgrind or
693 gdb get fooled by the optimization done by the compiler. But you
694 should remove these flag when everything works before going in
695 production (before launching your 1252135 experiments), or everything
696 will run only one half of the true SimGrid potential.
698 \subsection faq_deadlock There is a deadlock in my code!!!
700 Unfortunately, we cannot debug every code written in SimGrid. We
701 furthermore believe that the framework provides ways enough
702 information to debug such informations yourself. If the textual output
703 is not enough, Make sure to check the \ref faq_visualization FAQ entry to see
704 how to get a graphical one.
706 Now, if you come up with a really simple example that deadlocks and
707 you're absolutely convinced that it should not, you can ask on the
708 list. Just be aware that you'll be severely punished if the mistake is
709 on your side... We have plenty of FAQ entries to redact and new
710 features to implement for the impenitents! ;)
712 \subsection faq_surf_network_latency I get weird timings when I play with the latencies.
714 OK, first of all, remember that units should be Bytes, Flops and
715 Seconds. If you don't use such units, some SimGrid constants (e.g. the
716 SG_TCP_CTE_GAMMA constant used in most network models) won't have the
717 right unit and you'll end up with weird results.
719 Here is what happens with a single transfer of size L on a link
720 (bw,lat) when nothing else happens.
723 0-----lat--------------------------------------------------t
724 |-----|**** real_bw =min(bw,SG_TCP_CTE_GAMMA/(2*lat)) *****|
727 In more complex situations, this min is the solution of a complex
728 max-min linear system. Have a look
729 <a href="http://lists.gforge.inria.fr/pipermail/simgrid-devel/2006-April/thread.html">here</a>
730 and read the two threads "Bug in SURF?" and "Surf bug not
731 fixed?". You'll have a few other examples of such computations. You
732 can also read "A Network Model for Simulation of Grid Application" by
733 Henri Casanova and Loris Marchal to have all the details. The fact
734 that the real_bw is smaller than bw is easy to understand. The fact
735 that real_bw is smaller than SG_TCP_CTE_GAMMA/(2*lat) is due to the
736 window-based congestion mechanism of TCP. With TCP, you can't exploit
737 your huge network capacity if you don't have a good round-trip-time
738 because of the acks...
740 Anyway, what you get is t=lat + L/min(bw,SG_TCP_CTE_GAMMA/(2*lat)).
742 * if I you set (bw,lat)=(100 000 000, 0.00001), you get t = 1.00001 (you fully
744 * if I you set (bw,lat)=(100 000 000, 0.0001), you get t = 1.0001 (you're on the
746 * if I you set (bw,lat)=(100 000 000, 0.001), you get t = 10.001 (ouch!)
748 This bound on the effective bandwidth of a flow is not the only thing
749 that may make your result be unexpected. For example, two flows
750 competing on a saturated link receive an amount of bandwidth inversely
751 proportional to their round trip time.
753 \subsection faq_bugrepport So I've found a bug in SimGrid. How to report it?
755 We do our best to make sure to hammer away any bugs of SimGrid, but this is
756 still an academic project so please be patient if/when you find bugs in it.
757 If you do, the best solution is to drop an email either on the simgrid-user
758 or the simgrid-devel mailing list and explain us about the issue. You can
759 also decide to open a formal bug report using the
760 <a href="https://gforge.inria.fr/tracker/?atid=165&group_id=12&func=browse">relevant
761 interface</a>. You need to login on the server to get the ability to submit
764 We will do our best to solve any problem repported, but you need to help us
765 finding the issue. Just telling "it segfault" isn't enough. Telling "It
766 segfaults when running the attached simulator" doesn't really help either.
767 You may find the following article interesting to see how to repport
768 informative bug repports:
769 http://www.chiark.greenend.org.uk/~sgtatham/bugs.html (it is not SimGrid
770 specific at all, but it's full of good advices).
772 \author Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
776 ******************************************************************
777 * OLD CRUFT NOT USED ANYMORE *
778 ******************************************************************
781 subsection faq_crosscompile Cross-compiling a Windows DLL of SimGrid from linux
783 At the moment, we do not distribute Windows pre-compiled version of SimGrid
784 because the support for this platform is still experimental. We know that
785 some parts of the GRAS environment do not work, and we think that the others
786 environments (MSG and SD) have good chances to work, but we didn't test
787 ourselves. This section explains how we generate the SimGrid DLL so that you
788 can build it for yourself. First of all, you need to have a version more
789 recent than 3.1 (ie, a SVN version as time of writting).
791 In order to cross-compile the package to windows from linux, you need to
792 install mingw32 (minimalist gnu win32). On Debian, you can do so by
793 installing the packages mingw32 (compiler), mingw32-binutils (linker and
794 so), mingw32-runtime.
796 You can use the VPATH support of configure to compile at the same time for
797 linux and windows without dupplicating the source nor cleaning the tree
798 between each. Just run bootstrap (if you use the SVN) to run the autotools.
799 Then, create a linux and a win directories. Then, type:
800 \verbatim cd linux; ../configure --srcdir=.. <usual configure flags>; make; cd ..
801 cd win; ../configure --srcdir=.. --host=i586-mingw32msvc <flags>; make; cd ..
803 The trick to VPATH builds is to call configure from another directory,
804 passing it an extra --srcdir argument to tell it where all the sources are.
805 It will understand you want to use VPATH. Then, the trick to cross-compile
806 is simply to add a --host argument specifying the target you want to build
807 for. The i586-mingw32msvc string is what you have to pass to use the mingw32
808 environment as distributed in Debian.
810 After that, you can run all make targets from both directories, and test
811 easily that what you change for one arch does not break the other one.
813 It is possible that this VPATH build thing breaks from time to time in the
814 SVN since it's quite fragile, but it's granted to work in any released
815 version. If you experience problems, drop us a mail.
817 Another possible source of issue is that at the moment, building the
818 examples request to use the gras_stub_generator tool, which is a compiled
819 program, not a script. In cross-compilation, you need to cross-execute with
820 wine for example, which is not really pleasant. We are working on this, but
821 in the meanwhile, simply don't build the examples in cross-compilation
822 (<tt>cd src</tt> before running make).
824 Program (cross-)compiled with mingw32 do request an extra DLL at run-time to be
825 usable. For example, if you want to test your build with wine, you should do
826 the following to put this library where wine looks for DLLs.
828 cp /usr/share/doc/mingw32-runtime/mingwm10.dll.gz ~/.wine/c/windows/system/
829 gunzip ~/.wine/c/windows/system/mingwm10.dll.gz
832 The DLL is built in src/.libs, and installed in the <i>prefix</i>/bin directory
833 when you run make install.
835 If you want to use it in a native project on windows, you need to use
836 simgrid.dll and mingwm10.dll. For each DLL, you need to build .def file
837 under linux (listing the defined symbols), and convert it into a .lib file
838 under windows (specifying this in a way that windows compilers like). To
839 generate the def files, run (under linux):
840 \verbatim echo "LIBRARY libsimgrid-0.dll" > simgrid.def
841 echo EXPORTS >> simgrid.def
842 nm libsimgrid-0.dll | grep ' T _' | sed 's/.* T _//' >> simgrid.def
843 nm libsimgrid-0.dll | grep ' D _' | sed 's/.* D _//' | sed 's/$/ DATA/' >> simgrid.def
845 echo "LIBRARY mingwm10.dll" > mingwm10.def
846 echo EXPORTS >> mingwm10.def
847 nm mingwm10.dll | grep ' T _' | sed 's/.* T _//' >> mingwm10.def
848 nm mingwm10.dll | grep ' D _' | sed 's/.* D _//' | sed 's/$/ DATA/' >> mingwm10.def
851 To create the import .lib files, use the <tt>lib</tt> windows tool (from
852 MSVC) the following way to produce simgrid.lib and mingwm10.lib
853 \verbatim lib /def:simgrid.def
854 lib /def:mingwm10.def
857 If you happen to use Borland C Builder, the right command line is the
858 following (note that you don't need any file.def to get this working).
859 \verbatim implib simgrid.lib libsimgrid-0.dll
860 implib mingwm10.lib mingwm10.dll
863 Then, set the following parameters in Visual C++ 2005:
864 Linker -> Input -> Additional dependencies = simgrid.lib mingwm10.lib
866 Just in case you wonder how to generate a DLL from libtool in another
867 project, we added -no-undefined to any lib*_la_LDFLAGS variables so that
868 libtool accepts to generate a dynamic library under windows. Then, to make
869 it true, we pass any dependencies (such as -lws2 under windows or -lpthread
870 on need) on the linking line. Passing such deps is a good idea anyway so
871 that they get noted in the library itself, avoiding the users to know about
872 our dependencies and put them manually on their compilation line. Then we
873 added the AC_LIBTOOL_WIN32_DLL macro just before AC_PROG_LIBTOOL in the
874 configure.ac. It means that we exported any symbols which need to be.
875 Nowadays, functions get automatically exported, so we don't need to load our
876 header files with tons of __declspec(dllexport) cruft. We only need to do so
877 for data, but there is no public data in SimGrid so we are good.