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/articles/simgrid-tutorial.pdf">the tutorial slides</a>
9 (or to these <a href="http://graal.ens-lyon.fr/~alegrand/articles/slides_g5k_simul.pdf">old slides</a>,
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 or on a real platform. For this,
40 there is two implementations of the GRAS interface, one for simulation, one
41 for real execution. So, you just have to relink your code to chose one of
44 \subsection faq_generic First steps with SimGrid
46 If you decide to go for the MSG interface, please read carefully the
47 \ref MSG_examples. You'll find in \ref MSG_ex_master_slave a very
48 simple consisting of a master (that owns a bunch of tasks and
49 distributes them) , some slaves (that process tasks whenever they
50 receive one) and some forwarder agents (that simply pass the tasks
51 they receive to some slaves).
53 If you decide to go for the GRAS interface, you should definitively
54 read the \ref GRAS_tut. The first section constitutes an introduction
55 to the tool and presents the model we use. The second section
56 constitutes a complete step-by-step tutorial building a distributed
57 application from the beginning and exemplifying most of the GRAS
58 features in the process. The last section groups some HOWTOS
59 highlighting a given feature of the framework in a more concise way.
61 If you decide to go for another interface, I'm afraid your only sources
62 of information will be the source code and the mailing lists...
64 \subsection faq_visualization Visualizing and analyzing the results
66 It is sometime convenient to "see" how the agents are behaving. If you
67 like colors, you can use <tt>tools/MSG_visualization/colorize.pl </tt>
68 as a filter to your MSG outputs. It works directly with INFO. Beware,
69 INFO() prints on stderr. Do not forget to redirect if you want to
70 filter (e.g. with bash):
72 ./msg_test small_platform.xml small_deployment.xml 2>&1 | ../../tools/MSG_visualization/colorize.pl
75 We also have a more graphical output. Have a look at MSG_paje_output(). It
76 generates an input to <a href="http://www-id.imag.fr/Logiciels/paje/">Paje</a>.
79 <a href="Paje_MSG_screenshot.jpg"><img src="Paje_MSG_screenshot_thn.jpg"></a>
83 Visualization with Paje can be seen as a kind of postmortem
84 analysis. However, as soon as you start playing with big simulations,
85 you'll realize that processing such output is kind of tricky. There is
86 so much generic information that it is hard to find the information
89 As a matter of fact, logging really depends on simulations (e.g. what
90 kind of events is important...). That is why we do not propose a big
91 dump of your whole simulation (it would slow everything down) but give
92 you neat tools to structure you logs. Have a look at \ref XBT_log. In
93 fact, rather than a post-mortem analysis, you may want to do it on the
94 fly. The process you are running can do whatever you want. Have you
95 thought about adding a global structure where you directly compute the
96 information that are really important rather than writing everything
97 down and then processing huge files?
99 \subsection faq_C Argh! Do I really have to code in C?
101 Up until now, there is no binding for other languages. If you use C++,
102 you should be able to use the SimGrid library as a standard C library
103 and everything should work fine (simply <i>link</i> against this
104 library; recompiling SimGrid with a C++ compiler won't work and it
105 wouldn't help if you could).
107 In fact, we are currently working on Java bindings of MSG to allow
108 all the undergrad students of the world to use this tool. This is a
109 little more tricky than I would have expected, but the work is moving
110 fast forward [2006/05/13]. More languages are evaluated, but for now,
111 we do not feel a real demand for any other language. Please speak up!
113 \section faq_installation Installing the SimGrid library
115 Many people have been asking me questions on how to use SimGrid. Quite
116 often, the questions were not really about SimGrid but on the
117 installation process. This section is intended to help people that are
118 not familiar with compiling C files under UNIX. If you follow these
119 instructions and still have some troubles, drop an e-mail to
120 <simgrid-user@lists.gforge.inria.fr>.
122 \subsection faq_compiling Compiling SimGrid from a stable archive
124 First of all, you need to download the latest version of SimGrid from
125 <a href="http://gforge.inria.fr/frs/?group_id=12">here</a>.
126 Suppose you have uncompressed SimGrid in some temporary location of
127 your home directory (say <tt>/home/joe/tmp/simgrid-3.0.1 </tt>). The
128 simplest way to use SimGrid is to install it in your home
129 directory. Change your directory to
130 <tt>/home/joe/tmp/simgrid-3.0.1</tt> and type
133 ./configure --prefix=$HOME
138 If at some point, something fails, check the section "\ref
139 faq_trouble_compil". If it does not help, you can report this problem to the
140 list but, please, avoid sending a laconic mail like "There is a problem. Is it
141 okay?". Send the config.log file which is automatically generated by
142 configure. Try to capture both the standard output and the error output of the
143 <tt>make</tt> command with <tt>script</tt>. There is no way for us to help you
144 without the relevant bits of information.
146 Now, the following directory should have been created :
148 \li <tt>/home/joe/doc/simgrid/html/</tt>
149 \li <tt>/home/joe/lib/</tt>
150 \li <tt>/home/joe/include/</tt>
152 SimGrid is not a binary, it is a library. Both a static and a dynamic
153 version are available. Here is what you can find if you try a <tt>ls
156 \verbatim libsimgrid.a libsimgrid.la libsimgrid.so libsimgrid.so.0 libsimgrid.so.0.0.1
159 Thus, there is two ways to link your program with SimGrid:
160 \li Either you use the static version, e.g
161 \verbatim gcc libsimgrid.a -o MainProgram MainProgram.c
163 In this case, all the SimGrid functions are directly
164 included in <tt>MainProgram</tt> (hence a bigger binary).
165 \li Either you use the dynamic version (the preferred method)
166 \verbatim gcc -lsimgrid -o MainProgram MainProgram.c
168 In this case, the SimGrid functions are not included in
169 <tt>MainProgram</tt> and you need to set your environment
170 variable in such a way that <tt>libsimgrid.so</tt> will be
171 found at runtime. This can be done by adding the following
172 line in your .bashrc (if you use bash and if you have
173 installed the SimGrid libraries in your home directory):
174 \verbatim export LD_LIBRARY_PATH=$HOME/lib/:$LD_LIBRARY_PATH
177 \subsection faq_compiling_snapshoot SimGrid development snapshots
179 We have very high standards on software quality, and we are reluctant releasing
180 a stable release as long as there is still some known bug in the code base. In
181 addition, we added quite an extensive test base, making sure that we correctly
182 test the most important parts of the tool.
184 As an unfortunate conclusion, there may be some time between the stable
185 releases. If you want to benefit from the most recent features we introduced,
186 but don't want to take the risk of an untested version from the SVN, then
187 development snapshots are done for you.
189 These are pre-releases of SimGrid that still fail some tests about features
190 that almost nobody use, or on platforms not being in our core target (which is
191 Linux, Mac, other Unixes and Windows, from the most important to the less
192 one). That means that using this development releases should be safe for most
195 These archives can be found on
196 <a href="http://www.loria.fr/~quinson/simgrid.html">this web page</a>. Once you
197 got the lastest archive, you can compile it just like any archive (see above).
199 \subsection faq_compiling_svn Compiling SimGrid from the SVN
201 The project development takes place in the SVN, where all changes are
202 committed when they happen. Then every once in a while, we make sure that the
203 code quality meets our standard and release an archive from the code in the
204 SVN. We afterward go back to the development in the SVN. So, if you need a
205 recently added feature and can afford some little problem with the stability
206 of the lastest features, you may want to use the SVN version instead of a
209 For that, you first need to get the "simgrid" module from
210 <a href="http://gforge.inria.fr/scm/?group_id=12">here</a>.
212 You won't find any <tt>configure</tt> and a few other things
213 (<tt>Makefile.in</tt>'s, documentation, ...) will be missing as well. The
214 reason for that is that all these files have to be regenerated using the
215 latest versions of <tt>autoconf</tt>, <tt>libtool</tt>, <tt>automake</tt>
216 (>1.9) and <tt>doxygen</tt> (>1.4). To generate the <tt>configure</tt> and
217 the <tt>Makefile.in</tt>'s, you just have to launch the <tt>bootstrap</tt>
218 command that resides in the top of the source tree. Then just follow the
219 instructions of Section \ref faq_compiling.
221 We insist on the fact that you really need the latest versions of
222 autoconf, automake and libtool. Doing this step on exotic architectures/systems
223 (i.e. anything different from a recent linux distribution) may be
224 ... uncertain. If you need to compile the SVN version on a machine where all these
225 dependencies are not met, the easiest is to do <tt>make dist</tt> in the SVN
226 directory of another machine where all dependencies are met. It will create an
227 archive you may deploy on other sites just as a regular stable release.
229 In summary, the following commands will checkout the SVN, regenerate the
230 configure script and friends, configure SimGrid and build it.
232 \verbatim svn checkout svn://scm.gforge.inria.fr/svn/simgrid/simgrid/trunk simgrid
235 ./configure --enable-maintainer-mode --prefix=<where to install SimGrid>
238 Then, if you want to install SimGrid on the current box, just do:
239 \verbatim make install \endverbatim
241 If you want to build an snapshot of the SVN to deploy it on another box (for
242 example because the other machine don't have the autotools), do:
243 \verbatim make dist \endverbatim
245 Moreover, you should never call the autotools manually since you must run
246 them in a specific order with specific arguments. Most of the times, the
247 makefiles will automatically call the tools for you. When it's not possible
248 (such as the first time you checkout the SVN), use the ./bootstrap command
249 to call them explicitly.
252 \subsection faq_setting_MSG Setting up your own MSG code
254 Do not build your simulator by modifying the SimGrid examples. Go
255 outside the SimGrid source tree and create your own working directory
256 (say <tt>/home/joe/SimGrid/MyFirstScheduler/</tt>).
258 Suppose your simulation has the following structure (remember it is
259 just an example to illustrate a possible way to compile everything;
260 feel free to organize it as you want).
262 \li <tt>sched.h</tt>: a description of the core of the
263 scheduler (i.e. which functions are can be used by the
264 agents). For example we could find the following functions
265 (master, forwarder, slave).
267 \li <tt>sched.c</tt>: a C file including <tt>sched.h</tt> and
268 implementing the core of the scheduler. Most of these
269 functions use the MSG functions defined in section \ref
272 \li <tt>masterslave.c</tt>: a C file with the main function, i.e.
273 the MSG initialization (MSG_global_init()), the platform
274 creation (e.g. with MSG_create_environment()), the
275 deployment phase (e.g. with MSG_function_register() and
276 MSG_launch_application()) and the call to
279 To compile such a program, we suggest to use the following
280 Makefile. It is a generic Makefile that we have used many times with
281 our students when we teach the C language.
285 masterslave: masterslave.o sched.o
287 INSTALL_PATH = $$HOME
289 PEDANTIC_PARANOID_FREAK = -O0 -Wshadow -Wcast-align \
290 -Waggregate-return -Wmissing-prototypes -Wmissing-declarations \
291 -Wstrict-prototypes -Wmissing-prototypes -Wmissing-declarations \
292 -Wmissing-noreturn -Wredundant-decls -Wnested-externs \
293 -Wpointer-arith -Wwrite-strings -finline-functions
294 REASONABLY_CAREFUL_DUDE = -Wall
295 NO_PRAYER_FOR_THE_WICKED = -w -O2
296 WARNINGS = $(REASONABLY_CAREFUL_DUDE)
297 CFLAGS = -g $(WARNINGS)
299 INCLUDES = -I$(INSTALL_PATH)/include
300 DEFS = -L$(INSTALL_PATH)/lib/
301 LDADD = -lm -lsimgrid
305 $(CC) $(INCLUDES) $(DEFS) $(CFLAGS) $^ $(LIBS) $(LDADD) -o $@
308 $(CC) $(INCLUDES) $(DEFS) $(CFLAGS) -c -o $@ $<
311 rm -f $(BIN_FILES) *.o *~
317 The first two lines indicates what should be build when typing make
318 (<tt>masterslave</tt>) and of which files it is to be made of
319 (<tt>masterslave.o</tt> and <tt>sched.o</tt>). This makefile assumes
320 that you have set up correctly your <tt>LD_LIBRARY_PATH</tt> variable
321 (look, there is a <tt>LDADD = -lm -lsimgrid</tt>). If you prefer using
322 the static version, remove the <tt>-lsimgrid</tt> and add a
323 <tt>$(INSTALL_PATH)/lib/libsimgrid.a</tt> on the next line, right
324 after the <tt>LIBS = </tt>.
326 More generally, if you have never written a Makefile by yourself, type
327 in a terminal : <tt>info make</tt> and read the introduction. The
328 previous example should be enough for a first try but you may want to
329 perform some more complex compilations...
331 \subsection faq_setting_GRAS Setting up your own GRAS code
333 If you use the GRAS interface instead of the MSG one, then previous section
334 is not the better source of information. Instead, you should check the GRAS
335 tutorial in general, and the \ref GRAS_tut_tour_setup in particular.
338 \section faq_howto Feature related questions
340 \subsection faq_MIA "Could you please add (your favorite feature here) to SimGrid?"
342 Here is the deal. The whole SimGrid project (MSG, SURF, GRAS, ...) is
343 meant to be kept as simple and generic as possible. We cannot add
344 functions for everybody's needs when these functions can easily be
345 built from the ones already in the API. Most of the time, it is
346 possible and when it was not possible we always have upgraded the API
347 accordingly. When somebody asks us a question like "How to do that?
348 Is there a function in the API to simply do this?", we're always glad
349 to answer and help. However if we don't need this code for our own
350 need, there is no chance we're going to write it... it's your job! :)
351 The counterpart to our answers is that once you come up with a neat
352 implementation of this feature (task duplication, RPC, thread
353 synchronization, ...), you should send it to us and we will be glad to
354 add it to the distribution. Thus, other people will take advantage of
355 it (and we don't have to answer this question again and again ;).
357 You'll find in this section a few "Missing In Action" features. Many
358 people have asked about it and we have given hints on how to simply do
359 it with MSG. Feel free to contribute...
361 \subsection faq_MIA_MSG MSG features
363 \subsubsection faq_MIA_examples I want some more complex MSG examples!
365 Many people have come to ask me a more complex example and each time,
366 they have realized afterward that the basics were in the previous three
369 Of course they have often been needing more complex functions like
370 MSG_process_suspend(), MSG_process_resume() and
371 MSG_process_isSuspended() (to perform synchronization), or
372 MSG_task_Iprobe() and MSG_process_sleep() (to avoid blocking
373 receptions), or even MSG_process_create() (to design asynchronous
374 communications or computations). But the examples are sufficient to
377 We know. We should add some more examples, but not really some more
378 complex ones... We should add some examples that illustrate some other
379 functionalists (like how to simply encode asynchronous
380 communications, RPC, process migrations, thread synchronization, ...)
381 and we will do it when we will have a little bit more time. We have
382 tried to document the examples so that they are understandable. Tell
383 us if something is not clear and once again feel free to participate!
386 \subsubsection faq_MIA_taskdup Missing in action: MSG Task duplication/replication
388 There is no task duplication in MSG. When you create a task, you can
389 process it or send it somewhere else. As soon as a process has sent
390 this task, he doesn't have this task anymore. It's gone. The receiver
391 process has got the task. However, you could decide upon receiving to
392 create a "copy" of a task but you have to handle by yourself the
393 semantic associated to this "duplication".
395 As we already told, we prefer keeping the API as simple as
396 possible. This kind of feature is rather easy to implement by users
397 and the semantic you associate really depends on people. Having a
398 *generic* task duplication mechanism is not that trivial (in
399 particular because of the data field). That is why I would recommand
400 that you write it by yourself even if I can give you advice on how to
403 You have the following functions to get informations about a task:
404 MSG_task_get_name(), MSG_task_get_compute_duration(),
405 MSG_task_get_remaining_computation(), MSG_task_get_data_size(),
406 and MSG_task_get_data().
408 You could use a dictionary (#xbt_dict_t) of dynars (#xbt_dynar_t). If
409 you still don't see how to do it, please come back to us...
411 \subsubsection faq_MIA_asynchronous I want to do asynchronous communications in MSG
413 Up until now, there is no asynchronous communications in MSG. However,
414 you can create as many process as you want so you should be able to do
415 whatever you want... I've written a queue module to help implementing
416 some asynchronous communications at low cost (creating thousands of
417 process only to handle communications may be problematic in term of
418 performance at some point). I'll add it in the distribution asap.
420 \subsubsection faq_MIA_thread_synchronization I need to synchronize my MSG processes
422 You obviously cannot use pthread_mutexes of pthread_conds. The best
423 thing would be to propose similar structures. Unfortunately, we
424 haven't found time to do it yet. However you can try to play with
425 MSG_process_suspend() and MSG_process_resume(). You can even do some
426 synchronization with fake communications (using MSG_task_get(),
427 MSG_task_put() and MSG_task_Iprobe()).
429 \subsubsection faq_MIA_host_load Where is the get_host_load function hidden in MSG?
431 There is no such thing because its semantic wouldn't be really
432 clear. Of course, it is something about the amount of host throughput,
433 but there is as many definition of "host load" as people asking for
434 this function. First, you have to remember that resource availability
435 may vary over time, which make any load notion harder to define.
437 It may be instantaneous value or an average one. Moreover it may be only the
438 power of the computer, or may take the background load into account, or may
439 even take the currently running tasks into account. In some SURF models,
440 communications have an influence on computational power. Should it be taken
443 First of all, it's near to impossible to predict the load beforehands in the
444 simulator since it depends on too much parameters (background load
445 variation, bandwidth sharing algorithmic complexity) some of them even being
446 not known beforehands (other task starting at the same time). So, getting
447 this information is really hard (just like in real life). It's not just that
448 we want MSG to be as painful as real life. But as it is in some way
449 realistic, we face some of the same problems as we would face in real life.
451 How would you do it for real? The most common option is to use something
452 like NWS that performs active probes. The best solution is probably to do
453 the same within MSG, as in next code snippet. It is very close from what you
454 would have to do out of the simulator, and thus gives you information that
455 you could also get in real settings to not hinder the realism of your
459 double get_host_load() {
460 m_task_t task = MSG_task_create("test", 0.001, 0, NULL);
461 double date = MSG_get_clock();
463 MSG_task_execute(task);
464 date = MSG_get_clock() - date;
465 MSG_task_destroy(task);
470 Of course, it may not match your personal definition of "host load". In this
471 case, please detail what you mean on the mailing list, and we will extend
472 this FAQ section to fit your taste if possible.
474 \subsubsection faq_MIA_communication_time How can I get the *real* communication time?
476 Communications are synchronous and thus if you simply get the time
477 before and after a communication, you'll only get the transmission
478 time and the time spent to really communicate (it will also take into
479 account the time spent waiting for the other party to be
480 ready). However, getting the *real* communication time is not really
481 hard either. The following solution is a good starting point.
486 m_task_t task = MSG_task_create("Task", task_comp_size, task_comm_size,
487 calloc(1,sizeof(double)));
488 *((double*) task->data) = MSG_get_clock();
489 MSG_task_put(task, slaves[i % slaves_count], PORT_22);
490 INFO0("Send completed");
495 m_task_t task = NULL;
498 time1 = MSG_get_clock();
499 a = MSG_task_get(&(task), PORT_22);
500 time2 = MSG_get_clock();
501 if(time1<*((double *)task->data))
502 time1 = *((double *) task->data);
503 INFO1("Communication time : \"%f\" ", time2-time1);
505 MSG_task_destroy(task);
510 \subsection faq_MIA_SimDag SimDag related questions
512 \subsubsection faq_SG_comm Implementing communication delays between tasks.
514 A classic question of SimDag newcomers is about how to express a
515 communication delay between tasks. The thing is that in SimDag, both
516 computation and communication are seen as tasks. So, if you want to
517 model a data dependency between two DAG tasks t1 and t2, you have to
518 create 3 SD_tasks: t1, t2 and c and add dependencies in the following
522 SD_task_dependency_add(NULL, NULL, t1, c);
523 SD_task_dependency_add(NULL, NULL, c, t2);
526 This way task t2 cannot start before the termination of communication c
527 which in turn cannot start before t1 ends.
529 When creating task c, you have to associate an amount of data (in bytes)
530 corresponding to what has to be sent by t1 to t2.
532 Finally to schedule the communication task c, you have to build a list
533 comprising the workstations on which t1 and t2 are scheduled (w1 and w2
534 for example) and build a communication matrix that should look like
537 \subsubsection faq_SG_DAG How to implement a distributed dynamic scheduler of DAGs.
539 Distributed is somehow "contagious". If you start making distributed
540 decisions, there is no way to handle DAGs directly anymore (unless I
541 am missing something). You have to encode your DAGs in term of
542 communicating process to make the whole scheduling process
543 distributed. Here is an example of how you could do that. Assume T1
544 has to be done before T2.
547 int your_agent(int argc, char *argv[] {
549 T1 = MSG_task_create(...);
550 T2 = MSG_task_create(...);
554 if(cond) MSG_task_execute(T1);
556 if((MSG_task_get_remaining_computation(T1)=0.0) && (you_re_in_a_good_mood))
559 /* do something else */
565 If you decide that the distributed part is not that much important and that
566 DAG is really the level of abstraction you want to work with, then you should
567 give a try to \ref SD_API.
569 \subsection faq_MIA_generic Generic features
571 \subsubsection faq_more_processes Increasing the amount of simulated processes
573 Here are a few tricks you can apply if you want to increase the amount
574 of processes in your simulations.
576 - <b>A few thousands of simulated processes</b> (soft tricks)\n
577 SimGrid can use either pthreads library or the UNIX98 contextes. On
578 most systems, the number of pthreads is limited and then your
579 simulation may be limited for a stupid reason. This is especially
580 true with the current linux pthreads, and I cannot get more than
581 2000 simulated processes with pthreads on my box. The UNIX98
582 contexts allow me to raise the limit to 25,000 simulated processes
584 The <tt>--with-context</tt> option of the <tt>./configure</tt>
585 script allows you to choose between UNIX98 contextes
586 (<tt>--with-context=ucontext</tt>) and the pthread version
587 (<tt>--with-context=pthread</tt>). The default value is ucontext
588 when the script detect a working UNIX98 context implementation. On
589 Windows boxes, the provided value is discarded and an adapted
590 version is picked up.\n\n
591 We experienced some issues with contextes on some rare systems
592 (solaris 8 and lower or old alpha linuxes comes to mind). The main
593 problem is that the configure script detect the contextes as being
594 functional when it's not true. If you happen to use such a system,
595 switch manually to the pthread version, and provide us with a good
596 patch for the configure script so that it is done automatically ;)
598 - <b>Hundred thousands of simulated processes</b> (hard-core tricks)\n
599 As explained above, SimGrid can use UNIX98 contextes to represent
600 and handle the simulated processes. Thanks to this, the main
601 limitation to the number of simulated processes becomes the
602 available memory.\n\n
603 Here are some tricks I had to use in order to run a token ring
604 between 25,000 processes on my laptop (1Gb memory, 1.5Gb swap).\n
605 - First of all, make sure your code runs for a few hundreds
606 processes before trying to push the limit. Make sure it's
607 valgrind-clean, ie that valgrind does not report neither memory
608 error nor memory leaks. Indeed, numerous simulated processes
609 result in *fat* simulation hindering debugging.
610 - It was really boring to write 25,000 entries in the deployment
611 file, so I wrote a little script
612 <tt>examples/gras/mutual_exclusion/simple_token/make_deployment.pl</tt>, which you may
613 want to adapt to your case. You could also think about hijacking
614 the SURFXML parser (have look at \ref faq_flexml_bypassing).
615 - The deployment file became quite big, so I had to do what is in
616 the FAQ entry \ref faq_flexml_limit
617 - Each UNIX98 context has its own stack entry. As debugging this is
618 quite hairly, the default value is a bit overestimated so that
619 user don't get into trouble about this. You want to tune this
620 size to increse the number of processes. This is the
621 <tt>STACK_SIZE</tt> define in
622 <tt>src/xbt/xbt_context_sysv.c</tt>, which is 128kb by default.
623 Reduce this as much as you can, but be warned that if this value
624 is too low, you'll get a segfault. The token ring example, which
625 is quite simple, runs with 40kb stacks.
626 - You may tweak the logs to reduce the stack size further. When
627 logging something, we try to build the string to display in a
628 char array on the stack. The size of this array is constant (and
629 equal to XBT_LOG_BUFF_SIZE, defined in include/xbt/log/h). If the
630 string is too large to fit this buffer, we move to a dynamically
631 sized buffer. In which case, we have to traverse one time the log
632 event arguments to compute the size we need for the buffer,
633 malloc it, and traverse the argument list again to do the actual
635 The idea here is to move XBT_LOG_BUFF_SIZE to 1, forcing the logs
636 to use a dynamic array each time. This allows us to lower further
637 the stack size at the price of some performance loss...\n
638 This allowed me to run the reduce the stack size to ... 4k. Ie,
639 on my 1Gb laptop, I can run more than 250,000 processes!
641 \subsubsection faq_MIA_batch_scheduler Is there a native support for batch schedulers in SimGrid?
643 No, there is no native support for batch schedulers and none is
644 planned because this is a very specific need (and doing it in a
645 generic way is thus very hard). However some people have implemented
646 their own batch schedulers. Vincent Garonne wrote one during his PhD
647 and put his code in the contrib directory of our SVN so that other can
648 keep working on it. You may find inspiring ideas in it.
650 \subsubsection faq_MIA_checkpointing I need a checkpointing thing
652 Actually, it depends on whether you want to checkpoint the simulation, or to
653 simulate checkpoints.
655 The first one could help if your simulation is a long standing process you
656 want to keep running even on hardware issues. It could also help to
657 <i>rewind</i> the simulation by jumping sometimes on an old checkpoint to
658 cancel recent calculations.\n
659 Unfortunately, such thing will probably never exist in SG. One would have to
660 duplicate all data structures because doing a rewind at the simulator level
661 is very very hard (not talking about the malloc free operations that might
662 have been done in between). Instead, you may be interested in the Libckpt
663 library (http://www.cs.utk.edu/~plank/plank/www/libckpt.html). This is the
664 checkpointing solution used in the condor project, for example. It makes it
665 easy to create checkpoints (at the OS level, creating something like core
666 files), and rerunning them on need.
668 If you want to simulate checkpoints instead, it means that you want the
669 state of an executing task (in particular, the progress made towards
670 completion) to be saved somewhere. So if a host (and the task executing on
671 it) fails (cf. #MSG_HOST_FAILURE), then the task can be restarted
672 from the last checkpoint.\n
674 Actually, such a thing does not exists in SimGrid either, but it's just
675 because we don't think it is fundamental and it may be done in the user code
676 at relatively low cost. You could for example use a watcher that
677 periodically get the remaining amount of things to do (using
678 MSG_task_get_remaining_computation()), or fragment the task in smaller
681 \subsection faq_platform Platform building and Dynamic resources
683 \subsubsection faq_platform_example Where can I find SimGrid platform files?
685 There is several little examples in the archive, in the examples/msg
686 directory. From time to time, we are asked for other files, but we
687 don't have much at hand right now.
689 You should refer to the Platform Description Archive
690 (http://pda.gforge.inria.fr) project to see the other platform file we
691 have available, as well as the Simulacrum simulator, meant to generate
692 SimGrid platforms using all classical generation algorithms.
694 \subsubsection faq_platform_alnem How can I automatically map an existing platform?
696 We are working on a project called ALNeM (Application-Level Network
697 Mapper) which goal is to automatically discover the topology of an
698 existing network. Its output will be a platform description file
699 following the SimGrid syntax, so everybody will get the ability to map
700 their own lab network (and contribute them to the catalog project).
701 This tool is not ready yet, but it move quite fast forward. Just stay
704 \subsubsection faq_platform_synthetic Generating synthetic but realistic platforms
706 The third possibility to get a platform file (after manual or
707 automatic mapping of real platforms) is to generate synthetic
708 platforms. Getting a realistic result is not a trivial task, and
709 moreover, nobody is really able to define what "realistic" means when
710 speaking of topology files. You can find some more thoughts on this
712 <a href="http://graal.ens-lyon.fr/~alegrand/articles/Simgrid-Introduction.pdf">slides</a>.
714 If you are looking for an actual tool, there we have a little tool to
715 annotate Tiers-generated topologies. This perl-script is in
716 <tt>tools/platform_generation/</tt> directory of the SVN. Dinda et Al.
717 released a very comparable tool, and called it GridG.
719 \subsubsection faq_SURF_dynamic Expressing dynamic resource availability in platform files
721 A nice feature of SimGrid is that it enables you to seamlessly have
722 resources whose availability change over time. When you build a
723 platform, you generally declare hosts like that:
726 <host id="host A" power="100.00"/>
729 If you want the availability of "host A" to change over time, the only
730 thing you have to do is change this definition like that:
733 <host id="host A" power="100.00" availability_file="trace_A.txt" state_file="trace_A_failure.txt"/>
736 For hosts, availability files are expressed in fraction of available
737 power. Let's have a look at what "trace_A.txt" may look like:
746 At time 0, our host will deliver 100 flop/s. At time 11.0, it will
747 deliver only 50 flop/s until time 20.0 where it will will start
748 delivering 90 flop/s. Last at time 21.0 (20.0 plus the periodicity
749 1.0), we'll be back to the beginning and it will deliver 100 flop/s.
751 Now let's look at the state file:
758 A negative value means "off" while a positive one means "on". At time
759 1.0, the host is on. At time 1.0, it is turned off and at time 2.0, it
760 is turned on again until time 12 (2.0 plus the periodicity 10.0). It
761 will be turned on again at time 13.0 until time 23.0, and so on.
763 Now, let's look how the same kind of thing can be done for network
764 links. A usual declaration looks like:
767 <link id="LinkA" bandwidth="10.0" latency="0.2"/>
770 You have at your disposal the following options: bandwidth_file,
771 latency_file and state_file. The only difference with hosts is that
772 bandwidth_file and latency_file do not express fraction of available
773 power but are expressed directly in bytes per seconds and seconds.
775 \subsubsection faq_platform_multipath How to express multipath routing in platform files?
777 It is unfortunately impossible to express the fact that there is more
778 than one routing path between two given hosts. Let's consider the
779 following platform file:
782 <route src="A" dst="B">
785 <route src="B" dst="C">
788 <route src="A" dst="C">
793 Although it is perfectly valid, it does not mean that data traveling
794 from A to C can either go directly (using link 3) or through B (using
795 links 1 and 2). It simply means that the routing on the graph is not
796 trivial, and that data do not following the shortest path in number of
797 hops on this graph. Another way to say it is that there is no implicit
798 in these routing descriptions. The system will only use the routes you
799 declare (such as <route src="A" dst="C"><link:ctn
800 id="3"/></route>), without trying to build new routes by aggregating
803 You are also free to declare platform where the routing is not
804 symmetric. For example, add the following to the previous file:
807 <route src="C" dst="A">
813 This makes sure that data from C to A go through B where data from A
814 to C go directly. Don't worry about realism of such settings since
815 we've seen ways more weird situation in real settings (in fact, that's
816 the realism of very regular platforms which is questionable, but
817 that's another story).
819 \subsubsection faq_flexml_bypassing Bypassing the XML parser with your own C functions
821 So you want to bypass the XML files parser, uh? Maybe doing some parameter
822 sweep experiments on your simulations or so? This is possible, and
823 it's not even really difficult (well. Such a brutal idea could be
824 harder to implement). Here is how it goes.
826 For this, you have to first remember that the XML parsing in SimGrid is done
827 using a tool called FleXML. Given a DTD, this gives a flex-based parser. If
828 you want to bypass the parser, you need to provide some code mimicking what
829 it does and replacing it in its interactions with the SURF code. So, let's
830 have a look at these interactions.
832 FleXML parser are close to classical SAX parsers. It means that a
833 well-formed SimGrid platform XML file might result in the following
836 - start "platform_description" with attribute version="2"
837 - start "host" with attributes id="host1" power="1.0"
839 - start "host" with attributes id="host2" power="2.0"
841 - start "link" with ...
843 - start "route" with ...
844 - start "link:ctn" with ...
847 - end "platform_description"
849 The communication from the parser to the SURF code uses two means:
850 Attributes get copied into some global variables, and a surf-provided
851 function gets called by the parser for each event. For example, the event
852 - start "host" with attributes id="host1" power="1.0"
854 let the parser do something roughly equivalent to:
856 strcpy(A_host_id,"host1");
861 In SURF, we attach callbacks to the different events by initializing the
862 pointer functions to some the right surf functions. Since there can be
863 more than one callback attached to the same event (if more than one
864 model is in use, for example), they are stored in a dynar. Example in
865 workstation_ptask_L07.c:
867 /* Adding callback functions */
868 surf_parse_reset_parser();
869 surfxml_add_callback(STag_surfxml_host_cb_list, &parse_cpu_init);
870 surfxml_add_callback(STag_surfxml_prop_cb_list, &parse_properties);
871 surfxml_add_callback(STag_surfxml_link_cb_list, &parse_link_init);
872 surfxml_add_callback(STag_surfxml_route_cb_list, &parse_route_set_endpoints);
873 surfxml_add_callback(ETag_surfxml_link_c_ctn_cb_list, &parse_route_elem);
874 surfxml_add_callback(ETag_surfxml_route_cb_list, &parse_route_set_route);
877 surf_parse_open(file);
878 xbt_assert1((!surf_parse()), "Parse error in %s", file);
882 So, to bypass the FleXML parser, you need to write your own version of the
883 surf_parse function, which should do the following:
884 - Fill the A_<tag>_<attribute> variables with the wanted values
885 - Call the corresponding STag_<tag>_fun function to simulate tag start
886 - Call the corresponding ETag_<tag>_fun function to simulate tag end
887 - (do the same for the next set of values, and loop)
889 Then, tell SimGrid that you want to use your own "parser" instead of the stock one:
891 surf_parse = surf_parse_bypass_environment;
892 MSG_create_environment(NULL);
893 surf_parse = surf_parse_bypass_application;
894 MSG_launch_application(NULL);
897 A set of macros are provided at the end of
898 include/surf/surfxml_parse.h to ease the writing of the bypass
899 functions. An example of this trick is distributed in the file
900 examples/msg/masterslave/masterslave_bypass.c
902 \subsection faq_simgrid_configuration Changing SimGrid's behavior
904 A number of options can be given at runtime to change the default
905 SimGrid behavior. In particular, you can change the default cpu and
908 \subsubsection faq_simgrid_configuration_gtnets Using GTNetS
910 It is possible to use a packet-level network simulator
911 instead of the default flow-based simulation. You may want to use such
912 an approach if you have doubts about the validity of the default model
913 or if you want to perform some validation experiments. At the moment,
914 we support the GTNetS simulator (it is still rather experimental
915 though, so leave us a message if you play with it).
919 To enable GTNetS model inside SimGrid it is needed to patch the GTNetS simulator source code
920 and build/install it from scratch
923 - <b>Download and enter the recent downloaded GTNetS directory</b>
926 svn checkout svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/GTNetS/
931 - <b>Use the following commands to unzip and patch GTNetS package to work within SimGrid.</b>
934 unzip gtnets-current.zip
935 tar zxvf gtnets-current-patch.tgz
937 cat ../00*.patch | patch -p1
940 - <b>OPTIONALLY</b> you can use a patch for itanium 64bit processor family.
943 cat ../AMD64-FATAL-Removed-DUL_SIZE_DIFF-Added-fPIC-compillin.patch | patch -p1
946 - <b>Compile GTNetS</b>
948 Due to portability issues it is possible that GTNetS does not compile in your architecture. The patches furnished in SimGrid SVN repository are intended for use in Linux architecture only. Unfortunately, we do not have the time, the money, neither the manpower to guarantee GTNetS portability. We advice you to use one of GTNetS communication channel to get more help in compiling GTNetS.
952 ln -sf Makefile.linux Makefile
958 - <b>NOTE</b> A lot of warnings are expected but the application should compile
959 just fine. If the makefile insists in compiling some QT libraries
960 please try a make clean before asking for help.
963 - <b>To compile optimized version</b>
970 - <b>Installing GTNetS</b>
972 Replace <userhome> by some path you have write access to.
975 ln -sf libgtsim-debug.so /<userhome>/usr/lib/libgtnets.so
976 export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/<userhome>/usr/lib/libgtnets.so
977 mkdir /<userhome>/usr/include/gtnets
978 cp -fr SRC/*.h /<userhome>/usr/include/gtnets
982 - <b>Enable GTNetS support in SimGrid</b>
985 ./configure --with-gtnets=/<userhome>/usr
988 - <b>Once you have followed all the instructions for compiling and
989 installing successfully you can activate this feature at
990 runntime with the following options:</b>
993 cd simgrid/example/msg/
999 - <b>Or try the GTNetS model dogbone example with</b>
1002 gtnets/gtnets gtnets/onelink-p.xml gtnets/onelink-d.xml --cfg=network_model:GTNets
1006 A long version of this <a href="http://gforge.inria.fr/docman/view.php/12/6283/GTNetS HowTo.html">HowTo</a> it is available
1009 More about GTNetS simulator at <a href="http://www.ece.gatech.edu/research/labs/MANIACS/GTNetS/index.html">GTNetS Website</a>
1013 The patches provided by us worked successfully with GTNetS found
1014 <a href="http://www.ece.gatech.edu/research/labs/MANIACS/GTNetS/software/gtnets-current.zip">here</a>,
1015 dated from 12th June 2008. Due to the discontinuing development of
1016 GTNetS it is impossible to precise a version number. We STRONGLY recommend you
1017 to download and install the GTNetS version found in SimGrid repository as explained above.
1022 \subsubsection faq_simgrid_configuration_alternate_network Using alternative flow models
1024 The default simgrid network model uses a max-min based approach as
1025 explained in the research report
1026 <a href="ftp://ftp.ens-lyon.fr/pub/LIP/Rapports/RR/RR2002/RR2002-40.ps.gz">A Network Model for Simulation of Grid Application</a>.
1027 Other models have been proposed and implemented since then (see for example
1028 <a href="http://mescal.imag.fr/membres/arnaud.legrand/articles/simutools09.pdf">Accuracy Study and Improvement of Network Simulation in the SimGrid Framework</a>)
1029 and can be activated at runtime. For example:
1031 ./mycode platform.xml deployment.xml --cfg=workstation_model:compound --cfg=network_model:LV08 -cfg=cpu_model:Cas01
1034 Possible models for the network are currently "Constant", "CM02",
1035 "LegrandVelho", "GTNets", Reno", "Reno2", "Vegas". Others will
1036 probably be added in the future and many of the previous ones are
1037 experimental and are likely to disappear without notice...
1039 \section faq_troubleshooting Troubleshooting
1041 \subsection faq_trouble_lib_compil SimGrid compilation and installation problems
1043 \subsubsection faq_trouble_lib_config ./configure fails!
1045 We know only one reason for the configure to fail:
1047 - <b>You are using a broken build environment</b>\n
1048 If symptom is that configure complains about gcc not being able to build
1049 executables, you are probably missing the libc6-dev package. Damn Ubuntu.
1051 If you experience other kind of issue, please get in touch with us. We are
1052 always interested in improving our portability to new systems.
1054 \subsubsection faq_trouble_distcheck Dude! "make check" fails on my machine!
1056 Don't assume we never run this target, because we do. Check
1057 http://bob.loria.fr:8010 if you don't believe us.
1059 There is several reasons which may cause the make check to fail on your
1062 - <b>You are using a broken libc (probably concerning the contextes)</b>.\n
1063 The symptom is that the "make check" fails within the examples/msg directory.\n
1064 By default, SimGrid uses something called ucontexts. This is part of the
1065 libc, but it's quite undertested. For example, some (old) versions of the
1066 glibc on alpha do not implement these functions, but provide the stubs
1067 (which return ENOSYS: not implemented). It may fool our detection mechanism
1068 and leads to segfaults. There is not much we can do to fix the bug.
1069 A workaround is to compile with --with-context=pthread to avoid
1070 ucontext completely. You'll be a bit more limited in the number
1071 of simulated processes you can start concurrently, but 5000
1072 processes is still enough for most purposes, isn't it?\n
1073 This limitation is the reason why we insist on using this piece of ...
1074 software even if it's so troublesome.\n
1075 <b>=> use --with-pthread on AMD64 architecture that do not have an
1076 ultra-recent libc.</b>
1078 - <b>There is a bug in SimGrid we aren't aware of</b>.\n
1079 If none of the above apply, please drop us a mail on the mailing list so
1080 that we can check it out. Make sure to read \ref faq_bugrepport
1083 \subsection faq_trouble_compil User code compilation problems
1085 \subsubsection faq_trouble_err_logcat "gcc: _simgrid_this_log_category_does_not_exist__??? undeclared (first use in this function)"
1087 This is because you are using the log mecanism, but you didn't created
1088 any default category in this file. You should refer to \ref XBT_log
1089 for all the details, but you simply forgot to call one of
1090 XBT_LOG_NEW_DEFAULT_CATEGORY() or XBT_LOG_NEW_DEFAULT_SUBCATEGORY().
1092 \subsubsection faq_trouble_pthreadstatic "gcc: undefined reference to pthread_key_create"
1094 This indicates that one of the library SimGrid depends on (libpthread
1095 here) was missing on the linking command line. Dependencies of
1096 libsimgrid are expressed directly in the dynamic library, so it's
1097 quite impossible that you see this message when doing dynamic linking.
1099 If you compile your code statically (and if you use a pthread version
1100 of SimGrid -- see \ref faq_more_processes), you must absolutely
1101 specify <tt>-lpthread</tt> on the linker command line. As usual, this should
1102 come after <tt>-lsimgrid</tt> on this command line.
1104 \subsection faq_trouble_errors Runtime error messages
1106 \subsubsection faq_flexml_limit "surf_parse_lex: Assertion `next limit' failed."
1108 This is because your platform file is too big for the parser.
1110 Actually, the message comes directly from FleXML, the technology on top of
1111 which the parser is built. FleXML has the bad idea of fetching the whole
1112 document in memory before parsing it. And moreover, the memory buffer size
1113 must be determined at compilation time.
1115 We use a value which seems big enough for our need without bloating the
1116 simulators footprints. But of course your mileage may vary. In this case,
1117 just edit src/surf/surfxml.l modify the definition of
1118 FLEXML_BUFFERSTACKSIZE. E.g.
1121 #define FLEXML_BUFFERSTACKSIZE 1000000000
1124 Then recompile and everything should be fine, provided that your version of
1125 Flex is recent enough (>= 2.5.31). If not the compilation process should
1128 A while ago, we worked on FleXML to reduce a bit its memory consumption, but
1129 these issues remain. There is two things we should do:
1131 - use a dynamic buffer instead of a static one so that the only limit
1132 becomes your memory, not a stupid constant fixed at compilation time
1133 (maybe not so difficult).
1134 - change the parser so that it does not need to get the whole file in
1135 memory before parsing
1136 (seems quite difficult, but I'm a complete newbe wrt flex stuff).
1138 These are changes to FleXML itself, not SimGrid. But since we kinda hijacked
1139 the development of FleXML, I can grant you that any patches would be really
1140 welcome and quickly integrated.
1142 <b>Update:</b> A new version of FleXML (1.7) was released. Most of the work
1143 was done by William Dowling, who use it in his own work. The good point is
1144 that it now use a dynamic buffer, and that the memory usage was greatly
1145 improved. The downside is that William also changed some things internally,
1146 and it breaks the hack we devised to bypass the parser, as explained in
1147 \ref faq_flexml_bypassing. Indeed, this is not a classical usage of the
1148 parser, and Will didn't imagine that we may have used (and even documented)
1149 such a crude usage of FleXML. So, we now have to repair the bypassing
1150 functionality to use the lastest FleXML version and fix the memory usage in
1153 \subsubsection faq_trouble_gras_transport GRAS spits networking error messages
1155 Gras, on real platforms, naturally use regular sockets to communicate. They
1156 are deeply hidden in the gras abstraction, but when things go wrong, you may
1157 get some weird error messages. Here are some example, with the probable
1160 - <b>Transport endpoint is not connected</b>: several processes try to open
1161 a server socket on the same port number of the same machine. This is
1162 naturally bad and each process should pick its own port number for this.\n
1163 Maybe, you just have some processes remaining from a previous experiment
1165 Killing them may help, but again if you kill -KILL them, you'll have to
1166 wait for a while: they didn't close there sockets properly and the system
1167 needs a while to notice that this port is free again.
1169 - <b>Socket closed by remote side</b>: if the remote process is not
1170 supposed to close the socket at this point, it may be dead.
1172 - <b>Connection reset by peer</b>: I found this on Internet about this
1173 error. I think it's what's happening here, too:\n
1174 <i>This basically means that a network error occurred while the client was
1175 receiving data from the server. But what is really happening is that the
1176 server actually accepts the connection, processes the request, and sends
1177 a reply to the client. However, when the server closes the socket, the
1178 client believes that the connection has been terminated abnormally
1179 because the socket implementation sends a TCP reset segment telling the
1180 client to throw away the data and report an error.\n
1181 Sometimes, this problem is caused by not properly closing the
1182 input/output streams and the socket connection. Make sure you close the
1183 input/output streams and socket connection properly. If everything is
1184 closed properly, however, and the problem persists, you can work around
1185 it by adding a one-second sleep before closing the streams and the
1186 socket. This technique, however, is not reliable and may not work on all
1188 Since GRAS sockets are closed properly (repeat after me: there is no bug
1189 in GRAS), it is either that you are closing your sockets on server side
1190 before the client get a chance to read them (use gras_os_sleep() to delay
1191 the server), or the server died awfully before the client got the data.
1193 \subsubsection faq_trouble_errors_big_fat_warning I'm told that my XML files are too old.
1195 The format of the XML platform description files is sometimes
1196 improved. For example, we decided to change the units used in SimGrid
1197 from MBytes, MFlops and seconds to Bytes, Flops and seconds to ease
1198 people exchanging small messages. We also reworked the route
1199 descriptions to allow more compact descriptions.
1201 That is why the XML files are versionned using the 'version' attribute
1202 of the root tag. Currently, it should read:
1204 <platform version="2">
1207 If your files are too old, you can use the simgrid_update_xml.pl
1208 script which can be found in the tools directory of the archive.
1210 \subsection faq_trouble_valgrind Valgrind-related and other debugger issues
1212 If you don't, you really should use valgrind to debug your code, it's
1215 \subsubsection faq_trouble_vg_longjmp longjmp madness in valgrind
1217 This is when valgrind starts complaining about longjmp things, just like:
1219 \verbatim ==21434== Conditional jump or move depends on uninitialised value(s)
1220 ==21434== at 0x420DBE5: longjmp (longjmp.c:33)
1222 ==21434== Use of uninitialised value of size 4
1223 ==21434== at 0x420DC3A: __longjmp (__longjmp.S:48)
1226 This is the sign that you didn't used the exception mecanism well. Most
1227 probably, you have a <tt>return;</tt> somewhere within a <tt>TRY{}</tt>
1228 block. This is <b>evil</b>, and you must not do this. Did you read the section
1231 \subsubsection faq_trouble_vg_libc Valgrind spits tons of errors about backtraces!
1233 It may happen that valgrind, the memory debugger beloved by any decent C
1234 programmer, spits tons of warnings like the following :
1235 \verbatim ==8414== Conditional jump or move depends on uninitialised value(s)
1236 ==8414== at 0x400882D: (within /lib/ld-2.3.6.so)
1237 ==8414== by 0x414EDE9: (within /lib/tls/i686/cmov/libc-2.3.6.so)
1238 ==8414== by 0x400B105: (within /lib/ld-2.3.6.so)
1239 ==8414== by 0x414F937: _dl_open (in /lib/tls/i686/cmov/libc-2.3.6.so)
1240 ==8414== by 0x4150F4C: (within /lib/tls/i686/cmov/libc-2.3.6.so)
1241 ==8414== by 0x400B105: (within /lib/ld-2.3.6.so)
1242 ==8414== by 0x415102D: __libc_dlopen_mode (in /lib/tls/i686/cmov/libc-2.3.6.so)
1243 ==8414== by 0x412D6B9: backtrace (in /lib/tls/i686/cmov/libc-2.3.6.so)
1244 ==8414== by 0x8076446: xbt_dictelm_get_ext (dict_elm.c:714)
1245 ==8414== by 0x80764C1: xbt_dictelm_get (dict_elm.c:732)
1246 ==8414== by 0x8079010: xbt_cfg_register (config.c:208)
1247 ==8414== by 0x806821B: MSG_config (msg_config.c:42)
1250 This problem is somewhere in the libc when using the backtraces and there is
1251 very few things we can do ourselves to fix it. Instead, here is how to tell
1252 valgrind to ignore the error. Add the following to your ~/.valgrind.supp (or
1253 create this file on need). Make sure to change the obj line according to
1254 your personnal mileage (change 2.3.6 to the actual version you are using,
1255 which you can retrieve with a simple "ls /lib/ld*.so").
1258 name: Backtrace madness
1260 obj:/lib/ld-2.3.6.so
1265 fun:__libc_dlopen_mode
1268 Then, you have to specify valgrind to use this suppression file by passing
1269 the <tt>--suppressions=$HOME/.valgrind.supp</tt> option on the command line.
1270 You can also add the following to your ~/.bashrc so that it gets passed
1271 automatically. Actually, it passes a bit more options to valgrind, and this
1272 happen to be my personnal settings. Check the valgrind documentation for
1275 \verbatim export VALGRIND_OPTS="--leak-check=yes --leak-resolution=high --num-callers=40 --tool=memcheck --suppressions=$HOME/.valgrind.supp" \endverbatim
1277 \subsubsection faq_trouble_backtraces Truncated backtraces
1279 When debugging SimGrid, it's easier to pass the
1280 --disable-compiler-optimization flag to the configure if valgrind or
1281 gdb get fooled by the optimization done by the compiler. But you
1282 should remove these flag when everything works before going in
1283 production (before launching your 1252135 experiments), or everything
1284 will run only one half of the true SimGrid potential.
1286 \subsection faq_deadlock There is a deadlock in my code!!!
1288 Unfortunately, we cannot debug every code written in SimGrid. We
1289 furthermore believe that the framework provides ways enough
1290 information to debug such informations yourself. If the textual output
1291 is not enough, Make sure to check the \ref faq_visualization FAQ entry to see
1292 how to get a graphical one.
1294 Now, if you come up with a really simple example that deadlocks and
1295 you're absolutely convinced that it should not, you can ask on the
1296 list. Just be aware that you'll be severely punished if the mistake is
1297 on your side... We have plenty of FAQ entries to redact and new
1298 features to implement for the impenitents! ;)
1300 \subsection faq_surf_network_latency I get weird timings when I play with the latencies.
1302 OK, first of all, remember that units should be Bytes, Flops and
1303 Seconds. If you don't use such units, some SimGrid constants (e.g. the
1304 SG_TCP_CTE_GAMMA constant used in most network models) won't have the
1305 right unit and you'll end up with weird results.
1307 Here is what happens with a single transfer of size L on a link
1308 (bw,lat) when nothing else happens.
1311 0-----lat--------------------------------------------------t
1312 |-----|**** real_bw =min(bw,SG_TCP_CTE_GAMMA/(2*lat)) *****|
1315 In more complex situations, this min is the solution of a complex
1316 max-min linear system. Have a look
1317 <a href="http://lists.gforge.inria.fr/pipermail/simgrid-devel/2006-April/thread.html">here</a>
1318 and read the two threads "Bug in SURF?" and "Surf bug not
1319 fixed?". You'll have a few other examples of such computations. You
1320 can also read "A Network Model for Simulation of Grid Application" by
1321 Henri Casanova and Loris Marchal to have all the details. The fact
1322 that the real_bw is smaller than bw is easy to understand. The fact
1323 that real_bw is smaller than SG_TCP_CTE_GAMMA/(2*lat) is due to the
1324 window-based congestion mechanism of TCP. With TCP, you can't exploit
1325 your huge network capacity if you don't have a good round-trip-time
1326 because of the acks...
1328 Anyway, what you get is t=lat + L/min(bw,SG_TCP_CTE_GAMMA/(2*lat)).
1330 * if I you set (bw,lat)=(100 000 000, 0.00001), you get t = 1.00001 (you fully
1332 * if I you set (bw,lat)=(100 000 000, 0.0001), you get t = 1.0001 (you're on the
1334 * if I you set (bw,lat)=(100 000 000, 0.001), you get t = 10.001 (ouch!)
1336 This bound on the effective bandwidth of a flow is not the only thing
1337 that may make your result be unexpected. For example, two flows
1338 competing on a saturated link receive an amount of bandwidth inversely
1339 proportional to their round trip time.
1341 \subsection faq_bugrepport So I've found a bug in SimGrid. How to report it?
1343 We do our best to make sure to hammer away any bugs of SimGrid, but this is
1344 still an academic project so please be patient if/when you find bugs in it.
1345 If you do, the best solution is to drop an email either on the simgrid-user
1346 or the simgrid-devel mailing list and explain us about the issue. You can
1347 also decide to open a formal bug report using the
1348 <a href="https://gforge.inria.fr/tracker/?atid=165&group_id=12&func=browse">relevant
1349 interface</a>. You need to login on the server to get the ability to submit
1352 We will do our best to solve any problem repported, but you need to help us
1353 finding the issue. Just telling "it segfault" isn't enough. Telling "It
1354 segfaults when running the attached simulator" doesn't really help either.
1355 You may find the following article interesting to see how to repport
1356 informative bug repports:
1357 http://www.chiark.greenend.org.uk/~sgtatham/bugs.html (it is not SimGrid
1358 specific at all, but it's full of good advices).
1360 \author Arnaud Legrand (arnaud.legrand::imag.fr)
1361 \author Martin Quinson (martin.quinson::loria.fr)
1366 ******************************************************************
1367 * OLD CRUFT NOT USED ANYMORE *
1368 ******************************************************************
1371 \subsection faq_crosscompile Cross-compiling a Windows DLL of SimGrid from linux
1373 At the moment, we do not distribute Windows pre-compiled version of SimGrid
1374 because the support for this platform is still experimental. We know that
1375 some parts of the GRAS environment do not work, and we think that the others
1376 environments (MSG and SD) have good chances to work, but we didn't test
1377 ourselves. This section explains how we generate the SimGrid DLL so that you
1378 can build it for yourself. First of all, you need to have a version more
1379 recent than 3.1 (ie, a SVN version as time of writting).
1381 In order to cross-compile the package to windows from linux, you need to
1382 install mingw32 (minimalist gnu win32). On Debian, you can do so by
1383 installing the packages mingw32 (compiler), mingw32-binutils (linker and
1384 so), mingw32-runtime.
1386 You can use the VPATH support of configure to compile at the same time for
1387 linux and windows without dupplicating the source nor cleaning the tree
1388 between each. Just run bootstrap (if you use the SVN) to run the autotools.
1389 Then, create a linux and a win directories. Then, type:
1390 \verbatim cd linux; ../configure --srcdir=.. <usual configure flags>; make; cd ..
1391 cd win; ../configure --srcdir=.. --host=i586-mingw32msvc <flags>; make; cd ..
1393 The trick to VPATH builds is to call configure from another directory,
1394 passing it an extra --srcdir argument to tell it where all the sources are.
1395 It will understand you want to use VPATH. Then, the trick to cross-compile
1396 is simply to add a --host argument specifying the target you want to build
1397 for. The i586-mingw32msvc string is what you have to pass to use the mingw32
1398 environment as distributed in Debian.
1400 After that, you can run all make targets from both directories, and test
1401 easily that what you change for one arch does not break the other one.
1403 It is possible that this VPATH build thing breaks from time to time in the
1404 SVN since it's quite fragile, but it's granted to work in any released
1405 version. If you experience problems, drop us a mail.
1407 Another possible source of issue is that at the moment, building the
1408 examples request to use the gras_stub_generator tool, which is a compiled
1409 program, not a script. In cross-compilation, you need to cross-execute with
1410 wine for example, which is not really pleasant. We are working on this, but
1411 in the meanwhile, simply don't build the examples in cross-compilation
1412 (<tt>cd src</tt> before running make).
1414 Program (cross-)compiled with mingw32 do request an extra DLL at run-time to be
1415 usable. For example, if you want to test your build with wine, you should do
1416 the following to put this library where wine looks for DLLs.
1418 cp /usr/share/doc/mingw32-runtime/mingwm10.dll.gz ~/.wine/c/windows/system/
1419 gunzip ~/.wine/c/windows/system/mingwm10.dll.gz
1422 The DLL is built in src/.libs, and installed in the <i>prefix</i>/bin directory
1423 when you run make install.
1425 If you want to use it in a native project on windows, you need to use
1426 simgrid.dll and mingwm10.dll. For each DLL, you need to build .def file
1427 under linux (listing the defined symbols), and convert it into a .lib file
1428 under windows (specifying this in a way that windows compilers like). To
1429 generate the def files, run (under linux):
1430 \verbatim echo "LIBRARY libsimgrid-0.dll" > simgrid.def
1431 echo EXPORTS >> simgrid.def
1432 nm libsimgrid-0.dll | grep ' T _' | sed 's/.* T _//' >> simgrid.def
1433 nm libsimgrid-0.dll | grep ' D _' | sed 's/.* D _//' | sed 's/$/ DATA/' >> simgrid.def
1435 echo "LIBRARY mingwm10.dll" > mingwm10.def
1436 echo EXPORTS >> mingwm10.def
1437 nm mingwm10.dll | grep ' T _' | sed 's/.* T _//' >> mingwm10.def
1438 nm mingwm10.dll | grep ' D _' | sed 's/.* D _//' | sed 's/$/ DATA/' >> mingwm10.def
1441 To create the import .lib files, use the <tt>lib</tt> windows tool (from
1442 MSVC) the following way to produce simgrid.lib and mingwm10.lib
1443 \verbatim lib /def:simgrid.def
1444 lib /def:mingwm10.def
1447 If you happen to use Borland C Builder, the right command line is the
1448 following (note that you don't need any file.def to get this working).
1449 \verbatim implib simgrid.lib libsimgrid-0.dll
1450 implib mingwm10.lib mingwm10.dll
1453 Then, set the following parameters in Visual C++ 2005:
1454 Linker -> Input -> Additional dependencies = simgrid.lib mingwm10.lib
1456 Just in case you wonder how to generate a DLL from libtool in another
1457 project, we added -no-undefined to any lib*_la_LDFLAGS variables so that
1458 libtool accepts to generate a dynamic library under windows. Then, to make
1459 it true, we pass any dependencies (such as -lws2 under windows or -lpthread
1460 on need) on the linking line. Passing such deps is a good idea anyway so
1461 that they get noted in the library itself, avoiding the users to know about
1462 our dependencies and put them manually on their compilation line. Then we
1463 added the AC_LIBTOOL_WIN32_DLL macro just before AC_PROG_LIBTOOL in the
1464 configure.ac. It means that we exported any symbols which need to be.
1465 Nowadays, functions get automatically exported, so we don't need to load our
1466 header files with tons of __declspec(dllexport) cruft. We only need to do so
1467 for data, but there is no public data in SimGrid so we are good.