1 SimGrid (3.5) unstable; urgency=low
4 * New function: xbt_dict_cursor_set_data()
5 * New function: xbt_fifo_get_last_item()
6 * Bug fix in xbt_dynar_shrink(): use the right element size
8 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
10 SimGrid (3.4.1) stable; urgency=low
12 The "Polishing easter eggs is probably a good idea" release.
13 This is a bug fixes release only.
17 * Fix a bug preventing the tasks from begin garbage collected.
20 * Fix a bug occuring when a host involved in a communication fails.
21 This was not detected properly by the other peer involved in the
22 communication. Now, it's reported as a network error.
25 * Warn the user about loop dependencies in data flow of DAX files
26 * Obey the control-flow dependencies of DAX files
29 * Add option "enable_smpi" allowing to not compile SMPI.
30 Probably useful for the (Mac) users experiencing a build error here
31 * Improve the detection of lua5.1 and ruby1.8
33 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
35 SimGrid (3.4) stable; urgency=low
37 The "Easter in Cargese" release. Also known as (major changes):
39 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
40 ~> bindings were greatly improved
41 ~> new tracing infrastructure for better visualization introduced
43 * the "Welcome to configury modernity" release.
44 ~> we switched from autotools to cmake, and improved our cdash
47 A more detailled list of changes follow (full detail in svn log).
49 Java Bindings: Various Cleanups
50 * (install java-gcj-compat-dev on debian-like to use them)
51 * Remove put/get: no need to export deprecated interface in Java
52 Use send/receive instead.
53 * Cleanup the examples and add a README per directory
54 * Remove example autoDestination (that's the only way to go now)
55 * Remove example explicitDestination (was a plain copy of basic)
56 * Make JniException a runtime exception, so that there is no need to
57 declare the fact that you may encounter such a beast. I guess that
58 nobody will ever want to survive such error.
59 * Create specific errors for each MSG case of failure:
60 host failure, transfer failure, timeout, task cancelled
61 * Cleanup the exceptions that may get thrown by each function
62 * Other internal cleanups in Java bindings. Performance still bad :/
63 Ruby and Lua Bindings: create them
64 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
65 * That's new and great, you should try them out.
66 Same functionalities than Java bindings, only even less polished
68 * Kill the useless "rate" argument of SD_task_get_execution_time()
69 Everyone used to provide -1 as a value, it was not used, and the
70 semantic of a possible use wasn't even clear.
71 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
72 either as comm costs or compute costs to mean that there is no
73 such thing for that specific task.
74 * Add a SD_task_set_name() function
75 * Fix SD_task_unschedule() on typed tasks
76 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
77 * In DAX loader, accept useless 'level' attributes to <job> since
78 LIGO DAGs have them (seem to be to ease graphical representation).
80 * Add an example masterslave_mailbox.c using send/receive and not
81 the deprecated put/get interface.
82 * Kill the MSG_paje_output() function. It's a noop since 2 years.
83 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
85 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
86 (declare MSG_USE_DEPRECATED to still have the old name)
87 * Add a MSG_task_set_data() function
88 * About trace replay (see examples/msg/actions):
90 - Allow to work with splitted trace files for each process
91 Give the specific trace file as argument of each process,
92 and call MSG_action_trace_run(NULL)
93 You can still have one merged file for all processes.
94 - Fix implementation of collective operations
95 * Allow task_execute() on 0-sized tasks (closes #10063)
97 * This is the first release of SimGrid where SMPI is not considered
98 beta anymore (even if some corners should still be improved)
99 * Port over the new SIMIX_network submodule (internal refactoring)
100 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
101 * Implement more missing elements of the standard:
103 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
105 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
106 MPI_Scan() MPI_Get_processor_name()
107 - Added implementation of missing case for Alltoall (warning: it's
108 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
109 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
110 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
111 SMPI_MPI_Allgatherv()
113 - MPI_Waitsome() was broken
114 - Allow relative includes in smpicc
115 - Command line cfg argument 'reference_speed' was ignored...
116 - Some functions did not properly lead to auto-benching of user code
117 - smpicc passes -O2 by default (just like openmpi one)
119 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
120 * Bug fixes about timeouts during communications
121 * add SIMIX_message_sizes_output() as a pimple to write to file the
122 amount of messages per size. Use gnuplot to get histogram.
123 Pimple because that's the only user-visible function of simix,
124 defined directly in xbt.h (irk, sorry)
126 - Add a SIMIX_sem_get_capacity() function
127 - Fix interactions with processe resume/suspende
128 - release_forever() was stupidly broken
129 - Fix SIMIX_display_process_status() for processes in a semaphore
130 - Make SIMIX_sem_block_onto() user-visible
131 * Refactoring context stuff:
132 - Use pseudo-OOP for better modularity
133 - reimplement SIMIX_process_kill() without process_schedule() so
134 that the latter can take as invariant that it is called from
136 - Merge context_start into context_new for sake of simplicity
138 * Add a Vivaldi network model, coded live during SUD'10 ;)
139 * Rename configuration variables to start a hierarchy:
140 o cpu_model -> cpu/model
141 o network_model -> network/model
142 o workstation_model -> workstation/model
143 * New configuration variables:
144 o network/bandwidth_factor: correction to bandwith
145 o network/latency_factor: correction to latency
146 o netwotk/weight_S: correction to the weight of competing streams
147 * Add a long description to the models, that users can see with such
148 argument on the command line: --cfg=cpu/model:help
149 * --help-models display the long description of all known models
151 * config: add the ability to set a default value after registration
152 Does not override any previously set value (e.g. from cmd line)
153 * dict: allow to have integer key and data.
154 When so, you need to use the following functions
155 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
156 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
157 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
158 In contrary to regular dicts, the key is not malloced before copy.
159 Mixing scalar and regular elements in the same dict is not tested
161 * Allow to use xbt_dynar_shrink() to expend the dynar instead
162 Tracing for Visualization:
163 * SimGrid is now instrumented in order to generate a trace file for
164 visualization analysis: to use it, need to compile SimGrid with the
165 "tracing" option enabled, and instrument the program using SimGrid with
166 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
167 (among other functions).
168 * The instrumentation only traces the platform utilization for now
169 * Documentation to use the tracing functions and how to analyze the
170 traces with the Triva tool is written.
171 * More information about: SimGrid FAQ (in the section Tracing Simulations
174 * We moved to cmake as default build system. Autotools support will
175 be dropped soon. Check the FAQ for more info about how to use it.
176 * Greatly improved our cdash/ctest interactions
177 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
178 * Added memory checking tests with valgrind; lot of memleak fixing.
179 This may be the first release of simgrid with so few memory issues
180 * Added code coverage tests.
181 Our coverage is still improvable, but at least we see it on cdash.
183 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
185 SimGrid (3.3.4) stable; urgency=low
187 The "Desktop Grid needs love too" release (also called Xmas release).
190 * Major speedup in the maxmin system solving by using lazy evaluation
191 Instead of solving completely the maxmin system at each iteration,
192 only invalidate (and recompute) the modified parts.
193 This new feature is enabled in default models but you can try to
194 turn it on with "--cfg:maxmin-selective-update=1" for other models.
195 * Cas01 IMproved as default CPU model
196 This CPU model is the same Cas01 model, but it uses the
197 maxmin-selective-update flag and a heap structure to manage
198 actions on SURF kernel.
199 It reduces the complexity to find the next action to finish and,
200 consequently, it's faster than the old Cas01.
201 This is the new default CPU model (Cas01).
202 * Rename the old Cas01 model to Cas01_fullupdate
203 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
204 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
205 * CpuTI (CPU Trace Integration)
206 A new CPU model whose objective is simulate faster when using
207 availability trace files.
208 Instead of using a full featured, over engineered maxmin system for
209 CPU modeling, this model does the pre-integration of traces files
210 to calculate the amount of CPU power available, and so, executes
211 faster than the old CPU models.
212 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
213 * Use LV08 as default network model since it gives better accuracy
214 for small messages and shouldn't change things for big ones.
215 Use --cfg=network_model:CM02 to get the previous behavior.
218 ******************************************
219 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
220 ******************************************
221 * The new CPU model may changes simulations!
222 The point is that events occurring at the exact same timestamp
223 are not scheduled in the same order with the old and new
224 version. This may be enough to completely change the execution
225 of simulations in some cases.
226 * The new network model will change simulations!
227 This new model is more realistic than the previous one, so you
228 should consider redoing your old experiments with this model.
229 Sorry for the inconvenience.
232 * Introduce the supernovae compilation mode
233 When compiled that way, the whole SimGrid (or almost) is put in a
234 single compilation unit and compiled in one shoot.
235 This is to help gcc which has difficulties to inline stuff from one
237 The speedup seem to be above 15%, althrough more tests are needed on
238 amd64 to confirm that gain.
241 * Port of MSG's mailbox on top of SIMIX network
242 The put/get mechanism was greatly simplified on the way.
245 * New SIMIX network module. Provides:
246 - Mailbox: rendez-vous mecanism to find with who you want to speak
247 - Synchronous send/recv: easier and hopefully faster since the
248 logic is handled in the maestro process directly now
249 - Asynchronous send/recv: you dreamt of it? It's here now
250 Too bad that nobody cared enough to propagate the change to MSG.
251 * Add semaphores as SIMIX synchronization mechanism.
254 * new function SD_daxload(char*) to load a DAX file
255 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
256 * Introduce typed tasks. Specify its kind and cost at creation.
257 At scheduling, just give where it should be placed, and the cost
258 for each involved resource is automatically computed.
259 Existing constructors so far (more to come of course):
260 - SD_task_create_comm_e2e() for end-to-end communication
261 - SD_task_create_comp_seq() for sequential computation
262 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
263 * new function SD_task_dump() for debuging display
264 * new function SD_task_dotty(task,FILE*) writing to file the info
265 about the task in dotty format
266 * SD_task_dependency_exists() can now cope with having one of its
267 arguments NULL. If so, it tests whether the other argument has any
269 * Add getters on list of preceding/following tasks:
270 SD_task_get_parents(task) and SD_task_get_children(task)
271 * Add getters on amount of workstations and list:
272 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
273 * Add getter on task kind: SD_task_get_kind(task)
274 * Update the start_time and finish_time of tasks on completion/failure
275 * Bugfix: Remove task from state swags when destroyed
278 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
281 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
282 dynar with one separate thread per value of the dynar.
283 * Change the prototype of xbt_thread_create(), sorry.
284 Added a boolean parameter indicating whether we want to join this
285 thread (used in SG only for now)
286 * Implement xbt_thread_join and xbt_thread_yield in SG also.
289 * GTNetS wrappers should now be usable again (and betterly tested too)
290 * Fix a major regression from 3.2 where the timeout provided to
291 MSG_task_put_with_timeout() was used as absolute time before which
292 the comm should be done.
293 * Start to fix the <cluster> tag.
294 - Internal links should be good now (beside of the loopback, which
295 use the private link instead)
296 - paths to the external world is still rather broken
297 - the <route:multi> tag is just broken. Actually that's brain-dead.
298 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
299 to make it less stupid
300 ** Check your platform with teshsuite/simdag/platforms/flatifier **
301 * Fix a source-level compatibility glitch from 3.2: after defining
302 MSG_USE_DEPRECATED, you can use the old name
303 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
304 * Allow to compile from the SVN with automake 1.11
305 * Fix some problems when using the "start_time" tag in deployment XMLs.
306 * Fix #8569: XBT/synchro.h has redundant declarations
307 * Fix #8563: MSG return values and exceptions
308 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
309 * Integrate patch #8636: Obey DESTDIR when installing documentation.
310 Thanks to Robson Peixoto.
311 * Fix a vicious bug in dictionaries inducing that some elements were
312 not freed on xbt_dict_free()
314 Portability report of this version:
315 * Main portability targets:
316 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
318 - mac leopard on i386
319 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
320 but nothing critical.
321 * Other platforms: windows, AIX and others were not tested for this release
323 Timing report of this version:
324 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
325 scenario parameters). From 8h to a few seconds in desktop grid settings.
326 * Supernovae brings about 25% speedup on i386.
328 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
330 SimGrid (3.3.3) stable; urgency=low
332 The "Need for Speed" release.
334 The timings done to validate the 3.3.2 were faulty.
335 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
337 The problem was a conversion from a manually handled vector to
338 xbt_dynar_t on the critical path.
339 xbt_dynar_foreach calls functions, inducing stack management crap.
341 We inlined these functions and xbt_dynar_foreach is now breath taking.
342 We also inlined xbt_swag_belong on the way.
344 Here are some approximate speedup measurements (on master/slaves
345 simulations lasting between 10s and 20s each):
346 3.3.1 -> 3.3.2: about same performance
347 3.3.2 -> 3.3.3: 40% speedup
348 3.3.1 -> 3.3.3: 40% speedup
349 3.3.1 with inline patch -> 3.3.3: 30% speedup
351 Our reading is that the refactoring which occurred in 3.3.2 made us
352 suffer much more from the xbt_dynar_foreach low performance, but
353 once we solved this, this refactoring proved to be very performance
354 effective. From the 40% speedup, somehow, 10% are due to the
355 inlining and 30% to the refactoring.
357 That's a pitty that gcc cannot inline functions placed in other files
358 alone. We have to choose between:
359 - break the encapsulation (by putting private data structures and
360 accessors in headers files to help gcc)
361 - live with low performance
362 - switch to a decent compiler such as icc (not quite possible).
364 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
366 SimGrid (3.3.2) stable; urgency=low
368 The "Simplicity does not preceed complexity, but follows it" release.
370 The main contributors of this release were (lexical order):
371 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
374 * Extract the routing logic into its own object.
375 (was dupplicated in network.c and workstation_LV07.c;
376 Allows to implement other ways of storing that info)
377 => kill now useless network_card concept
378 - Use dynar to represent routes (instead of void** + int*)
379 - kill link_set (use surf_network_model->resource_set instead)
380 - Add a command-line option to choose the routing schema to use
381 - Add three new models:
382 * Floyd (shortest path computed at initialization)
383 * Dijikstra (shortest path recomputed all the time)
384 * Cached Dijikstra (shortest path computed on need)
385 All these models where contributed by Silas De Munck, and are
386 described in his ICCS09 paper.
388 * Simplify model declaration
389 (less redirections, less function to write when defining a model)
390 - Factorize stuff between models:
393 surf_model_resource_set(model)
394 surf_model_resource_by_name(model, name)
395 - Unify the types of models in s_surf_model_t (using an union)
396 - Embeed fields of common_public directly into s_surf_model_t
397 - Rename model methods:
398 action_free ~> action_unref
399 action_change_state ~> action_state_set
400 action_get_state ~> action_state_get
401 - Change model methods into functions :
402 (model)->common_public->action_use ~> surf_action_ref
404 * Implement a generic resource; use it as ancestor to specific ones
405 (allows to kill duplicated code in models)
406 Drawback: timer command don't need no name nor properties;
407 workstation_CLM03 don't need no properties
408 (but I guess we can live with those few bytes wasted)
410 * Improve the action object model
411 - implement a constructor avoiding dupplicated code about field
412 initialization in generic_action part.
414 * Kill the SDP model: it has an external dependency, is deprecated
415 in flavor of modern lmm models, and didn't compile since a while
418 * Relocation of the context module from XBT to SIMIX.
419 (the context were decoupled from the simix processes, duplicating a lot of code)
420 => a lot of code was factorized
421 - less overhead is introduced during scheduling
422 - simpler API for the context factory
423 - the logic for process creation,destruction and manipulation was simplified
424 * Simplification of the s_smx_process_t data structure.
425 => accesing the simix level data associated to a process is faster now,
426 and the code is a lot more readable.
429 * Implement some more MPI primitives:
430 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
431 -implementation: Bcast: flat or 2-ary tree (default),
434 Allreduce: Reduce then Bcast
435 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
436 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
437 Alltoallv: flat tree, like ompi
439 * Add support for optimized collectives (Bcast is now binomial by default)
440 * Port smpirun and smpicc to OS X
443 * Kill SD_link_get_properties: hard to maintain and makes very little sense
444 Shout out if you used it.
447 * Display the list of still queued messages in SG mode when existing
451 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
452 * Add xbt_graph_node_get_outedges() [Silas De Munck]
453 * Add xbt_str_from_file(FILE*)
454 * Add xbt_dict_get_key achieving a linear reverse search
455 * Remove the context module
457 Portability report of this version:
458 * Main portability targets:
459 - Linux(debian)/x86/context
460 - Linux(debian)/x86/pthreads
461 - Linux(debian)/amd64/context
462 - Linux(debian)/amd64/pthreads
463 On these, we still have the eratic breakages of gras/pmm and
464 amok/saturate_sg reported in previous version. We still think
465 that the tests are the cause of the fault, not the tested code.
467 - Mac OSX Leopard/x86/context
468 Still false negative in tesh autotesting.
469 Smpi still fails, but this time because readlink does not accept -f
470 Everything seems to work properly beside of that.
473 - AIX version 5.3 (only tested contexts this time)
474 Smpi still fails there because mktemp is not installed.
475 Everything seems to work properly beside of that.
477 I managed to compile it for the first time, but several breakages.
478 Won't delay the release for this exotic platform.
480 * Windows: it's still lagging behind. If you want to help, please
483 Timing report of this version:
484 This version seem to be more than 5% faster than 3.3.1 (on linux
485 64bits with contextes). The gain is less than expected, we are
486 investigating this for next release.
488 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
490 SimGrid (3.3.1) stable; urgency=low
493 * Implement a --cfg-help to show existing configuration variables
494 * Build chain do not require doxygen in maintainer mode
497 * fix a bug on struct sizeof computation, which prevented the
498 exchange of arrays of structs in some conditions
499 - added a regression test about this in datadesc_usage
500 * Allow the exchange of 0-long dynamic vectors.
501 - for that, use -1 as indicator of dynamic size instead of 0
502 - This implied to change any size from unsigned long to long,
503 reducing a bit communication abilities, but I guess that with
504 64bits being quite common, this is more than enough.
505 - This also induce a protocol change, thus bumping network protocol
506 version from 0 to 1 (if we have external users, we have to get
507 clean on that point too ;)
508 - added two regression tests about this in datadesc_usage
509 * Be more verbose when propagating local exceptions
510 This helps debugging.
511 * Display the status of simulated processes when receiving SIGINT in
515 * Allow to control the simulation from a trace file.
516 New functions MSG_action_register() and MSG_action_trace_run()
517 The first one allows to associate a function execution to each
518 kind of action while the second one parses a trace file and
519 triggers the corresponding actions within the system.
520 For now, only a toy example is provided in examples/msg/actions
521 * Add an exemple of process migration in examples/msg/migration
522 * Fix a bug in task exchange which broke MSG_task_get_sender()
523 Add a teshsuite regression test for that.
524 [Bug: if MSG_task_get_sender() is called after sender exit,
526 * Fix a bug which prevented suspend/resume to work properly
527 * Display the status of simulated processes when receiving SIGINT
528 This fixes a regression of v3.3. due to the introduction of SIMIX
529 * Bug fixing in failure management:
530 - trace could not start by a failure at time 0
531 - failure during communications were not working
534 * Add SIMIX_process_set_name() to change the name of the current
535 process in the log messages.
536 * Store smx_hosts in a dict since we only retrieve them by name
537 * Move the configuration infrastructure to surf
540 * Move the configuration infrastructure to surf
543 * Massive internal cleanups:
544 - Store internal structures on processes instead of hosts (allows
545 to have more than one process per host, in addition of being more
547 - Cleanup the initialization/finalization process
548 - Kill a whole bunch of unneeded synchronization:
549 processes run in exclusive manner within the simulator
550 - Move queues from global tables to process data fields
552 - now accept -platform and -hostfile arguments
553 - Pass the right rank value to processes according to the hostfile
554 * Compile the examples by default, and use them as regression tests
555 * Implement MPI_Wtime()
556 * Change the reference speed to a command line option
559 * TCP_gamma can now be specified as command line option using
560 --cfg=TCP_gamma:10000000.0
561 * Change the --surf-path cmd line option into --cfg=path:
564 * Also include strbuff from xbt.h public header
565 * xbt_ex_display(): do not free the exception after displaying
566 This allows to do more with the given exception afterward.
567 Users should call xbt_ex_free() themselves.
571 Portability report of this version:
572 * Main portability targets:
573 - Linux(debian)/x86/context
574 - Linux(debian)/x86/pthreads
575 - Linux(debian)/amd64/context
576 - Linux(debian)/amd64/pthreads
577 These targets fail about 1/10 of times on gras/pmm, but we believe
578 that this is because of the test, not because of simgrid.
579 amok/saturate_sg fails even more rarely, and the test may not be
582 - Mac OSX Leopard/x86/context
583 The test suite still spits tons of errors because some obscure
584 force prevents us from removing the temporary directories
585 arguing that they still contain some metadata I've never heard of.
586 Smpi fails because seq is not installed.
587 Everything seems to work properly beside of that.
590 - AIX version 5.3 (both contexts and pthread)
591 Smpi still fails there because mktemp is not installed.
592 XML inclusions seems rosty on AIX.
594 * Windows: it's still lagging behind. If you want to help, please
597 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
599 SimGrid (3.3) stable; urgency=high
603 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
606 * Introduce the SIMIX module: factorize code between MSG and GRAS.
609 Until now, GRAS were using MSG as an interface to SURF. It was
610 quite difficult because both interface have several differences
611 (MSG channels vs GRAS sockets were the most notable point).
613 This also opens the gate to SMPI (which should occur soon) and speed
614 up simulations by to 40% (even if it were not the main goal).
616 **************************************
617 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
618 **************************************
619 The point is that events occuring at the exact same timestamp are
620 not scheduled in the same order with the old and new version. This
621 may be enough to completely change the execution of simulations in
622 some cases. Sorry for the inconvenience.
624 * Cleanup and upgrade the XML format to push further scalability
625 issues (check http://hal.inria.fr/inria-00256883/ for more info)
627 * Improve the testing infrastructure with tesh. Now a very large part of
628 the code is tested not only by being run but also by checking that the
629 output match an expected output [Mt].
631 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
632 is really less memory-demanding, which should allow you to use
633 larger files in SimGrid [AL].
635 * Inform valgrind about our contextes, so that it becomes usable
636 with the default (and more effecient) version of SimGrid
637 [contributed by Sékou Diakite, many thanks]
640 * Introduce a listener thread in charge of receiving incomming
641 messages from the network. It allows to overlap communication and
642 computation but most notably, it removes some stupid deadlocks due
643 to the fact that so far, a process could not send and receive at
644 the same time. This made most non trivial communication schema
646 * Convert the PIDs from long int to int to match the MSG ones (and
648 * New function: gras_agent_spawn() to launch a new process on
649 current host. Only working in simulation for now. [Mt]
650 * New function: gras_os_hostport() returning a constant form (ie,
651 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
654 * Make the backtrace of exceptions more human readable [Mt]
655 * New module: xbt/str [Mt]
656 a ton of string utility functions (split, join, printf to a newly
657 allocated buffer, trim, etc)
658 * New module: xbt/hash [Mt]
659 SHA1 hashing algorithm (more to come if needed)
660 * New module: xbt/synchro [Mt]
661 synchronization tools (mutex and conditions) working the same way
662 in simulation and in real life (mainly useful for GRAS, but not
664 * New module: xbt/queue [Mt]
665 classical producer/consumer synchronization scheme
666 * xbt_dynar_new_sync() creates a synchronized dynar. All access
667 (using the classical functions will get serialized) [Mt]
668 * Make dictionary internal table dynamic. No need to specify its size
669 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
671 * Make sure the log channels are organized as a tree under windows
672 (because of ANSI C compatibility issue, any channel were child of
676 * Cleaned many thing in surf and fixed a few bugs [AL].
677 * Add a nice command line configuration mechanism to compose models [AL].
678 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
679 the previous one (KCCFLN05). It relies on something that looks like
680 a max-min sharing mechanism but cannot be written as such. A new solver
681 was thus designed [AL].
682 * Add a new solver to lmm. Based on Lagrange optimization and
683 gradient-based descent, it enables to efficiently maximise systems s.a
685 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
687 This solver enables to propose two new network models for TCP Reno and
688 TCP Vegas based on Low's work. These models still need to be fully
689 tested though [Pedro Velho].
692 * Bug fix in SD_simulate. Now the time bound given as argument is
694 * Use the new parallel task model (ptask_L07) as default.
695 * Use the SURF command line configuration mechanism.
696 * 0-size tasks (for synchronization) should now work.
698 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
700 SimGrid (3.2) stable; urgency=high
704 We still experience issues on this platform, but we believe that at
707 GRAS API BREAKAGE (for simplification purpose, sorry):
708 * the gras_msgtype_by_name is not used anymore. Instead of
709 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
710 you can write (and must)
711 gras_msg_send(toserver, "request", &request);
712 - If you still want to pass a gras_msgtype_t to the function (to cache
713 the type and avoid the lookup time), use the gras_msg_send_() variant.
714 - Impacted functions:
715 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
716 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
717 * The callbacks are now expected to return 0 when everything went well
718 (just like the main() function)
720 GRAS new features and improvements:
721 * New module mecanism where user code can use per process globals [Mt]
722 This is similar to gras_userdata_*() functions, but for libraries. It
723 factorize some code developped over and over in the examples and AMOK.
724 It has still to be documented and used (only amok/peermanagement is
726 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
727 loss when they were small enough to fit into the buffer and sent quickly
728 enough so that they can all get received in one shoot.
729 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
730 if not found. Use the *_or_null() variant for the old semantic.
731 * In gras_msg_handle, do not discard messages without callback.
732 They are probably messages to be explicitly awaited later (ie, proofs of
733 mis-synchronization in userland since they are sent before being awaited)
735 * gras_socket_meas_send/recv: semantic changed!
736 The numerical arguments used to be (1) the total amount of data to send
737 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
738 messages. This was need for the fool willing to send more than MAXINT
739 bytes on quite fat pipes.
742 * Do really rename the hostmanagement module to peermanagement. [Mt]
743 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
744 but this is rather new and this was documented in the module
745 documentation (poor excuses, I admit)
746 * Bandwidth measurement semantic changed! This follows the changes to
747 gras_socket_meas_send/recv explained above.
750 * A sequential mode has been added to the workstations. When a workstation
751 is in sequential mode, it can execute only one task, and the other tasks
752 are waiting in a FIFO. [Christophe Thiery]
755 * The KCCFLN05 workstation model now handles parallel tasks. It is the
756 model for SIMDAG. [Christophe Thiery]
757 * Bug fix in the maxmin solver: Some values were close to 0 instead of
758 equal to 0, which caused some bad behaviors in
759 saturated_constraint_set_update. I now use a threshold mechanism like in
763 * When running manually src/testall, you select specific units [Mt]
764 testall is the result of our cunit mecanism, and should replace all
765 the scripty thingy around since bash don't run easily on billware.
767 * A mallocator system has been added. [Christophe Thiery]
768 Mallocators allow you to recycle your unused objects instead of freeing them
769 and allocating new ones.
771 Documentation update:
772 * FAQ reworking + New FAQs:
773 - "Valgrind spits tons of errors!" [Mt]
774 - "How to repport bugs" [Mt]
775 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
776 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
777 - Communication time measurement within MSG [AL]
778 - I experience weird communication times when I change the latency [AL]
781 - an introduction to the framework and to the used communication model
782 - an initiatic tour introducing the most proheminent features:
784 . Lesson 0: Installing GRAS
785 . Lesson 1: Setting up your own project
786 o Part 2: Message passing
787 . Lesson 2: Exchanging simple messages
788 . Lesson 3: Passing arguments to the processes (in SG)
789 . Lesson 4: Attaching callbacks to messages
790 . Lesson 5: Using globals in processes
791 . Lesson 6: Logging informations properly
792 . Lesson 7: Using internal timers
793 . Lesson 8: Handling errors through exceptions
794 . Lesson 9: Exchanging simple data
795 . Lesson 10: Remote Procedure Calling (RPC)
796 . Lesson 11: Explicitely waiting for messages
797 . Recapping of message passing features in GRAS
798 - A HOWTO section containing:
799 o HOWTO design a GRAS application
800 More are due, of course. They will come latter. In the meanwhile, you can
801 check the examples which are still here.
803 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
805 SimGrid (3.1) stable; urgency=high
809 There was a stack corruption somewhere, visible only when optimizing
810 with these versions. [Vince]
813 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
814 module that enabled to play with Directed Acyclic Graphs. It is built
815 directly on top of SURF and provides an API rather close to the old
816 SG. Some old codes using SG are currently under rewrite to check that
817 all needful functions are provided. [Christophe Thiery]
820 * Complete rewrite of the KCCFLN05 workstation model. It is now an
821 extension of the classical CLM03 model that gracefully handles
822 failures. This is now the default model for MSG and GRAS. It doesn't
823 handle parallel tasks yet though. [AL]
824 * Bug fix: Weights were not correctly set in the network part.
825 WARNING: This may have resulted in incorrect results with simulations
826 where there are more than one flow on a given link. [AL]
829 * After a (long ?) discussion on simgrid-devel, we have decided that the
830 convention we had on units was stupid. That is why it has been decided
831 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
832 WARNING : This means that all previous platform files will not work as
833 such with this version! A warning is issued to ask users to update
835 A conversion script can be found in the contrib module of the CVS, under
836 the name contrib/platform_generation/surfxml_update.pl [MQ]
839 * Bug fix: Processes were started in reverse order, wrt deployment file.
840 WARNING: if your code relies on this bug, please fix it. [AL]
841 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
842 being executed on two different locations. [AL]
843 * Bug fix: Failures are now better supported thanks to Derrick's tests
844 (there was many failure situations I hadn't thought of and that weren't
845 correctly handled). [AL]
846 * New function: MSG_host_is_avail indicates you whether a given m_host_t
850 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
851 Exception occurring on server-side are propagated back to client (!).
853 API CHANGE: the callback changed their prototype. Change:
854 int my_handler(gras_socket_t expeditor, void *payload_data) {
856 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
857 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
859 * New! function: gras_msg_handleall to deal with all messages arriving
860 within a given period.
861 * New! function: gras_socket_server_range to get a server socket in a
862 range of port numbers (ease to avoid port number conflicts) [MQ]
863 * New! gras processes display their backtrace when they get a SIGUSR1
864 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
865 Sweet to debug RL processes [MQ]
869 - Do not force experiment sizes to be expressed in kb, or it becomes
870 impossible to measure the latency this way (needs one byte-long tests)
871 WARNING: this changes the amok_bw_* function semantic. [MQ]
872 - Implements the link saturation stuff. [MQ]
873 * Peer management module:
874 New! module factorizing code that we wrote over and over [MQ].
877 * New module: cunit (my jUnit implementation in ansi C) [MQ]
878 - Test units are placed directly into the library code, they get extracted
879 automatically and placed into the src/testall binary.
880 - Convert most of the XBT tests to this system.
881 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
882 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
883 live in the same C code. This required to change a little bit the API
884 of surfxml parsing but shouldn't be an issue for end-users. [AL]
885 * New module: sparse graph structure with basic algorithms (this is work
886 in progress and the API is not considered to be frozen yet). [AL]
887 * Display more information on backtraces: source line & function names are
888 now displayed just like valgrind does (rely on addr2line tool) [MQ]
889 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
890 * Reworked a little bit some #include statements to load only required
891 headers. Some user code that relied on SimGrid to include stdlib or
892 stdio may need to include it by themselves. [AL]
893 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
895 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
896 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
897 but I guess that almost nobody used those parts. [MQ]
899 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
901 SimGrid (3.0.1) stable; urgency=low
904 * Unfortunately, I had missed 5 misnamed functions:
905 xbt_fifo_item_t xbt_fifo_newitem(void);
906 void xbt_fifo_freeitem(xbt_fifo_item_t);
907 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
908 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
909 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
910 They're now deprecated. Please use their new versions:
911 xbt_fifo_item_t xbt_fifo_new_item(void);
912 void xbt_fifo_free_item(xbt_fifo_item_t);
913 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
914 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
915 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
917 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
918 * Documentation: xbt_log module unmercifully reworked [MQ]
919 * Bugfix: there was a problem with the ending of contexts with
920 the pthread backend. It caused some weird deadlock or behavior
921 depending on the pthread implementation. [AL]
922 * Bugfix: get the exceptions raised in the simulator repport where
923 and why they come from when they are not catched in time [AL, MQ]
926 * Bugfix: Do repport the error when two non-connected hosts try to
927 exchange data (Thanks to Flavien for stumbling into this one) [AL]
930 * Add additionnal checkings on communications. Assert that two
931 communicating hosts are connected by a set of links... [AL]
934 * Add additionnal checkings on channel values in communication [AL]
935 * New: MSG_task_get_source to see on which host a task was generated [HC]
936 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
937 the number of tasks waiting to be received on channel and sent
939 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
940 waits for the first task coming from a given host.. [AL]
942 GRAS new functionnalities: [MQ]
943 * Enhance the parsing macro to allow the size of multidimentional objects
944 to be given thru annotations.
945 * New example (and documentation): Matrix Multiplication a la RPC
946 (as when I was young!) and fix a bunch of bugs found on the way.
948 GRAS performance improvements: [MQ]
950 * Reduce the amount of cbps creation/destruction by making it static to
951 datadesc_send/recv() and using a (newly created) cbps_reset (based on
954 * Change libdata to a set so that we can search for stuff by ID (and thus
955 reduce the insane amount of dict lookups)
958 * Actually implement gras_datadesc_copy() so that we don't have to mimick
959 RL communication on top of SG since it's so uneffective.
960 It may also be used for inter-thread communication in RL, one day.
961 * Use gras_datadesc_copy() to exchange messages on top of SG
963 - improve message exchange performance on top of SG
964 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
965 * Don't exchange on the network the size of the used part of buffer,
966 instead, specify the possible buffer size to read().
968 - reduces the amount of read/write calls (one pair per exchange)
969 - reduces the amount of exchanged data (the size)
970 - allows to retrieve all arrived data on receiver side, if we don't need
971 it right now (subsequent read will peek the buffer)
972 - allows the receiver to proceed with the begining of the stream before
973 everything is arrived
974 - make it possible to build an iov transport (using readv/writev)
976 - take care of the data with non-stable storage (like stacked data),
978 * If possible, TCP send uses vector I/O (when writev() is here)
979 - Don't use it for receive since we send data sizes and data on the
980 same stream, so we wouldn't be able to chain large amount of chunks
981 before having to flush the stuff to read the size.
982 * Rework the transport plugin mecanism to simplify it and reduce the
983 amount of pointer dereferencement when searching for the right function
986 * I guess that now, we do almost as few system calls as possible while
987 doing as few data copy as possible.
989 To improve it further, we could try to send all the sizes first and then
990 all the data (to use iov on receiving size), but it's only a partial
991 solution: when you have 2 dimensional data, the sizes of the second
992 dimension is data of the first dimension, so you need 3 streams.
994 I'm not sure the potential performance gains justify the coding burden.
996 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
998 SimGrid (3.00) stable; urgency=high
1001 * New! Give the possibility to hijack the surf parser and thus bypass
1002 MSG_create_environment and MSG_launch_application. Have a look at
1003 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1005 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1007 SimGrid (2.96) unstable; urgency=low
1012 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1013 This deprecates the xbt_error_t mecanisms.
1014 It modifies (simplifies) all XBT and GRAS API.
1015 MSG API keeps unchanged (exceptions raised by XBT are catched from
1016 within MSG and masked with existing error handling facilities)
1019 * New! Add a FATPIPE model. [AL]
1020 * New! Add a parallel task model. [AL]
1021 * New! Add automatically a loopback interface (in the default
1022 network model) if none was precised.
1025 * Bugfix: MSG_process_resume now works with the current running process.
1027 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1028 * Modification of MSG_task_get_compute_duration. Once a task has been
1029 processed, the value returned by this function is now equal to 0. [AL]
1030 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1031 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1032 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1033 of a m_task. For now, MSG_task_cancel only works with computation
1035 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1036 of the processor (in Mflop/s) regardless of the current load on the
1038 * API Change: use proper naming convention for MSG_getClock and
1039 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1041 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1042 This function changes the priority of a computation task. This priority
1043 doesn't affect the transfer rate. A priority of 2 will make a task
1044 receive two times more cpu power than the other ones. This function
1045 has been added to suit the needs of Nguyen The Loc and hasn't been that
1046 much tested yet. So if it fails, please report it and send me your code.
1048 * API Change: removed all functions and types that were marked "deprecated"
1049 since many months. Renamed MSG_global_init_args to MSG_global_init.
1051 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1053 SimGrid (2.95) unstable; urgency=low
1056 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1057 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1058 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1061 * Bugfix: Allow absolute paths for platform description files [MQ]
1062 * Bugfix: do free the variables after use. Leads to drastic performance
1064 * Implement max_duration (ie, timeouts) on resources [AL]
1067 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1069 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1070 the needed support (timeouts) [AL]
1073 * Implement measurement sockets. You can now get the bandwidth between two
1074 hosts thanks to AMOK (see below). [MQ]
1075 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1076 dynar over the network (yeah) [MQ]
1077 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1078 * simple/static token-ring example. [Alexandre Colucci and MQ]
1079 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1080 in place before [MQ]
1083 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1084 type parsing. "short" and "long" modifiers were also ignored; other
1085 modifier (such as reference level) are still ignored. [MQ]
1086 * Embeed the buffer size within the buffer itself on SG. [MQ]
1087 That way, send() are atomic and cannot get intermixed anymore (at least
1088 the ones which are less than 100k; bigger messages still have the issue)
1089 * Array size pushed by the field, not by the field type (or each
1090 and every long int will push stuff to the cbps) [MQ]
1091 * use select() to sleep since it allows to portably sleep less than one
1094 GRAS (minor cleanups)
1095 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1098 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1099 * type category 'ignored' killed as it was never used and were difficult
1101 * whether a type can cycle or not is now a flag, leaving room for more
1102 flags (as "ignored", if we feel the need one day ;)
1103 * Rename raw sockets to measurement sockets since "raw" has another
1104 meaning in networking community.
1107 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1108 and offers features not belonging to GRAS but that most applications
1109 need. One day, it may be a set of plugins loadable at runtime.
1110 * New module: bandwidth
1111 bandwidth measurement between arbitrary nodes running this module. [MQ]
1113 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1115 SimGrid (2.94) unstable; urgency=low
1117 The first beta release of SimGrid 3 !
1121 * Update the main page and the FAQ. Adding references to gforge.
1124 * Add a gras_os_getpid function.
1127 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1128 * Extend the logs so that they also print PID, hostname, date, ... if
1130 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1131 and kill the old version which were in testsuite/
1132 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1136 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1137 for users. Instead, each programming environment (GRAS, MSG,...) use it
1138 when needed to provide such a feature to users.
1139 Don't shortcut the mecanism or you will also shortcut the virtualization
1140 you need on the simulator.
1144 * Cleanups in configury with regard to compile optimization/warning flags.
1145 Also add -fno-loop-optimize to any powerpc since it's the optimization
1146 killing gcc (< 3.4.0).
1147 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1148 needed by MSG examples complications
1149 * Borrow configury beautifications from PHP
1152 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1153 warning (thanks loris for stumbling into this one).
1154 * Bugfix: stop loading private headers (gras_config.h) from the public
1158 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1159 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1160 RL would imply a DNS resolver, which is *hard* to do in a portable way
1161 (and therefore delayed).
1162 * Implement a real timer mecanism and use it in timing macros. This allows
1163 to avoid rounding errors and get a 0.000005 sec precision in timing
1164 macros. While I was at it, various cleanups:
1165 - allow to declare more than one timed section per file (fix a stupid bug)
1166 - move some private declaration to the right place
1167 - merge conditional execution and timing macros into emulation module
1168 - document the module
1169 - make sure the module cleanups its mess on gras_exit
1170 * Documentation improvements:
1171 - (new) how to compile applications using GRAS
1172 - (new) emulation support (timing macros)
1174 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1176 SimGrid (2.93) unstable; urgency=low
1178 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1181 - Use Paje properly where used. Still to be sanitized properly.
1182 - Portability fix: Add an implementation of the contexts using pthread
1186 - Add xbt_procname(): returns the name of the current process.
1187 Use it to show the current process's name in all logging.
1189 - fix detection of older flex version and the reaction, since we do
1190 depend on modern ones (we use lex_destroy)
1191 - Better separation of SG and RL in the libs: remove all simulation code
1192 from libgras. As a result, this lib is now only 200k when stripped.
1193 Some of the xbt modules may also be duplicated (two sets and such) and
1194 should be cleaned/killed before SG3.
1195 - Insist on using xlC on AIX because of weird problems involving gcc there.
1196 - Cleanup the make remote stuff. This is now done by scripts
1197 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1198 mainly for our private use, but we're working on changing them to user
1201 - Bugfix: flush the socket on close only if there is some *output*.
1202 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1204 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1207 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1209 SimGrid (2.92) unstable; urgency=low
1211 Alpha 3 on the path to SimGrid 3
1215 - New! First try of benchmarking macros.
1216 - New! First try so that gras_stub_generator generate deployment and
1217 remote compilation helpers.
1219 - Bugfix: Initialization fix in msg_test.
1223 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1225 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1227 - API Change: gras_os_sleep to take the amount of seconds as a double.
1228 Accepting an int was error prone since it was the only location where
1229 seconds were coded as such. It leaded to damn rounding errors.
1230 - Bugfix: Hard to belive that timers ever worked before this.
1232 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1234 SimGrid (2.91) unstable; urgency=low
1236 Alpha 2 on the path to SimGrid 3
1240 - Bug fix in the lmm_solver.
1242 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1243 through the function MSG_paje_output.
1244 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1245 - It is possible to bound the rate of a communication in MSG with
1246 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1247 to write it in the changelog).
1248 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1253 - Various cleanups to the autotools stuff
1254 - Begin to move Gras examples to examples/gras/
1255 - Let make distcheck work again (yeah!)
1257 - documentation overhauled using doxygen.
1258 gtk-doc-tools is dead in SimGrid now.
1259 - Automatically extract all existing logging categories, and add the list
1260 to the documentation (long standing one, to say the less)
1262 - Cleanup the known architecture table. Reorder the entries to group what
1263 should be, and use a more consistent naming scheme.
1264 (some of the test dataset are still to be regenerated)
1265 - New! Allow library to register globals on each process just as userdata
1267 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1268 lookup time (for now).
1269 Use it in msg and trp.
1270 This cleans a lot the internals and helps enforcing privacy of the
1271 headers between the gras components.
1272 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1273 - Bugfix: gras_os_time was delirious in RL.
1274 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1276 - Reenable GRAS now that it works.
1278 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1280 SimGrid (2.90) unstable; urgency=low
1282 Alpha 1 on the path to SimGrid 3
1284 * It is a long time since the last release of SimGrid. I'm sorry about
1285 that but as I had told you, I was rewriting a lot of things. I apologize
1286 to those who had been reporting bugs to me and that I had not answered.
1287 If your bug is still in the new version, please tell me. Here is a
1288 summary of the main changes.
1290 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1291 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1292 portability, speed, and a lot more but you'll figure it out later.
1293 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1294 I wanted to release the new MSG as soon as possible and I have
1295 broken GRAS, which is the reason why, for now, only MSG is fully
1296 functional. A laconic description of these projects is available
1297 in the documentation.
1299 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1300 kernel optimized for our needs (called SURF but only the developers
1301 should use it). Hence, MSG is now roughly 30 times faster and I think
1302 that by rewriting a little bit MSG, I could event speed it up a little
1303 bit more. Beside the gain in speed, it is also much easier to encode a
1304 new platform model with SURF than it was with SG. More to come...
1306 * REVOLUTION 3: I have tried to change a little as possible the API of
1307 MSG but a few things really had to disappear. The main differences
1308 with the previous version are :
1309 1) no more m_links_t and the corresponding functions. Platforms are
1310 directly read from a XML description and cannot be hard-coded
1311 anymore. The same format is used for application deployment
1312 description. The new format is described in the documentation.
1313 Have a look in tools/platform_generation. There is a tiny script
1314 that converts from the old platform format to the new one. Concerning
1315 the application deployment format, parsing the old one is tricky.
1316 I think most of you should however be able to convert your files. If
1317 it is really an issue, I can write a C code that does the conversion.
1319 2) the toolbox tbx does not exist anymore. We now have a library
1320 with much more data-structures but without the hash-tables (we have
1321 dictionaries that are much faster).
1323 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1325 *****************************************************************************
1326 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1327 * don't feel like converting the oldies. (Mt) *
1328 *****************************************************************************
1331 Version 2.90: "the long awaited one"
1332 - Finished rewriting and debugging MSG. Rewrote the documentation.
1333 - disable GRAS for now since it needs to be ported to the newest SG
1336 - Finish the port to windows (using mingw32 for cross-compile)
1339 - Main loop and datastructures of SURF. A cpu resource object is
1340 functional. Surf can thus be used to create cpu's with variable
1341 performance on which you can execute some actions.
1343 2004-11-15 Martin Quinson
1344 - Port to ARM. Simply added the alignment and size descriptions. Should
1345 work, but the ARM machines are so slow that I didn't had the opportunity
1346 to 'make check' over there yet.
1348 2004-11-15 Arnaud Legrand
1349 - Trace manager now written. It uses a heap structure and is therefore
1350 expected to be efficient. It may however be speeded up (particularly
1351 when many events occur at the same date) by using red and black
1352 trees. One day maybe...
1353 - Max-min linear system solver written. It uses a sparse matrix
1354 structure taking advantage of its expected use. Most operations are
1355 O(1) and free/calloc are called as few as possible. The computation of
1356 the minimum could however be improved by using a red and black tree
1359 2004-11-03 Arnaud Legrand
1360 - Rename every gras_* function that was in xbt/ to its xbt_
1362 - Add a heap and a doubly-linked list to xbt
1363 - Added a dichotomy to the dictionaries. make check works as well before
1364 so I assume that the patch is correct. I do not know however if things
1365 run effectively faster than before now. :)
1367 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1368 SimGrid, and the version number is bumped to 2.x
1370 2004-10-29 Martin Quinson
1371 - Introduction of the remote errors.
1372 They are the result of a RMI/RPC on the remote machine.
1373 ErrCodes being scalar values, you can't get the host on which those
1374 errors did happen. Extending the error mechanism as in Gnome is possible.
1375 No idea yet whether it is a good idea.
1377 2004-10-28 Martin Quinson
1378 - Interface revolution: the Starred Structure Eradication.
1379 I used to do typedef struct {} toto_t; and then handle *toto_t.
1380 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1381 - ???_t is a valid type (builded with typedef)
1382 - s_toto_t is a structure (access to fields with .)
1383 - s_toto is a structure needing 'struct' keyword to be used
1384 - e_toto_t is an enum
1385 - toto_t is an 'object' (struct*)
1387 typedef struct s_toto {} s_toto_t, *toto_t;
1388 typedef enum {} e_toto_t;
1389 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1390 s_toto_t) is private.
1392 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1393 so that it can set the variable to NULL. It was so for dicts and sets,
1394 it changed for dynars.
1396 - Fix a bunch of memleaks in dict_remove
1397 - Fix a bug in sg/server_socket opening: it failed all the time.
1399 2004-10-07 Martin Quinson
1400 - Speed up dynar lookup operation a bit.
1402 gras_dynar_get is dead.
1404 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1405 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1406 (which gives you the address of the stored data).
1408 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1409 copy of the data using an affectation to do the job and not a memcpy.
1411 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1412 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1414 It does not really speedup the dynar test because they are
1415 setting elements all the time (and look them seldom). But the dict does
1416 far more lookup than setting.
1418 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1420 2004-10-05 Martin Quinson
1421 - Allow to (en/dis)able the cycle detection at run time.
1423 Whether we should check for cycle or not is now a property of each
1424 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1425 datadesc_cycle_unset allow to remove this property when previously set.
1427 Note that the cycle detection is off by default since it impacts the
1428 performance. Watch the data you feed GRAS with ;)
1430 This property is hereditary. Any element embedded in a structure having it
1431 set have it set for the time of this data exchange.
1433 You should set it both on sender and receiver side. If you don't set it on
1434 sender side, it will enter an endless loop. If you forget on receiver
1435 side, the cycles won't be recreated after communication.
1437 - Header reorganization.
1438 Kill gras_private.h, each submodule must load the headers it needs.
1440 2004-10-04 Martin Quinson
1441 - Interface revolution: do not try to survive to malloc failure.
1443 Now, gras_malloc and friends call gras_abort() on failure.
1444 As a conclusion, malloc_error is not a valid error anymore, and all
1445 functions for which it was the only gras_error_t return value are
1446 changed. They now return void, or there result directly.
1447 This simplify the API a lot.
1449 2004-09-29 Martin Quinson
1450 - Re-enable raw sockets.
1451 Created by gras_socket_{client,server}_ext;
1452 Used with gras_raw_{send,recv}
1455 It should allow to kill the last bits of gras first version soon.
1457 This is not completely satisfactory yet (duplicate code with
1458 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1461 - Simplify transport plugin (internal) interface by not passing any
1462 argument to _server and _client, but embedding them in the socket
1465 2004-09-28 Martin Quinson
1466 - Finish the port to AIX.
1467 autoconf was my problem (segfault within the malloc replacement
1468 function. No idea why)
1470 2004-09-16 Martin Quinson
1471 - Fix some size_t madness on 64bit architectures.
1473 2004-09-08 Martin Quinson
1474 - Reduce the number of system headers loaded, overload some more system
1475 calls (such as malloc to cast the result of the system one, and work
1477 - Fix and reintroduce the config support
1479 2004-09-07 Martin Quinson
1480 - Source code reorganization to allow Arnaud to surf all over there.
1481 - Allow to document the logging categories.
1482 - Remove all uppercase from logging categories and useless cleanup in names.
1484 2004-08-18 Martin Quinson
1485 Version 0.6.2 (protocol not changed; API changed)
1486 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1487 gras_error_t), and NULL when not found. Functions expecting a msgtype
1488 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1489 hopefully usefull message.
1490 - Portability to prehistoric sparcs again
1492 2004-08-17 Martin Quinson
1493 Version 0.6.1 (protocol not changed; ABI not changed)
1494 - prealloc some buffers to speed things up
1496 2004-08-11 Martin Quinson
1497 Version 0.6 (protocol not changed; ABI expended)
1498 - The parsing macro can deal with the references, provided that you add
1499 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1501 2004-08-09 Martin Quinson
1502 Version 0.5 (protocol not changed; ABI changed)
1503 - Allow to off turn the cycle detection code in data exchange at
1504 compilation time. It should be at run time, but I'm short of time (and
1505 the config stuff is still broken). That way, we keep dict out of the
1506 critical path, which is good because the performance is poor:
1507 - search not dichotomial yet
1508 - dynar give no way to access their content and memcpy everytime
1509 - In composed data description (struct, ref and so on), stop foolness of
1510 keeping the subtype's ID, but store the type itself. This keeps sets out
1511 of the critical path, which is good since they rely on dynar and
1512 dictionnaries. The only loose of that is that we cannot detect the
1513 redeclaration of a structure/union with another content (but I'm not sure
1514 the code detected well this error before anyway). We still can detect
1515 the redefinition discrepancy for the other types.
1516 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1517 it breaks the code because of type-punning used all over the place).
1518 This breaks on all non-gcc architectures (for now).
1520 All those changes (plus the buffer of last time) allow me to gain 2 order
1521 of magnitude on cruel tests consisting of 800000 array of integers on two
1522 level of a hierarchical structure (200 secondes -> 4 secondes)
1525 - the selector of reference must now return the type it points to, not
1526 the ID of this type.
1528 2004-08-06 Martin Quinson
1529 Version 0.4 (protocol changed; ABI not changed)
1530 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1531 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1532 - Implement a buffer transport, and use it by default (it relies on tcp in
1533 real life and on sg in simulation).
1534 That's a bit hackish since I had a new field to the structure to store
1535 its data without interfering with the subtype ones. Inheritance
1536 is tricky in C. And that's a kind of reverse inheritance with one class
1537 derivating two classes. Or maybe a game with java interfaces. Anyway,
1538 that's damn hard in C (at least).
1539 Moreover, I got tired while trying to ensure plugin separation and
1540 genericity in SG mode. MSG wants me to do weird things, so let's go for
1541 cruel hacks (temporarily of course ;).
1542 See comment in transport_private.h:71
1543 - do not include all the _interface headers in private but in the files
1544 which really need them (to cut the compilation time when they are
1547 2004-07-26 Martin Quinson
1548 Version 0.3 (protocol not changed; ABI changed)
1549 - Major overhault of the datadesc interface to simplify it:
1550 - shorted the function names:
1551 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1552 - add a trivial way to push/pop integers into the cbps without malloc.
1553 This allows to make really generic sub_type description, which simply
1554 pop their size of the stack.
1555 - add a function gras_datadesc_ref_pop_arr() which does what users want
1556 most of the time: Declare a dynamic array (which pops its size of the
1557 stack) and declare a reference to it. Poor name, but anyway.
1558 - kill the post-send callback, add a post-receive one
1560 2004-07-23 Martin Quinson
1561 Version 0.2 (protocol changed; ABI changed)
1562 - add some testing for cpbs in the test cases, and fix some more bugs.
1563 This invalidate again the little64 data file, since I cannot regenerate
1565 - remove an awfull optimization in the logging stuff, allowing me to:
1566 - understand it again
1567 - learn gcc how to check that the argument match the provided format
1568 - fix all errors revealed by gcc after that
1569 - internal keys of dict are not \0 terminated. Deal with it properly in
1570 loggings instead of segfaulting when the user want to see the logs :-/
1572 2004-07-22 Martin Quinson
1573 - Fix some stupid bug preventing cbps (callback postit) from working
1575 2004-07-21 Martin Quinson
1576 - Some documentation cleanups
1577 - remove the useless last argument of msgtype_declare
1578 - rename the Virtu functions to fit into the 'os' namespace
1579 - move headers src/include -> src/include/gras/ and stop fooling with
1580 gras -> . symbolic link
1581 - make distcheck is now successful
1583 2004-07-19 Martin Quinson
1585 - Build shared library also
1586 - Install html doc to the right location
1587 - stop removing maintainer files in make clean
1588 - build tests only on make check
1590 2004-07-13 Martin Quinson
1592 - No major issue in previous version => change versionning schema
1593 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1594 corresponding dataset.
1596 2004-07-11 Martin Quinson
1598 - Get it working with any kind of structure (we can compute the padding
1599 bytes remotely for all the architectures I have access to)
1600 - Implement the structure parsing macro (still not quite robust/complete)
1601 - Improvement to the remote testing toysuite
1603 2004-07-10 Martin Quinson
1604 [autoconf mechanism]
1605 - get ride of a bunch of deprecated macros
1606 - actually run the test for two-compliment, not only compile it :-/
1607 - test whether the structures get packed (and bail out if yes. Damn.
1608 Alignment is a serious matter)
1609 - test whether the structures get compacted (but respecting the alignment
1610 constraints of each types)
1611 - test whether the array fields of structures can straddle alignment boundaries
1613 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1614 (took me 2 hours to find that bug, looking at the wrong place)
1615 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1616 before sending/receiving and are used to compute the offsets of fields
1617 - Given that padding size depend even on compiler options, keep track of
1618 alignment and aligned_size only for the current architecture. Not a big
1619 deal since we send structure fields one after the other (seems
1621 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1622 program, let it work on linux/gcc/little32. portability todo.
1624 2004-07-08 Martin Quinson
1625 - import and improve remote compilation support from FAST
1626 - make sure make check works on half a dozen of machines out there
1628 2004-07-07 Martin Quinson
1629 Let's say it's version 0.0.3 ;)
1630 - Implement conversions (yuhu!)
1631 - Let it work on solaris (beside conversion, of course)
1632 - Stupid me, using rand() to generate the conversion datatests in not wise.
1634 2004-07-06 Martin Quinson
1635 - Let make dist work, since I'm gonna need it to compile on remote hosts
1636 - Let Tests/datadesc_usage write the architecture on which the file was
1637 generated as first byte.
1638 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1639 architecture descriptions.
1640 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1641 on those architectures.
1642 - Optimization: send/recv array of scalar in one shoot
1644 2004-07-05 Martin Quinson
1645 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1647 - Plug a whole bunch of memleaks
1648 - each process now have to call gras_{init,exit}. One day, their log
1649 settings will be separated
1650 - Continue the code factorisation between SG, RL and common in Transport.
1652 2004-07-04 Martin Quinson
1654 - Redistribution between SG and RL.
1655 We wanna have to accept in SG, so move accepted related parts of RL in
1656 the common part. (more precisely, the dynar of all known sockets is no
1657 more a static in transport.c, but part of the process_data)
1659 [gras_stub_generator]
1660 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1662 2004-07-03 Martin Quinson
1663 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1664 them all up in one shot)
1666 - Fix the ugly case of reference to dynamic array.
1667 - New (semi-public) function gras_datadesc_size to allow the messaging
1668 layer to malloc the needed space for the buffer.
1670 - gras_socket_close now expect the socket to close (and not its address to
1671 put NULL in it after it). This is because the socket passed to handlers
1672 is one of their argument (=> not writable).
1674 - propagate the interface cleanup from last week in datadesc, ie remove a
1675 superfluous level of indirection. User pass adress of variable
1676 containing data (both when sending and receiving), and not of a variable
1677 being a pointer to the data. Let's say that I like it better ;)
1678 The price for that is constructs like "int msg=*(int*)payload" in
1679 handlers, but it's a fine price, IMHO.
1681 - Let it work in RL (yuhu)
1683 2004-06-21 Martin Quinson
1685 - porting SG plugin and SG select to new standards (works almost).
1686 - plug memleaks and fix bugs around.
1689 - cleanup the prototype of data recv and force users to specify when they
1690 want to handle references to objects. Test case working even for cycles.
1691 - plug memleaks. Valgrind is perfectly ok with this.
1693 2004-06-12 Martin Quinson
1695 - cleanup the separation between plugin and main code in plugin creation
1697 2004-06-11 Martin Quinson
1699 - Reput hook for raw sockets, needed for BW experiments
1700 - kill a few lines of dead code
1701 [Data description] Interface cleanup
1702 - gras_datadesc_by_name returns the searched type or NULL.
1703 That way, no variable is needed to use a type desc once, which makes
1705 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1706 two parameters were strings (field name, type name), leading to
1708 [Dicos] Interface cleanup
1709 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1710 This is consistant with the dynar API.
1712 2004-04-21 Martin Quinson
1714 - Porting to new standards.
1716 - interface cleanup.
1717 There is no bag anymore, no need to take extra provision to mask the
1718 pointers behind "ID".
1719 Better splitup of functions between files create/exchange/convert.
1720 This is still a bit artificial since convert and receive are so
1721 interleaved, but anyway.
1723 - add a queued message list to procdata (the ones not matching criteria
1725 - factorize some more code between SG and RL wrt procdata
1727 - use gras_exit in example to track memleaks
1728 - get rid of gs_example now that GS is properly integrated into gras
1729 - update run_test to integrate the lastest tests (datadesc)
1731 - rename WARNINGn macros to WARNn since it prooved error-prone
1733 2004-04-19 Martin Quinson
1735 - register init/exit functions within gras module mechanism
1736 - send/receive function.
1737 Convertion is not implemented, but short-cutted if not needed.
1738 struct/array elements are sent one by one (instead of block-wise), but
1739 nobody really cares (yet). Get a prototype before optimizing.
1740 - tests (using a file socket) for DD send/receive on:
1741 - base types: int, float
1742 - array: fixed size, string (ie ref to dynamic string)
1743 - structure: homogeneous, heterogeneous
1744 - chained list, graph with cycle
1745 Believe it or not, valgrind is not too unhappy with the results. The
1746 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1750 - Bugfix when using a filename explicitely (instead of '-')
1752 2004-04-09 Martin Quinson
1754 - factorize more code between RL and SG in socket creation
1755 - Complete the implementation and tests of:
1757 o file (only in RL, and mainly for debugging)
1759 I lost 3 days to design a portable address resolver, and then decided
1760 that the prototype mainly have to run on my box.
1761 Addressing portability too early may be like optimizing too early :-/
1763 - use gras_init in the Tests instead of the crappy parse_log_opt
1764 (the latter function is removed)
1765 [Conditional execution]
1766 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1767 [Code reorganisation]
1768 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1769 Build examples against the RL library, since there is no way to disable
1770 its creation for now.
1772 For information, the beginning of coding on GRAS was back in june
1773 2003. I guess that every line has been rewritten at least twice since