1 SimGrid (3.5) unstable; urgency=low
4 * New data container: setset (set of sets of elements)
5 * New module: mmalloc (mapped malloc)
6 * New function: xbt_dict_cursor_set_data()
7 * New function: xbt_fifo_get_last_item()
8 * Bug fix in xbt_dynar_shrink(): use the right element size
9 * Use library init/fini functions for our initialization.
10 - you can use logs and other feature as soon as you want in your
11 code (even before the xbt_init / MSG_init)
12 - xbt_exit is now a no-op and produce a warning when used.
14 Build chain: bug fixes and further polishing
15 * Fix the 'make install' target.
16 No need to use 'make install-simgrid' anymore
17 * Introduce a 'make dist' target compiling a *source* archive
18 'make package' compiles a binary archive (not well tested)
19 * Compile java files only on need
21 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
23 SimGrid (3.4.1) stable; urgency=low
25 The "Polishing easter eggs is probably a good idea" release.
26 This is a bug fixes release only.
30 * Fix a bug preventing the tasks from begin garbage collected.
33 * Fix a bug occuring when a host involved in a communication fails.
34 This was not detected properly by the other peer involved in the
35 communication. Now, it's reported as a network error.
38 * Warn the user about loop dependencies in data flow of DAX files
39 * Obey the control-flow dependencies of DAX files
40 * Change the way a task is considered as ready. Instead of removing
41 dependencies when a task is done, a counter is decreased. This way,
42 it is always possible to reach ancestors thanks to the
43 SD_taks_get_parents function (even after the end of the simulation.)
44 * Change the level of some logs in sd_global from INFO to VERB. The
45 corresponding tests are now less verbose
48 * Add option "enable_smpi" allowing to not compile SMPI.
49 Probably useful for the (Mac) users experiencing a build error here
50 * Improve the detection of lua5.1 and ruby1.8
52 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
54 SimGrid (3.4) stable; urgency=low
56 The "Easter in Cargese" release. Also known as (major changes):
58 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
59 ~> bindings were greatly improved
60 ~> new tracing infrastructure for better visualization introduced
62 * the "Welcome to configury modernity" release.
63 ~> we switched from autotools to cmake, and improved our cdash
66 A more detailled list of changes follow (full detail in svn log).
68 Java Bindings: Various Cleanups
69 * (install java-gcj-compat-dev on debian-like to use them)
70 * Remove put/get: no need to export deprecated interface in Java
71 Use send/receive instead.
72 * Cleanup the examples and add a README per directory
73 * Remove example autoDestination (that's the only way to go now)
74 * Remove example explicitDestination (was a plain copy of basic)
75 * Make JniException a runtime exception, so that there is no need to
76 declare the fact that you may encounter such a beast. I guess that
77 nobody will ever want to survive such error.
78 * Create specific errors for each MSG case of failure:
79 host failure, transfer failure, timeout, task cancelled
80 * Cleanup the exceptions that may get thrown by each function
81 * Other internal cleanups in Java bindings. Performance still bad :/
82 Ruby and Lua Bindings: create them
83 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
84 * That's new and great, you should try them out.
85 Same functionalities than Java bindings, only even less polished
87 * Kill the useless "rate" argument of SD_task_get_execution_time()
88 Everyone used to provide -1 as a value, it was not used, and the
89 semantic of a possible use wasn't even clear.
90 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
91 either as comm costs or compute costs to mean that there is no
92 such thing for that specific task.
93 * Add a SD_task_set_name() function
94 * Fix SD_task_unschedule() on typed tasks
95 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
96 * In DAX loader, accept useless 'level' attributes to <job> since
97 LIGO DAGs have them (seem to be to ease graphical representation).
99 * Add an example masterslave_mailbox.c using send/receive and not
100 the deprecated put/get interface.
101 * Kill the MSG_paje_output() function. It's a noop since 2 years.
102 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
104 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
105 (declare MSG_USE_DEPRECATED to still have the old name)
106 * Add a MSG_task_set_data() function
107 * About trace replay (see examples/msg/actions):
109 - Allow to work with splitted trace files for each process
110 Give the specific trace file as argument of each process,
111 and call MSG_action_trace_run(NULL)
112 You can still have one merged file for all processes.
113 - Fix implementation of collective operations
114 * Allow task_execute() on 0-sized tasks (closes #10063)
116 * This is the first release of SimGrid where SMPI is not considered
117 beta anymore (even if some corners should still be improved)
118 * Port over the new SIMIX_network submodule (internal refactoring)
119 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
120 * Implement more missing elements of the standard:
122 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
124 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
125 MPI_Scan() MPI_Get_processor_name()
126 - Added implementation of missing case for Alltoall (warning: it's
127 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
128 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
129 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
130 SMPI_MPI_Allgatherv()
132 - MPI_Waitsome() was broken
133 - Allow relative includes in smpicc
134 - Command line cfg argument 'reference_speed' was ignored...
135 - Some functions did not properly lead to auto-benching of user code
136 - smpicc passes -O2 by default (just like openmpi one)
138 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
139 * Bug fixes about timeouts during communications
140 * add SIMIX_message_sizes_output() as a pimple to write to file the
141 amount of messages per size. Use gnuplot to get histogram.
142 Pimple because that's the only user-visible function of simix,
143 defined directly in xbt.h (irk, sorry)
145 - Add a SIMIX_sem_get_capacity() function
146 - Fix interactions with processe resume/suspende
147 - release_forever() was stupidly broken
148 - Fix SIMIX_display_process_status() for processes in a semaphore
149 - Make SIMIX_sem_block_onto() user-visible
150 * Refactoring context stuff:
151 - Use pseudo-OOP for better modularity
152 - reimplement SIMIX_process_kill() without process_schedule() so
153 that the latter can take as invariant that it is called from
155 - Merge context_start into context_new for sake of simplicity
157 * Add a Vivaldi network model, coded live during SUD'10 ;)
158 * Rename configuration variables to start a hierarchy:
159 o cpu_model -> cpu/model
160 o network_model -> network/model
161 o workstation_model -> workstation/model
162 * New configuration variables:
163 o network/bandwidth_factor: correction to bandwith
164 o network/latency_factor: correction to latency
165 o netwotk/weight_S: correction to the weight of competing streams
166 * Add a long description to the models, that users can see with such
167 argument on the command line: --cfg=cpu/model:help
168 * --help-models display the long description of all known models
170 * config: add the ability to set a default value after registration
171 Does not override any previously set value (e.g. from cmd line)
172 * dict: allow to have integer key and data.
173 When so, you need to use the following functions
174 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
175 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
176 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
177 In contrary to regular dicts, the key is not malloced before copy.
178 Mixing scalar and regular elements in the same dict is not tested
180 * Allow to use xbt_dynar_shrink() to expend the dynar instead
181 Tracing for Visualization:
182 * SimGrid is now instrumented in order to generate a trace file for
183 visualization analysis: to use it, need to compile SimGrid with the
184 "tracing" option enabled, and instrument the program using SimGrid with
185 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
186 (among other functions).
187 * The instrumentation only traces the platform utilization for now
188 * Documentation to use the tracing functions and how to analyze the
189 traces with the Triva tool is written.
190 * More information about: SimGrid FAQ (in the section Tracing Simulations
193 * We moved to cmake as default build system. Autotools support will
194 be dropped soon. Check the FAQ for more info about how to use it.
195 * Greatly improved our cdash/ctest interactions
196 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
197 * Added memory checking tests with valgrind; lot of memleak fixing.
198 This may be the first release of simgrid with so few memory issues
199 * Added code coverage tests.
200 Our coverage is still improvable, but at least we see it on cdash.
202 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
204 SimGrid (3.3.4) stable; urgency=low
206 The "Desktop Grid needs love too" release (also called Xmas release).
209 * Major speedup in the maxmin system solving by using lazy evaluation
210 Instead of solving completely the maxmin system at each iteration,
211 only invalidate (and recompute) the modified parts.
212 This new feature is enabled in default models but you can try to
213 turn it on with "--cfg:maxmin-selective-update=1" for other models.
214 * Cas01 IMproved as default CPU model
215 This CPU model is the same Cas01 model, but it uses the
216 maxmin-selective-update flag and a heap structure to manage
217 actions on SURF kernel.
218 It reduces the complexity to find the next action to finish and,
219 consequently, it's faster than the old Cas01.
220 This is the new default CPU model (Cas01).
221 * Rename the old Cas01 model to Cas01_fullupdate
222 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
223 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
224 * CpuTI (CPU Trace Integration)
225 A new CPU model whose objective is simulate faster when using
226 availability trace files.
227 Instead of using a full featured, over engineered maxmin system for
228 CPU modeling, this model does the pre-integration of traces files
229 to calculate the amount of CPU power available, and so, executes
230 faster than the old CPU models.
231 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
232 * Use LV08 as default network model since it gives better accuracy
233 for small messages and shouldn't change things for big ones.
234 Use --cfg=network_model:CM02 to get the previous behavior.
237 ******************************************
238 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
239 ******************************************
240 * The new CPU model may changes simulations!
241 The point is that events occurring at the exact same timestamp
242 are not scheduled in the same order with the old and new
243 version. This may be enough to completely change the execution
244 of simulations in some cases.
245 * The new network model will change simulations!
246 This new model is more realistic than the previous one, so you
247 should consider redoing your old experiments with this model.
248 Sorry for the inconvenience.
251 * Introduce the supernovae compilation mode
252 When compiled that way, the whole SimGrid (or almost) is put in a
253 single compilation unit and compiled in one shoot.
254 This is to help gcc which has difficulties to inline stuff from one
256 The speedup seem to be above 15%, althrough more tests are needed on
257 amd64 to confirm that gain.
260 * Port of MSG's mailbox on top of SIMIX network
261 The put/get mechanism was greatly simplified on the way.
264 * New SIMIX network module. Provides:
265 - Mailbox: rendez-vous mecanism to find with who you want to speak
266 - Synchronous send/recv: easier and hopefully faster since the
267 logic is handled in the maestro process directly now
268 - Asynchronous send/recv: you dreamt of it? It's here now
269 Too bad that nobody cared enough to propagate the change to MSG.
270 * Add semaphores as SIMIX synchronization mechanism.
273 * new function SD_daxload(char*) to load a DAX file
274 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
275 * Introduce typed tasks. Specify its kind and cost at creation.
276 At scheduling, just give where it should be placed, and the cost
277 for each involved resource is automatically computed.
278 Existing constructors so far (more to come of course):
279 - SD_task_create_comm_e2e() for end-to-end communication
280 - SD_task_create_comp_seq() for sequential computation
281 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
282 * new function SD_task_dump() for debuging display
283 * new function SD_task_dotty(task,FILE*) writing to file the info
284 about the task in dotty format
285 * SD_task_dependency_exists() can now cope with having one of its
286 arguments NULL. If so, it tests whether the other argument has any
288 * Add getters on list of preceding/following tasks:
289 SD_task_get_parents(task) and SD_task_get_children(task)
290 * Add getters on amount of workstations and list:
291 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
292 * Add getter on task kind: SD_task_get_kind(task)
293 * Update the start_time and finish_time of tasks on completion/failure
294 * Bugfix: Remove task from state swags when destroyed
297 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
300 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
301 dynar with one separate thread per value of the dynar.
302 * Change the prototype of xbt_thread_create(), sorry.
303 Added a boolean parameter indicating whether we want to join this
304 thread (used in SG only for now)
305 * Implement xbt_thread_join and xbt_thread_yield in SG also.
308 * GTNetS wrappers should now be usable again (and betterly tested too)
309 * Fix a major regression from 3.2 where the timeout provided to
310 MSG_task_put_with_timeout() was used as absolute time before which
311 the comm should be done.
312 * Start to fix the <cluster> tag.
313 - Internal links should be good now (beside of the loopback, which
314 use the private link instead)
315 - paths to the external world is still rather broken
316 - the <route:multi> tag is just broken. Actually that's brain-dead.
317 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
318 to make it less stupid
319 ** Check your platform with teshsuite/simdag/platforms/flatifier **
320 * Fix a source-level compatibility glitch from 3.2: after defining
321 MSG_USE_DEPRECATED, you can use the old name
322 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
323 * Allow to compile from the SVN with automake 1.11
324 * Fix some problems when using the "start_time" tag in deployment XMLs.
325 * Fix #8569: XBT/synchro.h has redundant declarations
326 * Fix #8563: MSG return values and exceptions
327 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
328 * Integrate patch #8636: Obey DESTDIR when installing documentation.
329 Thanks to Robson Peixoto.
330 * Fix a vicious bug in dictionaries inducing that some elements were
331 not freed on xbt_dict_free()
333 Portability report of this version:
334 * Main portability targets:
335 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
337 - mac leopard on i386
338 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
339 but nothing critical.
340 * Other platforms: windows, AIX and others were not tested for this release
342 Timing report of this version:
343 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
344 scenario parameters). From 8h to a few seconds in desktop grid settings.
345 * Supernovae brings about 25% speedup on i386.
347 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
349 SimGrid (3.3.3) stable; urgency=low
351 The "Need for Speed" release.
353 The timings done to validate the 3.3.2 were faulty.
354 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
356 The problem was a conversion from a manually handled vector to
357 xbt_dynar_t on the critical path.
358 xbt_dynar_foreach calls functions, inducing stack management crap.
360 We inlined these functions and xbt_dynar_foreach is now breath taking.
361 We also inlined xbt_swag_belong on the way.
363 Here are some approximate speedup measurements (on master/slaves
364 simulations lasting between 10s and 20s each):
365 3.3.1 -> 3.3.2: about same performance
366 3.3.2 -> 3.3.3: 40% speedup
367 3.3.1 -> 3.3.3: 40% speedup
368 3.3.1 with inline patch -> 3.3.3: 30% speedup
370 Our reading is that the refactoring which occurred in 3.3.2 made us
371 suffer much more from the xbt_dynar_foreach low performance, but
372 once we solved this, this refactoring proved to be very performance
373 effective. From the 40% speedup, somehow, 10% are due to the
374 inlining and 30% to the refactoring.
376 That's a pitty that gcc cannot inline functions placed in other files
377 alone. We have to choose between:
378 - break the encapsulation (by putting private data structures and
379 accessors in headers files to help gcc)
380 - live with low performance
381 - switch to a decent compiler such as icc (not quite possible).
383 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
385 SimGrid (3.3.2) stable; urgency=low
387 The "Simplicity does not preceed complexity, but follows it" release.
389 The main contributors of this release were (lexical order):
390 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
393 * Extract the routing logic into its own object.
394 (was dupplicated in network.c and workstation_LV07.c;
395 Allows to implement other ways of storing that info)
396 => kill now useless network_card concept
397 - Use dynar to represent routes (instead of void** + int*)
398 - kill link_set (use surf_network_model->resource_set instead)
399 - Add a command-line option to choose the routing schema to use
400 - Add three new models:
401 * Floyd (shortest path computed at initialization)
402 * Dijikstra (shortest path recomputed all the time)
403 * Cached Dijikstra (shortest path computed on need)
404 All these models where contributed by Silas De Munck, and are
405 described in his ICCS09 paper.
407 * Simplify model declaration
408 (less redirections, less function to write when defining a model)
409 - Factorize stuff between models:
412 surf_model_resource_set(model)
413 surf_model_resource_by_name(model, name)
414 - Unify the types of models in s_surf_model_t (using an union)
415 - Embeed fields of common_public directly into s_surf_model_t
416 - Rename model methods:
417 action_free ~> action_unref
418 action_change_state ~> action_state_set
419 action_get_state ~> action_state_get
420 - Change model methods into functions :
421 (model)->common_public->action_use ~> surf_action_ref
423 * Implement a generic resource; use it as ancestor to specific ones
424 (allows to kill duplicated code in models)
425 Drawback: timer command don't need no name nor properties;
426 workstation_CLM03 don't need no properties
427 (but I guess we can live with those few bytes wasted)
429 * Improve the action object model
430 - implement a constructor avoiding dupplicated code about field
431 initialization in generic_action part.
433 * Kill the SDP model: it has an external dependency, is deprecated
434 in flavor of modern lmm models, and didn't compile since a while
437 * Relocation of the context module from XBT to SIMIX.
438 (the context were decoupled from the simix processes, duplicating a lot of code)
439 => a lot of code was factorized
440 - less overhead is introduced during scheduling
441 - simpler API for the context factory
442 - the logic for process creation,destruction and manipulation was simplified
443 * Simplification of the s_smx_process_t data structure.
444 => accesing the simix level data associated to a process is faster now,
445 and the code is a lot more readable.
448 * Implement some more MPI primitives:
449 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
450 -implementation: Bcast: flat or 2-ary tree (default),
453 Allreduce: Reduce then Bcast
454 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
455 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
456 Alltoallv: flat tree, like ompi
458 * Add support for optimized collectives (Bcast is now binomial by default)
459 * Port smpirun and smpicc to OS X
462 * Kill SD_link_get_properties: hard to maintain and makes very little sense
463 Shout out if you used it.
466 * Display the list of still queued messages in SG mode when existing
470 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
471 * Add xbt_graph_node_get_outedges() [Silas De Munck]
472 * Add xbt_str_from_file(FILE*)
473 * Add xbt_dict_get_key achieving a linear reverse search
474 * Remove the context module
476 Portability report of this version:
477 * Main portability targets:
478 - Linux(debian)/x86/context
479 - Linux(debian)/x86/pthreads
480 - Linux(debian)/amd64/context
481 - Linux(debian)/amd64/pthreads
482 On these, we still have the eratic breakages of gras/pmm and
483 amok/saturate_sg reported in previous version. We still think
484 that the tests are the cause of the fault, not the tested code.
486 - Mac OSX Leopard/x86/context
487 Still false negative in tesh autotesting.
488 Smpi still fails, but this time because readlink does not accept -f
489 Everything seems to work properly beside of that.
492 - AIX version 5.3 (only tested contexts this time)
493 Smpi still fails there because mktemp is not installed.
494 Everything seems to work properly beside of that.
496 I managed to compile it for the first time, but several breakages.
497 Won't delay the release for this exotic platform.
499 * Windows: it's still lagging behind. If you want to help, please
502 Timing report of this version:
503 This version seem to be more than 5% faster than 3.3.1 (on linux
504 64bits with contextes). The gain is less than expected, we are
505 investigating this for next release.
507 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
509 SimGrid (3.3.1) stable; urgency=low
512 * Implement a --cfg-help to show existing configuration variables
513 * Build chain do not require doxygen in maintainer mode
516 * fix a bug on struct sizeof computation, which prevented the
517 exchange of arrays of structs in some conditions
518 - added a regression test about this in datadesc_usage
519 * Allow the exchange of 0-long dynamic vectors.
520 - for that, use -1 as indicator of dynamic size instead of 0
521 - This implied to change any size from unsigned long to long,
522 reducing a bit communication abilities, but I guess that with
523 64bits being quite common, this is more than enough.
524 - This also induce a protocol change, thus bumping network protocol
525 version from 0 to 1 (if we have external users, we have to get
526 clean on that point too ;)
527 - added two regression tests about this in datadesc_usage
528 * Be more verbose when propagating local exceptions
529 This helps debugging.
530 * Display the status of simulated processes when receiving SIGINT in
534 * Allow to control the simulation from a trace file.
535 New functions MSG_action_register() and MSG_action_trace_run()
536 The first one allows to associate a function execution to each
537 kind of action while the second one parses a trace file and
538 triggers the corresponding actions within the system.
539 For now, only a toy example is provided in examples/msg/actions
540 * Add an exemple of process migration in examples/msg/migration
541 * Fix a bug in task exchange which broke MSG_task_get_sender()
542 Add a teshsuite regression test for that.
543 [Bug: if MSG_task_get_sender() is called after sender exit,
545 * Fix a bug which prevented suspend/resume to work properly
546 * Display the status of simulated processes when receiving SIGINT
547 This fixes a regression of v3.3. due to the introduction of SIMIX
548 * Bug fixing in failure management:
549 - trace could not start by a failure at time 0
550 - failure during communications were not working
553 * Add SIMIX_process_set_name() to change the name of the current
554 process in the log messages.
555 * Store smx_hosts in a dict since we only retrieve them by name
556 * Move the configuration infrastructure to surf
559 * Move the configuration infrastructure to surf
562 * Massive internal cleanups:
563 - Store internal structures on processes instead of hosts (allows
564 to have more than one process per host, in addition of being more
566 - Cleanup the initialization/finalization process
567 - Kill a whole bunch of unneeded synchronization:
568 processes run in exclusive manner within the simulator
569 - Move queues from global tables to process data fields
571 - now accept -platform and -hostfile arguments
572 - Pass the right rank value to processes according to the hostfile
573 * Compile the examples by default, and use them as regression tests
574 * Implement MPI_Wtime()
575 * Change the reference speed to a command line option
578 * TCP_gamma can now be specified as command line option using
579 --cfg=TCP_gamma:10000000.0
580 * Change the --surf-path cmd line option into --cfg=path:
583 * Also include strbuff from xbt.h public header
584 * xbt_ex_display(): do not free the exception after displaying
585 This allows to do more with the given exception afterward.
586 Users should call xbt_ex_free() themselves.
590 Portability report of this version:
591 * Main portability targets:
592 - Linux(debian)/x86/context
593 - Linux(debian)/x86/pthreads
594 - Linux(debian)/amd64/context
595 - Linux(debian)/amd64/pthreads
596 These targets fail about 1/10 of times on gras/pmm, but we believe
597 that this is because of the test, not because of simgrid.
598 amok/saturate_sg fails even more rarely, and the test may not be
601 - Mac OSX Leopard/x86/context
602 The test suite still spits tons of errors because some obscure
603 force prevents us from removing the temporary directories
604 arguing that they still contain some metadata I've never heard of.
605 Smpi fails because seq is not installed.
606 Everything seems to work properly beside of that.
609 - AIX version 5.3 (both contexts and pthread)
610 Smpi still fails there because mktemp is not installed.
611 XML inclusions seems rosty on AIX.
613 * Windows: it's still lagging behind. If you want to help, please
616 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
618 SimGrid (3.3) stable; urgency=high
622 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
625 * Introduce the SIMIX module: factorize code between MSG and GRAS.
628 Until now, GRAS were using MSG as an interface to SURF. It was
629 quite difficult because both interface have several differences
630 (MSG channels vs GRAS sockets were the most notable point).
632 This also opens the gate to SMPI (which should occur soon) and speed
633 up simulations by to 40% (even if it were not the main goal).
635 **************************************
636 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
637 **************************************
638 The point is that events occuring at the exact same timestamp are
639 not scheduled in the same order with the old and new version. This
640 may be enough to completely change the execution of simulations in
641 some cases. Sorry for the inconvenience.
643 * Cleanup and upgrade the XML format to push further scalability
644 issues (check http://hal.inria.fr/inria-00256883/ for more info)
646 * Improve the testing infrastructure with tesh. Now a very large part of
647 the code is tested not only by being run but also by checking that the
648 output match an expected output [Mt].
650 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
651 is really less memory-demanding, which should allow you to use
652 larger files in SimGrid [AL].
654 * Inform valgrind about our contextes, so that it becomes usable
655 with the default (and more effecient) version of SimGrid
656 [contributed by Sékou Diakite, many thanks]
659 * Introduce a listener thread in charge of receiving incomming
660 messages from the network. It allows to overlap communication and
661 computation but most notably, it removes some stupid deadlocks due
662 to the fact that so far, a process could not send and receive at
663 the same time. This made most non trivial communication schema
665 * Convert the PIDs from long int to int to match the MSG ones (and
667 * New function: gras_agent_spawn() to launch a new process on
668 current host. Only working in simulation for now. [Mt]
669 * New function: gras_os_hostport() returning a constant form (ie,
670 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
673 * Make the backtrace of exceptions more human readable [Mt]
674 * New module: xbt/str [Mt]
675 a ton of string utility functions (split, join, printf to a newly
676 allocated buffer, trim, etc)
677 * New module: xbt/hash [Mt]
678 SHA1 hashing algorithm (more to come if needed)
679 * New module: xbt/synchro [Mt]
680 synchronization tools (mutex and conditions) working the same way
681 in simulation and in real life (mainly useful for GRAS, but not
683 * New module: xbt/queue [Mt]
684 classical producer/consumer synchronization scheme
685 * xbt_dynar_new_sync() creates a synchronized dynar. All access
686 (using the classical functions will get serialized) [Mt]
687 * Make dictionary internal table dynamic. No need to specify its size
688 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
690 * Make sure the log channels are organized as a tree under windows
691 (because of ANSI C compatibility issue, any channel were child of
695 * Cleaned many thing in surf and fixed a few bugs [AL].
696 * Add a nice command line configuration mechanism to compose models [AL].
697 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
698 the previous one (KCCFLN05). It relies on something that looks like
699 a max-min sharing mechanism but cannot be written as such. A new solver
700 was thus designed [AL].
701 * Add a new solver to lmm. Based on Lagrange optimization and
702 gradient-based descent, it enables to efficiently maximise systems s.a
704 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
706 This solver enables to propose two new network models for TCP Reno and
707 TCP Vegas based on Low's work. These models still need to be fully
708 tested though [Pedro Velho].
711 * Bug fix in SD_simulate. Now the time bound given as argument is
713 * Use the new parallel task model (ptask_L07) as default.
714 * Use the SURF command line configuration mechanism.
715 * 0-size tasks (for synchronization) should now work.
717 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
719 SimGrid (3.2) stable; urgency=high
723 We still experience issues on this platform, but we believe that at
726 GRAS API BREAKAGE (for simplification purpose, sorry):
727 * the gras_msgtype_by_name is not used anymore. Instead of
728 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
729 you can write (and must)
730 gras_msg_send(toserver, "request", &request);
731 - If you still want to pass a gras_msgtype_t to the function (to cache
732 the type and avoid the lookup time), use the gras_msg_send_() variant.
733 - Impacted functions:
734 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
735 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
736 * The callbacks are now expected to return 0 when everything went well
737 (just like the main() function)
739 GRAS new features and improvements:
740 * New module mecanism where user code can use per process globals [Mt]
741 This is similar to gras_userdata_*() functions, but for libraries. It
742 factorize some code developped over and over in the examples and AMOK.
743 It has still to be documented and used (only amok/peermanagement is
745 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
746 loss when they were small enough to fit into the buffer and sent quickly
747 enough so that they can all get received in one shoot.
748 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
749 if not found. Use the *_or_null() variant for the old semantic.
750 * In gras_msg_handle, do not discard messages without callback.
751 They are probably messages to be explicitly awaited later (ie, proofs of
752 mis-synchronization in userland since they are sent before being awaited)
754 * gras_socket_meas_send/recv: semantic changed!
755 The numerical arguments used to be (1) the total amount of data to send
756 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
757 messages. This was need for the fool willing to send more than MAXINT
758 bytes on quite fat pipes.
761 * Do really rename the hostmanagement module to peermanagement. [Mt]
762 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
763 but this is rather new and this was documented in the module
764 documentation (poor excuses, I admit)
765 * Bandwidth measurement semantic changed! This follows the changes to
766 gras_socket_meas_send/recv explained above.
769 * A sequential mode has been added to the workstations. When a workstation
770 is in sequential mode, it can execute only one task, and the other tasks
771 are waiting in a FIFO. [Christophe Thiery]
774 * The KCCFLN05 workstation model now handles parallel tasks. It is the
775 model for SIMDAG. [Christophe Thiery]
776 * Bug fix in the maxmin solver: Some values were close to 0 instead of
777 equal to 0, which caused some bad behaviors in
778 saturated_constraint_set_update. I now use a threshold mechanism like in
782 * When running manually src/testall, you select specific units [Mt]
783 testall is the result of our cunit mecanism, and should replace all
784 the scripty thingy around since bash don't run easily on billware.
786 * A mallocator system has been added. [Christophe Thiery]
787 Mallocators allow you to recycle your unused objects instead of freeing them
788 and allocating new ones.
790 Documentation update:
791 * FAQ reworking + New FAQs:
792 - "Valgrind spits tons of errors!" [Mt]
793 - "How to repport bugs" [Mt]
794 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
795 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
796 - Communication time measurement within MSG [AL]
797 - I experience weird communication times when I change the latency [AL]
800 - an introduction to the framework and to the used communication model
801 - an initiatic tour introducing the most proheminent features:
803 . Lesson 0: Installing GRAS
804 . Lesson 1: Setting up your own project
805 o Part 2: Message passing
806 . Lesson 2: Exchanging simple messages
807 . Lesson 3: Passing arguments to the processes (in SG)
808 . Lesson 4: Attaching callbacks to messages
809 . Lesson 5: Using globals in processes
810 . Lesson 6: Logging informations properly
811 . Lesson 7: Using internal timers
812 . Lesson 8: Handling errors through exceptions
813 . Lesson 9: Exchanging simple data
814 . Lesson 10: Remote Procedure Calling (RPC)
815 . Lesson 11: Explicitely waiting for messages
816 . Recapping of message passing features in GRAS
817 - A HOWTO section containing:
818 o HOWTO design a GRAS application
819 More are due, of course. They will come latter. In the meanwhile, you can
820 check the examples which are still here.
822 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
824 SimGrid (3.1) stable; urgency=high
828 There was a stack corruption somewhere, visible only when optimizing
829 with these versions. [Vince]
832 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
833 module that enabled to play with Directed Acyclic Graphs. It is built
834 directly on top of SURF and provides an API rather close to the old
835 SG. Some old codes using SG are currently under rewrite to check that
836 all needful functions are provided. [Christophe Thiery]
839 * Complete rewrite of the KCCFLN05 workstation model. It is now an
840 extension of the classical CLM03 model that gracefully handles
841 failures. This is now the default model for MSG and GRAS. It doesn't
842 handle parallel tasks yet though. [AL]
843 * Bug fix: Weights were not correctly set in the network part.
844 WARNING: This may have resulted in incorrect results with simulations
845 where there are more than one flow on a given link. [AL]
848 * After a (long ?) discussion on simgrid-devel, we have decided that the
849 convention we had on units was stupid. That is why it has been decided
850 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
851 WARNING : This means that all previous platform files will not work as
852 such with this version! A warning is issued to ask users to update
854 A conversion script can be found in the contrib module of the CVS, under
855 the name contrib/platform_generation/surfxml_update.pl [MQ]
858 * Bug fix: Processes were started in reverse order, wrt deployment file.
859 WARNING: if your code relies on this bug, please fix it. [AL]
860 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
861 being executed on two different locations. [AL]
862 * Bug fix: Failures are now better supported thanks to Derrick's tests
863 (there was many failure situations I hadn't thought of and that weren't
864 correctly handled). [AL]
865 * New function: MSG_host_is_avail indicates you whether a given m_host_t
869 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
870 Exception occurring on server-side are propagated back to client (!).
872 API CHANGE: the callback changed their prototype. Change:
873 int my_handler(gras_socket_t expeditor, void *payload_data) {
875 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
876 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
878 * New! function: gras_msg_handleall to deal with all messages arriving
879 within a given period.
880 * New! function: gras_socket_server_range to get a server socket in a
881 range of port numbers (ease to avoid port number conflicts) [MQ]
882 * New! gras processes display their backtrace when they get a SIGUSR1
883 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
884 Sweet to debug RL processes [MQ]
888 - Do not force experiment sizes to be expressed in kb, or it becomes
889 impossible to measure the latency this way (needs one byte-long tests)
890 WARNING: this changes the amok_bw_* function semantic. [MQ]
891 - Implements the link saturation stuff. [MQ]
892 * Peer management module:
893 New! module factorizing code that we wrote over and over [MQ].
896 * New module: cunit (my jUnit implementation in ansi C) [MQ]
897 - Test units are placed directly into the library code, they get extracted
898 automatically and placed into the src/testall binary.
899 - Convert most of the XBT tests to this system.
900 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
901 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
902 live in the same C code. This required to change a little bit the API
903 of surfxml parsing but shouldn't be an issue for end-users. [AL]
904 * New module: sparse graph structure with basic algorithms (this is work
905 in progress and the API is not considered to be frozen yet). [AL]
906 * Display more information on backtraces: source line & function names are
907 now displayed just like valgrind does (rely on addr2line tool) [MQ]
908 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
909 * Reworked a little bit some #include statements to load only required
910 headers. Some user code that relied on SimGrid to include stdlib or
911 stdio may need to include it by themselves. [AL]
912 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
914 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
915 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
916 but I guess that almost nobody used those parts. [MQ]
918 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
920 SimGrid (3.0.1) stable; urgency=low
923 * Unfortunately, I had missed 5 misnamed functions:
924 xbt_fifo_item_t xbt_fifo_newitem(void);
925 void xbt_fifo_freeitem(xbt_fifo_item_t);
926 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
927 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
928 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
929 They're now deprecated. Please use their new versions:
930 xbt_fifo_item_t xbt_fifo_new_item(void);
931 void xbt_fifo_free_item(xbt_fifo_item_t);
932 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
933 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
934 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
936 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
937 * Documentation: xbt_log module unmercifully reworked [MQ]
938 * Bugfix: there was a problem with the ending of contexts with
939 the pthread backend. It caused some weird deadlock or behavior
940 depending on the pthread implementation. [AL]
941 * Bugfix: get the exceptions raised in the simulator repport where
942 and why they come from when they are not catched in time [AL, MQ]
945 * Bugfix: Do repport the error when two non-connected hosts try to
946 exchange data (Thanks to Flavien for stumbling into this one) [AL]
949 * Add additionnal checkings on communications. Assert that two
950 communicating hosts are connected by a set of links... [AL]
953 * Add additionnal checkings on channel values in communication [AL]
954 * New: MSG_task_get_source to see on which host a task was generated [HC]
955 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
956 the number of tasks waiting to be received on channel and sent
958 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
959 waits for the first task coming from a given host.. [AL]
961 GRAS new functionnalities: [MQ]
962 * Enhance the parsing macro to allow the size of multidimentional objects
963 to be given thru annotations.
964 * New example (and documentation): Matrix Multiplication a la RPC
965 (as when I was young!) and fix a bunch of bugs found on the way.
967 GRAS performance improvements: [MQ]
969 * Reduce the amount of cbps creation/destruction by making it static to
970 datadesc_send/recv() and using a (newly created) cbps_reset (based on
973 * Change libdata to a set so that we can search for stuff by ID (and thus
974 reduce the insane amount of dict lookups)
977 * Actually implement gras_datadesc_copy() so that we don't have to mimick
978 RL communication on top of SG since it's so uneffective.
979 It may also be used for inter-thread communication in RL, one day.
980 * Use gras_datadesc_copy() to exchange messages on top of SG
982 - improve message exchange performance on top of SG
983 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
984 * Don't exchange on the network the size of the used part of buffer,
985 instead, specify the possible buffer size to read().
987 - reduces the amount of read/write calls (one pair per exchange)
988 - reduces the amount of exchanged data (the size)
989 - allows to retrieve all arrived data on receiver side, if we don't need
990 it right now (subsequent read will peek the buffer)
991 - allows the receiver to proceed with the begining of the stream before
992 everything is arrived
993 - make it possible to build an iov transport (using readv/writev)
995 - take care of the data with non-stable storage (like stacked data),
997 * If possible, TCP send uses vector I/O (when writev() is here)
998 - Don't use it for receive since we send data sizes and data on the
999 same stream, so we wouldn't be able to chain large amount of chunks
1000 before having to flush the stuff to read the size.
1001 * Rework the transport plugin mecanism to simplify it and reduce the
1002 amount of pointer dereferencement when searching for the right function
1005 * I guess that now, we do almost as few system calls as possible while
1006 doing as few data copy as possible.
1008 To improve it further, we could try to send all the sizes first and then
1009 all the data (to use iov on receiving size), but it's only a partial
1010 solution: when you have 2 dimensional data, the sizes of the second
1011 dimension is data of the first dimension, so you need 3 streams.
1013 I'm not sure the potential performance gains justify the coding burden.
1015 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1017 SimGrid (3.00) stable; urgency=high
1020 * New! Give the possibility to hijack the surf parser and thus bypass
1021 MSG_create_environment and MSG_launch_application. Have a look at
1022 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1024 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1026 SimGrid (2.96) unstable; urgency=low
1031 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1032 This deprecates the xbt_error_t mecanisms.
1033 It modifies (simplifies) all XBT and GRAS API.
1034 MSG API keeps unchanged (exceptions raised by XBT are catched from
1035 within MSG and masked with existing error handling facilities)
1038 * New! Add a FATPIPE model. [AL]
1039 * New! Add a parallel task model. [AL]
1040 * New! Add automatically a loopback interface (in the default
1041 network model) if none was precised.
1044 * Bugfix: MSG_process_resume now works with the current running process.
1046 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1047 * Modification of MSG_task_get_compute_duration. Once a task has been
1048 processed, the value returned by this function is now equal to 0. [AL]
1049 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1050 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1051 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1052 of a m_task. For now, MSG_task_cancel only works with computation
1054 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1055 of the processor (in Mflop/s) regardless of the current load on the
1057 * API Change: use proper naming convention for MSG_getClock and
1058 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1060 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1061 This function changes the priority of a computation task. This priority
1062 doesn't affect the transfer rate. A priority of 2 will make a task
1063 receive two times more cpu power than the other ones. This function
1064 has been added to suit the needs of Nguyen The Loc and hasn't been that
1065 much tested yet. So if it fails, please report it and send me your code.
1067 * API Change: removed all functions and types that were marked "deprecated"
1068 since many months. Renamed MSG_global_init_args to MSG_global_init.
1070 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1072 SimGrid (2.95) unstable; urgency=low
1075 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1076 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1077 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1080 * Bugfix: Allow absolute paths for platform description files [MQ]
1081 * Bugfix: do free the variables after use. Leads to drastic performance
1083 * Implement max_duration (ie, timeouts) on resources [AL]
1086 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1088 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1089 the needed support (timeouts) [AL]
1092 * Implement measurement sockets. You can now get the bandwidth between two
1093 hosts thanks to AMOK (see below). [MQ]
1094 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1095 dynar over the network (yeah) [MQ]
1096 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1097 * simple/static token-ring example. [Alexandre Colucci and MQ]
1098 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1099 in place before [MQ]
1102 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1103 type parsing. "short" and "long" modifiers were also ignored; other
1104 modifier (such as reference level) are still ignored. [MQ]
1105 * Embeed the buffer size within the buffer itself on SG. [MQ]
1106 That way, send() are atomic and cannot get intermixed anymore (at least
1107 the ones which are less than 100k; bigger messages still have the issue)
1108 * Array size pushed by the field, not by the field type (or each
1109 and every long int will push stuff to the cbps) [MQ]
1110 * use select() to sleep since it allows to portably sleep less than one
1113 GRAS (minor cleanups)
1114 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1117 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1118 * type category 'ignored' killed as it was never used and were difficult
1120 * whether a type can cycle or not is now a flag, leaving room for more
1121 flags (as "ignored", if we feel the need one day ;)
1122 * Rename raw sockets to measurement sockets since "raw" has another
1123 meaning in networking community.
1126 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1127 and offers features not belonging to GRAS but that most applications
1128 need. One day, it may be a set of plugins loadable at runtime.
1129 * New module: bandwidth
1130 bandwidth measurement between arbitrary nodes running this module. [MQ]
1132 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1134 SimGrid (2.94) unstable; urgency=low
1136 The first beta release of SimGrid 3 !
1140 * Update the main page and the FAQ. Adding references to gforge.
1143 * Add a gras_os_getpid function.
1146 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1147 * Extend the logs so that they also print PID, hostname, date, ... if
1149 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1150 and kill the old version which were in testsuite/
1151 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1155 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1156 for users. Instead, each programming environment (GRAS, MSG,...) use it
1157 when needed to provide such a feature to users.
1158 Don't shortcut the mecanism or you will also shortcut the virtualization
1159 you need on the simulator.
1163 * Cleanups in configury with regard to compile optimization/warning flags.
1164 Also add -fno-loop-optimize to any powerpc since it's the optimization
1165 killing gcc (< 3.4.0).
1166 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1167 needed by MSG examples complications
1168 * Borrow configury beautifications from PHP
1171 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1172 warning (thanks loris for stumbling into this one).
1173 * Bugfix: stop loading private headers (gras_config.h) from the public
1177 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1178 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1179 RL would imply a DNS resolver, which is *hard* to do in a portable way
1180 (and therefore delayed).
1181 * Implement a real timer mecanism and use it in timing macros. This allows
1182 to avoid rounding errors and get a 0.000005 sec precision in timing
1183 macros. While I was at it, various cleanups:
1184 - allow to declare more than one timed section per file (fix a stupid bug)
1185 - move some private declaration to the right place
1186 - merge conditional execution and timing macros into emulation module
1187 - document the module
1188 - make sure the module cleanups its mess on gras_exit
1189 * Documentation improvements:
1190 - (new) how to compile applications using GRAS
1191 - (new) emulation support (timing macros)
1193 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1195 SimGrid (2.93) unstable; urgency=low
1197 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1200 - Use Paje properly where used. Still to be sanitized properly.
1201 - Portability fix: Add an implementation of the contexts using pthread
1205 - Add xbt_procname(): returns the name of the current process.
1206 Use it to show the current process's name in all logging.
1208 - fix detection of older flex version and the reaction, since we do
1209 depend on modern ones (we use lex_destroy)
1210 - Better separation of SG and RL in the libs: remove all simulation code
1211 from libgras. As a result, this lib is now only 200k when stripped.
1212 Some of the xbt modules may also be duplicated (two sets and such) and
1213 should be cleaned/killed before SG3.
1214 - Insist on using xlC on AIX because of weird problems involving gcc there.
1215 - Cleanup the make remote stuff. This is now done by scripts
1216 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1217 mainly for our private use, but we're working on changing them to user
1220 - Bugfix: flush the socket on close only if there is some *output*.
1221 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1223 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1226 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1228 SimGrid (2.92) unstable; urgency=low
1230 Alpha 3 on the path to SimGrid 3
1234 - New! First try of benchmarking macros.
1235 - New! First try so that gras_stub_generator generate deployment and
1236 remote compilation helpers.
1238 - Bugfix: Initialization fix in msg_test.
1242 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1244 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1246 - API Change: gras_os_sleep to take the amount of seconds as a double.
1247 Accepting an int was error prone since it was the only location where
1248 seconds were coded as such. It leaded to damn rounding errors.
1249 - Bugfix: Hard to belive that timers ever worked before this.
1251 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1253 SimGrid (2.91) unstable; urgency=low
1255 Alpha 2 on the path to SimGrid 3
1259 - Bug fix in the lmm_solver.
1261 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1262 through the function MSG_paje_output.
1263 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1264 - It is possible to bound the rate of a communication in MSG with
1265 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1266 to write it in the changelog).
1267 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1272 - Various cleanups to the autotools stuff
1273 - Begin to move Gras examples to examples/gras/
1274 - Let make distcheck work again (yeah!)
1276 - documentation overhauled using doxygen.
1277 gtk-doc-tools is dead in SimGrid now.
1278 - Automatically extract all existing logging categories, and add the list
1279 to the documentation (long standing one, to say the less)
1281 - Cleanup the known architecture table. Reorder the entries to group what
1282 should be, and use a more consistent naming scheme.
1283 (some of the test dataset are still to be regenerated)
1284 - New! Allow library to register globals on each process just as userdata
1286 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1287 lookup time (for now).
1288 Use it in msg and trp.
1289 This cleans a lot the internals and helps enforcing privacy of the
1290 headers between the gras components.
1291 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1292 - Bugfix: gras_os_time was delirious in RL.
1293 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1295 - Reenable GRAS now that it works.
1297 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1299 SimGrid (2.90) unstable; urgency=low
1301 Alpha 1 on the path to SimGrid 3
1303 * It is a long time since the last release of SimGrid. I'm sorry about
1304 that but as I had told you, I was rewriting a lot of things. I apologize
1305 to those who had been reporting bugs to me and that I had not answered.
1306 If your bug is still in the new version, please tell me. Here is a
1307 summary of the main changes.
1309 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1310 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1311 portability, speed, and a lot more but you'll figure it out later.
1312 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1313 I wanted to release the new MSG as soon as possible and I have
1314 broken GRAS, which is the reason why, for now, only MSG is fully
1315 functional. A laconic description of these projects is available
1316 in the documentation.
1318 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1319 kernel optimized for our needs (called SURF but only the developers
1320 should use it). Hence, MSG is now roughly 30 times faster and I think
1321 that by rewriting a little bit MSG, I could event speed it up a little
1322 bit more. Beside the gain in speed, it is also much easier to encode a
1323 new platform model with SURF than it was with SG. More to come...
1325 * REVOLUTION 3: I have tried to change a little as possible the API of
1326 MSG but a few things really had to disappear. The main differences
1327 with the previous version are :
1328 1) no more m_links_t and the corresponding functions. Platforms are
1329 directly read from a XML description and cannot be hard-coded
1330 anymore. The same format is used for application deployment
1331 description. The new format is described in the documentation.
1332 Have a look in tools/platform_generation. There is a tiny script
1333 that converts from the old platform format to the new one. Concerning
1334 the application deployment format, parsing the old one is tricky.
1335 I think most of you should however be able to convert your files. If
1336 it is really an issue, I can write a C code that does the conversion.
1338 2) the toolbox tbx does not exist anymore. We now have a library
1339 with much more data-structures but without the hash-tables (we have
1340 dictionaries that are much faster).
1342 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1344 *****************************************************************************
1345 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1346 * don't feel like converting the oldies. (Mt) *
1347 *****************************************************************************
1350 Version 2.90: "the long awaited one"
1351 - Finished rewriting and debugging MSG. Rewrote the documentation.
1352 - disable GRAS for now since it needs to be ported to the newest SG
1355 - Finish the port to windows (using mingw32 for cross-compile)
1358 - Main loop and datastructures of SURF. A cpu resource object is
1359 functional. Surf can thus be used to create cpu's with variable
1360 performance on which you can execute some actions.
1362 2004-11-15 Martin Quinson
1363 - Port to ARM. Simply added the alignment and size descriptions. Should
1364 work, but the ARM machines are so slow that I didn't had the opportunity
1365 to 'make check' over there yet.
1367 2004-11-15 Arnaud Legrand
1368 - Trace manager now written. It uses a heap structure and is therefore
1369 expected to be efficient. It may however be speeded up (particularly
1370 when many events occur at the same date) by using red and black
1371 trees. One day maybe...
1372 - Max-min linear system solver written. It uses a sparse matrix
1373 structure taking advantage of its expected use. Most operations are
1374 O(1) and free/calloc are called as few as possible. The computation of
1375 the minimum could however be improved by using a red and black tree
1378 2004-11-03 Arnaud Legrand
1379 - Rename every gras_* function that was in xbt/ to its xbt_
1381 - Add a heap and a doubly-linked list to xbt
1382 - Added a dichotomy to the dictionaries. make check works as well before
1383 so I assume that the patch is correct. I do not know however if things
1384 run effectively faster than before now. :)
1386 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1387 SimGrid, and the version number is bumped to 2.x
1389 2004-10-29 Martin Quinson
1390 - Introduction of the remote errors.
1391 They are the result of a RMI/RPC on the remote machine.
1392 ErrCodes being scalar values, you can't get the host on which those
1393 errors did happen. Extending the error mechanism as in Gnome is possible.
1394 No idea yet whether it is a good idea.
1396 2004-10-28 Martin Quinson
1397 - Interface revolution: the Starred Structure Eradication.
1398 I used to do typedef struct {} toto_t; and then handle *toto_t.
1399 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1400 - ???_t is a valid type (builded with typedef)
1401 - s_toto_t is a structure (access to fields with .)
1402 - s_toto is a structure needing 'struct' keyword to be used
1403 - e_toto_t is an enum
1404 - toto_t is an 'object' (struct*)
1406 typedef struct s_toto {} s_toto_t, *toto_t;
1407 typedef enum {} e_toto_t;
1408 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1409 s_toto_t) is private.
1411 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1412 so that it can set the variable to NULL. It was so for dicts and sets,
1413 it changed for dynars.
1415 - Fix a bunch of memleaks in dict_remove
1416 - Fix a bug in sg/server_socket opening: it failed all the time.
1418 2004-10-07 Martin Quinson
1419 - Speed up dynar lookup operation a bit.
1421 gras_dynar_get is dead.
1423 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1424 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1425 (which gives you the address of the stored data).
1427 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1428 copy of the data using an affectation to do the job and not a memcpy.
1430 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1431 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1433 It does not really speedup the dynar test because they are
1434 setting elements all the time (and look them seldom). But the dict does
1435 far more lookup than setting.
1437 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1439 2004-10-05 Martin Quinson
1440 - Allow to (en/dis)able the cycle detection at run time.
1442 Whether we should check for cycle or not is now a property of each
1443 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1444 datadesc_cycle_unset allow to remove this property when previously set.
1446 Note that the cycle detection is off by default since it impacts the
1447 performance. Watch the data you feed GRAS with ;)
1449 This property is hereditary. Any element embedded in a structure having it
1450 set have it set for the time of this data exchange.
1452 You should set it both on sender and receiver side. If you don't set it on
1453 sender side, it will enter an endless loop. If you forget on receiver
1454 side, the cycles won't be recreated after communication.
1456 - Header reorganization.
1457 Kill gras_private.h, each submodule must load the headers it needs.
1459 2004-10-04 Martin Quinson
1460 - Interface revolution: do not try to survive to malloc failure.
1462 Now, gras_malloc and friends call gras_abort() on failure.
1463 As a conclusion, malloc_error is not a valid error anymore, and all
1464 functions for which it was the only gras_error_t return value are
1465 changed. They now return void, or there result directly.
1466 This simplify the API a lot.
1468 2004-09-29 Martin Quinson
1469 - Re-enable raw sockets.
1470 Created by gras_socket_{client,server}_ext;
1471 Used with gras_raw_{send,recv}
1474 It should allow to kill the last bits of gras first version soon.
1476 This is not completely satisfactory yet (duplicate code with
1477 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1480 - Simplify transport plugin (internal) interface by not passing any
1481 argument to _server and _client, but embedding them in the socket
1484 2004-09-28 Martin Quinson
1485 - Finish the port to AIX.
1486 autoconf was my problem (segfault within the malloc replacement
1487 function. No idea why)
1489 2004-09-16 Martin Quinson
1490 - Fix some size_t madness on 64bit architectures.
1492 2004-09-08 Martin Quinson
1493 - Reduce the number of system headers loaded, overload some more system
1494 calls (such as malloc to cast the result of the system one, and work
1496 - Fix and reintroduce the config support
1498 2004-09-07 Martin Quinson
1499 - Source code reorganization to allow Arnaud to surf all over there.
1500 - Allow to document the logging categories.
1501 - Remove all uppercase from logging categories and useless cleanup in names.
1503 2004-08-18 Martin Quinson
1504 Version 0.6.2 (protocol not changed; API changed)
1505 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1506 gras_error_t), and NULL when not found. Functions expecting a msgtype
1507 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1508 hopefully usefull message.
1509 - Portability to prehistoric sparcs again
1511 2004-08-17 Martin Quinson
1512 Version 0.6.1 (protocol not changed; ABI not changed)
1513 - prealloc some buffers to speed things up
1515 2004-08-11 Martin Quinson
1516 Version 0.6 (protocol not changed; ABI expended)
1517 - The parsing macro can deal with the references, provided that you add
1518 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1520 2004-08-09 Martin Quinson
1521 Version 0.5 (protocol not changed; ABI changed)
1522 - Allow to off turn the cycle detection code in data exchange at
1523 compilation time. It should be at run time, but I'm short of time (and
1524 the config stuff is still broken). That way, we keep dict out of the
1525 critical path, which is good because the performance is poor:
1526 - search not dichotomial yet
1527 - dynar give no way to access their content and memcpy everytime
1528 - In composed data description (struct, ref and so on), stop foolness of
1529 keeping the subtype's ID, but store the type itself. This keeps sets out
1530 of the critical path, which is good since they rely on dynar and
1531 dictionnaries. The only loose of that is that we cannot detect the
1532 redeclaration of a structure/union with another content (but I'm not sure
1533 the code detected well this error before anyway). We still can detect
1534 the redefinition discrepancy for the other types.
1535 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1536 it breaks the code because of type-punning used all over the place).
1537 This breaks on all non-gcc architectures (for now).
1539 All those changes (plus the buffer of last time) allow me to gain 2 order
1540 of magnitude on cruel tests consisting of 800000 array of integers on two
1541 level of a hierarchical structure (200 secondes -> 4 secondes)
1544 - the selector of reference must now return the type it points to, not
1545 the ID of this type.
1547 2004-08-06 Martin Quinson
1548 Version 0.4 (protocol changed; ABI not changed)
1549 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1550 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1551 - Implement a buffer transport, and use it by default (it relies on tcp in
1552 real life and on sg in simulation).
1553 That's a bit hackish since I had a new field to the structure to store
1554 its data without interfering with the subtype ones. Inheritance
1555 is tricky in C. And that's a kind of reverse inheritance with one class
1556 derivating two classes. Or maybe a game with java interfaces. Anyway,
1557 that's damn hard in C (at least).
1558 Moreover, I got tired while trying to ensure plugin separation and
1559 genericity in SG mode. MSG wants me to do weird things, so let's go for
1560 cruel hacks (temporarily of course ;).
1561 See comment in transport_private.h:71
1562 - do not include all the _interface headers in private but in the files
1563 which really need them (to cut the compilation time when they are
1566 2004-07-26 Martin Quinson
1567 Version 0.3 (protocol not changed; ABI changed)
1568 - Major overhault of the datadesc interface to simplify it:
1569 - shorted the function names:
1570 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1571 - add a trivial way to push/pop integers into the cbps without malloc.
1572 This allows to make really generic sub_type description, which simply
1573 pop their size of the stack.
1574 - add a function gras_datadesc_ref_pop_arr() which does what users want
1575 most of the time: Declare a dynamic array (which pops its size of the
1576 stack) and declare a reference to it. Poor name, but anyway.
1577 - kill the post-send callback, add a post-receive one
1579 2004-07-23 Martin Quinson
1580 Version 0.2 (protocol changed; ABI changed)
1581 - add some testing for cpbs in the test cases, and fix some more bugs.
1582 This invalidate again the little64 data file, since I cannot regenerate
1584 - remove an awfull optimization in the logging stuff, allowing me to:
1585 - understand it again
1586 - learn gcc how to check that the argument match the provided format
1587 - fix all errors revealed by gcc after that
1588 - internal keys of dict are not \0 terminated. Deal with it properly in
1589 loggings instead of segfaulting when the user want to see the logs :-/
1591 2004-07-22 Martin Quinson
1592 - Fix some stupid bug preventing cbps (callback postit) from working
1594 2004-07-21 Martin Quinson
1595 - Some documentation cleanups
1596 - remove the useless last argument of msgtype_declare
1597 - rename the Virtu functions to fit into the 'os' namespace
1598 - move headers src/include -> src/include/gras/ and stop fooling with
1599 gras -> . symbolic link
1600 - make distcheck is now successful
1602 2004-07-19 Martin Quinson
1604 - Build shared library also
1605 - Install html doc to the right location
1606 - stop removing maintainer files in make clean
1607 - build tests only on make check
1609 2004-07-13 Martin Quinson
1611 - No major issue in previous version => change versionning schema
1612 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1613 corresponding dataset.
1615 2004-07-11 Martin Quinson
1617 - Get it working with any kind of structure (we can compute the padding
1618 bytes remotely for all the architectures I have access to)
1619 - Implement the structure parsing macro (still not quite robust/complete)
1620 - Improvement to the remote testing toysuite
1622 2004-07-10 Martin Quinson
1623 [autoconf mechanism]
1624 - get ride of a bunch of deprecated macros
1625 - actually run the test for two-compliment, not only compile it :-/
1626 - test whether the structures get packed (and bail out if yes. Damn.
1627 Alignment is a serious matter)
1628 - test whether the structures get compacted (but respecting the alignment
1629 constraints of each types)
1630 - test whether the array fields of structures can straddle alignment boundaries
1632 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1633 (took me 2 hours to find that bug, looking at the wrong place)
1634 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1635 before sending/receiving and are used to compute the offsets of fields
1636 - Given that padding size depend even on compiler options, keep track of
1637 alignment and aligned_size only for the current architecture. Not a big
1638 deal since we send structure fields one after the other (seems
1640 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1641 program, let it work on linux/gcc/little32. portability todo.
1643 2004-07-08 Martin Quinson
1644 - import and improve remote compilation support from FAST
1645 - make sure make check works on half a dozen of machines out there
1647 2004-07-07 Martin Quinson
1648 Let's say it's version 0.0.3 ;)
1649 - Implement conversions (yuhu!)
1650 - Let it work on solaris (beside conversion, of course)
1651 - Stupid me, using rand() to generate the conversion datatests in not wise.
1653 2004-07-06 Martin Quinson
1654 - Let make dist work, since I'm gonna need it to compile on remote hosts
1655 - Let Tests/datadesc_usage write the architecture on which the file was
1656 generated as first byte.
1657 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1658 architecture descriptions.
1659 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1660 on those architectures.
1661 - Optimization: send/recv array of scalar in one shoot
1663 2004-07-05 Martin Quinson
1664 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1666 - Plug a whole bunch of memleaks
1667 - each process now have to call gras_{init,exit}. One day, their log
1668 settings will be separated
1669 - Continue the code factorisation between SG, RL and common in Transport.
1671 2004-07-04 Martin Quinson
1673 - Redistribution between SG and RL.
1674 We wanna have to accept in SG, so move accepted related parts of RL in
1675 the common part. (more precisely, the dynar of all known sockets is no
1676 more a static in transport.c, but part of the process_data)
1678 [gras_stub_generator]
1679 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1681 2004-07-03 Martin Quinson
1682 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1683 them all up in one shot)
1685 - Fix the ugly case of reference to dynamic array.
1686 - New (semi-public) function gras_datadesc_size to allow the messaging
1687 layer to malloc the needed space for the buffer.
1689 - gras_socket_close now expect the socket to close (and not its address to
1690 put NULL in it after it). This is because the socket passed to handlers
1691 is one of their argument (=> not writable).
1693 - propagate the interface cleanup from last week in datadesc, ie remove a
1694 superfluous level of indirection. User pass adress of variable
1695 containing data (both when sending and receiving), and not of a variable
1696 being a pointer to the data. Let's say that I like it better ;)
1697 The price for that is constructs like "int msg=*(int*)payload" in
1698 handlers, but it's a fine price, IMHO.
1700 - Let it work in RL (yuhu)
1702 2004-06-21 Martin Quinson
1704 - porting SG plugin and SG select to new standards (works almost).
1705 - plug memleaks and fix bugs around.
1708 - cleanup the prototype of data recv and force users to specify when they
1709 want to handle references to objects. Test case working even for cycles.
1710 - plug memleaks. Valgrind is perfectly ok with this.
1712 2004-06-12 Martin Quinson
1714 - cleanup the separation between plugin and main code in plugin creation
1716 2004-06-11 Martin Quinson
1718 - Reput hook for raw sockets, needed for BW experiments
1719 - kill a few lines of dead code
1720 [Data description] Interface cleanup
1721 - gras_datadesc_by_name returns the searched type or NULL.
1722 That way, no variable is needed to use a type desc once, which makes
1724 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1725 two parameters were strings (field name, type name), leading to
1727 [Dicos] Interface cleanup
1728 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1729 This is consistant with the dynar API.
1731 2004-04-21 Martin Quinson
1733 - Porting to new standards.
1735 - interface cleanup.
1736 There is no bag anymore, no need to take extra provision to mask the
1737 pointers behind "ID".
1738 Better splitup of functions between files create/exchange/convert.
1739 This is still a bit artificial since convert and receive are so
1740 interleaved, but anyway.
1742 - add a queued message list to procdata (the ones not matching criteria
1744 - factorize some more code between SG and RL wrt procdata
1746 - use gras_exit in example to track memleaks
1747 - get rid of gs_example now that GS is properly integrated into gras
1748 - update run_test to integrate the lastest tests (datadesc)
1750 - rename WARNINGn macros to WARNn since it prooved error-prone
1752 2004-04-19 Martin Quinson
1754 - register init/exit functions within gras module mechanism
1755 - send/receive function.
1756 Convertion is not implemented, but short-cutted if not needed.
1757 struct/array elements are sent one by one (instead of block-wise), but
1758 nobody really cares (yet). Get a prototype before optimizing.
1759 - tests (using a file socket) for DD send/receive on:
1760 - base types: int, float
1761 - array: fixed size, string (ie ref to dynamic string)
1762 - structure: homogeneous, heterogeneous
1763 - chained list, graph with cycle
1764 Believe it or not, valgrind is not too unhappy with the results. The
1765 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1769 - Bugfix when using a filename explicitely (instead of '-')
1771 2004-04-09 Martin Quinson
1773 - factorize more code between RL and SG in socket creation
1774 - Complete the implementation and tests of:
1776 o file (only in RL, and mainly for debugging)
1778 I lost 3 days to design a portable address resolver, and then decided
1779 that the prototype mainly have to run on my box.
1780 Addressing portability too early may be like optimizing too early :-/
1782 - use gras_init in the Tests instead of the crappy parse_log_opt
1783 (the latter function is removed)
1784 [Conditional execution]
1785 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1786 [Code reorganisation]
1787 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1788 Build examples against the RL library, since there is no way to disable
1789 its creation for now.
1791 For information, the beginning of coding on GRAS was back in june
1792 2003. I guess that every line has been rewritten at least twice since