1 SimGrid (3.5) unstable; urgency=low
4 * Change the way a task is considered as ready. Instead of removing
5 dependencies when a task is done, a counter is decreased. This way,
6 it is always possible to reach ancestors thanks to the
7 SD_taks_get_parents function (even after the end of the simulation.)
8 * Change the level of some logs in sd_global from INFO to VERB. The
9 corresponding tests are now less verbose
10 * Change the return type of SD_Simulate from (SD_task_t*) into
11 xbt_dynar_t. This function was in handling a dynar internally and
12 converted it into a NULL terminated array for historical reasons.
13 * Rename the SD_READY (all dependencies are satisfied and task is
14 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
15 dependencies are satisfied) state.
16 This prevents a confusion between the notion of "ready to schedule"
17 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
18 simulated" (SD_RUNNABLE) used by the simulation kernel.
19 * Add an example that scheduling a DAX on an heterogeneous platform
20 using a Min-Min strategy
21 * new function SD_dotload(char*) to load a DAG described in dot
22 format. This loader and the corresponding examples require the
23 installation of the graphviz library.
26 * Change the underlying waiting queue in semaphores so that a process
27 can wait on several of them simultaneously (as in waitany).
28 * Fix the way to handle tokens in semaphores so that all access patterns
29 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
32 * New data container: setset (set of sets of elements)
33 * New module: mmalloc (mapped malloc)
34 * New function: xbt_dict_cursor_set_data()
35 * New function: xbt_dynar_sort()
36 * New function: xbt_dynar_is_empty()
37 * New function: xbt_fifo_get_last_item()
38 * Bug fix in xbt_dynar_shrink(): use the right element size
39 * Use library init/fini functions for our initialization.
40 - you can use logs and other feature as soon as you want in your
41 code (even before the xbt_init / MSG_init)
42 - xbt_exit is now a no-op and produce a warning when used.
45 * New function: gras_msgtype_get_name()
47 Build chain: bug fixes and further polishing
48 * Fix the 'make install' target.
49 No need to use 'make install-simgrid' anymore
50 * Introduce a 'make dist' target compiling a *source* archive
51 'make package' compiles a binary archive (not well tested)
52 * Compile java files only on need
54 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
56 SimGrid (3.4.1) stable; urgency=low
58 The "Polishing easter eggs is probably a good idea" release.
59 This is a bug fixes release only.
63 * Fix a bug preventing the tasks from begin garbage collected.
66 * Fix a bug occuring when a host involved in a communication fails.
67 This was not detected properly by the other peer involved in the
68 communication. Now, it's reported as a network error.
71 * Warn the user about loop dependencies in data flow of DAX files
72 * Obey the control-flow dependencies of DAX files
75 * Add option "enable_smpi" allowing to not compile SMPI.
76 Probably useful for the (Mac) users experiencing a build error here
77 * Improve the detection of lua5.1 and ruby1.8
79 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
81 SimGrid (3.4) stable; urgency=low
83 The "Easter in Cargese" release. Also known as (major changes):
85 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
86 ~> bindings were greatly improved
87 ~> new tracing infrastructure for better visualization introduced
89 * the "Welcome to configury modernity" release.
90 ~> we switched from autotools to cmake, and improved our cdash
93 A more detailled list of changes follow (full detail in svn log).
95 Java Bindings: Various Cleanups
96 * (install java-gcj-compat-dev on debian-like to use them)
97 * Remove put/get: no need to export deprecated interface in Java
98 Use send/receive instead.
99 * Cleanup the examples and add a README per directory
100 * Remove example autoDestination (that's the only way to go now)
101 * Remove example explicitDestination (was a plain copy of basic)
102 * Make JniException a runtime exception, so that there is no need to
103 declare the fact that you may encounter such a beast. I guess that
104 nobody will ever want to survive such error.
105 * Create specific errors for each MSG case of failure:
106 host failure, transfer failure, timeout, task cancelled
107 * Cleanup the exceptions that may get thrown by each function
108 * Other internal cleanups in Java bindings. Performance still bad :/
109 Ruby and Lua Bindings: create them
110 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
111 * That's new and great, you should try them out.
112 Same functionalities than Java bindings, only even less polished
114 * Kill the useless "rate" argument of SD_task_get_execution_time()
115 Everyone used to provide -1 as a value, it was not used, and the
116 semantic of a possible use wasn't even clear.
117 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
118 either as comm costs or compute costs to mean that there is no
119 such thing for that specific task.
120 * Add a SD_task_set_name() function
121 * Fix SD_task_unschedule() on typed tasks
122 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
123 * In DAX loader, accept useless 'level' attributes to <job> since
124 LIGO DAGs have them (seem to be to ease graphical representation).
126 * Add an example masterslave_mailbox.c using send/receive and not
127 the deprecated put/get interface.
128 * Kill the MSG_paje_output() function. It's a noop since 2 years.
129 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
131 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
132 (declare MSG_USE_DEPRECATED to still have the old name)
133 * Add a MSG_task_set_data() function
134 * About trace replay (see examples/msg/actions):
136 - Allow to work with splitted trace files for each process
137 Give the specific trace file as argument of each process,
138 and call MSG_action_trace_run(NULL)
139 You can still have one merged file for all processes.
140 - Fix implementation of collective operations
141 * Allow task_execute() on 0-sized tasks (closes #10063)
143 * This is the first release of SimGrid where SMPI is not considered
144 beta anymore (even if some corners should still be improved)
145 * Port over the new SIMIX_network submodule (internal refactoring)
146 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
147 * Implement more missing elements of the standard:
149 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
151 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
152 MPI_Scan() MPI_Get_processor_name()
153 - Added implementation of missing case for Alltoall (warning: it's
154 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
155 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
156 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
157 SMPI_MPI_Allgatherv()
159 - MPI_Waitsome() was broken
160 - Allow relative includes in smpicc
161 - Command line cfg argument 'reference_speed' was ignored...
162 - Some functions did not properly lead to auto-benching of user code
163 - smpicc passes -O2 by default (just like openmpi one)
165 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
166 * Bug fixes about timeouts during communications
167 * add SIMIX_message_sizes_output() as a pimple to write to file the
168 amount of messages per size. Use gnuplot to get histogram.
169 Pimple because that's the only user-visible function of simix,
170 defined directly in xbt.h (irk, sorry)
172 - Add a SIMIX_sem_get_capacity() function
173 - Fix interactions with processe resume/suspende
174 - release_forever() was stupidly broken
175 - Fix SIMIX_display_process_status() for processes in a semaphore
176 - Make SIMIX_sem_block_onto() user-visible
177 * Refactoring context stuff:
178 - Use pseudo-OOP for better modularity
179 - reimplement SIMIX_process_kill() without process_schedule() so
180 that the latter can take as invariant that it is called from
182 - Merge context_start into context_new for sake of simplicity
184 * Add a Vivaldi network model, coded live during SUD'10 ;)
185 * Rename configuration variables to start a hierarchy:
186 o cpu_model -> cpu/model
187 o network_model -> network/model
188 o workstation_model -> workstation/model
189 * New configuration variables:
190 o network/bandwidth_factor: correction to bandwith
191 o network/latency_factor: correction to latency
192 o netwotk/weight_S: correction to the weight of competing streams
193 * Add a long description to the models, that users can see with such
194 argument on the command line: --cfg=cpu/model:help
195 * --help-models display the long description of all known models
197 * config: add the ability to set a default value after registration
198 Does not override any previously set value (e.g. from cmd line)
199 * dict: allow to have integer key and data.
200 When so, you need to use the following functions
201 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
202 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
203 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
204 In contrary to regular dicts, the key is not malloced before copy.
205 Mixing scalar and regular elements in the same dict is not tested
207 * Allow to use xbt_dynar_shrink() to expend the dynar instead
208 Tracing for Visualization:
209 * SimGrid is now instrumented in order to generate a trace file for
210 visualization analysis: to use it, need to compile SimGrid with the
211 "tracing" option enabled, and instrument the program using SimGrid with
212 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
213 (among other functions).
214 * The instrumentation only traces the platform utilization for now
215 * Documentation to use the tracing functions and how to analyze the
216 traces with the Triva tool is written.
217 * More information about: SimGrid FAQ (in the section Tracing Simulations
220 * We moved to cmake as default build system. Autotools support will
221 be dropped soon. Check the FAQ for more info about how to use it.
222 * Greatly improved our cdash/ctest interactions
223 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
224 * Added memory checking tests with valgrind; lot of memleak fixing.
225 This may be the first release of simgrid with so few memory issues
226 * Added code coverage tests.
227 Our coverage is still improvable, but at least we see it on cdash.
229 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
231 SimGrid (3.3.4) stable; urgency=low
233 The "Desktop Grid needs love too" release (also called Xmas release).
236 * Major speedup in the maxmin system solving by using lazy evaluation
237 Instead of solving completely the maxmin system at each iteration,
238 only invalidate (and recompute) the modified parts.
239 This new feature is enabled in default models but you can try to
240 turn it on with "--cfg:maxmin-selective-update=1" for other models.
241 * Cas01 IMproved as default CPU model
242 This CPU model is the same Cas01 model, but it uses the
243 maxmin-selective-update flag and a heap structure to manage
244 actions on SURF kernel.
245 It reduces the complexity to find the next action to finish and,
246 consequently, it's faster than the old Cas01.
247 This is the new default CPU model (Cas01).
248 * Rename the old Cas01 model to Cas01_fullupdate
249 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
250 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
251 * CpuTI (CPU Trace Integration)
252 A new CPU model whose objective is simulate faster when using
253 availability trace files.
254 Instead of using a full featured, over engineered maxmin system for
255 CPU modeling, this model does the pre-integration of traces files
256 to calculate the amount of CPU power available, and so, executes
257 faster than the old CPU models.
258 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
259 * Use LV08 as default network model since it gives better accuracy
260 for small messages and shouldn't change things for big ones.
261 Use --cfg=network_model:CM02 to get the previous behavior.
264 ******************************************
265 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
266 ******************************************
267 * The new CPU model may changes simulations!
268 The point is that events occurring at the exact same timestamp
269 are not scheduled in the same order with the old and new
270 version. This may be enough to completely change the execution
271 of simulations in some cases.
272 * The new network model will change simulations!
273 This new model is more realistic than the previous one, so you
274 should consider redoing your old experiments with this model.
275 Sorry for the inconvenience.
278 * Introduce the supernovae compilation mode
279 When compiled that way, the whole SimGrid (or almost) is put in a
280 single compilation unit and compiled in one shoot.
281 This is to help gcc which has difficulties to inline stuff from one
283 The speedup seem to be above 15%, althrough more tests are needed on
284 amd64 to confirm that gain.
287 * Port of MSG's mailbox on top of SIMIX network
288 The put/get mechanism was greatly simplified on the way.
291 * New SIMIX network module. Provides:
292 - Mailbox: rendez-vous mecanism to find with who you want to speak
293 - Synchronous send/recv: easier and hopefully faster since the
294 logic is handled in the maestro process directly now
295 - Asynchronous send/recv: you dreamt of it? It's here now
296 Too bad that nobody cared enough to propagate the change to MSG.
297 * Add semaphores as SIMIX synchronization mechanism.
300 * new function SD_daxload(char*) to load a DAX file
301 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
302 * Introduce typed tasks. Specify its kind and cost at creation.
303 At scheduling, just give where it should be placed, and the cost
304 for each involved resource is automatically computed.
305 Existing constructors so far (more to come of course):
306 - SD_task_create_comm_e2e() for end-to-end communication
307 - SD_task_create_comp_seq() for sequential computation
308 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
309 * new function SD_task_dump() for debuging display
310 * new function SD_task_dotty(task,FILE*) writing to file the info
311 about the task in dotty format
312 * SD_task_dependency_exists() can now cope with having one of its
313 arguments NULL. If so, it tests whether the other argument has any
315 * Add getters on list of preceding/following tasks:
316 SD_task_get_parents(task) and SD_task_get_children(task)
317 * Add getters on amount of workstations and list:
318 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
319 * Add getter on task kind: SD_task_get_kind(task)
320 * Update the start_time and finish_time of tasks on completion/failure
321 * Bugfix: Remove task from state swags when destroyed
324 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
327 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
328 dynar with one separate thread per value of the dynar.
329 * Change the prototype of xbt_thread_create(), sorry.
330 Added a boolean parameter indicating whether we want to join this
331 thread (used in SG only for now)
332 * Implement xbt_thread_join and xbt_thread_yield in SG also.
335 * GTNetS wrappers should now be usable again (and betterly tested too)
336 * Fix a major regression from 3.2 where the timeout provided to
337 MSG_task_put_with_timeout() was used as absolute time before which
338 the comm should be done.
339 * Start to fix the <cluster> tag.
340 - Internal links should be good now (beside of the loopback, which
341 use the private link instead)
342 - paths to the external world is still rather broken
343 - the <route:multi> tag is just broken. Actually that's brain-dead.
344 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
345 to make it less stupid
346 ** Check your platform with teshsuite/simdag/platforms/flatifier **
347 * Fix a source-level compatibility glitch from 3.2: after defining
348 MSG_USE_DEPRECATED, you can use the old name
349 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
350 * Allow to compile from the SVN with automake 1.11
351 * Fix some problems when using the "start_time" tag in deployment XMLs.
352 * Fix #8569: XBT/synchro.h has redundant declarations
353 * Fix #8563: MSG return values and exceptions
354 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
355 * Integrate patch #8636: Obey DESTDIR when installing documentation.
356 Thanks to Robson Peixoto.
357 * Fix a vicious bug in dictionaries inducing that some elements were
358 not freed on xbt_dict_free()
360 Portability report of this version:
361 * Main portability targets:
362 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
364 - mac leopard on i386
365 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
366 but nothing critical.
367 * Other platforms: windows, AIX and others were not tested for this release
369 Timing report of this version:
370 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
371 scenario parameters). From 8h to a few seconds in desktop grid settings.
372 * Supernovae brings about 25% speedup on i386.
374 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
376 SimGrid (3.3.3) stable; urgency=low
378 The "Need for Speed" release.
380 The timings done to validate the 3.3.2 were faulty.
381 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
383 The problem was a conversion from a manually handled vector to
384 xbt_dynar_t on the critical path.
385 xbt_dynar_foreach calls functions, inducing stack management crap.
387 We inlined these functions and xbt_dynar_foreach is now breath taking.
388 We also inlined xbt_swag_belong on the way.
390 Here are some approximate speedup measurements (on master/slaves
391 simulations lasting between 10s and 20s each):
392 3.3.1 -> 3.3.2: about same performance
393 3.3.2 -> 3.3.3: 40% speedup
394 3.3.1 -> 3.3.3: 40% speedup
395 3.3.1 with inline patch -> 3.3.3: 30% speedup
397 Our reading is that the refactoring which occurred in 3.3.2 made us
398 suffer much more from the xbt_dynar_foreach low performance, but
399 once we solved this, this refactoring proved to be very performance
400 effective. From the 40% speedup, somehow, 10% are due to the
401 inlining and 30% to the refactoring.
403 That's a pitty that gcc cannot inline functions placed in other files
404 alone. We have to choose between:
405 - break the encapsulation (by putting private data structures and
406 accessors in headers files to help gcc)
407 - live with low performance
408 - switch to a decent compiler such as icc (not quite possible).
410 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
412 SimGrid (3.3.2) stable; urgency=low
414 The "Simplicity does not preceed complexity, but follows it" release.
416 The main contributors of this release were (lexical order):
417 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
420 * Extract the routing logic into its own object.
421 (was dupplicated in network.c and workstation_LV07.c;
422 Allows to implement other ways of storing that info)
423 => kill now useless network_card concept
424 - Use dynar to represent routes (instead of void** + int*)
425 - kill link_set (use surf_network_model->resource_set instead)
426 - Add a command-line option to choose the routing schema to use
427 - Add three new models:
428 * Floyd (shortest path computed at initialization)
429 * Dijikstra (shortest path recomputed all the time)
430 * Cached Dijikstra (shortest path computed on need)
431 All these models where contributed by Silas De Munck, and are
432 described in his ICCS09 paper.
434 * Simplify model declaration
435 (less redirections, less function to write when defining a model)
436 - Factorize stuff between models:
439 surf_model_resource_set(model)
440 surf_model_resource_by_name(model, name)
441 - Unify the types of models in s_surf_model_t (using an union)
442 - Embeed fields of common_public directly into s_surf_model_t
443 - Rename model methods:
444 action_free ~> action_unref
445 action_change_state ~> action_state_set
446 action_get_state ~> action_state_get
447 - Change model methods into functions :
448 (model)->common_public->action_use ~> surf_action_ref
450 * Implement a generic resource; use it as ancestor to specific ones
451 (allows to kill duplicated code in models)
452 Drawback: timer command don't need no name nor properties;
453 workstation_CLM03 don't need no properties
454 (but I guess we can live with those few bytes wasted)
456 * Improve the action object model
457 - implement a constructor avoiding dupplicated code about field
458 initialization in generic_action part.
460 * Kill the SDP model: it has an external dependency, is deprecated
461 in flavor of modern lmm models, and didn't compile since a while
464 * Relocation of the context module from XBT to SIMIX.
465 (the context were decoupled from the simix processes, duplicating a lot of code)
466 => a lot of code was factorized
467 - less overhead is introduced during scheduling
468 - simpler API for the context factory
469 - the logic for process creation,destruction and manipulation was simplified
470 * Simplification of the s_smx_process_t data structure.
471 => accesing the simix level data associated to a process is faster now,
472 and the code is a lot more readable.
475 * Implement some more MPI primitives:
476 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
477 -implementation: Bcast: flat or 2-ary tree (default),
480 Allreduce: Reduce then Bcast
481 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
482 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
483 Alltoallv: flat tree, like ompi
485 * Add support for optimized collectives (Bcast is now binomial by default)
486 * Port smpirun and smpicc to OS X
489 * Kill SD_link_get_properties: hard to maintain and makes very little sense
490 Shout out if you used it.
493 * Display the list of still queued messages in SG mode when existing
497 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
498 * Add xbt_graph_node_get_outedges() [Silas De Munck]
499 * Add xbt_str_from_file(FILE*)
500 * Add xbt_dict_get_key achieving a linear reverse search
501 * Remove the context module
503 Portability report of this version:
504 * Main portability targets:
505 - Linux(debian)/x86/context
506 - Linux(debian)/x86/pthreads
507 - Linux(debian)/amd64/context
508 - Linux(debian)/amd64/pthreads
509 On these, we still have the eratic breakages of gras/pmm and
510 amok/saturate_sg reported in previous version. We still think
511 that the tests are the cause of the fault, not the tested code.
513 - Mac OSX Leopard/x86/context
514 Still false negative in tesh autotesting.
515 Smpi still fails, but this time because readlink does not accept -f
516 Everything seems to work properly beside of that.
519 - AIX version 5.3 (only tested contexts this time)
520 Smpi still fails there because mktemp is not installed.
521 Everything seems to work properly beside of that.
523 I managed to compile it for the first time, but several breakages.
524 Won't delay the release for this exotic platform.
526 * Windows: it's still lagging behind. If you want to help, please
529 Timing report of this version:
530 This version seem to be more than 5% faster than 3.3.1 (on linux
531 64bits with contextes). The gain is less than expected, we are
532 investigating this for next release.
534 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
536 SimGrid (3.3.1) stable; urgency=low
539 * Implement a --cfg-help to show existing configuration variables
540 * Build chain do not require doxygen in maintainer mode
543 * fix a bug on struct sizeof computation, which prevented the
544 exchange of arrays of structs in some conditions
545 - added a regression test about this in datadesc_usage
546 * Allow the exchange of 0-long dynamic vectors.
547 - for that, use -1 as indicator of dynamic size instead of 0
548 - This implied to change any size from unsigned long to long,
549 reducing a bit communication abilities, but I guess that with
550 64bits being quite common, this is more than enough.
551 - This also induce a protocol change, thus bumping network protocol
552 version from 0 to 1 (if we have external users, we have to get
553 clean on that point too ;)
554 - added two regression tests about this in datadesc_usage
555 * Be more verbose when propagating local exceptions
556 This helps debugging.
557 * Display the status of simulated processes when receiving SIGINT in
561 * Allow to control the simulation from a trace file.
562 New functions MSG_action_register() and MSG_action_trace_run()
563 The first one allows to associate a function execution to each
564 kind of action while the second one parses a trace file and
565 triggers the corresponding actions within the system.
566 For now, only a toy example is provided in examples/msg/actions
567 * Add an exemple of process migration in examples/msg/migration
568 * Fix a bug in task exchange which broke MSG_task_get_sender()
569 Add a teshsuite regression test for that.
570 [Bug: if MSG_task_get_sender() is called after sender exit,
572 * Fix a bug which prevented suspend/resume to work properly
573 * Display the status of simulated processes when receiving SIGINT
574 This fixes a regression of v3.3. due to the introduction of SIMIX
575 * Bug fixing in failure management:
576 - trace could not start by a failure at time 0
577 - failure during communications were not working
580 * Add SIMIX_process_set_name() to change the name of the current
581 process in the log messages.
582 * Store smx_hosts in a dict since we only retrieve them by name
583 * Move the configuration infrastructure to surf
586 * Move the configuration infrastructure to surf
589 * Massive internal cleanups:
590 - Store internal structures on processes instead of hosts (allows
591 to have more than one process per host, in addition of being more
593 - Cleanup the initialization/finalization process
594 - Kill a whole bunch of unneeded synchronization:
595 processes run in exclusive manner within the simulator
596 - Move queues from global tables to process data fields
598 - now accept -platform and -hostfile arguments
599 - Pass the right rank value to processes according to the hostfile
600 * Compile the examples by default, and use them as regression tests
601 * Implement MPI_Wtime()
602 * Change the reference speed to a command line option
605 * TCP_gamma can now be specified as command line option using
606 --cfg=TCP_gamma:10000000.0
607 * Change the --surf-path cmd line option into --cfg=path:
610 * Also include strbuff from xbt.h public header
611 * xbt_ex_display(): do not free the exception after displaying
612 This allows to do more with the given exception afterward.
613 Users should call xbt_ex_free() themselves.
617 Portability report of this version:
618 * Main portability targets:
619 - Linux(debian)/x86/context
620 - Linux(debian)/x86/pthreads
621 - Linux(debian)/amd64/context
622 - Linux(debian)/amd64/pthreads
623 These targets fail about 1/10 of times on gras/pmm, but we believe
624 that this is because of the test, not because of simgrid.
625 amok/saturate_sg fails even more rarely, and the test may not be
628 - Mac OSX Leopard/x86/context
629 The test suite still spits tons of errors because some obscure
630 force prevents us from removing the temporary directories
631 arguing that they still contain some metadata I've never heard of.
632 Smpi fails because seq is not installed.
633 Everything seems to work properly beside of that.
636 - AIX version 5.3 (both contexts and pthread)
637 Smpi still fails there because mktemp is not installed.
638 XML inclusions seems rosty on AIX.
640 * Windows: it's still lagging behind. If you want to help, please
643 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
645 SimGrid (3.3) stable; urgency=high
649 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
652 * Introduce the SIMIX module: factorize code between MSG and GRAS.
655 Until now, GRAS were using MSG as an interface to SURF. It was
656 quite difficult because both interface have several differences
657 (MSG channels vs GRAS sockets were the most notable point).
659 This also opens the gate to SMPI (which should occur soon) and speed
660 up simulations by to 40% (even if it were not the main goal).
662 **************************************
663 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
664 **************************************
665 The point is that events occuring at the exact same timestamp are
666 not scheduled in the same order with the old and new version. This
667 may be enough to completely change the execution of simulations in
668 some cases. Sorry for the inconvenience.
670 * Cleanup and upgrade the XML format to push further scalability
671 issues (check http://hal.inria.fr/inria-00256883/ for more info)
673 * Improve the testing infrastructure with tesh. Now a very large part of
674 the code is tested not only by being run but also by checking that the
675 output match an expected output [Mt].
677 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
678 is really less memory-demanding, which should allow you to use
679 larger files in SimGrid [AL].
681 * Inform valgrind about our contextes, so that it becomes usable
682 with the default (and more effecient) version of SimGrid
683 [contributed by Sékou Diakite, many thanks]
686 * Introduce a listener thread in charge of receiving incomming
687 messages from the network. It allows to overlap communication and
688 computation but most notably, it removes some stupid deadlocks due
689 to the fact that so far, a process could not send and receive at
690 the same time. This made most non trivial communication schema
692 * Convert the PIDs from long int to int to match the MSG ones (and
694 * New function: gras_agent_spawn() to launch a new process on
695 current host. Only working in simulation for now. [Mt]
696 * New function: gras_os_hostport() returning a constant form (ie,
697 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
700 * Make the backtrace of exceptions more human readable [Mt]
701 * New module: xbt/str [Mt]
702 a ton of string utility functions (split, join, printf to a newly
703 allocated buffer, trim, etc)
704 * New module: xbt/hash [Mt]
705 SHA1 hashing algorithm (more to come if needed)
706 * New module: xbt/synchro [Mt]
707 synchronization tools (mutex and conditions) working the same way
708 in simulation and in real life (mainly useful for GRAS, but not
710 * New module: xbt/queue [Mt]
711 classical producer/consumer synchronization scheme
712 * xbt_dynar_new_sync() creates a synchronized dynar. All access
713 (using the classical functions will get serialized) [Mt]
714 * Make dictionary internal table dynamic. No need to specify its size
715 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
717 * Make sure the log channels are organized as a tree under windows
718 (because of ANSI C compatibility issue, any channel were child of
722 * Cleaned many thing in surf and fixed a few bugs [AL].
723 * Add a nice command line configuration mechanism to compose models [AL].
724 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
725 the previous one (KCCFLN05). It relies on something that looks like
726 a max-min sharing mechanism but cannot be written as such. A new solver
727 was thus designed [AL].
728 * Add a new solver to lmm. Based on Lagrange optimization and
729 gradient-based descent, it enables to efficiently maximise systems s.a
731 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
733 This solver enables to propose two new network models for TCP Reno and
734 TCP Vegas based on Low's work. These models still need to be fully
735 tested though [Pedro Velho].
738 * Bug fix in SD_simulate. Now the time bound given as argument is
740 * Use the new parallel task model (ptask_L07) as default.
741 * Use the SURF command line configuration mechanism.
742 * 0-size tasks (for synchronization) should now work.
744 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
746 SimGrid (3.2) stable; urgency=high
750 We still experience issues on this platform, but we believe that at
753 GRAS API BREAKAGE (for simplification purpose, sorry):
754 * the gras_msgtype_by_name is not used anymore. Instead of
755 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
756 you can write (and must)
757 gras_msg_send(toserver, "request", &request);
758 - If you still want to pass a gras_msgtype_t to the function (to cache
759 the type and avoid the lookup time), use the gras_msg_send_() variant.
760 - Impacted functions:
761 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
762 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
763 * The callbacks are now expected to return 0 when everything went well
764 (just like the main() function)
766 GRAS new features and improvements:
767 * New module mecanism where user code can use per process globals [Mt]
768 This is similar to gras_userdata_*() functions, but for libraries. It
769 factorize some code developped over and over in the examples and AMOK.
770 It has still to be documented and used (only amok/peermanagement is
772 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
773 loss when they were small enough to fit into the buffer and sent quickly
774 enough so that they can all get received in one shoot.
775 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
776 if not found. Use the *_or_null() variant for the old semantic.
777 * In gras_msg_handle, do not discard messages without callback.
778 They are probably messages to be explicitly awaited later (ie, proofs of
779 mis-synchronization in userland since they are sent before being awaited)
781 * gras_socket_meas_send/recv: semantic changed!
782 The numerical arguments used to be (1) the total amount of data to send
783 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
784 messages. This was need for the fool willing to send more than MAXINT
785 bytes on quite fat pipes.
788 * Do really rename the hostmanagement module to peermanagement. [Mt]
789 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
790 but this is rather new and this was documented in the module
791 documentation (poor excuses, I admit)
792 * Bandwidth measurement semantic changed! This follows the changes to
793 gras_socket_meas_send/recv explained above.
796 * A sequential mode has been added to the workstations. When a workstation
797 is in sequential mode, it can execute only one task, and the other tasks
798 are waiting in a FIFO. [Christophe Thiery]
801 * The KCCFLN05 workstation model now handles parallel tasks. It is the
802 model for SIMDAG. [Christophe Thiery]
803 * Bug fix in the maxmin solver: Some values were close to 0 instead of
804 equal to 0, which caused some bad behaviors in
805 saturated_constraint_set_update. I now use a threshold mechanism like in
809 * When running manually src/testall, you select specific units [Mt]
810 testall is the result of our cunit mecanism, and should replace all
811 the scripty thingy around since bash don't run easily on billware.
813 * A mallocator system has been added. [Christophe Thiery]
814 Mallocators allow you to recycle your unused objects instead of freeing them
815 and allocating new ones.
817 Documentation update:
818 * FAQ reworking + New FAQs:
819 - "Valgrind spits tons of errors!" [Mt]
820 - "How to repport bugs" [Mt]
821 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
822 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
823 - Communication time measurement within MSG [AL]
824 - I experience weird communication times when I change the latency [AL]
827 - an introduction to the framework and to the used communication model
828 - an initiatic tour introducing the most proheminent features:
830 . Lesson 0: Installing GRAS
831 . Lesson 1: Setting up your own project
832 o Part 2: Message passing
833 . Lesson 2: Exchanging simple messages
834 . Lesson 3: Passing arguments to the processes (in SG)
835 . Lesson 4: Attaching callbacks to messages
836 . Lesson 5: Using globals in processes
837 . Lesson 6: Logging informations properly
838 . Lesson 7: Using internal timers
839 . Lesson 8: Handling errors through exceptions
840 . Lesson 9: Exchanging simple data
841 . Lesson 10: Remote Procedure Calling (RPC)
842 . Lesson 11: Explicitely waiting for messages
843 . Recapping of message passing features in GRAS
844 - A HOWTO section containing:
845 o HOWTO design a GRAS application
846 More are due, of course. They will come latter. In the meanwhile, you can
847 check the examples which are still here.
849 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
851 SimGrid (3.1) stable; urgency=high
855 There was a stack corruption somewhere, visible only when optimizing
856 with these versions. [Vince]
859 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
860 module that enabled to play with Directed Acyclic Graphs. It is built
861 directly on top of SURF and provides an API rather close to the old
862 SG. Some old codes using SG are currently under rewrite to check that
863 all needful functions are provided. [Christophe Thiery]
866 * Complete rewrite of the KCCFLN05 workstation model. It is now an
867 extension of the classical CLM03 model that gracefully handles
868 failures. This is now the default model for MSG and GRAS. It doesn't
869 handle parallel tasks yet though. [AL]
870 * Bug fix: Weights were not correctly set in the network part.
871 WARNING: This may have resulted in incorrect results with simulations
872 where there are more than one flow on a given link. [AL]
875 * After a (long ?) discussion on simgrid-devel, we have decided that the
876 convention we had on units was stupid. That is why it has been decided
877 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
878 WARNING : This means that all previous platform files will not work as
879 such with this version! A warning is issued to ask users to update
881 A conversion script can be found in the contrib module of the CVS, under
882 the name contrib/platform_generation/surfxml_update.pl [MQ]
885 * Bug fix: Processes were started in reverse order, wrt deployment file.
886 WARNING: if your code relies on this bug, please fix it. [AL]
887 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
888 being executed on two different locations. [AL]
889 * Bug fix: Failures are now better supported thanks to Derrick's tests
890 (there was many failure situations I hadn't thought of and that weren't
891 correctly handled). [AL]
892 * New function: MSG_host_is_avail indicates you whether a given m_host_t
896 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
897 Exception occurring on server-side are propagated back to client (!).
899 API CHANGE: the callback changed their prototype. Change:
900 int my_handler(gras_socket_t expeditor, void *payload_data) {
902 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
903 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
905 * New! function: gras_msg_handleall to deal with all messages arriving
906 within a given period.
907 * New! function: gras_socket_server_range to get a server socket in a
908 range of port numbers (ease to avoid port number conflicts) [MQ]
909 * New! gras processes display their backtrace when they get a SIGUSR1
910 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
911 Sweet to debug RL processes [MQ]
915 - Do not force experiment sizes to be expressed in kb, or it becomes
916 impossible to measure the latency this way (needs one byte-long tests)
917 WARNING: this changes the amok_bw_* function semantic. [MQ]
918 - Implements the link saturation stuff. [MQ]
919 * Peer management module:
920 New! module factorizing code that we wrote over and over [MQ].
923 * New module: cunit (my jUnit implementation in ansi C) [MQ]
924 - Test units are placed directly into the library code, they get extracted
925 automatically and placed into the src/testall binary.
926 - Convert most of the XBT tests to this system.
927 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
928 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
929 live in the same C code. This required to change a little bit the API
930 of surfxml parsing but shouldn't be an issue for end-users. [AL]
931 * New module: sparse graph structure with basic algorithms (this is work
932 in progress and the API is not considered to be frozen yet). [AL]
933 * Display more information on backtraces: source line & function names are
934 now displayed just like valgrind does (rely on addr2line tool) [MQ]
935 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
936 * Reworked a little bit some #include statements to load only required
937 headers. Some user code that relied on SimGrid to include stdlib or
938 stdio may need to include it by themselves. [AL]
939 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
941 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
942 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
943 but I guess that almost nobody used those parts. [MQ]
945 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
947 SimGrid (3.0.1) stable; urgency=low
950 * Unfortunately, I had missed 5 misnamed functions:
951 xbt_fifo_item_t xbt_fifo_newitem(void);
952 void xbt_fifo_freeitem(xbt_fifo_item_t);
953 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
954 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
955 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
956 They're now deprecated. Please use their new versions:
957 xbt_fifo_item_t xbt_fifo_new_item(void);
958 void xbt_fifo_free_item(xbt_fifo_item_t);
959 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
960 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
961 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
963 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
964 * Documentation: xbt_log module unmercifully reworked [MQ]
965 * Bugfix: there was a problem with the ending of contexts with
966 the pthread backend. It caused some weird deadlock or behavior
967 depending on the pthread implementation. [AL]
968 * Bugfix: get the exceptions raised in the simulator repport where
969 and why they come from when they are not catched in time [AL, MQ]
972 * Bugfix: Do repport the error when two non-connected hosts try to
973 exchange data (Thanks to Flavien for stumbling into this one) [AL]
976 * Add additionnal checkings on communications. Assert that two
977 communicating hosts are connected by a set of links... [AL]
980 * Add additionnal checkings on channel values in communication [AL]
981 * New: MSG_task_get_source to see on which host a task was generated [HC]
982 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
983 the number of tasks waiting to be received on channel and sent
985 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
986 waits for the first task coming from a given host.. [AL]
988 GRAS new functionnalities: [MQ]
989 * Enhance the parsing macro to allow the size of multidimentional objects
990 to be given thru annotations.
991 * New example (and documentation): Matrix Multiplication a la RPC
992 (as when I was young!) and fix a bunch of bugs found on the way.
994 GRAS performance improvements: [MQ]
996 * Reduce the amount of cbps creation/destruction by making it static to
997 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1000 * Change libdata to a set so that we can search for stuff by ID (and thus
1001 reduce the insane amount of dict lookups)
1004 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1005 RL communication on top of SG since it's so uneffective.
1006 It may also be used for inter-thread communication in RL, one day.
1007 * Use gras_datadesc_copy() to exchange messages on top of SG
1009 - improve message exchange performance on top of SG
1010 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1011 * Don't exchange on the network the size of the used part of buffer,
1012 instead, specify the possible buffer size to read().
1014 - reduces the amount of read/write calls (one pair per exchange)
1015 - reduces the amount of exchanged data (the size)
1016 - allows to retrieve all arrived data on receiver side, if we don't need
1017 it right now (subsequent read will peek the buffer)
1018 - allows the receiver to proceed with the begining of the stream before
1019 everything is arrived
1020 - make it possible to build an iov transport (using readv/writev)
1022 - take care of the data with non-stable storage (like stacked data),
1024 * If possible, TCP send uses vector I/O (when writev() is here)
1025 - Don't use it for receive since we send data sizes and data on the
1026 same stream, so we wouldn't be able to chain large amount of chunks
1027 before having to flush the stuff to read the size.
1028 * Rework the transport plugin mecanism to simplify it and reduce the
1029 amount of pointer dereferencement when searching for the right function
1032 * I guess that now, we do almost as few system calls as possible while
1033 doing as few data copy as possible.
1035 To improve it further, we could try to send all the sizes first and then
1036 all the data (to use iov on receiving size), but it's only a partial
1037 solution: when you have 2 dimensional data, the sizes of the second
1038 dimension is data of the first dimension, so you need 3 streams.
1040 I'm not sure the potential performance gains justify the coding burden.
1042 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1044 SimGrid (3.00) stable; urgency=high
1047 * New! Give the possibility to hijack the surf parser and thus bypass
1048 MSG_create_environment and MSG_launch_application. Have a look at
1049 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1051 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1053 SimGrid (2.96) unstable; urgency=low
1058 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1059 This deprecates the xbt_error_t mecanisms.
1060 It modifies (simplifies) all XBT and GRAS API.
1061 MSG API keeps unchanged (exceptions raised by XBT are catched from
1062 within MSG and masked with existing error handling facilities)
1065 * New! Add a FATPIPE model. [AL]
1066 * New! Add a parallel task model. [AL]
1067 * New! Add automatically a loopback interface (in the default
1068 network model) if none was precised.
1071 * Bugfix: MSG_process_resume now works with the current running process.
1073 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1074 * Modification of MSG_task_get_compute_duration. Once a task has been
1075 processed, the value returned by this function is now equal to 0. [AL]
1076 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1077 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1078 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1079 of a m_task. For now, MSG_task_cancel only works with computation
1081 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1082 of the processor (in Mflop/s) regardless of the current load on the
1084 * API Change: use proper naming convention for MSG_getClock and
1085 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1087 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1088 This function changes the priority of a computation task. This priority
1089 doesn't affect the transfer rate. A priority of 2 will make a task
1090 receive two times more cpu power than the other ones. This function
1091 has been added to suit the needs of Nguyen The Loc and hasn't been that
1092 much tested yet. So if it fails, please report it and send me your code.
1094 * API Change: removed all functions and types that were marked "deprecated"
1095 since many months. Renamed MSG_global_init_args to MSG_global_init.
1097 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1099 SimGrid (2.95) unstable; urgency=low
1102 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1103 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1104 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1107 * Bugfix: Allow absolute paths for platform description files [MQ]
1108 * Bugfix: do free the variables after use. Leads to drastic performance
1110 * Implement max_duration (ie, timeouts) on resources [AL]
1113 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1115 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1116 the needed support (timeouts) [AL]
1119 * Implement measurement sockets. You can now get the bandwidth between two
1120 hosts thanks to AMOK (see below). [MQ]
1121 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1122 dynar over the network (yeah) [MQ]
1123 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1124 * simple/static token-ring example. [Alexandre Colucci and MQ]
1125 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1126 in place before [MQ]
1129 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1130 type parsing. "short" and "long" modifiers were also ignored; other
1131 modifier (such as reference level) are still ignored. [MQ]
1132 * Embeed the buffer size within the buffer itself on SG. [MQ]
1133 That way, send() are atomic and cannot get intermixed anymore (at least
1134 the ones which are less than 100k; bigger messages still have the issue)
1135 * Array size pushed by the field, not by the field type (or each
1136 and every long int will push stuff to the cbps) [MQ]
1137 * use select() to sleep since it allows to portably sleep less than one
1140 GRAS (minor cleanups)
1141 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1144 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1145 * type category 'ignored' killed as it was never used and were difficult
1147 * whether a type can cycle or not is now a flag, leaving room for more
1148 flags (as "ignored", if we feel the need one day ;)
1149 * Rename raw sockets to measurement sockets since "raw" has another
1150 meaning in networking community.
1153 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1154 and offers features not belonging to GRAS but that most applications
1155 need. One day, it may be a set of plugins loadable at runtime.
1156 * New module: bandwidth
1157 bandwidth measurement between arbitrary nodes running this module. [MQ]
1159 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1161 SimGrid (2.94) unstable; urgency=low
1163 The first beta release of SimGrid 3 !
1167 * Update the main page and the FAQ. Adding references to gforge.
1170 * Add a gras_os_getpid function.
1173 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1174 * Extend the logs so that they also print PID, hostname, date, ... if
1176 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1177 and kill the old version which were in testsuite/
1178 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1182 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1183 for users. Instead, each programming environment (GRAS, MSG,...) use it
1184 when needed to provide such a feature to users.
1185 Don't shortcut the mecanism or you will also shortcut the virtualization
1186 you need on the simulator.
1190 * Cleanups in configury with regard to compile optimization/warning flags.
1191 Also add -fno-loop-optimize to any powerpc since it's the optimization
1192 killing gcc (< 3.4.0).
1193 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1194 needed by MSG examples complications
1195 * Borrow configury beautifications from PHP
1198 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1199 warning (thanks loris for stumbling into this one).
1200 * Bugfix: stop loading private headers (gras_config.h) from the public
1204 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1205 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1206 RL would imply a DNS resolver, which is *hard* to do in a portable way
1207 (and therefore delayed).
1208 * Implement a real timer mecanism and use it in timing macros. This allows
1209 to avoid rounding errors and get a 0.000005 sec precision in timing
1210 macros. While I was at it, various cleanups:
1211 - allow to declare more than one timed section per file (fix a stupid bug)
1212 - move some private declaration to the right place
1213 - merge conditional execution and timing macros into emulation module
1214 - document the module
1215 - make sure the module cleanups its mess on gras_exit
1216 * Documentation improvements:
1217 - (new) how to compile applications using GRAS
1218 - (new) emulation support (timing macros)
1220 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1222 SimGrid (2.93) unstable; urgency=low
1224 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1227 - Use Paje properly where used. Still to be sanitized properly.
1228 - Portability fix: Add an implementation of the contexts using pthread
1232 - Add xbt_procname(): returns the name of the current process.
1233 Use it to show the current process's name in all logging.
1235 - fix detection of older flex version and the reaction, since we do
1236 depend on modern ones (we use lex_destroy)
1237 - Better separation of SG and RL in the libs: remove all simulation code
1238 from libgras. As a result, this lib is now only 200k when stripped.
1239 Some of the xbt modules may also be duplicated (two sets and such) and
1240 should be cleaned/killed before SG3.
1241 - Insist on using xlC on AIX because of weird problems involving gcc there.
1242 - Cleanup the make remote stuff. This is now done by scripts
1243 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1244 mainly for our private use, but we're working on changing them to user
1247 - Bugfix: flush the socket on close only if there is some *output*.
1248 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1250 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1253 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1255 SimGrid (2.92) unstable; urgency=low
1257 Alpha 3 on the path to SimGrid 3
1261 - New! First try of benchmarking macros.
1262 - New! First try so that gras_stub_generator generate deployment and
1263 remote compilation helpers.
1265 - Bugfix: Initialization fix in msg_test.
1269 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1271 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1273 - API Change: gras_os_sleep to take the amount of seconds as a double.
1274 Accepting an int was error prone since it was the only location where
1275 seconds were coded as such. It leaded to damn rounding errors.
1276 - Bugfix: Hard to belive that timers ever worked before this.
1278 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1280 SimGrid (2.91) unstable; urgency=low
1282 Alpha 2 on the path to SimGrid 3
1286 - Bug fix in the lmm_solver.
1288 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1289 through the function MSG_paje_output.
1290 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1291 - It is possible to bound the rate of a communication in MSG with
1292 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1293 to write it in the changelog).
1294 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1299 - Various cleanups to the autotools stuff
1300 - Begin to move Gras examples to examples/gras/
1301 - Let make distcheck work again (yeah!)
1303 - documentation overhauled using doxygen.
1304 gtk-doc-tools is dead in SimGrid now.
1305 - Automatically extract all existing logging categories, and add the list
1306 to the documentation (long standing one, to say the less)
1308 - Cleanup the known architecture table. Reorder the entries to group what
1309 should be, and use a more consistent naming scheme.
1310 (some of the test dataset are still to be regenerated)
1311 - New! Allow library to register globals on each process just as userdata
1313 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1314 lookup time (for now).
1315 Use it in msg and trp.
1316 This cleans a lot the internals and helps enforcing privacy of the
1317 headers between the gras components.
1318 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1319 - Bugfix: gras_os_time was delirious in RL.
1320 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1322 - Reenable GRAS now that it works.
1324 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1326 SimGrid (2.90) unstable; urgency=low
1328 Alpha 1 on the path to SimGrid 3
1330 * It is a long time since the last release of SimGrid. I'm sorry about
1331 that but as I had told you, I was rewriting a lot of things. I apologize
1332 to those who had been reporting bugs to me and that I had not answered.
1333 If your bug is still in the new version, please tell me. Here is a
1334 summary of the main changes.
1336 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1337 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1338 portability, speed, and a lot more but you'll figure it out later.
1339 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1340 I wanted to release the new MSG as soon as possible and I have
1341 broken GRAS, which is the reason why, for now, only MSG is fully
1342 functional. A laconic description of these projects is available
1343 in the documentation.
1345 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1346 kernel optimized for our needs (called SURF but only the developers
1347 should use it). Hence, MSG is now roughly 30 times faster and I think
1348 that by rewriting a little bit MSG, I could event speed it up a little
1349 bit more. Beside the gain in speed, it is also much easier to encode a
1350 new platform model with SURF than it was with SG. More to come...
1352 * REVOLUTION 3: I have tried to change a little as possible the API of
1353 MSG but a few things really had to disappear. The main differences
1354 with the previous version are :
1355 1) no more m_links_t and the corresponding functions. Platforms are
1356 directly read from a XML description and cannot be hard-coded
1357 anymore. The same format is used for application deployment
1358 description. The new format is described in the documentation.
1359 Have a look in tools/platform_generation. There is a tiny script
1360 that converts from the old platform format to the new one. Concerning
1361 the application deployment format, parsing the old one is tricky.
1362 I think most of you should however be able to convert your files. If
1363 it is really an issue, I can write a C code that does the conversion.
1365 2) the toolbox tbx does not exist anymore. We now have a library
1366 with much more data-structures but without the hash-tables (we have
1367 dictionaries that are much faster).
1369 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1371 *****************************************************************************
1372 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1373 * don't feel like converting the oldies. (Mt) *
1374 *****************************************************************************
1377 Version 2.90: "the long awaited one"
1378 - Finished rewriting and debugging MSG. Rewrote the documentation.
1379 - disable GRAS for now since it needs to be ported to the newest SG
1382 - Finish the port to windows (using mingw32 for cross-compile)
1385 - Main loop and datastructures of SURF. A cpu resource object is
1386 functional. Surf can thus be used to create cpu's with variable
1387 performance on which you can execute some actions.
1389 2004-11-15 Martin Quinson
1390 - Port to ARM. Simply added the alignment and size descriptions. Should
1391 work, but the ARM machines are so slow that I didn't had the opportunity
1392 to 'make check' over there yet.
1394 2004-11-15 Arnaud Legrand
1395 - Trace manager now written. It uses a heap structure and is therefore
1396 expected to be efficient. It may however be speeded up (particularly
1397 when many events occur at the same date) by using red and black
1398 trees. One day maybe...
1399 - Max-min linear system solver written. It uses a sparse matrix
1400 structure taking advantage of its expected use. Most operations are
1401 O(1) and free/calloc are called as few as possible. The computation of
1402 the minimum could however be improved by using a red and black tree
1405 2004-11-03 Arnaud Legrand
1406 - Rename every gras_* function that was in xbt/ to its xbt_
1408 - Add a heap and a doubly-linked list to xbt
1409 - Added a dichotomy to the dictionaries. make check works as well before
1410 so I assume that the patch is correct. I do not know however if things
1411 run effectively faster than before now. :)
1413 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1414 SimGrid, and the version number is bumped to 2.x
1416 2004-10-29 Martin Quinson
1417 - Introduction of the remote errors.
1418 They are the result of a RMI/RPC on the remote machine.
1419 ErrCodes being scalar values, you can't get the host on which those
1420 errors did happen. Extending the error mechanism as in Gnome is possible.
1421 No idea yet whether it is a good idea.
1423 2004-10-28 Martin Quinson
1424 - Interface revolution: the Starred Structure Eradication.
1425 I used to do typedef struct {} toto_t; and then handle *toto_t.
1426 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1427 - ???_t is a valid type (builded with typedef)
1428 - s_toto_t is a structure (access to fields with .)
1429 - s_toto is a structure needing 'struct' keyword to be used
1430 - e_toto_t is an enum
1431 - toto_t is an 'object' (struct*)
1433 typedef struct s_toto {} s_toto_t, *toto_t;
1434 typedef enum {} e_toto_t;
1435 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1436 s_toto_t) is private.
1438 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1439 so that it can set the variable to NULL. It was so for dicts and sets,
1440 it changed for dynars.
1442 - Fix a bunch of memleaks in dict_remove
1443 - Fix a bug in sg/server_socket opening: it failed all the time.
1445 2004-10-07 Martin Quinson
1446 - Speed up dynar lookup operation a bit.
1448 gras_dynar_get is dead.
1450 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1451 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1452 (which gives you the address of the stored data).
1454 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1455 copy of the data using an affectation to do the job and not a memcpy.
1457 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1458 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1460 It does not really speedup the dynar test because they are
1461 setting elements all the time (and look them seldom). But the dict does
1462 far more lookup than setting.
1464 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1466 2004-10-05 Martin Quinson
1467 - Allow to (en/dis)able the cycle detection at run time.
1469 Whether we should check for cycle or not is now a property of each
1470 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1471 datadesc_cycle_unset allow to remove this property when previously set.
1473 Note that the cycle detection is off by default since it impacts the
1474 performance. Watch the data you feed GRAS with ;)
1476 This property is hereditary. Any element embedded in a structure having it
1477 set have it set for the time of this data exchange.
1479 You should set it both on sender and receiver side. If you don't set it on
1480 sender side, it will enter an endless loop. If you forget on receiver
1481 side, the cycles won't be recreated after communication.
1483 - Header reorganization.
1484 Kill gras_private.h, each submodule must load the headers it needs.
1486 2004-10-04 Martin Quinson
1487 - Interface revolution: do not try to survive to malloc failure.
1489 Now, gras_malloc and friends call gras_abort() on failure.
1490 As a conclusion, malloc_error is not a valid error anymore, and all
1491 functions for which it was the only gras_error_t return value are
1492 changed. They now return void, or there result directly.
1493 This simplify the API a lot.
1495 2004-09-29 Martin Quinson
1496 - Re-enable raw sockets.
1497 Created by gras_socket_{client,server}_ext;
1498 Used with gras_raw_{send,recv}
1501 It should allow to kill the last bits of gras first version soon.
1503 This is not completely satisfactory yet (duplicate code with
1504 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1507 - Simplify transport plugin (internal) interface by not passing any
1508 argument to _server and _client, but embedding them in the socket
1511 2004-09-28 Martin Quinson
1512 - Finish the port to AIX.
1513 autoconf was my problem (segfault within the malloc replacement
1514 function. No idea why)
1516 2004-09-16 Martin Quinson
1517 - Fix some size_t madness on 64bit architectures.
1519 2004-09-08 Martin Quinson
1520 - Reduce the number of system headers loaded, overload some more system
1521 calls (such as malloc to cast the result of the system one, and work
1523 - Fix and reintroduce the config support
1525 2004-09-07 Martin Quinson
1526 - Source code reorganization to allow Arnaud to surf all over there.
1527 - Allow to document the logging categories.
1528 - Remove all uppercase from logging categories and useless cleanup in names.
1530 2004-08-18 Martin Quinson
1531 Version 0.6.2 (protocol not changed; API changed)
1532 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1533 gras_error_t), and NULL when not found. Functions expecting a msgtype
1534 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1535 hopefully usefull message.
1536 - Portability to prehistoric sparcs again
1538 2004-08-17 Martin Quinson
1539 Version 0.6.1 (protocol not changed; ABI not changed)
1540 - prealloc some buffers to speed things up
1542 2004-08-11 Martin Quinson
1543 Version 0.6 (protocol not changed; ABI expended)
1544 - The parsing macro can deal with the references, provided that you add
1545 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1547 2004-08-09 Martin Quinson
1548 Version 0.5 (protocol not changed; ABI changed)
1549 - Allow to off turn the cycle detection code in data exchange at
1550 compilation time. It should be at run time, but I'm short of time (and
1551 the config stuff is still broken). That way, we keep dict out of the
1552 critical path, which is good because the performance is poor:
1553 - search not dichotomial yet
1554 - dynar give no way to access their content and memcpy everytime
1555 - In composed data description (struct, ref and so on), stop foolness of
1556 keeping the subtype's ID, but store the type itself. This keeps sets out
1557 of the critical path, which is good since they rely on dynar and
1558 dictionnaries. The only loose of that is that we cannot detect the
1559 redeclaration of a structure/union with another content (but I'm not sure
1560 the code detected well this error before anyway). We still can detect
1561 the redefinition discrepancy for the other types.
1562 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1563 it breaks the code because of type-punning used all over the place).
1564 This breaks on all non-gcc architectures (for now).
1566 All those changes (plus the buffer of last time) allow me to gain 2 order
1567 of magnitude on cruel tests consisting of 800000 array of integers on two
1568 level of a hierarchical structure (200 secondes -> 4 secondes)
1571 - the selector of reference must now return the type it points to, not
1572 the ID of this type.
1574 2004-08-06 Martin Quinson
1575 Version 0.4 (protocol changed; ABI not changed)
1576 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1577 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1578 - Implement a buffer transport, and use it by default (it relies on tcp in
1579 real life and on sg in simulation).
1580 That's a bit hackish since I had a new field to the structure to store
1581 its data without interfering with the subtype ones. Inheritance
1582 is tricky in C. And that's a kind of reverse inheritance with one class
1583 derivating two classes. Or maybe a game with java interfaces. Anyway,
1584 that's damn hard in C (at least).
1585 Moreover, I got tired while trying to ensure plugin separation and
1586 genericity in SG mode. MSG wants me to do weird things, so let's go for
1587 cruel hacks (temporarily of course ;).
1588 See comment in transport_private.h:71
1589 - do not include all the _interface headers in private but in the files
1590 which really need them (to cut the compilation time when they are
1593 2004-07-26 Martin Quinson
1594 Version 0.3 (protocol not changed; ABI changed)
1595 - Major overhault of the datadesc interface to simplify it:
1596 - shorted the function names:
1597 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1598 - add a trivial way to push/pop integers into the cbps without malloc.
1599 This allows to make really generic sub_type description, which simply
1600 pop their size of the stack.
1601 - add a function gras_datadesc_ref_pop_arr() which does what users want
1602 most of the time: Declare a dynamic array (which pops its size of the
1603 stack) and declare a reference to it. Poor name, but anyway.
1604 - kill the post-send callback, add a post-receive one
1606 2004-07-23 Martin Quinson
1607 Version 0.2 (protocol changed; ABI changed)
1608 - add some testing for cpbs in the test cases, and fix some more bugs.
1609 This invalidate again the little64 data file, since I cannot regenerate
1611 - remove an awfull optimization in the logging stuff, allowing me to:
1612 - understand it again
1613 - learn gcc how to check that the argument match the provided format
1614 - fix all errors revealed by gcc after that
1615 - internal keys of dict are not \0 terminated. Deal with it properly in
1616 loggings instead of segfaulting when the user want to see the logs :-/
1618 2004-07-22 Martin Quinson
1619 - Fix some stupid bug preventing cbps (callback postit) from working
1621 2004-07-21 Martin Quinson
1622 - Some documentation cleanups
1623 - remove the useless last argument of msgtype_declare
1624 - rename the Virtu functions to fit into the 'os' namespace
1625 - move headers src/include -> src/include/gras/ and stop fooling with
1626 gras -> . symbolic link
1627 - make distcheck is now successful
1629 2004-07-19 Martin Quinson
1631 - Build shared library also
1632 - Install html doc to the right location
1633 - stop removing maintainer files in make clean
1634 - build tests only on make check
1636 2004-07-13 Martin Quinson
1638 - No major issue in previous version => change versionning schema
1639 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1640 corresponding dataset.
1642 2004-07-11 Martin Quinson
1644 - Get it working with any kind of structure (we can compute the padding
1645 bytes remotely for all the architectures I have access to)
1646 - Implement the structure parsing macro (still not quite robust/complete)
1647 - Improvement to the remote testing toysuite
1649 2004-07-10 Martin Quinson
1650 [autoconf mechanism]
1651 - get ride of a bunch of deprecated macros
1652 - actually run the test for two-compliment, not only compile it :-/
1653 - test whether the structures get packed (and bail out if yes. Damn.
1654 Alignment is a serious matter)
1655 - test whether the structures get compacted (but respecting the alignment
1656 constraints of each types)
1657 - test whether the array fields of structures can straddle alignment boundaries
1659 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1660 (took me 2 hours to find that bug, looking at the wrong place)
1661 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1662 before sending/receiving and are used to compute the offsets of fields
1663 - Given that padding size depend even on compiler options, keep track of
1664 alignment and aligned_size only for the current architecture. Not a big
1665 deal since we send structure fields one after the other (seems
1667 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1668 program, let it work on linux/gcc/little32. portability todo.
1670 2004-07-08 Martin Quinson
1671 - import and improve remote compilation support from FAST
1672 - make sure make check works on half a dozen of machines out there
1674 2004-07-07 Martin Quinson
1675 Let's say it's version 0.0.3 ;)
1676 - Implement conversions (yuhu!)
1677 - Let it work on solaris (beside conversion, of course)
1678 - Stupid me, using rand() to generate the conversion datatests in not wise.
1680 2004-07-06 Martin Quinson
1681 - Let make dist work, since I'm gonna need it to compile on remote hosts
1682 - Let Tests/datadesc_usage write the architecture on which the file was
1683 generated as first byte.
1684 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1685 architecture descriptions.
1686 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1687 on those architectures.
1688 - Optimization: send/recv array of scalar in one shoot
1690 2004-07-05 Martin Quinson
1691 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1693 - Plug a whole bunch of memleaks
1694 - each process now have to call gras_{init,exit}. One day, their log
1695 settings will be separated
1696 - Continue the code factorisation between SG, RL and common in Transport.
1698 2004-07-04 Martin Quinson
1700 - Redistribution between SG and RL.
1701 We wanna have to accept in SG, so move accepted related parts of RL in
1702 the common part. (more precisely, the dynar of all known sockets is no
1703 more a static in transport.c, but part of the process_data)
1705 [gras_stub_generator]
1706 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1708 2004-07-03 Martin Quinson
1709 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1710 them all up in one shot)
1712 - Fix the ugly case of reference to dynamic array.
1713 - New (semi-public) function gras_datadesc_size to allow the messaging
1714 layer to malloc the needed space for the buffer.
1716 - gras_socket_close now expect the socket to close (and not its address to
1717 put NULL in it after it). This is because the socket passed to handlers
1718 is one of their argument (=> not writable).
1720 - propagate the interface cleanup from last week in datadesc, ie remove a
1721 superfluous level of indirection. User pass adress of variable
1722 containing data (both when sending and receiving), and not of a variable
1723 being a pointer to the data. Let's say that I like it better ;)
1724 The price for that is constructs like "int msg=*(int*)payload" in
1725 handlers, but it's a fine price, IMHO.
1727 - Let it work in RL (yuhu)
1729 2004-06-21 Martin Quinson
1731 - porting SG plugin and SG select to new standards (works almost).
1732 - plug memleaks and fix bugs around.
1735 - cleanup the prototype of data recv and force users to specify when they
1736 want to handle references to objects. Test case working even for cycles.
1737 - plug memleaks. Valgrind is perfectly ok with this.
1739 2004-06-12 Martin Quinson
1741 - cleanup the separation between plugin and main code in plugin creation
1743 2004-06-11 Martin Quinson
1745 - Reput hook for raw sockets, needed for BW experiments
1746 - kill a few lines of dead code
1747 [Data description] Interface cleanup
1748 - gras_datadesc_by_name returns the searched type or NULL.
1749 That way, no variable is needed to use a type desc once, which makes
1751 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1752 two parameters were strings (field name, type name), leading to
1754 [Dicos] Interface cleanup
1755 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1756 This is consistant with the dynar API.
1758 2004-04-21 Martin Quinson
1760 - Porting to new standards.
1762 - interface cleanup.
1763 There is no bag anymore, no need to take extra provision to mask the
1764 pointers behind "ID".
1765 Better splitup of functions between files create/exchange/convert.
1766 This is still a bit artificial since convert and receive are so
1767 interleaved, but anyway.
1769 - add a queued message list to procdata (the ones not matching criteria
1771 - factorize some more code between SG and RL wrt procdata
1773 - use gras_exit in example to track memleaks
1774 - get rid of gs_example now that GS is properly integrated into gras
1775 - update run_test to integrate the lastest tests (datadesc)
1777 - rename WARNINGn macros to WARNn since it prooved error-prone
1779 2004-04-19 Martin Quinson
1781 - register init/exit functions within gras module mechanism
1782 - send/receive function.
1783 Convertion is not implemented, but short-cutted if not needed.
1784 struct/array elements are sent one by one (instead of block-wise), but
1785 nobody really cares (yet). Get a prototype before optimizing.
1786 - tests (using a file socket) for DD send/receive on:
1787 - base types: int, float
1788 - array: fixed size, string (ie ref to dynamic string)
1789 - structure: homogeneous, heterogeneous
1790 - chained list, graph with cycle
1791 Believe it or not, valgrind is not too unhappy with the results. The
1792 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1796 - Bugfix when using a filename explicitely (instead of '-')
1798 2004-04-09 Martin Quinson
1800 - factorize more code between RL and SG in socket creation
1801 - Complete the implementation and tests of:
1803 o file (only in RL, and mainly for debugging)
1805 I lost 3 days to design a portable address resolver, and then decided
1806 that the prototype mainly have to run on my box.
1807 Addressing portability too early may be like optimizing too early :-/
1809 - use gras_init in the Tests instead of the crappy parse_log_opt
1810 (the latter function is removed)
1811 [Conditional execution]
1812 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1813 [Code reorganisation]
1814 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1815 Build examples against the RL library, since there is no way to disable
1816 its creation for now.
1818 For information, the beginning of coding on GRAS was back in june
1819 2003. I guess that every line has been rewritten at least twice since