1 SimGrid (3.5) unstable; urgency=low
5 * Change the way a task is considered as ready. Instead of removing
6 dependencies when a task is done, a counter is decreased. This way,
7 it is always possible to reach ancestors thanks to the
8 SD_taks_get_parents function (even after the end of the simulation.)
9 * Change the level of some logs in sd_global from INFO to VERB. The
10 corresponding tests are now less verbose
11 * Change the return type of SD_Simulate from (SD_task_t*) into
12 xbt_dynar_t. This function was in handling a dynar internally and
13 converted it into a NULL terminated array for historical reasons.
14 * Rename the SD_READY (all dependencies are satisfied and task is
15 scheduled) state in SD_RUNNABLE. This prevents a confusion with
16 the notion of "ready" used in DAG scheduling (all dependencies are
20 * New data container: setset (set of sets of elements)
21 * New module: mmalloc (mapped malloc)
22 * New function: xbt_dict_cursor_set_data()
23 * New function: xbt_dynar_sort()
24 * New function: xbt_fifo_get_last_item()
25 * Bug fix in xbt_dynar_shrink(): use the right element size
26 * Use library init/fini functions for our initialization.
27 - you can use logs and other feature as soon as you want in your
28 code (even before the xbt_init / MSG_init)
29 - xbt_exit is now a no-op and produce a warning when used.
31 Build chain: bug fixes and further polishing
32 * Fix the 'make install' target.
33 No need to use 'make install-simgrid' anymore
34 * Introduce a 'make dist' target compiling a *source* archive
35 'make package' compiles a binary archive (not well tested)
36 * Compile java files only on need
38 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
40 SimGrid (3.4.1) stable; urgency=low
42 The "Polishing easter eggs is probably a good idea" release.
43 This is a bug fixes release only.
47 * Fix a bug preventing the tasks from begin garbage collected.
50 * Fix a bug occuring when a host involved in a communication fails.
51 This was not detected properly by the other peer involved in the
52 communication. Now, it's reported as a network error.
55 * Warn the user about loop dependencies in data flow of DAX files
56 * Obey the control-flow dependencies of DAX files
59 * Add option "enable_smpi" allowing to not compile SMPI.
60 Probably useful for the (Mac) users experiencing a build error here
61 * Improve the detection of lua5.1 and ruby1.8
63 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
65 SimGrid (3.4) stable; urgency=low
67 The "Easter in Cargese" release. Also known as (major changes):
69 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
70 ~> bindings were greatly improved
71 ~> new tracing infrastructure for better visualization introduced
73 * the "Welcome to configury modernity" release.
74 ~> we switched from autotools to cmake, and improved our cdash
77 A more detailled list of changes follow (full detail in svn log).
79 Java Bindings: Various Cleanups
80 * (install java-gcj-compat-dev on debian-like to use them)
81 * Remove put/get: no need to export deprecated interface in Java
82 Use send/receive instead.
83 * Cleanup the examples and add a README per directory
84 * Remove example autoDestination (that's the only way to go now)
85 * Remove example explicitDestination (was a plain copy of basic)
86 * Make JniException a runtime exception, so that there is no need to
87 declare the fact that you may encounter such a beast. I guess that
88 nobody will ever want to survive such error.
89 * Create specific errors for each MSG case of failure:
90 host failure, transfer failure, timeout, task cancelled
91 * Cleanup the exceptions that may get thrown by each function
92 * Other internal cleanups in Java bindings. Performance still bad :/
93 Ruby and Lua Bindings: create them
94 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
95 * That's new and great, you should try them out.
96 Same functionalities than Java bindings, only even less polished
98 * Kill the useless "rate" argument of SD_task_get_execution_time()
99 Everyone used to provide -1 as a value, it was not used, and the
100 semantic of a possible use wasn't even clear.
101 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
102 either as comm costs or compute costs to mean that there is no
103 such thing for that specific task.
104 * Add a SD_task_set_name() function
105 * Fix SD_task_unschedule() on typed tasks
106 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
107 * In DAX loader, accept useless 'level' attributes to <job> since
108 LIGO DAGs have them (seem to be to ease graphical representation).
110 * Add an example masterslave_mailbox.c using send/receive and not
111 the deprecated put/get interface.
112 * Kill the MSG_paje_output() function. It's a noop since 2 years.
113 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
115 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
116 (declare MSG_USE_DEPRECATED to still have the old name)
117 * Add a MSG_task_set_data() function
118 * About trace replay (see examples/msg/actions):
120 - Allow to work with splitted trace files for each process
121 Give the specific trace file as argument of each process,
122 and call MSG_action_trace_run(NULL)
123 You can still have one merged file for all processes.
124 - Fix implementation of collective operations
125 * Allow task_execute() on 0-sized tasks (closes #10063)
127 * This is the first release of SimGrid where SMPI is not considered
128 beta anymore (even if some corners should still be improved)
129 * Port over the new SIMIX_network submodule (internal refactoring)
130 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
131 * Implement more missing elements of the standard:
133 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
135 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
136 MPI_Scan() MPI_Get_processor_name()
137 - Added implementation of missing case for Alltoall (warning: it's
138 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
139 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
140 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
141 SMPI_MPI_Allgatherv()
143 - MPI_Waitsome() was broken
144 - Allow relative includes in smpicc
145 - Command line cfg argument 'reference_speed' was ignored...
146 - Some functions did not properly lead to auto-benching of user code
147 - smpicc passes -O2 by default (just like openmpi one)
149 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
150 * Bug fixes about timeouts during communications
151 * add SIMIX_message_sizes_output() as a pimple to write to file the
152 amount of messages per size. Use gnuplot to get histogram.
153 Pimple because that's the only user-visible function of simix,
154 defined directly in xbt.h (irk, sorry)
156 - Add a SIMIX_sem_get_capacity() function
157 - Fix interactions with processe resume/suspende
158 - release_forever() was stupidly broken
159 - Fix SIMIX_display_process_status() for processes in a semaphore
160 - Make SIMIX_sem_block_onto() user-visible
161 * Refactoring context stuff:
162 - Use pseudo-OOP for better modularity
163 - reimplement SIMIX_process_kill() without process_schedule() so
164 that the latter can take as invariant that it is called from
166 - Merge context_start into context_new for sake of simplicity
168 * Add a Vivaldi network model, coded live during SUD'10 ;)
169 * Rename configuration variables to start a hierarchy:
170 o cpu_model -> cpu/model
171 o network_model -> network/model
172 o workstation_model -> workstation/model
173 * New configuration variables:
174 o network/bandwidth_factor: correction to bandwith
175 o network/latency_factor: correction to latency
176 o netwotk/weight_S: correction to the weight of competing streams
177 * Add a long description to the models, that users can see with such
178 argument on the command line: --cfg=cpu/model:help
179 * --help-models display the long description of all known models
181 * config: add the ability to set a default value after registration
182 Does not override any previously set value (e.g. from cmd line)
183 * dict: allow to have integer key and data.
184 When so, you need to use the following functions
185 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
186 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
187 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
188 In contrary to regular dicts, the key is not malloced before copy.
189 Mixing scalar and regular elements in the same dict is not tested
191 * Allow to use xbt_dynar_shrink() to expend the dynar instead
192 Tracing for Visualization:
193 * SimGrid is now instrumented in order to generate a trace file for
194 visualization analysis: to use it, need to compile SimGrid with the
195 "tracing" option enabled, and instrument the program using SimGrid with
196 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
197 (among other functions).
198 * The instrumentation only traces the platform utilization for now
199 * Documentation to use the tracing functions and how to analyze the
200 traces with the Triva tool is written.
201 * More information about: SimGrid FAQ (in the section Tracing Simulations
204 * We moved to cmake as default build system. Autotools support will
205 be dropped soon. Check the FAQ for more info about how to use it.
206 * Greatly improved our cdash/ctest interactions
207 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
208 * Added memory checking tests with valgrind; lot of memleak fixing.
209 This may be the first release of simgrid with so few memory issues
210 * Added code coverage tests.
211 Our coverage is still improvable, but at least we see it on cdash.
213 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
215 SimGrid (3.3.4) stable; urgency=low
217 The "Desktop Grid needs love too" release (also called Xmas release).
220 * Major speedup in the maxmin system solving by using lazy evaluation
221 Instead of solving completely the maxmin system at each iteration,
222 only invalidate (and recompute) the modified parts.
223 This new feature is enabled in default models but you can try to
224 turn it on with "--cfg:maxmin-selective-update=1" for other models.
225 * Cas01 IMproved as default CPU model
226 This CPU model is the same Cas01 model, but it uses the
227 maxmin-selective-update flag and a heap structure to manage
228 actions on SURF kernel.
229 It reduces the complexity to find the next action to finish and,
230 consequently, it's faster than the old Cas01.
231 This is the new default CPU model (Cas01).
232 * Rename the old Cas01 model to Cas01_fullupdate
233 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
234 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
235 * CpuTI (CPU Trace Integration)
236 A new CPU model whose objective is simulate faster when using
237 availability trace files.
238 Instead of using a full featured, over engineered maxmin system for
239 CPU modeling, this model does the pre-integration of traces files
240 to calculate the amount of CPU power available, and so, executes
241 faster than the old CPU models.
242 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
243 * Use LV08 as default network model since it gives better accuracy
244 for small messages and shouldn't change things for big ones.
245 Use --cfg=network_model:CM02 to get the previous behavior.
248 ******************************************
249 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
250 ******************************************
251 * The new CPU model may changes simulations!
252 The point is that events occurring at the exact same timestamp
253 are not scheduled in the same order with the old and new
254 version. This may be enough to completely change the execution
255 of simulations in some cases.
256 * The new network model will change simulations!
257 This new model is more realistic than the previous one, so you
258 should consider redoing your old experiments with this model.
259 Sorry for the inconvenience.
262 * Introduce the supernovae compilation mode
263 When compiled that way, the whole SimGrid (or almost) is put in a
264 single compilation unit and compiled in one shoot.
265 This is to help gcc which has difficulties to inline stuff from one
267 The speedup seem to be above 15%, althrough more tests are needed on
268 amd64 to confirm that gain.
271 * Port of MSG's mailbox on top of SIMIX network
272 The put/get mechanism was greatly simplified on the way.
275 * New SIMIX network module. Provides:
276 - Mailbox: rendez-vous mecanism to find with who you want to speak
277 - Synchronous send/recv: easier and hopefully faster since the
278 logic is handled in the maestro process directly now
279 - Asynchronous send/recv: you dreamt of it? It's here now
280 Too bad that nobody cared enough to propagate the change to MSG.
281 * Add semaphores as SIMIX synchronization mechanism.
284 * new function SD_daxload(char*) to load a DAX file
285 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
286 * Introduce typed tasks. Specify its kind and cost at creation.
287 At scheduling, just give where it should be placed, and the cost
288 for each involved resource is automatically computed.
289 Existing constructors so far (more to come of course):
290 - SD_task_create_comm_e2e() for end-to-end communication
291 - SD_task_create_comp_seq() for sequential computation
292 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
293 * new function SD_task_dump() for debuging display
294 * new function SD_task_dotty(task,FILE*) writing to file the info
295 about the task in dotty format
296 * SD_task_dependency_exists() can now cope with having one of its
297 arguments NULL. If so, it tests whether the other argument has any
299 * Add getters on list of preceding/following tasks:
300 SD_task_get_parents(task) and SD_task_get_children(task)
301 * Add getters on amount of workstations and list:
302 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
303 * Add getter on task kind: SD_task_get_kind(task)
304 * Update the start_time and finish_time of tasks on completion/failure
305 * Bugfix: Remove task from state swags when destroyed
308 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
311 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
312 dynar with one separate thread per value of the dynar.
313 * Change the prototype of xbt_thread_create(), sorry.
314 Added a boolean parameter indicating whether we want to join this
315 thread (used in SG only for now)
316 * Implement xbt_thread_join and xbt_thread_yield in SG also.
319 * GTNetS wrappers should now be usable again (and betterly tested too)
320 * Fix a major regression from 3.2 where the timeout provided to
321 MSG_task_put_with_timeout() was used as absolute time before which
322 the comm should be done.
323 * Start to fix the <cluster> tag.
324 - Internal links should be good now (beside of the loopback, which
325 use the private link instead)
326 - paths to the external world is still rather broken
327 - the <route:multi> tag is just broken. Actually that's brain-dead.
328 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
329 to make it less stupid
330 ** Check your platform with teshsuite/simdag/platforms/flatifier **
331 * Fix a source-level compatibility glitch from 3.2: after defining
332 MSG_USE_DEPRECATED, you can use the old name
333 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
334 * Allow to compile from the SVN with automake 1.11
335 * Fix some problems when using the "start_time" tag in deployment XMLs.
336 * Fix #8569: XBT/synchro.h has redundant declarations
337 * Fix #8563: MSG return values and exceptions
338 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
339 * Integrate patch #8636: Obey DESTDIR when installing documentation.
340 Thanks to Robson Peixoto.
341 * Fix a vicious bug in dictionaries inducing that some elements were
342 not freed on xbt_dict_free()
344 Portability report of this version:
345 * Main portability targets:
346 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
348 - mac leopard on i386
349 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
350 but nothing critical.
351 * Other platforms: windows, AIX and others were not tested for this release
353 Timing report of this version:
354 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
355 scenario parameters). From 8h to a few seconds in desktop grid settings.
356 * Supernovae brings about 25% speedup on i386.
358 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
360 SimGrid (3.3.3) stable; urgency=low
362 The "Need for Speed" release.
364 The timings done to validate the 3.3.2 were faulty.
365 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
367 The problem was a conversion from a manually handled vector to
368 xbt_dynar_t on the critical path.
369 xbt_dynar_foreach calls functions, inducing stack management crap.
371 We inlined these functions and xbt_dynar_foreach is now breath taking.
372 We also inlined xbt_swag_belong on the way.
374 Here are some approximate speedup measurements (on master/slaves
375 simulations lasting between 10s and 20s each):
376 3.3.1 -> 3.3.2: about same performance
377 3.3.2 -> 3.3.3: 40% speedup
378 3.3.1 -> 3.3.3: 40% speedup
379 3.3.1 with inline patch -> 3.3.3: 30% speedup
381 Our reading is that the refactoring which occurred in 3.3.2 made us
382 suffer much more from the xbt_dynar_foreach low performance, but
383 once we solved this, this refactoring proved to be very performance
384 effective. From the 40% speedup, somehow, 10% are due to the
385 inlining and 30% to the refactoring.
387 That's a pitty that gcc cannot inline functions placed in other files
388 alone. We have to choose between:
389 - break the encapsulation (by putting private data structures and
390 accessors in headers files to help gcc)
391 - live with low performance
392 - switch to a decent compiler such as icc (not quite possible).
394 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
396 SimGrid (3.3.2) stable; urgency=low
398 The "Simplicity does not preceed complexity, but follows it" release.
400 The main contributors of this release were (lexical order):
401 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
404 * Extract the routing logic into its own object.
405 (was dupplicated in network.c and workstation_LV07.c;
406 Allows to implement other ways of storing that info)
407 => kill now useless network_card concept
408 - Use dynar to represent routes (instead of void** + int*)
409 - kill link_set (use surf_network_model->resource_set instead)
410 - Add a command-line option to choose the routing schema to use
411 - Add three new models:
412 * Floyd (shortest path computed at initialization)
413 * Dijikstra (shortest path recomputed all the time)
414 * Cached Dijikstra (shortest path computed on need)
415 All these models where contributed by Silas De Munck, and are
416 described in his ICCS09 paper.
418 * Simplify model declaration
419 (less redirections, less function to write when defining a model)
420 - Factorize stuff between models:
423 surf_model_resource_set(model)
424 surf_model_resource_by_name(model, name)
425 - Unify the types of models in s_surf_model_t (using an union)
426 - Embeed fields of common_public directly into s_surf_model_t
427 - Rename model methods:
428 action_free ~> action_unref
429 action_change_state ~> action_state_set
430 action_get_state ~> action_state_get
431 - Change model methods into functions :
432 (model)->common_public->action_use ~> surf_action_ref
434 * Implement a generic resource; use it as ancestor to specific ones
435 (allows to kill duplicated code in models)
436 Drawback: timer command don't need no name nor properties;
437 workstation_CLM03 don't need no properties
438 (but I guess we can live with those few bytes wasted)
440 * Improve the action object model
441 - implement a constructor avoiding dupplicated code about field
442 initialization in generic_action part.
444 * Kill the SDP model: it has an external dependency, is deprecated
445 in flavor of modern lmm models, and didn't compile since a while
448 * Relocation of the context module from XBT to SIMIX.
449 (the context were decoupled from the simix processes, duplicating a lot of code)
450 => a lot of code was factorized
451 - less overhead is introduced during scheduling
452 - simpler API for the context factory
453 - the logic for process creation,destruction and manipulation was simplified
454 * Simplification of the s_smx_process_t data structure.
455 => accesing the simix level data associated to a process is faster now,
456 and the code is a lot more readable.
459 * Implement some more MPI primitives:
460 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
461 -implementation: Bcast: flat or 2-ary tree (default),
464 Allreduce: Reduce then Bcast
465 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
466 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
467 Alltoallv: flat tree, like ompi
469 * Add support for optimized collectives (Bcast is now binomial by default)
470 * Port smpirun and smpicc to OS X
473 * Kill SD_link_get_properties: hard to maintain and makes very little sense
474 Shout out if you used it.
477 * Display the list of still queued messages in SG mode when existing
481 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
482 * Add xbt_graph_node_get_outedges() [Silas De Munck]
483 * Add xbt_str_from_file(FILE*)
484 * Add xbt_dict_get_key achieving a linear reverse search
485 * Remove the context module
487 Portability report of this version:
488 * Main portability targets:
489 - Linux(debian)/x86/context
490 - Linux(debian)/x86/pthreads
491 - Linux(debian)/amd64/context
492 - Linux(debian)/amd64/pthreads
493 On these, we still have the eratic breakages of gras/pmm and
494 amok/saturate_sg reported in previous version. We still think
495 that the tests are the cause of the fault, not the tested code.
497 - Mac OSX Leopard/x86/context
498 Still false negative in tesh autotesting.
499 Smpi still fails, but this time because readlink does not accept -f
500 Everything seems to work properly beside of that.
503 - AIX version 5.3 (only tested contexts this time)
504 Smpi still fails there because mktemp is not installed.
505 Everything seems to work properly beside of that.
507 I managed to compile it for the first time, but several breakages.
508 Won't delay the release for this exotic platform.
510 * Windows: it's still lagging behind. If you want to help, please
513 Timing report of this version:
514 This version seem to be more than 5% faster than 3.3.1 (on linux
515 64bits with contextes). The gain is less than expected, we are
516 investigating this for next release.
518 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
520 SimGrid (3.3.1) stable; urgency=low
523 * Implement a --cfg-help to show existing configuration variables
524 * Build chain do not require doxygen in maintainer mode
527 * fix a bug on struct sizeof computation, which prevented the
528 exchange of arrays of structs in some conditions
529 - added a regression test about this in datadesc_usage
530 * Allow the exchange of 0-long dynamic vectors.
531 - for that, use -1 as indicator of dynamic size instead of 0
532 - This implied to change any size from unsigned long to long,
533 reducing a bit communication abilities, but I guess that with
534 64bits being quite common, this is more than enough.
535 - This also induce a protocol change, thus bumping network protocol
536 version from 0 to 1 (if we have external users, we have to get
537 clean on that point too ;)
538 - added two regression tests about this in datadesc_usage
539 * Be more verbose when propagating local exceptions
540 This helps debugging.
541 * Display the status of simulated processes when receiving SIGINT in
545 * Allow to control the simulation from a trace file.
546 New functions MSG_action_register() and MSG_action_trace_run()
547 The first one allows to associate a function execution to each
548 kind of action while the second one parses a trace file and
549 triggers the corresponding actions within the system.
550 For now, only a toy example is provided in examples/msg/actions
551 * Add an exemple of process migration in examples/msg/migration
552 * Fix a bug in task exchange which broke MSG_task_get_sender()
553 Add a teshsuite regression test for that.
554 [Bug: if MSG_task_get_sender() is called after sender exit,
556 * Fix a bug which prevented suspend/resume to work properly
557 * Display the status of simulated processes when receiving SIGINT
558 This fixes a regression of v3.3. due to the introduction of SIMIX
559 * Bug fixing in failure management:
560 - trace could not start by a failure at time 0
561 - failure during communications were not working
564 * Add SIMIX_process_set_name() to change the name of the current
565 process in the log messages.
566 * Store smx_hosts in a dict since we only retrieve them by name
567 * Move the configuration infrastructure to surf
570 * Move the configuration infrastructure to surf
573 * Massive internal cleanups:
574 - Store internal structures on processes instead of hosts (allows
575 to have more than one process per host, in addition of being more
577 - Cleanup the initialization/finalization process
578 - Kill a whole bunch of unneeded synchronization:
579 processes run in exclusive manner within the simulator
580 - Move queues from global tables to process data fields
582 - now accept -platform and -hostfile arguments
583 - Pass the right rank value to processes according to the hostfile
584 * Compile the examples by default, and use them as regression tests
585 * Implement MPI_Wtime()
586 * Change the reference speed to a command line option
589 * TCP_gamma can now be specified as command line option using
590 --cfg=TCP_gamma:10000000.0
591 * Change the --surf-path cmd line option into --cfg=path:
594 * Also include strbuff from xbt.h public header
595 * xbt_ex_display(): do not free the exception after displaying
596 This allows to do more with the given exception afterward.
597 Users should call xbt_ex_free() themselves.
601 Portability report of this version:
602 * Main portability targets:
603 - Linux(debian)/x86/context
604 - Linux(debian)/x86/pthreads
605 - Linux(debian)/amd64/context
606 - Linux(debian)/amd64/pthreads
607 These targets fail about 1/10 of times on gras/pmm, but we believe
608 that this is because of the test, not because of simgrid.
609 amok/saturate_sg fails even more rarely, and the test may not be
612 - Mac OSX Leopard/x86/context
613 The test suite still spits tons of errors because some obscure
614 force prevents us from removing the temporary directories
615 arguing that they still contain some metadata I've never heard of.
616 Smpi fails because seq is not installed.
617 Everything seems to work properly beside of that.
620 - AIX version 5.3 (both contexts and pthread)
621 Smpi still fails there because mktemp is not installed.
622 XML inclusions seems rosty on AIX.
624 * Windows: it's still lagging behind. If you want to help, please
627 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
629 SimGrid (3.3) stable; urgency=high
633 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
636 * Introduce the SIMIX module: factorize code between MSG and GRAS.
639 Until now, GRAS were using MSG as an interface to SURF. It was
640 quite difficult because both interface have several differences
641 (MSG channels vs GRAS sockets were the most notable point).
643 This also opens the gate to SMPI (which should occur soon) and speed
644 up simulations by to 40% (even if it were not the main goal).
646 **************************************
647 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
648 **************************************
649 The point is that events occuring at the exact same timestamp are
650 not scheduled in the same order with the old and new version. This
651 may be enough to completely change the execution of simulations in
652 some cases. Sorry for the inconvenience.
654 * Cleanup and upgrade the XML format to push further scalability
655 issues (check http://hal.inria.fr/inria-00256883/ for more info)
657 * Improve the testing infrastructure with tesh. Now a very large part of
658 the code is tested not only by being run but also by checking that the
659 output match an expected output [Mt].
661 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
662 is really less memory-demanding, which should allow you to use
663 larger files in SimGrid [AL].
665 * Inform valgrind about our contextes, so that it becomes usable
666 with the default (and more effecient) version of SimGrid
667 [contributed by Sékou Diakite, many thanks]
670 * Introduce a listener thread in charge of receiving incomming
671 messages from the network. It allows to overlap communication and
672 computation but most notably, it removes some stupid deadlocks due
673 to the fact that so far, a process could not send and receive at
674 the same time. This made most non trivial communication schema
676 * Convert the PIDs from long int to int to match the MSG ones (and
678 * New function: gras_agent_spawn() to launch a new process on
679 current host. Only working in simulation for now. [Mt]
680 * New function: gras_os_hostport() returning a constant form (ie,
681 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
684 * Make the backtrace of exceptions more human readable [Mt]
685 * New module: xbt/str [Mt]
686 a ton of string utility functions (split, join, printf to a newly
687 allocated buffer, trim, etc)
688 * New module: xbt/hash [Mt]
689 SHA1 hashing algorithm (more to come if needed)
690 * New module: xbt/synchro [Mt]
691 synchronization tools (mutex and conditions) working the same way
692 in simulation and in real life (mainly useful for GRAS, but not
694 * New module: xbt/queue [Mt]
695 classical producer/consumer synchronization scheme
696 * xbt_dynar_new_sync() creates a synchronized dynar. All access
697 (using the classical functions will get serialized) [Mt]
698 * Make dictionary internal table dynamic. No need to specify its size
699 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
701 * Make sure the log channels are organized as a tree under windows
702 (because of ANSI C compatibility issue, any channel were child of
706 * Cleaned many thing in surf and fixed a few bugs [AL].
707 * Add a nice command line configuration mechanism to compose models [AL].
708 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
709 the previous one (KCCFLN05). It relies on something that looks like
710 a max-min sharing mechanism but cannot be written as such. A new solver
711 was thus designed [AL].
712 * Add a new solver to lmm. Based on Lagrange optimization and
713 gradient-based descent, it enables to efficiently maximise systems s.a
715 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
717 This solver enables to propose two new network models for TCP Reno and
718 TCP Vegas based on Low's work. These models still need to be fully
719 tested though [Pedro Velho].
722 * Bug fix in SD_simulate. Now the time bound given as argument is
724 * Use the new parallel task model (ptask_L07) as default.
725 * Use the SURF command line configuration mechanism.
726 * 0-size tasks (for synchronization) should now work.
728 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
730 SimGrid (3.2) stable; urgency=high
734 We still experience issues on this platform, but we believe that at
737 GRAS API BREAKAGE (for simplification purpose, sorry):
738 * the gras_msgtype_by_name is not used anymore. Instead of
739 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
740 you can write (and must)
741 gras_msg_send(toserver, "request", &request);
742 - If you still want to pass a gras_msgtype_t to the function (to cache
743 the type and avoid the lookup time), use the gras_msg_send_() variant.
744 - Impacted functions:
745 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
746 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
747 * The callbacks are now expected to return 0 when everything went well
748 (just like the main() function)
750 GRAS new features and improvements:
751 * New module mecanism where user code can use per process globals [Mt]
752 This is similar to gras_userdata_*() functions, but for libraries. It
753 factorize some code developped over and over in the examples and AMOK.
754 It has still to be documented and used (only amok/peermanagement is
756 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
757 loss when they were small enough to fit into the buffer and sent quickly
758 enough so that they can all get received in one shoot.
759 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
760 if not found. Use the *_or_null() variant for the old semantic.
761 * In gras_msg_handle, do not discard messages without callback.
762 They are probably messages to be explicitly awaited later (ie, proofs of
763 mis-synchronization in userland since they are sent before being awaited)
765 * gras_socket_meas_send/recv: semantic changed!
766 The numerical arguments used to be (1) the total amount of data to send
767 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
768 messages. This was need for the fool willing to send more than MAXINT
769 bytes on quite fat pipes.
772 * Do really rename the hostmanagement module to peermanagement. [Mt]
773 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
774 but this is rather new and this was documented in the module
775 documentation (poor excuses, I admit)
776 * Bandwidth measurement semantic changed! This follows the changes to
777 gras_socket_meas_send/recv explained above.
780 * A sequential mode has been added to the workstations. When a workstation
781 is in sequential mode, it can execute only one task, and the other tasks
782 are waiting in a FIFO. [Christophe Thiery]
785 * The KCCFLN05 workstation model now handles parallel tasks. It is the
786 model for SIMDAG. [Christophe Thiery]
787 * Bug fix in the maxmin solver: Some values were close to 0 instead of
788 equal to 0, which caused some bad behaviors in
789 saturated_constraint_set_update. I now use a threshold mechanism like in
793 * When running manually src/testall, you select specific units [Mt]
794 testall is the result of our cunit mecanism, and should replace all
795 the scripty thingy around since bash don't run easily on billware.
797 * A mallocator system has been added. [Christophe Thiery]
798 Mallocators allow you to recycle your unused objects instead of freeing them
799 and allocating new ones.
801 Documentation update:
802 * FAQ reworking + New FAQs:
803 - "Valgrind spits tons of errors!" [Mt]
804 - "How to repport bugs" [Mt]
805 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
806 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
807 - Communication time measurement within MSG [AL]
808 - I experience weird communication times when I change the latency [AL]
811 - an introduction to the framework and to the used communication model
812 - an initiatic tour introducing the most proheminent features:
814 . Lesson 0: Installing GRAS
815 . Lesson 1: Setting up your own project
816 o Part 2: Message passing
817 . Lesson 2: Exchanging simple messages
818 . Lesson 3: Passing arguments to the processes (in SG)
819 . Lesson 4: Attaching callbacks to messages
820 . Lesson 5: Using globals in processes
821 . Lesson 6: Logging informations properly
822 . Lesson 7: Using internal timers
823 . Lesson 8: Handling errors through exceptions
824 . Lesson 9: Exchanging simple data
825 . Lesson 10: Remote Procedure Calling (RPC)
826 . Lesson 11: Explicitely waiting for messages
827 . Recapping of message passing features in GRAS
828 - A HOWTO section containing:
829 o HOWTO design a GRAS application
830 More are due, of course. They will come latter. In the meanwhile, you can
831 check the examples which are still here.
833 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
835 SimGrid (3.1) stable; urgency=high
839 There was a stack corruption somewhere, visible only when optimizing
840 with these versions. [Vince]
843 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
844 module that enabled to play with Directed Acyclic Graphs. It is built
845 directly on top of SURF and provides an API rather close to the old
846 SG. Some old codes using SG are currently under rewrite to check that
847 all needful functions are provided. [Christophe Thiery]
850 * Complete rewrite of the KCCFLN05 workstation model. It is now an
851 extension of the classical CLM03 model that gracefully handles
852 failures. This is now the default model for MSG and GRAS. It doesn't
853 handle parallel tasks yet though. [AL]
854 * Bug fix: Weights were not correctly set in the network part.
855 WARNING: This may have resulted in incorrect results with simulations
856 where there are more than one flow on a given link. [AL]
859 * After a (long ?) discussion on simgrid-devel, we have decided that the
860 convention we had on units was stupid. That is why it has been decided
861 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
862 WARNING : This means that all previous platform files will not work as
863 such with this version! A warning is issued to ask users to update
865 A conversion script can be found in the contrib module of the CVS, under
866 the name contrib/platform_generation/surfxml_update.pl [MQ]
869 * Bug fix: Processes were started in reverse order, wrt deployment file.
870 WARNING: if your code relies on this bug, please fix it. [AL]
871 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
872 being executed on two different locations. [AL]
873 * Bug fix: Failures are now better supported thanks to Derrick's tests
874 (there was many failure situations I hadn't thought of and that weren't
875 correctly handled). [AL]
876 * New function: MSG_host_is_avail indicates you whether a given m_host_t
880 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
881 Exception occurring on server-side are propagated back to client (!).
883 API CHANGE: the callback changed their prototype. Change:
884 int my_handler(gras_socket_t expeditor, void *payload_data) {
886 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
887 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
889 * New! function: gras_msg_handleall to deal with all messages arriving
890 within a given period.
891 * New! function: gras_socket_server_range to get a server socket in a
892 range of port numbers (ease to avoid port number conflicts) [MQ]
893 * New! gras processes display their backtrace when they get a SIGUSR1
894 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
895 Sweet to debug RL processes [MQ]
899 - Do not force experiment sizes to be expressed in kb, or it becomes
900 impossible to measure the latency this way (needs one byte-long tests)
901 WARNING: this changes the amok_bw_* function semantic. [MQ]
902 - Implements the link saturation stuff. [MQ]
903 * Peer management module:
904 New! module factorizing code that we wrote over and over [MQ].
907 * New module: cunit (my jUnit implementation in ansi C) [MQ]
908 - Test units are placed directly into the library code, they get extracted
909 automatically and placed into the src/testall binary.
910 - Convert most of the XBT tests to this system.
911 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
912 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
913 live in the same C code. This required to change a little bit the API
914 of surfxml parsing but shouldn't be an issue for end-users. [AL]
915 * New module: sparse graph structure with basic algorithms (this is work
916 in progress and the API is not considered to be frozen yet). [AL]
917 * Display more information on backtraces: source line & function names are
918 now displayed just like valgrind does (rely on addr2line tool) [MQ]
919 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
920 * Reworked a little bit some #include statements to load only required
921 headers. Some user code that relied on SimGrid to include stdlib or
922 stdio may need to include it by themselves. [AL]
923 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
925 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
926 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
927 but I guess that almost nobody used those parts. [MQ]
929 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
931 SimGrid (3.0.1) stable; urgency=low
934 * Unfortunately, I had missed 5 misnamed functions:
935 xbt_fifo_item_t xbt_fifo_newitem(void);
936 void xbt_fifo_freeitem(xbt_fifo_item_t);
937 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
938 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
939 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
940 They're now deprecated. Please use their new versions:
941 xbt_fifo_item_t xbt_fifo_new_item(void);
942 void xbt_fifo_free_item(xbt_fifo_item_t);
943 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
944 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
945 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
947 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
948 * Documentation: xbt_log module unmercifully reworked [MQ]
949 * Bugfix: there was a problem with the ending of contexts with
950 the pthread backend. It caused some weird deadlock or behavior
951 depending on the pthread implementation. [AL]
952 * Bugfix: get the exceptions raised in the simulator repport where
953 and why they come from when they are not catched in time [AL, MQ]
956 * Bugfix: Do repport the error when two non-connected hosts try to
957 exchange data (Thanks to Flavien for stumbling into this one) [AL]
960 * Add additionnal checkings on communications. Assert that two
961 communicating hosts are connected by a set of links... [AL]
964 * Add additionnal checkings on channel values in communication [AL]
965 * New: MSG_task_get_source to see on which host a task was generated [HC]
966 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
967 the number of tasks waiting to be received on channel and sent
969 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
970 waits for the first task coming from a given host.. [AL]
972 GRAS new functionnalities: [MQ]
973 * Enhance the parsing macro to allow the size of multidimentional objects
974 to be given thru annotations.
975 * New example (and documentation): Matrix Multiplication a la RPC
976 (as when I was young!) and fix a bunch of bugs found on the way.
978 GRAS performance improvements: [MQ]
980 * Reduce the amount of cbps creation/destruction by making it static to
981 datadesc_send/recv() and using a (newly created) cbps_reset (based on
984 * Change libdata to a set so that we can search for stuff by ID (and thus
985 reduce the insane amount of dict lookups)
988 * Actually implement gras_datadesc_copy() so that we don't have to mimick
989 RL communication on top of SG since it's so uneffective.
990 It may also be used for inter-thread communication in RL, one day.
991 * Use gras_datadesc_copy() to exchange messages on top of SG
993 - improve message exchange performance on top of SG
994 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
995 * Don't exchange on the network the size of the used part of buffer,
996 instead, specify the possible buffer size to read().
998 - reduces the amount of read/write calls (one pair per exchange)
999 - reduces the amount of exchanged data (the size)
1000 - allows to retrieve all arrived data on receiver side, if we don't need
1001 it right now (subsequent read will peek the buffer)
1002 - allows the receiver to proceed with the begining of the stream before
1003 everything is arrived
1004 - make it possible to build an iov transport (using readv/writev)
1006 - take care of the data with non-stable storage (like stacked data),
1008 * If possible, TCP send uses vector I/O (when writev() is here)
1009 - Don't use it for receive since we send data sizes and data on the
1010 same stream, so we wouldn't be able to chain large amount of chunks
1011 before having to flush the stuff to read the size.
1012 * Rework the transport plugin mecanism to simplify it and reduce the
1013 amount of pointer dereferencement when searching for the right function
1016 * I guess that now, we do almost as few system calls as possible while
1017 doing as few data copy as possible.
1019 To improve it further, we could try to send all the sizes first and then
1020 all the data (to use iov on receiving size), but it's only a partial
1021 solution: when you have 2 dimensional data, the sizes of the second
1022 dimension is data of the first dimension, so you need 3 streams.
1024 I'm not sure the potential performance gains justify the coding burden.
1026 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1028 SimGrid (3.00) stable; urgency=high
1031 * New! Give the possibility to hijack the surf parser and thus bypass
1032 MSG_create_environment and MSG_launch_application. Have a look at
1033 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1035 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1037 SimGrid (2.96) unstable; urgency=low
1042 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1043 This deprecates the xbt_error_t mecanisms.
1044 It modifies (simplifies) all XBT and GRAS API.
1045 MSG API keeps unchanged (exceptions raised by XBT are catched from
1046 within MSG and masked with existing error handling facilities)
1049 * New! Add a FATPIPE model. [AL]
1050 * New! Add a parallel task model. [AL]
1051 * New! Add automatically a loopback interface (in the default
1052 network model) if none was precised.
1055 * Bugfix: MSG_process_resume now works with the current running process.
1057 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1058 * Modification of MSG_task_get_compute_duration. Once a task has been
1059 processed, the value returned by this function is now equal to 0. [AL]
1060 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1061 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1062 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1063 of a m_task. For now, MSG_task_cancel only works with computation
1065 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1066 of the processor (in Mflop/s) regardless of the current load on the
1068 * API Change: use proper naming convention for MSG_getClock and
1069 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1071 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1072 This function changes the priority of a computation task. This priority
1073 doesn't affect the transfer rate. A priority of 2 will make a task
1074 receive two times more cpu power than the other ones. This function
1075 has been added to suit the needs of Nguyen The Loc and hasn't been that
1076 much tested yet. So if it fails, please report it and send me your code.
1078 * API Change: removed all functions and types that were marked "deprecated"
1079 since many months. Renamed MSG_global_init_args to MSG_global_init.
1081 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1083 SimGrid (2.95) unstable; urgency=low
1086 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1087 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1088 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1091 * Bugfix: Allow absolute paths for platform description files [MQ]
1092 * Bugfix: do free the variables after use. Leads to drastic performance
1094 * Implement max_duration (ie, timeouts) on resources [AL]
1097 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1099 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1100 the needed support (timeouts) [AL]
1103 * Implement measurement sockets. You can now get the bandwidth between two
1104 hosts thanks to AMOK (see below). [MQ]
1105 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1106 dynar over the network (yeah) [MQ]
1107 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1108 * simple/static token-ring example. [Alexandre Colucci and MQ]
1109 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1110 in place before [MQ]
1113 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1114 type parsing. "short" and "long" modifiers were also ignored; other
1115 modifier (such as reference level) are still ignored. [MQ]
1116 * Embeed the buffer size within the buffer itself on SG. [MQ]
1117 That way, send() are atomic and cannot get intermixed anymore (at least
1118 the ones which are less than 100k; bigger messages still have the issue)
1119 * Array size pushed by the field, not by the field type (or each
1120 and every long int will push stuff to the cbps) [MQ]
1121 * use select() to sleep since it allows to portably sleep less than one
1124 GRAS (minor cleanups)
1125 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1128 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1129 * type category 'ignored' killed as it was never used and were difficult
1131 * whether a type can cycle or not is now a flag, leaving room for more
1132 flags (as "ignored", if we feel the need one day ;)
1133 * Rename raw sockets to measurement sockets since "raw" has another
1134 meaning in networking community.
1137 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1138 and offers features not belonging to GRAS but that most applications
1139 need. One day, it may be a set of plugins loadable at runtime.
1140 * New module: bandwidth
1141 bandwidth measurement between arbitrary nodes running this module. [MQ]
1143 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1145 SimGrid (2.94) unstable; urgency=low
1147 The first beta release of SimGrid 3 !
1151 * Update the main page and the FAQ. Adding references to gforge.
1154 * Add a gras_os_getpid function.
1157 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1158 * Extend the logs so that they also print PID, hostname, date, ... if
1160 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1161 and kill the old version which were in testsuite/
1162 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1166 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1167 for users. Instead, each programming environment (GRAS, MSG,...) use it
1168 when needed to provide such a feature to users.
1169 Don't shortcut the mecanism or you will also shortcut the virtualization
1170 you need on the simulator.
1174 * Cleanups in configury with regard to compile optimization/warning flags.
1175 Also add -fno-loop-optimize to any powerpc since it's the optimization
1176 killing gcc (< 3.4.0).
1177 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1178 needed by MSG examples complications
1179 * Borrow configury beautifications from PHP
1182 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1183 warning (thanks loris for stumbling into this one).
1184 * Bugfix: stop loading private headers (gras_config.h) from the public
1188 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1189 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1190 RL would imply a DNS resolver, which is *hard* to do in a portable way
1191 (and therefore delayed).
1192 * Implement a real timer mecanism and use it in timing macros. This allows
1193 to avoid rounding errors and get a 0.000005 sec precision in timing
1194 macros. While I was at it, various cleanups:
1195 - allow to declare more than one timed section per file (fix a stupid bug)
1196 - move some private declaration to the right place
1197 - merge conditional execution and timing macros into emulation module
1198 - document the module
1199 - make sure the module cleanups its mess on gras_exit
1200 * Documentation improvements:
1201 - (new) how to compile applications using GRAS
1202 - (new) emulation support (timing macros)
1204 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1206 SimGrid (2.93) unstable; urgency=low
1208 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1211 - Use Paje properly where used. Still to be sanitized properly.
1212 - Portability fix: Add an implementation of the contexts using pthread
1216 - Add xbt_procname(): returns the name of the current process.
1217 Use it to show the current process's name in all logging.
1219 - fix detection of older flex version and the reaction, since we do
1220 depend on modern ones (we use lex_destroy)
1221 - Better separation of SG and RL in the libs: remove all simulation code
1222 from libgras. As a result, this lib is now only 200k when stripped.
1223 Some of the xbt modules may also be duplicated (two sets and such) and
1224 should be cleaned/killed before SG3.
1225 - Insist on using xlC on AIX because of weird problems involving gcc there.
1226 - Cleanup the make remote stuff. This is now done by scripts
1227 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1228 mainly for our private use, but we're working on changing them to user
1231 - Bugfix: flush the socket on close only if there is some *output*.
1232 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1234 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1237 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1239 SimGrid (2.92) unstable; urgency=low
1241 Alpha 3 on the path to SimGrid 3
1245 - New! First try of benchmarking macros.
1246 - New! First try so that gras_stub_generator generate deployment and
1247 remote compilation helpers.
1249 - Bugfix: Initialization fix in msg_test.
1253 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1255 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1257 - API Change: gras_os_sleep to take the amount of seconds as a double.
1258 Accepting an int was error prone since it was the only location where
1259 seconds were coded as such. It leaded to damn rounding errors.
1260 - Bugfix: Hard to belive that timers ever worked before this.
1262 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1264 SimGrid (2.91) unstable; urgency=low
1266 Alpha 2 on the path to SimGrid 3
1270 - Bug fix in the lmm_solver.
1272 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1273 through the function MSG_paje_output.
1274 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1275 - It is possible to bound the rate of a communication in MSG with
1276 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1277 to write it in the changelog).
1278 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1283 - Various cleanups to the autotools stuff
1284 - Begin to move Gras examples to examples/gras/
1285 - Let make distcheck work again (yeah!)
1287 - documentation overhauled using doxygen.
1288 gtk-doc-tools is dead in SimGrid now.
1289 - Automatically extract all existing logging categories, and add the list
1290 to the documentation (long standing one, to say the less)
1292 - Cleanup the known architecture table. Reorder the entries to group what
1293 should be, and use a more consistent naming scheme.
1294 (some of the test dataset are still to be regenerated)
1295 - New! Allow library to register globals on each process just as userdata
1297 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1298 lookup time (for now).
1299 Use it in msg and trp.
1300 This cleans a lot the internals and helps enforcing privacy of the
1301 headers between the gras components.
1302 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1303 - Bugfix: gras_os_time was delirious in RL.
1304 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1306 - Reenable GRAS now that it works.
1308 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1310 SimGrid (2.90) unstable; urgency=low
1312 Alpha 1 on the path to SimGrid 3
1314 * It is a long time since the last release of SimGrid. I'm sorry about
1315 that but as I had told you, I was rewriting a lot of things. I apologize
1316 to those who had been reporting bugs to me and that I had not answered.
1317 If your bug is still in the new version, please tell me. Here is a
1318 summary of the main changes.
1320 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1321 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1322 portability, speed, and a lot more but you'll figure it out later.
1323 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1324 I wanted to release the new MSG as soon as possible and I have
1325 broken GRAS, which is the reason why, for now, only MSG is fully
1326 functional. A laconic description of these projects is available
1327 in the documentation.
1329 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1330 kernel optimized for our needs (called SURF but only the developers
1331 should use it). Hence, MSG is now roughly 30 times faster and I think
1332 that by rewriting a little bit MSG, I could event speed it up a little
1333 bit more. Beside the gain in speed, it is also much easier to encode a
1334 new platform model with SURF than it was with SG. More to come...
1336 * REVOLUTION 3: I have tried to change a little as possible the API of
1337 MSG but a few things really had to disappear. The main differences
1338 with the previous version are :
1339 1) no more m_links_t and the corresponding functions. Platforms are
1340 directly read from a XML description and cannot be hard-coded
1341 anymore. The same format is used for application deployment
1342 description. The new format is described in the documentation.
1343 Have a look in tools/platform_generation. There is a tiny script
1344 that converts from the old platform format to the new one. Concerning
1345 the application deployment format, parsing the old one is tricky.
1346 I think most of you should however be able to convert your files. If
1347 it is really an issue, I can write a C code that does the conversion.
1349 2) the toolbox tbx does not exist anymore. We now have a library
1350 with much more data-structures but without the hash-tables (we have
1351 dictionaries that are much faster).
1353 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1355 *****************************************************************************
1356 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1357 * don't feel like converting the oldies. (Mt) *
1358 *****************************************************************************
1361 Version 2.90: "the long awaited one"
1362 - Finished rewriting and debugging MSG. Rewrote the documentation.
1363 - disable GRAS for now since it needs to be ported to the newest SG
1366 - Finish the port to windows (using mingw32 for cross-compile)
1369 - Main loop and datastructures of SURF. A cpu resource object is
1370 functional. Surf can thus be used to create cpu's with variable
1371 performance on which you can execute some actions.
1373 2004-11-15 Martin Quinson
1374 - Port to ARM. Simply added the alignment and size descriptions. Should
1375 work, but the ARM machines are so slow that I didn't had the opportunity
1376 to 'make check' over there yet.
1378 2004-11-15 Arnaud Legrand
1379 - Trace manager now written. It uses a heap structure and is therefore
1380 expected to be efficient. It may however be speeded up (particularly
1381 when many events occur at the same date) by using red and black
1382 trees. One day maybe...
1383 - Max-min linear system solver written. It uses a sparse matrix
1384 structure taking advantage of its expected use. Most operations are
1385 O(1) and free/calloc are called as few as possible. The computation of
1386 the minimum could however be improved by using a red and black tree
1389 2004-11-03 Arnaud Legrand
1390 - Rename every gras_* function that was in xbt/ to its xbt_
1392 - Add a heap and a doubly-linked list to xbt
1393 - Added a dichotomy to the dictionaries. make check works as well before
1394 so I assume that the patch is correct. I do not know however if things
1395 run effectively faster than before now. :)
1397 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1398 SimGrid, and the version number is bumped to 2.x
1400 2004-10-29 Martin Quinson
1401 - Introduction of the remote errors.
1402 They are the result of a RMI/RPC on the remote machine.
1403 ErrCodes being scalar values, you can't get the host on which those
1404 errors did happen. Extending the error mechanism as in Gnome is possible.
1405 No idea yet whether it is a good idea.
1407 2004-10-28 Martin Quinson
1408 - Interface revolution: the Starred Structure Eradication.
1409 I used to do typedef struct {} toto_t; and then handle *toto_t.
1410 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1411 - ???_t is a valid type (builded with typedef)
1412 - s_toto_t is a structure (access to fields with .)
1413 - s_toto is a structure needing 'struct' keyword to be used
1414 - e_toto_t is an enum
1415 - toto_t is an 'object' (struct*)
1417 typedef struct s_toto {} s_toto_t, *toto_t;
1418 typedef enum {} e_toto_t;
1419 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1420 s_toto_t) is private.
1422 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1423 so that it can set the variable to NULL. It was so for dicts and sets,
1424 it changed for dynars.
1426 - Fix a bunch of memleaks in dict_remove
1427 - Fix a bug in sg/server_socket opening: it failed all the time.
1429 2004-10-07 Martin Quinson
1430 - Speed up dynar lookup operation a bit.
1432 gras_dynar_get is dead.
1434 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1435 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1436 (which gives you the address of the stored data).
1438 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1439 copy of the data using an affectation to do the job and not a memcpy.
1441 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1442 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1444 It does not really speedup the dynar test because they are
1445 setting elements all the time (and look them seldom). But the dict does
1446 far more lookup than setting.
1448 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1450 2004-10-05 Martin Quinson
1451 - Allow to (en/dis)able the cycle detection at run time.
1453 Whether we should check for cycle or not is now a property of each
1454 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1455 datadesc_cycle_unset allow to remove this property when previously set.
1457 Note that the cycle detection is off by default since it impacts the
1458 performance. Watch the data you feed GRAS with ;)
1460 This property is hereditary. Any element embedded in a structure having it
1461 set have it set for the time of this data exchange.
1463 You should set it both on sender and receiver side. If you don't set it on
1464 sender side, it will enter an endless loop. If you forget on receiver
1465 side, the cycles won't be recreated after communication.
1467 - Header reorganization.
1468 Kill gras_private.h, each submodule must load the headers it needs.
1470 2004-10-04 Martin Quinson
1471 - Interface revolution: do not try to survive to malloc failure.
1473 Now, gras_malloc and friends call gras_abort() on failure.
1474 As a conclusion, malloc_error is not a valid error anymore, and all
1475 functions for which it was the only gras_error_t return value are
1476 changed. They now return void, or there result directly.
1477 This simplify the API a lot.
1479 2004-09-29 Martin Quinson
1480 - Re-enable raw sockets.
1481 Created by gras_socket_{client,server}_ext;
1482 Used with gras_raw_{send,recv}
1485 It should allow to kill the last bits of gras first version soon.
1487 This is not completely satisfactory yet (duplicate code with
1488 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1491 - Simplify transport plugin (internal) interface by not passing any
1492 argument to _server and _client, but embedding them in the socket
1495 2004-09-28 Martin Quinson
1496 - Finish the port to AIX.
1497 autoconf was my problem (segfault within the malloc replacement
1498 function. No idea why)
1500 2004-09-16 Martin Quinson
1501 - Fix some size_t madness on 64bit architectures.
1503 2004-09-08 Martin Quinson
1504 - Reduce the number of system headers loaded, overload some more system
1505 calls (such as malloc to cast the result of the system one, and work
1507 - Fix and reintroduce the config support
1509 2004-09-07 Martin Quinson
1510 - Source code reorganization to allow Arnaud to surf all over there.
1511 - Allow to document the logging categories.
1512 - Remove all uppercase from logging categories and useless cleanup in names.
1514 2004-08-18 Martin Quinson
1515 Version 0.6.2 (protocol not changed; API changed)
1516 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1517 gras_error_t), and NULL when not found. Functions expecting a msgtype
1518 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1519 hopefully usefull message.
1520 - Portability to prehistoric sparcs again
1522 2004-08-17 Martin Quinson
1523 Version 0.6.1 (protocol not changed; ABI not changed)
1524 - prealloc some buffers to speed things up
1526 2004-08-11 Martin Quinson
1527 Version 0.6 (protocol not changed; ABI expended)
1528 - The parsing macro can deal with the references, provided that you add
1529 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1531 2004-08-09 Martin Quinson
1532 Version 0.5 (protocol not changed; ABI changed)
1533 - Allow to off turn the cycle detection code in data exchange at
1534 compilation time. It should be at run time, but I'm short of time (and
1535 the config stuff is still broken). That way, we keep dict out of the
1536 critical path, which is good because the performance is poor:
1537 - search not dichotomial yet
1538 - dynar give no way to access their content and memcpy everytime
1539 - In composed data description (struct, ref and so on), stop foolness of
1540 keeping the subtype's ID, but store the type itself. This keeps sets out
1541 of the critical path, which is good since they rely on dynar and
1542 dictionnaries. The only loose of that is that we cannot detect the
1543 redeclaration of a structure/union with another content (but I'm not sure
1544 the code detected well this error before anyway). We still can detect
1545 the redefinition discrepancy for the other types.
1546 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1547 it breaks the code because of type-punning used all over the place).
1548 This breaks on all non-gcc architectures (for now).
1550 All those changes (plus the buffer of last time) allow me to gain 2 order
1551 of magnitude on cruel tests consisting of 800000 array of integers on two
1552 level of a hierarchical structure (200 secondes -> 4 secondes)
1555 - the selector of reference must now return the type it points to, not
1556 the ID of this type.
1558 2004-08-06 Martin Quinson
1559 Version 0.4 (protocol changed; ABI not changed)
1560 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1561 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1562 - Implement a buffer transport, and use it by default (it relies on tcp in
1563 real life and on sg in simulation).
1564 That's a bit hackish since I had a new field to the structure to store
1565 its data without interfering with the subtype ones. Inheritance
1566 is tricky in C. And that's a kind of reverse inheritance with one class
1567 derivating two classes. Or maybe a game with java interfaces. Anyway,
1568 that's damn hard in C (at least).
1569 Moreover, I got tired while trying to ensure plugin separation and
1570 genericity in SG mode. MSG wants me to do weird things, so let's go for
1571 cruel hacks (temporarily of course ;).
1572 See comment in transport_private.h:71
1573 - do not include all the _interface headers in private but in the files
1574 which really need them (to cut the compilation time when they are
1577 2004-07-26 Martin Quinson
1578 Version 0.3 (protocol not changed; ABI changed)
1579 - Major overhault of the datadesc interface to simplify it:
1580 - shorted the function names:
1581 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1582 - add a trivial way to push/pop integers into the cbps without malloc.
1583 This allows to make really generic sub_type description, which simply
1584 pop their size of the stack.
1585 - add a function gras_datadesc_ref_pop_arr() which does what users want
1586 most of the time: Declare a dynamic array (which pops its size of the
1587 stack) and declare a reference to it. Poor name, but anyway.
1588 - kill the post-send callback, add a post-receive one
1590 2004-07-23 Martin Quinson
1591 Version 0.2 (protocol changed; ABI changed)
1592 - add some testing for cpbs in the test cases, and fix some more bugs.
1593 This invalidate again the little64 data file, since I cannot regenerate
1595 - remove an awfull optimization in the logging stuff, allowing me to:
1596 - understand it again
1597 - learn gcc how to check that the argument match the provided format
1598 - fix all errors revealed by gcc after that
1599 - internal keys of dict are not \0 terminated. Deal with it properly in
1600 loggings instead of segfaulting when the user want to see the logs :-/
1602 2004-07-22 Martin Quinson
1603 - Fix some stupid bug preventing cbps (callback postit) from working
1605 2004-07-21 Martin Quinson
1606 - Some documentation cleanups
1607 - remove the useless last argument of msgtype_declare
1608 - rename the Virtu functions to fit into the 'os' namespace
1609 - move headers src/include -> src/include/gras/ and stop fooling with
1610 gras -> . symbolic link
1611 - make distcheck is now successful
1613 2004-07-19 Martin Quinson
1615 - Build shared library also
1616 - Install html doc to the right location
1617 - stop removing maintainer files in make clean
1618 - build tests only on make check
1620 2004-07-13 Martin Quinson
1622 - No major issue in previous version => change versionning schema
1623 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1624 corresponding dataset.
1626 2004-07-11 Martin Quinson
1628 - Get it working with any kind of structure (we can compute the padding
1629 bytes remotely for all the architectures I have access to)
1630 - Implement the structure parsing macro (still not quite robust/complete)
1631 - Improvement to the remote testing toysuite
1633 2004-07-10 Martin Quinson
1634 [autoconf mechanism]
1635 - get ride of a bunch of deprecated macros
1636 - actually run the test for two-compliment, not only compile it :-/
1637 - test whether the structures get packed (and bail out if yes. Damn.
1638 Alignment is a serious matter)
1639 - test whether the structures get compacted (but respecting the alignment
1640 constraints of each types)
1641 - test whether the array fields of structures can straddle alignment boundaries
1643 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1644 (took me 2 hours to find that bug, looking at the wrong place)
1645 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1646 before sending/receiving and are used to compute the offsets of fields
1647 - Given that padding size depend even on compiler options, keep track of
1648 alignment and aligned_size only for the current architecture. Not a big
1649 deal since we send structure fields one after the other (seems
1651 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1652 program, let it work on linux/gcc/little32. portability todo.
1654 2004-07-08 Martin Quinson
1655 - import and improve remote compilation support from FAST
1656 - make sure make check works on half a dozen of machines out there
1658 2004-07-07 Martin Quinson
1659 Let's say it's version 0.0.3 ;)
1660 - Implement conversions (yuhu!)
1661 - Let it work on solaris (beside conversion, of course)
1662 - Stupid me, using rand() to generate the conversion datatests in not wise.
1664 2004-07-06 Martin Quinson
1665 - Let make dist work, since I'm gonna need it to compile on remote hosts
1666 - Let Tests/datadesc_usage write the architecture on which the file was
1667 generated as first byte.
1668 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1669 architecture descriptions.
1670 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1671 on those architectures.
1672 - Optimization: send/recv array of scalar in one shoot
1674 2004-07-05 Martin Quinson
1675 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1677 - Plug a whole bunch of memleaks
1678 - each process now have to call gras_{init,exit}. One day, their log
1679 settings will be separated
1680 - Continue the code factorisation between SG, RL and common in Transport.
1682 2004-07-04 Martin Quinson
1684 - Redistribution between SG and RL.
1685 We wanna have to accept in SG, so move accepted related parts of RL in
1686 the common part. (more precisely, the dynar of all known sockets is no
1687 more a static in transport.c, but part of the process_data)
1689 [gras_stub_generator]
1690 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1692 2004-07-03 Martin Quinson
1693 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1694 them all up in one shot)
1696 - Fix the ugly case of reference to dynamic array.
1697 - New (semi-public) function gras_datadesc_size to allow the messaging
1698 layer to malloc the needed space for the buffer.
1700 - gras_socket_close now expect the socket to close (and not its address to
1701 put NULL in it after it). This is because the socket passed to handlers
1702 is one of their argument (=> not writable).
1704 - propagate the interface cleanup from last week in datadesc, ie remove a
1705 superfluous level of indirection. User pass adress of variable
1706 containing data (both when sending and receiving), and not of a variable
1707 being a pointer to the data. Let's say that I like it better ;)
1708 The price for that is constructs like "int msg=*(int*)payload" in
1709 handlers, but it's a fine price, IMHO.
1711 - Let it work in RL (yuhu)
1713 2004-06-21 Martin Quinson
1715 - porting SG plugin and SG select to new standards (works almost).
1716 - plug memleaks and fix bugs around.
1719 - cleanup the prototype of data recv and force users to specify when they
1720 want to handle references to objects. Test case working even for cycles.
1721 - plug memleaks. Valgrind is perfectly ok with this.
1723 2004-06-12 Martin Quinson
1725 - cleanup the separation between plugin and main code in plugin creation
1727 2004-06-11 Martin Quinson
1729 - Reput hook for raw sockets, needed for BW experiments
1730 - kill a few lines of dead code
1731 [Data description] Interface cleanup
1732 - gras_datadesc_by_name returns the searched type or NULL.
1733 That way, no variable is needed to use a type desc once, which makes
1735 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1736 two parameters were strings (field name, type name), leading to
1738 [Dicos] Interface cleanup
1739 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1740 This is consistant with the dynar API.
1742 2004-04-21 Martin Quinson
1744 - Porting to new standards.
1746 - interface cleanup.
1747 There is no bag anymore, no need to take extra provision to mask the
1748 pointers behind "ID".
1749 Better splitup of functions between files create/exchange/convert.
1750 This is still a bit artificial since convert and receive are so
1751 interleaved, but anyway.
1753 - add a queued message list to procdata (the ones not matching criteria
1755 - factorize some more code between SG and RL wrt procdata
1757 - use gras_exit in example to track memleaks
1758 - get rid of gs_example now that GS is properly integrated into gras
1759 - update run_test to integrate the lastest tests (datadesc)
1761 - rename WARNINGn macros to WARNn since it prooved error-prone
1763 2004-04-19 Martin Quinson
1765 - register init/exit functions within gras module mechanism
1766 - send/receive function.
1767 Convertion is not implemented, but short-cutted if not needed.
1768 struct/array elements are sent one by one (instead of block-wise), but
1769 nobody really cares (yet). Get a prototype before optimizing.
1770 - tests (using a file socket) for DD send/receive on:
1771 - base types: int, float
1772 - array: fixed size, string (ie ref to dynamic string)
1773 - structure: homogeneous, heterogeneous
1774 - chained list, graph with cycle
1775 Believe it or not, valgrind is not too unhappy with the results. The
1776 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1780 - Bugfix when using a filename explicitely (instead of '-')
1782 2004-04-09 Martin Quinson
1784 - factorize more code between RL and SG in socket creation
1785 - Complete the implementation and tests of:
1787 o file (only in RL, and mainly for debugging)
1789 I lost 3 days to design a portable address resolver, and then decided
1790 that the prototype mainly have to run on my box.
1791 Addressing portability too early may be like optimizing too early :-/
1793 - use gras_init in the Tests instead of the crappy parse_log_opt
1794 (the latter function is removed)
1795 [Conditional execution]
1796 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1797 [Code reorganisation]
1798 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1799 Build examples against the RL library, since there is no way to disable
1800 its creation for now.
1802 For information, the beginning of coding on GRAS was back in june
1803 2003. I guess that every line has been rewritten at least twice since