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 and define a new SD_SCHEDULABLE (all
16 dependencies are satisfied) state.
17 This prevents a confusion between the notion of "ready to schedule"
18 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
19 simulated" (SD_RUNNABLE) used by the simulation kernel.
20 * Add an example that scheduling a DAX on an heterogeneous platform
21 using a Min-Min strategy
24 * New data container: setset (set of sets of elements)
25 * New module: mmalloc (mapped malloc)
26 * New function: xbt_dict_cursor_set_data()
27 * New function: xbt_dynar_sort()
28 * New function: xbt_dynar_is_empty()
29 * New function: xbt_fifo_get_last_item()
30 * Bug fix in xbt_dynar_shrink(): use the right element size
31 * Use library init/fini functions for our initialization.
32 - you can use logs and other feature as soon as you want in your
33 code (even before the xbt_init / MSG_init)
34 - xbt_exit is now a no-op and produce a warning when used.
36 Build chain: bug fixes and further polishing
37 * Fix the 'make install' target.
38 No need to use 'make install-simgrid' anymore
39 * Introduce a 'make dist' target compiling a *source* archive
40 'make package' compiles a binary archive (not well tested)
41 * Compile java files only on need
43 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
45 SimGrid (3.4.1) stable; urgency=low
47 The "Polishing easter eggs is probably a good idea" release.
48 This is a bug fixes release only.
52 * Fix a bug preventing the tasks from begin garbage collected.
55 * Fix a bug occuring when a host involved in a communication fails.
56 This was not detected properly by the other peer involved in the
57 communication. Now, it's reported as a network error.
60 * Warn the user about loop dependencies in data flow of DAX files
61 * Obey the control-flow dependencies of DAX files
64 * Add option "enable_smpi" allowing to not compile SMPI.
65 Probably useful for the (Mac) users experiencing a build error here
66 * Improve the detection of lua5.1 and ruby1.8
68 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
70 SimGrid (3.4) stable; urgency=low
72 The "Easter in Cargese" release. Also known as (major changes):
74 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
75 ~> bindings were greatly improved
76 ~> new tracing infrastructure for better visualization introduced
78 * the "Welcome to configury modernity" release.
79 ~> we switched from autotools to cmake, and improved our cdash
82 A more detailled list of changes follow (full detail in svn log).
84 Java Bindings: Various Cleanups
85 * (install java-gcj-compat-dev on debian-like to use them)
86 * Remove put/get: no need to export deprecated interface in Java
87 Use send/receive instead.
88 * Cleanup the examples and add a README per directory
89 * Remove example autoDestination (that's the only way to go now)
90 * Remove example explicitDestination (was a plain copy of basic)
91 * Make JniException a runtime exception, so that there is no need to
92 declare the fact that you may encounter such a beast. I guess that
93 nobody will ever want to survive such error.
94 * Create specific errors for each MSG case of failure:
95 host failure, transfer failure, timeout, task cancelled
96 * Cleanup the exceptions that may get thrown by each function
97 * Other internal cleanups in Java bindings. Performance still bad :/
98 Ruby and Lua Bindings: create them
99 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
100 * That's new and great, you should try them out.
101 Same functionalities than Java bindings, only even less polished
103 * Kill the useless "rate" argument of SD_task_get_execution_time()
104 Everyone used to provide -1 as a value, it was not used, and the
105 semantic of a possible use wasn't even clear.
106 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
107 either as comm costs or compute costs to mean that there is no
108 such thing for that specific task.
109 * Add a SD_task_set_name() function
110 * Fix SD_task_unschedule() on typed tasks
111 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
112 * In DAX loader, accept useless 'level' attributes to <job> since
113 LIGO DAGs have them (seem to be to ease graphical representation).
115 * Add an example masterslave_mailbox.c using send/receive and not
116 the deprecated put/get interface.
117 * Kill the MSG_paje_output() function. It's a noop since 2 years.
118 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
120 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
121 (declare MSG_USE_DEPRECATED to still have the old name)
122 * Add a MSG_task_set_data() function
123 * About trace replay (see examples/msg/actions):
125 - Allow to work with splitted trace files for each process
126 Give the specific trace file as argument of each process,
127 and call MSG_action_trace_run(NULL)
128 You can still have one merged file for all processes.
129 - Fix implementation of collective operations
130 * Allow task_execute() on 0-sized tasks (closes #10063)
132 * This is the first release of SimGrid where SMPI is not considered
133 beta anymore (even if some corners should still be improved)
134 * Port over the new SIMIX_network submodule (internal refactoring)
135 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
136 * Implement more missing elements of the standard:
138 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
140 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
141 MPI_Scan() MPI_Get_processor_name()
142 - Added implementation of missing case for Alltoall (warning: it's
143 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
144 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
145 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
146 SMPI_MPI_Allgatherv()
148 - MPI_Waitsome() was broken
149 - Allow relative includes in smpicc
150 - Command line cfg argument 'reference_speed' was ignored...
151 - Some functions did not properly lead to auto-benching of user code
152 - smpicc passes -O2 by default (just like openmpi one)
154 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
155 * Bug fixes about timeouts during communications
156 * add SIMIX_message_sizes_output() as a pimple to write to file the
157 amount of messages per size. Use gnuplot to get histogram.
158 Pimple because that's the only user-visible function of simix,
159 defined directly in xbt.h (irk, sorry)
161 - Add a SIMIX_sem_get_capacity() function
162 - Fix interactions with processe resume/suspende
163 - release_forever() was stupidly broken
164 - Fix SIMIX_display_process_status() for processes in a semaphore
165 - Make SIMIX_sem_block_onto() user-visible
166 * Refactoring context stuff:
167 - Use pseudo-OOP for better modularity
168 - reimplement SIMIX_process_kill() without process_schedule() so
169 that the latter can take as invariant that it is called from
171 - Merge context_start into context_new for sake of simplicity
173 * Add a Vivaldi network model, coded live during SUD'10 ;)
174 * Rename configuration variables to start a hierarchy:
175 o cpu_model -> cpu/model
176 o network_model -> network/model
177 o workstation_model -> workstation/model
178 * New configuration variables:
179 o network/bandwidth_factor: correction to bandwith
180 o network/latency_factor: correction to latency
181 o netwotk/weight_S: correction to the weight of competing streams
182 * Add a long description to the models, that users can see with such
183 argument on the command line: --cfg=cpu/model:help
184 * --help-models display the long description of all known models
186 * config: add the ability to set a default value after registration
187 Does not override any previously set value (e.g. from cmd line)
188 * dict: allow to have integer key and data.
189 When so, you need to use the following functions
190 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
191 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
192 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
193 In contrary to regular dicts, the key is not malloced before copy.
194 Mixing scalar and regular elements in the same dict is not tested
196 * Allow to use xbt_dynar_shrink() to expend the dynar instead
197 Tracing for Visualization:
198 * SimGrid is now instrumented in order to generate a trace file for
199 visualization analysis: to use it, need to compile SimGrid with the
200 "tracing" option enabled, and instrument the program using SimGrid with
201 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
202 (among other functions).
203 * The instrumentation only traces the platform utilization for now
204 * Documentation to use the tracing functions and how to analyze the
205 traces with the Triva tool is written.
206 * More information about: SimGrid FAQ (in the section Tracing Simulations
209 * We moved to cmake as default build system. Autotools support will
210 be dropped soon. Check the FAQ for more info about how to use it.
211 * Greatly improved our cdash/ctest interactions
212 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
213 * Added memory checking tests with valgrind; lot of memleak fixing.
214 This may be the first release of simgrid with so few memory issues
215 * Added code coverage tests.
216 Our coverage is still improvable, but at least we see it on cdash.
218 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
220 SimGrid (3.3.4) stable; urgency=low
222 The "Desktop Grid needs love too" release (also called Xmas release).
225 * Major speedup in the maxmin system solving by using lazy evaluation
226 Instead of solving completely the maxmin system at each iteration,
227 only invalidate (and recompute) the modified parts.
228 This new feature is enabled in default models but you can try to
229 turn it on with "--cfg:maxmin-selective-update=1" for other models.
230 * Cas01 IMproved as default CPU model
231 This CPU model is the same Cas01 model, but it uses the
232 maxmin-selective-update flag and a heap structure to manage
233 actions on SURF kernel.
234 It reduces the complexity to find the next action to finish and,
235 consequently, it's faster than the old Cas01.
236 This is the new default CPU model (Cas01).
237 * Rename the old Cas01 model to Cas01_fullupdate
238 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
239 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
240 * CpuTI (CPU Trace Integration)
241 A new CPU model whose objective is simulate faster when using
242 availability trace files.
243 Instead of using a full featured, over engineered maxmin system for
244 CPU modeling, this model does the pre-integration of traces files
245 to calculate the amount of CPU power available, and so, executes
246 faster than the old CPU models.
247 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
248 * Use LV08 as default network model since it gives better accuracy
249 for small messages and shouldn't change things for big ones.
250 Use --cfg=network_model:CM02 to get the previous behavior.
253 ******************************************
254 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
255 ******************************************
256 * The new CPU model may changes simulations!
257 The point is that events occurring at the exact same timestamp
258 are not scheduled in the same order with the old and new
259 version. This may be enough to completely change the execution
260 of simulations in some cases.
261 * The new network model will change simulations!
262 This new model is more realistic than the previous one, so you
263 should consider redoing your old experiments with this model.
264 Sorry for the inconvenience.
267 * Introduce the supernovae compilation mode
268 When compiled that way, the whole SimGrid (or almost) is put in a
269 single compilation unit and compiled in one shoot.
270 This is to help gcc which has difficulties to inline stuff from one
272 The speedup seem to be above 15%, althrough more tests are needed on
273 amd64 to confirm that gain.
276 * Port of MSG's mailbox on top of SIMIX network
277 The put/get mechanism was greatly simplified on the way.
280 * New SIMIX network module. Provides:
281 - Mailbox: rendez-vous mecanism to find with who you want to speak
282 - Synchronous send/recv: easier and hopefully faster since the
283 logic is handled in the maestro process directly now
284 - Asynchronous send/recv: you dreamt of it? It's here now
285 Too bad that nobody cared enough to propagate the change to MSG.
286 * Add semaphores as SIMIX synchronization mechanism.
289 * new function SD_daxload(char*) to load a DAX file
290 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
291 * Introduce typed tasks. Specify its kind and cost at creation.
292 At scheduling, just give where it should be placed, and the cost
293 for each involved resource is automatically computed.
294 Existing constructors so far (more to come of course):
295 - SD_task_create_comm_e2e() for end-to-end communication
296 - SD_task_create_comp_seq() for sequential computation
297 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
298 * new function SD_task_dump() for debuging display
299 * new function SD_task_dotty(task,FILE*) writing to file the info
300 about the task in dotty format
301 * SD_task_dependency_exists() can now cope with having one of its
302 arguments NULL. If so, it tests whether the other argument has any
304 * Add getters on list of preceding/following tasks:
305 SD_task_get_parents(task) and SD_task_get_children(task)
306 * Add getters on amount of workstations and list:
307 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
308 * Add getter on task kind: SD_task_get_kind(task)
309 * Update the start_time and finish_time of tasks on completion/failure
310 * Bugfix: Remove task from state swags when destroyed
313 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
316 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
317 dynar with one separate thread per value of the dynar.
318 * Change the prototype of xbt_thread_create(), sorry.
319 Added a boolean parameter indicating whether we want to join this
320 thread (used in SG only for now)
321 * Implement xbt_thread_join and xbt_thread_yield in SG also.
324 * GTNetS wrappers should now be usable again (and betterly tested too)
325 * Fix a major regression from 3.2 where the timeout provided to
326 MSG_task_put_with_timeout() was used as absolute time before which
327 the comm should be done.
328 * Start to fix the <cluster> tag.
329 - Internal links should be good now (beside of the loopback, which
330 use the private link instead)
331 - paths to the external world is still rather broken
332 - the <route:multi> tag is just broken. Actually that's brain-dead.
333 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
334 to make it less stupid
335 ** Check your platform with teshsuite/simdag/platforms/flatifier **
336 * Fix a source-level compatibility glitch from 3.2: after defining
337 MSG_USE_DEPRECATED, you can use the old name
338 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
339 * Allow to compile from the SVN with automake 1.11
340 * Fix some problems when using the "start_time" tag in deployment XMLs.
341 * Fix #8569: XBT/synchro.h has redundant declarations
342 * Fix #8563: MSG return values and exceptions
343 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
344 * Integrate patch #8636: Obey DESTDIR when installing documentation.
345 Thanks to Robson Peixoto.
346 * Fix a vicious bug in dictionaries inducing that some elements were
347 not freed on xbt_dict_free()
349 Portability report of this version:
350 * Main portability targets:
351 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
353 - mac leopard on i386
354 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
355 but nothing critical.
356 * Other platforms: windows, AIX and others were not tested for this release
358 Timing report of this version:
359 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
360 scenario parameters). From 8h to a few seconds in desktop grid settings.
361 * Supernovae brings about 25% speedup on i386.
363 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
365 SimGrid (3.3.3) stable; urgency=low
367 The "Need for Speed" release.
369 The timings done to validate the 3.3.2 were faulty.
370 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
372 The problem was a conversion from a manually handled vector to
373 xbt_dynar_t on the critical path.
374 xbt_dynar_foreach calls functions, inducing stack management crap.
376 We inlined these functions and xbt_dynar_foreach is now breath taking.
377 We also inlined xbt_swag_belong on the way.
379 Here are some approximate speedup measurements (on master/slaves
380 simulations lasting between 10s and 20s each):
381 3.3.1 -> 3.3.2: about same performance
382 3.3.2 -> 3.3.3: 40% speedup
383 3.3.1 -> 3.3.3: 40% speedup
384 3.3.1 with inline patch -> 3.3.3: 30% speedup
386 Our reading is that the refactoring which occurred in 3.3.2 made us
387 suffer much more from the xbt_dynar_foreach low performance, but
388 once we solved this, this refactoring proved to be very performance
389 effective. From the 40% speedup, somehow, 10% are due to the
390 inlining and 30% to the refactoring.
392 That's a pitty that gcc cannot inline functions placed in other files
393 alone. We have to choose between:
394 - break the encapsulation (by putting private data structures and
395 accessors in headers files to help gcc)
396 - live with low performance
397 - switch to a decent compiler such as icc (not quite possible).
399 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
401 SimGrid (3.3.2) stable; urgency=low
403 The "Simplicity does not preceed complexity, but follows it" release.
405 The main contributors of this release were (lexical order):
406 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
409 * Extract the routing logic into its own object.
410 (was dupplicated in network.c and workstation_LV07.c;
411 Allows to implement other ways of storing that info)
412 => kill now useless network_card concept
413 - Use dynar to represent routes (instead of void** + int*)
414 - kill link_set (use surf_network_model->resource_set instead)
415 - Add a command-line option to choose the routing schema to use
416 - Add three new models:
417 * Floyd (shortest path computed at initialization)
418 * Dijikstra (shortest path recomputed all the time)
419 * Cached Dijikstra (shortest path computed on need)
420 All these models where contributed by Silas De Munck, and are
421 described in his ICCS09 paper.
423 * Simplify model declaration
424 (less redirections, less function to write when defining a model)
425 - Factorize stuff between models:
428 surf_model_resource_set(model)
429 surf_model_resource_by_name(model, name)
430 - Unify the types of models in s_surf_model_t (using an union)
431 - Embeed fields of common_public directly into s_surf_model_t
432 - Rename model methods:
433 action_free ~> action_unref
434 action_change_state ~> action_state_set
435 action_get_state ~> action_state_get
436 - Change model methods into functions :
437 (model)->common_public->action_use ~> surf_action_ref
439 * Implement a generic resource; use it as ancestor to specific ones
440 (allows to kill duplicated code in models)
441 Drawback: timer command don't need no name nor properties;
442 workstation_CLM03 don't need no properties
443 (but I guess we can live with those few bytes wasted)
445 * Improve the action object model
446 - implement a constructor avoiding dupplicated code about field
447 initialization in generic_action part.
449 * Kill the SDP model: it has an external dependency, is deprecated
450 in flavor of modern lmm models, and didn't compile since a while
453 * Relocation of the context module from XBT to SIMIX.
454 (the context were decoupled from the simix processes, duplicating a lot of code)
455 => a lot of code was factorized
456 - less overhead is introduced during scheduling
457 - simpler API for the context factory
458 - the logic for process creation,destruction and manipulation was simplified
459 * Simplification of the s_smx_process_t data structure.
460 => accesing the simix level data associated to a process is faster now,
461 and the code is a lot more readable.
464 * Implement some more MPI primitives:
465 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
466 -implementation: Bcast: flat or 2-ary tree (default),
469 Allreduce: Reduce then Bcast
470 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
471 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
472 Alltoallv: flat tree, like ompi
474 * Add support for optimized collectives (Bcast is now binomial by default)
475 * Port smpirun and smpicc to OS X
478 * Kill SD_link_get_properties: hard to maintain and makes very little sense
479 Shout out if you used it.
482 * Display the list of still queued messages in SG mode when existing
486 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
487 * Add xbt_graph_node_get_outedges() [Silas De Munck]
488 * Add xbt_str_from_file(FILE*)
489 * Add xbt_dict_get_key achieving a linear reverse search
490 * Remove the context module
492 Portability report of this version:
493 * Main portability targets:
494 - Linux(debian)/x86/context
495 - Linux(debian)/x86/pthreads
496 - Linux(debian)/amd64/context
497 - Linux(debian)/amd64/pthreads
498 On these, we still have the eratic breakages of gras/pmm and
499 amok/saturate_sg reported in previous version. We still think
500 that the tests are the cause of the fault, not the tested code.
502 - Mac OSX Leopard/x86/context
503 Still false negative in tesh autotesting.
504 Smpi still fails, but this time because readlink does not accept -f
505 Everything seems to work properly beside of that.
508 - AIX version 5.3 (only tested contexts this time)
509 Smpi still fails there because mktemp is not installed.
510 Everything seems to work properly beside of that.
512 I managed to compile it for the first time, but several breakages.
513 Won't delay the release for this exotic platform.
515 * Windows: it's still lagging behind. If you want to help, please
518 Timing report of this version:
519 This version seem to be more than 5% faster than 3.3.1 (on linux
520 64bits with contextes). The gain is less than expected, we are
521 investigating this for next release.
523 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
525 SimGrid (3.3.1) stable; urgency=low
528 * Implement a --cfg-help to show existing configuration variables
529 * Build chain do not require doxygen in maintainer mode
532 * fix a bug on struct sizeof computation, which prevented the
533 exchange of arrays of structs in some conditions
534 - added a regression test about this in datadesc_usage
535 * Allow the exchange of 0-long dynamic vectors.
536 - for that, use -1 as indicator of dynamic size instead of 0
537 - This implied to change any size from unsigned long to long,
538 reducing a bit communication abilities, but I guess that with
539 64bits being quite common, this is more than enough.
540 - This also induce a protocol change, thus bumping network protocol
541 version from 0 to 1 (if we have external users, we have to get
542 clean on that point too ;)
543 - added two regression tests about this in datadesc_usage
544 * Be more verbose when propagating local exceptions
545 This helps debugging.
546 * Display the status of simulated processes when receiving SIGINT in
550 * Allow to control the simulation from a trace file.
551 New functions MSG_action_register() and MSG_action_trace_run()
552 The first one allows to associate a function execution to each
553 kind of action while the second one parses a trace file and
554 triggers the corresponding actions within the system.
555 For now, only a toy example is provided in examples/msg/actions
556 * Add an exemple of process migration in examples/msg/migration
557 * Fix a bug in task exchange which broke MSG_task_get_sender()
558 Add a teshsuite regression test for that.
559 [Bug: if MSG_task_get_sender() is called after sender exit,
561 * Fix a bug which prevented suspend/resume to work properly
562 * Display the status of simulated processes when receiving SIGINT
563 This fixes a regression of v3.3. due to the introduction of SIMIX
564 * Bug fixing in failure management:
565 - trace could not start by a failure at time 0
566 - failure during communications were not working
569 * Add SIMIX_process_set_name() to change the name of the current
570 process in the log messages.
571 * Store smx_hosts in a dict since we only retrieve them by name
572 * Move the configuration infrastructure to surf
575 * Move the configuration infrastructure to surf
578 * Massive internal cleanups:
579 - Store internal structures on processes instead of hosts (allows
580 to have more than one process per host, in addition of being more
582 - Cleanup the initialization/finalization process
583 - Kill a whole bunch of unneeded synchronization:
584 processes run in exclusive manner within the simulator
585 - Move queues from global tables to process data fields
587 - now accept -platform and -hostfile arguments
588 - Pass the right rank value to processes according to the hostfile
589 * Compile the examples by default, and use them as regression tests
590 * Implement MPI_Wtime()
591 * Change the reference speed to a command line option
594 * TCP_gamma can now be specified as command line option using
595 --cfg=TCP_gamma:10000000.0
596 * Change the --surf-path cmd line option into --cfg=path:
599 * Also include strbuff from xbt.h public header
600 * xbt_ex_display(): do not free the exception after displaying
601 This allows to do more with the given exception afterward.
602 Users should call xbt_ex_free() themselves.
606 Portability report of this version:
607 * Main portability targets:
608 - Linux(debian)/x86/context
609 - Linux(debian)/x86/pthreads
610 - Linux(debian)/amd64/context
611 - Linux(debian)/amd64/pthreads
612 These targets fail about 1/10 of times on gras/pmm, but we believe
613 that this is because of the test, not because of simgrid.
614 amok/saturate_sg fails even more rarely, and the test may not be
617 - Mac OSX Leopard/x86/context
618 The test suite still spits tons of errors because some obscure
619 force prevents us from removing the temporary directories
620 arguing that they still contain some metadata I've never heard of.
621 Smpi fails because seq is not installed.
622 Everything seems to work properly beside of that.
625 - AIX version 5.3 (both contexts and pthread)
626 Smpi still fails there because mktemp is not installed.
627 XML inclusions seems rosty on AIX.
629 * Windows: it's still lagging behind. If you want to help, please
632 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
634 SimGrid (3.3) stable; urgency=high
638 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
641 * Introduce the SIMIX module: factorize code between MSG and GRAS.
644 Until now, GRAS were using MSG as an interface to SURF. It was
645 quite difficult because both interface have several differences
646 (MSG channels vs GRAS sockets were the most notable point).
648 This also opens the gate to SMPI (which should occur soon) and speed
649 up simulations by to 40% (even if it were not the main goal).
651 **************************************
652 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
653 **************************************
654 The point is that events occuring at the exact same timestamp are
655 not scheduled in the same order with the old and new version. This
656 may be enough to completely change the execution of simulations in
657 some cases. Sorry for the inconvenience.
659 * Cleanup and upgrade the XML format to push further scalability
660 issues (check http://hal.inria.fr/inria-00256883/ for more info)
662 * Improve the testing infrastructure with tesh. Now a very large part of
663 the code is tested not only by being run but also by checking that the
664 output match an expected output [Mt].
666 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
667 is really less memory-demanding, which should allow you to use
668 larger files in SimGrid [AL].
670 * Inform valgrind about our contextes, so that it becomes usable
671 with the default (and more effecient) version of SimGrid
672 [contributed by Sékou Diakite, many thanks]
675 * Introduce a listener thread in charge of receiving incomming
676 messages from the network. It allows to overlap communication and
677 computation but most notably, it removes some stupid deadlocks due
678 to the fact that so far, a process could not send and receive at
679 the same time. This made most non trivial communication schema
681 * Convert the PIDs from long int to int to match the MSG ones (and
683 * New function: gras_agent_spawn() to launch a new process on
684 current host. Only working in simulation for now. [Mt]
685 * New function: gras_os_hostport() returning a constant form (ie,
686 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
689 * Make the backtrace of exceptions more human readable [Mt]
690 * New module: xbt/str [Mt]
691 a ton of string utility functions (split, join, printf to a newly
692 allocated buffer, trim, etc)
693 * New module: xbt/hash [Mt]
694 SHA1 hashing algorithm (more to come if needed)
695 * New module: xbt/synchro [Mt]
696 synchronization tools (mutex and conditions) working the same way
697 in simulation and in real life (mainly useful for GRAS, but not
699 * New module: xbt/queue [Mt]
700 classical producer/consumer synchronization scheme
701 * xbt_dynar_new_sync() creates a synchronized dynar. All access
702 (using the classical functions will get serialized) [Mt]
703 * Make dictionary internal table dynamic. No need to specify its size
704 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
706 * Make sure the log channels are organized as a tree under windows
707 (because of ANSI C compatibility issue, any channel were child of
711 * Cleaned many thing in surf and fixed a few bugs [AL].
712 * Add a nice command line configuration mechanism to compose models [AL].
713 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
714 the previous one (KCCFLN05). It relies on something that looks like
715 a max-min sharing mechanism but cannot be written as such. A new solver
716 was thus designed [AL].
717 * Add a new solver to lmm. Based on Lagrange optimization and
718 gradient-based descent, it enables to efficiently maximise systems s.a
720 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
722 This solver enables to propose two new network models for TCP Reno and
723 TCP Vegas based on Low's work. These models still need to be fully
724 tested though [Pedro Velho].
727 * Bug fix in SD_simulate. Now the time bound given as argument is
729 * Use the new parallel task model (ptask_L07) as default.
730 * Use the SURF command line configuration mechanism.
731 * 0-size tasks (for synchronization) should now work.
733 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
735 SimGrid (3.2) stable; urgency=high
739 We still experience issues on this platform, but we believe that at
742 GRAS API BREAKAGE (for simplification purpose, sorry):
743 * the gras_msgtype_by_name is not used anymore. Instead of
744 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
745 you can write (and must)
746 gras_msg_send(toserver, "request", &request);
747 - If you still want to pass a gras_msgtype_t to the function (to cache
748 the type and avoid the lookup time), use the gras_msg_send_() variant.
749 - Impacted functions:
750 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
751 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
752 * The callbacks are now expected to return 0 when everything went well
753 (just like the main() function)
755 GRAS new features and improvements:
756 * New module mecanism where user code can use per process globals [Mt]
757 This is similar to gras_userdata_*() functions, but for libraries. It
758 factorize some code developped over and over in the examples and AMOK.
759 It has still to be documented and used (only amok/peermanagement is
761 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
762 loss when they were small enough to fit into the buffer and sent quickly
763 enough so that they can all get received in one shoot.
764 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
765 if not found. Use the *_or_null() variant for the old semantic.
766 * In gras_msg_handle, do not discard messages without callback.
767 They are probably messages to be explicitly awaited later (ie, proofs of
768 mis-synchronization in userland since they are sent before being awaited)
770 * gras_socket_meas_send/recv: semantic changed!
771 The numerical arguments used to be (1) the total amount of data to send
772 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
773 messages. This was need for the fool willing to send more than MAXINT
774 bytes on quite fat pipes.
777 * Do really rename the hostmanagement module to peermanagement. [Mt]
778 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
779 but this is rather new and this was documented in the module
780 documentation (poor excuses, I admit)
781 * Bandwidth measurement semantic changed! This follows the changes to
782 gras_socket_meas_send/recv explained above.
785 * A sequential mode has been added to the workstations. When a workstation
786 is in sequential mode, it can execute only one task, and the other tasks
787 are waiting in a FIFO. [Christophe Thiery]
790 * The KCCFLN05 workstation model now handles parallel tasks. It is the
791 model for SIMDAG. [Christophe Thiery]
792 * Bug fix in the maxmin solver: Some values were close to 0 instead of
793 equal to 0, which caused some bad behaviors in
794 saturated_constraint_set_update. I now use a threshold mechanism like in
798 * When running manually src/testall, you select specific units [Mt]
799 testall is the result of our cunit mecanism, and should replace all
800 the scripty thingy around since bash don't run easily on billware.
802 * A mallocator system has been added. [Christophe Thiery]
803 Mallocators allow you to recycle your unused objects instead of freeing them
804 and allocating new ones.
806 Documentation update:
807 * FAQ reworking + New FAQs:
808 - "Valgrind spits tons of errors!" [Mt]
809 - "How to repport bugs" [Mt]
810 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
811 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
812 - Communication time measurement within MSG [AL]
813 - I experience weird communication times when I change the latency [AL]
816 - an introduction to the framework and to the used communication model
817 - an initiatic tour introducing the most proheminent features:
819 . Lesson 0: Installing GRAS
820 . Lesson 1: Setting up your own project
821 o Part 2: Message passing
822 . Lesson 2: Exchanging simple messages
823 . Lesson 3: Passing arguments to the processes (in SG)
824 . Lesson 4: Attaching callbacks to messages
825 . Lesson 5: Using globals in processes
826 . Lesson 6: Logging informations properly
827 . Lesson 7: Using internal timers
828 . Lesson 8: Handling errors through exceptions
829 . Lesson 9: Exchanging simple data
830 . Lesson 10: Remote Procedure Calling (RPC)
831 . Lesson 11: Explicitely waiting for messages
832 . Recapping of message passing features in GRAS
833 - A HOWTO section containing:
834 o HOWTO design a GRAS application
835 More are due, of course. They will come latter. In the meanwhile, you can
836 check the examples which are still here.
838 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
840 SimGrid (3.1) stable; urgency=high
844 There was a stack corruption somewhere, visible only when optimizing
845 with these versions. [Vince]
848 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
849 module that enabled to play with Directed Acyclic Graphs. It is built
850 directly on top of SURF and provides an API rather close to the old
851 SG. Some old codes using SG are currently under rewrite to check that
852 all needful functions are provided. [Christophe Thiery]
855 * Complete rewrite of the KCCFLN05 workstation model. It is now an
856 extension of the classical CLM03 model that gracefully handles
857 failures. This is now the default model for MSG and GRAS. It doesn't
858 handle parallel tasks yet though. [AL]
859 * Bug fix: Weights were not correctly set in the network part.
860 WARNING: This may have resulted in incorrect results with simulations
861 where there are more than one flow on a given link. [AL]
864 * After a (long ?) discussion on simgrid-devel, we have decided that the
865 convention we had on units was stupid. That is why it has been decided
866 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
867 WARNING : This means that all previous platform files will not work as
868 such with this version! A warning is issued to ask users to update
870 A conversion script can be found in the contrib module of the CVS, under
871 the name contrib/platform_generation/surfxml_update.pl [MQ]
874 * Bug fix: Processes were started in reverse order, wrt deployment file.
875 WARNING: if your code relies on this bug, please fix it. [AL]
876 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
877 being executed on two different locations. [AL]
878 * Bug fix: Failures are now better supported thanks to Derrick's tests
879 (there was many failure situations I hadn't thought of and that weren't
880 correctly handled). [AL]
881 * New function: MSG_host_is_avail indicates you whether a given m_host_t
885 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
886 Exception occurring on server-side are propagated back to client (!).
888 API CHANGE: the callback changed their prototype. Change:
889 int my_handler(gras_socket_t expeditor, void *payload_data) {
891 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
892 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
894 * New! function: gras_msg_handleall to deal with all messages arriving
895 within a given period.
896 * New! function: gras_socket_server_range to get a server socket in a
897 range of port numbers (ease to avoid port number conflicts) [MQ]
898 * New! gras processes display their backtrace when they get a SIGUSR1
899 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
900 Sweet to debug RL processes [MQ]
904 - Do not force experiment sizes to be expressed in kb, or it becomes
905 impossible to measure the latency this way (needs one byte-long tests)
906 WARNING: this changes the amok_bw_* function semantic. [MQ]
907 - Implements the link saturation stuff. [MQ]
908 * Peer management module:
909 New! module factorizing code that we wrote over and over [MQ].
912 * New module: cunit (my jUnit implementation in ansi C) [MQ]
913 - Test units are placed directly into the library code, they get extracted
914 automatically and placed into the src/testall binary.
915 - Convert most of the XBT tests to this system.
916 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
917 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
918 live in the same C code. This required to change a little bit the API
919 of surfxml parsing but shouldn't be an issue for end-users. [AL]
920 * New module: sparse graph structure with basic algorithms (this is work
921 in progress and the API is not considered to be frozen yet). [AL]
922 * Display more information on backtraces: source line & function names are
923 now displayed just like valgrind does (rely on addr2line tool) [MQ]
924 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
925 * Reworked a little bit some #include statements to load only required
926 headers. Some user code that relied on SimGrid to include stdlib or
927 stdio may need to include it by themselves. [AL]
928 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
930 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
931 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
932 but I guess that almost nobody used those parts. [MQ]
934 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
936 SimGrid (3.0.1) stable; urgency=low
939 * Unfortunately, I had missed 5 misnamed functions:
940 xbt_fifo_item_t xbt_fifo_newitem(void);
941 void xbt_fifo_freeitem(xbt_fifo_item_t);
942 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
943 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
944 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
945 They're now deprecated. Please use their new versions:
946 xbt_fifo_item_t xbt_fifo_new_item(void);
947 void xbt_fifo_free_item(xbt_fifo_item_t);
948 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
949 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
950 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
952 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
953 * Documentation: xbt_log module unmercifully reworked [MQ]
954 * Bugfix: there was a problem with the ending of contexts with
955 the pthread backend. It caused some weird deadlock or behavior
956 depending on the pthread implementation. [AL]
957 * Bugfix: get the exceptions raised in the simulator repport where
958 and why they come from when they are not catched in time [AL, MQ]
961 * Bugfix: Do repport the error when two non-connected hosts try to
962 exchange data (Thanks to Flavien for stumbling into this one) [AL]
965 * Add additionnal checkings on communications. Assert that two
966 communicating hosts are connected by a set of links... [AL]
969 * Add additionnal checkings on channel values in communication [AL]
970 * New: MSG_task_get_source to see on which host a task was generated [HC]
971 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
972 the number of tasks waiting to be received on channel and sent
974 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
975 waits for the first task coming from a given host.. [AL]
977 GRAS new functionnalities: [MQ]
978 * Enhance the parsing macro to allow the size of multidimentional objects
979 to be given thru annotations.
980 * New example (and documentation): Matrix Multiplication a la RPC
981 (as when I was young!) and fix a bunch of bugs found on the way.
983 GRAS performance improvements: [MQ]
985 * Reduce the amount of cbps creation/destruction by making it static to
986 datadesc_send/recv() and using a (newly created) cbps_reset (based on
989 * Change libdata to a set so that we can search for stuff by ID (and thus
990 reduce the insane amount of dict lookups)
993 * Actually implement gras_datadesc_copy() so that we don't have to mimick
994 RL communication on top of SG since it's so uneffective.
995 It may also be used for inter-thread communication in RL, one day.
996 * Use gras_datadesc_copy() to exchange messages on top of SG
998 - improve message exchange performance on top of SG
999 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1000 * Don't exchange on the network the size of the used part of buffer,
1001 instead, specify the possible buffer size to read().
1003 - reduces the amount of read/write calls (one pair per exchange)
1004 - reduces the amount of exchanged data (the size)
1005 - allows to retrieve all arrived data on receiver side, if we don't need
1006 it right now (subsequent read will peek the buffer)
1007 - allows the receiver to proceed with the begining of the stream before
1008 everything is arrived
1009 - make it possible to build an iov transport (using readv/writev)
1011 - take care of the data with non-stable storage (like stacked data),
1013 * If possible, TCP send uses vector I/O (when writev() is here)
1014 - Don't use it for receive since we send data sizes and data on the
1015 same stream, so we wouldn't be able to chain large amount of chunks
1016 before having to flush the stuff to read the size.
1017 * Rework the transport plugin mecanism to simplify it and reduce the
1018 amount of pointer dereferencement when searching for the right function
1021 * I guess that now, we do almost as few system calls as possible while
1022 doing as few data copy as possible.
1024 To improve it further, we could try to send all the sizes first and then
1025 all the data (to use iov on receiving size), but it's only a partial
1026 solution: when you have 2 dimensional data, the sizes of the second
1027 dimension is data of the first dimension, so you need 3 streams.
1029 I'm not sure the potential performance gains justify the coding burden.
1031 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1033 SimGrid (3.00) stable; urgency=high
1036 * New! Give the possibility to hijack the surf parser and thus bypass
1037 MSG_create_environment and MSG_launch_application. Have a look at
1038 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1040 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1042 SimGrid (2.96) unstable; urgency=low
1047 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1048 This deprecates the xbt_error_t mecanisms.
1049 It modifies (simplifies) all XBT and GRAS API.
1050 MSG API keeps unchanged (exceptions raised by XBT are catched from
1051 within MSG and masked with existing error handling facilities)
1054 * New! Add a FATPIPE model. [AL]
1055 * New! Add a parallel task model. [AL]
1056 * New! Add automatically a loopback interface (in the default
1057 network model) if none was precised.
1060 * Bugfix: MSG_process_resume now works with the current running process.
1062 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1063 * Modification of MSG_task_get_compute_duration. Once a task has been
1064 processed, the value returned by this function is now equal to 0. [AL]
1065 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1066 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1067 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1068 of a m_task. For now, MSG_task_cancel only works with computation
1070 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1071 of the processor (in Mflop/s) regardless of the current load on the
1073 * API Change: use proper naming convention for MSG_getClock and
1074 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1076 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1077 This function changes the priority of a computation task. This priority
1078 doesn't affect the transfer rate. A priority of 2 will make a task
1079 receive two times more cpu power than the other ones. This function
1080 has been added to suit the needs of Nguyen The Loc and hasn't been that
1081 much tested yet. So if it fails, please report it and send me your code.
1083 * API Change: removed all functions and types that were marked "deprecated"
1084 since many months. Renamed MSG_global_init_args to MSG_global_init.
1086 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1088 SimGrid (2.95) unstable; urgency=low
1091 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1092 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1093 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1096 * Bugfix: Allow absolute paths for platform description files [MQ]
1097 * Bugfix: do free the variables after use. Leads to drastic performance
1099 * Implement max_duration (ie, timeouts) on resources [AL]
1102 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1104 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1105 the needed support (timeouts) [AL]
1108 * Implement measurement sockets. You can now get the bandwidth between two
1109 hosts thanks to AMOK (see below). [MQ]
1110 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1111 dynar over the network (yeah) [MQ]
1112 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1113 * simple/static token-ring example. [Alexandre Colucci and MQ]
1114 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1115 in place before [MQ]
1118 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1119 type parsing. "short" and "long" modifiers were also ignored; other
1120 modifier (such as reference level) are still ignored. [MQ]
1121 * Embeed the buffer size within the buffer itself on SG. [MQ]
1122 That way, send() are atomic and cannot get intermixed anymore (at least
1123 the ones which are less than 100k; bigger messages still have the issue)
1124 * Array size pushed by the field, not by the field type (or each
1125 and every long int will push stuff to the cbps) [MQ]
1126 * use select() to sleep since it allows to portably sleep less than one
1129 GRAS (minor cleanups)
1130 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1133 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1134 * type category 'ignored' killed as it was never used and were difficult
1136 * whether a type can cycle or not is now a flag, leaving room for more
1137 flags (as "ignored", if we feel the need one day ;)
1138 * Rename raw sockets to measurement sockets since "raw" has another
1139 meaning in networking community.
1142 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1143 and offers features not belonging to GRAS but that most applications
1144 need. One day, it may be a set of plugins loadable at runtime.
1145 * New module: bandwidth
1146 bandwidth measurement between arbitrary nodes running this module. [MQ]
1148 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1150 SimGrid (2.94) unstable; urgency=low
1152 The first beta release of SimGrid 3 !
1156 * Update the main page and the FAQ. Adding references to gforge.
1159 * Add a gras_os_getpid function.
1162 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1163 * Extend the logs so that they also print PID, hostname, date, ... if
1165 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1166 and kill the old version which were in testsuite/
1167 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1171 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1172 for users. Instead, each programming environment (GRAS, MSG,...) use it
1173 when needed to provide such a feature to users.
1174 Don't shortcut the mecanism or you will also shortcut the virtualization
1175 you need on the simulator.
1179 * Cleanups in configury with regard to compile optimization/warning flags.
1180 Also add -fno-loop-optimize to any powerpc since it's the optimization
1181 killing gcc (< 3.4.0).
1182 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1183 needed by MSG examples complications
1184 * Borrow configury beautifications from PHP
1187 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1188 warning (thanks loris for stumbling into this one).
1189 * Bugfix: stop loading private headers (gras_config.h) from the public
1193 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1194 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1195 RL would imply a DNS resolver, which is *hard* to do in a portable way
1196 (and therefore delayed).
1197 * Implement a real timer mecanism and use it in timing macros. This allows
1198 to avoid rounding errors and get a 0.000005 sec precision in timing
1199 macros. While I was at it, various cleanups:
1200 - allow to declare more than one timed section per file (fix a stupid bug)
1201 - move some private declaration to the right place
1202 - merge conditional execution and timing macros into emulation module
1203 - document the module
1204 - make sure the module cleanups its mess on gras_exit
1205 * Documentation improvements:
1206 - (new) how to compile applications using GRAS
1207 - (new) emulation support (timing macros)
1209 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1211 SimGrid (2.93) unstable; urgency=low
1213 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1216 - Use Paje properly where used. Still to be sanitized properly.
1217 - Portability fix: Add an implementation of the contexts using pthread
1221 - Add xbt_procname(): returns the name of the current process.
1222 Use it to show the current process's name in all logging.
1224 - fix detection of older flex version and the reaction, since we do
1225 depend on modern ones (we use lex_destroy)
1226 - Better separation of SG and RL in the libs: remove all simulation code
1227 from libgras. As a result, this lib is now only 200k when stripped.
1228 Some of the xbt modules may also be duplicated (two sets and such) and
1229 should be cleaned/killed before SG3.
1230 - Insist on using xlC on AIX because of weird problems involving gcc there.
1231 - Cleanup the make remote stuff. This is now done by scripts
1232 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1233 mainly for our private use, but we're working on changing them to user
1236 - Bugfix: flush the socket on close only if there is some *output*.
1237 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1239 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1242 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1244 SimGrid (2.92) unstable; urgency=low
1246 Alpha 3 on the path to SimGrid 3
1250 - New! First try of benchmarking macros.
1251 - New! First try so that gras_stub_generator generate deployment and
1252 remote compilation helpers.
1254 - Bugfix: Initialization fix in msg_test.
1258 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1260 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1262 - API Change: gras_os_sleep to take the amount of seconds as a double.
1263 Accepting an int was error prone since it was the only location where
1264 seconds were coded as such. It leaded to damn rounding errors.
1265 - Bugfix: Hard to belive that timers ever worked before this.
1267 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1269 SimGrid (2.91) unstable; urgency=low
1271 Alpha 2 on the path to SimGrid 3
1275 - Bug fix in the lmm_solver.
1277 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1278 through the function MSG_paje_output.
1279 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1280 - It is possible to bound the rate of a communication in MSG with
1281 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1282 to write it in the changelog).
1283 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1288 - Various cleanups to the autotools stuff
1289 - Begin to move Gras examples to examples/gras/
1290 - Let make distcheck work again (yeah!)
1292 - documentation overhauled using doxygen.
1293 gtk-doc-tools is dead in SimGrid now.
1294 - Automatically extract all existing logging categories, and add the list
1295 to the documentation (long standing one, to say the less)
1297 - Cleanup the known architecture table. Reorder the entries to group what
1298 should be, and use a more consistent naming scheme.
1299 (some of the test dataset are still to be regenerated)
1300 - New! Allow library to register globals on each process just as userdata
1302 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1303 lookup time (for now).
1304 Use it in msg and trp.
1305 This cleans a lot the internals and helps enforcing privacy of the
1306 headers between the gras components.
1307 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1308 - Bugfix: gras_os_time was delirious in RL.
1309 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1311 - Reenable GRAS now that it works.
1313 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1315 SimGrid (2.90) unstable; urgency=low
1317 Alpha 1 on the path to SimGrid 3
1319 * It is a long time since the last release of SimGrid. I'm sorry about
1320 that but as I had told you, I was rewriting a lot of things. I apologize
1321 to those who had been reporting bugs to me and that I had not answered.
1322 If your bug is still in the new version, please tell me. Here is a
1323 summary of the main changes.
1325 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1326 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1327 portability, speed, and a lot more but you'll figure it out later.
1328 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1329 I wanted to release the new MSG as soon as possible and I have
1330 broken GRAS, which is the reason why, for now, only MSG is fully
1331 functional. A laconic description of these projects is available
1332 in the documentation.
1334 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1335 kernel optimized for our needs (called SURF but only the developers
1336 should use it). Hence, MSG is now roughly 30 times faster and I think
1337 that by rewriting a little bit MSG, I could event speed it up a little
1338 bit more. Beside the gain in speed, it is also much easier to encode a
1339 new platform model with SURF than it was with SG. More to come...
1341 * REVOLUTION 3: I have tried to change a little as possible the API of
1342 MSG but a few things really had to disappear. The main differences
1343 with the previous version are :
1344 1) no more m_links_t and the corresponding functions. Platforms are
1345 directly read from a XML description and cannot be hard-coded
1346 anymore. The same format is used for application deployment
1347 description. The new format is described in the documentation.
1348 Have a look in tools/platform_generation. There is a tiny script
1349 that converts from the old platform format to the new one. Concerning
1350 the application deployment format, parsing the old one is tricky.
1351 I think most of you should however be able to convert your files. If
1352 it is really an issue, I can write a C code that does the conversion.
1354 2) the toolbox tbx does not exist anymore. We now have a library
1355 with much more data-structures but without the hash-tables (we have
1356 dictionaries that are much faster).
1358 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1360 *****************************************************************************
1361 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1362 * don't feel like converting the oldies. (Mt) *
1363 *****************************************************************************
1366 Version 2.90: "the long awaited one"
1367 - Finished rewriting and debugging MSG. Rewrote the documentation.
1368 - disable GRAS for now since it needs to be ported to the newest SG
1371 - Finish the port to windows (using mingw32 for cross-compile)
1374 - Main loop and datastructures of SURF. A cpu resource object is
1375 functional. Surf can thus be used to create cpu's with variable
1376 performance on which you can execute some actions.
1378 2004-11-15 Martin Quinson
1379 - Port to ARM. Simply added the alignment and size descriptions. Should
1380 work, but the ARM machines are so slow that I didn't had the opportunity
1381 to 'make check' over there yet.
1383 2004-11-15 Arnaud Legrand
1384 - Trace manager now written. It uses a heap structure and is therefore
1385 expected to be efficient. It may however be speeded up (particularly
1386 when many events occur at the same date) by using red and black
1387 trees. One day maybe...
1388 - Max-min linear system solver written. It uses a sparse matrix
1389 structure taking advantage of its expected use. Most operations are
1390 O(1) and free/calloc are called as few as possible. The computation of
1391 the minimum could however be improved by using a red and black tree
1394 2004-11-03 Arnaud Legrand
1395 - Rename every gras_* function that was in xbt/ to its xbt_
1397 - Add a heap and a doubly-linked list to xbt
1398 - Added a dichotomy to the dictionaries. make check works as well before
1399 so I assume that the patch is correct. I do not know however if things
1400 run effectively faster than before now. :)
1402 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1403 SimGrid, and the version number is bumped to 2.x
1405 2004-10-29 Martin Quinson
1406 - Introduction of the remote errors.
1407 They are the result of a RMI/RPC on the remote machine.
1408 ErrCodes being scalar values, you can't get the host on which those
1409 errors did happen. Extending the error mechanism as in Gnome is possible.
1410 No idea yet whether it is a good idea.
1412 2004-10-28 Martin Quinson
1413 - Interface revolution: the Starred Structure Eradication.
1414 I used to do typedef struct {} toto_t; and then handle *toto_t.
1415 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1416 - ???_t is a valid type (builded with typedef)
1417 - s_toto_t is a structure (access to fields with .)
1418 - s_toto is a structure needing 'struct' keyword to be used
1419 - e_toto_t is an enum
1420 - toto_t is an 'object' (struct*)
1422 typedef struct s_toto {} s_toto_t, *toto_t;
1423 typedef enum {} e_toto_t;
1424 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1425 s_toto_t) is private.
1427 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1428 so that it can set the variable to NULL. It was so for dicts and sets,
1429 it changed for dynars.
1431 - Fix a bunch of memleaks in dict_remove
1432 - Fix a bug in sg/server_socket opening: it failed all the time.
1434 2004-10-07 Martin Quinson
1435 - Speed up dynar lookup operation a bit.
1437 gras_dynar_get is dead.
1439 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1440 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1441 (which gives you the address of the stored data).
1443 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1444 copy of the data using an affectation to do the job and not a memcpy.
1446 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1447 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1449 It does not really speedup the dynar test because they are
1450 setting elements all the time (and look them seldom). But the dict does
1451 far more lookup than setting.
1453 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1455 2004-10-05 Martin Quinson
1456 - Allow to (en/dis)able the cycle detection at run time.
1458 Whether we should check for cycle or not is now a property of each
1459 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1460 datadesc_cycle_unset allow to remove this property when previously set.
1462 Note that the cycle detection is off by default since it impacts the
1463 performance. Watch the data you feed GRAS with ;)
1465 This property is hereditary. Any element embedded in a structure having it
1466 set have it set for the time of this data exchange.
1468 You should set it both on sender and receiver side. If you don't set it on
1469 sender side, it will enter an endless loop. If you forget on receiver
1470 side, the cycles won't be recreated after communication.
1472 - Header reorganization.
1473 Kill gras_private.h, each submodule must load the headers it needs.
1475 2004-10-04 Martin Quinson
1476 - Interface revolution: do not try to survive to malloc failure.
1478 Now, gras_malloc and friends call gras_abort() on failure.
1479 As a conclusion, malloc_error is not a valid error anymore, and all
1480 functions for which it was the only gras_error_t return value are
1481 changed. They now return void, or there result directly.
1482 This simplify the API a lot.
1484 2004-09-29 Martin Quinson
1485 - Re-enable raw sockets.
1486 Created by gras_socket_{client,server}_ext;
1487 Used with gras_raw_{send,recv}
1490 It should allow to kill the last bits of gras first version soon.
1492 This is not completely satisfactory yet (duplicate code with
1493 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1496 - Simplify transport plugin (internal) interface by not passing any
1497 argument to _server and _client, but embedding them in the socket
1500 2004-09-28 Martin Quinson
1501 - Finish the port to AIX.
1502 autoconf was my problem (segfault within the malloc replacement
1503 function. No idea why)
1505 2004-09-16 Martin Quinson
1506 - Fix some size_t madness on 64bit architectures.
1508 2004-09-08 Martin Quinson
1509 - Reduce the number of system headers loaded, overload some more system
1510 calls (such as malloc to cast the result of the system one, and work
1512 - Fix and reintroduce the config support
1514 2004-09-07 Martin Quinson
1515 - Source code reorganization to allow Arnaud to surf all over there.
1516 - Allow to document the logging categories.
1517 - Remove all uppercase from logging categories and useless cleanup in names.
1519 2004-08-18 Martin Quinson
1520 Version 0.6.2 (protocol not changed; API changed)
1521 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1522 gras_error_t), and NULL when not found. Functions expecting a msgtype
1523 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1524 hopefully usefull message.
1525 - Portability to prehistoric sparcs again
1527 2004-08-17 Martin Quinson
1528 Version 0.6.1 (protocol not changed; ABI not changed)
1529 - prealloc some buffers to speed things up
1531 2004-08-11 Martin Quinson
1532 Version 0.6 (protocol not changed; ABI expended)
1533 - The parsing macro can deal with the references, provided that you add
1534 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1536 2004-08-09 Martin Quinson
1537 Version 0.5 (protocol not changed; ABI changed)
1538 - Allow to off turn the cycle detection code in data exchange at
1539 compilation time. It should be at run time, but I'm short of time (and
1540 the config stuff is still broken). That way, we keep dict out of the
1541 critical path, which is good because the performance is poor:
1542 - search not dichotomial yet
1543 - dynar give no way to access their content and memcpy everytime
1544 - In composed data description (struct, ref and so on), stop foolness of
1545 keeping the subtype's ID, but store the type itself. This keeps sets out
1546 of the critical path, which is good since they rely on dynar and
1547 dictionnaries. The only loose of that is that we cannot detect the
1548 redeclaration of a structure/union with another content (but I'm not sure
1549 the code detected well this error before anyway). We still can detect
1550 the redefinition discrepancy for the other types.
1551 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1552 it breaks the code because of type-punning used all over the place).
1553 This breaks on all non-gcc architectures (for now).
1555 All those changes (plus the buffer of last time) allow me to gain 2 order
1556 of magnitude on cruel tests consisting of 800000 array of integers on two
1557 level of a hierarchical structure (200 secondes -> 4 secondes)
1560 - the selector of reference must now return the type it points to, not
1561 the ID of this type.
1563 2004-08-06 Martin Quinson
1564 Version 0.4 (protocol changed; ABI not changed)
1565 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1566 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1567 - Implement a buffer transport, and use it by default (it relies on tcp in
1568 real life and on sg in simulation).
1569 That's a bit hackish since I had a new field to the structure to store
1570 its data without interfering with the subtype ones. Inheritance
1571 is tricky in C. And that's a kind of reverse inheritance with one class
1572 derivating two classes. Or maybe a game with java interfaces. Anyway,
1573 that's damn hard in C (at least).
1574 Moreover, I got tired while trying to ensure plugin separation and
1575 genericity in SG mode. MSG wants me to do weird things, so let's go for
1576 cruel hacks (temporarily of course ;).
1577 See comment in transport_private.h:71
1578 - do not include all the _interface headers in private but in the files
1579 which really need them (to cut the compilation time when they are
1582 2004-07-26 Martin Quinson
1583 Version 0.3 (protocol not changed; ABI changed)
1584 - Major overhault of the datadesc interface to simplify it:
1585 - shorted the function names:
1586 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1587 - add a trivial way to push/pop integers into the cbps without malloc.
1588 This allows to make really generic sub_type description, which simply
1589 pop their size of the stack.
1590 - add a function gras_datadesc_ref_pop_arr() which does what users want
1591 most of the time: Declare a dynamic array (which pops its size of the
1592 stack) and declare a reference to it. Poor name, but anyway.
1593 - kill the post-send callback, add a post-receive one
1595 2004-07-23 Martin Quinson
1596 Version 0.2 (protocol changed; ABI changed)
1597 - add some testing for cpbs in the test cases, and fix some more bugs.
1598 This invalidate again the little64 data file, since I cannot regenerate
1600 - remove an awfull optimization in the logging stuff, allowing me to:
1601 - understand it again
1602 - learn gcc how to check that the argument match the provided format
1603 - fix all errors revealed by gcc after that
1604 - internal keys of dict are not \0 terminated. Deal with it properly in
1605 loggings instead of segfaulting when the user want to see the logs :-/
1607 2004-07-22 Martin Quinson
1608 - Fix some stupid bug preventing cbps (callback postit) from working
1610 2004-07-21 Martin Quinson
1611 - Some documentation cleanups
1612 - remove the useless last argument of msgtype_declare
1613 - rename the Virtu functions to fit into the 'os' namespace
1614 - move headers src/include -> src/include/gras/ and stop fooling with
1615 gras -> . symbolic link
1616 - make distcheck is now successful
1618 2004-07-19 Martin Quinson
1620 - Build shared library also
1621 - Install html doc to the right location
1622 - stop removing maintainer files in make clean
1623 - build tests only on make check
1625 2004-07-13 Martin Quinson
1627 - No major issue in previous version => change versionning schema
1628 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1629 corresponding dataset.
1631 2004-07-11 Martin Quinson
1633 - Get it working with any kind of structure (we can compute the padding
1634 bytes remotely for all the architectures I have access to)
1635 - Implement the structure parsing macro (still not quite robust/complete)
1636 - Improvement to the remote testing toysuite
1638 2004-07-10 Martin Quinson
1639 [autoconf mechanism]
1640 - get ride of a bunch of deprecated macros
1641 - actually run the test for two-compliment, not only compile it :-/
1642 - test whether the structures get packed (and bail out if yes. Damn.
1643 Alignment is a serious matter)
1644 - test whether the structures get compacted (but respecting the alignment
1645 constraints of each types)
1646 - test whether the array fields of structures can straddle alignment boundaries
1648 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1649 (took me 2 hours to find that bug, looking at the wrong place)
1650 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1651 before sending/receiving and are used to compute the offsets of fields
1652 - Given that padding size depend even on compiler options, keep track of
1653 alignment and aligned_size only for the current architecture. Not a big
1654 deal since we send structure fields one after the other (seems
1656 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1657 program, let it work on linux/gcc/little32. portability todo.
1659 2004-07-08 Martin Quinson
1660 - import and improve remote compilation support from FAST
1661 - make sure make check works on half a dozen of machines out there
1663 2004-07-07 Martin Quinson
1664 Let's say it's version 0.0.3 ;)
1665 - Implement conversions (yuhu!)
1666 - Let it work on solaris (beside conversion, of course)
1667 - Stupid me, using rand() to generate the conversion datatests in not wise.
1669 2004-07-06 Martin Quinson
1670 - Let make dist work, since I'm gonna need it to compile on remote hosts
1671 - Let Tests/datadesc_usage write the architecture on which the file was
1672 generated as first byte.
1673 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1674 architecture descriptions.
1675 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1676 on those architectures.
1677 - Optimization: send/recv array of scalar in one shoot
1679 2004-07-05 Martin Quinson
1680 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1682 - Plug a whole bunch of memleaks
1683 - each process now have to call gras_{init,exit}. One day, their log
1684 settings will be separated
1685 - Continue the code factorisation between SG, RL and common in Transport.
1687 2004-07-04 Martin Quinson
1689 - Redistribution between SG and RL.
1690 We wanna have to accept in SG, so move accepted related parts of RL in
1691 the common part. (more precisely, the dynar of all known sockets is no
1692 more a static in transport.c, but part of the process_data)
1694 [gras_stub_generator]
1695 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1697 2004-07-03 Martin Quinson
1698 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1699 them all up in one shot)
1701 - Fix the ugly case of reference to dynamic array.
1702 - New (semi-public) function gras_datadesc_size to allow the messaging
1703 layer to malloc the needed space for the buffer.
1705 - gras_socket_close now expect the socket to close (and not its address to
1706 put NULL in it after it). This is because the socket passed to handlers
1707 is one of their argument (=> not writable).
1709 - propagate the interface cleanup from last week in datadesc, ie remove a
1710 superfluous level of indirection. User pass adress of variable
1711 containing data (both when sending and receiving), and not of a variable
1712 being a pointer to the data. Let's say that I like it better ;)
1713 The price for that is constructs like "int msg=*(int*)payload" in
1714 handlers, but it's a fine price, IMHO.
1716 - Let it work in RL (yuhu)
1718 2004-06-21 Martin Quinson
1720 - porting SG plugin and SG select to new standards (works almost).
1721 - plug memleaks and fix bugs around.
1724 - cleanup the prototype of data recv and force users to specify when they
1725 want to handle references to objects. Test case working even for cycles.
1726 - plug memleaks. Valgrind is perfectly ok with this.
1728 2004-06-12 Martin Quinson
1730 - cleanup the separation between plugin and main code in plugin creation
1732 2004-06-11 Martin Quinson
1734 - Reput hook for raw sockets, needed for BW experiments
1735 - kill a few lines of dead code
1736 [Data description] Interface cleanup
1737 - gras_datadesc_by_name returns the searched type or NULL.
1738 That way, no variable is needed to use a type desc once, which makes
1740 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1741 two parameters were strings (field name, type name), leading to
1743 [Dicos] Interface cleanup
1744 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1745 This is consistant with the dynar API.
1747 2004-04-21 Martin Quinson
1749 - Porting to new standards.
1751 - interface cleanup.
1752 There is no bag anymore, no need to take extra provision to mask the
1753 pointers behind "ID".
1754 Better splitup of functions between files create/exchange/convert.
1755 This is still a bit artificial since convert and receive are so
1756 interleaved, but anyway.
1758 - add a queued message list to procdata (the ones not matching criteria
1760 - factorize some more code between SG and RL wrt procdata
1762 - use gras_exit in example to track memleaks
1763 - get rid of gs_example now that GS is properly integrated into gras
1764 - update run_test to integrate the lastest tests (datadesc)
1766 - rename WARNINGn macros to WARNn since it prooved error-prone
1768 2004-04-19 Martin Quinson
1770 - register init/exit functions within gras module mechanism
1771 - send/receive function.
1772 Convertion is not implemented, but short-cutted if not needed.
1773 struct/array elements are sent one by one (instead of block-wise), but
1774 nobody really cares (yet). Get a prototype before optimizing.
1775 - tests (using a file socket) for DD send/receive on:
1776 - base types: int, float
1777 - array: fixed size, string (ie ref to dynamic string)
1778 - structure: homogeneous, heterogeneous
1779 - chained list, graph with cycle
1780 Believe it or not, valgrind is not too unhappy with the results. The
1781 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1785 - Bugfix when using a filename explicitely (instead of '-')
1787 2004-04-09 Martin Quinson
1789 - factorize more code between RL and SG in socket creation
1790 - Complete the implementation and tests of:
1792 o file (only in RL, and mainly for debugging)
1794 I lost 3 days to design a portable address resolver, and then decided
1795 that the prototype mainly have to run on my box.
1796 Addressing portability too early may be like optimizing too early :-/
1798 - use gras_init in the Tests instead of the crappy parse_log_opt
1799 (the latter function is removed)
1800 [Conditional execution]
1801 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1802 [Code reorganisation]
1803 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1804 Build examples against the RL library, since there is no way to disable
1805 its creation for now.
1807 For information, the beginning of coding on GRAS was back in june
1808 2003. I guess that every line has been rewritten at least twice since