1 SimGrid (3.5) unstable; urgency=low
4 * New function: MSG_load_platform_script()
5 to make possible using a lua script instead of XML files to set up platforms
6 * New function: MSG_set_function
7 to associate functions to processes, used when bypassing the parser
9 * add layer to set up environment and bypass the XML surf parser
10 * modify the entry point, to take into consideration when lua is used only as a platform generator
12 * add new callbacks to the CPU & netwok models to create resources
13 no need to pass through the parser callbacks to do so.
14 * add new function to create routes table via the routing model
15 * add generic functions in the public interface
16 that allows the user to call SURF 'create_resource' methods from a lua script.
18 * New function: SIMIX_process_set_function()
19 called by MSG_set_function
21 * Change the way a task is considered as ready. Instead of removing
22 dependencies when a task is done, a counter is decreased. This way,
23 it is always possible to reach ancestors thanks to the
24 SD_taks_get_parents function (even after the end of the simulation.)
25 * Change the level of some logs in sd_global from INFO to VERB. The
26 corresponding tests are now less verbose
27 * Change the return type of SD_Simulate from (SD_task_t*) into
28 xbt_dynar_t. This function was in handling a dynar internally and
29 converted it into a NULL terminated array for historical reasons.
30 * Rename the SD_READY (all dependencies are satisfied and task is
31 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
32 dependencies are satisfied) state.
33 This prevents a confusion between the notion of "ready to schedule"
34 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
35 simulated" (SD_RUNNABLE) used by the simulation kernel.
36 * Add an example that scheduling a DAX on an heterogeneous platform
37 using a Min-Min strategy
38 * new function SD_dotload(char*) to load a DAG described in dot
39 format. This loader and the corresponding examples require the
40 installation of the graphviz library.
43 * Change the underlying waiting queue in semaphores so that a process
44 can wait on several of them simultaneously (as in waitany).
45 * Fix the way to handle tokens in semaphores so that all access patterns
46 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
49 * New data container: setset (set of sets of elements)
50 * New module: mmalloc (mapped malloc)
51 * New function: xbt_dict_cursor_set_data()
52 * New function: xbt_dynar_sort()
53 * New function: xbt_dynar_is_empty()
54 * New function: xbt_fifo_get_last_item()
55 * Bug fix in xbt_dynar_shrink(): use the right element size
56 * Use library init/fini functions for our initialization.
57 - you can use logs and other feature as soon as you want in your
58 code (even before the xbt_init / MSG_init)
59 - xbt_exit is now a no-op and produce a warning when used.
62 * New function: gras_msgtype_get_name()
64 Build chain: bug fixes and further polishing
65 * Fix the 'make install' target.
66 No need to use 'make install-simgrid' anymore
67 * Introduce a 'make dist' target compiling a *source* archive
68 'make package' compiles a binary archive (not well tested)
69 * Compile java files only on need
71 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
73 SimGrid (3.4.1) stable; urgency=low
75 The "Polishing easter eggs is probably a good idea" release.
76 This is a bug fixes release only.
80 * Fix a bug preventing the tasks from begin garbage collected.
83 * Fix a bug occuring when a host involved in a communication fails.
84 This was not detected properly by the other peer involved in the
85 communication. Now, it's reported as a network error.
88 * Warn the user about loop dependencies in data flow of DAX files
89 * Obey the control-flow dependencies of DAX files
92 * Add option "enable_smpi" allowing to not compile SMPI.
93 Probably useful for the (Mac) users experiencing a build error here
94 * Improve the detection of lua5.1 and ruby1.8
96 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
98 SimGrid (3.4) stable; urgency=low
100 The "Easter in Cargese" release. Also known as (major changes):
102 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
103 ~> bindings were greatly improved
104 ~> new tracing infrastructure for better visualization introduced
106 * the "Welcome to configury modernity" release.
107 ~> we switched from autotools to cmake, and improved our cdash
110 A more detailled list of changes follow (full detail in svn log).
112 Java Bindings: Various Cleanups
113 * (install java-gcj-compat-dev on debian-like to use them)
114 * Remove put/get: no need to export deprecated interface in Java
115 Use send/receive instead.
116 * Cleanup the examples and add a README per directory
117 * Remove example autoDestination (that's the only way to go now)
118 * Remove example explicitDestination (was a plain copy of basic)
119 * Make JniException a runtime exception, so that there is no need to
120 declare the fact that you may encounter such a beast. I guess that
121 nobody will ever want to survive such error.
122 * Create specific errors for each MSG case of failure:
123 host failure, transfer failure, timeout, task cancelled
124 * Cleanup the exceptions that may get thrown by each function
125 * Other internal cleanups in Java bindings. Performance still bad :/
126 Ruby and Lua Bindings: create them
127 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
128 * That's new and great, you should try them out.
129 Same functionalities than Java bindings, only even less polished
131 * Kill the useless "rate" argument of SD_task_get_execution_time()
132 Everyone used to provide -1 as a value, it was not used, and the
133 semantic of a possible use wasn't even clear.
134 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
135 either as comm costs or compute costs to mean that there is no
136 such thing for that specific task.
137 * Add a SD_task_set_name() function
138 * Fix SD_task_unschedule() on typed tasks
139 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
140 * In DAX loader, accept useless 'level' attributes to <job> since
141 LIGO DAGs have them (seem to be to ease graphical representation).
143 * Add an example masterslave_mailbox.c using send/receive and not
144 the deprecated put/get interface.
145 * Kill the MSG_paje_output() function. It's a noop since 2 years.
146 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
148 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
149 (declare MSG_USE_DEPRECATED to still have the old name)
150 * Add a MSG_task_set_data() function
151 * About trace replay (see examples/msg/actions):
153 - Allow to work with splitted trace files for each process
154 Give the specific trace file as argument of each process,
155 and call MSG_action_trace_run(NULL)
156 You can still have one merged file for all processes.
157 - Fix implementation of collective operations
158 * Allow task_execute() on 0-sized tasks (closes #10063)
160 * This is the first release of SimGrid where SMPI is not considered
161 beta anymore (even if some corners should still be improved)
162 * Port over the new SIMIX_network submodule (internal refactoring)
163 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
164 * Implement more missing elements of the standard:
166 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
168 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
169 MPI_Scan() MPI_Get_processor_name()
170 - Added implementation of missing case for Alltoall (warning: it's
171 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
172 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
173 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
174 SMPI_MPI_Allgatherv()
176 - MPI_Waitsome() was broken
177 - Allow relative includes in smpicc
178 - Command line cfg argument 'reference_speed' was ignored...
179 - Some functions did not properly lead to auto-benching of user code
180 - smpicc passes -O2 by default (just like openmpi one)
182 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
183 * Bug fixes about timeouts during communications
184 * add SIMIX_message_sizes_output() as a pimple to write to file the
185 amount of messages per size. Use gnuplot to get histogram.
186 Pimple because that's the only user-visible function of simix,
187 defined directly in xbt.h (irk, sorry)
189 - Add a SIMIX_sem_get_capacity() function
190 - Fix interactions with processe resume/suspende
191 - release_forever() was stupidly broken
192 - Fix SIMIX_display_process_status() for processes in a semaphore
193 - Make SIMIX_sem_block_onto() user-visible
194 * Refactoring context stuff:
195 - Use pseudo-OOP for better modularity
196 - reimplement SIMIX_process_kill() without process_schedule() so
197 that the latter can take as invariant that it is called from
199 - Merge context_start into context_new for sake of simplicity
201 * Add a Vivaldi network model, coded live during SUD'10 ;)
202 * Rename configuration variables to start a hierarchy:
203 o cpu_model -> cpu/model
204 o network_model -> network/model
205 o workstation_model -> workstation/model
206 * New configuration variables:
207 o network/bandwidth_factor: correction to bandwith
208 o network/latency_factor: correction to latency
209 o netwotk/weight_S: correction to the weight of competing streams
210 * Add a long description to the models, that users can see with such
211 argument on the command line: --cfg=cpu/model:help
212 * --help-models display the long description of all known models
214 * config: add the ability to set a default value after registration
215 Does not override any previously set value (e.g. from cmd line)
216 * dict: allow to have integer key and data.
217 When so, you need to use the following functions
218 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
219 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
220 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
221 In contrary to regular dicts, the key is not malloced before copy.
222 Mixing scalar and regular elements in the same dict is not tested
224 * Allow to use xbt_dynar_shrink() to expend the dynar instead
225 Tracing for Visualization:
226 * SimGrid is now instrumented in order to generate a trace file for
227 visualization analysis: to use it, need to compile SimGrid with the
228 "tracing" option enabled, and instrument the program using SimGrid with
229 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
230 (among other functions).
231 * The instrumentation only traces the platform utilization for now
232 * Documentation to use the tracing functions and how to analyze the
233 traces with the Triva tool is written.
234 * More information about: SimGrid FAQ (in the section Tracing Simulations
237 * We moved to cmake as default build system. Autotools support will
238 be dropped soon. Check the FAQ for more info about how to use it.
239 * Greatly improved our cdash/ctest interactions
240 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
241 * Added memory checking tests with valgrind; lot of memleak fixing.
242 This may be the first release of simgrid with so few memory issues
243 * Added code coverage tests.
244 Our coverage is still improvable, but at least we see it on cdash.
246 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
248 SimGrid (3.3.4) stable; urgency=low
250 The "Desktop Grid needs love too" release (also called Xmas release).
253 * Major speedup in the maxmin system solving by using lazy evaluation
254 Instead of solving completely the maxmin system at each iteration,
255 only invalidate (and recompute) the modified parts.
256 This new feature is enabled in default models but you can try to
257 turn it on with "--cfg:maxmin-selective-update=1" for other models.
258 * Cas01 IMproved as default CPU model
259 This CPU model is the same Cas01 model, but it uses the
260 maxmin-selective-update flag and a heap structure to manage
261 actions on SURF kernel.
262 It reduces the complexity to find the next action to finish and,
263 consequently, it's faster than the old Cas01.
264 This is the new default CPU model (Cas01).
265 * Rename the old Cas01 model to Cas01_fullupdate
266 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
267 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
268 * CpuTI (CPU Trace Integration)
269 A new CPU model whose objective is simulate faster when using
270 availability trace files.
271 Instead of using a full featured, over engineered maxmin system for
272 CPU modeling, this model does the pre-integration of traces files
273 to calculate the amount of CPU power available, and so, executes
274 faster than the old CPU models.
275 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
276 * Use LV08 as default network model since it gives better accuracy
277 for small messages and shouldn't change things for big ones.
278 Use --cfg=network_model:CM02 to get the previous behavior.
281 ******************************************
282 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
283 ******************************************
284 * The new CPU model may changes simulations!
285 The point is that events occurring at the exact same timestamp
286 are not scheduled in the same order with the old and new
287 version. This may be enough to completely change the execution
288 of simulations in some cases.
289 * The new network model will change simulations!
290 This new model is more realistic than the previous one, so you
291 should consider redoing your old experiments with this model.
292 Sorry for the inconvenience.
295 * Introduce the supernovae compilation mode
296 When compiled that way, the whole SimGrid (or almost) is put in a
297 single compilation unit and compiled in one shoot.
298 This is to help gcc which has difficulties to inline stuff from one
300 The speedup seem to be above 15%, althrough more tests are needed on
301 amd64 to confirm that gain.
304 * Port of MSG's mailbox on top of SIMIX network
305 The put/get mechanism was greatly simplified on the way.
308 * New SIMIX network module. Provides:
309 - Mailbox: rendez-vous mecanism to find with who you want to speak
310 - Synchronous send/recv: easier and hopefully faster since the
311 logic is handled in the maestro process directly now
312 - Asynchronous send/recv: you dreamt of it? It's here now
313 Too bad that nobody cared enough to propagate the change to MSG.
314 * Add semaphores as SIMIX synchronization mechanism.
317 * new function SD_daxload(char*) to load a DAX file
318 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
319 * Introduce typed tasks. Specify its kind and cost at creation.
320 At scheduling, just give where it should be placed, and the cost
321 for each involved resource is automatically computed.
322 Existing constructors so far (more to come of course):
323 - SD_task_create_comm_e2e() for end-to-end communication
324 - SD_task_create_comp_seq() for sequential computation
325 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
326 * new function SD_task_dump() for debuging display
327 * new function SD_task_dotty(task,FILE*) writing to file the info
328 about the task in dotty format
329 * SD_task_dependency_exists() can now cope with having one of its
330 arguments NULL. If so, it tests whether the other argument has any
332 * Add getters on list of preceding/following tasks:
333 SD_task_get_parents(task) and SD_task_get_children(task)
334 * Add getters on amount of workstations and list:
335 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
336 * Add getter on task kind: SD_task_get_kind(task)
337 * Update the start_time and finish_time of tasks on completion/failure
338 * Bugfix: Remove task from state swags when destroyed
341 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
344 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
345 dynar with one separate thread per value of the dynar.
346 * Change the prototype of xbt_thread_create(), sorry.
347 Added a boolean parameter indicating whether we want to join this
348 thread (used in SG only for now)
349 * Implement xbt_thread_join and xbt_thread_yield in SG also.
352 * GTNetS wrappers should now be usable again (and betterly tested too)
353 * Fix a major regression from 3.2 where the timeout provided to
354 MSG_task_put_with_timeout() was used as absolute time before which
355 the comm should be done.
356 * Start to fix the <cluster> tag.
357 - Internal links should be good now (beside of the loopback, which
358 use the private link instead)
359 - paths to the external world is still rather broken
360 - the <route:multi> tag is just broken. Actually that's brain-dead.
361 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
362 to make it less stupid
363 ** Check your platform with teshsuite/simdag/platforms/flatifier **
364 * Fix a source-level compatibility glitch from 3.2: after defining
365 MSG_USE_DEPRECATED, you can use the old name
366 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
367 * Allow to compile from the SVN with automake 1.11
368 * Fix some problems when using the "start_time" tag in deployment XMLs.
369 * Fix #8569: XBT/synchro.h has redundant declarations
370 * Fix #8563: MSG return values and exceptions
371 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
372 * Integrate patch #8636: Obey DESTDIR when installing documentation.
373 Thanks to Robson Peixoto.
374 * Fix a vicious bug in dictionaries inducing that some elements were
375 not freed on xbt_dict_free()
377 Portability report of this version:
378 * Main portability targets:
379 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
381 - mac leopard on i386
382 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
383 but nothing critical.
384 * Other platforms: windows, AIX and others were not tested for this release
386 Timing report of this version:
387 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
388 scenario parameters). From 8h to a few seconds in desktop grid settings.
389 * Supernovae brings about 25% speedup on i386.
391 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
393 SimGrid (3.3.3) stable; urgency=low
395 The "Need for Speed" release.
397 The timings done to validate the 3.3.2 were faulty.
398 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
400 The problem was a conversion from a manually handled vector to
401 xbt_dynar_t on the critical path.
402 xbt_dynar_foreach calls functions, inducing stack management crap.
404 We inlined these functions and xbt_dynar_foreach is now breath taking.
405 We also inlined xbt_swag_belong on the way.
407 Here are some approximate speedup measurements (on master/slaves
408 simulations lasting between 10s and 20s each):
409 3.3.1 -> 3.3.2: about same performance
410 3.3.2 -> 3.3.3: 40% speedup
411 3.3.1 -> 3.3.3: 40% speedup
412 3.3.1 with inline patch -> 3.3.3: 30% speedup
414 Our reading is that the refactoring which occurred in 3.3.2 made us
415 suffer much more from the xbt_dynar_foreach low performance, but
416 once we solved this, this refactoring proved to be very performance
417 effective. From the 40% speedup, somehow, 10% are due to the
418 inlining and 30% to the refactoring.
420 That's a pitty that gcc cannot inline functions placed in other files
421 alone. We have to choose between:
422 - break the encapsulation (by putting private data structures and
423 accessors in headers files to help gcc)
424 - live with low performance
425 - switch to a decent compiler such as icc (not quite possible).
427 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
429 SimGrid (3.3.2) stable; urgency=low
431 The "Simplicity does not preceed complexity, but follows it" release.
433 The main contributors of this release were (lexical order):
434 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
437 * Extract the routing logic into its own object.
438 (was dupplicated in network.c and workstation_LV07.c;
439 Allows to implement other ways of storing that info)
440 => kill now useless network_card concept
441 - Use dynar to represent routes (instead of void** + int*)
442 - kill link_set (use surf_network_model->resource_set instead)
443 - Add a command-line option to choose the routing schema to use
444 - Add three new models:
445 * Floyd (shortest path computed at initialization)
446 * Dijikstra (shortest path recomputed all the time)
447 * Cached Dijikstra (shortest path computed on need)
448 All these models where contributed by Silas De Munck, and are
449 described in his ICCS09 paper.
451 * Simplify model declaration
452 (less redirections, less function to write when defining a model)
453 - Factorize stuff between models:
456 surf_model_resource_set(model)
457 surf_model_resource_by_name(model, name)
458 - Unify the types of models in s_surf_model_t (using an union)
459 - Embeed fields of common_public directly into s_surf_model_t
460 - Rename model methods:
461 action_free ~> action_unref
462 action_change_state ~> action_state_set
463 action_get_state ~> action_state_get
464 - Change model methods into functions :
465 (model)->common_public->action_use ~> surf_action_ref
467 * Implement a generic resource; use it as ancestor to specific ones
468 (allows to kill duplicated code in models)
469 Drawback: timer command don't need no name nor properties;
470 workstation_CLM03 don't need no properties
471 (but I guess we can live with those few bytes wasted)
473 * Improve the action object model
474 - implement a constructor avoiding dupplicated code about field
475 initialization in generic_action part.
477 * Kill the SDP model: it has an external dependency, is deprecated
478 in flavor of modern lmm models, and didn't compile since a while
481 * Relocation of the context module from XBT to SIMIX.
482 (the context were decoupled from the simix processes, duplicating a lot of code)
483 => a lot of code was factorized
484 - less overhead is introduced during scheduling
485 - simpler API for the context factory
486 - the logic for process creation,destruction and manipulation was simplified
487 * Simplification of the s_smx_process_t data structure.
488 => accesing the simix level data associated to a process is faster now,
489 and the code is a lot more readable.
492 * Implement some more MPI primitives:
493 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
494 -implementation: Bcast: flat or 2-ary tree (default),
497 Allreduce: Reduce then Bcast
498 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
499 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
500 Alltoallv: flat tree, like ompi
502 * Add support for optimized collectives (Bcast is now binomial by default)
503 * Port smpirun and smpicc to OS X
506 * Kill SD_link_get_properties: hard to maintain and makes very little sense
507 Shout out if you used it.
510 * Display the list of still queued messages in SG mode when existing
514 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
515 * Add xbt_graph_node_get_outedges() [Silas De Munck]
516 * Add xbt_str_from_file(FILE*)
517 * Add xbt_dict_get_key achieving a linear reverse search
518 * Remove the context module
520 Portability report of this version:
521 * Main portability targets:
522 - Linux(debian)/x86/context
523 - Linux(debian)/x86/pthreads
524 - Linux(debian)/amd64/context
525 - Linux(debian)/amd64/pthreads
526 On these, we still have the eratic breakages of gras/pmm and
527 amok/saturate_sg reported in previous version. We still think
528 that the tests are the cause of the fault, not the tested code.
530 - Mac OSX Leopard/x86/context
531 Still false negative in tesh autotesting.
532 Smpi still fails, but this time because readlink does not accept -f
533 Everything seems to work properly beside of that.
536 - AIX version 5.3 (only tested contexts this time)
537 Smpi still fails there because mktemp is not installed.
538 Everything seems to work properly beside of that.
540 I managed to compile it for the first time, but several breakages.
541 Won't delay the release for this exotic platform.
543 * Windows: it's still lagging behind. If you want to help, please
546 Timing report of this version:
547 This version seem to be more than 5% faster than 3.3.1 (on linux
548 64bits with contextes). The gain is less than expected, we are
549 investigating this for next release.
551 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
553 SimGrid (3.3.1) stable; urgency=low
556 * Implement a --cfg-help to show existing configuration variables
557 * Build chain do not require doxygen in maintainer mode
560 * fix a bug on struct sizeof computation, which prevented the
561 exchange of arrays of structs in some conditions
562 - added a regression test about this in datadesc_usage
563 * Allow the exchange of 0-long dynamic vectors.
564 - for that, use -1 as indicator of dynamic size instead of 0
565 - This implied to change any size from unsigned long to long,
566 reducing a bit communication abilities, but I guess that with
567 64bits being quite common, this is more than enough.
568 - This also induce a protocol change, thus bumping network protocol
569 version from 0 to 1 (if we have external users, we have to get
570 clean on that point too ;)
571 - added two regression tests about this in datadesc_usage
572 * Be more verbose when propagating local exceptions
573 This helps debugging.
574 * Display the status of simulated processes when receiving SIGINT in
578 * Allow to control the simulation from a trace file.
579 New functions MSG_action_register() and MSG_action_trace_run()
580 The first one allows to associate a function execution to each
581 kind of action while the second one parses a trace file and
582 triggers the corresponding actions within the system.
583 For now, only a toy example is provided in examples/msg/actions
584 * Add an exemple of process migration in examples/msg/migration
585 * Fix a bug in task exchange which broke MSG_task_get_sender()
586 Add a teshsuite regression test for that.
587 [Bug: if MSG_task_get_sender() is called after sender exit,
589 * Fix a bug which prevented suspend/resume to work properly
590 * Display the status of simulated processes when receiving SIGINT
591 This fixes a regression of v3.3. due to the introduction of SIMIX
592 * Bug fixing in failure management:
593 - trace could not start by a failure at time 0
594 - failure during communications were not working
597 * Add SIMIX_process_set_name() to change the name of the current
598 process in the log messages.
599 * Store smx_hosts in a dict since we only retrieve them by name
600 * Move the configuration infrastructure to surf
603 * Move the configuration infrastructure to surf
606 * Massive internal cleanups:
607 - Store internal structures on processes instead of hosts (allows
608 to have more than one process per host, in addition of being more
610 - Cleanup the initialization/finalization process
611 - Kill a whole bunch of unneeded synchronization:
612 processes run in exclusive manner within the simulator
613 - Move queues from global tables to process data fields
615 - now accept -platform and -hostfile arguments
616 - Pass the right rank value to processes according to the hostfile
617 * Compile the examples by default, and use them as regression tests
618 * Implement MPI_Wtime()
619 * Change the reference speed to a command line option
622 * TCP_gamma can now be specified as command line option using
623 --cfg=TCP_gamma:10000000.0
624 * Change the --surf-path cmd line option into --cfg=path:
627 * Also include strbuff from xbt.h public header
628 * xbt_ex_display(): do not free the exception after displaying
629 This allows to do more with the given exception afterward.
630 Users should call xbt_ex_free() themselves.
634 Portability report of this version:
635 * Main portability targets:
636 - Linux(debian)/x86/context
637 - Linux(debian)/x86/pthreads
638 - Linux(debian)/amd64/context
639 - Linux(debian)/amd64/pthreads
640 These targets fail about 1/10 of times on gras/pmm, but we believe
641 that this is because of the test, not because of simgrid.
642 amok/saturate_sg fails even more rarely, and the test may not be
645 - Mac OSX Leopard/x86/context
646 The test suite still spits tons of errors because some obscure
647 force prevents us from removing the temporary directories
648 arguing that they still contain some metadata I've never heard of.
649 Smpi fails because seq is not installed.
650 Everything seems to work properly beside of that.
653 - AIX version 5.3 (both contexts and pthread)
654 Smpi still fails there because mktemp is not installed.
655 XML inclusions seems rosty on AIX.
657 * Windows: it's still lagging behind. If you want to help, please
660 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
662 SimGrid (3.3) stable; urgency=high
666 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
669 * Introduce the SIMIX module: factorize code between MSG and GRAS.
672 Until now, GRAS were using MSG as an interface to SURF. It was
673 quite difficult because both interface have several differences
674 (MSG channels vs GRAS sockets were the most notable point).
676 This also opens the gate to SMPI (which should occur soon) and speed
677 up simulations by to 40% (even if it were not the main goal).
679 **************************************
680 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
681 **************************************
682 The point is that events occuring at the exact same timestamp are
683 not scheduled in the same order with the old and new version. This
684 may be enough to completely change the execution of simulations in
685 some cases. Sorry for the inconvenience.
687 * Cleanup and upgrade the XML format to push further scalability
688 issues (check http://hal.inria.fr/inria-00256883/ for more info)
690 * Improve the testing infrastructure with tesh. Now a very large part of
691 the code is tested not only by being run but also by checking that the
692 output match an expected output [Mt].
694 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
695 is really less memory-demanding, which should allow you to use
696 larger files in SimGrid [AL].
698 * Inform valgrind about our contextes, so that it becomes usable
699 with the default (and more effecient) version of SimGrid
700 [contributed by Sékou Diakite, many thanks]
703 * Introduce a listener thread in charge of receiving incomming
704 messages from the network. It allows to overlap communication and
705 computation but most notably, it removes some stupid deadlocks due
706 to the fact that so far, a process could not send and receive at
707 the same time. This made most non trivial communication schema
709 * Convert the PIDs from long int to int to match the MSG ones (and
711 * New function: gras_agent_spawn() to launch a new process on
712 current host. Only working in simulation for now. [Mt]
713 * New function: gras_os_hostport() returning a constant form (ie,
714 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
717 * Make the backtrace of exceptions more human readable [Mt]
718 * New module: xbt/str [Mt]
719 a ton of string utility functions (split, join, printf to a newly
720 allocated buffer, trim, etc)
721 * New module: xbt/hash [Mt]
722 SHA1 hashing algorithm (more to come if needed)
723 * New module: xbt/synchro [Mt]
724 synchronization tools (mutex and conditions) working the same way
725 in simulation and in real life (mainly useful for GRAS, but not
727 * New module: xbt/queue [Mt]
728 classical producer/consumer synchronization scheme
729 * xbt_dynar_new_sync() creates a synchronized dynar. All access
730 (using the classical functions will get serialized) [Mt]
731 * Make dictionary internal table dynamic. No need to specify its size
732 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
734 * Make sure the log channels are organized as a tree under windows
735 (because of ANSI C compatibility issue, any channel were child of
739 * Cleaned many thing in surf and fixed a few bugs [AL].
740 * Add a nice command line configuration mechanism to compose models [AL].
741 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
742 the previous one (KCCFLN05). It relies on something that looks like
743 a max-min sharing mechanism but cannot be written as such. A new solver
744 was thus designed [AL].
745 * Add a new solver to lmm. Based on Lagrange optimization and
746 gradient-based descent, it enables to efficiently maximise systems s.a
748 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
750 This solver enables to propose two new network models for TCP Reno and
751 TCP Vegas based on Low's work. These models still need to be fully
752 tested though [Pedro Velho].
755 * Bug fix in SD_simulate. Now the time bound given as argument is
757 * Use the new parallel task model (ptask_L07) as default.
758 * Use the SURF command line configuration mechanism.
759 * 0-size tasks (for synchronization) should now work.
761 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
763 SimGrid (3.2) stable; urgency=high
767 We still experience issues on this platform, but we believe that at
770 GRAS API BREAKAGE (for simplification purpose, sorry):
771 * the gras_msgtype_by_name is not used anymore. Instead of
772 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
773 you can write (and must)
774 gras_msg_send(toserver, "request", &request);
775 - If you still want to pass a gras_msgtype_t to the function (to cache
776 the type and avoid the lookup time), use the gras_msg_send_() variant.
777 - Impacted functions:
778 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
779 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
780 * The callbacks are now expected to return 0 when everything went well
781 (just like the main() function)
783 GRAS new features and improvements:
784 * New module mecanism where user code can use per process globals [Mt]
785 This is similar to gras_userdata_*() functions, but for libraries. It
786 factorize some code developped over and over in the examples and AMOK.
787 It has still to be documented and used (only amok/peermanagement is
789 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
790 loss when they were small enough to fit into the buffer and sent quickly
791 enough so that they can all get received in one shoot.
792 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
793 if not found. Use the *_or_null() variant for the old semantic.
794 * In gras_msg_handle, do not discard messages without callback.
795 They are probably messages to be explicitly awaited later (ie, proofs of
796 mis-synchronization in userland since they are sent before being awaited)
798 * gras_socket_meas_send/recv: semantic changed!
799 The numerical arguments used to be (1) the total amount of data to send
800 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
801 messages. This was need for the fool willing to send more than MAXINT
802 bytes on quite fat pipes.
805 * Do really rename the hostmanagement module to peermanagement. [Mt]
806 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
807 but this is rather new and this was documented in the module
808 documentation (poor excuses, I admit)
809 * Bandwidth measurement semantic changed! This follows the changes to
810 gras_socket_meas_send/recv explained above.
813 * A sequential mode has been added to the workstations. When a workstation
814 is in sequential mode, it can execute only one task, and the other tasks
815 are waiting in a FIFO. [Christophe Thiery]
818 * The KCCFLN05 workstation model now handles parallel tasks. It is the
819 model for SIMDAG. [Christophe Thiery]
820 * Bug fix in the maxmin solver: Some values were close to 0 instead of
821 equal to 0, which caused some bad behaviors in
822 saturated_constraint_set_update. I now use a threshold mechanism like in
826 * When running manually src/testall, you select specific units [Mt]
827 testall is the result of our cunit mecanism, and should replace all
828 the scripty thingy around since bash don't run easily on billware.
830 * A mallocator system has been added. [Christophe Thiery]
831 Mallocators allow you to recycle your unused objects instead of freeing them
832 and allocating new ones.
834 Documentation update:
835 * FAQ reworking + New FAQs:
836 - "Valgrind spits tons of errors!" [Mt]
837 - "How to repport bugs" [Mt]
838 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
839 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
840 - Communication time measurement within MSG [AL]
841 - I experience weird communication times when I change the latency [AL]
844 - an introduction to the framework and to the used communication model
845 - an initiatic tour introducing the most proheminent features:
847 . Lesson 0: Installing GRAS
848 . Lesson 1: Setting up your own project
849 o Part 2: Message passing
850 . Lesson 2: Exchanging simple messages
851 . Lesson 3: Passing arguments to the processes (in SG)
852 . Lesson 4: Attaching callbacks to messages
853 . Lesson 5: Using globals in processes
854 . Lesson 6: Logging informations properly
855 . Lesson 7: Using internal timers
856 . Lesson 8: Handling errors through exceptions
857 . Lesson 9: Exchanging simple data
858 . Lesson 10: Remote Procedure Calling (RPC)
859 . Lesson 11: Explicitely waiting for messages
860 . Recapping of message passing features in GRAS
861 - A HOWTO section containing:
862 o HOWTO design a GRAS application
863 More are due, of course. They will come latter. In the meanwhile, you can
864 check the examples which are still here.
866 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
868 SimGrid (3.1) stable; urgency=high
872 There was a stack corruption somewhere, visible only when optimizing
873 with these versions. [Vince]
876 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
877 module that enabled to play with Directed Acyclic Graphs. It is built
878 directly on top of SURF and provides an API rather close to the old
879 SG. Some old codes using SG are currently under rewrite to check that
880 all needful functions are provided. [Christophe Thiery]
883 * Complete rewrite of the KCCFLN05 workstation model. It is now an
884 extension of the classical CLM03 model that gracefully handles
885 failures. This is now the default model for MSG and GRAS. It doesn't
886 handle parallel tasks yet though. [AL]
887 * Bug fix: Weights were not correctly set in the network part.
888 WARNING: This may have resulted in incorrect results with simulations
889 where there are more than one flow on a given link. [AL]
892 * After a (long ?) discussion on simgrid-devel, we have decided that the
893 convention we had on units was stupid. That is why it has been decided
894 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
895 WARNING : This means that all previous platform files will not work as
896 such with this version! A warning is issued to ask users to update
898 A conversion script can be found in the contrib module of the CVS, under
899 the name contrib/platform_generation/surfxml_update.pl [MQ]
902 * Bug fix: Processes were started in reverse order, wrt deployment file.
903 WARNING: if your code relies on this bug, please fix it. [AL]
904 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
905 being executed on two different locations. [AL]
906 * Bug fix: Failures are now better supported thanks to Derrick's tests
907 (there was many failure situations I hadn't thought of and that weren't
908 correctly handled). [AL]
909 * New function: MSG_host_is_avail indicates you whether a given m_host_t
913 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
914 Exception occurring on server-side are propagated back to client (!).
916 API CHANGE: the callback changed their prototype. Change:
917 int my_handler(gras_socket_t expeditor, void *payload_data) {
919 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
920 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
922 * New! function: gras_msg_handleall to deal with all messages arriving
923 within a given period.
924 * New! function: gras_socket_server_range to get a server socket in a
925 range of port numbers (ease to avoid port number conflicts) [MQ]
926 * New! gras processes display their backtrace when they get a SIGUSR1
927 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
928 Sweet to debug RL processes [MQ]
932 - Do not force experiment sizes to be expressed in kb, or it becomes
933 impossible to measure the latency this way (needs one byte-long tests)
934 WARNING: this changes the amok_bw_* function semantic. [MQ]
935 - Implements the link saturation stuff. [MQ]
936 * Peer management module:
937 New! module factorizing code that we wrote over and over [MQ].
940 * New module: cunit (my jUnit implementation in ansi C) [MQ]
941 - Test units are placed directly into the library code, they get extracted
942 automatically and placed into the src/testall binary.
943 - Convert most of the XBT tests to this system.
944 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
945 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
946 live in the same C code. This required to change a little bit the API
947 of surfxml parsing but shouldn't be an issue for end-users. [AL]
948 * New module: sparse graph structure with basic algorithms (this is work
949 in progress and the API is not considered to be frozen yet). [AL]
950 * Display more information on backtraces: source line & function names are
951 now displayed just like valgrind does (rely on addr2line tool) [MQ]
952 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
953 * Reworked a little bit some #include statements to load only required
954 headers. Some user code that relied on SimGrid to include stdlib or
955 stdio may need to include it by themselves. [AL]
956 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
958 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
959 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
960 but I guess that almost nobody used those parts. [MQ]
962 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
964 SimGrid (3.0.1) stable; urgency=low
967 * Unfortunately, I had missed 5 misnamed functions:
968 xbt_fifo_item_t xbt_fifo_newitem(void);
969 void xbt_fifo_freeitem(xbt_fifo_item_t);
970 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
971 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
972 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
973 They're now deprecated. Please use their new versions:
974 xbt_fifo_item_t xbt_fifo_new_item(void);
975 void xbt_fifo_free_item(xbt_fifo_item_t);
976 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
977 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
978 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
980 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
981 * Documentation: xbt_log module unmercifully reworked [MQ]
982 * Bugfix: there was a problem with the ending of contexts with
983 the pthread backend. It caused some weird deadlock or behavior
984 depending on the pthread implementation. [AL]
985 * Bugfix: get the exceptions raised in the simulator repport where
986 and why they come from when they are not catched in time [AL, MQ]
989 * Bugfix: Do repport the error when two non-connected hosts try to
990 exchange data (Thanks to Flavien for stumbling into this one) [AL]
993 * Add additionnal checkings on communications. Assert that two
994 communicating hosts are connected by a set of links... [AL]
997 * Add additionnal checkings on channel values in communication [AL]
998 * New: MSG_task_get_source to see on which host a task was generated [HC]
999 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1000 the number of tasks waiting to be received on channel and sent
1002 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1003 waits for the first task coming from a given host.. [AL]
1005 GRAS new functionnalities: [MQ]
1006 * Enhance the parsing macro to allow the size of multidimentional objects
1007 to be given thru annotations.
1008 * New example (and documentation): Matrix Multiplication a la RPC
1009 (as when I was young!) and fix a bunch of bugs found on the way.
1011 GRAS performance improvements: [MQ]
1013 * Reduce the amount of cbps creation/destruction by making it static to
1014 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1017 * Change libdata to a set so that we can search for stuff by ID (and thus
1018 reduce the insane amount of dict lookups)
1021 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1022 RL communication on top of SG since it's so uneffective.
1023 It may also be used for inter-thread communication in RL, one day.
1024 * Use gras_datadesc_copy() to exchange messages on top of SG
1026 - improve message exchange performance on top of SG
1027 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1028 * Don't exchange on the network the size of the used part of buffer,
1029 instead, specify the possible buffer size to read().
1031 - reduces the amount of read/write calls (one pair per exchange)
1032 - reduces the amount of exchanged data (the size)
1033 - allows to retrieve all arrived data on receiver side, if we don't need
1034 it right now (subsequent read will peek the buffer)
1035 - allows the receiver to proceed with the begining of the stream before
1036 everything is arrived
1037 - make it possible to build an iov transport (using readv/writev)
1039 - take care of the data with non-stable storage (like stacked data),
1041 * If possible, TCP send uses vector I/O (when writev() is here)
1042 - Don't use it for receive since we send data sizes and data on the
1043 same stream, so we wouldn't be able to chain large amount of chunks
1044 before having to flush the stuff to read the size.
1045 * Rework the transport plugin mecanism to simplify it and reduce the
1046 amount of pointer dereferencement when searching for the right function
1049 * I guess that now, we do almost as few system calls as possible while
1050 doing as few data copy as possible.
1052 To improve it further, we could try to send all the sizes first and then
1053 all the data (to use iov on receiving size), but it's only a partial
1054 solution: when you have 2 dimensional data, the sizes of the second
1055 dimension is data of the first dimension, so you need 3 streams.
1057 I'm not sure the potential performance gains justify the coding burden.
1059 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1061 SimGrid (3.00) stable; urgency=high
1064 * New! Give the possibility to hijack the surf parser and thus bypass
1065 MSG_create_environment and MSG_launch_application. Have a look at
1066 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1068 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1070 SimGrid (2.96) unstable; urgency=low
1075 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1076 This deprecates the xbt_error_t mecanisms.
1077 It modifies (simplifies) all XBT and GRAS API.
1078 MSG API keeps unchanged (exceptions raised by XBT are catched from
1079 within MSG and masked with existing error handling facilities)
1082 * New! Add a FATPIPE model. [AL]
1083 * New! Add a parallel task model. [AL]
1084 * New! Add automatically a loopback interface (in the default
1085 network model) if none was precised.
1088 * Bugfix: MSG_process_resume now works with the current running process.
1090 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1091 * Modification of MSG_task_get_compute_duration. Once a task has been
1092 processed, the value returned by this function is now equal to 0. [AL]
1093 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1094 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1095 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1096 of a m_task. For now, MSG_task_cancel only works with computation
1098 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1099 of the processor (in Mflop/s) regardless of the current load on the
1101 * API Change: use proper naming convention for MSG_getClock and
1102 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1104 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1105 This function changes the priority of a computation task. This priority
1106 doesn't affect the transfer rate. A priority of 2 will make a task
1107 receive two times more cpu power than the other ones. This function
1108 has been added to suit the needs of Nguyen The Loc and hasn't been that
1109 much tested yet. So if it fails, please report it and send me your code.
1111 * API Change: removed all functions and types that were marked "deprecated"
1112 since many months. Renamed MSG_global_init_args to MSG_global_init.
1114 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1116 SimGrid (2.95) unstable; urgency=low
1119 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1120 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1121 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1124 * Bugfix: Allow absolute paths for platform description files [MQ]
1125 * Bugfix: do free the variables after use. Leads to drastic performance
1127 * Implement max_duration (ie, timeouts) on resources [AL]
1130 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1132 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1133 the needed support (timeouts) [AL]
1136 * Implement measurement sockets. You can now get the bandwidth between two
1137 hosts thanks to AMOK (see below). [MQ]
1138 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1139 dynar over the network (yeah) [MQ]
1140 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1141 * simple/static token-ring example. [Alexandre Colucci and MQ]
1142 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1143 in place before [MQ]
1146 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1147 type parsing. "short" and "long" modifiers were also ignored; other
1148 modifier (such as reference level) are still ignored. [MQ]
1149 * Embeed the buffer size within the buffer itself on SG. [MQ]
1150 That way, send() are atomic and cannot get intermixed anymore (at least
1151 the ones which are less than 100k; bigger messages still have the issue)
1152 * Array size pushed by the field, not by the field type (or each
1153 and every long int will push stuff to the cbps) [MQ]
1154 * use select() to sleep since it allows to portably sleep less than one
1157 GRAS (minor cleanups)
1158 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1161 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1162 * type category 'ignored' killed as it was never used and were difficult
1164 * whether a type can cycle or not is now a flag, leaving room for more
1165 flags (as "ignored", if we feel the need one day ;)
1166 * Rename raw sockets to measurement sockets since "raw" has another
1167 meaning in networking community.
1170 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1171 and offers features not belonging to GRAS but that most applications
1172 need. One day, it may be a set of plugins loadable at runtime.
1173 * New module: bandwidth
1174 bandwidth measurement between arbitrary nodes running this module. [MQ]
1176 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1178 SimGrid (2.94) unstable; urgency=low
1180 The first beta release of SimGrid 3 !
1184 * Update the main page and the FAQ. Adding references to gforge.
1187 * Add a gras_os_getpid function.
1190 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1191 * Extend the logs so that they also print PID, hostname, date, ... if
1193 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1194 and kill the old version which were in testsuite/
1195 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1199 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1200 for users. Instead, each programming environment (GRAS, MSG,...) use it
1201 when needed to provide such a feature to users.
1202 Don't shortcut the mecanism or you will also shortcut the virtualization
1203 you need on the simulator.
1207 * Cleanups in configury with regard to compile optimization/warning flags.
1208 Also add -fno-loop-optimize to any powerpc since it's the optimization
1209 killing gcc (< 3.4.0).
1210 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1211 needed by MSG examples complications
1212 * Borrow configury beautifications from PHP
1215 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1216 warning (thanks loris for stumbling into this one).
1217 * Bugfix: stop loading private headers (gras_config.h) from the public
1221 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1222 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1223 RL would imply a DNS resolver, which is *hard* to do in a portable way
1224 (and therefore delayed).
1225 * Implement a real timer mecanism and use it in timing macros. This allows
1226 to avoid rounding errors and get a 0.000005 sec precision in timing
1227 macros. While I was at it, various cleanups:
1228 - allow to declare more than one timed section per file (fix a stupid bug)
1229 - move some private declaration to the right place
1230 - merge conditional execution and timing macros into emulation module
1231 - document the module
1232 - make sure the module cleanups its mess on gras_exit
1233 * Documentation improvements:
1234 - (new) how to compile applications using GRAS
1235 - (new) emulation support (timing macros)
1237 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1239 SimGrid (2.93) unstable; urgency=low
1241 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1244 - Use Paje properly where used. Still to be sanitized properly.
1245 - Portability fix: Add an implementation of the contexts using pthread
1249 - Add xbt_procname(): returns the name of the current process.
1250 Use it to show the current process's name in all logging.
1252 - fix detection of older flex version and the reaction, since we do
1253 depend on modern ones (we use lex_destroy)
1254 - Better separation of SG and RL in the libs: remove all simulation code
1255 from libgras. As a result, this lib is now only 200k when stripped.
1256 Some of the xbt modules may also be duplicated (two sets and such) and
1257 should be cleaned/killed before SG3.
1258 - Insist on using xlC on AIX because of weird problems involving gcc there.
1259 - Cleanup the make remote stuff. This is now done by scripts
1260 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1261 mainly for our private use, but we're working on changing them to user
1264 - Bugfix: flush the socket on close only if there is some *output*.
1265 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1267 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1270 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1272 SimGrid (2.92) unstable; urgency=low
1274 Alpha 3 on the path to SimGrid 3
1278 - New! First try of benchmarking macros.
1279 - New! First try so that gras_stub_generator generate deployment and
1280 remote compilation helpers.
1282 - Bugfix: Initialization fix in msg_test.
1286 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1288 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1290 - API Change: gras_os_sleep to take the amount of seconds as a double.
1291 Accepting an int was error prone since it was the only location where
1292 seconds were coded as such. It leaded to damn rounding errors.
1293 - Bugfix: Hard to belive that timers ever worked before this.
1295 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1297 SimGrid (2.91) unstable; urgency=low
1299 Alpha 2 on the path to SimGrid 3
1303 - Bug fix in the lmm_solver.
1305 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1306 through the function MSG_paje_output.
1307 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1308 - It is possible to bound the rate of a communication in MSG with
1309 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1310 to write it in the changelog).
1311 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1316 - Various cleanups to the autotools stuff
1317 - Begin to move Gras examples to examples/gras/
1318 - Let make distcheck work again (yeah!)
1320 - documentation overhauled using doxygen.
1321 gtk-doc-tools is dead in SimGrid now.
1322 - Automatically extract all existing logging categories, and add the list
1323 to the documentation (long standing one, to say the less)
1325 - Cleanup the known architecture table. Reorder the entries to group what
1326 should be, and use a more consistent naming scheme.
1327 (some of the test dataset are still to be regenerated)
1328 - New! Allow library to register globals on each process just as userdata
1330 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1331 lookup time (for now).
1332 Use it in msg and trp.
1333 This cleans a lot the internals and helps enforcing privacy of the
1334 headers between the gras components.
1335 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1336 - Bugfix: gras_os_time was delirious in RL.
1337 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1339 - Reenable GRAS now that it works.
1341 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1343 SimGrid (2.90) unstable; urgency=low
1345 Alpha 1 on the path to SimGrid 3
1347 * It is a long time since the last release of SimGrid. I'm sorry about
1348 that but as I had told you, I was rewriting a lot of things. I apologize
1349 to those who had been reporting bugs to me and that I had not answered.
1350 If your bug is still in the new version, please tell me. Here is a
1351 summary of the main changes.
1353 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1354 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1355 portability, speed, and a lot more but you'll figure it out later.
1356 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1357 I wanted to release the new MSG as soon as possible and I have
1358 broken GRAS, which is the reason why, for now, only MSG is fully
1359 functional. A laconic description of these projects is available
1360 in the documentation.
1362 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1363 kernel optimized for our needs (called SURF but only the developers
1364 should use it). Hence, MSG is now roughly 30 times faster and I think
1365 that by rewriting a little bit MSG, I could event speed it up a little
1366 bit more. Beside the gain in speed, it is also much easier to encode a
1367 new platform model with SURF than it was with SG. More to come...
1369 * REVOLUTION 3: I have tried to change a little as possible the API of
1370 MSG but a few things really had to disappear. The main differences
1371 with the previous version are :
1372 1) no more m_links_t and the corresponding functions. Platforms are
1373 directly read from a XML description and cannot be hard-coded
1374 anymore. The same format is used for application deployment
1375 description. The new format is described in the documentation.
1376 Have a look in tools/platform_generation. There is a tiny script
1377 that converts from the old platform format to the new one. Concerning
1378 the application deployment format, parsing the old one is tricky.
1379 I think most of you should however be able to convert your files. If
1380 it is really an issue, I can write a C code that does the conversion.
1382 2) the toolbox tbx does not exist anymore. We now have a library
1383 with much more data-structures but without the hash-tables (we have
1384 dictionaries that are much faster).
1386 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1388 *****************************************************************************
1389 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1390 * don't feel like converting the oldies. (Mt) *
1391 *****************************************************************************
1394 Version 2.90: "the long awaited one"
1395 - Finished rewriting and debugging MSG. Rewrote the documentation.
1396 - disable GRAS for now since it needs to be ported to the newest SG
1399 - Finish the port to windows (using mingw32 for cross-compile)
1402 - Main loop and datastructures of SURF. A cpu resource object is
1403 functional. Surf can thus be used to create cpu's with variable
1404 performance on which you can execute some actions.
1406 2004-11-15 Martin Quinson
1407 - Port to ARM. Simply added the alignment and size descriptions. Should
1408 work, but the ARM machines are so slow that I didn't had the opportunity
1409 to 'make check' over there yet.
1411 2004-11-15 Arnaud Legrand
1412 - Trace manager now written. It uses a heap structure and is therefore
1413 expected to be efficient. It may however be speeded up (particularly
1414 when many events occur at the same date) by using red and black
1415 trees. One day maybe...
1416 - Max-min linear system solver written. It uses a sparse matrix
1417 structure taking advantage of its expected use. Most operations are
1418 O(1) and free/calloc are called as few as possible. The computation of
1419 the minimum could however be improved by using a red and black tree
1422 2004-11-03 Arnaud Legrand
1423 - Rename every gras_* function that was in xbt/ to its xbt_
1425 - Add a heap and a doubly-linked list to xbt
1426 - Added a dichotomy to the dictionaries. make check works as well before
1427 so I assume that the patch is correct. I do not know however if things
1428 run effectively faster than before now. :)
1430 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1431 SimGrid, and the version number is bumped to 2.x
1433 2004-10-29 Martin Quinson
1434 - Introduction of the remote errors.
1435 They are the result of a RMI/RPC on the remote machine.
1436 ErrCodes being scalar values, you can't get the host on which those
1437 errors did happen. Extending the error mechanism as in Gnome is possible.
1438 No idea yet whether it is a good idea.
1440 2004-10-28 Martin Quinson
1441 - Interface revolution: the Starred Structure Eradication.
1442 I used to do typedef struct {} toto_t; and then handle *toto_t.
1443 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1444 - ???_t is a valid type (builded with typedef)
1445 - s_toto_t is a structure (access to fields with .)
1446 - s_toto is a structure needing 'struct' keyword to be used
1447 - e_toto_t is an enum
1448 - toto_t is an 'object' (struct*)
1450 typedef struct s_toto {} s_toto_t, *toto_t;
1451 typedef enum {} e_toto_t;
1452 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1453 s_toto_t) is private.
1455 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1456 so that it can set the variable to NULL. It was so for dicts and sets,
1457 it changed for dynars.
1459 - Fix a bunch of memleaks in dict_remove
1460 - Fix a bug in sg/server_socket opening: it failed all the time.
1462 2004-10-07 Martin Quinson
1463 - Speed up dynar lookup operation a bit.
1465 gras_dynar_get is dead.
1467 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1468 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1469 (which gives you the address of the stored data).
1471 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1472 copy of the data using an affectation to do the job and not a memcpy.
1474 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1475 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1477 It does not really speedup the dynar test because they are
1478 setting elements all the time (and look them seldom). But the dict does
1479 far more lookup than setting.
1481 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1483 2004-10-05 Martin Quinson
1484 - Allow to (en/dis)able the cycle detection at run time.
1486 Whether we should check for cycle or not is now a property of each
1487 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1488 datadesc_cycle_unset allow to remove this property when previously set.
1490 Note that the cycle detection is off by default since it impacts the
1491 performance. Watch the data you feed GRAS with ;)
1493 This property is hereditary. Any element embedded in a structure having it
1494 set have it set for the time of this data exchange.
1496 You should set it both on sender and receiver side. If you don't set it on
1497 sender side, it will enter an endless loop. If you forget on receiver
1498 side, the cycles won't be recreated after communication.
1500 - Header reorganization.
1501 Kill gras_private.h, each submodule must load the headers it needs.
1503 2004-10-04 Martin Quinson
1504 - Interface revolution: do not try to survive to malloc failure.
1506 Now, gras_malloc and friends call gras_abort() on failure.
1507 As a conclusion, malloc_error is not a valid error anymore, and all
1508 functions for which it was the only gras_error_t return value are
1509 changed. They now return void, or there result directly.
1510 This simplify the API a lot.
1512 2004-09-29 Martin Quinson
1513 - Re-enable raw sockets.
1514 Created by gras_socket_{client,server}_ext;
1515 Used with gras_raw_{send,recv}
1518 It should allow to kill the last bits of gras first version soon.
1520 This is not completely satisfactory yet (duplicate code with
1521 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1524 - Simplify transport plugin (internal) interface by not passing any
1525 argument to _server and _client, but embedding them in the socket
1528 2004-09-28 Martin Quinson
1529 - Finish the port to AIX.
1530 autoconf was my problem (segfault within the malloc replacement
1531 function. No idea why)
1533 2004-09-16 Martin Quinson
1534 - Fix some size_t madness on 64bit architectures.
1536 2004-09-08 Martin Quinson
1537 - Reduce the number of system headers loaded, overload some more system
1538 calls (such as malloc to cast the result of the system one, and work
1540 - Fix and reintroduce the config support
1542 2004-09-07 Martin Quinson
1543 - Source code reorganization to allow Arnaud to surf all over there.
1544 - Allow to document the logging categories.
1545 - Remove all uppercase from logging categories and useless cleanup in names.
1547 2004-08-18 Martin Quinson
1548 Version 0.6.2 (protocol not changed; API changed)
1549 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1550 gras_error_t), and NULL when not found. Functions expecting a msgtype
1551 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1552 hopefully usefull message.
1553 - Portability to prehistoric sparcs again
1555 2004-08-17 Martin Quinson
1556 Version 0.6.1 (protocol not changed; ABI not changed)
1557 - prealloc some buffers to speed things up
1559 2004-08-11 Martin Quinson
1560 Version 0.6 (protocol not changed; ABI expended)
1561 - The parsing macro can deal with the references, provided that you add
1562 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1564 2004-08-09 Martin Quinson
1565 Version 0.5 (protocol not changed; ABI changed)
1566 - Allow to off turn the cycle detection code in data exchange at
1567 compilation time. It should be at run time, but I'm short of time (and
1568 the config stuff is still broken). That way, we keep dict out of the
1569 critical path, which is good because the performance is poor:
1570 - search not dichotomial yet
1571 - dynar give no way to access their content and memcpy everytime
1572 - In composed data description (struct, ref and so on), stop foolness of
1573 keeping the subtype's ID, but store the type itself. This keeps sets out
1574 of the critical path, which is good since they rely on dynar and
1575 dictionnaries. The only loose of that is that we cannot detect the
1576 redeclaration of a structure/union with another content (but I'm not sure
1577 the code detected well this error before anyway). We still can detect
1578 the redefinition discrepancy for the other types.
1579 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1580 it breaks the code because of type-punning used all over the place).
1581 This breaks on all non-gcc architectures (for now).
1583 All those changes (plus the buffer of last time) allow me to gain 2 order
1584 of magnitude on cruel tests consisting of 800000 array of integers on two
1585 level of a hierarchical structure (200 secondes -> 4 secondes)
1588 - the selector of reference must now return the type it points to, not
1589 the ID of this type.
1591 2004-08-06 Martin Quinson
1592 Version 0.4 (protocol changed; ABI not changed)
1593 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1594 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1595 - Implement a buffer transport, and use it by default (it relies on tcp in
1596 real life and on sg in simulation).
1597 That's a bit hackish since I had a new field to the structure to store
1598 its data without interfering with the subtype ones. Inheritance
1599 is tricky in C. And that's a kind of reverse inheritance with one class
1600 derivating two classes. Or maybe a game with java interfaces. Anyway,
1601 that's damn hard in C (at least).
1602 Moreover, I got tired while trying to ensure plugin separation and
1603 genericity in SG mode. MSG wants me to do weird things, so let's go for
1604 cruel hacks (temporarily of course ;).
1605 See comment in transport_private.h:71
1606 - do not include all the _interface headers in private but in the files
1607 which really need them (to cut the compilation time when they are
1610 2004-07-26 Martin Quinson
1611 Version 0.3 (protocol not changed; ABI changed)
1612 - Major overhault of the datadesc interface to simplify it:
1613 - shorted the function names:
1614 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1615 - add a trivial way to push/pop integers into the cbps without malloc.
1616 This allows to make really generic sub_type description, which simply
1617 pop their size of the stack.
1618 - add a function gras_datadesc_ref_pop_arr() which does what users want
1619 most of the time: Declare a dynamic array (which pops its size of the
1620 stack) and declare a reference to it. Poor name, but anyway.
1621 - kill the post-send callback, add a post-receive one
1623 2004-07-23 Martin Quinson
1624 Version 0.2 (protocol changed; ABI changed)
1625 - add some testing for cpbs in the test cases, and fix some more bugs.
1626 This invalidate again the little64 data file, since I cannot regenerate
1628 - remove an awfull optimization in the logging stuff, allowing me to:
1629 - understand it again
1630 - learn gcc how to check that the argument match the provided format
1631 - fix all errors revealed by gcc after that
1632 - internal keys of dict are not \0 terminated. Deal with it properly in
1633 loggings instead of segfaulting when the user want to see the logs :-/
1635 2004-07-22 Martin Quinson
1636 - Fix some stupid bug preventing cbps (callback postit) from working
1638 2004-07-21 Martin Quinson
1639 - Some documentation cleanups
1640 - remove the useless last argument of msgtype_declare
1641 - rename the Virtu functions to fit into the 'os' namespace
1642 - move headers src/include -> src/include/gras/ and stop fooling with
1643 gras -> . symbolic link
1644 - make distcheck is now successful
1646 2004-07-19 Martin Quinson
1648 - Build shared library also
1649 - Install html doc to the right location
1650 - stop removing maintainer files in make clean
1651 - build tests only on make check
1653 2004-07-13 Martin Quinson
1655 - No major issue in previous version => change versionning schema
1656 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1657 corresponding dataset.
1659 2004-07-11 Martin Quinson
1661 - Get it working with any kind of structure (we can compute the padding
1662 bytes remotely for all the architectures I have access to)
1663 - Implement the structure parsing macro (still not quite robust/complete)
1664 - Improvement to the remote testing toysuite
1666 2004-07-10 Martin Quinson
1667 [autoconf mechanism]
1668 - get ride of a bunch of deprecated macros
1669 - actually run the test for two-compliment, not only compile it :-/
1670 - test whether the structures get packed (and bail out if yes. Damn.
1671 Alignment is a serious matter)
1672 - test whether the structures get compacted (but respecting the alignment
1673 constraints of each types)
1674 - test whether the array fields of structures can straddle alignment boundaries
1676 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1677 (took me 2 hours to find that bug, looking at the wrong place)
1678 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1679 before sending/receiving and are used to compute the offsets of fields
1680 - Given that padding size depend even on compiler options, keep track of
1681 alignment and aligned_size only for the current architecture. Not a big
1682 deal since we send structure fields one after the other (seems
1684 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1685 program, let it work on linux/gcc/little32. portability todo.
1687 2004-07-08 Martin Quinson
1688 - import and improve remote compilation support from FAST
1689 - make sure make check works on half a dozen of machines out there
1691 2004-07-07 Martin Quinson
1692 Let's say it's version 0.0.3 ;)
1693 - Implement conversions (yuhu!)
1694 - Let it work on solaris (beside conversion, of course)
1695 - Stupid me, using rand() to generate the conversion datatests in not wise.
1697 2004-07-06 Martin Quinson
1698 - Let make dist work, since I'm gonna need it to compile on remote hosts
1699 - Let Tests/datadesc_usage write the architecture on which the file was
1700 generated as first byte.
1701 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1702 architecture descriptions.
1703 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1704 on those architectures.
1705 - Optimization: send/recv array of scalar in one shoot
1707 2004-07-05 Martin Quinson
1708 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1710 - Plug a whole bunch of memleaks
1711 - each process now have to call gras_{init,exit}. One day, their log
1712 settings will be separated
1713 - Continue the code factorisation between SG, RL and common in Transport.
1715 2004-07-04 Martin Quinson
1717 - Redistribution between SG and RL.
1718 We wanna have to accept in SG, so move accepted related parts of RL in
1719 the common part. (more precisely, the dynar of all known sockets is no
1720 more a static in transport.c, but part of the process_data)
1722 [gras_stub_generator]
1723 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1725 2004-07-03 Martin Quinson
1726 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1727 them all up in one shot)
1729 - Fix the ugly case of reference to dynamic array.
1730 - New (semi-public) function gras_datadesc_size to allow the messaging
1731 layer to malloc the needed space for the buffer.
1733 - gras_socket_close now expect the socket to close (and not its address to
1734 put NULL in it after it). This is because the socket passed to handlers
1735 is one of their argument (=> not writable).
1737 - propagate the interface cleanup from last week in datadesc, ie remove a
1738 superfluous level of indirection. User pass adress of variable
1739 containing data (both when sending and receiving), and not of a variable
1740 being a pointer to the data. Let's say that I like it better ;)
1741 The price for that is constructs like "int msg=*(int*)payload" in
1742 handlers, but it's a fine price, IMHO.
1744 - Let it work in RL (yuhu)
1746 2004-06-21 Martin Quinson
1748 - porting SG plugin and SG select to new standards (works almost).
1749 - plug memleaks and fix bugs around.
1752 - cleanup the prototype of data recv and force users to specify when they
1753 want to handle references to objects. Test case working even for cycles.
1754 - plug memleaks. Valgrind is perfectly ok with this.
1756 2004-06-12 Martin Quinson
1758 - cleanup the separation between plugin and main code in plugin creation
1760 2004-06-11 Martin Quinson
1762 - Reput hook for raw sockets, needed for BW experiments
1763 - kill a few lines of dead code
1764 [Data description] Interface cleanup
1765 - gras_datadesc_by_name returns the searched type or NULL.
1766 That way, no variable is needed to use a type desc once, which makes
1768 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1769 two parameters were strings (field name, type name), leading to
1771 [Dicos] Interface cleanup
1772 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1773 This is consistant with the dynar API.
1775 2004-04-21 Martin Quinson
1777 - Porting to new standards.
1779 - interface cleanup.
1780 There is no bag anymore, no need to take extra provision to mask the
1781 pointers behind "ID".
1782 Better splitup of functions between files create/exchange/convert.
1783 This is still a bit artificial since convert and receive are so
1784 interleaved, but anyway.
1786 - add a queued message list to procdata (the ones not matching criteria
1788 - factorize some more code between SG and RL wrt procdata
1790 - use gras_exit in example to track memleaks
1791 - get rid of gs_example now that GS is properly integrated into gras
1792 - update run_test to integrate the lastest tests (datadesc)
1794 - rename WARNINGn macros to WARNn since it prooved error-prone
1796 2004-04-19 Martin Quinson
1798 - register init/exit functions within gras module mechanism
1799 - send/receive function.
1800 Convertion is not implemented, but short-cutted if not needed.
1801 struct/array elements are sent one by one (instead of block-wise), but
1802 nobody really cares (yet). Get a prototype before optimizing.
1803 - tests (using a file socket) for DD send/receive on:
1804 - base types: int, float
1805 - array: fixed size, string (ie ref to dynamic string)
1806 - structure: homogeneous, heterogeneous
1807 - chained list, graph with cycle
1808 Believe it or not, valgrind is not too unhappy with the results. The
1809 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1813 - Bugfix when using a filename explicitely (instead of '-')
1815 2004-04-09 Martin Quinson
1817 - factorize more code between RL and SG in socket creation
1818 - Complete the implementation and tests of:
1820 o file (only in RL, and mainly for debugging)
1822 I lost 3 days to design a portable address resolver, and then decided
1823 that the prototype mainly have to run on my box.
1824 Addressing portability too early may be like optimizing too early :-/
1826 - use gras_init in the Tests instead of the crappy parse_log_opt
1827 (the latter function is removed)
1828 [Conditional execution]
1829 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1830 [Code reorganisation]
1831 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1832 Build examples against the RL library, since there is no way to disable
1833 its creation for now.
1835 For information, the beginning of coding on GRAS was back in june
1836 2003. I guess that every line has been rewritten at least twice since