1 SimGrid (3.4.1-svn) unstable; urgency=low
4 * Warn the user about loop dependencies in data flow of DAX files
5 * Obey the control-flow dependencies of DAX files
8 * Fix a bug preventing the tasks from begin garbage collected.
11 * Add option "enable_smpi" for fix mac problem with smpi
12 * Add verification of lua5.1 and ruby1.8
14 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
16 SimGrid (3.4) stable; urgency=low
18 The "Easter in Cargese" release. Also known as (major changes):
20 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
21 ~> bindings were greatly improved
22 ~> new tracing infrastructure for better visualization introduced
24 * the "Welcome to configury modernity" release.
25 ~> we switched from autotools to cmake, and improved our cdash
28 A more detailled list of changes follow (full detail in svn log).
30 Java Bindings: Various Cleanups
31 * (install java-gcj-compat-dev on debian-like to use them)
32 * Remove put/get: no need to export deprecated interface in Java
33 Use send/receive instead.
34 * Cleanup the examples and add a README per directory
35 * Remove example autoDestination (that's the only way to go now)
36 * Remove example explicitDestination (was a plain copy of basic)
37 * Make JniException a runtime exception, so that there is no need to
38 declare the fact that you may encounter such a beast. I guess that
39 nobody will ever want to survive such error.
40 * Create specific errors for each MSG case of failure:
41 host failure, transfer failure, timeout, task cancelled
42 * Cleanup the exceptions that may get thrown by each function
43 * Other internal cleanups in Java bindings. Performance still bad :/
44 Ruby and Lua Bindings: create them
45 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
46 * That's new and great, you should try them out.
47 Same functionalities than Java bindings, only even less polished
49 * Kill the useless "rate" argument of SD_task_get_execution_time()
50 Everyone used to provide -1 as a value, it was not used, and the
51 semantic of a possible use wasn't even clear.
52 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
53 either as comm costs or compute costs to mean that there is no
54 such thing for that specific task.
55 * Add a SD_task_set_name() function
56 * Fix SD_task_unschedule() on typed tasks
57 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
58 * In DAX loader, accept useless 'level' attributes to <job> since
59 LIGO DAGs have them (seem to be to ease graphical representation).
61 * Add an example masterslave_mailbox.c using send/receive and not
62 the deprecated put/get interface.
63 * Kill the MSG_paje_output() function. It's a noop since 2 years.
64 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
66 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
67 (declare MSG_USE_DEPRECATED to still have the old name)
68 * Add a MSG_task_set_data() function
69 * About trace replay (see examples/msg/actions):
71 - Allow to work with splitted trace files for each process
72 Give the specific trace file as argument of each process,
73 and call MSG_action_trace_run(NULL)
74 You can still have one merged file for all processes.
75 - Fix implementation of collective operations
76 * Allow task_execute() on 0-sized tasks (closes #10063)
78 * This is the first release of SimGrid where SMPI is not considered
79 beta anymore (even if some corners should still be improved)
80 * Port over the new SIMIX_network submodule (internal refactoring)
81 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
82 * Implement more missing elements of the standard:
84 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
86 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
87 MPI_Scan() MPI_Get_processor_name()
88 - Added implementation of missing case for Alltoall (warning: it's
89 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
90 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
91 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
94 - MPI_Waitsome() was broken
95 - Allow relative includes in smpicc
96 - Command line cfg argument 'reference_speed' was ignored...
97 - Some functions did not properly lead to auto-benching of user code
98 - smpicc passes -O2 by default (just like openmpi one)
100 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
101 * Bug fixes about timeouts during communications
102 * add SIMIX_message_sizes_output() as a pimple to write to file the
103 amount of messages per size. Use gnuplot to get histogram.
104 Pimple because that's the only user-visible function of simix,
105 defined directly in xbt.h (irk, sorry)
107 - Add a SIMIX_sem_get_capacity() function
108 - Fix interactions with processe resume/suspende
109 - release_forever() was stupidly broken
110 - Fix SIMIX_display_process_status() for processes in a semaphore
111 - Make SIMIX_sem_block_onto() user-visible
112 * Refactoring context stuff:
113 - Use pseudo-OOP for better modularity
114 - reimplement SIMIX_process_kill() without process_schedule() so
115 that the latter can take as invariant that it is called from
117 - Merge context_start into context_new for sake of simplicity
119 * Add a Vivaldi network model, coded live during SUD'10 ;)
120 * Rename configuration variables to start a hierarchy:
121 o cpu_model -> cpu/model
122 o network_model -> network/model
123 o workstation_model -> workstation/model
124 * New configuration variables:
125 o network/bandwidth_factor: correction to bandwith
126 o network/latency_factor: correction to latency
127 o netwotk/weight_S: correction to the weight of competing streams
128 * Add a long description to the models, that users can see with such
129 argument on the command line: --cfg=cpu/model:help
130 * --help-models display the long description of all known models
132 * config: add the ability to set a default value after registration
133 Does not override any previously set value (e.g. from cmd line)
134 * dict: allow to have integer key and data.
135 When so, you need to use the following functions
136 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
137 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
138 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
139 In contrary to regular dicts, the key is not malloced before copy.
140 Mixing scalar and regular elements in the same dict is not tested
142 * Allow to use xbt_dynar_shrink() to expend the dynar instead
143 Tracing for Visualization:
144 * SimGrid is now instrumented in order to generate a trace file for
145 visualization analysis: to use it, need to compile SimGrid with the
146 "tracing" option enabled, and instrument the program using SimGrid with
147 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
148 (among other functions).
149 * The instrumentation only traces the platform utilization for now
150 * Documentation to use the tracing functions and how to analyze the
151 traces with the Triva tool is written.
152 * More information about: SimGrid FAQ (in the section Tracing Simulations
155 * We moved to cmake as default build system. Autotools support will
156 be dropped soon. Check the FAQ for more info about how to use it.
157 * Greatly improved our cdash/ctest interactions
158 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
159 * Added memory checking tests with valgrind; lot of memleak fixing.
160 This may be the first release of simgrid with so few memory issues
161 * Added code coverage tests.
162 Our coverage is still improvable, but at least we see it on cdash.
164 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
166 SimGrid (3.3.4) stable; urgency=low
168 The "Desktop Grid needs love too" release (also called Xmas release).
171 * Major speedup in the maxmin system solving by using lazy evaluation
172 Instead of solving completely the maxmin system at each iteration,
173 only invalidate (and recompute) the modified parts.
174 This new feature is enabled in default models but you can try to
175 turn it on with "--cfg:maxmin-selective-update=1" for other models.
176 * Cas01 IMproved as default CPU model
177 This CPU model is the same Cas01 model, but it uses the
178 maxmin-selective-update flag and a heap structure to manage
179 actions on SURF kernel.
180 It reduces the complexity to find the next action to finish and,
181 consequently, it's faster than the old Cas01.
182 This is the new default CPU model (Cas01).
183 * Rename the old Cas01 model to Cas01_fullupdate
184 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
185 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
186 * CpuTI (CPU Trace Integration)
187 A new CPU model whose objective is simulate faster when using
188 availability trace files.
189 Instead of using a full featured, over engineered maxmin system for
190 CPU modeling, this model does the pre-integration of traces files
191 to calculate the amount of CPU power available, and so, executes
192 faster than the old CPU models.
193 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
194 * Use LV08 as default network model since it gives better accuracy
195 for small messages and shouldn't change things for big ones.
196 Use --cfg=network_model:CM02 to get the previous behavior.
199 ******************************************
200 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
201 ******************************************
202 * The new CPU model may changes simulations!
203 The point is that events occurring at the exact same timestamp
204 are not scheduled in the same order with the old and new
205 version. This may be enough to completely change the execution
206 of simulations in some cases.
207 * The new network model will change simulations!
208 This new model is more realistic than the previous one, so you
209 should consider redoing your old experiments with this model.
210 Sorry for the inconvenience.
213 * Introduce the supernovae compilation mode
214 When compiled that way, the whole SimGrid (or almost) is put in a
215 single compilation unit and compiled in one shoot.
216 This is to help gcc which has difficulties to inline stuff from one
218 The speedup seem to be above 15%, althrough more tests are needed on
219 amd64 to confirm that gain.
222 * Port of MSG's mailbox on top of SIMIX network
223 The put/get mechanism was greatly simplified on the way.
226 * New SIMIX network module. Provides:
227 - Mailbox: rendez-vous mecanism to find with who you want to speak
228 - Synchronous send/recv: easier and hopefully faster since the
229 logic is handled in the maestro process directly now
230 - Asynchronous send/recv: you dreamt of it? It's here now
231 Too bad that nobody cared enough to propagate the change to MSG.
232 * Add semaphores as SIMIX synchronization mechanism.
235 * new function SD_daxload(char*) to load a DAX file
236 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
237 * Introduce typed tasks. Specify its kind and cost at creation.
238 At scheduling, just give where it should be placed, and the cost
239 for each involved resource is automatically computed.
240 Existing constructors so far (more to come of course):
241 - SD_task_create_comm_e2e() for end-to-end communication
242 - SD_task_create_comp_seq() for sequential computation
243 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
244 * new function SD_task_dump() for debuging display
245 * new function SD_task_dotty(task,FILE*) writing to file the info
246 about the task in dotty format
247 * SD_task_dependency_exists() can now cope with having one of its
248 arguments NULL. If so, it tests whether the other argument has any
250 * Add getters on list of preceding/following tasks:
251 SD_task_get_parents(task) and SD_task_get_children(task)
252 * Add getters on amount of workstations and list:
253 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
254 * Add getter on task kind: SD_task_get_kind(task)
255 * Update the start_time and finish_time of tasks on completion/failure
256 * Bugfix: Remove task from state swags when destroyed
259 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
262 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
263 dynar with one separate thread per value of the dynar.
264 * Change the prototype of xbt_thread_create(), sorry.
265 Added a boolean parameter indicating whether we want to join this
266 thread (used in SG only for now)
267 * Implement xbt_thread_join and xbt_thread_yield in SG also.
270 * GTNetS wrappers should now be usable again (and betterly tested too)
271 * Fix a major regression from 3.2 where the timeout provided to
272 MSG_task_put_with_timeout() was used as absolute time before which
273 the comm should be done.
274 * Start to fix the <cluster> tag.
275 - Internal links should be good now (beside of the loopback, which
276 use the private link instead)
277 - paths to the external world is still rather broken
278 - the <route:multi> tag is just broken. Actually that's brain-dead.
279 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
280 to make it less stupid
281 ** Check your platform with teshsuite/simdag/platforms/flatifier **
282 * Fix a source-level compatibility glitch from 3.2: after defining
283 MSG_USE_DEPRECATED, you can use the old name
284 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
285 * Allow to compile from the SVN with automake 1.11
286 * Fix some problems when using the "start_time" tag in deployment XMLs.
287 * Fix #8569: XBT/synchro.h has redundant declarations
288 * Fix #8563: MSG return values and exceptions
289 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
290 * Integrate patch #8636: Obey DESTDIR when installing documentation.
291 Thanks to Robson Peixoto.
292 * Fix a vicious bug in dictionaries inducing that some elements were
293 not freed on xbt_dict_free()
295 Portability report of this version:
296 * Main portability targets:
297 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
299 - mac leopard on i386
300 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
301 but nothing critical.
302 * Other platforms: windows, AIX and others were not tested for this release
304 Timing report of this version:
305 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
306 scenario parameters). From 8h to a few seconds in desktop grid settings.
307 * Supernovae brings about 25% speedup on i386.
309 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
311 SimGrid (3.3.3) stable; urgency=low
313 The "Need for Speed" release.
315 The timings done to validate the 3.3.2 were faulty.
316 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
318 The problem was a conversion from a manually handled vector to
319 xbt_dynar_t on the critical path.
320 xbt_dynar_foreach calls functions, inducing stack management crap.
322 We inlined these functions and xbt_dynar_foreach is now breath taking.
323 We also inlined xbt_swag_belong on the way.
325 Here are some approximate speedup measurements (on master/slaves
326 simulations lasting between 10s and 20s each):
327 3.3.1 -> 3.3.2: about same performance
328 3.3.2 -> 3.3.3: 40% speedup
329 3.3.1 -> 3.3.3: 40% speedup
330 3.3.1 with inline patch -> 3.3.3: 30% speedup
332 Our reading is that the refactoring which occurred in 3.3.2 made us
333 suffer much more from the xbt_dynar_foreach low performance, but
334 once we solved this, this refactoring proved to be very performance
335 effective. From the 40% speedup, somehow, 10% are due to the
336 inlining and 30% to the refactoring.
338 That's a pitty that gcc cannot inline functions placed in other files
339 alone. We have to choose between:
340 - break the encapsulation (by putting private data structures and
341 accessors in headers files to help gcc)
342 - live with low performance
343 - switch to a decent compiler such as icc (not quite possible).
345 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
347 SimGrid (3.3.2) stable; urgency=low
349 The "Simplicity does not preceed complexity, but follows it" release.
351 The main contributors of this release were (lexical order):
352 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
355 * Extract the routing logic into its own object.
356 (was dupplicated in network.c and workstation_LV07.c;
357 Allows to implement other ways of storing that info)
358 => kill now useless network_card concept
359 - Use dynar to represent routes (instead of void** + int*)
360 - kill link_set (use surf_network_model->resource_set instead)
361 - Add a command-line option to choose the routing schema to use
362 - Add three new models:
363 * Floyd (shortest path computed at initialization)
364 * Dijikstra (shortest path recomputed all the time)
365 * Cached Dijikstra (shortest path computed on need)
366 All these models where contributed by Silas De Munck, and are
367 described in his ICCS09 paper.
369 * Simplify model declaration
370 (less redirections, less function to write when defining a model)
371 - Factorize stuff between models:
374 surf_model_resource_set(model)
375 surf_model_resource_by_name(model, name)
376 - Unify the types of models in s_surf_model_t (using an union)
377 - Embeed fields of common_public directly into s_surf_model_t
378 - Rename model methods:
379 action_free ~> action_unref
380 action_change_state ~> action_state_set
381 action_get_state ~> action_state_get
382 - Change model methods into functions :
383 (model)->common_public->action_use ~> surf_action_ref
385 * Implement a generic resource; use it as ancestor to specific ones
386 (allows to kill duplicated code in models)
387 Drawback: timer command don't need no name nor properties;
388 workstation_CLM03 don't need no properties
389 (but I guess we can live with those few bytes wasted)
391 * Improve the action object model
392 - implement a constructor avoiding dupplicated code about field
393 initialization in generic_action part.
395 * Kill the SDP model: it has an external dependency, is deprecated
396 in flavor of modern lmm models, and didn't compile since a while
399 * Relocation of the context module from XBT to SIMIX.
400 (the context were decoupled from the simix processes, duplicating a lot of code)
401 => a lot of code was factorized
402 - less overhead is introduced during scheduling
403 - simpler API for the context factory
404 - the logic for process creation,destruction and manipulation was simplified
405 * Simplification of the s_smx_process_t data structure.
406 => accesing the simix level data associated to a process is faster now,
407 and the code is a lot more readable.
410 * Implement some more MPI primitives:
411 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
412 -implementation: Bcast: flat or 2-ary tree (default),
415 Allreduce: Reduce then Bcast
416 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
417 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
418 Alltoallv: flat tree, like ompi
420 * Add support for optimized collectives (Bcast is now binomial by default)
421 * Port smpirun and smpicc to OS X
424 * Kill SD_link_get_properties: hard to maintain and makes very little sense
425 Shout out if you used it.
428 * Display the list of still queued messages in SG mode when existing
432 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
433 * Add xbt_graph_node_get_outedges() [Silas De Munck]
434 * Add xbt_str_from_file(FILE*)
435 * Add xbt_dict_get_key achieving a linear reverse search
436 * Remove the context module
438 Portability report of this version:
439 * Main portability targets:
440 - Linux(debian)/x86/context
441 - Linux(debian)/x86/pthreads
442 - Linux(debian)/amd64/context
443 - Linux(debian)/amd64/pthreads
444 On these, we still have the eratic breakages of gras/pmm and
445 amok/saturate_sg reported in previous version. We still think
446 that the tests are the cause of the fault, not the tested code.
448 - Mac OSX Leopard/x86/context
449 Still false negative in tesh autotesting.
450 Smpi still fails, but this time because readlink does not accept -f
451 Everything seems to work properly beside of that.
454 - AIX version 5.3 (only tested contexts this time)
455 Smpi still fails there because mktemp is not installed.
456 Everything seems to work properly beside of that.
458 I managed to compile it for the first time, but several breakages.
459 Won't delay the release for this exotic platform.
461 * Windows: it's still lagging behind. If you want to help, please
464 Timing report of this version:
465 This version seem to be more than 5% faster than 3.3.1 (on linux
466 64bits with contextes). The gain is less than expected, we are
467 investigating this for next release.
469 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
471 SimGrid (3.3.1) stable; urgency=low
474 * Implement a --cfg-help to show existing configuration variables
475 * Build chain do not require doxygen in maintainer mode
478 * fix a bug on struct sizeof computation, which prevented the
479 exchange of arrays of structs in some conditions
480 - added a regression test about this in datadesc_usage
481 * Allow the exchange of 0-long dynamic vectors.
482 - for that, use -1 as indicator of dynamic size instead of 0
483 - This implied to change any size from unsigned long to long,
484 reducing a bit communication abilities, but I guess that with
485 64bits being quite common, this is more than enough.
486 - This also induce a protocol change, thus bumping network protocol
487 version from 0 to 1 (if we have external users, we have to get
488 clean on that point too ;)
489 - added two regression tests about this in datadesc_usage
490 * Be more verbose when propagating local exceptions
491 This helps debugging.
492 * Display the status of simulated processes when receiving SIGINT in
496 * Allow to control the simulation from a trace file.
497 New functions MSG_action_register() and MSG_action_trace_run()
498 The first one allows to associate a function execution to each
499 kind of action while the second one parses a trace file and
500 triggers the corresponding actions within the system.
501 For now, only a toy example is provided in examples/msg/actions
502 * Add an exemple of process migration in examples/msg/migration
503 * Fix a bug in task exchange which broke MSG_task_get_sender()
504 Add a teshsuite regression test for that.
505 [Bug: if MSG_task_get_sender() is called after sender exit,
507 * Fix a bug which prevented suspend/resume to work properly
508 * Display the status of simulated processes when receiving SIGINT
509 This fixes a regression of v3.3. due to the introduction of SIMIX
510 * Bug fixing in failure management:
511 - trace could not start by a failure at time 0
512 - failure during communications were not working
515 * Add SIMIX_process_set_name() to change the name of the current
516 process in the log messages.
517 * Store smx_hosts in a dict since we only retrieve them by name
518 * Move the configuration infrastructure to surf
521 * Move the configuration infrastructure to surf
524 * Massive internal cleanups:
525 - Store internal structures on processes instead of hosts (allows
526 to have more than one process per host, in addition of being more
528 - Cleanup the initialization/finalization process
529 - Kill a whole bunch of unneeded synchronization:
530 processes run in exclusive manner within the simulator
531 - Move queues from global tables to process data fields
533 - now accept -platform and -hostfile arguments
534 - Pass the right rank value to processes according to the hostfile
535 * Compile the examples by default, and use them as regression tests
536 * Implement MPI_Wtime()
537 * Change the reference speed to a command line option
540 * TCP_gamma can now be specified as command line option using
541 --cfg=TCP_gamma:10000000.0
542 * Change the --surf-path cmd line option into --cfg=path:
545 * Also include strbuff from xbt.h public header
546 * xbt_ex_display(): do not free the exception after displaying
547 This allows to do more with the given exception afterward.
548 Users should call xbt_ex_free() themselves.
552 Portability report of this version:
553 * Main portability targets:
554 - Linux(debian)/x86/context
555 - Linux(debian)/x86/pthreads
556 - Linux(debian)/amd64/context
557 - Linux(debian)/amd64/pthreads
558 These targets fail about 1/10 of times on gras/pmm, but we believe
559 that this is because of the test, not because of simgrid.
560 amok/saturate_sg fails even more rarely, and the test may not be
563 - Mac OSX Leopard/x86/context
564 The test suite still spits tons of errors because some obscure
565 force prevents us from removing the temporary directories
566 arguing that they still contain some metadata I've never heard of.
567 Smpi fails because seq is not installed.
568 Everything seems to work properly beside of that.
571 - AIX version 5.3 (both contexts and pthread)
572 Smpi still fails there because mktemp is not installed.
573 XML inclusions seems rosty on AIX.
575 * Windows: it's still lagging behind. If you want to help, please
578 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
580 SimGrid (3.3) stable; urgency=high
584 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
587 * Introduce the SIMIX module: factorize code between MSG and GRAS.
590 Until now, GRAS were using MSG as an interface to SURF. It was
591 quite difficult because both interface have several differences
592 (MSG channels vs GRAS sockets were the most notable point).
594 This also opens the gate to SMPI (which should occur soon) and speed
595 up simulations by to 40% (even if it were not the main goal).
597 **************************************
598 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
599 **************************************
600 The point is that events occuring at the exact same timestamp are
601 not scheduled in the same order with the old and new version. This
602 may be enough to completely change the execution of simulations in
603 some cases. Sorry for the inconvenience.
605 * Cleanup and upgrade the XML format to push further scalability
606 issues (check http://hal.inria.fr/inria-00256883/ for more info)
608 * Improve the testing infrastructure with tesh. Now a very large part of
609 the code is tested not only by being run but also by checking that the
610 output match an expected output [Mt].
612 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
613 is really less memory-demanding, which should allow you to use
614 larger files in SimGrid [AL].
616 * Inform valgrind about our contextes, so that it becomes usable
617 with the default (and more effecient) version of SimGrid
618 [contributed by Sékou Diakite, many thanks]
621 * Introduce a listener thread in charge of receiving incomming
622 messages from the network. It allows to overlap communication and
623 computation but most notably, it removes some stupid deadlocks due
624 to the fact that so far, a process could not send and receive at
625 the same time. This made most non trivial communication schema
627 * Convert the PIDs from long int to int to match the MSG ones (and
629 * New function: gras_agent_spawn() to launch a new process on
630 current host. Only working in simulation for now. [Mt]
631 * New function: gras_os_hostport() returning a constant form (ie,
632 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
635 * Make the backtrace of exceptions more human readable [Mt]
636 * New module: xbt/str [Mt]
637 a ton of string utility functions (split, join, printf to a newly
638 allocated buffer, trim, etc)
639 * New module: xbt/hash [Mt]
640 SHA1 hashing algorithm (more to come if needed)
641 * New module: xbt/synchro [Mt]
642 synchronization tools (mutex and conditions) working the same way
643 in simulation and in real life (mainly useful for GRAS, but not
645 * New module: xbt/queue [Mt]
646 classical producer/consumer synchronization scheme
647 * xbt_dynar_new_sync() creates a synchronized dynar. All access
648 (using the classical functions will get serialized) [Mt]
649 * Make dictionary internal table dynamic. No need to specify its size
650 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
652 * Make sure the log channels are organized as a tree under windows
653 (because of ANSI C compatibility issue, any channel were child of
657 * Cleaned many thing in surf and fixed a few bugs [AL].
658 * Add a nice command line configuration mechanism to compose models [AL].
659 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
660 the previous one (KCCFLN05). It relies on something that looks like
661 a max-min sharing mechanism but cannot be written as such. A new solver
662 was thus designed [AL].
663 * Add a new solver to lmm. Based on Lagrange optimization and
664 gradient-based descent, it enables to efficiently maximise systems s.a
666 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
668 This solver enables to propose two new network models for TCP Reno and
669 TCP Vegas based on Low's work. These models still need to be fully
670 tested though [Pedro Velho].
673 * Bug fix in SD_simulate. Now the time bound given as argument is
675 * Use the new parallel task model (ptask_L07) as default.
676 * Use the SURF command line configuration mechanism.
677 * 0-size tasks (for synchronization) should now work.
679 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
681 SimGrid (3.2) stable; urgency=high
685 We still experience issues on this platform, but we believe that at
688 GRAS API BREAKAGE (for simplification purpose, sorry):
689 * the gras_msgtype_by_name is not used anymore. Instead of
690 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
691 you can write (and must)
692 gras_msg_send(toserver, "request", &request);
693 - If you still want to pass a gras_msgtype_t to the function (to cache
694 the type and avoid the lookup time), use the gras_msg_send_() variant.
695 - Impacted functions:
696 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
697 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
698 * The callbacks are now expected to return 0 when everything went well
699 (just like the main() function)
701 GRAS new features and improvements:
702 * New module mecanism where user code can use per process globals [Mt]
703 This is similar to gras_userdata_*() functions, but for libraries. It
704 factorize some code developped over and over in the examples and AMOK.
705 It has still to be documented and used (only amok/peermanagement is
707 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
708 loss when they were small enough to fit into the buffer and sent quickly
709 enough so that they can all get received in one shoot.
710 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
711 if not found. Use the *_or_null() variant for the old semantic.
712 * In gras_msg_handle, do not discard messages without callback.
713 They are probably messages to be explicitly awaited later (ie, proofs of
714 mis-synchronization in userland since they are sent before being awaited)
716 * gras_socket_meas_send/recv: semantic changed!
717 The numerical arguments used to be (1) the total amount of data to send
718 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
719 messages. This was need for the fool willing to send more than MAXINT
720 bytes on quite fat pipes.
723 * Do really rename the hostmanagement module to peermanagement. [Mt]
724 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
725 but this is rather new and this was documented in the module
726 documentation (poor excuses, I admit)
727 * Bandwidth measurement semantic changed! This follows the changes to
728 gras_socket_meas_send/recv explained above.
731 * A sequential mode has been added to the workstations. When a workstation
732 is in sequential mode, it can execute only one task, and the other tasks
733 are waiting in a FIFO. [Christophe Thiery]
736 * The KCCFLN05 workstation model now handles parallel tasks. It is the
737 model for SIMDAG. [Christophe Thiery]
738 * Bug fix in the maxmin solver: Some values were close to 0 instead of
739 equal to 0, which caused some bad behaviors in
740 saturated_constraint_set_update. I now use a threshold mechanism like in
744 * When running manually src/testall, you select specific units [Mt]
745 testall is the result of our cunit mecanism, and should replace all
746 the scripty thingy around since bash don't run easily on billware.
748 * A mallocator system has been added. [Christophe Thiery]
749 Mallocators allow you to recycle your unused objects instead of freeing them
750 and allocating new ones.
752 Documentation update:
753 * FAQ reworking + New FAQs:
754 - "Valgrind spits tons of errors!" [Mt]
755 - "How to repport bugs" [Mt]
756 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
757 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
758 - Communication time measurement within MSG [AL]
759 - I experience weird communication times when I change the latency [AL]
762 - an introduction to the framework and to the used communication model
763 - an initiatic tour introducing the most proheminent features:
765 . Lesson 0: Installing GRAS
766 . Lesson 1: Setting up your own project
767 o Part 2: Message passing
768 . Lesson 2: Exchanging simple messages
769 . Lesson 3: Passing arguments to the processes (in SG)
770 . Lesson 4: Attaching callbacks to messages
771 . Lesson 5: Using globals in processes
772 . Lesson 6: Logging informations properly
773 . Lesson 7: Using internal timers
774 . Lesson 8: Handling errors through exceptions
775 . Lesson 9: Exchanging simple data
776 . Lesson 10: Remote Procedure Calling (RPC)
777 . Lesson 11: Explicitely waiting for messages
778 . Recapping of message passing features in GRAS
779 - A HOWTO section containing:
780 o HOWTO design a GRAS application
781 More are due, of course. They will come latter. In the meanwhile, you can
782 check the examples which are still here.
784 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
786 SimGrid (3.1) stable; urgency=high
790 There was a stack corruption somewhere, visible only when optimizing
791 with these versions. [Vince]
794 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
795 module that enabled to play with Directed Acyclic Graphs. It is built
796 directly on top of SURF and provides an API rather close to the old
797 SG. Some old codes using SG are currently under rewrite to check that
798 all needful functions are provided. [Christophe Thiery]
801 * Complete rewrite of the KCCFLN05 workstation model. It is now an
802 extension of the classical CLM03 model that gracefully handles
803 failures. This is now the default model for MSG and GRAS. It doesn't
804 handle parallel tasks yet though. [AL]
805 * Bug fix: Weights were not correctly set in the network part.
806 WARNING: This may have resulted in incorrect results with simulations
807 where there are more than one flow on a given link. [AL]
810 * After a (long ?) discussion on simgrid-devel, we have decided that the
811 convention we had on units was stupid. That is why it has been decided
812 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
813 WARNING : This means that all previous platform files will not work as
814 such with this version! A warning is issued to ask users to update
816 A conversion script can be found in the contrib module of the CVS, under
817 the name contrib/platform_generation/surfxml_update.pl [MQ]
820 * Bug fix: Processes were started in reverse order, wrt deployment file.
821 WARNING: if your code relies on this bug, please fix it. [AL]
822 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
823 being executed on two different locations. [AL]
824 * Bug fix: Failures are now better supported thanks to Derrick's tests
825 (there was many failure situations I hadn't thought of and that weren't
826 correctly handled). [AL]
827 * New function: MSG_host_is_avail indicates you whether a given m_host_t
831 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
832 Exception occurring on server-side are propagated back to client (!).
834 API CHANGE: the callback changed their prototype. Change:
835 int my_handler(gras_socket_t expeditor, void *payload_data) {
837 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
838 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
840 * New! function: gras_msg_handleall to deal with all messages arriving
841 within a given period.
842 * New! function: gras_socket_server_range to get a server socket in a
843 range of port numbers (ease to avoid port number conflicts) [MQ]
844 * New! gras processes display their backtrace when they get a SIGUSR1
845 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
846 Sweet to debug RL processes [MQ]
850 - Do not force experiment sizes to be expressed in kb, or it becomes
851 impossible to measure the latency this way (needs one byte-long tests)
852 WARNING: this changes the amok_bw_* function semantic. [MQ]
853 - Implements the link saturation stuff. [MQ]
854 * Peer management module:
855 New! module factorizing code that we wrote over and over [MQ].
858 * New module: cunit (my jUnit implementation in ansi C) [MQ]
859 - Test units are placed directly into the library code, they get extracted
860 automatically and placed into the src/testall binary.
861 - Convert most of the XBT tests to this system.
862 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
863 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
864 live in the same C code. This required to change a little bit the API
865 of surfxml parsing but shouldn't be an issue for end-users. [AL]
866 * New module: sparse graph structure with basic algorithms (this is work
867 in progress and the API is not considered to be frozen yet). [AL]
868 * Display more information on backtraces: source line & function names are
869 now displayed just like valgrind does (rely on addr2line tool) [MQ]
870 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
871 * Reworked a little bit some #include statements to load only required
872 headers. Some user code that relied on SimGrid to include stdlib or
873 stdio may need to include it by themselves. [AL]
874 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
876 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
877 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
878 but I guess that almost nobody used those parts. [MQ]
880 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
882 SimGrid (3.0.1) stable; urgency=low
885 * Unfortunately, I had missed 5 misnamed functions:
886 xbt_fifo_item_t xbt_fifo_newitem(void);
887 void xbt_fifo_freeitem(xbt_fifo_item_t);
888 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
889 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
890 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
891 They're now deprecated. Please use their new versions:
892 xbt_fifo_item_t xbt_fifo_new_item(void);
893 void xbt_fifo_free_item(xbt_fifo_item_t);
894 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
895 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
896 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
898 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
899 * Documentation: xbt_log module unmercifully reworked [MQ]
900 * Bugfix: there was a problem with the ending of contexts with
901 the pthread backend. It caused some weird deadlock or behavior
902 depending on the pthread implementation. [AL]
903 * Bugfix: get the exceptions raised in the simulator repport where
904 and why they come from when they are not catched in time [AL, MQ]
907 * Bugfix: Do repport the error when two non-connected hosts try to
908 exchange data (Thanks to Flavien for stumbling into this one) [AL]
911 * Add additionnal checkings on communications. Assert that two
912 communicating hosts are connected by a set of links... [AL]
915 * Add additionnal checkings on channel values in communication [AL]
916 * New: MSG_task_get_source to see on which host a task was generated [HC]
917 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
918 the number of tasks waiting to be received on channel and sent
920 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
921 waits for the first task coming from a given host.. [AL]
923 GRAS new functionnalities: [MQ]
924 * Enhance the parsing macro to allow the size of multidimentional objects
925 to be given thru annotations.
926 * New example (and documentation): Matrix Multiplication a la RPC
927 (as when I was young!) and fix a bunch of bugs found on the way.
929 GRAS performance improvements: [MQ]
931 * Reduce the amount of cbps creation/destruction by making it static to
932 datadesc_send/recv() and using a (newly created) cbps_reset (based on
935 * Change libdata to a set so that we can search for stuff by ID (and thus
936 reduce the insane amount of dict lookups)
939 * Actually implement gras_datadesc_copy() so that we don't have to mimick
940 RL communication on top of SG since it's so uneffective.
941 It may also be used for inter-thread communication in RL, one day.
942 * Use gras_datadesc_copy() to exchange messages on top of SG
944 - improve message exchange performance on top of SG
945 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
946 * Don't exchange on the network the size of the used part of buffer,
947 instead, specify the possible buffer size to read().
949 - reduces the amount of read/write calls (one pair per exchange)
950 - reduces the amount of exchanged data (the size)
951 - allows to retrieve all arrived data on receiver side, if we don't need
952 it right now (subsequent read will peek the buffer)
953 - allows the receiver to proceed with the begining of the stream before
954 everything is arrived
955 - make it possible to build an iov transport (using readv/writev)
957 - take care of the data with non-stable storage (like stacked data),
959 * If possible, TCP send uses vector I/O (when writev() is here)
960 - Don't use it for receive since we send data sizes and data on the
961 same stream, so we wouldn't be able to chain large amount of chunks
962 before having to flush the stuff to read the size.
963 * Rework the transport plugin mecanism to simplify it and reduce the
964 amount of pointer dereferencement when searching for the right function
967 * I guess that now, we do almost as few system calls as possible while
968 doing as few data copy as possible.
970 To improve it further, we could try to send all the sizes first and then
971 all the data (to use iov on receiving size), but it's only a partial
972 solution: when you have 2 dimensional data, the sizes of the second
973 dimension is data of the first dimension, so you need 3 streams.
975 I'm not sure the potential performance gains justify the coding burden.
977 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
979 SimGrid (3.00) stable; urgency=high
982 * New! Give the possibility to hijack the surf parser and thus bypass
983 MSG_create_environment and MSG_launch_application. Have a look at
984 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
986 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
988 SimGrid (2.96) unstable; urgency=low
993 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
994 This deprecates the xbt_error_t mecanisms.
995 It modifies (simplifies) all XBT and GRAS API.
996 MSG API keeps unchanged (exceptions raised by XBT are catched from
997 within MSG and masked with existing error handling facilities)
1000 * New! Add a FATPIPE model. [AL]
1001 * New! Add a parallel task model. [AL]
1002 * New! Add automatically a loopback interface (in the default
1003 network model) if none was precised.
1006 * Bugfix: MSG_process_resume now works with the current running process.
1008 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1009 * Modification of MSG_task_get_compute_duration. Once a task has been
1010 processed, the value returned by this function is now equal to 0. [AL]
1011 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1012 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1013 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1014 of a m_task. For now, MSG_task_cancel only works with computation
1016 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1017 of the processor (in Mflop/s) regardless of the current load on the
1019 * API Change: use proper naming convention for MSG_getClock and
1020 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1022 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1023 This function changes the priority of a computation task. This priority
1024 doesn't affect the transfer rate. A priority of 2 will make a task
1025 receive two times more cpu power than the other ones. This function
1026 has been added to suit the needs of Nguyen The Loc and hasn't been that
1027 much tested yet. So if it fails, please report it and send me your code.
1029 * API Change: removed all functions and types that were marked "deprecated"
1030 since many months. Renamed MSG_global_init_args to MSG_global_init.
1032 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1034 SimGrid (2.95) unstable; urgency=low
1037 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1038 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1039 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1042 * Bugfix: Allow absolute paths for platform description files [MQ]
1043 * Bugfix: do free the variables after use. Leads to drastic performance
1045 * Implement max_duration (ie, timeouts) on resources [AL]
1048 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1050 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1051 the needed support (timeouts) [AL]
1054 * Implement measurement sockets. You can now get the bandwidth between two
1055 hosts thanks to AMOK (see below). [MQ]
1056 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1057 dynar over the network (yeah) [MQ]
1058 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1059 * simple/static token-ring example. [Alexandre Colucci and MQ]
1060 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1061 in place before [MQ]
1064 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1065 type parsing. "short" and "long" modifiers were also ignored; other
1066 modifier (such as reference level) are still ignored. [MQ]
1067 * Embeed the buffer size within the buffer itself on SG. [MQ]
1068 That way, send() are atomic and cannot get intermixed anymore (at least
1069 the ones which are less than 100k; bigger messages still have the issue)
1070 * Array size pushed by the field, not by the field type (or each
1071 and every long int will push stuff to the cbps) [MQ]
1072 * use select() to sleep since it allows to portably sleep less than one
1075 GRAS (minor cleanups)
1076 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1079 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1080 * type category 'ignored' killed as it was never used and were difficult
1082 * whether a type can cycle or not is now a flag, leaving room for more
1083 flags (as "ignored", if we feel the need one day ;)
1084 * Rename raw sockets to measurement sockets since "raw" has another
1085 meaning in networking community.
1088 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1089 and offers features not belonging to GRAS but that most applications
1090 need. One day, it may be a set of plugins loadable at runtime.
1091 * New module: bandwidth
1092 bandwidth measurement between arbitrary nodes running this module. [MQ]
1094 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1096 SimGrid (2.94) unstable; urgency=low
1098 The first beta release of SimGrid 3 !
1102 * Update the main page and the FAQ. Adding references to gforge.
1105 * Add a gras_os_getpid function.
1108 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1109 * Extend the logs so that they also print PID, hostname, date, ... if
1111 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1112 and kill the old version which were in testsuite/
1113 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1117 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1118 for users. Instead, each programming environment (GRAS, MSG,...) use it
1119 when needed to provide such a feature to users.
1120 Don't shortcut the mecanism or you will also shortcut the virtualization
1121 you need on the simulator.
1125 * Cleanups in configury with regard to compile optimization/warning flags.
1126 Also add -fno-loop-optimize to any powerpc since it's the optimization
1127 killing gcc (< 3.4.0).
1128 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1129 needed by MSG examples complications
1130 * Borrow configury beautifications from PHP
1133 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1134 warning (thanks loris for stumbling into this one).
1135 * Bugfix: stop loading private headers (gras_config.h) from the public
1139 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1140 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1141 RL would imply a DNS resolver, which is *hard* to do in a portable way
1142 (and therefore delayed).
1143 * Implement a real timer mecanism and use it in timing macros. This allows
1144 to avoid rounding errors and get a 0.000005 sec precision in timing
1145 macros. While I was at it, various cleanups:
1146 - allow to declare more than one timed section per file (fix a stupid bug)
1147 - move some private declaration to the right place
1148 - merge conditional execution and timing macros into emulation module
1149 - document the module
1150 - make sure the module cleanups its mess on gras_exit
1151 * Documentation improvements:
1152 - (new) how to compile applications using GRAS
1153 - (new) emulation support (timing macros)
1155 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1157 SimGrid (2.93) unstable; urgency=low
1159 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1162 - Use Paje properly where used. Still to be sanitized properly.
1163 - Portability fix: Add an implementation of the contexts using pthread
1167 - Add xbt_procname(): returns the name of the current process.
1168 Use it to show the current process's name in all logging.
1170 - fix detection of older flex version and the reaction, since we do
1171 depend on modern ones (we use lex_destroy)
1172 - Better separation of SG and RL in the libs: remove all simulation code
1173 from libgras. As a result, this lib is now only 200k when stripped.
1174 Some of the xbt modules may also be duplicated (two sets and such) and
1175 should be cleaned/killed before SG3.
1176 - Insist on using xlC on AIX because of weird problems involving gcc there.
1177 - Cleanup the make remote stuff. This is now done by scripts
1178 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1179 mainly for our private use, but we're working on changing them to user
1182 - Bugfix: flush the socket on close only if there is some *output*.
1183 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1185 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1188 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1190 SimGrid (2.92) unstable; urgency=low
1192 Alpha 3 on the path to SimGrid 3
1196 - New! First try of benchmarking macros.
1197 - New! First try so that gras_stub_generator generate deployment and
1198 remote compilation helpers.
1200 - Bugfix: Initialization fix in msg_test.
1204 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1206 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1208 - API Change: gras_os_sleep to take the amount of seconds as a double.
1209 Accepting an int was error prone since it was the only location where
1210 seconds were coded as such. It leaded to damn rounding errors.
1211 - Bugfix: Hard to belive that timers ever worked before this.
1213 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1215 SimGrid (2.91) unstable; urgency=low
1217 Alpha 2 on the path to SimGrid 3
1221 - Bug fix in the lmm_solver.
1223 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1224 through the function MSG_paje_output.
1225 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1226 - It is possible to bound the rate of a communication in MSG with
1227 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1228 to write it in the changelog).
1229 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1234 - Various cleanups to the autotools stuff
1235 - Begin to move Gras examples to examples/gras/
1236 - Let make distcheck work again (yeah!)
1238 - documentation overhauled using doxygen.
1239 gtk-doc-tools is dead in SimGrid now.
1240 - Automatically extract all existing logging categories, and add the list
1241 to the documentation (long standing one, to say the less)
1243 - Cleanup the known architecture table. Reorder the entries to group what
1244 should be, and use a more consistent naming scheme.
1245 (some of the test dataset are still to be regenerated)
1246 - New! Allow library to register globals on each process just as userdata
1248 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1249 lookup time (for now).
1250 Use it in msg and trp.
1251 This cleans a lot the internals and helps enforcing privacy of the
1252 headers between the gras components.
1253 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1254 - Bugfix: gras_os_time was delirious in RL.
1255 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1257 - Reenable GRAS now that it works.
1259 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1261 SimGrid (2.90) unstable; urgency=low
1263 Alpha 1 on the path to SimGrid 3
1265 * It is a long time since the last release of SimGrid. I'm sorry about
1266 that but as I had told you, I was rewriting a lot of things. I apologize
1267 to those who had been reporting bugs to me and that I had not answered.
1268 If your bug is still in the new version, please tell me. Here is a
1269 summary of the main changes.
1271 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1272 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1273 portability, speed, and a lot more but you'll figure it out later.
1274 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1275 I wanted to release the new MSG as soon as possible and I have
1276 broken GRAS, which is the reason why, for now, only MSG is fully
1277 functional. A laconic description of these projects is available
1278 in the documentation.
1280 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1281 kernel optimized for our needs (called SURF but only the developers
1282 should use it). Hence, MSG is now roughly 30 times faster and I think
1283 that by rewriting a little bit MSG, I could event speed it up a little
1284 bit more. Beside the gain in speed, it is also much easier to encode a
1285 new platform model with SURF than it was with SG. More to come...
1287 * REVOLUTION 3: I have tried to change a little as possible the API of
1288 MSG but a few things really had to disappear. The main differences
1289 with the previous version are :
1290 1) no more m_links_t and the corresponding functions. Platforms are
1291 directly read from a XML description and cannot be hard-coded
1292 anymore. The same format is used for application deployment
1293 description. The new format is described in the documentation.
1294 Have a look in tools/platform_generation. There is a tiny script
1295 that converts from the old platform format to the new one. Concerning
1296 the application deployment format, parsing the old one is tricky.
1297 I think most of you should however be able to convert your files. If
1298 it is really an issue, I can write a C code that does the conversion.
1300 2) the toolbox tbx does not exist anymore. We now have a library
1301 with much more data-structures but without the hash-tables (we have
1302 dictionaries that are much faster).
1304 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1306 *****************************************************************************
1307 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1308 * don't feel like converting the oldies. (Mt) *
1309 *****************************************************************************
1312 Version 2.90: "the long awaited one"
1313 - Finished rewriting and debugging MSG. Rewrote the documentation.
1314 - disable GRAS for now since it needs to be ported to the newest SG
1317 - Finish the port to windows (using mingw32 for cross-compile)
1320 - Main loop and datastructures of SURF. A cpu resource object is
1321 functional. Surf can thus be used to create cpu's with variable
1322 performance on which you can execute some actions.
1324 2004-11-15 Martin Quinson
1325 - Port to ARM. Simply added the alignment and size descriptions. Should
1326 work, but the ARM machines are so slow that I didn't had the opportunity
1327 to 'make check' over there yet.
1329 2004-11-15 Arnaud Legrand
1330 - Trace manager now written. It uses a heap structure and is therefore
1331 expected to be efficient. It may however be speeded up (particularly
1332 when many events occur at the same date) by using red and black
1333 trees. One day maybe...
1334 - Max-min linear system solver written. It uses a sparse matrix
1335 structure taking advantage of its expected use. Most operations are
1336 O(1) and free/calloc are called as few as possible. The computation of
1337 the minimum could however be improved by using a red and black tree
1340 2004-11-03 Arnaud Legrand
1341 - Rename every gras_* function that was in xbt/ to its xbt_
1343 - Add a heap and a doubly-linked list to xbt
1344 - Added a dichotomy to the dictionaries. make check works as well before
1345 so I assume that the patch is correct. I do not know however if things
1346 run effectively faster than before now. :)
1348 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1349 SimGrid, and the version number is bumped to 2.x
1351 2004-10-29 Martin Quinson
1352 - Introduction of the remote errors.
1353 They are the result of a RMI/RPC on the remote machine.
1354 ErrCodes being scalar values, you can't get the host on which those
1355 errors did happen. Extending the error mechanism as in Gnome is possible.
1356 No idea yet whether it is a good idea.
1358 2004-10-28 Martin Quinson
1359 - Interface revolution: the Starred Structure Eradication.
1360 I used to do typedef struct {} toto_t; and then handle *toto_t.
1361 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1362 - ???_t is a valid type (builded with typedef)
1363 - s_toto_t is a structure (access to fields with .)
1364 - s_toto is a structure needing 'struct' keyword to be used
1365 - e_toto_t is an enum
1366 - toto_t is an 'object' (struct*)
1368 typedef struct s_toto {} s_toto_t, *toto_t;
1369 typedef enum {} e_toto_t;
1370 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1371 s_toto_t) is private.
1373 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1374 so that it can set the variable to NULL. It was so for dicts and sets,
1375 it changed for dynars.
1377 - Fix a bunch of memleaks in dict_remove
1378 - Fix a bug in sg/server_socket opening: it failed all the time.
1380 2004-10-07 Martin Quinson
1381 - Speed up dynar lookup operation a bit.
1383 gras_dynar_get is dead.
1385 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1386 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1387 (which gives you the address of the stored data).
1389 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1390 copy of the data using an affectation to do the job and not a memcpy.
1392 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1393 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1395 It does not really speedup the dynar test because they are
1396 setting elements all the time (and look them seldom). But the dict does
1397 far more lookup than setting.
1399 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1401 2004-10-05 Martin Quinson
1402 - Allow to (en/dis)able the cycle detection at run time.
1404 Whether we should check for cycle or not is now a property of each
1405 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1406 datadesc_cycle_unset allow to remove this property when previously set.
1408 Note that the cycle detection is off by default since it impacts the
1409 performance. Watch the data you feed GRAS with ;)
1411 This property is hereditary. Any element embedded in a structure having it
1412 set have it set for the time of this data exchange.
1414 You should set it both on sender and receiver side. If you don't set it on
1415 sender side, it will enter an endless loop. If you forget on receiver
1416 side, the cycles won't be recreated after communication.
1418 - Header reorganization.
1419 Kill gras_private.h, each submodule must load the headers it needs.
1421 2004-10-04 Martin Quinson
1422 - Interface revolution: do not try to survive to malloc failure.
1424 Now, gras_malloc and friends call gras_abort() on failure.
1425 As a conclusion, malloc_error is not a valid error anymore, and all
1426 functions for which it was the only gras_error_t return value are
1427 changed. They now return void, or there result directly.
1428 This simplify the API a lot.
1430 2004-09-29 Martin Quinson
1431 - Re-enable raw sockets.
1432 Created by gras_socket_{client,server}_ext;
1433 Used with gras_raw_{send,recv}
1436 It should allow to kill the last bits of gras first version soon.
1438 This is not completely satisfactory yet (duplicate code with
1439 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1442 - Simplify transport plugin (internal) interface by not passing any
1443 argument to _server and _client, but embedding them in the socket
1446 2004-09-28 Martin Quinson
1447 - Finish the port to AIX.
1448 autoconf was my problem (segfault within the malloc replacement
1449 function. No idea why)
1451 2004-09-16 Martin Quinson
1452 - Fix some size_t madness on 64bit architectures.
1454 2004-09-08 Martin Quinson
1455 - Reduce the number of system headers loaded, overload some more system
1456 calls (such as malloc to cast the result of the system one, and work
1458 - Fix and reintroduce the config support
1460 2004-09-07 Martin Quinson
1461 - Source code reorganization to allow Arnaud to surf all over there.
1462 - Allow to document the logging categories.
1463 - Remove all uppercase from logging categories and useless cleanup in names.
1465 2004-08-18 Martin Quinson
1466 Version 0.6.2 (protocol not changed; API changed)
1467 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1468 gras_error_t), and NULL when not found. Functions expecting a msgtype
1469 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1470 hopefully usefull message.
1471 - Portability to prehistoric sparcs again
1473 2004-08-17 Martin Quinson
1474 Version 0.6.1 (protocol not changed; ABI not changed)
1475 - prealloc some buffers to speed things up
1477 2004-08-11 Martin Quinson
1478 Version 0.6 (protocol not changed; ABI expended)
1479 - The parsing macro can deal with the references, provided that you add
1480 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1482 2004-08-09 Martin Quinson
1483 Version 0.5 (protocol not changed; ABI changed)
1484 - Allow to off turn the cycle detection code in data exchange at
1485 compilation time. It should be at run time, but I'm short of time (and
1486 the config stuff is still broken). That way, we keep dict out of the
1487 critical path, which is good because the performance is poor:
1488 - search not dichotomial yet
1489 - dynar give no way to access their content and memcpy everytime
1490 - In composed data description (struct, ref and so on), stop foolness of
1491 keeping the subtype's ID, but store the type itself. This keeps sets out
1492 of the critical path, which is good since they rely on dynar and
1493 dictionnaries. The only loose of that is that we cannot detect the
1494 redeclaration of a structure/union with another content (but I'm not sure
1495 the code detected well this error before anyway). We still can detect
1496 the redefinition discrepancy for the other types.
1497 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1498 it breaks the code because of type-punning used all over the place).
1499 This breaks on all non-gcc architectures (for now).
1501 All those changes (plus the buffer of last time) allow me to gain 2 order
1502 of magnitude on cruel tests consisting of 800000 array of integers on two
1503 level of a hierarchical structure (200 secondes -> 4 secondes)
1506 - the selector of reference must now return the type it points to, not
1507 the ID of this type.
1509 2004-08-06 Martin Quinson
1510 Version 0.4 (protocol changed; ABI not changed)
1511 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1512 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1513 - Implement a buffer transport, and use it by default (it relies on tcp in
1514 real life and on sg in simulation).
1515 That's a bit hackish since I had a new field to the structure to store
1516 its data without interfering with the subtype ones. Inheritance
1517 is tricky in C. And that's a kind of reverse inheritance with one class
1518 derivating two classes. Or maybe a game with java interfaces. Anyway,
1519 that's damn hard in C (at least).
1520 Moreover, I got tired while trying to ensure plugin separation and
1521 genericity in SG mode. MSG wants me to do weird things, so let's go for
1522 cruel hacks (temporarily of course ;).
1523 See comment in transport_private.h:71
1524 - do not include all the _interface headers in private but in the files
1525 which really need them (to cut the compilation time when they are
1528 2004-07-26 Martin Quinson
1529 Version 0.3 (protocol not changed; ABI changed)
1530 - Major overhault of the datadesc interface to simplify it:
1531 - shorted the function names:
1532 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1533 - add a trivial way to push/pop integers into the cbps without malloc.
1534 This allows to make really generic sub_type description, which simply
1535 pop their size of the stack.
1536 - add a function gras_datadesc_ref_pop_arr() which does what users want
1537 most of the time: Declare a dynamic array (which pops its size of the
1538 stack) and declare a reference to it. Poor name, but anyway.
1539 - kill the post-send callback, add a post-receive one
1541 2004-07-23 Martin Quinson
1542 Version 0.2 (protocol changed; ABI changed)
1543 - add some testing for cpbs in the test cases, and fix some more bugs.
1544 This invalidate again the little64 data file, since I cannot regenerate
1546 - remove an awfull optimization in the logging stuff, allowing me to:
1547 - understand it again
1548 - learn gcc how to check that the argument match the provided format
1549 - fix all errors revealed by gcc after that
1550 - internal keys of dict are not \0 terminated. Deal with it properly in
1551 loggings instead of segfaulting when the user want to see the logs :-/
1553 2004-07-22 Martin Quinson
1554 - Fix some stupid bug preventing cbps (callback postit) from working
1556 2004-07-21 Martin Quinson
1557 - Some documentation cleanups
1558 - remove the useless last argument of msgtype_declare
1559 - rename the Virtu functions to fit into the 'os' namespace
1560 - move headers src/include -> src/include/gras/ and stop fooling with
1561 gras -> . symbolic link
1562 - make distcheck is now successful
1564 2004-07-19 Martin Quinson
1566 - Build shared library also
1567 - Install html doc to the right location
1568 - stop removing maintainer files in make clean
1569 - build tests only on make check
1571 2004-07-13 Martin Quinson
1573 - No major issue in previous version => change versionning schema
1574 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1575 corresponding dataset.
1577 2004-07-11 Martin Quinson
1579 - Get it working with any kind of structure (we can compute the padding
1580 bytes remotely for all the architectures I have access to)
1581 - Implement the structure parsing macro (still not quite robust/complete)
1582 - Improvement to the remote testing toysuite
1584 2004-07-10 Martin Quinson
1585 [autoconf mechanism]
1586 - get ride of a bunch of deprecated macros
1587 - actually run the test for two-compliment, not only compile it :-/
1588 - test whether the structures get packed (and bail out if yes. Damn.
1589 Alignment is a serious matter)
1590 - test whether the structures get compacted (but respecting the alignment
1591 constraints of each types)
1592 - test whether the array fields of structures can straddle alignment boundaries
1594 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1595 (took me 2 hours to find that bug, looking at the wrong place)
1596 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1597 before sending/receiving and are used to compute the offsets of fields
1598 - Given that padding size depend even on compiler options, keep track of
1599 alignment and aligned_size only for the current architecture. Not a big
1600 deal since we send structure fields one after the other (seems
1602 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1603 program, let it work on linux/gcc/little32. portability todo.
1605 2004-07-08 Martin Quinson
1606 - import and improve remote compilation support from FAST
1607 - make sure make check works on half a dozen of machines out there
1609 2004-07-07 Martin Quinson
1610 Let's say it's version 0.0.3 ;)
1611 - Implement conversions (yuhu!)
1612 - Let it work on solaris (beside conversion, of course)
1613 - Stupid me, using rand() to generate the conversion datatests in not wise.
1615 2004-07-06 Martin Quinson
1616 - Let make dist work, since I'm gonna need it to compile on remote hosts
1617 - Let Tests/datadesc_usage write the architecture on which the file was
1618 generated as first byte.
1619 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1620 architecture descriptions.
1621 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1622 on those architectures.
1623 - Optimization: send/recv array of scalar in one shoot
1625 2004-07-05 Martin Quinson
1626 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1628 - Plug a whole bunch of memleaks
1629 - each process now have to call gras_{init,exit}. One day, their log
1630 settings will be separated
1631 - Continue the code factorisation between SG, RL and common in Transport.
1633 2004-07-04 Martin Quinson
1635 - Redistribution between SG and RL.
1636 We wanna have to accept in SG, so move accepted related parts of RL in
1637 the common part. (more precisely, the dynar of all known sockets is no
1638 more a static in transport.c, but part of the process_data)
1640 [gras_stub_generator]
1641 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1643 2004-07-03 Martin Quinson
1644 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1645 them all up in one shot)
1647 - Fix the ugly case of reference to dynamic array.
1648 - New (semi-public) function gras_datadesc_size to allow the messaging
1649 layer to malloc the needed space for the buffer.
1651 - gras_socket_close now expect the socket to close (and not its address to
1652 put NULL in it after it). This is because the socket passed to handlers
1653 is one of their argument (=> not writable).
1655 - propagate the interface cleanup from last week in datadesc, ie remove a
1656 superfluous level of indirection. User pass adress of variable
1657 containing data (both when sending and receiving), and not of a variable
1658 being a pointer to the data. Let's say that I like it better ;)
1659 The price for that is constructs like "int msg=*(int*)payload" in
1660 handlers, but it's a fine price, IMHO.
1662 - Let it work in RL (yuhu)
1664 2004-06-21 Martin Quinson
1666 - porting SG plugin and SG select to new standards (works almost).
1667 - plug memleaks and fix bugs around.
1670 - cleanup the prototype of data recv and force users to specify when they
1671 want to handle references to objects. Test case working even for cycles.
1672 - plug memleaks. Valgrind is perfectly ok with this.
1674 2004-06-12 Martin Quinson
1676 - cleanup the separation between plugin and main code in plugin creation
1678 2004-06-11 Martin Quinson
1680 - Reput hook for raw sockets, needed for BW experiments
1681 - kill a few lines of dead code
1682 [Data description] Interface cleanup
1683 - gras_datadesc_by_name returns the searched type or NULL.
1684 That way, no variable is needed to use a type desc once, which makes
1686 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1687 two parameters were strings (field name, type name), leading to
1689 [Dicos] Interface cleanup
1690 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1691 This is consistant with the dynar API.
1693 2004-04-21 Martin Quinson
1695 - Porting to new standards.
1697 - interface cleanup.
1698 There is no bag anymore, no need to take extra provision to mask the
1699 pointers behind "ID".
1700 Better splitup of functions between files create/exchange/convert.
1701 This is still a bit artificial since convert and receive are so
1702 interleaved, but anyway.
1704 - add a queued message list to procdata (the ones not matching criteria
1706 - factorize some more code between SG and RL wrt procdata
1708 - use gras_exit in example to track memleaks
1709 - get rid of gs_example now that GS is properly integrated into gras
1710 - update run_test to integrate the lastest tests (datadesc)
1712 - rename WARNINGn macros to WARNn since it prooved error-prone
1714 2004-04-19 Martin Quinson
1716 - register init/exit functions within gras module mechanism
1717 - send/receive function.
1718 Convertion is not implemented, but short-cutted if not needed.
1719 struct/array elements are sent one by one (instead of block-wise), but
1720 nobody really cares (yet). Get a prototype before optimizing.
1721 - tests (using a file socket) for DD send/receive on:
1722 - base types: int, float
1723 - array: fixed size, string (ie ref to dynamic string)
1724 - structure: homogeneous, heterogeneous
1725 - chained list, graph with cycle
1726 Believe it or not, valgrind is not too unhappy with the results. The
1727 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1731 - Bugfix when using a filename explicitely (instead of '-')
1733 2004-04-09 Martin Quinson
1735 - factorize more code between RL and SG in socket creation
1736 - Complete the implementation and tests of:
1738 o file (only in RL, and mainly for debugging)
1740 I lost 3 days to design a portable address resolver, and then decided
1741 that the prototype mainly have to run on my box.
1742 Addressing portability too early may be like optimizing too early :-/
1744 - use gras_init in the Tests instead of the crappy parse_log_opt
1745 (the latter function is removed)
1746 [Conditional execution]
1747 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1748 [Code reorganisation]
1749 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1750 Build examples against the RL library, since there is no way to disable
1751 its creation for now.
1753 For information, the beginning of coding on GRAS was back in june
1754 2003. I guess that every line has been rewritten at least twice since