1 SimGrid (3.4.1-svn) unstable; urgency=low
4 * Warn the user about loop dependencies in data flow of DAX files
7 * Fix a bug preventing the tasks from begin garbage collected.
9 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
11 SimGrid (3.4) stable; urgency=low
13 The "Easter in Cargese" release. Also known as (major changes):
15 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
16 ~> bindings were greatly improved
17 ~> new tracing infrastructure for better visualization introduced
19 * the "Welcome to configury modernity" release.
20 ~> we switched from autotools to cmake, and improved our cdash
23 A more detailled list of changes follow (full detail in svn log).
25 Java Bindings: Various Cleanups
26 * (install java-gcj-compat-dev on debian-like to use them)
27 * Remove put/get: no need to export deprecated interface in Java
28 Use send/receive instead.
29 * Cleanup the examples and add a README per directory
30 * Remove example autoDestination (that's the only way to go now)
31 * Remove example explicitDestination (was a plain copy of basic)
32 * Make JniException a runtime exception, so that there is no need to
33 declare the fact that you may encounter such a beast. I guess that
34 nobody will ever want to survive such error.
35 * Create specific errors for each MSG case of failure:
36 host failure, transfer failure, timeout, task cancelled
37 * Cleanup the exceptions that may get thrown by each function
38 * Other internal cleanups in Java bindings. Performance still bad :/
39 Ruby and Lua Bindings: create them
40 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
41 * That's new and great, you should try them out.
42 Same functionalities than Java bindings, only even less polished
44 * Kill the useless "rate" argument of SD_task_get_execution_time()
45 Everyone used to provide -1 as a value, it was not used, and the
46 semantic of a possible use wasn't even clear.
47 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
48 either as comm costs or compute costs to mean that there is no
49 such thing for that specific task.
50 * Add a SD_task_set_name() function
51 * Fix SD_task_unschedule() on typed tasks
52 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
53 * In DAX loader, accept useless 'level' attributes to <job> since
54 LIGO DAGs have them (seem to be to ease graphical representation).
56 * Add an example masterslave_mailbox.c using send/receive and not
57 the deprecated put/get interface.
58 * Kill the MSG_paje_output() function. It's a noop since 2 years.
59 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
61 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
62 (declare MSG_USE_DEPRECATED to still have the old name)
63 * Add a MSG_task_set_data() function
64 * About trace replay (see examples/msg/actions):
66 - Allow to work with splitted trace files for each process
67 Give the specific trace file as argument of each process,
68 and call MSG_action_trace_run(NULL)
69 You can still have one merged file for all processes.
70 - Fix implementation of collective operations
71 * Allow task_execute() on 0-sized tasks (closes #10063)
73 * This is the first release of SimGrid where SMPI is not considered
74 beta anymore (even if some corners should still be improved)
75 * Port over the new SIMIX_network submodule (internal refactoring)
76 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
77 * Implement more missing elements of the standard:
79 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
81 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
82 MPI_Scan() MPI_Get_processor_name()
83 - Added implementation of missing case for Alltoall (warning: it's
84 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
85 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
86 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
89 - MPI_Waitsome() was broken
90 - Allow relative includes in smpicc
91 - Command line cfg argument 'reference_speed' was ignored...
92 - Some functions did not properly lead to auto-benching of user code
93 - smpicc passes -O2 by default (just like openmpi one)
95 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
96 * Bug fixes about timeouts during communications
97 * add SIMIX_message_sizes_output() as a pimple to write to file the
98 amount of messages per size. Use gnuplot to get histogram.
99 Pimple because that's the only user-visible function of simix,
100 defined directly in xbt.h (irk, sorry)
102 - Add a SIMIX_sem_get_capacity() function
103 - Fix interactions with processe resume/suspende
104 - release_forever() was stupidly broken
105 - Fix SIMIX_display_process_status() for processes in a semaphore
106 - Make SIMIX_sem_block_onto() user-visible
107 * Refactoring context stuff:
108 - Use pseudo-OOP for better modularity
109 - reimplement SIMIX_process_kill() without process_schedule() so
110 that the latter can take as invariant that it is called from
112 - Merge context_start into context_new for sake of simplicity
114 * Add a Vivaldi network model, coded live during SUD'10 ;)
115 * Rename configuration variables to start a hierarchy:
116 o cpu_model -> cpu/model
117 o network_model -> network/model
118 o workstation_model -> workstation/model
119 * New configuration variables:
120 o network/bandwidth_factor: correction to bandwith
121 o network/latency_factor: correction to latency
122 o netwotk/weight_S: correction to the weight of competing streams
123 * Add a long description to the models, that users can see with such
124 argument on the command line: --cfg=cpu/model:help
125 * --help-models display the long description of all known models
127 * config: add the ability to set a default value after registration
128 Does not override any previously set value (e.g. from cmd line)
129 * dict: allow to have integer key and data.
130 When so, you need to use the following functions
131 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
132 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
133 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
134 In contrary to regular dicts, the key is not malloced before copy.
135 Mixing scalar and regular elements in the same dict is not tested
137 * Allow to use xbt_dynar_shrink() to expend the dynar instead
138 Tracing for Visualization:
139 * SimGrid is now instrumented in order to generate a trace file for
140 visualization analysis: to use it, need to compile SimGrid with the
141 "tracing" option enabled, and instrument the program using SimGrid with
142 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
143 (among other functions).
144 * The instrumentation only traces the platform utilization for now
145 * Documentation to use the tracing functions and how to analyze the
146 traces with the Triva tool is written.
147 * More information about: SimGrid FAQ (in the section Tracing Simulations
150 * We moved to cmake as default build system. Autotools support will
151 be dropped soon. Check the FAQ for more info about how to use it.
152 * Greatly improved our cdash/ctest interactions
153 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
154 * Added memory checking tests with valgrind; lot of memleak fixing.
155 This may be the first release of simgrid with so few memory issues
156 * Added code coverage tests.
157 Our coverage is still improvable, but at least we see it on cdash.
159 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
161 SimGrid (3.3.4) stable; urgency=low
163 The "Desktop Grid needs love too" release (also called Xmas release).
166 * Major speedup in the maxmin system solving by using lazy evaluation
167 Instead of solving completely the maxmin system at each iteration,
168 only invalidate (and recompute) the modified parts.
169 This new feature is enabled in default models but you can try to
170 turn it on with "--cfg:maxmin-selective-update=1" for other models.
171 * Cas01 IMproved as default CPU model
172 This CPU model is the same Cas01 model, but it uses the
173 maxmin-selective-update flag and a heap structure to manage
174 actions on SURF kernel.
175 It reduces the complexity to find the next action to finish and,
176 consequently, it's faster than the old Cas01.
177 This is the new default CPU model (Cas01).
178 * Rename the old Cas01 model to Cas01_fullupdate
179 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
180 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
181 * CpuTI (CPU Trace Integration)
182 A new CPU model whose objective is simulate faster when using
183 availability trace files.
184 Instead of using a full featured, over engineered maxmin system for
185 CPU modeling, this model does the pre-integration of traces files
186 to calculate the amount of CPU power available, and so, executes
187 faster than the old CPU models.
188 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
189 * Use LV08 as default network model since it gives better accuracy
190 for small messages and shouldn't change things for big ones.
191 Use --cfg=network_model:CM02 to get the previous behavior.
194 ******************************************
195 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
196 ******************************************
197 * The new CPU model may changes simulations!
198 The point is that events occurring at the exact same timestamp
199 are not scheduled in the same order with the old and new
200 version. This may be enough to completely change the execution
201 of simulations in some cases.
202 * The new network model will change simulations!
203 This new model is more realistic than the previous one, so you
204 should consider redoing your old experiments with this model.
205 Sorry for the inconvenience.
208 * Introduce the supernovae compilation mode
209 When compiled that way, the whole SimGrid (or almost) is put in a
210 single compilation unit and compiled in one shoot.
211 This is to help gcc which has difficulties to inline stuff from one
213 The speedup seem to be above 15%, althrough more tests are needed on
214 amd64 to confirm that gain.
217 * Port of MSG's mailbox on top of SIMIX network
218 The put/get mechanism was greatly simplified on the way.
221 * New SIMIX network module. Provides:
222 - Mailbox: rendez-vous mecanism to find with who you want to speak
223 - Synchronous send/recv: easier and hopefully faster since the
224 logic is handled in the maestro process directly now
225 - Asynchronous send/recv: you dreamt of it? It's here now
226 Too bad that nobody cared enough to propagate the change to MSG.
227 * Add semaphores as SIMIX synchronization mechanism.
230 * new function SD_daxload(char*) to load a DAX file
231 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
232 * Introduce typed tasks. Specify its kind and cost at creation.
233 At scheduling, just give where it should be placed, and the cost
234 for each involved resource is automatically computed.
235 Existing constructors so far (more to come of course):
236 - SD_task_create_comm_e2e() for end-to-end communication
237 - SD_task_create_comp_seq() for sequential computation
238 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
239 * new function SD_task_dump() for debuging display
240 * new function SD_task_dotty(task,FILE*) writing to file the info
241 about the task in dotty format
242 * SD_task_dependency_exists() can now cope with having one of its
243 arguments NULL. If so, it tests whether the other argument has any
245 * Add getters on list of preceding/following tasks:
246 SD_task_get_parents(task) and SD_task_get_children(task)
247 * Add getters on amount of workstations and list:
248 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
249 * Add getter on task kind: SD_task_get_kind(task)
250 * Update the start_time and finish_time of tasks on completion/failure
251 * Bugfix: Remove task from state swags when destroyed
254 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
257 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
258 dynar with one separate thread per value of the dynar.
259 * Change the prototype of xbt_thread_create(), sorry.
260 Added a boolean parameter indicating whether we want to join this
261 thread (used in SG only for now)
262 * Implement xbt_thread_join and xbt_thread_yield in SG also.
265 * GTNetS wrappers should now be usable again (and betterly tested too)
266 * Fix a major regression from 3.2 where the timeout provided to
267 MSG_task_put_with_timeout() was used as absolute time before which
268 the comm should be done.
269 * Start to fix the <cluster> tag.
270 - Internal links should be good now (beside of the loopback, which
271 use the private link instead)
272 - paths to the external world is still rather broken
273 - the <route:multi> tag is just broken. Actually that's brain-dead.
274 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
275 to make it less stupid
276 ** Check your platform with teshsuite/simdag/platforms/flatifier **
277 * Fix a source-level compatibility glitch from 3.2: after defining
278 MSG_USE_DEPRECATED, you can use the old name
279 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
280 * Allow to compile from the SVN with automake 1.11
281 * Fix some problems when using the "start_time" tag in deployment XMLs.
282 * Fix #8569: XBT/synchro.h has redundant declarations
283 * Fix #8563: MSG return values and exceptions
284 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
285 * Integrate patch #8636: Obey DESTDIR when installing documentation.
286 Thanks to Robson Peixoto.
287 * Fix a vicious bug in dictionaries inducing that some elements were
288 not freed on xbt_dict_free()
290 Portability report of this version:
291 * Main portability targets:
292 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
294 - mac leopard on i386
295 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
296 but nothing critical.
297 * Other platforms: windows, AIX and others were not tested for this release
299 Timing report of this version:
300 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
301 scenario parameters). From 8h to a few seconds in desktop grid settings.
302 * Supernovae brings about 25% speedup on i386.
304 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
306 SimGrid (3.3.3) stable; urgency=low
308 The "Need for Speed" release.
310 The timings done to validate the 3.3.2 were faulty.
311 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
313 The problem was a conversion from a manually handled vector to
314 xbt_dynar_t on the critical path.
315 xbt_dynar_foreach calls functions, inducing stack management crap.
317 We inlined these functions and xbt_dynar_foreach is now breath taking.
318 We also inlined xbt_swag_belong on the way.
320 Here are some approximate speedup measurements (on master/slaves
321 simulations lasting between 10s and 20s each):
322 3.3.1 -> 3.3.2: about same performance
323 3.3.2 -> 3.3.3: 40% speedup
324 3.3.1 -> 3.3.3: 40% speedup
325 3.3.1 with inline patch -> 3.3.3: 30% speedup
327 Our reading is that the refactoring which occurred in 3.3.2 made us
328 suffer much more from the xbt_dynar_foreach low performance, but
329 once we solved this, this refactoring proved to be very performance
330 effective. From the 40% speedup, somehow, 10% are due to the
331 inlining and 30% to the refactoring.
333 That's a pitty that gcc cannot inline functions placed in other files
334 alone. We have to choose between:
335 - break the encapsulation (by putting private data structures and
336 accessors in headers files to help gcc)
337 - live with low performance
338 - switch to a decent compiler such as icc (not quite possible).
340 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
342 SimGrid (3.3.2) stable; urgency=low
344 The "Simplicity does not preceed complexity, but follows it" release.
346 The main contributors of this release were (lexical order):
347 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
350 * Extract the routing logic into its own object.
351 (was dupplicated in network.c and workstation_LV07.c;
352 Allows to implement other ways of storing that info)
353 => kill now useless network_card concept
354 - Use dynar to represent routes (instead of void** + int*)
355 - kill link_set (use surf_network_model->resource_set instead)
356 - Add a command-line option to choose the routing schema to use
357 - Add three new models:
358 * Floyd (shortest path computed at initialization)
359 * Dijikstra (shortest path recomputed all the time)
360 * Cached Dijikstra (shortest path computed on need)
361 All these models where contributed by Silas De Munck, and are
362 described in his ICCS09 paper.
364 * Simplify model declaration
365 (less redirections, less function to write when defining a model)
366 - Factorize stuff between models:
369 surf_model_resource_set(model)
370 surf_model_resource_by_name(model, name)
371 - Unify the types of models in s_surf_model_t (using an union)
372 - Embeed fields of common_public directly into s_surf_model_t
373 - Rename model methods:
374 action_free ~> action_unref
375 action_change_state ~> action_state_set
376 action_get_state ~> action_state_get
377 - Change model methods into functions :
378 (model)->common_public->action_use ~> surf_action_ref
380 * Implement a generic resource; use it as ancestor to specific ones
381 (allows to kill duplicated code in models)
382 Drawback: timer command don't need no name nor properties;
383 workstation_CLM03 don't need no properties
384 (but I guess we can live with those few bytes wasted)
386 * Improve the action object model
387 - implement a constructor avoiding dupplicated code about field
388 initialization in generic_action part.
390 * Kill the SDP model: it has an external dependency, is deprecated
391 in flavor of modern lmm models, and didn't compile since a while
394 * Relocation of the context module from XBT to SIMIX.
395 (the context were decoupled from the simix processes, duplicating a lot of code)
396 => a lot of code was factorized
397 - less overhead is introduced during scheduling
398 - simpler API for the context factory
399 - the logic for process creation,destruction and manipulation was simplified
400 * Simplification of the s_smx_process_t data structure.
401 => accesing the simix level data associated to a process is faster now,
402 and the code is a lot more readable.
405 * Implement some more MPI primitives:
406 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
407 -implementation: Bcast: flat or 2-ary tree (default),
410 Allreduce: Reduce then Bcast
411 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
412 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
413 Alltoallv: flat tree, like ompi
415 * Add support for optimized collectives (Bcast is now binomial by default)
416 * Port smpirun and smpicc to OS X
419 * Kill SD_link_get_properties: hard to maintain and makes very little sense
420 Shout out if you used it.
423 * Display the list of still queued messages in SG mode when existing
427 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
428 * Add xbt_graph_node_get_outedges() [Silas De Munck]
429 * Add xbt_str_from_file(FILE*)
430 * Add xbt_dict_get_key achieving a linear reverse search
431 * Remove the context module
433 Portability report of this version:
434 * Main portability targets:
435 - Linux(debian)/x86/context
436 - Linux(debian)/x86/pthreads
437 - Linux(debian)/amd64/context
438 - Linux(debian)/amd64/pthreads
439 On these, we still have the eratic breakages of gras/pmm and
440 amok/saturate_sg reported in previous version. We still think
441 that the tests are the cause of the fault, not the tested code.
443 - Mac OSX Leopard/x86/context
444 Still false negative in tesh autotesting.
445 Smpi still fails, but this time because readlink does not accept -f
446 Everything seems to work properly beside of that.
449 - AIX version 5.3 (only tested contexts this time)
450 Smpi still fails there because mktemp is not installed.
451 Everything seems to work properly beside of that.
453 I managed to compile it for the first time, but several breakages.
454 Won't delay the release for this exotic platform.
456 * Windows: it's still lagging behind. If you want to help, please
459 Timing report of this version:
460 This version seem to be more than 5% faster than 3.3.1 (on linux
461 64bits with contextes). The gain is less than expected, we are
462 investigating this for next release.
464 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
466 SimGrid (3.3.1) stable; urgency=low
469 * Implement a --cfg-help to show existing configuration variables
470 * Build chain do not require doxygen in maintainer mode
473 * fix a bug on struct sizeof computation, which prevented the
474 exchange of arrays of structs in some conditions
475 - added a regression test about this in datadesc_usage
476 * Allow the exchange of 0-long dynamic vectors.
477 - for that, use -1 as indicator of dynamic size instead of 0
478 - This implied to change any size from unsigned long to long,
479 reducing a bit communication abilities, but I guess that with
480 64bits being quite common, this is more than enough.
481 - This also induce a protocol change, thus bumping network protocol
482 version from 0 to 1 (if we have external users, we have to get
483 clean on that point too ;)
484 - added two regression tests about this in datadesc_usage
485 * Be more verbose when propagating local exceptions
486 This helps debugging.
487 * Display the status of simulated processes when receiving SIGINT in
491 * Allow to control the simulation from a trace file.
492 New functions MSG_action_register() and MSG_action_trace_run()
493 The first one allows to associate a function execution to each
494 kind of action while the second one parses a trace file and
495 triggers the corresponding actions within the system.
496 For now, only a toy example is provided in examples/msg/actions
497 * Add an exemple of process migration in examples/msg/migration
498 * Fix a bug in task exchange which broke MSG_task_get_sender()
499 Add a teshsuite regression test for that.
500 [Bug: if MSG_task_get_sender() is called after sender exit,
502 * Fix a bug which prevented suspend/resume to work properly
503 * Display the status of simulated processes when receiving SIGINT
504 This fixes a regression of v3.3. due to the introduction of SIMIX
505 * Bug fixing in failure management:
506 - trace could not start by a failure at time 0
507 - failure during communications were not working
510 * Add SIMIX_process_set_name() to change the name of the current
511 process in the log messages.
512 * Store smx_hosts in a dict since we only retrieve them by name
513 * Move the configuration infrastructure to surf
516 * Move the configuration infrastructure to surf
519 * Massive internal cleanups:
520 - Store internal structures on processes instead of hosts (allows
521 to have more than one process per host, in addition of being more
523 - Cleanup the initialization/finalization process
524 - Kill a whole bunch of unneeded synchronization:
525 processes run in exclusive manner within the simulator
526 - Move queues from global tables to process data fields
528 - now accept -platform and -hostfile arguments
529 - Pass the right rank value to processes according to the hostfile
530 * Compile the examples by default, and use them as regression tests
531 * Implement MPI_Wtime()
532 * Change the reference speed to a command line option
535 * TCP_gamma can now be specified as command line option using
536 --cfg=TCP_gamma:10000000.0
537 * Change the --surf-path cmd line option into --cfg=path:
540 * Also include strbuff from xbt.h public header
541 * xbt_ex_display(): do not free the exception after displaying
542 This allows to do more with the given exception afterward.
543 Users should call xbt_ex_free() themselves.
547 Portability report of this version:
548 * Main portability targets:
549 - Linux(debian)/x86/context
550 - Linux(debian)/x86/pthreads
551 - Linux(debian)/amd64/context
552 - Linux(debian)/amd64/pthreads
553 These targets fail about 1/10 of times on gras/pmm, but we believe
554 that this is because of the test, not because of simgrid.
555 amok/saturate_sg fails even more rarely, and the test may not be
558 - Mac OSX Leopard/x86/context
559 The test suite still spits tons of errors because some obscure
560 force prevents us from removing the temporary directories
561 arguing that they still contain some metadata I've never heard of.
562 Smpi fails because seq is not installed.
563 Everything seems to work properly beside of that.
566 - AIX version 5.3 (both contexts and pthread)
567 Smpi still fails there because mktemp is not installed.
568 XML inclusions seems rosty on AIX.
570 * Windows: it's still lagging behind. If you want to help, please
573 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
575 SimGrid (3.3) stable; urgency=high
579 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
582 * Introduce the SIMIX module: factorize code between MSG and GRAS.
585 Until now, GRAS were using MSG as an interface to SURF. It was
586 quite difficult because both interface have several differences
587 (MSG channels vs GRAS sockets were the most notable point).
589 This also opens the gate to SMPI (which should occur soon) and speed
590 up simulations by to 40% (even if it were not the main goal).
592 **************************************
593 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
594 **************************************
595 The point is that events occuring at the exact same timestamp are
596 not scheduled in the same order with the old and new version. This
597 may be enough to completely change the execution of simulations in
598 some cases. Sorry for the inconvenience.
600 * Cleanup and upgrade the XML format to push further scalability
601 issues (check http://hal.inria.fr/inria-00256883/ for more info)
603 * Improve the testing infrastructure with tesh. Now a very large part of
604 the code is tested not only by being run but also by checking that the
605 output match an expected output [Mt].
607 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
608 is really less memory-demanding, which should allow you to use
609 larger files in SimGrid [AL].
611 * Inform valgrind about our contextes, so that it becomes usable
612 with the default (and more effecient) version of SimGrid
613 [contributed by Sékou Diakite, many thanks]
616 * Introduce a listener thread in charge of receiving incomming
617 messages from the network. It allows to overlap communication and
618 computation but most notably, it removes some stupid deadlocks due
619 to the fact that so far, a process could not send and receive at
620 the same time. This made most non trivial communication schema
622 * Convert the PIDs from long int to int to match the MSG ones (and
624 * New function: gras_agent_spawn() to launch a new process on
625 current host. Only working in simulation for now. [Mt]
626 * New function: gras_os_hostport() returning a constant form (ie,
627 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
630 * Make the backtrace of exceptions more human readable [Mt]
631 * New module: xbt/str [Mt]
632 a ton of string utility functions (split, join, printf to a newly
633 allocated buffer, trim, etc)
634 * New module: xbt/hash [Mt]
635 SHA1 hashing algorithm (more to come if needed)
636 * New module: xbt/synchro [Mt]
637 synchronization tools (mutex and conditions) working the same way
638 in simulation and in real life (mainly useful for GRAS, but not
640 * New module: xbt/queue [Mt]
641 classical producer/consumer synchronization scheme
642 * xbt_dynar_new_sync() creates a synchronized dynar. All access
643 (using the classical functions will get serialized) [Mt]
644 * Make dictionary internal table dynamic. No need to specify its size
645 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
647 * Make sure the log channels are organized as a tree under windows
648 (because of ANSI C compatibility issue, any channel were child of
652 * Cleaned many thing in surf and fixed a few bugs [AL].
653 * Add a nice command line configuration mechanism to compose models [AL].
654 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
655 the previous one (KCCFLN05). It relies on something that looks like
656 a max-min sharing mechanism but cannot be written as such. A new solver
657 was thus designed [AL].
658 * Add a new solver to lmm. Based on Lagrange optimization and
659 gradient-based descent, it enables to efficiently maximise systems s.a
661 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
663 This solver enables to propose two new network models for TCP Reno and
664 TCP Vegas based on Low's work. These models still need to be fully
665 tested though [Pedro Velho].
668 * Bug fix in SD_simulate. Now the time bound given as argument is
670 * Use the new parallel task model (ptask_L07) as default.
671 * Use the SURF command line configuration mechanism.
672 * 0-size tasks (for synchronization) should now work.
674 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
676 SimGrid (3.2) stable; urgency=high
680 We still experience issues on this platform, but we believe that at
683 GRAS API BREAKAGE (for simplification purpose, sorry):
684 * the gras_msgtype_by_name is not used anymore. Instead of
685 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
686 you can write (and must)
687 gras_msg_send(toserver, "request", &request);
688 - If you still want to pass a gras_msgtype_t to the function (to cache
689 the type and avoid the lookup time), use the gras_msg_send_() variant.
690 - Impacted functions:
691 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
692 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
693 * The callbacks are now expected to return 0 when everything went well
694 (just like the main() function)
696 GRAS new features and improvements:
697 * New module mecanism where user code can use per process globals [Mt]
698 This is similar to gras_userdata_*() functions, but for libraries. It
699 factorize some code developped over and over in the examples and AMOK.
700 It has still to be documented and used (only amok/peermanagement is
702 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
703 loss when they were small enough to fit into the buffer and sent quickly
704 enough so that they can all get received in one shoot.
705 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
706 if not found. Use the *_or_null() variant for the old semantic.
707 * In gras_msg_handle, do not discard messages without callback.
708 They are probably messages to be explicitly awaited later (ie, proofs of
709 mis-synchronization in userland since they are sent before being awaited)
711 * gras_socket_meas_send/recv: semantic changed!
712 The numerical arguments used to be (1) the total amount of data to send
713 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
714 messages. This was need for the fool willing to send more than MAXINT
715 bytes on quite fat pipes.
718 * Do really rename the hostmanagement module to peermanagement. [Mt]
719 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
720 but this is rather new and this was documented in the module
721 documentation (poor excuses, I admit)
722 * Bandwidth measurement semantic changed! This follows the changes to
723 gras_socket_meas_send/recv explained above.
726 * A sequential mode has been added to the workstations. When a workstation
727 is in sequential mode, it can execute only one task, and the other tasks
728 are waiting in a FIFO. [Christophe Thiery]
731 * The KCCFLN05 workstation model now handles parallel tasks. It is the
732 model for SIMDAG. [Christophe Thiery]
733 * Bug fix in the maxmin solver: Some values were close to 0 instead of
734 equal to 0, which caused some bad behaviors in
735 saturated_constraint_set_update. I now use a threshold mechanism like in
739 * When running manually src/testall, you select specific units [Mt]
740 testall is the result of our cunit mecanism, and should replace all
741 the scripty thingy around since bash don't run easily on billware.
743 * A mallocator system has been added. [Christophe Thiery]
744 Mallocators allow you to recycle your unused objects instead of freeing them
745 and allocating new ones.
747 Documentation update:
748 * FAQ reworking + New FAQs:
749 - "Valgrind spits tons of errors!" [Mt]
750 - "How to repport bugs" [Mt]
751 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
752 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
753 - Communication time measurement within MSG [AL]
754 - I experience weird communication times when I change the latency [AL]
757 - an introduction to the framework and to the used communication model
758 - an initiatic tour introducing the most proheminent features:
760 . Lesson 0: Installing GRAS
761 . Lesson 1: Setting up your own project
762 o Part 2: Message passing
763 . Lesson 2: Exchanging simple messages
764 . Lesson 3: Passing arguments to the processes (in SG)
765 . Lesson 4: Attaching callbacks to messages
766 . Lesson 5: Using globals in processes
767 . Lesson 6: Logging informations properly
768 . Lesson 7: Using internal timers
769 . Lesson 8: Handling errors through exceptions
770 . Lesson 9: Exchanging simple data
771 . Lesson 10: Remote Procedure Calling (RPC)
772 . Lesson 11: Explicitely waiting for messages
773 . Recapping of message passing features in GRAS
774 - A HOWTO section containing:
775 o HOWTO design a GRAS application
776 More are due, of course. They will come latter. In the meanwhile, you can
777 check the examples which are still here.
779 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
781 SimGrid (3.1) stable; urgency=high
785 There was a stack corruption somewhere, visible only when optimizing
786 with these versions. [Vince]
789 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
790 module that enabled to play with Directed Acyclic Graphs. It is built
791 directly on top of SURF and provides an API rather close to the old
792 SG. Some old codes using SG are currently under rewrite to check that
793 all needful functions are provided. [Christophe Thiery]
796 * Complete rewrite of the KCCFLN05 workstation model. It is now an
797 extension of the classical CLM03 model that gracefully handles
798 failures. This is now the default model for MSG and GRAS. It doesn't
799 handle parallel tasks yet though. [AL]
800 * Bug fix: Weights were not correctly set in the network part.
801 WARNING: This may have resulted in incorrect results with simulations
802 where there are more than one flow on a given link. [AL]
805 * After a (long ?) discussion on simgrid-devel, we have decided that the
806 convention we had on units was stupid. That is why it has been decided
807 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
808 WARNING : This means that all previous platform files will not work as
809 such with this version! A warning is issued to ask users to update
811 A conversion script can be found in the contrib module of the CVS, under
812 the name contrib/platform_generation/surfxml_update.pl [MQ]
815 * Bug fix: Processes were started in reverse order, wrt deployment file.
816 WARNING: if your code relies on this bug, please fix it. [AL]
817 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
818 being executed on two different locations. [AL]
819 * Bug fix: Failures are now better supported thanks to Derrick's tests
820 (there was many failure situations I hadn't thought of and that weren't
821 correctly handled). [AL]
822 * New function: MSG_host_is_avail indicates you whether a given m_host_t
826 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
827 Exception occurring on server-side are propagated back to client (!).
829 API CHANGE: the callback changed their prototype. Change:
830 int my_handler(gras_socket_t expeditor, void *payload_data) {
832 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
833 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
835 * New! function: gras_msg_handleall to deal with all messages arriving
836 within a given period.
837 * New! function: gras_socket_server_range to get a server socket in a
838 range of port numbers (ease to avoid port number conflicts) [MQ]
839 * New! gras processes display their backtrace when they get a SIGUSR1
840 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
841 Sweet to debug RL processes [MQ]
845 - Do not force experiment sizes to be expressed in kb, or it becomes
846 impossible to measure the latency this way (needs one byte-long tests)
847 WARNING: this changes the amok_bw_* function semantic. [MQ]
848 - Implements the link saturation stuff. [MQ]
849 * Peer management module:
850 New! module factorizing code that we wrote over and over [MQ].
853 * New module: cunit (my jUnit implementation in ansi C) [MQ]
854 - Test units are placed directly into the library code, they get extracted
855 automatically and placed into the src/testall binary.
856 - Convert most of the XBT tests to this system.
857 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
858 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
859 live in the same C code. This required to change a little bit the API
860 of surfxml parsing but shouldn't be an issue for end-users. [AL]
861 * New module: sparse graph structure with basic algorithms (this is work
862 in progress and the API is not considered to be frozen yet). [AL]
863 * Display more information on backtraces: source line & function names are
864 now displayed just like valgrind does (rely on addr2line tool) [MQ]
865 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
866 * Reworked a little bit some #include statements to load only required
867 headers. Some user code that relied on SimGrid to include stdlib or
868 stdio may need to include it by themselves. [AL]
869 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
871 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
872 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
873 but I guess that almost nobody used those parts. [MQ]
875 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
877 SimGrid (3.0.1) stable; urgency=low
880 * Unfortunately, I had missed 5 misnamed functions:
881 xbt_fifo_item_t xbt_fifo_newitem(void);
882 void xbt_fifo_freeitem(xbt_fifo_item_t);
883 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
884 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
885 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
886 They're now deprecated. Please use their new versions:
887 xbt_fifo_item_t xbt_fifo_new_item(void);
888 void xbt_fifo_free_item(xbt_fifo_item_t);
889 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
890 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
891 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
893 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
894 * Documentation: xbt_log module unmercifully reworked [MQ]
895 * Bugfix: there was a problem with the ending of contexts with
896 the pthread backend. It caused some weird deadlock or behavior
897 depending on the pthread implementation. [AL]
898 * Bugfix: get the exceptions raised in the simulator repport where
899 and why they come from when they are not catched in time [AL, MQ]
902 * Bugfix: Do repport the error when two non-connected hosts try to
903 exchange data (Thanks to Flavien for stumbling into this one) [AL]
906 * Add additionnal checkings on communications. Assert that two
907 communicating hosts are connected by a set of links... [AL]
910 * Add additionnal checkings on channel values in communication [AL]
911 * New: MSG_task_get_source to see on which host a task was generated [HC]
912 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
913 the number of tasks waiting to be received on channel and sent
915 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
916 waits for the first task coming from a given host.. [AL]
918 GRAS new functionnalities: [MQ]
919 * Enhance the parsing macro to allow the size of multidimentional objects
920 to be given thru annotations.
921 * New example (and documentation): Matrix Multiplication a la RPC
922 (as when I was young!) and fix a bunch of bugs found on the way.
924 GRAS performance improvements: [MQ]
926 * Reduce the amount of cbps creation/destruction by making it static to
927 datadesc_send/recv() and using a (newly created) cbps_reset (based on
930 * Change libdata to a set so that we can search for stuff by ID (and thus
931 reduce the insane amount of dict lookups)
934 * Actually implement gras_datadesc_copy() so that we don't have to mimick
935 RL communication on top of SG since it's so uneffective.
936 It may also be used for inter-thread communication in RL, one day.
937 * Use gras_datadesc_copy() to exchange messages on top of SG
939 - improve message exchange performance on top of SG
940 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
941 * Don't exchange on the network the size of the used part of buffer,
942 instead, specify the possible buffer size to read().
944 - reduces the amount of read/write calls (one pair per exchange)
945 - reduces the amount of exchanged data (the size)
946 - allows to retrieve all arrived data on receiver side, if we don't need
947 it right now (subsequent read will peek the buffer)
948 - allows the receiver to proceed with the begining of the stream before
949 everything is arrived
950 - make it possible to build an iov transport (using readv/writev)
952 - take care of the data with non-stable storage (like stacked data),
954 * If possible, TCP send uses vector I/O (when writev() is here)
955 - Don't use it for receive since we send data sizes and data on the
956 same stream, so we wouldn't be able to chain large amount of chunks
957 before having to flush the stuff to read the size.
958 * Rework the transport plugin mecanism to simplify it and reduce the
959 amount of pointer dereferencement when searching for the right function
962 * I guess that now, we do almost as few system calls as possible while
963 doing as few data copy as possible.
965 To improve it further, we could try to send all the sizes first and then
966 all the data (to use iov on receiving size), but it's only a partial
967 solution: when you have 2 dimensional data, the sizes of the second
968 dimension is data of the first dimension, so you need 3 streams.
970 I'm not sure the potential performance gains justify the coding burden.
972 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
974 SimGrid (3.00) stable; urgency=high
977 * New! Give the possibility to hijack the surf parser and thus bypass
978 MSG_create_environment and MSG_launch_application. Have a look at
979 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
981 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
983 SimGrid (2.96) unstable; urgency=low
988 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
989 This deprecates the xbt_error_t mecanisms.
990 It modifies (simplifies) all XBT and GRAS API.
991 MSG API keeps unchanged (exceptions raised by XBT are catched from
992 within MSG and masked with existing error handling facilities)
995 * New! Add a FATPIPE model. [AL]
996 * New! Add a parallel task model. [AL]
997 * New! Add automatically a loopback interface (in the default
998 network model) if none was precised.
1001 * Bugfix: MSG_process_resume now works with the current running process.
1003 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1004 * Modification of MSG_task_get_compute_duration. Once a task has been
1005 processed, the value returned by this function is now equal to 0. [AL]
1006 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1007 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1008 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1009 of a m_task. For now, MSG_task_cancel only works with computation
1011 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1012 of the processor (in Mflop/s) regardless of the current load on the
1014 * API Change: use proper naming convention for MSG_getClock and
1015 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1017 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1018 This function changes the priority of a computation task. This priority
1019 doesn't affect the transfer rate. A priority of 2 will make a task
1020 receive two times more cpu power than the other ones. This function
1021 has been added to suit the needs of Nguyen The Loc and hasn't been that
1022 much tested yet. So if it fails, please report it and send me your code.
1024 * API Change: removed all functions and types that were marked "deprecated"
1025 since many months. Renamed MSG_global_init_args to MSG_global_init.
1027 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1029 SimGrid (2.95) unstable; urgency=low
1032 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1033 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1034 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1037 * Bugfix: Allow absolute paths for platform description files [MQ]
1038 * Bugfix: do free the variables after use. Leads to drastic performance
1040 * Implement max_duration (ie, timeouts) on resources [AL]
1043 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1045 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1046 the needed support (timeouts) [AL]
1049 * Implement measurement sockets. You can now get the bandwidth between two
1050 hosts thanks to AMOK (see below). [MQ]
1051 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1052 dynar over the network (yeah) [MQ]
1053 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1054 * simple/static token-ring example. [Alexandre Colucci and MQ]
1055 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1056 in place before [MQ]
1059 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1060 type parsing. "short" and "long" modifiers were also ignored; other
1061 modifier (such as reference level) are still ignored. [MQ]
1062 * Embeed the buffer size within the buffer itself on SG. [MQ]
1063 That way, send() are atomic and cannot get intermixed anymore (at least
1064 the ones which are less than 100k; bigger messages still have the issue)
1065 * Array size pushed by the field, not by the field type (or each
1066 and every long int will push stuff to the cbps) [MQ]
1067 * use select() to sleep since it allows to portably sleep less than one
1070 GRAS (minor cleanups)
1071 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1074 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1075 * type category 'ignored' killed as it was never used and were difficult
1077 * whether a type can cycle or not is now a flag, leaving room for more
1078 flags (as "ignored", if we feel the need one day ;)
1079 * Rename raw sockets to measurement sockets since "raw" has another
1080 meaning in networking community.
1083 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1084 and offers features not belonging to GRAS but that most applications
1085 need. One day, it may be a set of plugins loadable at runtime.
1086 * New module: bandwidth
1087 bandwidth measurement between arbitrary nodes running this module. [MQ]
1089 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1091 SimGrid (2.94) unstable; urgency=low
1093 The first beta release of SimGrid 3 !
1097 * Update the main page and the FAQ. Adding references to gforge.
1100 * Add a gras_os_getpid function.
1103 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1104 * Extend the logs so that they also print PID, hostname, date, ... if
1106 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1107 and kill the old version which were in testsuite/
1108 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1112 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1113 for users. Instead, each programming environment (GRAS, MSG,...) use it
1114 when needed to provide such a feature to users.
1115 Don't shortcut the mecanism or you will also shortcut the virtualization
1116 you need on the simulator.
1120 * Cleanups in configury with regard to compile optimization/warning flags.
1121 Also add -fno-loop-optimize to any powerpc since it's the optimization
1122 killing gcc (< 3.4.0).
1123 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1124 needed by MSG examples complications
1125 * Borrow configury beautifications from PHP
1128 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1129 warning (thanks loris for stumbling into this one).
1130 * Bugfix: stop loading private headers (gras_config.h) from the public
1134 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1135 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1136 RL would imply a DNS resolver, which is *hard* to do in a portable way
1137 (and therefore delayed).
1138 * Implement a real timer mecanism and use it in timing macros. This allows
1139 to avoid rounding errors and get a 0.000005 sec precision in timing
1140 macros. While I was at it, various cleanups:
1141 - allow to declare more than one timed section per file (fix a stupid bug)
1142 - move some private declaration to the right place
1143 - merge conditional execution and timing macros into emulation module
1144 - document the module
1145 - make sure the module cleanups its mess on gras_exit
1146 * Documentation improvements:
1147 - (new) how to compile applications using GRAS
1148 - (new) emulation support (timing macros)
1150 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1152 SimGrid (2.93) unstable; urgency=low
1154 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1157 - Use Paje properly where used. Still to be sanitized properly.
1158 - Portability fix: Add an implementation of the contexts using pthread
1162 - Add xbt_procname(): returns the name of the current process.
1163 Use it to show the current process's name in all logging.
1165 - fix detection of older flex version and the reaction, since we do
1166 depend on modern ones (we use lex_destroy)
1167 - Better separation of SG and RL in the libs: remove all simulation code
1168 from libgras. As a result, this lib is now only 200k when stripped.
1169 Some of the xbt modules may also be duplicated (two sets and such) and
1170 should be cleaned/killed before SG3.
1171 - Insist on using xlC on AIX because of weird problems involving gcc there.
1172 - Cleanup the make remote stuff. This is now done by scripts
1173 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1174 mainly for our private use, but we're working on changing them to user
1177 - Bugfix: flush the socket on close only if there is some *output*.
1178 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1180 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1183 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1185 SimGrid (2.92) unstable; urgency=low
1187 Alpha 3 on the path to SimGrid 3
1191 - New! First try of benchmarking macros.
1192 - New! First try so that gras_stub_generator generate deployment and
1193 remote compilation helpers.
1195 - Bugfix: Initialization fix in msg_test.
1199 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1201 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1203 - API Change: gras_os_sleep to take the amount of seconds as a double.
1204 Accepting an int was error prone since it was the only location where
1205 seconds were coded as such. It leaded to damn rounding errors.
1206 - Bugfix: Hard to belive that timers ever worked before this.
1208 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1210 SimGrid (2.91) unstable; urgency=low
1212 Alpha 2 on the path to SimGrid 3
1216 - Bug fix in the lmm_solver.
1218 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1219 through the function MSG_paje_output.
1220 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1221 - It is possible to bound the rate of a communication in MSG with
1222 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1223 to write it in the changelog).
1224 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1229 - Various cleanups to the autotools stuff
1230 - Begin to move Gras examples to examples/gras/
1231 - Let make distcheck work again (yeah!)
1233 - documentation overhauled using doxygen.
1234 gtk-doc-tools is dead in SimGrid now.
1235 - Automatically extract all existing logging categories, and add the list
1236 to the documentation (long standing one, to say the less)
1238 - Cleanup the known architecture table. Reorder the entries to group what
1239 should be, and use a more consistent naming scheme.
1240 (some of the test dataset are still to be regenerated)
1241 - New! Allow library to register globals on each process just as userdata
1243 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1244 lookup time (for now).
1245 Use it in msg and trp.
1246 This cleans a lot the internals and helps enforcing privacy of the
1247 headers between the gras components.
1248 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1249 - Bugfix: gras_os_time was delirious in RL.
1250 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1252 - Reenable GRAS now that it works.
1254 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1256 SimGrid (2.90) unstable; urgency=low
1258 Alpha 1 on the path to SimGrid 3
1260 * It is a long time since the last release of SimGrid. I'm sorry about
1261 that but as I had told you, I was rewriting a lot of things. I apologize
1262 to those who had been reporting bugs to me and that I had not answered.
1263 If your bug is still in the new version, please tell me. Here is a
1264 summary of the main changes.
1266 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1267 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1268 portability, speed, and a lot more but you'll figure it out later.
1269 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1270 I wanted to release the new MSG as soon as possible and I have
1271 broken GRAS, which is the reason why, for now, only MSG is fully
1272 functional. A laconic description of these projects is available
1273 in the documentation.
1275 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1276 kernel optimized for our needs (called SURF but only the developers
1277 should use it). Hence, MSG is now roughly 30 times faster and I think
1278 that by rewriting a little bit MSG, I could event speed it up a little
1279 bit more. Beside the gain in speed, it is also much easier to encode a
1280 new platform model with SURF than it was with SG. More to come...
1282 * REVOLUTION 3: I have tried to change a little as possible the API of
1283 MSG but a few things really had to disappear. The main differences
1284 with the previous version are :
1285 1) no more m_links_t and the corresponding functions. Platforms are
1286 directly read from a XML description and cannot be hard-coded
1287 anymore. The same format is used for application deployment
1288 description. The new format is described in the documentation.
1289 Have a look in tools/platform_generation. There is a tiny script
1290 that converts from the old platform format to the new one. Concerning
1291 the application deployment format, parsing the old one is tricky.
1292 I think most of you should however be able to convert your files. If
1293 it is really an issue, I can write a C code that does the conversion.
1295 2) the toolbox tbx does not exist anymore. We now have a library
1296 with much more data-structures but without the hash-tables (we have
1297 dictionaries that are much faster).
1299 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1301 *****************************************************************************
1302 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1303 * don't feel like converting the oldies. (Mt) *
1304 *****************************************************************************
1307 Version 2.90: "the long awaited one"
1308 - Finished rewriting and debugging MSG. Rewrote the documentation.
1309 - disable GRAS for now since it needs to be ported to the newest SG
1312 - Finish the port to windows (using mingw32 for cross-compile)
1315 - Main loop and datastructures of SURF. A cpu resource object is
1316 functional. Surf can thus be used to create cpu's with variable
1317 performance on which you can execute some actions.
1319 2004-11-15 Martin Quinson
1320 - Port to ARM. Simply added the alignment and size descriptions. Should
1321 work, but the ARM machines are so slow that I didn't had the opportunity
1322 to 'make check' over there yet.
1324 2004-11-15 Arnaud Legrand
1325 - Trace manager now written. It uses a heap structure and is therefore
1326 expected to be efficient. It may however be speeded up (particularly
1327 when many events occur at the same date) by using red and black
1328 trees. One day maybe...
1329 - Max-min linear system solver written. It uses a sparse matrix
1330 structure taking advantage of its expected use. Most operations are
1331 O(1) and free/calloc are called as few as possible. The computation of
1332 the minimum could however be improved by using a red and black tree
1335 2004-11-03 Arnaud Legrand
1336 - Rename every gras_* function that was in xbt/ to its xbt_
1338 - Add a heap and a doubly-linked list to xbt
1339 - Added a dichotomy to the dictionaries. make check works as well before
1340 so I assume that the patch is correct. I do not know however if things
1341 run effectively faster than before now. :)
1343 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1344 SimGrid, and the version number is bumped to 2.x
1346 2004-10-29 Martin Quinson
1347 - Introduction of the remote errors.
1348 They are the result of a RMI/RPC on the remote machine.
1349 ErrCodes being scalar values, you can't get the host on which those
1350 errors did happen. Extending the error mechanism as in Gnome is possible.
1351 No idea yet whether it is a good idea.
1353 2004-10-28 Martin Quinson
1354 - Interface revolution: the Starred Structure Eradication.
1355 I used to do typedef struct {} toto_t; and then handle *toto_t.
1356 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1357 - ???_t is a valid type (builded with typedef)
1358 - s_toto_t is a structure (access to fields with .)
1359 - s_toto is a structure needing 'struct' keyword to be used
1360 - e_toto_t is an enum
1361 - toto_t is an 'object' (struct*)
1363 typedef struct s_toto {} s_toto_t, *toto_t;
1364 typedef enum {} e_toto_t;
1365 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1366 s_toto_t) is private.
1368 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1369 so that it can set the variable to NULL. It was so for dicts and sets,
1370 it changed for dynars.
1372 - Fix a bunch of memleaks in dict_remove
1373 - Fix a bug in sg/server_socket opening: it failed all the time.
1375 2004-10-07 Martin Quinson
1376 - Speed up dynar lookup operation a bit.
1378 gras_dynar_get is dead.
1380 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1381 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1382 (which gives you the address of the stored data).
1384 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1385 copy of the data using an affectation to do the job and not a memcpy.
1387 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1388 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1390 It does not really speedup the dynar test because they are
1391 setting elements all the time (and look them seldom). But the dict does
1392 far more lookup than setting.
1394 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1396 2004-10-05 Martin Quinson
1397 - Allow to (en/dis)able the cycle detection at run time.
1399 Whether we should check for cycle or not is now a property of each
1400 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1401 datadesc_cycle_unset allow to remove this property when previously set.
1403 Note that the cycle detection is off by default since it impacts the
1404 performance. Watch the data you feed GRAS with ;)
1406 This property is hereditary. Any element embedded in a structure having it
1407 set have it set for the time of this data exchange.
1409 You should set it both on sender and receiver side. If you don't set it on
1410 sender side, it will enter an endless loop. If you forget on receiver
1411 side, the cycles won't be recreated after communication.
1413 - Header reorganization.
1414 Kill gras_private.h, each submodule must load the headers it needs.
1416 2004-10-04 Martin Quinson
1417 - Interface revolution: do not try to survive to malloc failure.
1419 Now, gras_malloc and friends call gras_abort() on failure.
1420 As a conclusion, malloc_error is not a valid error anymore, and all
1421 functions for which it was the only gras_error_t return value are
1422 changed. They now return void, or there result directly.
1423 This simplify the API a lot.
1425 2004-09-29 Martin Quinson
1426 - Re-enable raw sockets.
1427 Created by gras_socket_{client,server}_ext;
1428 Used with gras_raw_{send,recv}
1431 It should allow to kill the last bits of gras first version soon.
1433 This is not completely satisfactory yet (duplicate code with
1434 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1437 - Simplify transport plugin (internal) interface by not passing any
1438 argument to _server and _client, but embedding them in the socket
1441 2004-09-28 Martin Quinson
1442 - Finish the port to AIX.
1443 autoconf was my problem (segfault within the malloc replacement
1444 function. No idea why)
1446 2004-09-16 Martin Quinson
1447 - Fix some size_t madness on 64bit architectures.
1449 2004-09-08 Martin Quinson
1450 - Reduce the number of system headers loaded, overload some more system
1451 calls (such as malloc to cast the result of the system one, and work
1453 - Fix and reintroduce the config support
1455 2004-09-07 Martin Quinson
1456 - Source code reorganization to allow Arnaud to surf all over there.
1457 - Allow to document the logging categories.
1458 - Remove all uppercase from logging categories and useless cleanup in names.
1460 2004-08-18 Martin Quinson
1461 Version 0.6.2 (protocol not changed; API changed)
1462 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1463 gras_error_t), and NULL when not found. Functions expecting a msgtype
1464 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1465 hopefully usefull message.
1466 - Portability to prehistoric sparcs again
1468 2004-08-17 Martin Quinson
1469 Version 0.6.1 (protocol not changed; ABI not changed)
1470 - prealloc some buffers to speed things up
1472 2004-08-11 Martin Quinson
1473 Version 0.6 (protocol not changed; ABI expended)
1474 - The parsing macro can deal with the references, provided that you add
1475 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1477 2004-08-09 Martin Quinson
1478 Version 0.5 (protocol not changed; ABI changed)
1479 - Allow to off turn the cycle detection code in data exchange at
1480 compilation time. It should be at run time, but I'm short of time (and
1481 the config stuff is still broken). That way, we keep dict out of the
1482 critical path, which is good because the performance is poor:
1483 - search not dichotomial yet
1484 - dynar give no way to access their content and memcpy everytime
1485 - In composed data description (struct, ref and so on), stop foolness of
1486 keeping the subtype's ID, but store the type itself. This keeps sets out
1487 of the critical path, which is good since they rely on dynar and
1488 dictionnaries. The only loose of that is that we cannot detect the
1489 redeclaration of a structure/union with another content (but I'm not sure
1490 the code detected well this error before anyway). We still can detect
1491 the redefinition discrepancy for the other types.
1492 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1493 it breaks the code because of type-punning used all over the place).
1494 This breaks on all non-gcc architectures (for now).
1496 All those changes (plus the buffer of last time) allow me to gain 2 order
1497 of magnitude on cruel tests consisting of 800000 array of integers on two
1498 level of a hierarchical structure (200 secondes -> 4 secondes)
1501 - the selector of reference must now return the type it points to, not
1502 the ID of this type.
1504 2004-08-06 Martin Quinson
1505 Version 0.4 (protocol changed; ABI not changed)
1506 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1507 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1508 - Implement a buffer transport, and use it by default (it relies on tcp in
1509 real life and on sg in simulation).
1510 That's a bit hackish since I had a new field to the structure to store
1511 its data without interfering with the subtype ones. Inheritance
1512 is tricky in C. And that's a kind of reverse inheritance with one class
1513 derivating two classes. Or maybe a game with java interfaces. Anyway,
1514 that's damn hard in C (at least).
1515 Moreover, I got tired while trying to ensure plugin separation and
1516 genericity in SG mode. MSG wants me to do weird things, so let's go for
1517 cruel hacks (temporarily of course ;).
1518 See comment in transport_private.h:71
1519 - do not include all the _interface headers in private but in the files
1520 which really need them (to cut the compilation time when they are
1523 2004-07-26 Martin Quinson
1524 Version 0.3 (protocol not changed; ABI changed)
1525 - Major overhault of the datadesc interface to simplify it:
1526 - shorted the function names:
1527 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1528 - add a trivial way to push/pop integers into the cbps without malloc.
1529 This allows to make really generic sub_type description, which simply
1530 pop their size of the stack.
1531 - add a function gras_datadesc_ref_pop_arr() which does what users want
1532 most of the time: Declare a dynamic array (which pops its size of the
1533 stack) and declare a reference to it. Poor name, but anyway.
1534 - kill the post-send callback, add a post-receive one
1536 2004-07-23 Martin Quinson
1537 Version 0.2 (protocol changed; ABI changed)
1538 - add some testing for cpbs in the test cases, and fix some more bugs.
1539 This invalidate again the little64 data file, since I cannot regenerate
1541 - remove an awfull optimization in the logging stuff, allowing me to:
1542 - understand it again
1543 - learn gcc how to check that the argument match the provided format
1544 - fix all errors revealed by gcc after that
1545 - internal keys of dict are not \0 terminated. Deal with it properly in
1546 loggings instead of segfaulting when the user want to see the logs :-/
1548 2004-07-22 Martin Quinson
1549 - Fix some stupid bug preventing cbps (callback postit) from working
1551 2004-07-21 Martin Quinson
1552 - Some documentation cleanups
1553 - remove the useless last argument of msgtype_declare
1554 - rename the Virtu functions to fit into the 'os' namespace
1555 - move headers src/include -> src/include/gras/ and stop fooling with
1556 gras -> . symbolic link
1557 - make distcheck is now successful
1559 2004-07-19 Martin Quinson
1561 - Build shared library also
1562 - Install html doc to the right location
1563 - stop removing maintainer files in make clean
1564 - build tests only on make check
1566 2004-07-13 Martin Quinson
1568 - No major issue in previous version => change versionning schema
1569 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1570 corresponding dataset.
1572 2004-07-11 Martin Quinson
1574 - Get it working with any kind of structure (we can compute the padding
1575 bytes remotely for all the architectures I have access to)
1576 - Implement the structure parsing macro (still not quite robust/complete)
1577 - Improvement to the remote testing toysuite
1579 2004-07-10 Martin Quinson
1580 [autoconf mechanism]
1581 - get ride of a bunch of deprecated macros
1582 - actually run the test for two-compliment, not only compile it :-/
1583 - test whether the structures get packed (and bail out if yes. Damn.
1584 Alignment is a serious matter)
1585 - test whether the structures get compacted (but respecting the alignment
1586 constraints of each types)
1587 - test whether the array fields of structures can straddle alignment boundaries
1589 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1590 (took me 2 hours to find that bug, looking at the wrong place)
1591 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1592 before sending/receiving and are used to compute the offsets of fields
1593 - Given that padding size depend even on compiler options, keep track of
1594 alignment and aligned_size only for the current architecture. Not a big
1595 deal since we send structure fields one after the other (seems
1597 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1598 program, let it work on linux/gcc/little32. portability todo.
1600 2004-07-08 Martin Quinson
1601 - import and improve remote compilation support from FAST
1602 - make sure make check works on half a dozen of machines out there
1604 2004-07-07 Martin Quinson
1605 Let's say it's version 0.0.3 ;)
1606 - Implement conversions (yuhu!)
1607 - Let it work on solaris (beside conversion, of course)
1608 - Stupid me, using rand() to generate the conversion datatests in not wise.
1610 2004-07-06 Martin Quinson
1611 - Let make dist work, since I'm gonna need it to compile on remote hosts
1612 - Let Tests/datadesc_usage write the architecture on which the file was
1613 generated as first byte.
1614 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1615 architecture descriptions.
1616 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1617 on those architectures.
1618 - Optimization: send/recv array of scalar in one shoot
1620 2004-07-05 Martin Quinson
1621 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1623 - Plug a whole bunch of memleaks
1624 - each process now have to call gras_{init,exit}. One day, their log
1625 settings will be separated
1626 - Continue the code factorisation between SG, RL and common in Transport.
1628 2004-07-04 Martin Quinson
1630 - Redistribution between SG and RL.
1631 We wanna have to accept in SG, so move accepted related parts of RL in
1632 the common part. (more precisely, the dynar of all known sockets is no
1633 more a static in transport.c, but part of the process_data)
1635 [gras_stub_generator]
1636 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1638 2004-07-03 Martin Quinson
1639 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1640 them all up in one shot)
1642 - Fix the ugly case of reference to dynamic array.
1643 - New (semi-public) function gras_datadesc_size to allow the messaging
1644 layer to malloc the needed space for the buffer.
1646 - gras_socket_close now expect the socket to close (and not its address to
1647 put NULL in it after it). This is because the socket passed to handlers
1648 is one of their argument (=> not writable).
1650 - propagate the interface cleanup from last week in datadesc, ie remove a
1651 superfluous level of indirection. User pass adress of variable
1652 containing data (both when sending and receiving), and not of a variable
1653 being a pointer to the data. Let's say that I like it better ;)
1654 The price for that is constructs like "int msg=*(int*)payload" in
1655 handlers, but it's a fine price, IMHO.
1657 - Let it work in RL (yuhu)
1659 2004-06-21 Martin Quinson
1661 - porting SG plugin and SG select to new standards (works almost).
1662 - plug memleaks and fix bugs around.
1665 - cleanup the prototype of data recv and force users to specify when they
1666 want to handle references to objects. Test case working even for cycles.
1667 - plug memleaks. Valgrind is perfectly ok with this.
1669 2004-06-12 Martin Quinson
1671 - cleanup the separation between plugin and main code in plugin creation
1673 2004-06-11 Martin Quinson
1675 - Reput hook for raw sockets, needed for BW experiments
1676 - kill a few lines of dead code
1677 [Data description] Interface cleanup
1678 - gras_datadesc_by_name returns the searched type or NULL.
1679 That way, no variable is needed to use a type desc once, which makes
1681 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1682 two parameters were strings (field name, type name), leading to
1684 [Dicos] Interface cleanup
1685 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1686 This is consistant with the dynar API.
1688 2004-04-21 Martin Quinson
1690 - Porting to new standards.
1692 - interface cleanup.
1693 There is no bag anymore, no need to take extra provision to mask the
1694 pointers behind "ID".
1695 Better splitup of functions between files create/exchange/convert.
1696 This is still a bit artificial since convert and receive are so
1697 interleaved, but anyway.
1699 - add a queued message list to procdata (the ones not matching criteria
1701 - factorize some more code between SG and RL wrt procdata
1703 - use gras_exit in example to track memleaks
1704 - get rid of gs_example now that GS is properly integrated into gras
1705 - update run_test to integrate the lastest tests (datadesc)
1707 - rename WARNINGn macros to WARNn since it prooved error-prone
1709 2004-04-19 Martin Quinson
1711 - register init/exit functions within gras module mechanism
1712 - send/receive function.
1713 Convertion is not implemented, but short-cutted if not needed.
1714 struct/array elements are sent one by one (instead of block-wise), but
1715 nobody really cares (yet). Get a prototype before optimizing.
1716 - tests (using a file socket) for DD send/receive on:
1717 - base types: int, float
1718 - array: fixed size, string (ie ref to dynamic string)
1719 - structure: homogeneous, heterogeneous
1720 - chained list, graph with cycle
1721 Believe it or not, valgrind is not too unhappy with the results. The
1722 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1726 - Bugfix when using a filename explicitely (instead of '-')
1728 2004-04-09 Martin Quinson
1730 - factorize more code between RL and SG in socket creation
1731 - Complete the implementation and tests of:
1733 o file (only in RL, and mainly for debugging)
1735 I lost 3 days to design a portable address resolver, and then decided
1736 that the prototype mainly have to run on my box.
1737 Addressing portability too early may be like optimizing too early :-/
1739 - use gras_init in the Tests instead of the crappy parse_log_opt
1740 (the latter function is removed)
1741 [Conditional execution]
1742 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1743 [Code reorganisation]
1744 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1745 Build examples against the RL library, since there is no way to disable
1746 its creation for now.
1748 For information, the beginning of coding on GRAS was back in june
1749 2003. I guess that every line has been rewritten at least twice since