1 SimGrid (3.20) NOT RELEASED YET (Release target: June 21. 2018 10:07 UTC)
4 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
5 behavior tracing is no longer limited to MSG
8 - Introduced new function simgrid::s4u::Host::get_actor_count. This function
9 returns the number of actors running on a specific host.
12 - Allow to run the Link energy plugin from the command line with
13 --cfg=plugin:link_energy
14 - Rename Energy plugin into host_energy
15 - Rename Load plugin into host_load
18 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
19 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
22 - Replay: The replay file has been re-written in C++.
23 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
24 supported. They are stored in the trace and used when replayed.
27 - Config: the C API is now deprecated (will be removed in 3.23), and
28 the C++ API has been slightly improved.
31 - Move simgrid_config.h to simgrid/config.h (old header still working)
34 - #143: Setting a breakpoint at a given time
35 - #258: daemonized actors hang after all non-daemonized actors have completed
36 - #267: Linker error on unit_tmgr
37 - #269: SMPI : tracing of MPI_Wait/all/any broken
39 ----------------------------------------------------------------------------
41 SimGrid (3.19.1) Released March 22 2018
43 The Fixed ABI Release.
45 As you may know, we are currently refactoring SimGrid in deep.
46 Upcoming SimGrid4 will be really different from SimGrid3: modular,
47 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
49 Our goal is to smooth this transition, with backward compatibility
50 and automatic update paths, while still progressing toward SimGrid4.
52 SimGrid remains open during works: The last pure SimGrid3 release was
53 v3.12 while all subsequent versions are usable alpha versions of
54 SimGrid4: Existing interfaces remain unchanged, but the new S4U
55 interface is budding and the internals are deeply reorganized.
57 Since 2015, we work hard to reduce the changes to public APIs. When
58 we need to rename a public library symbol in S4U, we let your
59 compiler issue an explicative warning when you use the deprecated
60 function. These messages remain for four releases, i.e. for one full
61 year, before turning into an error. Starting with v3.15, your can
62 also adapt to API changes with the SIMGRID_VERSION macro, that is
63 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
65 Starting with this v3.19.1, our commitment to reduce the changes to
66 the public interfaces is extended from the API to the ABI: a program
67 using only MSG or SimDag and compiled against a given version of
68 simgrid can probably be used with a later version of SimGrid without
69 recompilation. We will do our best... but don't expect too much of
70 it, that's a really difficult goal during such profund refactoring.
72 The difference between v3.19 and v3.19.1 is that the former was
73 accidentally breaking the ABI of MSG, while the later is restoring
76 S4U and kernel APIs will still evolve until SimGrid4, with one-year
77 deprecation warnings as currently. In fact, cleaning up these
78 interfaces and converting them to snake_case() is one release goal of
79 v3.20. But don't worry, we are working to smooth this upgrade path.
81 In summary, new projects should start with S4U to benefit of the
82 future, but old MSG projects should still be usable with no change.
84 ----------------------------------------------------------------------------
86 SimGrid (3.19) Released March 20 2018
88 The Moscovitly-cold Spring release.
90 SMPI user-visible improvements:
91 - Support of MPI_Type_create_subarray
92 - Unimplemented MPI calls now abort (instead of issuing a warning)
93 Some harmless calls around errhandlers still only issue a warning.
94 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
95 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
96 is now possible (mmap privatization was refactored to this end).
97 Adding new SMPI instances should also be technically doable now.
99 SMPI internal cleanups:
100 - Removed index notion from SMPI -- no more "getPid() - 1"!
101 In tracing files, actors are called rank-<PROCESS-ID> but these are
102 NOT the real ranks (tracing several communicators is not supported yet).
103 - Deprecate smpi_process_index() (will be removed in 3.22)
104 - Better testing on classical proxy apps, through an external project:
105 https://github.com/simgrid/SMPI-proxy-apps/
108 - Use a graphical TOC to make it easier to find the documentation you need
109 - Revamp the MSG tutorial
112 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
113 - Remove parameter of MSG_process_killall().
114 Resetting the PID was bogus anyway (several actors could have the same PID).
117 - Execution->setHost() can be called after start(), to migrate it.
118 - Comm::test_any() is now implemented.
119 - s4u::Actor now has onCreation() and onDestruction() signals.
120 - Install some kernel header files for the users' plugins and more.
121 Warning, their API is really not stable yet.
122 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
123 include/simgrid/kernel/routing/*Zone.hpp
126 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
128 Build System and other cleanups:
129 - Remove unused run-time parameter "tracing/onelink-only".
130 - SimGrid now works with intel compilers (MC still buggy)
131 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
132 - tesh kills subprocesses on timeout (but not on windows)
135 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
138 - #194: Feature request: simgrid::s4u::Comm::test_any()
139 - #245: migrating an actor does not migrate its execution
140 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
141 - #256: Modernize FindSimGrid.cmake
142 - #257: Fix (ab)use of CMake install
144 ----------------------------------------------------------------------------
146 SimGrid (3.18) Released December 24 2017
148 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
150 MSG IS NOW DEPRECATED. Please use S4U for new projects.
151 - Support for MSG should not be removed from SimGrid before 2020,
152 but future evolutions will be limited to the integration of
153 user-provided patches.
156 - s4u::onDeadlock() signal, run before stopping the simulation.
157 - s4u::Exec: asynchronous executions (abstraction of background
159 - s4u::Host->getLoad() returns the achieved speed in flops/s
162 - Link::name() is deprecated and will be removed in v3.21, use
163 Link::getCname() instead.
164 - Mailbox::getName() changed to return a std::string, use
165 Mailbox::getCname() instead to get a char*.
166 - Storage::getName() changed to return a std::string, use
167 Storage::getCname() instead to get a char*.
168 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
169 to have both allocation and deallocation of the map in user space.
172 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
173 The amount of flops remaining to do is a vector, not a scalar.
174 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
175 name implies on both sequential and parallel tasks.
176 - Both changes fix GitHub's #223 using PR #237 as a basis.
177 Thanks Michael Mercier.
178 - Most examples were converted to S4U and hidden elsewhere at tests.
179 Rationale: we still want MSG to work; we want newcomers to use S4U.
182 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
183 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
186 - Switch to the faster dlopen privatization mechanism by default
187 - Documentation improvments
190 - Live migration is getting moved to a plugin. Dirty page tracking is
191 the first part of this plugin. This imply that VM migration is now
192 only possible if one this function is called :
193 - C/MSG: MSG_vm_live_migration_plugin_init()
194 - C/C++: sg_vm_live_migration_plugin_init()
195 - Java: Msg.liveMigrationInit()
196 For C and C++, "simgrid/plugins/live_migration.h" also has to be
200 - Define class simgrid::xbt::Path to manage file names.
201 - Removed unused functions:
202 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
203 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
204 xbt_graph_export_graphviz()
205 - xbt/str.h: xbt_str_join()
206 - Remove unused datatypes:
207 - xbt/heap.h: use std::priority_queue or boost::heap instead
208 - xbt/swag.h: use boost::intrusive::list instead
211 - New link_energy plugin for the consumption of the links.
212 - All of the operations on files and storage contents have been
213 packaged into a plugin (src/plugins/file_system). The current
214 public interface can be found in
215 include/simgrid/plugins/file_system.h
216 To use these functions you now have to initialize the plugin by
217 calling MSG_storage_file_system_init() just after calling
218 MSG_init() or sg_storage_file_system_init() just after creating
222 - Remove the undocumented/untested tag <include>
225 - Remove viva specific tracing as the tool is no longer maintained
228 - #248: Kill of finished processes leads to segfault
229 - #240: xbt_cond_wait_timeout should gracefully return for C apps
230 - #239: please implement signal s4u::onDeadlock()
231 - #230: segfaults when exit() without run()
232 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
233 - #223: MSG_task_get_flops_amount() not working with parallel tasks
234 - #222: Actor::kill() doesn't really kill and segfaults
235 - #221: odd LMM warning when killing an actor
236 - #120: Memory leak when the processes are forcefully killed
238 ----------------------------------------------------------------------------
240 SimGrid (3.17) Released October 8 2017
242 The Drained Leaks release: (almost) no known leaks despite the tests.
244 Even very long simulations will preserve your memory: our testsuite
245 stresses SimGrid for over 45mn on fast machines for a coverage of
246 over 80%, and there is only one single known leak, of about 4kb.
250 - Comm.detach(): start and forget about asynchronous emission. A cleanup
251 handler may be given to free resources if the comm cannot be completed.
252 - this_actor::send(mailbox) is now mailbox->put()
253 - New: simgrid::s4u::Comm::wait_all()
254 - New: Host.execute() for remote executions.
257 - Unused option network/sender-gap is removed.
260 - Deprecate MSG_task_isend_with_matching(): This unused feature
261 really complicates our internals. Will be removed in v3.20.
264 - Improved context termination. It is now possible to free resources, even
265 when a process is forcibly killed.
266 - ContextBoost: add support for Boost versions above 1.61.
269 - Bring back run-time option --cfg=exception/cutpath to remove exception
271 - Removed unused functions:
272 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
273 xbt_str_rtrim(), xbt_str_trim().
274 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
277 - Removed header files obsolete since SimGrid 3.12:
278 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
280 - #3: SD_exit should be made optional
281 - #120: Memory leak when the processes are forcefully killed
282 - #159: Threading test regression in Actor refcounting
283 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
284 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
285 - #186: Actor::killAll() segfaults if some process is blocked on wait()
286 - #191: VM migration and pstate
287 - #192: Updating the energy consumptions of all hosts crashes with VMs
288 - #195: All actors have PID=0 in the logs
289 - #204: Sometimes segfault with thread contexts and mmap privatization
290 - #222: Actor::kill() doesn't really kill and segfaults
291 - #225: Actor::kill() doesn't really kill when victims are doing a join()
293 ----------------------------------------------------------------------------
295 SimGrid (3.16) Released June 22. 2017.
297 The Blooming Spring Release: developments are budding.
300 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
301 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
302 - Java JAR file should be finally fully working on Mac OSX too.
304 XML platforms: Switch to platform v4.1 format.
305 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
306 - <zone> can be used as a synonym for the now deprecated <as>
307 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
308 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
309 - <actor> can be used as a synonym for the now deprecated <process>
310 - state_file and avail_file periodicity is now easier to express
311 (check the documentation) the old behavior should still work.
312 * Storage: not backward compatible, but it seems that nobody ever used it
313 - Remove attribute 'content_type' of <storage_type>: was never used
314 - Make attribute 'model' of <storage_type> optional: for future usage
315 - Remove Bconnection model property: was never (in)validated. Replaced by
316 the maximum of the read and write bandwidth as a resource constraint.
319 - New and Backwards Compatibility break:
320 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
321 When one wants to get the list of tasks whose states have changed during a
322 simulation round, s/he has to allocate and free a dynar and use it as
323 argument to this function. The former SD_simulate (double how_long)
327 - Allow multicore VMs, with the correct sharing (unless you overcommit)
328 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
331 - New function to update the consumption of all hosts at once.
332 - Fix the model for multi-core hosts, linear in the amount of busy
333 cores with an abnormality for pIdle (see paper).
334 BUG: Still not satisfactory for ptask on mono-cores.
337 - The netzone are now available from the MSG API.
338 The old names still work, but are now deprecated.
341 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
342 - New option: smpi/keep-temps to not cleanup temp files
343 - New option : smpi/shared-malloc-blocksize . Relevant only when global shared
344 mallocs mode is used, allows to change the size of the fake file used
345 (default 1MB), to potentially limit the number of mappings for large runs.
346 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
347 - Fortran ifort and flang compilers support
348 - New RMA calls supported (experimental) :
349 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
350 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
351 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
352 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
353 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
354 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
357 - New: this_actor::isMaestro() returns whether we are in kernel mode.
358 - Behind the scene, ActivityImpl are now automatically refcounted.
359 Removing this nasty bottleneck will greatly simplify our internals.
362 - Replay: New function xbt_replay_action_get():
363 Retrieve the function previously associated to an event type.
364 - DROPPED FUNCTION: xbt_str_varsubst()
365 - DROPPED MODULE: strbuff. We don't need it anymore.
366 - DROPPED MODULE: matrix. We don't need it anymore.
367 - DROPPED MODULE: lib. We don't need it anymore.
369 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
371 SimGrid (3.15) Released March 22. 2017
373 The Spring Release: continuous integration servers become green
374 We fixed even the transient bugs on all target architectures:
375 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
376 Windows, FreeBSD, NetBSD.
378 - Rename NetCards to NetPoints.
379 This was intended to help NS3 users, but that's not a netcard.
380 That's a point in the routing algorithm, let's avoid wrong simplifications.
383 - New: MSG_process_yield(). Stop and yield to other processes.
384 - New: MSG_process_daemon(). Daemon processes are automatically killed
385 when the last non-daemon process terminates
386 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
387 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
388 to make room for the upcoming network energy plugin.
389 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
392 - Ensure that an actor can kill itself with Process::exit()
393 - Kill the obscure NativeException. Nobody want to survive the issues
394 it denotes, so use JniException that is a RuntimeException (not to
395 be caught explicitly).
396 - Partial bug fix in initialization. SimGrid flags on command line were
397 consumed at C level but stayed in the original Java String[] args.
398 This could mess users' args[i] if SG flags were not put at the end of
400 The SimGrid flags are now removed from the Java arguments. However,
401 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
402 if args.length is greater than the number of YOUR OWN ARGUMENTS.
403 It might be if you have --log or --cfg flags in the command line.
404 - Fix numerous memleaks all around the place. In particular, around VMs.
408 - s4u::onPlatformCreated: right before the simulation starts
409 - s4u::onSimulationEnd: right after the main simulation loop
410 - s4u::onTimeAdvance: right after a clock change (time is discrete)
411 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
412 event from the availability_file changes the avail speed.
413 - Links are now usable from s4u
414 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
415 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
416 - New examples: The conversion of MSG examples to S4U has begun
417 - Actors: create, kill, migration, and suspend
418 - Applications: master-worker and token-ring
419 - Action replay: communications and storage
420 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
423 - Backwards Compatibility breaks
424 - The SD_route_get_list and SD_route_get_size functions have been removed.
425 They are replaced by the unique
426 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
427 The route from src to dst is built in the links dynar whose size can
428 be obtained with xbt_dynar_length.
429 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
430 sg_host_route_bandwidth and sg_host_route_latency.
431 Macros ensure the backwards compatibility, but you should fix your code
434 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
435 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
436 - Bump our claim of support from MPI 1.1 to MPI 2.2.
437 We don't support 100% of it, but it should be enough. Contact us if not.
438 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
439 naive and probably inaccurate)
440 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
441 mapped onto a unique small file using some system magic.
442 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
443 - Fortran: cleanups, fixes, support of user-added operations
444 - RMA: MPI_Accumulate are applied in correct order.
445 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
446 - Better support for MPI_IN_PLACE
447 - Support for MPI_Win attrs and keyvals.
448 - Support MPI_Comm_group_create, MPI_Type_size_x
449 - MPI ops should be properly applied to all allowed types
450 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
451 algorithm by a non failing one from mpich.
454 - Kill the fifo data container: we don't use it anymore.
457 - A <cluster> can now be created with different speed values to represent pstates.
458 see examples/platforms/energy_cluster.xml for an example.
460 SimGrid (3.14.159) Released December 28. 2016
462 The previous release was only a crude approximation of a Pi release;
463 this one should be a bit better.
465 - Revert a last minute change that broke on Mac OSX.
466 - Fix the changelog and NEWS files.
467 - Improve the documentation, in particular of the routing module.
468 - Integrate some patches flying around in Debian.
470 SimGrid (3.14) Released December 24. 2016
473 * The whole documentation was reworked and reorganized. There is
474 still much room for improvement here, but we're on it.
477 * We now need python3 for our testing infrastructure.
478 * Model checking needs libevent
481 * Now works on FreeBSD
484 * Privatization now works on FreeBSD
486 * Privatization is now activated by default in smpirun
487 You can turn it off with -no-privatize if something goes wrong.
489 * Call-location tracing for SMPI.
490 You can add the exact location (filename / linenumber) of an MPI call to
491 your trace files and slow down or speed up the simulation between two
492 consecutive calls by using an adjustment file (see the documentation).
494 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
495 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
496 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
497 values were ignored in some cases. The timings of these functions can now
498 be significantly different.
500 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
501 smpi/running-power is renamed to smpi/host-speed
503 * smpi/grow-injected-times option to enable or disable multiplication of the
504 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
505 make simulation less precise (but also much faster).
507 * smpirun script should be (much) faster for large deployments.
509 * SMPI tracing : fixed issue with poor matching of send/receives.
511 * Replay : Fix broken waitall
513 New functions and features
514 * MSG_parallel_task_execute_with_timeout, to timeout computations.
516 Dropped / renamed functions and features
517 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
518 - MSG_mailbox_is_empty() -> Mailbox::empty()
519 - MSG_mailbox_front() -> Mailbox::front()
520 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
521 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
522 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
523 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
524 * Don't pass the free_f parameter to property related functions:
525 - MSG_host_set_property_value()
526 - MSG_as_router_set_property_value()
527 - MSG_storage_set_property_value()
528 * VM properties. Since msg_vm_t are msg_host_t, just use
529 MSG_host_get_property_value() and friends
530 * VM I/O related things:
531 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
532 - Unimplemented save/restore methods
533 * MSG_as_router_get_property_value() was redundent with
534 MSG_environment_as_get_property_value().
535 - Removed MSG_as_router_*propert*() functions
536 - Added MSG_environment_as_set_property_value() for consistency
537 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
538 Well, they are still there for now, but deprecated with a warning.
539 Please switch to xbt_dict_new_homogeneous() before this is removed
541 * Task affinity. Its intended behavior (that was very badly tested
542 and probably not really working) was deceiving what most users
543 would have hoped here.
544 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
545 * Option network/coordinates is now useless and should be dropped.
548 * Added option storage/max_file_descriptors to allow more than 1024 files opened
551 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
552 * Added Dragonfly topology support
555 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
556 EiBps, EBps, Eibps, Ebps for bandwidth.
557 They may become useful to some lucky ones.
560 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
561 * ex/app_tokenring: new example, very similar to the MSG Token Ring
562 * ex/async_waitAll: new example, on asynchronous communications
565 * Memory usage should be decreased for simulations with a large number
566 of processes. This also helps for SMPI.
568 SimGrid (3.13) stable; urgency=low
570 The Half Release, a.k.a. the Zealous Easter Trim.
572 - We removed half of the lines, that were mostly experimental cruft.
573 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
574 (not counting blanks and comments -- according to openhub.net)
575 - The internals are now compiled in C++ (and will soon be clean C++)
576 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
577 59 klines of C, etc. We added only 29 klines of C++ in replacement.
579 * Backwards Compatibility breaks
580 - Removed Lua simulation bindings (switch to C or Java for that).
581 Lua can still be used to describe platforms
582 - Removed Java kernel plug-ins.
583 Will be reintroduced after the ongoing major internals reorg.
585 - the following functions were removed.
586 They were too specific and should be reimplemented in a generic
587 way, with filter function.
588 - MSG_task_listen_from_host
589 - MSG_mailbox_get_count_host_waiting_tasks
590 - MSG_mailbox_put_with_timeout was removed.
591 Please use MSG_task_send_with_timeout instead.
593 - the SD_application_reinit function was removed. It has been a noop for a while.
594 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
595 replaced by a more flexible mechanism.
596 - The following functions thus do not exist anymore
597 - SD_workstation_get_access_mode
598 - SD_workstation_set_access_mode
599 - SD_workstation_get_current_task
600 - Basic estimation functions have been removed but can easily be replaced
601 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
602 - SD_workstation_get_computation_time => amount / sg_host_speed()
604 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
605 Use VM.getSpeed()*load/100 for the legacy behavior.
607 - option enable_tracing was removed. It was not doing anything for a while.
608 - In the ModelChecker:
609 - the model-checker now ptraces the model-checked process which means
610 you cannot use a debugger on the latter anymore (we might make this
611 optional in the feature);
612 - removed soft-dirty page tracking;
613 - remove model-checked side snapshot management,
614 MC_snapshot() and MC_compare_snapshot();
615 - keep the MC_cut() function as a stub (it was not really working
616 in the previous release).
619 * All options are consistently in kebab-case. Old names are kept as alias.
622 * Switch to platform v4 format.
623 - Rename from 'power' to 'speed' the attributes describing the amount of
624 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
625 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
626 - In <host> and <link>, attributes availability and state are gone.
627 It was redundent with state and availability traces, and with peak values.
628 - In <cluster>, attributes availability_file and state_file are gone.
629 It was too complex and unused.
630 - Kill <gpu>. Was not doing anything.
631 - The DOCTYPE points to the right URL:
632 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
633 (the file at this address now documents the changelog since its v1)
634 - A warning is emitted for unit-less values (they are still accepted).
635 - speed. Default: 'f' or 'flops'. Also defined:
636 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
637 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
638 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
639 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
640 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
641 - latency. Default: 's' second. Also defined:
642 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
644 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
645 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
648 * s4u::Host is now the preferred public interface to the Host features.
649 sg_host_* functions are C bindings to the exact same behavior
650 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
653 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
654 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
655 They were deprecated since a few years, and probably did not even compile anymore.
658 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
659 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
660 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
661 encouraged to update their codes. A list of the modified functions can be found at the end of
662 include/simgrid/simdag.h
665 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
666 the lambda or closure passed as a parameter will run in kernel mode.
667 All the callback functions should be rewritten to that interface at some point.
670 * Reorganizing and cleaning the internals all around the place.
673 * Remove old default barrier/bcast buggy algorithms (see #18407)
674 * Various bug fixes to handle more codes
675 * Remove the need for the --foreground option of smpirun (it is still
676 accepted for backward compatibility).
679 * Kill the setset data container: MC don't use it anymore.
680 * Kill the queue data container: it made more sense with GRAS.
681 * Kill the xbt_peer_t data type: it's useless without GRAS.
682 * Kill rm_cb feature of config sets: it was never useful.
683 * Kill graphxml parsing feature. It was not used.
684 * Kill the deprecated code protected by XBT_USE_DEPRECATED
686 - xbt_dynar_sort_strings(), when the content is char*
687 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
688 They throw exceptions on invalid input;
689 * C++ support for declaring CLI flags (simgrid::config::Flag);
690 * class for abstracting different signal backends (simgrid::xbt::signal).
691 with no external dependencies (we need very simple signals).
694 * refactoring and cleanup of the code;
695 * ongoing process to cleanly separate the model-checking algorithms
696 from the code model-checking support.
698 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
700 SimGrid (3.12) stable; urgency=low
702 The Facelift Release.
705 * Require g++ v4.7 at least to not speak prehistorical C++.
706 * Require Boost 1.48 (for signal2 component).
707 * Java must be version 7 at least when activated.
708 * Builds on Windows again (including Java bindings).
709 * Tracing is now always enabled (no way to turn it out)
710 * Remove GTNetS. It was not working anyway.
711 * Various cleanups in the cmake scripts.
712 * Move headers around to sort them out on installed systems:
713 - instr/instr.h -> simgrid/instr.h
714 - instr/jedule/* -> simgrid/jedule
715 - simdag/datatypes.h was removed
716 - simdag/simdag.h -> simgrid/simdag.h
717 - msg/datatypes.h was removed
718 - msg/msg.h -> simgrid/msg.h
721 * Interface improvement:
722 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
723 - Sanitize the interface in MSG_task_ module:
724 - Merge two functions that were close enough but misleading:
725 set_compute_duration(t) -> set_flops_amount(t)
726 get_remaining_computation(t) -> get_flops_amount(t)
727 - set_data_size(t) -> set_bytes_amount(t)
728 get_data_size(t) -> get_bytes_amount(t)
729 - Massive cleanups in the functions related to the energy
730 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
731 - New: MSG_host_get_pstate()
733 - msg/energy/onoff: switching hosts on and off
735 * Interface improvement:
736 - Rename Host.isAvail() to Host.isOn()
737 - Rename Process.currentProcess() to Process.getCurrentProcess()
738 - Rename Task.setDataSize() to Task.setBytesAmount()
739 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
741 - #18874: Actually allows the GC to reclaim tasks
745 - SIMIX_process_throw: raises an exception in a remote process
746 * Refactoring: Separate sub-modules
747 - libsmx: the public interface, as libc in a real system
748 - popping: the strange dance that converts a user request into a kernel handling
749 - smx_context_*: the virtualization mechanisms that embed the user code
750 - smx_*: the handling of each simcalls
752 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
753 * Rename smx_action_t into smx_synchro_t, making explicit that these
754 things are used to synchronize processes with their environment.
755 For example, a communication is a sort of synchronization involving
756 the communicating processes (that may block until the exchange) and
757 the platform. The same can be said from computations, etc.
759 - #18888: segfault when a process ends before its kill_time
763 - Onesided early support for : MPI_Win_(create, free, fence, get_name, set_name, get_group), MPI_Get, MPI_Put, MPI_Accumulate, MPI_Alloc_mem, MPI_Free_mem.
764 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
765 - MPI_Type_set_name, MPI_Type_get_name
766 - MPI_*_c2f and MPI_*_f2c functions
767 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
768 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
769 - Activate a lot of new tests from the mpich 3 testsuite
771 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
772 - InfiniBand network model added : Based on the works of Jerome Vienne
773 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
774 - When smpi/display_timing is set, also display global simulation time and application times
775 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
776 * Collective communications
777 - SMP-aware algorithms are now dynamically handled. An internal communicator is created for each node, and an external one to handle communications between "leaders" of each node
778 - MVAPICH2 (1.9) collective algorithms selector : normal and SMP algorithms are handled, and selection logic is based on the one used on TACC's Stampede cluster (https://www.tacc.utexas.edu/stampede/).
779 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
781 - Replay now uses algorithms from wanted collective selector
782 - Replay can be used with SMP-aware algorithms
783 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
784 - Replay can now replay several traces at the same time (check examples/smpi/replay_multiple example), to simulate interactions between several applications on a given platform. User can specify the start time of each instance. This should also allow replay + actual applications to run.
786 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
787 - Correctly use loopback on fat-tree clusters
788 - Asynchronous small messages shouldn't trigger deadlocks anymore
789 * Energy/DVFS cleanup and improvement
790 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
791 - new: smpi_get_host_pstate()
794 - "Full" network optimization flag was broken since Surf++
795 - Better handling of precision flags in maxmin
796 - Fix bug causing sometimes "Impossible" errors
797 - Properly pass cluster properties to included hosts
798 * Improvement of the Energy plugin.
799 - Always update the consumption before returning that value
800 - New property: watt_off to denote the dissipation when the host is off
801 - New functions getWattMinAt and getWattMaxAt to retrieve the
802 dissipation of pstates that we are not currently at.
803 * Java: class NetworkLink renamed to Link
804 * New function: simcall_process_get_kill_time()
805 * Massive rename s/workstation/host/
806 - That's intrusive, but that's good for the project consistency. Sorry.
807 - Change config option "workstation/model" into "host/model"
810 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
811 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
812 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
813 - xbt_dirname and xbt_basename for non-POSIX systems
815 * The model checker now runs as a separate process.
816 * The model checker runs must now be launched with the new simgrid-mc program.
817 * Record/Replay: the MC can display a textual representation of a path in the
818 execution graph. It can then be replayed outside of the model checker.
820 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
822 SimGrid (3.11) stable; urgency=low
827 * Normalizing pointers addresses tool for better diff between logs
830 * Add cloud examples using new VMs
831 - examples/msg/cloud/two_tasks_vm.tesh
832 - examples/msg/cloud/simple_vm.tesh
833 - examples/java/cloud/cloud.tesh
834 - examples/java/cloud/migration/migration.tesh
835 * Add java surf examples:
836 - examples/java/surfPlugin/surf_plugin.tesh
837 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
838 - examples/java/surfCpuModel/surf_cpu_model.tesh
839 * Add SMPI+MSG example:
840 - examples/smpi/smpi_msg_masterslave/
845 - msg task destroy cancel test
846 - msg_host on/off test
847 * Move all tests in testsuite to teshsuite (adding tesh files)
848 * Restructure teshsuites
849 - one folder for each kind of test
850 * Restructure AddTests.cmake
852 - structure the order of tests (with sections)
855 * Add virtual machine
856 - creation of a VM on a PM
857 - migration of a VM from a PM to another PM
859 - MSG_process_join(msg_process_t process, double timeout)
860 - msg_bar_t MSG_barrier_init(unsigned int count)
861 - int MSG_barrier_wait(msg_bar_t barrier)
862 - void MSG_barrier_destroy(msg_bar_t barrier)
863 - msg_as_t MSG_environment_as_get_by_name(const char * name)
864 * New option "msg/debug_multiple_use" to help debugging when a task is used
867 - Improvements and finalization of MSG_storage, MSG_file APIs and their
869 - Increase code coverage in test suites
873 * Protect context stacks against stack overflow. The number of protected memory
874 pages allocated on the top of each stack (1 by default) can be configured
875 with the new command line option --cfg=contexts/guard_size:<value>.
876 * Simcalls are now generated by a python script that
877 - generates files included by SimGrid
878 - checks that all the functions exist, and proposes prototypes
880 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
881 host_set_data, host_get_data
883 - simcall_process_join(smx_process_t process, double timeout)
884 * Fix bug where sleeping processing could not be suspended.
887 * Translate surf models from C to C++
888 - Generic classes for all models: Model, Resource, Action
889 - A generic interface for each kind of model (CPU, Network, Storage
890 Workstation, WorkstationVM)
892 * Translate surf routings from C to C++
893 * Add callbacks using sigc++ or boost::signals2
894 - Add callback functions for resource creation/destruction
895 - Add callback functions for action state change
896 - Handle Energy as a plugin
897 * Replace swag by boost::intrusive
898 * Add new routing models for clusters. For documentation, see
899 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
900 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
901 parameters for cluster tag
902 - Fat trees, with topology="FAT_TREE" and
903 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
904 - see examples/platforms/torus_cluster.xml and
905 examples/platforms/fat_tree_cluster.xml
909 * Hostfiles support host:nb_processes construct to deploy several processes on
911 * Collective communication algorithms should not crash if used with
912 improper number of nodes and report the error.
913 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
914 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
915 MPI_Dims_create, MPI_Cart_sub are supported.
916 * New interface to use SMPI programmatically (still depends on MSG for
917 some parts, see examples/smpi/smpi_msg_masterslave) :
918 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
922 * Global variables privatization in MPI executables is now performed at runtime
923 with the option smpi/privatize_global_variables (default:no).
924 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
925 dynamic libraries loaded by the application are not privatized (static
926 linking with these libraries is advised in this case)
929 * Options defined in XML work correctly now.
932 * New cmake option, enable_lib_in_jar, to control whether native libraries are
933 copied into simgrid.jar or not (ON by default). Use this option if you want
934 to reduce the size of the installed simgrid.jar, *and* the native libraries
935 are kept installed elsewhere.
936 * Surf binding with SWIG (code generated in maintainer mode only):
937 - plugin to handle callbacks
938 - CPU model only for the moment
941 * Supernovae build mode is definitively removed. It was used to improve
942 inlining and inter-module optimizations. It is nowadays superseded by
943 link-time optimizations commonly available in compilers.
944 * Update ns-3 find lib. Bindings for ns-3 should work again now.
945 * Add boost dependency for surf++
946 * Add new macro for tests
947 - ADD_TESH(name <tesh_args>)
948 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
952 - xbt_bar_t XBT_barrier_init(unsigned int count)
953 - int XBT_barrier_wait(xbt_bar_t barrier)
954 - void XBT_barrier_destroy(xbt_bar_t barrier)
955 * Make the xbt_os_time module public
957 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
959 SimGrid (3.10) stable; urgency=low
961 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
964 * Reintegrate Java to the main archive as desynchronizing these
965 package is not acceptable anymore (Java is now considered stable)
966 * Add explicit synchronization facilities through semaphores
967 * Bug fix: Task.setDataSize() only changed the C world, not the value
968 cached in the Java world
971 * Dramatically change the way files are handled. API and internals changed, but
972 this part of MSG was not considered as production grade either.
973 * Add explicit synchronization facilities through semaphores
974 * Add a new function MSG_host_get_process_list()
975 * Preliminary DVFS support (see examples/msg/energy/ for details)
978 * SMPI is now included directly in the libsimgrid as the windows
979 linker doesn't force us on splitting it anymore.
980 * Improvements of the SMPI replay tool:
981 - Most of the collective communications are now rooted in the same process as
982 in the original application.
983 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
984 used until now). Multiple data types can now be used in a trace.
985 - The replay tool now supports traces produce either by TAU or a modified
987 - Bug Fix: the compute part of the reduce action is now taken into account.
988 - Gatherv collective is now supported
989 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
990 of smpirun outputs time independent traces for the current run. One file
991 is created per process. If too many processes are simulated, this behavior
992 can be changed to one file for all processes by using the
993 tracing/smpi/format/ti_one_file flag
994 * smpirun generates the host file if needed (with given host count and platform)
995 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
996 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
997 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
998 --cfg=smpi/coll_selector:(mpich/ompi)
999 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
1001 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1002 happening inside a collective SMPI call.
1003 * Fix the behavior of complex data types handling.
1004 * Make MPI_Wtime another synchronization point to take computations into
1006 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1007 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1008 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1009 (no privatization of global variables yet)
1010 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1013 * Verification of liveness properties is now available for SMPI applications
1014 (in addition to MSG applications)
1015 * Bugged examples using SMPI in examples/smpi/mc/
1016 * Add --cfg=model-check/visited option. Allows the verification of infinite
1017 programs. Detection of loops in the execution thanks to the system state
1018 comparison and reduction of the state space to explore. Can be combined with
1019 DPOR for safety properties.
1022 * Allow to change SimGrid configuration (see --help) within the code
1023 thanks to SD_config() as it can be done in MSG.
1024 * Add a new function SD_task_set_amount() upon user request.
1027 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1028 * Remove rule based routing (no more PCRE dependency)
1029 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1030 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1031 * Add a loopback_bw and loopback_lat options to cluster tag.
1032 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1033 routing. Now peers are encapsulated in an AS and have their own private
1034 router but this is transparent.
1037 * Our own implementation of getline is renamed xbt_getline, and gets
1038 used even if the OS provide a getline(). This should reduce the
1039 configuration complexity by using the same code on all platforms.
1040 * New type: xbt_cfg_elm_boolean.
1041 * Allow to use yes/no for boolean configuration options in the command line.
1042 * Allow to disable SimGrid cleanups at exit from command line option.
1043 There are situations where one may want a simulation to end with an exit.
1044 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1045 annoying when scripting around the simulator. Adding a
1046 --cfg=clean_atexit:no allows to circumvent this issue.
1049 * Lots of memory leaks were corrected in this release.
1050 There is no known memory leaks anymore, in all of our 600+ tests.
1051 * New command line option --version, to get SimGrid version information.
1052 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1054 * Supernovae builds are deprecated, and expected to be removed in the next
1057 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1059 SimGrid (3.9) stable; urgency=low
1061 The Grasgory release: GRAS is really dead now.
1063 * Complete overhaul of the internal host structures scheme.
1066 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1067 was considered as experimental and badly maintained since then.
1068 * Keeping it was thus a trap to our potential users, that could take
1069 it instead of MSG or SMPI by mistake despite is pity state.
1070 * GRAS seems to have very few users (if any), and no one volunteered
1071 to maintain it further. It also induces a lot of XBT code (for
1072 portability sake), that must be maintained too.
1073 * For all these reasons, we killed GRAS. If someone wants to revive it
1074 in the future, don't cry, our git history still remembers of GRAS.
1077 * Major overhaul. Merge our documentation again as time proved that
1078 splitting it was really not helping our users.
1079 * Further improve the developer documentation to help newcomers
1080 hacking on SimGrid itself. The user documentation (and in
1081 particular, the beginner documentation) is still in a sorry state.
1084 * Now works on Windows too!
1085 * Much more extensive test suite, from MPICH
1088 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1089 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1090 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1091 * Bug fix: let task be scheduled when the last dependency to be solved is
1092 a control dependency.
1093 * Remove SD_load_environment_script function.
1094 Use the C sg_platf function if you want to declare a platform
1098 * New function: MSG_process_get_number()
1099 * Old function documented: MSG_config()
1100 * Remove MSG_load_platform_script function
1101 Use the C sg_platf function if you want to declare a platform
1105 * Change the default value of the TCP_gamma constant (maximal size of TCP
1106 congestion window) to a more realistic 4MiB value. If you notice changes in
1107 your simulation results, you can fall back to the previous 20k tiny window
1108 by adding --cfg=network/TCP_gamma:20000 on command line.
1109 * (Hopefully) fix a bug wrt periodic availability and state traces
1110 * Bug fix: use default values at start when first event in availability/state
1111 trace is not at time 0.
1114 * remove the "new_" part of function name sg_platf_new_trace_connect
1115 (resulting in sg_platf_trace_connect), since it does not create
1119 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1120 use case where it's really mandatory, and maintaining it was a pain in
1122 * New: xbt_fifo_search(), search an item with a user-provided
1123 comparison function instead of dumb pointer comparison.
1126 * Fix the lua deployment:
1127 Use `simgrid.init_application()` before deployment instead of
1128 `simgrid.msg_register_application()` after.
1131 * Transfer the tracing files into the corresponding modules.
1133 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1135 SimGrid (3.8.1) stable; urgency=low
1137 The "we are told that some people want to also *install* the simgrid
1140 * Add missing file "tesh.1" to the archive.
1142 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1144 SimGrid (3.8) stable; urgency=low
1146 The Psssshiiiit release: SimGrid jumps into the Cloud.
1149 * Add an experimental interface to manipulate VMs. They are mainly
1150 process groups with very few intrinsic semantic, but they should
1151 allow you to build the semantic you want easily.
1152 * New function: MSG_host_set_property_value()
1153 * New function: MSG_process_on_exit(). To clean memory in all cases.
1154 * Bug fixes that made the host (and link) failures unusable.
1155 * Add a way to auto-restart process when the host in which they are
1156 executing comes back (ON_FAILURE="RESTART" on deployment file,
1157 MSG_process_auto_restart_set).
1158 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1159 please stop using the old ones, they are DEPRECATED.
1161 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1162 Please use MSG_init() instead. (reducing the amount of entry
1163 points in the library helps us).
1164 * Make it impossible to link against the wrong version of the lib
1165 * Deprecate MSG_clean(). No need to call it anymore.
1166 * Function MSG_get_host_number() is not deprecated anymore.
1169 * Split the doc into a user guide and a reference guide.
1170 * Start a developper guide to help people hacking on SimGrid.
1173 * Enable tracing by default. This modules rocks you should use it.
1174 * Remove option custom_flags. Now use environment variables CFLAGS
1176 * Use default cmake things to detect lua instead of home grown ones.
1177 * New option "enable_mallocators" to disable mallocators, for debugging
1178 purpose ("on" by default).
1181 * Bug fixes around the resource failures: don't let the processes
1182 survive the host they are running onto.
1183 * Add an interface to auto-restart processes when the host in which they are
1184 executing comes back.
1185 * Ensures that SIMIX_clean is called automatically. It's not part of
1186 the public interface anymore (bindings should be updated).
1189 * Bug fix for when SD_Simulate is called with a positive value: be careful
1190 when comparing doubles. Sometimes they are different for non significant
1192 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1193 parallel task whose initial work is distributed among host according
1194 to the Amdahl's law. Such tasks are created with a parameter alpha
1195 that corresponds to the non-parallelizable part of the computation.
1196 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1197 between two sets of workstations assuming a 1D block distribution (each
1198 workstation owns a similar share of data) on both sides.
1200 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1201 Data redistribution will be automatically scheduled once parent and child
1202 are both scheduled. The filling of computation_amount and
1203 communication_amount structures is now done seamlessly thanks to the chosen
1205 * New function SD_workstation_dump to display various information
1206 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1207 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1208 bandwidth on the route onto which the task is scheduled and the amount of
1210 To divide the nominal bandwidth by 2, the rate then has to be :
1211 rate = bandwidth/(2*amount)
1212 * Compute tasks that have failed can now be rescheduled and executed again
1213 (from their beginning)
1214 * Increasing source code coverage (src/simdag is now covered at 95.8%
1218 * Re-implement time-independent trace replay using SMPI (at the
1219 smpi_smp_* level) instead of MSG. This should replace
1220 examples/msg/actions/actions.c
1221 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1222 hindexed and structs)
1223 * Implement the exchange of non-contiguous data.
1224 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1225 * Correct behavior of smpi/sender_gap and set its default value to 0
1226 * Add option to asynchronously send small messages to allow better
1227 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1228 specifies the size in bytes under which messages will be asynchronously sent.
1229 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1230 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1233 * New C interface to define a platform: XML is now optional.
1234 For more info, please check include/simgrid/platf.h.
1235 * New interface to define random platforms from the C:
1236 For more info, please check include/simgrid/platf_generator.h and
1237 examples/msg/masterslave/masterslave_platfgen.c
1238 * Export a sg_cmdline dynar containing all the arguments we got from
1242 * Two new tracing options for adding comments to trace file so you
1243 can track your experiments (see --help-tracing for details).
1244 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1245 * Adding the SimGrid version that generated the trace file as a comment.
1246 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1247 * Fix to avoid key clashes on Paje links
1248 * Other minor fixes related to the Paje specification
1251 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1252 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1253 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1254 * Greatly improve the robustness of mmalloc to user errors (such as
1255 using an area after freeing it, or freeing it twice)
1257 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1259 SimGrid-java (3.8.1) stable; urgency=low
1261 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1262 Warning: all methods are visible, but only some of them are
1263 implemented so far. Check the source (src/jtrace.c)
1264 for further information.
1265 * New module: org.simgrid.msg.File (SimGrid File management functions)
1266 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1267 * Change the meaning of Process.restart: now restart the process from
1268 the begining, like MSG_process_restart in C.
1269 * Add Process.setAutoRestart: handling of process restart when failed
1271 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1272 you to retrieve the properties of the processes/hosts
1273 * Deprecate Msg.clean(): you can just forget about it now.
1274 * New function Process.getCount(), that only works when compiling
1275 with the not yet released version 3.9 of the C library.
1277 * New context factory based on Coroutines. It mandates a modified JVM
1278 but then, the simulations run about five times faster, and there is
1279 no limit to the amount of processes (beside of the available memory).
1281 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1283 SimGrid (3.7.1) stable; urgency=low
1286 * Restore the prototype of MSG_process_create_with_environment() to
1287 the pre-3.7 situation by removing the kill_time argument.
1288 * Add a MSG_process_set_kill_time() function instead.
1291 * Fix weird behaviors when dealing with parallel tasks.
1294 * Simgrid is now built as a dll.
1295 * Simgrid-java now works on Windows.
1296 * Simgrid-Java is now included into Windows package.
1299 * First pre-build package for MacOSX.
1302 * Fix compilation when using MSG_USE_DEPRECATED.
1303 * Fix some compilation issues on Macs and Windows.
1304 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1305 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1307 We discovered that the Lua console is broken, but we are missing the
1308 manpower to fix it right now. The problem existed in 3.7 too, so we
1309 are not blocking the release for that. Sorry if you depended on this
1310 feature, any help would be really welcome.
1312 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1315 SimGrid-java (3.7.1) stable; urgency=low
1317 The "Java aint got to be bloated and slow" release
1320 * Various internal cleanups and performance improvement
1321 Simulations are expected to run up to twice faster or so
1322 * Make Process.kill(process) an instance method, not a static one
1323 * User processes are not java.lang.Thread subclasses.
1324 This breaks the compatibility (sorry), but previous API was
1325 brain-dead, making it impossible to have non-trivial
1326 initializations in the process constructor.
1327 * Require a full constructor per Process sub-class.
1328 Kinda breaks the compatibility (sorry), but this allows a much more
1329 efficient way to launch the processes at simulation startup.
1330 * Do not embeed our version of semaphores, java 1.5 can be considered
1331 as sufficiently prevalent for us to not dupplicate its features.
1335 * Add examples for almost every part of the API
1336 We spotted and fixed a lot of bugs in the process
1338 * New module: asynchronous communication API
1339 * New function: Process.sleep()
1340 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1341 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1342 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1344 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1346 SimGrid (3.7) stable; urgency=low
1348 The "spring cleanups (before the next Big Project kicks in)" release.
1351 * We can specify the SMPI latency/bandwidth factor with command line
1352 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1353 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1354 You can also use the "config tag" from platform file by adding this line
1355 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1356 (see "example/platforms/tag_config.xml" to use "config tag").
1357 Note that the command line supersedes the platform file configuration.
1358 * Change the correction factors used in LMM model, according to
1359 the latest experiments described in INRIA RR-7821.
1360 Accuracy should be improved this way.
1361 * Use the partial invalidation optimization by default for the
1362 network too. Should produce the exact same results, only faster.
1363 * Major cleanup in surf to merge models and split some optimization
1364 mechanisms from the core of the models. As a result you can now
1365 specify which model to use (e.g., --cfg=network/model:LV08
1366 --cfg=cpu/model:Cas01) and which optimization mode to use
1367 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1368 Incompatible combinations should err at initialization. See
1369 --help-models for the list of all models and optimization modes.
1370 * The CLM03 workstation model was dropped for simplicity because it
1371 used the deprecated CM02 network model. Use default instead.
1372 * Rename the TCP_gamma configuration option to network/TCP_gamma
1373 * Rename the coordinates configuration option to
1374 network/coordinates, and document it
1375 * Use now crosstraffic keyword instead of the terribly misleading
1376 fullduplex keyword. It is activated by default now in the current
1377 default model, use --cfg=network/crosstraffic:0 to turn it off.
1378 * Ongoing refactoring the model parsing to make XML files optional
1379 See include/simgrid/platf.h for details (still to be completed)
1382 * Major overhaul of the documentation. Almost instructive now :/
1383 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1384 functions and friends. This interface was considered as
1385 deprecated since over 2 years, it's time to inform our users that it is.
1386 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1387 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1388 use these (crappy) functions in your code.
1389 These functions will be removed soon. Stop using them now.
1390 * Deprecate MSG_get_host_{table,number}
1391 Implement MSG_hosts_as_dynar() instead.
1392 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1393 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1396 * Stabilize the parallel execution mode of user contexts
1397 * Introduce configuration variables to control parallel execution:
1398 - contexts/synchro: Synchronization mode to use when running
1399 contexts in parallel (either futex, posix or busy_wait)
1400 - contexts/parallel_threshold: Minimal number of user contexts
1401 that must be part of a scheduling round to switch to parallel
1402 execution mode (raw contexts only)
1403 * Fix bugs that prevented to use suspend/resume along with
1404 synchronization structures.
1405 * Fix bugs in process termination that lead to invalid memory access
1406 in very specific conditions.
1409 * Introduce a parallel mode for the models (controlled by surf/nthreads
1410 configuration item). In our tests, running the models in parallel
1411 never lead to any speedups because they are so fast that the gain
1412 of computing each model in parallel does not amortizes the
1413 synchronization costs, even when ultra fast futexes are used.
1414 This is released anyway because YMMV.
1417 * Performance boost by using a swag internally to compute the set of
1418 tasks that are finished and should constitute the return value of
1422 * Enable it by default now that it is considered rather stable.
1425 * Documentation of the tracing functions.
1426 * Performance gains when tracing categorized/uncategorized resource
1427 utilization by avoiding calls to get route when updating resource
1428 variables. LMM constraints are being used instead.
1429 * API changed to set task categories. Use MSG_task_set_category instead
1430 of TRACE_msg_set_task_category, and SD_task_set_category instead
1431 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1432 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1433 * Examples for link user variables added (see at examples/msg/tracing/)
1434 * Deprecated function TRACE_msg_set_process_category completely removed
1435 * Trace header updated according to the latest Paje file format
1436 * Tracing network lazy updates, no longer obligate users to use full updates
1437 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1438 * Experimental: let user code declare/set/push/pop application states for hosts
1439 * API changed to allow the manual creation of graph configuration files
1440 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1443 * Improve the API of Lua MSG bindings, using the Lua spirit.
1444 * Each simulated process now lives in its own Lua world (globals are
1445 automatically duplicated). It helps writing simulators. Will allow
1446 to run Splay programs within SimGrid in the future.
1447 * Add a Chord example in Lua, equivalent to the MSG one.
1450 * Start the implementation of a solution to express temporal
1451 properties, not only local assertions. This is still an
1452 experimental work in progress, stay clear from it to be safe.
1456 - Add new runtime parameters --help-logs and --help-log-categories
1457 to display information about supported logging parameters and
1459 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1460 don't exists anymore.
1461 * Mallocators: allow value NULL for the reset function.
1463 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1464 create homogeneous dictionaries, where all the elements share the
1465 same free function. Non homogeneous dictionaries will be
1466 deprecated in the next release.
1467 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1468 - Multi-level dictionaries are deprecated.
1470 - new function xbt_dynar_search_or_negative() that is useful when
1471 you have less than 2 million elements in your dynar and don't
1472 want of the extra complexity of catching exceptions when the
1473 element is not found.
1475 - Make xbt_os_thread module (for thread portability) public.
1476 Documentation is still to come, sorry.
1478 - Cleanups and simplifications to make it maintainable again.
1479 - Exotic features (such as memalign and valloc) were removed.
1480 - The metadata were extended and improved so that the
1481 model-checker becomes able to explore and inspect the heaps.
1482 - This may induce a performance drop when enable_model-checking is
1483 ON in cmake (even if it's not used in the simulation), but it is
1484 necessary at this point to get MC working.
1486 Turn model-checking OFF if simulation performance matters to you.
1487 Not enabling it at runtime is not enough, disable it in cmake.
1489 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1492 SimGrid (3.6.2) stable; urgency=low
1494 The "Not coding new stuff allows to polish old things" release.
1497 * New bindings to the NS3 packet level simulator (experimental)
1498 * Use the raw (efficient) execution contexts instead of the sysv
1499 (portable) ones when possible.
1500 * libpcre is now mandatory in any cases since not using it led to
1501 severe performance loss and possibly other issues
1502 * Update the XML platforms:
1503 - G5K: include the latest machine in Nancy
1504 - GridPP and LCG: new platforms
1505 * Documentation was partially updated, at least (more to come)
1507 Bug fixes, cosmetics and small improvements
1508 * Free terminated processes before the end of the simulation to avoid
1509 exhausting the memory of users having very dynamic amount of
1511 * Bug fix and cosmetics about canceling non-running tasks
1512 * Bug fix about the dot loader's issues when using libcgraph
1515 * Create an installer for windows with nsis (amd64 and win32)
1516 - Add an hello world project to illustrate simgrid project creation.
1517 - Embed libpcre into the Simgrid installer to avoid
1518 its compilation burden
1519 * The raw execution contexts should work on Apple now
1520 * Port to Windows 64 bits
1521 - Sysv contexts now have an implementation for this arch
1522 - GRAS communication features now support this arch
1523 * Drop support for borland compiler on windows
1524 - this code was not maintained, and we kinda depend on gcc nowadays
1525 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1526 * Fix portability issue on unstable ubuntu: linker became picky on
1529 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1531 SimGrid (3.6.1) stable; urgency=low
1533 The "Oops, we broke Macs too" release
1536 * Fixed contexts detection so that raw ones are used when possible
1537 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1538 leads to a strange error, with user code segfaulting sometimes when
1539 the generated code is not perfectly aligned (which is not
1540 controllable from the user side, depends on the amount of code)
1543 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1545 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1547 SimGrid-java (3.6) unstable; urgency=low
1550 * Split of every thing from simgrid v3.5 into a separate package.
1552 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1554 SimGrid (3.6) stable; urgency=medium
1556 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1559 * Bindings now constitute their own package, separated from the main one.
1560 Rationale: reduce our maintenance nightmare by reducing the module coupling
1561 They will soon be released on their own on gforge.
1563 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1564 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1566 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1567 * It's not quite deprecated for now because we have no replacement,
1568 but it may soon become the case.
1571 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1572 * Fortran: New user-level cache variable to store the rank of the running
1573 process. This improves performance by an order of magnitude.
1574 * C: New coccinelle script to automatically locate and modify global and
1575 local static variables.
1576 * Improved SMPI network model with a sender-side gap to account for multiple
1580 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1581 only say if a communication is finished, no matter whether it succeeded or
1582 failed. You can call MSG_comm_get_status() to know the status of a finished
1584 * New function MSG_task_dsend() to send a task and detach it. When a
1585 communication is detached, you are never notified of its success or failure
1586 and the memory is released automatically once it is finished. This function
1587 is useful when you don't care about the end nor the success of a
1589 * Change the prototypes of action replay. Sorry for inconvenience,
1590 but this is really more efficient this way (and to adapt your code,
1591 you just have to fix the initialization, that shouldn't be too long)
1592 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1593 ever managed to do anything useful with it.
1594 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1595 takes a dynar of communications. It returns immediately and gives the
1596 index of a finished communication (if any).
1597 * New example: a basic implementation of the Chord P2P algorithm.
1600 * New model for multi-core CPUs. You can now use the core attribute to
1601 precise the number of cores of a host. This is a basic model. Every
1602 process running on the host receives at most the power provided in
1603 the DTD (throughput<=power). Total throughput of process cannot exceed
1605 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1606 router linked by an up-link and a down-link (possibly asymmetrical).
1607 This kind of pattern allows to easily build last-mile model style platforms.
1608 Aggregating such patterns in a rule-based AS is thus the technique of
1609 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1610 * New model for Vivaldi routing. We transformed the Vivaldi network model
1611 into a Vivaldi routing model (based on the rule-based model). This allows to
1612 combine Vivaldi based latencies with last-mile platforms.
1615 * Added a check for NaN of IEEE754 infinite in the double entries of
1617 * Introduce a new context factory "raw", highly inspirated from the
1618 ucontext factory, but using manually crafted functions in assembly to
1619 do the work in an efficient manner.
1620 * Allow to change the used context factory at run time, not only at
1621 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1622 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1623 contexts at runtime (only with raw contexts or ucontexts).
1624 * Completely rewrote this module to allow parallel execution of user
1625 processes. Use --cfg=contexts/nthreads:N to execute user processes
1626 with N parallel threads (the default is 1, meaning no parallelism).
1627 * Allow to decide dynamically between sequential and parallel modes.
1628 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1629 processes in parallel only when their number is greater than or equal to P
1631 * Added a check for NaN of IEEE754 infinite in the double entries of
1635 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1636 won't output the state of processes when interrupted with Ctrl-C
1637 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1638 NULL-terminated array. This may solve backward compatibility issues
1639 due to the change to return type of SD_simulate. See also:
1640 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1641 * Add new macros with variable number of arguments.
1642 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1643 - in xbt/asserts.h: xbt_assert
1644 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1645 - in xbt/ex.h: THROWF and RETHROWF.
1646 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1648 * Change xbt_die() to accept a format string with arguments, just like printf.
1649 * New data structure: xbt_lib_t, like a dict but more general and with better
1653 * New configuration options
1654 Options triva/categorized and triva/uncategorized can be used to generate
1655 graph configuration files for Triva visualization tool.
1656 * Configuration option tracing/platform is renamed to tracing/categorized
1657 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1658 * New TRACE_user_link_variable interface:
1659 User provides the name of the link and the tracing variable to attach to it
1660 * the declaration of tracing categories must be done after the environment
1662 * simpler tracing interface, just one way to declare categories
1663 TRACE_category or TRACE_category_with_color, it is up to you
1664 * links in the trace file are again identified by their names
1665 * trace contains the full platform hierarchy exactly as declared using the ASes
1666 * Options tracing/msg/[task|process]:1 groups the process by hosts
1667 for both cases, tasks and processes must have names that are unique during the simulation
1668 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1669 * The experimental option tracing/msg/volume is deprecated
1670 its functionality may be reincorporated if needed
1672 The tracing generates a trace file with unordered timestamped events,
1673 because of the way the core simulator (surf) works. A script available
1674 at the tools directory (fix-paje-trace.sh) can be used to put the events
1675 in order. We have changed the tracing so it can generate ordered timestamped
1676 events in the final trace, but depending on the simulator (and how much time
1677 is simulated) that can lead to a huge memory utilization. It is deactivated
1678 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1680 Build Infrastructure
1681 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1682 - We are trying hard to keep the API stable, but it may happen that
1683 some things change (we're a research project after all, not a
1684 nuclear plant operating system). If such things should happen, you
1685 could rely on that macro to adapt.
1686 - current value: 30600 for 3.06.00, aka 3.6
1687 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1688 a number that can be compared with SIMGRID_VERSION.
1689 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1690 assertions and verbose/debug logging events are disabled at compile time.
1692 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1694 SimGrid (3.5) stable; urgency=medium
1697 * New feature to any SimGrid-based simulator: Model-Checking
1698 Check SIN#1 for more details.
1701 * New Model SMPI (three-interval linear regression for correction factors)
1702 See RR-7426, available at http://hal.inria.fr/inria-00527150
1703 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1704 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1705 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1706 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1707 * See also src/smpi/README
1712 - Tracing API changes: TRACE_start and TRACE_end should not be called
1713 by user-code. They are automatically called by simulators created
1714 with SimDAG, MSG and SMPI if the toolkit is compiled with
1715 tracing_enabled=ON. Categories declaration and utilization remain the
1716 same for MSG and SimDag.
1717 - A function was added to the tracing API to declare categories with
1719 - TRACE_category_with_color (char *category, char *color)
1720 where color must be in the following format
1721 "%f %f %f", red, green, blue
1722 and red, green, blue are float values in the interval [0, 1]
1723 - User can specify NULL as color parameter, or continue calling
1724 TRACE_category (cat)
1725 On that case, the tracing system will define random colors
1726 - The following command-line options are supported:
1727 --cfg=tracing/filename:msg.trace
1728 --cfg=tracing:1 (activate tracing, needed to use others)
1729 --cfg=tracing/platform:1 (categorized resource use)
1730 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1731 --cfg=tracing/msg/task:1 (task creation)
1732 --cfg=tracing/msg/process:1 (process creation, migration)
1733 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1734 --cfg=tracing/smpi:1 (SMPI interface tracing)
1735 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1736 - examples of examples/msg/tracing updated
1738 - DAXLoader and DOTLoader functions can generate tasks with categories
1739 - A new function to attribute a category to SD tasks:
1740 TRACE_sd_set_task_category (SD_task_t task, char *category)
1741 * Tracing the MPI interface implemented by SMPI
1742 - Collective operations are traced with states
1743 - Point-to-Point operations are traced with states/links
1744 - Tracing activated by a parameter "-trace filename" passed
1745 to smpirun during execution (considering that simgrid
1746 is compiled with tracing enabled)
1747 - To run the simulation with gdb, the simulator
1748 accepts --cfg=tracing/smpi:1 to trace SMPI
1749 - tesh files to check if smpi tracing is ok
1750 - See examples/smpi/NAS/DT-trace
1751 * GTNetS tracing re-worked
1752 - adaptation to the tracing system of GTNets to cope
1753 with modifications regarding the fullduplex mode
1754 - new tesh files to check if gtnets tracing is ok
1757 * Asynchronous communications through the functions:
1758 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1759 * New function: MSG_load_platform_script()
1760 to make possible using a lua script instead of XML files to set up platforms
1761 * New function: MSG_set_function
1762 to associate functions to processes, used when bypassing the parser
1763 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1765 Platforms: Add some more examples in examples/platforms
1766 * Grid'5000: see www.grid5000.fr
1767 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1770 * Change the XML format. This is a very important modification. SimGrid
1771 3.5 introduces a new hierarchical format based on the notion of
1772 Autonomous Systems. Compatibility with old format is ensured
1773 through the perl script provided in the install bin directory
1774 bin/simgrid_update_xml.
1775 It is now possible to build platforms with specific routing
1776 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1777 connect several platforms together. We will try to provide soon
1778 set of realistic platforms exploiting these properties (have a
1779 look at examples/platforms/ for the moment).
1780 * Take the opportunity of the XML format change to be a good XML citizen:
1781 rename link:ctn to link_ctn and similar changes (also dealt with by
1783 * Add a new routing scheme (rule-based) using regular expressions. It
1784 enables to have an extremely low memory footprint when the
1785 underlying routing is simple and can be compactly described. You
1786 need to have libpcre4-dev (perl regular expressions) installed if
1787 you want to use this routing scheme.
1788 * Revive the cluster TAG and allow to easily and efficiently (both in
1789 term of memory and speed) connect clusters together. Have a look
1790 at teshsuite/simdag/platforms/ to see how this can be done. With
1791 this tag, you can create clusters with thousands of tasks at no
1792 cost (have a look at examples/platforms/).
1793 Note: clusters are implemented as ASes, so there is no need for an
1794 enclosing AS tag if you have only one cluster in your platform.
1795 * Add new generic functions in the public interface that allows the user
1796 to call SURF 'create_resource' methods from your code (same
1797 functionality as the XML bypass mechanism but with a much lighter
1799 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1800 that uses a piecewise linear approximation to produce better
1801 results when exchanging small messages.
1802 * Add a new parameter to handle correctly full duplex link and account
1803 for interferences between uplink and downlink communications
1804 (activate with --cfg=fullduplex:1).
1807 * Rename the SD_READY (all dependencies are satisfied and task is
1808 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1809 dependencies are satisfied) state.
1810 This prevents a confusion between the notion of "ready to schedule"
1811 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1812 simulated" (SD_RUNNABLE) used by the simulation kernel.
1813 * Change the way a task is considered as ready. Instead of removing
1814 dependencies when a task is done, a counter is decreased. This way,
1815 it is always possible to reach ancestors thanks to the
1816 SD_taks_get_parents function (even after the end of the simulation.)
1817 * Change the return type of SD_Simulate from (SD_task_t*) into
1818 xbt_dynar_t. This function was in handling a dynar internally and
1819 converted it into a NULL terminated array for historical reasons.
1820 * New function SD_dotload(char*) to load a DAG described in dot
1821 format. This loader and the corresponding examples require the
1822 installation of the graphviz library.
1823 * Fix a bug in the management of tasks of size 0 in the surf network
1824 models. This problem was only visible with SIMDAG and you should
1825 thus disregard results produced with earlier versions if you
1826 relied on this feature (some tasks were blocked because of this).
1827 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1828 even though your applications were doing only point-to-point
1829 communications and sequential computations. Now you can really use any
1830 model you want (of course, if you create real parallel tasks, which are
1831 not implemented in most models beside ptaskL07, this will abort).
1832 * Add an example that schedules a DAX on an heterogeneous platform
1833 using a Min-Min strategy.
1834 * New function SD_workstation_get_current_task() that returns the kind
1835 of task currently running on a workstation in the sequential access
1837 * Raise some warnings when unexecuted tasks remains at the end of the
1838 simulation. This is usually caused by cycles in the DAG.
1841 * New function: SIMIX_process_set_function() called by MSG_set_function
1842 * Change the underlying waiting queue in semaphores so that a process
1843 can wait on several of them simultaneously (as in waitany).
1844 * Fix the way to handle tokens in semaphores so that all access patterns
1845 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1846 * kill the dirty pimple SIMIX_message_sizes_output()
1847 Please use (proper) visualization instead
1850 * New data container: setset (set of sets of elements)
1851 * New module: mmalloc (mapped malloc, allowing to have several
1852 independent segments of malloc)
1853 * New function: xbt_dict_cursor_set_data()
1854 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1855 * New function: xbt_dynar_is_empty()
1856 * New function: xbt_fifo_get_last_item()
1857 * Fix xbt_dynar_shrink(): use the right element size.
1858 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1859 uninitialized areas during expand.
1860 * Fix semaphores: previous implementation was severely broken.
1861 * Use library init/fini functions for our initialization.
1862 - you can use logs and other feature as soon as you want in your
1863 code (even before the xbt_init / MSG_init)
1864 - xbt_exit is now a no-op and produce a warning when used.
1867 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1868 benefit from the latest improvement to the simulation kernel.
1869 * Kill measurement sockets for now. If you rely on them, sorry. This
1870 release is not for you. This feature will be reintroduced in the
1871 future, but we cannot delay the release any further.
1872 * New function: gras_msgtype_get_name().
1873 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1874 * Fix (at last) the pmm example: it should not randomly fail anymore.
1876 Build chain: bug fixes and overall polishing
1877 * Cmake is now stable enough. Hence, we killed the autotools.
1878 * Port to windows ( TM :)
1879 * Fix the 'make install' target.
1880 No need to use 'make install-simgrid' anymore
1881 * Introduce a 'make dist' target compiling a *source* archive
1882 'make package' compiles a binary archive
1883 * Compile java files only on need
1884 * Add --cd and --setenv command line options to tesh
1885 * Out of source builds are not fully supported yet, but we are close
1886 * Enable supernovae and optimization flags by default for our users
1889 * Add layer to set up environment directly from lua, without XML.
1890 * The effect of gras_stub_generator can be achieved through
1891 lua too (check examples/gras/console/ping_generator.lua)
1893 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1895 SimGrid (3.4.1) stable; urgency=low
1897 The "Polishing easter eggs is probably a good idea" release.
1898 This is a bug fixes release only.
1902 * Fix a bug preventing the tasks from begin garbage collected.
1905 * Fix a bug occuring when a host involved in a communication fails.
1906 This was not detected properly by the other peer involved in the
1907 communication. Now, it's reported as a network error.
1910 * Warn the user about loop dependencies in data flow of DAX files
1911 * Obey the control-flow dependencies of DAX files
1914 * Add option "enable_smpi" allowing to not compile SMPI.
1915 Probably useful for the (Mac) users experiencing a build error here
1916 * Improve the detection of lua5.1 and ruby1.8
1918 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1920 SimGrid (3.4) stable; urgency=low
1922 The "Easter in Cargese" release. Also known as (major changes):
1924 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1925 ~> bindings were greatly improved
1926 ~> new tracing infrastructure for better visualization introduced
1928 * the "Welcome to configury modernity" release.
1929 ~> we switched from autotools to cmake, and improved our cdash
1932 A more detailled list of changes follow (full detail in svn log).
1934 Java Bindings: Various Cleanups
1935 * (install java-gcj-compat-dev on debian-like to use them)
1936 * Remove put/get: no need to export deprecated interface in Java
1937 Use send/receive instead.
1938 * Cleanup the examples and add a README per directory
1939 * Remove example autoDestination (that's the only way to go now)
1940 * Remove example explicitDestination (was a plain copy of basic)
1941 * Make JniException a runtime exception, so that there is no need to
1942 declare the fact that you may encounter such a beast. I guess that
1943 nobody will ever want to survive such error.
1944 * Create specific errors for each MSG case of failure:
1945 host failure, transfer failure, timeout, task cancelled
1946 * Cleanup the exceptions that may get thrown by each function
1947 * Other internal cleanups in Java bindings. Performance still bad :/
1948 Ruby and Lua Bindings: create them
1949 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1950 * That's new and great, you should try them out.
1951 Same functionalities than Java bindings, only even less polished
1953 * Kill the useless "rate" argument of SD_task_get_execution_time()
1954 Everyone used to provide -1 as a value, it was not used, and the
1955 semantic of a possible use wasn't even clear.
1956 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1957 either as comm costs or compute costs to mean that there is no
1958 such thing for that specific task.
1959 * Add a SD_task_set_name() function
1960 * Fix SD_task_unschedule() on typed tasks
1961 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1962 * In DAX loader, accept useless 'level' attributes to <job> since
1963 LIGO DAGs have them (seem to be to ease graphical representation).
1965 * Add an example masterslave_mailbox.c using send/receive and not
1966 the deprecated put/get interface.
1967 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1968 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1970 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1971 (declare MSG_USE_DEPRECATED to still have the old name)
1972 * Add a MSG_task_set_data() function
1973 * About trace replay (see examples/msg/actions):
1975 - Allow to work with splitted trace files for each process
1976 Give the specific trace file as argument of each process,
1977 and call MSG_action_trace_run(NULL)
1978 You can still have one merged file for all processes.
1979 - Fix implementation of collective operations
1980 * Allow task_execute() on 0-sized tasks (closes #10063)
1982 * This is the first release of SimGrid where SMPI is not considered
1983 beta anymore (even if some corners should still be improved)
1984 * Port over the new SIMIX_network submodule (internal refactoring)
1985 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1986 * Implement more missing elements of the standard:
1988 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1990 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1991 MPI_Scan() MPI_Get_processor_name()
1992 - Added implementation of missing case for Alltoall (warning: it's
1993 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1994 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1995 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1996 SMPI_MPI_Allgatherv()
1997 * Bug fixes include:
1998 - MPI_Waitsome() was broken
1999 - Allow relative includes in smpicc
2000 - Command line cfg argument 'reference_speed' was ignored...
2001 - Some functions did not properly lead to auto-benching of user code
2002 - smpicc passes -O2 by default (just like openmpi one)
2004 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2005 * Bug fixes about timeouts during communications
2006 * add SIMIX_message_sizes_output() as a pimple to write to file the
2007 amount of messages per size. Use gnuplot to get histogram.
2008 Pimple because that's the only user-visible function of simix,
2009 defined directly in xbt.h (irk, sorry)
2011 - Add a SIMIX_sem_get_capacity() function
2012 - Fix interactions with processe resume/suspende
2013 - release_forever() was stupidly broken
2014 - Fix SIMIX_display_process_status() for processes in a semaphore
2015 - Make SIMIX_sem_block_onto() user-visible
2016 * Refactoring context stuff:
2017 - Use pseudo-OOP for better modularity
2018 - reimplement SIMIX_process_kill() without process_schedule() so
2019 that the latter can take as invariant that it is called from
2021 - Merge context_start into context_new for sake of simplicity
2023 * Add a Vivaldi network model, coded live during SUD'10 ;)
2024 * Rename configuration variables to start a hierarchy:
2025 o cpu_model -> cpu/model
2026 o network_model -> network/model
2027 o workstation_model -> workstation/model
2028 * New configuration variables:
2029 o network/bandwidth_factor: correction to bandwith
2030 o network/latency_factor: correction to latency
2031 o netwotk/weight_S: correction to the weight of competing streams
2032 * Add a long description to the models, that users can see with such
2033 argument on the command line: --cfg=cpu/model:help
2034 * --help-models display the long description of all known models
2036 * config: add the ability to set a default value after registration
2037 Does not override any previously set value (e.g. from cmd line)
2038 * dict: allow to have integer key and data.
2039 When so, you need to use the following functions
2040 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2041 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2042 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2043 In contrary to regular dicts, the key is not malloced before copy.
2044 Mixing scalar and regular elements in the same dict is not tested
2046 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2047 Tracing for Visualization:
2048 * SimGrid is now instrumented in order to generate a trace file for
2049 visualization analysis: to use it, need to compile SimGrid with the
2050 "tracing" option enabled, and instrument the program using SimGrid with
2051 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2052 (among other functions).
2053 * The instrumentation only traces the platform utilization for now
2054 * Documentation to use the tracing functions and how to analyze the
2055 traces with the Triva tool is written.
2056 * More information about: SimGrid FAQ (in the section Tracing Simulations
2059 * We moved to cmake as default build system. Autotools support will
2060 be dropped soon. Check the FAQ for more info about how to use it.
2061 * Greatly improved our cdash/ctest interactions
2062 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2063 * Added memory checking tests with valgrind; lot of memleak fixing.
2064 This may be the first release of simgrid with so few memory issues
2065 * Added code coverage tests.
2066 Our coverage is still improvable, but at least we see it on cdash.
2068 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2070 SimGrid (3.3.4) stable; urgency=low
2072 The "Desktop Grid needs love too" release (also called Xmas release).
2074 Models improvements:
2075 * Major speedup in the maxmin system solving by using lazy evaluation
2076 Instead of solving completely the maxmin system at each iteration,
2077 only invalidate (and recompute) the modified parts.
2078 This new feature is enabled in default models but you can try to
2079 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2080 * Cas01 IMproved as default CPU model
2081 This CPU model is the same Cas01 model, but it uses the
2082 maxmin-selective-update flag and a heap structure to manage
2083 actions on SURF kernel.
2084 It reduces the complexity to find the next action to finish and,
2085 consequently, it's faster than the old Cas01.
2086 This is the new default CPU model (Cas01).
2087 * Rename the old Cas01 model to Cas01_fullupdate
2088 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2089 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2090 * CpuTI (CPU Trace Integration)
2091 A new CPU model whose objective is simulate faster when using
2092 availability trace files.
2093 Instead of using a full featured, over engineered maxmin system for
2094 CPU modeling, this model does the pre-integration of traces files
2095 to calculate the amount of CPU power available, and so, executes
2096 faster than the old CPU models.
2097 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2098 * Use LV08 as default network model since it gives better accuracy
2099 for small messages and shouldn't change things for big ones.
2100 Use --cfg=network_model:CM02 to get the previous behavior.
2103 ******************************************
2104 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2105 ******************************************
2106 * The new CPU model may changes simulations!
2107 The point is that events occurring at the exact same timestamp
2108 are not scheduled in the same order with the old and new
2109 version. This may be enough to completely change the execution
2110 of simulations in some cases.
2111 * The new network model will change simulations!
2112 This new model is more realistic than the previous one, so you
2113 should consider redoing your old experiments with this model.
2114 Sorry for the inconvenience.
2117 * Introduce the supernovae compilation mode
2118 When compiled that way, the whole SimGrid (or almost) is put in a
2119 single compilation unit and compiled in one shoot.
2120 This is to help gcc which has difficulties to inline stuff from one
2122 The speedup seem to be above 15%, althrough more tests are needed on
2123 amd64 to confirm that gain.
2126 * Port of MSG's mailbox on top of SIMIX network
2127 The put/get mechanism was greatly simplified on the way.
2130 * New SIMIX network module. Provides:
2131 - Mailbox: rendez-vous mechanism to find with who you want to speak
2132 - Synchronous send/recv: easier and hopefully faster since the
2133 logic is handled in the maestro process directly now
2134 - Asynchronous send/recv: you dreamt of it? It's here now
2135 Too bad that nobody cared enough to propagate the change to MSG.
2136 * Add semaphores as SIMIX synchronization mechanism.
2139 * new function SD_daxload(char*) to load a DAX file
2140 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2141 * Introduce typed tasks. Specify its kind and cost at creation.
2142 At scheduling, just give where it should be placed, and the cost
2143 for each involved resource is automatically computed.
2144 Existing constructors so far (more to come of course):
2145 - SD_task_create_comm_e2e() for end-to-end communication
2146 - SD_task_create_comp_seq() for sequential computation
2147 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2148 * new function SD_task_dump() for debuging display
2149 * new function SD_task_dotty(task,FILE*) writing to file the info
2150 about the task in dotty format
2151 * SD_task_dependency_exists() can now cope with having one of its
2152 arguments NULL. If so, it tests whether the other argument has any
2154 * Add getters on list of preceding/following tasks:
2155 SD_task_get_parents(task) and SD_task_get_children(task)
2156 * Add getters on amount of workstations and list:
2157 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2158 * Add getter on task kind: SD_task_get_kind(task)
2159 * Update the start_time and finish_time of tasks on completion/failure
2160 * Bugfix: Remove task from state swags when destroyed
2163 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2166 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2167 dynar with one separate thread per value of the dynar.
2168 * Change the prototype of xbt_thread_create(), sorry.
2169 Added a boolean parameter indicating whether we want to join this
2170 thread (used in SG only for now)
2171 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2174 * GTNetS wrappers should now be usable again (and betterly tested too)
2175 * Fix a major regression from 3.2 where the timeout provided to
2176 MSG_task_put_with_timeout() was used as absolute time before which
2177 the comm should be done.
2178 * Start to fix the <cluster> tag.
2179 - Internal links should be good now (beside of the loopback, which
2180 use the private link instead)
2181 - paths to the external world is still rather broken
2182 - the <route:multi> tag is just broken. Actually that's brain-dead.
2183 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2184 to make it less stupid
2185 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2186 * Fix a source-level compatibility glitch from 3.2: after defining
2187 MSG_USE_DEPRECATED, you can use the old name
2188 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2189 * Allow to compile from the SVN with automake 1.11
2190 * Fix some problems when using the "start_time" tag in deployment XMLs.
2191 * Fix #8569: XBT/synchro.h has redundant declarations
2192 * Fix #8563: MSG return values and exceptions
2193 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2194 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2195 Thanks to Robson Peixoto.
2196 * Fix a vicious bug in dictionaries inducing that some elements were
2197 not freed on xbt_dict_free()
2199 Portability report of this version:
2200 * Main portability targets:
2201 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2202 on (amd64/i386/ia64)
2203 - mac leopard on i386
2204 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2205 but nothing critical.
2206 * Other platforms: windows, AIX and others were not tested for this release
2208 Timing report of this version:
2209 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2210 scenario parameters). From 8h to a few seconds in desktop grid settings.
2211 * Supernovae brings about 25% speedup on i386.
2213 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2215 SimGrid (3.3.3) stable; urgency=low
2217 The "Need for Speed" release.
2219 The timings done to validate the 3.3.2 were faulty.
2220 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2222 The problem was a conversion from a manually handled vector to
2223 xbt_dynar_t on the critical path.
2224 xbt_dynar_foreach calls functions, inducing stack management crap.
2226 We inlined these functions and xbt_dynar_foreach is now breath taking.
2227 We also inlined xbt_swag_belong on the way.
2229 Here are some approximate speedup measurements (on master/slaves
2230 simulations lasting between 10s and 20s each):
2231 3.3.1 -> 3.3.2: about same performance
2232 3.3.2 -> 3.3.3: 40% speedup
2233 3.3.1 -> 3.3.3: 40% speedup
2234 3.3.1 with inline patch -> 3.3.3: 30% speedup
2236 Our reading is that the refactoring which occurred in 3.3.2 made us
2237 suffer much more from the xbt_dynar_foreach low performance, but
2238 once we solved this, this refactoring proved to be very performance
2239 effective. From the 40% speedup, somehow, 10% are due to the
2240 inlining and 30% to the refactoring.
2242 That's a pitty that gcc cannot inline functions placed in other files
2243 alone. We have to choose between:
2244 - break the encapsulation (by putting private data structures and
2245 accessors in headers files to help gcc)
2246 - live with low performance
2247 - switch to a decent compiler such as icc (not quite possible).
2249 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2251 SimGrid (3.3.2) stable; urgency=low
2253 The "Simplicity does not preceed complexity, but follows it" release.
2255 The main contributors of this release were (lexical order):
2256 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2259 * Extract the routing logic into its own object.
2260 (was dupplicated in network.c and workstation_LV07.c;
2261 Allows to implement other ways of storing that info)
2262 => kill now useless network_card concept
2263 - Use dynar to represent routes (instead of void** + int*)
2264 - kill link_set (use surf_network_model->resource_set instead)
2265 - Add a command-line option to choose the routing schema to use
2266 - Add three new models:
2267 * Floyd (shortest path computed at initialization)
2268 * Dijikstra (shortest path recomputed all the time)
2269 * Cached Dijikstra (shortest path computed on need)
2270 All these models where contributed by Silas De Munck, and are
2271 described in his ICCS09 paper.
2273 * Simplify model declaration
2274 (less redirections, less function to write when defining a model)
2275 - Factorize stuff between models:
2278 surf_model_resource_set(model)
2279 surf_model_resource_by_name(model, name)
2280 - Unify the types of models in s_surf_model_t (using an union)
2281 - Embeed fields of common_public directly into s_surf_model_t
2282 - Rename model methods:
2283 action_free ~> action_unref
2284 action_change_state ~> action_state_set
2285 action_get_state ~> action_state_get
2286 - Change model methods into functions :
2287 (model)->common_public->action_use ~> surf_action_ref
2289 * Implement a generic resource; use it as ancestor to specific ones
2290 (allows to kill duplicated code in models)
2291 Drawback: timer command don't need no name nor properties;
2292 workstation_CLM03 don't need no properties
2293 (but I guess we can live with those few bytes wasted)
2295 * Improve the action object model
2296 - implement a constructor avoiding dupplicated code about field
2297 initialization in generic_action part.
2299 * Kill the SDP model: it has an external dependency, is deprecated
2300 in flavor of modern lmm models, and didn't compile since a while
2303 * Relocation of the context module from XBT to SIMIX.
2304 (the context were decoupled from the simix processes, duplicating a lot of code)
2305 => a lot of code was factorized
2306 - less overhead is introduced during scheduling
2307 - simpler API for the context factory
2308 - the logic for process creation,destruction and manipulation was simplified
2309 * Simplification of the s_smx_process_t data structure.
2310 => accesing the simix level data associated to a process is faster now,
2311 and the code is a lot more readable.
2314 * Implement some more MPI primitives:
2315 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2316 -implementation: Bcast: flat or 2-ary tree (default),
2317 Barrier: 4-ary tree,
2319 Allreduce: Reduce then Bcast
2320 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2321 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2322 Alltoallv: flat tree, like ompi
2324 * Add support for optimized collectives (Bcast is now binomial by default)
2325 * Port smpirun and smpicc to OS X
2328 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2329 Shout out if you used it.
2332 * Display the list of still queued messages in SG mode when existing
2336 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2337 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2338 * Add xbt_str_from_file(FILE*)
2339 * Add xbt_dict_get_key achieving a linear reverse search
2340 * Remove the context module
2342 Portability report of this version:
2343 * Main portability targets:
2344 - Linux(debian)/x86/context
2345 - Linux(debian)/x86/pthreads
2346 - Linux(debian)/amd64/context
2347 - Linux(debian)/amd64/pthreads
2348 On these, we still have the eratic breakages of gras/pmm and
2349 amok/saturate_sg reported in previous version. We still think
2350 that the tests are the cause of the fault, not the tested code.
2352 - Mac OSX Leopard/x86/context
2353 Still false negative in tesh autotesting.
2354 Smpi still fails, but this time because readlink does not accept -f
2355 Everything seems to work properly beside of that.
2358 - AIX version 5.3 (only tested contexts this time)
2359 Smpi still fails there because mktemp is not installed.
2360 Everything seems to work properly beside of that.
2362 I managed to compile it for the first time, but several breakages.
2363 Won't delay the release for this exotic platform.
2365 * Windows: it's still lagging behind. If you want to help, please
2368 Timing report of this version:
2369 This version seem to be more than 5% faster than 3.3.1 (on linux
2370 64bits with contextes). The gain is less than expected, we are
2371 investigating this for next release.
2373 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2375 SimGrid (3.3.1) stable; urgency=low
2378 * Implement a --cfg-help to show existing configuration variables
2379 * Build chain do not require doxygen in maintainer mode
2382 * fix a bug on struct sizeof computation, which prevented the
2383 exchange of arrays of structs in some conditions
2384 - added a regression test about this in datadesc_usage
2385 * Allow the exchange of 0-long dynamic vectors.
2386 - for that, use -1 as indicator of dynamic size instead of 0
2387 - This implied to change any size from unsigned long to long,
2388 reducing a bit communication abilities, but I guess that with
2389 64bits being quite common, this is more than enough.
2390 - This also induce a protocol change, thus bumping network protocol
2391 version from 0 to 1 (if we have external users, we have to get
2392 clean on that point too ;)
2393 - added two regression tests about this in datadesc_usage
2394 * Be more verbose when propagating local exceptions
2395 This helps debugging.
2396 * Display the status of simulated processes when receiving SIGINT in
2400 * Allow to control the simulation from a trace file.
2401 New functions MSG_action_register() and MSG_action_trace_run()
2402 The first one allows to associate a function execution to each
2403 kind of action while the second one parses a trace file and
2404 triggers the corresponding actions within the system.
2405 For now, only a toy example is provided in examples/msg/actions
2406 * Add an exemple of process migration in examples/msg/migration
2407 * Fix a bug in task exchange which broke MSG_task_get_sender()
2408 Add a teshsuite regression test for that.
2409 [Bug: if MSG_task_get_sender() is called after sender exit,
2411 * Fix a bug which prevented suspend/resume to work properly
2412 * Display the status of simulated processes when receiving SIGINT
2413 This fixes a regression of v3.3. due to the introduction of SIMIX
2414 * Bug fixing in failure management:
2415 - trace could not start by a failure at time 0
2416 - failure during communications were not working
2419 * Add SIMIX_process_set_name() to change the name of the current
2420 process in the log messages.
2421 * Store smx_hosts in a dict since we only retrieve them by name
2422 * Move the configuration infrastructure to surf
2425 * Move the configuration infrastructure to surf
2428 * Massive internal cleanups:
2429 - Store internal structures on processes instead of hosts (allows
2430 to have more than one process per host, in addition of being more
2432 - Cleanup the initialization/finalization process
2433 - Kill a whole bunch of unneeded synchronization:
2434 processes run in exclusive manner within the simulator
2435 - Move queues from global tables to process data fields
2437 - now accept -platform and -hostfile arguments
2438 - Pass the right rank value to processes according to the hostfile
2439 * Compile the examples by default, and use them as regression tests
2440 * Implement MPI_Wtime()
2441 * Change the reference speed to a command line option
2444 * TCP_gamma can now be specified as command line option using
2445 --cfg=TCP_gamma:10000000.0
2446 * Change the --surf-path cmd line option into --cfg=path:
2449 * Also include strbuff from xbt.h public header
2450 * xbt_ex_display(): do not free the exception after displaying
2451 This allows to do more with the given exception afterward.
2452 Users should call xbt_ex_free() themselves.
2456 Portability report of this version:
2457 * Main portability targets:
2458 - Linux(debian)/x86/context
2459 - Linux(debian)/x86/pthreads
2460 - Linux(debian)/amd64/context
2461 - Linux(debian)/amd64/pthreads
2462 These targets fail about 1/10 of times on gras/pmm, but we believe
2463 that this is because of the test, not because of simgrid.
2464 amok/saturate_sg fails even more rarely, and the test may not be
2467 - Mac OSX Leopard/x86/context
2468 The test suite still spits tons of errors because some obscure
2469 force prevents us from removing the temporary directories
2470 arguing that they still contain some metadata I've never heard of.
2471 Smpi fails because seq is not installed.
2472 Everything seems to work properly beside of that.
2475 - AIX version 5.3 (both contexts and pthread)
2476 Smpi still fails there because mktemp is not installed.
2477 XML inclusions seems rosty on AIX.
2479 * Windows: it's still lagging behind. If you want to help, please
2482 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2484 SimGrid (3.3) stable; urgency=high
2488 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2489 [Malek Cherier & Mt]
2491 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2494 Until now, GRAS were using MSG as an interface to SURF. It was
2495 quite difficult because both interface have several differences
2496 (MSG channels vs GRAS sockets were the most notable point).
2498 This also opens the gate to SMPI (which should occur soon) and speed
2499 up simulations by to 40% (even if it were not the main goal).
2501 **************************************
2502 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2503 **************************************
2504 The point is that events occuring at the exact same timestamp are
2505 not scheduled in the same order with the old and new version. This
2506 may be enough to completely change the execution of simulations in
2507 some cases. Sorry for the inconvenience.
2509 * Cleanup and upgrade the XML format to push further scalability
2510 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2512 * Improve the testing infrastructure with tesh. Now a very large part of
2513 the code is tested not only by being run but also by checking that the
2514 output match an expected output [Mt].
2516 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2517 is really less memory-demanding, which should allow you to use
2518 larger files in SimGrid [AL].
2520 * Inform valgrind about our contextes, so that it becomes usable
2521 with the default (and more effecient) version of SimGrid
2522 [contributed by Sékou Diakite, many thanks]
2525 * Introduce a listener thread in charge of receiving incoming
2526 messages from the network. It allows to overlap communication and
2527 computation but most notably, it removes some stupid deadlocks due
2528 to the fact that so far, a process could not send and receive at
2529 the same time. This made most non trivial communication schema
2531 * Convert the PIDs from long int to int to match the MSG ones (and
2532 linux ones too) [Mt]
2533 * New function: gras_agent_spawn() to launch a new process on
2534 current host. Only working in simulation for now. [Mt]
2535 * New function: gras_os_hostport() returning a constant form (ie,
2536 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2539 * Make the backtrace of exceptions more human readable [Mt]
2540 * New module: xbt/str [Mt]
2541 a ton of string utility functions (split, join, printf to a newly
2542 allocated buffer, trim, etc)
2543 * New module: xbt/hash [Mt]
2544 SHA1 hashing algorithm (more to come if needed)
2545 * New module: xbt/synchro [Mt]
2546 synchronization tools (mutex and conditions) working the same way
2547 in simulation and in real life (mainly useful for GRAS, but not
2549 * New module: xbt/queue [Mt]
2550 classical producer/consumer synchronization scheme
2551 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2552 (using the classical functions will get serialized) [Mt]
2553 * Make dictionary internal table dynamic. No need to specify its size
2554 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2556 * Make sure the log channels are organized as a tree under windows
2557 (because of ANSI C compatibility issue, any channel were child of
2558 root directly) [Mt].
2561 * Cleaned many thing in surf and fixed a few bugs [AL].
2562 * Add a nice command line configuration mechanism to compose models [AL].
2563 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2564 the previous one (KCCFLN05). It relies on something that looks like
2565 a max-min sharing mechanism but cannot be written as such. A new solver
2566 was thus designed [AL].
2567 * Add a new solver to lmm. Based on Lagrange optimization and
2568 gradient-based descent, it enables to efficiently maximise systems s.a
2570 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2572 This solver enables to propose two new network models for TCP Reno and
2573 TCP Vegas based on Low's work. These models still need to be fully
2574 tested though [Pedro Velho].
2577 * Bug fix in SD_simulate. Now the time bound given as argument is
2579 * Use the new parallel task model (ptask_L07) as default.
2580 * Use the SURF command line configuration mechanism.
2581 * 0-size tasks (for synchronization) should now work.
2583 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2585 SimGrid (3.2) stable; urgency=high
2589 We still experience issues on this platform, but we believe that at
2590 least MSG is usable.
2592 GRAS API BREAKAGE (for simplification purpose, sorry):
2593 * the gras_msgtype_by_name is not used anymore. Instead of
2594 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2595 you can write (and must)
2596 gras_msg_send(toserver, "request", &request);
2597 - If you still want to pass a gras_msgtype_t to the function (to cache
2598 the type and avoid the lookup time), use the gras_msg_send_() variant.
2599 - Impacted functions:
2600 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2601 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2602 * The callbacks are now expected to return 0 when everything went well
2603 (just like the main() function)
2605 GRAS new features and improvements:
2606 * New module mechanism where user code can use per process globals [Mt]
2607 This is similar to gras_userdata_*() functions, but for libraries. It
2608 factorize some code developped over and over in the examples and AMOK.
2609 It has still to be documented and used (only amok/peermanagement is
2611 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2612 loss when they were small enough to fit into the buffer and sent quickly
2613 enough so that they can all get received in one shoot.
2614 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2615 if not found. Use the *_or_null() variant for the old semantic.
2616 * In gras_msg_handle, do not discard messages without callback.
2617 They are probably messages to be explicitly awaited later (ie, proofs of
2618 mis-synchronization in userland since they are sent before being awaited)
2619 No big deal usually.
2620 * gras_socket_meas_send/recv: semantic changed!
2621 The numerical arguments used to be (1) the total amount of data to send
2622 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2623 messages. This was need for the fool willing to send more than MAXINT
2624 bytes on quite fat pipes.
2627 * Do really rename the hostmanagement module to peermanagement. [Mt]
2628 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2629 but this is rather new and this was documented in the module
2630 documentation (poor excuses, I admit)
2631 * Bandwidth measurement semantic changed! This follows the changes to
2632 gras_socket_meas_send/recv explained above.
2635 * A sequential mode has been added to the workstations. When a workstation
2636 is in sequential mode, it can execute only one task, and the other tasks
2637 are waiting in a FIFO. [Christophe Thiery]
2640 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2641 model for SIMDAG. [Christophe Thiery]
2642 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2643 equal to 0, which caused some bad behaviors in
2644 saturated_constraint_set_update. I now use a threshold mechanism like in
2648 * When running manually src/testall, you select specific units [Mt]
2649 testall is the result of our cunit mechanism, and should replace all
2650 the scripty thingy around since bash don't run easily on billware.
2652 * A mallocator system has been added. [Christophe Thiery]
2653 Mallocators allow you to recycle your unused objects instead of freeing them
2654 and allocating new ones.
2656 Documentation update:
2657 * FAQ reworking + New FAQs:
2658 - "Valgrind spits tons of errors!" [Mt]
2659 - "How to repport bugs" [Mt]
2660 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2661 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2662 - Communication time measurement within MSG [AL]
2663 - I experience weird communication times when I change the latency [AL]
2664 * GRAS tutorial [Mt]
2666 - an introduction to the framework and to the used communication model
2667 - an initiatic tour introducing the most proheminent features:
2669 . Lesson 0: Installing GRAS
2670 . Lesson 1: Setting up your own project
2671 o Part 2: Message passing
2672 . Lesson 2: Exchanging simple messages
2673 . Lesson 3: Passing arguments to the processes (in SG)
2674 . Lesson 4: Attaching callbacks to messages
2675 . Lesson 5: Using globals in processes
2676 . Lesson 6: Logging information properly
2677 . Lesson 7: Using internal timers
2678 . Lesson 8: Handling errors through exceptions
2679 . Lesson 9: Exchanging simple data
2680 . Lesson 10: Remote Procedure Calling (RPC)
2681 . Lesson 11: Explicitely waiting for messages
2682 . Recapping of message passing features in GRAS
2683 - A HOWTO section containing:
2684 o HOWTO design a GRAS application
2685 More are due, of course. They will come latter. In the meanwhile, you can
2686 check the examples which are still here.
2688 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2690 SimGrid (3.1) stable; urgency=high
2694 There was a stack corruption somewhere, visible only when optimizing
2695 with these versions. [Vince]
2698 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2699 module that enabled to play with Directed Acyclic Graphs. It is built
2700 directly on top of SURF and provides an API rather close to the old
2701 SG. Some old codes using SG are currently under rewrite to check that
2702 all needful functions are provided. [Christophe Thiery]
2705 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2706 extension of the classical CLM03 model that gracefully handles
2707 failures. This is now the default model for MSG and GRAS. It doesn't
2708 handle parallel tasks yet though. [AL]
2709 * Bug fix: Weights were not correctly set in the network part.
2710 WARNING: This may have resulted in incorrect results with simulations
2711 where there are more than one flow on a given link. [AL]
2714 * After a (long ?) discussion on simgrid-devel, we have decided that the
2715 convention we had on units was stupid. That is why it has been decided
2716 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2717 WARNING : This means that all previous platform files will not work as
2718 such with this version! A warning is issued to ask users to update
2720 A conversion script can be found in the contrib module of the CVS, under
2721 the name contrib/platform_generation/surfxml_update.pl [MQ]
2724 * Bug fix: Processes were started in reverse order, wrt deployment file.
2725 WARNING: if your code relies on this bug, please fix it. [AL]
2726 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2727 being executed on two different locations. [AL]
2728 * Bug fix: Failures are now better supported thanks to Derrick's tests
2729 (there was many failure situations I hadn't thought of and that weren't
2730 correctly handled). [AL]
2731 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2735 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2736 Exception occurring on server-side are propagated back to client (!).
2738 API CHANGE: the callback changed their prototype. Change:
2739 int my_handler(gras_socket_t expeditor, void *payload_data) {
2741 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2742 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2744 * New! function: gras_msg_handleall to deal with all messages arriving
2745 within a given period.
2746 * New! function: gras_socket_server_range to get a server socket in a
2747 range of port numbers (ease to avoid port number conflicts) [MQ]
2748 * New! gras processes display their backtrace when they get a SIGUSR1
2749 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2750 Sweet to debug RL processes [MQ]
2754 - Do not force experiment sizes to be expressed in kb, or it becomes
2755 impossible to measure the latency this way (needs one byte-long tests)
2756 WARNING: this changes the amok_bw_* function semantic. [MQ]
2757 - Implements the link saturation stuff. [MQ]
2758 * Peer management module:
2759 New! module factorizing code that we wrote over and over [MQ].
2762 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2763 - Test units are placed directly into the library code, they get extracted
2764 automatically and placed into the src/testall binary.
2765 - Convert most of the XBT tests to this system.
2766 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2767 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2768 live in the same C code. This required to change a little bit the API
2769 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2770 * New module: sparse graph structure with basic algorithms (this is work
2771 in progress and the API is not considered to be frozen yet). [AL]
2772 * Display more information on backtraces: source line & function names are
2773 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2774 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2775 * Reworked a little bit some #include statements to load only required
2776 headers. Some user code that relied on SimGrid to include stdlib or
2777 stdio may need to include it by themselves. [AL]
2778 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2780 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2781 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2782 but I guess that almost nobody used those parts. [MQ]
2784 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2786 SimGrid (3.0.1) stable; urgency=low
2789 * Unfortunately, I had missed 5 misnamed functions:
2790 xbt_fifo_item_t xbt_fifo_newitem(void);
2791 void xbt_fifo_freeitem(xbt_fifo_item_t);
2792 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2793 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2794 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2795 They're now deprecated. Please use their new versions:
2796 xbt_fifo_item_t xbt_fifo_new_item(void);
2797 void xbt_fifo_free_item(xbt_fifo_item_t);
2798 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2799 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2800 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2802 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2803 * Documentation: xbt_log module unmercifully reworked [MQ]
2804 * Bugfix: there was a problem with the ending of contexts with
2805 the pthread backend. It caused some weird deadlock or behavior
2806 depending on the pthread implementation. [AL]
2807 * Bugfix: get the exceptions raised in the simulator repport where
2808 and why they come from when they are not catched in time [AL, MQ]
2811 * Bugfix: Do repport the error when two non-connected hosts try to
2812 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2815 * Add additionnal checkings on communications. Assert that two
2816 communicating hosts are connected by a set of links... [AL]
2819 * Add additionnal checkings on channel values in communication [AL]
2820 * New: MSG_task_get_source to see on which host a task was generated [HC]
2821 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2822 the number of tasks waiting to be received on channel and sent
2824 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2825 waits for the first task coming from a given host.. [AL]
2827 GRAS new functionnalities: [MQ]
2828 * Enhance the parsing macro to allow the size of multidimentional objects
2829 to be given thru annotations.
2830 * New example (and documentation): Matrix Multiplication a la RPC
2831 (as when I was young!) and fix a bunch of bugs found on the way.
2833 GRAS performance improvements: [MQ]
2835 * Reduce the amount of cbps creation/destruction by making it static to
2836 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2839 * Change libdata to a set so that we can search for stuff by ID (and thus
2840 reduce the insane amount of dict lookups)
2843 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2844 RL communication on top of SG since it's so uneffective.
2845 It may also be used for inter-thread communication in RL, one day.
2846 * Use gras_datadesc_copy() to exchange messages on top of SG
2848 - improve message exchange performance on top of SG
2849 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2850 * Don't exchange on the network the size of the used part of buffer,
2851 instead, specify the possible buffer size to read().
2853 - reduces the amount of read/write calls (one pair per exchange)
2854 - reduces the amount of exchanged data (the size)
2855 - allows to retrieve all arrived data on receiver side, if we don't need
2856 it right now (subsequent read will peek the buffer)
2857 - allows the receiver to proceed with the begining of the stream before
2858 everything is arrived
2859 - make it possible to build an iov transport (using readv/writev)
2861 - take care of the data with non-stable storage (like stacked data),
2863 * If possible, TCP send uses vector I/O (when writev() is here)
2864 - Don't use it for receive since we send data sizes and data on the
2865 same stream, so we wouldn't be able to chain large amount of chunks
2866 before having to flush the stuff to read the size.
2867 * Rework the transport plugin mechanism to simplify it and reduce the
2868 amount of pointer dereferencement when searching for the right function
2871 * I guess that now, we do almost as few system calls as possible while
2872 doing as few data copy as possible.
2874 To improve it further, we could try to send all the sizes first and then
2875 all the data (to use iov on receiving size), but it's only a partial
2876 solution: when you have 2 dimensional data, the sizes of the second
2877 dimension is data of the first dimension, so you need 3 streams.
2879 I'm not sure the potential performance gains justify the coding burden.
2881 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2883 SimGrid (3.00) stable; urgency=high
2886 * New! Give the possibility to hijack the surf parser and thus bypass
2887 MSG_create_environment and MSG_launch_application. Have a look at
2888 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2890 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2892 SimGrid (2.96) unstable; urgency=low
2897 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2898 This deprecates the xbt_error_t mechanisms.
2899 It modifies (simplifies) all XBT and GRAS API.
2900 MSG API keeps unchanged (exceptions raised by XBT are catched from
2901 within MSG and masked with existing error handling facilities)
2904 * New! Add a FATPIPE model. [AL]
2905 * New! Add a parallel task model. [AL]
2906 * New! Add automatically a loopback interface (in the default
2907 network model) if none was precised.
2910 * Bugfix: MSG_process_resume now works with the current running process.
2912 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2913 * Modification of MSG_task_get_compute_duration. Once a task has been
2914 processed, the value returned by this function is now equal to 0. [AL]
2915 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2916 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2917 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2918 of a m_task. For now, MSG_task_cancel only works with computation
2920 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2921 of the processor (in Mflop/s) regardless of the current load on the
2923 * API Change: use proper naming convention for MSG_getClock and
2924 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2926 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2927 This function changes the priority of a computation task. This priority
2928 doesn't affect the transfer rate. A priority of 2 will make a task
2929 receive two times more cpu power than the other ones. This function
2930 has been added to suit the needs of Nguyen The Loc and hasn't been that
2931 much tested yet. So if it fails, please report it and send me your code.
2933 * API Change: removed all functions and types that were marked "deprecated"
2934 since many months. Renamed MSG_global_init_args to MSG_global_init.
2936 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2938 SimGrid (2.95) unstable; urgency=low
2941 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2942 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2943 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2946 * Bugfix: Allow absolute paths for platform description files [MQ]
2947 * Bugfix: do free the variables after use. Leads to drastic performance
2949 * Implement max_duration (ie, timeouts) on resources [AL]
2952 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2954 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2955 the needed support (timeouts) [AL]
2958 * Implement measurement sockets. You can now get the bandwidth between two
2959 hosts thanks to AMOK (see below). [MQ]
2960 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2961 dynar over the network (yeah) [MQ]
2962 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2963 * simple/static token-ring example. [Alexandre Colucci and MQ]
2964 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2965 in place before [MQ]
2968 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2969 type parsing. "short" and "long" modifiers were also ignored; other
2970 modifier (such as reference level) are still ignored. [MQ]
2971 * Embeed the buffer size within the buffer itself on SG. [MQ]
2972 That way, send() are atomic and cannot get intermixed anymore (at least
2973 the ones which are less than 100k; bigger messages still have the issue)
2974 * Array size pushed by the field, not by the field type (or each
2975 and every long int will push stuff to the cbps) [MQ]
2976 * use select() to sleep since it allows to portably sleep less than one
2979 GRAS (minor cleanups)
2980 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2983 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2984 * type category 'ignored' killed as it was never used and were difficult
2986 * whether a type can cycle or not is now a flag, leaving room for more
2987 flags (as "ignored", if we feel the need one day ;)
2988 * Rename raw sockets to measurement sockets since "raw" has another
2989 meaning in networking community.
2992 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2993 and offers features not belonging to GRAS but that most applications
2994 need. One day, it may be a set of plugins loadable at runtime.
2995 * New module: bandwidth
2996 bandwidth measurement between arbitrary nodes running this module. [MQ]
2998 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3000 SimGrid (2.94) unstable; urgency=low
3002 The first beta release of SimGrid 3 !
3006 * Update the main page and the FAQ. Adding references to gforge.
3009 * Add a gras_os_getpid function.
3012 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3013 * Extend the logs so that they also print PID, hostname, date, ... if
3015 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3016 and kill the old version which were in testsuite/
3017 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3021 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3022 for users. Instead, each programming environment (GRAS, MSG,...) use it
3023 when needed to provide such a feature to users.
3024 Don't shortcut the mechanism or you will also shortcut the virtualization
3025 you need on the simulator.
3029 * Cleanups in configury with regard to compile optimization/warning flags.
3030 Also add -fno-loop-optimize to any powerpc since it's the optimization
3031 killing gcc (< 3.4.0).
3032 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3033 needed by MSG examples complications
3034 * Borrow configury beautifications from PHP
3037 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3038 warning (thanks loris for stumbling into this one).
3039 * Bugfix: stop loading private headers (gras_config.h) from the public
3043 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3044 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3045 RL would imply a DNS resolver, which is *hard* to do in a portable way
3046 (and therefore delayed).
3047 * Implement a real timer mechanism and use it in timing macros. This allows
3048 to avoid rounding errors and get a 0.000005 sec precision in timing
3049 macros. While I was at it, various cleanups:
3050 - allow to declare more than one timed section per file (fix a stupid bug)
3051 - move some private declaration to the right place
3052 - merge conditional execution and timing macros into emulation module
3053 - document the module
3054 - make sure the module cleanups its mess on gras_exit
3055 * Documentation improvements:
3056 - (new) how to compile applications using GRAS
3057 - (new) emulation support (timing macros)
3059 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3061 SimGrid (2.93) unstable; urgency=low
3063 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3066 - Use Paje properly where used. Still to be sanitized properly.
3067 - Portability fix: Add an implementation of the contexts using pthread
3071 - Add xbt_procname(): returns the name of the current process.
3072 Use it to show the current process's name in all logging.
3074 - fix detection of older flex version and the reaction, since we do
3075 depend on modern ones (we use lex_destroy)
3076 - Better separation of SG and RL in the libs: remove all simulation code
3077 from libgras. As a result, this lib is now only 200k when stripped.
3078 Some of the xbt modules may also be duplicated (two sets and such) and
3079 should be cleaned/killed before SG3.
3080 - Insist on using xlC on AIX because of weird problems involving gcc there.
3081 - Cleanup the make remote stuff. This is now done by scripts
3082 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3083 mainly for our private use, but we're working on changing them to user
3086 - Bugfix: flush the socket on close only if there is some *output*.
3087 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3089 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3092 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3094 SimGrid (2.92) unstable; urgency=low
3096 Alpha 3 on the path to SimGrid 3
3100 - New! First try of benchmarking macros.
3101 - New! First try so that gras_stub_generator generate deployment and
3102 remote compilation helpers.
3104 - Bugfix: Initialization fix in msg_test.
3108 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3110 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3112 - API Change: gras_os_sleep to take the amount of seconds as a double.
3113 Accepting an int was error prone since it was the only location where
3114 seconds were coded as such. It leaded to damn rounding errors.
3115 - Bugfix: Hard to belive that timers ever worked before this.
3117 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3119 SimGrid (2.91) unstable; urgency=low
3121 Alpha 2 on the path to SimGrid 3
3125 - Bug fix in the lmm_solver.
3127 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3128 through the function MSG_paje_output.
3129 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3130 - It is possible to bound the rate of a communication in MSG with
3131 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3132 to write it in the changelog).
3133 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3138 - Various cleanups to the autotools stuff
3139 - Begin to move Gras examples to examples/gras/
3140 - Let make distcheck work again (yeah!)
3142 - documentation overhauled using doxygen.
3143 gtk-doc-tools is dead in SimGrid now.
3144 - Automatically extract all existing logging categories, and add the list
3145 to the documentation (long standing one, to say the less)
3147 - Cleanup the known architecture table. Reorder the entries to group what
3148 should be, and use a more consistent naming scheme.
3149 (some of the test dataset are still to be regenerated)
3150 - New! Allow library to register globals on each process just as userdata
3152 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3153 lookup time (for now).
3154 Use it in msg and trp.
3155 This cleans a lot the internals and helps enforcing privacy of the
3156 headers between the gras components.
3157 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3158 - Bugfix: gras_os_time was delirious in RL.
3159 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3161 - Reenable GRAS now that it works.
3163 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3165 SimGrid (2.90) unstable; urgency=low
3167 Alpha 1 on the path to SimGrid 3
3169 * It is a long time since the last release of SimGrid. I'm sorry about
3170 that but as I had told you, I was rewriting a lot of things. I apologize
3171 to those who had been reporting bugs to me and that I had not answered.
3172 If your bug is still in the new version, please tell me. Here is a
3173 summary of the main changes.
3175 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3176 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3177 portability, speed, and a lot more but you'll figure it out later.
3178 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
3179 I wanted to release the new MSG as soon as possible and I have
3180 broken GRAS, which is the reason why, for now, only MSG is fully
3181 functional. A laconic description of these projects is available
3182 in the documentation.
3184 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3185 kernel optimized for our needs (called SURF but only the developers
3186 should use it). Hence, MSG is now roughly 30 times faster and I think
3187 that by rewriting a little bit MSG, I could event speed it up a little
3188 bit more. Beside the gain in speed, it is also much easier to encode a
3189 new platform model with SURF than it was with SG. More to come...
3191 * REVOLUTION 3: I have tried to change a little as possible the API of
3192 MSG but a few things really had to disappear. The main differences
3193 with the previous version are :
3194 1) no more m_links_t and the corresponding functions. Platforms are
3195 directly read from a XML description and cannot be hard-coded
3196 anymore. The same format is used for application deployment
3197 description. The new format is described in the documentation.
3198 Have a look in tools/platform_generation. There is a tiny script
3199 that converts from the old platform format to the new one. Concerning
3200 the application deployment format, parsing the old one is tricky.
3201 I think most of you should however be able to convert your files. If
3202 it is really an issue, I can write a C code that does the conversion.
3204 2) the toolbox tbx does not exist anymore. We now have a library
3205 with much more data-structures but without the hash-tables (we have
3206 dictionaries that are much faster).
3208 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3210 *****************************************************************************
3211 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3212 * don't feel like converting the oldies. (Mt) *
3213 *****************************************************************************
3216 Version 2.90: "the long awaited one"
3217 - Finished rewriting and debugging MSG. Rewrote the documentation.
3218 - disable GRAS for now since it needs to be ported to the newest SG
3221 - Finish the port to windows (using mingw32 for cross-compile)
3224 - Main loop and datastructures of SURF. A cpu resource object is
3225 functional. Surf can thus be used to create cpu's with variable
3226 performance on which you can execute some actions.
3228 2004-11-15 Martin Quinson
3229 - Port to ARM. Simply added the alignment and size descriptions. Should
3230 work, but the ARM machines are so slow that I didn't had the opportunity
3231 to 'make check' over there yet.
3233 2004-11-15 Arnaud Legrand
3234 - Trace manager now written. It uses a heap structure and is therefore
3235 expected to be efficient. It may however be speeded up (particularly
3236 when many events occur at the same date) by using red and black
3237 trees. One day maybe...
3238 - Max-min linear system solver written. It uses a sparse matrix
3239 structure taking advantage of its expected use. Most operations are
3240 O(1) and free/calloc are called as few as possible. The computation of
3241 the minimum could however be improved by using a red and black tree
3244 2004-11-03 Arnaud Legrand
3245 - Rename every gras_* function that was in xbt/ to its xbt_
3247 - Add a heap and a doubly-linked list to xbt
3248 - Added a dichotomy to the dictionaries. make check works as well before
3249 so I assume that the patch is correct. I do not know however if things
3250 run effectively faster than before now. :)
3252 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3253 SimGrid, and the version number is bumped to 2.x
3255 2004-10-29 Martin Quinson
3256 - Introduction of the remote errors.
3257 They are the result of a RMI/RPC on the remote machine.
3258 ErrCodes being scalar values, you can't get the host on which those
3259 errors did happen. Extending the error mechanism as in Gnome is possible.
3260 No idea yet whether it is a good idea.
3262 2004-10-28 Martin Quinson
3263 - Interface revolution: the Starred Structure Eradication.
3264 I used to do typedef struct {} toto_t; and then handle *toto_t.
3265 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3266 - ???_t is a valid type (builded with typedef)
3267 - s_toto_t is a structure (access to fields with .)
3268 - s_toto is a structure needing 'struct' keyword to be used
3269 - e_toto_t is an enum
3270 - toto_t is an 'object' (struct*)
3272 typedef struct s_toto {} s_toto_t, *toto_t;
3273 typedef enum {} e_toto_t;
3274 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3275 s_toto_t) is private.
3277 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3278 so that it can set the variable to NULL. It was so for dicts and sets,
3279 it changed for dynars.
3281 - Fix a bunch of memleaks in dict_remove
3282 - Fix a bug in sg/server_socket opening: it failed all the time.
3284 2004-10-07 Martin Quinson
3285 - Speed up dynar lookup operation a bit.
3287 gras_dynar_get is dead.
3289 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3290 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3291 (which gives you the address of the stored data).
3293 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3294 copy of the data using an affectation to do the job and not a memcpy.
3296 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3297 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3299 It does not really speedup the dynar test because they are
3300 setting elements all the time (and look them seldom). But the dict does
3301 far more lookup than setting.
3303 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3305 2004-10-05 Martin Quinson
3306 - Allow to (en/dis)able the cycle detection at run time.
3308 Whether we should check for cycle or not is now a property of each
3309 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3310 datadesc_cycle_unset allow to remove this property when previously set.
3312 Note that the cycle detection is off by default since it impacts the
3313 performance. Watch the data you feed GRAS with ;)
3315 This property is hereditary. Any element embedded in a structure having it
3316 set have it set for the time of this data exchange.
3318 You should set it both on sender and receiver side. If you don't set it on
3319 sender side, it will enter an endless loop. If you forget on receiver
3320 side, the cycles won't be recreated after communication.
3322 - Header reorganization.
3323 Kill gras_private.h, each submodule must load the headers it needs.
3325 2004-10-04 Martin Quinson
3326 - Interface revolution: do not try to survive to malloc failure.
3328 Now, gras_malloc and friends call gras_abort() on failure.
3329 As a conclusion, malloc_error is not a valid error anymore, and all
3330 functions for which it was the only gras_error_t return value are
3331 changed. They now return void, or there result directly.
3332 This simplify the API a lot.
3334 2004-09-29 Martin Quinson
3335 - Re-enable raw sockets.
3336 Created by gras_socket_{client,server}_ext;
3337 Used with gras_raw_{send,recv}
3340 It should allow to kill the last bits of gras first version soon.
3342 This is not completely satisfactory yet (duplicate code with
3343 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3346 - Simplify transport plugin (internal) interface by not passing any
3347 argument to _server and _client, but embedding them in the socket
3350 2004-09-28 Martin Quinson
3351 - Finish the port to AIX.
3352 autoconf was my problem (segfault within the malloc replacement
3353 function. No idea why)
3355 2004-09-16 Martin Quinson
3356 - Fix some size_t madness on 64bit architectures.
3358 2004-09-08 Martin Quinson
3359 - Reduce the number of system headers loaded, overload some more system
3360 calls (such as malloc to cast the result of the system one, and work
3362 - Fix and reintroduce the config support
3364 2004-09-07 Martin Quinson
3365 - Source code reorganization to allow Arnaud to surf all over there.
3366 - Allow to document the logging categories.
3367 - Remove all uppercase from logging categories and useless cleanup in names.
3369 2004-08-18 Martin Quinson
3370 Version 0.6.2 (protocol not changed; API changed)
3371 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3372 gras_error_t), and NULL when not found. Functions expecting a msgtype
3373 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3374 hopefully usefull message.
3375 - Portability to prehistoric sparcs again
3377 2004-08-17 Martin Quinson
3378 Version 0.6.1 (protocol not changed; ABI not changed)
3379 - prealloc some buffers to speed things up
3381 2004-08-11 Martin Quinson
3382 Version 0.6 (protocol not changed; ABI expended)
3383 - The parsing macro can deal with the references, provided that you add
3384 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3386 2004-08-09 Martin Quinson
3387 Version 0.5 (protocol not changed; ABI changed)
3388 - Allow to off turn the cycle detection code in data exchange at
3389 compilation time. It should be at run time, but I'm short of time (and
3390 the config stuff is still broken). That way, we keep dict out of the
3391 critical path, which is good because the performance is poor:
3392 - search not dichotomial yet
3393 - dynar give no way to access their content and memcpy everytime
3394 - In composed data description (struct, ref and so on), stop foolness of
3395 keeping the subtype's ID, but store the type itself. This keeps sets out
3396 of the critical path, which is good since they rely on dynar and
3397 dictionnaries. The only loose of that is that we cannot detect the
3398 redeclaration of a structure/union with another content (but I'm not sure
3399 the code detected well this error before anyway). We still can detect
3400 the redefinition discrepancy for the other types.
3401 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3402 it breaks the code because of type-punning used all over the place).
3403 This breaks on all non-gcc architectures (for now).
3405 All those changes (plus the buffer of last time) allow me to gain 2 order
3406 of magnitude on cruel tests consisting of 800000 array of integers on two
3407 level of a hierarchical structure (200 secondes -> 4 secondes)
3410 - the selector of reference must now return the type it points to, not
3411 the ID of this type.
3413 2004-08-06 Martin Quinson
3414 Version 0.4 (protocol changed; ABI not changed)
3415 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3416 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3417 - Implement a buffer transport, and use it by default (it relies on tcp in
3418 real life and on sg in simulation).
3419 That's a bit hackish since I had a new field to the structure to store
3420 its data without interfering with the subtype ones. Inheritance
3421 is tricky in C. And that's a kind of reverse inheritance with one class
3422 derivating two classes. Or maybe a game with java interfaces. Anyway,
3423 that's damn hard in C (at least).
3424 Moreover, I got tired while trying to ensure plugin separation and
3425 genericity in SG mode. MSG wants me to do weird things, so let's go for
3426 cruel hacks (temporarily of course ;).
3427 See comment in transport_private.h:71
3428 - do not include all the _interface headers in private but in the files
3429 which really need them (to cut the compilation time when they are
3432 2004-07-26 Martin Quinson
3433 Version 0.3 (protocol not changed; ABI changed)
3434 - Major overhault of the datadesc interface to simplify it:
3435 - shorted the function names:
3436 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3437 - add a trivial way to push/pop integers into the cbps without malloc.
3438 This allows to make really generic sub_type description, which simply
3439 pop their size of the stack.
3440 - add a function gras_datadesc_ref_pop_arr() which does what users want
3441 most of the time: Declare a dynamic array (which pops its size of the
3442 stack) and declare a reference to it. Poor name, but anyway.
3443 - kill the post-send callback, add a post-receive one
3445 2004-07-23 Martin Quinson
3446 Version 0.2 (protocol changed; ABI changed)
3447 - add some testing for cpbs in the test cases, and fix some more bugs.
3448 This invalidate again the little64 data file, since I cannot regenerate
3450 - remove an awfull optimization in the logging stuff, allowing me to:
3451 - understand it again
3452 - learn gcc how to check that the argument match the provided format
3453 - fix all errors revealed by gcc after that
3454 - internal keys of dict are not \0 terminated. Deal with it properly in
3455 loggings instead of segfaulting when the user want to see the logs :-/
3457 2004-07-22 Martin Quinson
3458 - Fix some stupid bug preventing cbps (callback postit) from working
3460 2004-07-21 Martin Quinson
3461 - Some documentation cleanups
3462 - remove the useless last argument of msgtype_declare
3463 - rename the Virtu functions to fit into the 'os' namespace
3464 - move headers src/include -> src/include/gras/ and stop fooling with
3465 gras -> . symbolic link
3466 - make distcheck is now successful
3468 2004-07-19 Martin Quinson
3470 - Build shared library also
3471 - Install html doc to the right location
3472 - stop removing maintainer files in make clean
3473 - build tests only on make check
3475 2004-07-13 Martin Quinson
3477 - No major issue in previous version => change versioning schema
3478 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3479 corresponding dataset.
3481 2004-07-11 Martin Quinson
3483 - Get it working with any kind of structure (we can compute the padding
3484 bytes remotely for all the architectures I have access to)
3485 - Implement the structure parsing macro (still not quite robust/complete)
3486 - Improvement to the remote testing toysuite
3488 2004-07-10 Martin Quinson
3489 [autoconf mechanism]
3490 - get ride of a bunch of deprecated macros
3491 - actually run the test for two-compliment, not only compile it :-/
3492 - test whether the structures get packed (and bail out if yes. Damn.
3493 Alignment is a serious matter)
3494 - test whether the structures get compacted (but respecting the alignment
3495 constraints of each types)
3496 - test whether the array fields of structures can straddle alignment boundaries
3498 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3499 (took me 2 hours to find that bug, looking at the wrong place)
3500 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3501 before sending/receiving and are used to compute the offsets of fields
3502 - Given that padding size depend even on compiler options, keep track of
3503 alignment and aligned_size only for the current architecture. Not a big
3504 deal since we send structure fields one after the other (seems
3506 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3507 program, let it work on linux/gcc/little32. portability todo.
3509 2004-07-08 Martin Quinson
3510 - import and improve remote compilation support from FAST
3511 - make sure make check works on half a dozen of machines out there
3513 2004-07-07 Martin Quinson
3514 Let's say it's version 0.0.3 ;)
3515 - Implement conversions (yuhu!)
3516 - Let it work on solaris (beside conversion, of course)
3517 - Stupid me, using rand() to generate the conversion datatests in not wise.
3519 2004-07-06 Martin Quinson
3520 - Let make dist work, since I'm gonna need it to compile on remote hosts
3521 - Let Tests/datadesc_usage write the architecture on which the file was
3522 generated as first byte.
3523 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3524 architecture descriptions.
3525 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3526 on those architectures.
3527 - Optimization: send/recv array of scalar in one shoot
3529 2004-07-05 Martin Quinson
3530 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3532 - Plug a whole bunch of memleaks
3533 - each process now have to call gras_{init,exit}. One day, their log
3534 settings will be separated
3535 - Continue the code factorisation between SG, RL and common in Transport.
3537 2004-07-04 Martin Quinson
3539 - Redistribution between SG and RL.
3540 We wanna have to accept in SG, so move accepted related parts of RL in
3541 the common part. (more precisely, the dynar of all known sockets is no
3542 more a static in transport.c, but part of the process_data)
3544 [gras_stub_generator]
3545 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3547 2004-07-03 Martin Quinson
3548 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3549 them all up in one shot)
3551 - Fix the ugly case of reference to dynamic array.
3552 - New (semi-public) function gras_datadesc_size to allow the messaging
3553 layer to malloc the needed space for the buffer.
3555 - gras_socket_close now expect the socket to close (and not its address to
3556 put NULL in it after it). This is because the socket passed to handlers
3557 is one of their argument (=> not writable).
3559 - propagate the interface cleanup from last week in datadesc, ie remove a
3560 superfluous level of indirection. User pass adress of variable
3561 containing data (both when sending and receiving), and not of a variable
3562 being a pointer to the data. Let's say that I like it better ;)
3563 The price for that is constructs like "int msg=*(int*)payload" in
3564 handlers, but it's a fine price, IMHO.
3566 - Let it work in RL (yuhu)
3568 2004-06-21 Martin Quinson
3570 - porting SG plugin and SG select to new standards (works almost).
3571 - plug memleaks and fix bugs around.
3574 - cleanup the prototype of data recv and force users to specify when they
3575 want to handle references to objects. Test case working even for cycles.
3576 - plug memleaks. Valgrind is perfectly ok with this.
3578 2004-06-12 Martin Quinson
3580 - cleanup the separation between plugin and main code in plugin creation
3582 2004-06-11 Martin Quinson
3584 - Reput hook for raw sockets, needed for BW experiments
3585 - kill a few lines of dead code
3586 [Data description] Interface cleanup
3587 - gras_datadesc_by_name returns the searched type or NULL.
3588 That way, no variable is needed to use a type desc once, which makes
3590 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3591 two parameters were strings (field name, type name), leading to
3593 [Dicos] Interface cleanup
3594 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3595 This is consistant with the dynar API.
3597 2004-04-21 Martin Quinson
3599 - Porting to new standards.
3601 - interface cleanup.
3602 There is no bag anymore, no need to take extra provision to mask the
3603 pointers behind "ID".
3604 Better splitup of functions between files create/exchange/convert.
3605 This is still a bit artificial since convert and receive are so
3606 interleaved, but anyway.
3608 - add a queued message list to procdata (the ones not matching criteria
3610 - factorize some more code between SG and RL wrt procdata
3612 - use gras_exit in example to track memleaks
3613 - get rid of gs_example now that GS is properly integrated into gras
3614 - update run_test to integrate the latest tests (datadesc)
3616 - rename WARNINGn macros to WARNn since it prooved error-prone
3618 2004-04-19 Martin Quinson
3620 - register init/exit functions within gras module mechanism
3621 - send/receive function.
3622 Convertion is not implemented, but short-cutted if not needed.
3623 struct/array elements are sent one by one (instead of block-wise), but
3624 nobody really cares (yet). Get a prototype before optimizing.
3625 - tests (using a file socket) for DD send/receive on:
3626 - base types: int, float
3627 - array: fixed size, string (ie ref to dynamic string)
3628 - structure: homogeneous, heterogeneous
3629 - chained list, graph with cycle
3630 Believe it or not, valgrind is not too unhappy with the results. The
3631 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3635 - Bugfix when using a filename explicitely (instead of '-')
3637 2004-04-09 Martin Quinson
3639 - factorize more code between RL and SG in socket creation
3640 - Complete the implementation and tests of:
3642 o file (only in RL, and mainly for debugging)
3644 I lost 3 days to design a portable address resolver, and then decided
3645 that the prototype mainly have to run on my box.
3646 Addressing portability too early may be like optimizing too early :-/
3648 - use gras_init in the Tests instead of the crappy parse_log_opt
3649 (the latter function is removed)
3650 [Conditional execution]
3651 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3652 [Code reorganisation]
3653 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3654 Build examples against the RL library, since there is no way to disable
3655 its creation for now.
3657 For information, the beginning of coding on GRAS was back in june
3658 2003. I guess that every line has been rewritten at least twice since