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 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
15 - Replay: The replay file has been re-written in C++.
16 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
17 supported. They are stored in the trace and used when replayed.
20 - Config: the C API is now deprecated (will be removed in 3.23), and
21 the C++ API has been slightly improved.
24 - #143: Setting a breakpoint at a given time
25 - #258: daemonized actors hang after all non-daemonized actors have completed
27 ----------------------------------------------------------------------------
29 SimGrid (3.19.1) Released March 22 2018
31 The Fixed ABI Release.
33 As you may know, we are currently refactoring SimGrid in deep.
34 Upcoming SimGrid4 will be really different from SimGrid3: modular,
35 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
37 Our goal is to smooth this transition, with backward compatibility
38 and automatic update paths, while still progressing toward SimGrid4.
40 SimGrid remains open during works: The last pure SimGrid3 release was
41 v3.12 while all subsequent versions are usable alpha versions of
42 SimGrid4: Existing interfaces remain unchanged, but the new S4U
43 interface is budding and the internals are deeply reorganized.
45 Since 2015, we work hard to reduce the changes to public APIs. When
46 we need to rename a public library symbol in S4U, we let your
47 compiler issue an explicative warning when you use the deprecated
48 function. These messages remain for four releases, i.e. for one full
49 year, before turning into an error. Starting with v3.15, your can
50 also adapt to API changes with the SIMGRID_VERSION macro, that is
51 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
53 Starting with this v3.19.1, our commitment to reduce the changes to
54 the public interfaces is extended from the API to the ABI: a program
55 using only MSG or SimDag and compiled against a given version of
56 simgrid can probably be used with a later version of SimGrid without
57 recompilation. We will do our best... but don't expect too much of
58 it, that's a really difficult goal during such profund refactoring.
60 The difference between v3.19 and v3.19.1 is that the former was
61 accidentally breaking the ABI of MSG, while the later is restoring
64 S4U and kernel APIs will still evolve until SimGrid4, with one-year
65 deprecation warnings as currently. In fact, cleaning up these
66 interfaces and converting them to snake_case() is one release goal of
67 v3.20. But don't worry, we are working to smooth this upgrade path.
69 In summary, new projects should start with S4U to benefit of the
70 future, but old MSG projects should still be usable with no change.
72 ----------------------------------------------------------------------------
74 SimGrid (3.19) Released March 20 2018
76 The Moscovitly-cold Spring release.
78 SMPI user-visible improvements:
79 - Support of MPI_Type_create_subarray
80 - Unimplemented MPI calls now abort (instead of issuing a warning)
81 Some harmless calls around errhandlers still only issue a warning.
82 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
83 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
84 is now possible (mmap privatization was refactored to this end).
85 Adding new SMPI instances should also be technically doable now.
87 SMPI internal cleanups:
88 - Removed index notion from SMPI -- no more "getPid() - 1"!
89 In tracing files, actors are called rank-<PROCESS-ID> but these are
90 NOT the real ranks (tracing several communicators is not supported yet).
91 - Deprecate smpi_process_index() (will be removed in 3.22)
92 - Better testing on classical proxy apps, through an external project:
93 https://github.com/simgrid/SMPI-proxy-apps/
96 - Use a graphical TOC to make it easier to find the documentation you need
97 - Revamp the MSG tutorial
100 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
101 - Remove parameter of MSG_process_killall().
102 Resetting the PID was bogus anyway (several actors could have the same PID).
105 - Execution->setHost() can be called after start(), to migrate it.
106 - Comm::test_any() is now implemented.
107 - s4u::Actor now has onCreation() and onDestruction() signals.
108 - Install some kernel header files for the users' plugins and more.
109 Warning, their API is really not stable yet.
110 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
111 include/simgrid/kernel/routing/*Zone.hpp
114 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
116 Build System and other cleanups:
117 - Remove unused run-time parameter "tracing/onelink-only".
118 - SimGrid now works with intel compilers (MC still buggy)
119 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
120 - tesh kills subprocesses on timeout (but not on windows)
123 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
126 - #194: Feature request: simgrid::s4u::Comm::test_any()
127 - #245: migrating an actor does not migrate its execution
128 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
129 - #256: Modernize FindSimGrid.cmake
130 - #257: Fix (ab)use of CMake install
132 ----------------------------------------------------------------------------
134 SimGrid (3.18) Released December 24 2017
136 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
138 MSG IS NOW DEPRECATED. Please use S4U for new projects.
139 - Support for MSG should not be removed from SimGrid before 2020,
140 but future evolutions will be limited to the integration of
141 user-provided patches.
144 - s4u::onDeadlock() signal, run before stopping the simulation.
145 - s4u::Exec: asynchronous executions (abstraction of background
147 - s4u::Host->getLoad() returns the achieved speed in flops/s
150 - Link::name() is deprecated and will be removed in v3.21, use
151 Link::getCname() instead.
152 - Mailbox::getName() changed to return a std::string, use
153 Mailbox::getCname() instead to get a char*.
154 - Storage::getName() changed to return a std::string, use
155 Storage::getCname() instead to get a char*.
156 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
157 to have both allocation and deallocation of the map in user space.
160 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
161 The amount of flops remaining to do is a vector, not a scalar.
162 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
163 name implies on both sequential and parallel tasks.
164 - Both changes fix GitHub's #223 using PR #237 as a basis.
165 Thanks Michael Mercier.
166 - Most examples were converted to S4U and hidden elsewhere at tests.
167 Rationale: we still want MSG to work; we want newcomers to use S4U.
170 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
171 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
174 - Switch to the faster dlopen privatization mechanism by default
175 - Documentation improvments
178 - Live migration is getting moved to a plugin. Dirty page tracking is
179 the first part of this plugin. This imply that VM migration is now
180 only possible if one this function is called :
181 - C/MSG: MSG_vm_live_migration_plugin_init()
182 - C/C++: sg_vm_live_migration_plugin_init()
183 - Java: Msg.liveMigrationInit()
184 For C and C++, "simgrid/plugins/live_migration.h" also has to be
188 - Define class simgrid::xbt::Path to manage file names.
189 - Removed unused functions:
190 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
191 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
192 xbt_graph_export_graphviz()
193 - xbt/str.h: xbt_str_join()
194 - Remove unused datatypes:
195 - xbt/heap.h: use std::priority_queue or boost::heap instead
196 - xbt/swag.h: use boost::intrusive::list instead
199 - New link_energy plugin for the consumption of the links.
200 - All of the operations on files and storage contents have been
201 packaged into a plugin (src/plugins/file_system). The current
202 public interface can be found in
203 include/simgrid/plugins/file_system.h
204 To use these functions you now have to initialize the plugin by
205 calling MSG_storage_file_system_init() just after calling
206 MSG_init() or sg_storage_file_system_init() just after creating
210 - Remove the undocumented/untested tag <include>
213 - Remove viva specific tracing as the tool is no longer maintained
216 - #248: Kill of finished processes leads to segfault
217 - #240: xbt_cond_wait_timeout should gracefully return for C apps
218 - #239: please implement signal s4u::onDeadlock()
219 - #230: segfaults when exit() without run()
220 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
221 - #223: MSG_task_get_flops_amount() not working with parallel tasks
222 - #222: Actor::kill() doesn't really kill and segfaults
223 - #221: odd LMM warning when killing an actor
224 - #120: Memory leak when the processes are forcefully killed
226 ----------------------------------------------------------------------------
228 SimGrid (3.17) Released October 8 2017
230 The Drained Leaks release: (almost) no known leaks despite the tests.
232 Even very long simulations will preserve your memory: our testsuite
233 stresses SimGrid for over 45mn on fast machines for a coverage of
234 over 80%, and there is only one single known leak, of about 4kb.
238 - Comm.detach(): start and forget about asynchronous emission. A cleanup
239 handler may be given to free resources if the comm cannot be completed.
240 - this_actor::send(mailbox) is now mailbox->put()
241 - New: simgrid::s4u::Comm::wait_all()
242 - New: Host.execute() for remote executions.
245 - Unused option network/sender-gap is removed.
248 - Deprecate MSG_task_isend_with_matching(): This unused feature
249 really complicates our internals. Will be removed in v3.20.
252 - Improved context termination. It is now possible to free resources, even
253 when a process is forcibly killed.
254 - ContextBoost: add support for Boost versions above 1.61.
257 - Bring back run-time option --cfg=exception/cutpath to remove exception
259 - Removed unused functions:
260 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
261 xbt_str_rtrim(), xbt_str_trim().
262 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
265 - Removed header files obsolete since SimGrid 3.12:
266 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
268 - #3: SD_exit should be made optional
269 - #120: Memory leak when the processes are forcefully killed
270 - #159: Threading test regression in Actor refcounting
271 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
272 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
273 - #186: Actor::killAll() segfaults if some process is blocked on wait()
274 - #191: VM migration and pstate
275 - #192: Updating the energy consumptions of all hosts crashes with VMs
276 - #195: All actors have PID=0 in the logs
277 - #204: Sometimes segfault with thread contexts and mmap privatization
278 - #222: Actor::kill() doesn't really kill and segfaults
279 - #225: Actor::kill() doesn't really kill when victims are doing a join()
281 ----------------------------------------------------------------------------
283 SimGrid (3.16) Released June 22. 2017.
285 The Blooming Spring Release: developments are budding.
288 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
289 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
290 - Java JAR file should be finally fully working on Mac OSX too.
292 XML platforms: Switch to platform v4.1 format.
293 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
294 - <zone> can be used as a synonym for the now deprecated <as>
295 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
296 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
297 - <actor> can be used as a synonym for the now deprecated <process>
298 - state_file and avail_file periodicity is now easier to express
299 (check the documentation) the old behavior should still work.
300 * Storage: not backward compatible, but it seems that nobody ever used it
301 - Remove attribute 'content_type' of <storage_type>: was never used
302 - Make attribute 'model' of <storage_type> optional: for future usage
303 - Remove Bconnection model property: was never (in)validated. Replaced by
304 the maximum of the read and write bandwidth as a resource constraint.
307 - New and Backwards Compatibility break:
308 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
309 When one wants to get the list of tasks whose states have changed during a
310 simulation round, s/he has to allocate and free a dynar and use it as
311 argument to this function. The former SD_simulate (double how_long)
315 - Allow multicore VMs, with the correct sharing (unless you overcommit)
316 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
319 - New function to update the consumption of all hosts at once.
320 - Fix the model for multi-core hosts, linear in the amount of busy
321 cores with an abnormality for pIdle (see paper).
322 BUG: Still not satisfactory for ptask on mono-cores.
325 - The netzone are now available from the MSG API.
326 The old names still work, but are now deprecated.
329 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
330 - New option: smpi/keep-temps to not cleanup temp files
331 - New option : smpi/shared-malloc-blocksize . Relevant only when global shared
332 mallocs mode is used, allows to change the size of the fake file used
333 (default 1MB), to potentially limit the number of mappings for large runs.
334 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
335 - Fortran ifort and flang compilers support
336 - New RMA calls supported (experimental) :
337 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
338 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
339 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
340 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
341 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
342 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
345 - New: this_actor::isMaestro() returns whether we are in kernel mode.
346 - Behind the scene, ActivityImpl are now automatically refcounted.
347 Removing this nasty bottleneck will greatly simplify our internals.
350 - Replay: New function xbt_replay_action_get():
351 Retrieve the function previously associated to an event type.
352 - DROPPED FUNCTION: xbt_str_varsubst()
353 - DROPPED MODULE: strbuff. We don't need it anymore.
354 - DROPPED MODULE: matrix. We don't need it anymore.
355 - DROPPED MODULE: lib. We don't need it anymore.
357 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
359 SimGrid (3.15) Released March 22. 2017
361 The Spring Release: continuous integration servers become green
362 We fixed even the transient bugs on all target architectures:
363 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
364 Windows, FreeBSD, NetBSD.
366 - Rename NetCards to NetPoints.
367 This was intended to help NS3 users, but that's not a netcard.
368 That's a point in the routing algorithm, let's avoid wrong simplifications.
371 - New: MSG_process_yield(). Stop and yield to other processes.
372 - New: MSG_process_daemon(). Daemon processes are automatically killed
373 when the last non-daemon process terminates
374 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
375 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
376 to make room for the upcoming network energy plugin.
377 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
380 - Ensure that an actor can kill itself with Process::exit()
381 - Kill the obscure NativeException. Nobody want to survive the issues
382 it denotes, so use JniException that is a RuntimeException (not to
383 be caught explicitly).
384 - Partial bug fix in initialization. SimGrid flags on command line were
385 consumed at C level but stayed in the original Java String[] args.
386 This could mess users' args[i] if SG flags were not put at the end of
388 The SimGrid flags are now removed from the Java arguments. However,
389 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
390 if args.length is greater than the number of YOUR OWN ARGUMENTS.
391 It might be if you have --log or --cfg flags in the command line.
392 - Fix numerous memleaks all around the place. In particular, around VMs.
396 - s4u::onPlatformCreated: right before the simulation starts
397 - s4u::onSimulationEnd: right after the main simulation loop
398 - s4u::onTimeAdvance: right after a clock change (time is discrete)
399 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
400 event from the availability_file changes the avail speed.
401 - Links are now usable from s4u
402 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
403 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
404 - New examples: The conversion of MSG examples to S4U has begun
405 - Actors: create, kill, migration, and suspend
406 - Applications: master-worker and token-ring
407 - Action replay: communications and storage
408 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
411 - Backwards Compatibility breaks
412 - The SD_route_get_list and SD_route_get_size functions have been removed.
413 They are replaced by the unique
414 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
415 The route from src to dst is built in the links dynar whose size can
416 be obtained with xbt_dynar_length.
417 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
418 sg_host_route_bandwidth and sg_host_route_latency.
419 Macros ensure the backwards compatibility, but you should fix your code
422 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
423 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
424 - Bump our claim of support from MPI 1.1 to MPI 2.2.
425 We don't support 100% of it, but it should be enough. Contact us if not.
426 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
427 naive and probably inaccurate)
428 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
429 mapped onto a unique small file using some system magic.
430 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
431 - Fortran: cleanups, fixes, support of user-added operations
432 - RMA: MPI_Accumulate are applied in correct order.
433 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
434 - Better support for MPI_IN_PLACE
435 - Support for MPI_Win attrs and keyvals.
436 - Support MPI_Comm_group_create, MPI_Type_size_x
437 - MPI ops should be properly applied to all allowed types
438 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
439 algorithm by a non failing one from mpich.
442 - Kill the fifo data container: we don't use it anymore.
445 - A <cluster> can now be created with different speed values to represent pstates.
446 see examples/platforms/energy_cluster.xml for an example.
448 SimGrid (3.14.159) Released December 28. 2016
450 The previous release was only a crude approximation of a Pi release;
451 this one should be a bit better.
453 - Revert a last minute change that broke on Mac OSX.
454 - Fix the changelog and NEWS files.
455 - Improve the documentation, in particular of the routing module.
456 - Integrate some patches flying around in Debian.
458 SimGrid (3.14) Released December 24. 2016
461 * The whole documentation was reworked and reorganized. There is
462 still much room for improvement here, but we're on it.
465 * We now need python3 for our testing infrastructure.
466 * Model checking needs libevent
469 * Now works on FreeBSD
472 * Privatization now works on FreeBSD
474 * Privatization is now activated by default in smpirun
475 You can turn it off with -no-privatize if something goes wrong.
477 * Call-location tracing for SMPI.
478 You can add the exact location (filename / linenumber) of an MPI call to
479 your trace files and slow down or speed up the simulation between two
480 consecutive calls by using an adjustment file (see the documentation).
482 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
483 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
484 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
485 values were ignored in some cases. The timings of these functions can now
486 be significantly different.
488 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
489 smpi/running-power is renamed to smpi/host-speed
491 * smpi/grow-injected-times option to enable or disable multiplication of the
492 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
493 make simulation less precise (but also much faster).
495 * smpirun script should be (much) faster for large deployments.
497 * SMPI tracing : fixed issue with poor matching of send/receives.
499 * Replay : Fix broken waitall
501 New functions and features
502 * MSG_parallel_task_execute_with_timeout, to timeout computations.
504 Dropped / renamed functions and features
505 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
506 - MSG_mailbox_is_empty() -> Mailbox::empty()
507 - MSG_mailbox_front() -> Mailbox::front()
508 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
509 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
510 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
511 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
512 * Don't pass the free_f parameter to property related functions:
513 - MSG_host_set_property_value()
514 - MSG_as_router_set_property_value()
515 - MSG_storage_set_property_value()
516 * VM properties. Since msg_vm_t are msg_host_t, just use
517 MSG_host_get_property_value() and friends
518 * VM I/O related things:
519 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
520 - Unimplemented save/restore methods
521 * MSG_as_router_get_property_value() was redundent with
522 MSG_environment_as_get_property_value().
523 - Removed MSG_as_router_*propert*() functions
524 - Added MSG_environment_as_set_property_value() for consistency
525 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
526 Well, they are still there for now, but deprecated with a warning.
527 Please switch to xbt_dict_new_homogeneous() before this is removed
529 * Task affinity. Its intended behavior (that was very badly tested
530 and probably not really working) was deceiving what most users
531 would have hoped here.
532 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
533 * Option network/coordinates is now useless and should be dropped.
536 * Added option storage/max_file_descriptors to allow more than 1024 files opened
539 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
540 * Added Dragonfly topology support
543 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
544 EiBps, EBps, Eibps, Ebps for bandwidth.
545 They may become useful to some lucky ones.
548 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
549 * ex/app_tokenring: new example, very similar to the MSG Token Ring
550 * ex/async_waitAll: new example, on asynchronous communications
553 * Memory usage should be decreased for simulations with a large number
554 of processes. This also helps for SMPI.
556 SimGrid (3.13) stable; urgency=low
558 The Half Release, a.k.a. the Zealous Easter Trim.
560 - We removed half of the lines, that were mostly experimental cruft.
561 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
562 (not counting blanks and comments -- according to openhub.net)
563 - The internals are now compiled in C++ (and will soon be clean C++)
564 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
565 59 klines of C, etc. We added only 29 klines of C++ in replacement.
567 * Backwards Compatibility breaks
568 - Removed Lua simulation bindings (switch to C or Java for that).
569 Lua can still be used to describe platforms
570 - Removed Java kernel plug-ins.
571 Will be reintroduced after the ongoing major internals reorg.
573 - the following functions were removed.
574 They were too specific and should be reimplemented in a generic
575 way, with filter function.
576 - MSG_task_listen_from_host
577 - MSG_mailbox_get_count_host_waiting_tasks
578 - MSG_mailbox_put_with_timeout was removed.
579 Please use MSG_task_send_with_timeout instead.
581 - the SD_application_reinit function was removed. It has been a noop for a while.
582 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
583 replaced by a more flexible mechanism.
584 - The following functions thus do not exist anymore
585 - SD_workstation_get_access_mode
586 - SD_workstation_set_access_mode
587 - SD_workstation_get_current_task
588 - Basic estimation functions have been removed but can easily be replaced
589 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
590 - SD_workstation_get_computation_time => amount / sg_host_speed()
592 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
593 Use VM.getSpeed()*load/100 for the legacy behavior.
595 - option enable_tracing was removed. It was not doing anything for a while.
596 - In the ModelChecker:
597 - the model-checker now ptraces the model-checked process which means
598 you cannot use a debugger on the latter anymore (we might make this
599 optional in the feature);
600 - removed soft-dirty page tracking;
601 - remove model-checked side snapshot management,
602 MC_snapshot() and MC_compare_snapshot();
603 - keep the MC_cut() function as a stub (it was not really working
604 in the previous release).
607 * All options are consistently in kebab-case. Old names are kept as alias.
610 * Switch to platform v4 format.
611 - Rename from 'power' to 'speed' the attributes describing the amount of
612 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
613 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
614 - In <host> and <link>, attributes availability and state are gone.
615 It was redundent with state and availability traces, and with peak values.
616 - In <cluster>, attributes availability_file and state_file are gone.
617 It was too complex and unused.
618 - Kill <gpu>. Was not doing anything.
619 - The DOCTYPE points to the right URL:
620 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
621 (the file at this address now documents the changelog since its v1)
622 - A warning is emitted for unit-less values (they are still accepted).
623 - speed. Default: 'f' or 'flops'. Also defined:
624 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
625 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
626 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
627 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
628 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
629 - latency. Default: 's' second. Also defined:
630 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
632 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
633 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
636 * s4u::Host is now the preferred public interface to the Host features.
637 sg_host_* functions are C bindings to the exact same behavior
638 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
641 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
642 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
643 They were deprecated since a few years, and probably did not even compile anymore.
646 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
647 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
648 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
649 encouraged to update their codes. A list of the modified functions can be found at the end of
650 include/simgrid/simdag.h
653 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
654 the lambda or closure passed as a parameter will run in kernel mode.
655 All the callback functions should be rewritten to that interface at some point.
658 * Reorganizing and cleaning the internals all around the place.
661 * Remove old default barrier/bcast buggy algorithms (see #18407)
662 * Various bug fixes to handle more codes
663 * Remove the need for the --foreground option of smpirun (it is still
664 accepted for backward compatibility).
667 * Kill the setset data container: MC don't use it anymore.
668 * Kill the queue data container: it made more sense with GRAS.
669 * Kill the xbt_peer_t data type: it's useless without GRAS.
670 * Kill rm_cb feature of config sets: it was never useful.
671 * Kill graphxml parsing feature. It was not used.
672 * Kill the deprecated code protected by XBT_USE_DEPRECATED
674 - xbt_dynar_sort_strings(), when the content is char*
675 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
676 They throw exceptions on invalid input;
677 * C++ support for declaring CLI flags (simgrid::config::Flag);
678 * class for abstracting different signal backends (simgrid::xbt::signal).
679 with no external dependencies (we need very simple signals).
682 * refactoring and cleanup of the code;
683 * ongoing process to cleanly separate the model-checking algorithms
684 from the code model-checking support.
686 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
688 SimGrid (3.12) stable; urgency=low
690 The Facelift Release.
693 * Require g++ v4.7 at least to not speak prehistorical C++.
694 * Require Boost 1.48 (for signal2 component).
695 * Java must be version 7 at least when activated.
696 * Builds on Windows again (including Java bindings).
697 * Tracing is now always enabled (no way to turn it out)
698 * Remove GTNetS. It was not working anyway.
699 * Various cleanups in the cmake scripts.
700 * Move headers around to sort them out on installed systems:
701 - instr/instr.h -> simgrid/instr.h
702 - instr/jedule/* -> simgrid/jedule
703 - simdag/datatypes.h was removed
704 - simdag/simdag.h -> simgrid/simdag.h
705 - msg/datatypes.h was removed
706 - msg/msg.h -> simgrid/msg.h
709 * Interface improvement:
710 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
711 - Sanitize the interface in MSG_task_ module:
712 - Merge two functions that were close enough but misleading:
713 set_compute_duration(t) -> set_flops_amount(t)
714 get_remaining_computation(t) -> get_flops_amount(t)
715 - set_data_size(t) -> set_bytes_amount(t)
716 get_data_size(t) -> get_bytes_amount(t)
717 - Massive cleanups in the functions related to the energy
718 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
719 - New: MSG_host_get_pstate()
721 - msg/energy/onoff: switching hosts on and off
723 * Interface improvement:
724 - Rename Host.isAvail() to Host.isOn()
725 - Rename Process.currentProcess() to Process.getCurrentProcess()
726 - Rename Task.setDataSize() to Task.setBytesAmount()
727 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
729 - #18874: Actually allows the GC to reclaim tasks
733 - SIMIX_process_throw: raises an exception in a remote process
734 * Refactoring: Separate sub-modules
735 - libsmx: the public interface, as libc in a real system
736 - popping: the strange dance that converts a user request into a kernel handling
737 - smx_context_*: the virtualization mechanisms that embed the user code
738 - smx_*: the handling of each simcalls
740 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
741 * Rename smx_action_t into smx_synchro_t, making explicit that these
742 things are used to synchronize processes with their environment.
743 For example, a communication is a sort of synchronization involving
744 the communicating processes (that may block until the exchange) and
745 the platform. The same can be said from computations, etc.
747 - #18888: segfault when a process ends before its kill_time
751 - 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.
752 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
753 - MPI_Type_set_name, MPI_Type_get_name
754 - MPI_*_c2f and MPI_*_f2c functions
755 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
756 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
757 - Activate a lot of new tests from the mpich 3 testsuite
759 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
760 - InfiniBand network model added : Based on the works of Jerome Vienne
761 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
762 - When smpi/display_timing is set, also display global simulation time and application times
763 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
764 * Collective communications
765 - 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
766 - 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/).
767 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
769 - Replay now uses algorithms from wanted collective selector
770 - Replay can be used with SMP-aware algorithms
771 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
772 - 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.
774 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
775 - Correctly use loopback on fat-tree clusters
776 - Asynchronous small messages shouldn't trigger deadlocks anymore
777 * Energy/DVFS cleanup and improvement
778 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
779 - new: smpi_get_host_pstate()
782 - "Full" network optimization flag was broken since Surf++
783 - Better handling of precision flags in maxmin
784 - Fix bug causing sometimes "Impossible" errors
785 - Properly pass cluster properties to included hosts
786 * Improvement of the Energy plugin.
787 - Always update the consumption before returning that value
788 - New property: watt_off to denote the dissipation when the host is off
789 - New functions getWattMinAt and getWattMaxAt to retrieve the
790 dissipation of pstates that we are not currently at.
791 * Java: class NetworkLink renamed to Link
792 * New function: simcall_process_get_kill_time()
793 * Massive rename s/workstation/host/
794 - That's intrusive, but that's good for the project consistency. Sorry.
795 - Change config option "workstation/model" into "host/model"
798 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
799 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
800 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
801 - xbt_dirname and xbt_basename for non-POSIX systems
803 * The model checker now runs as a separate process.
804 * The model checker runs must now be launched with the new simgrid-mc program.
805 * Record/Replay: the MC can display a textual representation of a path in the
806 execution graph. It can then be replayed outside of the model checker.
808 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
810 SimGrid (3.11) stable; urgency=low
815 * Normalizing pointers addresses tool for better diff between logs
818 * Add cloud examples using new VMs
819 - examples/msg/cloud/two_tasks_vm.tesh
820 - examples/msg/cloud/simple_vm.tesh
821 - examples/java/cloud/cloud.tesh
822 - examples/java/cloud/migration/migration.tesh
823 * Add java surf examples:
824 - examples/java/surfPlugin/surf_plugin.tesh
825 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
826 - examples/java/surfCpuModel/surf_cpu_model.tesh
827 * Add SMPI+MSG example:
828 - examples/smpi/smpi_msg_masterslave/
833 - msg task destroy cancel test
834 - msg_host on/off test
835 * Move all tests in testsuite to teshsuite (adding tesh files)
836 * Restructure teshsuites
837 - one folder for each kind of test
838 * Restructure AddTests.cmake
840 - structure the order of tests (with sections)
843 * Add virtual machine
844 - creation of a VM on a PM
845 - migration of a VM from a PM to another PM
847 - MSG_process_join(msg_process_t process, double timeout)
848 - msg_bar_t MSG_barrier_init(unsigned int count)
849 - int MSG_barrier_wait(msg_bar_t barrier)
850 - void MSG_barrier_destroy(msg_bar_t barrier)
851 - msg_as_t MSG_environment_as_get_by_name(const char * name)
852 * New option "msg/debug_multiple_use" to help debugging when a task is used
855 - Improvements and finalization of MSG_storage, MSG_file APIs and their
857 - Increase code coverage in test suites
861 * Protect context stacks against stack overflow. The number of protected memory
862 pages allocated on the top of each stack (1 by default) can be configured
863 with the new command line option --cfg=contexts/guard_size:<value>.
864 * Simcalls are now generated by a python script that
865 - generates files included by SimGrid
866 - checks that all the functions exist, and proposes prototypes
868 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
869 host_set_data, host_get_data
871 - simcall_process_join(smx_process_t process, double timeout)
872 * Fix bug where sleeping processing could not be suspended.
875 * Translate surf models from C to C++
876 - Generic classes for all models: Model, Resource, Action
877 - A generic interface for each kind of model (CPU, Network, Storage
878 Workstation, WorkstationVM)
880 * Translate surf routings from C to C++
881 * Add callbacks using sigc++ or boost::signals2
882 - Add callback functions for resource creation/destruction
883 - Add callback functions for action state change
884 - Handle Energy as a plugin
885 * Replace swag by boost::intrusive
886 * Add new routing models for clusters. For documentation, see
887 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
888 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
889 parameters for cluster tag
890 - Fat trees, with topology="FAT_TREE" and
891 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
892 - see examples/platforms/torus_cluster.xml and
893 examples/platforms/fat_tree_cluster.xml
897 * Hostfiles support host:nb_processes construct to deploy several processes on
899 * Collective communication algorithms should not crash if used with
900 improper number of nodes and report the error.
901 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
902 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
903 MPI_Dims_create, MPI_Cart_sub are supported.
904 * New interface to use SMPI programmatically (still depends on MSG for
905 some parts, see examples/smpi/smpi_msg_masterslave) :
906 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
910 * Global variables privatization in MPI executables is now performed at runtime
911 with the option smpi/privatize_global_variables (default:no).
912 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
913 dynamic libraries loaded by the application are not privatized (static
914 linking with these libraries is advised in this case)
917 * Options defined in XML work correctly now.
920 * New cmake option, enable_lib_in_jar, to control whether native libraries are
921 copied into simgrid.jar or not (ON by default). Use this option if you want
922 to reduce the size of the installed simgrid.jar, *and* the native libraries
923 are kept installed elsewhere.
924 * Surf binding with SWIG (code generated in maintainer mode only):
925 - plugin to handle callbacks
926 - CPU model only for the moment
929 * Supernovae build mode is definitively removed. It was used to improve
930 inlining and inter-module optimizations. It is nowadays superseded by
931 link-time optimizations commonly available in compilers.
932 * Update ns-3 find lib. Bindings for ns-3 should work again now.
933 * Add boost dependency for surf++
934 * Add new macro for tests
935 - ADD_TESH(name <tesh_args>)
936 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
940 - xbt_bar_t XBT_barrier_init(unsigned int count)
941 - int XBT_barrier_wait(xbt_bar_t barrier)
942 - void XBT_barrier_destroy(xbt_bar_t barrier)
943 * Make the xbt_os_time module public
945 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
947 SimGrid (3.10) stable; urgency=low
949 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
952 * Reintegrate Java to the main archive as desynchronizing these
953 package is not acceptable anymore (Java is now considered stable)
954 * Add explicit synchronization facilities through semaphores
955 * Bug fix: Task.setDataSize() only changed the C world, not the value
956 cached in the Java world
959 * Dramatically change the way files are handled. API and internals changed, but
960 this part of MSG was not considered as production grade either.
961 * Add explicit synchronization facilities through semaphores
962 * Add a new function MSG_host_get_process_list()
963 * Preliminary DVFS support (see examples/msg/energy/ for details)
966 * SMPI is now included directly in the libsimgrid as the windows
967 linker doesn't force us on splitting it anymore.
968 * Improvements of the SMPI replay tool:
969 - Most of the collective communications are now rooted in the same process as
970 in the original application.
971 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
972 used until now). Multiple data types can now be used in a trace.
973 - The replay tool now supports traces produce either by TAU or a modified
975 - Bug Fix: the compute part of the reduce action is now taken into account.
976 - Gatherv collective is now supported
977 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
978 of smpirun outputs time independent traces for the current run. One file
979 is created per process. If too many processes are simulated, this behavior
980 can be changed to one file for all processes by using the
981 tracing/smpi/format/ti_one_file flag
982 * smpirun generates the host file if needed (with given host count and platform)
983 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
984 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
985 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
986 --cfg=smpi/coll_selector:(mpich/ompi)
987 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
989 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
990 happening inside a collective SMPI call.
991 * Fix the behavior of complex data types handling.
992 * Make MPI_Wtime another synchronization point to take computations into
994 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
995 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
996 * Add all missing Fortran bindings, SMPI should work with Fortran 90
997 (no privatization of global variables yet)
998 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1001 * Verification of liveness properties is now available for SMPI applications
1002 (in addition to MSG applications)
1003 * Bugged examples using SMPI in examples/smpi/mc/
1004 * Add --cfg=model-check/visited option. Allows the verification of infinite
1005 programs. Detection of loops in the execution thanks to the system state
1006 comparison and reduction of the state space to explore. Can be combined with
1007 DPOR for safety properties.
1010 * Allow to change SimGrid configuration (see --help) within the code
1011 thanks to SD_config() as it can be done in MSG.
1012 * Add a new function SD_task_set_amount() upon user request.
1015 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1016 * Remove rule based routing (no more PCRE dependency)
1017 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1018 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1019 * Add a loopback_bw and loopback_lat options to cluster tag.
1020 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1021 routing. Now peers are encapsulated in an AS and have their own private
1022 router but this is transparent.
1025 * Our own implementation of getline is renamed xbt_getline, and gets
1026 used even if the OS provide a getline(). This should reduce the
1027 configuration complexity by using the same code on all platforms.
1028 * New type: xbt_cfg_elm_boolean.
1029 * Allow to use yes/no for boolean configuration options in the command line.
1030 * Allow to disable SimGrid cleanups at exit from command line option.
1031 There are situations where one may want a simulation to end with an exit.
1032 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1033 annoying when scripting around the simulator. Adding a
1034 --cfg=clean_atexit:no allows to circumvent this issue.
1037 * Lots of memory leaks were corrected in this release.
1038 There is no known memory leaks anymore, in all of our 600+ tests.
1039 * New command line option --version, to get SimGrid version information.
1040 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1042 * Supernovae builds are deprecated, and expected to be removed in the next
1045 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1047 SimGrid (3.9) stable; urgency=low
1049 The Grasgory release: GRAS is really dead now.
1051 * Complete overhaul of the internal host structures scheme.
1054 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1055 was considered as experimental and badly maintained since then.
1056 * Keeping it was thus a trap to our potential users, that could take
1057 it instead of MSG or SMPI by mistake despite is pity state.
1058 * GRAS seems to have very few users (if any), and no one volunteered
1059 to maintain it further. It also induces a lot of XBT code (for
1060 portability sake), that must be maintained too.
1061 * For all these reasons, we killed GRAS. If someone wants to revive it
1062 in the future, don't cry, our git history still remembers of GRAS.
1065 * Major overhaul. Merge our documentation again as time proved that
1066 splitting it was really not helping our users.
1067 * Further improve the developer documentation to help newcomers
1068 hacking on SimGrid itself. The user documentation (and in
1069 particular, the beginner documentation) is still in a sorry state.
1072 * Now works on Windows too!
1073 * Much more extensive test suite, from MPICH
1076 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1077 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1078 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1079 * Bug fix: let task be scheduled when the last dependency to be solved is
1080 a control dependency.
1081 * Remove SD_load_environment_script function.
1082 Use the C sg_platf function if you want to declare a platform
1086 * New function: MSG_process_get_number()
1087 * Old function documented: MSG_config()
1088 * Remove MSG_load_platform_script function
1089 Use the C sg_platf function if you want to declare a platform
1093 * Change the default value of the TCP_gamma constant (maximal size of TCP
1094 congestion window) to a more realistic 4MiB value. If you notice changes in
1095 your simulation results, you can fall back to the previous 20k tiny window
1096 by adding --cfg=network/TCP_gamma:20000 on command line.
1097 * (Hopefully) fix a bug wrt periodic availability and state traces
1098 * Bug fix: use default values at start when first event in availability/state
1099 trace is not at time 0.
1102 * remove the "new_" part of function name sg_platf_new_trace_connect
1103 (resulting in sg_platf_trace_connect), since it does not create
1107 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1108 use case where it's really mandatory, and maintaining it was a pain in
1110 * New: xbt_fifo_search(), search an item with a user-provided
1111 comparison function instead of dumb pointer comparison.
1114 * Fix the lua deployment:
1115 Use `simgrid.init_application()` before deployment instead of
1116 `simgrid.msg_register_application()` after.
1119 * Transfer the tracing files into the corresponding modules.
1121 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1123 SimGrid (3.8.1) stable; urgency=low
1125 The "we are told that some people want to also *install* the simgrid
1128 * Add missing file "tesh.1" to the archive.
1130 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1132 SimGrid (3.8) stable; urgency=low
1134 The Psssshiiiit release: SimGrid jumps into the Cloud.
1137 * Add an experimental interface to manipulate VMs. They are mainly
1138 process groups with very few intrinsic semantic, but they should
1139 allow you to build the semantic you want easily.
1140 * New function: MSG_host_set_property_value()
1141 * New function: MSG_process_on_exit(). To clean memory in all cases.
1142 * Bug fixes that made the host (and link) failures unusable.
1143 * Add a way to auto-restart process when the host in which they are
1144 executing comes back (ON_FAILURE="RESTART" on deployment file,
1145 MSG_process_auto_restart_set).
1146 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1147 please stop using the old ones, they are DEPRECATED.
1149 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1150 Please use MSG_init() instead. (reducing the amount of entry
1151 points in the library helps us).
1152 * Make it impossible to link against the wrong version of the lib
1153 * Deprecate MSG_clean(). No need to call it anymore.
1154 * Function MSG_get_host_number() is not deprecated anymore.
1157 * Split the doc into a user guide and a reference guide.
1158 * Start a developper guide to help people hacking on SimGrid.
1161 * Enable tracing by default. This modules rocks you should use it.
1162 * Remove option custom_flags. Now use environment variables CFLAGS
1164 * Use default cmake things to detect lua instead of home grown ones.
1165 * New option "enable_mallocators" to disable mallocators, for debugging
1166 purpose ("on" by default).
1169 * Bug fixes around the resource failures: don't let the processes
1170 survive the host they are running onto.
1171 * Add an interface to auto-restart processes when the host in which they are
1172 executing comes back.
1173 * Ensures that SIMIX_clean is called automatically. It's not part of
1174 the public interface anymore (bindings should be updated).
1177 * Bug fix for when SD_Simulate is called with a positive value: be careful
1178 when comparing doubles. Sometimes they are different for non significant
1180 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1181 parallel task whose initial work is distributed among host according
1182 to the Amdahl's law. Such tasks are created with a parameter alpha
1183 that corresponds to the non-parallelizable part of the computation.
1184 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1185 between two sets of workstations assuming a 1D block distribution (each
1186 workstation owns a similar share of data) on both sides.
1188 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1189 Data redistribution will be automatically scheduled once parent and child
1190 are both scheduled. The filling of computation_amount and
1191 communication_amount structures is now done seamlessly thanks to the chosen
1193 * New function SD_workstation_dump to display various information
1194 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1195 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1196 bandwidth on the route onto which the task is scheduled and the amount of
1198 To divide the nominal bandwidth by 2, the rate then has to be :
1199 rate = bandwidth/(2*amount)
1200 * Compute tasks that have failed can now be rescheduled and executed again
1201 (from their beginning)
1202 * Increasing source code coverage (src/simdag is now covered at 95.8%
1206 * Re-implement time-independent trace replay using SMPI (at the
1207 smpi_smp_* level) instead of MSG. This should replace
1208 examples/msg/actions/actions.c
1209 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1210 hindexed and structs)
1211 * Implement the exchange of non-contiguous data.
1212 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1213 * Correct behavior of smpi/sender_gap and set its default value to 0
1214 * Add option to asynchronously send small messages to allow better
1215 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1216 specifies the size in bytes under which messages will be asynchronously sent.
1217 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1218 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1221 * New C interface to define a platform: XML is now optional.
1222 For more info, please check include/simgrid/platf.h.
1223 * New interface to define random platforms from the C:
1224 For more info, please check include/simgrid/platf_generator.h and
1225 examples/msg/masterslave/masterslave_platfgen.c
1226 * Export a sg_cmdline dynar containing all the arguments we got from
1230 * Two new tracing options for adding comments to trace file so you
1231 can track your experiments (see --help-tracing for details).
1232 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1233 * Adding the SimGrid version that generated the trace file as a comment.
1234 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1235 * Fix to avoid key clashes on Paje links
1236 * Other minor fixes related to the Paje specification
1239 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1240 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1241 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1242 * Greatly improve the robustness of mmalloc to user errors (such as
1243 using an area after freeing it, or freeing it twice)
1245 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1247 SimGrid-java (3.8.1) stable; urgency=low
1249 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1250 Warning: all methods are visible, but only some of them are
1251 implemented so far. Check the source (src/jtrace.c)
1252 for further information.
1253 * New module: org.simgrid.msg.File (SimGrid File management functions)
1254 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1255 * Change the meaning of Process.restart: now restart the process from
1256 the begining, like MSG_process_restart in C.
1257 * Add Process.setAutoRestart: handling of process restart when failed
1259 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1260 you to retrieve the properties of the processes/hosts
1261 * Deprecate Msg.clean(): you can just forget about it now.
1262 * New function Process.getCount(), that only works when compiling
1263 with the not yet released version 3.9 of the C library.
1265 * New context factory based on Coroutines. It mandates a modified JVM
1266 but then, the simulations run about five times faster, and there is
1267 no limit to the amount of processes (beside of the available memory).
1269 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1271 SimGrid (3.7.1) stable; urgency=low
1274 * Restore the prototype of MSG_process_create_with_environment() to
1275 the pre-3.7 situation by removing the kill_time argument.
1276 * Add a MSG_process_set_kill_time() function instead.
1279 * Fix weird behaviors when dealing with parallel tasks.
1282 * Simgrid is now built as a dll.
1283 * Simgrid-java now works on Windows.
1284 * Simgrid-Java is now included into Windows package.
1287 * First pre-build package for MacOSX.
1290 * Fix compilation when using MSG_USE_DEPRECATED.
1291 * Fix some compilation issues on Macs and Windows.
1292 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1293 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1295 We discovered that the Lua console is broken, but we are missing the
1296 manpower to fix it right now. The problem existed in 3.7 too, so we
1297 are not blocking the release for that. Sorry if you depended on this
1298 feature, any help would be really welcome.
1300 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1303 SimGrid-java (3.7.1) stable; urgency=low
1305 The "Java aint got to be bloated and slow" release
1308 * Various internal cleanups and performance improvement
1309 Simulations are expected to run up to twice faster or so
1310 * Make Process.kill(process) an instance method, not a static one
1311 * User processes are not java.lang.Thread subclasses.
1312 This breaks the compatibility (sorry), but previous API was
1313 brain-dead, making it impossible to have non-trivial
1314 initializations in the process constructor.
1315 * Require a full constructor per Process sub-class.
1316 Kinda breaks the compatibility (sorry), but this allows a much more
1317 efficient way to launch the processes at simulation startup.
1318 * Do not embeed our version of semaphores, java 1.5 can be considered
1319 as sufficiently prevalent for us to not dupplicate its features.
1323 * Add examples for almost every part of the API
1324 We spotted and fixed a lot of bugs in the process
1326 * New module: asynchronous communication API
1327 * New function: Process.sleep()
1328 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1329 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1330 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1332 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1334 SimGrid (3.7) stable; urgency=low
1336 The "spring cleanups (before the next Big Project kicks in)" release.
1339 * We can specify the SMPI latency/bandwidth factor with command line
1340 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1341 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1342 You can also use the "config tag" from platform file by adding this line
1343 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1344 (see "example/platforms/tag_config.xml" to use "config tag").
1345 Note that the command line supersedes the platform file configuration.
1346 * Change the correction factors used in LMM model, according to
1347 the latest experiments described in INRIA RR-7821.
1348 Accuracy should be improved this way.
1349 * Use the partial invalidation optimization by default for the
1350 network too. Should produce the exact same results, only faster.
1351 * Major cleanup in surf to merge models and split some optimization
1352 mechanisms from the core of the models. As a result you can now
1353 specify which model to use (e.g., --cfg=network/model:LV08
1354 --cfg=cpu/model:Cas01) and which optimization mode to use
1355 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1356 Incompatible combinations should err at initialization. See
1357 --help-models for the list of all models and optimization modes.
1358 * The CLM03 workstation model was dropped for simplicity because it
1359 used the deprecated CM02 network model. Use default instead.
1360 * Rename the TCP_gamma configuration option to network/TCP_gamma
1361 * Rename the coordinates configuration option to
1362 network/coordinates, and document it
1363 * Use now crosstraffic keyword instead of the terribly misleading
1364 fullduplex keyword. It is activated by default now in the current
1365 default model, use --cfg=network/crosstraffic:0 to turn it off.
1366 * Ongoing refactoring the model parsing to make XML files optional
1367 See include/simgrid/platf.h for details (still to be completed)
1370 * Major overhaul of the documentation. Almost instructive now :/
1371 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1372 functions and friends. This interface was considered as
1373 deprecated since over 2 years, it's time to inform our users that it is.
1374 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1375 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1376 use these (crappy) functions in your code.
1377 These functions will be removed soon. Stop using them now.
1378 * Deprecate MSG_get_host_{table,number}
1379 Implement MSG_hosts_as_dynar() instead.
1380 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1381 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1384 * Stabilize the parallel execution mode of user contexts
1385 * Introduce configuration variables to control parallel execution:
1386 - contexts/synchro: Synchronization mode to use when running
1387 contexts in parallel (either futex, posix or busy_wait)
1388 - contexts/parallel_threshold: Minimal number of user contexts
1389 that must be part of a scheduling round to switch to parallel
1390 execution mode (raw contexts only)
1391 * Fix bugs that prevented to use suspend/resume along with
1392 synchronization structures.
1393 * Fix bugs in process termination that lead to invalid memory access
1394 in very specific conditions.
1397 * Introduce a parallel mode for the models (controlled by surf/nthreads
1398 configuration item). In our tests, running the models in parallel
1399 never lead to any speedups because they are so fast that the gain
1400 of computing each model in parallel does not amortizes the
1401 synchronization costs, even when ultra fast futexes are used.
1402 This is released anyway because YMMV.
1405 * Performance boost by using a swag internally to compute the set of
1406 tasks that are finished and should constitute the return value of
1410 * Enable it by default now that it is considered rather stable.
1413 * Documentation of the tracing functions.
1414 * Performance gains when tracing categorized/uncategorized resource
1415 utilization by avoiding calls to get route when updating resource
1416 variables. LMM constraints are being used instead.
1417 * API changed to set task categories. Use MSG_task_set_category instead
1418 of TRACE_msg_set_task_category, and SD_task_set_category instead
1419 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1420 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1421 * Examples for link user variables added (see at examples/msg/tracing/)
1422 * Deprecated function TRACE_msg_set_process_category completely removed
1423 * Trace header updated according to the latest Paje file format
1424 * Tracing network lazy updates, no longer obligate users to use full updates
1425 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1426 * Experimental: let user code declare/set/push/pop application states for hosts
1427 * API changed to allow the manual creation of graph configuration files
1428 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1431 * Improve the API of Lua MSG bindings, using the Lua spirit.
1432 * Each simulated process now lives in its own Lua world (globals are
1433 automatically duplicated). It helps writing simulators. Will allow
1434 to run Splay programs within SimGrid in the future.
1435 * Add a Chord example in Lua, equivalent to the MSG one.
1438 * Start the implementation of a solution to express temporal
1439 properties, not only local assertions. This is still an
1440 experimental work in progress, stay clear from it to be safe.
1444 - Add new runtime parameters --help-logs and --help-log-categories
1445 to display information about supported logging parameters and
1447 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1448 don't exists anymore.
1449 * Mallocators: allow value NULL for the reset function.
1451 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1452 create homogeneous dictionaries, where all the elements share the
1453 same free function. Non homogeneous dictionaries will be
1454 deprecated in the next release.
1455 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1456 - Multi-level dictionaries are deprecated.
1458 - new function xbt_dynar_search_or_negative() that is useful when
1459 you have less than 2 million elements in your dynar and don't
1460 want of the extra complexity of catching exceptions when the
1461 element is not found.
1463 - Make xbt_os_thread module (for thread portability) public.
1464 Documentation is still to come, sorry.
1466 - Cleanups and simplifications to make it maintainable again.
1467 - Exotic features (such as memalign and valloc) were removed.
1468 - The metadata were extended and improved so that the
1469 model-checker becomes able to explore and inspect the heaps.
1470 - This may induce a performance drop when enable_model-checking is
1471 ON in cmake (even if it's not used in the simulation), but it is
1472 necessary at this point to get MC working.
1474 Turn model-checking OFF if simulation performance matters to you.
1475 Not enabling it at runtime is not enough, disable it in cmake.
1477 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1480 SimGrid (3.6.2) stable; urgency=low
1482 The "Not coding new stuff allows to polish old things" release.
1485 * New bindings to the NS3 packet level simulator (experimental)
1486 * Use the raw (efficient) execution contexts instead of the sysv
1487 (portable) ones when possible.
1488 * libpcre is now mandatory in any cases since not using it led to
1489 severe performance loss and possibly other issues
1490 * Update the XML platforms:
1491 - G5K: include the latest machine in Nancy
1492 - GridPP and LCG: new platforms
1493 * Documentation was partially updated, at least (more to come)
1495 Bug fixes, cosmetics and small improvements
1496 * Free terminated processes before the end of the simulation to avoid
1497 exhausting the memory of users having very dynamic amount of
1499 * Bug fix and cosmetics about canceling non-running tasks
1500 * Bug fix about the dot loader's issues when using libcgraph
1503 * Create an installer for windows with nsis (amd64 and win32)
1504 - Add an hello world project to illustrate simgrid project creation.
1505 - Embed libpcre into the Simgrid installer to avoid
1506 its compilation burden
1507 * The raw execution contexts should work on Apple now
1508 * Port to Windows 64 bits
1509 - Sysv contexts now have an implementation for this arch
1510 - GRAS communication features now support this arch
1511 * Drop support for borland compiler on windows
1512 - this code was not maintained, and we kinda depend on gcc nowadays
1513 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1514 * Fix portability issue on unstable ubuntu: linker became picky on
1517 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1519 SimGrid (3.6.1) stable; urgency=low
1521 The "Oops, we broke Macs too" release
1524 * Fixed contexts detection so that raw ones are used when possible
1525 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1526 leads to a strange error, with user code segfaulting sometimes when
1527 the generated code is not perfectly aligned (which is not
1528 controllable from the user side, depends on the amount of code)
1531 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1533 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1535 SimGrid-java (3.6) unstable; urgency=low
1538 * Split of every thing from simgrid v3.5 into a separate package.
1540 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1542 SimGrid (3.6) stable; urgency=medium
1544 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1547 * Bindings now constitute their own package, separated from the main one.
1548 Rationale: reduce our maintenance nightmare by reducing the module coupling
1549 They will soon be released on their own on gforge.
1551 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1552 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1554 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1555 * It's not quite deprecated for now because we have no replacement,
1556 but it may soon become the case.
1559 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1560 * Fortran: New user-level cache variable to store the rank of the running
1561 process. This improves performance by an order of magnitude.
1562 * C: New coccinelle script to automatically locate and modify global and
1563 local static variables.
1564 * Improved SMPI network model with a sender-side gap to account for multiple
1568 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1569 only say if a communication is finished, no matter whether it succeeded or
1570 failed. You can call MSG_comm_get_status() to know the status of a finished
1572 * New function MSG_task_dsend() to send a task and detach it. When a
1573 communication is detached, you are never notified of its success or failure
1574 and the memory is released automatically once it is finished. This function
1575 is useful when you don't care about the end nor the success of a
1577 * Change the prototypes of action replay. Sorry for inconvenience,
1578 but this is really more efficient this way (and to adapt your code,
1579 you just have to fix the initialization, that shouldn't be too long)
1580 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1581 ever managed to do anything useful with it.
1582 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1583 takes a dynar of communications. It returns immediately and gives the
1584 index of a finished communication (if any).
1585 * New example: a basic implementation of the Chord P2P algorithm.
1588 * New model for multi-core CPUs. You can now use the core attribute to
1589 precise the number of cores of a host. This is a basic model. Every
1590 process running on the host receives at most the power provided in
1591 the DTD (throughput<=power). Total throughput of process cannot exceed
1593 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1594 router linked by an up-link and a down-link (possibly asymmetrical).
1595 This kind of pattern allows to easily build last-mile model style platforms.
1596 Aggregating such patterns in a rule-based AS is thus the technique of
1597 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1598 * New model for Vivaldi routing. We transformed the Vivaldi network model
1599 into a Vivaldi routing model (based on the rule-based model). This allows to
1600 combine Vivaldi based latencies with last-mile platforms.
1603 * Added a check for NaN of IEEE754 infinite in the double entries of
1605 * Introduce a new context factory "raw", highly inspirated from the
1606 ucontext factory, but using manually crafted functions in assembly to
1607 do the work in an efficient manner.
1608 * Allow to change the used context factory at run time, not only at
1609 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1610 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1611 contexts at runtime (only with raw contexts or ucontexts).
1612 * Completely rewrote this module to allow parallel execution of user
1613 processes. Use --cfg=contexts/nthreads:N to execute user processes
1614 with N parallel threads (the default is 1, meaning no parallelism).
1615 * Allow to decide dynamically between sequential and parallel modes.
1616 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1617 processes in parallel only when their number is greater than or equal to P
1619 * Added a check for NaN of IEEE754 infinite in the double entries of
1623 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1624 won't output the state of processes when interrupted with Ctrl-C
1625 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1626 NULL-terminated array. This may solve backward compatibility issues
1627 due to the change to return type of SD_simulate. See also:
1628 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1629 * Add new macros with variable number of arguments.
1630 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1631 - in xbt/asserts.h: xbt_assert
1632 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1633 - in xbt/ex.h: THROWF and RETHROWF.
1634 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1636 * Change xbt_die() to accept a format string with arguments, just like printf.
1637 * New data structure: xbt_lib_t, like a dict but more general and with better
1641 * New configuration options
1642 Options triva/categorized and triva/uncategorized can be used to generate
1643 graph configuration files for Triva visualization tool.
1644 * Configuration option tracing/platform is renamed to tracing/categorized
1645 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1646 * New TRACE_user_link_variable interface:
1647 User provides the name of the link and the tracing variable to attach to it
1648 * the declaration of tracing categories must be done after the environment
1650 * simpler tracing interface, just one way to declare categories
1651 TRACE_category or TRACE_category_with_color, it is up to you
1652 * links in the trace file are again identified by their names
1653 * trace contains the full platform hierarchy exactly as declared using the ASes
1654 * Options tracing/msg/[task|process]:1 groups the process by hosts
1655 for both cases, tasks and processes must have names that are unique during the simulation
1656 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1657 * The experimental option tracing/msg/volume is deprecated
1658 its functionality may be reincorporated if needed
1660 The tracing generates a trace file with unordered timestamped events,
1661 because of the way the core simulator (surf) works. A script available
1662 at the tools directory (fix-paje-trace.sh) can be used to put the events
1663 in order. We have changed the tracing so it can generate ordered timestamped
1664 events in the final trace, but depending on the simulator (and how much time
1665 is simulated) that can lead to a huge memory utilization. It is deactivated
1666 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1668 Build Infrastructure
1669 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1670 - We are trying hard to keep the API stable, but it may happen that
1671 some things change (we're a research project after all, not a
1672 nuclear plant operating system). If such things should happen, you
1673 could rely on that macro to adapt.
1674 - current value: 30600 for 3.06.00, aka 3.6
1675 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1676 a number that can be compared with SIMGRID_VERSION.
1677 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1678 assertions and verbose/debug logging events are disabled at compile time.
1680 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1682 SimGrid (3.5) stable; urgency=medium
1685 * New feature to any SimGrid-based simulator: Model-Checking
1686 Check SIN#1 for more details.
1689 * New Model SMPI (three-interval linear regression for correction factors)
1690 See RR-7426, available at http://hal.inria.fr/inria-00527150
1691 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1692 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1693 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1694 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1695 * See also src/smpi/README
1700 - Tracing API changes: TRACE_start and TRACE_end should not be called
1701 by user-code. They are automatically called by simulators created
1702 with SimDAG, MSG and SMPI if the toolkit is compiled with
1703 tracing_enabled=ON. Categories declaration and utilization remain the
1704 same for MSG and SimDag.
1705 - A function was added to the tracing API to declare categories with
1707 - TRACE_category_with_color (char *category, char *color)
1708 where color must be in the following format
1709 "%f %f %f", red, green, blue
1710 and red, green, blue are float values in the interval [0, 1]
1711 - User can specify NULL as color parameter, or continue calling
1712 TRACE_category (cat)
1713 On that case, the tracing system will define random colors
1714 - The following command-line options are supported:
1715 --cfg=tracing/filename:msg.trace
1716 --cfg=tracing:1 (activate tracing, needed to use others)
1717 --cfg=tracing/platform:1 (categorized resource use)
1718 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1719 --cfg=tracing/msg/task:1 (task creation)
1720 --cfg=tracing/msg/process:1 (process creation, migration)
1721 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1722 --cfg=tracing/smpi:1 (SMPI interface tracing)
1723 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1724 - examples of examples/msg/tracing updated
1726 - DAXLoader and DOTLoader functions can generate tasks with categories
1727 - A new function to attribute a category to SD tasks:
1728 TRACE_sd_set_task_category (SD_task_t task, char *category)
1729 * Tracing the MPI interface implemented by SMPI
1730 - Collective operations are traced with states
1731 - Point-to-Point operations are traced with states/links
1732 - Tracing activated by a parameter "-trace filename" passed
1733 to smpirun during execution (considering that simgrid
1734 is compiled with tracing enabled)
1735 - To run the simulation with gdb, the simulator
1736 accepts --cfg=tracing/smpi:1 to trace SMPI
1737 - tesh files to check if smpi tracing is ok
1738 - See examples/smpi/NAS/DT-trace
1739 * GTNetS tracing re-worked
1740 - adaptation to the tracing system of GTNets to cope
1741 with modifications regarding the fullduplex mode
1742 - new tesh files to check if gtnets tracing is ok
1745 * Asynchronous communications through the functions:
1746 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1747 * New function: MSG_load_platform_script()
1748 to make possible using a lua script instead of XML files to set up platforms
1749 * New function: MSG_set_function
1750 to associate functions to processes, used when bypassing the parser
1751 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1753 Platforms: Add some more examples in examples/platforms
1754 * Grid'5000: see www.grid5000.fr
1755 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1758 * Change the XML format. This is a very important modification. SimGrid
1759 3.5 introduces a new hierarchical format based on the notion of
1760 Autonomous Systems. Compatibility with old format is ensured
1761 through the perl script provided in the install bin directory
1762 bin/simgrid_update_xml.
1763 It is now possible to build platforms with specific routing
1764 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1765 connect several platforms together. We will try to provide soon
1766 set of realistic platforms exploiting these properties (have a
1767 look at examples/platforms/ for the moment).
1768 * Take the opportunity of the XML format change to be a good XML citizen:
1769 rename link:ctn to link_ctn and similar changes (also dealt with by
1771 * Add a new routing scheme (rule-based) using regular expressions. It
1772 enables to have an extremely low memory footprint when the
1773 underlying routing is simple and can be compactly described. You
1774 need to have libpcre4-dev (perl regular expressions) installed if
1775 you want to use this routing scheme.
1776 * Revive the cluster TAG and allow to easily and efficiently (both in
1777 term of memory and speed) connect clusters together. Have a look
1778 at teshsuite/simdag/platforms/ to see how this can be done. With
1779 this tag, you can create clusters with thousands of tasks at no
1780 cost (have a look at examples/platforms/).
1781 Note: clusters are implemented as ASes, so there is no need for an
1782 enclosing AS tag if you have only one cluster in your platform.
1783 * Add new generic functions in the public interface that allows the user
1784 to call SURF 'create_resource' methods from your code (same
1785 functionality as the XML bypass mechanism but with a much lighter
1787 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1788 that uses a piecewise linear approximation to produce better
1789 results when exchanging small messages.
1790 * Add a new parameter to handle correctly full duplex link and account
1791 for interferences between uplink and downlink communications
1792 (activate with --cfg=fullduplex:1).
1795 * Rename the SD_READY (all dependencies are satisfied and task is
1796 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1797 dependencies are satisfied) state.
1798 This prevents a confusion between the notion of "ready to schedule"
1799 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1800 simulated" (SD_RUNNABLE) used by the simulation kernel.
1801 * Change the way a task is considered as ready. Instead of removing
1802 dependencies when a task is done, a counter is decreased. This way,
1803 it is always possible to reach ancestors thanks to the
1804 SD_taks_get_parents function (even after the end of the simulation.)
1805 * Change the return type of SD_Simulate from (SD_task_t*) into
1806 xbt_dynar_t. This function was in handling a dynar internally and
1807 converted it into a NULL terminated array for historical reasons.
1808 * New function SD_dotload(char*) to load a DAG described in dot
1809 format. This loader and the corresponding examples require the
1810 installation of the graphviz library.
1811 * Fix a bug in the management of tasks of size 0 in the surf network
1812 models. This problem was only visible with SIMDAG and you should
1813 thus disregard results produced with earlier versions if you
1814 relied on this feature (some tasks were blocked because of this).
1815 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1816 even though your applications were doing only point-to-point
1817 communications and sequential computations. Now you can really use any
1818 model you want (of course, if you create real parallel tasks, which are
1819 not implemented in most models beside ptaskL07, this will abort).
1820 * Add an example that schedules a DAX on an heterogeneous platform
1821 using a Min-Min strategy.
1822 * New function SD_workstation_get_current_task() that returns the kind
1823 of task currently running on a workstation in the sequential access
1825 * Raise some warnings when unexecuted tasks remains at the end of the
1826 simulation. This is usually caused by cycles in the DAG.
1829 * New function: SIMIX_process_set_function() called by MSG_set_function
1830 * Change the underlying waiting queue in semaphores so that a process
1831 can wait on several of them simultaneously (as in waitany).
1832 * Fix the way to handle tokens in semaphores so that all access patterns
1833 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1834 * kill the dirty pimple SIMIX_message_sizes_output()
1835 Please use (proper) visualization instead
1838 * New data container: setset (set of sets of elements)
1839 * New module: mmalloc (mapped malloc, allowing to have several
1840 independent segments of malloc)
1841 * New function: xbt_dict_cursor_set_data()
1842 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1843 * New function: xbt_dynar_is_empty()
1844 * New function: xbt_fifo_get_last_item()
1845 * Fix xbt_dynar_shrink(): use the right element size.
1846 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1847 uninitialized areas during expand.
1848 * Fix semaphores: previous implementation was severely broken.
1849 * Use library init/fini functions for our initialization.
1850 - you can use logs and other feature as soon as you want in your
1851 code (even before the xbt_init / MSG_init)
1852 - xbt_exit is now a no-op and produce a warning when used.
1855 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1856 benefit from the latest improvement to the simulation kernel.
1857 * Kill measurement sockets for now. If you rely on them, sorry. This
1858 release is not for you. This feature will be reintroduced in the
1859 future, but we cannot delay the release any further.
1860 * New function: gras_msgtype_get_name().
1861 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1862 * Fix (at last) the pmm example: it should not randomly fail anymore.
1864 Build chain: bug fixes and overall polishing
1865 * Cmake is now stable enough. Hence, we killed the autotools.
1866 * Port to windows ( TM :)
1867 * Fix the 'make install' target.
1868 No need to use 'make install-simgrid' anymore
1869 * Introduce a 'make dist' target compiling a *source* archive
1870 'make package' compiles a binary archive
1871 * Compile java files only on need
1872 * Add --cd and --setenv command line options to tesh
1873 * Out of source builds are not fully supported yet, but we are close
1874 * Enable supernovae and optimization flags by default for our users
1877 * Add layer to set up environment directly from lua, without XML.
1878 * The effect of gras_stub_generator can be achieved through
1879 lua too (check examples/gras/console/ping_generator.lua)
1881 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1883 SimGrid (3.4.1) stable; urgency=low
1885 The "Polishing easter eggs is probably a good idea" release.
1886 This is a bug fixes release only.
1890 * Fix a bug preventing the tasks from begin garbage collected.
1893 * Fix a bug occuring when a host involved in a communication fails.
1894 This was not detected properly by the other peer involved in the
1895 communication. Now, it's reported as a network error.
1898 * Warn the user about loop dependencies in data flow of DAX files
1899 * Obey the control-flow dependencies of DAX files
1902 * Add option "enable_smpi" allowing to not compile SMPI.
1903 Probably useful for the (Mac) users experiencing a build error here
1904 * Improve the detection of lua5.1 and ruby1.8
1906 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1908 SimGrid (3.4) stable; urgency=low
1910 The "Easter in Cargese" release. Also known as (major changes):
1912 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1913 ~> bindings were greatly improved
1914 ~> new tracing infrastructure for better visualization introduced
1916 * the "Welcome to configury modernity" release.
1917 ~> we switched from autotools to cmake, and improved our cdash
1920 A more detailled list of changes follow (full detail in svn log).
1922 Java Bindings: Various Cleanups
1923 * (install java-gcj-compat-dev on debian-like to use them)
1924 * Remove put/get: no need to export deprecated interface in Java
1925 Use send/receive instead.
1926 * Cleanup the examples and add a README per directory
1927 * Remove example autoDestination (that's the only way to go now)
1928 * Remove example explicitDestination (was a plain copy of basic)
1929 * Make JniException a runtime exception, so that there is no need to
1930 declare the fact that you may encounter such a beast. I guess that
1931 nobody will ever want to survive such error.
1932 * Create specific errors for each MSG case of failure:
1933 host failure, transfer failure, timeout, task cancelled
1934 * Cleanup the exceptions that may get thrown by each function
1935 * Other internal cleanups in Java bindings. Performance still bad :/
1936 Ruby and Lua Bindings: create them
1937 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1938 * That's new and great, you should try them out.
1939 Same functionalities than Java bindings, only even less polished
1941 * Kill the useless "rate" argument of SD_task_get_execution_time()
1942 Everyone used to provide -1 as a value, it was not used, and the
1943 semantic of a possible use wasn't even clear.
1944 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1945 either as comm costs or compute costs to mean that there is no
1946 such thing for that specific task.
1947 * Add a SD_task_set_name() function
1948 * Fix SD_task_unschedule() on typed tasks
1949 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1950 * In DAX loader, accept useless 'level' attributes to <job> since
1951 LIGO DAGs have them (seem to be to ease graphical representation).
1953 * Add an example masterslave_mailbox.c using send/receive and not
1954 the deprecated put/get interface.
1955 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1956 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1958 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1959 (declare MSG_USE_DEPRECATED to still have the old name)
1960 * Add a MSG_task_set_data() function
1961 * About trace replay (see examples/msg/actions):
1963 - Allow to work with splitted trace files for each process
1964 Give the specific trace file as argument of each process,
1965 and call MSG_action_trace_run(NULL)
1966 You can still have one merged file for all processes.
1967 - Fix implementation of collective operations
1968 * Allow task_execute() on 0-sized tasks (closes #10063)
1970 * This is the first release of SimGrid where SMPI is not considered
1971 beta anymore (even if some corners should still be improved)
1972 * Port over the new SIMIX_network submodule (internal refactoring)
1973 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1974 * Implement more missing elements of the standard:
1976 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1978 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1979 MPI_Scan() MPI_Get_processor_name()
1980 - Added implementation of missing case for Alltoall (warning: it's
1981 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1982 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1983 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1984 SMPI_MPI_Allgatherv()
1985 * Bug fixes include:
1986 - MPI_Waitsome() was broken
1987 - Allow relative includes in smpicc
1988 - Command line cfg argument 'reference_speed' was ignored...
1989 - Some functions did not properly lead to auto-benching of user code
1990 - smpicc passes -O2 by default (just like openmpi one)
1992 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1993 * Bug fixes about timeouts during communications
1994 * add SIMIX_message_sizes_output() as a pimple to write to file the
1995 amount of messages per size. Use gnuplot to get histogram.
1996 Pimple because that's the only user-visible function of simix,
1997 defined directly in xbt.h (irk, sorry)
1999 - Add a SIMIX_sem_get_capacity() function
2000 - Fix interactions with processe resume/suspende
2001 - release_forever() was stupidly broken
2002 - Fix SIMIX_display_process_status() for processes in a semaphore
2003 - Make SIMIX_sem_block_onto() user-visible
2004 * Refactoring context stuff:
2005 - Use pseudo-OOP for better modularity
2006 - reimplement SIMIX_process_kill() without process_schedule() so
2007 that the latter can take as invariant that it is called from
2009 - Merge context_start into context_new for sake of simplicity
2011 * Add a Vivaldi network model, coded live during SUD'10 ;)
2012 * Rename configuration variables to start a hierarchy:
2013 o cpu_model -> cpu/model
2014 o network_model -> network/model
2015 o workstation_model -> workstation/model
2016 * New configuration variables:
2017 o network/bandwidth_factor: correction to bandwith
2018 o network/latency_factor: correction to latency
2019 o netwotk/weight_S: correction to the weight of competing streams
2020 * Add a long description to the models, that users can see with such
2021 argument on the command line: --cfg=cpu/model:help
2022 * --help-models display the long description of all known models
2024 * config: add the ability to set a default value after registration
2025 Does not override any previously set value (e.g. from cmd line)
2026 * dict: allow to have integer key and data.
2027 When so, you need to use the following functions
2028 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2029 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2030 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2031 In contrary to regular dicts, the key is not malloced before copy.
2032 Mixing scalar and regular elements in the same dict is not tested
2034 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2035 Tracing for Visualization:
2036 * SimGrid is now instrumented in order to generate a trace file for
2037 visualization analysis: to use it, need to compile SimGrid with the
2038 "tracing" option enabled, and instrument the program using SimGrid with
2039 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2040 (among other functions).
2041 * The instrumentation only traces the platform utilization for now
2042 * Documentation to use the tracing functions and how to analyze the
2043 traces with the Triva tool is written.
2044 * More information about: SimGrid FAQ (in the section Tracing Simulations
2047 * We moved to cmake as default build system. Autotools support will
2048 be dropped soon. Check the FAQ for more info about how to use it.
2049 * Greatly improved our cdash/ctest interactions
2050 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2051 * Added memory checking tests with valgrind; lot of memleak fixing.
2052 This may be the first release of simgrid with so few memory issues
2053 * Added code coverage tests.
2054 Our coverage is still improvable, but at least we see it on cdash.
2056 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2058 SimGrid (3.3.4) stable; urgency=low
2060 The "Desktop Grid needs love too" release (also called Xmas release).
2062 Models improvements:
2063 * Major speedup in the maxmin system solving by using lazy evaluation
2064 Instead of solving completely the maxmin system at each iteration,
2065 only invalidate (and recompute) the modified parts.
2066 This new feature is enabled in default models but you can try to
2067 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2068 * Cas01 IMproved as default CPU model
2069 This CPU model is the same Cas01 model, but it uses the
2070 maxmin-selective-update flag and a heap structure to manage
2071 actions on SURF kernel.
2072 It reduces the complexity to find the next action to finish and,
2073 consequently, it's faster than the old Cas01.
2074 This is the new default CPU model (Cas01).
2075 * Rename the old Cas01 model to Cas01_fullupdate
2076 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2077 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2078 * CpuTI (CPU Trace Integration)
2079 A new CPU model whose objective is simulate faster when using
2080 availability trace files.
2081 Instead of using a full featured, over engineered maxmin system for
2082 CPU modeling, this model does the pre-integration of traces files
2083 to calculate the amount of CPU power available, and so, executes
2084 faster than the old CPU models.
2085 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2086 * Use LV08 as default network model since it gives better accuracy
2087 for small messages and shouldn't change things for big ones.
2088 Use --cfg=network_model:CM02 to get the previous behavior.
2091 ******************************************
2092 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2093 ******************************************
2094 * The new CPU model may changes simulations!
2095 The point is that events occurring at the exact same timestamp
2096 are not scheduled in the same order with the old and new
2097 version. This may be enough to completely change the execution
2098 of simulations in some cases.
2099 * The new network model will change simulations!
2100 This new model is more realistic than the previous one, so you
2101 should consider redoing your old experiments with this model.
2102 Sorry for the inconvenience.
2105 * Introduce the supernovae compilation mode
2106 When compiled that way, the whole SimGrid (or almost) is put in a
2107 single compilation unit and compiled in one shoot.
2108 This is to help gcc which has difficulties to inline stuff from one
2110 The speedup seem to be above 15%, althrough more tests are needed on
2111 amd64 to confirm that gain.
2114 * Port of MSG's mailbox on top of SIMIX network
2115 The put/get mechanism was greatly simplified on the way.
2118 * New SIMIX network module. Provides:
2119 - Mailbox: rendez-vous mechanism to find with who you want to speak
2120 - Synchronous send/recv: easier and hopefully faster since the
2121 logic is handled in the maestro process directly now
2122 - Asynchronous send/recv: you dreamt of it? It's here now
2123 Too bad that nobody cared enough to propagate the change to MSG.
2124 * Add semaphores as SIMIX synchronization mechanism.
2127 * new function SD_daxload(char*) to load a DAX file
2128 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2129 * Introduce typed tasks. Specify its kind and cost at creation.
2130 At scheduling, just give where it should be placed, and the cost
2131 for each involved resource is automatically computed.
2132 Existing constructors so far (more to come of course):
2133 - SD_task_create_comm_e2e() for end-to-end communication
2134 - SD_task_create_comp_seq() for sequential computation
2135 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2136 * new function SD_task_dump() for debuging display
2137 * new function SD_task_dotty(task,FILE*) writing to file the info
2138 about the task in dotty format
2139 * SD_task_dependency_exists() can now cope with having one of its
2140 arguments NULL. If so, it tests whether the other argument has any
2142 * Add getters on list of preceding/following tasks:
2143 SD_task_get_parents(task) and SD_task_get_children(task)
2144 * Add getters on amount of workstations and list:
2145 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2146 * Add getter on task kind: SD_task_get_kind(task)
2147 * Update the start_time and finish_time of tasks on completion/failure
2148 * Bugfix: Remove task from state swags when destroyed
2151 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2154 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2155 dynar with one separate thread per value of the dynar.
2156 * Change the prototype of xbt_thread_create(), sorry.
2157 Added a boolean parameter indicating whether we want to join this
2158 thread (used in SG only for now)
2159 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2162 * GTNetS wrappers should now be usable again (and betterly tested too)
2163 * Fix a major regression from 3.2 where the timeout provided to
2164 MSG_task_put_with_timeout() was used as absolute time before which
2165 the comm should be done.
2166 * Start to fix the <cluster> tag.
2167 - Internal links should be good now (beside of the loopback, which
2168 use the private link instead)
2169 - paths to the external world is still rather broken
2170 - the <route:multi> tag is just broken. Actually that's brain-dead.
2171 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2172 to make it less stupid
2173 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2174 * Fix a source-level compatibility glitch from 3.2: after defining
2175 MSG_USE_DEPRECATED, you can use the old name
2176 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2177 * Allow to compile from the SVN with automake 1.11
2178 * Fix some problems when using the "start_time" tag in deployment XMLs.
2179 * Fix #8569: XBT/synchro.h has redundant declarations
2180 * Fix #8563: MSG return values and exceptions
2181 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2182 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2183 Thanks to Robson Peixoto.
2184 * Fix a vicious bug in dictionaries inducing that some elements were
2185 not freed on xbt_dict_free()
2187 Portability report of this version:
2188 * Main portability targets:
2189 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2190 on (amd64/i386/ia64)
2191 - mac leopard on i386
2192 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2193 but nothing critical.
2194 * Other platforms: windows, AIX and others were not tested for this release
2196 Timing report of this version:
2197 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2198 scenario parameters). From 8h to a few seconds in desktop grid settings.
2199 * Supernovae brings about 25% speedup on i386.
2201 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2203 SimGrid (3.3.3) stable; urgency=low
2205 The "Need for Speed" release.
2207 The timings done to validate the 3.3.2 were faulty.
2208 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2210 The problem was a conversion from a manually handled vector to
2211 xbt_dynar_t on the critical path.
2212 xbt_dynar_foreach calls functions, inducing stack management crap.
2214 We inlined these functions and xbt_dynar_foreach is now breath taking.
2215 We also inlined xbt_swag_belong on the way.
2217 Here are some approximate speedup measurements (on master/slaves
2218 simulations lasting between 10s and 20s each):
2219 3.3.1 -> 3.3.2: about same performance
2220 3.3.2 -> 3.3.3: 40% speedup
2221 3.3.1 -> 3.3.3: 40% speedup
2222 3.3.1 with inline patch -> 3.3.3: 30% speedup
2224 Our reading is that the refactoring which occurred in 3.3.2 made us
2225 suffer much more from the xbt_dynar_foreach low performance, but
2226 once we solved this, this refactoring proved to be very performance
2227 effective. From the 40% speedup, somehow, 10% are due to the
2228 inlining and 30% to the refactoring.
2230 That's a pitty that gcc cannot inline functions placed in other files
2231 alone. We have to choose between:
2232 - break the encapsulation (by putting private data structures and
2233 accessors in headers files to help gcc)
2234 - live with low performance
2235 - switch to a decent compiler such as icc (not quite possible).
2237 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2239 SimGrid (3.3.2) stable; urgency=low
2241 The "Simplicity does not preceed complexity, but follows it" release.
2243 The main contributors of this release were (lexical order):
2244 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2247 * Extract the routing logic into its own object.
2248 (was dupplicated in network.c and workstation_LV07.c;
2249 Allows to implement other ways of storing that info)
2250 => kill now useless network_card concept
2251 - Use dynar to represent routes (instead of void** + int*)
2252 - kill link_set (use surf_network_model->resource_set instead)
2253 - Add a command-line option to choose the routing schema to use
2254 - Add three new models:
2255 * Floyd (shortest path computed at initialization)
2256 * Dijikstra (shortest path recomputed all the time)
2257 * Cached Dijikstra (shortest path computed on need)
2258 All these models where contributed by Silas De Munck, and are
2259 described in his ICCS09 paper.
2261 * Simplify model declaration
2262 (less redirections, less function to write when defining a model)
2263 - Factorize stuff between models:
2266 surf_model_resource_set(model)
2267 surf_model_resource_by_name(model, name)
2268 - Unify the types of models in s_surf_model_t (using an union)
2269 - Embeed fields of common_public directly into s_surf_model_t
2270 - Rename model methods:
2271 action_free ~> action_unref
2272 action_change_state ~> action_state_set
2273 action_get_state ~> action_state_get
2274 - Change model methods into functions :
2275 (model)->common_public->action_use ~> surf_action_ref
2277 * Implement a generic resource; use it as ancestor to specific ones
2278 (allows to kill duplicated code in models)
2279 Drawback: timer command don't need no name nor properties;
2280 workstation_CLM03 don't need no properties
2281 (but I guess we can live with those few bytes wasted)
2283 * Improve the action object model
2284 - implement a constructor avoiding dupplicated code about field
2285 initialization in generic_action part.
2287 * Kill the SDP model: it has an external dependency, is deprecated
2288 in flavor of modern lmm models, and didn't compile since a while
2291 * Relocation of the context module from XBT to SIMIX.
2292 (the context were decoupled from the simix processes, duplicating a lot of code)
2293 => a lot of code was factorized
2294 - less overhead is introduced during scheduling
2295 - simpler API for the context factory
2296 - the logic for process creation,destruction and manipulation was simplified
2297 * Simplification of the s_smx_process_t data structure.
2298 => accesing the simix level data associated to a process is faster now,
2299 and the code is a lot more readable.
2302 * Implement some more MPI primitives:
2303 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2304 -implementation: Bcast: flat or 2-ary tree (default),
2305 Barrier: 4-ary tree,
2307 Allreduce: Reduce then Bcast
2308 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2309 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2310 Alltoallv: flat tree, like ompi
2312 * Add support for optimized collectives (Bcast is now binomial by default)
2313 * Port smpirun and smpicc to OS X
2316 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2317 Shout out if you used it.
2320 * Display the list of still queued messages in SG mode when existing
2324 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2325 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2326 * Add xbt_str_from_file(FILE*)
2327 * Add xbt_dict_get_key achieving a linear reverse search
2328 * Remove the context module
2330 Portability report of this version:
2331 * Main portability targets:
2332 - Linux(debian)/x86/context
2333 - Linux(debian)/x86/pthreads
2334 - Linux(debian)/amd64/context
2335 - Linux(debian)/amd64/pthreads
2336 On these, we still have the eratic breakages of gras/pmm and
2337 amok/saturate_sg reported in previous version. We still think
2338 that the tests are the cause of the fault, not the tested code.
2340 - Mac OSX Leopard/x86/context
2341 Still false negative in tesh autotesting.
2342 Smpi still fails, but this time because readlink does not accept -f
2343 Everything seems to work properly beside of that.
2346 - AIX version 5.3 (only tested contexts this time)
2347 Smpi still fails there because mktemp is not installed.
2348 Everything seems to work properly beside of that.
2350 I managed to compile it for the first time, but several breakages.
2351 Won't delay the release for this exotic platform.
2353 * Windows: it's still lagging behind. If you want to help, please
2356 Timing report of this version:
2357 This version seem to be more than 5% faster than 3.3.1 (on linux
2358 64bits with contextes). The gain is less than expected, we are
2359 investigating this for next release.
2361 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2363 SimGrid (3.3.1) stable; urgency=low
2366 * Implement a --cfg-help to show existing configuration variables
2367 * Build chain do not require doxygen in maintainer mode
2370 * fix a bug on struct sizeof computation, which prevented the
2371 exchange of arrays of structs in some conditions
2372 - added a regression test about this in datadesc_usage
2373 * Allow the exchange of 0-long dynamic vectors.
2374 - for that, use -1 as indicator of dynamic size instead of 0
2375 - This implied to change any size from unsigned long to long,
2376 reducing a bit communication abilities, but I guess that with
2377 64bits being quite common, this is more than enough.
2378 - This also induce a protocol change, thus bumping network protocol
2379 version from 0 to 1 (if we have external users, we have to get
2380 clean on that point too ;)
2381 - added two regression tests about this in datadesc_usage
2382 * Be more verbose when propagating local exceptions
2383 This helps debugging.
2384 * Display the status of simulated processes when receiving SIGINT in
2388 * Allow to control the simulation from a trace file.
2389 New functions MSG_action_register() and MSG_action_trace_run()
2390 The first one allows to associate a function execution to each
2391 kind of action while the second one parses a trace file and
2392 triggers the corresponding actions within the system.
2393 For now, only a toy example is provided in examples/msg/actions
2394 * Add an exemple of process migration in examples/msg/migration
2395 * Fix a bug in task exchange which broke MSG_task_get_sender()
2396 Add a teshsuite regression test for that.
2397 [Bug: if MSG_task_get_sender() is called after sender exit,
2399 * Fix a bug which prevented suspend/resume to work properly
2400 * Display the status of simulated processes when receiving SIGINT
2401 This fixes a regression of v3.3. due to the introduction of SIMIX
2402 * Bug fixing in failure management:
2403 - trace could not start by a failure at time 0
2404 - failure during communications were not working
2407 * Add SIMIX_process_set_name() to change the name of the current
2408 process in the log messages.
2409 * Store smx_hosts in a dict since we only retrieve them by name
2410 * Move the configuration infrastructure to surf
2413 * Move the configuration infrastructure to surf
2416 * Massive internal cleanups:
2417 - Store internal structures on processes instead of hosts (allows
2418 to have more than one process per host, in addition of being more
2420 - Cleanup the initialization/finalization process
2421 - Kill a whole bunch of unneeded synchronization:
2422 processes run in exclusive manner within the simulator
2423 - Move queues from global tables to process data fields
2425 - now accept -platform and -hostfile arguments
2426 - Pass the right rank value to processes according to the hostfile
2427 * Compile the examples by default, and use them as regression tests
2428 * Implement MPI_Wtime()
2429 * Change the reference speed to a command line option
2432 * TCP_gamma can now be specified as command line option using
2433 --cfg=TCP_gamma:10000000.0
2434 * Change the --surf-path cmd line option into --cfg=path:
2437 * Also include strbuff from xbt.h public header
2438 * xbt_ex_display(): do not free the exception after displaying
2439 This allows to do more with the given exception afterward.
2440 Users should call xbt_ex_free() themselves.
2444 Portability report of this version:
2445 * Main portability targets:
2446 - Linux(debian)/x86/context
2447 - Linux(debian)/x86/pthreads
2448 - Linux(debian)/amd64/context
2449 - Linux(debian)/amd64/pthreads
2450 These targets fail about 1/10 of times on gras/pmm, but we believe
2451 that this is because of the test, not because of simgrid.
2452 amok/saturate_sg fails even more rarely, and the test may not be
2455 - Mac OSX Leopard/x86/context
2456 The test suite still spits tons of errors because some obscure
2457 force prevents us from removing the temporary directories
2458 arguing that they still contain some metadata I've never heard of.
2459 Smpi fails because seq is not installed.
2460 Everything seems to work properly beside of that.
2463 - AIX version 5.3 (both contexts and pthread)
2464 Smpi still fails there because mktemp is not installed.
2465 XML inclusions seems rosty on AIX.
2467 * Windows: it's still lagging behind. If you want to help, please
2470 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2472 SimGrid (3.3) stable; urgency=high
2476 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2477 [Malek Cherier & Mt]
2479 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2482 Until now, GRAS were using MSG as an interface to SURF. It was
2483 quite difficult because both interface have several differences
2484 (MSG channels vs GRAS sockets were the most notable point).
2486 This also opens the gate to SMPI (which should occur soon) and speed
2487 up simulations by to 40% (even if it were not the main goal).
2489 **************************************
2490 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2491 **************************************
2492 The point is that events occuring at the exact same timestamp are
2493 not scheduled in the same order with the old and new version. This
2494 may be enough to completely change the execution of simulations in
2495 some cases. Sorry for the inconvenience.
2497 * Cleanup and upgrade the XML format to push further scalability
2498 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2500 * Improve the testing infrastructure with tesh. Now a very large part of
2501 the code is tested not only by being run but also by checking that the
2502 output match an expected output [Mt].
2504 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2505 is really less memory-demanding, which should allow you to use
2506 larger files in SimGrid [AL].
2508 * Inform valgrind about our contextes, so that it becomes usable
2509 with the default (and more effecient) version of SimGrid
2510 [contributed by Sékou Diakite, many thanks]
2513 * Introduce a listener thread in charge of receiving incoming
2514 messages from the network. It allows to overlap communication and
2515 computation but most notably, it removes some stupid deadlocks due
2516 to the fact that so far, a process could not send and receive at
2517 the same time. This made most non trivial communication schema
2519 * Convert the PIDs from long int to int to match the MSG ones (and
2520 linux ones too) [Mt]
2521 * New function: gras_agent_spawn() to launch a new process on
2522 current host. Only working in simulation for now. [Mt]
2523 * New function: gras_os_hostport() returning a constant form (ie,
2524 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2527 * Make the backtrace of exceptions more human readable [Mt]
2528 * New module: xbt/str [Mt]
2529 a ton of string utility functions (split, join, printf to a newly
2530 allocated buffer, trim, etc)
2531 * New module: xbt/hash [Mt]
2532 SHA1 hashing algorithm (more to come if needed)
2533 * New module: xbt/synchro [Mt]
2534 synchronization tools (mutex and conditions) working the same way
2535 in simulation and in real life (mainly useful for GRAS, but not
2537 * New module: xbt/queue [Mt]
2538 classical producer/consumer synchronization scheme
2539 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2540 (using the classical functions will get serialized) [Mt]
2541 * Make dictionary internal table dynamic. No need to specify its size
2542 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2544 * Make sure the log channels are organized as a tree under windows
2545 (because of ANSI C compatibility issue, any channel were child of
2546 root directly) [Mt].
2549 * Cleaned many thing in surf and fixed a few bugs [AL].
2550 * Add a nice command line configuration mechanism to compose models [AL].
2551 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2552 the previous one (KCCFLN05). It relies on something that looks like
2553 a max-min sharing mechanism but cannot be written as such. A new solver
2554 was thus designed [AL].
2555 * Add a new solver to lmm. Based on Lagrange optimization and
2556 gradient-based descent, it enables to efficiently maximise systems s.a
2558 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2560 This solver enables to propose two new network models for TCP Reno and
2561 TCP Vegas based on Low's work. These models still need to be fully
2562 tested though [Pedro Velho].
2565 * Bug fix in SD_simulate. Now the time bound given as argument is
2567 * Use the new parallel task model (ptask_L07) as default.
2568 * Use the SURF command line configuration mechanism.
2569 * 0-size tasks (for synchronization) should now work.
2571 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2573 SimGrid (3.2) stable; urgency=high
2577 We still experience issues on this platform, but we believe that at
2578 least MSG is usable.
2580 GRAS API BREAKAGE (for simplification purpose, sorry):
2581 * the gras_msgtype_by_name is not used anymore. Instead of
2582 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2583 you can write (and must)
2584 gras_msg_send(toserver, "request", &request);
2585 - If you still want to pass a gras_msgtype_t to the function (to cache
2586 the type and avoid the lookup time), use the gras_msg_send_() variant.
2587 - Impacted functions:
2588 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2589 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2590 * The callbacks are now expected to return 0 when everything went well
2591 (just like the main() function)
2593 GRAS new features and improvements:
2594 * New module mechanism where user code can use per process globals [Mt]
2595 This is similar to gras_userdata_*() functions, but for libraries. It
2596 factorize some code developped over and over in the examples and AMOK.
2597 It has still to be documented and used (only amok/peermanagement is
2599 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2600 loss when they were small enough to fit into the buffer and sent quickly
2601 enough so that they can all get received in one shoot.
2602 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2603 if not found. Use the *_or_null() variant for the old semantic.
2604 * In gras_msg_handle, do not discard messages without callback.
2605 They are probably messages to be explicitly awaited later (ie, proofs of
2606 mis-synchronization in userland since they are sent before being awaited)
2607 No big deal usually.
2608 * gras_socket_meas_send/recv: semantic changed!
2609 The numerical arguments used to be (1) the total amount of data to send
2610 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2611 messages. This was need for the fool willing to send more than MAXINT
2612 bytes on quite fat pipes.
2615 * Do really rename the hostmanagement module to peermanagement. [Mt]
2616 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2617 but this is rather new and this was documented in the module
2618 documentation (poor excuses, I admit)
2619 * Bandwidth measurement semantic changed! This follows the changes to
2620 gras_socket_meas_send/recv explained above.
2623 * A sequential mode has been added to the workstations. When a workstation
2624 is in sequential mode, it can execute only one task, and the other tasks
2625 are waiting in a FIFO. [Christophe Thiery]
2628 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2629 model for SIMDAG. [Christophe Thiery]
2630 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2631 equal to 0, which caused some bad behaviors in
2632 saturated_constraint_set_update. I now use a threshold mechanism like in
2636 * When running manually src/testall, you select specific units [Mt]
2637 testall is the result of our cunit mechanism, and should replace all
2638 the scripty thingy around since bash don't run easily on billware.
2640 * A mallocator system has been added. [Christophe Thiery]
2641 Mallocators allow you to recycle your unused objects instead of freeing them
2642 and allocating new ones.
2644 Documentation update:
2645 * FAQ reworking + New FAQs:
2646 - "Valgrind spits tons of errors!" [Mt]
2647 - "How to repport bugs" [Mt]
2648 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2649 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2650 - Communication time measurement within MSG [AL]
2651 - I experience weird communication times when I change the latency [AL]
2652 * GRAS tutorial [Mt]
2654 - an introduction to the framework and to the used communication model
2655 - an initiatic tour introducing the most proheminent features:
2657 . Lesson 0: Installing GRAS
2658 . Lesson 1: Setting up your own project
2659 o Part 2: Message passing
2660 . Lesson 2: Exchanging simple messages
2661 . Lesson 3: Passing arguments to the processes (in SG)
2662 . Lesson 4: Attaching callbacks to messages
2663 . Lesson 5: Using globals in processes
2664 . Lesson 6: Logging information properly
2665 . Lesson 7: Using internal timers
2666 . Lesson 8: Handling errors through exceptions
2667 . Lesson 9: Exchanging simple data
2668 . Lesson 10: Remote Procedure Calling (RPC)
2669 . Lesson 11: Explicitely waiting for messages
2670 . Recapping of message passing features in GRAS
2671 - A HOWTO section containing:
2672 o HOWTO design a GRAS application
2673 More are due, of course. They will come latter. In the meanwhile, you can
2674 check the examples which are still here.
2676 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2678 SimGrid (3.1) stable; urgency=high
2682 There was a stack corruption somewhere, visible only when optimizing
2683 with these versions. [Vince]
2686 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2687 module that enabled to play with Directed Acyclic Graphs. It is built
2688 directly on top of SURF and provides an API rather close to the old
2689 SG. Some old codes using SG are currently under rewrite to check that
2690 all needful functions are provided. [Christophe Thiery]
2693 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2694 extension of the classical CLM03 model that gracefully handles
2695 failures. This is now the default model for MSG and GRAS. It doesn't
2696 handle parallel tasks yet though. [AL]
2697 * Bug fix: Weights were not correctly set in the network part.
2698 WARNING: This may have resulted in incorrect results with simulations
2699 where there are more than one flow on a given link. [AL]
2702 * After a (long ?) discussion on simgrid-devel, we have decided that the
2703 convention we had on units was stupid. That is why it has been decided
2704 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2705 WARNING : This means that all previous platform files will not work as
2706 such with this version! A warning is issued to ask users to update
2708 A conversion script can be found in the contrib module of the CVS, under
2709 the name contrib/platform_generation/surfxml_update.pl [MQ]
2712 * Bug fix: Processes were started in reverse order, wrt deployment file.
2713 WARNING: if your code relies on this bug, please fix it. [AL]
2714 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2715 being executed on two different locations. [AL]
2716 * Bug fix: Failures are now better supported thanks to Derrick's tests
2717 (there was many failure situations I hadn't thought of and that weren't
2718 correctly handled). [AL]
2719 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2723 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2724 Exception occurring on server-side are propagated back to client (!).
2726 API CHANGE: the callback changed their prototype. Change:
2727 int my_handler(gras_socket_t expeditor, void *payload_data) {
2729 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2730 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2732 * New! function: gras_msg_handleall to deal with all messages arriving
2733 within a given period.
2734 * New! function: gras_socket_server_range to get a server socket in a
2735 range of port numbers (ease to avoid port number conflicts) [MQ]
2736 * New! gras processes display their backtrace when they get a SIGUSR1
2737 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2738 Sweet to debug RL processes [MQ]
2742 - Do not force experiment sizes to be expressed in kb, or it becomes
2743 impossible to measure the latency this way (needs one byte-long tests)
2744 WARNING: this changes the amok_bw_* function semantic. [MQ]
2745 - Implements the link saturation stuff. [MQ]
2746 * Peer management module:
2747 New! module factorizing code that we wrote over and over [MQ].
2750 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2751 - Test units are placed directly into the library code, they get extracted
2752 automatically and placed into the src/testall binary.
2753 - Convert most of the XBT tests to this system.
2754 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2755 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2756 live in the same C code. This required to change a little bit the API
2757 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2758 * New module: sparse graph structure with basic algorithms (this is work
2759 in progress and the API is not considered to be frozen yet). [AL]
2760 * Display more information on backtraces: source line & function names are
2761 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2762 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2763 * Reworked a little bit some #include statements to load only required
2764 headers. Some user code that relied on SimGrid to include stdlib or
2765 stdio may need to include it by themselves. [AL]
2766 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2768 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2769 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2770 but I guess that almost nobody used those parts. [MQ]
2772 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2774 SimGrid (3.0.1) stable; urgency=low
2777 * Unfortunately, I had missed 5 misnamed functions:
2778 xbt_fifo_item_t xbt_fifo_newitem(void);
2779 void xbt_fifo_freeitem(xbt_fifo_item_t);
2780 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2781 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2782 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2783 They're now deprecated. Please use their new versions:
2784 xbt_fifo_item_t xbt_fifo_new_item(void);
2785 void xbt_fifo_free_item(xbt_fifo_item_t);
2786 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2787 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2788 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2790 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2791 * Documentation: xbt_log module unmercifully reworked [MQ]
2792 * Bugfix: there was a problem with the ending of contexts with
2793 the pthread backend. It caused some weird deadlock or behavior
2794 depending on the pthread implementation. [AL]
2795 * Bugfix: get the exceptions raised in the simulator repport where
2796 and why they come from when they are not catched in time [AL, MQ]
2799 * Bugfix: Do repport the error when two non-connected hosts try to
2800 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2803 * Add additionnal checkings on communications. Assert that two
2804 communicating hosts are connected by a set of links... [AL]
2807 * Add additionnal checkings on channel values in communication [AL]
2808 * New: MSG_task_get_source to see on which host a task was generated [HC]
2809 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2810 the number of tasks waiting to be received on channel and sent
2812 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2813 waits for the first task coming from a given host.. [AL]
2815 GRAS new functionnalities: [MQ]
2816 * Enhance the parsing macro to allow the size of multidimentional objects
2817 to be given thru annotations.
2818 * New example (and documentation): Matrix Multiplication a la RPC
2819 (as when I was young!) and fix a bunch of bugs found on the way.
2821 GRAS performance improvements: [MQ]
2823 * Reduce the amount of cbps creation/destruction by making it static to
2824 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2827 * Change libdata to a set so that we can search for stuff by ID (and thus
2828 reduce the insane amount of dict lookups)
2831 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2832 RL communication on top of SG since it's so uneffective.
2833 It may also be used for inter-thread communication in RL, one day.
2834 * Use gras_datadesc_copy() to exchange messages on top of SG
2836 - improve message exchange performance on top of SG
2837 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2838 * Don't exchange on the network the size of the used part of buffer,
2839 instead, specify the possible buffer size to read().
2841 - reduces the amount of read/write calls (one pair per exchange)
2842 - reduces the amount of exchanged data (the size)
2843 - allows to retrieve all arrived data on receiver side, if we don't need
2844 it right now (subsequent read will peek the buffer)
2845 - allows the receiver to proceed with the begining of the stream before
2846 everything is arrived
2847 - make it possible to build an iov transport (using readv/writev)
2849 - take care of the data with non-stable storage (like stacked data),
2851 * If possible, TCP send uses vector I/O (when writev() is here)
2852 - Don't use it for receive since we send data sizes and data on the
2853 same stream, so we wouldn't be able to chain large amount of chunks
2854 before having to flush the stuff to read the size.
2855 * Rework the transport plugin mechanism to simplify it and reduce the
2856 amount of pointer dereferencement when searching for the right function
2859 * I guess that now, we do almost as few system calls as possible while
2860 doing as few data copy as possible.
2862 To improve it further, we could try to send all the sizes first and then
2863 all the data (to use iov on receiving size), but it's only a partial
2864 solution: when you have 2 dimensional data, the sizes of the second
2865 dimension is data of the first dimension, so you need 3 streams.
2867 I'm not sure the potential performance gains justify the coding burden.
2869 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2871 SimGrid (3.00) stable; urgency=high
2874 * New! Give the possibility to hijack the surf parser and thus bypass
2875 MSG_create_environment and MSG_launch_application. Have a look at
2876 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2878 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2880 SimGrid (2.96) unstable; urgency=low
2885 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2886 This deprecates the xbt_error_t mechanisms.
2887 It modifies (simplifies) all XBT and GRAS API.
2888 MSG API keeps unchanged (exceptions raised by XBT are catched from
2889 within MSG and masked with existing error handling facilities)
2892 * New! Add a FATPIPE model. [AL]
2893 * New! Add a parallel task model. [AL]
2894 * New! Add automatically a loopback interface (in the default
2895 network model) if none was precised.
2898 * Bugfix: MSG_process_resume now works with the current running process.
2900 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2901 * Modification of MSG_task_get_compute_duration. Once a task has been
2902 processed, the value returned by this function is now equal to 0. [AL]
2903 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2904 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2905 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2906 of a m_task. For now, MSG_task_cancel only works with computation
2908 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2909 of the processor (in Mflop/s) regardless of the current load on the
2911 * API Change: use proper naming convention for MSG_getClock and
2912 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2914 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2915 This function changes the priority of a computation task. This priority
2916 doesn't affect the transfer rate. A priority of 2 will make a task
2917 receive two times more cpu power than the other ones. This function
2918 has been added to suit the needs of Nguyen The Loc and hasn't been that
2919 much tested yet. So if it fails, please report it and send me your code.
2921 * API Change: removed all functions and types that were marked "deprecated"
2922 since many months. Renamed MSG_global_init_args to MSG_global_init.
2924 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2926 SimGrid (2.95) unstable; urgency=low
2929 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2930 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2931 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2934 * Bugfix: Allow absolute paths for platform description files [MQ]
2935 * Bugfix: do free the variables after use. Leads to drastic performance
2937 * Implement max_duration (ie, timeouts) on resources [AL]
2940 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2942 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2943 the needed support (timeouts) [AL]
2946 * Implement measurement sockets. You can now get the bandwidth between two
2947 hosts thanks to AMOK (see below). [MQ]
2948 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2949 dynar over the network (yeah) [MQ]
2950 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2951 * simple/static token-ring example. [Alexandre Colucci and MQ]
2952 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2953 in place before [MQ]
2956 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2957 type parsing. "short" and "long" modifiers were also ignored; other
2958 modifier (such as reference level) are still ignored. [MQ]
2959 * Embeed the buffer size within the buffer itself on SG. [MQ]
2960 That way, send() are atomic and cannot get intermixed anymore (at least
2961 the ones which are less than 100k; bigger messages still have the issue)
2962 * Array size pushed by the field, not by the field type (or each
2963 and every long int will push stuff to the cbps) [MQ]
2964 * use select() to sleep since it allows to portably sleep less than one
2967 GRAS (minor cleanups)
2968 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2971 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2972 * type category 'ignored' killed as it was never used and were difficult
2974 * whether a type can cycle or not is now a flag, leaving room for more
2975 flags (as "ignored", if we feel the need one day ;)
2976 * Rename raw sockets to measurement sockets since "raw" has another
2977 meaning in networking community.
2980 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2981 and offers features not belonging to GRAS but that most applications
2982 need. One day, it may be a set of plugins loadable at runtime.
2983 * New module: bandwidth
2984 bandwidth measurement between arbitrary nodes running this module. [MQ]
2986 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2988 SimGrid (2.94) unstable; urgency=low
2990 The first beta release of SimGrid 3 !
2994 * Update the main page and the FAQ. Adding references to gforge.
2997 * Add a gras_os_getpid function.
3000 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3001 * Extend the logs so that they also print PID, hostname, date, ... if
3003 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3004 and kill the old version which were in testsuite/
3005 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3009 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3010 for users. Instead, each programming environment (GRAS, MSG,...) use it
3011 when needed to provide such a feature to users.
3012 Don't shortcut the mechanism or you will also shortcut the virtualization
3013 you need on the simulator.
3017 * Cleanups in configury with regard to compile optimization/warning flags.
3018 Also add -fno-loop-optimize to any powerpc since it's the optimization
3019 killing gcc (< 3.4.0).
3020 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3021 needed by MSG examples complications
3022 * Borrow configury beautifications from PHP
3025 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3026 warning (thanks loris for stumbling into this one).
3027 * Bugfix: stop loading private headers (gras_config.h) from the public
3031 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3032 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3033 RL would imply a DNS resolver, which is *hard* to do in a portable way
3034 (and therefore delayed).
3035 * Implement a real timer mechanism and use it in timing macros. This allows
3036 to avoid rounding errors and get a 0.000005 sec precision in timing
3037 macros. While I was at it, various cleanups:
3038 - allow to declare more than one timed section per file (fix a stupid bug)
3039 - move some private declaration to the right place
3040 - merge conditional execution and timing macros into emulation module
3041 - document the module
3042 - make sure the module cleanups its mess on gras_exit
3043 * Documentation improvements:
3044 - (new) how to compile applications using GRAS
3045 - (new) emulation support (timing macros)
3047 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3049 SimGrid (2.93) unstable; urgency=low
3051 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3054 - Use Paje properly where used. Still to be sanitized properly.
3055 - Portability fix: Add an implementation of the contexts using pthread
3059 - Add xbt_procname(): returns the name of the current process.
3060 Use it to show the current process's name in all logging.
3062 - fix detection of older flex version and the reaction, since we do
3063 depend on modern ones (we use lex_destroy)
3064 - Better separation of SG and RL in the libs: remove all simulation code
3065 from libgras. As a result, this lib is now only 200k when stripped.
3066 Some of the xbt modules may also be duplicated (two sets and such) and
3067 should be cleaned/killed before SG3.
3068 - Insist on using xlC on AIX because of weird problems involving gcc there.
3069 - Cleanup the make remote stuff. This is now done by scripts
3070 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3071 mainly for our private use, but we're working on changing them to user
3074 - Bugfix: flush the socket on close only if there is some *output*.
3075 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3077 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3080 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3082 SimGrid (2.92) unstable; urgency=low
3084 Alpha 3 on the path to SimGrid 3
3088 - New! First try of benchmarking macros.
3089 - New! First try so that gras_stub_generator generate deployment and
3090 remote compilation helpers.
3092 - Bugfix: Initialization fix in msg_test.
3096 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3098 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3100 - API Change: gras_os_sleep to take the amount of seconds as a double.
3101 Accepting an int was error prone since it was the only location where
3102 seconds were coded as such. It leaded to damn rounding errors.
3103 - Bugfix: Hard to belive that timers ever worked before this.
3105 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3107 SimGrid (2.91) unstable; urgency=low
3109 Alpha 2 on the path to SimGrid 3
3113 - Bug fix in the lmm_solver.
3115 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3116 through the function MSG_paje_output.
3117 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3118 - It is possible to bound the rate of a communication in MSG with
3119 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3120 to write it in the changelog).
3121 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3126 - Various cleanups to the autotools stuff
3127 - Begin to move Gras examples to examples/gras/
3128 - Let make distcheck work again (yeah!)
3130 - documentation overhauled using doxygen.
3131 gtk-doc-tools is dead in SimGrid now.
3132 - Automatically extract all existing logging categories, and add the list
3133 to the documentation (long standing one, to say the less)
3135 - Cleanup the known architecture table. Reorder the entries to group what
3136 should be, and use a more consistent naming scheme.
3137 (some of the test dataset are still to be regenerated)
3138 - New! Allow library to register globals on each process just as userdata
3140 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3141 lookup time (for now).
3142 Use it in msg and trp.
3143 This cleans a lot the internals and helps enforcing privacy of the
3144 headers between the gras components.
3145 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3146 - Bugfix: gras_os_time was delirious in RL.
3147 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3149 - Reenable GRAS now that it works.
3151 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3153 SimGrid (2.90) unstable; urgency=low
3155 Alpha 1 on the path to SimGrid 3
3157 * It is a long time since the last release of SimGrid. I'm sorry about
3158 that but as I had told you, I was rewriting a lot of things. I apologize
3159 to those who had been reporting bugs to me and that I had not answered.
3160 If your bug is still in the new version, please tell me. Here is a
3161 summary of the main changes.
3163 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3164 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3165 portability, speed, and a lot more but you'll figure it out later.
3166 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
3167 I wanted to release the new MSG as soon as possible and I have
3168 broken GRAS, which is the reason why, for now, only MSG is fully
3169 functional. A laconic description of these projects is available
3170 in the documentation.
3172 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3173 kernel optimized for our needs (called SURF but only the developers
3174 should use it). Hence, MSG is now roughly 30 times faster and I think
3175 that by rewriting a little bit MSG, I could event speed it up a little
3176 bit more. Beside the gain in speed, it is also much easier to encode a
3177 new platform model with SURF than it was with SG. More to come...
3179 * REVOLUTION 3: I have tried to change a little as possible the API of
3180 MSG but a few things really had to disappear. The main differences
3181 with the previous version are :
3182 1) no more m_links_t and the corresponding functions. Platforms are
3183 directly read from a XML description and cannot be hard-coded
3184 anymore. The same format is used for application deployment
3185 description. The new format is described in the documentation.
3186 Have a look in tools/platform_generation. There is a tiny script
3187 that converts from the old platform format to the new one. Concerning
3188 the application deployment format, parsing the old one is tricky.
3189 I think most of you should however be able to convert your files. If
3190 it is really an issue, I can write a C code that does the conversion.
3192 2) the toolbox tbx does not exist anymore. We now have a library
3193 with much more data-structures but without the hash-tables (we have
3194 dictionaries that are much faster).
3196 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3198 *****************************************************************************
3199 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3200 * don't feel like converting the oldies. (Mt) *
3201 *****************************************************************************
3204 Version 2.90: "the long awaited one"
3205 - Finished rewriting and debugging MSG. Rewrote the documentation.
3206 - disable GRAS for now since it needs to be ported to the newest SG
3209 - Finish the port to windows (using mingw32 for cross-compile)
3212 - Main loop and datastructures of SURF. A cpu resource object is
3213 functional. Surf can thus be used to create cpu's with variable
3214 performance on which you can execute some actions.
3216 2004-11-15 Martin Quinson
3217 - Port to ARM. Simply added the alignment and size descriptions. Should
3218 work, but the ARM machines are so slow that I didn't had the opportunity
3219 to 'make check' over there yet.
3221 2004-11-15 Arnaud Legrand
3222 - Trace manager now written. It uses a heap structure and is therefore
3223 expected to be efficient. It may however be speeded up (particularly
3224 when many events occur at the same date) by using red and black
3225 trees. One day maybe...
3226 - Max-min linear system solver written. It uses a sparse matrix
3227 structure taking advantage of its expected use. Most operations are
3228 O(1) and free/calloc are called as few as possible. The computation of
3229 the minimum could however be improved by using a red and black tree
3232 2004-11-03 Arnaud Legrand
3233 - Rename every gras_* function that was in xbt/ to its xbt_
3235 - Add a heap and a doubly-linked list to xbt
3236 - Added a dichotomy to the dictionaries. make check works as well before
3237 so I assume that the patch is correct. I do not know however if things
3238 run effectively faster than before now. :)
3240 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3241 SimGrid, and the version number is bumped to 2.x
3243 2004-10-29 Martin Quinson
3244 - Introduction of the remote errors.
3245 They are the result of a RMI/RPC on the remote machine.
3246 ErrCodes being scalar values, you can't get the host on which those
3247 errors did happen. Extending the error mechanism as in Gnome is possible.
3248 No idea yet whether it is a good idea.
3250 2004-10-28 Martin Quinson
3251 - Interface revolution: the Starred Structure Eradication.
3252 I used to do typedef struct {} toto_t; and then handle *toto_t.
3253 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3254 - ???_t is a valid type (builded with typedef)
3255 - s_toto_t is a structure (access to fields with .)
3256 - s_toto is a structure needing 'struct' keyword to be used
3257 - e_toto_t is an enum
3258 - toto_t is an 'object' (struct*)
3260 typedef struct s_toto {} s_toto_t, *toto_t;
3261 typedef enum {} e_toto_t;
3262 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3263 s_toto_t) is private.
3265 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3266 so that it can set the variable to NULL. It was so for dicts and sets,
3267 it changed for dynars.
3269 - Fix a bunch of memleaks in dict_remove
3270 - Fix a bug in sg/server_socket opening: it failed all the time.
3272 2004-10-07 Martin Quinson
3273 - Speed up dynar lookup operation a bit.
3275 gras_dynar_get is dead.
3277 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3278 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3279 (which gives you the address of the stored data).
3281 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3282 copy of the data using an affectation to do the job and not a memcpy.
3284 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3285 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3287 It does not really speedup the dynar test because they are
3288 setting elements all the time (and look them seldom). But the dict does
3289 far more lookup than setting.
3291 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3293 2004-10-05 Martin Quinson
3294 - Allow to (en/dis)able the cycle detection at run time.
3296 Whether we should check for cycle or not is now a property of each
3297 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3298 datadesc_cycle_unset allow to remove this property when previously set.
3300 Note that the cycle detection is off by default since it impacts the
3301 performance. Watch the data you feed GRAS with ;)
3303 This property is hereditary. Any element embedded in a structure having it
3304 set have it set for the time of this data exchange.
3306 You should set it both on sender and receiver side. If you don't set it on
3307 sender side, it will enter an endless loop. If you forget on receiver
3308 side, the cycles won't be recreated after communication.
3310 - Header reorganization.
3311 Kill gras_private.h, each submodule must load the headers it needs.
3313 2004-10-04 Martin Quinson
3314 - Interface revolution: do not try to survive to malloc failure.
3316 Now, gras_malloc and friends call gras_abort() on failure.
3317 As a conclusion, malloc_error is not a valid error anymore, and all
3318 functions for which it was the only gras_error_t return value are
3319 changed. They now return void, or there result directly.
3320 This simplify the API a lot.
3322 2004-09-29 Martin Quinson
3323 - Re-enable raw sockets.
3324 Created by gras_socket_{client,server}_ext;
3325 Used with gras_raw_{send,recv}
3328 It should allow to kill the last bits of gras first version soon.
3330 This is not completely satisfactory yet (duplicate code with
3331 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3334 - Simplify transport plugin (internal) interface by not passing any
3335 argument to _server and _client, but embedding them in the socket
3338 2004-09-28 Martin Quinson
3339 - Finish the port to AIX.
3340 autoconf was my problem (segfault within the malloc replacement
3341 function. No idea why)
3343 2004-09-16 Martin Quinson
3344 - Fix some size_t madness on 64bit architectures.
3346 2004-09-08 Martin Quinson
3347 - Reduce the number of system headers loaded, overload some more system
3348 calls (such as malloc to cast the result of the system one, and work
3350 - Fix and reintroduce the config support
3352 2004-09-07 Martin Quinson
3353 - Source code reorganization to allow Arnaud to surf all over there.
3354 - Allow to document the logging categories.
3355 - Remove all uppercase from logging categories and useless cleanup in names.
3357 2004-08-18 Martin Quinson
3358 Version 0.6.2 (protocol not changed; API changed)
3359 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3360 gras_error_t), and NULL when not found. Functions expecting a msgtype
3361 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3362 hopefully usefull message.
3363 - Portability to prehistoric sparcs again
3365 2004-08-17 Martin Quinson
3366 Version 0.6.1 (protocol not changed; ABI not changed)
3367 - prealloc some buffers to speed things up
3369 2004-08-11 Martin Quinson
3370 Version 0.6 (protocol not changed; ABI expended)
3371 - The parsing macro can deal with the references, provided that you add
3372 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3374 2004-08-09 Martin Quinson
3375 Version 0.5 (protocol not changed; ABI changed)
3376 - Allow to off turn the cycle detection code in data exchange at
3377 compilation time. It should be at run time, but I'm short of time (and
3378 the config stuff is still broken). That way, we keep dict out of the
3379 critical path, which is good because the performance is poor:
3380 - search not dichotomial yet
3381 - dynar give no way to access their content and memcpy everytime
3382 - In composed data description (struct, ref and so on), stop foolness of
3383 keeping the subtype's ID, but store the type itself. This keeps sets out
3384 of the critical path, which is good since they rely on dynar and
3385 dictionnaries. The only loose of that is that we cannot detect the
3386 redeclaration of a structure/union with another content (but I'm not sure
3387 the code detected well this error before anyway). We still can detect
3388 the redefinition discrepancy for the other types.
3389 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3390 it breaks the code because of type-punning used all over the place).
3391 This breaks on all non-gcc architectures (for now).
3393 All those changes (plus the buffer of last time) allow me to gain 2 order
3394 of magnitude on cruel tests consisting of 800000 array of integers on two
3395 level of a hierarchical structure (200 secondes -> 4 secondes)
3398 - the selector of reference must now return the type it points to, not
3399 the ID of this type.
3401 2004-08-06 Martin Quinson
3402 Version 0.4 (protocol changed; ABI not changed)
3403 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3404 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3405 - Implement a buffer transport, and use it by default (it relies on tcp in
3406 real life and on sg in simulation).
3407 That's a bit hackish since I had a new field to the structure to store
3408 its data without interfering with the subtype ones. Inheritance
3409 is tricky in C. And that's a kind of reverse inheritance with one class
3410 derivating two classes. Or maybe a game with java interfaces. Anyway,
3411 that's damn hard in C (at least).
3412 Moreover, I got tired while trying to ensure plugin separation and
3413 genericity in SG mode. MSG wants me to do weird things, so let's go for
3414 cruel hacks (temporarily of course ;).
3415 See comment in transport_private.h:71
3416 - do not include all the _interface headers in private but in the files
3417 which really need them (to cut the compilation time when they are
3420 2004-07-26 Martin Quinson
3421 Version 0.3 (protocol not changed; ABI changed)
3422 - Major overhault of the datadesc interface to simplify it:
3423 - shorted the function names:
3424 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3425 - add a trivial way to push/pop integers into the cbps without malloc.
3426 This allows to make really generic sub_type description, which simply
3427 pop their size of the stack.
3428 - add a function gras_datadesc_ref_pop_arr() which does what users want
3429 most of the time: Declare a dynamic array (which pops its size of the
3430 stack) and declare a reference to it. Poor name, but anyway.
3431 - kill the post-send callback, add a post-receive one
3433 2004-07-23 Martin Quinson
3434 Version 0.2 (protocol changed; ABI changed)
3435 - add some testing for cpbs in the test cases, and fix some more bugs.
3436 This invalidate again the little64 data file, since I cannot regenerate
3438 - remove an awfull optimization in the logging stuff, allowing me to:
3439 - understand it again
3440 - learn gcc how to check that the argument match the provided format
3441 - fix all errors revealed by gcc after that
3442 - internal keys of dict are not \0 terminated. Deal with it properly in
3443 loggings instead of segfaulting when the user want to see the logs :-/
3445 2004-07-22 Martin Quinson
3446 - Fix some stupid bug preventing cbps (callback postit) from working
3448 2004-07-21 Martin Quinson
3449 - Some documentation cleanups
3450 - remove the useless last argument of msgtype_declare
3451 - rename the Virtu functions to fit into the 'os' namespace
3452 - move headers src/include -> src/include/gras/ and stop fooling with
3453 gras -> . symbolic link
3454 - make distcheck is now successful
3456 2004-07-19 Martin Quinson
3458 - Build shared library also
3459 - Install html doc to the right location
3460 - stop removing maintainer files in make clean
3461 - build tests only on make check
3463 2004-07-13 Martin Quinson
3465 - No major issue in previous version => change versioning schema
3466 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3467 corresponding dataset.
3469 2004-07-11 Martin Quinson
3471 - Get it working with any kind of structure (we can compute the padding
3472 bytes remotely for all the architectures I have access to)
3473 - Implement the structure parsing macro (still not quite robust/complete)
3474 - Improvement to the remote testing toysuite
3476 2004-07-10 Martin Quinson
3477 [autoconf mechanism]
3478 - get ride of a bunch of deprecated macros
3479 - actually run the test for two-compliment, not only compile it :-/
3480 - test whether the structures get packed (and bail out if yes. Damn.
3481 Alignment is a serious matter)
3482 - test whether the structures get compacted (but respecting the alignment
3483 constraints of each types)
3484 - test whether the array fields of structures can straddle alignment boundaries
3486 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3487 (took me 2 hours to find that bug, looking at the wrong place)
3488 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3489 before sending/receiving and are used to compute the offsets of fields
3490 - Given that padding size depend even on compiler options, keep track of
3491 alignment and aligned_size only for the current architecture. Not a big
3492 deal since we send structure fields one after the other (seems
3494 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3495 program, let it work on linux/gcc/little32. portability todo.
3497 2004-07-08 Martin Quinson
3498 - import and improve remote compilation support from FAST
3499 - make sure make check works on half a dozen of machines out there
3501 2004-07-07 Martin Quinson
3502 Let's say it's version 0.0.3 ;)
3503 - Implement conversions (yuhu!)
3504 - Let it work on solaris (beside conversion, of course)
3505 - Stupid me, using rand() to generate the conversion datatests in not wise.
3507 2004-07-06 Martin Quinson
3508 - Let make dist work, since I'm gonna need it to compile on remote hosts
3509 - Let Tests/datadesc_usage write the architecture on which the file was
3510 generated as first byte.
3511 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3512 architecture descriptions.
3513 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3514 on those architectures.
3515 - Optimization: send/recv array of scalar in one shoot
3517 2004-07-05 Martin Quinson
3518 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3520 - Plug a whole bunch of memleaks
3521 - each process now have to call gras_{init,exit}. One day, their log
3522 settings will be separated
3523 - Continue the code factorisation between SG, RL and common in Transport.
3525 2004-07-04 Martin Quinson
3527 - Redistribution between SG and RL.
3528 We wanna have to accept in SG, so move accepted related parts of RL in
3529 the common part. (more precisely, the dynar of all known sockets is no
3530 more a static in transport.c, but part of the process_data)
3532 [gras_stub_generator]
3533 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3535 2004-07-03 Martin Quinson
3536 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3537 them all up in one shot)
3539 - Fix the ugly case of reference to dynamic array.
3540 - New (semi-public) function gras_datadesc_size to allow the messaging
3541 layer to malloc the needed space for the buffer.
3543 - gras_socket_close now expect the socket to close (and not its address to
3544 put NULL in it after it). This is because the socket passed to handlers
3545 is one of their argument (=> not writable).
3547 - propagate the interface cleanup from last week in datadesc, ie remove a
3548 superfluous level of indirection. User pass adress of variable
3549 containing data (both when sending and receiving), and not of a variable
3550 being a pointer to the data. Let's say that I like it better ;)
3551 The price for that is constructs like "int msg=*(int*)payload" in
3552 handlers, but it's a fine price, IMHO.
3554 - Let it work in RL (yuhu)
3556 2004-06-21 Martin Quinson
3558 - porting SG plugin and SG select to new standards (works almost).
3559 - plug memleaks and fix bugs around.
3562 - cleanup the prototype of data recv and force users to specify when they
3563 want to handle references to objects. Test case working even for cycles.
3564 - plug memleaks. Valgrind is perfectly ok with this.
3566 2004-06-12 Martin Quinson
3568 - cleanup the separation between plugin and main code in plugin creation
3570 2004-06-11 Martin Quinson
3572 - Reput hook for raw sockets, needed for BW experiments
3573 - kill a few lines of dead code
3574 [Data description] Interface cleanup
3575 - gras_datadesc_by_name returns the searched type or NULL.
3576 That way, no variable is needed to use a type desc once, which makes
3578 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3579 two parameters were strings (field name, type name), leading to
3581 [Dicos] Interface cleanup
3582 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3583 This is consistant with the dynar API.
3585 2004-04-21 Martin Quinson
3587 - Porting to new standards.
3589 - interface cleanup.
3590 There is no bag anymore, no need to take extra provision to mask the
3591 pointers behind "ID".
3592 Better splitup of functions between files create/exchange/convert.
3593 This is still a bit artificial since convert and receive are so
3594 interleaved, but anyway.
3596 - add a queued message list to procdata (the ones not matching criteria
3598 - factorize some more code between SG and RL wrt procdata
3600 - use gras_exit in example to track memleaks
3601 - get rid of gs_example now that GS is properly integrated into gras
3602 - update run_test to integrate the latest tests (datadesc)
3604 - rename WARNINGn macros to WARNn since it prooved error-prone
3606 2004-04-19 Martin Quinson
3608 - register init/exit functions within gras module mechanism
3609 - send/receive function.
3610 Convertion is not implemented, but short-cutted if not needed.
3611 struct/array elements are sent one by one (instead of block-wise), but
3612 nobody really cares (yet). Get a prototype before optimizing.
3613 - tests (using a file socket) for DD send/receive on:
3614 - base types: int, float
3615 - array: fixed size, string (ie ref to dynamic string)
3616 - structure: homogeneous, heterogeneous
3617 - chained list, graph with cycle
3618 Believe it or not, valgrind is not too unhappy with the results. The
3619 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3623 - Bugfix when using a filename explicitely (instead of '-')
3625 2004-04-09 Martin Quinson
3627 - factorize more code between RL and SG in socket creation
3628 - Complete the implementation and tests of:
3630 o file (only in RL, and mainly for debugging)
3632 I lost 3 days to design a portable address resolver, and then decided
3633 that the prototype mainly have to run on my box.
3634 Addressing portability too early may be like optimizing too early :-/
3636 - use gras_init in the Tests instead of the crappy parse_log_opt
3637 (the latter function is removed)
3638 [Conditional execution]
3639 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3640 [Code reorganisation]
3641 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3642 Build examples against the RL library, since there is no way to disable
3643 its creation for now.
3645 For information, the beginning of coding on GRAS was back in june
3646 2003. I guess that every line has been rewritten at least twice since