1 SimGrid (3.22) NOT RELEASED (Release Target: December 21. 2018, 22:23 UTC)
3 SimGrid (3.21) October 3. 2018
5 The Restarting Documentation (TRD) Release.
8 - Convert the user manual to Sphinx for improved usability.
9 Unstable version now hosted on https://simgrid.frama.io/simgrid
10 Many glitches remain; stay tuned.
11 - Restrict the Doxygen reference API to the installed header files.
12 (documenting all internals this way is overkill + doxygen scales badly)
13 - New tutorial on S4U (highly inspired from the old MSG one, but with
14 a git repository to fork as a starting point, and a docker image)
15 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
16 - The Developper manual is still to be converted (not compiled until then)
17 - Some parts are still missing in this conversion (such as the
18 platform and deployment sections) while others were blindly converted
19 and would need more love (such as the configuration flags).
20 Things will certainly further improve in the future releases.
23 - Fully reimplement auto_restart mechanism that was utterly broken
24 (fix #22, #131 and #281 that were all very related)
25 - Implement semaphores (fix #303 and #291)
26 - When creating an actor from a function and its parameters,
27 move-only parameters are not allowed anymore, as it would prevent
28 the actor to be restartable if its parameters are consumed on run.
29 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
30 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
31 IO operations that can be started, waited for, or canceled as a regular
33 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
37 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
38 - New host variable: 'core_count'
41 - Remove xbt_os_thread_specific features
42 - Remove portability wrapper to condition variables
43 - Remove xbt_os_thread_yield()
46 - MPICH collective selector now mimics MPICH 3.3b
47 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
48 - MPI_Init(NULL,NULL) is now allowed
49 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
50 - Command-line arguments should now be handled properly in Fortran simulations
51 - MPI Attributes and related callbacks should work in Fortran
52 - Apps using C + Fortran should now work
53 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
54 - Support MPI_Error_String
57 - Due to an internal bug, Msg.run() must now be your last line.
58 We hope to fix it in a future release, and we are sorry for the inconvenience.
61 - #22: Process autorestart seem to only work with CAS01 cpus
62 - #56: Feature request: dynamic SMPI replay
63 - #93: simgrid should not eat --help
64 - #111: How to change the loopback link characteristics?
65 - #116: Platform cluster radical
66 - #117: Platform cluster/host order
67 - #129: Replay_multiple on steroid
68 - #131: Java : autostart does not work
69 - #143: Setting a breakpoint at a given time
70 - #235: Network energy models should be integrated
71 - #264: Add ptask L07 resource tracing
72 - #271: Dynamic and manual replay of SMPI traces
73 - #279: Breakpoints option uses time deltas instead of absolute time
74 - #281: Daemonized actor no longer daemonized after an auto-restart
75 - #282: TIT convertor erase traces when using absolute path in the trace list file
76 - #285: segfault when a s4u actor kills itself with msg process tracing activated
77 - #286: Pajé traces are not exposing the number of cores
78 - #287: Command-line parsing should stop when encountering '--'
79 - #288: MPI_Init(NULL, NULL)
80 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
81 - #290: Method to check if a mailbox has a message ready to be consumed
82 - #291: [s4u] semaphores do not seem available yet
83 - #294: Alltoall collective implementation of mpich
84 - #295: Starting a Host as OFF is not supported
85 - #296: DTD too permissive
86 - #299: Add a s4u barrier example
88 ----------------------------------------------------------------------------
90 SimGrid (3.20) Released June 24. 2018
92 The proxy snake_case() release.
95 - The whole API is now using snake_case() instead of mixing it with camelCase.
96 Compatibility wrappers in place for one year.
97 - We took the opportunity to increase the consistency of the API.
98 Please report any remaining glitches.
99 - New function simgrid::s4u::Host::get_actor_count:
100 Returns the number of actors running on a specific host.
103 - SMPI is now tested with ~45 proxy apps from various sources, with none or
104 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
105 - Replay: The replay file has been re-written in C++.
106 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
107 supported. They are stored in the trace and used when replayed.
108 - Basic support of MPI_Cancel. Robustness not guaranteed.
109 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
110 (only for MPI_COMM_TYPE_SHARED).
111 - New option: smpi/privatize-libs, to add external shared libs to be privatized
112 by SMPI. They will be copied locally and loaded separately by each process.
113 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
114 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
115 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
116 with internal index optind with multiple processes. Only works if MPI_Init has
118 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
119 syntax without preprocessing tricks.
122 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
123 behavior tracing is no longer limited to MSG
124 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
125 this means that traces now need to be updated or re-obtained; both Irecv
126 and Isend lines in your traces have now in total 5 mandatory fields:
127 <rankid> <command> <to/from rankid> <tag> <size>
128 To update your traces, it suffices to add a 0 for the tag here.
129 - TIT now also supports waiting for a distinct request via MPI_Wait.
130 Wait/Test now wait for a specific request, not just the last one that was
131 issued. This unfortunately means another update, because we need to
132 identify which request you want to wait for. We do this via the
133 triplet (sender, receiver, tag), which needs to be added:
134 <rankid> <command> <sender> <receiver> <tag>
135 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
136 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
139 - The deprecation of MSG is ongoing (but this should not impact you).
140 Many MSG functions are now simple wrappers on the C API of S4U. If
141 you wish to convert your code to S4U, find the S4U counterparts of
142 your MSG calls in src/msg/msg_legacy.cpp.
143 - MSG can still be used, but won't evolve anymore.
146 - Allow to run the Link energy plugin from the command line with
147 --cfg=plugin:link_energy
148 - Rename Energy plugin into host_energy
149 - Rename Load plugin into host_load
152 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
153 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
154 - Improve performance when handling timeouts of simix synchros.
157 - Config: the C API is now deprecated (will be removed in 3.23), and
158 the C++ API has been slightly improved.
161 - Fix several build issues on OSX.
162 - Move simgrid_config.h to simgrid/config.h (old header still working)
165 - #143: Setting a breakpoint at a given time
166 - #150: Inconsistent event names in SMPI replay
167 - #258: daemonized actors hang after all non-daemonized actors have completed
168 - #267: Linker error on unit_tmgr
169 - #269: SMPI: tracing of MPI_Wait/all/any broken
170 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
172 ----------------------------------------------------------------------------
174 SimGrid (3.19.1) Released March 22 2018
176 The Fixed ABI Release.
178 As you may know, we are currently refactoring SimGrid in deep.
179 Upcoming SimGrid4 will be really different from SimGrid3: modular,
180 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
182 Our goal is to smooth this transition, with backward compatibility
183 and automatic update paths, while still progressing toward SimGrid4.
185 SimGrid remains open during works: The last pure SimGrid3 release was
186 v3.12 while all subsequent versions are usable alpha versions of
187 SimGrid4: Existing interfaces remain unchanged, but the new S4U
188 interface is budding and the internals are deeply reorganized.
190 Since 2015, we work hard to reduce the changes to public APIs. When
191 we need to rename a public library symbol in S4U, we let your
192 compiler issue an explicative warning when you use the deprecated
193 function. These messages remain for four releases, i.e. for one full
194 year, before turning into an error. Starting with v3.15, your can
195 also adapt to API changes with the SIMGRID_VERSION macro, that is
196 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
198 Starting with this v3.19.1, our commitment to reduce the changes to
199 the public interfaces is extended from the API to the ABI: a program
200 using only MSG or SimDag and compiled against a given version of
201 simgrid can probably be used with a later version of SimGrid without
202 recompilation. We will do our best... but don't expect too much of
203 it, that's a really difficult goal during such profund refactoring.
205 The difference between v3.19 and v3.19.1 is that the former was
206 accidentally breaking the ABI of MSG, while the later is restoring
209 S4U and kernel APIs will still evolve until SimGrid4, with one-year
210 deprecation warnings as currently. In fact, cleaning up these
211 interfaces and converting them to snake_case() is one release goal of
212 v3.20. But don't worry, we are working to smooth this upgrade path.
214 In summary, new projects should start with S4U to benefit of the
215 future, but old MSG projects should still be usable with no change.
217 ----------------------------------------------------------------------------
219 SimGrid (3.19) Released March 20 2018
221 The Moscovitly-cold Spring release.
223 SMPI user-visible improvements:
224 - Support of MPI_Type_create_subarray
225 - Unimplemented MPI calls now abort (instead of issuing a warning)
226 Some harmless calls around errhandlers still only issue a warning.
227 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
228 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
229 is now possible (mmap privatization was refactored to this end).
230 Adding new SMPI instances should also be technically doable now.
232 SMPI internal cleanups:
233 - Removed index notion from SMPI -- no more "getPid() - 1"!
234 In tracing files, actors are called rank-<PROCESS-ID> but these are
235 NOT the real ranks (tracing several communicators is not supported yet).
236 - Deprecate smpi_process_index() (will be removed in 3.22)
237 - Better testing on classical proxy apps, through an external project:
238 https://github.com/simgrid/SMPI-proxy-apps/
241 - Use a graphical TOC to make it easier to find the documentation you need
242 - Revamp the MSG tutorial
245 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
246 - Remove parameter of MSG_process_killall().
247 Resetting the PID was bogus anyway (several actors could have the same PID).
250 - Execution->setHost() can be called after start(), to migrate it.
251 - Comm::test_any() is now implemented.
252 - s4u::Actor now has onCreation() and onDestruction() signals.
253 - Install some kernel header files for the users' plugins and more.
254 Warning, their API is really not stable yet.
255 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
256 include/simgrid/kernel/routing/*Zone.hpp
259 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
261 Build System and other cleanups:
262 - Remove unused run-time parameter "tracing/onelink-only".
263 - SimGrid now works with intel compilers (MC still buggy)
264 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
265 - tesh kills subprocesses on timeout (but not on windows)
268 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
271 - #194: Feature request: simgrid::s4u::Comm::test_any()
272 - #245: migrating an actor does not migrate its execution
273 - #253: Feature Request: expose clusters as objects
274 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
275 - #255: Tesh broken on Windows
276 - #256: Modernize FindSimGrid.cmake
277 - #257: Fix (ab)use of CMake install
279 ----------------------------------------------------------------------------
281 SimGrid (3.18) Released December 24 2017
283 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
285 MSG IS NOW DEPRECATED. Please use S4U for new projects.
286 - Support for MSG should not be removed from SimGrid before 2020,
287 but future evolutions will be limited to the integration of
288 user-provided patches.
291 - s4u::onDeadlock() signal, run before stopping the simulation.
292 - s4u::Exec: asynchronous executions (abstraction of background
294 - s4u::Host->getLoad() returns the achieved speed in flops/s
297 - Link::name() is deprecated and will be removed in v3.21, use
298 Link::getCname() instead.
299 - Mailbox::getName() changed to return a std::string, use
300 Mailbox::getCname() instead to get a char*.
301 - Storage::getName() changed to return a std::string, use
302 Storage::getCname() instead to get a char*.
303 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
304 to have both allocation and deallocation of the map in user space.
307 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
308 The amount of flops remaining to do is a vector, not a scalar.
309 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
310 name implies on both sequential and parallel tasks.
311 - Both changes fix GitHub's #223 using PR #237 as a basis.
312 Thanks Michael Mercier.
313 - Most examples were converted to S4U and hidden elsewhere at tests.
314 Rationale: we still want MSG to work; we want newcomers to use S4U.
317 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
318 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
321 - Switch to the faster dlopen privatization mechanism by default
322 - Documentation improvments
325 - Live migration is getting moved to a plugin. Dirty page tracking is
326 the first part of this plugin. This imply that VM migration is now
327 only possible if one this function is called:
328 - C/MSG: MSG_vm_live_migration_plugin_init()
329 - C/C++: sg_vm_live_migration_plugin_init()
330 - Java: Msg.liveMigrationInit()
331 For C and C++, "simgrid/plugins/live_migration.h" also has to be
335 - Define class simgrid::xbt::Path to manage file names.
336 - Removed unused functions:
337 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
338 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
339 xbt_graph_export_graphviz()
340 - xbt/str.h: xbt_str_join()
341 - Remove unused datatypes:
342 - xbt/heap.h: use std::priority_queue or boost::heap instead
343 - xbt/swag.h: use boost::intrusive::list instead
346 - New link_energy plugin for the consumption of the links.
347 - All of the operations on files and storage contents have been
348 packaged into a plugin (src/plugins/file_system). The current
349 public interface can be found in
350 include/simgrid/plugins/file_system.h
351 To use these functions you now have to initialize the plugin by
352 calling MSG_storage_file_system_init() just after calling
353 MSG_init() or sg_storage_file_system_init() just after creating
357 - Remove the undocumented/untested tag <include>
360 - Remove viva specific tracing as the tool is no longer maintained
363 - #248: Kill of finished processes leads to segfault
364 - #240: xbt_cond_wait_timeout should gracefully return for C apps
365 - #239: please implement signal s4u::onDeadlock()
366 - #230: segfaults when exit() without run()
367 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
368 - #223: MSG_task_get_flops_amount() not working with parallel tasks
369 - #222: Actor::kill() doesn't really kill and segfaults
370 - #221: odd LMM warning when killing an actor
371 - #120: Memory leak when the processes are forcefully killed
373 ----------------------------------------------------------------------------
375 SimGrid (3.17) Released October 8 2017
377 The Drained Leaks release: (almost) no known leaks despite the tests.
379 Even very long simulations will preserve your memory: our testsuite
380 stresses SimGrid for over 45mn on fast machines for a coverage of
381 over 80%, and there is only one single known leak, of about 4kb.
385 - Comm.detach(): start and forget about asynchronous emission. A cleanup
386 handler may be given to free resources if the comm cannot be completed.
387 - this_actor::send(mailbox) is now mailbox->put()
388 - New: simgrid::s4u::Comm::wait_all()
389 - New: Host.execute() for remote executions.
392 - Unused option network/sender-gap is removed.
395 - Deprecate MSG_task_isend_with_matching(): This unused feature
396 really complicates our internals. Will be removed in v3.20.
399 - Improved context termination. It is now possible to free resources, even
400 when a process is forcibly killed.
401 - ContextBoost: add support for Boost versions above 1.61.
404 - Bring back run-time option --cfg=exception/cutpath to remove exception
406 - Removed unused functions:
407 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
408 xbt_str_rtrim(), xbt_str_trim().
409 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
412 - Removed header files obsolete since SimGrid 3.12:
413 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
415 - #3: SD_exit should be made optional
416 - #120: Memory leak when the processes are forcefully killed
417 - #159: Threading test regression in Actor refcounting
418 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
419 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
420 - #186: Actor::killAll() segfaults if some process is blocked on wait()
421 - #191: VM migration and pstate
422 - #192: Updating the energy consumptions of all hosts crashes with VMs
423 - #195: All actors have PID=0 in the logs
424 - #204: Sometimes segfault with thread contexts and mmap privatization
425 - #222: Actor::kill() doesn't really kill and segfaults
426 - #225: Actor::kill() doesn't really kill when victims are doing a join()
428 ----------------------------------------------------------------------------
430 SimGrid (3.16) Released June 22. 2017.
432 The Blooming Spring Release: developments are budding.
435 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
436 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
437 - Java JAR file should be finally fully working on Mac OSX too.
439 XML platforms: Switch to platform v4.1 format.
440 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
441 - <zone> can be used as a synonym for the now deprecated <as>
442 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
443 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
444 - <actor> can be used as a synonym for the now deprecated <process>
445 - state_file and avail_file periodicity is now easier to express
446 (check the documentation) the old behavior should still work.
447 * Storage: not backward compatible, but it seems that nobody ever used it
448 - Remove attribute 'content_type' of <storage_type>: was never used
449 - Make attribute 'model' of <storage_type> optional: for future usage
450 - Remove Bconnection model property: was never (in)validated. Replaced by
451 the maximum of the read and write bandwidth as a resource constraint.
454 - New and Backwards Compatibility break:
455 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
456 When one wants to get the list of tasks whose states have changed during a
457 simulation round, s/he has to allocate and free a dynar and use it as
458 argument to this function. The former SD_simulate (double how_long)
462 - Allow multicore VMs, with the correct sharing (unless you overcommit)
463 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
466 - New function to update the consumption of all hosts at once.
467 - Fix the model for multi-core hosts, linear in the amount of busy
468 cores with an abnormality for pIdle (see paper).
469 BUG: Still not satisfactory for ptask on mono-cores.
472 - The netzone are now available from the MSG API.
473 The old names still work, but are now deprecated.
476 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
477 - New option: smpi/keep-temps to not cleanup temp files
478 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
479 mallocs mode is used, allows to change the size of the fake file used
480 (default 1MB), to potentially limit the number of mappings for large runs.
481 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
482 - Fortran ifort and flang compilers support
483 - New RMA calls supported (experimental):
484 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
485 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
486 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
487 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
488 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
489 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
492 - New: this_actor::isMaestro() returns whether we are in kernel mode.
493 - Behind the scene, ActivityImpl are now automatically refcounted.
494 Removing this nasty bottleneck will greatly simplify our internals.
497 - Replay: New function xbt_replay_action_get():
498 Retrieve the function previously associated to an event type.
499 - DROPPED FUNCTION: xbt_str_varsubst()
500 - DROPPED MODULE: strbuff. We don't need it anymore.
501 - DROPPED MODULE: matrix. We don't need it anymore.
502 - DROPPED MODULE: lib. We don't need it anymore.
504 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
506 SimGrid (3.15) Released March 22. 2017
508 The Spring Release: continuous integration servers become green
509 We fixed even the transient bugs on all target architectures:
510 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
511 Windows, FreeBSD, NetBSD.
513 - Rename NetCards to NetPoints.
514 This was intended to help NS3 users, but that's not a netcard.
515 That's a point in the routing algorithm, let's avoid wrong simplifications.
518 - New: MSG_process_yield(). Stop and yield to other processes.
519 - New: MSG_process_daemon(). Daemon processes are automatically killed
520 when the last non-daemon process terminates
521 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
522 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
523 to make room for the upcoming network energy plugin.
524 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
527 - Ensure that an actor can kill itself with Process::exit()
528 - Kill the obscure NativeException. Nobody want to survive the issues
529 it denotes, so use JniException that is a RuntimeException (not to
530 be caught explicitly).
531 - Partial bug fix in initialization. SimGrid flags on command line were
532 consumed at C level but stayed in the original Java String[] args.
533 This could mess users' args[i] if SG flags were not put at the end of
535 The SimGrid flags are now removed from the Java arguments. However,
536 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
537 if args.length is greater than the number of YOUR OWN ARGUMENTS.
538 It might be if you have --log or --cfg flags in the command line.
539 - Fix numerous memleaks all around the place. In particular, around VMs.
543 - s4u::onPlatformCreated: right before the simulation starts
544 - s4u::onSimulationEnd: right after the main simulation loop
545 - s4u::onTimeAdvance: right after a clock change (time is discrete)
546 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
547 event from the availability_file changes the avail speed.
548 - Links are now usable from s4u
549 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
550 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
551 - New examples: The conversion of MSG examples to S4U has begun
552 - Actors: create, kill, migration, and suspend
553 - Applications: master-worker and token-ring
554 - Action replay: communications and storage
555 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
558 - Backwards Compatibility breaks
559 - The SD_route_get_list and SD_route_get_size functions have been removed.
560 They are replaced by the unique
561 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
562 The route from src to dst is built in the links dynar whose size can
563 be obtained with xbt_dynar_length.
564 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
565 sg_host_route_bandwidth and sg_host_route_latency.
566 Macros ensure the backwards compatibility, but you should fix your code
569 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
570 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
571 - Bump our claim of support from MPI 1.1 to MPI 2.2.
572 We don't support 100% of it, but it should be enough. Contact us if not.
573 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
574 naive and probably inaccurate)
575 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
576 mapped onto a unique small file using some system magic.
577 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
578 - Fortran: cleanups, fixes, support of user-added operations
579 - RMA: MPI_Accumulate are applied in correct order.
580 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
581 - Better support for MPI_IN_PLACE
582 - Support for MPI_Win attrs and keyvals.
583 - Support MPI_Comm_group_create, MPI_Type_size_x
584 - MPI ops should be properly applied to all allowed types
585 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
586 algorithm by a non failing one from mpich.
589 - Kill the fifo data container: we don't use it anymore.
592 - A <cluster> can now be created with different speed values to represent pstates.
593 see examples/platforms/energy_cluster.xml for an example.
595 SimGrid (3.14.159) Released December 28. 2016
597 The previous release was only a crude approximation of a Pi release;
598 this one should be a bit better.
600 - Revert a last minute change that broke on Mac OSX.
601 - Fix the changelog and NEWS files.
602 - Improve the documentation, in particular of the routing module.
603 - Integrate some patches flying around in Debian.
605 SimGrid (3.14) Released December 24. 2016
608 * The whole documentation was reworked and reorganized. There is
609 still much room for improvement here, but we're on it.
612 * We now need python3 for our testing infrastructure.
613 * Model checking needs libevent
616 * Now works on FreeBSD
619 * Privatization now works on FreeBSD
621 * Privatization is now activated by default in smpirun
622 You can turn it off with -no-privatize if something goes wrong.
624 * Call-location tracing for SMPI.
625 You can add the exact location (filename / linenumber) of an MPI call to
626 your trace files and slow down or speed up the simulation between two
627 consecutive calls by using an adjustment file (see the documentation).
629 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
630 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
631 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
632 values were ignored in some cases. The timings of these functions can now
633 be significantly different.
635 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
636 smpi/running-power is renamed to smpi/host-speed
638 * smpi/grow-injected-times option to enable or disable multiplication of the
639 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
640 make simulation less precise (but also much faster).
642 * smpirun script should be (much) faster for large deployments.
644 * SMPI tracing: fixed issue with poor matching of send/receives.
646 * Replay: Fix broken waitall
648 New functions and features
649 * MSG_parallel_task_execute_with_timeout, to timeout computations.
651 Dropped / renamed functions and features
652 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
653 - MSG_mailbox_is_empty() -> Mailbox::empty()
654 - MSG_mailbox_front() -> Mailbox::front()
655 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
656 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
657 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
658 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
659 * Don't pass the free_f parameter to property related functions:
660 - MSG_host_set_property_value()
661 - MSG_as_router_set_property_value()
662 - MSG_storage_set_property_value()
663 * VM properties. Since msg_vm_t are msg_host_t, just use
664 MSG_host_get_property_value() and friends
665 * VM I/O related things:
666 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
667 - Unimplemented save/restore methods
668 * MSG_as_router_get_property_value() was redundent with
669 MSG_environment_as_get_property_value().
670 - Removed MSG_as_router_*propert*() functions
671 - Added MSG_environment_as_set_property_value() for consistency
672 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
673 Well, they are still there for now, but deprecated with a warning.
674 Please switch to xbt_dict_new_homogeneous() before this is removed
676 * Task affinity. Its intended behavior (that was very badly tested
677 and probably not really working) was deceiving what most users
678 would have hoped here.
679 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
680 * Option network/coordinates is now useless and should be dropped.
683 * Added option storage/max_file_descriptors to allow more than 1024 files opened
686 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
687 * Added Dragonfly topology support
690 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
691 EiBps, EBps, Eibps, Ebps for bandwidth.
692 They may become useful to some lucky ones.
695 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
696 * ex/app_tokenring: new example, very similar to the MSG Token Ring
697 * ex/async_waitAll: new example, on asynchronous communications
700 * Memory usage should be decreased for simulations with a large number
701 of processes. This also helps for SMPI.
703 SimGrid (3.13) stable; urgency=low
705 The Half Release, a.k.a. the Zealous Easter Trim.
707 - We removed half of the lines, that were mostly experimental cruft.
708 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
709 (not counting blanks and comments -- according to openhub.net)
710 - The internals are now compiled in C++ (and will soon be clean C++)
711 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
712 59 klines of C, etc. We added only 29 klines of C++ in replacement.
714 * Backwards Compatibility breaks
715 - Removed Lua simulation bindings (switch to C or Java for that).
716 Lua can still be used to describe platforms
717 - Removed Java kernel plug-ins.
718 Will be reintroduced after the ongoing major internals reorg.
720 - the following functions were removed.
721 They were too specific and should be reimplemented in a generic
722 way, with filter function.
723 - MSG_task_listen_from_host
724 - MSG_mailbox_get_count_host_waiting_tasks
725 - MSG_mailbox_put_with_timeout was removed.
726 Please use MSG_task_send_with_timeout instead.
728 - the SD_application_reinit function was removed. It has been a noop for a while.
729 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
730 replaced by a more flexible mechanism.
731 - The following functions thus do not exist anymore
732 - SD_workstation_get_access_mode
733 - SD_workstation_set_access_mode
734 - SD_workstation_get_current_task
735 - Basic estimation functions have been removed but can easily be replaced
736 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
737 - SD_workstation_get_computation_time => amount / sg_host_speed()
739 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
740 Use VM.getSpeed()*load/100 for the legacy behavior.
742 - option enable_tracing was removed. It was not doing anything for a while.
743 - In the ModelChecker:
744 - the model-checker now ptraces the model-checked process which means
745 you cannot use a debugger on the latter anymore (we might make this
746 optional in the feature);
747 - removed soft-dirty page tracking;
748 - remove model-checked side snapshot management,
749 MC_snapshot() and MC_compare_snapshot();
750 - keep the MC_cut() function as a stub (it was not really working
751 in the previous release).
754 * All options are consistently in kebab-case. Old names are kept as alias.
757 * Switch to platform v4 format.
758 - Rename from 'power' to 'speed' the attributes describing the amount of
759 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
760 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
761 - In <host> and <link>, attributes availability and state are gone.
762 It was redundent with state and availability traces, and with peak values.
763 - In <cluster>, attributes availability_file and state_file are gone.
764 It was too complex and unused.
765 - Kill <gpu>. Was not doing anything.
766 - The DOCTYPE points to the right URL:
767 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
768 (the file at this address now documents the changelog since its v1)
769 - A warning is emitted for unit-less values (they are still accepted).
770 - speed. Default: 'f' or 'flops'. Also defined:
771 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
772 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
773 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
774 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
775 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
776 - latency. Default: 's' second. Also defined:
777 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
779 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
780 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
783 * s4u::Host is now the preferred public interface to the Host features.
784 sg_host_* functions are C bindings to the exact same behavior
785 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
788 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
789 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
790 They were deprecated since a few years, and probably did not even compile anymore.
793 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
794 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
795 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
796 encouraged to update their codes. A list of the modified functions can be found at the end of
797 include/simgrid/simdag.h
800 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
801 the lambda or closure passed as a parameter will run in kernel mode.
802 All the callback functions should be rewritten to that interface at some point.
805 * Reorganizing and cleaning the internals all around the place.
808 * Remove old default barrier/bcast buggy algorithms (see #18407)
809 * Various bug fixes to handle more codes
810 * Remove the need for the --foreground option of smpirun (it is still
811 accepted for backward compatibility).
814 * Kill the setset data container: MC don't use it anymore.
815 * Kill the queue data container: it made more sense with GRAS.
816 * Kill the xbt_peer_t data type: it's useless without GRAS.
817 * Kill rm_cb feature of config sets: it was never useful.
818 * Kill graphxml parsing feature. It was not used.
819 * Kill the deprecated code protected by XBT_USE_DEPRECATED
821 - xbt_dynar_sort_strings(), when the content is char*
822 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
823 They throw exceptions on invalid input;
824 * C++ support for declaring CLI flags (simgrid::config::Flag);
825 * class for abstracting different signal backends (simgrid::xbt::signal).
826 with no external dependencies (we need very simple signals).
829 * refactoring and cleanup of the code;
830 * ongoing process to cleanly separate the model-checking algorithms
831 from the code model-checking support.
833 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
835 SimGrid (3.12) stable; urgency=low
837 The Facelift Release.
840 * Require g++ v4.7 at least to not speak prehistorical C++.
841 * Require Boost 1.48 (for signal2 component).
842 * Java must be version 7 at least when activated.
843 * Builds on Windows again (including Java bindings).
844 * Tracing is now always enabled (no way to turn it out)
845 * Remove GTNetS. It was not working anyway.
846 * Various cleanups in the cmake scripts.
847 * Move headers around to sort them out on installed systems:
848 - instr/instr.h -> simgrid/instr.h
849 - instr/jedule/* -> simgrid/jedule
850 - simdag/datatypes.h was removed
851 - simdag/simdag.h -> simgrid/simdag.h
852 - msg/datatypes.h was removed
853 - msg/msg.h -> simgrid/msg.h
856 * Interface improvement:
857 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
858 - Sanitize the interface in MSG_task_ module:
859 - Merge two functions that were close enough but misleading:
860 set_compute_duration(t) -> set_flops_amount(t)
861 get_remaining_computation(t) -> get_flops_amount(t)
862 - set_data_size(t) -> set_bytes_amount(t)
863 get_data_size(t) -> get_bytes_amount(t)
864 - Massive cleanups in the functions related to the energy
865 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
866 - New: MSG_host_get_pstate()
868 - msg/energy/onoff: switching hosts on and off
870 * Interface improvement:
871 - Rename Host.isAvail() to Host.isOn()
872 - Rename Process.currentProcess() to Process.getCurrentProcess()
873 - Rename Task.setDataSize() to Task.setBytesAmount()
874 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
876 - #18874: Actually allows the GC to reclaim tasks
880 - SIMIX_process_throw: raises an exception in a remote process
881 * Refactoring: Separate sub-modules
882 - libsmx: the public interface, as libc in a real system
883 - popping: the strange dance that converts a user request into a kernel handling
884 - smx_context_*: the virtualization mechanisms that embed the user code
885 - smx_*: the handling of each simcalls
887 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
888 * Rename smx_action_t into smx_synchro_t, making explicit that these
889 things are used to synchronize processes with their environment.
890 For example, a communication is a sort of synchronization involving
891 the communicating processes (that may block until the exchange) and
892 the platform. The same can be said from computations, etc.
894 - #18888: segfault when a process ends before its kill_time
898 - 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.
899 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
900 - MPI_Type_set_name, MPI_Type_get_name
901 - MPI_*_c2f and MPI_*_f2c functions
902 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
903 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
904 - Activate a lot of new tests from the mpich 3 testsuite
906 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
907 - InfiniBand network model added: Based on the works of Jerome Vienne
908 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
909 - When smpi/display_timing is set, also display global simulation time and application times
910 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
911 * Collective communications
912 - 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
913 - 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/).
914 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
916 - Replay now uses algorithms from wanted collective selector
917 - Replay can be used with SMP-aware algorithms
918 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
919 - 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.
921 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
922 - Correctly use loopback on fat-tree clusters
923 - Asynchronous small messages shouldn't trigger deadlocks anymore
924 * Energy/DVFS cleanup and improvement
925 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
926 - new: smpi_get_host_pstate()
929 - "Full" network optimization flag was broken since Surf++
930 - Better handling of precision flags in maxmin
931 - Fix bug causing sometimes "Impossible" errors
932 - Properly pass cluster properties to included hosts
933 * Improvement of the Energy plugin.
934 - Always update the consumption before returning that value
935 - New property: watt_off to denote the dissipation when the host is off
936 - New functions getWattMinAt and getWattMaxAt to retrieve the
937 dissipation of pstates that we are not currently at.
938 * Java: class NetworkLink renamed to Link
939 * New function: simcall_process_get_kill_time()
940 * Massive rename s/workstation/host/
941 - That's intrusive, but that's good for the project consistency. Sorry.
942 - Change config option "workstation/model" into "host/model"
945 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
946 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
947 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
948 - xbt_dirname and xbt_basename for non-POSIX systems
950 * The model checker now runs as a separate process.
951 * The model checker runs must now be launched with the new simgrid-mc program.
952 * Record/Replay: the MC can display a textual representation of a path in the
953 execution graph. It can then be replayed outside of the model checker.
955 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
957 SimGrid (3.11) stable; urgency=low
962 * Normalizing pointers addresses tool for better diff between logs
965 * Add cloud examples using new VMs
966 - examples/msg/cloud/two_tasks_vm.tesh
967 - examples/msg/cloud/simple_vm.tesh
968 - examples/java/cloud/cloud.tesh
969 - examples/java/cloud/migration/migration.tesh
970 * Add java surf examples:
971 - examples/java/surfPlugin/surf_plugin.tesh
972 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
973 - examples/java/surfCpuModel/surf_cpu_model.tesh
974 * Add SMPI+MSG example:
975 - examples/smpi/smpi_msg_masterslave/
980 - msg task destroy cancel test
981 - msg_host on/off test
982 * Move all tests in testsuite to teshsuite (adding tesh files)
983 * Restructure teshsuites
984 - one folder for each kind of test
985 * Restructure AddTests.cmake
987 - structure the order of tests (with sections)
990 * Add virtual machine
991 - creation of a VM on a PM
992 - migration of a VM from a PM to another PM
994 - MSG_process_join(msg_process_t process, double timeout)
995 - msg_bar_t MSG_barrier_init(unsigned int count)
996 - int MSG_barrier_wait(msg_bar_t barrier)
997 - void MSG_barrier_destroy(msg_bar_t barrier)
998 - msg_as_t MSG_environment_as_get_by_name(const char * name)
999 * New option "msg/debug_multiple_use" to help debugging when a task is used
1002 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1004 - Increase code coverage in test suites
1008 * Protect context stacks against stack overflow. The number of protected memory
1009 pages allocated on the top of each stack (1 by default) can be configured
1010 with the new command line option --cfg=contexts/guard_size:<value>.
1011 * Simcalls are now generated by a python script that
1012 - generates files included by SimGrid
1013 - checks that all the functions exist, and proposes prototypes
1015 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1016 host_set_data, host_get_data
1018 - simcall_process_join(smx_process_t process, double timeout)
1019 * Fix bug where sleeping processing could not be suspended.
1022 * Translate surf models from C to C++
1023 - Generic classes for all models: Model, Resource, Action
1024 - A generic interface for each kind of model (CPU, Network, Storage
1025 Workstation, WorkstationVM)
1027 * Translate surf routings from C to C++
1028 * Add callbacks using sigc++ or boost::signals2
1029 - Add callback functions for resource creation/destruction
1030 - Add callback functions for action state change
1031 - Handle Energy as a plugin
1032 * Replace swag by boost::intrusive
1033 * Add new routing models for clusters. For documentation, see
1034 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1035 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1036 parameters for cluster tag
1037 - Fat trees, with topology="FAT_TREE" and
1038 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1039 - see examples/platforms/torus_cluster.xml and
1040 examples/platforms/fat_tree_cluster.xml
1041 * More documentation
1044 * Hostfiles support host:nb_processes construct to deploy several processes on
1046 * Collective communication algorithms should not crash if used with
1047 improper number of nodes and report the error.
1048 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1049 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1050 MPI_Dims_create, MPI_Cart_sub are supported.
1051 * New interface to use SMPI programmatically (still depends on MSG for
1052 some parts, see examples/smpi/smpi_msg_masterslave):
1053 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1057 * Global variables privatization in MPI executables is now performed at runtime
1058 with the option smpi/privatize_global_variables (default:no).
1059 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1060 dynamic libraries loaded by the application are not privatized (static
1061 linking with these libraries is advised in this case)
1064 * Options defined in XML work correctly now.
1067 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1068 copied into simgrid.jar or not (ON by default). Use this option if you want
1069 to reduce the size of the installed simgrid.jar, *and* the native libraries
1070 are kept installed elsewhere.
1071 * Surf binding with SWIG (code generated in maintainer mode only):
1072 - plugin to handle callbacks
1073 - CPU model only for the moment
1076 * Supernovae build mode is definitively removed. It was used to improve
1077 inlining and inter-module optimizations. It is nowadays superseded by
1078 link-time optimizations commonly available in compilers.
1079 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1080 * Add boost dependency for surf++
1081 * Add new macro for tests
1082 - ADD_TESH(name <tesh_args>)
1083 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1087 - xbt_bar_t XBT_barrier_init(unsigned int count)
1088 - int XBT_barrier_wait(xbt_bar_t barrier)
1089 - void XBT_barrier_destroy(xbt_bar_t barrier)
1090 * Make the xbt_os_time module public
1092 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1094 SimGrid (3.10) stable; urgency=low
1096 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1099 * Reintegrate Java to the main archive as desynchronizing these
1100 package is not acceptable anymore (Java is now considered stable)
1101 * Add explicit synchronization facilities through semaphores
1102 * Bug fix: Task.setDataSize() only changed the C world, not the value
1103 cached in the Java world
1106 * Dramatically change the way files are handled. API and internals changed, but
1107 this part of MSG was not considered as production grade either.
1108 * Add explicit synchronization facilities through semaphores
1109 * Add a new function MSG_host_get_process_list()
1110 * Preliminary DVFS support (see examples/msg/energy/ for details)
1113 * SMPI is now included directly in the libsimgrid as the windows
1114 linker doesn't force us on splitting it anymore.
1115 * Improvements of the SMPI replay tool:
1116 - Most of the collective communications are now rooted in the same process as
1117 in the original application.
1118 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1119 used until now). Multiple data types can now be used in a trace.
1120 - The replay tool now supports traces produce either by TAU or a modified
1122 - Bug Fix: the compute part of the reduce action is now taken into account.
1123 - Gatherv collective is now supported
1124 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1125 of smpirun outputs time independent traces for the current run. One file
1126 is created per process. If too many processes are simulated, this behavior
1127 can be changed to one file for all processes by using the
1128 tracing/smpi/format/ti_one_file flag
1129 * smpirun generates the host file if needed (with given host count and platform)
1130 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1131 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1132 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1133 --cfg=smpi/coll_selector:(mpich/ompi)
1134 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1136 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1137 happening inside a collective SMPI call.
1138 * Fix the behavior of complex data types handling.
1139 * Make MPI_Wtime another synchronization point to take computations into
1141 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1142 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1143 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1144 (no privatization of global variables yet)
1145 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1148 * Verification of liveness properties is now available for SMPI applications
1149 (in addition to MSG applications)
1150 * Bugged examples using SMPI in examples/smpi/mc/
1151 * Add --cfg=model-check/visited option. Allows the verification of infinite
1152 programs. Detection of loops in the execution thanks to the system state
1153 comparison and reduction of the state space to explore. Can be combined with
1154 DPOR for safety properties.
1157 * Allow to change SimGrid configuration (see --help) within the code
1158 thanks to SD_config() as it can be done in MSG.
1159 * Add a new function SD_task_set_amount() upon user request.
1162 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1163 * Remove rule based routing (no more PCRE dependency)
1164 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1165 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1166 * Add a loopback_bw and loopback_lat options to cluster tag.
1167 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1168 routing. Now peers are encapsulated in an AS and have their own private
1169 router but this is transparent.
1172 * Our own implementation of getline is renamed xbt_getline, and gets
1173 used even if the OS provide a getline(). This should reduce the
1174 configuration complexity by using the same code on all platforms.
1175 * New type: xbt_cfg_elm_boolean.
1176 * Allow to use yes/no for boolean configuration options in the command line.
1177 * Allow to disable SimGrid cleanups at exit from command line option.
1178 There are situations where one may want a simulation to end with an exit.
1179 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1180 annoying when scripting around the simulator. Adding a
1181 --cfg=clean_atexit:no allows to circumvent this issue.
1184 * Lots of memory leaks were corrected in this release.
1185 There is no known memory leaks anymore, in all of our 600+ tests.
1186 * New command line option --version, to get SimGrid version information.
1187 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1189 * Supernovae builds are deprecated, and expected to be removed in the next
1192 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1194 SimGrid (3.9) stable; urgency=low
1196 The Grasgory release: GRAS is really dead now.
1198 * Complete overhaul of the internal host structures scheme.
1201 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1202 was considered as experimental and badly maintained since then.
1203 * Keeping it was thus a trap to our potential users, that could take
1204 it instead of MSG or SMPI by mistake despite is pity state.
1205 * GRAS seems to have very few users (if any), and no one volunteered
1206 to maintain it further. It also induces a lot of XBT code (for
1207 portability sake), that must be maintained too.
1208 * For all these reasons, we killed GRAS. If someone wants to revive it
1209 in the future, don't cry, our git history still remembers of GRAS.
1212 * Major overhaul. Merge our documentation again as time proved that
1213 splitting it was really not helping our users.
1214 * Further improve the developer documentation to help newcomers
1215 hacking on SimGrid itself. The user documentation (and in
1216 particular, the beginner documentation) is still in a sorry state.
1219 * Now works on Windows too!
1220 * Much more extensive test suite, from MPICH
1223 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1224 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1225 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1226 * Bug fix: let task be scheduled when the last dependency to be solved is
1227 a control dependency.
1228 * Remove SD_load_environment_script function.
1229 Use the C sg_platf function if you want to declare a platform
1233 * New function: MSG_process_get_number()
1234 * Old function documented: MSG_config()
1235 * Remove MSG_load_platform_script function
1236 Use the C sg_platf function if you want to declare a platform
1240 * Change the default value of the TCP_gamma constant (maximal size of TCP
1241 congestion window) to a more realistic 4MiB value. If you notice changes in
1242 your simulation results, you can fall back to the previous 20k tiny window
1243 by adding --cfg=network/TCP_gamma:20000 on command line.
1244 * (Hopefully) fix a bug wrt periodic availability and state traces
1245 * Bug fix: use default values at start when first event in availability/state
1246 trace is not at time 0.
1249 * remove the "new_" part of function name sg_platf_new_trace_connect
1250 (resulting in sg_platf_trace_connect), since it does not create
1254 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1255 use case where it's really mandatory, and maintaining it was a pain in
1257 * New: xbt_fifo_search(), search an item with a user-provided
1258 comparison function instead of dumb pointer comparison.
1261 * Fix the lua deployment:
1262 Use `simgrid.init_application()` before deployment instead of
1263 `simgrid.msg_register_application()` after.
1266 * Transfer the tracing files into the corresponding modules.
1268 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1270 SimGrid (3.8.1) stable; urgency=low
1272 The "we are told that some people want to also *install* the simgrid
1275 * Add missing file "tesh.1" to the archive.
1277 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1279 SimGrid (3.8) stable; urgency=low
1281 The Psssshiiiit release: SimGrid jumps into the Cloud.
1284 * Add an experimental interface to manipulate VMs. They are mainly
1285 process groups with very few intrinsic semantic, but they should
1286 allow you to build the semantic you want easily.
1287 * New function: MSG_host_set_property_value()
1288 * New function: MSG_process_on_exit(). To clean memory in all cases.
1289 * Bug fixes that made the host (and link) failures unusable.
1290 * Add a way to auto-restart process when the host in which they are
1291 executing comes back (ON_FAILURE="RESTART" on deployment file,
1292 MSG_process_auto_restart_set).
1293 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1294 please stop using the old ones, they are DEPRECATED.
1296 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1297 Please use MSG_init() instead. (reducing the amount of entry
1298 points in the library helps us).
1299 * Make it impossible to link against the wrong version of the lib
1300 * Deprecate MSG_clean(). No need to call it anymore.
1301 * Function MSG_get_host_number() is not deprecated anymore.
1304 * Split the doc into a user guide and a reference guide.
1305 * Start a developper guide to help people hacking on SimGrid.
1308 * Enable tracing by default. This modules rocks you should use it.
1309 * Remove option custom_flags. Now use environment variables CFLAGS
1311 * Use default cmake things to detect lua instead of home grown ones.
1312 * New option "enable_mallocators" to disable mallocators, for debugging
1313 purpose ("on" by default).
1316 * Bug fixes around the resource failures: don't let the processes
1317 survive the host they are running onto.
1318 * Add an interface to auto-restart processes when the host in which they are
1319 executing comes back.
1320 * Ensures that SIMIX_clean is called automatically. It's not part of
1321 the public interface anymore (bindings should be updated).
1324 * Bug fix for when SD_Simulate is called with a positive value: be careful
1325 when comparing doubles. Sometimes they are different for non significant
1327 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1328 parallel task whose initial work is distributed among host according
1329 to the Amdahl's law. Such tasks are created with a parameter alpha
1330 that corresponds to the non-parallelizable part of the computation.
1331 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1332 between two sets of workstations assuming a 1D block distribution (each
1333 workstation owns a similar share of data) on both sides.
1335 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1336 Data redistribution will be automatically scheduled once parent and child
1337 are both scheduled. The filling of computation_amount and
1338 communication_amount structures is now done seamlessly thanks to the chosen
1340 * New function SD_workstation_dump to display various information
1341 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1342 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1343 bandwidth on the route onto which the task is scheduled and the amount of
1345 To divide the nominal bandwidth by 2, the rate then has to be:
1346 rate = bandwidth/(2*amount)
1347 * Compute tasks that have failed can now be rescheduled and executed again
1348 (from their beginning)
1349 * Increasing source code coverage (src/simdag is now covered at 95.8%
1353 * Re-implement time-independent trace replay using SMPI (at the
1354 smpi_smp_* level) instead of MSG. This should replace
1355 examples/msg/actions/actions.c
1356 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1357 hindexed and structs)
1358 * Implement the exchange of non-contiguous data.
1359 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1360 * Correct behavior of smpi/sender_gap and set its default value to 0
1361 * Add option to asynchronously send small messages to allow better
1362 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1363 specifies the size in bytes under which messages will be asynchronously sent.
1364 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1365 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1368 * New C interface to define a platform: XML is now optional.
1369 For more info, please check include/simgrid/platf.h.
1370 * New interface to define random platforms from the C:
1371 For more info, please check include/simgrid/platf_generator.h and
1372 examples/msg/masterslave/masterslave_platfgen.c
1373 * Export a sg_cmdline dynar containing all the arguments we got from
1377 * Two new tracing options for adding comments to trace file so you
1378 can track your experiments (see --help-tracing for details).
1379 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1380 * Adding the SimGrid version that generated the trace file as a comment.
1381 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1382 * Fix to avoid key clashes on Paje links
1383 * Other minor fixes related to the Paje specification
1386 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1387 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1388 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1389 * Greatly improve the robustness of mmalloc to user errors (such as
1390 using an area after freeing it, or freeing it twice)
1392 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1394 SimGrid-java (3.8.1) stable; urgency=low
1396 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1397 Warning: all methods are visible, but only some of them are
1398 implemented so far. Check the source (src/jtrace.c)
1399 for further information.
1400 * New module: org.simgrid.msg.File (SimGrid File management functions)
1401 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1402 * Change the meaning of Process.restart: now restart the process from
1403 the begining, like MSG_process_restart in C.
1404 * Add Process.setAutoRestart: handling of process restart when failed
1406 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1407 you to retrieve the properties of the processes/hosts
1408 * Deprecate Msg.clean(): you can just forget about it now.
1409 * New function Process.getCount(), that only works when compiling
1410 with the not yet released version 3.9 of the C library.
1412 * New context factory based on Coroutines. It mandates a modified JVM
1413 but then, the simulations run about five times faster, and there is
1414 no limit to the amount of processes (beside of the available memory).
1416 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1418 SimGrid (3.7.1) stable; urgency=low
1421 * Restore the prototype of MSG_process_create_with_environment() to
1422 the pre-3.7 situation by removing the kill_time argument.
1423 * Add a MSG_process_set_kill_time() function instead.
1426 * Fix weird behaviors when dealing with parallel tasks.
1429 * Simgrid is now built as a dll.
1430 * Simgrid-java now works on Windows.
1431 * Simgrid-Java is now included into Windows package.
1434 * First pre-build package for MacOSX.
1437 * Fix compilation when using MSG_USE_DEPRECATED.
1438 * Fix some compilation issues on Macs and Windows.
1439 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1440 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1442 We discovered that the Lua console is broken, but we are missing the
1443 manpower to fix it right now. The problem existed in 3.7 too, so we
1444 are not blocking the release for that. Sorry if you depended on this
1445 feature, any help would be really welcome.
1447 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1450 SimGrid-java (3.7.1) stable; urgency=low
1452 The "Java aint got to be bloated and slow" release
1455 * Various internal cleanups and performance improvement
1456 Simulations are expected to run up to twice faster or so
1457 * Make Process.kill(process) an instance method, not a static one
1458 * User processes are not java.lang.Thread subclasses.
1459 This breaks the compatibility (sorry), but previous API was
1460 brain-dead, making it impossible to have non-trivial
1461 initializations in the process constructor.
1462 * Require a full constructor per Process sub-class.
1463 Kinda breaks the compatibility (sorry), but this allows a much more
1464 efficient way to launch the processes at simulation startup.
1465 * Do not embeed our version of semaphores, java 1.5 can be considered
1466 as sufficiently prevalent for us to not dupplicate its features.
1470 * Add examples for almost every part of the API
1471 We spotted and fixed a lot of bugs in the process
1473 * New module: asynchronous communication API
1474 * New function: Process.sleep()
1475 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1476 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1477 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1479 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1481 SimGrid (3.7) stable; urgency=low
1483 The "spring cleanups (before the next Big Project kicks in)" release.
1486 * We can specify the SMPI latency/bandwidth factor with command line
1487 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1488 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1489 You can also use the "config tag" from platform file by adding this line
1490 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1491 (see "example/platforms/tag_config.xml" to use "config tag").
1492 Note that the command line supersedes the platform file configuration.
1493 * Change the correction factors used in LMM model, according to
1494 the latest experiments described in INRIA RR-7821.
1495 Accuracy should be improved this way.
1496 * Use the partial invalidation optimization by default for the
1497 network too. Should produce the exact same results, only faster.
1498 * Major cleanup in surf to merge models and split some optimization
1499 mechanisms from the core of the models. As a result you can now
1500 specify which model to use (e.g., --cfg=network/model:LV08
1501 --cfg=cpu/model:Cas01) and which optimization mode to use
1502 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1503 Incompatible combinations should err at initialization. See
1504 --help-models for the list of all models and optimization modes.
1505 * The CLM03 workstation model was dropped for simplicity because it
1506 used the deprecated CM02 network model. Use default instead.
1507 * Rename the TCP_gamma configuration option to network/TCP_gamma
1508 * Rename the coordinates configuration option to
1509 network/coordinates, and document it
1510 * Use now crosstraffic keyword instead of the terribly misleading
1511 fullduplex keyword. It is activated by default now in the current
1512 default model, use --cfg=network/crosstraffic:0 to turn it off.
1513 * Ongoing refactoring the model parsing to make XML files optional
1514 See include/simgrid/platf.h for details (still to be completed)
1517 * Major overhaul of the documentation. Almost instructive now :/
1518 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1519 functions and friends. This interface was considered as
1520 deprecated since over 2 years, it's time to inform our users that it is.
1521 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1522 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1523 use these (crappy) functions in your code.
1524 These functions will be removed soon. Stop using them now.
1525 * Deprecate MSG_get_host_{table,number}
1526 Implement MSG_hosts_as_dynar() instead.
1527 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1528 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1531 * Stabilize the parallel execution mode of user contexts
1532 * Introduce configuration variables to control parallel execution:
1533 - contexts/synchro: Synchronization mode to use when running
1534 contexts in parallel (either futex, posix or busy_wait)
1535 - contexts/parallel_threshold: Minimal number of user contexts
1536 that must be part of a scheduling round to switch to parallel
1537 execution mode (raw contexts only)
1538 * Fix bugs that prevented to use suspend/resume along with
1539 synchronization structures.
1540 * Fix bugs in process termination that lead to invalid memory access
1541 in very specific conditions.
1544 * Introduce a parallel mode for the models (controlled by surf/nthreads
1545 configuration item). In our tests, running the models in parallel
1546 never lead to any speedups because they are so fast that the gain
1547 of computing each model in parallel does not amortizes the
1548 synchronization costs, even when ultra fast futexes are used.
1549 This is released anyway because YMMV.
1552 * Performance boost by using a swag internally to compute the set of
1553 tasks that are finished and should constitute the return value of
1557 * Enable it by default now that it is considered rather stable.
1560 * Documentation of the tracing functions.
1561 * Performance gains when tracing categorized/uncategorized resource
1562 utilization by avoiding calls to get route when updating resource
1563 variables. LMM constraints are being used instead.
1564 * API changed to set task categories. Use MSG_task_set_category instead
1565 of TRACE_msg_set_task_category, and SD_task_set_category instead
1566 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1567 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1568 * Examples for link user variables added (see at examples/msg/tracing/)
1569 * Deprecated function TRACE_msg_set_process_category completely removed
1570 * Trace header updated according to the latest Paje file format
1571 * Tracing network lazy updates, no longer obligate users to use full updates
1572 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1573 * Experimental: let user code declare/set/push/pop application states for hosts
1574 * API changed to allow the manual creation of graph configuration files
1575 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1578 * Improve the API of Lua MSG bindings, using the Lua spirit.
1579 * Each simulated process now lives in its own Lua world (globals are
1580 automatically duplicated). It helps writing simulators. Will allow
1581 to run Splay programs within SimGrid in the future.
1582 * Add a Chord example in Lua, equivalent to the MSG one.
1585 * Start the implementation of a solution to express temporal
1586 properties, not only local assertions. This is still an
1587 experimental work in progress, stay clear from it to be safe.
1591 - Add new runtime parameters --help-logs and --help-log-categories
1592 to display information about supported logging parameters and
1594 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1595 don't exists anymore.
1596 * Mallocators: allow value NULL for the reset function.
1598 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1599 create homogeneous dictionaries, where all the elements share the
1600 same free function. Non homogeneous dictionaries will be
1601 deprecated in the next release.
1602 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1603 - Multi-level dictionaries are deprecated.
1605 - new function xbt_dynar_search_or_negative() that is useful when
1606 you have less than 2 million elements in your dynar and don't
1607 want of the extra complexity of catching exceptions when the
1608 element is not found.
1610 - Make xbt_os_thread module (for thread portability) public.
1611 Documentation is still to come, sorry.
1613 - Cleanups and simplifications to make it maintainable again.
1614 - Exotic features (such as memalign and valloc) were removed.
1615 - The metadata were extended and improved so that the
1616 model-checker becomes able to explore and inspect the heaps.
1617 - This may induce a performance drop when enable_model-checking is
1618 ON in cmake (even if it's not used in the simulation), but it is
1619 necessary at this point to get MC working.
1621 Turn model-checking OFF if simulation performance matters to you.
1622 Not enabling it at runtime is not enough, disable it in cmake.
1624 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1627 SimGrid (3.6.2) stable; urgency=low
1629 The "Not coding new stuff allows to polish old things" release.
1632 * New bindings to the NS3 packet level simulator (experimental)
1633 * Use the raw (efficient) execution contexts instead of the sysv
1634 (portable) ones when possible.
1635 * libpcre is now mandatory in any cases since not using it led to
1636 severe performance loss and possibly other issues
1637 * Update the XML platforms:
1638 - G5K: include the latest machine in Nancy
1639 - GridPP and LCG: new platforms
1640 * Documentation was partially updated, at least (more to come)
1642 Bug fixes, cosmetics and small improvements
1643 * Free terminated processes before the end of the simulation to avoid
1644 exhausting the memory of users having very dynamic amount of
1646 * Bug fix and cosmetics about canceling non-running tasks
1647 * Bug fix about the dot loader's issues when using libcgraph
1650 * Create an installer for windows with nsis (amd64 and win32)
1651 - Add an hello world project to illustrate simgrid project creation.
1652 - Embed libpcre into the Simgrid installer to avoid
1653 its compilation burden
1654 * The raw execution contexts should work on Apple now
1655 * Port to Windows 64 bits
1656 - Sysv contexts now have an implementation for this arch
1657 - GRAS communication features now support this arch
1658 * Drop support for borland compiler on windows
1659 - this code was not maintained, and we kinda depend on gcc nowadays
1660 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1661 * Fix portability issue on unstable ubuntu: linker became picky on
1664 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1666 SimGrid (3.6.1) stable; urgency=low
1668 The "Oops, we broke Macs too" release
1671 * Fixed contexts detection so that raw ones are used when possible
1672 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1673 leads to a strange error, with user code segfaulting sometimes when
1674 the generated code is not perfectly aligned (which is not
1675 controllable from the user side, depends on the amount of code)
1678 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1680 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1682 SimGrid-java (3.6) unstable; urgency=low
1685 * Split of every thing from simgrid v3.5 into a separate package.
1687 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1689 SimGrid (3.6) stable; urgency=medium
1691 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1694 * Bindings now constitute their own package, separated from the main one.
1695 Rationale: reduce our maintenance nightmare by reducing the module coupling
1696 They will soon be released on their own on gforge.
1698 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1699 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1701 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1702 * It's not quite deprecated for now because we have no replacement,
1703 but it may soon become the case.
1706 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1707 * Fortran: New user-level cache variable to store the rank of the running
1708 process. This improves performance by an order of magnitude.
1709 * C: New coccinelle script to automatically locate and modify global and
1710 local static variables.
1711 * Improved SMPI network model with a sender-side gap to account for multiple
1715 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1716 only say if a communication is finished, no matter whether it succeeded or
1717 failed. You can call MSG_comm_get_status() to know the status of a finished
1719 * New function MSG_task_dsend() to send a task and detach it. When a
1720 communication is detached, you are never notified of its success or failure
1721 and the memory is released automatically once it is finished. This function
1722 is useful when you don't care about the end nor the success of a
1724 * Change the prototypes of action replay. Sorry for inconvenience,
1725 but this is really more efficient this way (and to adapt your code,
1726 you just have to fix the initialization, that shouldn't be too long)
1727 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1728 ever managed to do anything useful with it.
1729 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1730 takes a dynar of communications. It returns immediately and gives the
1731 index of a finished communication (if any).
1732 * New example: a basic implementation of the Chord P2P algorithm.
1735 * New model for multi-core CPUs. You can now use the core attribute to
1736 precise the number of cores of a host. This is a basic model. Every
1737 process running on the host receives at most the power provided in
1738 the DTD (throughput<=power). Total throughput of process cannot exceed
1740 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1741 router linked by an up-link and a down-link (possibly asymmetrical).
1742 This kind of pattern allows to easily build last-mile model style platforms.
1743 Aggregating such patterns in a rule-based AS is thus the technique of
1744 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1745 * New model for Vivaldi routing. We transformed the Vivaldi network model
1746 into a Vivaldi routing model (based on the rule-based model). This allows to
1747 combine Vivaldi based latencies with last-mile platforms.
1750 * Added a check for NaN of IEEE754 infinite in the double entries of
1752 * Introduce a new context factory "raw", highly inspirated from the
1753 ucontext factory, but using manually crafted functions in assembly to
1754 do the work in an efficient manner.
1755 * Allow to change the used context factory at run time, not only at
1756 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1757 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1758 contexts at runtime (only with raw contexts or ucontexts).
1759 * Completely rewrote this module to allow parallel execution of user
1760 processes. Use --cfg=contexts/nthreads:N to execute user processes
1761 with N parallel threads (the default is 1, meaning no parallelism).
1762 * Allow to decide dynamically between sequential and parallel modes.
1763 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1764 processes in parallel only when their number is greater than or equal to P
1766 * Added a check for NaN of IEEE754 infinite in the double entries of
1770 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1771 won't output the state of processes when interrupted with Ctrl-C
1772 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1773 NULL-terminated array. This may solve backward compatibility issues
1774 due to the change to return type of SD_simulate. See also:
1775 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1776 * Add new macros with variable number of arguments.
1777 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1778 - in xbt/asserts.h: xbt_assert
1779 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1780 - in xbt/ex.h: THROWF and RETHROWF.
1781 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1783 * Change xbt_die() to accept a format string with arguments, just like printf.
1784 * New data structure: xbt_lib_t, like a dict but more general and with better
1788 * New configuration options
1789 Options triva/categorized and triva/uncategorized can be used to generate
1790 graph configuration files for Triva visualization tool.
1791 * Configuration option tracing/platform is renamed to tracing/categorized
1792 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1793 * New TRACE_user_link_variable interface:
1794 User provides the name of the link and the tracing variable to attach to it
1795 * the declaration of tracing categories must be done after the environment
1797 * simpler tracing interface, just one way to declare categories
1798 TRACE_category or TRACE_category_with_color, it is up to you
1799 * links in the trace file are again identified by their names
1800 * trace contains the full platform hierarchy exactly as declared using the ASes
1801 * Options tracing/msg/[task|process]:1 groups the process by hosts
1802 for both cases, tasks and processes must have names that are unique during the simulation
1803 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1804 * The experimental option tracing/msg/volume is deprecated
1805 its functionality may be reincorporated if needed
1807 The tracing generates a trace file with unordered timestamped events,
1808 because of the way the core simulator (surf) works. A script available
1809 at the tools directory (fix-paje-trace.sh) can be used to put the events
1810 in order. We have changed the tracing so it can generate ordered timestamped
1811 events in the final trace, but depending on the simulator (and how much time
1812 is simulated) that can lead to a huge memory utilization. It is deactivated
1813 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1815 Build Infrastructure
1816 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1817 - We are trying hard to keep the API stable, but it may happen that
1818 some things change (we're a research project after all, not a
1819 nuclear plant operating system). If such things should happen, you
1820 could rely on that macro to adapt.
1821 - current value: 30600 for 3.06.00, aka 3.6
1822 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1823 a number that can be compared with SIMGRID_VERSION.
1824 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1825 assertions and verbose/debug logging events are disabled at compile time.
1827 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1829 SimGrid (3.5) stable; urgency=medium
1832 * New feature to any SimGrid-based simulator: Model-Checking
1833 Check SIN#1 for more details.
1836 * New Model SMPI (three-interval linear regression for correction factors)
1837 See RR-7426, available at http://hal.inria.fr/inria-00527150
1838 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1839 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1840 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1841 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1842 * See also src/smpi/README
1847 - Tracing API changes: TRACE_start and TRACE_end should not be called
1848 by user-code. They are automatically called by simulators created
1849 with SimDAG, MSG and SMPI if the toolkit is compiled with
1850 tracing_enabled=ON. Categories declaration and utilization remain the
1851 same for MSG and SimDag.
1852 - A function was added to the tracing API to declare categories with
1854 - TRACE_category_with_color (char *category, char *color)
1855 where color must be in the following format
1856 "%f %f %f", red, green, blue
1857 and red, green, blue are float values in the interval [0, 1]
1858 - User can specify NULL as color parameter, or continue calling
1859 TRACE_category (cat)
1860 On that case, the tracing system will define random colors
1861 - The following command-line options are supported:
1862 --cfg=tracing/filename:msg.trace
1863 --cfg=tracing:1 (activate tracing, needed to use others)
1864 --cfg=tracing/platform:1 (categorized resource use)
1865 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1866 --cfg=tracing/msg/task:1 (task creation)
1867 --cfg=tracing/msg/process:1 (process creation, migration)
1868 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1869 --cfg=tracing/smpi:1 (SMPI interface tracing)
1870 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1871 - examples of examples/msg/tracing updated
1873 - DAXLoader and DOTLoader functions can generate tasks with categories
1874 - A new function to attribute a category to SD tasks:
1875 TRACE_sd_set_task_category (SD_task_t task, char *category)
1876 * Tracing the MPI interface implemented by SMPI
1877 - Collective operations are traced with states
1878 - Point-to-Point operations are traced with states/links
1879 - Tracing activated by a parameter "-trace filename" passed
1880 to smpirun during execution (considering that simgrid
1881 is compiled with tracing enabled)
1882 - To run the simulation with gdb, the simulator
1883 accepts --cfg=tracing/smpi:1 to trace SMPI
1884 - tesh files to check if smpi tracing is ok
1885 - See examples/smpi/NAS/DT-trace
1886 * GTNetS tracing re-worked
1887 - adaptation to the tracing system of GTNets to cope
1888 with modifications regarding the fullduplex mode
1889 - new tesh files to check if gtnets tracing is ok
1892 * Asynchronous communications through the functions:
1893 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1894 * New function: MSG_load_platform_script()
1895 to make possible using a lua script instead of XML files to set up platforms
1896 * New function: MSG_set_function
1897 to associate functions to processes, used when bypassing the parser
1898 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1900 Platforms: Add some more examples in examples/platforms
1901 * Grid'5000: see www.grid5000.fr
1902 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1905 * Change the XML format. This is a very important modification. SimGrid
1906 3.5 introduces a new hierarchical format based on the notion of
1907 Autonomous Systems. Compatibility with old format is ensured
1908 through the perl script provided in the install bin directory
1909 bin/simgrid_update_xml.
1910 It is now possible to build platforms with specific routing
1911 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1912 connect several platforms together. We will try to provide soon
1913 set of realistic platforms exploiting these properties (have a
1914 look at examples/platforms/ for the moment).
1915 * Take the opportunity of the XML format change to be a good XML citizen:
1916 rename link:ctn to link_ctn and similar changes (also dealt with by
1918 * Add a new routing scheme (rule-based) using regular expressions. It
1919 enables to have an extremely low memory footprint when the
1920 underlying routing is simple and can be compactly described. You
1921 need to have libpcre4-dev (perl regular expressions) installed if
1922 you want to use this routing scheme.
1923 * Revive the cluster TAG and allow to easily and efficiently (both in
1924 term of memory and speed) connect clusters together. Have a look
1925 at teshsuite/simdag/platforms/ to see how this can be done. With
1926 this tag, you can create clusters with thousands of tasks at no
1927 cost (have a look at examples/platforms/).
1928 Note: clusters are implemented as ASes, so there is no need for an
1929 enclosing AS tag if you have only one cluster in your platform.
1930 * Add new generic functions in the public interface that allows the user
1931 to call SURF 'create_resource' methods from your code (same
1932 functionality as the XML bypass mechanism but with a much lighter
1934 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1935 that uses a piecewise linear approximation to produce better
1936 results when exchanging small messages.
1937 * Add a new parameter to handle correctly full duplex link and account
1938 for interferences between uplink and downlink communications
1939 (activate with --cfg=fullduplex:1).
1942 * Rename the SD_READY (all dependencies are satisfied and task is
1943 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1944 dependencies are satisfied) state.
1945 This prevents a confusion between the notion of "ready to schedule"
1946 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1947 simulated" (SD_RUNNABLE) used by the simulation kernel.
1948 * Change the way a task is considered as ready. Instead of removing
1949 dependencies when a task is done, a counter is decreased. This way,
1950 it is always possible to reach ancestors thanks to the
1951 SD_taks_get_parents function (even after the end of the simulation.)
1952 * Change the return type of SD_Simulate from (SD_task_t*) into
1953 xbt_dynar_t. This function was in handling a dynar internally and
1954 converted it into a NULL terminated array for historical reasons.
1955 * New function SD_dotload(char*) to load a DAG described in dot
1956 format. This loader and the corresponding examples require the
1957 installation of the graphviz library.
1958 * Fix a bug in the management of tasks of size 0 in the surf network
1959 models. This problem was only visible with SIMDAG and you should
1960 thus disregard results produced with earlier versions if you
1961 relied on this feature (some tasks were blocked because of this).
1962 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1963 even though your applications were doing only point-to-point
1964 communications and sequential computations. Now you can really use any
1965 model you want (of course, if you create real parallel tasks, which are
1966 not implemented in most models beside ptaskL07, this will abort).
1967 * Add an example that schedules a DAX on an heterogeneous platform
1968 using a Min-Min strategy.
1969 * New function SD_workstation_get_current_task() that returns the kind
1970 of task currently running on a workstation in the sequential access
1972 * Raise some warnings when unexecuted tasks remains at the end of the
1973 simulation. This is usually caused by cycles in the DAG.
1976 * New function: SIMIX_process_set_function() called by MSG_set_function
1977 * Change the underlying waiting queue in semaphores so that a process
1978 can wait on several of them simultaneously (as in waitany).
1979 * Fix the way to handle tokens in semaphores so that all access patterns
1980 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1981 * kill the dirty pimple SIMIX_message_sizes_output()
1982 Please use (proper) visualization instead
1985 * New data container: setset (set of sets of elements)
1986 * New module: mmalloc (mapped malloc, allowing to have several
1987 independent segments of malloc)
1988 * New function: xbt_dict_cursor_set_data()
1989 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1990 * New function: xbt_dynar_is_empty()
1991 * New function: xbt_fifo_get_last_item()
1992 * Fix xbt_dynar_shrink(): use the right element size.
1993 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1994 uninitialized areas during expand.
1995 * Fix semaphores: previous implementation was severely broken.
1996 * Use library init/fini functions for our initialization.
1997 - you can use logs and other feature as soon as you want in your
1998 code (even before the xbt_init / MSG_init)
1999 - xbt_exit is now a no-op and produce a warning when used.
2002 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2003 benefit from the latest improvement to the simulation kernel.
2004 * Kill measurement sockets for now. If you rely on them, sorry. This
2005 release is not for you. This feature will be reintroduced in the
2006 future, but we cannot delay the release any further.
2007 * New function: gras_msgtype_get_name().
2008 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2009 * Fix (at last) the pmm example: it should not randomly fail anymore.
2011 Build chain: bug fixes and overall polishing
2012 * Cmake is now stable enough. Hence, we killed the autotools.
2013 * Port to windows ( TM :)
2014 * Fix the 'make install' target.
2015 No need to use 'make install-simgrid' anymore
2016 * Introduce a 'make dist' target compiling a *source* archive
2017 'make package' compiles a binary archive
2018 * Compile java files only on need
2019 * Add --cd and --setenv command line options to tesh
2020 * Out of source builds are not fully supported yet, but we are close
2021 * Enable supernovae and optimization flags by default for our users
2024 * Add layer to set up environment directly from lua, without XML.
2025 * The effect of gras_stub_generator can be achieved through
2026 lua too (check examples/gras/console/ping_generator.lua)
2028 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2030 SimGrid (3.4.1) stable; urgency=low
2032 The "Polishing easter eggs is probably a good idea" release.
2033 This is a bug fixes release only.
2037 * Fix a bug preventing the tasks from begin garbage collected.
2040 * Fix a bug occuring when a host involved in a communication fails.
2041 This was not detected properly by the other peer involved in the
2042 communication. Now, it's reported as a network error.
2045 * Warn the user about loop dependencies in data flow of DAX files
2046 * Obey the control-flow dependencies of DAX files
2049 * Add option "enable_smpi" allowing to not compile SMPI.
2050 Probably useful for the (Mac) users experiencing a build error here
2051 * Improve the detection of lua5.1 and ruby1.8
2053 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2055 SimGrid (3.4) stable; urgency=low
2057 The "Easter in Cargese" release. Also known as (major changes):
2059 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2060 ~> bindings were greatly improved
2061 ~> new tracing infrastructure for better visualization introduced
2063 * the "Welcome to configury modernity" release.
2064 ~> we switched from autotools to cmake, and improved our cdash
2067 A more detailled list of changes follow (full detail in svn log).
2069 Java Bindings: Various Cleanups
2070 * (install java-gcj-compat-dev on debian-like to use them)
2071 * Remove put/get: no need to export deprecated interface in Java
2072 Use send/receive instead.
2073 * Cleanup the examples and add a README per directory
2074 * Remove example autoDestination (that's the only way to go now)
2075 * Remove example explicitDestination (was a plain copy of basic)
2076 * Make JniException a runtime exception, so that there is no need to
2077 declare the fact that you may encounter such a beast. I guess that
2078 nobody will ever want to survive such error.
2079 * Create specific errors for each MSG case of failure:
2080 host failure, transfer failure, timeout, task cancelled
2081 * Cleanup the exceptions that may get thrown by each function
2082 * Other internal cleanups in Java bindings. Performance still bad :/
2083 Ruby and Lua Bindings: create them
2084 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2085 * That's new and great, you should try them out.
2086 Same functionalities than Java bindings, only even less polished
2088 * Kill the useless "rate" argument of SD_task_get_execution_time()
2089 Everyone used to provide -1 as a value, it was not used, and the
2090 semantic of a possible use wasn't even clear.
2091 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2092 either as comm costs or compute costs to mean that there is no
2093 such thing for that specific task.
2094 * Add a SD_task_set_name() function
2095 * Fix SD_task_unschedule() on typed tasks
2096 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2097 * In DAX loader, accept useless 'level' attributes to <job> since
2098 LIGO DAGs have them (seem to be to ease graphical representation).
2100 * Add an example masterslave_mailbox.c using send/receive and not
2101 the deprecated put/get interface.
2102 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2103 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2105 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2106 (declare MSG_USE_DEPRECATED to still have the old name)
2107 * Add a MSG_task_set_data() function
2108 * About trace replay (see examples/msg/actions):
2110 - Allow to work with splitted trace files for each process
2111 Give the specific trace file as argument of each process,
2112 and call MSG_action_trace_run(NULL)
2113 You can still have one merged file for all processes.
2114 - Fix implementation of collective operations
2115 * Allow task_execute() on 0-sized tasks (closes #10063)
2117 * This is the first release of SimGrid where SMPI is not considered
2118 beta anymore (even if some corners should still be improved)
2119 * Port over the new SIMIX_network submodule (internal refactoring)
2120 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2121 * Implement more missing elements of the standard:
2123 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2125 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2126 MPI_Scan() MPI_Get_processor_name()
2127 - Added implementation of missing case for Alltoall (warning: it's
2128 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2129 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2130 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2131 SMPI_MPI_Allgatherv()
2132 * Bug fixes include:
2133 - MPI_Waitsome() was broken
2134 - Allow relative includes in smpicc
2135 - Command line cfg argument 'reference_speed' was ignored...
2136 - Some functions did not properly lead to auto-benching of user code
2137 - smpicc passes -O2 by default (just like openmpi one)
2139 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2140 * Bug fixes about timeouts during communications
2141 * add SIMIX_message_sizes_output() as a pimple to write to file the
2142 amount of messages per size. Use gnuplot to get histogram.
2143 Pimple because that's the only user-visible function of simix,
2144 defined directly in xbt.h (irk, sorry)
2146 - Add a SIMIX_sem_get_capacity() function
2147 - Fix interactions with processe resume/suspende
2148 - release_forever() was stupidly broken
2149 - Fix SIMIX_display_process_status() for processes in a semaphore
2150 - Make SIMIX_sem_block_onto() user-visible
2151 * Refactoring context stuff:
2152 - Use pseudo-OOP for better modularity
2153 - reimplement SIMIX_process_kill() without process_schedule() so
2154 that the latter can take as invariant that it is called from
2156 - Merge context_start into context_new for sake of simplicity
2158 * Add a Vivaldi network model, coded live during SUD'10 ;)
2159 * Rename configuration variables to start a hierarchy:
2160 o cpu_model -> cpu/model
2161 o network_model -> network/model
2162 o workstation_model -> workstation/model
2163 * New configuration variables:
2164 o network/bandwidth_factor: correction to bandwith
2165 o network/latency_factor: correction to latency
2166 o netwotk/weight_S: correction to the weight of competing streams
2167 * Add a long description to the models, that users can see with such
2168 argument on the command line: --cfg=cpu/model:help
2169 * --help-models display the long description of all known models
2171 * config: add the ability to set a default value after registration
2172 Does not override any previously set value (e.g. from cmd line)
2173 * dict: allow to have integer key and data.
2174 When so, you need to use the following functions
2175 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2176 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2177 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2178 In contrary to regular dicts, the key is not malloced before copy.
2179 Mixing scalar and regular elements in the same dict is not tested
2181 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2182 Tracing for Visualization:
2183 * SimGrid is now instrumented in order to generate a trace file for
2184 visualization analysis: to use it, need to compile SimGrid with the
2185 "tracing" option enabled, and instrument the program using SimGrid with
2186 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2187 (among other functions).
2188 * The instrumentation only traces the platform utilization for now
2189 * Documentation to use the tracing functions and how to analyze the
2190 traces with the Triva tool is written.
2191 * More information about: SimGrid FAQ (in the section Tracing Simulations
2194 * We moved to cmake as default build system. Autotools support will
2195 be dropped soon. Check the FAQ for more info about how to use it.
2196 * Greatly improved our cdash/ctest interactions
2197 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2198 * Added memory checking tests with valgrind; lot of memleak fixing.
2199 This may be the first release of simgrid with so few memory issues
2200 * Added code coverage tests.
2201 Our coverage is still improvable, but at least we see it on cdash.
2203 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2205 SimGrid (3.3.4) stable; urgency=low
2207 The "Desktop Grid needs love too" release (also called Xmas release).
2209 Models improvements:
2210 * Major speedup in the maxmin system solving by using lazy evaluation
2211 Instead of solving completely the maxmin system at each iteration,
2212 only invalidate (and recompute) the modified parts.
2213 This new feature is enabled in default models but you can try to
2214 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2215 * Cas01 IMproved as default CPU model
2216 This CPU model is the same Cas01 model, but it uses the
2217 maxmin-selective-update flag and a heap structure to manage
2218 actions on SURF kernel.
2219 It reduces the complexity to find the next action to finish and,
2220 consequently, it's faster than the old Cas01.
2221 This is the new default CPU model (Cas01).
2222 * Rename the old Cas01 model to Cas01_fullupdate
2223 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2224 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2225 * CpuTI (CPU Trace Integration)
2226 A new CPU model whose objective is simulate faster when using
2227 availability trace files.
2228 Instead of using a full featured, over engineered maxmin system for
2229 CPU modeling, this model does the pre-integration of traces files
2230 to calculate the amount of CPU power available, and so, executes
2231 faster than the old CPU models.
2232 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2233 * Use LV08 as default network model since it gives better accuracy
2234 for small messages and shouldn't change things for big ones.
2235 Use --cfg=network_model:CM02 to get the previous behavior.
2238 ******************************************
2239 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2240 ******************************************
2241 * The new CPU model may changes simulations!
2242 The point is that events occurring at the exact same timestamp
2243 are not scheduled in the same order with the old and new
2244 version. This may be enough to completely change the execution
2245 of simulations in some cases.
2246 * The new network model will change simulations!
2247 This new model is more realistic than the previous one, so you
2248 should consider redoing your old experiments with this model.
2249 Sorry for the inconvenience.
2252 * Introduce the supernovae compilation mode
2253 When compiled that way, the whole SimGrid (or almost) is put in a
2254 single compilation unit and compiled in one shoot.
2255 This is to help gcc which has difficulties to inline stuff from one
2257 The speedup seem to be above 15%, althrough more tests are needed on
2258 amd64 to confirm that gain.
2261 * Port of MSG's mailbox on top of SIMIX network
2262 The put/get mechanism was greatly simplified on the way.
2265 * New SIMIX network module. Provides:
2266 - Mailbox: rendez-vous mechanism to find with who you want to speak
2267 - Synchronous send/recv: easier and hopefully faster since the
2268 logic is handled in the maestro process directly now
2269 - Asynchronous send/recv: you dreamt of it? It's here now
2270 Too bad that nobody cared enough to propagate the change to MSG.
2271 * Add semaphores as SIMIX synchronization mechanism.
2274 * new function SD_daxload(char*) to load a DAX file
2275 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2276 * Introduce typed tasks. Specify its kind and cost at creation.
2277 At scheduling, just give where it should be placed, and the cost
2278 for each involved resource is automatically computed.
2279 Existing constructors so far (more to come of course):
2280 - SD_task_create_comm_e2e() for end-to-end communication
2281 - SD_task_create_comp_seq() for sequential computation
2282 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2283 * new function SD_task_dump() for debuging display
2284 * new function SD_task_dotty(task,FILE*) writing to file the info
2285 about the task in dotty format
2286 * SD_task_dependency_exists() can now cope with having one of its
2287 arguments NULL. If so, it tests whether the other argument has any
2289 * Add getters on list of preceding/following tasks:
2290 SD_task_get_parents(task) and SD_task_get_children(task)
2291 * Add getters on amount of workstations and list:
2292 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2293 * Add getter on task kind: SD_task_get_kind(task)
2294 * Update the start_time and finish_time of tasks on completion/failure
2295 * Bugfix: Remove task from state swags when destroyed
2298 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2301 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2302 dynar with one separate thread per value of the dynar.
2303 * Change the prototype of xbt_thread_create(), sorry.
2304 Added a boolean parameter indicating whether we want to join this
2305 thread (used in SG only for now)
2306 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2309 * GTNetS wrappers should now be usable again (and betterly tested too)
2310 * Fix a major regression from 3.2 where the timeout provided to
2311 MSG_task_put_with_timeout() was used as absolute time before which
2312 the comm should be done.
2313 * Start to fix the <cluster> tag.
2314 - Internal links should be good now (beside of the loopback, which
2315 use the private link instead)
2316 - paths to the external world is still rather broken
2317 - the <route:multi> tag is just broken. Actually that's brain-dead.
2318 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2319 to make it less stupid
2320 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2321 * Fix a source-level compatibility glitch from 3.2: after defining
2322 MSG_USE_DEPRECATED, you can use the old name
2323 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2324 * Allow to compile from the SVN with automake 1.11
2325 * Fix some problems when using the "start_time" tag in deployment XMLs.
2326 * Fix #8569: XBT/synchro.h has redundant declarations
2327 * Fix #8563: MSG return values and exceptions
2328 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2329 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2330 Thanks to Robson Peixoto.
2331 * Fix a vicious bug in dictionaries inducing that some elements were
2332 not freed on xbt_dict_free()
2334 Portability report of this version:
2335 * Main portability targets:
2336 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2337 on (amd64/i386/ia64)
2338 - mac leopard on i386
2339 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2340 but nothing critical.
2341 * Other platforms: windows, AIX and others were not tested for this release
2343 Timing report of this version:
2344 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2345 scenario parameters). From 8h to a few seconds in desktop grid settings.
2346 * Supernovae brings about 25% speedup on i386.
2348 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2350 SimGrid (3.3.3) stable; urgency=low
2352 The "Need for Speed" release.
2354 The timings done to validate the 3.3.2 were faulty.
2355 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2357 The problem was a conversion from a manually handled vector to
2358 xbt_dynar_t on the critical path.
2359 xbt_dynar_foreach calls functions, inducing stack management crap.
2361 We inlined these functions and xbt_dynar_foreach is now breath taking.
2362 We also inlined xbt_swag_belong on the way.
2364 Here are some approximate speedup measurements (on master/slaves
2365 simulations lasting between 10s and 20s each):
2366 3.3.1 -> 3.3.2: about same performance
2367 3.3.2 -> 3.3.3: 40% speedup
2368 3.3.1 -> 3.3.3: 40% speedup
2369 3.3.1 with inline patch -> 3.3.3: 30% speedup
2371 Our reading is that the refactoring which occurred in 3.3.2 made us
2372 suffer much more from the xbt_dynar_foreach low performance, but
2373 once we solved this, this refactoring proved to be very performance
2374 effective. From the 40% speedup, somehow, 10% are due to the
2375 inlining and 30% to the refactoring.
2377 That's a pitty that gcc cannot inline functions placed in other files
2378 alone. We have to choose between:
2379 - break the encapsulation (by putting private data structures and
2380 accessors in headers files to help gcc)
2381 - live with low performance
2382 - switch to a decent compiler such as icc (not quite possible).
2384 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2386 SimGrid (3.3.2) stable; urgency=low
2388 The "Simplicity does not preceed complexity, but follows it" release.
2390 The main contributors of this release were (lexical order):
2391 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2394 * Extract the routing logic into its own object.
2395 (was dupplicated in network.c and workstation_LV07.c;
2396 Allows to implement other ways of storing that info)
2397 => kill now useless network_card concept
2398 - Use dynar to represent routes (instead of void** + int*)
2399 - kill link_set (use surf_network_model->resource_set instead)
2400 - Add a command-line option to choose the routing schema to use
2401 - Add three new models:
2402 * Floyd (shortest path computed at initialization)
2403 * Dijikstra (shortest path recomputed all the time)
2404 * Cached Dijikstra (shortest path computed on need)
2405 All these models where contributed by Silas De Munck, and are
2406 described in his ICCS09 paper.
2408 * Simplify model declaration
2409 (less redirections, less function to write when defining a model)
2410 - Factorize stuff between models:
2413 surf_model_resource_set(model)
2414 surf_model_resource_by_name(model, name)
2415 - Unify the types of models in s_surf_model_t (using an union)
2416 - Embeed fields of common_public directly into s_surf_model_t
2417 - Rename model methods:
2418 action_free ~> action_unref
2419 action_change_state ~> action_state_set
2420 action_get_state ~> action_state_get
2421 - Change model methods into functions:
2422 (model)->common_public->action_use ~> surf_action_ref
2424 * Implement a generic resource; use it as ancestor to specific ones
2425 (allows to kill duplicated code in models)
2426 Drawback: timer command don't need no name nor properties;
2427 workstation_CLM03 don't need no properties
2428 (but I guess we can live with those few bytes wasted)
2430 * Improve the action object model
2431 - implement a constructor avoiding dupplicated code about field
2432 initialization in generic_action part.
2434 * Kill the SDP model: it has an external dependency, is deprecated
2435 in flavor of modern lmm models, and didn't compile since a while
2438 * Relocation of the context module from XBT to SIMIX.
2439 (the context were decoupled from the simix processes, duplicating a lot of code)
2440 => a lot of code was factorized
2441 - less overhead is introduced during scheduling
2442 - simpler API for the context factory
2443 - the logic for process creation,destruction and manipulation was simplified
2444 * Simplification of the s_smx_process_t data structure.
2445 => accesing the simix level data associated to a process is faster now,
2446 and the code is a lot more readable.
2449 * Implement some more MPI primitives:
2450 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2451 -implementation: Bcast: flat or 2-ary tree (default),
2452 Barrier: 4-ary tree,
2454 Allreduce: Reduce then Bcast
2455 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2456 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2457 Alltoallv: flat tree, like ompi
2459 * Add support for optimized collectives (Bcast is now binomial by default)
2460 * Port smpirun and smpicc to OS X
2463 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2464 Shout out if you used it.
2467 * Display the list of still queued messages in SG mode when existing
2471 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2472 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2473 * Add xbt_str_from_file(FILE*)
2474 * Add xbt_dict_get_key achieving a linear reverse search
2475 * Remove the context module
2477 Portability report of this version:
2478 * Main portability targets:
2479 - Linux(debian)/x86/context
2480 - Linux(debian)/x86/pthreads
2481 - Linux(debian)/amd64/context
2482 - Linux(debian)/amd64/pthreads
2483 On these, we still have the eratic breakages of gras/pmm and
2484 amok/saturate_sg reported in previous version. We still think
2485 that the tests are the cause of the fault, not the tested code.
2487 - Mac OSX Leopard/x86/context
2488 Still false negative in tesh autotesting.
2489 Smpi still fails, but this time because readlink does not accept -f
2490 Everything seems to work properly beside of that.
2493 - AIX version 5.3 (only tested contexts this time)
2494 Smpi still fails there because mktemp is not installed.
2495 Everything seems to work properly beside of that.
2497 I managed to compile it for the first time, but several breakages.
2498 Won't delay the release for this exotic platform.
2500 * Windows: it's still lagging behind. If you want to help, please
2503 Timing report of this version:
2504 This version seem to be more than 5% faster than 3.3.1 (on linux
2505 64bits with contextes). The gain is less than expected, we are
2506 investigating this for next release.
2508 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2510 SimGrid (3.3.1) stable; urgency=low
2513 * Implement a --cfg-help to show existing configuration variables
2514 * Build chain do not require doxygen in maintainer mode
2517 * fix a bug on struct sizeof computation, which prevented the
2518 exchange of arrays of structs in some conditions
2519 - added a regression test about this in datadesc_usage
2520 * Allow the exchange of 0-long dynamic vectors.
2521 - for that, use -1 as indicator of dynamic size instead of 0
2522 - This implied to change any size from unsigned long to long,
2523 reducing a bit communication abilities, but I guess that with
2524 64bits being quite common, this is more than enough.
2525 - This also induce a protocol change, thus bumping network protocol
2526 version from 0 to 1 (if we have external users, we have to get
2527 clean on that point too ;)
2528 - added two regression tests about this in datadesc_usage
2529 * Be more verbose when propagating local exceptions
2530 This helps debugging.
2531 * Display the status of simulated processes when receiving SIGINT in
2535 * Allow to control the simulation from a trace file.
2536 New functions MSG_action_register() and MSG_action_trace_run()
2537 The first one allows to associate a function execution to each
2538 kind of action while the second one parses a trace file and
2539 triggers the corresponding actions within the system.
2540 For now, only a toy example is provided in examples/msg/actions
2541 * Add an exemple of process migration in examples/msg/migration
2542 * Fix a bug in task exchange which broke MSG_task_get_sender()
2543 Add a teshsuite regression test for that.
2544 [Bug: if MSG_task_get_sender() is called after sender exit,
2546 * Fix a bug which prevented suspend/resume to work properly
2547 * Display the status of simulated processes when receiving SIGINT
2548 This fixes a regression of v3.3. due to the introduction of SIMIX
2549 * Bug fixing in failure management:
2550 - trace could not start by a failure at time 0
2551 - failure during communications were not working
2554 * Add SIMIX_process_set_name() to change the name of the current
2555 process in the log messages.
2556 * Store smx_hosts in a dict since we only retrieve them by name
2557 * Move the configuration infrastructure to surf
2560 * Move the configuration infrastructure to surf
2563 * Massive internal cleanups:
2564 - Store internal structures on processes instead of hosts (allows
2565 to have more than one process per host, in addition of being more
2567 - Cleanup the initialization/finalization process
2568 - Kill a whole bunch of unneeded synchronization:
2569 processes run in exclusive manner within the simulator
2570 - Move queues from global tables to process data fields
2572 - now accept -platform and -hostfile arguments
2573 - Pass the right rank value to processes according to the hostfile
2574 * Compile the examples by default, and use them as regression tests
2575 * Implement MPI_Wtime()
2576 * Change the reference speed to a command line option
2579 * TCP_gamma can now be specified as command line option using
2580 --cfg=TCP_gamma:10000000.0
2581 * Change the --surf-path cmd line option into --cfg=path:
2584 * Also include strbuff from xbt.h public header
2585 * xbt_ex_display(): do not free the exception after displaying
2586 This allows to do more with the given exception afterward.
2587 Users should call xbt_ex_free() themselves.
2591 Portability report of this version:
2592 * Main portability targets:
2593 - Linux(debian)/x86/context
2594 - Linux(debian)/x86/pthreads
2595 - Linux(debian)/amd64/context
2596 - Linux(debian)/amd64/pthreads
2597 These targets fail about 1/10 of times on gras/pmm, but we believe
2598 that this is because of the test, not because of simgrid.
2599 amok/saturate_sg fails even more rarely, and the test may not be
2602 - Mac OSX Leopard/x86/context
2603 The test suite still spits tons of errors because some obscure
2604 force prevents us from removing the temporary directories
2605 arguing that they still contain some metadata I've never heard of.
2606 Smpi fails because seq is not installed.
2607 Everything seems to work properly beside of that.
2610 - AIX version 5.3 (both contexts and pthread)
2611 Smpi still fails there because mktemp is not installed.
2612 XML inclusions seems rosty on AIX.
2614 * Windows: it's still lagging behind. If you want to help, please
2617 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2619 SimGrid (3.3) stable; urgency=high
2623 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2624 [Malek Cherier & Mt]
2626 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2629 Until now, GRAS were using MSG as an interface to SURF. It was
2630 quite difficult because both interface have several differences
2631 (MSG channels vs GRAS sockets were the most notable point).
2633 This also opens the gate to SMPI (which should occur soon) and speed
2634 up simulations by to 40% (even if it were not the main goal).
2636 **************************************
2637 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2638 **************************************
2639 The point is that events occuring at the exact same timestamp are
2640 not scheduled in the same order with the old and new version. This
2641 may be enough to completely change the execution of simulations in
2642 some cases. Sorry for the inconvenience.
2644 * Cleanup and upgrade the XML format to push further scalability
2645 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2647 * Improve the testing infrastructure with tesh. Now a very large part of
2648 the code is tested not only by being run but also by checking that the
2649 output match an expected output [Mt].
2651 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2652 is really less memory-demanding, which should allow you to use
2653 larger files in SimGrid [AL].
2655 * Inform valgrind about our contextes, so that it becomes usable
2656 with the default (and more effecient) version of SimGrid
2657 [contributed by Sékou Diakite, many thanks]
2660 * Introduce a listener thread in charge of receiving incoming
2661 messages from the network. It allows to overlap communication and
2662 computation but most notably, it removes some stupid deadlocks due
2663 to the fact that so far, a process could not send and receive at
2664 the same time. This made most non trivial communication schema
2666 * Convert the PIDs from long int to int to match the MSG ones (and
2667 linux ones too) [Mt]
2668 * New function: gras_agent_spawn() to launch a new process on
2669 current host. Only working in simulation for now. [Mt]
2670 * New function: gras_os_hostport() returning a constant form (ie,
2671 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2674 * Make the backtrace of exceptions more human readable [Mt]
2675 * New module: xbt/str [Mt]
2676 a ton of string utility functions (split, join, printf to a newly
2677 allocated buffer, trim, etc)
2678 * New module: xbt/hash [Mt]
2679 SHA1 hashing algorithm (more to come if needed)
2680 * New module: xbt/synchro [Mt]
2681 synchronization tools (mutex and conditions) working the same way
2682 in simulation and in real life (mainly useful for GRAS, but not
2684 * New module: xbt/queue [Mt]
2685 classical producer/consumer synchronization scheme
2686 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2687 (using the classical functions will get serialized) [Mt]
2688 * Make dictionary internal table dynamic. No need to specify its size
2689 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2691 * Make sure the log channels are organized as a tree under windows
2692 (because of ANSI C compatibility issue, any channel were child of
2693 root directly) [Mt].
2696 * Cleaned many thing in surf and fixed a few bugs [AL].
2697 * Add a nice command line configuration mechanism to compose models [AL].
2698 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2699 the previous one (KCCFLN05). It relies on something that looks like
2700 a max-min sharing mechanism but cannot be written as such. A new solver
2701 was thus designed [AL].
2702 * Add a new solver to lmm. Based on Lagrange optimization and
2703 gradient-based descent, it enables to efficiently maximise systems s.a
2705 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2707 This solver enables to propose two new network models for TCP Reno and
2708 TCP Vegas based on Low's work. These models still need to be fully
2709 tested though [Pedro Velho].
2712 * Bug fix in SD_simulate. Now the time bound given as argument is
2714 * Use the new parallel task model (ptask_L07) as default.
2715 * Use the SURF command line configuration mechanism.
2716 * 0-size tasks (for synchronization) should now work.
2718 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2720 SimGrid (3.2) stable; urgency=high
2724 We still experience issues on this platform, but we believe that at
2725 least MSG is usable.
2727 GRAS API BREAKAGE (for simplification purpose, sorry):
2728 * the gras_msgtype_by_name is not used anymore. Instead of
2729 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2730 you can write (and must)
2731 gras_msg_send(toserver, "request", &request);
2732 - If you still want to pass a gras_msgtype_t to the function (to cache
2733 the type and avoid the lookup time), use the gras_msg_send_() variant.
2734 - Impacted functions:
2735 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2736 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2737 * The callbacks are now expected to return 0 when everything went well
2738 (just like the main() function)
2740 GRAS new features and improvements:
2741 * New module mechanism where user code can use per process globals [Mt]
2742 This is similar to gras_userdata_*() functions, but for libraries. It
2743 factorize some code developped over and over in the examples and AMOK.
2744 It has still to be documented and used (only amok/peermanagement is
2746 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2747 loss when they were small enough to fit into the buffer and sent quickly
2748 enough so that they can all get received in one shoot.
2749 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2750 if not found. Use the *_or_null() variant for the old semantic.
2751 * In gras_msg_handle, do not discard messages without callback.
2752 They are probably messages to be explicitly awaited later (ie, proofs of
2753 mis-synchronization in userland since they are sent before being awaited)
2754 No big deal usually.
2755 * gras_socket_meas_send/recv: semantic changed!
2756 The numerical arguments used to be (1) the total amount of data to send
2757 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2758 messages. This was need for the fool willing to send more than MAXINT
2759 bytes on quite fat pipes.
2762 * Do really rename the hostmanagement module to peermanagement. [Mt]
2763 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2764 but this is rather new and this was documented in the module
2765 documentation (poor excuses, I admit)
2766 * Bandwidth measurement semantic changed! This follows the changes to
2767 gras_socket_meas_send/recv explained above.
2770 * A sequential mode has been added to the workstations. When a workstation
2771 is in sequential mode, it can execute only one task, and the other tasks
2772 are waiting in a FIFO. [Christophe Thiery]
2775 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2776 model for SIMDAG. [Christophe Thiery]
2777 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2778 equal to 0, which caused some bad behaviors in
2779 saturated_constraint_set_update. I now use a threshold mechanism like in
2783 * When running manually src/testall, you select specific units [Mt]
2784 testall is the result of our cunit mechanism, and should replace all
2785 the scripty thingy around since bash don't run easily on billware.
2787 * A mallocator system has been added. [Christophe Thiery]
2788 Mallocators allow you to recycle your unused objects instead of freeing them
2789 and allocating new ones.
2791 Documentation update:
2792 * FAQ reworking + New FAQs:
2793 - "Valgrind spits tons of errors!" [Mt]
2794 - "How to repport bugs" [Mt]
2795 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2796 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2797 - Communication time measurement within MSG [AL]
2798 - I experience weird communication times when I change the latency [AL]
2799 * GRAS tutorial [Mt]
2801 - an introduction to the framework and to the used communication model
2802 - an initiatic tour introducing the most proheminent features:
2804 . Lesson 0: Installing GRAS
2805 . Lesson 1: Setting up your own project
2806 o Part 2: Message passing
2807 . Lesson 2: Exchanging simple messages
2808 . Lesson 3: Passing arguments to the processes (in SG)
2809 . Lesson 4: Attaching callbacks to messages
2810 . Lesson 5: Using globals in processes
2811 . Lesson 6: Logging information properly
2812 . Lesson 7: Using internal timers
2813 . Lesson 8: Handling errors through exceptions
2814 . Lesson 9: Exchanging simple data
2815 . Lesson 10: Remote Procedure Calling (RPC)
2816 . Lesson 11: Explicitely waiting for messages
2817 . Recapping of message passing features in GRAS
2818 - A HOWTO section containing:
2819 o HOWTO design a GRAS application
2820 More are due, of course. They will come latter. In the meanwhile, you can
2821 check the examples which are still here.
2823 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2825 SimGrid (3.1) stable; urgency=high
2829 There was a stack corruption somewhere, visible only when optimizing
2830 with these versions. [Vince]
2833 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2834 module that enabled to play with Directed Acyclic Graphs. It is built
2835 directly on top of SURF and provides an API rather close to the old
2836 SG. Some old codes using SG are currently under rewrite to check that
2837 all needful functions are provided. [Christophe Thiery]
2840 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2841 extension of the classical CLM03 model that gracefully handles
2842 failures. This is now the default model for MSG and GRAS. It doesn't
2843 handle parallel tasks yet though. [AL]
2844 * Bug fix: Weights were not correctly set in the network part.
2845 WARNING: This may have resulted in incorrect results with simulations
2846 where there are more than one flow on a given link. [AL]
2849 * After a (long ?) discussion on simgrid-devel, we have decided that the
2850 convention we had on units was stupid. That is why it has been decided
2851 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2852 WARNING: This means that all previous platform files will not work as
2853 such with this version! A warning is issued to ask users to update
2855 A conversion script can be found in the contrib module of the CVS, under
2856 the name contrib/platform_generation/surfxml_update.pl [MQ]
2859 * Bug fix: Processes were started in reverse order, wrt deployment file.
2860 WARNING: if your code relies on this bug, please fix it. [AL]
2861 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2862 being executed on two different locations. [AL]
2863 * Bug fix: Failures are now better supported thanks to Derrick's tests
2864 (there was many failure situations I hadn't thought of and that weren't
2865 correctly handled). [AL]
2866 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2870 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2871 Exception occurring on server-side are propagated back to client (!).
2873 API CHANGE: the callback changed their prototype. Change:
2874 int my_handler(gras_socket_t expeditor, void *payload_data) {
2876 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2877 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2879 * New! function: gras_msg_handleall to deal with all messages arriving
2880 within a given period.
2881 * New! function: gras_socket_server_range to get a server socket in a
2882 range of port numbers (ease to avoid port number conflicts) [MQ]
2883 * New! gras processes display their backtrace when they get a SIGUSR1
2884 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2885 Sweet to debug RL processes [MQ]
2889 - Do not force experiment sizes to be expressed in kb, or it becomes
2890 impossible to measure the latency this way (needs one byte-long tests)
2891 WARNING: this changes the amok_bw_* function semantic. [MQ]
2892 - Implements the link saturation stuff. [MQ]
2893 * Peer management module:
2894 New! module factorizing code that we wrote over and over [MQ].
2897 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2898 - Test units are placed directly into the library code, they get extracted
2899 automatically and placed into the src/testall binary.
2900 - Convert most of the XBT tests to this system.
2901 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2902 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2903 live in the same C code. This required to change a little bit the API
2904 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2905 * New module: sparse graph structure with basic algorithms (this is work
2906 in progress and the API is not considered to be frozen yet). [AL]
2907 * Display more information on backtraces: source line & function names are
2908 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2909 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2910 * Reworked a little bit some #include statements to load only required
2911 headers. Some user code that relied on SimGrid to include stdlib or
2912 stdio may need to include it by themselves. [AL]
2913 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2915 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2916 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2917 but I guess that almost nobody used those parts. [MQ]
2919 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2921 SimGrid (3.0.1) stable; urgency=low
2924 * Unfortunately, I had missed 5 misnamed functions:
2925 xbt_fifo_item_t xbt_fifo_newitem(void);
2926 void xbt_fifo_freeitem(xbt_fifo_item_t);
2927 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2928 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2929 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2930 They're now deprecated. Please use their new versions:
2931 xbt_fifo_item_t xbt_fifo_new_item(void);
2932 void xbt_fifo_free_item(xbt_fifo_item_t);
2933 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2934 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2935 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2937 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2938 * Documentation: xbt_log module unmercifully reworked [MQ]
2939 * Bugfix: there was a problem with the ending of contexts with
2940 the pthread backend. It caused some weird deadlock or behavior
2941 depending on the pthread implementation. [AL]
2942 * Bugfix: get the exceptions raised in the simulator repport where
2943 and why they come from when they are not catched in time [AL, MQ]
2946 * Bugfix: Do repport the error when two non-connected hosts try to
2947 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2950 * Add additionnal checkings on communications. Assert that two
2951 communicating hosts are connected by a set of links... [AL]
2954 * Add additionnal checkings on channel values in communication [AL]
2955 * New: MSG_task_get_source to see on which host a task was generated [HC]
2956 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2957 the number of tasks waiting to be received on channel and sent
2959 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2960 waits for the first task coming from a given host.. [AL]
2962 GRAS new functionnalities: [MQ]
2963 * Enhance the parsing macro to allow the size of multidimentional objects
2964 to be given thru annotations.
2965 * New example (and documentation): Matrix Multiplication a la RPC
2966 (as when I was young!) and fix a bunch of bugs found on the way.
2968 GRAS performance improvements: [MQ]
2970 * Reduce the amount of cbps creation/destruction by making it static to
2971 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2974 * Change libdata to a set so that we can search for stuff by ID (and thus
2975 reduce the insane amount of dict lookups)
2978 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2979 RL communication on top of SG since it's so uneffective.
2980 It may also be used for inter-thread communication in RL, one day.
2981 * Use gras_datadesc_copy() to exchange messages on top of SG
2983 - improve message exchange performance on top of SG
2984 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2985 * Don't exchange on the network the size of the used part of buffer,
2986 instead, specify the possible buffer size to read().
2988 - reduces the amount of read/write calls (one pair per exchange)
2989 - reduces the amount of exchanged data (the size)
2990 - allows to retrieve all arrived data on receiver side, if we don't need
2991 it right now (subsequent read will peek the buffer)
2992 - allows the receiver to proceed with the begining of the stream before
2993 everything is arrived
2994 - make it possible to build an iov transport (using readv/writev)
2996 - take care of the data with non-stable storage (like stacked data),
2998 * If possible, TCP send uses vector I/O (when writev() is here)
2999 - Don't use it for receive since we send data sizes and data on the
3000 same stream, so we wouldn't be able to chain large amount of chunks
3001 before having to flush the stuff to read the size.
3002 * Rework the transport plugin mechanism to simplify it and reduce the
3003 amount of pointer dereferencement when searching for the right function
3006 * I guess that now, we do almost as few system calls as possible while
3007 doing as few data copy as possible.
3009 To improve it further, we could try to send all the sizes first and then
3010 all the data (to use iov on receiving size), but it's only a partial
3011 solution: when you have 2 dimensional data, the sizes of the second
3012 dimension is data of the first dimension, so you need 3 streams.
3014 I'm not sure the potential performance gains justify the coding burden.
3016 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3018 SimGrid (3.00) stable; urgency=high
3021 * New! Give the possibility to hijack the surf parser and thus bypass
3022 MSG_create_environment and MSG_launch_application. Have a look at
3023 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3025 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3027 SimGrid (2.96) unstable; urgency=low
3032 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3033 This deprecates the xbt_error_t mechanisms.
3034 It modifies (simplifies) all XBT and GRAS API.
3035 MSG API keeps unchanged (exceptions raised by XBT are catched from
3036 within MSG and masked with existing error handling facilities)
3039 * New! Add a FATPIPE model. [AL]
3040 * New! Add a parallel task model. [AL]
3041 * New! Add automatically a loopback interface (in the default
3042 network model) if none was precised.
3045 * Bugfix: MSG_process_resume now works with the current running process.
3047 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3048 * Modification of MSG_task_get_compute_duration. Once a task has been
3049 processed, the value returned by this function is now equal to 0. [AL]
3050 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3051 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3052 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3053 of a m_task. For now, MSG_task_cancel only works with computation
3055 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3056 of the processor (in Mflop/s) regardless of the current load on the
3058 * API Change: use proper naming convention for MSG_getClock and
3059 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3061 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3062 This function changes the priority of a computation task. This priority
3063 doesn't affect the transfer rate. A priority of 2 will make a task
3064 receive two times more cpu power than the other ones. This function
3065 has been added to suit the needs of Nguyen The Loc and hasn't been that
3066 much tested yet. So if it fails, please report it and send me your code.
3068 * API Change: removed all functions and types that were marked "deprecated"
3069 since many months. Renamed MSG_global_init_args to MSG_global_init.
3071 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3073 SimGrid (2.95) unstable; urgency=low
3076 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3077 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3078 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3081 * Bugfix: Allow absolute paths for platform description files [MQ]
3082 * Bugfix: do free the variables after use. Leads to drastic performance
3084 * Implement max_duration (ie, timeouts) on resources [AL]
3087 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3089 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3090 the needed support (timeouts) [AL]
3093 * Implement measurement sockets. You can now get the bandwidth between two
3094 hosts thanks to AMOK (see below). [MQ]
3095 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3096 dynar over the network (yeah) [MQ]
3097 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3098 * simple/static token-ring example. [Alexandre Colucci and MQ]
3099 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3100 in place before [MQ]
3103 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3104 type parsing. "short" and "long" modifiers were also ignored; other
3105 modifier (such as reference level) are still ignored. [MQ]
3106 * Embeed the buffer size within the buffer itself on SG. [MQ]
3107 That way, send() are atomic and cannot get intermixed anymore (at least
3108 the ones which are less than 100k; bigger messages still have the issue)
3109 * Array size pushed by the field, not by the field type (or each
3110 and every long int will push stuff to the cbps) [MQ]
3111 * use select() to sleep since it allows to portably sleep less than one
3114 GRAS (minor cleanups)
3115 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3118 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3119 * type category 'ignored' killed as it was never used and were difficult
3121 * whether a type can cycle or not is now a flag, leaving room for more
3122 flags (as "ignored", if we feel the need one day ;)
3123 * Rename raw sockets to measurement sockets since "raw" has another
3124 meaning in networking community.
3127 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3128 and offers features not belonging to GRAS but that most applications
3129 need. One day, it may be a set of plugins loadable at runtime.
3130 * New module: bandwidth
3131 bandwidth measurement between arbitrary nodes running this module. [MQ]
3133 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3135 SimGrid (2.94) unstable; urgency=low
3137 The first beta release of SimGrid 3 !
3141 * Update the main page and the FAQ. Adding references to gforge.
3144 * Add a gras_os_getpid function.
3147 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3148 * Extend the logs so that they also print PID, hostname, date, ... if
3150 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3151 and kill the old version which were in testsuite/
3152 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3156 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3157 for users. Instead, each programming environment (GRAS, MSG,...) use it
3158 when needed to provide such a feature to users.
3159 Don't shortcut the mechanism or you will also shortcut the virtualization
3160 you need on the simulator.
3164 * Cleanups in configury with regard to compile optimization/warning flags.
3165 Also add -fno-loop-optimize to any powerpc since it's the optimization
3166 killing gcc (< 3.4.0).
3167 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3168 needed by MSG examples complications
3169 * Borrow configury beautifications from PHP
3172 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3173 warning (thanks loris for stumbling into this one).
3174 * Bugfix: stop loading private headers (gras_config.h) from the public
3178 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3179 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3180 RL would imply a DNS resolver, which is *hard* to do in a portable way
3181 (and therefore delayed).
3182 * Implement a real timer mechanism and use it in timing macros. This allows
3183 to avoid rounding errors and get a 0.000005 sec precision in timing
3184 macros. While I was at it, various cleanups:
3185 - allow to declare more than one timed section per file (fix a stupid bug)
3186 - move some private declaration to the right place
3187 - merge conditional execution and timing macros into emulation module
3188 - document the module
3189 - make sure the module cleanups its mess on gras_exit
3190 * Documentation improvements:
3191 - (new) how to compile applications using GRAS
3192 - (new) emulation support (timing macros)
3194 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3196 SimGrid (2.93) unstable; urgency=low
3198 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3201 - Use Paje properly where used. Still to be sanitized properly.
3202 - Portability fix: Add an implementation of the contexts using pthread
3206 - Add xbt_procname(): returns the name of the current process.
3207 Use it to show the current process's name in all logging.
3209 - fix detection of older flex version and the reaction, since we do
3210 depend on modern ones (we use lex_destroy)
3211 - Better separation of SG and RL in the libs: remove all simulation code
3212 from libgras. As a result, this lib is now only 200k when stripped.
3213 Some of the xbt modules may also be duplicated (two sets and such) and
3214 should be cleaned/killed before SG3.
3215 - Insist on using xlC on AIX because of weird problems involving gcc there.
3216 - Cleanup the make remote stuff. This is now done by scripts
3217 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3218 mainly for our private use, but we're working on changing them to user
3221 - Bugfix: flush the socket on close only if there is some *output*.
3222 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3224 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3227 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3229 SimGrid (2.92) unstable; urgency=low
3231 Alpha 3 on the path to SimGrid 3
3235 - New! First try of benchmarking macros.
3236 - New! First try so that gras_stub_generator generate deployment and
3237 remote compilation helpers.
3239 - Bugfix: Initialization fix in msg_test.
3243 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3245 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3247 - API Change: gras_os_sleep to take the amount of seconds as a double.
3248 Accepting an int was error prone since it was the only location where
3249 seconds were coded as such. It leaded to damn rounding errors.
3250 - Bugfix: Hard to belive that timers ever worked before this.
3252 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3254 SimGrid (2.91) unstable; urgency=low
3256 Alpha 2 on the path to SimGrid 3
3260 - Bug fix in the lmm_solver.
3262 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3263 through the function MSG_paje_output.
3264 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3265 - It is possible to bound the rate of a communication in MSG with
3266 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3267 to write it in the changelog).
3268 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3273 - Various cleanups to the autotools stuff
3274 - Begin to move Gras examples to examples/gras/
3275 - Let make distcheck work again (yeah!)
3277 - documentation overhauled using doxygen.
3278 gtk-doc-tools is dead in SimGrid now.
3279 - Automatically extract all existing logging categories, and add the list
3280 to the documentation (long standing one, to say the less)
3282 - Cleanup the known architecture table. Reorder the entries to group what
3283 should be, and use a more consistent naming scheme.
3284 (some of the test dataset are still to be regenerated)
3285 - New! Allow library to register globals on each process just as userdata
3287 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3288 lookup time (for now).
3289 Use it in msg and trp.
3290 This cleans a lot the internals and helps enforcing privacy of the
3291 headers between the gras components.
3292 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3293 - Bugfix: gras_os_time was delirious in RL.
3294 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3296 - Reenable GRAS now that it works.
3298 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3300 SimGrid (2.90) unstable; urgency=low
3302 Alpha 1 on the path to SimGrid 3
3304 * It is a long time since the last release of SimGrid. I'm sorry about
3305 that but as I had told you, I was rewriting a lot of things. I apologize
3306 to those who had been reporting bugs to me and that I had not answered.
3307 If your bug is still in the new version, please tell me. Here is a
3308 summary of the main changes.
3310 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3311 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3312 portability, speed, and a lot more but you'll figure it out later.
3313 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3314 I wanted to release the new MSG as soon as possible and I have
3315 broken GRAS, which is the reason why, for now, only MSG is fully
3316 functional. A laconic description of these projects is available
3317 in the documentation.
3319 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3320 kernel optimized for our needs (called SURF but only the developers
3321 should use it). Hence, MSG is now roughly 30 times faster and I think
3322 that by rewriting a little bit MSG, I could event speed it up a little
3323 bit more. Beside the gain in speed, it is also much easier to encode a
3324 new platform model with SURF than it was with SG. More to come...
3326 * REVOLUTION 3: I have tried to change a little as possible the API of
3327 MSG but a few things really had to disappear. The main differences
3328 with the previous version are:
3329 1) no more m_links_t and the corresponding functions. Platforms are
3330 directly read from a XML description and cannot be hard-coded
3331 anymore. The same format is used for application deployment
3332 description. The new format is described in the documentation.
3333 Have a look in tools/platform_generation. There is a tiny script
3334 that converts from the old platform format to the new one. Concerning
3335 the application deployment format, parsing the old one is tricky.
3336 I think most of you should however be able to convert your files. If
3337 it is really an issue, I can write a C code that does the conversion.
3339 2) the toolbox tbx does not exist anymore. We now have a library
3340 with much more data-structures but without the hash-tables (we have
3341 dictionaries that are much faster).
3343 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3345 *****************************************************************************
3346 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3347 * don't feel like converting the oldies. (Mt) *
3348 *****************************************************************************
3351 Version 2.90: "the long awaited one"
3352 - Finished rewriting and debugging MSG. Rewrote the documentation.
3353 - disable GRAS for now since it needs to be ported to the newest SG
3356 - Finish the port to windows (using mingw32 for cross-compile)
3359 - Main loop and datastructures of SURF. A cpu resource object is
3360 functional. Surf can thus be used to create cpu's with variable
3361 performance on which you can execute some actions.
3363 2004-11-15 Martin Quinson
3364 - Port to ARM. Simply added the alignment and size descriptions. Should
3365 work, but the ARM machines are so slow that I didn't had the opportunity
3366 to 'make check' over there yet.
3368 2004-11-15 Arnaud Legrand
3369 - Trace manager now written. It uses a heap structure and is therefore
3370 expected to be efficient. It may however be speeded up (particularly
3371 when many events occur at the same date) by using red and black
3372 trees. One day maybe...
3373 - Max-min linear system solver written. It uses a sparse matrix
3374 structure taking advantage of its expected use. Most operations are
3375 O(1) and free/calloc are called as few as possible. The computation of
3376 the minimum could however be improved by using a red and black tree
3379 2004-11-03 Arnaud Legrand
3380 - Rename every gras_* function that was in xbt/ to its xbt_
3382 - Add a heap and a doubly-linked list to xbt
3383 - Added a dichotomy to the dictionaries. make check works as well before
3384 so I assume that the patch is correct. I do not know however if things
3385 run effectively faster than before now. :)
3387 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3388 SimGrid, and the version number is bumped to 2.x
3390 2004-10-29 Martin Quinson
3391 - Introduction of the remote errors.
3392 They are the result of a RMI/RPC on the remote machine.
3393 ErrCodes being scalar values, you can't get the host on which those
3394 errors did happen. Extending the error mechanism as in Gnome is possible.
3395 No idea yet whether it is a good idea.
3397 2004-10-28 Martin Quinson
3398 - Interface revolution: the Starred Structure Eradication.
3399 I used to do typedef struct {} toto_t; and then handle *toto_t.
3400 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3401 - ???_t is a valid type (builded with typedef)
3402 - s_toto_t is a structure (access to fields with .)
3403 - s_toto is a structure needing 'struct' keyword to be used
3404 - e_toto_t is an enum
3405 - toto_t is an 'object' (struct*)
3407 typedef struct s_toto {} s_toto_t, *toto_t;
3408 typedef enum {} e_toto_t;
3409 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3410 s_toto_t) is private.
3412 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3413 so that it can set the variable to NULL. It was so for dicts and sets,
3414 it changed for dynars.
3416 - Fix a bunch of memleaks in dict_remove
3417 - Fix a bug in sg/server_socket opening: it failed all the time.
3419 2004-10-07 Martin Quinson
3420 - Speed up dynar lookup operation a bit.
3422 gras_dynar_get is dead.
3424 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3425 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3426 (which gives you the address of the stored data).
3428 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3429 copy of the data using an affectation to do the job and not a memcpy.
3431 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3432 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3434 It does not really speedup the dynar test because they are
3435 setting elements all the time (and look them seldom). But the dict does
3436 far more lookup than setting.
3438 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3440 2004-10-05 Martin Quinson
3441 - Allow to (en/dis)able the cycle detection at run time.
3443 Whether we should check for cycle or not is now a property of each
3444 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3445 datadesc_cycle_unset allow to remove this property when previously set.
3447 Note that the cycle detection is off by default since it impacts the
3448 performance. Watch the data you feed GRAS with ;)
3450 This property is hereditary. Any element embedded in a structure having it
3451 set have it set for the time of this data exchange.
3453 You should set it both on sender and receiver side. If you don't set it on
3454 sender side, it will enter an endless loop. If you forget on receiver
3455 side, the cycles won't be recreated after communication.
3457 - Header reorganization.
3458 Kill gras_private.h, each submodule must load the headers it needs.
3460 2004-10-04 Martin Quinson
3461 - Interface revolution: do not try to survive to malloc failure.
3463 Now, gras_malloc and friends call gras_abort() on failure.
3464 As a conclusion, malloc_error is not a valid error anymore, and all
3465 functions for which it was the only gras_error_t return value are
3466 changed. They now return void, or there result directly.
3467 This simplify the API a lot.
3469 2004-09-29 Martin Quinson
3470 - Re-enable raw sockets.
3471 Created by gras_socket_{client,server}_ext;
3472 Used with gras_raw_{send,recv}
3475 It should allow to kill the last bits of gras first version soon.
3477 This is not completely satisfactory yet (duplicate code with
3478 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3481 - Simplify transport plugin (internal) interface by not passing any
3482 argument to _server and _client, but embedding them in the socket
3485 2004-09-28 Martin Quinson
3486 - Finish the port to AIX.
3487 autoconf was my problem (segfault within the malloc replacement
3488 function. No idea why)
3490 2004-09-16 Martin Quinson
3491 - Fix some size_t madness on 64bit architectures.
3493 2004-09-08 Martin Quinson
3494 - Reduce the number of system headers loaded, overload some more system
3495 calls (such as malloc to cast the result of the system one, and work
3497 - Fix and reintroduce the config support
3499 2004-09-07 Martin Quinson
3500 - Source code reorganization to allow Arnaud to surf all over there.
3501 - Allow to document the logging categories.
3502 - Remove all uppercase from logging categories and useless cleanup in names.
3504 2004-08-18 Martin Quinson
3505 Version 0.6.2 (protocol not changed; API changed)
3506 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3507 gras_error_t), and NULL when not found. Functions expecting a msgtype
3508 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3509 hopefully usefull message.
3510 - Portability to prehistoric sparcs again
3512 2004-08-17 Martin Quinson
3513 Version 0.6.1 (protocol not changed; ABI not changed)
3514 - prealloc some buffers to speed things up
3516 2004-08-11 Martin Quinson
3517 Version 0.6 (protocol not changed; ABI expended)
3518 - The parsing macro can deal with the references, provided that you add
3519 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3521 2004-08-09 Martin Quinson
3522 Version 0.5 (protocol not changed; ABI changed)
3523 - Allow to off turn the cycle detection code in data exchange at
3524 compilation time. It should be at run time, but I'm short of time (and
3525 the config stuff is still broken). That way, we keep dict out of the
3526 critical path, which is good because the performance is poor:
3527 - search not dichotomial yet
3528 - dynar give no way to access their content and memcpy everytime
3529 - In composed data description (struct, ref and so on), stop foolness of
3530 keeping the subtype's ID, but store the type itself. This keeps sets out
3531 of the critical path, which is good since they rely on dynar and
3532 dictionnaries. The only loose of that is that we cannot detect the
3533 redeclaration of a structure/union with another content (but I'm not sure
3534 the code detected well this error before anyway). We still can detect
3535 the redefinition discrepancy for the other types.
3536 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3537 it breaks the code because of type-punning used all over the place).
3538 This breaks on all non-gcc architectures (for now).
3540 All those changes (plus the buffer of last time) allow me to gain 2 order
3541 of magnitude on cruel tests consisting of 800000 array of integers on two
3542 level of a hierarchical structure (200 secondes -> 4 secondes)
3545 - the selector of reference must now return the type it points to, not
3546 the ID of this type.
3548 2004-08-06 Martin Quinson
3549 Version 0.4 (protocol changed; ABI not changed)
3550 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3551 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3552 - Implement a buffer transport, and use it by default (it relies on tcp in
3553 real life and on sg in simulation).
3554 That's a bit hackish since I had a new field to the structure to store
3555 its data without interfering with the subtype ones. Inheritance
3556 is tricky in C. And that's a kind of reverse inheritance with one class
3557 derivating two classes. Or maybe a game with java interfaces. Anyway,
3558 that's damn hard in C (at least).
3559 Moreover, I got tired while trying to ensure plugin separation and
3560 genericity in SG mode. MSG wants me to do weird things, so let's go for
3561 cruel hacks (temporarily of course ;).
3562 See comment in transport_private.h:71
3563 - do not include all the _interface headers in private but in the files
3564 which really need them (to cut the compilation time when they are
3567 2004-07-26 Martin Quinson
3568 Version 0.3 (protocol not changed; ABI changed)
3569 - Major overhault of the datadesc interface to simplify it:
3570 - shorted the function names:
3571 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3572 - add a trivial way to push/pop integers into the cbps without malloc.
3573 This allows to make really generic sub_type description, which simply
3574 pop their size of the stack.
3575 - add a function gras_datadesc_ref_pop_arr() which does what users want
3576 most of the time: Declare a dynamic array (which pops its size of the
3577 stack) and declare a reference to it. Poor name, but anyway.
3578 - kill the post-send callback, add a post-receive one
3580 2004-07-23 Martin Quinson
3581 Version 0.2 (protocol changed; ABI changed)
3582 - add some testing for cpbs in the test cases, and fix some more bugs.
3583 This invalidate again the little64 data file, since I cannot regenerate
3585 - remove an awfull optimization in the logging stuff, allowing me to:
3586 - understand it again
3587 - learn gcc how to check that the argument match the provided format
3588 - fix all errors revealed by gcc after that
3589 - internal keys of dict are not \0 terminated. Deal with it properly in
3590 loggings instead of segfaulting when the user want to see the logs :-/
3592 2004-07-22 Martin Quinson
3593 - Fix some stupid bug preventing cbps (callback postit) from working
3595 2004-07-21 Martin Quinson
3596 - Some documentation cleanups
3597 - remove the useless last argument of msgtype_declare
3598 - rename the Virtu functions to fit into the 'os' namespace
3599 - move headers src/include -> src/include/gras/ and stop fooling with
3600 gras -> . symbolic link
3601 - make distcheck is now successful
3603 2004-07-19 Martin Quinson
3605 - Build shared library also
3606 - Install html doc to the right location
3607 - stop removing maintainer files in make clean
3608 - build tests only on make check
3610 2004-07-13 Martin Quinson
3612 - No major issue in previous version => change versioning schema
3613 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3614 corresponding dataset.
3616 2004-07-11 Martin Quinson
3618 - Get it working with any kind of structure (we can compute the padding
3619 bytes remotely for all the architectures I have access to)
3620 - Implement the structure parsing macro (still not quite robust/complete)
3621 - Improvement to the remote testing toysuite
3623 2004-07-10 Martin Quinson
3624 [autoconf mechanism]
3625 - get ride of a bunch of deprecated macros
3626 - actually run the test for two-compliment, not only compile it :-/
3627 - test whether the structures get packed (and bail out if yes. Damn.
3628 Alignment is a serious matter)
3629 - test whether the structures get compacted (but respecting the alignment
3630 constraints of each types)
3631 - test whether the array fields of structures can straddle alignment boundaries
3633 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3634 (took me 2 hours to find that bug, looking at the wrong place)
3635 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3636 before sending/receiving and are used to compute the offsets of fields
3637 - Given that padding size depend even on compiler options, keep track of
3638 alignment and aligned_size only for the current architecture. Not a big
3639 deal since we send structure fields one after the other (seems
3641 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3642 program, let it work on linux/gcc/little32. portability todo.
3644 2004-07-08 Martin Quinson
3645 - import and improve remote compilation support from FAST
3646 - make sure make check works on half a dozen of machines out there
3648 2004-07-07 Martin Quinson
3649 Let's say it's version 0.0.3 ;)
3650 - Implement conversions (yuhu!)
3651 - Let it work on solaris (beside conversion, of course)
3652 - Stupid me, using rand() to generate the conversion datatests in not wise.
3654 2004-07-06 Martin Quinson
3655 - Let make dist work, since I'm gonna need it to compile on remote hosts
3656 - Let Tests/datadesc_usage write the architecture on which the file was
3657 generated as first byte.
3658 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3659 architecture descriptions.
3660 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3661 on those architectures.
3662 - Optimization: send/recv array of scalar in one shoot
3664 2004-07-05 Martin Quinson
3665 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3667 - Plug a whole bunch of memleaks
3668 - each process now have to call gras_{init,exit}. One day, their log
3669 settings will be separated
3670 - Continue the code factorisation between SG, RL and common in Transport.
3672 2004-07-04 Martin Quinson
3674 - Redistribution between SG and RL.
3675 We wanna have to accept in SG, so move accepted related parts of RL in
3676 the common part. (more precisely, the dynar of all known sockets is no
3677 more a static in transport.c, but part of the process_data)
3679 [gras_stub_generator]
3680 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3682 2004-07-03 Martin Quinson
3683 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3684 them all up in one shot)
3686 - Fix the ugly case of reference to dynamic array.
3687 - New (semi-public) function gras_datadesc_size to allow the messaging
3688 layer to malloc the needed space for the buffer.
3690 - gras_socket_close now expect the socket to close (and not its address to
3691 put NULL in it after it). This is because the socket passed to handlers
3692 is one of their argument (=> not writable).
3694 - propagate the interface cleanup from last week in datadesc, ie remove a
3695 superfluous level of indirection. User pass adress of variable
3696 containing data (both when sending and receiving), and not of a variable
3697 being a pointer to the data. Let's say that I like it better ;)
3698 The price for that is constructs like "int msg=*(int*)payload" in
3699 handlers, but it's a fine price, IMHO.
3701 - Let it work in RL (yuhu)
3703 2004-06-21 Martin Quinson
3705 - porting SG plugin and SG select to new standards (works almost).
3706 - plug memleaks and fix bugs around.
3709 - cleanup the prototype of data recv and force users to specify when they
3710 want to handle references to objects. Test case working even for cycles.
3711 - plug memleaks. Valgrind is perfectly ok with this.
3713 2004-06-12 Martin Quinson
3715 - cleanup the separation between plugin and main code in plugin creation
3717 2004-06-11 Martin Quinson
3719 - Reput hook for raw sockets, needed for BW experiments
3720 - kill a few lines of dead code
3721 [Data description] Interface cleanup
3722 - gras_datadesc_by_name returns the searched type or NULL.
3723 That way, no variable is needed to use a type desc once, which makes
3725 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3726 two parameters were strings (field name, type name), leading to
3728 [Dicos] Interface cleanup
3729 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3730 This is consistant with the dynar API.
3732 2004-04-21 Martin Quinson
3734 - Porting to new standards.
3736 - interface cleanup.
3737 There is no bag anymore, no need to take extra provision to mask the
3738 pointers behind "ID".
3739 Better splitup of functions between files create/exchange/convert.
3740 This is still a bit artificial since convert and receive are so
3741 interleaved, but anyway.
3743 - add a queued message list to procdata (the ones not matching criteria
3745 - factorize some more code between SG and RL wrt procdata
3747 - use gras_exit in example to track memleaks
3748 - get rid of gs_example now that GS is properly integrated into gras
3749 - update run_test to integrate the latest tests (datadesc)
3751 - rename WARNINGn macros to WARNn since it prooved error-prone
3753 2004-04-19 Martin Quinson
3755 - register init/exit functions within gras module mechanism
3756 - send/receive function.
3757 Convertion is not implemented, but short-cutted if not needed.
3758 struct/array elements are sent one by one (instead of block-wise), but
3759 nobody really cares (yet). Get a prototype before optimizing.
3760 - tests (using a file socket) for DD send/receive on:
3761 - base types: int, float
3762 - array: fixed size, string (ie ref to dynamic string)
3763 - structure: homogeneous, heterogeneous
3764 - chained list, graph with cycle
3765 Believe it or not, valgrind is not too unhappy with the results. The
3766 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3770 - Bugfix when using a filename explicitely (instead of '-')
3772 2004-04-09 Martin Quinson
3774 - factorize more code between RL and SG in socket creation
3775 - Complete the implementation and tests of:
3777 o file (only in RL, and mainly for debugging)
3779 I lost 3 days to design a portable address resolver, and then decided
3780 that the prototype mainly have to run on my box.
3781 Addressing portability too early may be like optimizing too early :-/
3783 - use gras_init in the Tests instead of the crappy parse_log_opt
3784 (the latter function is removed)
3785 [Conditional execution]
3786 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3787 [Code reorganisation]
3788 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3789 Build examples against the RL library, since there is no way to disable
3790 its creation for now.
3792 For information, the beginning of coding on GRAS was back in june
3793 2003. I guess that every line has been rewritten at least twice since