1 SimGrid (3.22) NOT RELEASED (Release Target: December 21. 2018, 22:23 UTC)
4 - #261: Document the parameters of parallel execution's constructor
6 ----------------------------------------------------------------------------
8 SimGrid (3.21) October 3. 2018
10 The Restarting Documentation (TRD) Release.
13 - Convert the user manual to Sphinx for improved usability.
14 Unstable version now hosted on https://simgrid.frama.io/simgrid
15 Many glitches remain; stay tuned.
16 - Restrict the Doxygen reference API to the installed header files.
17 (documenting all internals this way is overkill + doxygen scales badly)
18 - New tutorial on S4U (highly inspired from the old MSG one, but with
19 a git repository to fork as a starting point, and a docker image)
20 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
21 - The Developper manual is still to be converted (not compiled until then)
22 - Some parts are still missing in this conversion (such as the
23 platform and deployment sections) while others were blindly converted
24 and would need more love (such as the configuration flags).
25 Things will certainly further improve in the future releases.
28 - Fully reimplement auto_restart mechanism that was utterly broken
29 (fix #22, #131 and #281 that were all very related)
30 - Implement semaphores (fix #303 and #291)
31 - When creating an actor from a function and its parameters,
32 move-only parameters are not allowed anymore, as it would prevent
33 the actor to be restartable if its parameters are consumed on run.
34 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
35 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
36 IO operations that can be started, waited for, or canceled as a regular
38 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
42 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
43 - New host variable: 'core_count'
46 - Remove xbt_os_thread_specific features
47 - Remove portability wrapper to condition variables
48 - Remove xbt_os_thread_yield()
51 - MPICH collective selector now mimics MPICH 3.3b
52 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
53 - MPI_Init(NULL,NULL) is now allowed
54 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
55 - Command-line arguments should now be handled properly in Fortran simulations
56 - MPI Attributes and related callbacks should work in Fortran
57 - Apps using C + Fortran should now work
58 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
59 - Support MPI_Error_String
62 - Due to an internal bug, Msg.run() must now be your last line.
63 We hope to fix it in a future release, and we are sorry for the inconvenience.
64 - Expose host load plugin (i.e. loadInit, getCurrentLoad, getComputedFlops,getAvgLoad)
67 - #22: Process autorestart seem to only work with CAS01 cpus
68 - #56: Feature request: dynamic SMPI replay
69 - #93: simgrid should not eat --help
70 - #111: How to change the loopback link characteristics?
71 - #116: Platform cluster radical
72 - #117: Platform cluster/host order
73 - #129: Replay_multiple on steroid
74 - #131: Java : autostart does not work
75 - #143: Setting a breakpoint at a given time
76 - #235: Network energy models should be integrated
77 - #264: Add ptask L07 resource tracing
78 - #271: Dynamic and manual replay of SMPI traces
79 - #279: Breakpoints option uses time deltas instead of absolute time
80 - #281: Daemonized actor no longer daemonized after an auto-restart
81 - #282: TIT convertor erase traces when using absolute path in the trace list file
82 - #285: segfault when a s4u actor kills itself with msg process tracing activated
83 - #286: Pajé traces are not exposing the number of cores
84 - #287: Command-line parsing should stop when encountering '--'
85 - #288: MPI_Init(NULL, NULL)
86 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
87 - #290: Method to check if a mailbox has a message ready to be consumed
88 - #291: [s4u] semaphores do not seem available yet
89 - #294: Alltoall collective implementation of mpich
90 - #295: Starting a Host as OFF is not supported
91 - #296: DTD too permissive
92 - #299: Add a s4u barrier example
94 ----------------------------------------------------------------------------
96 SimGrid (3.20) Released June 24. 2018
98 The proxy snake_case() release.
101 - The whole API is now using snake_case() instead of mixing it with camelCase.
102 Compatibility wrappers in place for one year.
103 - We took the opportunity to increase the consistency of the API.
104 Please report any remaining glitches.
105 - New function simgrid::s4u::Host::get_actor_count:
106 Returns the number of actors running on a specific host.
109 - SMPI is now tested with ~45 proxy apps from various sources, with none or
110 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
111 - Replay: The replay file has been re-written in C++.
112 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
113 supported. They are stored in the trace and used when replayed.
114 - Basic support of MPI_Cancel. Robustness not guaranteed.
115 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
116 (only for MPI_COMM_TYPE_SHARED).
117 - New option: smpi/privatize-libs, to add external shared libs to be privatized
118 by SMPI. They will be copied locally and loaded separately by each process.
119 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
120 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
121 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
122 with internal index optind with multiple processes. Only works if MPI_Init has
124 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
125 syntax without preprocessing tricks.
128 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
129 behavior tracing is no longer limited to MSG
130 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
131 this means that traces now need to be updated or re-obtained; both Irecv
132 and Isend lines in your traces have now in total 5 mandatory fields:
133 <rankid> <command> <to/from rankid> <tag> <size>
134 To update your traces, it suffices to add a 0 for the tag here.
135 - TIT now also supports waiting for a distinct request via MPI_Wait.
136 Wait/Test now wait for a specific request, not just the last one that was
137 issued. This unfortunately means another update, because we need to
138 identify which request you want to wait for. We do this via the
139 triplet (sender, receiver, tag), which needs to be added:
140 <rankid> <command> <sender> <receiver> <tag>
141 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
142 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
145 - The deprecation of MSG is ongoing (but this should not impact you).
146 Many MSG functions are now simple wrappers on the C API of S4U. If
147 you wish to convert your code to S4U, find the S4U counterparts of
148 your MSG calls in src/msg/msg_legacy.cpp.
149 - MSG can still be used, but won't evolve anymore.
152 - Allow to run the Link energy plugin from the command line with
153 --cfg=plugin:link_energy
154 - Rename Energy plugin into host_energy
155 - Rename Load plugin into host_load
158 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
159 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
160 - Improve performance when handling timeouts of simix synchros.
163 - Config: the C API is now deprecated (will be removed in 3.23), and
164 the C++ API has been slightly improved.
167 - Fix several build issues on OSX.
168 - Move simgrid_config.h to simgrid/config.h (old header still working)
171 - #143: Setting a breakpoint at a given time
172 - #150: Inconsistent event names in SMPI replay
173 - #258: daemonized actors hang after all non-daemonized actors have completed
174 - #267: Linker error on unit_tmgr
175 - #269: SMPI: tracing of MPI_Wait/all/any broken
176 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
178 ----------------------------------------------------------------------------
180 SimGrid (3.19.1) Released March 22 2018
182 The Fixed ABI Release.
184 As you may know, we are currently refactoring SimGrid in deep.
185 Upcoming SimGrid4 will be really different from SimGrid3: modular,
186 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
188 Our goal is to smooth this transition, with backward compatibility
189 and automatic update paths, while still progressing toward SimGrid4.
191 SimGrid remains open during works: The last pure SimGrid3 release was
192 v3.12 while all subsequent versions are usable alpha versions of
193 SimGrid4: Existing interfaces remain unchanged, but the new S4U
194 interface is budding and the internals are deeply reorganized.
196 Since 2015, we work hard to reduce the changes to public APIs. When
197 we need to rename a public library symbol in S4U, we let your
198 compiler issue an explicative warning when you use the deprecated
199 function. These messages remain for four releases, i.e. for one full
200 year, before turning into an error. Starting with v3.15, your can
201 also adapt to API changes with the SIMGRID_VERSION macro, that is
202 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
204 Starting with this v3.19.1, our commitment to reduce the changes to
205 the public interfaces is extended from the API to the ABI: a program
206 using only MSG or SimDag and compiled against a given version of
207 simgrid can probably be used with a later version of SimGrid without
208 recompilation. We will do our best... but don't expect too much of
209 it, that's a really difficult goal during such profund refactoring.
211 The difference between v3.19 and v3.19.1 is that the former was
212 accidentally breaking the ABI of MSG, while the later is restoring
215 S4U and kernel APIs will still evolve until SimGrid4, with one-year
216 deprecation warnings as currently. In fact, cleaning up these
217 interfaces and converting them to snake_case() is one release goal of
218 v3.20. But don't worry, we are working to smooth this upgrade path.
220 In summary, new projects should start with S4U to benefit of the
221 future, but old MSG projects should still be usable with no change.
223 ----------------------------------------------------------------------------
225 SimGrid (3.19) Released March 20 2018
227 The Moscovitly-cold Spring release.
229 SMPI user-visible improvements:
230 - Support of MPI_Type_create_subarray
231 - Unimplemented MPI calls now abort (instead of issuing a warning)
232 Some harmless calls around errhandlers still only issue a warning.
233 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
234 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
235 is now possible (mmap privatization was refactored to this end).
236 Adding new SMPI instances should also be technically doable now.
238 SMPI internal cleanups:
239 - Removed index notion from SMPI -- no more "getPid() - 1"!
240 In tracing files, actors are called rank-<PROCESS-ID> but these are
241 NOT the real ranks (tracing several communicators is not supported yet).
242 - Deprecate smpi_process_index() (will be removed in 3.22)
243 - Better testing on classical proxy apps, through an external project:
244 https://github.com/simgrid/SMPI-proxy-apps/
247 - Use a graphical TOC to make it easier to find the documentation you need
248 - Revamp the MSG tutorial
251 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
252 - Remove parameter of MSG_process_killall().
253 Resetting the PID was bogus anyway (several actors could have the same PID).
256 - Execution->setHost() can be called after start(), to migrate it.
257 - Comm::test_any() is now implemented.
258 - s4u::Actor now has onCreation() and onDestruction() signals.
259 - Install some kernel header files for the users' plugins and more.
260 Warning, their API is really not stable yet.
261 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
262 include/simgrid/kernel/routing/*Zone.hpp
265 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
267 Build System and other cleanups:
268 - Remove unused run-time parameter "tracing/onelink-only".
269 - SimGrid now works with intel compilers (MC still buggy)
270 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
271 - tesh kills subprocesses on timeout (but not on windows)
274 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
277 - #194: Feature request: simgrid::s4u::Comm::test_any()
278 - #245: migrating an actor does not migrate its execution
279 - #253: Feature Request: expose clusters as objects
280 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
281 - #255: Tesh broken on Windows
282 - #256: Modernize FindSimGrid.cmake
283 - #257: Fix (ab)use of CMake install
285 ----------------------------------------------------------------------------
287 SimGrid (3.18) Released December 24 2017
289 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
291 MSG IS NOW DEPRECATED. Please use S4U for new projects.
292 - Support for MSG should not be removed from SimGrid before 2020,
293 but future evolutions will be limited to the integration of
294 user-provided patches.
297 - s4u::onDeadlock() signal, run before stopping the simulation.
298 - s4u::Exec: asynchronous executions (abstraction of background
300 - s4u::Host->getLoad() returns the achieved speed in flops/s
303 - Link::name() is deprecated and will be removed in v3.21, use
304 Link::getCname() instead.
305 - Mailbox::getName() changed to return a std::string, use
306 Mailbox::getCname() instead to get a char*.
307 - Storage::getName() changed to return a std::string, use
308 Storage::getCname() instead to get a char*.
309 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
310 to have both allocation and deallocation of the map in user space.
313 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
314 The amount of flops remaining to do is a vector, not a scalar.
315 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
316 name implies on both sequential and parallel tasks.
317 - Both changes fix GitHub's #223 using PR #237 as a basis.
318 Thanks Michael Mercier.
319 - Most examples were converted to S4U and hidden elsewhere at tests.
320 Rationale: we still want MSG to work; we want newcomers to use S4U.
323 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
324 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
327 - Switch to the faster dlopen privatization mechanism by default
328 - Documentation improvments
331 - Live migration is getting moved to a plugin. Dirty page tracking is
332 the first part of this plugin. This imply that VM migration is now
333 only possible if one this function is called:
334 - C/MSG: MSG_vm_live_migration_plugin_init()
335 - C/C++: sg_vm_live_migration_plugin_init()
336 - Java: Msg.liveMigrationInit()
337 For C and C++, "simgrid/plugins/live_migration.h" also has to be
341 - Define class simgrid::xbt::Path to manage file names.
342 - Removed unused functions:
343 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
344 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
345 xbt_graph_export_graphviz()
346 - xbt/str.h: xbt_str_join()
347 - Remove unused datatypes:
348 - xbt/heap.h: use std::priority_queue or boost::heap instead
349 - xbt/swag.h: use boost::intrusive::list instead
352 - New link_energy plugin for the consumption of the links.
353 - All of the operations on files and storage contents have been
354 packaged into a plugin (src/plugins/file_system). The current
355 public interface can be found in
356 include/simgrid/plugins/file_system.h
357 To use these functions you now have to initialize the plugin by
358 calling MSG_storage_file_system_init() just after calling
359 MSG_init() or sg_storage_file_system_init() just after creating
363 - Remove the undocumented/untested tag <include>
366 - Remove viva specific tracing as the tool is no longer maintained
369 - #248: Kill of finished processes leads to segfault
370 - #240: xbt_cond_wait_timeout should gracefully return for C apps
371 - #239: please implement signal s4u::onDeadlock()
372 - #230: segfaults when exit() without run()
373 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
374 - #223: MSG_task_get_flops_amount() not working with parallel tasks
375 - #222: Actor::kill() doesn't really kill and segfaults
376 - #221: odd LMM warning when killing an actor
377 - #120: Memory leak when the processes are forcefully killed
379 ----------------------------------------------------------------------------
381 SimGrid (3.17) Released October 8 2017
383 The Drained Leaks release: (almost) no known leaks despite the tests.
385 Even very long simulations will preserve your memory: our testsuite
386 stresses SimGrid for over 45mn on fast machines for a coverage of
387 over 80%, and there is only one single known leak, of about 4kb.
391 - Comm.detach(): start and forget about asynchronous emission. A cleanup
392 handler may be given to free resources if the comm cannot be completed.
393 - this_actor::send(mailbox) is now mailbox->put()
394 - New: simgrid::s4u::Comm::wait_all()
395 - New: Host.execute() for remote executions.
398 - Unused option network/sender-gap is removed.
401 - Deprecate MSG_task_isend_with_matching(): This unused feature
402 really complicates our internals. Will be removed in v3.20.
405 - Improved context termination. It is now possible to free resources, even
406 when a process is forcibly killed.
407 - ContextBoost: add support for Boost versions above 1.61.
410 - Bring back run-time option --cfg=exception/cutpath to remove exception
412 - Removed unused functions:
413 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
414 xbt_str_rtrim(), xbt_str_trim().
415 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
418 - Removed header files obsolete since SimGrid 3.12:
419 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
421 - #3: SD_exit should be made optional
422 - #120: Memory leak when the processes are forcefully killed
423 - #159: Threading test regression in Actor refcounting
424 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
425 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
426 - #186: Actor::killAll() segfaults if some process is blocked on wait()
427 - #191: VM migration and pstate
428 - #192: Updating the energy consumptions of all hosts crashes with VMs
429 - #195: All actors have PID=0 in the logs
430 - #204: Sometimes segfault with thread contexts and mmap privatization
431 - #222: Actor::kill() doesn't really kill and segfaults
432 - #225: Actor::kill() doesn't really kill when victims are doing a join()
434 ----------------------------------------------------------------------------
436 SimGrid (3.16) Released June 22. 2017.
438 The Blooming Spring Release: developments are budding.
441 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
442 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
443 - Java JAR file should be finally fully working on Mac OSX too.
445 XML platforms: Switch to platform v4.1 format.
446 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
447 - <zone> can be used as a synonym for the now deprecated <as>
448 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
449 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
450 - <actor> can be used as a synonym for the now deprecated <process>
451 - state_file and avail_file periodicity is now easier to express
452 (check the documentation) the old behavior should still work.
453 * Storage: not backward compatible, but it seems that nobody ever used it
454 - Remove attribute 'content_type' of <storage_type>: was never used
455 - Make attribute 'model' of <storage_type> optional: for future usage
456 - Remove Bconnection model property: was never (in)validated. Replaced by
457 the maximum of the read and write bandwidth as a resource constraint.
460 - New and Backwards Compatibility break:
461 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
462 When one wants to get the list of tasks whose states have changed during a
463 simulation round, s/he has to allocate and free a dynar and use it as
464 argument to this function. The former SD_simulate (double how_long)
468 - Allow multicore VMs, with the correct sharing (unless you overcommit)
469 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
472 - New function to update the consumption of all hosts at once.
473 - Fix the model for multi-core hosts, linear in the amount of busy
474 cores with an abnormality for pIdle (see paper).
475 BUG: Still not satisfactory for ptask on mono-cores.
478 - The netzone are now available from the MSG API.
479 The old names still work, but are now deprecated.
482 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
483 - New option: smpi/keep-temps to not cleanup temp files
484 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
485 mallocs mode is used, allows to change the size of the fake file used
486 (default 1MB), to potentially limit the number of mappings for large runs.
487 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
488 - Fortran ifort and flang compilers support
489 - New RMA calls supported (experimental):
490 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
491 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
492 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
493 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
494 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
495 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
498 - New: this_actor::isMaestro() returns whether we are in kernel mode.
499 - Behind the scene, ActivityImpl are now automatically refcounted.
500 Removing this nasty bottleneck will greatly simplify our internals.
503 - Replay: New function xbt_replay_action_get():
504 Retrieve the function previously associated to an event type.
505 - DROPPED FUNCTION: xbt_str_varsubst()
506 - DROPPED MODULE: strbuff. We don't need it anymore.
507 - DROPPED MODULE: matrix. We don't need it anymore.
508 - DROPPED MODULE: lib. We don't need it anymore.
510 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
512 SimGrid (3.15) Released March 22. 2017
514 The Spring Release: continuous integration servers become green
515 We fixed even the transient bugs on all target architectures:
516 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
517 Windows, FreeBSD, NetBSD.
519 - Rename NetCards to NetPoints.
520 This was intended to help NS3 users, but that's not a netcard.
521 That's a point in the routing algorithm, let's avoid wrong simplifications.
524 - New: MSG_process_yield(). Stop and yield to other processes.
525 - New: MSG_process_daemon(). Daemon processes are automatically killed
526 when the last non-daemon process terminates
527 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
528 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
529 to make room for the upcoming network energy plugin.
530 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
533 - Ensure that an actor can kill itself with Process::exit()
534 - Kill the obscure NativeException. Nobody want to survive the issues
535 it denotes, so use JniException that is a RuntimeException (not to
536 be caught explicitly).
537 - Partial bug fix in initialization. SimGrid flags on command line were
538 consumed at C level but stayed in the original Java String[] args.
539 This could mess users' args[i] if SG flags were not put at the end of
541 The SimGrid flags are now removed from the Java arguments. However,
542 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
543 if args.length is greater than the number of YOUR OWN ARGUMENTS.
544 It might be if you have --log or --cfg flags in the command line.
545 - Fix numerous memleaks all around the place. In particular, around VMs.
549 - s4u::onPlatformCreated: right before the simulation starts
550 - s4u::onSimulationEnd: right after the main simulation loop
551 - s4u::onTimeAdvance: right after a clock change (time is discrete)
552 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
553 event from the availability_file changes the avail speed.
554 - Links are now usable from s4u
555 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
556 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
557 - New examples: The conversion of MSG examples to S4U has begun
558 - Actors: create, kill, migration, and suspend
559 - Applications: master-worker and token-ring
560 - Action replay: communications and storage
561 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
564 - Backwards Compatibility breaks
565 - The SD_route_get_list and SD_route_get_size functions have been removed.
566 They are replaced by the unique
567 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
568 The route from src to dst is built in the links dynar whose size can
569 be obtained with xbt_dynar_length.
570 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
571 sg_host_route_bandwidth and sg_host_route_latency.
572 Macros ensure the backwards compatibility, but you should fix your code
575 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
576 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
577 - Bump our claim of support from MPI 1.1 to MPI 2.2.
578 We don't support 100% of it, but it should be enough. Contact us if not.
579 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
580 naive and probably inaccurate)
581 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
582 mapped onto a unique small file using some system magic.
583 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
584 - Fortran: cleanups, fixes, support of user-added operations
585 - RMA: MPI_Accumulate are applied in correct order.
586 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
587 - Better support for MPI_IN_PLACE
588 - Support for MPI_Win attrs and keyvals.
589 - Support MPI_Comm_group_create, MPI_Type_size_x
590 - MPI ops should be properly applied to all allowed types
591 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
592 algorithm by a non failing one from mpich.
595 - Kill the fifo data container: we don't use it anymore.
598 - A <cluster> can now be created with different speed values to represent pstates.
599 see examples/platforms/energy_cluster.xml for an example.
601 SimGrid (3.14.159) Released December 28. 2016
603 The previous release was only a crude approximation of a Pi release;
604 this one should be a bit better.
606 - Revert a last minute change that broke on Mac OSX.
607 - Fix the changelog and NEWS files.
608 - Improve the documentation, in particular of the routing module.
609 - Integrate some patches flying around in Debian.
611 SimGrid (3.14) Released December 24. 2016
614 * The whole documentation was reworked and reorganized. There is
615 still much room for improvement here, but we're on it.
618 * We now need python3 for our testing infrastructure.
619 * Model checking needs libevent
622 * Now works on FreeBSD
625 * Privatization now works on FreeBSD
627 * Privatization is now activated by default in smpirun
628 You can turn it off with -no-privatize if something goes wrong.
630 * Call-location tracing for SMPI.
631 You can add the exact location (filename / linenumber) of an MPI call to
632 your trace files and slow down or speed up the simulation between two
633 consecutive calls by using an adjustment file (see the documentation).
635 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
636 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
637 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
638 values were ignored in some cases. The timings of these functions can now
639 be significantly different.
641 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
642 smpi/running-power is renamed to smpi/host-speed
644 * smpi/grow-injected-times option to enable or disable multiplication of the
645 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
646 make simulation less precise (but also much faster).
648 * smpirun script should be (much) faster for large deployments.
650 * SMPI tracing: fixed issue with poor matching of send/receives.
652 * Replay: Fix broken waitall
654 New functions and features
655 * MSG_parallel_task_execute_with_timeout, to timeout computations.
657 Dropped / renamed functions and features
658 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
659 - MSG_mailbox_is_empty() -> Mailbox::empty()
660 - MSG_mailbox_front() -> Mailbox::front()
661 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
662 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
663 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
664 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
665 * Don't pass the free_f parameter to property related functions:
666 - MSG_host_set_property_value()
667 - MSG_as_router_set_property_value()
668 - MSG_storage_set_property_value()
669 * VM properties. Since msg_vm_t are msg_host_t, just use
670 MSG_host_get_property_value() and friends
671 * VM I/O related things:
672 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
673 - Unimplemented save/restore methods
674 * MSG_as_router_get_property_value() was redundent with
675 MSG_environment_as_get_property_value().
676 - Removed MSG_as_router_*propert*() functions
677 - Added MSG_environment_as_set_property_value() for consistency
678 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
679 Well, they are still there for now, but deprecated with a warning.
680 Please switch to xbt_dict_new_homogeneous() before this is removed
682 * Task affinity. Its intended behavior (that was very badly tested
683 and probably not really working) was deceiving what most users
684 would have hoped here.
685 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
686 * Option network/coordinates is now useless and should be dropped.
689 * Added option storage/max_file_descriptors to allow more than 1024 files opened
692 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
693 * Added Dragonfly topology support
696 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
697 EiBps, EBps, Eibps, Ebps for bandwidth.
698 They may become useful to some lucky ones.
701 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
702 * ex/app_tokenring: new example, very similar to the MSG Token Ring
703 * ex/async_waitAll: new example, on asynchronous communications
706 * Memory usage should be decreased for simulations with a large number
707 of processes. This also helps for SMPI.
709 SimGrid (3.13) stable; urgency=low
711 The Half Release, a.k.a. the Zealous Easter Trim.
713 - We removed half of the lines, that were mostly experimental cruft.
714 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
715 (not counting blanks and comments -- according to openhub.net)
716 - The internals are now compiled in C++ (and will soon be clean C++)
717 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
718 59 klines of C, etc. We added only 29 klines of C++ in replacement.
720 * Backwards Compatibility breaks
721 - Removed Lua simulation bindings (switch to C or Java for that).
722 Lua can still be used to describe platforms
723 - Removed Java kernel plug-ins.
724 Will be reintroduced after the ongoing major internals reorg.
726 - the following functions were removed.
727 They were too specific and should be reimplemented in a generic
728 way, with filter function.
729 - MSG_task_listen_from_host
730 - MSG_mailbox_get_count_host_waiting_tasks
731 - MSG_mailbox_put_with_timeout was removed.
732 Please use MSG_task_send_with_timeout instead.
734 - the SD_application_reinit function was removed. It has been a noop for a while.
735 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
736 replaced by a more flexible mechanism.
737 - The following functions thus do not exist anymore
738 - SD_workstation_get_access_mode
739 - SD_workstation_set_access_mode
740 - SD_workstation_get_current_task
741 - Basic estimation functions have been removed but can easily be replaced
742 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
743 - SD_workstation_get_computation_time => amount / sg_host_speed()
745 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
746 Use VM.getSpeed()*load/100 for the legacy behavior.
748 - option enable_tracing was removed. It was not doing anything for a while.
749 - In the ModelChecker:
750 - the model-checker now ptraces the model-checked process which means
751 you cannot use a debugger on the latter anymore (we might make this
752 optional in the feature);
753 - removed soft-dirty page tracking;
754 - remove model-checked side snapshot management,
755 MC_snapshot() and MC_compare_snapshot();
756 - keep the MC_cut() function as a stub (it was not really working
757 in the previous release).
760 * All options are consistently in kebab-case. Old names are kept as alias.
763 * Switch to platform v4 format.
764 - Rename from 'power' to 'speed' the attributes describing the amount of
765 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
766 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
767 - In <host> and <link>, attributes availability and state are gone.
768 It was redundent with state and availability traces, and with peak values.
769 - In <cluster>, attributes availability_file and state_file are gone.
770 It was too complex and unused.
771 - Kill <gpu>. Was not doing anything.
772 - The DOCTYPE points to the right URL:
773 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
774 (the file at this address now documents the changelog since its v1)
775 - A warning is emitted for unit-less values (they are still accepted).
776 - speed. Default: 'f' or 'flops'. Also defined:
777 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
778 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
779 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
780 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
781 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
782 - latency. Default: 's' second. Also defined:
783 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
785 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
786 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
789 * s4u::Host is now the preferred public interface to the Host features.
790 sg_host_* functions are C bindings to the exact same behavior
791 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
794 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
795 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
796 They were deprecated since a few years, and probably did not even compile anymore.
799 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
800 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
801 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
802 encouraged to update their codes. A list of the modified functions can be found at the end of
803 include/simgrid/simdag.h
806 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
807 the lambda or closure passed as a parameter will run in kernel mode.
808 All the callback functions should be rewritten to that interface at some point.
811 * Reorganizing and cleaning the internals all around the place.
814 * Remove old default barrier/bcast buggy algorithms (see #18407)
815 * Various bug fixes to handle more codes
816 * Remove the need for the --foreground option of smpirun (it is still
817 accepted for backward compatibility).
820 * Kill the setset data container: MC don't use it anymore.
821 * Kill the queue data container: it made more sense with GRAS.
822 * Kill the xbt_peer_t data type: it's useless without GRAS.
823 * Kill rm_cb feature of config sets: it was never useful.
824 * Kill graphxml parsing feature. It was not used.
825 * Kill the deprecated code protected by XBT_USE_DEPRECATED
827 - xbt_dynar_sort_strings(), when the content is char*
828 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
829 They throw exceptions on invalid input;
830 * C++ support for declaring CLI flags (simgrid::config::Flag);
831 * class for abstracting different signal backends (simgrid::xbt::signal).
832 with no external dependencies (we need very simple signals).
835 * refactoring and cleanup of the code;
836 * ongoing process to cleanly separate the model-checking algorithms
837 from the code model-checking support.
839 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
841 SimGrid (3.12) stable; urgency=low
843 The Facelift Release.
846 * Require g++ v4.7 at least to not speak prehistorical C++.
847 * Require Boost 1.48 (for signal2 component).
848 * Java must be version 7 at least when activated.
849 * Builds on Windows again (including Java bindings).
850 * Tracing is now always enabled (no way to turn it out)
851 * Remove GTNetS. It was not working anyway.
852 * Various cleanups in the cmake scripts.
853 * Move headers around to sort them out on installed systems:
854 - instr/instr.h -> simgrid/instr.h
855 - instr/jedule/* -> simgrid/jedule
856 - simdag/datatypes.h was removed
857 - simdag/simdag.h -> simgrid/simdag.h
858 - msg/datatypes.h was removed
859 - msg/msg.h -> simgrid/msg.h
862 * Interface improvement:
863 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
864 - Sanitize the interface in MSG_task_ module:
865 - Merge two functions that were close enough but misleading:
866 set_compute_duration(t) -> set_flops_amount(t)
867 get_remaining_computation(t) -> get_flops_amount(t)
868 - set_data_size(t) -> set_bytes_amount(t)
869 get_data_size(t) -> get_bytes_amount(t)
870 - Massive cleanups in the functions related to the energy
871 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
872 - New: MSG_host_get_pstate()
874 - msg/energy/onoff: switching hosts on and off
876 * Interface improvement:
877 - Rename Host.isAvail() to Host.isOn()
878 - Rename Process.currentProcess() to Process.getCurrentProcess()
879 - Rename Task.setDataSize() to Task.setBytesAmount()
880 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
882 - #18874: Actually allows the GC to reclaim tasks
886 - SIMIX_process_throw: raises an exception in a remote process
887 * Refactoring: Separate sub-modules
888 - libsmx: the public interface, as libc in a real system
889 - popping: the strange dance that converts a user request into a kernel handling
890 - smx_context_*: the virtualization mechanisms that embed the user code
891 - smx_*: the handling of each simcalls
893 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
894 * Rename smx_action_t into smx_synchro_t, making explicit that these
895 things are used to synchronize processes with their environment.
896 For example, a communication is a sort of synchronization involving
897 the communicating processes (that may block until the exchange) and
898 the platform. The same can be said from computations, etc.
900 - #18888: segfault when a process ends before its kill_time
904 - 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.
905 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
906 - MPI_Type_set_name, MPI_Type_get_name
907 - MPI_*_c2f and MPI_*_f2c functions
908 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
909 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
910 - Activate a lot of new tests from the mpich 3 testsuite
912 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
913 - InfiniBand network model added: Based on the works of Jerome Vienne
914 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
915 - When smpi/display_timing is set, also display global simulation time and application times
916 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
917 * Collective communications
918 - 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
919 - 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/).
920 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
922 - Replay now uses algorithms from wanted collective selector
923 - Replay can be used with SMP-aware algorithms
924 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
925 - 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.
927 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
928 - Correctly use loopback on fat-tree clusters
929 - Asynchronous small messages shouldn't trigger deadlocks anymore
930 * Energy/DVFS cleanup and improvement
931 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
932 - new: smpi_get_host_pstate()
935 - "Full" network optimization flag was broken since Surf++
936 - Better handling of precision flags in maxmin
937 - Fix bug causing sometimes "Impossible" errors
938 - Properly pass cluster properties to included hosts
939 * Improvement of the Energy plugin.
940 - Always update the consumption before returning that value
941 - New property: watt_off to denote the dissipation when the host is off
942 - New functions getWattMinAt and getWattMaxAt to retrieve the
943 dissipation of pstates that we are not currently at.
944 * Java: class NetworkLink renamed to Link
945 * New function: simcall_process_get_kill_time()
946 * Massive rename s/workstation/host/
947 - That's intrusive, but that's good for the project consistency. Sorry.
948 - Change config option "workstation/model" into "host/model"
951 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
952 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
953 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
954 - xbt_dirname and xbt_basename for non-POSIX systems
956 * The model checker now runs as a separate process.
957 * The model checker runs must now be launched with the new simgrid-mc program.
958 * Record/Replay: the MC can display a textual representation of a path in the
959 execution graph. It can then be replayed outside of the model checker.
961 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
963 SimGrid (3.11) stable; urgency=low
968 * Normalizing pointers addresses tool for better diff between logs
971 * Add cloud examples using new VMs
972 - examples/msg/cloud/two_tasks_vm.tesh
973 - examples/msg/cloud/simple_vm.tesh
974 - examples/java/cloud/cloud.tesh
975 - examples/java/cloud/migration/migration.tesh
976 * Add java surf examples:
977 - examples/java/surfPlugin/surf_plugin.tesh
978 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
979 - examples/java/surfCpuModel/surf_cpu_model.tesh
980 * Add SMPI+MSG example:
981 - examples/smpi/smpi_msg_masterslave/
986 - msg task destroy cancel test
987 - msg_host on/off test
988 * Move all tests in testsuite to teshsuite (adding tesh files)
989 * Restructure teshsuites
990 - one folder for each kind of test
991 * Restructure AddTests.cmake
993 - structure the order of tests (with sections)
996 * Add virtual machine
997 - creation of a VM on a PM
998 - migration of a VM from a PM to another PM
1000 - MSG_process_join(msg_process_t process, double timeout)
1001 - msg_bar_t MSG_barrier_init(unsigned int count)
1002 - int MSG_barrier_wait(msg_bar_t barrier)
1003 - void MSG_barrier_destroy(msg_bar_t barrier)
1004 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1005 * New option "msg/debug_multiple_use" to help debugging when a task is used
1008 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1010 - Increase code coverage in test suites
1014 * Protect context stacks against stack overflow. The number of protected memory
1015 pages allocated on the top of each stack (1 by default) can be configured
1016 with the new command line option --cfg=contexts/guard_size:<value>.
1017 * Simcalls are now generated by a python script that
1018 - generates files included by SimGrid
1019 - checks that all the functions exist, and proposes prototypes
1021 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1022 host_set_data, host_get_data
1024 - simcall_process_join(smx_process_t process, double timeout)
1025 * Fix bug where sleeping processing could not be suspended.
1028 * Translate surf models from C to C++
1029 - Generic classes for all models: Model, Resource, Action
1030 - A generic interface for each kind of model (CPU, Network, Storage
1031 Workstation, WorkstationVM)
1033 * Translate surf routings from C to C++
1034 * Add callbacks using sigc++ or boost::signals2
1035 - Add callback functions for resource creation/destruction
1036 - Add callback functions for action state change
1037 - Handle Energy as a plugin
1038 * Replace swag by boost::intrusive
1039 * Add new routing models for clusters. For documentation, see
1040 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1041 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1042 parameters for cluster tag
1043 - Fat trees, with topology="FAT_TREE" and
1044 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1045 - see examples/platforms/torus_cluster.xml and
1046 examples/platforms/fat_tree_cluster.xml
1047 * More documentation
1050 * Hostfiles support host:nb_processes construct to deploy several processes on
1052 * Collective communication algorithms should not crash if used with
1053 improper number of nodes and report the error.
1054 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1055 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1056 MPI_Dims_create, MPI_Cart_sub are supported.
1057 * New interface to use SMPI programmatically (still depends on MSG for
1058 some parts, see examples/smpi/smpi_msg_masterslave):
1059 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1063 * Global variables privatization in MPI executables is now performed at runtime
1064 with the option smpi/privatize_global_variables (default:no).
1065 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1066 dynamic libraries loaded by the application are not privatized (static
1067 linking with these libraries is advised in this case)
1070 * Options defined in XML work correctly now.
1073 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1074 copied into simgrid.jar or not (ON by default). Use this option if you want
1075 to reduce the size of the installed simgrid.jar, *and* the native libraries
1076 are kept installed elsewhere.
1077 * Surf binding with SWIG (code generated in maintainer mode only):
1078 - plugin to handle callbacks
1079 - CPU model only for the moment
1082 * Supernovae build mode is definitively removed. It was used to improve
1083 inlining and inter-module optimizations. It is nowadays superseded by
1084 link-time optimizations commonly available in compilers.
1085 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1086 * Add boost dependency for surf++
1087 * Add new macro for tests
1088 - ADD_TESH(name <tesh_args>)
1089 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1093 - xbt_bar_t XBT_barrier_init(unsigned int count)
1094 - int XBT_barrier_wait(xbt_bar_t barrier)
1095 - void XBT_barrier_destroy(xbt_bar_t barrier)
1096 * Make the xbt_os_time module public
1098 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1100 SimGrid (3.10) stable; urgency=low
1102 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1105 * Reintegrate Java to the main archive as desynchronizing these
1106 package is not acceptable anymore (Java is now considered stable)
1107 * Add explicit synchronization facilities through semaphores
1108 * Bug fix: Task.setDataSize() only changed the C world, not the value
1109 cached in the Java world
1112 * Dramatically change the way files are handled. API and internals changed, but
1113 this part of MSG was not considered as production grade either.
1114 * Add explicit synchronization facilities through semaphores
1115 * Add a new function MSG_host_get_process_list()
1116 * Preliminary DVFS support (see examples/msg/energy/ for details)
1119 * SMPI is now included directly in the libsimgrid as the windows
1120 linker doesn't force us on splitting it anymore.
1121 * Improvements of the SMPI replay tool:
1122 - Most of the collective communications are now rooted in the same process as
1123 in the original application.
1124 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1125 used until now). Multiple data types can now be used in a trace.
1126 - The replay tool now supports traces produce either by TAU or a modified
1128 - Bug Fix: the compute part of the reduce action is now taken into account.
1129 - Gatherv collective is now supported
1130 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1131 of smpirun outputs time independent traces for the current run. One file
1132 is created per process. If too many processes are simulated, this behavior
1133 can be changed to one file for all processes by using the
1134 tracing/smpi/format/ti_one_file flag
1135 * smpirun generates the host file if needed (with given host count and platform)
1136 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1137 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1138 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1139 --cfg=smpi/coll_selector:(mpich/ompi)
1140 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1142 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1143 happening inside a collective SMPI call.
1144 * Fix the behavior of complex data types handling.
1145 * Make MPI_Wtime another synchronization point to take computations into
1147 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1148 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1149 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1150 (no privatization of global variables yet)
1151 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1154 * Verification of liveness properties is now available for SMPI applications
1155 (in addition to MSG applications)
1156 * Bugged examples using SMPI in examples/smpi/mc/
1157 * Add --cfg=model-check/visited option. Allows the verification of infinite
1158 programs. Detection of loops in the execution thanks to the system state
1159 comparison and reduction of the state space to explore. Can be combined with
1160 DPOR for safety properties.
1163 * Allow to change SimGrid configuration (see --help) within the code
1164 thanks to SD_config() as it can be done in MSG.
1165 * Add a new function SD_task_set_amount() upon user request.
1168 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1169 * Remove rule based routing (no more PCRE dependency)
1170 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1171 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1172 * Add a loopback_bw and loopback_lat options to cluster tag.
1173 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1174 routing. Now peers are encapsulated in an AS and have their own private
1175 router but this is transparent.
1178 * Our own implementation of getline is renamed xbt_getline, and gets
1179 used even if the OS provide a getline(). This should reduce the
1180 configuration complexity by using the same code on all platforms.
1181 * New type: xbt_cfg_elm_boolean.
1182 * Allow to use yes/no for boolean configuration options in the command line.
1183 * Allow to disable SimGrid cleanups at exit from command line option.
1184 There are situations where one may want a simulation to end with an exit.
1185 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1186 annoying when scripting around the simulator. Adding a
1187 --cfg=clean_atexit:no allows to circumvent this issue.
1190 * Lots of memory leaks were corrected in this release.
1191 There is no known memory leaks anymore, in all of our 600+ tests.
1192 * New command line option --version, to get SimGrid version information.
1193 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1195 * Supernovae builds are deprecated, and expected to be removed in the next
1198 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1200 SimGrid (3.9) stable; urgency=low
1202 The Grasgory release: GRAS is really dead now.
1204 * Complete overhaul of the internal host structures scheme.
1207 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1208 was considered as experimental and badly maintained since then.
1209 * Keeping it was thus a trap to our potential users, that could take
1210 it instead of MSG or SMPI by mistake despite is pity state.
1211 * GRAS seems to have very few users (if any), and no one volunteered
1212 to maintain it further. It also induces a lot of XBT code (for
1213 portability sake), that must be maintained too.
1214 * For all these reasons, we killed GRAS. If someone wants to revive it
1215 in the future, don't cry, our git history still remembers of GRAS.
1218 * Major overhaul. Merge our documentation again as time proved that
1219 splitting it was really not helping our users.
1220 * Further improve the developer documentation to help newcomers
1221 hacking on SimGrid itself. The user documentation (and in
1222 particular, the beginner documentation) is still in a sorry state.
1225 * Now works on Windows too!
1226 * Much more extensive test suite, from MPICH
1229 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1230 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1231 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1232 * Bug fix: let task be scheduled when the last dependency to be solved is
1233 a control dependency.
1234 * Remove SD_load_environment_script function.
1235 Use the C sg_platf function if you want to declare a platform
1239 * New function: MSG_process_get_number()
1240 * Old function documented: MSG_config()
1241 * Remove MSG_load_platform_script function
1242 Use the C sg_platf function if you want to declare a platform
1246 * Change the default value of the TCP_gamma constant (maximal size of TCP
1247 congestion window) to a more realistic 4MiB value. If you notice changes in
1248 your simulation results, you can fall back to the previous 20k tiny window
1249 by adding --cfg=network/TCP_gamma:20000 on command line.
1250 * (Hopefully) fix a bug wrt periodic availability and state traces
1251 * Bug fix: use default values at start when first event in availability/state
1252 trace is not at time 0.
1255 * remove the "new_" part of function name sg_platf_new_trace_connect
1256 (resulting in sg_platf_trace_connect), since it does not create
1260 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1261 use case where it's really mandatory, and maintaining it was a pain in
1263 * New: xbt_fifo_search(), search an item with a user-provided
1264 comparison function instead of dumb pointer comparison.
1267 * Fix the lua deployment:
1268 Use `simgrid.init_application()` before deployment instead of
1269 `simgrid.msg_register_application()` after.
1272 * Transfer the tracing files into the corresponding modules.
1274 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1276 SimGrid (3.8.1) stable; urgency=low
1278 The "we are told that some people want to also *install* the simgrid
1281 * Add missing file "tesh.1" to the archive.
1283 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1285 SimGrid (3.8) stable; urgency=low
1287 The Psssshiiiit release: SimGrid jumps into the Cloud.
1290 * Add an experimental interface to manipulate VMs. They are mainly
1291 process groups with very few intrinsic semantic, but they should
1292 allow you to build the semantic you want easily.
1293 * New function: MSG_host_set_property_value()
1294 * New function: MSG_process_on_exit(). To clean memory in all cases.
1295 * Bug fixes that made the host (and link) failures unusable.
1296 * Add a way to auto-restart process when the host in which they are
1297 executing comes back (ON_FAILURE="RESTART" on deployment file,
1298 MSG_process_auto_restart_set).
1299 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1300 please stop using the old ones, they are DEPRECATED.
1302 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1303 Please use MSG_init() instead. (reducing the amount of entry
1304 points in the library helps us).
1305 * Make it impossible to link against the wrong version of the lib
1306 * Deprecate MSG_clean(). No need to call it anymore.
1307 * Function MSG_get_host_number() is not deprecated anymore.
1310 * Split the doc into a user guide and a reference guide.
1311 * Start a developper guide to help people hacking on SimGrid.
1314 * Enable tracing by default. This modules rocks you should use it.
1315 * Remove option custom_flags. Now use environment variables CFLAGS
1317 * Use default cmake things to detect lua instead of home grown ones.
1318 * New option "enable_mallocators" to disable mallocators, for debugging
1319 purpose ("on" by default).
1322 * Bug fixes around the resource failures: don't let the processes
1323 survive the host they are running onto.
1324 * Add an interface to auto-restart processes when the host in which they are
1325 executing comes back.
1326 * Ensures that SIMIX_clean is called automatically. It's not part of
1327 the public interface anymore (bindings should be updated).
1330 * Bug fix for when SD_Simulate is called with a positive value: be careful
1331 when comparing doubles. Sometimes they are different for non significant
1333 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1334 parallel task whose initial work is distributed among host according
1335 to the Amdahl's law. Such tasks are created with a parameter alpha
1336 that corresponds to the non-parallelizable part of the computation.
1337 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1338 between two sets of workstations assuming a 1D block distribution (each
1339 workstation owns a similar share of data) on both sides.
1341 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1342 Data redistribution will be automatically scheduled once parent and child
1343 are both scheduled. The filling of computation_amount and
1344 communication_amount structures is now done seamlessly thanks to the chosen
1346 * New function SD_workstation_dump to display various information
1347 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1348 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1349 bandwidth on the route onto which the task is scheduled and the amount of
1351 To divide the nominal bandwidth by 2, the rate then has to be:
1352 rate = bandwidth/(2*amount)
1353 * Compute tasks that have failed can now be rescheduled and executed again
1354 (from their beginning)
1355 * Increasing source code coverage (src/simdag is now covered at 95.8%
1359 * Re-implement time-independent trace replay using SMPI (at the
1360 smpi_smp_* level) instead of MSG. This should replace
1361 examples/msg/actions/actions.c
1362 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1363 hindexed and structs)
1364 * Implement the exchange of non-contiguous data.
1365 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1366 * Correct behavior of smpi/sender_gap and set its default value to 0
1367 * Add option to asynchronously send small messages to allow better
1368 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1369 specifies the size in bytes under which messages will be asynchronously sent.
1370 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1371 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1374 * New C interface to define a platform: XML is now optional.
1375 For more info, please check include/simgrid/platf.h.
1376 * New interface to define random platforms from the C:
1377 For more info, please check include/simgrid/platf_generator.h and
1378 examples/msg/masterslave/masterslave_platfgen.c
1379 * Export a sg_cmdline dynar containing all the arguments we got from
1383 * Two new tracing options for adding comments to trace file so you
1384 can track your experiments (see --help-tracing for details).
1385 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1386 * Adding the SimGrid version that generated the trace file as a comment.
1387 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1388 * Fix to avoid key clashes on Paje links
1389 * Other minor fixes related to the Paje specification
1392 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1393 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1394 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1395 * Greatly improve the robustness of mmalloc to user errors (such as
1396 using an area after freeing it, or freeing it twice)
1398 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1400 SimGrid-java (3.8.1) stable; urgency=low
1402 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1403 Warning: all methods are visible, but only some of them are
1404 implemented so far. Check the source (src/jtrace.c)
1405 for further information.
1406 * New module: org.simgrid.msg.File (SimGrid File management functions)
1407 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1408 * Change the meaning of Process.restart: now restart the process from
1409 the begining, like MSG_process_restart in C.
1410 * Add Process.setAutoRestart: handling of process restart when failed
1412 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1413 you to retrieve the properties of the processes/hosts
1414 * Deprecate Msg.clean(): you can just forget about it now.
1415 * New function Process.getCount(), that only works when compiling
1416 with the not yet released version 3.9 of the C library.
1418 * New context factory based on Coroutines. It mandates a modified JVM
1419 but then, the simulations run about five times faster, and there is
1420 no limit to the amount of processes (beside of the available memory).
1422 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1424 SimGrid (3.7.1) stable; urgency=low
1427 * Restore the prototype of MSG_process_create_with_environment() to
1428 the pre-3.7 situation by removing the kill_time argument.
1429 * Add a MSG_process_set_kill_time() function instead.
1432 * Fix weird behaviors when dealing with parallel tasks.
1435 * Simgrid is now built as a dll.
1436 * Simgrid-java now works on Windows.
1437 * Simgrid-Java is now included into Windows package.
1440 * First pre-build package for MacOSX.
1443 * Fix compilation when using MSG_USE_DEPRECATED.
1444 * Fix some compilation issues on Macs and Windows.
1445 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1446 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1448 We discovered that the Lua console is broken, but we are missing the
1449 manpower to fix it right now. The problem existed in 3.7 too, so we
1450 are not blocking the release for that. Sorry if you depended on this
1451 feature, any help would be really welcome.
1453 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1456 SimGrid-java (3.7.1) stable; urgency=low
1458 The "Java aint got to be bloated and slow" release
1461 * Various internal cleanups and performance improvement
1462 Simulations are expected to run up to twice faster or so
1463 * Make Process.kill(process) an instance method, not a static one
1464 * User processes are not java.lang.Thread subclasses.
1465 This breaks the compatibility (sorry), but previous API was
1466 brain-dead, making it impossible to have non-trivial
1467 initializations in the process constructor.
1468 * Require a full constructor per Process sub-class.
1469 Kinda breaks the compatibility (sorry), but this allows a much more
1470 efficient way to launch the processes at simulation startup.
1471 * Do not embeed our version of semaphores, java 1.5 can be considered
1472 as sufficiently prevalent for us to not dupplicate its features.
1476 * Add examples for almost every part of the API
1477 We spotted and fixed a lot of bugs in the process
1479 * New module: asynchronous communication API
1480 * New function: Process.sleep()
1481 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1482 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1483 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1485 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1487 SimGrid (3.7) stable; urgency=low
1489 The "spring cleanups (before the next Big Project kicks in)" release.
1492 * We can specify the SMPI latency/bandwidth factor with command line
1493 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1494 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1495 You can also use the "config tag" from platform file by adding this line
1496 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1497 (see "example/platforms/tag_config.xml" to use "config tag").
1498 Note that the command line supersedes the platform file configuration.
1499 * Change the correction factors used in LMM model, according to
1500 the latest experiments described in INRIA RR-7821.
1501 Accuracy should be improved this way.
1502 * Use the partial invalidation optimization by default for the
1503 network too. Should produce the exact same results, only faster.
1504 * Major cleanup in surf to merge models and split some optimization
1505 mechanisms from the core of the models. As a result you can now
1506 specify which model to use (e.g., --cfg=network/model:LV08
1507 --cfg=cpu/model:Cas01) and which optimization mode to use
1508 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1509 Incompatible combinations should err at initialization. See
1510 --help-models for the list of all models and optimization modes.
1511 * The CLM03 workstation model was dropped for simplicity because it
1512 used the deprecated CM02 network model. Use default instead.
1513 * Rename the TCP_gamma configuration option to network/TCP_gamma
1514 * Rename the coordinates configuration option to
1515 network/coordinates, and document it
1516 * Use now crosstraffic keyword instead of the terribly misleading
1517 fullduplex keyword. It is activated by default now in the current
1518 default model, use --cfg=network/crosstraffic:0 to turn it off.
1519 * Ongoing refactoring the model parsing to make XML files optional
1520 See include/simgrid/platf.h for details (still to be completed)
1523 * Major overhaul of the documentation. Almost instructive now :/
1524 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1525 functions and friends. This interface was considered as
1526 deprecated since over 2 years, it's time to inform our users that it is.
1527 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1528 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1529 use these (crappy) functions in your code.
1530 These functions will be removed soon. Stop using them now.
1531 * Deprecate MSG_get_host_{table,number}
1532 Implement MSG_hosts_as_dynar() instead.
1533 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1534 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1537 * Stabilize the parallel execution mode of user contexts
1538 * Introduce configuration variables to control parallel execution:
1539 - contexts/synchro: Synchronization mode to use when running
1540 contexts in parallel (either futex, posix or busy_wait)
1541 - contexts/parallel_threshold: Minimal number of user contexts
1542 that must be part of a scheduling round to switch to parallel
1543 execution mode (raw contexts only)
1544 * Fix bugs that prevented to use suspend/resume along with
1545 synchronization structures.
1546 * Fix bugs in process termination that lead to invalid memory access
1547 in very specific conditions.
1550 * Introduce a parallel mode for the models (controlled by surf/nthreads
1551 configuration item). In our tests, running the models in parallel
1552 never lead to any speedups because they are so fast that the gain
1553 of computing each model in parallel does not amortizes the
1554 synchronization costs, even when ultra fast futexes are used.
1555 This is released anyway because YMMV.
1558 * Performance boost by using a swag internally to compute the set of
1559 tasks that are finished and should constitute the return value of
1563 * Enable it by default now that it is considered rather stable.
1566 * Documentation of the tracing functions.
1567 * Performance gains when tracing categorized/uncategorized resource
1568 utilization by avoiding calls to get route when updating resource
1569 variables. LMM constraints are being used instead.
1570 * API changed to set task categories. Use MSG_task_set_category instead
1571 of TRACE_msg_set_task_category, and SD_task_set_category instead
1572 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1573 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1574 * Examples for link user variables added (see at examples/msg/tracing/)
1575 * Deprecated function TRACE_msg_set_process_category completely removed
1576 * Trace header updated according to the latest Paje file format
1577 * Tracing network lazy updates, no longer obligate users to use full updates
1578 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1579 * Experimental: let user code declare/set/push/pop application states for hosts
1580 * API changed to allow the manual creation of graph configuration files
1581 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1584 * Improve the API of Lua MSG bindings, using the Lua spirit.
1585 * Each simulated process now lives in its own Lua world (globals are
1586 automatically duplicated). It helps writing simulators. Will allow
1587 to run Splay programs within SimGrid in the future.
1588 * Add a Chord example in Lua, equivalent to the MSG one.
1591 * Start the implementation of a solution to express temporal
1592 properties, not only local assertions. This is still an
1593 experimental work in progress, stay clear from it to be safe.
1597 - Add new runtime parameters --help-logs and --help-log-categories
1598 to display information about supported logging parameters and
1600 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1601 don't exists anymore.
1602 * Mallocators: allow value NULL for the reset function.
1604 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1605 create homogeneous dictionaries, where all the elements share the
1606 same free function. Non homogeneous dictionaries will be
1607 deprecated in the next release.
1608 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1609 - Multi-level dictionaries are deprecated.
1611 - new function xbt_dynar_search_or_negative() that is useful when
1612 you have less than 2 million elements in your dynar and don't
1613 want of the extra complexity of catching exceptions when the
1614 element is not found.
1616 - Make xbt_os_thread module (for thread portability) public.
1617 Documentation is still to come, sorry.
1619 - Cleanups and simplifications to make it maintainable again.
1620 - Exotic features (such as memalign and valloc) were removed.
1621 - The metadata were extended and improved so that the
1622 model-checker becomes able to explore and inspect the heaps.
1623 - This may induce a performance drop when enable_model-checking is
1624 ON in cmake (even if it's not used in the simulation), but it is
1625 necessary at this point to get MC working.
1627 Turn model-checking OFF if simulation performance matters to you.
1628 Not enabling it at runtime is not enough, disable it in cmake.
1630 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1633 SimGrid (3.6.2) stable; urgency=low
1635 The "Not coding new stuff allows to polish old things" release.
1638 * New bindings to the NS3 packet level simulator (experimental)
1639 * Use the raw (efficient) execution contexts instead of the sysv
1640 (portable) ones when possible.
1641 * libpcre is now mandatory in any cases since not using it led to
1642 severe performance loss and possibly other issues
1643 * Update the XML platforms:
1644 - G5K: include the latest machine in Nancy
1645 - GridPP and LCG: new platforms
1646 * Documentation was partially updated, at least (more to come)
1648 Bug fixes, cosmetics and small improvements
1649 * Free terminated processes before the end of the simulation to avoid
1650 exhausting the memory of users having very dynamic amount of
1652 * Bug fix and cosmetics about canceling non-running tasks
1653 * Bug fix about the dot loader's issues when using libcgraph
1656 * Create an installer for windows with nsis (amd64 and win32)
1657 - Add an hello world project to illustrate simgrid project creation.
1658 - Embed libpcre into the Simgrid installer to avoid
1659 its compilation burden
1660 * The raw execution contexts should work on Apple now
1661 * Port to Windows 64 bits
1662 - Sysv contexts now have an implementation for this arch
1663 - GRAS communication features now support this arch
1664 * Drop support for borland compiler on windows
1665 - this code was not maintained, and we kinda depend on gcc nowadays
1666 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1667 * Fix portability issue on unstable ubuntu: linker became picky on
1670 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1672 SimGrid (3.6.1) stable; urgency=low
1674 The "Oops, we broke Macs too" release
1677 * Fixed contexts detection so that raw ones are used when possible
1678 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1679 leads to a strange error, with user code segfaulting sometimes when
1680 the generated code is not perfectly aligned (which is not
1681 controllable from the user side, depends on the amount of code)
1684 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1686 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1688 SimGrid-java (3.6) unstable; urgency=low
1691 * Split of every thing from simgrid v3.5 into a separate package.
1693 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1695 SimGrid (3.6) stable; urgency=medium
1697 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1700 * Bindings now constitute their own package, separated from the main one.
1701 Rationale: reduce our maintenance nightmare by reducing the module coupling
1702 They will soon be released on their own on gforge.
1704 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1705 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1707 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1708 * It's not quite deprecated for now because we have no replacement,
1709 but it may soon become the case.
1712 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1713 * Fortran: New user-level cache variable to store the rank of the running
1714 process. This improves performance by an order of magnitude.
1715 * C: New coccinelle script to automatically locate and modify global and
1716 local static variables.
1717 * Improved SMPI network model with a sender-side gap to account for multiple
1721 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1722 only say if a communication is finished, no matter whether it succeeded or
1723 failed. You can call MSG_comm_get_status() to know the status of a finished
1725 * New function MSG_task_dsend() to send a task and detach it. When a
1726 communication is detached, you are never notified of its success or failure
1727 and the memory is released automatically once it is finished. This function
1728 is useful when you don't care about the end nor the success of a
1730 * Change the prototypes of action replay. Sorry for inconvenience,
1731 but this is really more efficient this way (and to adapt your code,
1732 you just have to fix the initialization, that shouldn't be too long)
1733 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1734 ever managed to do anything useful with it.
1735 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1736 takes a dynar of communications. It returns immediately and gives the
1737 index of a finished communication (if any).
1738 * New example: a basic implementation of the Chord P2P algorithm.
1741 * New model for multi-core CPUs. You can now use the core attribute to
1742 precise the number of cores of a host. This is a basic model. Every
1743 process running on the host receives at most the power provided in
1744 the DTD (throughput<=power). Total throughput of process cannot exceed
1746 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1747 router linked by an up-link and a down-link (possibly asymmetrical).
1748 This kind of pattern allows to easily build last-mile model style platforms.
1749 Aggregating such patterns in a rule-based AS is thus the technique of
1750 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1751 * New model for Vivaldi routing. We transformed the Vivaldi network model
1752 into a Vivaldi routing model (based on the rule-based model). This allows to
1753 combine Vivaldi based latencies with last-mile platforms.
1756 * Added a check for NaN of IEEE754 infinite in the double entries of
1758 * Introduce a new context factory "raw", highly inspirated from the
1759 ucontext factory, but using manually crafted functions in assembly to
1760 do the work in an efficient manner.
1761 * Allow to change the used context factory at run time, not only at
1762 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1763 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1764 contexts at runtime (only with raw contexts or ucontexts).
1765 * Completely rewrote this module to allow parallel execution of user
1766 processes. Use --cfg=contexts/nthreads:N to execute user processes
1767 with N parallel threads (the default is 1, meaning no parallelism).
1768 * Allow to decide dynamically between sequential and parallel modes.
1769 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1770 processes in parallel only when their number is greater than or equal to P
1772 * Added a check for NaN of IEEE754 infinite in the double entries of
1776 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1777 won't output the state of processes when interrupted with Ctrl-C
1778 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1779 NULL-terminated array. This may solve backward compatibility issues
1780 due to the change to return type of SD_simulate. See also:
1781 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1782 * Add new macros with variable number of arguments.
1783 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1784 - in xbt/asserts.h: xbt_assert
1785 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1786 - in xbt/ex.h: THROWF and RETHROWF.
1787 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1789 * Change xbt_die() to accept a format string with arguments, just like printf.
1790 * New data structure: xbt_lib_t, like a dict but more general and with better
1794 * New configuration options
1795 Options triva/categorized and triva/uncategorized can be used to generate
1796 graph configuration files for Triva visualization tool.
1797 * Configuration option tracing/platform is renamed to tracing/categorized
1798 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1799 * New TRACE_user_link_variable interface:
1800 User provides the name of the link and the tracing variable to attach to it
1801 * the declaration of tracing categories must be done after the environment
1803 * simpler tracing interface, just one way to declare categories
1804 TRACE_category or TRACE_category_with_color, it is up to you
1805 * links in the trace file are again identified by their names
1806 * trace contains the full platform hierarchy exactly as declared using the ASes
1807 * Options tracing/msg/[task|process]:1 groups the process by hosts
1808 for both cases, tasks and processes must have names that are unique during the simulation
1809 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1810 * The experimental option tracing/msg/volume is deprecated
1811 its functionality may be reincorporated if needed
1813 The tracing generates a trace file with unordered timestamped events,
1814 because of the way the core simulator (surf) works. A script available
1815 at the tools directory (fix-paje-trace.sh) can be used to put the events
1816 in order. We have changed the tracing so it can generate ordered timestamped
1817 events in the final trace, but depending on the simulator (and how much time
1818 is simulated) that can lead to a huge memory utilization. It is deactivated
1819 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1821 Build Infrastructure
1822 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1823 - We are trying hard to keep the API stable, but it may happen that
1824 some things change (we're a research project after all, not a
1825 nuclear plant operating system). If such things should happen, you
1826 could rely on that macro to adapt.
1827 - current value: 30600 for 3.06.00, aka 3.6
1828 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1829 a number that can be compared with SIMGRID_VERSION.
1830 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1831 assertions and verbose/debug logging events are disabled at compile time.
1833 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1835 SimGrid (3.5) stable; urgency=medium
1838 * New feature to any SimGrid-based simulator: Model-Checking
1839 Check SIN#1 for more details.
1842 * New Model SMPI (three-interval linear regression for correction factors)
1843 See RR-7426, available at http://hal.inria.fr/inria-00527150
1844 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1845 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1846 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1847 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1848 * See also src/smpi/README
1853 - Tracing API changes: TRACE_start and TRACE_end should not be called
1854 by user-code. They are automatically called by simulators created
1855 with SimDAG, MSG and SMPI if the toolkit is compiled with
1856 tracing_enabled=ON. Categories declaration and utilization remain the
1857 same for MSG and SimDag.
1858 - A function was added to the tracing API to declare categories with
1860 - TRACE_category_with_color (char *category, char *color)
1861 where color must be in the following format
1862 "%f %f %f", red, green, blue
1863 and red, green, blue are float values in the interval [0, 1]
1864 - User can specify NULL as color parameter, or continue calling
1865 TRACE_category (cat)
1866 On that case, the tracing system will define random colors
1867 - The following command-line options are supported:
1868 --cfg=tracing/filename:msg.trace
1869 --cfg=tracing:1 (activate tracing, needed to use others)
1870 --cfg=tracing/platform:1 (categorized resource use)
1871 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1872 --cfg=tracing/msg/task:1 (task creation)
1873 --cfg=tracing/msg/process:1 (process creation, migration)
1874 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1875 --cfg=tracing/smpi:1 (SMPI interface tracing)
1876 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1877 - examples of examples/msg/tracing updated
1879 - DAXLoader and DOTLoader functions can generate tasks with categories
1880 - A new function to attribute a category to SD tasks:
1881 TRACE_sd_set_task_category (SD_task_t task, char *category)
1882 * Tracing the MPI interface implemented by SMPI
1883 - Collective operations are traced with states
1884 - Point-to-Point operations are traced with states/links
1885 - Tracing activated by a parameter "-trace filename" passed
1886 to smpirun during execution (considering that simgrid
1887 is compiled with tracing enabled)
1888 - To run the simulation with gdb, the simulator
1889 accepts --cfg=tracing/smpi:1 to trace SMPI
1890 - tesh files to check if smpi tracing is ok
1891 - See examples/smpi/NAS/DT-trace
1892 * GTNetS tracing re-worked
1893 - adaptation to the tracing system of GTNets to cope
1894 with modifications regarding the fullduplex mode
1895 - new tesh files to check if gtnets tracing is ok
1898 * Asynchronous communications through the functions:
1899 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1900 * New function: MSG_load_platform_script()
1901 to make possible using a lua script instead of XML files to set up platforms
1902 * New function: MSG_set_function
1903 to associate functions to processes, used when bypassing the parser
1904 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1906 Platforms: Add some more examples in examples/platforms
1907 * Grid'5000: see www.grid5000.fr
1908 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1911 * Change the XML format. This is a very important modification. SimGrid
1912 3.5 introduces a new hierarchical format based on the notion of
1913 Autonomous Systems. Compatibility with old format is ensured
1914 through the perl script provided in the install bin directory
1915 bin/simgrid_update_xml.
1916 It is now possible to build platforms with specific routing
1917 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1918 connect several platforms together. We will try to provide soon
1919 set of realistic platforms exploiting these properties (have a
1920 look at examples/platforms/ for the moment).
1921 * Take the opportunity of the XML format change to be a good XML citizen:
1922 rename link:ctn to link_ctn and similar changes (also dealt with by
1924 * Add a new routing scheme (rule-based) using regular expressions. It
1925 enables to have an extremely low memory footprint when the
1926 underlying routing is simple and can be compactly described. You
1927 need to have libpcre4-dev (perl regular expressions) installed if
1928 you want to use this routing scheme.
1929 * Revive the cluster TAG and allow to easily and efficiently (both in
1930 term of memory and speed) connect clusters together. Have a look
1931 at teshsuite/simdag/platforms/ to see how this can be done. With
1932 this tag, you can create clusters with thousands of tasks at no
1933 cost (have a look at examples/platforms/).
1934 Note: clusters are implemented as ASes, so there is no need for an
1935 enclosing AS tag if you have only one cluster in your platform.
1936 * Add new generic functions in the public interface that allows the user
1937 to call SURF 'create_resource' methods from your code (same
1938 functionality as the XML bypass mechanism but with a much lighter
1940 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1941 that uses a piecewise linear approximation to produce better
1942 results when exchanging small messages.
1943 * Add a new parameter to handle correctly full duplex link and account
1944 for interferences between uplink and downlink communications
1945 (activate with --cfg=fullduplex:1).
1948 * Rename the SD_READY (all dependencies are satisfied and task is
1949 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1950 dependencies are satisfied) state.
1951 This prevents a confusion between the notion of "ready to schedule"
1952 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1953 simulated" (SD_RUNNABLE) used by the simulation kernel.
1954 * Change the way a task is considered as ready. Instead of removing
1955 dependencies when a task is done, a counter is decreased. This way,
1956 it is always possible to reach ancestors thanks to the
1957 SD_taks_get_parents function (even after the end of the simulation.)
1958 * Change the return type of SD_Simulate from (SD_task_t*) into
1959 xbt_dynar_t. This function was in handling a dynar internally and
1960 converted it into a NULL terminated array for historical reasons.
1961 * New function SD_dotload(char*) to load a DAG described in dot
1962 format. This loader and the corresponding examples require the
1963 installation of the graphviz library.
1964 * Fix a bug in the management of tasks of size 0 in the surf network
1965 models. This problem was only visible with SIMDAG and you should
1966 thus disregard results produced with earlier versions if you
1967 relied on this feature (some tasks were blocked because of this).
1968 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1969 even though your applications were doing only point-to-point
1970 communications and sequential computations. Now you can really use any
1971 model you want (of course, if you create real parallel tasks, which are
1972 not implemented in most models beside ptaskL07, this will abort).
1973 * Add an example that schedules a DAX on an heterogeneous platform
1974 using a Min-Min strategy.
1975 * New function SD_workstation_get_current_task() that returns the kind
1976 of task currently running on a workstation in the sequential access
1978 * Raise some warnings when unexecuted tasks remains at the end of the
1979 simulation. This is usually caused by cycles in the DAG.
1982 * New function: SIMIX_process_set_function() called by MSG_set_function
1983 * Change the underlying waiting queue in semaphores so that a process
1984 can wait on several of them simultaneously (as in waitany).
1985 * Fix the way to handle tokens in semaphores so that all access patterns
1986 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1987 * kill the dirty pimple SIMIX_message_sizes_output()
1988 Please use (proper) visualization instead
1991 * New data container: setset (set of sets of elements)
1992 * New module: mmalloc (mapped malloc, allowing to have several
1993 independent segments of malloc)
1994 * New function: xbt_dict_cursor_set_data()
1995 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1996 * New function: xbt_dynar_is_empty()
1997 * New function: xbt_fifo_get_last_item()
1998 * Fix xbt_dynar_shrink(): use the right element size.
1999 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2000 uninitialized areas during expand.
2001 * Fix semaphores: previous implementation was severely broken.
2002 * Use library init/fini functions for our initialization.
2003 - you can use logs and other feature as soon as you want in your
2004 code (even before the xbt_init / MSG_init)
2005 - xbt_exit is now a no-op and produce a warning when used.
2008 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2009 benefit from the latest improvement to the simulation kernel.
2010 * Kill measurement sockets for now. If you rely on them, sorry. This
2011 release is not for you. This feature will be reintroduced in the
2012 future, but we cannot delay the release any further.
2013 * New function: gras_msgtype_get_name().
2014 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2015 * Fix (at last) the pmm example: it should not randomly fail anymore.
2017 Build chain: bug fixes and overall polishing
2018 * Cmake is now stable enough. Hence, we killed the autotools.
2019 * Port to windows ( TM :)
2020 * Fix the 'make install' target.
2021 No need to use 'make install-simgrid' anymore
2022 * Introduce a 'make dist' target compiling a *source* archive
2023 'make package' compiles a binary archive
2024 * Compile java files only on need
2025 * Add --cd and --setenv command line options to tesh
2026 * Out of source builds are not fully supported yet, but we are close
2027 * Enable supernovae and optimization flags by default for our users
2030 * Add layer to set up environment directly from lua, without XML.
2031 * The effect of gras_stub_generator can be achieved through
2032 lua too (check examples/gras/console/ping_generator.lua)
2034 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2036 SimGrid (3.4.1) stable; urgency=low
2038 The "Polishing easter eggs is probably a good idea" release.
2039 This is a bug fixes release only.
2043 * Fix a bug preventing the tasks from begin garbage collected.
2046 * Fix a bug occuring when a host involved in a communication fails.
2047 This was not detected properly by the other peer involved in the
2048 communication. Now, it's reported as a network error.
2051 * Warn the user about loop dependencies in data flow of DAX files
2052 * Obey the control-flow dependencies of DAX files
2055 * Add option "enable_smpi" allowing to not compile SMPI.
2056 Probably useful for the (Mac) users experiencing a build error here
2057 * Improve the detection of lua5.1 and ruby1.8
2059 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2061 SimGrid (3.4) stable; urgency=low
2063 The "Easter in Cargese" release. Also known as (major changes):
2065 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2066 ~> bindings were greatly improved
2067 ~> new tracing infrastructure for better visualization introduced
2069 * the "Welcome to configury modernity" release.
2070 ~> we switched from autotools to cmake, and improved our cdash
2073 A more detailled list of changes follow (full detail in svn log).
2075 Java Bindings: Various Cleanups
2076 * (install java-gcj-compat-dev on debian-like to use them)
2077 * Remove put/get: no need to export deprecated interface in Java
2078 Use send/receive instead.
2079 * Cleanup the examples and add a README per directory
2080 * Remove example autoDestination (that's the only way to go now)
2081 * Remove example explicitDestination (was a plain copy of basic)
2082 * Make JniException a runtime exception, so that there is no need to
2083 declare the fact that you may encounter such a beast. I guess that
2084 nobody will ever want to survive such error.
2085 * Create specific errors for each MSG case of failure:
2086 host failure, transfer failure, timeout, task cancelled
2087 * Cleanup the exceptions that may get thrown by each function
2088 * Other internal cleanups in Java bindings. Performance still bad :/
2089 Ruby and Lua Bindings: create them
2090 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2091 * That's new and great, you should try them out.
2092 Same functionalities than Java bindings, only even less polished
2094 * Kill the useless "rate" argument of SD_task_get_execution_time()
2095 Everyone used to provide -1 as a value, it was not used, and the
2096 semantic of a possible use wasn't even clear.
2097 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2098 either as comm costs or compute costs to mean that there is no
2099 such thing for that specific task.
2100 * Add a SD_task_set_name() function
2101 * Fix SD_task_unschedule() on typed tasks
2102 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2103 * In DAX loader, accept useless 'level' attributes to <job> since
2104 LIGO DAGs have them (seem to be to ease graphical representation).
2106 * Add an example masterslave_mailbox.c using send/receive and not
2107 the deprecated put/get interface.
2108 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2109 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2111 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2112 (declare MSG_USE_DEPRECATED to still have the old name)
2113 * Add a MSG_task_set_data() function
2114 * About trace replay (see examples/msg/actions):
2116 - Allow to work with splitted trace files for each process
2117 Give the specific trace file as argument of each process,
2118 and call MSG_action_trace_run(NULL)
2119 You can still have one merged file for all processes.
2120 - Fix implementation of collective operations
2121 * Allow task_execute() on 0-sized tasks (closes #10063)
2123 * This is the first release of SimGrid where SMPI is not considered
2124 beta anymore (even if some corners should still be improved)
2125 * Port over the new SIMIX_network submodule (internal refactoring)
2126 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2127 * Implement more missing elements of the standard:
2129 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2131 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2132 MPI_Scan() MPI_Get_processor_name()
2133 - Added implementation of missing case for Alltoall (warning: it's
2134 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2135 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2136 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2137 SMPI_MPI_Allgatherv()
2138 * Bug fixes include:
2139 - MPI_Waitsome() was broken
2140 - Allow relative includes in smpicc
2141 - Command line cfg argument 'reference_speed' was ignored...
2142 - Some functions did not properly lead to auto-benching of user code
2143 - smpicc passes -O2 by default (just like openmpi one)
2145 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2146 * Bug fixes about timeouts during communications
2147 * add SIMIX_message_sizes_output() as a pimple to write to file the
2148 amount of messages per size. Use gnuplot to get histogram.
2149 Pimple because that's the only user-visible function of simix,
2150 defined directly in xbt.h (irk, sorry)
2152 - Add a SIMIX_sem_get_capacity() function
2153 - Fix interactions with processe resume/suspende
2154 - release_forever() was stupidly broken
2155 - Fix SIMIX_display_process_status() for processes in a semaphore
2156 - Make SIMIX_sem_block_onto() user-visible
2157 * Refactoring context stuff:
2158 - Use pseudo-OOP for better modularity
2159 - reimplement SIMIX_process_kill() without process_schedule() so
2160 that the latter can take as invariant that it is called from
2162 - Merge context_start into context_new for sake of simplicity
2164 * Add a Vivaldi network model, coded live during SUD'10 ;)
2165 * Rename configuration variables to start a hierarchy:
2166 o cpu_model -> cpu/model
2167 o network_model -> network/model
2168 o workstation_model -> workstation/model
2169 * New configuration variables:
2170 o network/bandwidth_factor: correction to bandwith
2171 o network/latency_factor: correction to latency
2172 o netwotk/weight_S: correction to the weight of competing streams
2173 * Add a long description to the models, that users can see with such
2174 argument on the command line: --cfg=cpu/model:help
2175 * --help-models display the long description of all known models
2177 * config: add the ability to set a default value after registration
2178 Does not override any previously set value (e.g. from cmd line)
2179 * dict: allow to have integer key and data.
2180 When so, you need to use the following functions
2181 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2182 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2183 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2184 In contrary to regular dicts, the key is not malloced before copy.
2185 Mixing scalar and regular elements in the same dict is not tested
2187 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2188 Tracing for Visualization:
2189 * SimGrid is now instrumented in order to generate a trace file for
2190 visualization analysis: to use it, need to compile SimGrid with the
2191 "tracing" option enabled, and instrument the program using SimGrid with
2192 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2193 (among other functions).
2194 * The instrumentation only traces the platform utilization for now
2195 * Documentation to use the tracing functions and how to analyze the
2196 traces with the Triva tool is written.
2197 * More information about: SimGrid FAQ (in the section Tracing Simulations
2200 * We moved to cmake as default build system. Autotools support will
2201 be dropped soon. Check the FAQ for more info about how to use it.
2202 * Greatly improved our cdash/ctest interactions
2203 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2204 * Added memory checking tests with valgrind; lot of memleak fixing.
2205 This may be the first release of simgrid with so few memory issues
2206 * Added code coverage tests.
2207 Our coverage is still improvable, but at least we see it on cdash.
2209 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2211 SimGrid (3.3.4) stable; urgency=low
2213 The "Desktop Grid needs love too" release (also called Xmas release).
2215 Models improvements:
2216 * Major speedup in the maxmin system solving by using lazy evaluation
2217 Instead of solving completely the maxmin system at each iteration,
2218 only invalidate (and recompute) the modified parts.
2219 This new feature is enabled in default models but you can try to
2220 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2221 * Cas01 IMproved as default CPU model
2222 This CPU model is the same Cas01 model, but it uses the
2223 maxmin-selective-update flag and a heap structure to manage
2224 actions on SURF kernel.
2225 It reduces the complexity to find the next action to finish and,
2226 consequently, it's faster than the old Cas01.
2227 This is the new default CPU model (Cas01).
2228 * Rename the old Cas01 model to Cas01_fullupdate
2229 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2230 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2231 * CpuTI (CPU Trace Integration)
2232 A new CPU model whose objective is simulate faster when using
2233 availability trace files.
2234 Instead of using a full featured, over engineered maxmin system for
2235 CPU modeling, this model does the pre-integration of traces files
2236 to calculate the amount of CPU power available, and so, executes
2237 faster than the old CPU models.
2238 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2239 * Use LV08 as default network model since it gives better accuracy
2240 for small messages and shouldn't change things for big ones.
2241 Use --cfg=network_model:CM02 to get the previous behavior.
2244 ******************************************
2245 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2246 ******************************************
2247 * The new CPU model may changes simulations!
2248 The point is that events occurring at the exact same timestamp
2249 are not scheduled in the same order with the old and new
2250 version. This may be enough to completely change the execution
2251 of simulations in some cases.
2252 * The new network model will change simulations!
2253 This new model is more realistic than the previous one, so you
2254 should consider redoing your old experiments with this model.
2255 Sorry for the inconvenience.
2258 * Introduce the supernovae compilation mode
2259 When compiled that way, the whole SimGrid (or almost) is put in a
2260 single compilation unit and compiled in one shoot.
2261 This is to help gcc which has difficulties to inline stuff from one
2263 The speedup seem to be above 15%, althrough more tests are needed on
2264 amd64 to confirm that gain.
2267 * Port of MSG's mailbox on top of SIMIX network
2268 The put/get mechanism was greatly simplified on the way.
2271 * New SIMIX network module. Provides:
2272 - Mailbox: rendez-vous mechanism to find with who you want to speak
2273 - Synchronous send/recv: easier and hopefully faster since the
2274 logic is handled in the maestro process directly now
2275 - Asynchronous send/recv: you dreamt of it? It's here now
2276 Too bad that nobody cared enough to propagate the change to MSG.
2277 * Add semaphores as SIMIX synchronization mechanism.
2280 * new function SD_daxload(char*) to load a DAX file
2281 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2282 * Introduce typed tasks. Specify its kind and cost at creation.
2283 At scheduling, just give where it should be placed, and the cost
2284 for each involved resource is automatically computed.
2285 Existing constructors so far (more to come of course):
2286 - SD_task_create_comm_e2e() for end-to-end communication
2287 - SD_task_create_comp_seq() for sequential computation
2288 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2289 * new function SD_task_dump() for debuging display
2290 * new function SD_task_dotty(task,FILE*) writing to file the info
2291 about the task in dotty format
2292 * SD_task_dependency_exists() can now cope with having one of its
2293 arguments NULL. If so, it tests whether the other argument has any
2295 * Add getters on list of preceding/following tasks:
2296 SD_task_get_parents(task) and SD_task_get_children(task)
2297 * Add getters on amount of workstations and list:
2298 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2299 * Add getter on task kind: SD_task_get_kind(task)
2300 * Update the start_time and finish_time of tasks on completion/failure
2301 * Bugfix: Remove task from state swags when destroyed
2304 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2307 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2308 dynar with one separate thread per value of the dynar.
2309 * Change the prototype of xbt_thread_create(), sorry.
2310 Added a boolean parameter indicating whether we want to join this
2311 thread (used in SG only for now)
2312 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2315 * GTNetS wrappers should now be usable again (and betterly tested too)
2316 * Fix a major regression from 3.2 where the timeout provided to
2317 MSG_task_put_with_timeout() was used as absolute time before which
2318 the comm should be done.
2319 * Start to fix the <cluster> tag.
2320 - Internal links should be good now (beside of the loopback, which
2321 use the private link instead)
2322 - paths to the external world is still rather broken
2323 - the <route:multi> tag is just broken. Actually that's brain-dead.
2324 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2325 to make it less stupid
2326 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2327 * Fix a source-level compatibility glitch from 3.2: after defining
2328 MSG_USE_DEPRECATED, you can use the old name
2329 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2330 * Allow to compile from the SVN with automake 1.11
2331 * Fix some problems when using the "start_time" tag in deployment XMLs.
2332 * Fix #8569: XBT/synchro.h has redundant declarations
2333 * Fix #8563: MSG return values and exceptions
2334 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2335 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2336 Thanks to Robson Peixoto.
2337 * Fix a vicious bug in dictionaries inducing that some elements were
2338 not freed on xbt_dict_free()
2340 Portability report of this version:
2341 * Main portability targets:
2342 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2343 on (amd64/i386/ia64)
2344 - mac leopard on i386
2345 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2346 but nothing critical.
2347 * Other platforms: windows, AIX and others were not tested for this release
2349 Timing report of this version:
2350 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2351 scenario parameters). From 8h to a few seconds in desktop grid settings.
2352 * Supernovae brings about 25% speedup on i386.
2354 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2356 SimGrid (3.3.3) stable; urgency=low
2358 The "Need for Speed" release.
2360 The timings done to validate the 3.3.2 were faulty.
2361 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2363 The problem was a conversion from a manually handled vector to
2364 xbt_dynar_t on the critical path.
2365 xbt_dynar_foreach calls functions, inducing stack management crap.
2367 We inlined these functions and xbt_dynar_foreach is now breath taking.
2368 We also inlined xbt_swag_belong on the way.
2370 Here are some approximate speedup measurements (on master/slaves
2371 simulations lasting between 10s and 20s each):
2372 3.3.1 -> 3.3.2: about same performance
2373 3.3.2 -> 3.3.3: 40% speedup
2374 3.3.1 -> 3.3.3: 40% speedup
2375 3.3.1 with inline patch -> 3.3.3: 30% speedup
2377 Our reading is that the refactoring which occurred in 3.3.2 made us
2378 suffer much more from the xbt_dynar_foreach low performance, but
2379 once we solved this, this refactoring proved to be very performance
2380 effective. From the 40% speedup, somehow, 10% are due to the
2381 inlining and 30% to the refactoring.
2383 That's a pitty that gcc cannot inline functions placed in other files
2384 alone. We have to choose between:
2385 - break the encapsulation (by putting private data structures and
2386 accessors in headers files to help gcc)
2387 - live with low performance
2388 - switch to a decent compiler such as icc (not quite possible).
2390 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2392 SimGrid (3.3.2) stable; urgency=low
2394 The "Simplicity does not preceed complexity, but follows it" release.
2396 The main contributors of this release were (lexical order):
2397 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2400 * Extract the routing logic into its own object.
2401 (was dupplicated in network.c and workstation_LV07.c;
2402 Allows to implement other ways of storing that info)
2403 => kill now useless network_card concept
2404 - Use dynar to represent routes (instead of void** + int*)
2405 - kill link_set (use surf_network_model->resource_set instead)
2406 - Add a command-line option to choose the routing schema to use
2407 - Add three new models:
2408 * Floyd (shortest path computed at initialization)
2409 * Dijikstra (shortest path recomputed all the time)
2410 * Cached Dijikstra (shortest path computed on need)
2411 All these models where contributed by Silas De Munck, and are
2412 described in his ICCS09 paper.
2414 * Simplify model declaration
2415 (less redirections, less function to write when defining a model)
2416 - Factorize stuff between models:
2419 surf_model_resource_set(model)
2420 surf_model_resource_by_name(model, name)
2421 - Unify the types of models in s_surf_model_t (using an union)
2422 - Embeed fields of common_public directly into s_surf_model_t
2423 - Rename model methods:
2424 action_free ~> action_unref
2425 action_change_state ~> action_state_set
2426 action_get_state ~> action_state_get
2427 - Change model methods into functions:
2428 (model)->common_public->action_use ~> surf_action_ref
2430 * Implement a generic resource; use it as ancestor to specific ones
2431 (allows to kill duplicated code in models)
2432 Drawback: timer command don't need no name nor properties;
2433 workstation_CLM03 don't need no properties
2434 (but I guess we can live with those few bytes wasted)
2436 * Improve the action object model
2437 - implement a constructor avoiding dupplicated code about field
2438 initialization in generic_action part.
2440 * Kill the SDP model: it has an external dependency, is deprecated
2441 in flavor of modern lmm models, and didn't compile since a while
2444 * Relocation of the context module from XBT to SIMIX.
2445 (the context were decoupled from the simix processes, duplicating a lot of code)
2446 => a lot of code was factorized
2447 - less overhead is introduced during scheduling
2448 - simpler API for the context factory
2449 - the logic for process creation,destruction and manipulation was simplified
2450 * Simplification of the s_smx_process_t data structure.
2451 => accesing the simix level data associated to a process is faster now,
2452 and the code is a lot more readable.
2455 * Implement some more MPI primitives:
2456 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2457 -implementation: Bcast: flat or 2-ary tree (default),
2458 Barrier: 4-ary tree,
2460 Allreduce: Reduce then Bcast
2461 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2462 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2463 Alltoallv: flat tree, like ompi
2465 * Add support for optimized collectives (Bcast is now binomial by default)
2466 * Port smpirun and smpicc to OS X
2469 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2470 Shout out if you used it.
2473 * Display the list of still queued messages in SG mode when existing
2477 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2478 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2479 * Add xbt_str_from_file(FILE*)
2480 * Add xbt_dict_get_key achieving a linear reverse search
2481 * Remove the context module
2483 Portability report of this version:
2484 * Main portability targets:
2485 - Linux(debian)/x86/context
2486 - Linux(debian)/x86/pthreads
2487 - Linux(debian)/amd64/context
2488 - Linux(debian)/amd64/pthreads
2489 On these, we still have the eratic breakages of gras/pmm and
2490 amok/saturate_sg reported in previous version. We still think
2491 that the tests are the cause of the fault, not the tested code.
2493 - Mac OSX Leopard/x86/context
2494 Still false negative in tesh autotesting.
2495 Smpi still fails, but this time because readlink does not accept -f
2496 Everything seems to work properly beside of that.
2499 - AIX version 5.3 (only tested contexts this time)
2500 Smpi still fails there because mktemp is not installed.
2501 Everything seems to work properly beside of that.
2503 I managed to compile it for the first time, but several breakages.
2504 Won't delay the release for this exotic platform.
2506 * Windows: it's still lagging behind. If you want to help, please
2509 Timing report of this version:
2510 This version seem to be more than 5% faster than 3.3.1 (on linux
2511 64bits with contextes). The gain is less than expected, we are
2512 investigating this for next release.
2514 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2516 SimGrid (3.3.1) stable; urgency=low
2519 * Implement a --cfg-help to show existing configuration variables
2520 * Build chain do not require doxygen in maintainer mode
2523 * fix a bug on struct sizeof computation, which prevented the
2524 exchange of arrays of structs in some conditions
2525 - added a regression test about this in datadesc_usage
2526 * Allow the exchange of 0-long dynamic vectors.
2527 - for that, use -1 as indicator of dynamic size instead of 0
2528 - This implied to change any size from unsigned long to long,
2529 reducing a bit communication abilities, but I guess that with
2530 64bits being quite common, this is more than enough.
2531 - This also induce a protocol change, thus bumping network protocol
2532 version from 0 to 1 (if we have external users, we have to get
2533 clean on that point too ;)
2534 - added two regression tests about this in datadesc_usage
2535 * Be more verbose when propagating local exceptions
2536 This helps debugging.
2537 * Display the status of simulated processes when receiving SIGINT in
2541 * Allow to control the simulation from a trace file.
2542 New functions MSG_action_register() and MSG_action_trace_run()
2543 The first one allows to associate a function execution to each
2544 kind of action while the second one parses a trace file and
2545 triggers the corresponding actions within the system.
2546 For now, only a toy example is provided in examples/msg/actions
2547 * Add an exemple of process migration in examples/msg/migration
2548 * Fix a bug in task exchange which broke MSG_task_get_sender()
2549 Add a teshsuite regression test for that.
2550 [Bug: if MSG_task_get_sender() is called after sender exit,
2552 * Fix a bug which prevented suspend/resume to work properly
2553 * Display the status of simulated processes when receiving SIGINT
2554 This fixes a regression of v3.3. due to the introduction of SIMIX
2555 * Bug fixing in failure management:
2556 - trace could not start by a failure at time 0
2557 - failure during communications were not working
2560 * Add SIMIX_process_set_name() to change the name of the current
2561 process in the log messages.
2562 * Store smx_hosts in a dict since we only retrieve them by name
2563 * Move the configuration infrastructure to surf
2566 * Move the configuration infrastructure to surf
2569 * Massive internal cleanups:
2570 - Store internal structures on processes instead of hosts (allows
2571 to have more than one process per host, in addition of being more
2573 - Cleanup the initialization/finalization process
2574 - Kill a whole bunch of unneeded synchronization:
2575 processes run in exclusive manner within the simulator
2576 - Move queues from global tables to process data fields
2578 - now accept -platform and -hostfile arguments
2579 - Pass the right rank value to processes according to the hostfile
2580 * Compile the examples by default, and use them as regression tests
2581 * Implement MPI_Wtime()
2582 * Change the reference speed to a command line option
2585 * TCP_gamma can now be specified as command line option using
2586 --cfg=TCP_gamma:10000000.0
2587 * Change the --surf-path cmd line option into --cfg=path:
2590 * Also include strbuff from xbt.h public header
2591 * xbt_ex_display(): do not free the exception after displaying
2592 This allows to do more with the given exception afterward.
2593 Users should call xbt_ex_free() themselves.
2597 Portability report of this version:
2598 * Main portability targets:
2599 - Linux(debian)/x86/context
2600 - Linux(debian)/x86/pthreads
2601 - Linux(debian)/amd64/context
2602 - Linux(debian)/amd64/pthreads
2603 These targets fail about 1/10 of times on gras/pmm, but we believe
2604 that this is because of the test, not because of simgrid.
2605 amok/saturate_sg fails even more rarely, and the test may not be
2608 - Mac OSX Leopard/x86/context
2609 The test suite still spits tons of errors because some obscure
2610 force prevents us from removing the temporary directories
2611 arguing that they still contain some metadata I've never heard of.
2612 Smpi fails because seq is not installed.
2613 Everything seems to work properly beside of that.
2616 - AIX version 5.3 (both contexts and pthread)
2617 Smpi still fails there because mktemp is not installed.
2618 XML inclusions seems rosty on AIX.
2620 * Windows: it's still lagging behind. If you want to help, please
2623 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2625 SimGrid (3.3) stable; urgency=high
2629 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2630 [Malek Cherier & Mt]
2632 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2635 Until now, GRAS were using MSG as an interface to SURF. It was
2636 quite difficult because both interface have several differences
2637 (MSG channels vs GRAS sockets were the most notable point).
2639 This also opens the gate to SMPI (which should occur soon) and speed
2640 up simulations by to 40% (even if it were not the main goal).
2642 **************************************
2643 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2644 **************************************
2645 The point is that events occuring at the exact same timestamp are
2646 not scheduled in the same order with the old and new version. This
2647 may be enough to completely change the execution of simulations in
2648 some cases. Sorry for the inconvenience.
2650 * Cleanup and upgrade the XML format to push further scalability
2651 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2653 * Improve the testing infrastructure with tesh. Now a very large part of
2654 the code is tested not only by being run but also by checking that the
2655 output match an expected output [Mt].
2657 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2658 is really less memory-demanding, which should allow you to use
2659 larger files in SimGrid [AL].
2661 * Inform valgrind about our contextes, so that it becomes usable
2662 with the default (and more effecient) version of SimGrid
2663 [contributed by Sékou Diakite, many thanks]
2666 * Introduce a listener thread in charge of receiving incoming
2667 messages from the network. It allows to overlap communication and
2668 computation but most notably, it removes some stupid deadlocks due
2669 to the fact that so far, a process could not send and receive at
2670 the same time. This made most non trivial communication schema
2672 * Convert the PIDs from long int to int to match the MSG ones (and
2673 linux ones too) [Mt]
2674 * New function: gras_agent_spawn() to launch a new process on
2675 current host. Only working in simulation for now. [Mt]
2676 * New function: gras_os_hostport() returning a constant form (ie,
2677 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2680 * Make the backtrace of exceptions more human readable [Mt]
2681 * New module: xbt/str [Mt]
2682 a ton of string utility functions (split, join, printf to a newly
2683 allocated buffer, trim, etc)
2684 * New module: xbt/hash [Mt]
2685 SHA1 hashing algorithm (more to come if needed)
2686 * New module: xbt/synchro [Mt]
2687 synchronization tools (mutex and conditions) working the same way
2688 in simulation and in real life (mainly useful for GRAS, but not
2690 * New module: xbt/queue [Mt]
2691 classical producer/consumer synchronization scheme
2692 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2693 (using the classical functions will get serialized) [Mt]
2694 * Make dictionary internal table dynamic. No need to specify its size
2695 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2697 * Make sure the log channels are organized as a tree under windows
2698 (because of ANSI C compatibility issue, any channel were child of
2699 root directly) [Mt].
2702 * Cleaned many thing in surf and fixed a few bugs [AL].
2703 * Add a nice command line configuration mechanism to compose models [AL].
2704 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2705 the previous one (KCCFLN05). It relies on something that looks like
2706 a max-min sharing mechanism but cannot be written as such. A new solver
2707 was thus designed [AL].
2708 * Add a new solver to lmm. Based on Lagrange optimization and
2709 gradient-based descent, it enables to efficiently maximise systems s.a
2711 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2713 This solver enables to propose two new network models for TCP Reno and
2714 TCP Vegas based on Low's work. These models still need to be fully
2715 tested though [Pedro Velho].
2718 * Bug fix in SD_simulate. Now the time bound given as argument is
2720 * Use the new parallel task model (ptask_L07) as default.
2721 * Use the SURF command line configuration mechanism.
2722 * 0-size tasks (for synchronization) should now work.
2724 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2726 SimGrid (3.2) stable; urgency=high
2730 We still experience issues on this platform, but we believe that at
2731 least MSG is usable.
2733 GRAS API BREAKAGE (for simplification purpose, sorry):
2734 * the gras_msgtype_by_name is not used anymore. Instead of
2735 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2736 you can write (and must)
2737 gras_msg_send(toserver, "request", &request);
2738 - If you still want to pass a gras_msgtype_t to the function (to cache
2739 the type and avoid the lookup time), use the gras_msg_send_() variant.
2740 - Impacted functions:
2741 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2742 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2743 * The callbacks are now expected to return 0 when everything went well
2744 (just like the main() function)
2746 GRAS new features and improvements:
2747 * New module mechanism where user code can use per process globals [Mt]
2748 This is similar to gras_userdata_*() functions, but for libraries. It
2749 factorize some code developped over and over in the examples and AMOK.
2750 It has still to be documented and used (only amok/peermanagement is
2752 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2753 loss when they were small enough to fit into the buffer and sent quickly
2754 enough so that they can all get received in one shoot.
2755 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2756 if not found. Use the *_or_null() variant for the old semantic.
2757 * In gras_msg_handle, do not discard messages without callback.
2758 They are probably messages to be explicitly awaited later (ie, proofs of
2759 mis-synchronization in userland since they are sent before being awaited)
2760 No big deal usually.
2761 * gras_socket_meas_send/recv: semantic changed!
2762 The numerical arguments used to be (1) the total amount of data to send
2763 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2764 messages. This was need for the fool willing to send more than MAXINT
2765 bytes on quite fat pipes.
2768 * Do really rename the hostmanagement module to peermanagement. [Mt]
2769 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2770 but this is rather new and this was documented in the module
2771 documentation (poor excuses, I admit)
2772 * Bandwidth measurement semantic changed! This follows the changes to
2773 gras_socket_meas_send/recv explained above.
2776 * A sequential mode has been added to the workstations. When a workstation
2777 is in sequential mode, it can execute only one task, and the other tasks
2778 are waiting in a FIFO. [Christophe Thiery]
2781 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2782 model for SIMDAG. [Christophe Thiery]
2783 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2784 equal to 0, which caused some bad behaviors in
2785 saturated_constraint_set_update. I now use a threshold mechanism like in
2789 * When running manually src/testall, you select specific units [Mt]
2790 testall is the result of our cunit mechanism, and should replace all
2791 the scripty thingy around since bash don't run easily on billware.
2793 * A mallocator system has been added. [Christophe Thiery]
2794 Mallocators allow you to recycle your unused objects instead of freeing them
2795 and allocating new ones.
2797 Documentation update:
2798 * FAQ reworking + New FAQs:
2799 - "Valgrind spits tons of errors!" [Mt]
2800 - "How to repport bugs" [Mt]
2801 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2802 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2803 - Communication time measurement within MSG [AL]
2804 - I experience weird communication times when I change the latency [AL]
2805 * GRAS tutorial [Mt]
2807 - an introduction to the framework and to the used communication model
2808 - an initiatic tour introducing the most proheminent features:
2810 . Lesson 0: Installing GRAS
2811 . Lesson 1: Setting up your own project
2812 o Part 2: Message passing
2813 . Lesson 2: Exchanging simple messages
2814 . Lesson 3: Passing arguments to the processes (in SG)
2815 . Lesson 4: Attaching callbacks to messages
2816 . Lesson 5: Using globals in processes
2817 . Lesson 6: Logging information properly
2818 . Lesson 7: Using internal timers
2819 . Lesson 8: Handling errors through exceptions
2820 . Lesson 9: Exchanging simple data
2821 . Lesson 10: Remote Procedure Calling (RPC)
2822 . Lesson 11: Explicitely waiting for messages
2823 . Recapping of message passing features in GRAS
2824 - A HOWTO section containing:
2825 o HOWTO design a GRAS application
2826 More are due, of course. They will come latter. In the meanwhile, you can
2827 check the examples which are still here.
2829 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2831 SimGrid (3.1) stable; urgency=high
2835 There was a stack corruption somewhere, visible only when optimizing
2836 with these versions. [Vince]
2839 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2840 module that enabled to play with Directed Acyclic Graphs. It is built
2841 directly on top of SURF and provides an API rather close to the old
2842 SG. Some old codes using SG are currently under rewrite to check that
2843 all needful functions are provided. [Christophe Thiery]
2846 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2847 extension of the classical CLM03 model that gracefully handles
2848 failures. This is now the default model for MSG and GRAS. It doesn't
2849 handle parallel tasks yet though. [AL]
2850 * Bug fix: Weights were not correctly set in the network part.
2851 WARNING: This may have resulted in incorrect results with simulations
2852 where there are more than one flow on a given link. [AL]
2855 * After a (long ?) discussion on simgrid-devel, we have decided that the
2856 convention we had on units was stupid. That is why it has been decided
2857 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2858 WARNING: This means that all previous platform files will not work as
2859 such with this version! A warning is issued to ask users to update
2861 A conversion script can be found in the contrib module of the CVS, under
2862 the name contrib/platform_generation/surfxml_update.pl [MQ]
2865 * Bug fix: Processes were started in reverse order, wrt deployment file.
2866 WARNING: if your code relies on this bug, please fix it. [AL]
2867 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2868 being executed on two different locations. [AL]
2869 * Bug fix: Failures are now better supported thanks to Derrick's tests
2870 (there was many failure situations I hadn't thought of and that weren't
2871 correctly handled). [AL]
2872 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2876 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2877 Exception occurring on server-side are propagated back to client (!).
2879 API CHANGE: the callback changed their prototype. Change:
2880 int my_handler(gras_socket_t expeditor, void *payload_data) {
2882 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2883 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2885 * New! function: gras_msg_handleall to deal with all messages arriving
2886 within a given period.
2887 * New! function: gras_socket_server_range to get a server socket in a
2888 range of port numbers (ease to avoid port number conflicts) [MQ]
2889 * New! gras processes display their backtrace when they get a SIGUSR1
2890 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2891 Sweet to debug RL processes [MQ]
2895 - Do not force experiment sizes to be expressed in kb, or it becomes
2896 impossible to measure the latency this way (needs one byte-long tests)
2897 WARNING: this changes the amok_bw_* function semantic. [MQ]
2898 - Implements the link saturation stuff. [MQ]
2899 * Peer management module:
2900 New! module factorizing code that we wrote over and over [MQ].
2903 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2904 - Test units are placed directly into the library code, they get extracted
2905 automatically and placed into the src/testall binary.
2906 - Convert most of the XBT tests to this system.
2907 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2908 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2909 live in the same C code. This required to change a little bit the API
2910 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2911 * New module: sparse graph structure with basic algorithms (this is work
2912 in progress and the API is not considered to be frozen yet). [AL]
2913 * Display more information on backtraces: source line & function names are
2914 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2915 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2916 * Reworked a little bit some #include statements to load only required
2917 headers. Some user code that relied on SimGrid to include stdlib or
2918 stdio may need to include it by themselves. [AL]
2919 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2921 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2922 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2923 but I guess that almost nobody used those parts. [MQ]
2925 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2927 SimGrid (3.0.1) stable; urgency=low
2930 * Unfortunately, I had missed 5 misnamed functions:
2931 xbt_fifo_item_t xbt_fifo_newitem(void);
2932 void xbt_fifo_freeitem(xbt_fifo_item_t);
2933 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2934 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2935 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2936 They're now deprecated. Please use their new versions:
2937 xbt_fifo_item_t xbt_fifo_new_item(void);
2938 void xbt_fifo_free_item(xbt_fifo_item_t);
2939 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2940 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2941 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2943 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2944 * Documentation: xbt_log module unmercifully reworked [MQ]
2945 * Bugfix: there was a problem with the ending of contexts with
2946 the pthread backend. It caused some weird deadlock or behavior
2947 depending on the pthread implementation. [AL]
2948 * Bugfix: get the exceptions raised in the simulator repport where
2949 and why they come from when they are not catched in time [AL, MQ]
2952 * Bugfix: Do repport the error when two non-connected hosts try to
2953 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2956 * Add additionnal checkings on communications. Assert that two
2957 communicating hosts are connected by a set of links... [AL]
2960 * Add additionnal checkings on channel values in communication [AL]
2961 * New: MSG_task_get_source to see on which host a task was generated [HC]
2962 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2963 the number of tasks waiting to be received on channel and sent
2965 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2966 waits for the first task coming from a given host.. [AL]
2968 GRAS new functionnalities: [MQ]
2969 * Enhance the parsing macro to allow the size of multidimentional objects
2970 to be given thru annotations.
2971 * New example (and documentation): Matrix Multiplication a la RPC
2972 (as when I was young!) and fix a bunch of bugs found on the way.
2974 GRAS performance improvements: [MQ]
2976 * Reduce the amount of cbps creation/destruction by making it static to
2977 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2980 * Change libdata to a set so that we can search for stuff by ID (and thus
2981 reduce the insane amount of dict lookups)
2984 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2985 RL communication on top of SG since it's so uneffective.
2986 It may also be used for inter-thread communication in RL, one day.
2987 * Use gras_datadesc_copy() to exchange messages on top of SG
2989 - improve message exchange performance on top of SG
2990 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2991 * Don't exchange on the network the size of the used part of buffer,
2992 instead, specify the possible buffer size to read().
2994 - reduces the amount of read/write calls (one pair per exchange)
2995 - reduces the amount of exchanged data (the size)
2996 - allows to retrieve all arrived data on receiver side, if we don't need
2997 it right now (subsequent read will peek the buffer)
2998 - allows the receiver to proceed with the begining of the stream before
2999 everything is arrived
3000 - make it possible to build an iov transport (using readv/writev)
3002 - take care of the data with non-stable storage (like stacked data),
3004 * If possible, TCP send uses vector I/O (when writev() is here)
3005 - Don't use it for receive since we send data sizes and data on the
3006 same stream, so we wouldn't be able to chain large amount of chunks
3007 before having to flush the stuff to read the size.
3008 * Rework the transport plugin mechanism to simplify it and reduce the
3009 amount of pointer dereferencement when searching for the right function
3012 * I guess that now, we do almost as few system calls as possible while
3013 doing as few data copy as possible.
3015 To improve it further, we could try to send all the sizes first and then
3016 all the data (to use iov on receiving size), but it's only a partial
3017 solution: when you have 2 dimensional data, the sizes of the second
3018 dimension is data of the first dimension, so you need 3 streams.
3020 I'm not sure the potential performance gains justify the coding burden.
3022 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3024 SimGrid (3.00) stable; urgency=high
3027 * New! Give the possibility to hijack the surf parser and thus bypass
3028 MSG_create_environment and MSG_launch_application. Have a look at
3029 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3031 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3033 SimGrid (2.96) unstable; urgency=low
3038 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3039 This deprecates the xbt_error_t mechanisms.
3040 It modifies (simplifies) all XBT and GRAS API.
3041 MSG API keeps unchanged (exceptions raised by XBT are catched from
3042 within MSG and masked with existing error handling facilities)
3045 * New! Add a FATPIPE model. [AL]
3046 * New! Add a parallel task model. [AL]
3047 * New! Add automatically a loopback interface (in the default
3048 network model) if none was precised.
3051 * Bugfix: MSG_process_resume now works with the current running process.
3053 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3054 * Modification of MSG_task_get_compute_duration. Once a task has been
3055 processed, the value returned by this function is now equal to 0. [AL]
3056 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3057 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3058 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3059 of a m_task. For now, MSG_task_cancel only works with computation
3061 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3062 of the processor (in Mflop/s) regardless of the current load on the
3064 * API Change: use proper naming convention for MSG_getClock and
3065 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3067 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3068 This function changes the priority of a computation task. This priority
3069 doesn't affect the transfer rate. A priority of 2 will make a task
3070 receive two times more cpu power than the other ones. This function
3071 has been added to suit the needs of Nguyen The Loc and hasn't been that
3072 much tested yet. So if it fails, please report it and send me your code.
3074 * API Change: removed all functions and types that were marked "deprecated"
3075 since many months. Renamed MSG_global_init_args to MSG_global_init.
3077 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3079 SimGrid (2.95) unstable; urgency=low
3082 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3083 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3084 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3087 * Bugfix: Allow absolute paths for platform description files [MQ]
3088 * Bugfix: do free the variables after use. Leads to drastic performance
3090 * Implement max_duration (ie, timeouts) on resources [AL]
3093 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3095 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3096 the needed support (timeouts) [AL]
3099 * Implement measurement sockets. You can now get the bandwidth between two
3100 hosts thanks to AMOK (see below). [MQ]
3101 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3102 dynar over the network (yeah) [MQ]
3103 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3104 * simple/static token-ring example. [Alexandre Colucci and MQ]
3105 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3106 in place before [MQ]
3109 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3110 type parsing. "short" and "long" modifiers were also ignored; other
3111 modifier (such as reference level) are still ignored. [MQ]
3112 * Embeed the buffer size within the buffer itself on SG. [MQ]
3113 That way, send() are atomic and cannot get intermixed anymore (at least
3114 the ones which are less than 100k; bigger messages still have the issue)
3115 * Array size pushed by the field, not by the field type (or each
3116 and every long int will push stuff to the cbps) [MQ]
3117 * use select() to sleep since it allows to portably sleep less than one
3120 GRAS (minor cleanups)
3121 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3124 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3125 * type category 'ignored' killed as it was never used and were difficult
3127 * whether a type can cycle or not is now a flag, leaving room for more
3128 flags (as "ignored", if we feel the need one day ;)
3129 * Rename raw sockets to measurement sockets since "raw" has another
3130 meaning in networking community.
3133 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3134 and offers features not belonging to GRAS but that most applications
3135 need. One day, it may be a set of plugins loadable at runtime.
3136 * New module: bandwidth
3137 bandwidth measurement between arbitrary nodes running this module. [MQ]
3139 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3141 SimGrid (2.94) unstable; urgency=low
3143 The first beta release of SimGrid 3 !
3147 * Update the main page and the FAQ. Adding references to gforge.
3150 * Add a gras_os_getpid function.
3153 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3154 * Extend the logs so that they also print PID, hostname, date, ... if
3156 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3157 and kill the old version which were in testsuite/
3158 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3162 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3163 for users. Instead, each programming environment (GRAS, MSG,...) use it
3164 when needed to provide such a feature to users.
3165 Don't shortcut the mechanism or you will also shortcut the virtualization
3166 you need on the simulator.
3170 * Cleanups in configury with regard to compile optimization/warning flags.
3171 Also add -fno-loop-optimize to any powerpc since it's the optimization
3172 killing gcc (< 3.4.0).
3173 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3174 needed by MSG examples complications
3175 * Borrow configury beautifications from PHP
3178 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3179 warning (thanks loris for stumbling into this one).
3180 * Bugfix: stop loading private headers (gras_config.h) from the public
3184 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3185 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3186 RL would imply a DNS resolver, which is *hard* to do in a portable way
3187 (and therefore delayed).
3188 * Implement a real timer mechanism and use it in timing macros. This allows
3189 to avoid rounding errors and get a 0.000005 sec precision in timing
3190 macros. While I was at it, various cleanups:
3191 - allow to declare more than one timed section per file (fix a stupid bug)
3192 - move some private declaration to the right place
3193 - merge conditional execution and timing macros into emulation module
3194 - document the module
3195 - make sure the module cleanups its mess on gras_exit
3196 * Documentation improvements:
3197 - (new) how to compile applications using GRAS
3198 - (new) emulation support (timing macros)
3200 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3202 SimGrid (2.93) unstable; urgency=low
3204 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3207 - Use Paje properly where used. Still to be sanitized properly.
3208 - Portability fix: Add an implementation of the contexts using pthread
3212 - Add xbt_procname(): returns the name of the current process.
3213 Use it to show the current process's name in all logging.
3215 - fix detection of older flex version and the reaction, since we do
3216 depend on modern ones (we use lex_destroy)
3217 - Better separation of SG and RL in the libs: remove all simulation code
3218 from libgras. As a result, this lib is now only 200k when stripped.
3219 Some of the xbt modules may also be duplicated (two sets and such) and
3220 should be cleaned/killed before SG3.
3221 - Insist on using xlC on AIX because of weird problems involving gcc there.
3222 - Cleanup the make remote stuff. This is now done by scripts
3223 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3224 mainly for our private use, but we're working on changing them to user
3227 - Bugfix: flush the socket on close only if there is some *output*.
3228 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3230 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3233 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3235 SimGrid (2.92) unstable; urgency=low
3237 Alpha 3 on the path to SimGrid 3
3241 - New! First try of benchmarking macros.
3242 - New! First try so that gras_stub_generator generate deployment and
3243 remote compilation helpers.
3245 - Bugfix: Initialization fix in msg_test.
3249 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3251 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3253 - API Change: gras_os_sleep to take the amount of seconds as a double.
3254 Accepting an int was error prone since it was the only location where
3255 seconds were coded as such. It leaded to damn rounding errors.
3256 - Bugfix: Hard to belive that timers ever worked before this.
3258 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3260 SimGrid (2.91) unstable; urgency=low
3262 Alpha 2 on the path to SimGrid 3
3266 - Bug fix in the lmm_solver.
3268 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3269 through the function MSG_paje_output.
3270 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3271 - It is possible to bound the rate of a communication in MSG with
3272 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3273 to write it in the changelog).
3274 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3279 - Various cleanups to the autotools stuff
3280 - Begin to move Gras examples to examples/gras/
3281 - Let make distcheck work again (yeah!)
3283 - documentation overhauled using doxygen.
3284 gtk-doc-tools is dead in SimGrid now.
3285 - Automatically extract all existing logging categories, and add the list
3286 to the documentation (long standing one, to say the less)
3288 - Cleanup the known architecture table. Reorder the entries to group what
3289 should be, and use a more consistent naming scheme.
3290 (some of the test dataset are still to be regenerated)
3291 - New! Allow library to register globals on each process just as userdata
3293 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3294 lookup time (for now).
3295 Use it in msg and trp.
3296 This cleans a lot the internals and helps enforcing privacy of the
3297 headers between the gras components.
3298 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3299 - Bugfix: gras_os_time was delirious in RL.
3300 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3302 - Reenable GRAS now that it works.
3304 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3306 SimGrid (2.90) unstable; urgency=low
3308 Alpha 1 on the path to SimGrid 3
3310 * It is a long time since the last release of SimGrid. I'm sorry about
3311 that but as I had told you, I was rewriting a lot of things. I apologize
3312 to those who had been reporting bugs to me and that I had not answered.
3313 If your bug is still in the new version, please tell me. Here is a
3314 summary of the main changes.
3316 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3317 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3318 portability, speed, and a lot more but you'll figure it out later.
3319 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3320 I wanted to release the new MSG as soon as possible and I have
3321 broken GRAS, which is the reason why, for now, only MSG is fully
3322 functional. A laconic description of these projects is available
3323 in the documentation.
3325 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3326 kernel optimized for our needs (called SURF but only the developers
3327 should use it). Hence, MSG is now roughly 30 times faster and I think
3328 that by rewriting a little bit MSG, I could event speed it up a little
3329 bit more. Beside the gain in speed, it is also much easier to encode a
3330 new platform model with SURF than it was with SG. More to come...
3332 * REVOLUTION 3: I have tried to change a little as possible the API of
3333 MSG but a few things really had to disappear. The main differences
3334 with the previous version are:
3335 1) no more m_links_t and the corresponding functions. Platforms are
3336 directly read from a XML description and cannot be hard-coded
3337 anymore. The same format is used for application deployment
3338 description. The new format is described in the documentation.
3339 Have a look in tools/platform_generation. There is a tiny script
3340 that converts from the old platform format to the new one. Concerning
3341 the application deployment format, parsing the old one is tricky.
3342 I think most of you should however be able to convert your files. If
3343 it is really an issue, I can write a C code that does the conversion.
3345 2) the toolbox tbx does not exist anymore. We now have a library
3346 with much more data-structures but without the hash-tables (we have
3347 dictionaries that are much faster).
3349 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3351 *****************************************************************************
3352 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3353 * don't feel like converting the oldies. (Mt) *
3354 *****************************************************************************
3357 Version 2.90: "the long awaited one"
3358 - Finished rewriting and debugging MSG. Rewrote the documentation.
3359 - disable GRAS for now since it needs to be ported to the newest SG
3362 - Finish the port to windows (using mingw32 for cross-compile)
3365 - Main loop and datastructures of SURF. A cpu resource object is
3366 functional. Surf can thus be used to create cpu's with variable
3367 performance on which you can execute some actions.
3369 2004-11-15 Martin Quinson
3370 - Port to ARM. Simply added the alignment and size descriptions. Should
3371 work, but the ARM machines are so slow that I didn't had the opportunity
3372 to 'make check' over there yet.
3374 2004-11-15 Arnaud Legrand
3375 - Trace manager now written. It uses a heap structure and is therefore
3376 expected to be efficient. It may however be speeded up (particularly
3377 when many events occur at the same date) by using red and black
3378 trees. One day maybe...
3379 - Max-min linear system solver written. It uses a sparse matrix
3380 structure taking advantage of its expected use. Most operations are
3381 O(1) and free/calloc are called as few as possible. The computation of
3382 the minimum could however be improved by using a red and black tree
3385 2004-11-03 Arnaud Legrand
3386 - Rename every gras_* function that was in xbt/ to its xbt_
3388 - Add a heap and a doubly-linked list to xbt
3389 - Added a dichotomy to the dictionaries. make check works as well before
3390 so I assume that the patch is correct. I do not know however if things
3391 run effectively faster than before now. :)
3393 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3394 SimGrid, and the version number is bumped to 2.x
3396 2004-10-29 Martin Quinson
3397 - Introduction of the remote errors.
3398 They are the result of a RMI/RPC on the remote machine.
3399 ErrCodes being scalar values, you can't get the host on which those
3400 errors did happen. Extending the error mechanism as in Gnome is possible.
3401 No idea yet whether it is a good idea.
3403 2004-10-28 Martin Quinson
3404 - Interface revolution: the Starred Structure Eradication.
3405 I used to do typedef struct {} toto_t; and then handle *toto_t.
3406 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3407 - ???_t is a valid type (builded with typedef)
3408 - s_toto_t is a structure (access to fields with .)
3409 - s_toto is a structure needing 'struct' keyword to be used
3410 - e_toto_t is an enum
3411 - toto_t is an 'object' (struct*)
3413 typedef struct s_toto {} s_toto_t, *toto_t;
3414 typedef enum {} e_toto_t;
3415 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3416 s_toto_t) is private.
3418 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3419 so that it can set the variable to NULL. It was so for dicts and sets,
3420 it changed for dynars.
3422 - Fix a bunch of memleaks in dict_remove
3423 - Fix a bug in sg/server_socket opening: it failed all the time.
3425 2004-10-07 Martin Quinson
3426 - Speed up dynar lookup operation a bit.
3428 gras_dynar_get is dead.
3430 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3431 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3432 (which gives you the address of the stored data).
3434 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3435 copy of the data using an affectation to do the job and not a memcpy.
3437 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3438 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3440 It does not really speedup the dynar test because they are
3441 setting elements all the time (and look them seldom). But the dict does
3442 far more lookup than setting.
3444 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3446 2004-10-05 Martin Quinson
3447 - Allow to (en/dis)able the cycle detection at run time.
3449 Whether we should check for cycle or not is now a property of each
3450 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3451 datadesc_cycle_unset allow to remove this property when previously set.
3453 Note that the cycle detection is off by default since it impacts the
3454 performance. Watch the data you feed GRAS with ;)
3456 This property is hereditary. Any element embedded in a structure having it
3457 set have it set for the time of this data exchange.
3459 You should set it both on sender and receiver side. If you don't set it on
3460 sender side, it will enter an endless loop. If you forget on receiver
3461 side, the cycles won't be recreated after communication.
3463 - Header reorganization.
3464 Kill gras_private.h, each submodule must load the headers it needs.
3466 2004-10-04 Martin Quinson
3467 - Interface revolution: do not try to survive to malloc failure.
3469 Now, gras_malloc and friends call gras_abort() on failure.
3470 As a conclusion, malloc_error is not a valid error anymore, and all
3471 functions for which it was the only gras_error_t return value are
3472 changed. They now return void, or there result directly.
3473 This simplify the API a lot.
3475 2004-09-29 Martin Quinson
3476 - Re-enable raw sockets.
3477 Created by gras_socket_{client,server}_ext;
3478 Used with gras_raw_{send,recv}
3481 It should allow to kill the last bits of gras first version soon.
3483 This is not completely satisfactory yet (duplicate code with
3484 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3487 - Simplify transport plugin (internal) interface by not passing any
3488 argument to _server and _client, but embedding them in the socket
3491 2004-09-28 Martin Quinson
3492 - Finish the port to AIX.
3493 autoconf was my problem (segfault within the malloc replacement
3494 function. No idea why)
3496 2004-09-16 Martin Quinson
3497 - Fix some size_t madness on 64bit architectures.
3499 2004-09-08 Martin Quinson
3500 - Reduce the number of system headers loaded, overload some more system
3501 calls (such as malloc to cast the result of the system one, and work
3503 - Fix and reintroduce the config support
3505 2004-09-07 Martin Quinson
3506 - Source code reorganization to allow Arnaud to surf all over there.
3507 - Allow to document the logging categories.
3508 - Remove all uppercase from logging categories and useless cleanup in names.
3510 2004-08-18 Martin Quinson
3511 Version 0.6.2 (protocol not changed; API changed)
3512 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3513 gras_error_t), and NULL when not found. Functions expecting a msgtype
3514 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3515 hopefully usefull message.
3516 - Portability to prehistoric sparcs again
3518 2004-08-17 Martin Quinson
3519 Version 0.6.1 (protocol not changed; ABI not changed)
3520 - prealloc some buffers to speed things up
3522 2004-08-11 Martin Quinson
3523 Version 0.6 (protocol not changed; ABI expended)
3524 - The parsing macro can deal with the references, provided that you add
3525 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3527 2004-08-09 Martin Quinson
3528 Version 0.5 (protocol not changed; ABI changed)
3529 - Allow to off turn the cycle detection code in data exchange at
3530 compilation time. It should be at run time, but I'm short of time (and
3531 the config stuff is still broken). That way, we keep dict out of the
3532 critical path, which is good because the performance is poor:
3533 - search not dichotomial yet
3534 - dynar give no way to access their content and memcpy everytime
3535 - In composed data description (struct, ref and so on), stop foolness of
3536 keeping the subtype's ID, but store the type itself. This keeps sets out
3537 of the critical path, which is good since they rely on dynar and
3538 dictionnaries. The only loose of that is that we cannot detect the
3539 redeclaration of a structure/union with another content (but I'm not sure
3540 the code detected well this error before anyway). We still can detect
3541 the redefinition discrepancy for the other types.
3542 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3543 it breaks the code because of type-punning used all over the place).
3544 This breaks on all non-gcc architectures (for now).
3546 All those changes (plus the buffer of last time) allow me to gain 2 order
3547 of magnitude on cruel tests consisting of 800000 array of integers on two
3548 level of a hierarchical structure (200 secondes -> 4 secondes)
3551 - the selector of reference must now return the type it points to, not
3552 the ID of this type.
3554 2004-08-06 Martin Quinson
3555 Version 0.4 (protocol changed; ABI not changed)
3556 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3557 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3558 - Implement a buffer transport, and use it by default (it relies on tcp in
3559 real life and on sg in simulation).
3560 That's a bit hackish since I had a new field to the structure to store
3561 its data without interfering with the subtype ones. Inheritance
3562 is tricky in C. And that's a kind of reverse inheritance with one class
3563 derivating two classes. Or maybe a game with java interfaces. Anyway,
3564 that's damn hard in C (at least).
3565 Moreover, I got tired while trying to ensure plugin separation and
3566 genericity in SG mode. MSG wants me to do weird things, so let's go for
3567 cruel hacks (temporarily of course ;).
3568 See comment in transport_private.h:71
3569 - do not include all the _interface headers in private but in the files
3570 which really need them (to cut the compilation time when they are
3573 2004-07-26 Martin Quinson
3574 Version 0.3 (protocol not changed; ABI changed)
3575 - Major overhault of the datadesc interface to simplify it:
3576 - shorted the function names:
3577 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3578 - add a trivial way to push/pop integers into the cbps without malloc.
3579 This allows to make really generic sub_type description, which simply
3580 pop their size of the stack.
3581 - add a function gras_datadesc_ref_pop_arr() which does what users want
3582 most of the time: Declare a dynamic array (which pops its size of the
3583 stack) and declare a reference to it. Poor name, but anyway.
3584 - kill the post-send callback, add a post-receive one
3586 2004-07-23 Martin Quinson
3587 Version 0.2 (protocol changed; ABI changed)
3588 - add some testing for cpbs in the test cases, and fix some more bugs.
3589 This invalidate again the little64 data file, since I cannot regenerate
3591 - remove an awfull optimization in the logging stuff, allowing me to:
3592 - understand it again
3593 - learn gcc how to check that the argument match the provided format
3594 - fix all errors revealed by gcc after that
3595 - internal keys of dict are not \0 terminated. Deal with it properly in
3596 loggings instead of segfaulting when the user want to see the logs :-/
3598 2004-07-22 Martin Quinson
3599 - Fix some stupid bug preventing cbps (callback postit) from working
3601 2004-07-21 Martin Quinson
3602 - Some documentation cleanups
3603 - remove the useless last argument of msgtype_declare
3604 - rename the Virtu functions to fit into the 'os' namespace
3605 - move headers src/include -> src/include/gras/ and stop fooling with
3606 gras -> . symbolic link
3607 - make distcheck is now successful
3609 2004-07-19 Martin Quinson
3611 - Build shared library also
3612 - Install html doc to the right location
3613 - stop removing maintainer files in make clean
3614 - build tests only on make check
3616 2004-07-13 Martin Quinson
3618 - No major issue in previous version => change versioning schema
3619 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3620 corresponding dataset.
3622 2004-07-11 Martin Quinson
3624 - Get it working with any kind of structure (we can compute the padding
3625 bytes remotely for all the architectures I have access to)
3626 - Implement the structure parsing macro (still not quite robust/complete)
3627 - Improvement to the remote testing toysuite
3629 2004-07-10 Martin Quinson
3630 [autoconf mechanism]
3631 - get ride of a bunch of deprecated macros
3632 - actually run the test for two-compliment, not only compile it :-/
3633 - test whether the structures get packed (and bail out if yes. Damn.
3634 Alignment is a serious matter)
3635 - test whether the structures get compacted (but respecting the alignment
3636 constraints of each types)
3637 - test whether the array fields of structures can straddle alignment boundaries
3639 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3640 (took me 2 hours to find that bug, looking at the wrong place)
3641 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3642 before sending/receiving and are used to compute the offsets of fields
3643 - Given that padding size depend even on compiler options, keep track of
3644 alignment and aligned_size only for the current architecture. Not a big
3645 deal since we send structure fields one after the other (seems
3647 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3648 program, let it work on linux/gcc/little32. portability todo.
3650 2004-07-08 Martin Quinson
3651 - import and improve remote compilation support from FAST
3652 - make sure make check works on half a dozen of machines out there
3654 2004-07-07 Martin Quinson
3655 Let's say it's version 0.0.3 ;)
3656 - Implement conversions (yuhu!)
3657 - Let it work on solaris (beside conversion, of course)
3658 - Stupid me, using rand() to generate the conversion datatests in not wise.
3660 2004-07-06 Martin Quinson
3661 - Let make dist work, since I'm gonna need it to compile on remote hosts
3662 - Let Tests/datadesc_usage write the architecture on which the file was
3663 generated as first byte.
3664 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3665 architecture descriptions.
3666 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3667 on those architectures.
3668 - Optimization: send/recv array of scalar in one shoot
3670 2004-07-05 Martin Quinson
3671 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3673 - Plug a whole bunch of memleaks
3674 - each process now have to call gras_{init,exit}. One day, their log
3675 settings will be separated
3676 - Continue the code factorisation between SG, RL and common in Transport.
3678 2004-07-04 Martin Quinson
3680 - Redistribution between SG and RL.
3681 We wanna have to accept in SG, so move accepted related parts of RL in
3682 the common part. (more precisely, the dynar of all known sockets is no
3683 more a static in transport.c, but part of the process_data)
3685 [gras_stub_generator]
3686 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3688 2004-07-03 Martin Quinson
3689 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3690 them all up in one shot)
3692 - Fix the ugly case of reference to dynamic array.
3693 - New (semi-public) function gras_datadesc_size to allow the messaging
3694 layer to malloc the needed space for the buffer.
3696 - gras_socket_close now expect the socket to close (and not its address to
3697 put NULL in it after it). This is because the socket passed to handlers
3698 is one of their argument (=> not writable).
3700 - propagate the interface cleanup from last week in datadesc, ie remove a
3701 superfluous level of indirection. User pass adress of variable
3702 containing data (both when sending and receiving), and not of a variable
3703 being a pointer to the data. Let's say that I like it better ;)
3704 The price for that is constructs like "int msg=*(int*)payload" in
3705 handlers, but it's a fine price, IMHO.
3707 - Let it work in RL (yuhu)
3709 2004-06-21 Martin Quinson
3711 - porting SG plugin and SG select to new standards (works almost).
3712 - plug memleaks and fix bugs around.
3715 - cleanup the prototype of data recv and force users to specify when they
3716 want to handle references to objects. Test case working even for cycles.
3717 - plug memleaks. Valgrind is perfectly ok with this.
3719 2004-06-12 Martin Quinson
3721 - cleanup the separation between plugin and main code in plugin creation
3723 2004-06-11 Martin Quinson
3725 - Reput hook for raw sockets, needed for BW experiments
3726 - kill a few lines of dead code
3727 [Data description] Interface cleanup
3728 - gras_datadesc_by_name returns the searched type or NULL.
3729 That way, no variable is needed to use a type desc once, which makes
3731 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3732 two parameters were strings (field name, type name), leading to
3734 [Dicos] Interface cleanup
3735 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3736 This is consistant with the dynar API.
3738 2004-04-21 Martin Quinson
3740 - Porting to new standards.
3742 - interface cleanup.
3743 There is no bag anymore, no need to take extra provision to mask the
3744 pointers behind "ID".
3745 Better splitup of functions between files create/exchange/convert.
3746 This is still a bit artificial since convert and receive are so
3747 interleaved, but anyway.
3749 - add a queued message list to procdata (the ones not matching criteria
3751 - factorize some more code between SG and RL wrt procdata
3753 - use gras_exit in example to track memleaks
3754 - get rid of gs_example now that GS is properly integrated into gras
3755 - update run_test to integrate the latest tests (datadesc)
3757 - rename WARNINGn macros to WARNn since it prooved error-prone
3759 2004-04-19 Martin Quinson
3761 - register init/exit functions within gras module mechanism
3762 - send/receive function.
3763 Convertion is not implemented, but short-cutted if not needed.
3764 struct/array elements are sent one by one (instead of block-wise), but
3765 nobody really cares (yet). Get a prototype before optimizing.
3766 - tests (using a file socket) for DD send/receive on:
3767 - base types: int, float
3768 - array: fixed size, string (ie ref to dynamic string)
3769 - structure: homogeneous, heterogeneous
3770 - chained list, graph with cycle
3771 Believe it or not, valgrind is not too unhappy with the results. The
3772 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3776 - Bugfix when using a filename explicitely (instead of '-')
3778 2004-04-09 Martin Quinson
3780 - factorize more code between RL and SG in socket creation
3781 - Complete the implementation and tests of:
3783 o file (only in RL, and mainly for debugging)
3785 I lost 3 days to design a portable address resolver, and then decided
3786 that the prototype mainly have to run on my box.
3787 Addressing portability too early may be like optimizing too early :-/
3789 - use gras_init in the Tests instead of the crappy parse_log_opt
3790 (the latter function is removed)
3791 [Conditional execution]
3792 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3793 [Code reorganisation]
3794 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3795 Build examples against the RL library, since there is no way to disable
3796 its creation for now.
3798 For information, the beginning of coding on GRAS was back in june
3799 2003. I guess that every line has been rewritten at least twice since