1 SimGrid (3.21) NOT RELEASED (Release Target: September 23. 2018, 1:54 UTC)
3 The Restarting Documentation (TRD) Release.
6 - Convert the user manual to Sphinx for improved usability.
7 Unstable version now hosted on https://simgrid.frama.io/simgrid
8 Many glitches remain; stay tuned.
9 - Restrict the Doxygen reference API to the installed header files.
10 (documenting all internals this way is overkill + doxygen scales badly)
11 - New tutorial on S4U (highly inspired from the old MSG one, but with
12 a git repository to fork as a starting point, and a docker image)
13 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
14 - The Developper manual is still to be converted (not compiled until then)
17 - Fully reimplement auto_restart mechanism that was utterly broken
18 (fix #22, #131 and #281 that were all very related)
19 - When creating an actor from a function and its parameters,
20 move-only parameters are not allowed anymore, as it would prevent
21 the actor to be restartable if its parameters are consumed on run.
22 - s4u::Io: IOs go asynchronous as activities. This comes with new methods in the
24 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
25 IO operations that can be started, waited for, or canceled as a regular
27 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
31 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
32 - New host variable: 'core_count'
35 - Remove xbt_os_thread_specific features
36 - Remove portability wrapper to condition variables
37 - Remove xbt_os_thread_yield()
40 - MPICH collective selector now mimics MPICH 3.3b
41 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
42 - MPI_Init(NULL,NULL) is now allowed
43 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
44 - Command-line arguments should now be handled properly in Fortran simulations
45 - MPI Attributes and related callbacks should work in Fortran
46 - Apps using C + Fortran should now work
47 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
48 - Support MPI_Error_String
51 - Due to an internal bug, Msg.run() must now be your last line.
52 We hope to fix it in a future release, and we are sorry for the inconvenience.
55 - #22: Process autorestart seem to only work with CAS01 cpus
56 - #56: Feature request: dynamic SMPI replay
57 - #93: simgrid should not eat --help
58 - #111: How to change the loopback link characteristics?
59 - #116: Platform cluster radical
60 - #117: Platform cluster/host order
61 - #129: Replay_multiple on steroid
62 - #131: Java : autostart does not work
63 - #143: Setting a breakpoint at a given time
64 - #235: Network energy models should be integrated
65 - #264: Add ptask L07 resource tracing
66 - #271: Dynamic and manual replay of SMPI traces
67 - #279: Breakpoints option uses time deltas instead of absolute time
68 - #281: Daemonized actor no longer daemonized after an auto-restart
69 - #282: TIT convertor erase traces when using absolute path in the trace list file
70 - #285: segfault when a s4u actor kills itself with msg process tracing activated
71 - #286: Pajé traces are not exposing the number of cores
72 - #287: Command-line parsing should stop when encountering '--'
73 - #288: MPI_Init(NULL, NULL)
74 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
75 - #290: Method to check if a mailbox has a message ready to be consumed
76 - #294: Alltoall collective implementation of mpich
77 - #295: Starting a Host as OFF is not supported
78 - #296: DTD too permissive
79 - #299: Add a s4u barrier example
81 ----------------------------------------------------------------------------
83 SimGrid (3.20) Released June 24. 2018
85 The proxy snake_case() release.
88 - The whole API is now using snake_case() instead of mixing it with camelCase.
89 Compatibility wrappers in place for one year.
90 - We took the opportunity to increase the consistency of the API.
91 Please report any remaining glitches.
92 - New function simgrid::s4u::Host::get_actor_count:
93 Returns the number of actors running on a specific host.
96 - SMPI is now tested with ~45 proxy apps from various sources, with none or
97 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
98 - Replay: The replay file has been re-written in C++.
99 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
100 supported. They are stored in the trace and used when replayed.
101 - Basic support of MPI_Cancel. Robustness not guaranteed.
102 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
103 (only for MPI_COMM_TYPE_SHARED).
104 - New option: smpi/privatize-libs, to add external shared libs to be privatized
105 by SMPI. They will be copied locally and loaded separately by each process.
106 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
107 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
108 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
109 with internal index optind with multiple processes. Only works if MPI_Init has
111 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
112 syntax without preprocessing tricks.
115 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
116 behavior tracing is no longer limited to MSG
117 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
118 this means that traces now need to be updated or re-obtained; both Irecv
119 and Isend lines in your traces have now in total 5 mandatory fields:
120 <rankid> <command> <to/from rankid> <tag> <size>
121 To update your traces, it suffices to add a 0 for the tag here.
122 - TIT now also supports waiting for a distinct request via MPI_Wait.
123 Wait/Test now wait for a specific request, not just the last one that was
124 issued. This unfortunately means another update, because we need to
125 identify which request you want to wait for. We do this via the
126 triplet (sender, receiver, tag), which needs to be added:
127 <rankid> <command> <sender> <receiver> <tag>
128 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
129 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
132 - The deprecation of MSG is ongoing (but this should not impact you).
133 Many MSG functions are now simple wrappers on the C API of S4U. If
134 you wish to convert your code to S4U, find the S4U counterparts of
135 your MSG calls in src/msg/msg_legacy.cpp.
136 - MSG can still be used, but won't evolve anymore.
139 - Allow to run the Link energy plugin from the command line with
140 --cfg=plugin:link_energy
141 - Rename Energy plugin into host_energy
142 - Rename Load plugin into host_load
145 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
146 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
147 - Improve performance when handling timeouts of simix synchros.
150 - Config: the C API is now deprecated (will be removed in 3.23), and
151 the C++ API has been slightly improved.
154 - Fix several build issues on OSX.
155 - Move simgrid_config.h to simgrid/config.h (old header still working)
158 - #143: Setting a breakpoint at a given time
159 - #150: Inconsistent event names in SMPI replay
160 - #258: daemonized actors hang after all non-daemonized actors have completed
161 - #267: Linker error on unit_tmgr
162 - #269: SMPI: tracing of MPI_Wait/all/any broken
163 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
165 ----------------------------------------------------------------------------
167 SimGrid (3.19.1) Released March 22 2018
169 The Fixed ABI Release.
171 As you may know, we are currently refactoring SimGrid in deep.
172 Upcoming SimGrid4 will be really different from SimGrid3: modular,
173 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
175 Our goal is to smooth this transition, with backward compatibility
176 and automatic update paths, while still progressing toward SimGrid4.
178 SimGrid remains open during works: The last pure SimGrid3 release was
179 v3.12 while all subsequent versions are usable alpha versions of
180 SimGrid4: Existing interfaces remain unchanged, but the new S4U
181 interface is budding and the internals are deeply reorganized.
183 Since 2015, we work hard to reduce the changes to public APIs. When
184 we need to rename a public library symbol in S4U, we let your
185 compiler issue an explicative warning when you use the deprecated
186 function. These messages remain for four releases, i.e. for one full
187 year, before turning into an error. Starting with v3.15, your can
188 also adapt to API changes with the SIMGRID_VERSION macro, that is
189 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
191 Starting with this v3.19.1, our commitment to reduce the changes to
192 the public interfaces is extended from the API to the ABI: a program
193 using only MSG or SimDag and compiled against a given version of
194 simgrid can probably be used with a later version of SimGrid without
195 recompilation. We will do our best... but don't expect too much of
196 it, that's a really difficult goal during such profund refactoring.
198 The difference between v3.19 and v3.19.1 is that the former was
199 accidentally breaking the ABI of MSG, while the later is restoring
202 S4U and kernel APIs will still evolve until SimGrid4, with one-year
203 deprecation warnings as currently. In fact, cleaning up these
204 interfaces and converting them to snake_case() is one release goal of
205 v3.20. But don't worry, we are working to smooth this upgrade path.
207 In summary, new projects should start with S4U to benefit of the
208 future, but old MSG projects should still be usable with no change.
210 ----------------------------------------------------------------------------
212 SimGrid (3.19) Released March 20 2018
214 The Moscovitly-cold Spring release.
216 SMPI user-visible improvements:
217 - Support of MPI_Type_create_subarray
218 - Unimplemented MPI calls now abort (instead of issuing a warning)
219 Some harmless calls around errhandlers still only issue a warning.
220 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
221 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
222 is now possible (mmap privatization was refactored to this end).
223 Adding new SMPI instances should also be technically doable now.
225 SMPI internal cleanups:
226 - Removed index notion from SMPI -- no more "getPid() - 1"!
227 In tracing files, actors are called rank-<PROCESS-ID> but these are
228 NOT the real ranks (tracing several communicators is not supported yet).
229 - Deprecate smpi_process_index() (will be removed in 3.22)
230 - Better testing on classical proxy apps, through an external project:
231 https://github.com/simgrid/SMPI-proxy-apps/
234 - Use a graphical TOC to make it easier to find the documentation you need
235 - Revamp the MSG tutorial
238 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
239 - Remove parameter of MSG_process_killall().
240 Resetting the PID was bogus anyway (several actors could have the same PID).
243 - Execution->setHost() can be called after start(), to migrate it.
244 - Comm::test_any() is now implemented.
245 - s4u::Actor now has onCreation() and onDestruction() signals.
246 - Install some kernel header files for the users' plugins and more.
247 Warning, their API is really not stable yet.
248 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
249 include/simgrid/kernel/routing/*Zone.hpp
252 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
254 Build System and other cleanups:
255 - Remove unused run-time parameter "tracing/onelink-only".
256 - SimGrid now works with intel compilers (MC still buggy)
257 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
258 - tesh kills subprocesses on timeout (but not on windows)
261 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
264 - #194: Feature request: simgrid::s4u::Comm::test_any()
265 - #245: migrating an actor does not migrate its execution
266 - #253: Feature Request: expose clusters as objects
267 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
268 - #255: Tesh broken on Windows
269 - #256: Modernize FindSimGrid.cmake
270 - #257: Fix (ab)use of CMake install
272 ----------------------------------------------------------------------------
274 SimGrid (3.18) Released December 24 2017
276 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
278 MSG IS NOW DEPRECATED. Please use S4U for new projects.
279 - Support for MSG should not be removed from SimGrid before 2020,
280 but future evolutions will be limited to the integration of
281 user-provided patches.
284 - s4u::onDeadlock() signal, run before stopping the simulation.
285 - s4u::Exec: asynchronous executions (abstraction of background
287 - s4u::Host->getLoad() returns the achieved speed in flops/s
290 - Link::name() is deprecated and will be removed in v3.21, use
291 Link::getCname() instead.
292 - Mailbox::getName() changed to return a std::string, use
293 Mailbox::getCname() instead to get a char*.
294 - Storage::getName() changed to return a std::string, use
295 Storage::getCname() instead to get a char*.
296 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
297 to have both allocation and deallocation of the map in user space.
300 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
301 The amount of flops remaining to do is a vector, not a scalar.
302 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
303 name implies on both sequential and parallel tasks.
304 - Both changes fix GitHub's #223 using PR #237 as a basis.
305 Thanks Michael Mercier.
306 - Most examples were converted to S4U and hidden elsewhere at tests.
307 Rationale: we still want MSG to work; we want newcomers to use S4U.
310 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
311 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
314 - Switch to the faster dlopen privatization mechanism by default
315 - Documentation improvments
318 - Live migration is getting moved to a plugin. Dirty page tracking is
319 the first part of this plugin. This imply that VM migration is now
320 only possible if one this function is called:
321 - C/MSG: MSG_vm_live_migration_plugin_init()
322 - C/C++: sg_vm_live_migration_plugin_init()
323 - Java: Msg.liveMigrationInit()
324 For C and C++, "simgrid/plugins/live_migration.h" also has to be
328 - Define class simgrid::xbt::Path to manage file names.
329 - Removed unused functions:
330 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
331 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
332 xbt_graph_export_graphviz()
333 - xbt/str.h: xbt_str_join()
334 - Remove unused datatypes:
335 - xbt/heap.h: use std::priority_queue or boost::heap instead
336 - xbt/swag.h: use boost::intrusive::list instead
339 - New link_energy plugin for the consumption of the links.
340 - All of the operations on files and storage contents have been
341 packaged into a plugin (src/plugins/file_system). The current
342 public interface can be found in
343 include/simgrid/plugins/file_system.h
344 To use these functions you now have to initialize the plugin by
345 calling MSG_storage_file_system_init() just after calling
346 MSG_init() or sg_storage_file_system_init() just after creating
350 - Remove the undocumented/untested tag <include>
353 - Remove viva specific tracing as the tool is no longer maintained
356 - #248: Kill of finished processes leads to segfault
357 - #240: xbt_cond_wait_timeout should gracefully return for C apps
358 - #239: please implement signal s4u::onDeadlock()
359 - #230: segfaults when exit() without run()
360 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
361 - #223: MSG_task_get_flops_amount() not working with parallel tasks
362 - #222: Actor::kill() doesn't really kill and segfaults
363 - #221: odd LMM warning when killing an actor
364 - #120: Memory leak when the processes are forcefully killed
366 ----------------------------------------------------------------------------
368 SimGrid (3.17) Released October 8 2017
370 The Drained Leaks release: (almost) no known leaks despite the tests.
372 Even very long simulations will preserve your memory: our testsuite
373 stresses SimGrid for over 45mn on fast machines for a coverage of
374 over 80%, and there is only one single known leak, of about 4kb.
378 - Comm.detach(): start and forget about asynchronous emission. A cleanup
379 handler may be given to free resources if the comm cannot be completed.
380 - this_actor::send(mailbox) is now mailbox->put()
381 - New: simgrid::s4u::Comm::wait_all()
382 - New: Host.execute() for remote executions.
385 - Unused option network/sender-gap is removed.
388 - Deprecate MSG_task_isend_with_matching(): This unused feature
389 really complicates our internals. Will be removed in v3.20.
392 - Improved context termination. It is now possible to free resources, even
393 when a process is forcibly killed.
394 - ContextBoost: add support for Boost versions above 1.61.
397 - Bring back run-time option --cfg=exception/cutpath to remove exception
399 - Removed unused functions:
400 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
401 xbt_str_rtrim(), xbt_str_trim().
402 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
405 - Removed header files obsolete since SimGrid 3.12:
406 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
408 - #3: SD_exit should be made optional
409 - #120: Memory leak when the processes are forcefully killed
410 - #159: Threading test regression in Actor refcounting
411 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
412 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
413 - #186: Actor::killAll() segfaults if some process is blocked on wait()
414 - #191: VM migration and pstate
415 - #192: Updating the energy consumptions of all hosts crashes with VMs
416 - #195: All actors have PID=0 in the logs
417 - #204: Sometimes segfault with thread contexts and mmap privatization
418 - #222: Actor::kill() doesn't really kill and segfaults
419 - #225: Actor::kill() doesn't really kill when victims are doing a join()
421 ----------------------------------------------------------------------------
423 SimGrid (3.16) Released June 22. 2017.
425 The Blooming Spring Release: developments are budding.
428 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
429 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
430 - Java JAR file should be finally fully working on Mac OSX too.
432 XML platforms: Switch to platform v4.1 format.
433 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
434 - <zone> can be used as a synonym for the now deprecated <as>
435 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
436 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
437 - <actor> can be used as a synonym for the now deprecated <process>
438 - state_file and avail_file periodicity is now easier to express
439 (check the documentation) the old behavior should still work.
440 * Storage: not backward compatible, but it seems that nobody ever used it
441 - Remove attribute 'content_type' of <storage_type>: was never used
442 - Make attribute 'model' of <storage_type> optional: for future usage
443 - Remove Bconnection model property: was never (in)validated. Replaced by
444 the maximum of the read and write bandwidth as a resource constraint.
447 - New and Backwards Compatibility break:
448 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
449 When one wants to get the list of tasks whose states have changed during a
450 simulation round, s/he has to allocate and free a dynar and use it as
451 argument to this function. The former SD_simulate (double how_long)
455 - Allow multicore VMs, with the correct sharing (unless you overcommit)
456 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
459 - New function to update the consumption of all hosts at once.
460 - Fix the model for multi-core hosts, linear in the amount of busy
461 cores with an abnormality for pIdle (see paper).
462 BUG: Still not satisfactory for ptask on mono-cores.
465 - The netzone are now available from the MSG API.
466 The old names still work, but are now deprecated.
469 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
470 - New option: smpi/keep-temps to not cleanup temp files
471 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
472 mallocs mode is used, allows to change the size of the fake file used
473 (default 1MB), to potentially limit the number of mappings for large runs.
474 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
475 - Fortran ifort and flang compilers support
476 - New RMA calls supported (experimental):
477 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
478 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
479 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
480 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
481 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
482 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
485 - New: this_actor::isMaestro() returns whether we are in kernel mode.
486 - Behind the scene, ActivityImpl are now automatically refcounted.
487 Removing this nasty bottleneck will greatly simplify our internals.
490 - Replay: New function xbt_replay_action_get():
491 Retrieve the function previously associated to an event type.
492 - DROPPED FUNCTION: xbt_str_varsubst()
493 - DROPPED MODULE: strbuff. We don't need it anymore.
494 - DROPPED MODULE: matrix. We don't need it anymore.
495 - DROPPED MODULE: lib. We don't need it anymore.
497 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
499 SimGrid (3.15) Released March 22. 2017
501 The Spring Release: continuous integration servers become green
502 We fixed even the transient bugs on all target architectures:
503 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
504 Windows, FreeBSD, NetBSD.
506 - Rename NetCards to NetPoints.
507 This was intended to help NS3 users, but that's not a netcard.
508 That's a point in the routing algorithm, let's avoid wrong simplifications.
511 - New: MSG_process_yield(). Stop and yield to other processes.
512 - New: MSG_process_daemon(). Daemon processes are automatically killed
513 when the last non-daemon process terminates
514 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
515 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
516 to make room for the upcoming network energy plugin.
517 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
520 - Ensure that an actor can kill itself with Process::exit()
521 - Kill the obscure NativeException. Nobody want to survive the issues
522 it denotes, so use JniException that is a RuntimeException (not to
523 be caught explicitly).
524 - Partial bug fix in initialization. SimGrid flags on command line were
525 consumed at C level but stayed in the original Java String[] args.
526 This could mess users' args[i] if SG flags were not put at the end of
528 The SimGrid flags are now removed from the Java arguments. However,
529 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
530 if args.length is greater than the number of YOUR OWN ARGUMENTS.
531 It might be if you have --log or --cfg flags in the command line.
532 - Fix numerous memleaks all around the place. In particular, around VMs.
536 - s4u::onPlatformCreated: right before the simulation starts
537 - s4u::onSimulationEnd: right after the main simulation loop
538 - s4u::onTimeAdvance: right after a clock change (time is discrete)
539 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
540 event from the availability_file changes the avail speed.
541 - Links are now usable from s4u
542 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
543 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
544 - New examples: The conversion of MSG examples to S4U has begun
545 - Actors: create, kill, migration, and suspend
546 - Applications: master-worker and token-ring
547 - Action replay: communications and storage
548 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
551 - Backwards Compatibility breaks
552 - The SD_route_get_list and SD_route_get_size functions have been removed.
553 They are replaced by the unique
554 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
555 The route from src to dst is built in the links dynar whose size can
556 be obtained with xbt_dynar_length.
557 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
558 sg_host_route_bandwidth and sg_host_route_latency.
559 Macros ensure the backwards compatibility, but you should fix your code
562 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
563 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
564 - Bump our claim of support from MPI 1.1 to MPI 2.2.
565 We don't support 100% of it, but it should be enough. Contact us if not.
566 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
567 naive and probably inaccurate)
568 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
569 mapped onto a unique small file using some system magic.
570 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
571 - Fortran: cleanups, fixes, support of user-added operations
572 - RMA: MPI_Accumulate are applied in correct order.
573 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
574 - Better support for MPI_IN_PLACE
575 - Support for MPI_Win attrs and keyvals.
576 - Support MPI_Comm_group_create, MPI_Type_size_x
577 - MPI ops should be properly applied to all allowed types
578 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
579 algorithm by a non failing one from mpich.
582 - Kill the fifo data container: we don't use it anymore.
585 - A <cluster> can now be created with different speed values to represent pstates.
586 see examples/platforms/energy_cluster.xml for an example.
588 SimGrid (3.14.159) Released December 28. 2016
590 The previous release was only a crude approximation of a Pi release;
591 this one should be a bit better.
593 - Revert a last minute change that broke on Mac OSX.
594 - Fix the changelog and NEWS files.
595 - Improve the documentation, in particular of the routing module.
596 - Integrate some patches flying around in Debian.
598 SimGrid (3.14) Released December 24. 2016
601 * The whole documentation was reworked and reorganized. There is
602 still much room for improvement here, but we're on it.
605 * We now need python3 for our testing infrastructure.
606 * Model checking needs libevent
609 * Now works on FreeBSD
612 * Privatization now works on FreeBSD
614 * Privatization is now activated by default in smpirun
615 You can turn it off with -no-privatize if something goes wrong.
617 * Call-location tracing for SMPI.
618 You can add the exact location (filename / linenumber) of an MPI call to
619 your trace files and slow down or speed up the simulation between two
620 consecutive calls by using an adjustment file (see the documentation).
622 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
623 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
624 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
625 values were ignored in some cases. The timings of these functions can now
626 be significantly different.
628 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
629 smpi/running-power is renamed to smpi/host-speed
631 * smpi/grow-injected-times option to enable or disable multiplication of the
632 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
633 make simulation less precise (but also much faster).
635 * smpirun script should be (much) faster for large deployments.
637 * SMPI tracing: fixed issue with poor matching of send/receives.
639 * Replay: Fix broken waitall
641 New functions and features
642 * MSG_parallel_task_execute_with_timeout, to timeout computations.
644 Dropped / renamed functions and features
645 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
646 - MSG_mailbox_is_empty() -> Mailbox::empty()
647 - MSG_mailbox_front() -> Mailbox::front()
648 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
649 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
650 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
651 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
652 * Don't pass the free_f parameter to property related functions:
653 - MSG_host_set_property_value()
654 - MSG_as_router_set_property_value()
655 - MSG_storage_set_property_value()
656 * VM properties. Since msg_vm_t are msg_host_t, just use
657 MSG_host_get_property_value() and friends
658 * VM I/O related things:
659 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
660 - Unimplemented save/restore methods
661 * MSG_as_router_get_property_value() was redundent with
662 MSG_environment_as_get_property_value().
663 - Removed MSG_as_router_*propert*() functions
664 - Added MSG_environment_as_set_property_value() for consistency
665 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
666 Well, they are still there for now, but deprecated with a warning.
667 Please switch to xbt_dict_new_homogeneous() before this is removed
669 * Task affinity. Its intended behavior (that was very badly tested
670 and probably not really working) was deceiving what most users
671 would have hoped here.
672 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
673 * Option network/coordinates is now useless and should be dropped.
676 * Added option storage/max_file_descriptors to allow more than 1024 files opened
679 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
680 * Added Dragonfly topology support
683 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
684 EiBps, EBps, Eibps, Ebps for bandwidth.
685 They may become useful to some lucky ones.
688 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
689 * ex/app_tokenring: new example, very similar to the MSG Token Ring
690 * ex/async_waitAll: new example, on asynchronous communications
693 * Memory usage should be decreased for simulations with a large number
694 of processes. This also helps for SMPI.
696 SimGrid (3.13) stable; urgency=low
698 The Half Release, a.k.a. the Zealous Easter Trim.
700 - We removed half of the lines, that were mostly experimental cruft.
701 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
702 (not counting blanks and comments -- according to openhub.net)
703 - The internals are now compiled in C++ (and will soon be clean C++)
704 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
705 59 klines of C, etc. We added only 29 klines of C++ in replacement.
707 * Backwards Compatibility breaks
708 - Removed Lua simulation bindings (switch to C or Java for that).
709 Lua can still be used to describe platforms
710 - Removed Java kernel plug-ins.
711 Will be reintroduced after the ongoing major internals reorg.
713 - the following functions were removed.
714 They were too specific and should be reimplemented in a generic
715 way, with filter function.
716 - MSG_task_listen_from_host
717 - MSG_mailbox_get_count_host_waiting_tasks
718 - MSG_mailbox_put_with_timeout was removed.
719 Please use MSG_task_send_with_timeout instead.
721 - the SD_application_reinit function was removed. It has been a noop for a while.
722 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
723 replaced by a more flexible mechanism.
724 - The following functions thus do not exist anymore
725 - SD_workstation_get_access_mode
726 - SD_workstation_set_access_mode
727 - SD_workstation_get_current_task
728 - Basic estimation functions have been removed but can easily be replaced
729 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
730 - SD_workstation_get_computation_time => amount / sg_host_speed()
732 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
733 Use VM.getSpeed()*load/100 for the legacy behavior.
735 - option enable_tracing was removed. It was not doing anything for a while.
736 - In the ModelChecker:
737 - the model-checker now ptraces the model-checked process which means
738 you cannot use a debugger on the latter anymore (we might make this
739 optional in the feature);
740 - removed soft-dirty page tracking;
741 - remove model-checked side snapshot management,
742 MC_snapshot() and MC_compare_snapshot();
743 - keep the MC_cut() function as a stub (it was not really working
744 in the previous release).
747 * All options are consistently in kebab-case. Old names are kept as alias.
750 * Switch to platform v4 format.
751 - Rename from 'power' to 'speed' the attributes describing the amount of
752 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
753 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
754 - In <host> and <link>, attributes availability and state are gone.
755 It was redundent with state and availability traces, and with peak values.
756 - In <cluster>, attributes availability_file and state_file are gone.
757 It was too complex and unused.
758 - Kill <gpu>. Was not doing anything.
759 - The DOCTYPE points to the right URL:
760 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
761 (the file at this address now documents the changelog since its v1)
762 - A warning is emitted for unit-less values (they are still accepted).
763 - speed. Default: 'f' or 'flops'. Also defined:
764 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
765 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
766 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
767 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
768 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
769 - latency. Default: 's' second. Also defined:
770 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
772 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
773 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
776 * s4u::Host is now the preferred public interface to the Host features.
777 sg_host_* functions are C bindings to the exact same behavior
778 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
781 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
782 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
783 They were deprecated since a few years, and probably did not even compile anymore.
786 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
787 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
788 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
789 encouraged to update their codes. A list of the modified functions can be found at the end of
790 include/simgrid/simdag.h
793 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
794 the lambda or closure passed as a parameter will run in kernel mode.
795 All the callback functions should be rewritten to that interface at some point.
798 * Reorganizing and cleaning the internals all around the place.
801 * Remove old default barrier/bcast buggy algorithms (see #18407)
802 * Various bug fixes to handle more codes
803 * Remove the need for the --foreground option of smpirun (it is still
804 accepted for backward compatibility).
807 * Kill the setset data container: MC don't use it anymore.
808 * Kill the queue data container: it made more sense with GRAS.
809 * Kill the xbt_peer_t data type: it's useless without GRAS.
810 * Kill rm_cb feature of config sets: it was never useful.
811 * Kill graphxml parsing feature. It was not used.
812 * Kill the deprecated code protected by XBT_USE_DEPRECATED
814 - xbt_dynar_sort_strings(), when the content is char*
815 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
816 They throw exceptions on invalid input;
817 * C++ support for declaring CLI flags (simgrid::config::Flag);
818 * class for abstracting different signal backends (simgrid::xbt::signal).
819 with no external dependencies (we need very simple signals).
822 * refactoring and cleanup of the code;
823 * ongoing process to cleanly separate the model-checking algorithms
824 from the code model-checking support.
826 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
828 SimGrid (3.12) stable; urgency=low
830 The Facelift Release.
833 * Require g++ v4.7 at least to not speak prehistorical C++.
834 * Require Boost 1.48 (for signal2 component).
835 * Java must be version 7 at least when activated.
836 * Builds on Windows again (including Java bindings).
837 * Tracing is now always enabled (no way to turn it out)
838 * Remove GTNetS. It was not working anyway.
839 * Various cleanups in the cmake scripts.
840 * Move headers around to sort them out on installed systems:
841 - instr/instr.h -> simgrid/instr.h
842 - instr/jedule/* -> simgrid/jedule
843 - simdag/datatypes.h was removed
844 - simdag/simdag.h -> simgrid/simdag.h
845 - msg/datatypes.h was removed
846 - msg/msg.h -> simgrid/msg.h
849 * Interface improvement:
850 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
851 - Sanitize the interface in MSG_task_ module:
852 - Merge two functions that were close enough but misleading:
853 set_compute_duration(t) -> set_flops_amount(t)
854 get_remaining_computation(t) -> get_flops_amount(t)
855 - set_data_size(t) -> set_bytes_amount(t)
856 get_data_size(t) -> get_bytes_amount(t)
857 - Massive cleanups in the functions related to the energy
858 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
859 - New: MSG_host_get_pstate()
861 - msg/energy/onoff: switching hosts on and off
863 * Interface improvement:
864 - Rename Host.isAvail() to Host.isOn()
865 - Rename Process.currentProcess() to Process.getCurrentProcess()
866 - Rename Task.setDataSize() to Task.setBytesAmount()
867 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
869 - #18874: Actually allows the GC to reclaim tasks
873 - SIMIX_process_throw: raises an exception in a remote process
874 * Refactoring: Separate sub-modules
875 - libsmx: the public interface, as libc in a real system
876 - popping: the strange dance that converts a user request into a kernel handling
877 - smx_context_*: the virtualization mechanisms that embed the user code
878 - smx_*: the handling of each simcalls
880 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
881 * Rename smx_action_t into smx_synchro_t, making explicit that these
882 things are used to synchronize processes with their environment.
883 For example, a communication is a sort of synchronization involving
884 the communicating processes (that may block until the exchange) and
885 the platform. The same can be said from computations, etc.
887 - #18888: segfault when a process ends before its kill_time
891 - 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.
892 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
893 - MPI_Type_set_name, MPI_Type_get_name
894 - MPI_*_c2f and MPI_*_f2c functions
895 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
896 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
897 - Activate a lot of new tests from the mpich 3 testsuite
899 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
900 - InfiniBand network model added: Based on the works of Jerome Vienne
901 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
902 - When smpi/display_timing is set, also display global simulation time and application times
903 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
904 * Collective communications
905 - 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
906 - 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/).
907 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
909 - Replay now uses algorithms from wanted collective selector
910 - Replay can be used with SMP-aware algorithms
911 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
912 - 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.
914 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
915 - Correctly use loopback on fat-tree clusters
916 - Asynchronous small messages shouldn't trigger deadlocks anymore
917 * Energy/DVFS cleanup and improvement
918 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
919 - new: smpi_get_host_pstate()
922 - "Full" network optimization flag was broken since Surf++
923 - Better handling of precision flags in maxmin
924 - Fix bug causing sometimes "Impossible" errors
925 - Properly pass cluster properties to included hosts
926 * Improvement of the Energy plugin.
927 - Always update the consumption before returning that value
928 - New property: watt_off to denote the dissipation when the host is off
929 - New functions getWattMinAt and getWattMaxAt to retrieve the
930 dissipation of pstates that we are not currently at.
931 * Java: class NetworkLink renamed to Link
932 * New function: simcall_process_get_kill_time()
933 * Massive rename s/workstation/host/
934 - That's intrusive, but that's good for the project consistency. Sorry.
935 - Change config option "workstation/model" into "host/model"
938 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
939 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
940 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
941 - xbt_dirname and xbt_basename for non-POSIX systems
943 * The model checker now runs as a separate process.
944 * The model checker runs must now be launched with the new simgrid-mc program.
945 * Record/Replay: the MC can display a textual representation of a path in the
946 execution graph. It can then be replayed outside of the model checker.
948 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
950 SimGrid (3.11) stable; urgency=low
955 * Normalizing pointers addresses tool for better diff between logs
958 * Add cloud examples using new VMs
959 - examples/msg/cloud/two_tasks_vm.tesh
960 - examples/msg/cloud/simple_vm.tesh
961 - examples/java/cloud/cloud.tesh
962 - examples/java/cloud/migration/migration.tesh
963 * Add java surf examples:
964 - examples/java/surfPlugin/surf_plugin.tesh
965 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
966 - examples/java/surfCpuModel/surf_cpu_model.tesh
967 * Add SMPI+MSG example:
968 - examples/smpi/smpi_msg_masterslave/
973 - msg task destroy cancel test
974 - msg_host on/off test
975 * Move all tests in testsuite to teshsuite (adding tesh files)
976 * Restructure teshsuites
977 - one folder for each kind of test
978 * Restructure AddTests.cmake
980 - structure the order of tests (with sections)
983 * Add virtual machine
984 - creation of a VM on a PM
985 - migration of a VM from a PM to another PM
987 - MSG_process_join(msg_process_t process, double timeout)
988 - msg_bar_t MSG_barrier_init(unsigned int count)
989 - int MSG_barrier_wait(msg_bar_t barrier)
990 - void MSG_barrier_destroy(msg_bar_t barrier)
991 - msg_as_t MSG_environment_as_get_by_name(const char * name)
992 * New option "msg/debug_multiple_use" to help debugging when a task is used
995 - Improvements and finalization of MSG_storage, MSG_file APIs and their
997 - Increase code coverage in test suites
1001 * Protect context stacks against stack overflow. The number of protected memory
1002 pages allocated on the top of each stack (1 by default) can be configured
1003 with the new command line option --cfg=contexts/guard_size:<value>.
1004 * Simcalls are now generated by a python script that
1005 - generates files included by SimGrid
1006 - checks that all the functions exist, and proposes prototypes
1008 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1009 host_set_data, host_get_data
1011 - simcall_process_join(smx_process_t process, double timeout)
1012 * Fix bug where sleeping processing could not be suspended.
1015 * Translate surf models from C to C++
1016 - Generic classes for all models: Model, Resource, Action
1017 - A generic interface for each kind of model (CPU, Network, Storage
1018 Workstation, WorkstationVM)
1020 * Translate surf routings from C to C++
1021 * Add callbacks using sigc++ or boost::signals2
1022 - Add callback functions for resource creation/destruction
1023 - Add callback functions for action state change
1024 - Handle Energy as a plugin
1025 * Replace swag by boost::intrusive
1026 * Add new routing models for clusters. For documentation, see
1027 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1028 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1029 parameters for cluster tag
1030 - Fat trees, with topology="FAT_TREE" and
1031 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1032 - see examples/platforms/torus_cluster.xml and
1033 examples/platforms/fat_tree_cluster.xml
1034 * More documentation
1037 * Hostfiles support host:nb_processes construct to deploy several processes on
1039 * Collective communication algorithms should not crash if used with
1040 improper number of nodes and report the error.
1041 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1042 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1043 MPI_Dims_create, MPI_Cart_sub are supported.
1044 * New interface to use SMPI programmatically (still depends on MSG for
1045 some parts, see examples/smpi/smpi_msg_masterslave):
1046 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1050 * Global variables privatization in MPI executables is now performed at runtime
1051 with the option smpi/privatize_global_variables (default:no).
1052 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1053 dynamic libraries loaded by the application are not privatized (static
1054 linking with these libraries is advised in this case)
1057 * Options defined in XML work correctly now.
1060 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1061 copied into simgrid.jar or not (ON by default). Use this option if you want
1062 to reduce the size of the installed simgrid.jar, *and* the native libraries
1063 are kept installed elsewhere.
1064 * Surf binding with SWIG (code generated in maintainer mode only):
1065 - plugin to handle callbacks
1066 - CPU model only for the moment
1069 * Supernovae build mode is definitively removed. It was used to improve
1070 inlining and inter-module optimizations. It is nowadays superseded by
1071 link-time optimizations commonly available in compilers.
1072 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1073 * Add boost dependency for surf++
1074 * Add new macro for tests
1075 - ADD_TESH(name <tesh_args>)
1076 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1080 - xbt_bar_t XBT_barrier_init(unsigned int count)
1081 - int XBT_barrier_wait(xbt_bar_t barrier)
1082 - void XBT_barrier_destroy(xbt_bar_t barrier)
1083 * Make the xbt_os_time module public
1085 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1087 SimGrid (3.10) stable; urgency=low
1089 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1092 * Reintegrate Java to the main archive as desynchronizing these
1093 package is not acceptable anymore (Java is now considered stable)
1094 * Add explicit synchronization facilities through semaphores
1095 * Bug fix: Task.setDataSize() only changed the C world, not the value
1096 cached in the Java world
1099 * Dramatically change the way files are handled. API and internals changed, but
1100 this part of MSG was not considered as production grade either.
1101 * Add explicit synchronization facilities through semaphores
1102 * Add a new function MSG_host_get_process_list()
1103 * Preliminary DVFS support (see examples/msg/energy/ for details)
1106 * SMPI is now included directly in the libsimgrid as the windows
1107 linker doesn't force us on splitting it anymore.
1108 * Improvements of the SMPI replay tool:
1109 - Most of the collective communications are now rooted in the same process as
1110 in the original application.
1111 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1112 used until now). Multiple data types can now be used in a trace.
1113 - The replay tool now supports traces produce either by TAU or a modified
1115 - Bug Fix: the compute part of the reduce action is now taken into account.
1116 - Gatherv collective is now supported
1117 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1118 of smpirun outputs time independent traces for the current run. One file
1119 is created per process. If too many processes are simulated, this behavior
1120 can be changed to one file for all processes by using the
1121 tracing/smpi/format/ti_one_file flag
1122 * smpirun generates the host file if needed (with given host count and platform)
1123 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1124 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1125 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1126 --cfg=smpi/coll_selector:(mpich/ompi)
1127 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1129 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1130 happening inside a collective SMPI call.
1131 * Fix the behavior of complex data types handling.
1132 * Make MPI_Wtime another synchronization point to take computations into
1134 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1135 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1136 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1137 (no privatization of global variables yet)
1138 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1141 * Verification of liveness properties is now available for SMPI applications
1142 (in addition to MSG applications)
1143 * Bugged examples using SMPI in examples/smpi/mc/
1144 * Add --cfg=model-check/visited option. Allows the verification of infinite
1145 programs. Detection of loops in the execution thanks to the system state
1146 comparison and reduction of the state space to explore. Can be combined with
1147 DPOR for safety properties.
1150 * Allow to change SimGrid configuration (see --help) within the code
1151 thanks to SD_config() as it can be done in MSG.
1152 * Add a new function SD_task_set_amount() upon user request.
1155 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1156 * Remove rule based routing (no more PCRE dependency)
1157 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1158 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1159 * Add a loopback_bw and loopback_lat options to cluster tag.
1160 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1161 routing. Now peers are encapsulated in an AS and have their own private
1162 router but this is transparent.
1165 * Our own implementation of getline is renamed xbt_getline, and gets
1166 used even if the OS provide a getline(). This should reduce the
1167 configuration complexity by using the same code on all platforms.
1168 * New type: xbt_cfg_elm_boolean.
1169 * Allow to use yes/no for boolean configuration options in the command line.
1170 * Allow to disable SimGrid cleanups at exit from command line option.
1171 There are situations where one may want a simulation to end with an exit.
1172 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1173 annoying when scripting around the simulator. Adding a
1174 --cfg=clean_atexit:no allows to circumvent this issue.
1177 * Lots of memory leaks were corrected in this release.
1178 There is no known memory leaks anymore, in all of our 600+ tests.
1179 * New command line option --version, to get SimGrid version information.
1180 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1182 * Supernovae builds are deprecated, and expected to be removed in the next
1185 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1187 SimGrid (3.9) stable; urgency=low
1189 The Grasgory release: GRAS is really dead now.
1191 * Complete overhaul of the internal host structures scheme.
1194 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1195 was considered as experimental and badly maintained since then.
1196 * Keeping it was thus a trap to our potential users, that could take
1197 it instead of MSG or SMPI by mistake despite is pity state.
1198 * GRAS seems to have very few users (if any), and no one volunteered
1199 to maintain it further. It also induces a lot of XBT code (for
1200 portability sake), that must be maintained too.
1201 * For all these reasons, we killed GRAS. If someone wants to revive it
1202 in the future, don't cry, our git history still remembers of GRAS.
1205 * Major overhaul. Merge our documentation again as time proved that
1206 splitting it was really not helping our users.
1207 * Further improve the developer documentation to help newcomers
1208 hacking on SimGrid itself. The user documentation (and in
1209 particular, the beginner documentation) is still in a sorry state.
1212 * Now works on Windows too!
1213 * Much more extensive test suite, from MPICH
1216 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1217 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1218 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1219 * Bug fix: let task be scheduled when the last dependency to be solved is
1220 a control dependency.
1221 * Remove SD_load_environment_script function.
1222 Use the C sg_platf function if you want to declare a platform
1226 * New function: MSG_process_get_number()
1227 * Old function documented: MSG_config()
1228 * Remove MSG_load_platform_script function
1229 Use the C sg_platf function if you want to declare a platform
1233 * Change the default value of the TCP_gamma constant (maximal size of TCP
1234 congestion window) to a more realistic 4MiB value. If you notice changes in
1235 your simulation results, you can fall back to the previous 20k tiny window
1236 by adding --cfg=network/TCP_gamma:20000 on command line.
1237 * (Hopefully) fix a bug wrt periodic availability and state traces
1238 * Bug fix: use default values at start when first event in availability/state
1239 trace is not at time 0.
1242 * remove the "new_" part of function name sg_platf_new_trace_connect
1243 (resulting in sg_platf_trace_connect), since it does not create
1247 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1248 use case where it's really mandatory, and maintaining it was a pain in
1250 * New: xbt_fifo_search(), search an item with a user-provided
1251 comparison function instead of dumb pointer comparison.
1254 * Fix the lua deployment:
1255 Use `simgrid.init_application()` before deployment instead of
1256 `simgrid.msg_register_application()` after.
1259 * Transfer the tracing files into the corresponding modules.
1261 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1263 SimGrid (3.8.1) stable; urgency=low
1265 The "we are told that some people want to also *install* the simgrid
1268 * Add missing file "tesh.1" to the archive.
1270 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1272 SimGrid (3.8) stable; urgency=low
1274 The Psssshiiiit release: SimGrid jumps into the Cloud.
1277 * Add an experimental interface to manipulate VMs. They are mainly
1278 process groups with very few intrinsic semantic, but they should
1279 allow you to build the semantic you want easily.
1280 * New function: MSG_host_set_property_value()
1281 * New function: MSG_process_on_exit(). To clean memory in all cases.
1282 * Bug fixes that made the host (and link) failures unusable.
1283 * Add a way to auto-restart process when the host in which they are
1284 executing comes back (ON_FAILURE="RESTART" on deployment file,
1285 MSG_process_auto_restart_set).
1286 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1287 please stop using the old ones, they are DEPRECATED.
1289 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1290 Please use MSG_init() instead. (reducing the amount of entry
1291 points in the library helps us).
1292 * Make it impossible to link against the wrong version of the lib
1293 * Deprecate MSG_clean(). No need to call it anymore.
1294 * Function MSG_get_host_number() is not deprecated anymore.
1297 * Split the doc into a user guide and a reference guide.
1298 * Start a developper guide to help people hacking on SimGrid.
1301 * Enable tracing by default. This modules rocks you should use it.
1302 * Remove option custom_flags. Now use environment variables CFLAGS
1304 * Use default cmake things to detect lua instead of home grown ones.
1305 * New option "enable_mallocators" to disable mallocators, for debugging
1306 purpose ("on" by default).
1309 * Bug fixes around the resource failures: don't let the processes
1310 survive the host they are running onto.
1311 * Add an interface to auto-restart processes when the host in which they are
1312 executing comes back.
1313 * Ensures that SIMIX_clean is called automatically. It's not part of
1314 the public interface anymore (bindings should be updated).
1317 * Bug fix for when SD_Simulate is called with a positive value: be careful
1318 when comparing doubles. Sometimes they are different for non significant
1320 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1321 parallel task whose initial work is distributed among host according
1322 to the Amdahl's law. Such tasks are created with a parameter alpha
1323 that corresponds to the non-parallelizable part of the computation.
1324 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1325 between two sets of workstations assuming a 1D block distribution (each
1326 workstation owns a similar share of data) on both sides.
1328 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1329 Data redistribution will be automatically scheduled once parent and child
1330 are both scheduled. The filling of computation_amount and
1331 communication_amount structures is now done seamlessly thanks to the chosen
1333 * New function SD_workstation_dump to display various information
1334 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1335 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1336 bandwidth on the route onto which the task is scheduled and the amount of
1338 To divide the nominal bandwidth by 2, the rate then has to be:
1339 rate = bandwidth/(2*amount)
1340 * Compute tasks that have failed can now be rescheduled and executed again
1341 (from their beginning)
1342 * Increasing source code coverage (src/simdag is now covered at 95.8%
1346 * Re-implement time-independent trace replay using SMPI (at the
1347 smpi_smp_* level) instead of MSG. This should replace
1348 examples/msg/actions/actions.c
1349 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1350 hindexed and structs)
1351 * Implement the exchange of non-contiguous data.
1352 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1353 * Correct behavior of smpi/sender_gap and set its default value to 0
1354 * Add option to asynchronously send small messages to allow better
1355 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1356 specifies the size in bytes under which messages will be asynchronously sent.
1357 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1358 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1361 * New C interface to define a platform: XML is now optional.
1362 For more info, please check include/simgrid/platf.h.
1363 * New interface to define random platforms from the C:
1364 For more info, please check include/simgrid/platf_generator.h and
1365 examples/msg/masterslave/masterslave_platfgen.c
1366 * Export a sg_cmdline dynar containing all the arguments we got from
1370 * Two new tracing options for adding comments to trace file so you
1371 can track your experiments (see --help-tracing for details).
1372 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1373 * Adding the SimGrid version that generated the trace file as a comment.
1374 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1375 * Fix to avoid key clashes on Paje links
1376 * Other minor fixes related to the Paje specification
1379 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1380 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1381 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1382 * Greatly improve the robustness of mmalloc to user errors (such as
1383 using an area after freeing it, or freeing it twice)
1385 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1387 SimGrid-java (3.8.1) stable; urgency=low
1389 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1390 Warning: all methods are visible, but only some of them are
1391 implemented so far. Check the source (src/jtrace.c)
1392 for further information.
1393 * New module: org.simgrid.msg.File (SimGrid File management functions)
1394 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1395 * Change the meaning of Process.restart: now restart the process from
1396 the begining, like MSG_process_restart in C.
1397 * Add Process.setAutoRestart: handling of process restart when failed
1399 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1400 you to retrieve the properties of the processes/hosts
1401 * Deprecate Msg.clean(): you can just forget about it now.
1402 * New function Process.getCount(), that only works when compiling
1403 with the not yet released version 3.9 of the C library.
1405 * New context factory based on Coroutines. It mandates a modified JVM
1406 but then, the simulations run about five times faster, and there is
1407 no limit to the amount of processes (beside of the available memory).
1409 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1411 SimGrid (3.7.1) stable; urgency=low
1414 * Restore the prototype of MSG_process_create_with_environment() to
1415 the pre-3.7 situation by removing the kill_time argument.
1416 * Add a MSG_process_set_kill_time() function instead.
1419 * Fix weird behaviors when dealing with parallel tasks.
1422 * Simgrid is now built as a dll.
1423 * Simgrid-java now works on Windows.
1424 * Simgrid-Java is now included into Windows package.
1427 * First pre-build package for MacOSX.
1430 * Fix compilation when using MSG_USE_DEPRECATED.
1431 * Fix some compilation issues on Macs and Windows.
1432 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1433 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1435 We discovered that the Lua console is broken, but we are missing the
1436 manpower to fix it right now. The problem existed in 3.7 too, so we
1437 are not blocking the release for that. Sorry if you depended on this
1438 feature, any help would be really welcome.
1440 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1443 SimGrid-java (3.7.1) stable; urgency=low
1445 The "Java aint got to be bloated and slow" release
1448 * Various internal cleanups and performance improvement
1449 Simulations are expected to run up to twice faster or so
1450 * Make Process.kill(process) an instance method, not a static one
1451 * User processes are not java.lang.Thread subclasses.
1452 This breaks the compatibility (sorry), but previous API was
1453 brain-dead, making it impossible to have non-trivial
1454 initializations in the process constructor.
1455 * Require a full constructor per Process sub-class.
1456 Kinda breaks the compatibility (sorry), but this allows a much more
1457 efficient way to launch the processes at simulation startup.
1458 * Do not embeed our version of semaphores, java 1.5 can be considered
1459 as sufficiently prevalent for us to not dupplicate its features.
1463 * Add examples for almost every part of the API
1464 We spotted and fixed a lot of bugs in the process
1466 * New module: asynchronous communication API
1467 * New function: Process.sleep()
1468 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1469 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1470 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1472 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1474 SimGrid (3.7) stable; urgency=low
1476 The "spring cleanups (before the next Big Project kicks in)" release.
1479 * We can specify the SMPI latency/bandwidth factor with command line
1480 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1481 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1482 You can also use the "config tag" from platform file by adding this line
1483 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1484 (see "example/platforms/tag_config.xml" to use "config tag").
1485 Note that the command line supersedes the platform file configuration.
1486 * Change the correction factors used in LMM model, according to
1487 the latest experiments described in INRIA RR-7821.
1488 Accuracy should be improved this way.
1489 * Use the partial invalidation optimization by default for the
1490 network too. Should produce the exact same results, only faster.
1491 * Major cleanup in surf to merge models and split some optimization
1492 mechanisms from the core of the models. As a result you can now
1493 specify which model to use (e.g., --cfg=network/model:LV08
1494 --cfg=cpu/model:Cas01) and which optimization mode to use
1495 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1496 Incompatible combinations should err at initialization. See
1497 --help-models for the list of all models and optimization modes.
1498 * The CLM03 workstation model was dropped for simplicity because it
1499 used the deprecated CM02 network model. Use default instead.
1500 * Rename the TCP_gamma configuration option to network/TCP_gamma
1501 * Rename the coordinates configuration option to
1502 network/coordinates, and document it
1503 * Use now crosstraffic keyword instead of the terribly misleading
1504 fullduplex keyword. It is activated by default now in the current
1505 default model, use --cfg=network/crosstraffic:0 to turn it off.
1506 * Ongoing refactoring the model parsing to make XML files optional
1507 See include/simgrid/platf.h for details (still to be completed)
1510 * Major overhaul of the documentation. Almost instructive now :/
1511 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1512 functions and friends. This interface was considered as
1513 deprecated since over 2 years, it's time to inform our users that it is.
1514 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1515 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1516 use these (crappy) functions in your code.
1517 These functions will be removed soon. Stop using them now.
1518 * Deprecate MSG_get_host_{table,number}
1519 Implement MSG_hosts_as_dynar() instead.
1520 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1521 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1524 * Stabilize the parallel execution mode of user contexts
1525 * Introduce configuration variables to control parallel execution:
1526 - contexts/synchro: Synchronization mode to use when running
1527 contexts in parallel (either futex, posix or busy_wait)
1528 - contexts/parallel_threshold: Minimal number of user contexts
1529 that must be part of a scheduling round to switch to parallel
1530 execution mode (raw contexts only)
1531 * Fix bugs that prevented to use suspend/resume along with
1532 synchronization structures.
1533 * Fix bugs in process termination that lead to invalid memory access
1534 in very specific conditions.
1537 * Introduce a parallel mode for the models (controlled by surf/nthreads
1538 configuration item). In our tests, running the models in parallel
1539 never lead to any speedups because they are so fast that the gain
1540 of computing each model in parallel does not amortizes the
1541 synchronization costs, even when ultra fast futexes are used.
1542 This is released anyway because YMMV.
1545 * Performance boost by using a swag internally to compute the set of
1546 tasks that are finished and should constitute the return value of
1550 * Enable it by default now that it is considered rather stable.
1553 * Documentation of the tracing functions.
1554 * Performance gains when tracing categorized/uncategorized resource
1555 utilization by avoiding calls to get route when updating resource
1556 variables. LMM constraints are being used instead.
1557 * API changed to set task categories. Use MSG_task_set_category instead
1558 of TRACE_msg_set_task_category, and SD_task_set_category instead
1559 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1560 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1561 * Examples for link user variables added (see at examples/msg/tracing/)
1562 * Deprecated function TRACE_msg_set_process_category completely removed
1563 * Trace header updated according to the latest Paje file format
1564 * Tracing network lazy updates, no longer obligate users to use full updates
1565 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1566 * Experimental: let user code declare/set/push/pop application states for hosts
1567 * API changed to allow the manual creation of graph configuration files
1568 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1571 * Improve the API of Lua MSG bindings, using the Lua spirit.
1572 * Each simulated process now lives in its own Lua world (globals are
1573 automatically duplicated). It helps writing simulators. Will allow
1574 to run Splay programs within SimGrid in the future.
1575 * Add a Chord example in Lua, equivalent to the MSG one.
1578 * Start the implementation of a solution to express temporal
1579 properties, not only local assertions. This is still an
1580 experimental work in progress, stay clear from it to be safe.
1584 - Add new runtime parameters --help-logs and --help-log-categories
1585 to display information about supported logging parameters and
1587 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1588 don't exists anymore.
1589 * Mallocators: allow value NULL for the reset function.
1591 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1592 create homogeneous dictionaries, where all the elements share the
1593 same free function. Non homogeneous dictionaries will be
1594 deprecated in the next release.
1595 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1596 - Multi-level dictionaries are deprecated.
1598 - new function xbt_dynar_search_or_negative() that is useful when
1599 you have less than 2 million elements in your dynar and don't
1600 want of the extra complexity of catching exceptions when the
1601 element is not found.
1603 - Make xbt_os_thread module (for thread portability) public.
1604 Documentation is still to come, sorry.
1606 - Cleanups and simplifications to make it maintainable again.
1607 - Exotic features (such as memalign and valloc) were removed.
1608 - The metadata were extended and improved so that the
1609 model-checker becomes able to explore and inspect the heaps.
1610 - This may induce a performance drop when enable_model-checking is
1611 ON in cmake (even if it's not used in the simulation), but it is
1612 necessary at this point to get MC working.
1614 Turn model-checking OFF if simulation performance matters to you.
1615 Not enabling it at runtime is not enough, disable it in cmake.
1617 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1620 SimGrid (3.6.2) stable; urgency=low
1622 The "Not coding new stuff allows to polish old things" release.
1625 * New bindings to the NS3 packet level simulator (experimental)
1626 * Use the raw (efficient) execution contexts instead of the sysv
1627 (portable) ones when possible.
1628 * libpcre is now mandatory in any cases since not using it led to
1629 severe performance loss and possibly other issues
1630 * Update the XML platforms:
1631 - G5K: include the latest machine in Nancy
1632 - GridPP and LCG: new platforms
1633 * Documentation was partially updated, at least (more to come)
1635 Bug fixes, cosmetics and small improvements
1636 * Free terminated processes before the end of the simulation to avoid
1637 exhausting the memory of users having very dynamic amount of
1639 * Bug fix and cosmetics about canceling non-running tasks
1640 * Bug fix about the dot loader's issues when using libcgraph
1643 * Create an installer for windows with nsis (amd64 and win32)
1644 - Add an hello world project to illustrate simgrid project creation.
1645 - Embed libpcre into the Simgrid installer to avoid
1646 its compilation burden
1647 * The raw execution contexts should work on Apple now
1648 * Port to Windows 64 bits
1649 - Sysv contexts now have an implementation for this arch
1650 - GRAS communication features now support this arch
1651 * Drop support for borland compiler on windows
1652 - this code was not maintained, and we kinda depend on gcc nowadays
1653 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1654 * Fix portability issue on unstable ubuntu: linker became picky on
1657 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1659 SimGrid (3.6.1) stable; urgency=low
1661 The "Oops, we broke Macs too" release
1664 * Fixed contexts detection so that raw ones are used when possible
1665 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1666 leads to a strange error, with user code segfaulting sometimes when
1667 the generated code is not perfectly aligned (which is not
1668 controllable from the user side, depends on the amount of code)
1671 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1673 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1675 SimGrid-java (3.6) unstable; urgency=low
1678 * Split of every thing from simgrid v3.5 into a separate package.
1680 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1682 SimGrid (3.6) stable; urgency=medium
1684 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1687 * Bindings now constitute their own package, separated from the main one.
1688 Rationale: reduce our maintenance nightmare by reducing the module coupling
1689 They will soon be released on their own on gforge.
1691 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1692 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1694 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1695 * It's not quite deprecated for now because we have no replacement,
1696 but it may soon become the case.
1699 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1700 * Fortran: New user-level cache variable to store the rank of the running
1701 process. This improves performance by an order of magnitude.
1702 * C: New coccinelle script to automatically locate and modify global and
1703 local static variables.
1704 * Improved SMPI network model with a sender-side gap to account for multiple
1708 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1709 only say if a communication is finished, no matter whether it succeeded or
1710 failed. You can call MSG_comm_get_status() to know the status of a finished
1712 * New function MSG_task_dsend() to send a task and detach it. When a
1713 communication is detached, you are never notified of its success or failure
1714 and the memory is released automatically once it is finished. This function
1715 is useful when you don't care about the end nor the success of a
1717 * Change the prototypes of action replay. Sorry for inconvenience,
1718 but this is really more efficient this way (and to adapt your code,
1719 you just have to fix the initialization, that shouldn't be too long)
1720 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1721 ever managed to do anything useful with it.
1722 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1723 takes a dynar of communications. It returns immediately and gives the
1724 index of a finished communication (if any).
1725 * New example: a basic implementation of the Chord P2P algorithm.
1728 * New model for multi-core CPUs. You can now use the core attribute to
1729 precise the number of cores of a host. This is a basic model. Every
1730 process running on the host receives at most the power provided in
1731 the DTD (throughput<=power). Total throughput of process cannot exceed
1733 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1734 router linked by an up-link and a down-link (possibly asymmetrical).
1735 This kind of pattern allows to easily build last-mile model style platforms.
1736 Aggregating such patterns in a rule-based AS is thus the technique of
1737 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1738 * New model for Vivaldi routing. We transformed the Vivaldi network model
1739 into a Vivaldi routing model (based on the rule-based model). This allows to
1740 combine Vivaldi based latencies with last-mile platforms.
1743 * Added a check for NaN of IEEE754 infinite in the double entries of
1745 * Introduce a new context factory "raw", highly inspirated from the
1746 ucontext factory, but using manually crafted functions in assembly to
1747 do the work in an efficient manner.
1748 * Allow to change the used context factory at run time, not only at
1749 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1750 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1751 contexts at runtime (only with raw contexts or ucontexts).
1752 * Completely rewrote this module to allow parallel execution of user
1753 processes. Use --cfg=contexts/nthreads:N to execute user processes
1754 with N parallel threads (the default is 1, meaning no parallelism).
1755 * Allow to decide dynamically between sequential and parallel modes.
1756 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1757 processes in parallel only when their number is greater than or equal to P
1759 * Added a check for NaN of IEEE754 infinite in the double entries of
1763 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1764 won't output the state of processes when interrupted with Ctrl-C
1765 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1766 NULL-terminated array. This may solve backward compatibility issues
1767 due to the change to return type of SD_simulate. See also:
1768 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1769 * Add new macros with variable number of arguments.
1770 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1771 - in xbt/asserts.h: xbt_assert
1772 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1773 - in xbt/ex.h: THROWF and RETHROWF.
1774 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1776 * Change xbt_die() to accept a format string with arguments, just like printf.
1777 * New data structure: xbt_lib_t, like a dict but more general and with better
1781 * New configuration options
1782 Options triva/categorized and triva/uncategorized can be used to generate
1783 graph configuration files for Triva visualization tool.
1784 * Configuration option tracing/platform is renamed to tracing/categorized
1785 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1786 * New TRACE_user_link_variable interface:
1787 User provides the name of the link and the tracing variable to attach to it
1788 * the declaration of tracing categories must be done after the environment
1790 * simpler tracing interface, just one way to declare categories
1791 TRACE_category or TRACE_category_with_color, it is up to you
1792 * links in the trace file are again identified by their names
1793 * trace contains the full platform hierarchy exactly as declared using the ASes
1794 * Options tracing/msg/[task|process]:1 groups the process by hosts
1795 for both cases, tasks and processes must have names that are unique during the simulation
1796 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1797 * The experimental option tracing/msg/volume is deprecated
1798 its functionality may be reincorporated if needed
1800 The tracing generates a trace file with unordered timestamped events,
1801 because of the way the core simulator (surf) works. A script available
1802 at the tools directory (fix-paje-trace.sh) can be used to put the events
1803 in order. We have changed the tracing so it can generate ordered timestamped
1804 events in the final trace, but depending on the simulator (and how much time
1805 is simulated) that can lead to a huge memory utilization. It is deactivated
1806 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1808 Build Infrastructure
1809 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1810 - We are trying hard to keep the API stable, but it may happen that
1811 some things change (we're a research project after all, not a
1812 nuclear plant operating system). If such things should happen, you
1813 could rely on that macro to adapt.
1814 - current value: 30600 for 3.06.00, aka 3.6
1815 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1816 a number that can be compared with SIMGRID_VERSION.
1817 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1818 assertions and verbose/debug logging events are disabled at compile time.
1820 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1822 SimGrid (3.5) stable; urgency=medium
1825 * New feature to any SimGrid-based simulator: Model-Checking
1826 Check SIN#1 for more details.
1829 * New Model SMPI (three-interval linear regression for correction factors)
1830 See RR-7426, available at http://hal.inria.fr/inria-00527150
1831 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1832 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1833 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1834 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1835 * See also src/smpi/README
1840 - Tracing API changes: TRACE_start and TRACE_end should not be called
1841 by user-code. They are automatically called by simulators created
1842 with SimDAG, MSG and SMPI if the toolkit is compiled with
1843 tracing_enabled=ON. Categories declaration and utilization remain the
1844 same for MSG and SimDag.
1845 - A function was added to the tracing API to declare categories with
1847 - TRACE_category_with_color (char *category, char *color)
1848 where color must be in the following format
1849 "%f %f %f", red, green, blue
1850 and red, green, blue are float values in the interval [0, 1]
1851 - User can specify NULL as color parameter, or continue calling
1852 TRACE_category (cat)
1853 On that case, the tracing system will define random colors
1854 - The following command-line options are supported:
1855 --cfg=tracing/filename:msg.trace
1856 --cfg=tracing:1 (activate tracing, needed to use others)
1857 --cfg=tracing/platform:1 (categorized resource use)
1858 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1859 --cfg=tracing/msg/task:1 (task creation)
1860 --cfg=tracing/msg/process:1 (process creation, migration)
1861 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1862 --cfg=tracing/smpi:1 (SMPI interface tracing)
1863 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1864 - examples of examples/msg/tracing updated
1866 - DAXLoader and DOTLoader functions can generate tasks with categories
1867 - A new function to attribute a category to SD tasks:
1868 TRACE_sd_set_task_category (SD_task_t task, char *category)
1869 * Tracing the MPI interface implemented by SMPI
1870 - Collective operations are traced with states
1871 - Point-to-Point operations are traced with states/links
1872 - Tracing activated by a parameter "-trace filename" passed
1873 to smpirun during execution (considering that simgrid
1874 is compiled with tracing enabled)
1875 - To run the simulation with gdb, the simulator
1876 accepts --cfg=tracing/smpi:1 to trace SMPI
1877 - tesh files to check if smpi tracing is ok
1878 - See examples/smpi/NAS/DT-trace
1879 * GTNetS tracing re-worked
1880 - adaptation to the tracing system of GTNets to cope
1881 with modifications regarding the fullduplex mode
1882 - new tesh files to check if gtnets tracing is ok
1885 * Asynchronous communications through the functions:
1886 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1887 * New function: MSG_load_platform_script()
1888 to make possible using a lua script instead of XML files to set up platforms
1889 * New function: MSG_set_function
1890 to associate functions to processes, used when bypassing the parser
1891 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1893 Platforms: Add some more examples in examples/platforms
1894 * Grid'5000: see www.grid5000.fr
1895 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1898 * Change the XML format. This is a very important modification. SimGrid
1899 3.5 introduces a new hierarchical format based on the notion of
1900 Autonomous Systems. Compatibility with old format is ensured
1901 through the perl script provided in the install bin directory
1902 bin/simgrid_update_xml.
1903 It is now possible to build platforms with specific routing
1904 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1905 connect several platforms together. We will try to provide soon
1906 set of realistic platforms exploiting these properties (have a
1907 look at examples/platforms/ for the moment).
1908 * Take the opportunity of the XML format change to be a good XML citizen:
1909 rename link:ctn to link_ctn and similar changes (also dealt with by
1911 * Add a new routing scheme (rule-based) using regular expressions. It
1912 enables to have an extremely low memory footprint when the
1913 underlying routing is simple and can be compactly described. You
1914 need to have libpcre4-dev (perl regular expressions) installed if
1915 you want to use this routing scheme.
1916 * Revive the cluster TAG and allow to easily and efficiently (both in
1917 term of memory and speed) connect clusters together. Have a look
1918 at teshsuite/simdag/platforms/ to see how this can be done. With
1919 this tag, you can create clusters with thousands of tasks at no
1920 cost (have a look at examples/platforms/).
1921 Note: clusters are implemented as ASes, so there is no need for an
1922 enclosing AS tag if you have only one cluster in your platform.
1923 * Add new generic functions in the public interface that allows the user
1924 to call SURF 'create_resource' methods from your code (same
1925 functionality as the XML bypass mechanism but with a much lighter
1927 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1928 that uses a piecewise linear approximation to produce better
1929 results when exchanging small messages.
1930 * Add a new parameter to handle correctly full duplex link and account
1931 for interferences between uplink and downlink communications
1932 (activate with --cfg=fullduplex:1).
1935 * Rename the SD_READY (all dependencies are satisfied and task is
1936 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1937 dependencies are satisfied) state.
1938 This prevents a confusion between the notion of "ready to schedule"
1939 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1940 simulated" (SD_RUNNABLE) used by the simulation kernel.
1941 * Change the way a task is considered as ready. Instead of removing
1942 dependencies when a task is done, a counter is decreased. This way,
1943 it is always possible to reach ancestors thanks to the
1944 SD_taks_get_parents function (even after the end of the simulation.)
1945 * Change the return type of SD_Simulate from (SD_task_t*) into
1946 xbt_dynar_t. This function was in handling a dynar internally and
1947 converted it into a NULL terminated array for historical reasons.
1948 * New function SD_dotload(char*) to load a DAG described in dot
1949 format. This loader and the corresponding examples require the
1950 installation of the graphviz library.
1951 * Fix a bug in the management of tasks of size 0 in the surf network
1952 models. This problem was only visible with SIMDAG and you should
1953 thus disregard results produced with earlier versions if you
1954 relied on this feature (some tasks were blocked because of this).
1955 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1956 even though your applications were doing only point-to-point
1957 communications and sequential computations. Now you can really use any
1958 model you want (of course, if you create real parallel tasks, which are
1959 not implemented in most models beside ptaskL07, this will abort).
1960 * Add an example that schedules a DAX on an heterogeneous platform
1961 using a Min-Min strategy.
1962 * New function SD_workstation_get_current_task() that returns the kind
1963 of task currently running on a workstation in the sequential access
1965 * Raise some warnings when unexecuted tasks remains at the end of the
1966 simulation. This is usually caused by cycles in the DAG.
1969 * New function: SIMIX_process_set_function() called by MSG_set_function
1970 * Change the underlying waiting queue in semaphores so that a process
1971 can wait on several of them simultaneously (as in waitany).
1972 * Fix the way to handle tokens in semaphores so that all access patterns
1973 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1974 * kill the dirty pimple SIMIX_message_sizes_output()
1975 Please use (proper) visualization instead
1978 * New data container: setset (set of sets of elements)
1979 * New module: mmalloc (mapped malloc, allowing to have several
1980 independent segments of malloc)
1981 * New function: xbt_dict_cursor_set_data()
1982 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1983 * New function: xbt_dynar_is_empty()
1984 * New function: xbt_fifo_get_last_item()
1985 * Fix xbt_dynar_shrink(): use the right element size.
1986 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1987 uninitialized areas during expand.
1988 * Fix semaphores: previous implementation was severely broken.
1989 * Use library init/fini functions for our initialization.
1990 - you can use logs and other feature as soon as you want in your
1991 code (even before the xbt_init / MSG_init)
1992 - xbt_exit is now a no-op and produce a warning when used.
1995 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1996 benefit from the latest improvement to the simulation kernel.
1997 * Kill measurement sockets for now. If you rely on them, sorry. This
1998 release is not for you. This feature will be reintroduced in the
1999 future, but we cannot delay the release any further.
2000 * New function: gras_msgtype_get_name().
2001 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2002 * Fix (at last) the pmm example: it should not randomly fail anymore.
2004 Build chain: bug fixes and overall polishing
2005 * Cmake is now stable enough. Hence, we killed the autotools.
2006 * Port to windows ( TM :)
2007 * Fix the 'make install' target.
2008 No need to use 'make install-simgrid' anymore
2009 * Introduce a 'make dist' target compiling a *source* archive
2010 'make package' compiles a binary archive
2011 * Compile java files only on need
2012 * Add --cd and --setenv command line options to tesh
2013 * Out of source builds are not fully supported yet, but we are close
2014 * Enable supernovae and optimization flags by default for our users
2017 * Add layer to set up environment directly from lua, without XML.
2018 * The effect of gras_stub_generator can be achieved through
2019 lua too (check examples/gras/console/ping_generator.lua)
2021 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2023 SimGrid (3.4.1) stable; urgency=low
2025 The "Polishing easter eggs is probably a good idea" release.
2026 This is a bug fixes release only.
2030 * Fix a bug preventing the tasks from begin garbage collected.
2033 * Fix a bug occuring when a host involved in a communication fails.
2034 This was not detected properly by the other peer involved in the
2035 communication. Now, it's reported as a network error.
2038 * Warn the user about loop dependencies in data flow of DAX files
2039 * Obey the control-flow dependencies of DAX files
2042 * Add option "enable_smpi" allowing to not compile SMPI.
2043 Probably useful for the (Mac) users experiencing a build error here
2044 * Improve the detection of lua5.1 and ruby1.8
2046 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2048 SimGrid (3.4) stable; urgency=low
2050 The "Easter in Cargese" release. Also known as (major changes):
2052 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2053 ~> bindings were greatly improved
2054 ~> new tracing infrastructure for better visualization introduced
2056 * the "Welcome to configury modernity" release.
2057 ~> we switched from autotools to cmake, and improved our cdash
2060 A more detailled list of changes follow (full detail in svn log).
2062 Java Bindings: Various Cleanups
2063 * (install java-gcj-compat-dev on debian-like to use them)
2064 * Remove put/get: no need to export deprecated interface in Java
2065 Use send/receive instead.
2066 * Cleanup the examples and add a README per directory
2067 * Remove example autoDestination (that's the only way to go now)
2068 * Remove example explicitDestination (was a plain copy of basic)
2069 * Make JniException a runtime exception, so that there is no need to
2070 declare the fact that you may encounter such a beast. I guess that
2071 nobody will ever want to survive such error.
2072 * Create specific errors for each MSG case of failure:
2073 host failure, transfer failure, timeout, task cancelled
2074 * Cleanup the exceptions that may get thrown by each function
2075 * Other internal cleanups in Java bindings. Performance still bad :/
2076 Ruby and Lua Bindings: create them
2077 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2078 * That's new and great, you should try them out.
2079 Same functionalities than Java bindings, only even less polished
2081 * Kill the useless "rate" argument of SD_task_get_execution_time()
2082 Everyone used to provide -1 as a value, it was not used, and the
2083 semantic of a possible use wasn't even clear.
2084 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2085 either as comm costs or compute costs to mean that there is no
2086 such thing for that specific task.
2087 * Add a SD_task_set_name() function
2088 * Fix SD_task_unschedule() on typed tasks
2089 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2090 * In DAX loader, accept useless 'level' attributes to <job> since
2091 LIGO DAGs have them (seem to be to ease graphical representation).
2093 * Add an example masterslave_mailbox.c using send/receive and not
2094 the deprecated put/get interface.
2095 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2096 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2098 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2099 (declare MSG_USE_DEPRECATED to still have the old name)
2100 * Add a MSG_task_set_data() function
2101 * About trace replay (see examples/msg/actions):
2103 - Allow to work with splitted trace files for each process
2104 Give the specific trace file as argument of each process,
2105 and call MSG_action_trace_run(NULL)
2106 You can still have one merged file for all processes.
2107 - Fix implementation of collective operations
2108 * Allow task_execute() on 0-sized tasks (closes #10063)
2110 * This is the first release of SimGrid where SMPI is not considered
2111 beta anymore (even if some corners should still be improved)
2112 * Port over the new SIMIX_network submodule (internal refactoring)
2113 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2114 * Implement more missing elements of the standard:
2116 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2118 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2119 MPI_Scan() MPI_Get_processor_name()
2120 - Added implementation of missing case for Alltoall (warning: it's
2121 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2122 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2123 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2124 SMPI_MPI_Allgatherv()
2125 * Bug fixes include:
2126 - MPI_Waitsome() was broken
2127 - Allow relative includes in smpicc
2128 - Command line cfg argument 'reference_speed' was ignored...
2129 - Some functions did not properly lead to auto-benching of user code
2130 - smpicc passes -O2 by default (just like openmpi one)
2132 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2133 * Bug fixes about timeouts during communications
2134 * add SIMIX_message_sizes_output() as a pimple to write to file the
2135 amount of messages per size. Use gnuplot to get histogram.
2136 Pimple because that's the only user-visible function of simix,
2137 defined directly in xbt.h (irk, sorry)
2139 - Add a SIMIX_sem_get_capacity() function
2140 - Fix interactions with processe resume/suspende
2141 - release_forever() was stupidly broken
2142 - Fix SIMIX_display_process_status() for processes in a semaphore
2143 - Make SIMIX_sem_block_onto() user-visible
2144 * Refactoring context stuff:
2145 - Use pseudo-OOP for better modularity
2146 - reimplement SIMIX_process_kill() without process_schedule() so
2147 that the latter can take as invariant that it is called from
2149 - Merge context_start into context_new for sake of simplicity
2151 * Add a Vivaldi network model, coded live during SUD'10 ;)
2152 * Rename configuration variables to start a hierarchy:
2153 o cpu_model -> cpu/model
2154 o network_model -> network/model
2155 o workstation_model -> workstation/model
2156 * New configuration variables:
2157 o network/bandwidth_factor: correction to bandwith
2158 o network/latency_factor: correction to latency
2159 o netwotk/weight_S: correction to the weight of competing streams
2160 * Add a long description to the models, that users can see with such
2161 argument on the command line: --cfg=cpu/model:help
2162 * --help-models display the long description of all known models
2164 * config: add the ability to set a default value after registration
2165 Does not override any previously set value (e.g. from cmd line)
2166 * dict: allow to have integer key and data.
2167 When so, you need to use the following functions
2168 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2169 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2170 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2171 In contrary to regular dicts, the key is not malloced before copy.
2172 Mixing scalar and regular elements in the same dict is not tested
2174 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2175 Tracing for Visualization:
2176 * SimGrid is now instrumented in order to generate a trace file for
2177 visualization analysis: to use it, need to compile SimGrid with the
2178 "tracing" option enabled, and instrument the program using SimGrid with
2179 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2180 (among other functions).
2181 * The instrumentation only traces the platform utilization for now
2182 * Documentation to use the tracing functions and how to analyze the
2183 traces with the Triva tool is written.
2184 * More information about: SimGrid FAQ (in the section Tracing Simulations
2187 * We moved to cmake as default build system. Autotools support will
2188 be dropped soon. Check the FAQ for more info about how to use it.
2189 * Greatly improved our cdash/ctest interactions
2190 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2191 * Added memory checking tests with valgrind; lot of memleak fixing.
2192 This may be the first release of simgrid with so few memory issues
2193 * Added code coverage tests.
2194 Our coverage is still improvable, but at least we see it on cdash.
2196 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2198 SimGrid (3.3.4) stable; urgency=low
2200 The "Desktop Grid needs love too" release (also called Xmas release).
2202 Models improvements:
2203 * Major speedup in the maxmin system solving by using lazy evaluation
2204 Instead of solving completely the maxmin system at each iteration,
2205 only invalidate (and recompute) the modified parts.
2206 This new feature is enabled in default models but you can try to
2207 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2208 * Cas01 IMproved as default CPU model
2209 This CPU model is the same Cas01 model, but it uses the
2210 maxmin-selective-update flag and a heap structure to manage
2211 actions on SURF kernel.
2212 It reduces the complexity to find the next action to finish and,
2213 consequently, it's faster than the old Cas01.
2214 This is the new default CPU model (Cas01).
2215 * Rename the old Cas01 model to Cas01_fullupdate
2216 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2217 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2218 * CpuTI (CPU Trace Integration)
2219 A new CPU model whose objective is simulate faster when using
2220 availability trace files.
2221 Instead of using a full featured, over engineered maxmin system for
2222 CPU modeling, this model does the pre-integration of traces files
2223 to calculate the amount of CPU power available, and so, executes
2224 faster than the old CPU models.
2225 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2226 * Use LV08 as default network model since it gives better accuracy
2227 for small messages and shouldn't change things for big ones.
2228 Use --cfg=network_model:CM02 to get the previous behavior.
2231 ******************************************
2232 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2233 ******************************************
2234 * The new CPU model may changes simulations!
2235 The point is that events occurring at the exact same timestamp
2236 are not scheduled in the same order with the old and new
2237 version. This may be enough to completely change the execution
2238 of simulations in some cases.
2239 * The new network model will change simulations!
2240 This new model is more realistic than the previous one, so you
2241 should consider redoing your old experiments with this model.
2242 Sorry for the inconvenience.
2245 * Introduce the supernovae compilation mode
2246 When compiled that way, the whole SimGrid (or almost) is put in a
2247 single compilation unit and compiled in one shoot.
2248 This is to help gcc which has difficulties to inline stuff from one
2250 The speedup seem to be above 15%, althrough more tests are needed on
2251 amd64 to confirm that gain.
2254 * Port of MSG's mailbox on top of SIMIX network
2255 The put/get mechanism was greatly simplified on the way.
2258 * New SIMIX network module. Provides:
2259 - Mailbox: rendez-vous mechanism to find with who you want to speak
2260 - Synchronous send/recv: easier and hopefully faster since the
2261 logic is handled in the maestro process directly now
2262 - Asynchronous send/recv: you dreamt of it? It's here now
2263 Too bad that nobody cared enough to propagate the change to MSG.
2264 * Add semaphores as SIMIX synchronization mechanism.
2267 * new function SD_daxload(char*) to load a DAX file
2268 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2269 * Introduce typed tasks. Specify its kind and cost at creation.
2270 At scheduling, just give where it should be placed, and the cost
2271 for each involved resource is automatically computed.
2272 Existing constructors so far (more to come of course):
2273 - SD_task_create_comm_e2e() for end-to-end communication
2274 - SD_task_create_comp_seq() for sequential computation
2275 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2276 * new function SD_task_dump() for debuging display
2277 * new function SD_task_dotty(task,FILE*) writing to file the info
2278 about the task in dotty format
2279 * SD_task_dependency_exists() can now cope with having one of its
2280 arguments NULL. If so, it tests whether the other argument has any
2282 * Add getters on list of preceding/following tasks:
2283 SD_task_get_parents(task) and SD_task_get_children(task)
2284 * Add getters on amount of workstations and list:
2285 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2286 * Add getter on task kind: SD_task_get_kind(task)
2287 * Update the start_time and finish_time of tasks on completion/failure
2288 * Bugfix: Remove task from state swags when destroyed
2291 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2294 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2295 dynar with one separate thread per value of the dynar.
2296 * Change the prototype of xbt_thread_create(), sorry.
2297 Added a boolean parameter indicating whether we want to join this
2298 thread (used in SG only for now)
2299 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2302 * GTNetS wrappers should now be usable again (and betterly tested too)
2303 * Fix a major regression from 3.2 where the timeout provided to
2304 MSG_task_put_with_timeout() was used as absolute time before which
2305 the comm should be done.
2306 * Start to fix the <cluster> tag.
2307 - Internal links should be good now (beside of the loopback, which
2308 use the private link instead)
2309 - paths to the external world is still rather broken
2310 - the <route:multi> tag is just broken. Actually that's brain-dead.
2311 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2312 to make it less stupid
2313 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2314 * Fix a source-level compatibility glitch from 3.2: after defining
2315 MSG_USE_DEPRECATED, you can use the old name
2316 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2317 * Allow to compile from the SVN with automake 1.11
2318 * Fix some problems when using the "start_time" tag in deployment XMLs.
2319 * Fix #8569: XBT/synchro.h has redundant declarations
2320 * Fix #8563: MSG return values and exceptions
2321 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2322 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2323 Thanks to Robson Peixoto.
2324 * Fix a vicious bug in dictionaries inducing that some elements were
2325 not freed on xbt_dict_free()
2327 Portability report of this version:
2328 * Main portability targets:
2329 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2330 on (amd64/i386/ia64)
2331 - mac leopard on i386
2332 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2333 but nothing critical.
2334 * Other platforms: windows, AIX and others were not tested for this release
2336 Timing report of this version:
2337 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2338 scenario parameters). From 8h to a few seconds in desktop grid settings.
2339 * Supernovae brings about 25% speedup on i386.
2341 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2343 SimGrid (3.3.3) stable; urgency=low
2345 The "Need for Speed" release.
2347 The timings done to validate the 3.3.2 were faulty.
2348 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2350 The problem was a conversion from a manually handled vector to
2351 xbt_dynar_t on the critical path.
2352 xbt_dynar_foreach calls functions, inducing stack management crap.
2354 We inlined these functions and xbt_dynar_foreach is now breath taking.
2355 We also inlined xbt_swag_belong on the way.
2357 Here are some approximate speedup measurements (on master/slaves
2358 simulations lasting between 10s and 20s each):
2359 3.3.1 -> 3.3.2: about same performance
2360 3.3.2 -> 3.3.3: 40% speedup
2361 3.3.1 -> 3.3.3: 40% speedup
2362 3.3.1 with inline patch -> 3.3.3: 30% speedup
2364 Our reading is that the refactoring which occurred in 3.3.2 made us
2365 suffer much more from the xbt_dynar_foreach low performance, but
2366 once we solved this, this refactoring proved to be very performance
2367 effective. From the 40% speedup, somehow, 10% are due to the
2368 inlining and 30% to the refactoring.
2370 That's a pitty that gcc cannot inline functions placed in other files
2371 alone. We have to choose between:
2372 - break the encapsulation (by putting private data structures and
2373 accessors in headers files to help gcc)
2374 - live with low performance
2375 - switch to a decent compiler such as icc (not quite possible).
2377 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2379 SimGrid (3.3.2) stable; urgency=low
2381 The "Simplicity does not preceed complexity, but follows it" release.
2383 The main contributors of this release were (lexical order):
2384 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2387 * Extract the routing logic into its own object.
2388 (was dupplicated in network.c and workstation_LV07.c;
2389 Allows to implement other ways of storing that info)
2390 => kill now useless network_card concept
2391 - Use dynar to represent routes (instead of void** + int*)
2392 - kill link_set (use surf_network_model->resource_set instead)
2393 - Add a command-line option to choose the routing schema to use
2394 - Add three new models:
2395 * Floyd (shortest path computed at initialization)
2396 * Dijikstra (shortest path recomputed all the time)
2397 * Cached Dijikstra (shortest path computed on need)
2398 All these models where contributed by Silas De Munck, and are
2399 described in his ICCS09 paper.
2401 * Simplify model declaration
2402 (less redirections, less function to write when defining a model)
2403 - Factorize stuff between models:
2406 surf_model_resource_set(model)
2407 surf_model_resource_by_name(model, name)
2408 - Unify the types of models in s_surf_model_t (using an union)
2409 - Embeed fields of common_public directly into s_surf_model_t
2410 - Rename model methods:
2411 action_free ~> action_unref
2412 action_change_state ~> action_state_set
2413 action_get_state ~> action_state_get
2414 - Change model methods into functions:
2415 (model)->common_public->action_use ~> surf_action_ref
2417 * Implement a generic resource; use it as ancestor to specific ones
2418 (allows to kill duplicated code in models)
2419 Drawback: timer command don't need no name nor properties;
2420 workstation_CLM03 don't need no properties
2421 (but I guess we can live with those few bytes wasted)
2423 * Improve the action object model
2424 - implement a constructor avoiding dupplicated code about field
2425 initialization in generic_action part.
2427 * Kill the SDP model: it has an external dependency, is deprecated
2428 in flavor of modern lmm models, and didn't compile since a while
2431 * Relocation of the context module from XBT to SIMIX.
2432 (the context were decoupled from the simix processes, duplicating a lot of code)
2433 => a lot of code was factorized
2434 - less overhead is introduced during scheduling
2435 - simpler API for the context factory
2436 - the logic for process creation,destruction and manipulation was simplified
2437 * Simplification of the s_smx_process_t data structure.
2438 => accesing the simix level data associated to a process is faster now,
2439 and the code is a lot more readable.
2442 * Implement some more MPI primitives:
2443 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2444 -implementation: Bcast: flat or 2-ary tree (default),
2445 Barrier: 4-ary tree,
2447 Allreduce: Reduce then Bcast
2448 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2449 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2450 Alltoallv: flat tree, like ompi
2452 * Add support for optimized collectives (Bcast is now binomial by default)
2453 * Port smpirun and smpicc to OS X
2456 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2457 Shout out if you used it.
2460 * Display the list of still queued messages in SG mode when existing
2464 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2465 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2466 * Add xbt_str_from_file(FILE*)
2467 * Add xbt_dict_get_key achieving a linear reverse search
2468 * Remove the context module
2470 Portability report of this version:
2471 * Main portability targets:
2472 - Linux(debian)/x86/context
2473 - Linux(debian)/x86/pthreads
2474 - Linux(debian)/amd64/context
2475 - Linux(debian)/amd64/pthreads
2476 On these, we still have the eratic breakages of gras/pmm and
2477 amok/saturate_sg reported in previous version. We still think
2478 that the tests are the cause of the fault, not the tested code.
2480 - Mac OSX Leopard/x86/context
2481 Still false negative in tesh autotesting.
2482 Smpi still fails, but this time because readlink does not accept -f
2483 Everything seems to work properly beside of that.
2486 - AIX version 5.3 (only tested contexts this time)
2487 Smpi still fails there because mktemp is not installed.
2488 Everything seems to work properly beside of that.
2490 I managed to compile it for the first time, but several breakages.
2491 Won't delay the release for this exotic platform.
2493 * Windows: it's still lagging behind. If you want to help, please
2496 Timing report of this version:
2497 This version seem to be more than 5% faster than 3.3.1 (on linux
2498 64bits with contextes). The gain is less than expected, we are
2499 investigating this for next release.
2501 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2503 SimGrid (3.3.1) stable; urgency=low
2506 * Implement a --cfg-help to show existing configuration variables
2507 * Build chain do not require doxygen in maintainer mode
2510 * fix a bug on struct sizeof computation, which prevented the
2511 exchange of arrays of structs in some conditions
2512 - added a regression test about this in datadesc_usage
2513 * Allow the exchange of 0-long dynamic vectors.
2514 - for that, use -1 as indicator of dynamic size instead of 0
2515 - This implied to change any size from unsigned long to long,
2516 reducing a bit communication abilities, but I guess that with
2517 64bits being quite common, this is more than enough.
2518 - This also induce a protocol change, thus bumping network protocol
2519 version from 0 to 1 (if we have external users, we have to get
2520 clean on that point too ;)
2521 - added two regression tests about this in datadesc_usage
2522 * Be more verbose when propagating local exceptions
2523 This helps debugging.
2524 * Display the status of simulated processes when receiving SIGINT in
2528 * Allow to control the simulation from a trace file.
2529 New functions MSG_action_register() and MSG_action_trace_run()
2530 The first one allows to associate a function execution to each
2531 kind of action while the second one parses a trace file and
2532 triggers the corresponding actions within the system.
2533 For now, only a toy example is provided in examples/msg/actions
2534 * Add an exemple of process migration in examples/msg/migration
2535 * Fix a bug in task exchange which broke MSG_task_get_sender()
2536 Add a teshsuite regression test for that.
2537 [Bug: if MSG_task_get_sender() is called after sender exit,
2539 * Fix a bug which prevented suspend/resume to work properly
2540 * Display the status of simulated processes when receiving SIGINT
2541 This fixes a regression of v3.3. due to the introduction of SIMIX
2542 * Bug fixing in failure management:
2543 - trace could not start by a failure at time 0
2544 - failure during communications were not working
2547 * Add SIMIX_process_set_name() to change the name of the current
2548 process in the log messages.
2549 * Store smx_hosts in a dict since we only retrieve them by name
2550 * Move the configuration infrastructure to surf
2553 * Move the configuration infrastructure to surf
2556 * Massive internal cleanups:
2557 - Store internal structures on processes instead of hosts (allows
2558 to have more than one process per host, in addition of being more
2560 - Cleanup the initialization/finalization process
2561 - Kill a whole bunch of unneeded synchronization:
2562 processes run in exclusive manner within the simulator
2563 - Move queues from global tables to process data fields
2565 - now accept -platform and -hostfile arguments
2566 - Pass the right rank value to processes according to the hostfile
2567 * Compile the examples by default, and use them as regression tests
2568 * Implement MPI_Wtime()
2569 * Change the reference speed to a command line option
2572 * TCP_gamma can now be specified as command line option using
2573 --cfg=TCP_gamma:10000000.0
2574 * Change the --surf-path cmd line option into --cfg=path:
2577 * Also include strbuff from xbt.h public header
2578 * xbt_ex_display(): do not free the exception after displaying
2579 This allows to do more with the given exception afterward.
2580 Users should call xbt_ex_free() themselves.
2584 Portability report of this version:
2585 * Main portability targets:
2586 - Linux(debian)/x86/context
2587 - Linux(debian)/x86/pthreads
2588 - Linux(debian)/amd64/context
2589 - Linux(debian)/amd64/pthreads
2590 These targets fail about 1/10 of times on gras/pmm, but we believe
2591 that this is because of the test, not because of simgrid.
2592 amok/saturate_sg fails even more rarely, and the test may not be
2595 - Mac OSX Leopard/x86/context
2596 The test suite still spits tons of errors because some obscure
2597 force prevents us from removing the temporary directories
2598 arguing that they still contain some metadata I've never heard of.
2599 Smpi fails because seq is not installed.
2600 Everything seems to work properly beside of that.
2603 - AIX version 5.3 (both contexts and pthread)
2604 Smpi still fails there because mktemp is not installed.
2605 XML inclusions seems rosty on AIX.
2607 * Windows: it's still lagging behind. If you want to help, please
2610 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2612 SimGrid (3.3) stable; urgency=high
2616 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2617 [Malek Cherier & Mt]
2619 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2622 Until now, GRAS were using MSG as an interface to SURF. It was
2623 quite difficult because both interface have several differences
2624 (MSG channels vs GRAS sockets were the most notable point).
2626 This also opens the gate to SMPI (which should occur soon) and speed
2627 up simulations by to 40% (even if it were not the main goal).
2629 **************************************
2630 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2631 **************************************
2632 The point is that events occuring at the exact same timestamp are
2633 not scheduled in the same order with the old and new version. This
2634 may be enough to completely change the execution of simulations in
2635 some cases. Sorry for the inconvenience.
2637 * Cleanup and upgrade the XML format to push further scalability
2638 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2640 * Improve the testing infrastructure with tesh. Now a very large part of
2641 the code is tested not only by being run but also by checking that the
2642 output match an expected output [Mt].
2644 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2645 is really less memory-demanding, which should allow you to use
2646 larger files in SimGrid [AL].
2648 * Inform valgrind about our contextes, so that it becomes usable
2649 with the default (and more effecient) version of SimGrid
2650 [contributed by Sékou Diakite, many thanks]
2653 * Introduce a listener thread in charge of receiving incoming
2654 messages from the network. It allows to overlap communication and
2655 computation but most notably, it removes some stupid deadlocks due
2656 to the fact that so far, a process could not send and receive at
2657 the same time. This made most non trivial communication schema
2659 * Convert the PIDs from long int to int to match the MSG ones (and
2660 linux ones too) [Mt]
2661 * New function: gras_agent_spawn() to launch a new process on
2662 current host. Only working in simulation for now. [Mt]
2663 * New function: gras_os_hostport() returning a constant form (ie,
2664 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2667 * Make the backtrace of exceptions more human readable [Mt]
2668 * New module: xbt/str [Mt]
2669 a ton of string utility functions (split, join, printf to a newly
2670 allocated buffer, trim, etc)
2671 * New module: xbt/hash [Mt]
2672 SHA1 hashing algorithm (more to come if needed)
2673 * New module: xbt/synchro [Mt]
2674 synchronization tools (mutex and conditions) working the same way
2675 in simulation and in real life (mainly useful for GRAS, but not
2677 * New module: xbt/queue [Mt]
2678 classical producer/consumer synchronization scheme
2679 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2680 (using the classical functions will get serialized) [Mt]
2681 * Make dictionary internal table dynamic. No need to specify its size
2682 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2684 * Make sure the log channels are organized as a tree under windows
2685 (because of ANSI C compatibility issue, any channel were child of
2686 root directly) [Mt].
2689 * Cleaned many thing in surf and fixed a few bugs [AL].
2690 * Add a nice command line configuration mechanism to compose models [AL].
2691 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2692 the previous one (KCCFLN05). It relies on something that looks like
2693 a max-min sharing mechanism but cannot be written as such. A new solver
2694 was thus designed [AL].
2695 * Add a new solver to lmm. Based on Lagrange optimization and
2696 gradient-based descent, it enables to efficiently maximise systems s.a
2698 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2700 This solver enables to propose two new network models for TCP Reno and
2701 TCP Vegas based on Low's work. These models still need to be fully
2702 tested though [Pedro Velho].
2705 * Bug fix in SD_simulate. Now the time bound given as argument is
2707 * Use the new parallel task model (ptask_L07) as default.
2708 * Use the SURF command line configuration mechanism.
2709 * 0-size tasks (for synchronization) should now work.
2711 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2713 SimGrid (3.2) stable; urgency=high
2717 We still experience issues on this platform, but we believe that at
2718 least MSG is usable.
2720 GRAS API BREAKAGE (for simplification purpose, sorry):
2721 * the gras_msgtype_by_name is not used anymore. Instead of
2722 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2723 you can write (and must)
2724 gras_msg_send(toserver, "request", &request);
2725 - If you still want to pass a gras_msgtype_t to the function (to cache
2726 the type and avoid the lookup time), use the gras_msg_send_() variant.
2727 - Impacted functions:
2728 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2729 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2730 * The callbacks are now expected to return 0 when everything went well
2731 (just like the main() function)
2733 GRAS new features and improvements:
2734 * New module mechanism where user code can use per process globals [Mt]
2735 This is similar to gras_userdata_*() functions, but for libraries. It
2736 factorize some code developped over and over in the examples and AMOK.
2737 It has still to be documented and used (only amok/peermanagement is
2739 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2740 loss when they were small enough to fit into the buffer and sent quickly
2741 enough so that they can all get received in one shoot.
2742 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2743 if not found. Use the *_or_null() variant for the old semantic.
2744 * In gras_msg_handle, do not discard messages without callback.
2745 They are probably messages to be explicitly awaited later (ie, proofs of
2746 mis-synchronization in userland since they are sent before being awaited)
2747 No big deal usually.
2748 * gras_socket_meas_send/recv: semantic changed!
2749 The numerical arguments used to be (1) the total amount of data to send
2750 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2751 messages. This was need for the fool willing to send more than MAXINT
2752 bytes on quite fat pipes.
2755 * Do really rename the hostmanagement module to peermanagement. [Mt]
2756 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2757 but this is rather new and this was documented in the module
2758 documentation (poor excuses, I admit)
2759 * Bandwidth measurement semantic changed! This follows the changes to
2760 gras_socket_meas_send/recv explained above.
2763 * A sequential mode has been added to the workstations. When a workstation
2764 is in sequential mode, it can execute only one task, and the other tasks
2765 are waiting in a FIFO. [Christophe Thiery]
2768 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2769 model for SIMDAG. [Christophe Thiery]
2770 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2771 equal to 0, which caused some bad behaviors in
2772 saturated_constraint_set_update. I now use a threshold mechanism like in
2776 * When running manually src/testall, you select specific units [Mt]
2777 testall is the result of our cunit mechanism, and should replace all
2778 the scripty thingy around since bash don't run easily on billware.
2780 * A mallocator system has been added. [Christophe Thiery]
2781 Mallocators allow you to recycle your unused objects instead of freeing them
2782 and allocating new ones.
2784 Documentation update:
2785 * FAQ reworking + New FAQs:
2786 - "Valgrind spits tons of errors!" [Mt]
2787 - "How to repport bugs" [Mt]
2788 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2789 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2790 - Communication time measurement within MSG [AL]
2791 - I experience weird communication times when I change the latency [AL]
2792 * GRAS tutorial [Mt]
2794 - an introduction to the framework and to the used communication model
2795 - an initiatic tour introducing the most proheminent features:
2797 . Lesson 0: Installing GRAS
2798 . Lesson 1: Setting up your own project
2799 o Part 2: Message passing
2800 . Lesson 2: Exchanging simple messages
2801 . Lesson 3: Passing arguments to the processes (in SG)
2802 . Lesson 4: Attaching callbacks to messages
2803 . Lesson 5: Using globals in processes
2804 . Lesson 6: Logging information properly
2805 . Lesson 7: Using internal timers
2806 . Lesson 8: Handling errors through exceptions
2807 . Lesson 9: Exchanging simple data
2808 . Lesson 10: Remote Procedure Calling (RPC)
2809 . Lesson 11: Explicitely waiting for messages
2810 . Recapping of message passing features in GRAS
2811 - A HOWTO section containing:
2812 o HOWTO design a GRAS application
2813 More are due, of course. They will come latter. In the meanwhile, you can
2814 check the examples which are still here.
2816 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2818 SimGrid (3.1) stable; urgency=high
2822 There was a stack corruption somewhere, visible only when optimizing
2823 with these versions. [Vince]
2826 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2827 module that enabled to play with Directed Acyclic Graphs. It is built
2828 directly on top of SURF and provides an API rather close to the old
2829 SG. Some old codes using SG are currently under rewrite to check that
2830 all needful functions are provided. [Christophe Thiery]
2833 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2834 extension of the classical CLM03 model that gracefully handles
2835 failures. This is now the default model for MSG and GRAS. It doesn't
2836 handle parallel tasks yet though. [AL]
2837 * Bug fix: Weights were not correctly set in the network part.
2838 WARNING: This may have resulted in incorrect results with simulations
2839 where there are more than one flow on a given link. [AL]
2842 * After a (long ?) discussion on simgrid-devel, we have decided that the
2843 convention we had on units was stupid. That is why it has been decided
2844 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2845 WARNING: This means that all previous platform files will not work as
2846 such with this version! A warning is issued to ask users to update
2848 A conversion script can be found in the contrib module of the CVS, under
2849 the name contrib/platform_generation/surfxml_update.pl [MQ]
2852 * Bug fix: Processes were started in reverse order, wrt deployment file.
2853 WARNING: if your code relies on this bug, please fix it. [AL]
2854 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2855 being executed on two different locations. [AL]
2856 * Bug fix: Failures are now better supported thanks to Derrick's tests
2857 (there was many failure situations I hadn't thought of and that weren't
2858 correctly handled). [AL]
2859 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2863 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2864 Exception occurring on server-side are propagated back to client (!).
2866 API CHANGE: the callback changed their prototype. Change:
2867 int my_handler(gras_socket_t expeditor, void *payload_data) {
2869 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2870 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2872 * New! function: gras_msg_handleall to deal with all messages arriving
2873 within a given period.
2874 * New! function: gras_socket_server_range to get a server socket in a
2875 range of port numbers (ease to avoid port number conflicts) [MQ]
2876 * New! gras processes display their backtrace when they get a SIGUSR1
2877 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2878 Sweet to debug RL processes [MQ]
2882 - Do not force experiment sizes to be expressed in kb, or it becomes
2883 impossible to measure the latency this way (needs one byte-long tests)
2884 WARNING: this changes the amok_bw_* function semantic. [MQ]
2885 - Implements the link saturation stuff. [MQ]
2886 * Peer management module:
2887 New! module factorizing code that we wrote over and over [MQ].
2890 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2891 - Test units are placed directly into the library code, they get extracted
2892 automatically and placed into the src/testall binary.
2893 - Convert most of the XBT tests to this system.
2894 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2895 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2896 live in the same C code. This required to change a little bit the API
2897 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2898 * New module: sparse graph structure with basic algorithms (this is work
2899 in progress and the API is not considered to be frozen yet). [AL]
2900 * Display more information on backtraces: source line & function names are
2901 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2902 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2903 * Reworked a little bit some #include statements to load only required
2904 headers. Some user code that relied on SimGrid to include stdlib or
2905 stdio may need to include it by themselves. [AL]
2906 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2908 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2909 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2910 but I guess that almost nobody used those parts. [MQ]
2912 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2914 SimGrid (3.0.1) stable; urgency=low
2917 * Unfortunately, I had missed 5 misnamed functions:
2918 xbt_fifo_item_t xbt_fifo_newitem(void);
2919 void xbt_fifo_freeitem(xbt_fifo_item_t);
2920 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2921 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2922 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2923 They're now deprecated. Please use their new versions:
2924 xbt_fifo_item_t xbt_fifo_new_item(void);
2925 void xbt_fifo_free_item(xbt_fifo_item_t);
2926 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2927 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2928 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2930 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2931 * Documentation: xbt_log module unmercifully reworked [MQ]
2932 * Bugfix: there was a problem with the ending of contexts with
2933 the pthread backend. It caused some weird deadlock or behavior
2934 depending on the pthread implementation. [AL]
2935 * Bugfix: get the exceptions raised in the simulator repport where
2936 and why they come from when they are not catched in time [AL, MQ]
2939 * Bugfix: Do repport the error when two non-connected hosts try to
2940 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2943 * Add additionnal checkings on communications. Assert that two
2944 communicating hosts are connected by a set of links... [AL]
2947 * Add additionnal checkings on channel values in communication [AL]
2948 * New: MSG_task_get_source to see on which host a task was generated [HC]
2949 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2950 the number of tasks waiting to be received on channel and sent
2952 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2953 waits for the first task coming from a given host.. [AL]
2955 GRAS new functionnalities: [MQ]
2956 * Enhance the parsing macro to allow the size of multidimentional objects
2957 to be given thru annotations.
2958 * New example (and documentation): Matrix Multiplication a la RPC
2959 (as when I was young!) and fix a bunch of bugs found on the way.
2961 GRAS performance improvements: [MQ]
2963 * Reduce the amount of cbps creation/destruction by making it static to
2964 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2967 * Change libdata to a set so that we can search for stuff by ID (and thus
2968 reduce the insane amount of dict lookups)
2971 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2972 RL communication on top of SG since it's so uneffective.
2973 It may also be used for inter-thread communication in RL, one day.
2974 * Use gras_datadesc_copy() to exchange messages on top of SG
2976 - improve message exchange performance on top of SG
2977 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2978 * Don't exchange on the network the size of the used part of buffer,
2979 instead, specify the possible buffer size to read().
2981 - reduces the amount of read/write calls (one pair per exchange)
2982 - reduces the amount of exchanged data (the size)
2983 - allows to retrieve all arrived data on receiver side, if we don't need
2984 it right now (subsequent read will peek the buffer)
2985 - allows the receiver to proceed with the begining of the stream before
2986 everything is arrived
2987 - make it possible to build an iov transport (using readv/writev)
2989 - take care of the data with non-stable storage (like stacked data),
2991 * If possible, TCP send uses vector I/O (when writev() is here)
2992 - Don't use it for receive since we send data sizes and data on the
2993 same stream, so we wouldn't be able to chain large amount of chunks
2994 before having to flush the stuff to read the size.
2995 * Rework the transport plugin mechanism to simplify it and reduce the
2996 amount of pointer dereferencement when searching for the right function
2999 * I guess that now, we do almost as few system calls as possible while
3000 doing as few data copy as possible.
3002 To improve it further, we could try to send all the sizes first and then
3003 all the data (to use iov on receiving size), but it's only a partial
3004 solution: when you have 2 dimensional data, the sizes of the second
3005 dimension is data of the first dimension, so you need 3 streams.
3007 I'm not sure the potential performance gains justify the coding burden.
3009 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3011 SimGrid (3.00) stable; urgency=high
3014 * New! Give the possibility to hijack the surf parser and thus bypass
3015 MSG_create_environment and MSG_launch_application. Have a look at
3016 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3018 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3020 SimGrid (2.96) unstable; urgency=low
3025 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3026 This deprecates the xbt_error_t mechanisms.
3027 It modifies (simplifies) all XBT and GRAS API.
3028 MSG API keeps unchanged (exceptions raised by XBT are catched from
3029 within MSG and masked with existing error handling facilities)
3032 * New! Add a FATPIPE model. [AL]
3033 * New! Add a parallel task model. [AL]
3034 * New! Add automatically a loopback interface (in the default
3035 network model) if none was precised.
3038 * Bugfix: MSG_process_resume now works with the current running process.
3040 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3041 * Modification of MSG_task_get_compute_duration. Once a task has been
3042 processed, the value returned by this function is now equal to 0. [AL]
3043 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3044 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3045 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3046 of a m_task. For now, MSG_task_cancel only works with computation
3048 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3049 of the processor (in Mflop/s) regardless of the current load on the
3051 * API Change: use proper naming convention for MSG_getClock and
3052 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3054 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3055 This function changes the priority of a computation task. This priority
3056 doesn't affect the transfer rate. A priority of 2 will make a task
3057 receive two times more cpu power than the other ones. This function
3058 has been added to suit the needs of Nguyen The Loc and hasn't been that
3059 much tested yet. So if it fails, please report it and send me your code.
3061 * API Change: removed all functions and types that were marked "deprecated"
3062 since many months. Renamed MSG_global_init_args to MSG_global_init.
3064 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3066 SimGrid (2.95) unstable; urgency=low
3069 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3070 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3071 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3074 * Bugfix: Allow absolute paths for platform description files [MQ]
3075 * Bugfix: do free the variables after use. Leads to drastic performance
3077 * Implement max_duration (ie, timeouts) on resources [AL]
3080 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3082 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3083 the needed support (timeouts) [AL]
3086 * Implement measurement sockets. You can now get the bandwidth between two
3087 hosts thanks to AMOK (see below). [MQ]
3088 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3089 dynar over the network (yeah) [MQ]
3090 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3091 * simple/static token-ring example. [Alexandre Colucci and MQ]
3092 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3093 in place before [MQ]
3096 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3097 type parsing. "short" and "long" modifiers were also ignored; other
3098 modifier (such as reference level) are still ignored. [MQ]
3099 * Embeed the buffer size within the buffer itself on SG. [MQ]
3100 That way, send() are atomic and cannot get intermixed anymore (at least
3101 the ones which are less than 100k; bigger messages still have the issue)
3102 * Array size pushed by the field, not by the field type (or each
3103 and every long int will push stuff to the cbps) [MQ]
3104 * use select() to sleep since it allows to portably sleep less than one
3107 GRAS (minor cleanups)
3108 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3111 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3112 * type category 'ignored' killed as it was never used and were difficult
3114 * whether a type can cycle or not is now a flag, leaving room for more
3115 flags (as "ignored", if we feel the need one day ;)
3116 * Rename raw sockets to measurement sockets since "raw" has another
3117 meaning in networking community.
3120 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3121 and offers features not belonging to GRAS but that most applications
3122 need. One day, it may be a set of plugins loadable at runtime.
3123 * New module: bandwidth
3124 bandwidth measurement between arbitrary nodes running this module. [MQ]
3126 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3128 SimGrid (2.94) unstable; urgency=low
3130 The first beta release of SimGrid 3 !
3134 * Update the main page and the FAQ. Adding references to gforge.
3137 * Add a gras_os_getpid function.
3140 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3141 * Extend the logs so that they also print PID, hostname, date, ... if
3143 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3144 and kill the old version which were in testsuite/
3145 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3149 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3150 for users. Instead, each programming environment (GRAS, MSG,...) use it
3151 when needed to provide such a feature to users.
3152 Don't shortcut the mechanism or you will also shortcut the virtualization
3153 you need on the simulator.
3157 * Cleanups in configury with regard to compile optimization/warning flags.
3158 Also add -fno-loop-optimize to any powerpc since it's the optimization
3159 killing gcc (< 3.4.0).
3160 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3161 needed by MSG examples complications
3162 * Borrow configury beautifications from PHP
3165 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3166 warning (thanks loris for stumbling into this one).
3167 * Bugfix: stop loading private headers (gras_config.h) from the public
3171 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3172 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3173 RL would imply a DNS resolver, which is *hard* to do in a portable way
3174 (and therefore delayed).
3175 * Implement a real timer mechanism and use it in timing macros. This allows
3176 to avoid rounding errors and get a 0.000005 sec precision in timing
3177 macros. While I was at it, various cleanups:
3178 - allow to declare more than one timed section per file (fix a stupid bug)
3179 - move some private declaration to the right place
3180 - merge conditional execution and timing macros into emulation module
3181 - document the module
3182 - make sure the module cleanups its mess on gras_exit
3183 * Documentation improvements:
3184 - (new) how to compile applications using GRAS
3185 - (new) emulation support (timing macros)
3187 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3189 SimGrid (2.93) unstable; urgency=low
3191 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3194 - Use Paje properly where used. Still to be sanitized properly.
3195 - Portability fix: Add an implementation of the contexts using pthread
3199 - Add xbt_procname(): returns the name of the current process.
3200 Use it to show the current process's name in all logging.
3202 - fix detection of older flex version and the reaction, since we do
3203 depend on modern ones (we use lex_destroy)
3204 - Better separation of SG and RL in the libs: remove all simulation code
3205 from libgras. As a result, this lib is now only 200k when stripped.
3206 Some of the xbt modules may also be duplicated (two sets and such) and
3207 should be cleaned/killed before SG3.
3208 - Insist on using xlC on AIX because of weird problems involving gcc there.
3209 - Cleanup the make remote stuff. This is now done by scripts
3210 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3211 mainly for our private use, but we're working on changing them to user
3214 - Bugfix: flush the socket on close only if there is some *output*.
3215 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3217 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3220 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3222 SimGrid (2.92) unstable; urgency=low
3224 Alpha 3 on the path to SimGrid 3
3228 - New! First try of benchmarking macros.
3229 - New! First try so that gras_stub_generator generate deployment and
3230 remote compilation helpers.
3232 - Bugfix: Initialization fix in msg_test.
3236 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3238 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3240 - API Change: gras_os_sleep to take the amount of seconds as a double.
3241 Accepting an int was error prone since it was the only location where
3242 seconds were coded as such. It leaded to damn rounding errors.
3243 - Bugfix: Hard to belive that timers ever worked before this.
3245 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3247 SimGrid (2.91) unstable; urgency=low
3249 Alpha 2 on the path to SimGrid 3
3253 - Bug fix in the lmm_solver.
3255 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3256 through the function MSG_paje_output.
3257 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3258 - It is possible to bound the rate of a communication in MSG with
3259 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3260 to write it in the changelog).
3261 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3266 - Various cleanups to the autotools stuff
3267 - Begin to move Gras examples to examples/gras/
3268 - Let make distcheck work again (yeah!)
3270 - documentation overhauled using doxygen.
3271 gtk-doc-tools is dead in SimGrid now.
3272 - Automatically extract all existing logging categories, and add the list
3273 to the documentation (long standing one, to say the less)
3275 - Cleanup the known architecture table. Reorder the entries to group what
3276 should be, and use a more consistent naming scheme.
3277 (some of the test dataset are still to be regenerated)
3278 - New! Allow library to register globals on each process just as userdata
3280 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3281 lookup time (for now).
3282 Use it in msg and trp.
3283 This cleans a lot the internals and helps enforcing privacy of the
3284 headers between the gras components.
3285 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3286 - Bugfix: gras_os_time was delirious in RL.
3287 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3289 - Reenable GRAS now that it works.
3291 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3293 SimGrid (2.90) unstable; urgency=low
3295 Alpha 1 on the path to SimGrid 3
3297 * It is a long time since the last release of SimGrid. I'm sorry about
3298 that but as I had told you, I was rewriting a lot of things. I apologize
3299 to those who had been reporting bugs to me and that I had not answered.
3300 If your bug is still in the new version, please tell me. Here is a
3301 summary of the main changes.
3303 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3304 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3305 portability, speed, and a lot more but you'll figure it out later.
3306 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3307 I wanted to release the new MSG as soon as possible and I have
3308 broken GRAS, which is the reason why, for now, only MSG is fully
3309 functional. A laconic description of these projects is available
3310 in the documentation.
3312 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3313 kernel optimized for our needs (called SURF but only the developers
3314 should use it). Hence, MSG is now roughly 30 times faster and I think
3315 that by rewriting a little bit MSG, I could event speed it up a little
3316 bit more. Beside the gain in speed, it is also much easier to encode a
3317 new platform model with SURF than it was with SG. More to come...
3319 * REVOLUTION 3: I have tried to change a little as possible the API of
3320 MSG but a few things really had to disappear. The main differences
3321 with the previous version are:
3322 1) no more m_links_t and the corresponding functions. Platforms are
3323 directly read from a XML description and cannot be hard-coded
3324 anymore. The same format is used for application deployment
3325 description. The new format is described in the documentation.
3326 Have a look in tools/platform_generation. There is a tiny script
3327 that converts from the old platform format to the new one. Concerning
3328 the application deployment format, parsing the old one is tricky.
3329 I think most of you should however be able to convert your files. If
3330 it is really an issue, I can write a C code that does the conversion.
3332 2) the toolbox tbx does not exist anymore. We now have a library
3333 with much more data-structures but without the hash-tables (we have
3334 dictionaries that are much faster).
3336 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3338 *****************************************************************************
3339 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3340 * don't feel like converting the oldies. (Mt) *
3341 *****************************************************************************
3344 Version 2.90: "the long awaited one"
3345 - Finished rewriting and debugging MSG. Rewrote the documentation.
3346 - disable GRAS for now since it needs to be ported to the newest SG
3349 - Finish the port to windows (using mingw32 for cross-compile)
3352 - Main loop and datastructures of SURF. A cpu resource object is
3353 functional. Surf can thus be used to create cpu's with variable
3354 performance on which you can execute some actions.
3356 2004-11-15 Martin Quinson
3357 - Port to ARM. Simply added the alignment and size descriptions. Should
3358 work, but the ARM machines are so slow that I didn't had the opportunity
3359 to 'make check' over there yet.
3361 2004-11-15 Arnaud Legrand
3362 - Trace manager now written. It uses a heap structure and is therefore
3363 expected to be efficient. It may however be speeded up (particularly
3364 when many events occur at the same date) by using red and black
3365 trees. One day maybe...
3366 - Max-min linear system solver written. It uses a sparse matrix
3367 structure taking advantage of its expected use. Most operations are
3368 O(1) and free/calloc are called as few as possible. The computation of
3369 the minimum could however be improved by using a red and black tree
3372 2004-11-03 Arnaud Legrand
3373 - Rename every gras_* function that was in xbt/ to its xbt_
3375 - Add a heap and a doubly-linked list to xbt
3376 - Added a dichotomy to the dictionaries. make check works as well before
3377 so I assume that the patch is correct. I do not know however if things
3378 run effectively faster than before now. :)
3380 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3381 SimGrid, and the version number is bumped to 2.x
3383 2004-10-29 Martin Quinson
3384 - Introduction of the remote errors.
3385 They are the result of a RMI/RPC on the remote machine.
3386 ErrCodes being scalar values, you can't get the host on which those
3387 errors did happen. Extending the error mechanism as in Gnome is possible.
3388 No idea yet whether it is a good idea.
3390 2004-10-28 Martin Quinson
3391 - Interface revolution: the Starred Structure Eradication.
3392 I used to do typedef struct {} toto_t; and then handle *toto_t.
3393 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3394 - ???_t is a valid type (builded with typedef)
3395 - s_toto_t is a structure (access to fields with .)
3396 - s_toto is a structure needing 'struct' keyword to be used
3397 - e_toto_t is an enum
3398 - toto_t is an 'object' (struct*)
3400 typedef struct s_toto {} s_toto_t, *toto_t;
3401 typedef enum {} e_toto_t;
3402 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3403 s_toto_t) is private.
3405 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3406 so that it can set the variable to NULL. It was so for dicts and sets,
3407 it changed for dynars.
3409 - Fix a bunch of memleaks in dict_remove
3410 - Fix a bug in sg/server_socket opening: it failed all the time.
3412 2004-10-07 Martin Quinson
3413 - Speed up dynar lookup operation a bit.
3415 gras_dynar_get is dead.
3417 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3418 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3419 (which gives you the address of the stored data).
3421 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3422 copy of the data using an affectation to do the job and not a memcpy.
3424 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3425 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3427 It does not really speedup the dynar test because they are
3428 setting elements all the time (and look them seldom). But the dict does
3429 far more lookup than setting.
3431 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3433 2004-10-05 Martin Quinson
3434 - Allow to (en/dis)able the cycle detection at run time.
3436 Whether we should check for cycle or not is now a property of each
3437 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3438 datadesc_cycle_unset allow to remove this property when previously set.
3440 Note that the cycle detection is off by default since it impacts the
3441 performance. Watch the data you feed GRAS with ;)
3443 This property is hereditary. Any element embedded in a structure having it
3444 set have it set for the time of this data exchange.
3446 You should set it both on sender and receiver side. If you don't set it on
3447 sender side, it will enter an endless loop. If you forget on receiver
3448 side, the cycles won't be recreated after communication.
3450 - Header reorganization.
3451 Kill gras_private.h, each submodule must load the headers it needs.
3453 2004-10-04 Martin Quinson
3454 - Interface revolution: do not try to survive to malloc failure.
3456 Now, gras_malloc and friends call gras_abort() on failure.
3457 As a conclusion, malloc_error is not a valid error anymore, and all
3458 functions for which it was the only gras_error_t return value are
3459 changed. They now return void, or there result directly.
3460 This simplify the API a lot.
3462 2004-09-29 Martin Quinson
3463 - Re-enable raw sockets.
3464 Created by gras_socket_{client,server}_ext;
3465 Used with gras_raw_{send,recv}
3468 It should allow to kill the last bits of gras first version soon.
3470 This is not completely satisfactory yet (duplicate code with
3471 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3474 - Simplify transport plugin (internal) interface by not passing any
3475 argument to _server and _client, but embedding them in the socket
3478 2004-09-28 Martin Quinson
3479 - Finish the port to AIX.
3480 autoconf was my problem (segfault within the malloc replacement
3481 function. No idea why)
3483 2004-09-16 Martin Quinson
3484 - Fix some size_t madness on 64bit architectures.
3486 2004-09-08 Martin Quinson
3487 - Reduce the number of system headers loaded, overload some more system
3488 calls (such as malloc to cast the result of the system one, and work
3490 - Fix and reintroduce the config support
3492 2004-09-07 Martin Quinson
3493 - Source code reorganization to allow Arnaud to surf all over there.
3494 - Allow to document the logging categories.
3495 - Remove all uppercase from logging categories and useless cleanup in names.
3497 2004-08-18 Martin Quinson
3498 Version 0.6.2 (protocol not changed; API changed)
3499 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3500 gras_error_t), and NULL when not found. Functions expecting a msgtype
3501 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3502 hopefully usefull message.
3503 - Portability to prehistoric sparcs again
3505 2004-08-17 Martin Quinson
3506 Version 0.6.1 (protocol not changed; ABI not changed)
3507 - prealloc some buffers to speed things up
3509 2004-08-11 Martin Quinson
3510 Version 0.6 (protocol not changed; ABI expended)
3511 - The parsing macro can deal with the references, provided that you add
3512 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3514 2004-08-09 Martin Quinson
3515 Version 0.5 (protocol not changed; ABI changed)
3516 - Allow to off turn the cycle detection code in data exchange at
3517 compilation time. It should be at run time, but I'm short of time (and
3518 the config stuff is still broken). That way, we keep dict out of the
3519 critical path, which is good because the performance is poor:
3520 - search not dichotomial yet
3521 - dynar give no way to access their content and memcpy everytime
3522 - In composed data description (struct, ref and so on), stop foolness of
3523 keeping the subtype's ID, but store the type itself. This keeps sets out
3524 of the critical path, which is good since they rely on dynar and
3525 dictionnaries. The only loose of that is that we cannot detect the
3526 redeclaration of a structure/union with another content (but I'm not sure
3527 the code detected well this error before anyway). We still can detect
3528 the redefinition discrepancy for the other types.
3529 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3530 it breaks the code because of type-punning used all over the place).
3531 This breaks on all non-gcc architectures (for now).
3533 All those changes (plus the buffer of last time) allow me to gain 2 order
3534 of magnitude on cruel tests consisting of 800000 array of integers on two
3535 level of a hierarchical structure (200 secondes -> 4 secondes)
3538 - the selector of reference must now return the type it points to, not
3539 the ID of this type.
3541 2004-08-06 Martin Quinson
3542 Version 0.4 (protocol changed; ABI not changed)
3543 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3544 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3545 - Implement a buffer transport, and use it by default (it relies on tcp in
3546 real life and on sg in simulation).
3547 That's a bit hackish since I had a new field to the structure to store
3548 its data without interfering with the subtype ones. Inheritance
3549 is tricky in C. And that's a kind of reverse inheritance with one class
3550 derivating two classes. Or maybe a game with java interfaces. Anyway,
3551 that's damn hard in C (at least).
3552 Moreover, I got tired while trying to ensure plugin separation and
3553 genericity in SG mode. MSG wants me to do weird things, so let's go for
3554 cruel hacks (temporarily of course ;).
3555 See comment in transport_private.h:71
3556 - do not include all the _interface headers in private but in the files
3557 which really need them (to cut the compilation time when they are
3560 2004-07-26 Martin Quinson
3561 Version 0.3 (protocol not changed; ABI changed)
3562 - Major overhault of the datadesc interface to simplify it:
3563 - shorted the function names:
3564 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3565 - add a trivial way to push/pop integers into the cbps without malloc.
3566 This allows to make really generic sub_type description, which simply
3567 pop their size of the stack.
3568 - add a function gras_datadesc_ref_pop_arr() which does what users want
3569 most of the time: Declare a dynamic array (which pops its size of the
3570 stack) and declare a reference to it. Poor name, but anyway.
3571 - kill the post-send callback, add a post-receive one
3573 2004-07-23 Martin Quinson
3574 Version 0.2 (protocol changed; ABI changed)
3575 - add some testing for cpbs in the test cases, and fix some more bugs.
3576 This invalidate again the little64 data file, since I cannot regenerate
3578 - remove an awfull optimization in the logging stuff, allowing me to:
3579 - understand it again
3580 - learn gcc how to check that the argument match the provided format
3581 - fix all errors revealed by gcc after that
3582 - internal keys of dict are not \0 terminated. Deal with it properly in
3583 loggings instead of segfaulting when the user want to see the logs :-/
3585 2004-07-22 Martin Quinson
3586 - Fix some stupid bug preventing cbps (callback postit) from working
3588 2004-07-21 Martin Quinson
3589 - Some documentation cleanups
3590 - remove the useless last argument of msgtype_declare
3591 - rename the Virtu functions to fit into the 'os' namespace
3592 - move headers src/include -> src/include/gras/ and stop fooling with
3593 gras -> . symbolic link
3594 - make distcheck is now successful
3596 2004-07-19 Martin Quinson
3598 - Build shared library also
3599 - Install html doc to the right location
3600 - stop removing maintainer files in make clean
3601 - build tests only on make check
3603 2004-07-13 Martin Quinson
3605 - No major issue in previous version => change versioning schema
3606 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3607 corresponding dataset.
3609 2004-07-11 Martin Quinson
3611 - Get it working with any kind of structure (we can compute the padding
3612 bytes remotely for all the architectures I have access to)
3613 - Implement the structure parsing macro (still not quite robust/complete)
3614 - Improvement to the remote testing toysuite
3616 2004-07-10 Martin Quinson
3617 [autoconf mechanism]
3618 - get ride of a bunch of deprecated macros
3619 - actually run the test for two-compliment, not only compile it :-/
3620 - test whether the structures get packed (and bail out if yes. Damn.
3621 Alignment is a serious matter)
3622 - test whether the structures get compacted (but respecting the alignment
3623 constraints of each types)
3624 - test whether the array fields of structures can straddle alignment boundaries
3626 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3627 (took me 2 hours to find that bug, looking at the wrong place)
3628 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3629 before sending/receiving and are used to compute the offsets of fields
3630 - Given that padding size depend even on compiler options, keep track of
3631 alignment and aligned_size only for the current architecture. Not a big
3632 deal since we send structure fields one after the other (seems
3634 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3635 program, let it work on linux/gcc/little32. portability todo.
3637 2004-07-08 Martin Quinson
3638 - import and improve remote compilation support from FAST
3639 - make sure make check works on half a dozen of machines out there
3641 2004-07-07 Martin Quinson
3642 Let's say it's version 0.0.3 ;)
3643 - Implement conversions (yuhu!)
3644 - Let it work on solaris (beside conversion, of course)
3645 - Stupid me, using rand() to generate the conversion datatests in not wise.
3647 2004-07-06 Martin Quinson
3648 - Let make dist work, since I'm gonna need it to compile on remote hosts
3649 - Let Tests/datadesc_usage write the architecture on which the file was
3650 generated as first byte.
3651 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3652 architecture descriptions.
3653 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3654 on those architectures.
3655 - Optimization: send/recv array of scalar in one shoot
3657 2004-07-05 Martin Quinson
3658 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3660 - Plug a whole bunch of memleaks
3661 - each process now have to call gras_{init,exit}. One day, their log
3662 settings will be separated
3663 - Continue the code factorisation between SG, RL and common in Transport.
3665 2004-07-04 Martin Quinson
3667 - Redistribution between SG and RL.
3668 We wanna have to accept in SG, so move accepted related parts of RL in
3669 the common part. (more precisely, the dynar of all known sockets is no
3670 more a static in transport.c, but part of the process_data)
3672 [gras_stub_generator]
3673 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3675 2004-07-03 Martin Quinson
3676 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3677 them all up in one shot)
3679 - Fix the ugly case of reference to dynamic array.
3680 - New (semi-public) function gras_datadesc_size to allow the messaging
3681 layer to malloc the needed space for the buffer.
3683 - gras_socket_close now expect the socket to close (and not its address to
3684 put NULL in it after it). This is because the socket passed to handlers
3685 is one of their argument (=> not writable).
3687 - propagate the interface cleanup from last week in datadesc, ie remove a
3688 superfluous level of indirection. User pass adress of variable
3689 containing data (both when sending and receiving), and not of a variable
3690 being a pointer to the data. Let's say that I like it better ;)
3691 The price for that is constructs like "int msg=*(int*)payload" in
3692 handlers, but it's a fine price, IMHO.
3694 - Let it work in RL (yuhu)
3696 2004-06-21 Martin Quinson
3698 - porting SG plugin and SG select to new standards (works almost).
3699 - plug memleaks and fix bugs around.
3702 - cleanup the prototype of data recv and force users to specify when they
3703 want to handle references to objects. Test case working even for cycles.
3704 - plug memleaks. Valgrind is perfectly ok with this.
3706 2004-06-12 Martin Quinson
3708 - cleanup the separation between plugin and main code in plugin creation
3710 2004-06-11 Martin Quinson
3712 - Reput hook for raw sockets, needed for BW experiments
3713 - kill a few lines of dead code
3714 [Data description] Interface cleanup
3715 - gras_datadesc_by_name returns the searched type or NULL.
3716 That way, no variable is needed to use a type desc once, which makes
3718 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3719 two parameters were strings (field name, type name), leading to
3721 [Dicos] Interface cleanup
3722 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3723 This is consistant with the dynar API.
3725 2004-04-21 Martin Quinson
3727 - Porting to new standards.
3729 - interface cleanup.
3730 There is no bag anymore, no need to take extra provision to mask the
3731 pointers behind "ID".
3732 Better splitup of functions between files create/exchange/convert.
3733 This is still a bit artificial since convert and receive are so
3734 interleaved, but anyway.
3736 - add a queued message list to procdata (the ones not matching criteria
3738 - factorize some more code between SG and RL wrt procdata
3740 - use gras_exit in example to track memleaks
3741 - get rid of gs_example now that GS is properly integrated into gras
3742 - update run_test to integrate the latest tests (datadesc)
3744 - rename WARNINGn macros to WARNn since it prooved error-prone
3746 2004-04-19 Martin Quinson
3748 - register init/exit functions within gras module mechanism
3749 - send/receive function.
3750 Convertion is not implemented, but short-cutted if not needed.
3751 struct/array elements are sent one by one (instead of block-wise), but
3752 nobody really cares (yet). Get a prototype before optimizing.
3753 - tests (using a file socket) for DD send/receive on:
3754 - base types: int, float
3755 - array: fixed size, string (ie ref to dynamic string)
3756 - structure: homogeneous, heterogeneous
3757 - chained list, graph with cycle
3758 Believe it or not, valgrind is not too unhappy with the results. The
3759 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3763 - Bugfix when using a filename explicitely (instead of '-')
3765 2004-04-09 Martin Quinson
3767 - factorize more code between RL and SG in socket creation
3768 - Complete the implementation and tests of:
3770 o file (only in RL, and mainly for debugging)
3772 I lost 3 days to design a portable address resolver, and then decided
3773 that the prototype mainly have to run on my box.
3774 Addressing portability too early may be like optimizing too early :-/
3776 - use gras_init in the Tests instead of the crappy parse_log_opt
3777 (the latter function is removed)
3778 [Conditional execution]
3779 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3780 [Code reorganisation]
3781 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3782 Build examples against the RL library, since there is no way to disable
3783 its creation for now.
3785 For information, the beginning of coding on GRAS was back in june
3786 2003. I guess that every line has been rewritten at least twice since