1 SimGrid (3.22) NOT RELEASED (Release Target: December 21. 2018, 22:23 UTC)
4 - Process termination which was broken at version 3.21 has been repaired.
5 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
6 - Hide the examples into examples/deprecated. New users should use Python.
9 - Replace our own code to display a backtrace (that was forking addr2line)
10 with the Boost.Stacktrace library.
11 You won't see your backtraces without this optional dependency.
12 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
15 - Drop MSG_process_create_from_stdfunc() from the API.
16 This C++ function was a pimple in the C API, made necessary at some
17 point by the Java bindings. This is fixed now.
18 - Hide the examples into examples/deprecated. New users should use S4U.
21 - Drop sg_cmdline. Please use xbt_cmdline instead.
22 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
23 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
24 OsSemaphore is implemented in a portable way with C++11 threads.
25 This should allow much more threads to be created at the same time,
26 allowing Mac and Java users to simulate many more actors.
27 - Implement the 'thread' factory with std::thread instead of xbt ones.
28 It is not possible to set the stack size with threads anymore, but
29 -fsplit-stack is the way to go nowadays when using threads.
30 - Drop the xbt_os_thread_t module (now unused)
31 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
32 - Drop cunit, use Catch2 instead.
35 - In <host> and <peer>, availability_file is now speed_file.
36 XML file version remains 4.2 since old files are still compatible.
39 - #261: Document the parameters of parallel execution's constructor
40 - #300: [s4u] BarrierPtr is missing
41 - #314: SMPI args internal cleanup
42 - #316: Fix a bug related to the CPU utilization of multi-core VM
43 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
45 ----------------------------------------------------------------------------
47 SimGrid (3.21) October 3. 2018
49 The Restarting Documentation (TRD) Release.
52 - Convert the user manual to Sphinx for improved usability.
53 Unstable version now hosted on https://simgrid.frama.io/simgrid
54 Many glitches remain; stay tuned.
55 - Restrict the Doxygen reference API to the installed header files.
56 (documenting all internals this way is overkill + doxygen scales badly)
57 - New tutorial on S4U (highly inspired from the old MSG one, but with
58 a git repository to fork as a starting point, and a docker image)
59 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
60 - The Developper manual is still to be converted (not compiled until then)
61 - Some parts are still missing in this conversion (such as the
62 platform and deployment sections) while others were blindly converted
63 and would need more love (such as the configuration flags).
64 Things will certainly further improve in the future releases.
67 - Fully reimplement auto_restart mechanism that was utterly broken
68 (fix #22, #131 and #281 that were all very related)
69 - Implement semaphores (fix #303 and #291)
70 - When creating an actor from a function and its parameters,
71 move-only parameters are not allowed anymore, as it would prevent
72 the actor to be restartable if its parameters are consumed on run.
73 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
74 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
75 IO operations that can be started, waited for, or canceled as a regular
77 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
81 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
82 - New host variable: 'core_count'
85 - Remove xbt_os_thread_specific features
86 - Remove portability wrapper to condition variables
87 - Remove xbt_os_thread_yield()
90 - MPICH collective selector now mimics MPICH 3.3b
91 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
92 - MPI_Init(NULL,NULL) is now allowed
93 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
94 - Command-line arguments should now be handled properly in Fortran simulations
95 - MPI Attributes and related callbacks should work in Fortran
96 - Apps using C + Fortran should now work
97 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
98 - Support MPI_Error_String
101 - Due to an internal bug, Msg.run() must now be your last line.
102 We hope to fix it in a future release, and we are sorry for the inconvenience.
105 - #22: Process autorestart seem to only work with CAS01 cpus
106 - #56: Feature request: dynamic SMPI replay
107 - #93: simgrid should not eat --help
108 - #111: How to change the loopback link characteristics?
109 - #116: Platform cluster radical
110 - #117: Platform cluster/host order
111 - #129: Replay_multiple on steroid
112 - #131: Java : autostart does not work
113 - #143: Setting a breakpoint at a given time
114 - #235: Network energy models should be integrated
115 - #264: Add ptask L07 resource tracing
116 - #271: Dynamic and manual replay of SMPI traces
117 - #279: Breakpoints option uses time deltas instead of absolute time
118 - #281: Daemonized actor no longer daemonized after an auto-restart
119 - #282: TIT convertor erase traces when using absolute path in the trace list file
120 - #285: segfault when a s4u actor kills itself with msg process tracing activated
121 - #286: Pajé traces are not exposing the number of cores
122 - #287: Command-line parsing should stop when encountering '--'
123 - #288: MPI_Init(NULL, NULL)
124 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
125 - #290: Method to check if a mailbox has a message ready to be consumed
126 - #291: [s4u] semaphores do not seem available yet
127 - #294: Alltoall collective implementation of mpich
128 - #295: Starting a Host as OFF is not supported
129 - #296: DTD too permissive
130 - #299: Add a s4u barrier example
132 ----------------------------------------------------------------------------
134 SimGrid (3.20) Released June 24. 2018
136 The proxy snake_case() release.
139 - The whole API is now using snake_case() instead of mixing it with camelCase.
140 Compatibility wrappers in place for one year.
141 - We took the opportunity to increase the consistency of the API.
142 Please report any remaining glitches.
143 - New function simgrid::s4u::Host::get_actor_count:
144 Returns the number of actors running on a specific host.
147 - SMPI is now tested with ~45 proxy apps from various sources, with none or
148 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
149 - Replay: The replay file has been re-written in C++.
150 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
151 supported. They are stored in the trace and used when replayed.
152 - Basic support of MPI_Cancel. Robustness not guaranteed.
153 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
154 (only for MPI_COMM_TYPE_SHARED).
155 - New option: smpi/privatize-libs, to add external shared libs to be privatized
156 by SMPI. They will be copied locally and loaded separately by each process.
157 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
158 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
159 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
160 with internal index optind with multiple processes. Only works if MPI_Init has
162 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
163 syntax without preprocessing tricks.
166 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
167 behavior tracing is no longer limited to MSG
168 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
169 this means that traces now need to be updated or re-obtained; both Irecv
170 and Isend lines in your traces have now in total 5 mandatory fields:
171 <rankid> <command> <to/from rankid> <tag> <size>
172 To update your traces, it suffices to add a 0 for the tag here.
173 - TIT now also supports waiting for a distinct request via MPI_Wait.
174 Wait/Test now wait for a specific request, not just the last one that was
175 issued. This unfortunately means another update, because we need to
176 identify which request you want to wait for. We do this via the
177 triplet (sender, receiver, tag), which needs to be added:
178 <rankid> <command> <sender> <receiver> <tag>
179 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
180 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
183 - The deprecation of MSG is ongoing (but this should not impact you).
184 Many MSG functions are now simple wrappers on the C API of S4U. If
185 you wish to convert your code to S4U, find the S4U counterparts of
186 your MSG calls in src/msg/msg_legacy.cpp.
187 - MSG can still be used, but won't evolve anymore.
190 - Allow to run the Link energy plugin from the command line with
191 --cfg=plugin:link_energy
192 - Rename Energy plugin into host_energy
193 - Rename Load plugin into host_load
196 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
197 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
198 - Improve performance when handling timeouts of simix synchros.
201 - Config: the C API is now deprecated (will be removed in 3.23), and
202 the C++ API has been slightly improved.
205 - Fix several build issues on OSX.
206 - Move simgrid_config.h to simgrid/config.h (old header still working)
209 - #143: Setting a breakpoint at a given time
210 - #150: Inconsistent event names in SMPI replay
211 - #258: daemonized actors hang after all non-daemonized actors have completed
212 - #267: Linker error on unit_tmgr
213 - #269: SMPI: tracing of MPI_Wait/all/any broken
214 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
216 ----------------------------------------------------------------------------
218 SimGrid (3.19.1) Released March 22 2018
220 The Fixed ABI Release.
222 As you may know, we are currently refactoring SimGrid in deep.
223 Upcoming SimGrid4 will be really different from SimGrid3: modular,
224 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
226 Our goal is to smooth this transition, with backward compatibility
227 and automatic update paths, while still progressing toward SimGrid4.
229 SimGrid remains open during works: The last pure SimGrid3 release was
230 v3.12 while all subsequent versions are usable alpha versions of
231 SimGrid4: Existing interfaces remain unchanged, but the new S4U
232 interface is budding and the internals are deeply reorganized.
234 Since 2015, we work hard to reduce the changes to public APIs. When
235 we need to rename a public library symbol in S4U, we let your
236 compiler issue an explicative warning when you use the deprecated
237 function. These messages remain for four releases, i.e. for one full
238 year, before turning into an error. Starting with v3.15, your can
239 also adapt to API changes with the SIMGRID_VERSION macro, that is
240 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
242 Starting with this v3.19.1, our commitment to reduce the changes to
243 the public interfaces is extended from the API to the ABI: a program
244 using only MSG or SimDag and compiled against a given version of
245 simgrid can probably be used with a later version of SimGrid without
246 recompilation. We will do our best... but don't expect too much of
247 it, that's a really difficult goal during such profund refactoring.
249 The difference between v3.19 and v3.19.1 is that the former was
250 accidentally breaking the ABI of MSG, while the later is restoring
253 S4U and kernel APIs will still evolve until SimGrid4, with one-year
254 deprecation warnings as currently. In fact, cleaning up these
255 interfaces and converting them to snake_case() is one release goal of
256 v3.20. But don't worry, we are working to smooth this upgrade path.
258 In summary, new projects should start with S4U to benefit of the
259 future, but old MSG projects should still be usable with no change.
261 ----------------------------------------------------------------------------
263 SimGrid (3.19) Released March 20 2018
265 The Moscovitly-cold Spring release.
267 SMPI user-visible improvements:
268 - Support of MPI_Type_create_subarray
269 - Unimplemented MPI calls now abort (instead of issuing a warning)
270 Some harmless calls around errhandlers still only issue a warning.
271 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
272 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
273 is now possible (mmap privatization was refactored to this end).
274 Adding new SMPI instances should also be technically doable now.
276 SMPI internal cleanups:
277 - Removed index notion from SMPI -- no more "getPid() - 1"!
278 In tracing files, actors are called rank-<PROCESS-ID> but these are
279 NOT the real ranks (tracing several communicators is not supported yet).
280 - Deprecate smpi_process_index() (will be removed in 3.22)
281 - Better testing on classical proxy apps, through an external project:
282 https://github.com/simgrid/SMPI-proxy-apps/
285 - Use a graphical TOC to make it easier to find the documentation you need
286 - Revamp the MSG tutorial
289 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
290 - Remove parameter of MSG_process_killall().
291 Resetting the PID was bogus anyway (several actors could have the same PID).
294 - Execution->setHost() can be called after start(), to migrate it.
295 - Comm::test_any() is now implemented.
296 - s4u::Actor now has onCreation() and onDestruction() signals.
297 - Install some kernel header files for the users' plugins and more.
298 Warning, their API is really not stable yet.
299 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
300 include/simgrid/kernel/routing/*Zone.hpp
303 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
305 Build System and other cleanups:
306 - Remove unused run-time parameter "tracing/onelink-only".
307 - SimGrid now works with intel compilers (MC still buggy)
308 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
309 - tesh kills subprocesses on timeout (but not on windows)
312 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
315 - #194: Feature request: simgrid::s4u::Comm::test_any()
316 - #245: migrating an actor does not migrate its execution
317 - #253: Feature Request: expose clusters as objects
318 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
319 - #255: Tesh broken on Windows
320 - #256: Modernize FindSimGrid.cmake
321 - #257: Fix (ab)use of CMake install
323 ----------------------------------------------------------------------------
325 SimGrid (3.18) Released December 24 2017
327 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
329 MSG IS NOW DEPRECATED. Please use S4U for new projects.
330 - Support for MSG should not be removed from SimGrid before 2020,
331 but future evolutions will be limited to the integration of
332 user-provided patches.
335 - s4u::onDeadlock() signal, run before stopping the simulation.
336 - s4u::Exec: asynchronous executions (abstraction of background
338 - s4u::Host->getLoad() returns the achieved speed in flops/s
341 - Link::name() is deprecated and will be removed in v3.21, use
342 Link::getCname() instead.
343 - Mailbox::getName() changed to return a std::string, use
344 Mailbox::getCname() instead to get a char*.
345 - Storage::getName() changed to return a std::string, use
346 Storage::getCname() instead to get a char*.
347 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
348 to have both allocation and deallocation of the map in user space.
351 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
352 The amount of flops remaining to do is a vector, not a scalar.
353 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
354 name implies on both sequential and parallel tasks.
355 - Both changes fix GitHub's #223 using PR #237 as a basis.
356 Thanks Michael Mercier.
357 - Most examples were converted to S4U and hidden elsewhere at tests.
358 Rationale: we still want MSG to work; we want newcomers to use S4U.
361 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
362 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
365 - Switch to the faster dlopen privatization mechanism by default
366 - Documentation improvments
369 - Live migration is getting moved to a plugin. Dirty page tracking is
370 the first part of this plugin. This imply that VM migration is now
371 only possible if one this function is called:
372 - C/MSG: MSG_vm_live_migration_plugin_init()
373 - C/C++: sg_vm_live_migration_plugin_init()
374 - Java: Msg.liveMigrationInit()
375 For C and C++, "simgrid/plugins/live_migration.h" also has to be
379 - Define class simgrid::xbt::Path to manage file names.
380 - Removed unused functions:
381 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
382 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
383 xbt_graph_export_graphviz()
384 - xbt/str.h: xbt_str_join()
385 - Remove unused datatypes:
386 - xbt/heap.h: use std::priority_queue or boost::heap instead
387 - xbt/swag.h: use boost::intrusive::list instead
390 - New link_energy plugin for the consumption of the links.
391 - All of the operations on files and storage contents have been
392 packaged into a plugin (src/plugins/file_system). The current
393 public interface can be found in
394 include/simgrid/plugins/file_system.h
395 To use these functions you now have to initialize the plugin by
396 calling MSG_storage_file_system_init() just after calling
397 MSG_init() or sg_storage_file_system_init() just after creating
401 - Remove the undocumented/untested tag <include>
404 - Remove viva specific tracing as the tool is no longer maintained
407 - #248: Kill of finished processes leads to segfault
408 - #240: xbt_cond_wait_timeout should gracefully return for C apps
409 - #239: please implement signal s4u::onDeadlock()
410 - #230: segfaults when exit() without run()
411 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
412 - #223: MSG_task_get_flops_amount() not working with parallel tasks
413 - #222: Actor::kill() doesn't really kill and segfaults
414 - #221: odd LMM warning when killing an actor
415 - #120: Memory leak when the processes are forcefully killed
417 ----------------------------------------------------------------------------
419 SimGrid (3.17) Released October 8 2017
421 The Drained Leaks release: (almost) no known leaks despite the tests.
423 Even very long simulations will preserve your memory: our testsuite
424 stresses SimGrid for over 45mn on fast machines for a coverage of
425 over 80%, and there is only one single known leak, of about 4kb.
429 - Comm.detach(): start and forget about asynchronous emission. A cleanup
430 handler may be given to free resources if the comm cannot be completed.
431 - this_actor::send(mailbox) is now mailbox->put()
432 - New: simgrid::s4u::Comm::wait_all()
433 - New: Host.execute() for remote executions.
436 - Unused option network/sender-gap is removed.
439 - Deprecate MSG_task_isend_with_matching(): This unused feature
440 really complicates our internals. Will be removed in v3.20.
443 - Improved context termination. It is now possible to free resources, even
444 when a process is forcibly killed.
445 - ContextBoost: add support for Boost versions above 1.61.
448 - Bring back run-time option --cfg=exception/cutpath to remove exception
450 - Removed unused functions:
451 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
452 xbt_str_rtrim(), xbt_str_trim().
453 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
456 - Removed header files obsolete since SimGrid 3.12:
457 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
459 - #3: SD_exit should be made optional
460 - #120: Memory leak when the processes are forcefully killed
461 - #159: Threading test regression in Actor refcounting
462 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
463 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
464 - #186: Actor::killAll() segfaults if some process is blocked on wait()
465 - #191: VM migration and pstate
466 - #192: Updating the energy consumptions of all hosts crashes with VMs
467 - #195: All actors have PID=0 in the logs
468 - #204: Sometimes segfault with thread contexts and mmap privatization
469 - #222: Actor::kill() doesn't really kill and segfaults
470 - #225: Actor::kill() doesn't really kill when victims are doing a join()
472 ----------------------------------------------------------------------------
474 SimGrid (3.16) Released June 22. 2017.
476 The Blooming Spring Release: developments are budding.
479 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
480 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
481 - Java JAR file should be finally fully working on Mac OSX too.
483 XML platforms: Switch to platform v4.1 format.
484 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
485 - <zone> can be used as a synonym for the now deprecated <as>
486 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
487 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
488 - <actor> can be used as a synonym for the now deprecated <process>
489 - state_file and avail_file periodicity is now easier to express
490 (check the documentation) the old behavior should still work.
491 * Storage: not backward compatible, but it seems that nobody ever used it
492 - Remove attribute 'content_type' of <storage_type>: was never used
493 - Make attribute 'model' of <storage_type> optional: for future usage
494 - Remove Bconnection model property: was never (in)validated. Replaced by
495 the maximum of the read and write bandwidth as a resource constraint.
498 - New and Backwards Compatibility break:
499 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
500 When one wants to get the list of tasks whose states have changed during a
501 simulation round, s/he has to allocate and free a dynar and use it as
502 argument to this function. The former SD_simulate (double how_long)
506 - Allow multicore VMs, with the correct sharing (unless you overcommit)
507 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
510 - New function to update the consumption of all hosts at once.
511 - Fix the model for multi-core hosts, linear in the amount of busy
512 cores with an abnormality for pIdle (see paper).
513 BUG: Still not satisfactory for ptask on mono-cores.
516 - The netzone are now available from the MSG API.
517 The old names still work, but are now deprecated.
520 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
521 - New option: smpi/keep-temps to not cleanup temp files
522 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
523 mallocs mode is used, allows to change the size of the fake file used
524 (default 1MB), to potentially limit the number of mappings for large runs.
525 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
526 - Fortran ifort and flang compilers support
527 - New RMA calls supported (experimental):
528 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
529 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
530 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
531 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
532 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
533 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
536 - New: this_actor::isMaestro() returns whether we are in kernel mode.
537 - Behind the scene, ActivityImpl are now automatically refcounted.
538 Removing this nasty bottleneck will greatly simplify our internals.
541 - Replay: New function xbt_replay_action_get():
542 Retrieve the function previously associated to an event type.
543 - DROPPED FUNCTION: xbt_str_varsubst()
544 - DROPPED MODULE: strbuff. We don't need it anymore.
545 - DROPPED MODULE: matrix. We don't need it anymore.
546 - DROPPED MODULE: lib. We don't need it anymore.
548 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
550 SimGrid (3.15) Released March 22. 2017
552 The Spring Release: continuous integration servers become green
553 We fixed even the transient bugs on all target architectures:
554 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
555 Windows, FreeBSD, NetBSD.
557 - Rename NetCards to NetPoints.
558 This was intended to help NS3 users, but that's not a netcard.
559 That's a point in the routing algorithm, let's avoid wrong simplifications.
562 - New: MSG_process_yield(). Stop and yield to other processes.
563 - New: MSG_process_daemon(). Daemon processes are automatically killed
564 when the last non-daemon process terminates
565 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
566 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
567 to make room for the upcoming network energy plugin.
568 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
571 - Ensure that an actor can kill itself with Process::exit()
572 - Kill the obscure NativeException. Nobody want to survive the issues
573 it denotes, so use JniException that is a RuntimeException (not to
574 be caught explicitly).
575 - Partial bug fix in initialization. SimGrid flags on command line were
576 consumed at C level but stayed in the original Java String[] args.
577 This could mess users' args[i] if SG flags were not put at the end of
579 The SimGrid flags are now removed from the Java arguments. However,
580 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
581 if args.length is greater than the number of YOUR OWN ARGUMENTS.
582 It might be if you have --log or --cfg flags in the command line.
583 - Fix numerous memleaks all around the place. In particular, around VMs.
587 - s4u::onPlatformCreated: right before the simulation starts
588 - s4u::onSimulationEnd: right after the main simulation loop
589 - s4u::onTimeAdvance: right after a clock change (time is discrete)
590 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
591 event from the availability_file changes the avail speed.
592 - Links are now usable from s4u
593 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
594 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
595 - New examples: The conversion of MSG examples to S4U has begun
596 - Actors: create, kill, migration, and suspend
597 - Applications: master-worker and token-ring
598 - Action replay: communications and storage
599 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
602 - Backwards Compatibility breaks
603 - The SD_route_get_list and SD_route_get_size functions have been removed.
604 They are replaced by the unique
605 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
606 The route from src to dst is built in the links dynar whose size can
607 be obtained with xbt_dynar_length.
608 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
609 sg_host_route_bandwidth and sg_host_route_latency.
610 Macros ensure the backwards compatibility, but you should fix your code
613 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
614 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
615 - Bump our claim of support from MPI 1.1 to MPI 2.2.
616 We don't support 100% of it, but it should be enough. Contact us if not.
617 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
618 naive and probably inaccurate)
619 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
620 mapped onto a unique small file using some system magic.
621 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
622 - Fortran: cleanups, fixes, support of user-added operations
623 - RMA: MPI_Accumulate are applied in correct order.
624 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
625 - Better support for MPI_IN_PLACE
626 - Support for MPI_Win attrs and keyvals.
627 - Support MPI_Comm_group_create, MPI_Type_size_x
628 - MPI ops should be properly applied to all allowed types
629 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
630 algorithm by a non failing one from mpich.
633 - Kill the fifo data container: we don't use it anymore.
636 - A <cluster> can now be created with different speed values to represent pstates.
637 see examples/platforms/energy_cluster.xml for an example.
639 SimGrid (3.14.159) Released December 28. 2016
641 The previous release was only a crude approximation of a Pi release;
642 this one should be a bit better.
644 - Revert a last minute change that broke on Mac OSX.
645 - Fix the changelog and NEWS files.
646 - Improve the documentation, in particular of the routing module.
647 - Integrate some patches flying around in Debian.
649 SimGrid (3.14) Released December 24. 2016
652 * The whole documentation was reworked and reorganized. There is
653 still much room for improvement here, but we're on it.
656 * We now need python3 for our testing infrastructure.
657 * Model checking needs libevent
660 * Now works on FreeBSD
663 * Privatization now works on FreeBSD
665 * Privatization is now activated by default in smpirun
666 You can turn it off with -no-privatize if something goes wrong.
668 * Call-location tracing for SMPI.
669 You can add the exact location (filename / linenumber) of an MPI call to
670 your trace files and slow down or speed up the simulation between two
671 consecutive calls by using an adjustment file (see the documentation).
673 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
674 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
675 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
676 values were ignored in some cases. The timings of these functions can now
677 be significantly different.
679 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
680 smpi/running-power is renamed to smpi/host-speed
682 * smpi/grow-injected-times option to enable or disable multiplication of the
683 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
684 make simulation less precise (but also much faster).
686 * smpirun script should be (much) faster for large deployments.
688 * SMPI tracing: fixed issue with poor matching of send/receives.
690 * Replay: Fix broken waitall
692 New functions and features
693 * MSG_parallel_task_execute_with_timeout, to timeout computations.
695 Dropped / renamed functions and features
696 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
697 - MSG_mailbox_is_empty() -> Mailbox::empty()
698 - MSG_mailbox_front() -> Mailbox::front()
699 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
700 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
701 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
702 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
703 * Don't pass the free_f parameter to property related functions:
704 - MSG_host_set_property_value()
705 - MSG_as_router_set_property_value()
706 - MSG_storage_set_property_value()
707 * VM properties. Since msg_vm_t are msg_host_t, just use
708 MSG_host_get_property_value() and friends
709 * VM I/O related things:
710 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
711 - Unimplemented save/restore methods
712 * MSG_as_router_get_property_value() was redundent with
713 MSG_environment_as_get_property_value().
714 - Removed MSG_as_router_*propert*() functions
715 - Added MSG_environment_as_set_property_value() for consistency
716 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
717 Well, they are still there for now, but deprecated with a warning.
718 Please switch to xbt_dict_new_homogeneous() before this is removed
720 * Task affinity. Its intended behavior (that was very badly tested
721 and probably not really working) was deceiving what most users
722 would have hoped here.
723 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
724 * Option network/coordinates is now useless and should be dropped.
727 * Added option storage/max_file_descriptors to allow more than 1024 files opened
730 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
731 * Added Dragonfly topology support
734 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
735 EiBps, EBps, Eibps, Ebps for bandwidth.
736 They may become useful to some lucky ones.
739 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
740 * ex/app_tokenring: new example, very similar to the MSG Token Ring
741 * ex/async_waitAll: new example, on asynchronous communications
744 * Memory usage should be decreased for simulations with a large number
745 of processes. This also helps for SMPI.
747 SimGrid (3.13) stable; urgency=low
749 The Half Release, a.k.a. the Zealous Easter Trim.
751 - We removed half of the lines, that were mostly experimental cruft.
752 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
753 (not counting blanks and comments -- according to openhub.net)
754 - The internals are now compiled in C++ (and will soon be clean C++)
755 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
756 59 klines of C, etc. We added only 29 klines of C++ in replacement.
758 * Backwards Compatibility breaks
759 - Removed Lua simulation bindings (switch to C or Java for that).
760 Lua can still be used to describe platforms
761 - Removed Java kernel plug-ins.
762 Will be reintroduced after the ongoing major internals reorg.
764 - the following functions were removed.
765 They were too specific and should be reimplemented in a generic
766 way, with filter function.
767 - MSG_task_listen_from_host
768 - MSG_mailbox_get_count_host_waiting_tasks
769 - MSG_mailbox_put_with_timeout was removed.
770 Please use MSG_task_send_with_timeout instead.
772 - the SD_application_reinit function was removed. It has been a noop for a while.
773 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
774 replaced by a more flexible mechanism.
775 - The following functions thus do not exist anymore
776 - SD_workstation_get_access_mode
777 - SD_workstation_set_access_mode
778 - SD_workstation_get_current_task
779 - Basic estimation functions have been removed but can easily be replaced
780 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
781 - SD_workstation_get_computation_time => amount / sg_host_speed()
783 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
784 Use VM.getSpeed()*load/100 for the legacy behavior.
786 - option enable_tracing was removed. It was not doing anything for a while.
787 - In the ModelChecker:
788 - the model-checker now ptraces the model-checked process which means
789 you cannot use a debugger on the latter anymore (we might make this
790 optional in the feature);
791 - removed soft-dirty page tracking;
792 - remove model-checked side snapshot management,
793 MC_snapshot() and MC_compare_snapshot();
794 - keep the MC_cut() function as a stub (it was not really working
795 in the previous release).
798 * All options are consistently in kebab-case. Old names are kept as alias.
801 * Switch to platform v4 format.
802 - Rename from 'power' to 'speed' the attributes describing the amount of
803 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
804 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
805 - In <host> and <link>, attributes availability and state are gone.
806 It was redundent with state and availability traces, and with peak values.
807 - In <cluster>, attributes availability_file and state_file are gone.
808 It was too complex and unused.
809 - Kill <gpu>. Was not doing anything.
810 - The DOCTYPE points to the right URL:
811 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
812 (the file at this address now documents the changelog since its v1)
813 - A warning is emitted for unit-less values (they are still accepted).
814 - speed. Default: 'f' or 'flops'. Also defined:
815 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
816 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
817 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
818 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
819 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
820 - latency. Default: 's' second. Also defined:
821 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
823 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
824 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
827 * s4u::Host is now the preferred public interface to the Host features.
828 sg_host_* functions are C bindings to the exact same behavior
829 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
832 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
833 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
834 They were deprecated since a few years, and probably did not even compile anymore.
837 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
838 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
839 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
840 encouraged to update their codes. A list of the modified functions can be found at the end of
841 include/simgrid/simdag.h
844 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
845 the lambda or closure passed as a parameter will run in kernel mode.
846 All the callback functions should be rewritten to that interface at some point.
849 * Reorganizing and cleaning the internals all around the place.
852 * Remove old default barrier/bcast buggy algorithms (see #18407)
853 * Various bug fixes to handle more codes
854 * Remove the need for the --foreground option of smpirun (it is still
855 accepted for backward compatibility).
858 * Kill the setset data container: MC don't use it anymore.
859 * Kill the queue data container: it made more sense with GRAS.
860 * Kill the xbt_peer_t data type: it's useless without GRAS.
861 * Kill rm_cb feature of config sets: it was never useful.
862 * Kill graphxml parsing feature. It was not used.
863 * Kill the deprecated code protected by XBT_USE_DEPRECATED
865 - xbt_dynar_sort_strings(), when the content is char*
866 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
867 They throw exceptions on invalid input;
868 * C++ support for declaring CLI flags (simgrid::config::Flag);
869 * class for abstracting different signal backends (simgrid::xbt::signal).
870 with no external dependencies (we need very simple signals).
873 * refactoring and cleanup of the code;
874 * ongoing process to cleanly separate the model-checking algorithms
875 from the code model-checking support.
877 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
879 SimGrid (3.12) stable; urgency=low
881 The Facelift Release.
884 * Require g++ v4.7 at least to not speak prehistorical C++.
885 * Require Boost 1.48 (for signal2 component).
886 * Java must be version 7 at least when activated.
887 * Builds on Windows again (including Java bindings).
888 * Tracing is now always enabled (no way to turn it out)
889 * Remove GTNetS. It was not working anyway.
890 * Various cleanups in the cmake scripts.
891 * Move headers around to sort them out on installed systems:
892 - instr/instr.h -> simgrid/instr.h
893 - instr/jedule/* -> simgrid/jedule
894 - simdag/datatypes.h was removed
895 - simdag/simdag.h -> simgrid/simdag.h
896 - msg/datatypes.h was removed
897 - msg/msg.h -> simgrid/msg.h
900 * Interface improvement:
901 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
902 - Sanitize the interface in MSG_task_ module:
903 - Merge two functions that were close enough but misleading:
904 set_compute_duration(t) -> set_flops_amount(t)
905 get_remaining_computation(t) -> get_flops_amount(t)
906 - set_data_size(t) -> set_bytes_amount(t)
907 get_data_size(t) -> get_bytes_amount(t)
908 - Massive cleanups in the functions related to the energy
909 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
910 - New: MSG_host_get_pstate()
912 - msg/energy/onoff: switching hosts on and off
914 * Interface improvement:
915 - Rename Host.isAvail() to Host.isOn()
916 - Rename Process.currentProcess() to Process.getCurrentProcess()
917 - Rename Task.setDataSize() to Task.setBytesAmount()
918 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
920 - #18874: Actually allows the GC to reclaim tasks
924 - SIMIX_process_throw: raises an exception in a remote process
925 * Refactoring: Separate sub-modules
926 - libsmx: the public interface, as libc in a real system
927 - popping: the strange dance that converts a user request into a kernel handling
928 - smx_context_*: the virtualization mechanisms that embed the user code
929 - smx_*: the handling of each simcalls
931 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
932 * Rename smx_action_t into smx_synchro_t, making explicit that these
933 things are used to synchronize processes with their environment.
934 For example, a communication is a sort of synchronization involving
935 the communicating processes (that may block until the exchange) and
936 the platform. The same can be said from computations, etc.
938 - #18888: segfault when a process ends before its kill_time
942 - 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.
943 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
944 - MPI_Type_set_name, MPI_Type_get_name
945 - MPI_*_c2f and MPI_*_f2c functions
946 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
947 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
948 - Activate a lot of new tests from the mpich 3 testsuite
950 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
951 - InfiniBand network model added: Based on the works of Jerome Vienne
952 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
953 - When smpi/display_timing is set, also display global simulation time and application times
954 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
955 * Collective communications
956 - 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
957 - 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/).
958 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
960 - Replay now uses algorithms from wanted collective selector
961 - Replay can be used with SMP-aware algorithms
962 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
963 - 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.
965 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
966 - Correctly use loopback on fat-tree clusters
967 - Asynchronous small messages shouldn't trigger deadlocks anymore
968 * Energy/DVFS cleanup and improvement
969 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
970 - new: smpi_get_host_pstate()
973 - "Full" network optimization flag was broken since Surf++
974 - Better handling of precision flags in maxmin
975 - Fix bug causing sometimes "Impossible" errors
976 - Properly pass cluster properties to included hosts
977 * Improvement of the Energy plugin.
978 - Always update the consumption before returning that value
979 - New property: watt_off to denote the dissipation when the host is off
980 - New functions getWattMinAt and getWattMaxAt to retrieve the
981 dissipation of pstates that we are not currently at.
982 * Java: class NetworkLink renamed to Link
983 * New function: simcall_process_get_kill_time()
984 * Massive rename s/workstation/host/
985 - That's intrusive, but that's good for the project consistency. Sorry.
986 - Change config option "workstation/model" into "host/model"
989 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
990 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
991 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
992 - xbt_dirname and xbt_basename for non-POSIX systems
994 * The model checker now runs as a separate process.
995 * The model checker runs must now be launched with the new simgrid-mc program.
996 * Record/Replay: the MC can display a textual representation of a path in the
997 execution graph. It can then be replayed outside of the model checker.
999 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1001 SimGrid (3.11) stable; urgency=low
1006 * Normalizing pointers addresses tool for better diff between logs
1009 * Add cloud examples using new VMs
1010 - examples/msg/cloud/two_tasks_vm.tesh
1011 - examples/msg/cloud/simple_vm.tesh
1012 - examples/java/cloud/cloud.tesh
1013 - examples/java/cloud/migration/migration.tesh
1014 * Add java surf examples:
1015 - examples/java/surfPlugin/surf_plugin.tesh
1016 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1017 - examples/java/surfCpuModel/surf_cpu_model.tesh
1018 * Add SMPI+MSG example:
1019 - examples/smpi/smpi_msg_masterslave/
1024 - msg task destroy cancel test
1025 - msg_host on/off test
1026 * Move all tests in testsuite to teshsuite (adding tesh files)
1027 * Restructure teshsuites
1028 - one folder for each kind of test
1029 * Restructure AddTests.cmake
1031 - structure the order of tests (with sections)
1034 * Add virtual machine
1035 - creation of a VM on a PM
1036 - migration of a VM from a PM to another PM
1038 - MSG_process_join(msg_process_t process, double timeout)
1039 - msg_bar_t MSG_barrier_init(unsigned int count)
1040 - int MSG_barrier_wait(msg_bar_t barrier)
1041 - void MSG_barrier_destroy(msg_bar_t barrier)
1042 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1043 * New option "msg/debug_multiple_use" to help debugging when a task is used
1046 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1048 - Increase code coverage in test suites
1052 * Protect context stacks against stack overflow. The number of protected memory
1053 pages allocated on the top of each stack (1 by default) can be configured
1054 with the new command line option --cfg=contexts/guard_size:<value>.
1055 * Simcalls are now generated by a python script that
1056 - generates files included by SimGrid
1057 - checks that all the functions exist, and proposes prototypes
1059 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1060 host_set_data, host_get_data
1062 - simcall_process_join(smx_process_t process, double timeout)
1063 * Fix bug where sleeping processing could not be suspended.
1066 * Translate surf models from C to C++
1067 - Generic classes for all models: Model, Resource, Action
1068 - A generic interface for each kind of model (CPU, Network, Storage
1069 Workstation, WorkstationVM)
1071 * Translate surf routings from C to C++
1072 * Add callbacks using sigc++ or boost::signals2
1073 - Add callback functions for resource creation/destruction
1074 - Add callback functions for action state change
1075 - Handle Energy as a plugin
1076 * Replace swag by boost::intrusive
1077 * Add new routing models for clusters. For documentation, see
1078 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1079 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1080 parameters for cluster tag
1081 - Fat trees, with topology="FAT_TREE" and
1082 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1083 - see examples/platforms/torus_cluster.xml and
1084 examples/platforms/fat_tree_cluster.xml
1085 * More documentation
1088 * Hostfiles support host:nb_processes construct to deploy several processes on
1090 * Collective communication algorithms should not crash if used with
1091 improper number of nodes and report the error.
1092 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1093 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1094 MPI_Dims_create, MPI_Cart_sub are supported.
1095 * New interface to use SMPI programmatically (still depends on MSG for
1096 some parts, see examples/smpi/smpi_msg_masterslave):
1097 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1101 * Global variables privatization in MPI executables is now performed at runtime
1102 with the option smpi/privatize_global_variables (default:no).
1103 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1104 dynamic libraries loaded by the application are not privatized (static
1105 linking with these libraries is advised in this case)
1108 * Options defined in XML work correctly now.
1111 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1112 copied into simgrid.jar or not (ON by default). Use this option if you want
1113 to reduce the size of the installed simgrid.jar, *and* the native libraries
1114 are kept installed elsewhere.
1115 * Surf binding with SWIG (code generated in maintainer mode only):
1116 - plugin to handle callbacks
1117 - CPU model only for the moment
1120 * Supernovae build mode is definitively removed. It was used to improve
1121 inlining and inter-module optimizations. It is nowadays superseded by
1122 link-time optimizations commonly available in compilers.
1123 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1124 * Add boost dependency for surf++
1125 * Add new macro for tests
1126 - ADD_TESH(name <tesh_args>)
1127 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1131 - xbt_bar_t XBT_barrier_init(unsigned int count)
1132 - int XBT_barrier_wait(xbt_bar_t barrier)
1133 - void XBT_barrier_destroy(xbt_bar_t barrier)
1134 * Make the xbt_os_time module public
1136 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1138 SimGrid (3.10) stable; urgency=low
1140 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1143 * Reintegrate Java to the main archive as desynchronizing these
1144 package is not acceptable anymore (Java is now considered stable)
1145 * Add explicit synchronization facilities through semaphores
1146 * Bug fix: Task.setDataSize() only changed the C world, not the value
1147 cached in the Java world
1150 * Dramatically change the way files are handled. API and internals changed, but
1151 this part of MSG was not considered as production grade either.
1152 * Add explicit synchronization facilities through semaphores
1153 * Add a new function MSG_host_get_process_list()
1154 * Preliminary DVFS support (see examples/msg/energy/ for details)
1157 * SMPI is now included directly in the libsimgrid as the windows
1158 linker doesn't force us on splitting it anymore.
1159 * Improvements of the SMPI replay tool:
1160 - Most of the collective communications are now rooted in the same process as
1161 in the original application.
1162 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1163 used until now). Multiple data types can now be used in a trace.
1164 - The replay tool now supports traces produce either by TAU or a modified
1166 - Bug Fix: the compute part of the reduce action is now taken into account.
1167 - Gatherv collective is now supported
1168 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1169 of smpirun outputs time independent traces for the current run. One file
1170 is created per process. If too many processes are simulated, this behavior
1171 can be changed to one file for all processes by using the
1172 tracing/smpi/format/ti_one_file flag
1173 * smpirun generates the host file if needed (with given host count and platform)
1174 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1175 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1176 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1177 --cfg=smpi/coll_selector:(mpich/ompi)
1178 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1180 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1181 happening inside a collective SMPI call.
1182 * Fix the behavior of complex data types handling.
1183 * Make MPI_Wtime another synchronization point to take computations into
1185 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1186 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1187 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1188 (no privatization of global variables yet)
1189 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1192 * Verification of liveness properties is now available for SMPI applications
1193 (in addition to MSG applications)
1194 * Bugged examples using SMPI in examples/smpi/mc/
1195 * Add --cfg=model-check/visited option. Allows the verification of infinite
1196 programs. Detection of loops in the execution thanks to the system state
1197 comparison and reduction of the state space to explore. Can be combined with
1198 DPOR for safety properties.
1201 * Allow to change SimGrid configuration (see --help) within the code
1202 thanks to SD_config() as it can be done in MSG.
1203 * Add a new function SD_task_set_amount() upon user request.
1206 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1207 * Remove rule based routing (no more PCRE dependency)
1208 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1209 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1210 * Add a loopback_bw and loopback_lat options to cluster tag.
1211 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1212 routing. Now peers are encapsulated in an AS and have their own private
1213 router but this is transparent.
1216 * Our own implementation of getline is renamed xbt_getline, and gets
1217 used even if the OS provide a getline(). This should reduce the
1218 configuration complexity by using the same code on all platforms.
1219 * New type: xbt_cfg_elm_boolean.
1220 * Allow to use yes/no for boolean configuration options in the command line.
1221 * Allow to disable SimGrid cleanups at exit from command line option.
1222 There are situations where one may want a simulation to end with an exit.
1223 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1224 annoying when scripting around the simulator. Adding a
1225 --cfg=clean_atexit:no allows to circumvent this issue.
1228 * Lots of memory leaks were corrected in this release.
1229 There is no known memory leaks anymore, in all of our 600+ tests.
1230 * New command line option --version, to get SimGrid version information.
1231 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1233 * Supernovae builds are deprecated, and expected to be removed in the next
1236 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1238 SimGrid (3.9) stable; urgency=low
1240 The Grasgory release: GRAS is really dead now.
1242 * Complete overhaul of the internal host structures scheme.
1245 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1246 was considered as experimental and badly maintained since then.
1247 * Keeping it was thus a trap to our potential users, that could take
1248 it instead of MSG or SMPI by mistake despite is pity state.
1249 * GRAS seems to have very few users (if any), and no one volunteered
1250 to maintain it further. It also induces a lot of XBT code (for
1251 portability sake), that must be maintained too.
1252 * For all these reasons, we killed GRAS. If someone wants to revive it
1253 in the future, don't cry, our git history still remembers of GRAS.
1256 * Major overhaul. Merge our documentation again as time proved that
1257 splitting it was really not helping our users.
1258 * Further improve the developer documentation to help newcomers
1259 hacking on SimGrid itself. The user documentation (and in
1260 particular, the beginner documentation) is still in a sorry state.
1263 * Now works on Windows too!
1264 * Much more extensive test suite, from MPICH
1267 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1268 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1269 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1270 * Bug fix: let task be scheduled when the last dependency to be solved is
1271 a control dependency.
1272 * Remove SD_load_environment_script function.
1273 Use the C sg_platf function if you want to declare a platform
1277 * New function: MSG_process_get_number()
1278 * Old function documented: MSG_config()
1279 * Remove MSG_load_platform_script function
1280 Use the C sg_platf function if you want to declare a platform
1284 * Change the default value of the TCP_gamma constant (maximal size of TCP
1285 congestion window) to a more realistic 4MiB value. If you notice changes in
1286 your simulation results, you can fall back to the previous 20k tiny window
1287 by adding --cfg=network/TCP_gamma:20000 on command line.
1288 * (Hopefully) fix a bug wrt periodic availability and state traces
1289 * Bug fix: use default values at start when first event in availability/state
1290 trace is not at time 0.
1293 * remove the "new_" part of function name sg_platf_new_trace_connect
1294 (resulting in sg_platf_trace_connect), since it does not create
1298 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1299 use case where it's really mandatory, and maintaining it was a pain in
1301 * New: xbt_fifo_search(), search an item with a user-provided
1302 comparison function instead of dumb pointer comparison.
1305 * Fix the lua deployment:
1306 Use `simgrid.init_application()` before deployment instead of
1307 `simgrid.msg_register_application()` after.
1310 * Transfer the tracing files into the corresponding modules.
1312 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1314 SimGrid (3.8.1) stable; urgency=low
1316 The "we are told that some people want to also *install* the simgrid
1319 * Add missing file "tesh.1" to the archive.
1321 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1323 SimGrid (3.8) stable; urgency=low
1325 The Psssshiiiit release: SimGrid jumps into the Cloud.
1328 * Add an experimental interface to manipulate VMs. They are mainly
1329 process groups with very few intrinsic semantic, but they should
1330 allow you to build the semantic you want easily.
1331 * New function: MSG_host_set_property_value()
1332 * New function: MSG_process_on_exit(). To clean memory in all cases.
1333 * Bug fixes that made the host (and link) failures unusable.
1334 * Add a way to auto-restart process when the host in which they are
1335 executing comes back (ON_FAILURE="RESTART" on deployment file,
1336 MSG_process_auto_restart_set).
1337 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1338 please stop using the old ones, they are DEPRECATED.
1340 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1341 Please use MSG_init() instead. (reducing the amount of entry
1342 points in the library helps us).
1343 * Make it impossible to link against the wrong version of the lib
1344 * Deprecate MSG_clean(). No need to call it anymore.
1345 * Function MSG_get_host_number() is not deprecated anymore.
1348 * Split the doc into a user guide and a reference guide.
1349 * Start a developper guide to help people hacking on SimGrid.
1352 * Enable tracing by default. This modules rocks you should use it.
1353 * Remove option custom_flags. Now use environment variables CFLAGS
1355 * Use default cmake things to detect lua instead of home grown ones.
1356 * New option "enable_mallocators" to disable mallocators, for debugging
1357 purpose ("on" by default).
1360 * Bug fixes around the resource failures: don't let the processes
1361 survive the host they are running onto.
1362 * Add an interface to auto-restart processes when the host in which they are
1363 executing comes back.
1364 * Ensures that SIMIX_clean is called automatically. It's not part of
1365 the public interface anymore (bindings should be updated).
1368 * Bug fix for when SD_Simulate is called with a positive value: be careful
1369 when comparing doubles. Sometimes they are different for non significant
1371 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1372 parallel task whose initial work is distributed among host according
1373 to the Amdahl's law. Such tasks are created with a parameter alpha
1374 that corresponds to the non-parallelizable part of the computation.
1375 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1376 between two sets of workstations assuming a 1D block distribution (each
1377 workstation owns a similar share of data) on both sides.
1379 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1380 Data redistribution will be automatically scheduled once parent and child
1381 are both scheduled. The filling of computation_amount and
1382 communication_amount structures is now done seamlessly thanks to the chosen
1384 * New function SD_workstation_dump to display various information
1385 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1386 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1387 bandwidth on the route onto which the task is scheduled and the amount of
1389 To divide the nominal bandwidth by 2, the rate then has to be:
1390 rate = bandwidth/(2*amount)
1391 * Compute tasks that have failed can now be rescheduled and executed again
1392 (from their beginning)
1393 * Increasing source code coverage (src/simdag is now covered at 95.8%
1397 * Re-implement time-independent trace replay using SMPI (at the
1398 smpi_smp_* level) instead of MSG. This should replace
1399 examples/msg/actions/actions.c
1400 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1401 hindexed and structs)
1402 * Implement the exchange of non-contiguous data.
1403 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1404 * Correct behavior of smpi/sender_gap and set its default value to 0
1405 * Add option to asynchronously send small messages to allow better
1406 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1407 specifies the size in bytes under which messages will be asynchronously sent.
1408 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1409 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1412 * New C interface to define a platform: XML is now optional.
1413 For more info, please check include/simgrid/platf.h.
1414 * New interface to define random platforms from the C:
1415 For more info, please check include/simgrid/platf_generator.h and
1416 examples/msg/masterslave/masterslave_platfgen.c
1417 * Export a sg_cmdline dynar containing all the arguments we got from
1421 * Two new tracing options for adding comments to trace file so you
1422 can track your experiments (see --help-tracing for details).
1423 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1424 * Adding the SimGrid version that generated the trace file as a comment.
1425 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1426 * Fix to avoid key clashes on Paje links
1427 * Other minor fixes related to the Paje specification
1430 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1431 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1432 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1433 * Greatly improve the robustness of mmalloc to user errors (such as
1434 using an area after freeing it, or freeing it twice)
1436 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1438 SimGrid-java (3.8.1) stable; urgency=low
1440 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1441 Warning: all methods are visible, but only some of them are
1442 implemented so far. Check the source (src/jtrace.c)
1443 for further information.
1444 * New module: org.simgrid.msg.File (SimGrid File management functions)
1445 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1446 * Change the meaning of Process.restart: now restart the process from
1447 the begining, like MSG_process_restart in C.
1448 * Add Process.setAutoRestart: handling of process restart when failed
1450 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1451 you to retrieve the properties of the processes/hosts
1452 * Deprecate Msg.clean(): you can just forget about it now.
1453 * New function Process.getCount(), that only works when compiling
1454 with the not yet released version 3.9 of the C library.
1456 * New context factory based on Coroutines. It mandates a modified JVM
1457 but then, the simulations run about five times faster, and there is
1458 no limit to the amount of processes (beside of the available memory).
1460 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1462 SimGrid (3.7.1) stable; urgency=low
1465 * Restore the prototype of MSG_process_create_with_environment() to
1466 the pre-3.7 situation by removing the kill_time argument.
1467 * Add a MSG_process_set_kill_time() function instead.
1470 * Fix weird behaviors when dealing with parallel tasks.
1473 * Simgrid is now built as a dll.
1474 * Simgrid-java now works on Windows.
1475 * Simgrid-Java is now included into Windows package.
1478 * First pre-build package for MacOSX.
1481 * Fix compilation when using MSG_USE_DEPRECATED.
1482 * Fix some compilation issues on Macs and Windows.
1483 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1484 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1486 We discovered that the Lua console is broken, but we are missing the
1487 manpower to fix it right now. The problem existed in 3.7 too, so we
1488 are not blocking the release for that. Sorry if you depended on this
1489 feature, any help would be really welcome.
1491 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1494 SimGrid-java (3.7.1) stable; urgency=low
1496 The "Java aint got to be bloated and slow" release
1499 * Various internal cleanups and performance improvement
1500 Simulations are expected to run up to twice faster or so
1501 * Make Process.kill(process) an instance method, not a static one
1502 * User processes are not java.lang.Thread subclasses.
1503 This breaks the compatibility (sorry), but previous API was
1504 brain-dead, making it impossible to have non-trivial
1505 initializations in the process constructor.
1506 * Require a full constructor per Process sub-class.
1507 Kinda breaks the compatibility (sorry), but this allows a much more
1508 efficient way to launch the processes at simulation startup.
1509 * Do not embeed our version of semaphores, java 1.5 can be considered
1510 as sufficiently prevalent for us to not dupplicate its features.
1514 * Add examples for almost every part of the API
1515 We spotted and fixed a lot of bugs in the process
1517 * New module: asynchronous communication API
1518 * New function: Process.sleep()
1519 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1520 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1521 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1523 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1525 SimGrid (3.7) stable; urgency=low
1527 The "spring cleanups (before the next Big Project kicks in)" release.
1530 * We can specify the SMPI latency/bandwidth factor with command line
1531 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1532 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1533 You can also use the "config tag" from platform file by adding this line
1534 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1535 (see "example/platforms/tag_config.xml" to use "config tag").
1536 Note that the command line supersedes the platform file configuration.
1537 * Change the correction factors used in LMM model, according to
1538 the latest experiments described in INRIA RR-7821.
1539 Accuracy should be improved this way.
1540 * Use the partial invalidation optimization by default for the
1541 network too. Should produce the exact same results, only faster.
1542 * Major cleanup in surf to merge models and split some optimization
1543 mechanisms from the core of the models. As a result you can now
1544 specify which model to use (e.g., --cfg=network/model:LV08
1545 --cfg=cpu/model:Cas01) and which optimization mode to use
1546 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1547 Incompatible combinations should err at initialization. See
1548 --help-models for the list of all models and optimization modes.
1549 * The CLM03 workstation model was dropped for simplicity because it
1550 used the deprecated CM02 network model. Use default instead.
1551 * Rename the TCP_gamma configuration option to network/TCP_gamma
1552 * Rename the coordinates configuration option to
1553 network/coordinates, and document it
1554 * Use now crosstraffic keyword instead of the terribly misleading
1555 fullduplex keyword. It is activated by default now in the current
1556 default model, use --cfg=network/crosstraffic:0 to turn it off.
1557 * Ongoing refactoring the model parsing to make XML files optional
1558 See include/simgrid/platf.h for details (still to be completed)
1561 * Major overhaul of the documentation. Almost instructive now :/
1562 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1563 functions and friends. This interface was considered as
1564 deprecated since over 2 years, it's time to inform our users that it is.
1565 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1566 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1567 use these (crappy) functions in your code.
1568 These functions will be removed soon. Stop using them now.
1569 * Deprecate MSG_get_host_{table,number}
1570 Implement MSG_hosts_as_dynar() instead.
1571 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1572 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1575 * Stabilize the parallel execution mode of user contexts
1576 * Introduce configuration variables to control parallel execution:
1577 - contexts/synchro: Synchronization mode to use when running
1578 contexts in parallel (either futex, posix or busy_wait)
1579 - contexts/parallel_threshold: Minimal number of user contexts
1580 that must be part of a scheduling round to switch to parallel
1581 execution mode (raw contexts only)
1582 * Fix bugs that prevented to use suspend/resume along with
1583 synchronization structures.
1584 * Fix bugs in process termination that lead to invalid memory access
1585 in very specific conditions.
1588 * Introduce a parallel mode for the models (controlled by surf/nthreads
1589 configuration item). In our tests, running the models in parallel
1590 never lead to any speedups because they are so fast that the gain
1591 of computing each model in parallel does not amortizes the
1592 synchronization costs, even when ultra fast futexes are used.
1593 This is released anyway because YMMV.
1596 * Performance boost by using a swag internally to compute the set of
1597 tasks that are finished and should constitute the return value of
1601 * Enable it by default now that it is considered rather stable.
1604 * Documentation of the tracing functions.
1605 * Performance gains when tracing categorized/uncategorized resource
1606 utilization by avoiding calls to get route when updating resource
1607 variables. LMM constraints are being used instead.
1608 * API changed to set task categories. Use MSG_task_set_category instead
1609 of TRACE_msg_set_task_category, and SD_task_set_category instead
1610 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1611 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1612 * Examples for link user variables added (see at examples/msg/tracing/)
1613 * Deprecated function TRACE_msg_set_process_category completely removed
1614 * Trace header updated according to the latest Paje file format
1615 * Tracing network lazy updates, no longer obligate users to use full updates
1616 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1617 * Experimental: let user code declare/set/push/pop application states for hosts
1618 * API changed to allow the manual creation of graph configuration files
1619 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1622 * Improve the API of Lua MSG bindings, using the Lua spirit.
1623 * Each simulated process now lives in its own Lua world (globals are
1624 automatically duplicated). It helps writing simulators. Will allow
1625 to run Splay programs within SimGrid in the future.
1626 * Add a Chord example in Lua, equivalent to the MSG one.
1629 * Start the implementation of a solution to express temporal
1630 properties, not only local assertions. This is still an
1631 experimental work in progress, stay clear from it to be safe.
1635 - Add new runtime parameters --help-logs and --help-log-categories
1636 to display information about supported logging parameters and
1638 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1639 don't exists anymore.
1640 * Mallocators: allow value NULL for the reset function.
1642 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1643 create homogeneous dictionaries, where all the elements share the
1644 same free function. Non homogeneous dictionaries will be
1645 deprecated in the next release.
1646 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1647 - Multi-level dictionaries are deprecated.
1649 - new function xbt_dynar_search_or_negative() that is useful when
1650 you have less than 2 million elements in your dynar and don't
1651 want of the extra complexity of catching exceptions when the
1652 element is not found.
1654 - Make xbt_os_thread module (for thread portability) public.
1655 Documentation is still to come, sorry.
1657 - Cleanups and simplifications to make it maintainable again.
1658 - Exotic features (such as memalign and valloc) were removed.
1659 - The metadata were extended and improved so that the
1660 model-checker becomes able to explore and inspect the heaps.
1661 - This may induce a performance drop when enable_model-checking is
1662 ON in cmake (even if it's not used in the simulation), but it is
1663 necessary at this point to get MC working.
1665 Turn model-checking OFF if simulation performance matters to you.
1666 Not enabling it at runtime is not enough, disable it in cmake.
1668 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1671 SimGrid (3.6.2) stable; urgency=low
1673 The "Not coding new stuff allows to polish old things" release.
1676 * New bindings to the NS3 packet level simulator (experimental)
1677 * Use the raw (efficient) execution contexts instead of the sysv
1678 (portable) ones when possible.
1679 * libpcre is now mandatory in any cases since not using it led to
1680 severe performance loss and possibly other issues
1681 * Update the XML platforms:
1682 - G5K: include the latest machine in Nancy
1683 - GridPP and LCG: new platforms
1684 * Documentation was partially updated, at least (more to come)
1686 Bug fixes, cosmetics and small improvements
1687 * Free terminated processes before the end of the simulation to avoid
1688 exhausting the memory of users having very dynamic amount of
1690 * Bug fix and cosmetics about canceling non-running tasks
1691 * Bug fix about the dot loader's issues when using libcgraph
1694 * Create an installer for windows with nsis (amd64 and win32)
1695 - Add an hello world project to illustrate simgrid project creation.
1696 - Embed libpcre into the Simgrid installer to avoid
1697 its compilation burden
1698 * The raw execution contexts should work on Apple now
1699 * Port to Windows 64 bits
1700 - Sysv contexts now have an implementation for this arch
1701 - GRAS communication features now support this arch
1702 * Drop support for borland compiler on windows
1703 - this code was not maintained, and we kinda depend on gcc nowadays
1704 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1705 * Fix portability issue on unstable ubuntu: linker became picky on
1708 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1710 SimGrid (3.6.1) stable; urgency=low
1712 The "Oops, we broke Macs too" release
1715 * Fixed contexts detection so that raw ones are used when possible
1716 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1717 leads to a strange error, with user code segfaulting sometimes when
1718 the generated code is not perfectly aligned (which is not
1719 controllable from the user side, depends on the amount of code)
1722 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1724 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1726 SimGrid-java (3.6) unstable; urgency=low
1729 * Split of every thing from simgrid v3.5 into a separate package.
1731 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1733 SimGrid (3.6) stable; urgency=medium
1735 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1738 * Bindings now constitute their own package, separated from the main one.
1739 Rationale: reduce our maintenance nightmare by reducing the module coupling
1740 They will soon be released on their own on gforge.
1742 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1743 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1745 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1746 * It's not quite deprecated for now because we have no replacement,
1747 but it may soon become the case.
1750 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1751 * Fortran: New user-level cache variable to store the rank of the running
1752 process. This improves performance by an order of magnitude.
1753 * C: New coccinelle script to automatically locate and modify global and
1754 local static variables.
1755 * Improved SMPI network model with a sender-side gap to account for multiple
1759 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1760 only say if a communication is finished, no matter whether it succeeded or
1761 failed. You can call MSG_comm_get_status() to know the status of a finished
1763 * New function MSG_task_dsend() to send a task and detach it. When a
1764 communication is detached, you are never notified of its success or failure
1765 and the memory is released automatically once it is finished. This function
1766 is useful when you don't care about the end nor the success of a
1768 * Change the prototypes of action replay. Sorry for inconvenience,
1769 but this is really more efficient this way (and to adapt your code,
1770 you just have to fix the initialization, that shouldn't be too long)
1771 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1772 ever managed to do anything useful with it.
1773 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1774 takes a dynar of communications. It returns immediately and gives the
1775 index of a finished communication (if any).
1776 * New example: a basic implementation of the Chord P2P algorithm.
1779 * New model for multi-core CPUs. You can now use the core attribute to
1780 precise the number of cores of a host. This is a basic model. Every
1781 process running on the host receives at most the power provided in
1782 the DTD (throughput<=power). Total throughput of process cannot exceed
1784 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1785 router linked by an up-link and a down-link (possibly asymmetrical).
1786 This kind of pattern allows to easily build last-mile model style platforms.
1787 Aggregating such patterns in a rule-based AS is thus the technique of
1788 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1789 * New model for Vivaldi routing. We transformed the Vivaldi network model
1790 into a Vivaldi routing model (based on the rule-based model). This allows to
1791 combine Vivaldi based latencies with last-mile platforms.
1794 * Added a check for NaN of IEEE754 infinite in the double entries of
1796 * Introduce a new context factory "raw", highly inspirated from the
1797 ucontext factory, but using manually crafted functions in assembly to
1798 do the work in an efficient manner.
1799 * Allow to change the used context factory at run time, not only at
1800 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1801 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1802 contexts at runtime (only with raw contexts or ucontexts).
1803 * Completely rewrote this module to allow parallel execution of user
1804 processes. Use --cfg=contexts/nthreads:N to execute user processes
1805 with N parallel threads (the default is 1, meaning no parallelism).
1806 * Allow to decide dynamically between sequential and parallel modes.
1807 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1808 processes in parallel only when their number is greater than or equal to P
1810 * Added a check for NaN of IEEE754 infinite in the double entries of
1814 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1815 won't output the state of processes when interrupted with Ctrl-C
1816 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1817 NULL-terminated array. This may solve backward compatibility issues
1818 due to the change to return type of SD_simulate. See also:
1819 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1820 * Add new macros with variable number of arguments.
1821 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1822 - in xbt/asserts.h: xbt_assert
1823 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1824 - in xbt/ex.h: THROWF and RETHROWF.
1825 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1827 * Change xbt_die() to accept a format string with arguments, just like printf.
1828 * New data structure: xbt_lib_t, like a dict but more general and with better
1832 * New configuration options
1833 Options triva/categorized and triva/uncategorized can be used to generate
1834 graph configuration files for Triva visualization tool.
1835 * Configuration option tracing/platform is renamed to tracing/categorized
1836 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1837 * New TRACE_user_link_variable interface:
1838 User provides the name of the link and the tracing variable to attach to it
1839 * the declaration of tracing categories must be done after the environment
1841 * simpler tracing interface, just one way to declare categories
1842 TRACE_category or TRACE_category_with_color, it is up to you
1843 * links in the trace file are again identified by their names
1844 * trace contains the full platform hierarchy exactly as declared using the ASes
1845 * Options tracing/msg/[task|process]:1 groups the process by hosts
1846 for both cases, tasks and processes must have names that are unique during the simulation
1847 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1848 * The experimental option tracing/msg/volume is deprecated
1849 its functionality may be reincorporated if needed
1851 The tracing generates a trace file with unordered timestamped events,
1852 because of the way the core simulator (surf) works. A script available
1853 at the tools directory (fix-paje-trace.sh) can be used to put the events
1854 in order. We have changed the tracing so it can generate ordered timestamped
1855 events in the final trace, but depending on the simulator (and how much time
1856 is simulated) that can lead to a huge memory utilization. It is deactivated
1857 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1859 Build Infrastructure
1860 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1861 - We are trying hard to keep the API stable, but it may happen that
1862 some things change (we're a research project after all, not a
1863 nuclear plant operating system). If such things should happen, you
1864 could rely on that macro to adapt.
1865 - current value: 30600 for 3.06.00, aka 3.6
1866 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1867 a number that can be compared with SIMGRID_VERSION.
1868 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1869 assertions and verbose/debug logging events are disabled at compile time.
1871 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1873 SimGrid (3.5) stable; urgency=medium
1876 * New feature to any SimGrid-based simulator: Model-Checking
1877 Check SIN#1 for more details.
1880 * New Model SMPI (three-interval linear regression for correction factors)
1881 See RR-7426, available at http://hal.inria.fr/inria-00527150
1882 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1883 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1884 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1885 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1886 * See also src/smpi/README
1891 - Tracing API changes: TRACE_start and TRACE_end should not be called
1892 by user-code. They are automatically called by simulators created
1893 with SimDAG, MSG and SMPI if the toolkit is compiled with
1894 tracing_enabled=ON. Categories declaration and utilization remain the
1895 same for MSG and SimDag.
1896 - A function was added to the tracing API to declare categories with
1898 - TRACE_category_with_color (char *category, char *color)
1899 where color must be in the following format
1900 "%f %f %f", red, green, blue
1901 and red, green, blue are float values in the interval [0, 1]
1902 - User can specify NULL as color parameter, or continue calling
1903 TRACE_category (cat)
1904 On that case, the tracing system will define random colors
1905 - The following command-line options are supported:
1906 --cfg=tracing/filename:msg.trace
1907 --cfg=tracing:1 (activate tracing, needed to use others)
1908 --cfg=tracing/platform:1 (categorized resource use)
1909 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1910 --cfg=tracing/msg/task:1 (task creation)
1911 --cfg=tracing/msg/process:1 (process creation, migration)
1912 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1913 --cfg=tracing/smpi:1 (SMPI interface tracing)
1914 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1915 - examples of examples/msg/tracing updated
1917 - DAXLoader and DOTLoader functions can generate tasks with categories
1918 - A new function to attribute a category to SD tasks:
1919 TRACE_sd_set_task_category (SD_task_t task, char *category)
1920 * Tracing the MPI interface implemented by SMPI
1921 - Collective operations are traced with states
1922 - Point-to-Point operations are traced with states/links
1923 - Tracing activated by a parameter "-trace filename" passed
1924 to smpirun during execution (considering that simgrid
1925 is compiled with tracing enabled)
1926 - To run the simulation with gdb, the simulator
1927 accepts --cfg=tracing/smpi:1 to trace SMPI
1928 - tesh files to check if smpi tracing is ok
1929 - See examples/smpi/NAS/DT-trace
1930 * GTNetS tracing re-worked
1931 - adaptation to the tracing system of GTNets to cope
1932 with modifications regarding the fullduplex mode
1933 - new tesh files to check if gtnets tracing is ok
1936 * Asynchronous communications through the functions:
1937 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1938 * New function: MSG_load_platform_script()
1939 to make possible using a lua script instead of XML files to set up platforms
1940 * New function: MSG_set_function
1941 to associate functions to processes, used when bypassing the parser
1942 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1944 Platforms: Add some more examples in examples/platforms
1945 * Grid'5000: see www.grid5000.fr
1946 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1949 * Change the XML format. This is a very important modification. SimGrid
1950 3.5 introduces a new hierarchical format based on the notion of
1951 Autonomous Systems. Compatibility with old format is ensured
1952 through the perl script provided in the install bin directory
1953 bin/simgrid_update_xml.
1954 It is now possible to build platforms with specific routing
1955 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1956 connect several platforms together. We will try to provide soon
1957 set of realistic platforms exploiting these properties (have a
1958 look at examples/platforms/ for the moment).
1959 * Take the opportunity of the XML format change to be a good XML citizen:
1960 rename link:ctn to link_ctn and similar changes (also dealt with by
1962 * Add a new routing scheme (rule-based) using regular expressions. It
1963 enables to have an extremely low memory footprint when the
1964 underlying routing is simple and can be compactly described. You
1965 need to have libpcre4-dev (perl regular expressions) installed if
1966 you want to use this routing scheme.
1967 * Revive the cluster TAG and allow to easily and efficiently (both in
1968 term of memory and speed) connect clusters together. Have a look
1969 at teshsuite/simdag/platforms/ to see how this can be done. With
1970 this tag, you can create clusters with thousands of tasks at no
1971 cost (have a look at examples/platforms/).
1972 Note: clusters are implemented as ASes, so there is no need for an
1973 enclosing AS tag if you have only one cluster in your platform.
1974 * Add new generic functions in the public interface that allows the user
1975 to call SURF 'create_resource' methods from your code (same
1976 functionality as the XML bypass mechanism but with a much lighter
1978 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1979 that uses a piecewise linear approximation to produce better
1980 results when exchanging small messages.
1981 * Add a new parameter to handle correctly full duplex link and account
1982 for interferences between uplink and downlink communications
1983 (activate with --cfg=fullduplex:1).
1986 * Rename the SD_READY (all dependencies are satisfied and task is
1987 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1988 dependencies are satisfied) state.
1989 This prevents a confusion between the notion of "ready to schedule"
1990 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1991 simulated" (SD_RUNNABLE) used by the simulation kernel.
1992 * Change the way a task is considered as ready. Instead of removing
1993 dependencies when a task is done, a counter is decreased. This way,
1994 it is always possible to reach ancestors thanks to the
1995 SD_taks_get_parents function (even after the end of the simulation.)
1996 * Change the return type of SD_Simulate from (SD_task_t*) into
1997 xbt_dynar_t. This function was in handling a dynar internally and
1998 converted it into a NULL terminated array for historical reasons.
1999 * New function SD_dotload(char*) to load a DAG described in dot
2000 format. This loader and the corresponding examples require the
2001 installation of the graphviz library.
2002 * Fix a bug in the management of tasks of size 0 in the surf network
2003 models. This problem was only visible with SIMDAG and you should
2004 thus disregard results produced with earlier versions if you
2005 relied on this feature (some tasks were blocked because of this).
2006 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2007 even though your applications were doing only point-to-point
2008 communications and sequential computations. Now you can really use any
2009 model you want (of course, if you create real parallel tasks, which are
2010 not implemented in most models beside ptaskL07, this will abort).
2011 * Add an example that schedules a DAX on an heterogeneous platform
2012 using a Min-Min strategy.
2013 * New function SD_workstation_get_current_task() that returns the kind
2014 of task currently running on a workstation in the sequential access
2016 * Raise some warnings when unexecuted tasks remains at the end of the
2017 simulation. This is usually caused by cycles in the DAG.
2020 * New function: SIMIX_process_set_function() called by MSG_set_function
2021 * Change the underlying waiting queue in semaphores so that a process
2022 can wait on several of them simultaneously (as in waitany).
2023 * Fix the way to handle tokens in semaphores so that all access patterns
2024 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2025 * kill the dirty pimple SIMIX_message_sizes_output()
2026 Please use (proper) visualization instead
2029 * New data container: setset (set of sets of elements)
2030 * New module: mmalloc (mapped malloc, allowing to have several
2031 independent segments of malloc)
2032 * New function: xbt_dict_cursor_set_data()
2033 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2034 * New function: xbt_dynar_is_empty()
2035 * New function: xbt_fifo_get_last_item()
2036 * Fix xbt_dynar_shrink(): use the right element size.
2037 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2038 uninitialized areas during expand.
2039 * Fix semaphores: previous implementation was severely broken.
2040 * Use library init/fini functions for our initialization.
2041 - you can use logs and other feature as soon as you want in your
2042 code (even before the xbt_init / MSG_init)
2043 - xbt_exit is now a no-op and produce a warning when used.
2046 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2047 benefit from the latest improvement to the simulation kernel.
2048 * Kill measurement sockets for now. If you rely on them, sorry. This
2049 release is not for you. This feature will be reintroduced in the
2050 future, but we cannot delay the release any further.
2051 * New function: gras_msgtype_get_name().
2052 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2053 * Fix (at last) the pmm example: it should not randomly fail anymore.
2055 Build chain: bug fixes and overall polishing
2056 * Cmake is now stable enough. Hence, we killed the autotools.
2057 * Port to windows ( TM :)
2058 * Fix the 'make install' target.
2059 No need to use 'make install-simgrid' anymore
2060 * Introduce a 'make dist' target compiling a *source* archive
2061 'make package' compiles a binary archive
2062 * Compile java files only on need
2063 * Add --cd and --setenv command line options to tesh
2064 * Out of source builds are not fully supported yet, but we are close
2065 * Enable supernovae and optimization flags by default for our users
2068 * Add layer to set up environment directly from lua, without XML.
2069 * The effect of gras_stub_generator can be achieved through
2070 lua too (check examples/gras/console/ping_generator.lua)
2072 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2074 SimGrid (3.4.1) stable; urgency=low
2076 The "Polishing easter eggs is probably a good idea" release.
2077 This is a bug fixes release only.
2081 * Fix a bug preventing the tasks from begin garbage collected.
2084 * Fix a bug occuring when a host involved in a communication fails.
2085 This was not detected properly by the other peer involved in the
2086 communication. Now, it's reported as a network error.
2089 * Warn the user about loop dependencies in data flow of DAX files
2090 * Obey the control-flow dependencies of DAX files
2093 * Add option "enable_smpi" allowing to not compile SMPI.
2094 Probably useful for the (Mac) users experiencing a build error here
2095 * Improve the detection of lua5.1 and ruby1.8
2097 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2099 SimGrid (3.4) stable; urgency=low
2101 The "Easter in Cargese" release. Also known as (major changes):
2103 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2104 ~> bindings were greatly improved
2105 ~> new tracing infrastructure for better visualization introduced
2107 * the "Welcome to configury modernity" release.
2108 ~> we switched from autotools to cmake, and improved our cdash
2111 A more detailled list of changes follow (full detail in svn log).
2113 Java Bindings: Various Cleanups
2114 * (install java-gcj-compat-dev on debian-like to use them)
2115 * Remove put/get: no need to export deprecated interface in Java
2116 Use send/receive instead.
2117 * Cleanup the examples and add a README per directory
2118 * Remove example autoDestination (that's the only way to go now)
2119 * Remove example explicitDestination (was a plain copy of basic)
2120 * Make JniException a runtime exception, so that there is no need to
2121 declare the fact that you may encounter such a beast. I guess that
2122 nobody will ever want to survive such error.
2123 * Create specific errors for each MSG case of failure:
2124 host failure, transfer failure, timeout, task cancelled
2125 * Cleanup the exceptions that may get thrown by each function
2126 * Other internal cleanups in Java bindings. Performance still bad :/
2127 Ruby and Lua Bindings: create them
2128 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2129 * That's new and great, you should try them out.
2130 Same functionalities than Java bindings, only even less polished
2132 * Kill the useless "rate" argument of SD_task_get_execution_time()
2133 Everyone used to provide -1 as a value, it was not used, and the
2134 semantic of a possible use wasn't even clear.
2135 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2136 either as comm costs or compute costs to mean that there is no
2137 such thing for that specific task.
2138 * Add a SD_task_set_name() function
2139 * Fix SD_task_unschedule() on typed tasks
2140 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2141 * In DAX loader, accept useless 'level' attributes to <job> since
2142 LIGO DAGs have them (seem to be to ease graphical representation).
2144 * Add an example masterslave_mailbox.c using send/receive and not
2145 the deprecated put/get interface.
2146 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2147 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2149 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2150 (declare MSG_USE_DEPRECATED to still have the old name)
2151 * Add a MSG_task_set_data() function
2152 * About trace replay (see examples/msg/actions):
2154 - Allow to work with splitted trace files for each process
2155 Give the specific trace file as argument of each process,
2156 and call MSG_action_trace_run(NULL)
2157 You can still have one merged file for all processes.
2158 - Fix implementation of collective operations
2159 * Allow task_execute() on 0-sized tasks (closes #10063)
2161 * This is the first release of SimGrid where SMPI is not considered
2162 beta anymore (even if some corners should still be improved)
2163 * Port over the new SIMIX_network submodule (internal refactoring)
2164 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2165 * Implement more missing elements of the standard:
2167 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2169 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2170 MPI_Scan() MPI_Get_processor_name()
2171 - Added implementation of missing case for Alltoall (warning: it's
2172 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2173 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2174 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2175 SMPI_MPI_Allgatherv()
2176 * Bug fixes include:
2177 - MPI_Waitsome() was broken
2178 - Allow relative includes in smpicc
2179 - Command line cfg argument 'reference_speed' was ignored...
2180 - Some functions did not properly lead to auto-benching of user code
2181 - smpicc passes -O2 by default (just like openmpi one)
2183 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2184 * Bug fixes about timeouts during communications
2185 * add SIMIX_message_sizes_output() as a pimple to write to file the
2186 amount of messages per size. Use gnuplot to get histogram.
2187 Pimple because that's the only user-visible function of simix,
2188 defined directly in xbt.h (irk, sorry)
2190 - Add a SIMIX_sem_get_capacity() function
2191 - Fix interactions with processe resume/suspende
2192 - release_forever() was stupidly broken
2193 - Fix SIMIX_display_process_status() for processes in a semaphore
2194 - Make SIMIX_sem_block_onto() user-visible
2195 * Refactoring context stuff:
2196 - Use pseudo-OOP for better modularity
2197 - reimplement SIMIX_process_kill() without process_schedule() so
2198 that the latter can take as invariant that it is called from
2200 - Merge context_start into context_new for sake of simplicity
2202 * Add a Vivaldi network model, coded live during SUD'10 ;)
2203 * Rename configuration variables to start a hierarchy:
2204 o cpu_model -> cpu/model
2205 o network_model -> network/model
2206 o workstation_model -> workstation/model
2207 * New configuration variables:
2208 o network/bandwidth_factor: correction to bandwith
2209 o network/latency_factor: correction to latency
2210 o netwotk/weight_S: correction to the weight of competing streams
2211 * Add a long description to the models, that users can see with such
2212 argument on the command line: --cfg=cpu/model:help
2213 * --help-models display the long description of all known models
2215 * config: add the ability to set a default value after registration
2216 Does not override any previously set value (e.g. from cmd line)
2217 * dict: allow to have integer key and data.
2218 When so, you need to use the following functions
2219 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2220 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2221 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2222 In contrary to regular dicts, the key is not malloced before copy.
2223 Mixing scalar and regular elements in the same dict is not tested
2225 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2226 Tracing for Visualization:
2227 * SimGrid is now instrumented in order to generate a trace file for
2228 visualization analysis: to use it, need to compile SimGrid with the
2229 "tracing" option enabled, and instrument the program using SimGrid with
2230 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2231 (among other functions).
2232 * The instrumentation only traces the platform utilization for now
2233 * Documentation to use the tracing functions and how to analyze the
2234 traces with the Triva tool is written.
2235 * More information about: SimGrid FAQ (in the section Tracing Simulations
2238 * We moved to cmake as default build system. Autotools support will
2239 be dropped soon. Check the FAQ for more info about how to use it.
2240 * Greatly improved our cdash/ctest interactions
2241 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2242 * Added memory checking tests with valgrind; lot of memleak fixing.
2243 This may be the first release of simgrid with so few memory issues
2244 * Added code coverage tests.
2245 Our coverage is still improvable, but at least we see it on cdash.
2247 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2249 SimGrid (3.3.4) stable; urgency=low
2251 The "Desktop Grid needs love too" release (also called Xmas release).
2253 Models improvements:
2254 * Major speedup in the maxmin system solving by using lazy evaluation
2255 Instead of solving completely the maxmin system at each iteration,
2256 only invalidate (and recompute) the modified parts.
2257 This new feature is enabled in default models but you can try to
2258 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2259 * Cas01 IMproved as default CPU model
2260 This CPU model is the same Cas01 model, but it uses the
2261 maxmin-selective-update flag and a heap structure to manage
2262 actions on SURF kernel.
2263 It reduces the complexity to find the next action to finish and,
2264 consequently, it's faster than the old Cas01.
2265 This is the new default CPU model (Cas01).
2266 * Rename the old Cas01 model to Cas01_fullupdate
2267 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2268 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2269 * CpuTI (CPU Trace Integration)
2270 A new CPU model whose objective is simulate faster when using
2271 availability trace files.
2272 Instead of using a full featured, over engineered maxmin system for
2273 CPU modeling, this model does the pre-integration of traces files
2274 to calculate the amount of CPU power available, and so, executes
2275 faster than the old CPU models.
2276 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2277 * Use LV08 as default network model since it gives better accuracy
2278 for small messages and shouldn't change things for big ones.
2279 Use --cfg=network_model:CM02 to get the previous behavior.
2282 ******************************************
2283 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2284 ******************************************
2285 * The new CPU model may changes simulations!
2286 The point is that events occurring at the exact same timestamp
2287 are not scheduled in the same order with the old and new
2288 version. This may be enough to completely change the execution
2289 of simulations in some cases.
2290 * The new network model will change simulations!
2291 This new model is more realistic than the previous one, so you
2292 should consider redoing your old experiments with this model.
2293 Sorry for the inconvenience.
2296 * Introduce the supernovae compilation mode
2297 When compiled that way, the whole SimGrid (or almost) is put in a
2298 single compilation unit and compiled in one shoot.
2299 This is to help gcc which has difficulties to inline stuff from one
2301 The speedup seem to be above 15%, althrough more tests are needed on
2302 amd64 to confirm that gain.
2305 * Port of MSG's mailbox on top of SIMIX network
2306 The put/get mechanism was greatly simplified on the way.
2309 * New SIMIX network module. Provides:
2310 - Mailbox: rendez-vous mechanism to find with who you want to speak
2311 - Synchronous send/recv: easier and hopefully faster since the
2312 logic is handled in the maestro process directly now
2313 - Asynchronous send/recv: you dreamt of it? It's here now
2314 Too bad that nobody cared enough to propagate the change to MSG.
2315 * Add semaphores as SIMIX synchronization mechanism.
2318 * new function SD_daxload(char*) to load a DAX file
2319 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2320 * Introduce typed tasks. Specify its kind and cost at creation.
2321 At scheduling, just give where it should be placed, and the cost
2322 for each involved resource is automatically computed.
2323 Existing constructors so far (more to come of course):
2324 - SD_task_create_comm_e2e() for end-to-end communication
2325 - SD_task_create_comp_seq() for sequential computation
2326 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2327 * new function SD_task_dump() for debuging display
2328 * new function SD_task_dotty(task,FILE*) writing to file the info
2329 about the task in dotty format
2330 * SD_task_dependency_exists() can now cope with having one of its
2331 arguments NULL. If so, it tests whether the other argument has any
2333 * Add getters on list of preceding/following tasks:
2334 SD_task_get_parents(task) and SD_task_get_children(task)
2335 * Add getters on amount of workstations and list:
2336 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2337 * Add getter on task kind: SD_task_get_kind(task)
2338 * Update the start_time and finish_time of tasks on completion/failure
2339 * Bugfix: Remove task from state swags when destroyed
2342 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2345 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2346 dynar with one separate thread per value of the dynar.
2347 * Change the prototype of xbt_thread_create(), sorry.
2348 Added a boolean parameter indicating whether we want to join this
2349 thread (used in SG only for now)
2350 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2353 * GTNetS wrappers should now be usable again (and betterly tested too)
2354 * Fix a major regression from 3.2 where the timeout provided to
2355 MSG_task_put_with_timeout() was used as absolute time before which
2356 the comm should be done.
2357 * Start to fix the <cluster> tag.
2358 - Internal links should be good now (beside of the loopback, which
2359 use the private link instead)
2360 - paths to the external world is still rather broken
2361 - the <route:multi> tag is just broken. Actually that's brain-dead.
2362 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2363 to make it less stupid
2364 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2365 * Fix a source-level compatibility glitch from 3.2: after defining
2366 MSG_USE_DEPRECATED, you can use the old name
2367 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2368 * Allow to compile from the SVN with automake 1.11
2369 * Fix some problems when using the "start_time" tag in deployment XMLs.
2370 * Fix #8569: XBT/synchro.h has redundant declarations
2371 * Fix #8563: MSG return values and exceptions
2372 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2373 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2374 Thanks to Robson Peixoto.
2375 * Fix a vicious bug in dictionaries inducing that some elements were
2376 not freed on xbt_dict_free()
2378 Portability report of this version:
2379 * Main portability targets:
2380 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2381 on (amd64/i386/ia64)
2382 - mac leopard on i386
2383 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2384 but nothing critical.
2385 * Other platforms: windows, AIX and others were not tested for this release
2387 Timing report of this version:
2388 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2389 scenario parameters). From 8h to a few seconds in desktop grid settings.
2390 * Supernovae brings about 25% speedup on i386.
2392 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2394 SimGrid (3.3.3) stable; urgency=low
2396 The "Need for Speed" release.
2398 The timings done to validate the 3.3.2 were faulty.
2399 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2401 The problem was a conversion from a manually handled vector to
2402 xbt_dynar_t on the critical path.
2403 xbt_dynar_foreach calls functions, inducing stack management crap.
2405 We inlined these functions and xbt_dynar_foreach is now breath taking.
2406 We also inlined xbt_swag_belong on the way.
2408 Here are some approximate speedup measurements (on master/slaves
2409 simulations lasting between 10s and 20s each):
2410 3.3.1 -> 3.3.2: about same performance
2411 3.3.2 -> 3.3.3: 40% speedup
2412 3.3.1 -> 3.3.3: 40% speedup
2413 3.3.1 with inline patch -> 3.3.3: 30% speedup
2415 Our reading is that the refactoring which occurred in 3.3.2 made us
2416 suffer much more from the xbt_dynar_foreach low performance, but
2417 once we solved this, this refactoring proved to be very performance
2418 effective. From the 40% speedup, somehow, 10% are due to the
2419 inlining and 30% to the refactoring.
2421 That's a pitty that gcc cannot inline functions placed in other files
2422 alone. We have to choose between:
2423 - break the encapsulation (by putting private data structures and
2424 accessors in headers files to help gcc)
2425 - live with low performance
2426 - switch to a decent compiler such as icc (not quite possible).
2428 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2430 SimGrid (3.3.2) stable; urgency=low
2432 The "Simplicity does not preceed complexity, but follows it" release.
2434 The main contributors of this release were (lexical order):
2435 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2438 * Extract the routing logic into its own object.
2439 (was dupplicated in network.c and workstation_LV07.c;
2440 Allows to implement other ways of storing that info)
2441 => kill now useless network_card concept
2442 - Use dynar to represent routes (instead of void** + int*)
2443 - kill link_set (use surf_network_model->resource_set instead)
2444 - Add a command-line option to choose the routing schema to use
2445 - Add three new models:
2446 * Floyd (shortest path computed at initialization)
2447 * Dijikstra (shortest path recomputed all the time)
2448 * Cached Dijikstra (shortest path computed on need)
2449 All these models where contributed by Silas De Munck, and are
2450 described in his ICCS09 paper.
2452 * Simplify model declaration
2453 (less redirections, less function to write when defining a model)
2454 - Factorize stuff between models:
2457 surf_model_resource_set(model)
2458 surf_model_resource_by_name(model, name)
2459 - Unify the types of models in s_surf_model_t (using an union)
2460 - Embeed fields of common_public directly into s_surf_model_t
2461 - Rename model methods:
2462 action_free ~> action_unref
2463 action_change_state ~> action_state_set
2464 action_get_state ~> action_state_get
2465 - Change model methods into functions:
2466 (model)->common_public->action_use ~> surf_action_ref
2468 * Implement a generic resource; use it as ancestor to specific ones
2469 (allows to kill duplicated code in models)
2470 Drawback: timer command don't need no name nor properties;
2471 workstation_CLM03 don't need no properties
2472 (but I guess we can live with those few bytes wasted)
2474 * Improve the action object model
2475 - implement a constructor avoiding dupplicated code about field
2476 initialization in generic_action part.
2478 * Kill the SDP model: it has an external dependency, is deprecated
2479 in flavor of modern lmm models, and didn't compile since a while
2482 * Relocation of the context module from XBT to SIMIX.
2483 (the context were decoupled from the simix processes, duplicating a lot of code)
2484 => a lot of code was factorized
2485 - less overhead is introduced during scheduling
2486 - simpler API for the context factory
2487 - the logic for process creation,destruction and manipulation was simplified
2488 * Simplification of the s_smx_process_t data structure.
2489 => accesing the simix level data associated to a process is faster now,
2490 and the code is a lot more readable.
2493 * Implement some more MPI primitives:
2494 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2495 -implementation: Bcast: flat or 2-ary tree (default),
2496 Barrier: 4-ary tree,
2498 Allreduce: Reduce then Bcast
2499 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2500 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2501 Alltoallv: flat tree, like ompi
2503 * Add support for optimized collectives (Bcast is now binomial by default)
2504 * Port smpirun and smpicc to OS X
2507 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2508 Shout out if you used it.
2511 * Display the list of still queued messages in SG mode when existing
2515 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2516 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2517 * Add xbt_str_from_file(FILE*)
2518 * Add xbt_dict_get_key achieving a linear reverse search
2519 * Remove the context module
2521 Portability report of this version:
2522 * Main portability targets:
2523 - Linux(debian)/x86/context
2524 - Linux(debian)/x86/pthreads
2525 - Linux(debian)/amd64/context
2526 - Linux(debian)/amd64/pthreads
2527 On these, we still have the eratic breakages of gras/pmm and
2528 amok/saturate_sg reported in previous version. We still think
2529 that the tests are the cause of the fault, not the tested code.
2531 - Mac OSX Leopard/x86/context
2532 Still false negative in tesh autotesting.
2533 Smpi still fails, but this time because readlink does not accept -f
2534 Everything seems to work properly beside of that.
2537 - AIX version 5.3 (only tested contexts this time)
2538 Smpi still fails there because mktemp is not installed.
2539 Everything seems to work properly beside of that.
2541 I managed to compile it for the first time, but several breakages.
2542 Won't delay the release for this exotic platform.
2544 * Windows: it's still lagging behind. If you want to help, please
2547 Timing report of this version:
2548 This version seem to be more than 5% faster than 3.3.1 (on linux
2549 64bits with contextes). The gain is less than expected, we are
2550 investigating this for next release.
2552 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2554 SimGrid (3.3.1) stable; urgency=low
2557 * Implement a --cfg-help to show existing configuration variables
2558 * Build chain do not require doxygen in maintainer mode
2561 * fix a bug on struct sizeof computation, which prevented the
2562 exchange of arrays of structs in some conditions
2563 - added a regression test about this in datadesc_usage
2564 * Allow the exchange of 0-long dynamic vectors.
2565 - for that, use -1 as indicator of dynamic size instead of 0
2566 - This implied to change any size from unsigned long to long,
2567 reducing a bit communication abilities, but I guess that with
2568 64bits being quite common, this is more than enough.
2569 - This also induce a protocol change, thus bumping network protocol
2570 version from 0 to 1 (if we have external users, we have to get
2571 clean on that point too ;)
2572 - added two regression tests about this in datadesc_usage
2573 * Be more verbose when propagating local exceptions
2574 This helps debugging.
2575 * Display the status of simulated processes when receiving SIGINT in
2579 * Allow to control the simulation from a trace file.
2580 New functions MSG_action_register() and MSG_action_trace_run()
2581 The first one allows to associate a function execution to each
2582 kind of action while the second one parses a trace file and
2583 triggers the corresponding actions within the system.
2584 For now, only a toy example is provided in examples/msg/actions
2585 * Add an exemple of process migration in examples/msg/migration
2586 * Fix a bug in task exchange which broke MSG_task_get_sender()
2587 Add a teshsuite regression test for that.
2588 [Bug: if MSG_task_get_sender() is called after sender exit,
2590 * Fix a bug which prevented suspend/resume to work properly
2591 * Display the status of simulated processes when receiving SIGINT
2592 This fixes a regression of v3.3. due to the introduction of SIMIX
2593 * Bug fixing in failure management:
2594 - trace could not start by a failure at time 0
2595 - failure during communications were not working
2598 * Add SIMIX_process_set_name() to change the name of the current
2599 process in the log messages.
2600 * Store smx_hosts in a dict since we only retrieve them by name
2601 * Move the configuration infrastructure to surf
2604 * Move the configuration infrastructure to surf
2607 * Massive internal cleanups:
2608 - Store internal structures on processes instead of hosts (allows
2609 to have more than one process per host, in addition of being more
2611 - Cleanup the initialization/finalization process
2612 - Kill a whole bunch of unneeded synchronization:
2613 processes run in exclusive manner within the simulator
2614 - Move queues from global tables to process data fields
2616 - now accept -platform and -hostfile arguments
2617 - Pass the right rank value to processes according to the hostfile
2618 * Compile the examples by default, and use them as regression tests
2619 * Implement MPI_Wtime()
2620 * Change the reference speed to a command line option
2623 * TCP_gamma can now be specified as command line option using
2624 --cfg=TCP_gamma:10000000.0
2625 * Change the --surf-path cmd line option into --cfg=path:
2628 * Also include strbuff from xbt.h public header
2629 * xbt_ex_display(): do not free the exception after displaying
2630 This allows to do more with the given exception afterward.
2631 Users should call xbt_ex_free() themselves.
2635 Portability report of this version:
2636 * Main portability targets:
2637 - Linux(debian)/x86/context
2638 - Linux(debian)/x86/pthreads
2639 - Linux(debian)/amd64/context
2640 - Linux(debian)/amd64/pthreads
2641 These targets fail about 1/10 of times on gras/pmm, but we believe
2642 that this is because of the test, not because of simgrid.
2643 amok/saturate_sg fails even more rarely, and the test may not be
2646 - Mac OSX Leopard/x86/context
2647 The test suite still spits tons of errors because some obscure
2648 force prevents us from removing the temporary directories
2649 arguing that they still contain some metadata I've never heard of.
2650 Smpi fails because seq is not installed.
2651 Everything seems to work properly beside of that.
2654 - AIX version 5.3 (both contexts and pthread)
2655 Smpi still fails there because mktemp is not installed.
2656 XML inclusions seems rosty on AIX.
2658 * Windows: it's still lagging behind. If you want to help, please
2661 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2663 SimGrid (3.3) stable; urgency=high
2667 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2668 [Malek Cherier & Mt]
2670 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2673 Until now, GRAS were using MSG as an interface to SURF. It was
2674 quite difficult because both interface have several differences
2675 (MSG channels vs GRAS sockets were the most notable point).
2677 This also opens the gate to SMPI (which should occur soon) and speed
2678 up simulations by to 40% (even if it were not the main goal).
2680 **************************************
2681 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2682 **************************************
2683 The point is that events occuring at the exact same timestamp are
2684 not scheduled in the same order with the old and new version. This
2685 may be enough to completely change the execution of simulations in
2686 some cases. Sorry for the inconvenience.
2688 * Cleanup and upgrade the XML format to push further scalability
2689 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2691 * Improve the testing infrastructure with tesh. Now a very large part of
2692 the code is tested not only by being run but also by checking that the
2693 output match an expected output [Mt].
2695 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2696 is really less memory-demanding, which should allow you to use
2697 larger files in SimGrid [AL].
2699 * Inform valgrind about our contextes, so that it becomes usable
2700 with the default (and more effecient) version of SimGrid
2701 [contributed by Sékou Diakite, many thanks]
2704 * Introduce a listener thread in charge of receiving incoming
2705 messages from the network. It allows to overlap communication and
2706 computation but most notably, it removes some stupid deadlocks due
2707 to the fact that so far, a process could not send and receive at
2708 the same time. This made most non trivial communication schema
2710 * Convert the PIDs from long int to int to match the MSG ones (and
2711 linux ones too) [Mt]
2712 * New function: gras_agent_spawn() to launch a new process on
2713 current host. Only working in simulation for now. [Mt]
2714 * New function: gras_os_hostport() returning a constant form (ie,
2715 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2718 * Make the backtrace of exceptions more human readable [Mt]
2719 * New module: xbt/str [Mt]
2720 a ton of string utility functions (split, join, printf to a newly
2721 allocated buffer, trim, etc)
2722 * New module: xbt/hash [Mt]
2723 SHA1 hashing algorithm (more to come if needed)
2724 * New module: xbt/synchro [Mt]
2725 synchronization tools (mutex and conditions) working the same way
2726 in simulation and in real life (mainly useful for GRAS, but not
2728 * New module: xbt/queue [Mt]
2729 classical producer/consumer synchronization scheme
2730 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2731 (using the classical functions will get serialized) [Mt]
2732 * Make dictionary internal table dynamic. No need to specify its size
2733 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2735 * Make sure the log channels are organized as a tree under windows
2736 (because of ANSI C compatibility issue, any channel were child of
2737 root directly) [Mt].
2740 * Cleaned many thing in surf and fixed a few bugs [AL].
2741 * Add a nice command line configuration mechanism to compose models [AL].
2742 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2743 the previous one (KCCFLN05). It relies on something that looks like
2744 a max-min sharing mechanism but cannot be written as such. A new solver
2745 was thus designed [AL].
2746 * Add a new solver to lmm. Based on Lagrange optimization and
2747 gradient-based descent, it enables to efficiently maximise systems s.a
2749 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2751 This solver enables to propose two new network models for TCP Reno and
2752 TCP Vegas based on Low's work. These models still need to be fully
2753 tested though [Pedro Velho].
2756 * Bug fix in SD_simulate. Now the time bound given as argument is
2758 * Use the new parallel task model (ptask_L07) as default.
2759 * Use the SURF command line configuration mechanism.
2760 * 0-size tasks (for synchronization) should now work.
2762 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2764 SimGrid (3.2) stable; urgency=high
2768 We still experience issues on this platform, but we believe that at
2769 least MSG is usable.
2771 GRAS API BREAKAGE (for simplification purpose, sorry):
2772 * the gras_msgtype_by_name is not used anymore. Instead of
2773 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2774 you can write (and must)
2775 gras_msg_send(toserver, "request", &request);
2776 - If you still want to pass a gras_msgtype_t to the function (to cache
2777 the type and avoid the lookup time), use the gras_msg_send_() variant.
2778 - Impacted functions:
2779 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2780 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2781 * The callbacks are now expected to return 0 when everything went well
2782 (just like the main() function)
2784 GRAS new features and improvements:
2785 * New module mechanism where user code can use per process globals [Mt]
2786 This is similar to gras_userdata_*() functions, but for libraries. It
2787 factorize some code developped over and over in the examples and AMOK.
2788 It has still to be documented and used (only amok/peermanagement is
2790 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2791 loss when they were small enough to fit into the buffer and sent quickly
2792 enough so that they can all get received in one shoot.
2793 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2794 if not found. Use the *_or_null() variant for the old semantic.
2795 * In gras_msg_handle, do not discard messages without callback.
2796 They are probably messages to be explicitly awaited later (ie, proofs of
2797 mis-synchronization in userland since they are sent before being awaited)
2798 No big deal usually.
2799 * gras_socket_meas_send/recv: semantic changed!
2800 The numerical arguments used to be (1) the total amount of data to send
2801 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2802 messages. This was need for the fool willing to send more than MAXINT
2803 bytes on quite fat pipes.
2806 * Do really rename the hostmanagement module to peermanagement. [Mt]
2807 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2808 but this is rather new and this was documented in the module
2809 documentation (poor excuses, I admit)
2810 * Bandwidth measurement semantic changed! This follows the changes to
2811 gras_socket_meas_send/recv explained above.
2814 * A sequential mode has been added to the workstations. When a workstation
2815 is in sequential mode, it can execute only one task, and the other tasks
2816 are waiting in a FIFO. [Christophe Thiery]
2819 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2820 model for SIMDAG. [Christophe Thiery]
2821 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2822 equal to 0, which caused some bad behaviors in
2823 saturated_constraint_set_update. I now use a threshold mechanism like in
2827 * When running manually src/testall, you select specific units [Mt]
2828 testall is the result of our cunit mechanism, and should replace all
2829 the scripty thingy around since bash don't run easily on billware.
2831 * A mallocator system has been added. [Christophe Thiery]
2832 Mallocators allow you to recycle your unused objects instead of freeing them
2833 and allocating new ones.
2835 Documentation update:
2836 * FAQ reworking + New FAQs:
2837 - "Valgrind spits tons of errors!" [Mt]
2838 - "How to repport bugs" [Mt]
2839 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2840 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2841 - Communication time measurement within MSG [AL]
2842 - I experience weird communication times when I change the latency [AL]
2843 * GRAS tutorial [Mt]
2845 - an introduction to the framework and to the used communication model
2846 - an initiatic tour introducing the most proheminent features:
2848 . Lesson 0: Installing GRAS
2849 . Lesson 1: Setting up your own project
2850 o Part 2: Message passing
2851 . Lesson 2: Exchanging simple messages
2852 . Lesson 3: Passing arguments to the processes (in SG)
2853 . Lesson 4: Attaching callbacks to messages
2854 . Lesson 5: Using globals in processes
2855 . Lesson 6: Logging information properly
2856 . Lesson 7: Using internal timers
2857 . Lesson 8: Handling errors through exceptions
2858 . Lesson 9: Exchanging simple data
2859 . Lesson 10: Remote Procedure Calling (RPC)
2860 . Lesson 11: Explicitely waiting for messages
2861 . Recapping of message passing features in GRAS
2862 - A HOWTO section containing:
2863 o HOWTO design a GRAS application
2864 More are due, of course. They will come latter. In the meanwhile, you can
2865 check the examples which are still here.
2867 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2869 SimGrid (3.1) stable; urgency=high
2873 There was a stack corruption somewhere, visible only when optimizing
2874 with these versions. [Vince]
2877 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2878 module that enabled to play with Directed Acyclic Graphs. It is built
2879 directly on top of SURF and provides an API rather close to the old
2880 SG. Some old codes using SG are currently under rewrite to check that
2881 all needful functions are provided. [Christophe Thiery]
2884 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2885 extension of the classical CLM03 model that gracefully handles
2886 failures. This is now the default model for MSG and GRAS. It doesn't
2887 handle parallel tasks yet though. [AL]
2888 * Bug fix: Weights were not correctly set in the network part.
2889 WARNING: This may have resulted in incorrect results with simulations
2890 where there are more than one flow on a given link. [AL]
2893 * After a (long ?) discussion on simgrid-devel, we have decided that the
2894 convention we had on units was stupid. That is why it has been decided
2895 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2896 WARNING: This means that all previous platform files will not work as
2897 such with this version! A warning is issued to ask users to update
2899 A conversion script can be found in the contrib module of the CVS, under
2900 the name contrib/platform_generation/surfxml_update.pl [MQ]
2903 * Bug fix: Processes were started in reverse order, wrt deployment file.
2904 WARNING: if your code relies on this bug, please fix it. [AL]
2905 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2906 being executed on two different locations. [AL]
2907 * Bug fix: Failures are now better supported thanks to Derrick's tests
2908 (there was many failure situations I hadn't thought of and that weren't
2909 correctly handled). [AL]
2910 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2914 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2915 Exception occurring on server-side are propagated back to client (!).
2917 API CHANGE: the callback changed their prototype. Change:
2918 int my_handler(gras_socket_t expeditor, void *payload_data) {
2920 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2921 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2923 * New! function: gras_msg_handleall to deal with all messages arriving
2924 within a given period.
2925 * New! function: gras_socket_server_range to get a server socket in a
2926 range of port numbers (ease to avoid port number conflicts) [MQ]
2927 * New! gras processes display their backtrace when they get a SIGUSR1
2928 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2929 Sweet to debug RL processes [MQ]
2933 - Do not force experiment sizes to be expressed in kb, or it becomes
2934 impossible to measure the latency this way (needs one byte-long tests)
2935 WARNING: this changes the amok_bw_* function semantic. [MQ]
2936 - Implements the link saturation stuff. [MQ]
2937 * Peer management module:
2938 New! module factorizing code that we wrote over and over [MQ].
2941 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2942 - Test units are placed directly into the library code, they get extracted
2943 automatically and placed into the src/testall binary.
2944 - Convert most of the XBT tests to this system.
2945 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2946 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2947 live in the same C code. This required to change a little bit the API
2948 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2949 * New module: sparse graph structure with basic algorithms (this is work
2950 in progress and the API is not considered to be frozen yet). [AL]
2951 * Display more information on backtraces: source line & function names are
2952 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2953 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2954 * Reworked a little bit some #include statements to load only required
2955 headers. Some user code that relied on SimGrid to include stdlib or
2956 stdio may need to include it by themselves. [AL]
2957 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2959 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2960 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2961 but I guess that almost nobody used those parts. [MQ]
2963 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2965 SimGrid (3.0.1) stable; urgency=low
2968 * Unfortunately, I had missed 5 misnamed functions:
2969 xbt_fifo_item_t xbt_fifo_newitem(void);
2970 void xbt_fifo_freeitem(xbt_fifo_item_t);
2971 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2972 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2973 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2974 They're now deprecated. Please use their new versions:
2975 xbt_fifo_item_t xbt_fifo_new_item(void);
2976 void xbt_fifo_free_item(xbt_fifo_item_t);
2977 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2978 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2979 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2981 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2982 * Documentation: xbt_log module unmercifully reworked [MQ]
2983 * Bugfix: there was a problem with the ending of contexts with
2984 the pthread backend. It caused some weird deadlock or behavior
2985 depending on the pthread implementation. [AL]
2986 * Bugfix: get the exceptions raised in the simulator repport where
2987 and why they come from when they are not catched in time [AL, MQ]
2990 * Bugfix: Do repport the error when two non-connected hosts try to
2991 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2994 * Add additionnal checkings on communications. Assert that two
2995 communicating hosts are connected by a set of links... [AL]
2998 * Add additionnal checkings on channel values in communication [AL]
2999 * New: MSG_task_get_source to see on which host a task was generated [HC]
3000 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3001 the number of tasks waiting to be received on channel and sent
3003 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3004 waits for the first task coming from a given host.. [AL]
3006 GRAS new functionnalities: [MQ]
3007 * Enhance the parsing macro to allow the size of multidimentional objects
3008 to be given thru annotations.
3009 * New example (and documentation): Matrix Multiplication a la RPC
3010 (as when I was young!) and fix a bunch of bugs found on the way.
3012 GRAS performance improvements: [MQ]
3014 * Reduce the amount of cbps creation/destruction by making it static to
3015 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3018 * Change libdata to a set so that we can search for stuff by ID (and thus
3019 reduce the insane amount of dict lookups)
3022 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3023 RL communication on top of SG since it's so uneffective.
3024 It may also be used for inter-thread communication in RL, one day.
3025 * Use gras_datadesc_copy() to exchange messages on top of SG
3027 - improve message exchange performance on top of SG
3028 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3029 * Don't exchange on the network the size of the used part of buffer,
3030 instead, specify the possible buffer size to read().
3032 - reduces the amount of read/write calls (one pair per exchange)
3033 - reduces the amount of exchanged data (the size)
3034 - allows to retrieve all arrived data on receiver side, if we don't need
3035 it right now (subsequent read will peek the buffer)
3036 - allows the receiver to proceed with the begining of the stream before
3037 everything is arrived
3038 - make it possible to build an iov transport (using readv/writev)
3040 - take care of the data with non-stable storage (like stacked data),
3042 * If possible, TCP send uses vector I/O (when writev() is here)
3043 - Don't use it for receive since we send data sizes and data on the
3044 same stream, so we wouldn't be able to chain large amount of chunks
3045 before having to flush the stuff to read the size.
3046 * Rework the transport plugin mechanism to simplify it and reduce the
3047 amount of pointer dereferencement when searching for the right function
3050 * I guess that now, we do almost as few system calls as possible while
3051 doing as few data copy as possible.
3053 To improve it further, we could try to send all the sizes first and then
3054 all the data (to use iov on receiving size), but it's only a partial
3055 solution: when you have 2 dimensional data, the sizes of the second
3056 dimension is data of the first dimension, so you need 3 streams.
3058 I'm not sure the potential performance gains justify the coding burden.
3060 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3062 SimGrid (3.00) stable; urgency=high
3065 * New! Give the possibility to hijack the surf parser and thus bypass
3066 MSG_create_environment and MSG_launch_application. Have a look at
3067 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3069 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3071 SimGrid (2.96) unstable; urgency=low
3076 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3077 This deprecates the xbt_error_t mechanisms.
3078 It modifies (simplifies) all XBT and GRAS API.
3079 MSG API keeps unchanged (exceptions raised by XBT are catched from
3080 within MSG and masked with existing error handling facilities)
3083 * New! Add a FATPIPE model. [AL]
3084 * New! Add a parallel task model. [AL]
3085 * New! Add automatically a loopback interface (in the default
3086 network model) if none was precised.
3089 * Bugfix: MSG_process_resume now works with the current running process.
3091 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3092 * Modification of MSG_task_get_compute_duration. Once a task has been
3093 processed, the value returned by this function is now equal to 0. [AL]
3094 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3095 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3096 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3097 of a m_task. For now, MSG_task_cancel only works with computation
3099 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3100 of the processor (in Mflop/s) regardless of the current load on the
3102 * API Change: use proper naming convention for MSG_getClock and
3103 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3105 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3106 This function changes the priority of a computation task. This priority
3107 doesn't affect the transfer rate. A priority of 2 will make a task
3108 receive two times more cpu power than the other ones. This function
3109 has been added to suit the needs of Nguyen The Loc and hasn't been that
3110 much tested yet. So if it fails, please report it and send me your code.
3112 * API Change: removed all functions and types that were marked "deprecated"
3113 since many months. Renamed MSG_global_init_args to MSG_global_init.
3115 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3117 SimGrid (2.95) unstable; urgency=low
3120 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3121 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3122 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3125 * Bugfix: Allow absolute paths for platform description files [MQ]
3126 * Bugfix: do free the variables after use. Leads to drastic performance
3128 * Implement max_duration (ie, timeouts) on resources [AL]
3131 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3133 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3134 the needed support (timeouts) [AL]
3137 * Implement measurement sockets. You can now get the bandwidth between two
3138 hosts thanks to AMOK (see below). [MQ]
3139 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3140 dynar over the network (yeah) [MQ]
3141 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3142 * simple/static token-ring example. [Alexandre Colucci and MQ]
3143 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3144 in place before [MQ]
3147 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3148 type parsing. "short" and "long" modifiers were also ignored; other
3149 modifier (such as reference level) are still ignored. [MQ]
3150 * Embeed the buffer size within the buffer itself on SG. [MQ]
3151 That way, send() are atomic and cannot get intermixed anymore (at least
3152 the ones which are less than 100k; bigger messages still have the issue)
3153 * Array size pushed by the field, not by the field type (or each
3154 and every long int will push stuff to the cbps) [MQ]
3155 * use select() to sleep since it allows to portably sleep less than one
3158 GRAS (minor cleanups)
3159 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3162 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3163 * type category 'ignored' killed as it was never used and were difficult
3165 * whether a type can cycle or not is now a flag, leaving room for more
3166 flags (as "ignored", if we feel the need one day ;)
3167 * Rename raw sockets to measurement sockets since "raw" has another
3168 meaning in networking community.
3171 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3172 and offers features not belonging to GRAS but that most applications
3173 need. One day, it may be a set of plugins loadable at runtime.
3174 * New module: bandwidth
3175 bandwidth measurement between arbitrary nodes running this module. [MQ]
3177 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3179 SimGrid (2.94) unstable; urgency=low
3181 The first beta release of SimGrid 3 !
3185 * Update the main page and the FAQ. Adding references to gforge.
3188 * Add a gras_os_getpid function.
3191 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3192 * Extend the logs so that they also print PID, hostname, date, ... if
3194 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3195 and kill the old version which were in testsuite/
3196 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3200 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3201 for users. Instead, each programming environment (GRAS, MSG,...) use it
3202 when needed to provide such a feature to users.
3203 Don't shortcut the mechanism or you will also shortcut the virtualization
3204 you need on the simulator.
3208 * Cleanups in configury with regard to compile optimization/warning flags.
3209 Also add -fno-loop-optimize to any powerpc since it's the optimization
3210 killing gcc (< 3.4.0).
3211 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3212 needed by MSG examples complications
3213 * Borrow configury beautifications from PHP
3216 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3217 warning (thanks loris for stumbling into this one).
3218 * Bugfix: stop loading private headers (gras_config.h) from the public
3222 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3223 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3224 RL would imply a DNS resolver, which is *hard* to do in a portable way
3225 (and therefore delayed).
3226 * Implement a real timer mechanism and use it in timing macros. This allows
3227 to avoid rounding errors and get a 0.000005 sec precision in timing
3228 macros. While I was at it, various cleanups:
3229 - allow to declare more than one timed section per file (fix a stupid bug)
3230 - move some private declaration to the right place
3231 - merge conditional execution and timing macros into emulation module
3232 - document the module
3233 - make sure the module cleanups its mess on gras_exit
3234 * Documentation improvements:
3235 - (new) how to compile applications using GRAS
3236 - (new) emulation support (timing macros)
3238 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3240 SimGrid (2.93) unstable; urgency=low
3242 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3245 - Use Paje properly where used. Still to be sanitized properly.
3246 - Portability fix: Add an implementation of the contexts using pthread
3250 - Add xbt_procname(): returns the name of the current process.
3251 Use it to show the current process's name in all logging.
3253 - fix detection of older flex version and the reaction, since we do
3254 depend on modern ones (we use lex_destroy)
3255 - Better separation of SG and RL in the libs: remove all simulation code
3256 from libgras. As a result, this lib is now only 200k when stripped.
3257 Some of the xbt modules may also be duplicated (two sets and such) and
3258 should be cleaned/killed before SG3.
3259 - Insist on using xlC on AIX because of weird problems involving gcc there.
3260 - Cleanup the make remote stuff. This is now done by scripts
3261 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3262 mainly for our private use, but we're working on changing them to user
3265 - Bugfix: flush the socket on close only if there is some *output*.
3266 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3268 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3271 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3273 SimGrid (2.92) unstable; urgency=low
3275 Alpha 3 on the path to SimGrid 3
3279 - New! First try of benchmarking macros.
3280 - New! First try so that gras_stub_generator generate deployment and
3281 remote compilation helpers.
3283 - Bugfix: Initialization fix in msg_test.
3287 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3289 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3291 - API Change: gras_os_sleep to take the amount of seconds as a double.
3292 Accepting an int was error prone since it was the only location where
3293 seconds were coded as such. It leaded to damn rounding errors.
3294 - Bugfix: Hard to belive that timers ever worked before this.
3296 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3298 SimGrid (2.91) unstable; urgency=low
3300 Alpha 2 on the path to SimGrid 3
3304 - Bug fix in the lmm_solver.
3306 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3307 through the function MSG_paje_output.
3308 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3309 - It is possible to bound the rate of a communication in MSG with
3310 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3311 to write it in the changelog).
3312 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3317 - Various cleanups to the autotools stuff
3318 - Begin to move Gras examples to examples/gras/
3319 - Let make distcheck work again (yeah!)
3321 - documentation overhauled using doxygen.
3322 gtk-doc-tools is dead in SimGrid now.
3323 - Automatically extract all existing logging categories, and add the list
3324 to the documentation (long standing one, to say the less)
3326 - Cleanup the known architecture table. Reorder the entries to group what
3327 should be, and use a more consistent naming scheme.
3328 (some of the test dataset are still to be regenerated)
3329 - New! Allow library to register globals on each process just as userdata
3331 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3332 lookup time (for now).
3333 Use it in msg and trp.
3334 This cleans a lot the internals and helps enforcing privacy of the
3335 headers between the gras components.
3336 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3337 - Bugfix: gras_os_time was delirious in RL.
3338 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3340 - Reenable GRAS now that it works.
3342 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3344 SimGrid (2.90) unstable; urgency=low
3346 Alpha 1 on the path to SimGrid 3
3348 * It is a long time since the last release of SimGrid. I'm sorry about
3349 that but as I had told you, I was rewriting a lot of things. I apologize
3350 to those who had been reporting bugs to me and that I had not answered.
3351 If your bug is still in the new version, please tell me. Here is a
3352 summary of the main changes.
3354 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3355 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3356 portability, speed, and a lot more but you'll figure it out later.
3357 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3358 I wanted to release the new MSG as soon as possible and I have
3359 broken GRAS, which is the reason why, for now, only MSG is fully
3360 functional. A laconic description of these projects is available
3361 in the documentation.
3363 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3364 kernel optimized for our needs (called SURF but only the developers
3365 should use it). Hence, MSG is now roughly 30 times faster and I think
3366 that by rewriting a little bit MSG, I could event speed it up a little
3367 bit more. Beside the gain in speed, it is also much easier to encode a
3368 new platform model with SURF than it was with SG. More to come...
3370 * REVOLUTION 3: I have tried to change a little as possible the API of
3371 MSG but a few things really had to disappear. The main differences
3372 with the previous version are:
3373 1) no more m_links_t and the corresponding functions. Platforms are
3374 directly read from a XML description and cannot be hard-coded
3375 anymore. The same format is used for application deployment
3376 description. The new format is described in the documentation.
3377 Have a look in tools/platform_generation. There is a tiny script
3378 that converts from the old platform format to the new one. Concerning
3379 the application deployment format, parsing the old one is tricky.
3380 I think most of you should however be able to convert your files. If
3381 it is really an issue, I can write a C code that does the conversion.
3383 2) the toolbox tbx does not exist anymore. We now have a library
3384 with much more data-structures but without the hash-tables (we have
3385 dictionaries that are much faster).
3387 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3389 *****************************************************************************
3390 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3391 * don't feel like converting the oldies. (Mt) *
3392 *****************************************************************************
3395 Version 2.90: "the long awaited one"
3396 - Finished rewriting and debugging MSG. Rewrote the documentation.
3397 - disable GRAS for now since it needs to be ported to the newest SG
3400 - Finish the port to windows (using mingw32 for cross-compile)
3403 - Main loop and datastructures of SURF. A cpu resource object is
3404 functional. Surf can thus be used to create cpu's with variable
3405 performance on which you can execute some actions.
3407 2004-11-15 Martin Quinson
3408 - Port to ARM. Simply added the alignment and size descriptions. Should
3409 work, but the ARM machines are so slow that I didn't had the opportunity
3410 to 'make check' over there yet.
3412 2004-11-15 Arnaud Legrand
3413 - Trace manager now written. It uses a heap structure and is therefore
3414 expected to be efficient. It may however be speeded up (particularly
3415 when many events occur at the same date) by using red and black
3416 trees. One day maybe...
3417 - Max-min linear system solver written. It uses a sparse matrix
3418 structure taking advantage of its expected use. Most operations are
3419 O(1) and free/calloc are called as few as possible. The computation of
3420 the minimum could however be improved by using a red and black tree
3423 2004-11-03 Arnaud Legrand
3424 - Rename every gras_* function that was in xbt/ to its xbt_
3426 - Add a heap and a doubly-linked list to xbt
3427 - Added a dichotomy to the dictionaries. make check works as well before
3428 so I assume that the patch is correct. I do not know however if things
3429 run effectively faster than before now. :)
3431 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3432 SimGrid, and the version number is bumped to 2.x
3434 2004-10-29 Martin Quinson
3435 - Introduction of the remote errors.
3436 They are the result of a RMI/RPC on the remote machine.
3437 ErrCodes being scalar values, you can't get the host on which those
3438 errors did happen. Extending the error mechanism as in Gnome is possible.
3439 No idea yet whether it is a good idea.
3441 2004-10-28 Martin Quinson
3442 - Interface revolution: the Starred Structure Eradication.
3443 I used to do typedef struct {} toto_t; and then handle *toto_t.
3444 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3445 - ???_t is a valid type (builded with typedef)
3446 - s_toto_t is a structure (access to fields with .)
3447 - s_toto is a structure needing 'struct' keyword to be used
3448 - e_toto_t is an enum
3449 - toto_t is an 'object' (struct*)
3451 typedef struct s_toto {} s_toto_t, *toto_t;
3452 typedef enum {} e_toto_t;
3453 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3454 s_toto_t) is private.
3456 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3457 so that it can set the variable to NULL. It was so for dicts and sets,
3458 it changed for dynars.
3460 - Fix a bunch of memleaks in dict_remove
3461 - Fix a bug in sg/server_socket opening: it failed all the time.
3463 2004-10-07 Martin Quinson
3464 - Speed up dynar lookup operation a bit.
3466 gras_dynar_get is dead.
3468 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3469 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3470 (which gives you the address of the stored data).
3472 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3473 copy of the data using an affectation to do the job and not a memcpy.
3475 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3476 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3478 It does not really speedup the dynar test because they are
3479 setting elements all the time (and look them seldom). But the dict does
3480 far more lookup than setting.
3482 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3484 2004-10-05 Martin Quinson
3485 - Allow to (en/dis)able the cycle detection at run time.
3487 Whether we should check for cycle or not is now a property of each
3488 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3489 datadesc_cycle_unset allow to remove this property when previously set.
3491 Note that the cycle detection is off by default since it impacts the
3492 performance. Watch the data you feed GRAS with ;)
3494 This property is hereditary. Any element embedded in a structure having it
3495 set have it set for the time of this data exchange.
3497 You should set it both on sender and receiver side. If you don't set it on
3498 sender side, it will enter an endless loop. If you forget on receiver
3499 side, the cycles won't be recreated after communication.
3501 - Header reorganization.
3502 Kill gras_private.h, each submodule must load the headers it needs.
3504 2004-10-04 Martin Quinson
3505 - Interface revolution: do not try to survive to malloc failure.
3507 Now, gras_malloc and friends call gras_abort() on failure.
3508 As a conclusion, malloc_error is not a valid error anymore, and all
3509 functions for which it was the only gras_error_t return value are
3510 changed. They now return void, or there result directly.
3511 This simplify the API a lot.
3513 2004-09-29 Martin Quinson
3514 - Re-enable raw sockets.
3515 Created by gras_socket_{client,server}_ext;
3516 Used with gras_raw_{send,recv}
3519 It should allow to kill the last bits of gras first version soon.
3521 This is not completely satisfactory yet (duplicate code with
3522 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3525 - Simplify transport plugin (internal) interface by not passing any
3526 argument to _server and _client, but embedding them in the socket
3529 2004-09-28 Martin Quinson
3530 - Finish the port to AIX.
3531 autoconf was my problem (segfault within the malloc replacement
3532 function. No idea why)
3534 2004-09-16 Martin Quinson
3535 - Fix some size_t madness on 64bit architectures.
3537 2004-09-08 Martin Quinson
3538 - Reduce the number of system headers loaded, overload some more system
3539 calls (such as malloc to cast the result of the system one, and work
3541 - Fix and reintroduce the config support
3543 2004-09-07 Martin Quinson
3544 - Source code reorganization to allow Arnaud to surf all over there.
3545 - Allow to document the logging categories.
3546 - Remove all uppercase from logging categories and useless cleanup in names.
3548 2004-08-18 Martin Quinson
3549 Version 0.6.2 (protocol not changed; API changed)
3550 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3551 gras_error_t), and NULL when not found. Functions expecting a msgtype
3552 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3553 hopefully usefull message.
3554 - Portability to prehistoric sparcs again
3556 2004-08-17 Martin Quinson
3557 Version 0.6.1 (protocol not changed; ABI not changed)
3558 - prealloc some buffers to speed things up
3560 2004-08-11 Martin Quinson
3561 Version 0.6 (protocol not changed; ABI expended)
3562 - The parsing macro can deal with the references, provided that you add
3563 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3565 2004-08-09 Martin Quinson
3566 Version 0.5 (protocol not changed; ABI changed)
3567 - Allow to off turn the cycle detection code in data exchange at
3568 compilation time. It should be at run time, but I'm short of time (and
3569 the config stuff is still broken). That way, we keep dict out of the
3570 critical path, which is good because the performance is poor:
3571 - search not dichotomial yet
3572 - dynar give no way to access their content and memcpy everytime
3573 - In composed data description (struct, ref and so on), stop foolness of
3574 keeping the subtype's ID, but store the type itself. This keeps sets out
3575 of the critical path, which is good since they rely on dynar and
3576 dictionnaries. The only loose of that is that we cannot detect the
3577 redeclaration of a structure/union with another content (but I'm not sure
3578 the code detected well this error before anyway). We still can detect
3579 the redefinition discrepancy for the other types.
3580 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3581 it breaks the code because of type-punning used all over the place).
3582 This breaks on all non-gcc architectures (for now).
3584 All those changes (plus the buffer of last time) allow me to gain 2 order
3585 of magnitude on cruel tests consisting of 800000 array of integers on two
3586 level of a hierarchical structure (200 secondes -> 4 secondes)
3589 - the selector of reference must now return the type it points to, not
3590 the ID of this type.
3592 2004-08-06 Martin Quinson
3593 Version 0.4 (protocol changed; ABI not changed)
3594 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3595 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3596 - Implement a buffer transport, and use it by default (it relies on tcp in
3597 real life and on sg in simulation).
3598 That's a bit hackish since I had a new field to the structure to store
3599 its data without interfering with the subtype ones. Inheritance
3600 is tricky in C. And that's a kind of reverse inheritance with one class
3601 derivating two classes. Or maybe a game with java interfaces. Anyway,
3602 that's damn hard in C (at least).
3603 Moreover, I got tired while trying to ensure plugin separation and
3604 genericity in SG mode. MSG wants me to do weird things, so let's go for
3605 cruel hacks (temporarily of course ;).
3606 See comment in transport_private.h:71
3607 - do not include all the _interface headers in private but in the files
3608 which really need them (to cut the compilation time when they are
3611 2004-07-26 Martin Quinson
3612 Version 0.3 (protocol not changed; ABI changed)
3613 - Major overhault of the datadesc interface to simplify it:
3614 - shorted the function names:
3615 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3616 - add a trivial way to push/pop integers into the cbps without malloc.
3617 This allows to make really generic sub_type description, which simply
3618 pop their size of the stack.
3619 - add a function gras_datadesc_ref_pop_arr() which does what users want
3620 most of the time: Declare a dynamic array (which pops its size of the
3621 stack) and declare a reference to it. Poor name, but anyway.
3622 - kill the post-send callback, add a post-receive one
3624 2004-07-23 Martin Quinson
3625 Version 0.2 (protocol changed; ABI changed)
3626 - add some testing for cpbs in the test cases, and fix some more bugs.
3627 This invalidate again the little64 data file, since I cannot regenerate
3629 - remove an awfull optimization in the logging stuff, allowing me to:
3630 - understand it again
3631 - learn gcc how to check that the argument match the provided format
3632 - fix all errors revealed by gcc after that
3633 - internal keys of dict are not \0 terminated. Deal with it properly in
3634 loggings instead of segfaulting when the user want to see the logs :-/
3636 2004-07-22 Martin Quinson
3637 - Fix some stupid bug preventing cbps (callback postit) from working
3639 2004-07-21 Martin Quinson
3640 - Some documentation cleanups
3641 - remove the useless last argument of msgtype_declare
3642 - rename the Virtu functions to fit into the 'os' namespace
3643 - move headers src/include -> src/include/gras/ and stop fooling with
3644 gras -> . symbolic link
3645 - make distcheck is now successful
3647 2004-07-19 Martin Quinson
3649 - Build shared library also
3650 - Install html doc to the right location
3651 - stop removing maintainer files in make clean
3652 - build tests only on make check
3654 2004-07-13 Martin Quinson
3656 - No major issue in previous version => change versioning schema
3657 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3658 corresponding dataset.
3660 2004-07-11 Martin Quinson
3662 - Get it working with any kind of structure (we can compute the padding
3663 bytes remotely for all the architectures I have access to)
3664 - Implement the structure parsing macro (still not quite robust/complete)
3665 - Improvement to the remote testing toysuite
3667 2004-07-10 Martin Quinson
3668 [autoconf mechanism]
3669 - get ride of a bunch of deprecated macros
3670 - actually run the test for two-compliment, not only compile it :-/
3671 - test whether the structures get packed (and bail out if yes. Damn.
3672 Alignment is a serious matter)
3673 - test whether the structures get compacted (but respecting the alignment
3674 constraints of each types)
3675 - test whether the array fields of structures can straddle alignment boundaries
3677 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3678 (took me 2 hours to find that bug, looking at the wrong place)
3679 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3680 before sending/receiving and are used to compute the offsets of fields
3681 - Given that padding size depend even on compiler options, keep track of
3682 alignment and aligned_size only for the current architecture. Not a big
3683 deal since we send structure fields one after the other (seems
3685 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3686 program, let it work on linux/gcc/little32. portability todo.
3688 2004-07-08 Martin Quinson
3689 - import and improve remote compilation support from FAST
3690 - make sure make check works on half a dozen of machines out there
3692 2004-07-07 Martin Quinson
3693 Let's say it's version 0.0.3 ;)
3694 - Implement conversions (yuhu!)
3695 - Let it work on solaris (beside conversion, of course)
3696 - Stupid me, using rand() to generate the conversion datatests in not wise.
3698 2004-07-06 Martin Quinson
3699 - Let make dist work, since I'm gonna need it to compile on remote hosts
3700 - Let Tests/datadesc_usage write the architecture on which the file was
3701 generated as first byte.
3702 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3703 architecture descriptions.
3704 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3705 on those architectures.
3706 - Optimization: send/recv array of scalar in one shoot
3708 2004-07-05 Martin Quinson
3709 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3711 - Plug a whole bunch of memleaks
3712 - each process now have to call gras_{init,exit}. One day, their log
3713 settings will be separated
3714 - Continue the code factorisation between SG, RL and common in Transport.
3716 2004-07-04 Martin Quinson
3718 - Redistribution between SG and RL.
3719 We wanna have to accept in SG, so move accepted related parts of RL in
3720 the common part. (more precisely, the dynar of all known sockets is no
3721 more a static in transport.c, but part of the process_data)
3723 [gras_stub_generator]
3724 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3726 2004-07-03 Martin Quinson
3727 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3728 them all up in one shot)
3730 - Fix the ugly case of reference to dynamic array.
3731 - New (semi-public) function gras_datadesc_size to allow the messaging
3732 layer to malloc the needed space for the buffer.
3734 - gras_socket_close now expect the socket to close (and not its address to
3735 put NULL in it after it). This is because the socket passed to handlers
3736 is one of their argument (=> not writable).
3738 - propagate the interface cleanup from last week in datadesc, ie remove a
3739 superfluous level of indirection. User pass adress of variable
3740 containing data (both when sending and receiving), and not of a variable
3741 being a pointer to the data. Let's say that I like it better ;)
3742 The price for that is constructs like "int msg=*(int*)payload" in
3743 handlers, but it's a fine price, IMHO.
3745 - Let it work in RL (yuhu)
3747 2004-06-21 Martin Quinson
3749 - porting SG plugin and SG select to new standards (works almost).
3750 - plug memleaks and fix bugs around.
3753 - cleanup the prototype of data recv and force users to specify when they
3754 want to handle references to objects. Test case working even for cycles.
3755 - plug memleaks. Valgrind is perfectly ok with this.
3757 2004-06-12 Martin Quinson
3759 - cleanup the separation between plugin and main code in plugin creation
3761 2004-06-11 Martin Quinson
3763 - Reput hook for raw sockets, needed for BW experiments
3764 - kill a few lines of dead code
3765 [Data description] Interface cleanup
3766 - gras_datadesc_by_name returns the searched type or NULL.
3767 That way, no variable is needed to use a type desc once, which makes
3769 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3770 two parameters were strings (field name, type name), leading to
3772 [Dicos] Interface cleanup
3773 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3774 This is consistant with the dynar API.
3776 2004-04-21 Martin Quinson
3778 - Porting to new standards.
3780 - interface cleanup.
3781 There is no bag anymore, no need to take extra provision to mask the
3782 pointers behind "ID".
3783 Better splitup of functions between files create/exchange/convert.
3784 This is still a bit artificial since convert and receive are so
3785 interleaved, but anyway.
3787 - add a queued message list to procdata (the ones not matching criteria
3789 - factorize some more code between SG and RL wrt procdata
3791 - use gras_exit in example to track memleaks
3792 - get rid of gs_example now that GS is properly integrated into gras
3793 - update run_test to integrate the latest tests (datadesc)
3795 - rename WARNINGn macros to WARNn since it prooved error-prone
3797 2004-04-19 Martin Quinson
3799 - register init/exit functions within gras module mechanism
3800 - send/receive function.
3801 Convertion is not implemented, but short-cutted if not needed.
3802 struct/array elements are sent one by one (instead of block-wise), but
3803 nobody really cares (yet). Get a prototype before optimizing.
3804 - tests (using a file socket) for DD send/receive on:
3805 - base types: int, float
3806 - array: fixed size, string (ie ref to dynamic string)
3807 - structure: homogeneous, heterogeneous
3808 - chained list, graph with cycle
3809 Believe it or not, valgrind is not too unhappy with the results. The
3810 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3814 - Bugfix when using a filename explicitely (instead of '-')
3816 2004-04-09 Martin Quinson
3818 - factorize more code between RL and SG in socket creation
3819 - Complete the implementation and tests of:
3821 o file (only in RL, and mainly for debugging)
3823 I lost 3 days to design a portable address resolver, and then decided
3824 that the prototype mainly have to run on my box.
3825 Addressing portability too early may be like optimizing too early :-/
3827 - use gras_init in the Tests instead of the crappy parse_log_opt
3828 (the latter function is removed)
3829 [Conditional execution]
3830 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3831 [Code reorganisation]
3832 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3833 Build examples against the RL library, since there is no way to disable
3834 its creation for now.
3836 For information, the beginning of coding on GRAS was back in june
3837 2003. I guess that every line has been rewritten at least twice since