1 SimGrid (3.22) NOT RELEASED (Release Target: December 21. 2018, 22:23 UTC)
6 - We are excited to introduce the SimGrid/S4U interface to your neighborhood
7 - Not complete yet: asynchronous activities (amongst others) are still missing
8 - Still ongoing: the interface may change in the future. We need more
9 testers! Please report any glitches.
10 - No new project using Java should start now. Please switch to Python.
13 - Some of the internal cleanups may lead to speed improvements:
14 - The hard limitation on the amount of simulated actors with
15 Java+Mac was removed. Now, the available memory is the only limit.
16 - Our refcounting was tidyied, leading to 10% speedups in some cases.
17 - We are still working on making our code robust to the actor kills
18 and hosts' churn. Things are improving, but it's not perfect yet.
19 - Replaced our own code to display a backtrace (that was forking addr2line)
20 with the Boost.Stacktrace library.
21 You won't see your backtraces without this optional dependency.
22 - Bump cmake dependency to 3.5 (provided by Ubuntu 16.04).
23 - Stop setting random seed with srand() at initialization.
26 - In <host> and <peer>, 'availability_file' is now 'speed_file'.
27 XML file version remains 4.2 since old files are still compatible.
30 - Process termination which was broken at version 3.21 has been repaired.
31 - Expose host load plugin: loadInit, getCurrentLoad, getComputedFlops, getAvgLoad
32 - Hide the examples into examples/deprecated. New users should use Python.
35 - Drop MSG_process_create_from_stdfunc() from the API.
36 This C++ function was a pimple in the C API, made necessary at some
37 point by the Java bindings. This is fixed now.
38 - Hide the examples into examples/deprecated. New users should use S4U.
39 - MSG_process_create and MSG_process_attach now crash if the host used
40 to run this process is off.
41 - Fix the protype of MSG_process_on_exit()
42 Now use (int,void*) callbacks instead of (void*,void*) ones.
43 The implementation was ways too messy to actually work, I guess.
46 - Change the way SMPI_SAMPLE_* macros work to avoid invalidating cache
47 too often and provide more accurate timings.
48 - Add -gdb, -lldb, and -vgdb shortcuts to help debug MPI codes with smpirun
49 - MPI_Alltoallw support
50 - Partial MPI nonblocking collectives implementation: MPI_Ibcast, MPI_Ibarrier,
51 MPI_Iallgather, MPI_Iallgatherv, MPI_Ialltoall, MPI_Ialltoallv, MPI_Igather,
52 MPI_Igatherv, MPI_Iscatter, MPI_Iscatterv, MPI_Ialltoallw.
53 - MPI_Request_get_status, MPI_Status_set_cancelled, MPI_Status_set_elements
55 - Basic implementation of generalized requests (SMPI doesn't
56 allow MPI_THREAD_MULTIPLE) : MPI_Grequest_complete, MPI_Grequest_start
59 - Drop sg_cmdline. Please use xbt_cmdline instead.
60 - Drop the C xbt_os_mutex_t; Use the C++11 std::mutex.
61 - Drop the C xbt_os_sem_t; Use the C++ xbt::OsSemaphore.
62 OsSemaphore is implemented in a portable way with C++11 threads.
63 This should allow much more threads to be created at the same time,
64 allowing Mac and Java users to simulate many more actors.
65 - Implement the 'thread' factory with std::thread instead of xbt ones.
66 It is not possible to set the stack size with threads anymore, but
67 -fsplit-stack is the way to go nowadays when using threads.
68 - Drop the xbt_os_thread_t module (now unused)
69 - Drop xbt_ex_display(), use simgrid::xbt::log_exception() instead.
70 - Drop xbt_str_join_array().
71 - Drop cunit, use Catch2 instead.
74 - Many cleanups in the kernel::activity namespace. This was long
75 overdue, and shall open the path to many future endeavors.
78 - #132: Java: a process can not shut down its own host
79 - #220: S4U: detached send causes exception if sender terminates soon after sending
80 - #261: Document the parameters of parallel execution's constructor
81 - #300: [s4u] BarrierPtr is missing
82 - #314: SMPI args internal cleanup
83 - #316: Fix a bug related to the CPU utilization of multi-core VM
84 - #318: Invalid trace file when using option --cfg=tracing/smpi/display-sizes:yes
85 - #324: S4U: Attempting to create an actor on turned off host segfaults instead of throwing
86 - #325: Turning off a host has different behavior on sleeping actors and computing actors
88 ----------------------------------------------------------------------------
90 SimGrid (3.21) October 3. 2018
92 The Restarting Documentation (TRD) Release.
95 - Convert the user manual to Sphinx for improved usability.
96 Unstable version now hosted on https://simgrid.frama.io/simgrid
97 Many glitches remain; stay tuned.
98 - Restrict the Doxygen reference API to the installed header files.
99 (documenting all internals this way is overkill + doxygen scales badly)
100 - New tutorial on S4U (highly inspired from the old MSG one, but with
101 a git repository to fork as a starting point, and a docker image)
102 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
103 - The Developper manual is still to be converted (not compiled until then)
104 - Some parts are still missing in this conversion (such as the
105 platform and deployment sections) while others were blindly converted
106 and would need more love (such as the configuration flags).
107 Things will certainly further improve in the future releases.
110 - Fully reimplement auto_restart mechanism that was utterly broken
111 (fix #22, #131 and #281 that were all very related)
112 - Implement semaphores (fix #303 and #291)
113 - When creating an actor from a function and its parameters,
114 move-only parameters are not allowed anymore, as it would prevent
115 the actor to be restartable if its parameters are consumed on run.
116 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
117 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
118 IO operations that can be started, waited for, or canceled as a regular
120 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
124 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
125 - New host variable: 'core_count'
128 - Remove xbt_os_thread_specific features
129 - Remove portability wrapper to condition variables
130 - Remove xbt_os_thread_yield()
133 - MPICH collective selector now mimics MPICH 3.3b
134 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
135 - MPI_Init(NULL,NULL) is now allowed
136 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
137 - Command-line arguments should now be handled properly in Fortran simulations
138 - MPI Attributes and related callbacks should work in Fortran
139 - Apps using C + Fortran should now work
140 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
141 - Support MPI_Error_String
144 - Due to an internal bug, Msg.run() must now be your last line.
145 We hope to fix it in a future release, and we are sorry for the inconvenience.
148 - #22: Process autorestart seem to only work with CAS01 cpus
149 - #56: Feature request: dynamic SMPI replay
150 - #93: simgrid should not eat --help
151 - #111: How to change the loopback link characteristics?
152 - #116: Platform cluster radical
153 - #117: Platform cluster/host order
154 - #129: Replay_multiple on steroid
155 - #131: Java : autostart does not work
156 - #143: Setting a breakpoint at a given time
157 - #235: Network energy models should be integrated
158 - #264: Add ptask L07 resource tracing
159 - #271: Dynamic and manual replay of SMPI traces
160 - #279: Breakpoints option uses time deltas instead of absolute time
161 - #281: Daemonized actor no longer daemonized after an auto-restart
162 - #282: TIT convertor erase traces when using absolute path in the trace list file
163 - #285: segfault when a s4u actor kills itself with msg process tracing activated
164 - #286: Pajé traces are not exposing the number of cores
165 - #287: Command-line parsing should stop when encountering '--'
166 - #288: MPI_Init(NULL, NULL)
167 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
168 - #290: Method to check if a mailbox has a message ready to be consumed
169 - #291: [s4u] semaphores do not seem available yet
170 - #294: Alltoall collective implementation of mpich
171 - #295: Starting a Host as OFF is not supported
172 - #296: DTD too permissive
173 - #299: Add a s4u barrier example
175 ----------------------------------------------------------------------------
177 SimGrid (3.20) Released June 24. 2018
179 The proxy snake_case() release.
182 - The whole API is now using snake_case() instead of mixing it with camelCase.
183 Compatibility wrappers in place for one year.
184 - We took the opportunity to increase the consistency of the API.
185 Please report any remaining glitches.
186 - New function simgrid::s4u::Host::get_actor_count:
187 Returns the number of actors running on a specific host.
190 - SMPI is now tested with ~45 proxy apps from various sources, with none or
191 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
192 - Replay: The replay file has been re-written in C++.
193 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
194 supported. They are stored in the trace and used when replayed.
195 - Basic support of MPI_Cancel. Robustness not guaranteed.
196 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
197 (only for MPI_COMM_TYPE_SHARED).
198 - New option: smpi/privatize-libs, to add external shared libs to be privatized
199 by SMPI. They will be copied locally and loaded separately by each process.
200 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
201 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
202 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
203 with internal index optind with multiple processes. Only works if MPI_Init has
205 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
206 syntax without preprocessing tricks.
209 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
210 behavior tracing is no longer limited to MSG
211 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
212 this means that traces now need to be updated or re-obtained; both Irecv
213 and Isend lines in your traces have now in total 5 mandatory fields:
214 <rankid> <command> <to/from rankid> <tag> <size>
215 To update your traces, it suffices to add a 0 for the tag here.
216 - TIT now also supports waiting for a distinct request via MPI_Wait.
217 Wait/Test now wait for a specific request, not just the last one that was
218 issued. This unfortunately means another update, because we need to
219 identify which request you want to wait for. We do this via the
220 triplet (sender, receiver, tag), which needs to be added:
221 <rankid> <command> <sender> <receiver> <tag>
222 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
223 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
226 - The deprecation of MSG is ongoing (but this should not impact you).
227 Many MSG functions are now simple wrappers on the C API of S4U. If
228 you wish to convert your code to S4U, find the S4U counterparts of
229 your MSG calls in src/msg/msg_legacy.cpp.
230 - MSG can still be used, but won't evolve anymore.
233 - Allow to run the Link energy plugin from the command line with
234 --cfg=plugin:link_energy
235 - Rename Energy plugin into host_energy
236 - Rename Load plugin into host_load
239 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
240 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
241 - Improve performance when handling timeouts of simix synchros.
244 - Config: the C API is now deprecated (will be removed in 3.23), and
245 the C++ API has been slightly improved.
248 - Fix several build issues on OSX.
249 - Move simgrid_config.h to simgrid/config.h (old header still working)
252 - #143: Setting a breakpoint at a given time
253 - #150: Inconsistent event names in SMPI replay
254 - #258: daemonized actors hang after all non-daemonized actors have completed
255 - #267: Linker error on unit_tmgr
256 - #269: SMPI: tracing of MPI_Wait/all/any broken
257 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
259 ----------------------------------------------------------------------------
261 SimGrid (3.19.1) Released March 22 2018
263 The Fixed ABI Release.
265 As you may know, we are currently refactoring SimGrid in deep.
266 Upcoming SimGrid4 will be really different from SimGrid3: modular,
267 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
269 Our goal is to smooth this transition, with backward compatibility
270 and automatic update paths, while still progressing toward SimGrid4.
272 SimGrid remains open during works: The last pure SimGrid3 release was
273 v3.12 while all subsequent versions are usable alpha versions of
274 SimGrid4: Existing interfaces remain unchanged, but the new S4U
275 interface is budding and the internals are deeply reorganized.
277 Since 2015, we work hard to reduce the changes to public APIs. When
278 we need to rename a public library symbol in S4U, we let your
279 compiler issue an explicative warning when you use the deprecated
280 function. These messages remain for four releases, i.e. for one full
281 year, before turning into an error. Starting with v3.15, your can
282 also adapt to API changes with the SIMGRID_VERSION macro, that is
283 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
285 Starting with this v3.19.1, our commitment to reduce the changes to
286 the public interfaces is extended from the API to the ABI: a program
287 using only MSG or SimDag and compiled against a given version of
288 simgrid can probably be used with a later version of SimGrid without
289 recompilation. We will do our best... but don't expect too much of
290 it, that's a really difficult goal during such profund refactoring.
292 The difference between v3.19 and v3.19.1 is that the former was
293 accidentally breaking the ABI of MSG, while the later is restoring
296 S4U and kernel APIs will still evolve until SimGrid4, with one-year
297 deprecation warnings as currently. In fact, cleaning up these
298 interfaces and converting them to snake_case() is one release goal of
299 v3.20. But don't worry, we are working to smooth this upgrade path.
301 In summary, new projects should start with S4U to benefit of the
302 future, but old MSG projects should still be usable with no change.
304 ----------------------------------------------------------------------------
306 SimGrid (3.19) Released March 20 2018
308 The Moscovitly-cold Spring release.
310 SMPI user-visible improvements:
311 - Support of MPI_Type_create_subarray
312 - Unimplemented MPI calls now abort (instead of issuing a warning)
313 Some harmless calls around errhandlers still only issue a warning.
314 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
315 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
316 is now possible (mmap privatization was refactored to this end).
317 Adding new SMPI instances should also be technically doable now.
319 SMPI internal cleanups:
320 - Removed index notion from SMPI -- no more "getPid() - 1"!
321 In tracing files, actors are called rank-<PROCESS-ID> but these are
322 NOT the real ranks (tracing several communicators is not supported yet).
323 - Deprecate smpi_process_index() (will be removed in 3.22)
324 - Better testing on classical proxy apps, through an external project:
325 https://github.com/simgrid/SMPI-proxy-apps/
328 - Use a graphical TOC to make it easier to find the documentation you need
329 - Revamp the MSG tutorial
332 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
333 - Remove parameter of MSG_process_killall().
334 Resetting the PID was bogus anyway (several actors could have the same PID).
337 - Execution->setHost() can be called after start(), to migrate it.
338 - Comm::test_any() is now implemented.
339 - s4u::Actor now has onCreation() and onDestruction() signals.
340 - Install some kernel header files for the users' plugins and more.
341 Warning, their API is really not stable yet.
342 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
343 include/simgrid/kernel/routing/*Zone.hpp
346 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
348 Build System and other cleanups:
349 - Remove unused run-time parameter "tracing/onelink-only".
350 - SimGrid now works with intel compilers (MC still buggy)
351 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
352 - tesh kills subprocesses on timeout (but not on windows)
355 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
358 - #194: Feature request: simgrid::s4u::Comm::test_any()
359 - #245: migrating an actor does not migrate its execution
360 - #253: Feature Request: expose clusters as objects
361 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
362 - #255: Tesh broken on Windows
363 - #256: Modernize FindSimGrid.cmake
364 - #257: Fix (ab)use of CMake install
366 ----------------------------------------------------------------------------
368 SimGrid (3.18) Released December 24 2017
370 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
372 MSG IS NOW DEPRECATED. Please use S4U for new projects.
373 - Support for MSG should not be removed from SimGrid before 2020,
374 but future evolutions will be limited to the integration of
375 user-provided patches.
378 - s4u::onDeadlock() signal, run before stopping the simulation.
379 - s4u::Exec: asynchronous executions (abstraction of background
381 - s4u::Host->getLoad() returns the achieved speed in flops/s
384 - Link::name() is deprecated and will be removed in v3.21, use
385 Link::getCname() instead.
386 - Mailbox::getName() changed to return a std::string, use
387 Mailbox::getCname() instead to get a char*.
388 - Storage::getName() changed to return a std::string, use
389 Storage::getCname() instead to get a char*.
390 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
391 to have both allocation and deallocation of the map in user space.
394 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
395 The amount of flops remaining to do is a vector, not a scalar.
396 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
397 name implies on both sequential and parallel tasks.
398 - Both changes fix GitHub's #223 using PR #237 as a basis.
399 Thanks Michael Mercier.
400 - Most examples were converted to S4U and hidden elsewhere at tests.
401 Rationale: we still want MSG to work; we want newcomers to use S4U.
404 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
405 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
408 - Switch to the faster dlopen privatization mechanism by default
409 - Documentation improvments
412 - Live migration is getting moved to a plugin. Dirty page tracking is
413 the first part of this plugin. This imply that VM migration is now
414 only possible if one this function is called:
415 - C/MSG: MSG_vm_live_migration_plugin_init()
416 - C/C++: sg_vm_live_migration_plugin_init()
417 - Java: Msg.liveMigrationInit()
418 For C and C++, "simgrid/plugins/live_migration.h" also has to be
422 - Define class simgrid::xbt::Path to manage file names.
423 - Removed unused functions:
424 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
425 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
426 xbt_graph_export_graphviz()
427 - xbt/str.h: xbt_str_join()
428 - Remove unused datatypes:
429 - xbt/heap.h: use std::priority_queue or boost::heap instead
430 - xbt/swag.h: use boost::intrusive::list instead
433 - New link_energy plugin for the consumption of the links.
434 - All of the operations on files and storage contents have been
435 packaged into a plugin (src/plugins/file_system). The current
436 public interface can be found in
437 include/simgrid/plugins/file_system.h
438 To use these functions you now have to initialize the plugin by
439 calling MSG_storage_file_system_init() just after calling
440 MSG_init() or sg_storage_file_system_init() just after creating
444 - Remove the undocumented/untested tag <include>
447 - Remove viva specific tracing as the tool is no longer maintained
450 - #248: Kill of finished processes leads to segfault
451 - #240: xbt_cond_wait_timeout should gracefully return for C apps
452 - #239: please implement signal s4u::onDeadlock()
453 - #230: segfaults when exit() without run()
454 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
455 - #223: MSG_task_get_flops_amount() not working with parallel tasks
456 - #222: Actor::kill() doesn't really kill and segfaults
457 - #221: odd LMM warning when killing an actor
458 - #120: Memory leak when the processes are forcefully killed
460 ----------------------------------------------------------------------------
462 SimGrid (3.17) Released October 8 2017
464 The Drained Leaks release: (almost) no known leaks despite the tests.
466 Even very long simulations will preserve your memory: our testsuite
467 stresses SimGrid for over 45mn on fast machines for a coverage of
468 over 80%, and there is only one single known leak, of about 4kb.
472 - Comm.detach(): start and forget about asynchronous emission. A cleanup
473 handler may be given to free resources if the comm cannot be completed.
474 - this_actor::send(mailbox) is now mailbox->put()
475 - New: simgrid::s4u::Comm::wait_all()
476 - New: Host.execute() for remote executions.
479 - Unused option network/sender-gap is removed.
482 - Deprecate MSG_task_isend_with_matching(): This unused feature
483 really complicates our internals. Will be removed in v3.20.
486 - Improved context termination. It is now possible to free resources, even
487 when a process is forcibly killed.
488 - ContextBoost: add support for Boost versions above 1.61.
491 - Bring back run-time option --cfg=exception/cutpath to remove exception
493 - Removed unused functions:
494 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
495 xbt_str_rtrim(), xbt_str_trim().
496 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
499 - Removed header files obsolete since SimGrid 3.12:
500 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
502 - #3: SD_exit should be made optional
503 - #120: Memory leak when the processes are forcefully killed
504 - #159: Threading test regression in Actor refcounting
505 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
506 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
507 - #186: Actor::killAll() segfaults if some process is blocked on wait()
508 - #191: VM migration and pstate
509 - #192: Updating the energy consumptions of all hosts crashes with VMs
510 - #195: All actors have PID=0 in the logs
511 - #204: Sometimes segfault with thread contexts and mmap privatization
512 - #222: Actor::kill() doesn't really kill and segfaults
513 - #225: Actor::kill() doesn't really kill when victims are doing a join()
515 ----------------------------------------------------------------------------
517 SimGrid (3.16) Released June 22. 2017.
519 The Blooming Spring Release: developments are budding.
522 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
523 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
524 - Java JAR file should be finally fully working on Mac OSX too.
526 XML platforms: Switch to platform v4.1 format.
527 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
528 - <zone> can be used as a synonym for the now deprecated <as>
529 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
530 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
531 - <actor> can be used as a synonym for the now deprecated <process>
532 - state_file and avail_file periodicity is now easier to express
533 (check the documentation) the old behavior should still work.
534 * Storage: not backward compatible, but it seems that nobody ever used it
535 - Remove attribute 'content_type' of <storage_type>: was never used
536 - Make attribute 'model' of <storage_type> optional: for future usage
537 - Remove Bconnection model property: was never (in)validated. Replaced by
538 the maximum of the read and write bandwidth as a resource constraint.
541 - New and Backwards Compatibility break:
542 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
543 When one wants to get the list of tasks whose states have changed during a
544 simulation round, s/he has to allocate and free a dynar and use it as
545 argument to this function. The former SD_simulate (double how_long)
549 - Allow multicore VMs, with the correct sharing (unless you overcommit)
550 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
553 - New function to update the consumption of all hosts at once.
554 - Fix the model for multi-core hosts, linear in the amount of busy
555 cores with an abnormality for pIdle (see paper).
556 BUG: Still not satisfactory for ptask on mono-cores.
559 - The netzone are now available from the MSG API.
560 The old names still work, but are now deprecated.
563 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
564 - New option: smpi/keep-temps to not cleanup temp files
565 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
566 mallocs mode is used, allows to change the size of the fake file used
567 (default 1MB), to potentially limit the number of mappings for large runs.
568 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
569 - Fortran ifort and flang compilers support
570 - New RMA calls supported (experimental):
571 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
572 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
573 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
574 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
575 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
576 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
579 - New: this_actor::isMaestro() returns whether we are in kernel mode.
580 - Behind the scene, ActivityImpl are now automatically refcounted.
581 Removing this nasty bottleneck will greatly simplify our internals.
584 - Replay: New function xbt_replay_action_get():
585 Retrieve the function previously associated to an event type.
586 - DROPPED FUNCTION: xbt_str_varsubst()
587 - DROPPED MODULE: strbuff. We don't need it anymore.
588 - DROPPED MODULE: matrix. We don't need it anymore.
589 - DROPPED MODULE: lib. We don't need it anymore.
591 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
593 SimGrid (3.15) Released March 22. 2017
595 The Spring Release: continuous integration servers become green
596 We fixed even the transient bugs on all target architectures:
597 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
598 Windows, FreeBSD, NetBSD.
600 - Rename NetCards to NetPoints.
601 This was intended to help NS3 users, but that's not a netcard.
602 That's a point in the routing algorithm, let's avoid wrong simplifications.
605 - New: MSG_process_yield(). Stop and yield to other processes.
606 - New: MSG_process_daemon(). Daemon processes are automatically killed
607 when the last non-daemon process terminates
608 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
609 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
610 to make room for the upcoming network energy plugin.
611 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
614 - Ensure that an actor can kill itself with Process::exit()
615 - Kill the obscure NativeException. Nobody want to survive the issues
616 it denotes, so use JniException that is a RuntimeException (not to
617 be caught explicitly).
618 - Partial bug fix in initialization. SimGrid flags on command line were
619 consumed at C level but stayed in the original Java String[] args.
620 This could mess users' args[i] if SG flags were not put at the end of
622 The SimGrid flags are now removed from the Java arguments. However,
623 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
624 if args.length is greater than the number of YOUR OWN ARGUMENTS.
625 It might be if you have --log or --cfg flags in the command line.
626 - Fix numerous memleaks all around the place. In particular, around VMs.
630 - s4u::onPlatformCreated: right before the simulation starts
631 - s4u::onSimulationEnd: right after the main simulation loop
632 - s4u::onTimeAdvance: right after a clock change (time is discrete)
633 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
634 event from the availability_file changes the avail speed.
635 - Links are now usable from s4u
636 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
637 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
638 - New examples: The conversion of MSG examples to S4U has begun
639 - Actors: create, kill, migration, and suspend
640 - Applications: master-worker and token-ring
641 - Action replay: communications and storage
642 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
645 - Backwards Compatibility breaks
646 - The SD_route_get_list and SD_route_get_size functions have been removed.
647 They are replaced by the unique
648 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
649 The route from src to dst is built in the links dynar whose size can
650 be obtained with xbt_dynar_length.
651 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
652 sg_host_route_bandwidth and sg_host_route_latency.
653 Macros ensure the backwards compatibility, but you should fix your code
656 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
657 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
658 - Bump our claim of support from MPI 1.1 to MPI 2.2.
659 We don't support 100% of it, but it should be enough. Contact us if not.
660 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
661 naive and probably inaccurate)
662 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
663 mapped onto a unique small file using some system magic.
664 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
665 - Fortran: cleanups, fixes, support of user-added operations
666 - RMA: MPI_Accumulate are applied in correct order.
667 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
668 - Better support for MPI_IN_PLACE
669 - Support for MPI_Win attrs and keyvals.
670 - Support MPI_Comm_group_create, MPI_Type_size_x
671 - MPI ops should be properly applied to all allowed types
672 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
673 algorithm by a non failing one from mpich.
676 - Kill the fifo data container: we don't use it anymore.
679 - A <cluster> can now be created with different speed values to represent pstates.
680 see examples/platforms/energy_cluster.xml for an example.
682 SimGrid (3.14.159) Released December 28. 2016
684 The previous release was only a crude approximation of a Pi release;
685 this one should be a bit better.
687 - Revert a last minute change that broke on Mac OSX.
688 - Fix the changelog and NEWS files.
689 - Improve the documentation, in particular of the routing module.
690 - Integrate some patches flying around in Debian.
692 SimGrid (3.14) Released December 24. 2016
695 * The whole documentation was reworked and reorganized. There is
696 still much room for improvement here, but we're on it.
699 * We now need python3 for our testing infrastructure.
700 * Model checking needs libevent
703 * Now works on FreeBSD
706 * Privatization now works on FreeBSD
708 * Privatization is now activated by default in smpirun
709 You can turn it off with -no-privatize if something goes wrong.
711 * Call-location tracing for SMPI.
712 You can add the exact location (filename / linenumber) of an MPI call to
713 your trace files and slow down or speed up the simulation between two
714 consecutive calls by using an adjustment file (see the documentation).
716 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
717 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
718 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
719 values were ignored in some cases. The timings of these functions can now
720 be significantly different.
722 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
723 smpi/running-power is renamed to smpi/host-speed
725 * smpi/grow-injected-times option to enable or disable multiplication of the
726 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
727 make simulation less precise (but also much faster).
729 * smpirun script should be (much) faster for large deployments.
731 * SMPI tracing: fixed issue with poor matching of send/receives.
733 * Replay: Fix broken waitall
735 New functions and features
736 * MSG_parallel_task_execute_with_timeout, to timeout computations.
738 Dropped / renamed functions and features
739 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
740 - MSG_mailbox_is_empty() -> Mailbox::empty()
741 - MSG_mailbox_front() -> Mailbox::front()
742 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
743 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
744 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
745 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
746 * Don't pass the free_f parameter to property related functions:
747 - MSG_host_set_property_value()
748 - MSG_as_router_set_property_value()
749 - MSG_storage_set_property_value()
750 * VM properties. Since msg_vm_t are msg_host_t, just use
751 MSG_host_get_property_value() and friends
752 * VM I/O related things:
753 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
754 - Unimplemented save/restore methods
755 * MSG_as_router_get_property_value() was redundant with
756 MSG_environment_as_get_property_value().
757 - Removed MSG_as_router_*propert*() functions
758 - Added MSG_environment_as_set_property_value() for consistency
759 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
760 Well, they are still there for now, but deprecated with a warning.
761 Please switch to xbt_dict_new_homogeneous() before this is removed
763 * Task affinity. Its intended behavior (that was very badly tested
764 and probably not really working) was deceiving what most users
765 would have hoped here.
766 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
767 * Option network/coordinates is now useless and should be dropped.
770 * Added option storage/max_file_descriptors to allow more than 1024 files opened
773 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
774 * Added Dragonfly topology support
777 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
778 EiBps, EBps, Eibps, Ebps for bandwidth.
779 They may become useful to some lucky ones.
782 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
783 * ex/app_tokenring: new example, very similar to the MSG Token Ring
784 * ex/async_waitAll: new example, on asynchronous communications
787 * Memory usage should be decreased for simulations with a large number
788 of processes. This also helps for SMPI.
790 SimGrid (3.13) stable; urgency=low
792 The Half Release, a.k.a. the Zealous Easter Trim.
794 - We removed half of the lines, that were mostly experimental cruft.
795 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
796 (not counting blanks and comments -- according to openhub.net)
797 - The internals are now compiled in C++ (and will soon be clean C++)
798 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
799 59 klines of C, etc. We added only 29 klines of C++ in replacement.
801 * Backwards Compatibility breaks
802 - Removed Lua simulation bindings (switch to C or Java for that).
803 Lua can still be used to describe platforms
804 - Removed Java kernel plug-ins.
805 Will be reintroduced after the ongoing major internals reorg.
807 - the following functions were removed.
808 They were too specific and should be reimplemented in a generic
809 way, with filter function.
810 - MSG_task_listen_from_host
811 - MSG_mailbox_get_count_host_waiting_tasks
812 - MSG_mailbox_put_with_timeout was removed.
813 Please use MSG_task_send_with_timeout instead.
815 - the SD_application_reinit function was removed. It has been a noop for a while.
816 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
817 replaced by a more flexible mechanism.
818 - The following functions thus do not exist anymore
819 - SD_workstation_get_access_mode
820 - SD_workstation_set_access_mode
821 - SD_workstation_get_current_task
822 - Basic estimation functions have been removed but can easily be replaced
823 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
824 - SD_workstation_get_computation_time => amount / sg_host_speed()
826 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
827 Use VM.getSpeed()*load/100 for the legacy behavior.
829 - option enable_tracing was removed. It was not doing anything for a while.
830 - In the ModelChecker:
831 - the model-checker now ptraces the model-checked process which means
832 you cannot use a debugger on the latter anymore (we might make this
833 optional in the feature);
834 - removed soft-dirty page tracking;
835 - remove model-checked side snapshot management,
836 MC_snapshot() and MC_compare_snapshot();
837 - keep the MC_cut() function as a stub (it was not really working
838 in the previous release).
841 * All options are consistently in kebab-case. Old names are kept as alias.
844 * Switch to platform v4 format.
845 - Rename from 'power' to 'speed' the attributes describing the amount of
846 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
847 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
848 - In <host> and <link>, attributes availability and state are gone.
849 It was redundant with state and availability traces, and with peak values.
850 - In <cluster>, attributes availability_file and state_file are gone.
851 It was too complex and unused.
852 - Kill <gpu>. Was not doing anything.
853 - The DOCTYPE points to the right URL:
854 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
855 (the file at this address now documents the changelog since its v1)
856 - A warning is emitted for unit-less values (they are still accepted).
857 - speed. Default: 'f' or 'flops'. Also defined:
858 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
859 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
860 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
861 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
862 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
863 - latency. Default: 's' second. Also defined:
864 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
866 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
867 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
870 * s4u::Host is now the preferred public interface to the Host features.
871 sg_host_* functions are C bindings to the exact same behavior
872 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
875 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
876 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
877 They were deprecated since a few years, and probably did not even compile anymore.
880 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
881 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
882 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
883 encouraged to update their codes. A list of the modified functions can be found at the end of
884 include/simgrid/simdag.h
887 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
888 the lambda or closure passed as a parameter will run in kernel mode.
889 All the callback functions should be rewritten to that interface at some point.
892 * Reorganizing and cleaning the internals all around the place.
895 * Remove old default barrier/bcast buggy algorithms (see #18407)
896 * Various bug fixes to handle more codes
897 * Remove the need for the --foreground option of smpirun (it is still
898 accepted for backward compatibility).
901 * Kill the setset data container: MC don't use it anymore.
902 * Kill the queue data container: it made more sense with GRAS.
903 * Kill the xbt_peer_t data type: it's useless without GRAS.
904 * Kill rm_cb feature of config sets: it was never useful.
905 * Kill graphxml parsing feature. It was not used.
906 * Kill the deprecated code protected by XBT_USE_DEPRECATED
908 - xbt_dynar_sort_strings(), when the content is char*
909 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
910 They throw exceptions on invalid input;
911 * C++ support for declaring CLI flags (simgrid::config::Flag);
912 * class for abstracting different signal backends (simgrid::xbt::signal).
913 with no external dependencies (we need very simple signals).
916 * refactoring and cleanup of the code;
917 * ongoing process to cleanly separate the model-checking algorithms
918 from the code model-checking support.
920 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
922 SimGrid (3.12) stable; urgency=low
924 The Facelift Release.
927 * Require g++ v4.7 at least to not speak prehistorical C++.
928 * Require Boost 1.48 (for signal2 component).
929 * Java must be version 7 at least when activated.
930 * Builds on Windows again (including Java bindings).
931 * Tracing is now always enabled (no way to turn it out)
932 * Remove GTNetS. It was not working anyway.
933 * Various cleanups in the cmake scripts.
934 * Move headers around to sort them out on installed systems:
935 - instr/instr.h -> simgrid/instr.h
936 - instr/jedule/* -> simgrid/jedule
937 - simdag/datatypes.h was removed
938 - simdag/simdag.h -> simgrid/simdag.h
939 - msg/datatypes.h was removed
940 - msg/msg.h -> simgrid/msg.h
943 * Interface improvement:
944 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
945 - Sanitize the interface in MSG_task_ module:
946 - Merge two functions that were close enough but misleading:
947 set_compute_duration(t) -> set_flops_amount(t)
948 get_remaining_computation(t) -> get_flops_amount(t)
949 - set_data_size(t) -> set_bytes_amount(t)
950 get_data_size(t) -> get_bytes_amount(t)
951 - Massive cleanups in the functions related to the energy
952 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
953 - New: MSG_host_get_pstate()
955 - msg/energy/onoff: switching hosts on and off
957 * Interface improvement:
958 - Rename Host.isAvail() to Host.isOn()
959 - Rename Process.currentProcess() to Process.getCurrentProcess()
960 - Rename Task.setDataSize() to Task.setBytesAmount()
961 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
963 - #18874: Actually allows the GC to reclaim tasks
967 - SIMIX_process_throw: raises an exception in a remote process
968 * Refactoring: Separate sub-modules
969 - libsmx: the public interface, as libc in a real system
970 - popping: the strange dance that converts a user request into a kernel handling
971 - smx_context_*: the virtualization mechanisms that embed the user code
972 - smx_*: the handling of each simcalls
974 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
975 * Rename smx_action_t into smx_synchro_t, making explicit that these
976 things are used to synchronize processes with their environment.
977 For example, a communication is a sort of synchronization involving
978 the communicating processes (that may block until the exchange) and
979 the platform. The same can be said from computations, etc.
981 - #18888: segfault when a process ends before its kill_time
985 - 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.
986 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
987 - MPI_Type_set_name, MPI_Type_get_name
988 - MPI_*_c2f and MPI_*_f2c functions
989 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
990 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
991 - Activate a lot of new tests from the mpich 3 testsuite
993 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
994 - InfiniBand network model added: Based on the works of Jerome Vienne
995 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
996 - When smpi/display_timing is set, also display global simulation time and application times
997 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
998 * Collective communications
999 - 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
1000 - 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/).
1001 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
1003 - Replay now uses algorithms from wanted collective selector
1004 - Replay can be used with SMP-aware algorithms
1005 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
1006 - 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.
1008 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
1009 - Correctly use loopback on fat-tree clusters
1010 - Asynchronous small messages shouldn't trigger deadlocks anymore
1011 * Energy/DVFS cleanup and improvement
1012 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
1013 - new: smpi_get_host_pstate()
1016 - "Full" network optimization flag was broken since Surf++
1017 - Better handling of precision flags in maxmin
1018 - Fix bug causing sometimes "Impossible" errors
1019 - Properly pass cluster properties to included hosts
1020 * Improvement of the Energy plugin.
1021 - Always update the consumption before returning that value
1022 - New property: watt_off to denote the dissipation when the host is off
1023 - New functions getWattMinAt and getWattMaxAt to retrieve the
1024 dissipation of pstates that we are not currently at.
1025 * Java: class NetworkLink renamed to Link
1026 * New function: simcall_process_get_kill_time()
1027 * Massive rename s/workstation/host/
1028 - That's intrusive, but that's good for the project consistency. Sorry.
1029 - Change config option "workstation/model" into "host/model"
1032 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
1033 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
1034 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
1035 - xbt_dirname and xbt_basename for non-POSIX systems
1037 * The model checker now runs as a separate process.
1038 * The model checker runs must now be launched with the new simgrid-mc program.
1039 * Record/Replay: the MC can display a textual representation of a path in the
1040 execution graph. It can then be replayed outside of the model checker.
1042 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1044 SimGrid (3.11) stable; urgency=low
1049 * Normalizing pointers addresses tool for better diff between logs
1052 * Add cloud examples using new VMs
1053 - examples/msg/cloud/two_tasks_vm.tesh
1054 - examples/msg/cloud/simple_vm.tesh
1055 - examples/java/cloud/cloud.tesh
1056 - examples/java/cloud/migration/migration.tesh
1057 * Add java surf examples:
1058 - examples/java/surfPlugin/surf_plugin.tesh
1059 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
1060 - examples/java/surfCpuModel/surf_cpu_model.tesh
1061 * Add SMPI+MSG example:
1062 - examples/smpi/smpi_msg_masterslave/
1067 - msg task destroy cancel test
1068 - msg_host on/off test
1069 * Move all tests in testsuite to teshsuite (adding tesh files)
1070 * Restructure teshsuites
1071 - one folder for each kind of test
1072 * Restructure AddTests.cmake
1074 - structure the order of tests (with sections)
1077 * Add virtual machine
1078 - creation of a VM on a PM
1079 - migration of a VM from a PM to another PM
1081 - MSG_process_join(msg_process_t process, double timeout)
1082 - msg_bar_t MSG_barrier_init(unsigned int count)
1083 - int MSG_barrier_wait(msg_bar_t barrier)
1084 - void MSG_barrier_destroy(msg_bar_t barrier)
1085 - msg_as_t MSG_environment_as_get_by_name(const char * name)
1086 * New option "msg/debug_multiple_use" to help debugging when a task is used
1089 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1091 - Increase code coverage in test suites
1095 * Protect context stacks against stack overflow. The number of protected memory
1096 pages allocated on the top of each stack (1 by default) can be configured
1097 with the new command line option --cfg=contexts/guard_size:<value>.
1098 * Simcalls are now generated by a python script that
1099 - generates files included by SimGrid
1100 - checks that all the functions exist, and proposes prototypes
1102 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1103 host_set_data, host_get_data
1105 - simcall_process_join(smx_process_t process, double timeout)
1106 * Fix bug where sleeping processing could not be suspended.
1109 * Translate surf models from C to C++
1110 - Generic classes for all models: Model, Resource, Action
1111 - A generic interface for each kind of model (CPU, Network, Storage
1112 Workstation, WorkstationVM)
1114 * Translate surf routings from C to C++
1115 * Add callbacks using sigc++ or boost::signals2
1116 - Add callback functions for resource creation/destruction
1117 - Add callback functions for action state change
1118 - Handle Energy as a plugin
1119 * Replace swag by boost::intrusive
1120 * Add new routing models for clusters. For documentation, see
1121 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1122 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1123 parameters for cluster tag
1124 - Fat trees, with topology="FAT_TREE" and
1125 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1126 - see examples/platforms/torus_cluster.xml and
1127 examples/platforms/fat_tree_cluster.xml
1128 * More documentation
1131 * Hostfiles support host:nb_processes construct to deploy several processes on
1133 * Collective communication algorithms should not crash if used with
1134 improper number of nodes and report the error.
1135 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1136 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1137 MPI_Dims_create, MPI_Cart_sub are supported.
1138 * New interface to use SMPI programmatically (still depends on MSG for
1139 some parts, see examples/smpi/smpi_msg_masterslave):
1140 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1144 * Global variables privatization in MPI executables is now performed at runtime
1145 with the option smpi/privatize_global_variables (default:no).
1146 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1147 dynamic libraries loaded by the application are not privatized (static
1148 linking with these libraries is advised in this case)
1151 * Options defined in XML work correctly now.
1154 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1155 copied into simgrid.jar or not (ON by default). Use this option if you want
1156 to reduce the size of the installed simgrid.jar, *and* the native libraries
1157 are kept installed elsewhere.
1158 * Surf binding with SWIG (code generated in maintainer mode only):
1159 - plugin to handle callbacks
1160 - CPU model only for the moment
1163 * Supernovae build mode is definitively removed. It was used to improve
1164 inlining and inter-module optimizations. It is nowadays superseded by
1165 link-time optimizations commonly available in compilers.
1166 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1167 * Add boost dependency for surf++
1168 * Add new macro for tests
1169 - ADD_TESH(name <tesh_args>)
1170 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1174 - xbt_bar_t XBT_barrier_init(unsigned int count)
1175 - int XBT_barrier_wait(xbt_bar_t barrier)
1176 - void XBT_barrier_destroy(xbt_bar_t barrier)
1177 * Make the xbt_os_time module public
1179 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1181 SimGrid (3.10) stable; urgency=low
1183 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1186 * Reintegrate Java to the main archive as desynchronizing these
1187 package is not acceptable anymore (Java is now considered stable)
1188 * Add explicit synchronization facilities through semaphores
1189 * Bug fix: Task.setDataSize() only changed the C world, not the value
1190 cached in the Java world
1193 * Dramatically change the way files are handled. API and internals changed, but
1194 this part of MSG was not considered as production grade either.
1195 * Add explicit synchronization facilities through semaphores
1196 * Add a new function MSG_host_get_process_list()
1197 * Preliminary DVFS support (see examples/msg/energy/ for details)
1200 * SMPI is now included directly in the libsimgrid as the windows
1201 linker doesn't force us on splitting it anymore.
1202 * Improvements of the SMPI replay tool:
1203 - Most of the collective communications are now rooted in the same process as
1204 in the original application.
1205 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1206 used until now). Multiple data types can now be used in a trace.
1207 - The replay tool now supports traces produce either by TAU or a modified
1209 - Bug Fix: the compute part of the reduce action is now taken into account.
1210 - Gatherv collective is now supported
1211 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1212 of smpirun outputs time independent traces for the current run. One file
1213 is created per process. If too many processes are simulated, this behavior
1214 can be changed to one file for all processes by using the
1215 tracing/smpi/format/ti_one_file flag
1216 * smpirun generates the host file if needed (with given host count and platform)
1217 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1218 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1219 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1220 --cfg=smpi/coll_selector:(mpich/ompi)
1221 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1223 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1224 happening inside a collective SMPI call.
1225 * Fix the behavior of complex data types handling.
1226 * Make MPI_Wtime another synchronization point to take computations into
1228 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1229 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1230 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1231 (no privatization of global variables yet)
1232 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1235 * Verification of liveness properties is now available for SMPI applications
1236 (in addition to MSG applications)
1237 * Bugged examples using SMPI in examples/smpi/mc/
1238 * Add --cfg=model-check/visited option. Allows the verification of infinite
1239 programs. Detection of loops in the execution thanks to the system state
1240 comparison and reduction of the state space to explore. Can be combined with
1241 DPOR for safety properties.
1244 * Allow to change SimGrid configuration (see --help) within the code
1245 thanks to SD_config() as it can be done in MSG.
1246 * Add a new function SD_task_set_amount() upon user request.
1249 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1250 * Remove rule based routing (no more PCRE dependency)
1251 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1252 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1253 * Add a loopback_bw and loopback_lat options to cluster tag.
1254 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1255 routing. Now peers are encapsulated in an AS and have their own private
1256 router but this is transparent.
1259 * Our own implementation of getline is renamed xbt_getline, and gets
1260 used even if the OS provide a getline(). This should reduce the
1261 configuration complexity by using the same code on all platforms.
1262 * New type: xbt_cfg_elm_boolean.
1263 * Allow to use yes/no for boolean configuration options in the command line.
1264 * Allow to disable SimGrid cleanups at exit from command line option.
1265 There are situations where one may want a simulation to end with an exit.
1266 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1267 annoying when scripting around the simulator. Adding a
1268 --cfg=clean_atexit:no allows to circumvent this issue.
1271 * Lots of memory leaks were corrected in this release.
1272 There is no known memory leaks anymore, in all of our 600+ tests.
1273 * New command line option --version, to get SimGrid version information.
1274 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1276 * Supernovae builds are deprecated, and expected to be removed in the next
1279 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1281 SimGrid (3.9) stable; urgency=low
1283 The Grasgory release: GRAS is really dead now.
1285 * Complete overhaul of the internal host structures scheme.
1288 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1289 was considered as experimental and badly maintained since then.
1290 * Keeping it was thus a trap to our potential users, that could take
1291 it instead of MSG or SMPI by mistake despite is pity state.
1292 * GRAS seems to have very few users (if any), and no one volunteered
1293 to maintain it further. It also induces a lot of XBT code (for
1294 portability sake), that must be maintained too.
1295 * For all these reasons, we killed GRAS. If someone wants to revive it
1296 in the future, don't cry, our git history still remembers of GRAS.
1299 * Major overhaul. Merge our documentation again as time proved that
1300 splitting it was really not helping our users.
1301 * Further improve the developer documentation to help newcomers
1302 hacking on SimGrid itself. The user documentation (and in
1303 particular, the beginner documentation) is still in a sorry state.
1306 * Now works on Windows too!
1307 * Much more extensive test suite, from MPICH
1310 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1311 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1312 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1313 * Bug fix: let task be scheduled when the last dependency to be solved is
1314 a control dependency.
1315 * Remove SD_load_environment_script function.
1316 Use the C sg_platf function if you want to declare a platform
1320 * New function: MSG_process_get_number()
1321 * Old function documented: MSG_config()
1322 * Remove MSG_load_platform_script function
1323 Use the C sg_platf function if you want to declare a platform
1327 * Change the default value of the TCP_gamma constant (maximal size of TCP
1328 congestion window) to a more realistic 4MiB value. If you notice changes in
1329 your simulation results, you can fall back to the previous 20k tiny window
1330 by adding --cfg=network/TCP_gamma:20000 on command line.
1331 * (Hopefully) fix a bug wrt periodic availability and state traces
1332 * Bug fix: use default values at start when first event in availability/state
1333 trace is not at time 0.
1336 * remove the "new_" part of function name sg_platf_new_trace_connect
1337 (resulting in sg_platf_trace_connect), since it does not create
1341 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1342 use case where it's really mandatory, and maintaining it was a pain in
1344 * New: xbt_fifo_search(), search an item with a user-provided
1345 comparison function instead of dumb pointer comparison.
1348 * Fix the lua deployment:
1349 Use `simgrid.init_application()` before deployment instead of
1350 `simgrid.msg_register_application()` after.
1353 * Transfer the tracing files into the corresponding modules.
1355 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1357 SimGrid (3.8.1) stable; urgency=low
1359 The "we are told that some people want to also *install* the simgrid
1362 * Add missing file "tesh.1" to the archive.
1364 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1366 SimGrid (3.8) stable; urgency=low
1368 The Psssshiiiit release: SimGrid jumps into the Cloud.
1371 * Add an experimental interface to manipulate VMs. They are mainly
1372 process groups with very few intrinsic semantic, but they should
1373 allow you to build the semantic you want easily.
1374 * New function: MSG_host_set_property_value()
1375 * New function: MSG_process_on_exit(). To clean memory in all cases.
1376 * Bug fixes that made the host (and link) failures unusable.
1377 * Add a way to auto-restart process when the host in which they are
1378 executing comes back (ON_FAILURE="RESTART" on deployment file,
1379 MSG_process_auto_restart_set).
1380 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1381 please stop using the old ones, they are DEPRECATED.
1383 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1384 Please use MSG_init() instead. (reducing the amount of entry
1385 points in the library helps us).
1386 * Make it impossible to link against the wrong version of the lib
1387 * Deprecate MSG_clean(). No need to call it anymore.
1388 * Function MSG_get_host_number() is not deprecated anymore.
1391 * Split the doc into a user guide and a reference guide.
1392 * Start a developper guide to help people hacking on SimGrid.
1395 * Enable tracing by default. This modules rocks you should use it.
1396 * Remove option custom_flags. Now use environment variables CFLAGS
1398 * Use default cmake things to detect lua instead of home grown ones.
1399 * New option "enable_mallocators" to disable mallocators, for debugging
1400 purpose ("on" by default).
1403 * Bug fixes around the resource failures: don't let the processes
1404 survive the host they are running onto.
1405 * Add an interface to auto-restart processes when the host in which they are
1406 executing comes back.
1407 * Ensures that SIMIX_clean is called automatically. It's not part of
1408 the public interface anymore (bindings should be updated).
1411 * Bug fix for when SD_Simulate is called with a positive value: be careful
1412 when comparing doubles. Sometimes they are different for non significant
1414 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1415 parallel task whose initial work is distributed among host according
1416 to the Amdahl's law. Such tasks are created with a parameter alpha
1417 that corresponds to the non-parallelizable part of the computation.
1418 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1419 between two sets of workstations assuming a 1D block distribution (each
1420 workstation owns a similar share of data) on both sides.
1422 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1423 Data redistribution will be automatically scheduled once parent and child
1424 are both scheduled. The filling of computation_amount and
1425 communication_amount structures is now done seamlessly thanks to the chosen
1427 * New function SD_workstation_dump to display various information
1428 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1429 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1430 bandwidth on the route onto which the task is scheduled and the amount of
1432 To divide the nominal bandwidth by 2, the rate then has to be:
1433 rate = bandwidth/(2*amount)
1434 * Compute tasks that have failed can now be rescheduled and executed again
1435 (from their beginning)
1436 * Increasing source code coverage (src/simdag is now covered at 95.8%
1440 * Re-implement time-independent trace replay using SMPI (at the
1441 smpi_smp_* level) instead of MSG. This should replace
1442 examples/msg/actions/actions.c
1443 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1444 hindexed and structs)
1445 * Implement the exchange of non-contiguous data.
1446 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1447 * Correct behavior of smpi/sender_gap and set its default value to 0
1448 * Add option to asynchronously send small messages to allow better
1449 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1450 specifies the size in bytes under which messages will be asynchronously sent.
1451 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1452 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1455 * New C interface to define a platform: XML is now optional.
1456 For more info, please check include/simgrid/platf.h.
1457 * New interface to define random platforms from the C:
1458 For more info, please check include/simgrid/platf_generator.h and
1459 examples/msg/masterslave/masterslave_platfgen.c
1460 * Export a sg_cmdline dynar containing all the arguments we got from
1464 * Two new tracing options for adding comments to trace file so you
1465 can track your experiments (see --help-tracing for details).
1466 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1467 * Adding the SimGrid version that generated the trace file as a comment.
1468 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1469 * Fix to avoid key clashes on Paje links
1470 * Other minor fixes related to the Paje specification
1473 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1474 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1475 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1476 * Greatly improve the robustness of mmalloc to user errors (such as
1477 using an area after freeing it, or freeing it twice)
1479 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1481 SimGrid-java (3.8.1) stable; urgency=low
1483 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1484 Warning: all methods are visible, but only some of them are
1485 implemented so far. Check the source (src/jtrace.c)
1486 for further information.
1487 * New module: org.simgrid.msg.File (SimGrid File management functions)
1488 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1489 * Change the meaning of Process.restart: now restart the process from
1490 the begining, like MSG_process_restart in C.
1491 * Add Process.setAutoRestart: handling of process restart when failed
1493 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1494 you to retrieve the properties of the processes/hosts
1495 * Deprecate Msg.clean(): you can just forget about it now.
1496 * New function Process.getCount(), that only works when compiling
1497 with the not yet released version 3.9 of the C library.
1499 * New context factory based on Coroutines. It mandates a modified JVM
1500 but then, the simulations run about five times faster, and there is
1501 no limit to the amount of processes (beside of the available memory).
1503 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1505 SimGrid (3.7.1) stable; urgency=low
1508 * Restore the prototype of MSG_process_create_with_environment() to
1509 the pre-3.7 situation by removing the kill_time argument.
1510 * Add a MSG_process_set_kill_time() function instead.
1513 * Fix weird behaviors when dealing with parallel tasks.
1516 * Simgrid is now built as a dll.
1517 * Simgrid-java now works on Windows.
1518 * Simgrid-Java is now included into Windows package.
1521 * First pre-build package for MacOSX.
1524 * Fix compilation when using MSG_USE_DEPRECATED.
1525 * Fix some compilation issues on Macs and Windows.
1526 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1527 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1529 We discovered that the Lua console is broken, but we are missing the
1530 manpower to fix it right now. The problem existed in 3.7 too, so we
1531 are not blocking the release for that. Sorry if you depended on this
1532 feature, any help would be really welcome.
1534 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1537 SimGrid-java (3.7.1) stable; urgency=low
1539 The "Java aint got to be bloated and slow" release
1542 * Various internal cleanups and performance improvement
1543 Simulations are expected to run up to twice faster or so
1544 * Make Process.kill(process) an instance method, not a static one
1545 * User processes are not java.lang.Thread subclasses.
1546 This breaks the compatibility (sorry), but previous API was
1547 brain-dead, making it impossible to have non-trivial
1548 initializations in the process constructor.
1549 * Require a full constructor per Process sub-class.
1550 Kinda breaks the compatibility (sorry), but this allows a much more
1551 efficient way to launch the processes at simulation startup.
1552 * Do not embeed our version of semaphores, java 1.5 can be considered
1553 as sufficiently prevalent for us to not dupplicate its features.
1557 * Add examples for almost every part of the API
1558 We spotted and fixed a lot of bugs in the process
1560 * New module: asynchronous communication API
1561 * New function: Process.sleep()
1562 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1563 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1564 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1566 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1568 SimGrid (3.7) stable; urgency=low
1570 The "spring cleanups (before the next Big Project kicks in)" release.
1573 * We can specify the SMPI latency/bandwidth factor with command line
1574 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1575 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1576 You can also use the "config tag" from platform file by adding this line
1577 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1578 (see "example/platforms/tag_config.xml" to use "config tag").
1579 Note that the command line supersedes the platform file configuration.
1580 * Change the correction factors used in LMM model, according to
1581 the latest experiments described in INRIA RR-7821.
1582 Accuracy should be improved this way.
1583 * Use the partial invalidation optimization by default for the
1584 network too. Should produce the exact same results, only faster.
1585 * Major cleanup in surf to merge models and split some optimization
1586 mechanisms from the core of the models. As a result you can now
1587 specify which model to use (e.g., --cfg=network/model:LV08
1588 --cfg=cpu/model:Cas01) and which optimization mode to use
1589 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1590 Incompatible combinations should err at initialization. See
1591 --help-models for the list of all models and optimization modes.
1592 * The CLM03 workstation model was dropped for simplicity because it
1593 used the deprecated CM02 network model. Use default instead.
1594 * Rename the TCP_gamma configuration option to network/TCP_gamma
1595 * Rename the coordinates configuration option to
1596 network/coordinates, and document it
1597 * Use now crosstraffic keyword instead of the terribly misleading
1598 fullduplex keyword. It is activated by default now in the current
1599 default model, use --cfg=network/crosstraffic:0 to turn it off.
1600 * Ongoing refactoring the model parsing to make XML files optional
1601 See include/simgrid/platf.h for details (still to be completed)
1604 * Major overhaul of the documentation. Almost instructive now :/
1605 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1606 functions and friends. This interface was considered as
1607 deprecated since over 2 years, it's time to inform our users that it is.
1608 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1609 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1610 use these (crappy) functions in your code.
1611 These functions will be removed soon. Stop using them now.
1612 * Deprecate MSG_get_host_{table,number}
1613 Implement MSG_hosts_as_dynar() instead.
1614 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1615 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1618 * Stabilize the parallel execution mode of user contexts
1619 * Introduce configuration variables to control parallel execution:
1620 - contexts/synchro: Synchronization mode to use when running
1621 contexts in parallel (either futex, posix or busy_wait)
1622 - contexts/parallel_threshold: Minimal number of user contexts
1623 that must be part of a scheduling round to switch to parallel
1624 execution mode (raw contexts only)
1625 * Fix bugs that prevented to use suspend/resume along with
1626 synchronization structures.
1627 * Fix bugs in process termination that lead to invalid memory access
1628 in very specific conditions.
1631 * Introduce a parallel mode for the models (controlled by surf/nthreads
1632 configuration item). In our tests, running the models in parallel
1633 never lead to any speedups because they are so fast that the gain
1634 of computing each model in parallel does not amortizes the
1635 synchronization costs, even when ultra fast futexes are used.
1636 This is released anyway because YMMV.
1639 * Performance boost by using a swag internally to compute the set of
1640 tasks that are finished and should constitute the return value of
1644 * Enable it by default now that it is considered rather stable.
1647 * Documentation of the tracing functions.
1648 * Performance gains when tracing categorized/uncategorized resource
1649 utilization by avoiding calls to get route when updating resource
1650 variables. LMM constraints are being used instead.
1651 * API changed to set task categories. Use MSG_task_set_category instead
1652 of TRACE_msg_set_task_category, and SD_task_set_category instead
1653 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1654 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1655 * Examples for link user variables added (see at examples/msg/tracing/)
1656 * Deprecated function TRACE_msg_set_process_category completely removed
1657 * Trace header updated according to the latest Paje file format
1658 * Tracing network lazy updates, no longer obligate users to use full updates
1659 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1660 * Experimental: let user code declare/set/push/pop application states for hosts
1661 * API changed to allow the manual creation of graph configuration files
1662 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1665 * Improve the API of Lua MSG bindings, using the Lua spirit.
1666 * Each simulated process now lives in its own Lua world (globals are
1667 automatically duplicated). It helps writing simulators. Will allow
1668 to run Splay programs within SimGrid in the future.
1669 * Add a Chord example in Lua, equivalent to the MSG one.
1672 * Start the implementation of a solution to express temporal
1673 properties, not only local assertions. This is still an
1674 experimental work in progress, stay clear from it to be safe.
1678 - Add new runtime parameters --help-logs and --help-log-categories
1679 to display information about supported logging parameters and
1681 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1682 don't exists anymore.
1683 * Mallocators: allow value NULL for the reset function.
1685 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1686 create homogeneous dictionaries, where all the elements share the
1687 same free function. Non homogeneous dictionaries will be
1688 deprecated in the next release.
1689 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1690 - Multi-level dictionaries are deprecated.
1692 - new function xbt_dynar_search_or_negative() that is useful when
1693 you have less than 2 million elements in your dynar and don't
1694 want of the extra complexity of catching exceptions when the
1695 element is not found.
1697 - Make xbt_os_thread module (for thread portability) public.
1698 Documentation is still to come, sorry.
1700 - Cleanups and simplifications to make it maintainable again.
1701 - Exotic features (such as memalign and valloc) were removed.
1702 - The metadata were extended and improved so that the
1703 model-checker becomes able to explore and inspect the heaps.
1704 - This may induce a performance drop when enable_model-checking is
1705 ON in cmake (even if it's not used in the simulation), but it is
1706 necessary at this point to get MC working.
1708 Turn model-checking OFF if simulation performance matters to you.
1709 Not enabling it at runtime is not enough, disable it in cmake.
1711 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1714 SimGrid (3.6.2) stable; urgency=low
1716 The "Not coding new stuff allows to polish old things" release.
1719 * New bindings to the NS3 packet level simulator (experimental)
1720 * Use the raw (efficient) execution contexts instead of the sysv
1721 (portable) ones when possible.
1722 * libpcre is now mandatory in any cases since not using it led to
1723 severe performance loss and possibly other issues
1724 * Update the XML platforms:
1725 - G5K: include the latest machine in Nancy
1726 - GridPP and LCG: new platforms
1727 * Documentation was partially updated, at least (more to come)
1729 Bug fixes, cosmetics and small improvements
1730 * Free terminated processes before the end of the simulation to avoid
1731 exhausting the memory of users having very dynamic amount of
1733 * Bug fix and cosmetics about canceling non-running tasks
1734 * Bug fix about the dot loader's issues when using libcgraph
1737 * Create an installer for windows with nsis (amd64 and win32)
1738 - Add an hello world project to illustrate simgrid project creation.
1739 - Embed libpcre into the Simgrid installer to avoid
1740 its compilation burden
1741 * The raw execution contexts should work on Apple now
1742 * Port to Windows 64 bits
1743 - Sysv contexts now have an implementation for this arch
1744 - GRAS communication features now support this arch
1745 * Drop support for borland compiler on windows
1746 - this code was not maintained, and we kinda depend on gcc nowadays
1747 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1748 * Fix portability issue on unstable ubuntu: linker became picky on
1751 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1753 SimGrid (3.6.1) stable; urgency=low
1755 The "Oops, we broke Macs too" release
1758 * Fixed contexts detection so that raw ones are used when possible
1759 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1760 leads to a strange error, with user code segfaulting sometimes when
1761 the generated code is not perfectly aligned (which is not
1762 controllable from the user side, depends on the amount of code)
1765 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1767 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1769 SimGrid-java (3.6) unstable; urgency=low
1772 * Split of every thing from simgrid v3.5 into a separate package.
1774 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1776 SimGrid (3.6) stable; urgency=medium
1778 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1781 * Bindings now constitute their own package, separated from the main one.
1782 Rationale: reduce our maintenance nightmare by reducing the module coupling
1783 They will soon be released on their own on gforge.
1785 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1786 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1788 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1789 * It's not quite deprecated for now because we have no replacement,
1790 but it may soon become the case.
1793 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1794 * Fortran: New user-level cache variable to store the rank of the running
1795 process. This improves performance by an order of magnitude.
1796 * C: New coccinelle script to automatically locate and modify global and
1797 local static variables.
1798 * Improved SMPI network model with a sender-side gap to account for multiple
1802 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1803 only say if a communication is finished, no matter whether it succeeded or
1804 failed. You can call MSG_comm_get_status() to know the status of a finished
1806 * New function MSG_task_dsend() to send a task and detach it. When a
1807 communication is detached, you are never notified of its success or failure
1808 and the memory is released automatically once it is finished. This function
1809 is useful when you don't care about the end nor the success of a
1811 * Change the prototypes of action replay. Sorry for inconvenience,
1812 but this is really more efficient this way (and to adapt your code,
1813 you just have to fix the initialization, that shouldn't be too long)
1814 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1815 ever managed to do anything useful with it.
1816 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1817 takes a dynar of communications. It returns immediately and gives the
1818 index of a finished communication (if any).
1819 * New example: a basic implementation of the Chord P2P algorithm.
1822 * New model for multi-core CPUs. You can now use the core attribute to
1823 precise the number of cores of a host. This is a basic model. Every
1824 process running on the host receives at most the power provided in
1825 the DTD (throughput<=power). Total throughput of process cannot exceed
1827 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1828 router linked by an up-link and a down-link (possibly asymmetrical).
1829 This kind of pattern allows to easily build last-mile model style platforms.
1830 Aggregating such patterns in a rule-based AS is thus the technique of
1831 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1832 * New model for Vivaldi routing. We transformed the Vivaldi network model
1833 into a Vivaldi routing model (based on the rule-based model). This allows to
1834 combine Vivaldi based latencies with last-mile platforms.
1837 * Added a check for NaN of IEEE754 infinite in the double entries of
1839 * Introduce a new context factory "raw", highly inspirated from the
1840 ucontext factory, but using manually crafted functions in assembly to
1841 do the work in an efficient manner.
1842 * Allow to change the used context factory at run time, not only at
1843 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1844 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1845 contexts at runtime (only with raw contexts or ucontexts).
1846 * Completely rewrote this module to allow parallel execution of user
1847 processes. Use --cfg=contexts/nthreads:N to execute user processes
1848 with N parallel threads (the default is 1, meaning no parallelism).
1849 * Allow to decide dynamically between sequential and parallel modes.
1850 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1851 processes in parallel only when their number is greater than or equal to P
1853 * Added a check for NaN of IEEE754 infinite in the double entries of
1857 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1858 won't output the state of processes when interrupted with Ctrl-C
1859 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1860 NULL-terminated array. This may solve backward compatibility issues
1861 due to the change to return type of SD_simulate. See also:
1862 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1863 * Add new macros with variable number of arguments.
1864 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1865 - in xbt/asserts.h: xbt_assert
1866 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1867 - in xbt/ex.h: THROWF and RETHROWF.
1868 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1870 * Change xbt_die() to accept a format string with arguments, just like printf.
1871 * New data structure: xbt_lib_t, like a dict but more general and with better
1875 * New configuration options
1876 Options triva/categorized and triva/uncategorized can be used to generate
1877 graph configuration files for Triva visualization tool.
1878 * Configuration option tracing/platform is renamed to tracing/categorized
1879 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1880 * New TRACE_user_link_variable interface:
1881 User provides the name of the link and the tracing variable to attach to it
1882 * the declaration of tracing categories must be done after the environment
1884 * simpler tracing interface, just one way to declare categories
1885 TRACE_category or TRACE_category_with_color, it is up to you
1886 * links in the trace file are again identified by their names
1887 * trace contains the full platform hierarchy exactly as declared using the ASes
1888 * Options tracing/msg/[task|process]:1 groups the process by hosts
1889 for both cases, tasks and processes must have names that are unique during the simulation
1890 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1891 * The experimental option tracing/msg/volume is deprecated
1892 its functionality may be reincorporated if needed
1894 The tracing generates a trace file with unordered timestamped events,
1895 because of the way the core simulator (surf) works. A script available
1896 at the tools directory (fix-paje-trace.sh) can be used to put the events
1897 in order. We have changed the tracing so it can generate ordered timestamped
1898 events in the final trace, but depending on the simulator (and how much time
1899 is simulated) that can lead to a huge memory utilization. It is deactivated
1900 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1902 Build Infrastructure
1903 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1904 - We are trying hard to keep the API stable, but it may happen that
1905 some things change (we're a research project after all, not a
1906 nuclear plant operating system). If such things should happen, you
1907 could rely on that macro to adapt.
1908 - current value: 30600 for 3.06.00, aka 3.6
1909 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1910 a number that can be compared with SIMGRID_VERSION.
1911 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1912 assertions and verbose/debug logging events are disabled at compile time.
1914 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1916 SimGrid (3.5) stable; urgency=medium
1919 * New feature to any SimGrid-based simulator: Model-Checking
1920 Check SIN#1 for more details.
1923 * New Model SMPI (three-interval linear regression for correction factors)
1924 See RR-7426, available at http://hal.inria.fr/inria-00527150
1925 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1926 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1927 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1928 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1929 * See also src/smpi/README
1934 - Tracing API changes: TRACE_start and TRACE_end should not be called
1935 by user-code. They are automatically called by simulators created
1936 with SimDAG, MSG and SMPI if the toolkit is compiled with
1937 tracing_enabled=ON. Categories declaration and utilization remain the
1938 same for MSG and SimDag.
1939 - A function was added to the tracing API to declare categories with
1941 - TRACE_category_with_color (char *category, char *color)
1942 where color must be in the following format
1943 "%f %f %f", red, green, blue
1944 and red, green, blue are float values in the interval [0, 1]
1945 - User can specify NULL as color parameter, or continue calling
1946 TRACE_category (cat)
1947 On that case, the tracing system will define random colors
1948 - The following command-line options are supported:
1949 --cfg=tracing/filename:msg.trace
1950 --cfg=tracing:1 (activate tracing, needed to use others)
1951 --cfg=tracing/platform:1 (categorized resource use)
1952 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1953 --cfg=tracing/msg/task:1 (task creation)
1954 --cfg=tracing/msg/process:1 (process creation, migration)
1955 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1956 --cfg=tracing/smpi:1 (SMPI interface tracing)
1957 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1958 - examples of examples/msg/tracing updated
1960 - DAXLoader and DOTLoader functions can generate tasks with categories
1961 - A new function to attribute a category to SD tasks:
1962 TRACE_sd_set_task_category (SD_task_t task, char *category)
1963 * Tracing the MPI interface implemented by SMPI
1964 - Collective operations are traced with states
1965 - Point-to-Point operations are traced with states/links
1966 - Tracing activated by a parameter "-trace filename" passed
1967 to smpirun during execution (considering that simgrid
1968 is compiled with tracing enabled)
1969 - To run the simulation with gdb, the simulator
1970 accepts --cfg=tracing/smpi:1 to trace SMPI
1971 - tesh files to check if smpi tracing is ok
1972 - See examples/smpi/NAS/DT-trace
1973 * GTNetS tracing re-worked
1974 - adaptation to the tracing system of GTNets to cope
1975 with modifications regarding the fullduplex mode
1976 - new tesh files to check if gtnets tracing is ok
1979 * Asynchronous communications through the functions:
1980 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1981 * New function: MSG_load_platform_script()
1982 to make possible using a lua script instead of XML files to set up platforms
1983 * New function: MSG_set_function
1984 to associate functions to processes, used when bypassing the parser
1985 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1987 Platforms: Add some more examples in examples/platforms
1988 * Grid'5000: see www.grid5000.fr
1989 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1992 * Change the XML format. This is a very important modification. SimGrid
1993 3.5 introduces a new hierarchical format based on the notion of
1994 Autonomous Systems. Compatibility with old format is ensured
1995 through the perl script provided in the install bin directory
1996 bin/simgrid_update_xml.
1997 It is now possible to build platforms with specific routing
1998 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1999 connect several platforms together. We will try to provide soon
2000 set of realistic platforms exploiting these properties (have a
2001 look at examples/platforms/ for the moment).
2002 * Take the opportunity of the XML format change to be a good XML citizen:
2003 rename link:ctn to link_ctn and similar changes (also dealt with by
2005 * Add a new routing scheme (rule-based) using regular expressions. It
2006 enables to have an extremely low memory footprint when the
2007 underlying routing is simple and can be compactly described. You
2008 need to have libpcre4-dev (perl regular expressions) installed if
2009 you want to use this routing scheme.
2010 * Revive the cluster TAG and allow to easily and efficiently (both in
2011 term of memory and speed) connect clusters together. Have a look
2012 at teshsuite/simdag/platforms/ to see how this can be done. With
2013 this tag, you can create clusters with thousands of tasks at no
2014 cost (have a look at examples/platforms/).
2015 Note: clusters are implemented as ASes, so there is no need for an
2016 enclosing AS tag if you have only one cluster in your platform.
2017 * Add new generic functions in the public interface that allows the user
2018 to call SURF 'create_resource' methods from your code (same
2019 functionality as the XML bypass mechanism but with a much lighter
2021 * Add a new model (enabled through command line --cfg=network/model:SMPI)
2022 that uses a piecewise linear approximation to produce better
2023 results when exchanging small messages.
2024 * Add a new parameter to handle correctly full duplex link and account
2025 for interferences between uplink and downlink communications
2026 (activate with --cfg=fullduplex:1).
2029 * Rename the SD_READY (all dependencies are satisfied and task is
2030 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
2031 dependencies are satisfied) state.
2032 This prevents a confusion between the notion of "ready to schedule"
2033 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
2034 simulated" (SD_RUNNABLE) used by the simulation kernel.
2035 * Change the way a task is considered as ready. Instead of removing
2036 dependencies when a task is done, a counter is decreased. This way,
2037 it is always possible to reach ancestors thanks to the
2038 SD_taks_get_parents function (even after the end of the simulation.)
2039 * Change the return type of SD_Simulate from (SD_task_t*) into
2040 xbt_dynar_t. This function was in handling a dynar internally and
2041 converted it into a NULL terminated array for historical reasons.
2042 * New function SD_dotload(char*) to load a DAG described in dot
2043 format. This loader and the corresponding examples require the
2044 installation of the graphviz library.
2045 * Fix a bug in the management of tasks of size 0 in the surf network
2046 models. This problem was only visible with SIMDAG and you should
2047 thus disregard results produced with earlier versions if you
2048 relied on this feature (some tasks were blocked because of this).
2049 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
2050 even though your applications were doing only point-to-point
2051 communications and sequential computations. Now you can really use any
2052 model you want (of course, if you create real parallel tasks, which are
2053 not implemented in most models beside ptaskL07, this will abort).
2054 * Add an example that schedules a DAX on an heterogeneous platform
2055 using a Min-Min strategy.
2056 * New function SD_workstation_get_current_task() that returns the kind
2057 of task currently running on a workstation in the sequential access
2059 * Raise some warnings when unexecuted tasks remains at the end of the
2060 simulation. This is usually caused by cycles in the DAG.
2063 * New function: SIMIX_process_set_function() called by MSG_set_function
2064 * Change the underlying waiting queue in semaphores so that a process
2065 can wait on several of them simultaneously (as in waitany).
2066 * Fix the way to handle tokens in semaphores so that all access patterns
2067 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
2068 * kill the dirty pimple SIMIX_message_sizes_output()
2069 Please use (proper) visualization instead
2072 * New data container: setset (set of sets of elements)
2073 * New module: mmalloc (mapped malloc, allowing to have several
2074 independent segments of malloc)
2075 * New function: xbt_dict_cursor_set_data()
2076 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
2077 * New function: xbt_dynar_is_empty()
2078 * New function: xbt_fifo_get_last_item()
2079 * Fix xbt_dynar_shrink(): use the right element size.
2080 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
2081 uninitialized areas during expand.
2082 * Fix semaphores: previous implementation was severely broken.
2083 * Use library init/fini functions for our initialization.
2084 - you can use logs and other feature as soon as you want in your
2085 code (even before the xbt_init / MSG_init)
2086 - xbt_exit is now a no-op and produce a warning when used.
2089 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2090 benefit from the latest improvement to the simulation kernel.
2091 * Kill measurement sockets for now. If you rely on them, sorry. This
2092 release is not for you. This feature will be reintroduced in the
2093 future, but we cannot delay the release any further.
2094 * New function: gras_msgtype_get_name().
2095 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2096 * Fix (at last) the pmm example: it should not randomly fail anymore.
2098 Build chain: bug fixes and overall polishing
2099 * Cmake is now stable enough. Hence, we killed the autotools.
2100 * Port to windows ( TM :)
2101 * Fix the 'make install' target.
2102 No need to use 'make install-simgrid' anymore
2103 * Introduce a 'make dist' target compiling a *source* archive
2104 'make package' compiles a binary archive
2105 * Compile java files only on need
2106 * Add --cd and --setenv command line options to tesh
2107 * Out of source builds are not fully supported yet, but we are close
2108 * Enable supernovae and optimization flags by default for our users
2111 * Add layer to set up environment directly from lua, without XML.
2112 * The effect of gras_stub_generator can be achieved through
2113 lua too (check examples/gras/console/ping_generator.lua)
2115 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2117 SimGrid (3.4.1) stable; urgency=low
2119 The "Polishing easter eggs is probably a good idea" release.
2120 This is a bug fixes release only.
2124 * Fix a bug preventing the tasks from begin garbage collected.
2127 * Fix a bug occuring when a host involved in a communication fails.
2128 This was not detected properly by the other peer involved in the
2129 communication. Now, it's reported as a network error.
2132 * Warn the user about loop dependencies in data flow of DAX files
2133 * Obey the control-flow dependencies of DAX files
2136 * Add option "enable_smpi" allowing to not compile SMPI.
2137 Probably useful for the (Mac) users experiencing a build error here
2138 * Improve the detection of lua5.1 and ruby1.8
2140 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2142 SimGrid (3.4) stable; urgency=low
2144 The "Easter in Cargese" release. Also known as (major changes):
2146 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2147 ~> bindings were greatly improved
2148 ~> new tracing infrastructure for better visualization introduced
2150 * the "Welcome to configury modernity" release.
2151 ~> we switched from autotools to cmake, and improved our cdash
2154 A more detailled list of changes follow (full detail in svn log).
2156 Java Bindings: Various Cleanups
2157 * (install java-gcj-compat-dev on debian-like to use them)
2158 * Remove put/get: no need to export deprecated interface in Java
2159 Use send/receive instead.
2160 * Cleanup the examples and add a README per directory
2161 * Remove example autoDestination (that's the only way to go now)
2162 * Remove example explicitDestination (was a plain copy of basic)
2163 * Make JniException a runtime exception, so that there is no need to
2164 declare the fact that you may encounter such a beast. I guess that
2165 nobody will ever want to survive such error.
2166 * Create specific errors for each MSG case of failure:
2167 host failure, transfer failure, timeout, task cancelled
2168 * Cleanup the exceptions that may get thrown by each function
2169 * Other internal cleanups in Java bindings. Performance still bad :/
2170 Ruby and Lua Bindings: create them
2171 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2172 * That's new and great, you should try them out.
2173 Same functionalities than Java bindings, only even less polished
2175 * Kill the useless "rate" argument of SD_task_get_execution_time()
2176 Everyone used to provide -1 as a value, it was not used, and the
2177 semantic of a possible use wasn't even clear.
2178 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2179 either as comm costs or compute costs to mean that there is no
2180 such thing for that specific task.
2181 * Add a SD_task_set_name() function
2182 * Fix SD_task_unschedule() on typed tasks
2183 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2184 * In DAX loader, accept useless 'level' attributes to <job> since
2185 LIGO DAGs have them (seem to be to ease graphical representation).
2187 * Add an example masterslave_mailbox.c using send/receive and not
2188 the deprecated put/get interface.
2189 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2190 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2192 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2193 (declare MSG_USE_DEPRECATED to still have the old name)
2194 * Add a MSG_task_set_data() function
2195 * About trace replay (see examples/msg/actions):
2197 - Allow to work with splitted trace files for each process
2198 Give the specific trace file as argument of each process,
2199 and call MSG_action_trace_run(NULL)
2200 You can still have one merged file for all processes.
2201 - Fix implementation of collective operations
2202 * Allow task_execute() on 0-sized tasks (closes #10063)
2204 * This is the first release of SimGrid where SMPI is not considered
2205 beta anymore (even if some corners should still be improved)
2206 * Port over the new SIMIX_network submodule (internal refactoring)
2207 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2208 * Implement more missing elements of the standard:
2210 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2212 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2213 MPI_Scan() MPI_Get_processor_name()
2214 - Added implementation of missing case for Alltoall (warning: it's
2215 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2216 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2217 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2218 SMPI_MPI_Allgatherv()
2219 * Bug fixes include:
2220 - MPI_Waitsome() was broken
2221 - Allow relative includes in smpicc
2222 - Command line cfg argument 'reference_speed' was ignored...
2223 - Some functions did not properly lead to auto-benching of user code
2224 - smpicc passes -O2 by default (just like openmpi one)
2226 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2227 * Bug fixes about timeouts during communications
2228 * add SIMIX_message_sizes_output() as a pimple to write to file the
2229 amount of messages per size. Use gnuplot to get histogram.
2230 Pimple because that's the only user-visible function of simix,
2231 defined directly in xbt.h (irk, sorry)
2233 - Add a SIMIX_sem_get_capacity() function
2234 - Fix interactions with processe resume/suspende
2235 - release_forever() was stupidly broken
2236 - Fix SIMIX_display_process_status() for processes in a semaphore
2237 - Make SIMIX_sem_block_onto() user-visible
2238 * Refactoring context stuff:
2239 - Use pseudo-OOP for better modularity
2240 - reimplement SIMIX_process_kill() without process_schedule() so
2241 that the latter can take as invariant that it is called from
2243 - Merge context_start into context_new for sake of simplicity
2245 * Add a Vivaldi network model, coded live during SUD'10 ;)
2246 * Rename configuration variables to start a hierarchy:
2247 o cpu_model -> cpu/model
2248 o network_model -> network/model
2249 o workstation_model -> workstation/model
2250 * New configuration variables:
2251 o network/bandwidth_factor: correction to bandwith
2252 o network/latency_factor: correction to latency
2253 o netwotk/weight_S: correction to the weight of competing streams
2254 * Add a long description to the models, that users can see with such
2255 argument on the command line: --cfg=cpu/model:help
2256 * --help-models display the long description of all known models
2258 * config: add the ability to set a default value after registration
2259 Does not override any previously set value (e.g. from cmd line)
2260 * dict: allow to have integer key and data.
2261 When so, you need to use the following functions
2262 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2263 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2264 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2265 In contrary to regular dicts, the key is not malloced before copy.
2266 Mixing scalar and regular elements in the same dict is not tested
2268 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2269 Tracing for Visualization:
2270 * SimGrid is now instrumented in order to generate a trace file for
2271 visualization analysis: to use it, need to compile SimGrid with the
2272 "tracing" option enabled, and instrument the program using SimGrid with
2273 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2274 (among other functions).
2275 * The instrumentation only traces the platform utilization for now
2276 * Documentation to use the tracing functions and how to analyze the
2277 traces with the Triva tool is written.
2278 * More information about: SimGrid FAQ (in the section Tracing Simulations
2281 * We moved to cmake as default build system. Autotools support will
2282 be dropped soon. Check the FAQ for more info about how to use it.
2283 * Greatly improved our cdash/ctest interactions
2284 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2285 * Added memory checking tests with valgrind; lot of memleak fixing.
2286 This may be the first release of simgrid with so few memory issues
2287 * Added code coverage tests.
2288 Our coverage is still improvable, but at least we see it on cdash.
2290 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2292 SimGrid (3.3.4) stable; urgency=low
2294 The "Desktop Grid needs love too" release (also called Xmas release).
2296 Models improvements:
2297 * Major speedup in the maxmin system solving by using lazy evaluation
2298 Instead of solving completely the maxmin system at each iteration,
2299 only invalidate (and recompute) the modified parts.
2300 This new feature is enabled in default models but you can try to
2301 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2302 * Cas01 IMproved as default CPU model
2303 This CPU model is the same Cas01 model, but it uses the
2304 maxmin-selective-update flag and a heap structure to manage
2305 actions on SURF kernel.
2306 It reduces the complexity to find the next action to finish and,
2307 consequently, it's faster than the old Cas01.
2308 This is the new default CPU model (Cas01).
2309 * Rename the old Cas01 model to Cas01_fullupdate
2310 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2311 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2312 * CpuTI (CPU Trace Integration)
2313 A new CPU model whose objective is simulate faster when using
2314 availability trace files.
2315 Instead of using a full featured, over engineered maxmin system for
2316 CPU modeling, this model does the pre-integration of traces files
2317 to calculate the amount of CPU power available, and so, executes
2318 faster than the old CPU models.
2319 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2320 * Use LV08 as default network model since it gives better accuracy
2321 for small messages and shouldn't change things for big ones.
2322 Use --cfg=network_model:CM02 to get the previous behavior.
2325 ******************************************
2326 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2327 ******************************************
2328 * The new CPU model may changes simulations!
2329 The point is that events occurring at the exact same timestamp
2330 are not scheduled in the same order with the old and new
2331 version. This may be enough to completely change the execution
2332 of simulations in some cases.
2333 * The new network model will change simulations!
2334 This new model is more realistic than the previous one, so you
2335 should consider redoing your old experiments with this model.
2336 Sorry for the inconvenience.
2339 * Introduce the supernovae compilation mode
2340 When compiled that way, the whole SimGrid (or almost) is put in a
2341 single compilation unit and compiled in one shoot.
2342 This is to help gcc which has difficulties to inline stuff from one
2344 The speedup seem to be above 15%, althrough more tests are needed on
2345 amd64 to confirm that gain.
2348 * Port of MSG's mailbox on top of SIMIX network
2349 The put/get mechanism was greatly simplified on the way.
2352 * New SIMIX network module. Provides:
2353 - Mailbox: rendez-vous mechanism to find with who you want to speak
2354 - Synchronous send/recv: easier and hopefully faster since the
2355 logic is handled in the maestro process directly now
2356 - Asynchronous send/recv: you dreamt of it? It's here now
2357 Too bad that nobody cared enough to propagate the change to MSG.
2358 * Add semaphores as SIMIX synchronization mechanism.
2361 * new function SD_daxload(char*) to load a DAX file
2362 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2363 * Introduce typed tasks. Specify its kind and cost at creation.
2364 At scheduling, just give where it should be placed, and the cost
2365 for each involved resource is automatically computed.
2366 Existing constructors so far (more to come of course):
2367 - SD_task_create_comm_e2e() for end-to-end communication
2368 - SD_task_create_comp_seq() for sequential computation
2369 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2370 * new function SD_task_dump() for debuging display
2371 * new function SD_task_dotty(task,FILE*) writing to file the info
2372 about the task in dotty format
2373 * SD_task_dependency_exists() can now cope with having one of its
2374 arguments NULL. If so, it tests whether the other argument has any
2376 * Add getters on list of preceding/following tasks:
2377 SD_task_get_parents(task) and SD_task_get_children(task)
2378 * Add getters on amount of workstations and list:
2379 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2380 * Add getter on task kind: SD_task_get_kind(task)
2381 * Update the start_time and finish_time of tasks on completion/failure
2382 * Bugfix: Remove task from state swags when destroyed
2385 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2388 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2389 dynar with one separate thread per value of the dynar.
2390 * Change the prototype of xbt_thread_create(), sorry.
2391 Added a boolean parameter indicating whether we want to join this
2392 thread (used in SG only for now)
2393 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2396 * GTNetS wrappers should now be usable again (and betterly tested too)
2397 * Fix a major regression from 3.2 where the timeout provided to
2398 MSG_task_put_with_timeout() was used as absolute time before which
2399 the comm should be done.
2400 * Start to fix the <cluster> tag.
2401 - Internal links should be good now (beside of the loopback, which
2402 use the private link instead)
2403 - paths to the external world is still rather broken
2404 - the <route:multi> tag is just broken. Actually that's brain-dead.
2405 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2406 to make it less stupid
2407 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2408 * Fix a source-level compatibility glitch from 3.2: after defining
2409 MSG_USE_DEPRECATED, you can use the old name
2410 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2411 * Allow to compile from the SVN with automake 1.11
2412 * Fix some problems when using the "start_time" tag in deployment XMLs.
2413 * Fix #8569: XBT/synchro.h has redundant declarations
2414 * Fix #8563: MSG return values and exceptions
2415 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2416 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2417 Thanks to Robson Peixoto.
2418 * Fix a vicious bug in dictionaries inducing that some elements were
2419 not freed on xbt_dict_free()
2421 Portability report of this version:
2422 * Main portability targets:
2423 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2424 on (amd64/i386/ia64)
2425 - mac leopard on i386
2426 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2427 but nothing critical.
2428 * Other platforms: windows, AIX and others were not tested for this release
2430 Timing report of this version:
2431 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2432 scenario parameters). From 8h to a few seconds in desktop grid settings.
2433 * Supernovae brings about 25% speedup on i386.
2435 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2437 SimGrid (3.3.3) stable; urgency=low
2439 The "Need for Speed" release.
2441 The timings done to validate the 3.3.2 were faulty.
2442 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2444 The problem was a conversion from a manually handled vector to
2445 xbt_dynar_t on the critical path.
2446 xbt_dynar_foreach calls functions, inducing stack management crap.
2448 We inlined these functions and xbt_dynar_foreach is now breath taking.
2449 We also inlined xbt_swag_belong on the way.
2451 Here are some approximate speedup measurements (on master/slaves
2452 simulations lasting between 10s and 20s each):
2453 3.3.1 -> 3.3.2: about same performance
2454 3.3.2 -> 3.3.3: 40% speedup
2455 3.3.1 -> 3.3.3: 40% speedup
2456 3.3.1 with inline patch -> 3.3.3: 30% speedup
2458 Our reading is that the refactoring which occurred in 3.3.2 made us
2459 suffer much more from the xbt_dynar_foreach low performance, but
2460 once we solved this, this refactoring proved to be very performance
2461 effective. From the 40% speedup, somehow, 10% are due to the
2462 inlining and 30% to the refactoring.
2464 That's a pitty that gcc cannot inline functions placed in other files
2465 alone. We have to choose between:
2466 - break the encapsulation (by putting private data structures and
2467 accessors in headers files to help gcc)
2468 - live with low performance
2469 - switch to a decent compiler such as icc (not quite possible).
2471 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2473 SimGrid (3.3.2) stable; urgency=low
2475 The "Simplicity does not preceed complexity, but follows it" release.
2477 The main contributors of this release were (lexical order):
2478 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2481 * Extract the routing logic into its own object.
2482 (was dupplicated in network.c and workstation_LV07.c;
2483 Allows to implement other ways of storing that info)
2484 => kill now useless network_card concept
2485 - Use dynar to represent routes (instead of void** + int*)
2486 - kill link_set (use surf_network_model->resource_set instead)
2487 - Add a command-line option to choose the routing schema to use
2488 - Add three new models:
2489 * Floyd (shortest path computed at initialization)
2490 * Dijikstra (shortest path recomputed all the time)
2491 * Cached Dijikstra (shortest path computed on need)
2492 All these models where contributed by Silas De Munck, and are
2493 described in his ICCS09 paper.
2495 * Simplify model declaration
2496 (less redirections, less function to write when defining a model)
2497 - Factorize stuff between models:
2500 surf_model_resource_set(model)
2501 surf_model_resource_by_name(model, name)
2502 - Unify the types of models in s_surf_model_t (using an union)
2503 - Embeed fields of common_public directly into s_surf_model_t
2504 - Rename model methods:
2505 action_free ~> action_unref
2506 action_change_state ~> action_state_set
2507 action_get_state ~> action_state_get
2508 - Change model methods into functions:
2509 (model)->common_public->action_use ~> surf_action_ref
2511 * Implement a generic resource; use it as ancestor to specific ones
2512 (allows to kill duplicated code in models)
2513 Drawback: timer command don't need no name nor properties;
2514 workstation_CLM03 don't need no properties
2515 (but I guess we can live with those few bytes wasted)
2517 * Improve the action object model
2518 - implement a constructor avoiding dupplicated code about field
2519 initialization in generic_action part.
2521 * Kill the SDP model: it has an external dependency, is deprecated
2522 in flavor of modern lmm models, and didn't compile since a while
2525 * Relocation of the context module from XBT to SIMIX.
2526 (the context were decoupled from the simix processes, duplicating a lot of code)
2527 => a lot of code was factorized
2528 - less overhead is introduced during scheduling
2529 - simpler API for the context factory
2530 - the logic for process creation,destruction and manipulation was simplified
2531 * Simplification of the s_smx_process_t data structure.
2532 => accesing the simix level data associated to a process is faster now,
2533 and the code is a lot more readable.
2536 * Implement some more MPI primitives:
2537 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2538 -implementation: Bcast: flat or 2-ary tree (default),
2539 Barrier: 4-ary tree,
2541 Allreduce: Reduce then Bcast
2542 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2543 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2544 Alltoallv: flat tree, like ompi
2546 * Add support for optimized collectives (Bcast is now binomial by default)
2547 * Port smpirun and smpicc to OS X
2550 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2551 Shout out if you used it.
2554 * Display the list of still queued messages in SG mode when existing
2558 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2559 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2560 * Add xbt_str_from_file(FILE*)
2561 * Add xbt_dict_get_key achieving a linear reverse search
2562 * Remove the context module
2564 Portability report of this version:
2565 * Main portability targets:
2566 - Linux(debian)/x86/context
2567 - Linux(debian)/x86/pthreads
2568 - Linux(debian)/amd64/context
2569 - Linux(debian)/amd64/pthreads
2570 On these, we still have the eratic breakages of gras/pmm and
2571 amok/saturate_sg reported in previous version. We still think
2572 that the tests are the cause of the fault, not the tested code.
2574 - Mac OSX Leopard/x86/context
2575 Still false negative in tesh autotesting.
2576 Smpi still fails, but this time because readlink does not accept -f
2577 Everything seems to work properly beside of that.
2580 - AIX version 5.3 (only tested contexts this time)
2581 Smpi still fails there because mktemp is not installed.
2582 Everything seems to work properly beside of that.
2584 I managed to compile it for the first time, but several breakages.
2585 Won't delay the release for this exotic platform.
2587 * Windows: it's still lagging behind. If you want to help, please
2590 Timing report of this version:
2591 This version seem to be more than 5% faster than 3.3.1 (on linux
2592 64bits with contextes). The gain is less than expected, we are
2593 investigating this for next release.
2595 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2597 SimGrid (3.3.1) stable; urgency=low
2600 * Implement a --cfg-help to show existing configuration variables
2601 * Build chain do not require doxygen in maintainer mode
2604 * fix a bug on struct sizeof computation, which prevented the
2605 exchange of arrays of structs in some conditions
2606 - added a regression test about this in datadesc_usage
2607 * Allow the exchange of 0-long dynamic vectors.
2608 - for that, use -1 as indicator of dynamic size instead of 0
2609 - This implied to change any size from unsigned long to long,
2610 reducing a bit communication abilities, but I guess that with
2611 64bits being quite common, this is more than enough.
2612 - This also induce a protocol change, thus bumping network protocol
2613 version from 0 to 1 (if we have external users, we have to get
2614 clean on that point too ;)
2615 - added two regression tests about this in datadesc_usage
2616 * Be more verbose when propagating local exceptions
2617 This helps debugging.
2618 * Display the status of simulated processes when receiving SIGINT in
2622 * Allow to control the simulation from a trace file.
2623 New functions MSG_action_register() and MSG_action_trace_run()
2624 The first one allows to associate a function execution to each
2625 kind of action while the second one parses a trace file and
2626 triggers the corresponding actions within the system.
2627 For now, only a toy example is provided in examples/msg/actions
2628 * Add an exemple of process migration in examples/msg/migration
2629 * Fix a bug in task exchange which broke MSG_task_get_sender()
2630 Add a teshsuite regression test for that.
2631 [Bug: if MSG_task_get_sender() is called after sender exit,
2633 * Fix a bug which prevented suspend/resume to work properly
2634 * Display the status of simulated processes when receiving SIGINT
2635 This fixes a regression of v3.3. due to the introduction of SIMIX
2636 * Bug fixing in failure management:
2637 - trace could not start by a failure at time 0
2638 - failure during communications were not working
2641 * Add SIMIX_process_set_name() to change the name of the current
2642 process in the log messages.
2643 * Store smx_hosts in a dict since we only retrieve them by name
2644 * Move the configuration infrastructure to surf
2647 * Move the configuration infrastructure to surf
2650 * Massive internal cleanups:
2651 - Store internal structures on processes instead of hosts (allows
2652 to have more than one process per host, in addition of being more
2654 - Cleanup the initialization/finalization process
2655 - Kill a whole bunch of unneeded synchronization:
2656 processes run in exclusive manner within the simulator
2657 - Move queues from global tables to process data fields
2659 - now accept -platform and -hostfile arguments
2660 - Pass the right rank value to processes according to the hostfile
2661 * Compile the examples by default, and use them as regression tests
2662 * Implement MPI_Wtime()
2663 * Change the reference speed to a command line option
2666 * TCP_gamma can now be specified as command line option using
2667 --cfg=TCP_gamma:10000000.0
2668 * Change the --surf-path cmd line option into --cfg=path:
2671 * Also include strbuff from xbt.h public header
2672 * xbt_ex_display(): do not free the exception after displaying
2673 This allows to do more with the given exception afterward.
2674 Users should call xbt_ex_free() themselves.
2678 Portability report of this version:
2679 * Main portability targets:
2680 - Linux(debian)/x86/context
2681 - Linux(debian)/x86/pthreads
2682 - Linux(debian)/amd64/context
2683 - Linux(debian)/amd64/pthreads
2684 These targets fail about 1/10 of times on gras/pmm, but we believe
2685 that this is because of the test, not because of simgrid.
2686 amok/saturate_sg fails even more rarely, and the test may not be
2689 - Mac OSX Leopard/x86/context
2690 The test suite still spits tons of errors because some obscure
2691 force prevents us from removing the temporary directories
2692 arguing that they still contain some metadata I've never heard of.
2693 Smpi fails because seq is not installed.
2694 Everything seems to work properly beside of that.
2697 - AIX version 5.3 (both contexts and pthread)
2698 Smpi still fails there because mktemp is not installed.
2699 XML inclusions seems rosty on AIX.
2701 * Windows: it's still lagging behind. If you want to help, please
2704 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2706 SimGrid (3.3) stable; urgency=high
2710 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2711 [Malek Cherier & Mt]
2713 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2716 Until now, GRAS were using MSG as an interface to SURF. It was
2717 quite difficult because both interface have several differences
2718 (MSG channels vs GRAS sockets were the most notable point).
2720 This also opens the gate to SMPI (which should occur soon) and speed
2721 up simulations by to 40% (even if it were not the main goal).
2723 **************************************
2724 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2725 **************************************
2726 The point is that events occuring at the exact same timestamp are
2727 not scheduled in the same order with the old and new version. This
2728 may be enough to completely change the execution of simulations in
2729 some cases. Sorry for the inconvenience.
2731 * Cleanup and upgrade the XML format to push further scalability
2732 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2734 * Improve the testing infrastructure with tesh. Now a very large part of
2735 the code is tested not only by being run but also by checking that the
2736 output match an expected output [Mt].
2738 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2739 is really less memory-demanding, which should allow you to use
2740 larger files in SimGrid [AL].
2742 * Inform valgrind about our contextes, so that it becomes usable
2743 with the default (and more effecient) version of SimGrid
2744 [contributed by Sékou Diakite, many thanks]
2747 * Introduce a listener thread in charge of receiving incoming
2748 messages from the network. It allows to overlap communication and
2749 computation but most notably, it removes some stupid deadlocks due
2750 to the fact that so far, a process could not send and receive at
2751 the same time. This made most non trivial communication schema
2753 * Convert the PIDs from long int to int to match the MSG ones (and
2754 linux ones too) [Mt]
2755 * New function: gras_agent_spawn() to launch a new process on
2756 current host. Only working in simulation for now. [Mt]
2757 * New function: gras_os_hostport() returning a constant form (ie,
2758 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2761 * Make the backtrace of exceptions more human readable [Mt]
2762 * New module: xbt/str [Mt]
2763 a ton of string utility functions (split, join, printf to a newly
2764 allocated buffer, trim, etc)
2765 * New module: xbt/hash [Mt]
2766 SHA1 hashing algorithm (more to come if needed)
2767 * New module: xbt/synchro [Mt]
2768 synchronization tools (mutex and conditions) working the same way
2769 in simulation and in real life (mainly useful for GRAS, but not
2771 * New module: xbt/queue [Mt]
2772 classical producer/consumer synchronization scheme
2773 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2774 (using the classical functions will get serialized) [Mt]
2775 * Make dictionary internal table dynamic. No need to specify its size
2776 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2778 * Make sure the log channels are organized as a tree under windows
2779 (because of ANSI C compatibility issue, any channel were child of
2780 root directly) [Mt].
2783 * Cleaned many thing in surf and fixed a few bugs [AL].
2784 * Add a nice command line configuration mechanism to compose models [AL].
2785 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2786 the previous one (KCCFLN05). It relies on something that looks like
2787 a max-min sharing mechanism but cannot be written as such. A new solver
2788 was thus designed [AL].
2789 * Add a new solver to lmm. Based on Lagrange optimization and
2790 gradient-based descent, it enables to efficiently maximise systems s.a
2792 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2794 This solver enables to propose two new network models for TCP Reno and
2795 TCP Vegas based on Low's work. These models still need to be fully
2796 tested though [Pedro Velho].
2799 * Bug fix in SD_simulate. Now the time bound given as argument is
2801 * Use the new parallel task model (ptask_L07) as default.
2802 * Use the SURF command line configuration mechanism.
2803 * 0-size tasks (for synchronization) should now work.
2805 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2807 SimGrid (3.2) stable; urgency=high
2811 We still experience issues on this platform, but we believe that at
2812 least MSG is usable.
2814 GRAS API BREAKAGE (for simplification purpose, sorry):
2815 * the gras_msgtype_by_name is not used anymore. Instead of
2816 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2817 you can write (and must)
2818 gras_msg_send(toserver, "request", &request);
2819 - If you still want to pass a gras_msgtype_t to the function (to cache
2820 the type and avoid the lookup time), use the gras_msg_send_() variant.
2821 - Impacted functions:
2822 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2823 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2824 * The callbacks are now expected to return 0 when everything went well
2825 (just like the main() function)
2827 GRAS new features and improvements:
2828 * New module mechanism where user code can use per process globals [Mt]
2829 This is similar to gras_userdata_*() functions, but for libraries. It
2830 factorize some code developped over and over in the examples and AMOK.
2831 It has still to be documented and used (only amok/peermanagement is
2833 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2834 loss when they were small enough to fit into the buffer and sent quickly
2835 enough so that they can all get received in one shoot.
2836 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2837 if not found. Use the *_or_null() variant for the old semantic.
2838 * In gras_msg_handle, do not discard messages without callback.
2839 They are probably messages to be explicitly awaited later (ie, proofs of
2840 mis-synchronization in userland since they are sent before being awaited)
2841 No big deal usually.
2842 * gras_socket_meas_send/recv: semantic changed!
2843 The numerical arguments used to be (1) the total amount of data to send
2844 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2845 messages. This was need for the fool willing to send more than MAXINT
2846 bytes on quite fat pipes.
2849 * Do really rename the hostmanagement module to peermanagement. [Mt]
2850 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2851 but this is rather new and this was documented in the module
2852 documentation (poor excuses, I admit)
2853 * Bandwidth measurement semantic changed! This follows the changes to
2854 gras_socket_meas_send/recv explained above.
2857 * A sequential mode has been added to the workstations. When a workstation
2858 is in sequential mode, it can execute only one task, and the other tasks
2859 are waiting in a FIFO. [Christophe Thiery]
2862 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2863 model for SIMDAG. [Christophe Thiery]
2864 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2865 equal to 0, which caused some bad behaviors in
2866 saturated_constraint_set_update. I now use a threshold mechanism like in
2870 * When running manually src/testall, you select specific units [Mt]
2871 testall is the result of our cunit mechanism, and should replace all
2872 the scripty thingy around since bash don't run easily on billware.
2874 * A mallocator system has been added. [Christophe Thiery]
2875 Mallocators allow you to recycle your unused objects instead of freeing them
2876 and allocating new ones.
2878 Documentation update:
2879 * FAQ reworking + New FAQs:
2880 - "Valgrind spits tons of errors!" [Mt]
2881 - "How to repport bugs" [Mt]
2882 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2883 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2884 - Communication time measurement within MSG [AL]
2885 - I experience weird communication times when I change the latency [AL]
2886 * GRAS tutorial [Mt]
2888 - an introduction to the framework and to the used communication model
2889 - an initiatic tour introducing the most proheminent features:
2891 . Lesson 0: Installing GRAS
2892 . Lesson 1: Setting up your own project
2893 o Part 2: Message passing
2894 . Lesson 2: Exchanging simple messages
2895 . Lesson 3: Passing arguments to the processes (in SG)
2896 . Lesson 4: Attaching callbacks to messages
2897 . Lesson 5: Using globals in processes
2898 . Lesson 6: Logging information properly
2899 . Lesson 7: Using internal timers
2900 . Lesson 8: Handling errors through exceptions
2901 . Lesson 9: Exchanging simple data
2902 . Lesson 10: Remote Procedure Calling (RPC)
2903 . Lesson 11: Explicitely waiting for messages
2904 . Recapping of message passing features in GRAS
2905 - A HOWTO section containing:
2906 o HOWTO design a GRAS application
2907 More are due, of course. They will come latter. In the meanwhile, you can
2908 check the examples which are still here.
2910 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2912 SimGrid (3.1) stable; urgency=high
2916 There was a stack corruption somewhere, visible only when optimizing
2917 with these versions. [Vince]
2920 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2921 module that enabled to play with Directed Acyclic Graphs. It is built
2922 directly on top of SURF and provides an API rather close to the old
2923 SG. Some old codes using SG are currently under rewrite to check that
2924 all needful functions are provided. [Christophe Thiery]
2927 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2928 extension of the classical CLM03 model that gracefully handles
2929 failures. This is now the default model for MSG and GRAS. It doesn't
2930 handle parallel tasks yet though. [AL]
2931 * Bug fix: Weights were not correctly set in the network part.
2932 WARNING: This may have resulted in incorrect results with simulations
2933 where there are more than one flow on a given link. [AL]
2936 * After a (long ?) discussion on simgrid-devel, we have decided that the
2937 convention we had on units was stupid. That is why it has been decided
2938 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2939 WARNING: This means that all previous platform files will not work as
2940 such with this version! A warning is issued to ask users to update
2942 A conversion script can be found in the contrib module of the CVS, under
2943 the name contrib/platform_generation/surfxml_update.pl [MQ]
2946 * Bug fix: Processes were started in reverse order, wrt deployment file.
2947 WARNING: if your code relies on this bug, please fix it. [AL]
2948 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2949 being executed on two different locations. [AL]
2950 * Bug fix: Failures are now better supported thanks to Derrick's tests
2951 (there was many failure situations I hadn't thought of and that weren't
2952 correctly handled). [AL]
2953 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2957 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2958 Exception occurring on server-side are propagated back to client (!).
2960 API CHANGE: the callback changed their prototype. Change:
2961 int my_handler(gras_socket_t expeditor, void *payload_data) {
2963 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2964 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2966 * New! function: gras_msg_handleall to deal with all messages arriving
2967 within a given period.
2968 * New! function: gras_socket_server_range to get a server socket in a
2969 range of port numbers (ease to avoid port number conflicts) [MQ]
2970 * New! gras processes display their backtrace when they get a SIGUSR1
2971 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2972 Sweet to debug RL processes [MQ]
2976 - Do not force experiment sizes to be expressed in kb, or it becomes
2977 impossible to measure the latency this way (needs one byte-long tests)
2978 WARNING: this changes the amok_bw_* function semantic. [MQ]
2979 - Implements the link saturation stuff. [MQ]
2980 * Peer management module:
2981 New! module factorizing code that we wrote over and over [MQ].
2984 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2985 - Test units are placed directly into the library code, they get extracted
2986 automatically and placed into the src/testall binary.
2987 - Convert most of the XBT tests to this system.
2988 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2989 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2990 live in the same C code. This required to change a little bit the API
2991 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2992 * New module: sparse graph structure with basic algorithms (this is work
2993 in progress and the API is not considered to be frozen yet). [AL]
2994 * Display more information on backtraces: source line & function names are
2995 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2996 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2997 * Reworked a little bit some #include statements to load only required
2998 headers. Some user code that relied on SimGrid to include stdlib or
2999 stdio may need to include it by themselves. [AL]
3000 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
3002 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
3003 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
3004 but I guess that almost nobody used those parts. [MQ]
3006 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
3008 SimGrid (3.0.1) stable; urgency=low
3011 * Unfortunately, I had missed 5 misnamed functions:
3012 xbt_fifo_item_t xbt_fifo_newitem(void);
3013 void xbt_fifo_freeitem(xbt_fifo_item_t);
3014 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
3015 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
3016 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
3017 They're now deprecated. Please use their new versions:
3018 xbt_fifo_item_t xbt_fifo_new_item(void);
3019 void xbt_fifo_free_item(xbt_fifo_item_t);
3020 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
3021 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
3022 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
3024 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
3025 * Documentation: xbt_log module unmercifully reworked [MQ]
3026 * Bugfix: there was a problem with the ending of contexts with
3027 the pthread backend. It caused some weird deadlock or behavior
3028 depending on the pthread implementation. [AL]
3029 * Bugfix: get the exceptions raised in the simulator repport where
3030 and why they come from when they are not catched in time [AL, MQ]
3033 * Bugfix: Do repport the error when two non-connected hosts try to
3034 exchange data (Thanks to Flavien for stumbling into this one) [AL]
3037 * Add additionnal checkings on communications. Assert that two
3038 communicating hosts are connected by a set of links... [AL]
3041 * Add additionnal checkings on channel values in communication [AL]
3042 * New: MSG_task_get_source to see on which host a task was generated [HC]
3043 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
3044 the number of tasks waiting to be received on channel and sent
3046 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
3047 waits for the first task coming from a given host.. [AL]
3049 GRAS new functionnalities: [MQ]
3050 * Enhance the parsing macro to allow the size of multidimentional objects
3051 to be given thru annotations.
3052 * New example (and documentation): Matrix Multiplication a la RPC
3053 (as when I was young!) and fix a bunch of bugs found on the way.
3055 GRAS performance improvements: [MQ]
3057 * Reduce the amount of cbps creation/destruction by making it static to
3058 datadesc_send/recv() and using a (newly created) cbps_reset (based on
3061 * Change libdata to a set so that we can search for stuff by ID (and thus
3062 reduce the insane amount of dict lookups)
3065 * Actually implement gras_datadesc_copy() so that we don't have to mimick
3066 RL communication on top of SG since it's so uneffective.
3067 It may also be used for inter-thread communication in RL, one day.
3068 * Use gras_datadesc_copy() to exchange messages on top of SG
3070 - improve message exchange performance on top of SG
3071 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
3072 * Don't exchange on the network the size of the used part of buffer,
3073 instead, specify the possible buffer size to read().
3075 - reduces the amount of read/write calls (one pair per exchange)
3076 - reduces the amount of exchanged data (the size)
3077 - allows to retrieve all arrived data on receiver side, if we don't need
3078 it right now (subsequent read will peek the buffer)
3079 - allows the receiver to proceed with the begining of the stream before
3080 everything is arrived
3081 - make it possible to build an iov transport (using readv/writev)
3083 - take care of the data with non-stable storage (like stacked data),
3085 * If possible, TCP send uses vector I/O (when writev() is here)
3086 - Don't use it for receive since we send data sizes and data on the
3087 same stream, so we wouldn't be able to chain large amount of chunks
3088 before having to flush the stuff to read the size.
3089 * Rework the transport plugin mechanism to simplify it and reduce the
3090 amount of pointer dereferencement when searching for the right function
3093 * I guess that now, we do almost as few system calls as possible while
3094 doing as few data copy as possible.
3096 To improve it further, we could try to send all the sizes first and then
3097 all the data (to use iov on receiving size), but it's only a partial
3098 solution: when you have 2 dimensional data, the sizes of the second
3099 dimension is data of the first dimension, so you need 3 streams.
3101 I'm not sure the potential performance gains justify the coding burden.
3103 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3105 SimGrid (3.00) stable; urgency=high
3108 * New! Give the possibility to hijack the surf parser and thus bypass
3109 MSG_create_environment and MSG_launch_application. Have a look at
3110 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3112 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3114 SimGrid (2.96) unstable; urgency=low
3119 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3120 This deprecates the xbt_error_t mechanisms.
3121 It modifies (simplifies) all XBT and GRAS API.
3122 MSG API keeps unchanged (exceptions raised by XBT are catched from
3123 within MSG and masked with existing error handling facilities)
3126 * New! Add a FATPIPE model. [AL]
3127 * New! Add a parallel task model. [AL]
3128 * New! Add automatically a loopback interface (in the default
3129 network model) if none was precised.
3132 * Bugfix: MSG_process_resume now works with the current running process.
3134 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3135 * Modification of MSG_task_get_compute_duration. Once a task has been
3136 processed, the value returned by this function is now equal to 0. [AL]
3137 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3138 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3139 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3140 of a m_task. For now, MSG_task_cancel only works with computation
3142 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3143 of the processor (in Mflop/s) regardless of the current load on the
3145 * API Change: use proper naming convention for MSG_getClock and
3146 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3148 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3149 This function changes the priority of a computation task. This priority
3150 doesn't affect the transfer rate. A priority of 2 will make a task
3151 receive two times more cpu power than the other ones. This function
3152 has been added to suit the needs of Nguyen The Loc and hasn't been that
3153 much tested yet. So if it fails, please report it and send me your code.
3155 * API Change: removed all functions and types that were marked "deprecated"
3156 since many months. Renamed MSG_global_init_args to MSG_global_init.
3158 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3160 SimGrid (2.95) unstable; urgency=low
3163 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3164 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3165 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3168 * Bugfix: Allow absolute paths for platform description files [MQ]
3169 * Bugfix: do free the variables after use. Leads to drastic performance
3171 * Implement max_duration (ie, timeouts) on resources [AL]
3174 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3176 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3177 the needed support (timeouts) [AL]
3180 * Implement measurement sockets. You can now get the bandwidth between two
3181 hosts thanks to AMOK (see below). [MQ]
3182 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3183 dynar over the network (yeah) [MQ]
3184 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3185 * simple/static token-ring example. [Alexandre Colucci and MQ]
3186 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3187 in place before [MQ]
3190 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3191 type parsing. "short" and "long" modifiers were also ignored; other
3192 modifier (such as reference level) are still ignored. [MQ]
3193 * Embeed the buffer size within the buffer itself on SG. [MQ]
3194 That way, send() are atomic and cannot get intermixed anymore (at least
3195 the ones which are less than 100k; bigger messages still have the issue)
3196 * Array size pushed by the field, not by the field type (or each
3197 and every long int will push stuff to the cbps) [MQ]
3198 * use select() to sleep since it allows to portably sleep less than one
3201 GRAS (minor cleanups)
3202 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3205 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3206 * type category 'ignored' killed as it was never used and were difficult
3208 * whether a type can cycle or not is now a flag, leaving room for more
3209 flags (as "ignored", if we feel the need one day ;)
3210 * Rename raw sockets to measurement sockets since "raw" has another
3211 meaning in networking community.
3214 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3215 and offers features not belonging to GRAS but that most applications
3216 need. One day, it may be a set of plugins loadable at runtime.
3217 * New module: bandwidth
3218 bandwidth measurement between arbitrary nodes running this module. [MQ]
3220 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3222 SimGrid (2.94) unstable; urgency=low
3224 The first beta release of SimGrid 3 !
3228 * Update the main page and the FAQ. Adding references to gforge.
3231 * Add a gras_os_getpid function.
3234 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3235 * Extend the logs so that they also print PID, hostname, date, ... if
3237 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3238 and kill the old version which were in testsuite/
3239 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3243 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3244 for users. Instead, each programming environment (GRAS, MSG,...) use it
3245 when needed to provide such a feature to users.
3246 Don't shortcut the mechanism or you will also shortcut the virtualization
3247 you need on the simulator.
3251 * Cleanups in configury with regard to compile optimization/warning flags.
3252 Also add -fno-loop-optimize to any powerpc since it's the optimization
3253 killing gcc (< 3.4.0).
3254 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3255 needed by MSG examples complications
3256 * Borrow configury beautifications from PHP
3259 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3260 warning (thanks loris for stumbling into this one).
3261 * Bugfix: stop loading private headers (gras_config.h) from the public
3265 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3266 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3267 RL would imply a DNS resolver, which is *hard* to do in a portable way
3268 (and therefore delayed).
3269 * Implement a real timer mechanism and use it in timing macros. This allows
3270 to avoid rounding errors and get a 0.000005 sec precision in timing
3271 macros. While I was at it, various cleanups:
3272 - allow to declare more than one timed section per file (fix a stupid bug)
3273 - move some private declaration to the right place
3274 - merge conditional execution and timing macros into emulation module
3275 - document the module
3276 - make sure the module cleanups its mess on gras_exit
3277 * Documentation improvements:
3278 - (new) how to compile applications using GRAS
3279 - (new) emulation support (timing macros)
3281 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3283 SimGrid (2.93) unstable; urgency=low
3285 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3288 - Use Paje properly where used. Still to be sanitized properly.
3289 - Portability fix: Add an implementation of the contexts using pthread
3293 - Add xbt_procname(): returns the name of the current process.
3294 Use it to show the current process's name in all logging.
3296 - fix detection of older flex version and the reaction, since we do
3297 depend on modern ones (we use lex_destroy)
3298 - Better separation of SG and RL in the libs: remove all simulation code
3299 from libgras. As a result, this lib is now only 200k when stripped.
3300 Some of the xbt modules may also be duplicated (two sets and such) and
3301 should be cleaned/killed before SG3.
3302 - Insist on using xlC on AIX because of weird problems involving gcc there.
3303 - Cleanup the make remote stuff. This is now done by scripts
3304 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3305 mainly for our private use, but we're working on changing them to user
3308 - Bugfix: flush the socket on close only if there is some *output*.
3309 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3311 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3314 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3316 SimGrid (2.92) unstable; urgency=low
3318 Alpha 3 on the path to SimGrid 3
3322 - New! First try of benchmarking macros.
3323 - New! First try so that gras_stub_generator generate deployment and
3324 remote compilation helpers.
3326 - Bugfix: Initialization fix in msg_test.
3330 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3332 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3334 - API Change: gras_os_sleep to take the amount of seconds as a double.
3335 Accepting an int was error prone since it was the only location where
3336 seconds were coded as such. It leaded to damn rounding errors.
3337 - Bugfix: Hard to belive that timers ever worked before this.
3339 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3341 SimGrid (2.91) unstable; urgency=low
3343 Alpha 2 on the path to SimGrid 3
3347 - Bug fix in the lmm_solver.
3349 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3350 through the function MSG_paje_output.
3351 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3352 - It is possible to bound the rate of a communication in MSG with
3353 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3354 to write it in the changelog).
3355 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3360 - Various cleanups to the autotools stuff
3361 - Begin to move Gras examples to examples/gras/
3362 - Let make distcheck work again (yeah!)
3364 - documentation overhauled using doxygen.
3365 gtk-doc-tools is dead in SimGrid now.
3366 - Automatically extract all existing logging categories, and add the list
3367 to the documentation (long standing one, to say the less)
3369 - Cleanup the known architecture table. Reorder the entries to group what
3370 should be, and use a more consistent naming scheme.
3371 (some of the test dataset are still to be regenerated)
3372 - New! Allow library to register globals on each process just as userdata
3374 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3375 lookup time (for now).
3376 Use it in msg and trp.
3377 This cleans a lot the internals and helps enforcing privacy of the
3378 headers between the gras components.
3379 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3380 - Bugfix: gras_os_time was delirious in RL.
3381 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3383 - Reenable GRAS now that it works.
3385 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3387 SimGrid (2.90) unstable; urgency=low
3389 Alpha 1 on the path to SimGrid 3
3391 * It is a long time since the last release of SimGrid. I'm sorry about
3392 that but as I had told you, I was rewriting a lot of things. I apologize
3393 to those who had been reporting bugs to me and that I had not answered.
3394 If your bug is still in the new version, please tell me. Here is a
3395 summary of the main changes.
3397 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3398 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3399 portability, speed, and a lot more but you'll figure it out later.
3400 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3401 I wanted to release the new MSG as soon as possible and I have
3402 broken GRAS, which is the reason why, for now, only MSG is fully
3403 functional. A laconic description of these projects is available
3404 in the documentation.
3406 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3407 kernel optimized for our needs (called SURF but only the developers
3408 should use it). Hence, MSG is now roughly 30 times faster and I think
3409 that by rewriting a little bit MSG, I could event speed it up a little
3410 bit more. Beside the gain in speed, it is also much easier to encode a
3411 new platform model with SURF than it was with SG. More to come...
3413 * REVOLUTION 3: I have tried to change a little as possible the API of
3414 MSG but a few things really had to disappear. The main differences
3415 with the previous version are:
3416 1) no more m_links_t and the corresponding functions. Platforms are
3417 directly read from a XML description and cannot be hard-coded
3418 anymore. The same format is used for application deployment
3419 description. The new format is described in the documentation.
3420 Have a look in tools/platform_generation. There is a tiny script
3421 that converts from the old platform format to the new one. Concerning
3422 the application deployment format, parsing the old one is tricky.
3423 I think most of you should however be able to convert your files. If
3424 it is really an issue, I can write a C code that does the conversion.
3426 2) the toolbox tbx does not exist anymore. We now have a library
3427 with much more data-structures but without the hash-tables (we have
3428 dictionaries that are much faster).
3430 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3432 *****************************************************************************
3433 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3434 * don't feel like converting the oldies. (Mt) *
3435 *****************************************************************************
3438 Version 2.90: "the long awaited one"
3439 - Finished rewriting and debugging MSG. Rewrote the documentation.
3440 - disable GRAS for now since it needs to be ported to the newest SG
3443 - Finish the port to windows (using mingw32 for cross-compile)
3446 - Main loop and datastructures of SURF. A cpu resource object is
3447 functional. Surf can thus be used to create cpu's with variable
3448 performance on which you can execute some actions.
3450 2004-11-15 Martin Quinson
3451 - Port to ARM. Simply added the alignment and size descriptions. Should
3452 work, but the ARM machines are so slow that I didn't had the opportunity
3453 to 'make check' over there yet.
3455 2004-11-15 Arnaud Legrand
3456 - Trace manager now written. It uses a heap structure and is therefore
3457 expected to be efficient. It may however be speeded up (particularly
3458 when many events occur at the same date) by using red and black
3459 trees. One day maybe...
3460 - Max-min linear system solver written. It uses a sparse matrix
3461 structure taking advantage of its expected use. Most operations are
3462 O(1) and free/calloc are called as few as possible. The computation of
3463 the minimum could however be improved by using a red and black tree
3466 2004-11-03 Arnaud Legrand
3467 - Rename every gras_* function that was in xbt/ to its xbt_
3469 - Add a heap and a doubly-linked list to xbt
3470 - Added a dichotomy to the dictionaries. make check works as well before
3471 so I assume that the patch is correct. I do not know however if things
3472 run effectively faster than before now. :)
3474 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3475 SimGrid, and the version number is bumped to 2.x
3477 2004-10-29 Martin Quinson
3478 - Introduction of the remote errors.
3479 They are the result of a RMI/RPC on the remote machine.
3480 ErrCodes being scalar values, you can't get the host on which those
3481 errors did happen. Extending the error mechanism as in Gnome is possible.
3482 No idea yet whether it is a good idea.
3484 2004-10-28 Martin Quinson
3485 - Interface revolution: the Starred Structure Eradication.
3486 I used to do typedef struct {} toto_t; and then handle *toto_t.
3487 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3488 - ???_t is a valid type (builded with typedef)
3489 - s_toto_t is a structure (access to fields with .)
3490 - s_toto is a structure needing 'struct' keyword to be used
3491 - e_toto_t is an enum
3492 - toto_t is an 'object' (struct*)
3494 typedef struct s_toto {} s_toto_t, *toto_t;
3495 typedef enum {} e_toto_t;
3496 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3497 s_toto_t) is private.
3499 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3500 so that it can set the variable to NULL. It was so for dicts and sets,
3501 it changed for dynars.
3503 - Fix a bunch of memleaks in dict_remove
3504 - Fix a bug in sg/server_socket opening: it failed all the time.
3506 2004-10-07 Martin Quinson
3507 - Speed up dynar lookup operation a bit.
3509 gras_dynar_get is dead.
3511 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3512 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3513 (which gives you the address of the stored data).
3515 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3516 copy of the data using an affectation to do the job and not a memcpy.
3518 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3519 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3521 It does not really speedup the dynar test because they are
3522 setting elements all the time (and look them seldom). But the dict does
3523 far more lookup than setting.
3525 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3527 2004-10-05 Martin Quinson
3528 - Allow to (en/dis)able the cycle detection at run time.
3530 Whether we should check for cycle or not is now a property of each
3531 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3532 datadesc_cycle_unset allow to remove this property when previously set.
3534 Note that the cycle detection is off by default since it impacts the
3535 performance. Watch the data you feed GRAS with ;)
3537 This property is hereditary. Any element embedded in a structure having it
3538 set have it set for the time of this data exchange.
3540 You should set it both on sender and receiver side. If you don't set it on
3541 sender side, it will enter an endless loop. If you forget on receiver
3542 side, the cycles won't be recreated after communication.
3544 - Header reorganization.
3545 Kill gras_private.h, each submodule must load the headers it needs.
3547 2004-10-04 Martin Quinson
3548 - Interface revolution: do not try to survive to malloc failure.
3550 Now, gras_malloc and friends call gras_abort() on failure.
3551 As a conclusion, malloc_error is not a valid error anymore, and all
3552 functions for which it was the only gras_error_t return value are
3553 changed. They now return void, or there result directly.
3554 This simplify the API a lot.
3556 2004-09-29 Martin Quinson
3557 - Re-enable raw sockets.
3558 Created by gras_socket_{client,server}_ext;
3559 Used with gras_raw_{send,recv}
3562 It should allow to kill the last bits of gras first version soon.
3564 This is not completely satisfactory yet (duplicate code with
3565 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3568 - Simplify transport plugin (internal) interface by not passing any
3569 argument to _server and _client, but embedding them in the socket
3572 2004-09-28 Martin Quinson
3573 - Finish the port to AIX.
3574 autoconf was my problem (segfault within the malloc replacement
3575 function. No idea why)
3577 2004-09-16 Martin Quinson
3578 - Fix some size_t madness on 64bit architectures.
3580 2004-09-08 Martin Quinson
3581 - Reduce the number of system headers loaded, overload some more system
3582 calls (such as malloc to cast the result of the system one, and work
3584 - Fix and reintroduce the config support
3586 2004-09-07 Martin Quinson
3587 - Source code reorganization to allow Arnaud to surf all over there.
3588 - Allow to document the logging categories.
3589 - Remove all uppercase from logging categories and useless cleanup in names.
3591 2004-08-18 Martin Quinson
3592 Version 0.6.2 (protocol not changed; API changed)
3593 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3594 gras_error_t), and NULL when not found. Functions expecting a msgtype
3595 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3596 hopefully usefull message.
3597 - Portability to prehistoric sparcs again
3599 2004-08-17 Martin Quinson
3600 Version 0.6.1 (protocol not changed; ABI not changed)
3601 - prealloc some buffers to speed things up
3603 2004-08-11 Martin Quinson
3604 Version 0.6 (protocol not changed; ABI expended)
3605 - The parsing macro can deal with the references, provided that you add
3606 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3608 2004-08-09 Martin Quinson
3609 Version 0.5 (protocol not changed; ABI changed)
3610 - Allow to off turn the cycle detection code in data exchange at
3611 compilation time. It should be at run time, but I'm short of time (and
3612 the config stuff is still broken). That way, we keep dict out of the
3613 critical path, which is good because the performance is poor:
3614 - search not dichotomial yet
3615 - dynar give no way to access their content and memcpy everytime
3616 - In composed data description (struct, ref and so on), stop foolness of
3617 keeping the subtype's ID, but store the type itself. This keeps sets out
3618 of the critical path, which is good since they rely on dynar and
3619 dictionnaries. The only loose of that is that we cannot detect the
3620 redeclaration of a structure/union with another content (but I'm not sure
3621 the code detected well this error before anyway). We still can detect
3622 the redefinition discrepancy for the other types.
3623 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3624 it breaks the code because of type-punning used all over the place).
3625 This breaks on all non-gcc architectures (for now).
3627 All those changes (plus the buffer of last time) allow me to gain 2 order
3628 of magnitude on cruel tests consisting of 800000 array of integers on two
3629 level of a hierarchical structure (200 secondes -> 4 secondes)
3632 - the selector of reference must now return the type it points to, not
3633 the ID of this type.
3635 2004-08-06 Martin Quinson
3636 Version 0.4 (protocol changed; ABI not changed)
3637 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3638 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3639 - Implement a buffer transport, and use it by default (it relies on tcp in
3640 real life and on sg in simulation).
3641 That's a bit hackish since I had a new field to the structure to store
3642 its data without interfering with the subtype ones. Inheritance
3643 is tricky in C. And that's a kind of reverse inheritance with one class
3644 derivating two classes. Or maybe a game with java interfaces. Anyway,
3645 that's damn hard in C (at least).
3646 Moreover, I got tired while trying to ensure plugin separation and
3647 genericity in SG mode. MSG wants me to do weird things, so let's go for
3648 cruel hacks (temporarily of course ;).
3649 See comment in transport_private.h:71
3650 - do not include all the _interface headers in private but in the files
3651 which really need them (to cut the compilation time when they are
3654 2004-07-26 Martin Quinson
3655 Version 0.3 (protocol not changed; ABI changed)
3656 - Major overhault of the datadesc interface to simplify it:
3657 - shorted the function names:
3658 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3659 - add a trivial way to push/pop integers into the cbps without malloc.
3660 This allows to make really generic sub_type description, which simply
3661 pop their size of the stack.
3662 - add a function gras_datadesc_ref_pop_arr() which does what users want
3663 most of the time: Declare a dynamic array (which pops its size of the
3664 stack) and declare a reference to it. Poor name, but anyway.
3665 - kill the post-send callback, add a post-receive one
3667 2004-07-23 Martin Quinson
3668 Version 0.2 (protocol changed; ABI changed)
3669 - add some testing for cpbs in the test cases, and fix some more bugs.
3670 This invalidate again the little64 data file, since I cannot regenerate
3672 - remove an awfull optimization in the logging stuff, allowing me to:
3673 - understand it again
3674 - learn gcc how to check that the argument match the provided format
3675 - fix all errors revealed by gcc after that
3676 - internal keys of dict are not \0 terminated. Deal with it properly in
3677 loggings instead of segfaulting when the user want to see the logs :-/
3679 2004-07-22 Martin Quinson
3680 - Fix some stupid bug preventing cbps (callback postit) from working
3682 2004-07-21 Martin Quinson
3683 - Some documentation cleanups
3684 - remove the useless last argument of msgtype_declare
3685 - rename the Virtu functions to fit into the 'os' namespace
3686 - move headers src/include -> src/include/gras/ and stop fooling with
3687 gras -> . symbolic link
3688 - make distcheck is now successful
3690 2004-07-19 Martin Quinson
3692 - Build shared library also
3693 - Install html doc to the right location
3694 - stop removing maintainer files in make clean
3695 - build tests only on make check
3697 2004-07-13 Martin Quinson
3699 - No major issue in previous version => change versioning schema
3700 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3701 corresponding dataset.
3703 2004-07-11 Martin Quinson
3705 - Get it working with any kind of structure (we can compute the padding
3706 bytes remotely for all the architectures I have access to)
3707 - Implement the structure parsing macro (still not quite robust/complete)
3708 - Improvement to the remote testing toysuite
3710 2004-07-10 Martin Quinson
3711 [autoconf mechanism]
3712 - get ride of a bunch of deprecated macros
3713 - actually run the test for two-compliment, not only compile it :-/
3714 - test whether the structures get packed (and bail out if yes. Damn.
3715 Alignment is a serious matter)
3716 - test whether the structures get compacted (but respecting the alignment
3717 constraints of each types)
3718 - test whether the array fields of structures can straddle alignment boundaries
3720 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3721 (took me 2 hours to find that bug, looking at the wrong place)
3722 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3723 before sending/receiving and are used to compute the offsets of fields
3724 - Given that padding size depend even on compiler options, keep track of
3725 alignment and aligned_size only for the current architecture. Not a big
3726 deal since we send structure fields one after the other (seems
3728 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3729 program, let it work on linux/gcc/little32. portability todo.
3731 2004-07-08 Martin Quinson
3732 - import and improve remote compilation support from FAST
3733 - make sure make check works on half a dozen of machines out there
3735 2004-07-07 Martin Quinson
3736 Let's say it's version 0.0.3 ;)
3737 - Implement conversions (yuhu!)
3738 - Let it work on solaris (beside conversion, of course)
3739 - Stupid me, using rand() to generate the conversion datatests in not wise.
3741 2004-07-06 Martin Quinson
3742 - Let make dist work, since I'm gonna need it to compile on remote hosts
3743 - Let Tests/datadesc_usage write the architecture on which the file was
3744 generated as first byte.
3745 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3746 architecture descriptions.
3747 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3748 on those architectures.
3749 - Optimization: send/recv array of scalar in one shoot
3751 2004-07-05 Martin Quinson
3752 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3754 - Plug a whole bunch of memleaks
3755 - each process now have to call gras_{init,exit}. One day, their log
3756 settings will be separated
3757 - Continue the code factorisation between SG, RL and common in Transport.
3759 2004-07-04 Martin Quinson
3761 - Redistribution between SG and RL.
3762 We wanna have to accept in SG, so move accepted related parts of RL in
3763 the common part. (more precisely, the dynar of all known sockets is no
3764 more a static in transport.c, but part of the process_data)
3766 [gras_stub_generator]
3767 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3769 2004-07-03 Martin Quinson
3770 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3771 them all up in one shot)
3773 - Fix the ugly case of reference to dynamic array.
3774 - New (semi-public) function gras_datadesc_size to allow the messaging
3775 layer to malloc the needed space for the buffer.
3777 - gras_socket_close now expect the socket to close (and not its address to
3778 put NULL in it after it). This is because the socket passed to handlers
3779 is one of their argument (=> not writable).
3781 - propagate the interface cleanup from last week in datadesc, ie remove a
3782 superfluous level of indirection. User pass adress of variable
3783 containing data (both when sending and receiving), and not of a variable
3784 being a pointer to the data. Let's say that I like it better ;)
3785 The price for that is constructs like "int msg=*(int*)payload" in
3786 handlers, but it's a fine price, IMHO.
3788 - Let it work in RL (yuhu)
3790 2004-06-21 Martin Quinson
3792 - porting SG plugin and SG select to new standards (works almost).
3793 - plug memleaks and fix bugs around.
3796 - cleanup the prototype of data recv and force users to specify when they
3797 want to handle references to objects. Test case working even for cycles.
3798 - plug memleaks. Valgrind is perfectly ok with this.
3800 2004-06-12 Martin Quinson
3802 - cleanup the separation between plugin and main code in plugin creation
3804 2004-06-11 Martin Quinson
3806 - Reput hook for raw sockets, needed for BW experiments
3807 - kill a few lines of dead code
3808 [Data description] Interface cleanup
3809 - gras_datadesc_by_name returns the searched type or NULL.
3810 That way, no variable is needed to use a type desc once, which makes
3812 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3813 two parameters were strings (field name, type name), leading to
3815 [Dicos] Interface cleanup
3816 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3817 This is consistant with the dynar API.
3819 2004-04-21 Martin Quinson
3821 - Porting to new standards.
3823 - interface cleanup.
3824 There is no bag anymore, no need to take extra provision to mask the
3825 pointers behind "ID".
3826 Better splitup of functions between files create/exchange/convert.
3827 This is still a bit artificial since convert and receive are so
3828 interleaved, but anyway.
3830 - add a queued message list to procdata (the ones not matching criteria
3832 - factorize some more code between SG and RL wrt procdata
3834 - use gras_exit in example to track memleaks
3835 - get rid of gs_example now that GS is properly integrated into gras
3836 - update run_test to integrate the latest tests (datadesc)
3838 - rename WARNINGn macros to WARNn since it prooved error-prone
3840 2004-04-19 Martin Quinson
3842 - register init/exit functions within gras module mechanism
3843 - send/receive function.
3844 Convertion is not implemented, but short-cutted if not needed.
3845 struct/array elements are sent one by one (instead of block-wise), but
3846 nobody really cares (yet). Get a prototype before optimizing.
3847 - tests (using a file socket) for DD send/receive on:
3848 - base types: int, float
3849 - array: fixed size, string (ie ref to dynamic string)
3850 - structure: homogeneous, heterogeneous
3851 - chained list, graph with cycle
3852 Believe it or not, valgrind is not too unhappy with the results. The
3853 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3857 - Bugfix when using a filename explicitely (instead of '-')
3859 2004-04-09 Martin Quinson
3861 - factorize more code between RL and SG in socket creation
3862 - Complete the implementation and tests of:
3864 o file (only in RL, and mainly for debugging)
3866 I lost 3 days to design a portable address resolver, and then decided
3867 that the prototype mainly have to run on my box.
3868 Addressing portability too early may be like optimizing too early :-/
3870 - use gras_init in the Tests instead of the crappy parse_log_opt
3871 (the latter function is removed)
3872 [Conditional execution]
3873 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3874 [Code reorganisation]
3875 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3876 Build examples against the RL library, since there is no way to disable
3877 its creation for now.
3879 For information, the beginning of coding on GRAS was back in june
3880 2003. I guess that every line has been rewritten at least twice since