1 SimGrid (3.21) October 3. 2018
3 The Restarting Documentation (TRD) Release.
6 - Convert the user manual to Sphinx for improved usability.
7 Unstable version now hosted on https://simgrid.frama.io/simgrid
8 Many glitches remain; stay tuned.
9 - Restrict the Doxygen reference API to the installed header files.
10 (documenting all internals this way is overkill + doxygen scales badly)
11 - New tutorial on S4U (highly inspired from the old MSG one, but with
12 a git repository to fork as a starting point, and a docker image)
13 - Started but not finished a SMPI tutorial inspired from the JLPC'14 one.
14 - The Developper manual is still to be converted (not compiled until then)
15 - Some parts are still missing in this conversion (such as the
16 platform and deployment sections) while others were blindly converted
17 and would need more love (such as the configuration flags).
18 Things will certainly further improve in the future releases.
21 - Fully reimplement auto_restart mechanism that was utterly broken
22 (fix #22, #131 and #281 that were all very related)
23 - Implement semaphores (fix #303 and #291)
24 - When creating an actor from a function and its parameters,
25 move-only parameters are not allowed anymore, as it would prevent
26 the actor to be restartable if its parameters are consumed on run.
27 - s4u::Io: IOs go asynchronous as activities. New methods in s4u::Storage:
28 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
29 IO operations that can be started, waited for, or canceled as a regular
31 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
35 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
36 - New host variable: 'core_count'
39 - Remove xbt_os_thread_specific features
40 - Remove portability wrapper to condition variables
41 - Remove xbt_os_thread_yield()
44 - MPICH collective selector now mimics MPICH 3.3b
45 - OpenMPI collective selector now mimics OpenMPI 3.1.2 (default "tuned" setting)
46 - MPI_Init(NULL,NULL) is now allowed
47 - smpi/wtime option now injects time in gettimeofday or clock_gettime calls
48 - Command-line arguments should now be handled properly in Fortran simulations
49 - MPI Attributes and related callbacks should work in Fortran
50 - Apps using C + Fortran should now work
51 - MPI_* calls now check for non MPI_SUCCESS return values and emit warnings
52 - Support MPI_Error_String
55 - Due to an internal bug, Msg.run() must now be your last line.
56 We hope to fix it in a future release, and we are sorry for the inconvenience.
59 - #22: Process autorestart seem to only work with CAS01 cpus
60 - #56: Feature request: dynamic SMPI replay
61 - #93: simgrid should not eat --help
62 - #111: How to change the loopback link characteristics?
63 - #116: Platform cluster radical
64 - #117: Platform cluster/host order
65 - #129: Replay_multiple on steroid
66 - #131: Java : autostart does not work
67 - #143: Setting a breakpoint at a given time
68 - #235: Network energy models should be integrated
69 - #264: Add ptask L07 resource tracing
70 - #271: Dynamic and manual replay of SMPI traces
71 - #279: Breakpoints option uses time deltas instead of absolute time
72 - #281: Daemonized actor no longer daemonized after an auto-restart
73 - #282: TIT convertor erase traces when using absolute path in the trace list file
74 - #285: segfault when a s4u actor kills itself with msg process tracing activated
75 - #286: Pajé traces are not exposing the number of cores
76 - #287: Command-line parsing should stop when encountering '--'
77 - #288: MPI_Init(NULL, NULL)
78 - #289: Improve documentation of Actor::on_destruction and this_actor::on_exit()
79 - #290: Method to check if a mailbox has a message ready to be consumed
80 - #291: [s4u] semaphores do not seem available yet
81 - #294: Alltoall collective implementation of mpich
82 - #295: Starting a Host as OFF is not supported
83 - #296: DTD too permissive
84 - #299: Add a s4u barrier example
86 ----------------------------------------------------------------------------
88 SimGrid (3.20) Released June 24. 2018
90 The proxy snake_case() release.
93 - The whole API is now using snake_case() instead of mixing it with camelCase.
94 Compatibility wrappers in place for one year.
95 - We took the opportunity to increase the consistency of the API.
96 Please report any remaining glitches.
97 - New function simgrid::s4u::Host::get_actor_count:
98 Returns the number of actors running on a specific host.
101 - SMPI is now tested with ~45 proxy apps from various sources, with none or
102 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
103 - Replay: The replay file has been re-written in C++.
104 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
105 supported. They are stored in the trace and used when replayed.
106 - Basic support of MPI_Cancel. Robustness not guaranteed.
107 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
108 (only for MPI_COMM_TYPE_SHARED).
109 - New option: smpi/privatize-libs, to add external shared libs to be privatized
110 by SMPI. They will be copied locally and loaded separately by each process.
111 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
112 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
113 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
114 with internal index optind with multiple processes. Only works if MPI_Init has
116 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
117 syntax without preprocessing tricks.
120 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
121 behavior tracing is no longer limited to MSG
122 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
123 this means that traces now need to be updated or re-obtained; both Irecv
124 and Isend lines in your traces have now in total 5 mandatory fields:
125 <rankid> <command> <to/from rankid> <tag> <size>
126 To update your traces, it suffices to add a 0 for the tag here.
127 - TIT now also supports waiting for a distinct request via MPI_Wait.
128 Wait/Test now wait for a specific request, not just the last one that was
129 issued. This unfortunately means another update, because we need to
130 identify which request you want to wait for. We do this via the
131 triplet (sender, receiver, tag), which needs to be added:
132 <rankid> <command> <sender> <receiver> <tag>
133 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
134 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
137 - The deprecation of MSG is ongoing (but this should not impact you).
138 Many MSG functions are now simple wrappers on the C API of S4U. If
139 you wish to convert your code to S4U, find the S4U counterparts of
140 your MSG calls in src/msg/msg_legacy.cpp.
141 - MSG can still be used, but won't evolve anymore.
144 - Allow to run the Link energy plugin from the command line with
145 --cfg=plugin:link_energy
146 - Rename Energy plugin into host_energy
147 - Rename Load plugin into host_load
150 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
151 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
152 - Improve performance when handling timeouts of simix synchros.
155 - Config: the C API is now deprecated (will be removed in 3.23), and
156 the C++ API has been slightly improved.
159 - Fix several build issues on OSX.
160 - Move simgrid_config.h to simgrid/config.h (old header still working)
163 - #143: Setting a breakpoint at a given time
164 - #150: Inconsistent event names in SMPI replay
165 - #258: daemonized actors hang after all non-daemonized actors have completed
166 - #267: Linker error on unit_tmgr
167 - #269: SMPI: tracing of MPI_Wait/all/any broken
168 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
170 ----------------------------------------------------------------------------
172 SimGrid (3.19.1) Released March 22 2018
174 The Fixed ABI Release.
176 As you may know, we are currently refactoring SimGrid in deep.
177 Upcoming SimGrid4 will be really different from SimGrid3: modular,
178 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
180 Our goal is to smooth this transition, with backward compatibility
181 and automatic update paths, while still progressing toward SimGrid4.
183 SimGrid remains open during works: The last pure SimGrid3 release was
184 v3.12 while all subsequent versions are usable alpha versions of
185 SimGrid4: Existing interfaces remain unchanged, but the new S4U
186 interface is budding and the internals are deeply reorganized.
188 Since 2015, we work hard to reduce the changes to public APIs. When
189 we need to rename a public library symbol in S4U, we let your
190 compiler issue an explicative warning when you use the deprecated
191 function. These messages remain for four releases, i.e. for one full
192 year, before turning into an error. Starting with v3.15, your can
193 also adapt to API changes with the SIMGRID_VERSION macro, that is
194 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
196 Starting with this v3.19.1, our commitment to reduce the changes to
197 the public interfaces is extended from the API to the ABI: a program
198 using only MSG or SimDag and compiled against a given version of
199 simgrid can probably be used with a later version of SimGrid without
200 recompilation. We will do our best... but don't expect too much of
201 it, that's a really difficult goal during such profund refactoring.
203 The difference between v3.19 and v3.19.1 is that the former was
204 accidentally breaking the ABI of MSG, while the later is restoring
207 S4U and kernel APIs will still evolve until SimGrid4, with one-year
208 deprecation warnings as currently. In fact, cleaning up these
209 interfaces and converting them to snake_case() is one release goal of
210 v3.20. But don't worry, we are working to smooth this upgrade path.
212 In summary, new projects should start with S4U to benefit of the
213 future, but old MSG projects should still be usable with no change.
215 ----------------------------------------------------------------------------
217 SimGrid (3.19) Released March 20 2018
219 The Moscovitly-cold Spring release.
221 SMPI user-visible improvements:
222 - Support of MPI_Type_create_subarray
223 - Unimplemented MPI calls now abort (instead of issuing a warning)
224 Some harmless calls around errhandlers still only issue a warning.
225 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
226 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
227 is now possible (mmap privatization was refactored to this end).
228 Adding new SMPI instances should also be technically doable now.
230 SMPI internal cleanups:
231 - Removed index notion from SMPI -- no more "getPid() - 1"!
232 In tracing files, actors are called rank-<PROCESS-ID> but these are
233 NOT the real ranks (tracing several communicators is not supported yet).
234 - Deprecate smpi_process_index() (will be removed in 3.22)
235 - Better testing on classical proxy apps, through an external project:
236 https://github.com/simgrid/SMPI-proxy-apps/
239 - Use a graphical TOC to make it easier to find the documentation you need
240 - Revamp the MSG tutorial
243 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
244 - Remove parameter of MSG_process_killall().
245 Resetting the PID was bogus anyway (several actors could have the same PID).
248 - Execution->setHost() can be called after start(), to migrate it.
249 - Comm::test_any() is now implemented.
250 - s4u::Actor now has onCreation() and onDestruction() signals.
251 - Install some kernel header files for the users' plugins and more.
252 Warning, their API is really not stable yet.
253 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
254 include/simgrid/kernel/routing/*Zone.hpp
257 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
259 Build System and other cleanups:
260 - Remove unused run-time parameter "tracing/onelink-only".
261 - SimGrid now works with intel compilers (MC still buggy)
262 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
263 - tesh kills subprocesses on timeout (but not on windows)
266 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
269 - #194: Feature request: simgrid::s4u::Comm::test_any()
270 - #245: migrating an actor does not migrate its execution
271 - #253: Feature Request: expose clusters as objects
272 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
273 - #255: Tesh broken on Windows
274 - #256: Modernize FindSimGrid.cmake
275 - #257: Fix (ab)use of CMake install
277 ----------------------------------------------------------------------------
279 SimGrid (3.18) Released December 24 2017
281 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
283 MSG IS NOW DEPRECATED. Please use S4U for new projects.
284 - Support for MSG should not be removed from SimGrid before 2020,
285 but future evolutions will be limited to the integration of
286 user-provided patches.
289 - s4u::onDeadlock() signal, run before stopping the simulation.
290 - s4u::Exec: asynchronous executions (abstraction of background
292 - s4u::Host->getLoad() returns the achieved speed in flops/s
295 - Link::name() is deprecated and will be removed in v3.21, use
296 Link::getCname() instead.
297 - Mailbox::getName() changed to return a std::string, use
298 Mailbox::getCname() instead to get a char*.
299 - Storage::getName() changed to return a std::string, use
300 Storage::getCname() instead to get a char*.
301 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
302 to have both allocation and deallocation of the map in user space.
305 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
306 The amount of flops remaining to do is a vector, not a scalar.
307 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
308 name implies on both sequential and parallel tasks.
309 - Both changes fix GitHub's #223 using PR #237 as a basis.
310 Thanks Michael Mercier.
311 - Most examples were converted to S4U and hidden elsewhere at tests.
312 Rationale: we still want MSG to work; we want newcomers to use S4U.
315 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
316 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
319 - Switch to the faster dlopen privatization mechanism by default
320 - Documentation improvments
323 - Live migration is getting moved to a plugin. Dirty page tracking is
324 the first part of this plugin. This imply that VM migration is now
325 only possible if one this function is called:
326 - C/MSG: MSG_vm_live_migration_plugin_init()
327 - C/C++: sg_vm_live_migration_plugin_init()
328 - Java: Msg.liveMigrationInit()
329 For C and C++, "simgrid/plugins/live_migration.h" also has to be
333 - Define class simgrid::xbt::Path to manage file names.
334 - Removed unused functions:
335 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
336 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
337 xbt_graph_export_graphviz()
338 - xbt/str.h: xbt_str_join()
339 - Remove unused datatypes:
340 - xbt/heap.h: use std::priority_queue or boost::heap instead
341 - xbt/swag.h: use boost::intrusive::list instead
344 - New link_energy plugin for the consumption of the links.
345 - All of the operations on files and storage contents have been
346 packaged into a plugin (src/plugins/file_system). The current
347 public interface can be found in
348 include/simgrid/plugins/file_system.h
349 To use these functions you now have to initialize the plugin by
350 calling MSG_storage_file_system_init() just after calling
351 MSG_init() or sg_storage_file_system_init() just after creating
355 - Remove the undocumented/untested tag <include>
358 - Remove viva specific tracing as the tool is no longer maintained
361 - #248: Kill of finished processes leads to segfault
362 - #240: xbt_cond_wait_timeout should gracefully return for C apps
363 - #239: please implement signal s4u::onDeadlock()
364 - #230: segfaults when exit() without run()
365 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
366 - #223: MSG_task_get_flops_amount() not working with parallel tasks
367 - #222: Actor::kill() doesn't really kill and segfaults
368 - #221: odd LMM warning when killing an actor
369 - #120: Memory leak when the processes are forcefully killed
371 ----------------------------------------------------------------------------
373 SimGrid (3.17) Released October 8 2017
375 The Drained Leaks release: (almost) no known leaks despite the tests.
377 Even very long simulations will preserve your memory: our testsuite
378 stresses SimGrid for over 45mn on fast machines for a coverage of
379 over 80%, and there is only one single known leak, of about 4kb.
383 - Comm.detach(): start and forget about asynchronous emission. A cleanup
384 handler may be given to free resources if the comm cannot be completed.
385 - this_actor::send(mailbox) is now mailbox->put()
386 - New: simgrid::s4u::Comm::wait_all()
387 - New: Host.execute() for remote executions.
390 - Unused option network/sender-gap is removed.
393 - Deprecate MSG_task_isend_with_matching(): This unused feature
394 really complicates our internals. Will be removed in v3.20.
397 - Improved context termination. It is now possible to free resources, even
398 when a process is forcibly killed.
399 - ContextBoost: add support for Boost versions above 1.61.
402 - Bring back run-time option --cfg=exception/cutpath to remove exception
404 - Removed unused functions:
405 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
406 xbt_str_rtrim(), xbt_str_trim().
407 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
410 - Removed header files obsolete since SimGrid 3.12:
411 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
413 - #3: SD_exit should be made optional
414 - #120: Memory leak when the processes are forcefully killed
415 - #159: Threading test regression in Actor refcounting
416 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
417 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
418 - #186: Actor::killAll() segfaults if some process is blocked on wait()
419 - #191: VM migration and pstate
420 - #192: Updating the energy consumptions of all hosts crashes with VMs
421 - #195: All actors have PID=0 in the logs
422 - #204: Sometimes segfault with thread contexts and mmap privatization
423 - #222: Actor::kill() doesn't really kill and segfaults
424 - #225: Actor::kill() doesn't really kill when victims are doing a join()
426 ----------------------------------------------------------------------------
428 SimGrid (3.16) Released June 22. 2017.
430 The Blooming Spring Release: developments are budding.
433 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
434 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
435 - Java JAR file should be finally fully working on Mac OSX too.
437 XML platforms: Switch to platform v4.1 format.
438 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
439 - <zone> can be used as a synonym for the now deprecated <as>
440 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
441 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
442 - <actor> can be used as a synonym for the now deprecated <process>
443 - state_file and avail_file periodicity is now easier to express
444 (check the documentation) the old behavior should still work.
445 * Storage: not backward compatible, but it seems that nobody ever used it
446 - Remove attribute 'content_type' of <storage_type>: was never used
447 - Make attribute 'model' of <storage_type> optional: for future usage
448 - Remove Bconnection model property: was never (in)validated. Replaced by
449 the maximum of the read and write bandwidth as a resource constraint.
452 - New and Backwards Compatibility break:
453 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
454 When one wants to get the list of tasks whose states have changed during a
455 simulation round, s/he has to allocate and free a dynar and use it as
456 argument to this function. The former SD_simulate (double how_long)
460 - Allow multicore VMs, with the correct sharing (unless you overcommit)
461 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
464 - New function to update the consumption of all hosts at once.
465 - Fix the model for multi-core hosts, linear in the amount of busy
466 cores with an abnormality for pIdle (see paper).
467 BUG: Still not satisfactory for ptask on mono-cores.
470 - The netzone are now available from the MSG API.
471 The old names still work, but are now deprecated.
474 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
475 - New option: smpi/keep-temps to not cleanup temp files
476 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
477 mallocs mode is used, allows to change the size of the fake file used
478 (default 1MB), to potentially limit the number of mappings for large runs.
479 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
480 - Fortran ifort and flang compilers support
481 - New RMA calls supported (experimental):
482 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
483 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
484 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
485 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
486 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
487 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
490 - New: this_actor::isMaestro() returns whether we are in kernel mode.
491 - Behind the scene, ActivityImpl are now automatically refcounted.
492 Removing this nasty bottleneck will greatly simplify our internals.
495 - Replay: New function xbt_replay_action_get():
496 Retrieve the function previously associated to an event type.
497 - DROPPED FUNCTION: xbt_str_varsubst()
498 - DROPPED MODULE: strbuff. We don't need it anymore.
499 - DROPPED MODULE: matrix. We don't need it anymore.
500 - DROPPED MODULE: lib. We don't need it anymore.
502 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
504 SimGrid (3.15) Released March 22. 2017
506 The Spring Release: continuous integration servers become green
507 We fixed even the transient bugs on all target architectures:
508 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
509 Windows, FreeBSD, NetBSD.
511 - Rename NetCards to NetPoints.
512 This was intended to help NS3 users, but that's not a netcard.
513 That's a point in the routing algorithm, let's avoid wrong simplifications.
516 - New: MSG_process_yield(). Stop and yield to other processes.
517 - New: MSG_process_daemon(). Daemon processes are automatically killed
518 when the last non-daemon process terminates
519 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
520 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
521 to make room for the upcoming network energy plugin.
522 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
525 - Ensure that an actor can kill itself with Process::exit()
526 - Kill the obscure NativeException. Nobody want to survive the issues
527 it denotes, so use JniException that is a RuntimeException (not to
528 be caught explicitly).
529 - Partial bug fix in initialization. SimGrid flags on command line were
530 consumed at C level but stayed in the original Java String[] args.
531 This could mess users' args[i] if SG flags were not put at the end of
533 The SimGrid flags are now removed from the Java arguments. However,
534 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
535 if args.length is greater than the number of YOUR OWN ARGUMENTS.
536 It might be if you have --log or --cfg flags in the command line.
537 - Fix numerous memleaks all around the place. In particular, around VMs.
541 - s4u::onPlatformCreated: right before the simulation starts
542 - s4u::onSimulationEnd: right after the main simulation loop
543 - s4u::onTimeAdvance: right after a clock change (time is discrete)
544 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
545 event from the availability_file changes the avail speed.
546 - Links are now usable from s4u
547 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
548 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
549 - New examples: The conversion of MSG examples to S4U has begun
550 - Actors: create, kill, migration, and suspend
551 - Applications: master-worker and token-ring
552 - Action replay: communications and storage
553 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
556 - Backwards Compatibility breaks
557 - The SD_route_get_list and SD_route_get_size functions have been removed.
558 They are replaced by the unique
559 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
560 The route from src to dst is built in the links dynar whose size can
561 be obtained with xbt_dynar_length.
562 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
563 sg_host_route_bandwidth and sg_host_route_latency.
564 Macros ensure the backwards compatibility, but you should fix your code
567 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
568 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
569 - Bump our claim of support from MPI 1.1 to MPI 2.2.
570 We don't support 100% of it, but it should be enough. Contact us if not.
571 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
572 naive and probably inaccurate)
573 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
574 mapped onto a unique small file using some system magic.
575 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
576 - Fortran: cleanups, fixes, support of user-added operations
577 - RMA: MPI_Accumulate are applied in correct order.
578 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
579 - Better support for MPI_IN_PLACE
580 - Support for MPI_Win attrs and keyvals.
581 - Support MPI_Comm_group_create, MPI_Type_size_x
582 - MPI ops should be properly applied to all allowed types
583 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
584 algorithm by a non failing one from mpich.
587 - Kill the fifo data container: we don't use it anymore.
590 - A <cluster> can now be created with different speed values to represent pstates.
591 see examples/platforms/energy_cluster.xml for an example.
593 SimGrid (3.14.159) Released December 28. 2016
595 The previous release was only a crude approximation of a Pi release;
596 this one should be a bit better.
598 - Revert a last minute change that broke on Mac OSX.
599 - Fix the changelog and NEWS files.
600 - Improve the documentation, in particular of the routing module.
601 - Integrate some patches flying around in Debian.
603 SimGrid (3.14) Released December 24. 2016
606 * The whole documentation was reworked and reorganized. There is
607 still much room for improvement here, but we're on it.
610 * We now need python3 for our testing infrastructure.
611 * Model checking needs libevent
614 * Now works on FreeBSD
617 * Privatization now works on FreeBSD
619 * Privatization is now activated by default in smpirun
620 You can turn it off with -no-privatize if something goes wrong.
622 * Call-location tracing for SMPI.
623 You can add the exact location (filename / linenumber) of an MPI call to
624 your trace files and slow down or speed up the simulation between two
625 consecutive calls by using an adjustment file (see the documentation).
627 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
628 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
629 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
630 values were ignored in some cases. The timings of these functions can now
631 be significantly different.
633 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
634 smpi/running-power is renamed to smpi/host-speed
636 * smpi/grow-injected-times option to enable or disable multiplication of the
637 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
638 make simulation less precise (but also much faster).
640 * smpirun script should be (much) faster for large deployments.
642 * SMPI tracing: fixed issue with poor matching of send/receives.
644 * Replay: Fix broken waitall
646 New functions and features
647 * MSG_parallel_task_execute_with_timeout, to timeout computations.
649 Dropped / renamed functions and features
650 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
651 - MSG_mailbox_is_empty() -> Mailbox::empty()
652 - MSG_mailbox_front() -> Mailbox::front()
653 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
654 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
655 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
656 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
657 * Don't pass the free_f parameter to property related functions:
658 - MSG_host_set_property_value()
659 - MSG_as_router_set_property_value()
660 - MSG_storage_set_property_value()
661 * VM properties. Since msg_vm_t are msg_host_t, just use
662 MSG_host_get_property_value() and friends
663 * VM I/O related things:
664 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
665 - Unimplemented save/restore methods
666 * MSG_as_router_get_property_value() was redundent with
667 MSG_environment_as_get_property_value().
668 - Removed MSG_as_router_*propert*() functions
669 - Added MSG_environment_as_set_property_value() for consistency
670 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
671 Well, they are still there for now, but deprecated with a warning.
672 Please switch to xbt_dict_new_homogeneous() before this is removed
674 * Task affinity. Its intended behavior (that was very badly tested
675 and probably not really working) was deceiving what most users
676 would have hoped here.
677 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
678 * Option network/coordinates is now useless and should be dropped.
681 * Added option storage/max_file_descriptors to allow more than 1024 files opened
684 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
685 * Added Dragonfly topology support
688 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
689 EiBps, EBps, Eibps, Ebps for bandwidth.
690 They may become useful to some lucky ones.
693 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
694 * ex/app_tokenring: new example, very similar to the MSG Token Ring
695 * ex/async_waitAll: new example, on asynchronous communications
698 * Memory usage should be decreased for simulations with a large number
699 of processes. This also helps for SMPI.
701 SimGrid (3.13) stable; urgency=low
703 The Half Release, a.k.a. the Zealous Easter Trim.
705 - We removed half of the lines, that were mostly experimental cruft.
706 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
707 (not counting blanks and comments -- according to openhub.net)
708 - The internals are now compiled in C++ (and will soon be clean C++)
709 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
710 59 klines of C, etc. We added only 29 klines of C++ in replacement.
712 * Backwards Compatibility breaks
713 - Removed Lua simulation bindings (switch to C or Java for that).
714 Lua can still be used to describe platforms
715 - Removed Java kernel plug-ins.
716 Will be reintroduced after the ongoing major internals reorg.
718 - the following functions were removed.
719 They were too specific and should be reimplemented in a generic
720 way, with filter function.
721 - MSG_task_listen_from_host
722 - MSG_mailbox_get_count_host_waiting_tasks
723 - MSG_mailbox_put_with_timeout was removed.
724 Please use MSG_task_send_with_timeout instead.
726 - the SD_application_reinit function was removed. It has been a noop for a while.
727 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
728 replaced by a more flexible mechanism.
729 - The following functions thus do not exist anymore
730 - SD_workstation_get_access_mode
731 - SD_workstation_set_access_mode
732 - SD_workstation_get_current_task
733 - Basic estimation functions have been removed but can easily be replaced
734 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
735 - SD_workstation_get_computation_time => amount / sg_host_speed()
737 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
738 Use VM.getSpeed()*load/100 for the legacy behavior.
740 - option enable_tracing was removed. It was not doing anything for a while.
741 - In the ModelChecker:
742 - the model-checker now ptraces the model-checked process which means
743 you cannot use a debugger on the latter anymore (we might make this
744 optional in the feature);
745 - removed soft-dirty page tracking;
746 - remove model-checked side snapshot management,
747 MC_snapshot() and MC_compare_snapshot();
748 - keep the MC_cut() function as a stub (it was not really working
749 in the previous release).
752 * All options are consistently in kebab-case. Old names are kept as alias.
755 * Switch to platform v4 format.
756 - Rename from 'power' to 'speed' the attributes describing the amount of
757 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
758 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
759 - In <host> and <link>, attributes availability and state are gone.
760 It was redundent with state and availability traces, and with peak values.
761 - In <cluster>, attributes availability_file and state_file are gone.
762 It was too complex and unused.
763 - Kill <gpu>. Was not doing anything.
764 - The DOCTYPE points to the right URL:
765 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
766 (the file at this address now documents the changelog since its v1)
767 - A warning is emitted for unit-less values (they are still accepted).
768 - speed. Default: 'f' or 'flops'. Also defined:
769 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
770 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
771 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
772 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
773 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
774 - latency. Default: 's' second. Also defined:
775 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
777 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
778 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
781 * s4u::Host is now the preferred public interface to the Host features.
782 sg_host_* functions are C bindings to the exact same behavior
783 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
786 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
787 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
788 They were deprecated since a few years, and probably did not even compile anymore.
791 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
792 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
793 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
794 encouraged to update their codes. A list of the modified functions can be found at the end of
795 include/simgrid/simdag.h
798 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
799 the lambda or closure passed as a parameter will run in kernel mode.
800 All the callback functions should be rewritten to that interface at some point.
803 * Reorganizing and cleaning the internals all around the place.
806 * Remove old default barrier/bcast buggy algorithms (see #18407)
807 * Various bug fixes to handle more codes
808 * Remove the need for the --foreground option of smpirun (it is still
809 accepted for backward compatibility).
812 * Kill the setset data container: MC don't use it anymore.
813 * Kill the queue data container: it made more sense with GRAS.
814 * Kill the xbt_peer_t data type: it's useless without GRAS.
815 * Kill rm_cb feature of config sets: it was never useful.
816 * Kill graphxml parsing feature. It was not used.
817 * Kill the deprecated code protected by XBT_USE_DEPRECATED
819 - xbt_dynar_sort_strings(), when the content is char*
820 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
821 They throw exceptions on invalid input;
822 * C++ support for declaring CLI flags (simgrid::config::Flag);
823 * class for abstracting different signal backends (simgrid::xbt::signal).
824 with no external dependencies (we need very simple signals).
827 * refactoring and cleanup of the code;
828 * ongoing process to cleanly separate the model-checking algorithms
829 from the code model-checking support.
831 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
833 SimGrid (3.12) stable; urgency=low
835 The Facelift Release.
838 * Require g++ v4.7 at least to not speak prehistorical C++.
839 * Require Boost 1.48 (for signal2 component).
840 * Java must be version 7 at least when activated.
841 * Builds on Windows again (including Java bindings).
842 * Tracing is now always enabled (no way to turn it out)
843 * Remove GTNetS. It was not working anyway.
844 * Various cleanups in the cmake scripts.
845 * Move headers around to sort them out on installed systems:
846 - instr/instr.h -> simgrid/instr.h
847 - instr/jedule/* -> simgrid/jedule
848 - simdag/datatypes.h was removed
849 - simdag/simdag.h -> simgrid/simdag.h
850 - msg/datatypes.h was removed
851 - msg/msg.h -> simgrid/msg.h
854 * Interface improvement:
855 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
856 - Sanitize the interface in MSG_task_ module:
857 - Merge two functions that were close enough but misleading:
858 set_compute_duration(t) -> set_flops_amount(t)
859 get_remaining_computation(t) -> get_flops_amount(t)
860 - set_data_size(t) -> set_bytes_amount(t)
861 get_data_size(t) -> get_bytes_amount(t)
862 - Massive cleanups in the functions related to the energy
863 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
864 - New: MSG_host_get_pstate()
866 - msg/energy/onoff: switching hosts on and off
868 * Interface improvement:
869 - Rename Host.isAvail() to Host.isOn()
870 - Rename Process.currentProcess() to Process.getCurrentProcess()
871 - Rename Task.setDataSize() to Task.setBytesAmount()
872 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
874 - #18874: Actually allows the GC to reclaim tasks
878 - SIMIX_process_throw: raises an exception in a remote process
879 * Refactoring: Separate sub-modules
880 - libsmx: the public interface, as libc in a real system
881 - popping: the strange dance that converts a user request into a kernel handling
882 - smx_context_*: the virtualization mechanisms that embed the user code
883 - smx_*: the handling of each simcalls
885 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
886 * Rename smx_action_t into smx_synchro_t, making explicit that these
887 things are used to synchronize processes with their environment.
888 For example, a communication is a sort of synchronization involving
889 the communicating processes (that may block until the exchange) and
890 the platform. The same can be said from computations, etc.
892 - #18888: segfault when a process ends before its kill_time
896 - 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.
897 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
898 - MPI_Type_set_name, MPI_Type_get_name
899 - MPI_*_c2f and MPI_*_f2c functions
900 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
901 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
902 - Activate a lot of new tests from the mpich 3 testsuite
904 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
905 - InfiniBand network model added: Based on the works of Jerome Vienne
906 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
907 - When smpi/display_timing is set, also display global simulation time and application times
908 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
909 * Collective communications
910 - 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
911 - 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/).
912 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
914 - Replay now uses algorithms from wanted collective selector
915 - Replay can be used with SMP-aware algorithms
916 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
917 - 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.
919 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
920 - Correctly use loopback on fat-tree clusters
921 - Asynchronous small messages shouldn't trigger deadlocks anymore
922 * Energy/DVFS cleanup and improvement
923 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
924 - new: smpi_get_host_pstate()
927 - "Full" network optimization flag was broken since Surf++
928 - Better handling of precision flags in maxmin
929 - Fix bug causing sometimes "Impossible" errors
930 - Properly pass cluster properties to included hosts
931 * Improvement of the Energy plugin.
932 - Always update the consumption before returning that value
933 - New property: watt_off to denote the dissipation when the host is off
934 - New functions getWattMinAt and getWattMaxAt to retrieve the
935 dissipation of pstates that we are not currently at.
936 * Java: class NetworkLink renamed to Link
937 * New function: simcall_process_get_kill_time()
938 * Massive rename s/workstation/host/
939 - That's intrusive, but that's good for the project consistency. Sorry.
940 - Change config option "workstation/model" into "host/model"
943 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
944 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
945 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
946 - xbt_dirname and xbt_basename for non-POSIX systems
948 * The model checker now runs as a separate process.
949 * The model checker runs must now be launched with the new simgrid-mc program.
950 * Record/Replay: the MC can display a textual representation of a path in the
951 execution graph. It can then be replayed outside of the model checker.
953 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
955 SimGrid (3.11) stable; urgency=low
960 * Normalizing pointers addresses tool for better diff between logs
963 * Add cloud examples using new VMs
964 - examples/msg/cloud/two_tasks_vm.tesh
965 - examples/msg/cloud/simple_vm.tesh
966 - examples/java/cloud/cloud.tesh
967 - examples/java/cloud/migration/migration.tesh
968 * Add java surf examples:
969 - examples/java/surfPlugin/surf_plugin.tesh
970 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
971 - examples/java/surfCpuModel/surf_cpu_model.tesh
972 * Add SMPI+MSG example:
973 - examples/smpi/smpi_msg_masterslave/
978 - msg task destroy cancel test
979 - msg_host on/off test
980 * Move all tests in testsuite to teshsuite (adding tesh files)
981 * Restructure teshsuites
982 - one folder for each kind of test
983 * Restructure AddTests.cmake
985 - structure the order of tests (with sections)
988 * Add virtual machine
989 - creation of a VM on a PM
990 - migration of a VM from a PM to another PM
992 - MSG_process_join(msg_process_t process, double timeout)
993 - msg_bar_t MSG_barrier_init(unsigned int count)
994 - int MSG_barrier_wait(msg_bar_t barrier)
995 - void MSG_barrier_destroy(msg_bar_t barrier)
996 - msg_as_t MSG_environment_as_get_by_name(const char * name)
997 * New option "msg/debug_multiple_use" to help debugging when a task is used
1000 - Improvements and finalization of MSG_storage, MSG_file APIs and their
1002 - Increase code coverage in test suites
1006 * Protect context stacks against stack overflow. The number of protected memory
1007 pages allocated on the top of each stack (1 by default) can be configured
1008 with the new command line option --cfg=contexts/guard_size:<value>.
1009 * Simcalls are now generated by a python script that
1010 - generates files included by SimGrid
1011 - checks that all the functions exist, and proposes prototypes
1013 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
1014 host_set_data, host_get_data
1016 - simcall_process_join(smx_process_t process, double timeout)
1017 * Fix bug where sleeping processing could not be suspended.
1020 * Translate surf models from C to C++
1021 - Generic classes for all models: Model, Resource, Action
1022 - A generic interface for each kind of model (CPU, Network, Storage
1023 Workstation, WorkstationVM)
1025 * Translate surf routings from C to C++
1026 * Add callbacks using sigc++ or boost::signals2
1027 - Add callback functions for resource creation/destruction
1028 - Add callback functions for action state change
1029 - Handle Energy as a plugin
1030 * Replace swag by boost::intrusive
1031 * Add new routing models for clusters. For documentation, see
1032 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
1033 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
1034 parameters for cluster tag
1035 - Fat trees, with topology="FAT_TREE" and
1036 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
1037 - see examples/platforms/torus_cluster.xml and
1038 examples/platforms/fat_tree_cluster.xml
1039 * More documentation
1042 * Hostfiles support host:nb_processes construct to deploy several processes on
1044 * Collective communication algorithms should not crash if used with
1045 improper number of nodes and report the error.
1046 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1047 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1048 MPI_Dims_create, MPI_Cart_sub are supported.
1049 * New interface to use SMPI programmatically (still depends on MSG for
1050 some parts, see examples/smpi/smpi_msg_masterslave):
1051 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1055 * Global variables privatization in MPI executables is now performed at runtime
1056 with the option smpi/privatize_global_variables (default:no).
1057 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1058 dynamic libraries loaded by the application are not privatized (static
1059 linking with these libraries is advised in this case)
1062 * Options defined in XML work correctly now.
1065 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1066 copied into simgrid.jar or not (ON by default). Use this option if you want
1067 to reduce the size of the installed simgrid.jar, *and* the native libraries
1068 are kept installed elsewhere.
1069 * Surf binding with SWIG (code generated in maintainer mode only):
1070 - plugin to handle callbacks
1071 - CPU model only for the moment
1074 * Supernovae build mode is definitively removed. It was used to improve
1075 inlining and inter-module optimizations. It is nowadays superseded by
1076 link-time optimizations commonly available in compilers.
1077 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1078 * Add boost dependency for surf++
1079 * Add new macro for tests
1080 - ADD_TESH(name <tesh_args>)
1081 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1085 - xbt_bar_t XBT_barrier_init(unsigned int count)
1086 - int XBT_barrier_wait(xbt_bar_t barrier)
1087 - void XBT_barrier_destroy(xbt_bar_t barrier)
1088 * Make the xbt_os_time module public
1090 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1092 SimGrid (3.10) stable; urgency=low
1094 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1097 * Reintegrate Java to the main archive as desynchronizing these
1098 package is not acceptable anymore (Java is now considered stable)
1099 * Add explicit synchronization facilities through semaphores
1100 * Bug fix: Task.setDataSize() only changed the C world, not the value
1101 cached in the Java world
1104 * Dramatically change the way files are handled. API and internals changed, but
1105 this part of MSG was not considered as production grade either.
1106 * Add explicit synchronization facilities through semaphores
1107 * Add a new function MSG_host_get_process_list()
1108 * Preliminary DVFS support (see examples/msg/energy/ for details)
1111 * SMPI is now included directly in the libsimgrid as the windows
1112 linker doesn't force us on splitting it anymore.
1113 * Improvements of the SMPI replay tool:
1114 - Most of the collective communications are now rooted in the same process as
1115 in the original application.
1116 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1117 used until now). Multiple data types can now be used in a trace.
1118 - The replay tool now supports traces produce either by TAU or a modified
1120 - Bug Fix: the compute part of the reduce action is now taken into account.
1121 - Gatherv collective is now supported
1122 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1123 of smpirun outputs time independent traces for the current run. One file
1124 is created per process. If too many processes are simulated, this behavior
1125 can be changed to one file for all processes by using the
1126 tracing/smpi/format/ti_one_file flag
1127 * smpirun generates the host file if needed (with given host count and platform)
1128 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1129 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1130 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1131 --cfg=smpi/coll_selector:(mpich/ompi)
1132 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1134 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1135 happening inside a collective SMPI call.
1136 * Fix the behavior of complex data types handling.
1137 * Make MPI_Wtime another synchronization point to take computations into
1139 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1140 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1141 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1142 (no privatization of global variables yet)
1143 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1146 * Verification of liveness properties is now available for SMPI applications
1147 (in addition to MSG applications)
1148 * Bugged examples using SMPI in examples/smpi/mc/
1149 * Add --cfg=model-check/visited option. Allows the verification of infinite
1150 programs. Detection of loops in the execution thanks to the system state
1151 comparison and reduction of the state space to explore. Can be combined with
1152 DPOR for safety properties.
1155 * Allow to change SimGrid configuration (see --help) within the code
1156 thanks to SD_config() as it can be done in MSG.
1157 * Add a new function SD_task_set_amount() upon user request.
1160 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1161 * Remove rule based routing (no more PCRE dependency)
1162 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1163 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1164 * Add a loopback_bw and loopback_lat options to cluster tag.
1165 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1166 routing. Now peers are encapsulated in an AS and have their own private
1167 router but this is transparent.
1170 * Our own implementation of getline is renamed xbt_getline, and gets
1171 used even if the OS provide a getline(). This should reduce the
1172 configuration complexity by using the same code on all platforms.
1173 * New type: xbt_cfg_elm_boolean.
1174 * Allow to use yes/no for boolean configuration options in the command line.
1175 * Allow to disable SimGrid cleanups at exit from command line option.
1176 There are situations where one may want a simulation to end with an exit.
1177 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1178 annoying when scripting around the simulator. Adding a
1179 --cfg=clean_atexit:no allows to circumvent this issue.
1182 * Lots of memory leaks were corrected in this release.
1183 There is no known memory leaks anymore, in all of our 600+ tests.
1184 * New command line option --version, to get SimGrid version information.
1185 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1187 * Supernovae builds are deprecated, and expected to be removed in the next
1190 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1192 SimGrid (3.9) stable; urgency=low
1194 The Grasgory release: GRAS is really dead now.
1196 * Complete overhaul of the internal host structures scheme.
1199 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1200 was considered as experimental and badly maintained since then.
1201 * Keeping it was thus a trap to our potential users, that could take
1202 it instead of MSG or SMPI by mistake despite is pity state.
1203 * GRAS seems to have very few users (if any), and no one volunteered
1204 to maintain it further. It also induces a lot of XBT code (for
1205 portability sake), that must be maintained too.
1206 * For all these reasons, we killed GRAS. If someone wants to revive it
1207 in the future, don't cry, our git history still remembers of GRAS.
1210 * Major overhaul. Merge our documentation again as time proved that
1211 splitting it was really not helping our users.
1212 * Further improve the developer documentation to help newcomers
1213 hacking on SimGrid itself. The user documentation (and in
1214 particular, the beginner documentation) is still in a sorry state.
1217 * Now works on Windows too!
1218 * Much more extensive test suite, from MPICH
1221 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1222 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1223 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1224 * Bug fix: let task be scheduled when the last dependency to be solved is
1225 a control dependency.
1226 * Remove SD_load_environment_script function.
1227 Use the C sg_platf function if you want to declare a platform
1231 * New function: MSG_process_get_number()
1232 * Old function documented: MSG_config()
1233 * Remove MSG_load_platform_script function
1234 Use the C sg_platf function if you want to declare a platform
1238 * Change the default value of the TCP_gamma constant (maximal size of TCP
1239 congestion window) to a more realistic 4MiB value. If you notice changes in
1240 your simulation results, you can fall back to the previous 20k tiny window
1241 by adding --cfg=network/TCP_gamma:20000 on command line.
1242 * (Hopefully) fix a bug wrt periodic availability and state traces
1243 * Bug fix: use default values at start when first event in availability/state
1244 trace is not at time 0.
1247 * remove the "new_" part of function name sg_platf_new_trace_connect
1248 (resulting in sg_platf_trace_connect), since it does not create
1252 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1253 use case where it's really mandatory, and maintaining it was a pain in
1255 * New: xbt_fifo_search(), search an item with a user-provided
1256 comparison function instead of dumb pointer comparison.
1259 * Fix the lua deployment:
1260 Use `simgrid.init_application()` before deployment instead of
1261 `simgrid.msg_register_application()` after.
1264 * Transfer the tracing files into the corresponding modules.
1266 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1268 SimGrid (3.8.1) stable; urgency=low
1270 The "we are told that some people want to also *install* the simgrid
1273 * Add missing file "tesh.1" to the archive.
1275 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1277 SimGrid (3.8) stable; urgency=low
1279 The Psssshiiiit release: SimGrid jumps into the Cloud.
1282 * Add an experimental interface to manipulate VMs. They are mainly
1283 process groups with very few intrinsic semantic, but they should
1284 allow you to build the semantic you want easily.
1285 * New function: MSG_host_set_property_value()
1286 * New function: MSG_process_on_exit(). To clean memory in all cases.
1287 * Bug fixes that made the host (and link) failures unusable.
1288 * Add a way to auto-restart process when the host in which they are
1289 executing comes back (ON_FAILURE="RESTART" on deployment file,
1290 MSG_process_auto_restart_set).
1291 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1292 please stop using the old ones, they are DEPRECATED.
1294 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1295 Please use MSG_init() instead. (reducing the amount of entry
1296 points in the library helps us).
1297 * Make it impossible to link against the wrong version of the lib
1298 * Deprecate MSG_clean(). No need to call it anymore.
1299 * Function MSG_get_host_number() is not deprecated anymore.
1302 * Split the doc into a user guide and a reference guide.
1303 * Start a developper guide to help people hacking on SimGrid.
1306 * Enable tracing by default. This modules rocks you should use it.
1307 * Remove option custom_flags. Now use environment variables CFLAGS
1309 * Use default cmake things to detect lua instead of home grown ones.
1310 * New option "enable_mallocators" to disable mallocators, for debugging
1311 purpose ("on" by default).
1314 * Bug fixes around the resource failures: don't let the processes
1315 survive the host they are running onto.
1316 * Add an interface to auto-restart processes when the host in which they are
1317 executing comes back.
1318 * Ensures that SIMIX_clean is called automatically. It's not part of
1319 the public interface anymore (bindings should be updated).
1322 * Bug fix for when SD_Simulate is called with a positive value: be careful
1323 when comparing doubles. Sometimes they are different for non significant
1325 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1326 parallel task whose initial work is distributed among host according
1327 to the Amdahl's law. Such tasks are created with a parameter alpha
1328 that corresponds to the non-parallelizable part of the computation.
1329 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1330 between two sets of workstations assuming a 1D block distribution (each
1331 workstation owns a similar share of data) on both sides.
1333 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1334 Data redistribution will be automatically scheduled once parent and child
1335 are both scheduled. The filling of computation_amount and
1336 communication_amount structures is now done seamlessly thanks to the chosen
1338 * New function SD_workstation_dump to display various information
1339 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1340 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1341 bandwidth on the route onto which the task is scheduled and the amount of
1343 To divide the nominal bandwidth by 2, the rate then has to be:
1344 rate = bandwidth/(2*amount)
1345 * Compute tasks that have failed can now be rescheduled and executed again
1346 (from their beginning)
1347 * Increasing source code coverage (src/simdag is now covered at 95.8%
1351 * Re-implement time-independent trace replay using SMPI (at the
1352 smpi_smp_* level) instead of MSG. This should replace
1353 examples/msg/actions/actions.c
1354 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1355 hindexed and structs)
1356 * Implement the exchange of non-contiguous data.
1357 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1358 * Correct behavior of smpi/sender_gap and set its default value to 0
1359 * Add option to asynchronously send small messages to allow better
1360 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1361 specifies the size in bytes under which messages will be asynchronously sent.
1362 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1363 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1366 * New C interface to define a platform: XML is now optional.
1367 For more info, please check include/simgrid/platf.h.
1368 * New interface to define random platforms from the C:
1369 For more info, please check include/simgrid/platf_generator.h and
1370 examples/msg/masterslave/masterslave_platfgen.c
1371 * Export a sg_cmdline dynar containing all the arguments we got from
1375 * Two new tracing options for adding comments to trace file so you
1376 can track your experiments (see --help-tracing for details).
1377 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1378 * Adding the SimGrid version that generated the trace file as a comment.
1379 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1380 * Fix to avoid key clashes on Paje links
1381 * Other minor fixes related to the Paje specification
1384 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1385 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1386 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1387 * Greatly improve the robustness of mmalloc to user errors (such as
1388 using an area after freeing it, or freeing it twice)
1390 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1392 SimGrid-java (3.8.1) stable; urgency=low
1394 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1395 Warning: all methods are visible, but only some of them are
1396 implemented so far. Check the source (src/jtrace.c)
1397 for further information.
1398 * New module: org.simgrid.msg.File (SimGrid File management functions)
1399 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1400 * Change the meaning of Process.restart: now restart the process from
1401 the begining, like MSG_process_restart in C.
1402 * Add Process.setAutoRestart: handling of process restart when failed
1404 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1405 you to retrieve the properties of the processes/hosts
1406 * Deprecate Msg.clean(): you can just forget about it now.
1407 * New function Process.getCount(), that only works when compiling
1408 with the not yet released version 3.9 of the C library.
1410 * New context factory based on Coroutines. It mandates a modified JVM
1411 but then, the simulations run about five times faster, and there is
1412 no limit to the amount of processes (beside of the available memory).
1414 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1416 SimGrid (3.7.1) stable; urgency=low
1419 * Restore the prototype of MSG_process_create_with_environment() to
1420 the pre-3.7 situation by removing the kill_time argument.
1421 * Add a MSG_process_set_kill_time() function instead.
1424 * Fix weird behaviors when dealing with parallel tasks.
1427 * Simgrid is now built as a dll.
1428 * Simgrid-java now works on Windows.
1429 * Simgrid-Java is now included into Windows package.
1432 * First pre-build package for MacOSX.
1435 * Fix compilation when using MSG_USE_DEPRECATED.
1436 * Fix some compilation issues on Macs and Windows.
1437 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1438 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1440 We discovered that the Lua console is broken, but we are missing the
1441 manpower to fix it right now. The problem existed in 3.7 too, so we
1442 are not blocking the release for that. Sorry if you depended on this
1443 feature, any help would be really welcome.
1445 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1448 SimGrid-java (3.7.1) stable; urgency=low
1450 The "Java aint got to be bloated and slow" release
1453 * Various internal cleanups and performance improvement
1454 Simulations are expected to run up to twice faster or so
1455 * Make Process.kill(process) an instance method, not a static one
1456 * User processes are not java.lang.Thread subclasses.
1457 This breaks the compatibility (sorry), but previous API was
1458 brain-dead, making it impossible to have non-trivial
1459 initializations in the process constructor.
1460 * Require a full constructor per Process sub-class.
1461 Kinda breaks the compatibility (sorry), but this allows a much more
1462 efficient way to launch the processes at simulation startup.
1463 * Do not embeed our version of semaphores, java 1.5 can be considered
1464 as sufficiently prevalent for us to not dupplicate its features.
1468 * Add examples for almost every part of the API
1469 We spotted and fixed a lot of bugs in the process
1471 * New module: asynchronous communication API
1472 * New function: Process.sleep()
1473 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1474 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1475 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1477 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1479 SimGrid (3.7) stable; urgency=low
1481 The "spring cleanups (before the next Big Project kicks in)" release.
1484 * We can specify the SMPI latency/bandwidth factor with command line
1485 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1486 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1487 You can also use the "config tag" from platform file by adding this line
1488 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1489 (see "example/platforms/tag_config.xml" to use "config tag").
1490 Note that the command line supersedes the platform file configuration.
1491 * Change the correction factors used in LMM model, according to
1492 the latest experiments described in INRIA RR-7821.
1493 Accuracy should be improved this way.
1494 * Use the partial invalidation optimization by default for the
1495 network too. Should produce the exact same results, only faster.
1496 * Major cleanup in surf to merge models and split some optimization
1497 mechanisms from the core of the models. As a result you can now
1498 specify which model to use (e.g., --cfg=network/model:LV08
1499 --cfg=cpu/model:Cas01) and which optimization mode to use
1500 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1501 Incompatible combinations should err at initialization. See
1502 --help-models for the list of all models and optimization modes.
1503 * The CLM03 workstation model was dropped for simplicity because it
1504 used the deprecated CM02 network model. Use default instead.
1505 * Rename the TCP_gamma configuration option to network/TCP_gamma
1506 * Rename the coordinates configuration option to
1507 network/coordinates, and document it
1508 * Use now crosstraffic keyword instead of the terribly misleading
1509 fullduplex keyword. It is activated by default now in the current
1510 default model, use --cfg=network/crosstraffic:0 to turn it off.
1511 * Ongoing refactoring the model parsing to make XML files optional
1512 See include/simgrid/platf.h for details (still to be completed)
1515 * Major overhaul of the documentation. Almost instructive now :/
1516 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1517 functions and friends. This interface was considered as
1518 deprecated since over 2 years, it's time to inform our users that it is.
1519 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1520 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1521 use these (crappy) functions in your code.
1522 These functions will be removed soon. Stop using them now.
1523 * Deprecate MSG_get_host_{table,number}
1524 Implement MSG_hosts_as_dynar() instead.
1525 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1526 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1529 * Stabilize the parallel execution mode of user contexts
1530 * Introduce configuration variables to control parallel execution:
1531 - contexts/synchro: Synchronization mode to use when running
1532 contexts in parallel (either futex, posix or busy_wait)
1533 - contexts/parallel_threshold: Minimal number of user contexts
1534 that must be part of a scheduling round to switch to parallel
1535 execution mode (raw contexts only)
1536 * Fix bugs that prevented to use suspend/resume along with
1537 synchronization structures.
1538 * Fix bugs in process termination that lead to invalid memory access
1539 in very specific conditions.
1542 * Introduce a parallel mode for the models (controlled by surf/nthreads
1543 configuration item). In our tests, running the models in parallel
1544 never lead to any speedups because they are so fast that the gain
1545 of computing each model in parallel does not amortizes the
1546 synchronization costs, even when ultra fast futexes are used.
1547 This is released anyway because YMMV.
1550 * Performance boost by using a swag internally to compute the set of
1551 tasks that are finished and should constitute the return value of
1555 * Enable it by default now that it is considered rather stable.
1558 * Documentation of the tracing functions.
1559 * Performance gains when tracing categorized/uncategorized resource
1560 utilization by avoiding calls to get route when updating resource
1561 variables. LMM constraints are being used instead.
1562 * API changed to set task categories. Use MSG_task_set_category instead
1563 of TRACE_msg_set_task_category, and SD_task_set_category instead
1564 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1565 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1566 * Examples for link user variables added (see at examples/msg/tracing/)
1567 * Deprecated function TRACE_msg_set_process_category completely removed
1568 * Trace header updated according to the latest Paje file format
1569 * Tracing network lazy updates, no longer obligate users to use full updates
1570 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1571 * Experimental: let user code declare/set/push/pop application states for hosts
1572 * API changed to allow the manual creation of graph configuration files
1573 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1576 * Improve the API of Lua MSG bindings, using the Lua spirit.
1577 * Each simulated process now lives in its own Lua world (globals are
1578 automatically duplicated). It helps writing simulators. Will allow
1579 to run Splay programs within SimGrid in the future.
1580 * Add a Chord example in Lua, equivalent to the MSG one.
1583 * Start the implementation of a solution to express temporal
1584 properties, not only local assertions. This is still an
1585 experimental work in progress, stay clear from it to be safe.
1589 - Add new runtime parameters --help-logs and --help-log-categories
1590 to display information about supported logging parameters and
1592 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1593 don't exists anymore.
1594 * Mallocators: allow value NULL for the reset function.
1596 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1597 create homogeneous dictionaries, where all the elements share the
1598 same free function. Non homogeneous dictionaries will be
1599 deprecated in the next release.
1600 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1601 - Multi-level dictionaries are deprecated.
1603 - new function xbt_dynar_search_or_negative() that is useful when
1604 you have less than 2 million elements in your dynar and don't
1605 want of the extra complexity of catching exceptions when the
1606 element is not found.
1608 - Make xbt_os_thread module (for thread portability) public.
1609 Documentation is still to come, sorry.
1611 - Cleanups and simplifications to make it maintainable again.
1612 - Exotic features (such as memalign and valloc) were removed.
1613 - The metadata were extended and improved so that the
1614 model-checker becomes able to explore and inspect the heaps.
1615 - This may induce a performance drop when enable_model-checking is
1616 ON in cmake (even if it's not used in the simulation), but it is
1617 necessary at this point to get MC working.
1619 Turn model-checking OFF if simulation performance matters to you.
1620 Not enabling it at runtime is not enough, disable it in cmake.
1622 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1625 SimGrid (3.6.2) stable; urgency=low
1627 The "Not coding new stuff allows to polish old things" release.
1630 * New bindings to the NS3 packet level simulator (experimental)
1631 * Use the raw (efficient) execution contexts instead of the sysv
1632 (portable) ones when possible.
1633 * libpcre is now mandatory in any cases since not using it led to
1634 severe performance loss and possibly other issues
1635 * Update the XML platforms:
1636 - G5K: include the latest machine in Nancy
1637 - GridPP and LCG: new platforms
1638 * Documentation was partially updated, at least (more to come)
1640 Bug fixes, cosmetics and small improvements
1641 * Free terminated processes before the end of the simulation to avoid
1642 exhausting the memory of users having very dynamic amount of
1644 * Bug fix and cosmetics about canceling non-running tasks
1645 * Bug fix about the dot loader's issues when using libcgraph
1648 * Create an installer for windows with nsis (amd64 and win32)
1649 - Add an hello world project to illustrate simgrid project creation.
1650 - Embed libpcre into the Simgrid installer to avoid
1651 its compilation burden
1652 * The raw execution contexts should work on Apple now
1653 * Port to Windows 64 bits
1654 - Sysv contexts now have an implementation for this arch
1655 - GRAS communication features now support this arch
1656 * Drop support for borland compiler on windows
1657 - this code was not maintained, and we kinda depend on gcc nowadays
1658 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1659 * Fix portability issue on unstable ubuntu: linker became picky on
1662 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1664 SimGrid (3.6.1) stable; urgency=low
1666 The "Oops, we broke Macs too" release
1669 * Fixed contexts detection so that raw ones are used when possible
1670 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1671 leads to a strange error, with user code segfaulting sometimes when
1672 the generated code is not perfectly aligned (which is not
1673 controllable from the user side, depends on the amount of code)
1676 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1678 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1680 SimGrid-java (3.6) unstable; urgency=low
1683 * Split of every thing from simgrid v3.5 into a separate package.
1685 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1687 SimGrid (3.6) stable; urgency=medium
1689 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1692 * Bindings now constitute their own package, separated from the main one.
1693 Rationale: reduce our maintenance nightmare by reducing the module coupling
1694 They will soon be released on their own on gforge.
1696 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1697 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1699 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1700 * It's not quite deprecated for now because we have no replacement,
1701 but it may soon become the case.
1704 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1705 * Fortran: New user-level cache variable to store the rank of the running
1706 process. This improves performance by an order of magnitude.
1707 * C: New coccinelle script to automatically locate and modify global and
1708 local static variables.
1709 * Improved SMPI network model with a sender-side gap to account for multiple
1713 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1714 only say if a communication is finished, no matter whether it succeeded or
1715 failed. You can call MSG_comm_get_status() to know the status of a finished
1717 * New function MSG_task_dsend() to send a task and detach it. When a
1718 communication is detached, you are never notified of its success or failure
1719 and the memory is released automatically once it is finished. This function
1720 is useful when you don't care about the end nor the success of a
1722 * Change the prototypes of action replay. Sorry for inconvenience,
1723 but this is really more efficient this way (and to adapt your code,
1724 you just have to fix the initialization, that shouldn't be too long)
1725 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1726 ever managed to do anything useful with it.
1727 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1728 takes a dynar of communications. It returns immediately and gives the
1729 index of a finished communication (if any).
1730 * New example: a basic implementation of the Chord P2P algorithm.
1733 * New model for multi-core CPUs. You can now use the core attribute to
1734 precise the number of cores of a host. This is a basic model. Every
1735 process running on the host receives at most the power provided in
1736 the DTD (throughput<=power). Total throughput of process cannot exceed
1738 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1739 router linked by an up-link and a down-link (possibly asymmetrical).
1740 This kind of pattern allows to easily build last-mile model style platforms.
1741 Aggregating such patterns in a rule-based AS is thus the technique of
1742 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1743 * New model for Vivaldi routing. We transformed the Vivaldi network model
1744 into a Vivaldi routing model (based on the rule-based model). This allows to
1745 combine Vivaldi based latencies with last-mile platforms.
1748 * Added a check for NaN of IEEE754 infinite in the double entries of
1750 * Introduce a new context factory "raw", highly inspirated from the
1751 ucontext factory, but using manually crafted functions in assembly to
1752 do the work in an efficient manner.
1753 * Allow to change the used context factory at run time, not only at
1754 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1755 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1756 contexts at runtime (only with raw contexts or ucontexts).
1757 * Completely rewrote this module to allow parallel execution of user
1758 processes. Use --cfg=contexts/nthreads:N to execute user processes
1759 with N parallel threads (the default is 1, meaning no parallelism).
1760 * Allow to decide dynamically between sequential and parallel modes.
1761 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1762 processes in parallel only when their number is greater than or equal to P
1764 * Added a check for NaN of IEEE754 infinite in the double entries of
1768 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1769 won't output the state of processes when interrupted with Ctrl-C
1770 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1771 NULL-terminated array. This may solve backward compatibility issues
1772 due to the change to return type of SD_simulate. See also:
1773 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1774 * Add new macros with variable number of arguments.
1775 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1776 - in xbt/asserts.h: xbt_assert
1777 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1778 - in xbt/ex.h: THROWF and RETHROWF.
1779 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1781 * Change xbt_die() to accept a format string with arguments, just like printf.
1782 * New data structure: xbt_lib_t, like a dict but more general and with better
1786 * New configuration options
1787 Options triva/categorized and triva/uncategorized can be used to generate
1788 graph configuration files for Triva visualization tool.
1789 * Configuration option tracing/platform is renamed to tracing/categorized
1790 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1791 * New TRACE_user_link_variable interface:
1792 User provides the name of the link and the tracing variable to attach to it
1793 * the declaration of tracing categories must be done after the environment
1795 * simpler tracing interface, just one way to declare categories
1796 TRACE_category or TRACE_category_with_color, it is up to you
1797 * links in the trace file are again identified by their names
1798 * trace contains the full platform hierarchy exactly as declared using the ASes
1799 * Options tracing/msg/[task|process]:1 groups the process by hosts
1800 for both cases, tasks and processes must have names that are unique during the simulation
1801 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1802 * The experimental option tracing/msg/volume is deprecated
1803 its functionality may be reincorporated if needed
1805 The tracing generates a trace file with unordered timestamped events,
1806 because of the way the core simulator (surf) works. A script available
1807 at the tools directory (fix-paje-trace.sh) can be used to put the events
1808 in order. We have changed the tracing so it can generate ordered timestamped
1809 events in the final trace, but depending on the simulator (and how much time
1810 is simulated) that can lead to a huge memory utilization. It is deactivated
1811 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1813 Build Infrastructure
1814 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1815 - We are trying hard to keep the API stable, but it may happen that
1816 some things change (we're a research project after all, not a
1817 nuclear plant operating system). If such things should happen, you
1818 could rely on that macro to adapt.
1819 - current value: 30600 for 3.06.00, aka 3.6
1820 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1821 a number that can be compared with SIMGRID_VERSION.
1822 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1823 assertions and verbose/debug logging events are disabled at compile time.
1825 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1827 SimGrid (3.5) stable; urgency=medium
1830 * New feature to any SimGrid-based simulator: Model-Checking
1831 Check SIN#1 for more details.
1834 * New Model SMPI (three-interval linear regression for correction factors)
1835 See RR-7426, available at http://hal.inria.fr/inria-00527150
1836 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1837 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1838 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1839 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1840 * See also src/smpi/README
1845 - Tracing API changes: TRACE_start and TRACE_end should not be called
1846 by user-code. They are automatically called by simulators created
1847 with SimDAG, MSG and SMPI if the toolkit is compiled with
1848 tracing_enabled=ON. Categories declaration and utilization remain the
1849 same for MSG and SimDag.
1850 - A function was added to the tracing API to declare categories with
1852 - TRACE_category_with_color (char *category, char *color)
1853 where color must be in the following format
1854 "%f %f %f", red, green, blue
1855 and red, green, blue are float values in the interval [0, 1]
1856 - User can specify NULL as color parameter, or continue calling
1857 TRACE_category (cat)
1858 On that case, the tracing system will define random colors
1859 - The following command-line options are supported:
1860 --cfg=tracing/filename:msg.trace
1861 --cfg=tracing:1 (activate tracing, needed to use others)
1862 --cfg=tracing/platform:1 (categorized resource use)
1863 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1864 --cfg=tracing/msg/task:1 (task creation)
1865 --cfg=tracing/msg/process:1 (process creation, migration)
1866 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1867 --cfg=tracing/smpi:1 (SMPI interface tracing)
1868 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1869 - examples of examples/msg/tracing updated
1871 - DAXLoader and DOTLoader functions can generate tasks with categories
1872 - A new function to attribute a category to SD tasks:
1873 TRACE_sd_set_task_category (SD_task_t task, char *category)
1874 * Tracing the MPI interface implemented by SMPI
1875 - Collective operations are traced with states
1876 - Point-to-Point operations are traced with states/links
1877 - Tracing activated by a parameter "-trace filename" passed
1878 to smpirun during execution (considering that simgrid
1879 is compiled with tracing enabled)
1880 - To run the simulation with gdb, the simulator
1881 accepts --cfg=tracing/smpi:1 to trace SMPI
1882 - tesh files to check if smpi tracing is ok
1883 - See examples/smpi/NAS/DT-trace
1884 * GTNetS tracing re-worked
1885 - adaptation to the tracing system of GTNets to cope
1886 with modifications regarding the fullduplex mode
1887 - new tesh files to check if gtnets tracing is ok
1890 * Asynchronous communications through the functions:
1891 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1892 * New function: MSG_load_platform_script()
1893 to make possible using a lua script instead of XML files to set up platforms
1894 * New function: MSG_set_function
1895 to associate functions to processes, used when bypassing the parser
1896 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1898 Platforms: Add some more examples in examples/platforms
1899 * Grid'5000: see www.grid5000.fr
1900 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1903 * Change the XML format. This is a very important modification. SimGrid
1904 3.5 introduces a new hierarchical format based on the notion of
1905 Autonomous Systems. Compatibility with old format is ensured
1906 through the perl script provided in the install bin directory
1907 bin/simgrid_update_xml.
1908 It is now possible to build platforms with specific routing
1909 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1910 connect several platforms together. We will try to provide soon
1911 set of realistic platforms exploiting these properties (have a
1912 look at examples/platforms/ for the moment).
1913 * Take the opportunity of the XML format change to be a good XML citizen:
1914 rename link:ctn to link_ctn and similar changes (also dealt with by
1916 * Add a new routing scheme (rule-based) using regular expressions. It
1917 enables to have an extremely low memory footprint when the
1918 underlying routing is simple and can be compactly described. You
1919 need to have libpcre4-dev (perl regular expressions) installed if
1920 you want to use this routing scheme.
1921 * Revive the cluster TAG and allow to easily and efficiently (both in
1922 term of memory and speed) connect clusters together. Have a look
1923 at teshsuite/simdag/platforms/ to see how this can be done. With
1924 this tag, you can create clusters with thousands of tasks at no
1925 cost (have a look at examples/platforms/).
1926 Note: clusters are implemented as ASes, so there is no need for an
1927 enclosing AS tag if you have only one cluster in your platform.
1928 * Add new generic functions in the public interface that allows the user
1929 to call SURF 'create_resource' methods from your code (same
1930 functionality as the XML bypass mechanism but with a much lighter
1932 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1933 that uses a piecewise linear approximation to produce better
1934 results when exchanging small messages.
1935 * Add a new parameter to handle correctly full duplex link and account
1936 for interferences between uplink and downlink communications
1937 (activate with --cfg=fullduplex:1).
1940 * Rename the SD_READY (all dependencies are satisfied and task is
1941 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1942 dependencies are satisfied) state.
1943 This prevents a confusion between the notion of "ready to schedule"
1944 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1945 simulated" (SD_RUNNABLE) used by the simulation kernel.
1946 * Change the way a task is considered as ready. Instead of removing
1947 dependencies when a task is done, a counter is decreased. This way,
1948 it is always possible to reach ancestors thanks to the
1949 SD_taks_get_parents function (even after the end of the simulation.)
1950 * Change the return type of SD_Simulate from (SD_task_t*) into
1951 xbt_dynar_t. This function was in handling a dynar internally and
1952 converted it into a NULL terminated array for historical reasons.
1953 * New function SD_dotload(char*) to load a DAG described in dot
1954 format. This loader and the corresponding examples require the
1955 installation of the graphviz library.
1956 * Fix a bug in the management of tasks of size 0 in the surf network
1957 models. This problem was only visible with SIMDAG and you should
1958 thus disregard results produced with earlier versions if you
1959 relied on this feature (some tasks were blocked because of this).
1960 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1961 even though your applications were doing only point-to-point
1962 communications and sequential computations. Now you can really use any
1963 model you want (of course, if you create real parallel tasks, which are
1964 not implemented in most models beside ptaskL07, this will abort).
1965 * Add an example that schedules a DAX on an heterogeneous platform
1966 using a Min-Min strategy.
1967 * New function SD_workstation_get_current_task() that returns the kind
1968 of task currently running on a workstation in the sequential access
1970 * Raise some warnings when unexecuted tasks remains at the end of the
1971 simulation. This is usually caused by cycles in the DAG.
1974 * New function: SIMIX_process_set_function() called by MSG_set_function
1975 * Change the underlying waiting queue in semaphores so that a process
1976 can wait on several of them simultaneously (as in waitany).
1977 * Fix the way to handle tokens in semaphores so that all access patterns
1978 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1979 * kill the dirty pimple SIMIX_message_sizes_output()
1980 Please use (proper) visualization instead
1983 * New data container: setset (set of sets of elements)
1984 * New module: mmalloc (mapped malloc, allowing to have several
1985 independent segments of malloc)
1986 * New function: xbt_dict_cursor_set_data()
1987 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1988 * New function: xbt_dynar_is_empty()
1989 * New function: xbt_fifo_get_last_item()
1990 * Fix xbt_dynar_shrink(): use the right element size.
1991 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1992 uninitialized areas during expand.
1993 * Fix semaphores: previous implementation was severely broken.
1994 * Use library init/fini functions for our initialization.
1995 - you can use logs and other feature as soon as you want in your
1996 code (even before the xbt_init / MSG_init)
1997 - xbt_exit is now a no-op and produce a warning when used.
2000 * Port GRAS to new SIMIX mechanisms. This allows gras users to
2001 benefit from the latest improvement to the simulation kernel.
2002 * Kill measurement sockets for now. If you rely on them, sorry. This
2003 release is not for you. This feature will be reintroduced in the
2004 future, but we cannot delay the release any further.
2005 * New function: gras_msgtype_get_name().
2006 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
2007 * Fix (at last) the pmm example: it should not randomly fail anymore.
2009 Build chain: bug fixes and overall polishing
2010 * Cmake is now stable enough. Hence, we killed the autotools.
2011 * Port to windows ( TM :)
2012 * Fix the 'make install' target.
2013 No need to use 'make install-simgrid' anymore
2014 * Introduce a 'make dist' target compiling a *source* archive
2015 'make package' compiles a binary archive
2016 * Compile java files only on need
2017 * Add --cd and --setenv command line options to tesh
2018 * Out of source builds are not fully supported yet, but we are close
2019 * Enable supernovae and optimization flags by default for our users
2022 * Add layer to set up environment directly from lua, without XML.
2023 * The effect of gras_stub_generator can be achieved through
2024 lua too (check examples/gras/console/ping_generator.lua)
2026 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
2028 SimGrid (3.4.1) stable; urgency=low
2030 The "Polishing easter eggs is probably a good idea" release.
2031 This is a bug fixes release only.
2035 * Fix a bug preventing the tasks from begin garbage collected.
2038 * Fix a bug occuring when a host involved in a communication fails.
2039 This was not detected properly by the other peer involved in the
2040 communication. Now, it's reported as a network error.
2043 * Warn the user about loop dependencies in data flow of DAX files
2044 * Obey the control-flow dependencies of DAX files
2047 * Add option "enable_smpi" allowing to not compile SMPI.
2048 Probably useful for the (Mac) users experiencing a build error here
2049 * Improve the detection of lua5.1 and ruby1.8
2051 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2053 SimGrid (3.4) stable; urgency=low
2055 The "Easter in Cargese" release. Also known as (major changes):
2057 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2058 ~> bindings were greatly improved
2059 ~> new tracing infrastructure for better visualization introduced
2061 * the "Welcome to configury modernity" release.
2062 ~> we switched from autotools to cmake, and improved our cdash
2065 A more detailled list of changes follow (full detail in svn log).
2067 Java Bindings: Various Cleanups
2068 * (install java-gcj-compat-dev on debian-like to use them)
2069 * Remove put/get: no need to export deprecated interface in Java
2070 Use send/receive instead.
2071 * Cleanup the examples and add a README per directory
2072 * Remove example autoDestination (that's the only way to go now)
2073 * Remove example explicitDestination (was a plain copy of basic)
2074 * Make JniException a runtime exception, so that there is no need to
2075 declare the fact that you may encounter such a beast. I guess that
2076 nobody will ever want to survive such error.
2077 * Create specific errors for each MSG case of failure:
2078 host failure, transfer failure, timeout, task cancelled
2079 * Cleanup the exceptions that may get thrown by each function
2080 * Other internal cleanups in Java bindings. Performance still bad :/
2081 Ruby and Lua Bindings: create them
2082 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2083 * That's new and great, you should try them out.
2084 Same functionalities than Java bindings, only even less polished
2086 * Kill the useless "rate" argument of SD_task_get_execution_time()
2087 Everyone used to provide -1 as a value, it was not used, and the
2088 semantic of a possible use wasn't even clear.
2089 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2090 either as comm costs or compute costs to mean that there is no
2091 such thing for that specific task.
2092 * Add a SD_task_set_name() function
2093 * Fix SD_task_unschedule() on typed tasks
2094 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2095 * In DAX loader, accept useless 'level' attributes to <job> since
2096 LIGO DAGs have them (seem to be to ease graphical representation).
2098 * Add an example masterslave_mailbox.c using send/receive and not
2099 the deprecated put/get interface.
2100 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2101 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2103 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2104 (declare MSG_USE_DEPRECATED to still have the old name)
2105 * Add a MSG_task_set_data() function
2106 * About trace replay (see examples/msg/actions):
2108 - Allow to work with splitted trace files for each process
2109 Give the specific trace file as argument of each process,
2110 and call MSG_action_trace_run(NULL)
2111 You can still have one merged file for all processes.
2112 - Fix implementation of collective operations
2113 * Allow task_execute() on 0-sized tasks (closes #10063)
2115 * This is the first release of SimGrid where SMPI is not considered
2116 beta anymore (even if some corners should still be improved)
2117 * Port over the new SIMIX_network submodule (internal refactoring)
2118 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2119 * Implement more missing elements of the standard:
2121 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2123 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2124 MPI_Scan() MPI_Get_processor_name()
2125 - Added implementation of missing case for Alltoall (warning: it's
2126 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2127 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2128 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2129 SMPI_MPI_Allgatherv()
2130 * Bug fixes include:
2131 - MPI_Waitsome() was broken
2132 - Allow relative includes in smpicc
2133 - Command line cfg argument 'reference_speed' was ignored...
2134 - Some functions did not properly lead to auto-benching of user code
2135 - smpicc passes -O2 by default (just like openmpi one)
2137 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2138 * Bug fixes about timeouts during communications
2139 * add SIMIX_message_sizes_output() as a pimple to write to file the
2140 amount of messages per size. Use gnuplot to get histogram.
2141 Pimple because that's the only user-visible function of simix,
2142 defined directly in xbt.h (irk, sorry)
2144 - Add a SIMIX_sem_get_capacity() function
2145 - Fix interactions with processe resume/suspende
2146 - release_forever() was stupidly broken
2147 - Fix SIMIX_display_process_status() for processes in a semaphore
2148 - Make SIMIX_sem_block_onto() user-visible
2149 * Refactoring context stuff:
2150 - Use pseudo-OOP for better modularity
2151 - reimplement SIMIX_process_kill() without process_schedule() so
2152 that the latter can take as invariant that it is called from
2154 - Merge context_start into context_new for sake of simplicity
2156 * Add a Vivaldi network model, coded live during SUD'10 ;)
2157 * Rename configuration variables to start a hierarchy:
2158 o cpu_model -> cpu/model
2159 o network_model -> network/model
2160 o workstation_model -> workstation/model
2161 * New configuration variables:
2162 o network/bandwidth_factor: correction to bandwith
2163 o network/latency_factor: correction to latency
2164 o netwotk/weight_S: correction to the weight of competing streams
2165 * Add a long description to the models, that users can see with such
2166 argument on the command line: --cfg=cpu/model:help
2167 * --help-models display the long description of all known models
2169 * config: add the ability to set a default value after registration
2170 Does not override any previously set value (e.g. from cmd line)
2171 * dict: allow to have integer key and data.
2172 When so, you need to use the following functions
2173 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2174 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2175 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2176 In contrary to regular dicts, the key is not malloced before copy.
2177 Mixing scalar and regular elements in the same dict is not tested
2179 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2180 Tracing for Visualization:
2181 * SimGrid is now instrumented in order to generate a trace file for
2182 visualization analysis: to use it, need to compile SimGrid with the
2183 "tracing" option enabled, and instrument the program using SimGrid with
2184 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2185 (among other functions).
2186 * The instrumentation only traces the platform utilization for now
2187 * Documentation to use the tracing functions and how to analyze the
2188 traces with the Triva tool is written.
2189 * More information about: SimGrid FAQ (in the section Tracing Simulations
2192 * We moved to cmake as default build system. Autotools support will
2193 be dropped soon. Check the FAQ for more info about how to use it.
2194 * Greatly improved our cdash/ctest interactions
2195 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2196 * Added memory checking tests with valgrind; lot of memleak fixing.
2197 This may be the first release of simgrid with so few memory issues
2198 * Added code coverage tests.
2199 Our coverage is still improvable, but at least we see it on cdash.
2201 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2203 SimGrid (3.3.4) stable; urgency=low
2205 The "Desktop Grid needs love too" release (also called Xmas release).
2207 Models improvements:
2208 * Major speedup in the maxmin system solving by using lazy evaluation
2209 Instead of solving completely the maxmin system at each iteration,
2210 only invalidate (and recompute) the modified parts.
2211 This new feature is enabled in default models but you can try to
2212 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2213 * Cas01 IMproved as default CPU model
2214 This CPU model is the same Cas01 model, but it uses the
2215 maxmin-selective-update flag and a heap structure to manage
2216 actions on SURF kernel.
2217 It reduces the complexity to find the next action to finish and,
2218 consequently, it's faster than the old Cas01.
2219 This is the new default CPU model (Cas01).
2220 * Rename the old Cas01 model to Cas01_fullupdate
2221 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2222 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2223 * CpuTI (CPU Trace Integration)
2224 A new CPU model whose objective is simulate faster when using
2225 availability trace files.
2226 Instead of using a full featured, over engineered maxmin system for
2227 CPU modeling, this model does the pre-integration of traces files
2228 to calculate the amount of CPU power available, and so, executes
2229 faster than the old CPU models.
2230 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2231 * Use LV08 as default network model since it gives better accuracy
2232 for small messages and shouldn't change things for big ones.
2233 Use --cfg=network_model:CM02 to get the previous behavior.
2236 ******************************************
2237 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2238 ******************************************
2239 * The new CPU model may changes simulations!
2240 The point is that events occurring at the exact same timestamp
2241 are not scheduled in the same order with the old and new
2242 version. This may be enough to completely change the execution
2243 of simulations in some cases.
2244 * The new network model will change simulations!
2245 This new model is more realistic than the previous one, so you
2246 should consider redoing your old experiments with this model.
2247 Sorry for the inconvenience.
2250 * Introduce the supernovae compilation mode
2251 When compiled that way, the whole SimGrid (or almost) is put in a
2252 single compilation unit and compiled in one shoot.
2253 This is to help gcc which has difficulties to inline stuff from one
2255 The speedup seem to be above 15%, althrough more tests are needed on
2256 amd64 to confirm that gain.
2259 * Port of MSG's mailbox on top of SIMIX network
2260 The put/get mechanism was greatly simplified on the way.
2263 * New SIMIX network module. Provides:
2264 - Mailbox: rendez-vous mechanism to find with who you want to speak
2265 - Synchronous send/recv: easier and hopefully faster since the
2266 logic is handled in the maestro process directly now
2267 - Asynchronous send/recv: you dreamt of it? It's here now
2268 Too bad that nobody cared enough to propagate the change to MSG.
2269 * Add semaphores as SIMIX synchronization mechanism.
2272 * new function SD_daxload(char*) to load a DAX file
2273 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2274 * Introduce typed tasks. Specify its kind and cost at creation.
2275 At scheduling, just give where it should be placed, and the cost
2276 for each involved resource is automatically computed.
2277 Existing constructors so far (more to come of course):
2278 - SD_task_create_comm_e2e() for end-to-end communication
2279 - SD_task_create_comp_seq() for sequential computation
2280 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2281 * new function SD_task_dump() for debuging display
2282 * new function SD_task_dotty(task,FILE*) writing to file the info
2283 about the task in dotty format
2284 * SD_task_dependency_exists() can now cope with having one of its
2285 arguments NULL. If so, it tests whether the other argument has any
2287 * Add getters on list of preceding/following tasks:
2288 SD_task_get_parents(task) and SD_task_get_children(task)
2289 * Add getters on amount of workstations and list:
2290 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2291 * Add getter on task kind: SD_task_get_kind(task)
2292 * Update the start_time and finish_time of tasks on completion/failure
2293 * Bugfix: Remove task from state swags when destroyed
2296 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2299 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2300 dynar with one separate thread per value of the dynar.
2301 * Change the prototype of xbt_thread_create(), sorry.
2302 Added a boolean parameter indicating whether we want to join this
2303 thread (used in SG only for now)
2304 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2307 * GTNetS wrappers should now be usable again (and betterly tested too)
2308 * Fix a major regression from 3.2 where the timeout provided to
2309 MSG_task_put_with_timeout() was used as absolute time before which
2310 the comm should be done.
2311 * Start to fix the <cluster> tag.
2312 - Internal links should be good now (beside of the loopback, which
2313 use the private link instead)
2314 - paths to the external world is still rather broken
2315 - the <route:multi> tag is just broken. Actually that's brain-dead.
2316 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2317 to make it less stupid
2318 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2319 * Fix a source-level compatibility glitch from 3.2: after defining
2320 MSG_USE_DEPRECATED, you can use the old name
2321 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2322 * Allow to compile from the SVN with automake 1.11
2323 * Fix some problems when using the "start_time" tag in deployment XMLs.
2324 * Fix #8569: XBT/synchro.h has redundant declarations
2325 * Fix #8563: MSG return values and exceptions
2326 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2327 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2328 Thanks to Robson Peixoto.
2329 * Fix a vicious bug in dictionaries inducing that some elements were
2330 not freed on xbt_dict_free()
2332 Portability report of this version:
2333 * Main portability targets:
2334 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2335 on (amd64/i386/ia64)
2336 - mac leopard on i386
2337 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2338 but nothing critical.
2339 * Other platforms: windows, AIX and others were not tested for this release
2341 Timing report of this version:
2342 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2343 scenario parameters). From 8h to a few seconds in desktop grid settings.
2344 * Supernovae brings about 25% speedup on i386.
2346 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2348 SimGrid (3.3.3) stable; urgency=low
2350 The "Need for Speed" release.
2352 The timings done to validate the 3.3.2 were faulty.
2353 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2355 The problem was a conversion from a manually handled vector to
2356 xbt_dynar_t on the critical path.
2357 xbt_dynar_foreach calls functions, inducing stack management crap.
2359 We inlined these functions and xbt_dynar_foreach is now breath taking.
2360 We also inlined xbt_swag_belong on the way.
2362 Here are some approximate speedup measurements (on master/slaves
2363 simulations lasting between 10s and 20s each):
2364 3.3.1 -> 3.3.2: about same performance
2365 3.3.2 -> 3.3.3: 40% speedup
2366 3.3.1 -> 3.3.3: 40% speedup
2367 3.3.1 with inline patch -> 3.3.3: 30% speedup
2369 Our reading is that the refactoring which occurred in 3.3.2 made us
2370 suffer much more from the xbt_dynar_foreach low performance, but
2371 once we solved this, this refactoring proved to be very performance
2372 effective. From the 40% speedup, somehow, 10% are due to the
2373 inlining and 30% to the refactoring.
2375 That's a pitty that gcc cannot inline functions placed in other files
2376 alone. We have to choose between:
2377 - break the encapsulation (by putting private data structures and
2378 accessors in headers files to help gcc)
2379 - live with low performance
2380 - switch to a decent compiler such as icc (not quite possible).
2382 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2384 SimGrid (3.3.2) stable; urgency=low
2386 The "Simplicity does not preceed complexity, but follows it" release.
2388 The main contributors of this release were (lexical order):
2389 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2392 * Extract the routing logic into its own object.
2393 (was dupplicated in network.c and workstation_LV07.c;
2394 Allows to implement other ways of storing that info)
2395 => kill now useless network_card concept
2396 - Use dynar to represent routes (instead of void** + int*)
2397 - kill link_set (use surf_network_model->resource_set instead)
2398 - Add a command-line option to choose the routing schema to use
2399 - Add three new models:
2400 * Floyd (shortest path computed at initialization)
2401 * Dijikstra (shortest path recomputed all the time)
2402 * Cached Dijikstra (shortest path computed on need)
2403 All these models where contributed by Silas De Munck, and are
2404 described in his ICCS09 paper.
2406 * Simplify model declaration
2407 (less redirections, less function to write when defining a model)
2408 - Factorize stuff between models:
2411 surf_model_resource_set(model)
2412 surf_model_resource_by_name(model, name)
2413 - Unify the types of models in s_surf_model_t (using an union)
2414 - Embeed fields of common_public directly into s_surf_model_t
2415 - Rename model methods:
2416 action_free ~> action_unref
2417 action_change_state ~> action_state_set
2418 action_get_state ~> action_state_get
2419 - Change model methods into functions:
2420 (model)->common_public->action_use ~> surf_action_ref
2422 * Implement a generic resource; use it as ancestor to specific ones
2423 (allows to kill duplicated code in models)
2424 Drawback: timer command don't need no name nor properties;
2425 workstation_CLM03 don't need no properties
2426 (but I guess we can live with those few bytes wasted)
2428 * Improve the action object model
2429 - implement a constructor avoiding dupplicated code about field
2430 initialization in generic_action part.
2432 * Kill the SDP model: it has an external dependency, is deprecated
2433 in flavor of modern lmm models, and didn't compile since a while
2436 * Relocation of the context module from XBT to SIMIX.
2437 (the context were decoupled from the simix processes, duplicating a lot of code)
2438 => a lot of code was factorized
2439 - less overhead is introduced during scheduling
2440 - simpler API for the context factory
2441 - the logic for process creation,destruction and manipulation was simplified
2442 * Simplification of the s_smx_process_t data structure.
2443 => accesing the simix level data associated to a process is faster now,
2444 and the code is a lot more readable.
2447 * Implement some more MPI primitives:
2448 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2449 -implementation: Bcast: flat or 2-ary tree (default),
2450 Barrier: 4-ary tree,
2452 Allreduce: Reduce then Bcast
2453 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2454 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2455 Alltoallv: flat tree, like ompi
2457 * Add support for optimized collectives (Bcast is now binomial by default)
2458 * Port smpirun and smpicc to OS X
2461 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2462 Shout out if you used it.
2465 * Display the list of still queued messages in SG mode when existing
2469 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2470 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2471 * Add xbt_str_from_file(FILE*)
2472 * Add xbt_dict_get_key achieving a linear reverse search
2473 * Remove the context module
2475 Portability report of this version:
2476 * Main portability targets:
2477 - Linux(debian)/x86/context
2478 - Linux(debian)/x86/pthreads
2479 - Linux(debian)/amd64/context
2480 - Linux(debian)/amd64/pthreads
2481 On these, we still have the eratic breakages of gras/pmm and
2482 amok/saturate_sg reported in previous version. We still think
2483 that the tests are the cause of the fault, not the tested code.
2485 - Mac OSX Leopard/x86/context
2486 Still false negative in tesh autotesting.
2487 Smpi still fails, but this time because readlink does not accept -f
2488 Everything seems to work properly beside of that.
2491 - AIX version 5.3 (only tested contexts this time)
2492 Smpi still fails there because mktemp is not installed.
2493 Everything seems to work properly beside of that.
2495 I managed to compile it for the first time, but several breakages.
2496 Won't delay the release for this exotic platform.
2498 * Windows: it's still lagging behind. If you want to help, please
2501 Timing report of this version:
2502 This version seem to be more than 5% faster than 3.3.1 (on linux
2503 64bits with contextes). The gain is less than expected, we are
2504 investigating this for next release.
2506 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2508 SimGrid (3.3.1) stable; urgency=low
2511 * Implement a --cfg-help to show existing configuration variables
2512 * Build chain do not require doxygen in maintainer mode
2515 * fix a bug on struct sizeof computation, which prevented the
2516 exchange of arrays of structs in some conditions
2517 - added a regression test about this in datadesc_usage
2518 * Allow the exchange of 0-long dynamic vectors.
2519 - for that, use -1 as indicator of dynamic size instead of 0
2520 - This implied to change any size from unsigned long to long,
2521 reducing a bit communication abilities, but I guess that with
2522 64bits being quite common, this is more than enough.
2523 - This also induce a protocol change, thus bumping network protocol
2524 version from 0 to 1 (if we have external users, we have to get
2525 clean on that point too ;)
2526 - added two regression tests about this in datadesc_usage
2527 * Be more verbose when propagating local exceptions
2528 This helps debugging.
2529 * Display the status of simulated processes when receiving SIGINT in
2533 * Allow to control the simulation from a trace file.
2534 New functions MSG_action_register() and MSG_action_trace_run()
2535 The first one allows to associate a function execution to each
2536 kind of action while the second one parses a trace file and
2537 triggers the corresponding actions within the system.
2538 For now, only a toy example is provided in examples/msg/actions
2539 * Add an exemple of process migration in examples/msg/migration
2540 * Fix a bug in task exchange which broke MSG_task_get_sender()
2541 Add a teshsuite regression test for that.
2542 [Bug: if MSG_task_get_sender() is called after sender exit,
2544 * Fix a bug which prevented suspend/resume to work properly
2545 * Display the status of simulated processes when receiving SIGINT
2546 This fixes a regression of v3.3. due to the introduction of SIMIX
2547 * Bug fixing in failure management:
2548 - trace could not start by a failure at time 0
2549 - failure during communications were not working
2552 * Add SIMIX_process_set_name() to change the name of the current
2553 process in the log messages.
2554 * Store smx_hosts in a dict since we only retrieve them by name
2555 * Move the configuration infrastructure to surf
2558 * Move the configuration infrastructure to surf
2561 * Massive internal cleanups:
2562 - Store internal structures on processes instead of hosts (allows
2563 to have more than one process per host, in addition of being more
2565 - Cleanup the initialization/finalization process
2566 - Kill a whole bunch of unneeded synchronization:
2567 processes run in exclusive manner within the simulator
2568 - Move queues from global tables to process data fields
2570 - now accept -platform and -hostfile arguments
2571 - Pass the right rank value to processes according to the hostfile
2572 * Compile the examples by default, and use them as regression tests
2573 * Implement MPI_Wtime()
2574 * Change the reference speed to a command line option
2577 * TCP_gamma can now be specified as command line option using
2578 --cfg=TCP_gamma:10000000.0
2579 * Change the --surf-path cmd line option into --cfg=path:
2582 * Also include strbuff from xbt.h public header
2583 * xbt_ex_display(): do not free the exception after displaying
2584 This allows to do more with the given exception afterward.
2585 Users should call xbt_ex_free() themselves.
2589 Portability report of this version:
2590 * Main portability targets:
2591 - Linux(debian)/x86/context
2592 - Linux(debian)/x86/pthreads
2593 - Linux(debian)/amd64/context
2594 - Linux(debian)/amd64/pthreads
2595 These targets fail about 1/10 of times on gras/pmm, but we believe
2596 that this is because of the test, not because of simgrid.
2597 amok/saturate_sg fails even more rarely, and the test may not be
2600 - Mac OSX Leopard/x86/context
2601 The test suite still spits tons of errors because some obscure
2602 force prevents us from removing the temporary directories
2603 arguing that they still contain some metadata I've never heard of.
2604 Smpi fails because seq is not installed.
2605 Everything seems to work properly beside of that.
2608 - AIX version 5.3 (both contexts and pthread)
2609 Smpi still fails there because mktemp is not installed.
2610 XML inclusions seems rosty on AIX.
2612 * Windows: it's still lagging behind. If you want to help, please
2615 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2617 SimGrid (3.3) stable; urgency=high
2621 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2622 [Malek Cherier & Mt]
2624 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2627 Until now, GRAS were using MSG as an interface to SURF. It was
2628 quite difficult because both interface have several differences
2629 (MSG channels vs GRAS sockets were the most notable point).
2631 This also opens the gate to SMPI (which should occur soon) and speed
2632 up simulations by to 40% (even if it were not the main goal).
2634 **************************************
2635 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2636 **************************************
2637 The point is that events occuring at the exact same timestamp are
2638 not scheduled in the same order with the old and new version. This
2639 may be enough to completely change the execution of simulations in
2640 some cases. Sorry for the inconvenience.
2642 * Cleanup and upgrade the XML format to push further scalability
2643 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2645 * Improve the testing infrastructure with tesh. Now a very large part of
2646 the code is tested not only by being run but also by checking that the
2647 output match an expected output [Mt].
2649 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2650 is really less memory-demanding, which should allow you to use
2651 larger files in SimGrid [AL].
2653 * Inform valgrind about our contextes, so that it becomes usable
2654 with the default (and more effecient) version of SimGrid
2655 [contributed by Sékou Diakite, many thanks]
2658 * Introduce a listener thread in charge of receiving incoming
2659 messages from the network. It allows to overlap communication and
2660 computation but most notably, it removes some stupid deadlocks due
2661 to the fact that so far, a process could not send and receive at
2662 the same time. This made most non trivial communication schema
2664 * Convert the PIDs from long int to int to match the MSG ones (and
2665 linux ones too) [Mt]
2666 * New function: gras_agent_spawn() to launch a new process on
2667 current host. Only working in simulation for now. [Mt]
2668 * New function: gras_os_hostport() returning a constant form (ie,
2669 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2672 * Make the backtrace of exceptions more human readable [Mt]
2673 * New module: xbt/str [Mt]
2674 a ton of string utility functions (split, join, printf to a newly
2675 allocated buffer, trim, etc)
2676 * New module: xbt/hash [Mt]
2677 SHA1 hashing algorithm (more to come if needed)
2678 * New module: xbt/synchro [Mt]
2679 synchronization tools (mutex and conditions) working the same way
2680 in simulation and in real life (mainly useful for GRAS, but not
2682 * New module: xbt/queue [Mt]
2683 classical producer/consumer synchronization scheme
2684 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2685 (using the classical functions will get serialized) [Mt]
2686 * Make dictionary internal table dynamic. No need to specify its size
2687 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2689 * Make sure the log channels are organized as a tree under windows
2690 (because of ANSI C compatibility issue, any channel were child of
2691 root directly) [Mt].
2694 * Cleaned many thing in surf and fixed a few bugs [AL].
2695 * Add a nice command line configuration mechanism to compose models [AL].
2696 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2697 the previous one (KCCFLN05). It relies on something that looks like
2698 a max-min sharing mechanism but cannot be written as such. A new solver
2699 was thus designed [AL].
2700 * Add a new solver to lmm. Based on Lagrange optimization and
2701 gradient-based descent, it enables to efficiently maximise systems s.a
2703 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2705 This solver enables to propose two new network models for TCP Reno and
2706 TCP Vegas based on Low's work. These models still need to be fully
2707 tested though [Pedro Velho].
2710 * Bug fix in SD_simulate. Now the time bound given as argument is
2712 * Use the new parallel task model (ptask_L07) as default.
2713 * Use the SURF command line configuration mechanism.
2714 * 0-size tasks (for synchronization) should now work.
2716 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2718 SimGrid (3.2) stable; urgency=high
2722 We still experience issues on this platform, but we believe that at
2723 least MSG is usable.
2725 GRAS API BREAKAGE (for simplification purpose, sorry):
2726 * the gras_msgtype_by_name is not used anymore. Instead of
2727 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2728 you can write (and must)
2729 gras_msg_send(toserver, "request", &request);
2730 - If you still want to pass a gras_msgtype_t to the function (to cache
2731 the type and avoid the lookup time), use the gras_msg_send_() variant.
2732 - Impacted functions:
2733 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2734 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2735 * The callbacks are now expected to return 0 when everything went well
2736 (just like the main() function)
2738 GRAS new features and improvements:
2739 * New module mechanism where user code can use per process globals [Mt]
2740 This is similar to gras_userdata_*() functions, but for libraries. It
2741 factorize some code developped over and over in the examples and AMOK.
2742 It has still to be documented and used (only amok/peermanagement is
2744 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2745 loss when they were small enough to fit into the buffer and sent quickly
2746 enough so that they can all get received in one shoot.
2747 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2748 if not found. Use the *_or_null() variant for the old semantic.
2749 * In gras_msg_handle, do not discard messages without callback.
2750 They are probably messages to be explicitly awaited later (ie, proofs of
2751 mis-synchronization in userland since they are sent before being awaited)
2752 No big deal usually.
2753 * gras_socket_meas_send/recv: semantic changed!
2754 The numerical arguments used to be (1) the total amount of data to send
2755 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2756 messages. This was need for the fool willing to send more than MAXINT
2757 bytes on quite fat pipes.
2760 * Do really rename the hostmanagement module to peermanagement. [Mt]
2761 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2762 but this is rather new and this was documented in the module
2763 documentation (poor excuses, I admit)
2764 * Bandwidth measurement semantic changed! This follows the changes to
2765 gras_socket_meas_send/recv explained above.
2768 * A sequential mode has been added to the workstations. When a workstation
2769 is in sequential mode, it can execute only one task, and the other tasks
2770 are waiting in a FIFO. [Christophe Thiery]
2773 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2774 model for SIMDAG. [Christophe Thiery]
2775 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2776 equal to 0, which caused some bad behaviors in
2777 saturated_constraint_set_update. I now use a threshold mechanism like in
2781 * When running manually src/testall, you select specific units [Mt]
2782 testall is the result of our cunit mechanism, and should replace all
2783 the scripty thingy around since bash don't run easily on billware.
2785 * A mallocator system has been added. [Christophe Thiery]
2786 Mallocators allow you to recycle your unused objects instead of freeing them
2787 and allocating new ones.
2789 Documentation update:
2790 * FAQ reworking + New FAQs:
2791 - "Valgrind spits tons of errors!" [Mt]
2792 - "How to repport bugs" [Mt]
2793 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2794 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2795 - Communication time measurement within MSG [AL]
2796 - I experience weird communication times when I change the latency [AL]
2797 * GRAS tutorial [Mt]
2799 - an introduction to the framework and to the used communication model
2800 - an initiatic tour introducing the most proheminent features:
2802 . Lesson 0: Installing GRAS
2803 . Lesson 1: Setting up your own project
2804 o Part 2: Message passing
2805 . Lesson 2: Exchanging simple messages
2806 . Lesson 3: Passing arguments to the processes (in SG)
2807 . Lesson 4: Attaching callbacks to messages
2808 . Lesson 5: Using globals in processes
2809 . Lesson 6: Logging information properly
2810 . Lesson 7: Using internal timers
2811 . Lesson 8: Handling errors through exceptions
2812 . Lesson 9: Exchanging simple data
2813 . Lesson 10: Remote Procedure Calling (RPC)
2814 . Lesson 11: Explicitely waiting for messages
2815 . Recapping of message passing features in GRAS
2816 - A HOWTO section containing:
2817 o HOWTO design a GRAS application
2818 More are due, of course. They will come latter. In the meanwhile, you can
2819 check the examples which are still here.
2821 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2823 SimGrid (3.1) stable; urgency=high
2827 There was a stack corruption somewhere, visible only when optimizing
2828 with these versions. [Vince]
2831 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2832 module that enabled to play with Directed Acyclic Graphs. It is built
2833 directly on top of SURF and provides an API rather close to the old
2834 SG. Some old codes using SG are currently under rewrite to check that
2835 all needful functions are provided. [Christophe Thiery]
2838 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2839 extension of the classical CLM03 model that gracefully handles
2840 failures. This is now the default model for MSG and GRAS. It doesn't
2841 handle parallel tasks yet though. [AL]
2842 * Bug fix: Weights were not correctly set in the network part.
2843 WARNING: This may have resulted in incorrect results with simulations
2844 where there are more than one flow on a given link. [AL]
2847 * After a (long ?) discussion on simgrid-devel, we have decided that the
2848 convention we had on units was stupid. That is why it has been decided
2849 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2850 WARNING: This means that all previous platform files will not work as
2851 such with this version! A warning is issued to ask users to update
2853 A conversion script can be found in the contrib module of the CVS, under
2854 the name contrib/platform_generation/surfxml_update.pl [MQ]
2857 * Bug fix: Processes were started in reverse order, wrt deployment file.
2858 WARNING: if your code relies on this bug, please fix it. [AL]
2859 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2860 being executed on two different locations. [AL]
2861 * Bug fix: Failures are now better supported thanks to Derrick's tests
2862 (there was many failure situations I hadn't thought of and that weren't
2863 correctly handled). [AL]
2864 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2868 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2869 Exception occurring on server-side are propagated back to client (!).
2871 API CHANGE: the callback changed their prototype. Change:
2872 int my_handler(gras_socket_t expeditor, void *payload_data) {
2874 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2875 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2877 * New! function: gras_msg_handleall to deal with all messages arriving
2878 within a given period.
2879 * New! function: gras_socket_server_range to get a server socket in a
2880 range of port numbers (ease to avoid port number conflicts) [MQ]
2881 * New! gras processes display their backtrace when they get a SIGUSR1
2882 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2883 Sweet to debug RL processes [MQ]
2887 - Do not force experiment sizes to be expressed in kb, or it becomes
2888 impossible to measure the latency this way (needs one byte-long tests)
2889 WARNING: this changes the amok_bw_* function semantic. [MQ]
2890 - Implements the link saturation stuff. [MQ]
2891 * Peer management module:
2892 New! module factorizing code that we wrote over and over [MQ].
2895 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2896 - Test units are placed directly into the library code, they get extracted
2897 automatically and placed into the src/testall binary.
2898 - Convert most of the XBT tests to this system.
2899 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2900 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2901 live in the same C code. This required to change a little bit the API
2902 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2903 * New module: sparse graph structure with basic algorithms (this is work
2904 in progress and the API is not considered to be frozen yet). [AL]
2905 * Display more information on backtraces: source line & function names are
2906 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2907 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2908 * Reworked a little bit some #include statements to load only required
2909 headers. Some user code that relied on SimGrid to include stdlib or
2910 stdio may need to include it by themselves. [AL]
2911 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2913 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2914 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2915 but I guess that almost nobody used those parts. [MQ]
2917 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2919 SimGrid (3.0.1) stable; urgency=low
2922 * Unfortunately, I had missed 5 misnamed functions:
2923 xbt_fifo_item_t xbt_fifo_newitem(void);
2924 void xbt_fifo_freeitem(xbt_fifo_item_t);
2925 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2926 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2927 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2928 They're now deprecated. Please use their new versions:
2929 xbt_fifo_item_t xbt_fifo_new_item(void);
2930 void xbt_fifo_free_item(xbt_fifo_item_t);
2931 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2932 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2933 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2935 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2936 * Documentation: xbt_log module unmercifully reworked [MQ]
2937 * Bugfix: there was a problem with the ending of contexts with
2938 the pthread backend. It caused some weird deadlock or behavior
2939 depending on the pthread implementation. [AL]
2940 * Bugfix: get the exceptions raised in the simulator repport where
2941 and why they come from when they are not catched in time [AL, MQ]
2944 * Bugfix: Do repport the error when two non-connected hosts try to
2945 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2948 * Add additionnal checkings on communications. Assert that two
2949 communicating hosts are connected by a set of links... [AL]
2952 * Add additionnal checkings on channel values in communication [AL]
2953 * New: MSG_task_get_source to see on which host a task was generated [HC]
2954 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2955 the number of tasks waiting to be received on channel and sent
2957 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2958 waits for the first task coming from a given host.. [AL]
2960 GRAS new functionnalities: [MQ]
2961 * Enhance the parsing macro to allow the size of multidimentional objects
2962 to be given thru annotations.
2963 * New example (and documentation): Matrix Multiplication a la RPC
2964 (as when I was young!) and fix a bunch of bugs found on the way.
2966 GRAS performance improvements: [MQ]
2968 * Reduce the amount of cbps creation/destruction by making it static to
2969 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2972 * Change libdata to a set so that we can search for stuff by ID (and thus
2973 reduce the insane amount of dict lookups)
2976 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2977 RL communication on top of SG since it's so uneffective.
2978 It may also be used for inter-thread communication in RL, one day.
2979 * Use gras_datadesc_copy() to exchange messages on top of SG
2981 - improve message exchange performance on top of SG
2982 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2983 * Don't exchange on the network the size of the used part of buffer,
2984 instead, specify the possible buffer size to read().
2986 - reduces the amount of read/write calls (one pair per exchange)
2987 - reduces the amount of exchanged data (the size)
2988 - allows to retrieve all arrived data on receiver side, if we don't need
2989 it right now (subsequent read will peek the buffer)
2990 - allows the receiver to proceed with the begining of the stream before
2991 everything is arrived
2992 - make it possible to build an iov transport (using readv/writev)
2994 - take care of the data with non-stable storage (like stacked data),
2996 * If possible, TCP send uses vector I/O (when writev() is here)
2997 - Don't use it for receive since we send data sizes and data on the
2998 same stream, so we wouldn't be able to chain large amount of chunks
2999 before having to flush the stuff to read the size.
3000 * Rework the transport plugin mechanism to simplify it and reduce the
3001 amount of pointer dereferencement when searching for the right function
3004 * I guess that now, we do almost as few system calls as possible while
3005 doing as few data copy as possible.
3007 To improve it further, we could try to send all the sizes first and then
3008 all the data (to use iov on receiving size), but it's only a partial
3009 solution: when you have 2 dimensional data, the sizes of the second
3010 dimension is data of the first dimension, so you need 3 streams.
3012 I'm not sure the potential performance gains justify the coding burden.
3014 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
3016 SimGrid (3.00) stable; urgency=high
3019 * New! Give the possibility to hijack the surf parser and thus bypass
3020 MSG_create_environment and MSG_launch_application. Have a look at
3021 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
3023 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
3025 SimGrid (2.96) unstable; urgency=low
3030 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
3031 This deprecates the xbt_error_t mechanisms.
3032 It modifies (simplifies) all XBT and GRAS API.
3033 MSG API keeps unchanged (exceptions raised by XBT are catched from
3034 within MSG and masked with existing error handling facilities)
3037 * New! Add a FATPIPE model. [AL]
3038 * New! Add a parallel task model. [AL]
3039 * New! Add automatically a loopback interface (in the default
3040 network model) if none was precised.
3043 * Bugfix: MSG_process_resume now works with the current running process.
3045 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3046 * Modification of MSG_task_get_compute_duration. Once a task has been
3047 processed, the value returned by this function is now equal to 0. [AL]
3048 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3049 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3050 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3051 of a m_task. For now, MSG_task_cancel only works with computation
3053 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3054 of the processor (in Mflop/s) regardless of the current load on the
3056 * API Change: use proper naming convention for MSG_getClock and
3057 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3059 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3060 This function changes the priority of a computation task. This priority
3061 doesn't affect the transfer rate. A priority of 2 will make a task
3062 receive two times more cpu power than the other ones. This function
3063 has been added to suit the needs of Nguyen The Loc and hasn't been that
3064 much tested yet. So if it fails, please report it and send me your code.
3066 * API Change: removed all functions and types that were marked "deprecated"
3067 since many months. Renamed MSG_global_init_args to MSG_global_init.
3069 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3071 SimGrid (2.95) unstable; urgency=low
3074 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3075 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3076 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3079 * Bugfix: Allow absolute paths for platform description files [MQ]
3080 * Bugfix: do free the variables after use. Leads to drastic performance
3082 * Implement max_duration (ie, timeouts) on resources [AL]
3085 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3087 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3088 the needed support (timeouts) [AL]
3091 * Implement measurement sockets. You can now get the bandwidth between two
3092 hosts thanks to AMOK (see below). [MQ]
3093 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3094 dynar over the network (yeah) [MQ]
3095 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3096 * simple/static token-ring example. [Alexandre Colucci and MQ]
3097 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3098 in place before [MQ]
3101 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3102 type parsing. "short" and "long" modifiers were also ignored; other
3103 modifier (such as reference level) are still ignored. [MQ]
3104 * Embeed the buffer size within the buffer itself on SG. [MQ]
3105 That way, send() are atomic and cannot get intermixed anymore (at least
3106 the ones which are less than 100k; bigger messages still have the issue)
3107 * Array size pushed by the field, not by the field type (or each
3108 and every long int will push stuff to the cbps) [MQ]
3109 * use select() to sleep since it allows to portably sleep less than one
3112 GRAS (minor cleanups)
3113 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3116 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3117 * type category 'ignored' killed as it was never used and were difficult
3119 * whether a type can cycle or not is now a flag, leaving room for more
3120 flags (as "ignored", if we feel the need one day ;)
3121 * Rename raw sockets to measurement sockets since "raw" has another
3122 meaning in networking community.
3125 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3126 and offers features not belonging to GRAS but that most applications
3127 need. One day, it may be a set of plugins loadable at runtime.
3128 * New module: bandwidth
3129 bandwidth measurement between arbitrary nodes running this module. [MQ]
3131 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3133 SimGrid (2.94) unstable; urgency=low
3135 The first beta release of SimGrid 3 !
3139 * Update the main page and the FAQ. Adding references to gforge.
3142 * Add a gras_os_getpid function.
3145 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3146 * Extend the logs so that they also print PID, hostname, date, ... if
3148 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3149 and kill the old version which were in testsuite/
3150 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3154 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3155 for users. Instead, each programming environment (GRAS, MSG,...) use it
3156 when needed to provide such a feature to users.
3157 Don't shortcut the mechanism or you will also shortcut the virtualization
3158 you need on the simulator.
3162 * Cleanups in configury with regard to compile optimization/warning flags.
3163 Also add -fno-loop-optimize to any powerpc since it's the optimization
3164 killing gcc (< 3.4.0).
3165 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3166 needed by MSG examples complications
3167 * Borrow configury beautifications from PHP
3170 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3171 warning (thanks loris for stumbling into this one).
3172 * Bugfix: stop loading private headers (gras_config.h) from the public
3176 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3177 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3178 RL would imply a DNS resolver, which is *hard* to do in a portable way
3179 (and therefore delayed).
3180 * Implement a real timer mechanism and use it in timing macros. This allows
3181 to avoid rounding errors and get a 0.000005 sec precision in timing
3182 macros. While I was at it, various cleanups:
3183 - allow to declare more than one timed section per file (fix a stupid bug)
3184 - move some private declaration to the right place
3185 - merge conditional execution and timing macros into emulation module
3186 - document the module
3187 - make sure the module cleanups its mess on gras_exit
3188 * Documentation improvements:
3189 - (new) how to compile applications using GRAS
3190 - (new) emulation support (timing macros)
3192 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3194 SimGrid (2.93) unstable; urgency=low
3196 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3199 - Use Paje properly where used. Still to be sanitized properly.
3200 - Portability fix: Add an implementation of the contexts using pthread
3204 - Add xbt_procname(): returns the name of the current process.
3205 Use it to show the current process's name in all logging.
3207 - fix detection of older flex version and the reaction, since we do
3208 depend on modern ones (we use lex_destroy)
3209 - Better separation of SG and RL in the libs: remove all simulation code
3210 from libgras. As a result, this lib is now only 200k when stripped.
3211 Some of the xbt modules may also be duplicated (two sets and such) and
3212 should be cleaned/killed before SG3.
3213 - Insist on using xlC on AIX because of weird problems involving gcc there.
3214 - Cleanup the make remote stuff. This is now done by scripts
3215 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3216 mainly for our private use, but we're working on changing them to user
3219 - Bugfix: flush the socket on close only if there is some *output*.
3220 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3222 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3225 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3227 SimGrid (2.92) unstable; urgency=low
3229 Alpha 3 on the path to SimGrid 3
3233 - New! First try of benchmarking macros.
3234 - New! First try so that gras_stub_generator generate deployment and
3235 remote compilation helpers.
3237 - Bugfix: Initialization fix in msg_test.
3241 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3243 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3245 - API Change: gras_os_sleep to take the amount of seconds as a double.
3246 Accepting an int was error prone since it was the only location where
3247 seconds were coded as such. It leaded to damn rounding errors.
3248 - Bugfix: Hard to belive that timers ever worked before this.
3250 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3252 SimGrid (2.91) unstable; urgency=low
3254 Alpha 2 on the path to SimGrid 3
3258 - Bug fix in the lmm_solver.
3260 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3261 through the function MSG_paje_output.
3262 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3263 - It is possible to bound the rate of a communication in MSG with
3264 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3265 to write it in the changelog).
3266 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3271 - Various cleanups to the autotools stuff
3272 - Begin to move Gras examples to examples/gras/
3273 - Let make distcheck work again (yeah!)
3275 - documentation overhauled using doxygen.
3276 gtk-doc-tools is dead in SimGrid now.
3277 - Automatically extract all existing logging categories, and add the list
3278 to the documentation (long standing one, to say the less)
3280 - Cleanup the known architecture table. Reorder the entries to group what
3281 should be, and use a more consistent naming scheme.
3282 (some of the test dataset are still to be regenerated)
3283 - New! Allow library to register globals on each process just as userdata
3285 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3286 lookup time (for now).
3287 Use it in msg and trp.
3288 This cleans a lot the internals and helps enforcing privacy of the
3289 headers between the gras components.
3290 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3291 - Bugfix: gras_os_time was delirious in RL.
3292 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3294 - Reenable GRAS now that it works.
3296 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3298 SimGrid (2.90) unstable; urgency=low
3300 Alpha 1 on the path to SimGrid 3
3302 * It is a long time since the last release of SimGrid. I'm sorry about
3303 that but as I had told you, I was rewriting a lot of things. I apologize
3304 to those who had been reporting bugs to me and that I had not answered.
3305 If your bug is still in the new version, please tell me. Here is a
3306 summary of the main changes.
3308 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3309 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3310 portability, speed, and a lot more but you'll figure it out later.
3311 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3312 I wanted to release the new MSG as soon as possible and I have
3313 broken GRAS, which is the reason why, for now, only MSG is fully
3314 functional. A laconic description of these projects is available
3315 in the documentation.
3317 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3318 kernel optimized for our needs (called SURF but only the developers
3319 should use it). Hence, MSG is now roughly 30 times faster and I think
3320 that by rewriting a little bit MSG, I could event speed it up a little
3321 bit more. Beside the gain in speed, it is also much easier to encode a
3322 new platform model with SURF than it was with SG. More to come...
3324 * REVOLUTION 3: I have tried to change a little as possible the API of
3325 MSG but a few things really had to disappear. The main differences
3326 with the previous version are:
3327 1) no more m_links_t and the corresponding functions. Platforms are
3328 directly read from a XML description and cannot be hard-coded
3329 anymore. The same format is used for application deployment
3330 description. The new format is described in the documentation.
3331 Have a look in tools/platform_generation. There is a tiny script
3332 that converts from the old platform format to the new one. Concerning
3333 the application deployment format, parsing the old one is tricky.
3334 I think most of you should however be able to convert your files. If
3335 it is really an issue, I can write a C code that does the conversion.
3337 2) the toolbox tbx does not exist anymore. We now have a library
3338 with much more data-structures but without the hash-tables (we have
3339 dictionaries that are much faster).
3341 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3343 *****************************************************************************
3344 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3345 * don't feel like converting the oldies. (Mt) *
3346 *****************************************************************************
3349 Version 2.90: "the long awaited one"
3350 - Finished rewriting and debugging MSG. Rewrote the documentation.
3351 - disable GRAS for now since it needs to be ported to the newest SG
3354 - Finish the port to windows (using mingw32 for cross-compile)
3357 - Main loop and datastructures of SURF. A cpu resource object is
3358 functional. Surf can thus be used to create cpu's with variable
3359 performance on which you can execute some actions.
3361 2004-11-15 Martin Quinson
3362 - Port to ARM. Simply added the alignment and size descriptions. Should
3363 work, but the ARM machines are so slow that I didn't had the opportunity
3364 to 'make check' over there yet.
3366 2004-11-15 Arnaud Legrand
3367 - Trace manager now written. It uses a heap structure and is therefore
3368 expected to be efficient. It may however be speeded up (particularly
3369 when many events occur at the same date) by using red and black
3370 trees. One day maybe...
3371 - Max-min linear system solver written. It uses a sparse matrix
3372 structure taking advantage of its expected use. Most operations are
3373 O(1) and free/calloc are called as few as possible. The computation of
3374 the minimum could however be improved by using a red and black tree
3377 2004-11-03 Arnaud Legrand
3378 - Rename every gras_* function that was in xbt/ to its xbt_
3380 - Add a heap and a doubly-linked list to xbt
3381 - Added a dichotomy to the dictionaries. make check works as well before
3382 so I assume that the patch is correct. I do not know however if things
3383 run effectively faster than before now. :)
3385 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3386 SimGrid, and the version number is bumped to 2.x
3388 2004-10-29 Martin Quinson
3389 - Introduction of the remote errors.
3390 They are the result of a RMI/RPC on the remote machine.
3391 ErrCodes being scalar values, you can't get the host on which those
3392 errors did happen. Extending the error mechanism as in Gnome is possible.
3393 No idea yet whether it is a good idea.
3395 2004-10-28 Martin Quinson
3396 - Interface revolution: the Starred Structure Eradication.
3397 I used to do typedef struct {} toto_t; and then handle *toto_t.
3398 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3399 - ???_t is a valid type (builded with typedef)
3400 - s_toto_t is a structure (access to fields with .)
3401 - s_toto is a structure needing 'struct' keyword to be used
3402 - e_toto_t is an enum
3403 - toto_t is an 'object' (struct*)
3405 typedef struct s_toto {} s_toto_t, *toto_t;
3406 typedef enum {} e_toto_t;
3407 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3408 s_toto_t) is private.
3410 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3411 so that it can set the variable to NULL. It was so for dicts and sets,
3412 it changed for dynars.
3414 - Fix a bunch of memleaks in dict_remove
3415 - Fix a bug in sg/server_socket opening: it failed all the time.
3417 2004-10-07 Martin Quinson
3418 - Speed up dynar lookup operation a bit.
3420 gras_dynar_get is dead.
3422 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3423 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3424 (which gives you the address of the stored data).
3426 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3427 copy of the data using an affectation to do the job and not a memcpy.
3429 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3430 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3432 It does not really speedup the dynar test because they are
3433 setting elements all the time (and look them seldom). But the dict does
3434 far more lookup than setting.
3436 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3438 2004-10-05 Martin Quinson
3439 - Allow to (en/dis)able the cycle detection at run time.
3441 Whether we should check for cycle or not is now a property of each
3442 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3443 datadesc_cycle_unset allow to remove this property when previously set.
3445 Note that the cycle detection is off by default since it impacts the
3446 performance. Watch the data you feed GRAS with ;)
3448 This property is hereditary. Any element embedded in a structure having it
3449 set have it set for the time of this data exchange.
3451 You should set it both on sender and receiver side. If you don't set it on
3452 sender side, it will enter an endless loop. If you forget on receiver
3453 side, the cycles won't be recreated after communication.
3455 - Header reorganization.
3456 Kill gras_private.h, each submodule must load the headers it needs.
3458 2004-10-04 Martin Quinson
3459 - Interface revolution: do not try to survive to malloc failure.
3461 Now, gras_malloc and friends call gras_abort() on failure.
3462 As a conclusion, malloc_error is not a valid error anymore, and all
3463 functions for which it was the only gras_error_t return value are
3464 changed. They now return void, or there result directly.
3465 This simplify the API a lot.
3467 2004-09-29 Martin Quinson
3468 - Re-enable raw sockets.
3469 Created by gras_socket_{client,server}_ext;
3470 Used with gras_raw_{send,recv}
3473 It should allow to kill the last bits of gras first version soon.
3475 This is not completely satisfactory yet (duplicate code with
3476 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3479 - Simplify transport plugin (internal) interface by not passing any
3480 argument to _server and _client, but embedding them in the socket
3483 2004-09-28 Martin Quinson
3484 - Finish the port to AIX.
3485 autoconf was my problem (segfault within the malloc replacement
3486 function. No idea why)
3488 2004-09-16 Martin Quinson
3489 - Fix some size_t madness on 64bit architectures.
3491 2004-09-08 Martin Quinson
3492 - Reduce the number of system headers loaded, overload some more system
3493 calls (such as malloc to cast the result of the system one, and work
3495 - Fix and reintroduce the config support
3497 2004-09-07 Martin Quinson
3498 - Source code reorganization to allow Arnaud to surf all over there.
3499 - Allow to document the logging categories.
3500 - Remove all uppercase from logging categories and useless cleanup in names.
3502 2004-08-18 Martin Quinson
3503 Version 0.6.2 (protocol not changed; API changed)
3504 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3505 gras_error_t), and NULL when not found. Functions expecting a msgtype
3506 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3507 hopefully usefull message.
3508 - Portability to prehistoric sparcs again
3510 2004-08-17 Martin Quinson
3511 Version 0.6.1 (protocol not changed; ABI not changed)
3512 - prealloc some buffers to speed things up
3514 2004-08-11 Martin Quinson
3515 Version 0.6 (protocol not changed; ABI expended)
3516 - The parsing macro can deal with the references, provided that you add
3517 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3519 2004-08-09 Martin Quinson
3520 Version 0.5 (protocol not changed; ABI changed)
3521 - Allow to off turn the cycle detection code in data exchange at
3522 compilation time. It should be at run time, but I'm short of time (and
3523 the config stuff is still broken). That way, we keep dict out of the
3524 critical path, which is good because the performance is poor:
3525 - search not dichotomial yet
3526 - dynar give no way to access their content and memcpy everytime
3527 - In composed data description (struct, ref and so on), stop foolness of
3528 keeping the subtype's ID, but store the type itself. This keeps sets out
3529 of the critical path, which is good since they rely on dynar and
3530 dictionnaries. The only loose of that is that we cannot detect the
3531 redeclaration of a structure/union with another content (but I'm not sure
3532 the code detected well this error before anyway). We still can detect
3533 the redefinition discrepancy for the other types.
3534 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3535 it breaks the code because of type-punning used all over the place).
3536 This breaks on all non-gcc architectures (for now).
3538 All those changes (plus the buffer of last time) allow me to gain 2 order
3539 of magnitude on cruel tests consisting of 800000 array of integers on two
3540 level of a hierarchical structure (200 secondes -> 4 secondes)
3543 - the selector of reference must now return the type it points to, not
3544 the ID of this type.
3546 2004-08-06 Martin Quinson
3547 Version 0.4 (protocol changed; ABI not changed)
3548 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3549 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3550 - Implement a buffer transport, and use it by default (it relies on tcp in
3551 real life and on sg in simulation).
3552 That's a bit hackish since I had a new field to the structure to store
3553 its data without interfering with the subtype ones. Inheritance
3554 is tricky in C. And that's a kind of reverse inheritance with one class
3555 derivating two classes. Or maybe a game with java interfaces. Anyway,
3556 that's damn hard in C (at least).
3557 Moreover, I got tired while trying to ensure plugin separation and
3558 genericity in SG mode. MSG wants me to do weird things, so let's go for
3559 cruel hacks (temporarily of course ;).
3560 See comment in transport_private.h:71
3561 - do not include all the _interface headers in private but in the files
3562 which really need them (to cut the compilation time when they are
3565 2004-07-26 Martin Quinson
3566 Version 0.3 (protocol not changed; ABI changed)
3567 - Major overhault of the datadesc interface to simplify it:
3568 - shorted the function names:
3569 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3570 - add a trivial way to push/pop integers into the cbps without malloc.
3571 This allows to make really generic sub_type description, which simply
3572 pop their size of the stack.
3573 - add a function gras_datadesc_ref_pop_arr() which does what users want
3574 most of the time: Declare a dynamic array (which pops its size of the
3575 stack) and declare a reference to it. Poor name, but anyway.
3576 - kill the post-send callback, add a post-receive one
3578 2004-07-23 Martin Quinson
3579 Version 0.2 (protocol changed; ABI changed)
3580 - add some testing for cpbs in the test cases, and fix some more bugs.
3581 This invalidate again the little64 data file, since I cannot regenerate
3583 - remove an awfull optimization in the logging stuff, allowing me to:
3584 - understand it again
3585 - learn gcc how to check that the argument match the provided format
3586 - fix all errors revealed by gcc after that
3587 - internal keys of dict are not \0 terminated. Deal with it properly in
3588 loggings instead of segfaulting when the user want to see the logs :-/
3590 2004-07-22 Martin Quinson
3591 - Fix some stupid bug preventing cbps (callback postit) from working
3593 2004-07-21 Martin Quinson
3594 - Some documentation cleanups
3595 - remove the useless last argument of msgtype_declare
3596 - rename the Virtu functions to fit into the 'os' namespace
3597 - move headers src/include -> src/include/gras/ and stop fooling with
3598 gras -> . symbolic link
3599 - make distcheck is now successful
3601 2004-07-19 Martin Quinson
3603 - Build shared library also
3604 - Install html doc to the right location
3605 - stop removing maintainer files in make clean
3606 - build tests only on make check
3608 2004-07-13 Martin Quinson
3610 - No major issue in previous version => change versioning schema
3611 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3612 corresponding dataset.
3614 2004-07-11 Martin Quinson
3616 - Get it working with any kind of structure (we can compute the padding
3617 bytes remotely for all the architectures I have access to)
3618 - Implement the structure parsing macro (still not quite robust/complete)
3619 - Improvement to the remote testing toysuite
3621 2004-07-10 Martin Quinson
3622 [autoconf mechanism]
3623 - get ride of a bunch of deprecated macros
3624 - actually run the test for two-compliment, not only compile it :-/
3625 - test whether the structures get packed (and bail out if yes. Damn.
3626 Alignment is a serious matter)
3627 - test whether the structures get compacted (but respecting the alignment
3628 constraints of each types)
3629 - test whether the array fields of structures can straddle alignment boundaries
3631 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3632 (took me 2 hours to find that bug, looking at the wrong place)
3633 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3634 before sending/receiving and are used to compute the offsets of fields
3635 - Given that padding size depend even on compiler options, keep track of
3636 alignment and aligned_size only for the current architecture. Not a big
3637 deal since we send structure fields one after the other (seems
3639 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3640 program, let it work on linux/gcc/little32. portability todo.
3642 2004-07-08 Martin Quinson
3643 - import and improve remote compilation support from FAST
3644 - make sure make check works on half a dozen of machines out there
3646 2004-07-07 Martin Quinson
3647 Let's say it's version 0.0.3 ;)
3648 - Implement conversions (yuhu!)
3649 - Let it work on solaris (beside conversion, of course)
3650 - Stupid me, using rand() to generate the conversion datatests in not wise.
3652 2004-07-06 Martin Quinson
3653 - Let make dist work, since I'm gonna need it to compile on remote hosts
3654 - Let Tests/datadesc_usage write the architecture on which the file was
3655 generated as first byte.
3656 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3657 architecture descriptions.
3658 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3659 on those architectures.
3660 - Optimization: send/recv array of scalar in one shoot
3662 2004-07-05 Martin Quinson
3663 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3665 - Plug a whole bunch of memleaks
3666 - each process now have to call gras_{init,exit}. One day, their log
3667 settings will be separated
3668 - Continue the code factorisation between SG, RL and common in Transport.
3670 2004-07-04 Martin Quinson
3672 - Redistribution between SG and RL.
3673 We wanna have to accept in SG, so move accepted related parts of RL in
3674 the common part. (more precisely, the dynar of all known sockets is no
3675 more a static in transport.c, but part of the process_data)
3677 [gras_stub_generator]
3678 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3680 2004-07-03 Martin Quinson
3681 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3682 them all up in one shot)
3684 - Fix the ugly case of reference to dynamic array.
3685 - New (semi-public) function gras_datadesc_size to allow the messaging
3686 layer to malloc the needed space for the buffer.
3688 - gras_socket_close now expect the socket to close (and not its address to
3689 put NULL in it after it). This is because the socket passed to handlers
3690 is one of their argument (=> not writable).
3692 - propagate the interface cleanup from last week in datadesc, ie remove a
3693 superfluous level of indirection. User pass adress of variable
3694 containing data (both when sending and receiving), and not of a variable
3695 being a pointer to the data. Let's say that I like it better ;)
3696 The price for that is constructs like "int msg=*(int*)payload" in
3697 handlers, but it's a fine price, IMHO.
3699 - Let it work in RL (yuhu)
3701 2004-06-21 Martin Quinson
3703 - porting SG plugin and SG select to new standards (works almost).
3704 - plug memleaks and fix bugs around.
3707 - cleanup the prototype of data recv and force users to specify when they
3708 want to handle references to objects. Test case working even for cycles.
3709 - plug memleaks. Valgrind is perfectly ok with this.
3711 2004-06-12 Martin Quinson
3713 - cleanup the separation between plugin and main code in plugin creation
3715 2004-06-11 Martin Quinson
3717 - Reput hook for raw sockets, needed for BW experiments
3718 - kill a few lines of dead code
3719 [Data description] Interface cleanup
3720 - gras_datadesc_by_name returns the searched type or NULL.
3721 That way, no variable is needed to use a type desc once, which makes
3723 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3724 two parameters were strings (field name, type name), leading to
3726 [Dicos] Interface cleanup
3727 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3728 This is consistant with the dynar API.
3730 2004-04-21 Martin Quinson
3732 - Porting to new standards.
3734 - interface cleanup.
3735 There is no bag anymore, no need to take extra provision to mask the
3736 pointers behind "ID".
3737 Better splitup of functions between files create/exchange/convert.
3738 This is still a bit artificial since convert and receive are so
3739 interleaved, but anyway.
3741 - add a queued message list to procdata (the ones not matching criteria
3743 - factorize some more code between SG and RL wrt procdata
3745 - use gras_exit in example to track memleaks
3746 - get rid of gs_example now that GS is properly integrated into gras
3747 - update run_test to integrate the latest tests (datadesc)
3749 - rename WARNINGn macros to WARNn since it prooved error-prone
3751 2004-04-19 Martin Quinson
3753 - register init/exit functions within gras module mechanism
3754 - send/receive function.
3755 Convertion is not implemented, but short-cutted if not needed.
3756 struct/array elements are sent one by one (instead of block-wise), but
3757 nobody really cares (yet). Get a prototype before optimizing.
3758 - tests (using a file socket) for DD send/receive on:
3759 - base types: int, float
3760 - array: fixed size, string (ie ref to dynamic string)
3761 - structure: homogeneous, heterogeneous
3762 - chained list, graph with cycle
3763 Believe it or not, valgrind is not too unhappy with the results. The
3764 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3768 - Bugfix when using a filename explicitely (instead of '-')
3770 2004-04-09 Martin Quinson
3772 - factorize more code between RL and SG in socket creation
3773 - Complete the implementation and tests of:
3775 o file (only in RL, and mainly for debugging)
3777 I lost 3 days to design a portable address resolver, and then decided
3778 that the prototype mainly have to run on my box.
3779 Addressing portability too early may be like optimizing too early :-/
3781 - use gras_init in the Tests instead of the crappy parse_log_opt
3782 (the latter function is removed)
3783 [Conditional execution]
3784 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3785 [Code reorganisation]
3786 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3787 Build examples against the RL library, since there is no way to disable
3788 its creation for now.
3790 For information, the beginning of coding on GRAS was back in june
3791 2003. I guess that every line has been rewritten at least twice since