1 SimGrid (3.21) NOT RELEASED (Release Target: September 23. 2018, 1:54 UTC)
4 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
5 - New host variable: 'core_count'
8 - Remove xbt_os_thread_specific features
9 - Remove portability wrapper to condition variables
10 - Remove xbt_os_thread_yield()
13 - #93: simgrid should not eat --help
14 - #264: Add ptask L07 resource tracing
15 - #271: Dynamic and manual replay of SMPI traces
16 - #279: Breakpoints option uses time deltas instead of absolute time
17 - #282: TIT convertor erase traces when using absolute path in the trace list file
18 - #285: segfault when a s4u actor kills itself with msg process tracing activated
19 - #286: Pajé traces are not exposing the number of cores
20 - #287: Command-line parsing should stop when encountering '--'
22 ----------------------------------------------------------------------------
24 SimGrid (3.20) Released June 24. 2018
26 The proxy snake_case() release.
29 - The whole API is now using snake_case() instead of mixing it with camelCase.
30 Compatibility wrappers in place for one year.
31 - We took the opportunity to increase the consistency of the API.
32 Please report any remaining glitches.
33 - New function simgrid::s4u::Host::get_actor_count:
34 Returns the number of actors running on a specific host.
37 - SMPI is now tested with ~45 proxy apps from various sources, with none or
38 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
39 - Replay: The replay file has been re-written in C++.
40 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
41 supported. They are stored in the trace and used when replayed.
42 - Basic support of MPI_Cancel. Robustness not guaranteed.
43 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
44 (only for MPI_COMM_TYPE_SHARED).
45 - New option: smpi/privatize-libs, to add external shared libs to be privatized
46 by SMPI. They will be copied locally and loaded separately by each process.
47 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
48 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
49 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
50 with internal index optind with multiple processes. Only works if MPI_Init has
52 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
53 syntax without preprocessing tricks.
56 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
57 behavior tracing is no longer limited to MSG
58 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
59 this means that traces now need to be updated or re-obtained; both Irecv
60 and Isend lines in your traces have now in total 5 mandatory fields:
61 <rankid> <command> <to/from rankid> <tag> <size>
62 To update your traces, it suffices to add a 0 for the tag here.
63 - TIT now also supports waiting for a distinct request via MPI_Wait.
64 Wait/Test now wait for a specific request, not just the last one that was
65 issued. This unfortunately means another update, because we need to
66 identify which request you want to wait for. We do this via the
67 triplet (sender, receiver, tag), which needs to be added:
68 <rankid> <command> <sender> <receiver> <tag>
69 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
70 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
73 - The deprecation of MSG is ongoing (but this should not impact you).
74 Many MSG functions are now simple wrappers on the C API of S4U. If
75 you wish to convert your code to S4U, find the S4U counterparts of
76 your MSG calls in src/msg/msg_legacy.cpp.
77 - MSG can still be used, but won't evolve anymore.
80 - Allow to run the Link energy plugin from the command line with
81 --cfg=plugin:link_energy
82 - Rename Energy plugin into host_energy
83 - Rename Load plugin into host_load
86 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
87 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
88 - Improve performance when handling timeouts of simix synchros.
91 - Config: the C API is now deprecated (will be removed in 3.23), and
92 the C++ API has been slightly improved.
95 - Fix several build issues on OSX.
96 - Move simgrid_config.h to simgrid/config.h (old header still working)
99 - #143: Setting a breakpoint at a given time
100 - #150: Inconsistent event names in SMPI replay
101 - #258: daemonized actors hang after all non-daemonized actors have completed
102 - #267: Linker error on unit_tmgr
103 - #269: SMPI: tracing of MPI_Wait/all/any broken
104 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
106 ----------------------------------------------------------------------------
108 SimGrid (3.19.1) Released March 22 2018
110 The Fixed ABI Release.
112 As you may know, we are currently refactoring SimGrid in deep.
113 Upcoming SimGrid4 will be really different from SimGrid3: modular,
114 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
116 Our goal is to smooth this transition, with backward compatibility
117 and automatic update paths, while still progressing toward SimGrid4.
119 SimGrid remains open during works: The last pure SimGrid3 release was
120 v3.12 while all subsequent versions are usable alpha versions of
121 SimGrid4: Existing interfaces remain unchanged, but the new S4U
122 interface is budding and the internals are deeply reorganized.
124 Since 2015, we work hard to reduce the changes to public APIs. When
125 we need to rename a public library symbol in S4U, we let your
126 compiler issue an explicative warning when you use the deprecated
127 function. These messages remain for four releases, i.e. for one full
128 year, before turning into an error. Starting with v3.15, your can
129 also adapt to API changes with the SIMGRID_VERSION macro, that is
130 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
132 Starting with this v3.19.1, our commitment to reduce the changes to
133 the public interfaces is extended from the API to the ABI: a program
134 using only MSG or SimDag and compiled against a given version of
135 simgrid can probably be used with a later version of SimGrid without
136 recompilation. We will do our best... but don't expect too much of
137 it, that's a really difficult goal during such profund refactoring.
139 The difference between v3.19 and v3.19.1 is that the former was
140 accidentally breaking the ABI of MSG, while the later is restoring
143 S4U and kernel APIs will still evolve until SimGrid4, with one-year
144 deprecation warnings as currently. In fact, cleaning up these
145 interfaces and converting them to snake_case() is one release goal of
146 v3.20. But don't worry, we are working to smooth this upgrade path.
148 In summary, new projects should start with S4U to benefit of the
149 future, but old MSG projects should still be usable with no change.
151 ----------------------------------------------------------------------------
153 SimGrid (3.19) Released March 20 2018
155 The Moscovitly-cold Spring release.
157 SMPI user-visible improvements:
158 - Support of MPI_Type_create_subarray
159 - Unimplemented MPI calls now abort (instead of issuing a warning)
160 Some harmless calls around errhandlers still only issue a warning.
161 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
162 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
163 is now possible (mmap privatization was refactored to this end).
164 Adding new SMPI instances should also be technically doable now.
166 SMPI internal cleanups:
167 - Removed index notion from SMPI -- no more "getPid() - 1"!
168 In tracing files, actors are called rank-<PROCESS-ID> but these are
169 NOT the real ranks (tracing several communicators is not supported yet).
170 - Deprecate smpi_process_index() (will be removed in 3.22)
171 - Better testing on classical proxy apps, through an external project:
172 https://github.com/simgrid/SMPI-proxy-apps/
175 - Use a graphical TOC to make it easier to find the documentation you need
176 - Revamp the MSG tutorial
179 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
180 - Remove parameter of MSG_process_killall().
181 Resetting the PID was bogus anyway (several actors could have the same PID).
184 - Execution->setHost() can be called after start(), to migrate it.
185 - Comm::test_any() is now implemented.
186 - s4u::Actor now has onCreation() and onDestruction() signals.
187 - Install some kernel header files for the users' plugins and more.
188 Warning, their API is really not stable yet.
189 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
190 include/simgrid/kernel/routing/*Zone.hpp
193 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
195 Build System and other cleanups:
196 - Remove unused run-time parameter "tracing/onelink-only".
197 - SimGrid now works with intel compilers (MC still buggy)
198 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
199 - tesh kills subprocesses on timeout (but not on windows)
202 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
205 - #194: Feature request: simgrid::s4u::Comm::test_any()
206 - #245: migrating an actor does not migrate its execution
207 - #253: Feature Request: expose clusters as objects
208 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
209 - #255: Tesh broken on Windows
210 - #256: Modernize FindSimGrid.cmake
211 - #257: Fix (ab)use of CMake install
213 ----------------------------------------------------------------------------
215 SimGrid (3.18) Released December 24 2017
217 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
219 MSG IS NOW DEPRECATED. Please use S4U for new projects.
220 - Support for MSG should not be removed from SimGrid before 2020,
221 but future evolutions will be limited to the integration of
222 user-provided patches.
225 - s4u::onDeadlock() signal, run before stopping the simulation.
226 - s4u::Exec: asynchronous executions (abstraction of background
228 - s4u::Host->getLoad() returns the achieved speed in flops/s
231 - Link::name() is deprecated and will be removed in v3.21, use
232 Link::getCname() instead.
233 - Mailbox::getName() changed to return a std::string, use
234 Mailbox::getCname() instead to get a char*.
235 - Storage::getName() changed to return a std::string, use
236 Storage::getCname() instead to get a char*.
237 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
238 to have both allocation and deallocation of the map in user space.
241 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
242 The amount of flops remaining to do is a vector, not a scalar.
243 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
244 name implies on both sequential and parallel tasks.
245 - Both changes fix GitHub's #223 using PR #237 as a basis.
246 Thanks Michael Mercier.
247 - Most examples were converted to S4U and hidden elsewhere at tests.
248 Rationale: we still want MSG to work; we want newcomers to use S4U.
251 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
252 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
255 - Switch to the faster dlopen privatization mechanism by default
256 - Documentation improvments
259 - Live migration is getting moved to a plugin. Dirty page tracking is
260 the first part of this plugin. This imply that VM migration is now
261 only possible if one this function is called:
262 - C/MSG: MSG_vm_live_migration_plugin_init()
263 - C/C++: sg_vm_live_migration_plugin_init()
264 - Java: Msg.liveMigrationInit()
265 For C and C++, "simgrid/plugins/live_migration.h" also has to be
269 - Define class simgrid::xbt::Path to manage file names.
270 - Removed unused functions:
271 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
272 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
273 xbt_graph_export_graphviz()
274 - xbt/str.h: xbt_str_join()
275 - Remove unused datatypes:
276 - xbt/heap.h: use std::priority_queue or boost::heap instead
277 - xbt/swag.h: use boost::intrusive::list instead
280 - New link_energy plugin for the consumption of the links.
281 - All of the operations on files and storage contents have been
282 packaged into a plugin (src/plugins/file_system). The current
283 public interface can be found in
284 include/simgrid/plugins/file_system.h
285 To use these functions you now have to initialize the plugin by
286 calling MSG_storage_file_system_init() just after calling
287 MSG_init() or sg_storage_file_system_init() just after creating
291 - Remove the undocumented/untested tag <include>
294 - Remove viva specific tracing as the tool is no longer maintained
297 - #248: Kill of finished processes leads to segfault
298 - #240: xbt_cond_wait_timeout should gracefully return for C apps
299 - #239: please implement signal s4u::onDeadlock()
300 - #230: segfaults when exit() without run()
301 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
302 - #223: MSG_task_get_flops_amount() not working with parallel tasks
303 - #222: Actor::kill() doesn't really kill and segfaults
304 - #221: odd LMM warning when killing an actor
305 - #120: Memory leak when the processes are forcefully killed
307 ----------------------------------------------------------------------------
309 SimGrid (3.17) Released October 8 2017
311 The Drained Leaks release: (almost) no known leaks despite the tests.
313 Even very long simulations will preserve your memory: our testsuite
314 stresses SimGrid for over 45mn on fast machines for a coverage of
315 over 80%, and there is only one single known leak, of about 4kb.
319 - Comm.detach(): start and forget about asynchronous emission. A cleanup
320 handler may be given to free resources if the comm cannot be completed.
321 - this_actor::send(mailbox) is now mailbox->put()
322 - New: simgrid::s4u::Comm::wait_all()
323 - New: Host.execute() for remote executions.
326 - Unused option network/sender-gap is removed.
329 - Deprecate MSG_task_isend_with_matching(): This unused feature
330 really complicates our internals. Will be removed in v3.20.
333 - Improved context termination. It is now possible to free resources, even
334 when a process is forcibly killed.
335 - ContextBoost: add support for Boost versions above 1.61.
338 - Bring back run-time option --cfg=exception/cutpath to remove exception
340 - Removed unused functions:
341 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
342 xbt_str_rtrim(), xbt_str_trim().
343 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
346 - Removed header files obsolete since SimGrid 3.12:
347 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
349 - #3: SD_exit should be made optional
350 - #120: Memory leak when the processes are forcefully killed
351 - #159: Threading test regression in Actor refcounting
352 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
353 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
354 - #186: Actor::killAll() segfaults if some process is blocked on wait()
355 - #191: VM migration and pstate
356 - #192: Updating the energy consumptions of all hosts crashes with VMs
357 - #195: All actors have PID=0 in the logs
358 - #204: Sometimes segfault with thread contexts and mmap privatization
359 - #222: Actor::kill() doesn't really kill and segfaults
360 - #225: Actor::kill() doesn't really kill when victims are doing a join()
362 ----------------------------------------------------------------------------
364 SimGrid (3.16) Released June 22. 2017.
366 The Blooming Spring Release: developments are budding.
369 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
370 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
371 - Java JAR file should be finally fully working on Mac OSX too.
373 XML platforms: Switch to platform v4.1 format.
374 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
375 - <zone> can be used as a synonym for the now deprecated <as>
376 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
377 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
378 - <actor> can be used as a synonym for the now deprecated <process>
379 - state_file and avail_file periodicity is now easier to express
380 (check the documentation) the old behavior should still work.
381 * Storage: not backward compatible, but it seems that nobody ever used it
382 - Remove attribute 'content_type' of <storage_type>: was never used
383 - Make attribute 'model' of <storage_type> optional: for future usage
384 - Remove Bconnection model property: was never (in)validated. Replaced by
385 the maximum of the read and write bandwidth as a resource constraint.
388 - New and Backwards Compatibility break:
389 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
390 When one wants to get the list of tasks whose states have changed during a
391 simulation round, s/he has to allocate and free a dynar and use it as
392 argument to this function. The former SD_simulate (double how_long)
396 - Allow multicore VMs, with the correct sharing (unless you overcommit)
397 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
400 - New function to update the consumption of all hosts at once.
401 - Fix the model for multi-core hosts, linear in the amount of busy
402 cores with an abnormality for pIdle (see paper).
403 BUG: Still not satisfactory for ptask on mono-cores.
406 - The netzone are now available from the MSG API.
407 The old names still work, but are now deprecated.
410 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
411 - New option: smpi/keep-temps to not cleanup temp files
412 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
413 mallocs mode is used, allows to change the size of the fake file used
414 (default 1MB), to potentially limit the number of mappings for large runs.
415 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
416 - Fortran ifort and flang compilers support
417 - New RMA calls supported (experimental):
418 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
419 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
420 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
421 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
422 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
423 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
426 - New: this_actor::isMaestro() returns whether we are in kernel mode.
427 - Behind the scene, ActivityImpl are now automatically refcounted.
428 Removing this nasty bottleneck will greatly simplify our internals.
431 - Replay: New function xbt_replay_action_get():
432 Retrieve the function previously associated to an event type.
433 - DROPPED FUNCTION: xbt_str_varsubst()
434 - DROPPED MODULE: strbuff. We don't need it anymore.
435 - DROPPED MODULE: matrix. We don't need it anymore.
436 - DROPPED MODULE: lib. We don't need it anymore.
438 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
440 SimGrid (3.15) Released March 22. 2017
442 The Spring Release: continuous integration servers become green
443 We fixed even the transient bugs on all target architectures:
444 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
445 Windows, FreeBSD, NetBSD.
447 - Rename NetCards to NetPoints.
448 This was intended to help NS3 users, but that's not a netcard.
449 That's a point in the routing algorithm, let's avoid wrong simplifications.
452 - New: MSG_process_yield(). Stop and yield to other processes.
453 - New: MSG_process_daemon(). Daemon processes are automatically killed
454 when the last non-daemon process terminates
455 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
456 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
457 to make room for the upcoming network energy plugin.
458 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
461 - Ensure that an actor can kill itself with Process::exit()
462 - Kill the obscure NativeException. Nobody want to survive the issues
463 it denotes, so use JniException that is a RuntimeException (not to
464 be caught explicitly).
465 - Partial bug fix in initialization. SimGrid flags on command line were
466 consumed at C level but stayed in the original Java String[] args.
467 This could mess users' args[i] if SG flags were not put at the end of
469 The SimGrid flags are now removed from the Java arguments. However,
470 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
471 if args.length is greater than the number of YOUR OWN ARGUMENTS.
472 It might be if you have --log or --cfg flags in the command line.
473 - Fix numerous memleaks all around the place. In particular, around VMs.
477 - s4u::onPlatformCreated: right before the simulation starts
478 - s4u::onSimulationEnd: right after the main simulation loop
479 - s4u::onTimeAdvance: right after a clock change (time is discrete)
480 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
481 event from the availability_file changes the avail speed.
482 - Links are now usable from s4u
483 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
484 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
485 - New examples: The conversion of MSG examples to S4U has begun
486 - Actors: create, kill, migration, and suspend
487 - Applications: master-worker and token-ring
488 - Action replay: communications and storage
489 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
492 - Backwards Compatibility breaks
493 - The SD_route_get_list and SD_route_get_size functions have been removed.
494 They are replaced by the unique
495 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
496 The route from src to dst is built in the links dynar whose size can
497 be obtained with xbt_dynar_length.
498 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
499 sg_host_route_bandwidth and sg_host_route_latency.
500 Macros ensure the backwards compatibility, but you should fix your code
503 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
504 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
505 - Bump our claim of support from MPI 1.1 to MPI 2.2.
506 We don't support 100% of it, but it should be enough. Contact us if not.
507 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
508 naive and probably inaccurate)
509 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
510 mapped onto a unique small file using some system magic.
511 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
512 - Fortran: cleanups, fixes, support of user-added operations
513 - RMA: MPI_Accumulate are applied in correct order.
514 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
515 - Better support for MPI_IN_PLACE
516 - Support for MPI_Win attrs and keyvals.
517 - Support MPI_Comm_group_create, MPI_Type_size_x
518 - MPI ops should be properly applied to all allowed types
519 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
520 algorithm by a non failing one from mpich.
523 - Kill the fifo data container: we don't use it anymore.
526 - A <cluster> can now be created with different speed values to represent pstates.
527 see examples/platforms/energy_cluster.xml for an example.
529 SimGrid (3.14.159) Released December 28. 2016
531 The previous release was only a crude approximation of a Pi release;
532 this one should be a bit better.
534 - Revert a last minute change that broke on Mac OSX.
535 - Fix the changelog and NEWS files.
536 - Improve the documentation, in particular of the routing module.
537 - Integrate some patches flying around in Debian.
539 SimGrid (3.14) Released December 24. 2016
542 * The whole documentation was reworked and reorganized. There is
543 still much room for improvement here, but we're on it.
546 * We now need python3 for our testing infrastructure.
547 * Model checking needs libevent
550 * Now works on FreeBSD
553 * Privatization now works on FreeBSD
555 * Privatization is now activated by default in smpirun
556 You can turn it off with -no-privatize if something goes wrong.
558 * Call-location tracing for SMPI.
559 You can add the exact location (filename / linenumber) of an MPI call to
560 your trace files and slow down or speed up the simulation between two
561 consecutive calls by using an adjustment file (see the documentation).
563 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
564 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
565 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
566 values were ignored in some cases. The timings of these functions can now
567 be significantly different.
569 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
570 smpi/running-power is renamed to smpi/host-speed
572 * smpi/grow-injected-times option to enable or disable multiplication of the
573 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
574 make simulation less precise (but also much faster).
576 * smpirun script should be (much) faster for large deployments.
578 * SMPI tracing: fixed issue with poor matching of send/receives.
580 * Replay: Fix broken waitall
582 New functions and features
583 * MSG_parallel_task_execute_with_timeout, to timeout computations.
585 Dropped / renamed functions and features
586 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
587 - MSG_mailbox_is_empty() -> Mailbox::empty()
588 - MSG_mailbox_front() -> Mailbox::front()
589 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
590 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
591 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
592 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
593 * Don't pass the free_f parameter to property related functions:
594 - MSG_host_set_property_value()
595 - MSG_as_router_set_property_value()
596 - MSG_storage_set_property_value()
597 * VM properties. Since msg_vm_t are msg_host_t, just use
598 MSG_host_get_property_value() and friends
599 * VM I/O related things:
600 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
601 - Unimplemented save/restore methods
602 * MSG_as_router_get_property_value() was redundent with
603 MSG_environment_as_get_property_value().
604 - Removed MSG_as_router_*propert*() functions
605 - Added MSG_environment_as_set_property_value() for consistency
606 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
607 Well, they are still there for now, but deprecated with a warning.
608 Please switch to xbt_dict_new_homogeneous() before this is removed
610 * Task affinity. Its intended behavior (that was very badly tested
611 and probably not really working) was deceiving what most users
612 would have hoped here.
613 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
614 * Option network/coordinates is now useless and should be dropped.
617 * Added option storage/max_file_descriptors to allow more than 1024 files opened
620 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
621 * Added Dragonfly topology support
624 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
625 EiBps, EBps, Eibps, Ebps for bandwidth.
626 They may become useful to some lucky ones.
629 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
630 * ex/app_tokenring: new example, very similar to the MSG Token Ring
631 * ex/async_waitAll: new example, on asynchronous communications
634 * Memory usage should be decreased for simulations with a large number
635 of processes. This also helps for SMPI.
637 SimGrid (3.13) stable; urgency=low
639 The Half Release, a.k.a. the Zealous Easter Trim.
641 - We removed half of the lines, that were mostly experimental cruft.
642 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
643 (not counting blanks and comments -- according to openhub.net)
644 - The internals are now compiled in C++ (and will soon be clean C++)
645 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
646 59 klines of C, etc. We added only 29 klines of C++ in replacement.
648 * Backwards Compatibility breaks
649 - Removed Lua simulation bindings (switch to C or Java for that).
650 Lua can still be used to describe platforms
651 - Removed Java kernel plug-ins.
652 Will be reintroduced after the ongoing major internals reorg.
654 - the following functions were removed.
655 They were too specific and should be reimplemented in a generic
656 way, with filter function.
657 - MSG_task_listen_from_host
658 - MSG_mailbox_get_count_host_waiting_tasks
659 - MSG_mailbox_put_with_timeout was removed.
660 Please use MSG_task_send_with_timeout instead.
662 - the SD_application_reinit function was removed. It has been a noop for a while.
663 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
664 replaced by a more flexible mechanism.
665 - The following functions thus do not exist anymore
666 - SD_workstation_get_access_mode
667 - SD_workstation_set_access_mode
668 - SD_workstation_get_current_task
669 - Basic estimation functions have been removed but can easily be replaced
670 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
671 - SD_workstation_get_computation_time => amount / sg_host_speed()
673 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
674 Use VM.getSpeed()*load/100 for the legacy behavior.
676 - option enable_tracing was removed. It was not doing anything for a while.
677 - In the ModelChecker:
678 - the model-checker now ptraces the model-checked process which means
679 you cannot use a debugger on the latter anymore (we might make this
680 optional in the feature);
681 - removed soft-dirty page tracking;
682 - remove model-checked side snapshot management,
683 MC_snapshot() and MC_compare_snapshot();
684 - keep the MC_cut() function as a stub (it was not really working
685 in the previous release).
688 * All options are consistently in kebab-case. Old names are kept as alias.
691 * Switch to platform v4 format.
692 - Rename from 'power' to 'speed' the attributes describing the amount of
693 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
694 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
695 - In <host> and <link>, attributes availability and state are gone.
696 It was redundent with state and availability traces, and with peak values.
697 - In <cluster>, attributes availability_file and state_file are gone.
698 It was too complex and unused.
699 - Kill <gpu>. Was not doing anything.
700 - The DOCTYPE points to the right URL:
701 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
702 (the file at this address now documents the changelog since its v1)
703 - A warning is emitted for unit-less values (they are still accepted).
704 - speed. Default: 'f' or 'flops'. Also defined:
705 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
706 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
707 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
708 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
709 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
710 - latency. Default: 's' second. Also defined:
711 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
713 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
714 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
717 * s4u::Host is now the preferred public interface to the Host features.
718 sg_host_* functions are C bindings to the exact same behavior
719 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
722 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
723 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
724 They were deprecated since a few years, and probably did not even compile anymore.
727 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
728 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
729 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
730 encouraged to update their codes. A list of the modified functions can be found at the end of
731 include/simgrid/simdag.h
734 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
735 the lambda or closure passed as a parameter will run in kernel mode.
736 All the callback functions should be rewritten to that interface at some point.
739 * Reorganizing and cleaning the internals all around the place.
742 * Remove old default barrier/bcast buggy algorithms (see #18407)
743 * Various bug fixes to handle more codes
744 * Remove the need for the --foreground option of smpirun (it is still
745 accepted for backward compatibility).
748 * Kill the setset data container: MC don't use it anymore.
749 * Kill the queue data container: it made more sense with GRAS.
750 * Kill the xbt_peer_t data type: it's useless without GRAS.
751 * Kill rm_cb feature of config sets: it was never useful.
752 * Kill graphxml parsing feature. It was not used.
753 * Kill the deprecated code protected by XBT_USE_DEPRECATED
755 - xbt_dynar_sort_strings(), when the content is char*
756 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
757 They throw exceptions on invalid input;
758 * C++ support for declaring CLI flags (simgrid::config::Flag);
759 * class for abstracting different signal backends (simgrid::xbt::signal).
760 with no external dependencies (we need very simple signals).
763 * refactoring and cleanup of the code;
764 * ongoing process to cleanly separate the model-checking algorithms
765 from the code model-checking support.
767 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
769 SimGrid (3.12) stable; urgency=low
771 The Facelift Release.
774 * Require g++ v4.7 at least to not speak prehistorical C++.
775 * Require Boost 1.48 (for signal2 component).
776 * Java must be version 7 at least when activated.
777 * Builds on Windows again (including Java bindings).
778 * Tracing is now always enabled (no way to turn it out)
779 * Remove GTNetS. It was not working anyway.
780 * Various cleanups in the cmake scripts.
781 * Move headers around to sort them out on installed systems:
782 - instr/instr.h -> simgrid/instr.h
783 - instr/jedule/* -> simgrid/jedule
784 - simdag/datatypes.h was removed
785 - simdag/simdag.h -> simgrid/simdag.h
786 - msg/datatypes.h was removed
787 - msg/msg.h -> simgrid/msg.h
790 * Interface improvement:
791 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
792 - Sanitize the interface in MSG_task_ module:
793 - Merge two functions that were close enough but misleading:
794 set_compute_duration(t) -> set_flops_amount(t)
795 get_remaining_computation(t) -> get_flops_amount(t)
796 - set_data_size(t) -> set_bytes_amount(t)
797 get_data_size(t) -> get_bytes_amount(t)
798 - Massive cleanups in the functions related to the energy
799 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
800 - New: MSG_host_get_pstate()
802 - msg/energy/onoff: switching hosts on and off
804 * Interface improvement:
805 - Rename Host.isAvail() to Host.isOn()
806 - Rename Process.currentProcess() to Process.getCurrentProcess()
807 - Rename Task.setDataSize() to Task.setBytesAmount()
808 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
810 - #18874: Actually allows the GC to reclaim tasks
814 - SIMIX_process_throw: raises an exception in a remote process
815 * Refactoring: Separate sub-modules
816 - libsmx: the public interface, as libc in a real system
817 - popping: the strange dance that converts a user request into a kernel handling
818 - smx_context_*: the virtualization mechanisms that embed the user code
819 - smx_*: the handling of each simcalls
821 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
822 * Rename smx_action_t into smx_synchro_t, making explicit that these
823 things are used to synchronize processes with their environment.
824 For example, a communication is a sort of synchronization involving
825 the communicating processes (that may block until the exchange) and
826 the platform. The same can be said from computations, etc.
828 - #18888: segfault when a process ends before its kill_time
832 - 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.
833 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
834 - MPI_Type_set_name, MPI_Type_get_name
835 - MPI_*_c2f and MPI_*_f2c functions
836 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
837 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
838 - Activate a lot of new tests from the mpich 3 testsuite
840 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
841 - InfiniBand network model added: Based on the works of Jerome Vienne
842 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
843 - When smpi/display_timing is set, also display global simulation time and application times
844 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
845 * Collective communications
846 - 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
847 - 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/).
848 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
850 - Replay now uses algorithms from wanted collective selector
851 - Replay can be used with SMP-aware algorithms
852 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
853 - 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.
855 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
856 - Correctly use loopback on fat-tree clusters
857 - Asynchronous small messages shouldn't trigger deadlocks anymore
858 * Energy/DVFS cleanup and improvement
859 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
860 - new: smpi_get_host_pstate()
863 - "Full" network optimization flag was broken since Surf++
864 - Better handling of precision flags in maxmin
865 - Fix bug causing sometimes "Impossible" errors
866 - Properly pass cluster properties to included hosts
867 * Improvement of the Energy plugin.
868 - Always update the consumption before returning that value
869 - New property: watt_off to denote the dissipation when the host is off
870 - New functions getWattMinAt and getWattMaxAt to retrieve the
871 dissipation of pstates that we are not currently at.
872 * Java: class NetworkLink renamed to Link
873 * New function: simcall_process_get_kill_time()
874 * Massive rename s/workstation/host/
875 - That's intrusive, but that's good for the project consistency. Sorry.
876 - Change config option "workstation/model" into "host/model"
879 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
880 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
881 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
882 - xbt_dirname and xbt_basename for non-POSIX systems
884 * The model checker now runs as a separate process.
885 * The model checker runs must now be launched with the new simgrid-mc program.
886 * Record/Replay: the MC can display a textual representation of a path in the
887 execution graph. It can then be replayed outside of the model checker.
889 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
891 SimGrid (3.11) stable; urgency=low
896 * Normalizing pointers addresses tool for better diff between logs
899 * Add cloud examples using new VMs
900 - examples/msg/cloud/two_tasks_vm.tesh
901 - examples/msg/cloud/simple_vm.tesh
902 - examples/java/cloud/cloud.tesh
903 - examples/java/cloud/migration/migration.tesh
904 * Add java surf examples:
905 - examples/java/surfPlugin/surf_plugin.tesh
906 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
907 - examples/java/surfCpuModel/surf_cpu_model.tesh
908 * Add SMPI+MSG example:
909 - examples/smpi/smpi_msg_masterslave/
914 - msg task destroy cancel test
915 - msg_host on/off test
916 * Move all tests in testsuite to teshsuite (adding tesh files)
917 * Restructure teshsuites
918 - one folder for each kind of test
919 * Restructure AddTests.cmake
921 - structure the order of tests (with sections)
924 * Add virtual machine
925 - creation of a VM on a PM
926 - migration of a VM from a PM to another PM
928 - MSG_process_join(msg_process_t process, double timeout)
929 - msg_bar_t MSG_barrier_init(unsigned int count)
930 - int MSG_barrier_wait(msg_bar_t barrier)
931 - void MSG_barrier_destroy(msg_bar_t barrier)
932 - msg_as_t MSG_environment_as_get_by_name(const char * name)
933 * New option "msg/debug_multiple_use" to help debugging when a task is used
936 - Improvements and finalization of MSG_storage, MSG_file APIs and their
938 - Increase code coverage in test suites
942 * Protect context stacks against stack overflow. The number of protected memory
943 pages allocated on the top of each stack (1 by default) can be configured
944 with the new command line option --cfg=contexts/guard_size:<value>.
945 * Simcalls are now generated by a python script that
946 - generates files included by SimGrid
947 - checks that all the functions exist, and proposes prototypes
949 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
950 host_set_data, host_get_data
952 - simcall_process_join(smx_process_t process, double timeout)
953 * Fix bug where sleeping processing could not be suspended.
956 * Translate surf models from C to C++
957 - Generic classes for all models: Model, Resource, Action
958 - A generic interface for each kind of model (CPU, Network, Storage
959 Workstation, WorkstationVM)
961 * Translate surf routings from C to C++
962 * Add callbacks using sigc++ or boost::signals2
963 - Add callback functions for resource creation/destruction
964 - Add callback functions for action state change
965 - Handle Energy as a plugin
966 * Replace swag by boost::intrusive
967 * Add new routing models for clusters. For documentation, see
968 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
969 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
970 parameters for cluster tag
971 - Fat trees, with topology="FAT_TREE" and
972 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
973 - see examples/platforms/torus_cluster.xml and
974 examples/platforms/fat_tree_cluster.xml
978 * Hostfiles support host:nb_processes construct to deploy several processes on
980 * Collective communication algorithms should not crash if used with
981 improper number of nodes and report the error.
982 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
983 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
984 MPI_Dims_create, MPI_Cart_sub are supported.
985 * New interface to use SMPI programmatically (still depends on MSG for
986 some parts, see examples/smpi/smpi_msg_masterslave):
987 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
991 * Global variables privatization in MPI executables is now performed at runtime
992 with the option smpi/privatize_global_variables (default:no).
993 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
994 dynamic libraries loaded by the application are not privatized (static
995 linking with these libraries is advised in this case)
998 * Options defined in XML work correctly now.
1001 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1002 copied into simgrid.jar or not (ON by default). Use this option if you want
1003 to reduce the size of the installed simgrid.jar, *and* the native libraries
1004 are kept installed elsewhere.
1005 * Surf binding with SWIG (code generated in maintainer mode only):
1006 - plugin to handle callbacks
1007 - CPU model only for the moment
1010 * Supernovae build mode is definitively removed. It was used to improve
1011 inlining and inter-module optimizations. It is nowadays superseded by
1012 link-time optimizations commonly available in compilers.
1013 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1014 * Add boost dependency for surf++
1015 * Add new macro for tests
1016 - ADD_TESH(name <tesh_args>)
1017 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1021 - xbt_bar_t XBT_barrier_init(unsigned int count)
1022 - int XBT_barrier_wait(xbt_bar_t barrier)
1023 - void XBT_barrier_destroy(xbt_bar_t barrier)
1024 * Make the xbt_os_time module public
1026 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1028 SimGrid (3.10) stable; urgency=low
1030 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1033 * Reintegrate Java to the main archive as desynchronizing these
1034 package is not acceptable anymore (Java is now considered stable)
1035 * Add explicit synchronization facilities through semaphores
1036 * Bug fix: Task.setDataSize() only changed the C world, not the value
1037 cached in the Java world
1040 * Dramatically change the way files are handled. API and internals changed, but
1041 this part of MSG was not considered as production grade either.
1042 * Add explicit synchronization facilities through semaphores
1043 * Add a new function MSG_host_get_process_list()
1044 * Preliminary DVFS support (see examples/msg/energy/ for details)
1047 * SMPI is now included directly in the libsimgrid as the windows
1048 linker doesn't force us on splitting it anymore.
1049 * Improvements of the SMPI replay tool:
1050 - Most of the collective communications are now rooted in the same process as
1051 in the original application.
1052 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1053 used until now). Multiple data types can now be used in a trace.
1054 - The replay tool now supports traces produce either by TAU or a modified
1056 - Bug Fix: the compute part of the reduce action is now taken into account.
1057 - Gatherv collective is now supported
1058 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1059 of smpirun outputs time independent traces for the current run. One file
1060 is created per process. If too many processes are simulated, this behavior
1061 can be changed to one file for all processes by using the
1062 tracing/smpi/format/ti_one_file flag
1063 * smpirun generates the host file if needed (with given host count and platform)
1064 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1065 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1066 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1067 --cfg=smpi/coll_selector:(mpich/ompi)
1068 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1070 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1071 happening inside a collective SMPI call.
1072 * Fix the behavior of complex data types handling.
1073 * Make MPI_Wtime another synchronization point to take computations into
1075 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1076 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1077 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1078 (no privatization of global variables yet)
1079 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1082 * Verification of liveness properties is now available for SMPI applications
1083 (in addition to MSG applications)
1084 * Bugged examples using SMPI in examples/smpi/mc/
1085 * Add --cfg=model-check/visited option. Allows the verification of infinite
1086 programs. Detection of loops in the execution thanks to the system state
1087 comparison and reduction of the state space to explore. Can be combined with
1088 DPOR for safety properties.
1091 * Allow to change SimGrid configuration (see --help) within the code
1092 thanks to SD_config() as it can be done in MSG.
1093 * Add a new function SD_task_set_amount() upon user request.
1096 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1097 * Remove rule based routing (no more PCRE dependency)
1098 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1099 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1100 * Add a loopback_bw and loopback_lat options to cluster tag.
1101 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1102 routing. Now peers are encapsulated in an AS and have their own private
1103 router but this is transparent.
1106 * Our own implementation of getline is renamed xbt_getline, and gets
1107 used even if the OS provide a getline(). This should reduce the
1108 configuration complexity by using the same code on all platforms.
1109 * New type: xbt_cfg_elm_boolean.
1110 * Allow to use yes/no for boolean configuration options in the command line.
1111 * Allow to disable SimGrid cleanups at exit from command line option.
1112 There are situations where one may want a simulation to end with an exit.
1113 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1114 annoying when scripting around the simulator. Adding a
1115 --cfg=clean_atexit:no allows to circumvent this issue.
1118 * Lots of memory leaks were corrected in this release.
1119 There is no known memory leaks anymore, in all of our 600+ tests.
1120 * New command line option --version, to get SimGrid version information.
1121 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1123 * Supernovae builds are deprecated, and expected to be removed in the next
1126 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1128 SimGrid (3.9) stable; urgency=low
1130 The Grasgory release: GRAS is really dead now.
1132 * Complete overhaul of the internal host structures scheme.
1135 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1136 was considered as experimental and badly maintained since then.
1137 * Keeping it was thus a trap to our potential users, that could take
1138 it instead of MSG or SMPI by mistake despite is pity state.
1139 * GRAS seems to have very few users (if any), and no one volunteered
1140 to maintain it further. It also induces a lot of XBT code (for
1141 portability sake), that must be maintained too.
1142 * For all these reasons, we killed GRAS. If someone wants to revive it
1143 in the future, don't cry, our git history still remembers of GRAS.
1146 * Major overhaul. Merge our documentation again as time proved that
1147 splitting it was really not helping our users.
1148 * Further improve the developer documentation to help newcomers
1149 hacking on SimGrid itself. The user documentation (and in
1150 particular, the beginner documentation) is still in a sorry state.
1153 * Now works on Windows too!
1154 * Much more extensive test suite, from MPICH
1157 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1158 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1159 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1160 * Bug fix: let task be scheduled when the last dependency to be solved is
1161 a control dependency.
1162 * Remove SD_load_environment_script function.
1163 Use the C sg_platf function if you want to declare a platform
1167 * New function: MSG_process_get_number()
1168 * Old function documented: MSG_config()
1169 * Remove MSG_load_platform_script function
1170 Use the C sg_platf function if you want to declare a platform
1174 * Change the default value of the TCP_gamma constant (maximal size of TCP
1175 congestion window) to a more realistic 4MiB value. If you notice changes in
1176 your simulation results, you can fall back to the previous 20k tiny window
1177 by adding --cfg=network/TCP_gamma:20000 on command line.
1178 * (Hopefully) fix a bug wrt periodic availability and state traces
1179 * Bug fix: use default values at start when first event in availability/state
1180 trace is not at time 0.
1183 * remove the "new_" part of function name sg_platf_new_trace_connect
1184 (resulting in sg_platf_trace_connect), since it does not create
1188 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1189 use case where it's really mandatory, and maintaining it was a pain in
1191 * New: xbt_fifo_search(), search an item with a user-provided
1192 comparison function instead of dumb pointer comparison.
1195 * Fix the lua deployment:
1196 Use `simgrid.init_application()` before deployment instead of
1197 `simgrid.msg_register_application()` after.
1200 * Transfer the tracing files into the corresponding modules.
1202 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1204 SimGrid (3.8.1) stable; urgency=low
1206 The "we are told that some people want to also *install* the simgrid
1209 * Add missing file "tesh.1" to the archive.
1211 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1213 SimGrid (3.8) stable; urgency=low
1215 The Psssshiiiit release: SimGrid jumps into the Cloud.
1218 * Add an experimental interface to manipulate VMs. They are mainly
1219 process groups with very few intrinsic semantic, but they should
1220 allow you to build the semantic you want easily.
1221 * New function: MSG_host_set_property_value()
1222 * New function: MSG_process_on_exit(). To clean memory in all cases.
1223 * Bug fixes that made the host (and link) failures unusable.
1224 * Add a way to auto-restart process when the host in which they are
1225 executing comes back (ON_FAILURE="RESTART" on deployment file,
1226 MSG_process_auto_restart_set).
1227 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1228 please stop using the old ones, they are DEPRECATED.
1230 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1231 Please use MSG_init() instead. (reducing the amount of entry
1232 points in the library helps us).
1233 * Make it impossible to link against the wrong version of the lib
1234 * Deprecate MSG_clean(). No need to call it anymore.
1235 * Function MSG_get_host_number() is not deprecated anymore.
1238 * Split the doc into a user guide and a reference guide.
1239 * Start a developper guide to help people hacking on SimGrid.
1242 * Enable tracing by default. This modules rocks you should use it.
1243 * Remove option custom_flags. Now use environment variables CFLAGS
1245 * Use default cmake things to detect lua instead of home grown ones.
1246 * New option "enable_mallocators" to disable mallocators, for debugging
1247 purpose ("on" by default).
1250 * Bug fixes around the resource failures: don't let the processes
1251 survive the host they are running onto.
1252 * Add an interface to auto-restart processes when the host in which they are
1253 executing comes back.
1254 * Ensures that SIMIX_clean is called automatically. It's not part of
1255 the public interface anymore (bindings should be updated).
1258 * Bug fix for when SD_Simulate is called with a positive value: be careful
1259 when comparing doubles. Sometimes they are different for non significant
1261 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1262 parallel task whose initial work is distributed among host according
1263 to the Amdahl's law. Such tasks are created with a parameter alpha
1264 that corresponds to the non-parallelizable part of the computation.
1265 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1266 between two sets of workstations assuming a 1D block distribution (each
1267 workstation owns a similar share of data) on both sides.
1269 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1270 Data redistribution will be automatically scheduled once parent and child
1271 are both scheduled. The filling of computation_amount and
1272 communication_amount structures is now done seamlessly thanks to the chosen
1274 * New function SD_workstation_dump to display various information
1275 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1276 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1277 bandwidth on the route onto which the task is scheduled and the amount of
1279 To divide the nominal bandwidth by 2, the rate then has to be:
1280 rate = bandwidth/(2*amount)
1281 * Compute tasks that have failed can now be rescheduled and executed again
1282 (from their beginning)
1283 * Increasing source code coverage (src/simdag is now covered at 95.8%
1287 * Re-implement time-independent trace replay using SMPI (at the
1288 smpi_smp_* level) instead of MSG. This should replace
1289 examples/msg/actions/actions.c
1290 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1291 hindexed and structs)
1292 * Implement the exchange of non-contiguous data.
1293 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1294 * Correct behavior of smpi/sender_gap and set its default value to 0
1295 * Add option to asynchronously send small messages to allow better
1296 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1297 specifies the size in bytes under which messages will be asynchronously sent.
1298 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1299 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1302 * New C interface to define a platform: XML is now optional.
1303 For more info, please check include/simgrid/platf.h.
1304 * New interface to define random platforms from the C:
1305 For more info, please check include/simgrid/platf_generator.h and
1306 examples/msg/masterslave/masterslave_platfgen.c
1307 * Export a sg_cmdline dynar containing all the arguments we got from
1311 * Two new tracing options for adding comments to trace file so you
1312 can track your experiments (see --help-tracing for details).
1313 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1314 * Adding the SimGrid version that generated the trace file as a comment.
1315 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1316 * Fix to avoid key clashes on Paje links
1317 * Other minor fixes related to the Paje specification
1320 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1321 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1322 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1323 * Greatly improve the robustness of mmalloc to user errors (such as
1324 using an area after freeing it, or freeing it twice)
1326 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1328 SimGrid-java (3.8.1) stable; urgency=low
1330 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1331 Warning: all methods are visible, but only some of them are
1332 implemented so far. Check the source (src/jtrace.c)
1333 for further information.
1334 * New module: org.simgrid.msg.File (SimGrid File management functions)
1335 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1336 * Change the meaning of Process.restart: now restart the process from
1337 the begining, like MSG_process_restart in C.
1338 * Add Process.setAutoRestart: handling of process restart when failed
1340 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1341 you to retrieve the properties of the processes/hosts
1342 * Deprecate Msg.clean(): you can just forget about it now.
1343 * New function Process.getCount(), that only works when compiling
1344 with the not yet released version 3.9 of the C library.
1346 * New context factory based on Coroutines. It mandates a modified JVM
1347 but then, the simulations run about five times faster, and there is
1348 no limit to the amount of processes (beside of the available memory).
1350 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1352 SimGrid (3.7.1) stable; urgency=low
1355 * Restore the prototype of MSG_process_create_with_environment() to
1356 the pre-3.7 situation by removing the kill_time argument.
1357 * Add a MSG_process_set_kill_time() function instead.
1360 * Fix weird behaviors when dealing with parallel tasks.
1363 * Simgrid is now built as a dll.
1364 * Simgrid-java now works on Windows.
1365 * Simgrid-Java is now included into Windows package.
1368 * First pre-build package for MacOSX.
1371 * Fix compilation when using MSG_USE_DEPRECATED.
1372 * Fix some compilation issues on Macs and Windows.
1373 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1374 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1376 We discovered that the Lua console is broken, but we are missing the
1377 manpower to fix it right now. The problem existed in 3.7 too, so we
1378 are not blocking the release for that. Sorry if you depended on this
1379 feature, any help would be really welcome.
1381 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1384 SimGrid-java (3.7.1) stable; urgency=low
1386 The "Java aint got to be bloated and slow" release
1389 * Various internal cleanups and performance improvement
1390 Simulations are expected to run up to twice faster or so
1391 * Make Process.kill(process) an instance method, not a static one
1392 * User processes are not java.lang.Thread subclasses.
1393 This breaks the compatibility (sorry), but previous API was
1394 brain-dead, making it impossible to have non-trivial
1395 initializations in the process constructor.
1396 * Require a full constructor per Process sub-class.
1397 Kinda breaks the compatibility (sorry), but this allows a much more
1398 efficient way to launch the processes at simulation startup.
1399 * Do not embeed our version of semaphores, java 1.5 can be considered
1400 as sufficiently prevalent for us to not dupplicate its features.
1404 * Add examples for almost every part of the API
1405 We spotted and fixed a lot of bugs in the process
1407 * New module: asynchronous communication API
1408 * New function: Process.sleep()
1409 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1410 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1411 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1413 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1415 SimGrid (3.7) stable; urgency=low
1417 The "spring cleanups (before the next Big Project kicks in)" release.
1420 * We can specify the SMPI latency/bandwidth factor with command line
1421 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1422 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1423 You can also use the "config tag" from platform file by adding this line
1424 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1425 (see "example/platforms/tag_config.xml" to use "config tag").
1426 Note that the command line supersedes the platform file configuration.
1427 * Change the correction factors used in LMM model, according to
1428 the latest experiments described in INRIA RR-7821.
1429 Accuracy should be improved this way.
1430 * Use the partial invalidation optimization by default for the
1431 network too. Should produce the exact same results, only faster.
1432 * Major cleanup in surf to merge models and split some optimization
1433 mechanisms from the core of the models. As a result you can now
1434 specify which model to use (e.g., --cfg=network/model:LV08
1435 --cfg=cpu/model:Cas01) and which optimization mode to use
1436 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1437 Incompatible combinations should err at initialization. See
1438 --help-models for the list of all models and optimization modes.
1439 * The CLM03 workstation model was dropped for simplicity because it
1440 used the deprecated CM02 network model. Use default instead.
1441 * Rename the TCP_gamma configuration option to network/TCP_gamma
1442 * Rename the coordinates configuration option to
1443 network/coordinates, and document it
1444 * Use now crosstraffic keyword instead of the terribly misleading
1445 fullduplex keyword. It is activated by default now in the current
1446 default model, use --cfg=network/crosstraffic:0 to turn it off.
1447 * Ongoing refactoring the model parsing to make XML files optional
1448 See include/simgrid/platf.h for details (still to be completed)
1451 * Major overhaul of the documentation. Almost instructive now :/
1452 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1453 functions and friends. This interface was considered as
1454 deprecated since over 2 years, it's time to inform our users that it is.
1455 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1456 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1457 use these (crappy) functions in your code.
1458 These functions will be removed soon. Stop using them now.
1459 * Deprecate MSG_get_host_{table,number}
1460 Implement MSG_hosts_as_dynar() instead.
1461 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1462 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1465 * Stabilize the parallel execution mode of user contexts
1466 * Introduce configuration variables to control parallel execution:
1467 - contexts/synchro: Synchronization mode to use when running
1468 contexts in parallel (either futex, posix or busy_wait)
1469 - contexts/parallel_threshold: Minimal number of user contexts
1470 that must be part of a scheduling round to switch to parallel
1471 execution mode (raw contexts only)
1472 * Fix bugs that prevented to use suspend/resume along with
1473 synchronization structures.
1474 * Fix bugs in process termination that lead to invalid memory access
1475 in very specific conditions.
1478 * Introduce a parallel mode for the models (controlled by surf/nthreads
1479 configuration item). In our tests, running the models in parallel
1480 never lead to any speedups because they are so fast that the gain
1481 of computing each model in parallel does not amortizes the
1482 synchronization costs, even when ultra fast futexes are used.
1483 This is released anyway because YMMV.
1486 * Performance boost by using a swag internally to compute the set of
1487 tasks that are finished and should constitute the return value of
1491 * Enable it by default now that it is considered rather stable.
1494 * Documentation of the tracing functions.
1495 * Performance gains when tracing categorized/uncategorized resource
1496 utilization by avoiding calls to get route when updating resource
1497 variables. LMM constraints are being used instead.
1498 * API changed to set task categories. Use MSG_task_set_category instead
1499 of TRACE_msg_set_task_category, and SD_task_set_category instead
1500 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1501 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1502 * Examples for link user variables added (see at examples/msg/tracing/)
1503 * Deprecated function TRACE_msg_set_process_category completely removed
1504 * Trace header updated according to the latest Paje file format
1505 * Tracing network lazy updates, no longer obligate users to use full updates
1506 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1507 * Experimental: let user code declare/set/push/pop application states for hosts
1508 * API changed to allow the manual creation of graph configuration files
1509 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1512 * Improve the API of Lua MSG bindings, using the Lua spirit.
1513 * Each simulated process now lives in its own Lua world (globals are
1514 automatically duplicated). It helps writing simulators. Will allow
1515 to run Splay programs within SimGrid in the future.
1516 * Add a Chord example in Lua, equivalent to the MSG one.
1519 * Start the implementation of a solution to express temporal
1520 properties, not only local assertions. This is still an
1521 experimental work in progress, stay clear from it to be safe.
1525 - Add new runtime parameters --help-logs and --help-log-categories
1526 to display information about supported logging parameters and
1528 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1529 don't exists anymore.
1530 * Mallocators: allow value NULL for the reset function.
1532 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1533 create homogeneous dictionaries, where all the elements share the
1534 same free function. Non homogeneous dictionaries will be
1535 deprecated in the next release.
1536 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1537 - Multi-level dictionaries are deprecated.
1539 - new function xbt_dynar_search_or_negative() that is useful when
1540 you have less than 2 million elements in your dynar and don't
1541 want of the extra complexity of catching exceptions when the
1542 element is not found.
1544 - Make xbt_os_thread module (for thread portability) public.
1545 Documentation is still to come, sorry.
1547 - Cleanups and simplifications to make it maintainable again.
1548 - Exotic features (such as memalign and valloc) were removed.
1549 - The metadata were extended and improved so that the
1550 model-checker becomes able to explore and inspect the heaps.
1551 - This may induce a performance drop when enable_model-checking is
1552 ON in cmake (even if it's not used in the simulation), but it is
1553 necessary at this point to get MC working.
1555 Turn model-checking OFF if simulation performance matters to you.
1556 Not enabling it at runtime is not enough, disable it in cmake.
1558 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1561 SimGrid (3.6.2) stable; urgency=low
1563 The "Not coding new stuff allows to polish old things" release.
1566 * New bindings to the NS3 packet level simulator (experimental)
1567 * Use the raw (efficient) execution contexts instead of the sysv
1568 (portable) ones when possible.
1569 * libpcre is now mandatory in any cases since not using it led to
1570 severe performance loss and possibly other issues
1571 * Update the XML platforms:
1572 - G5K: include the latest machine in Nancy
1573 - GridPP and LCG: new platforms
1574 * Documentation was partially updated, at least (more to come)
1576 Bug fixes, cosmetics and small improvements
1577 * Free terminated processes before the end of the simulation to avoid
1578 exhausting the memory of users having very dynamic amount of
1580 * Bug fix and cosmetics about canceling non-running tasks
1581 * Bug fix about the dot loader's issues when using libcgraph
1584 * Create an installer for windows with nsis (amd64 and win32)
1585 - Add an hello world project to illustrate simgrid project creation.
1586 - Embed libpcre into the Simgrid installer to avoid
1587 its compilation burden
1588 * The raw execution contexts should work on Apple now
1589 * Port to Windows 64 bits
1590 - Sysv contexts now have an implementation for this arch
1591 - GRAS communication features now support this arch
1592 * Drop support for borland compiler on windows
1593 - this code was not maintained, and we kinda depend on gcc nowadays
1594 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1595 * Fix portability issue on unstable ubuntu: linker became picky on
1598 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1600 SimGrid (3.6.1) stable; urgency=low
1602 The "Oops, we broke Macs too" release
1605 * Fixed contexts detection so that raw ones are used when possible
1606 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1607 leads to a strange error, with user code segfaulting sometimes when
1608 the generated code is not perfectly aligned (which is not
1609 controllable from the user side, depends on the amount of code)
1612 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1614 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1616 SimGrid-java (3.6) unstable; urgency=low
1619 * Split of every thing from simgrid v3.5 into a separate package.
1621 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1623 SimGrid (3.6) stable; urgency=medium
1625 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1628 * Bindings now constitute their own package, separated from the main one.
1629 Rationale: reduce our maintenance nightmare by reducing the module coupling
1630 They will soon be released on their own on gforge.
1632 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1633 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1635 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1636 * It's not quite deprecated for now because we have no replacement,
1637 but it may soon become the case.
1640 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1641 * Fortran: New user-level cache variable to store the rank of the running
1642 process. This improves performance by an order of magnitude.
1643 * C: New coccinelle script to automatically locate and modify global and
1644 local static variables.
1645 * Improved SMPI network model with a sender-side gap to account for multiple
1649 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1650 only say if a communication is finished, no matter whether it succeeded or
1651 failed. You can call MSG_comm_get_status() to know the status of a finished
1653 * New function MSG_task_dsend() to send a task and detach it. When a
1654 communication is detached, you are never notified of its success or failure
1655 and the memory is released automatically once it is finished. This function
1656 is useful when you don't care about the end nor the success of a
1658 * Change the prototypes of action replay. Sorry for inconvenience,
1659 but this is really more efficient this way (and to adapt your code,
1660 you just have to fix the initialization, that shouldn't be too long)
1661 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1662 ever managed to do anything useful with it.
1663 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1664 takes a dynar of communications. It returns immediately and gives the
1665 index of a finished communication (if any).
1666 * New example: a basic implementation of the Chord P2P algorithm.
1669 * New model for multi-core CPUs. You can now use the core attribute to
1670 precise the number of cores of a host. This is a basic model. Every
1671 process running on the host receives at most the power provided in
1672 the DTD (throughput<=power). Total throughput of process cannot exceed
1674 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1675 router linked by an up-link and a down-link (possibly asymmetrical).
1676 This kind of pattern allows to easily build last-mile model style platforms.
1677 Aggregating such patterns in a rule-based AS is thus the technique of
1678 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1679 * New model for Vivaldi routing. We transformed the Vivaldi network model
1680 into a Vivaldi routing model (based on the rule-based model). This allows to
1681 combine Vivaldi based latencies with last-mile platforms.
1684 * Added a check for NaN of IEEE754 infinite in the double entries of
1686 * Introduce a new context factory "raw", highly inspirated from the
1687 ucontext factory, but using manually crafted functions in assembly to
1688 do the work in an efficient manner.
1689 * Allow to change the used context factory at run time, not only at
1690 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1691 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1692 contexts at runtime (only with raw contexts or ucontexts).
1693 * Completely rewrote this module to allow parallel execution of user
1694 processes. Use --cfg=contexts/nthreads:N to execute user processes
1695 with N parallel threads (the default is 1, meaning no parallelism).
1696 * Allow to decide dynamically between sequential and parallel modes.
1697 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1698 processes in parallel only when their number is greater than or equal to P
1700 * Added a check for NaN of IEEE754 infinite in the double entries of
1704 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1705 won't output the state of processes when interrupted with Ctrl-C
1706 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1707 NULL-terminated array. This may solve backward compatibility issues
1708 due to the change to return type of SD_simulate. See also:
1709 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1710 * Add new macros with variable number of arguments.
1711 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1712 - in xbt/asserts.h: xbt_assert
1713 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1714 - in xbt/ex.h: THROWF and RETHROWF.
1715 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1717 * Change xbt_die() to accept a format string with arguments, just like printf.
1718 * New data structure: xbt_lib_t, like a dict but more general and with better
1722 * New configuration options
1723 Options triva/categorized and triva/uncategorized can be used to generate
1724 graph configuration files for Triva visualization tool.
1725 * Configuration option tracing/platform is renamed to tracing/categorized
1726 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1727 * New TRACE_user_link_variable interface:
1728 User provides the name of the link and the tracing variable to attach to it
1729 * the declaration of tracing categories must be done after the environment
1731 * simpler tracing interface, just one way to declare categories
1732 TRACE_category or TRACE_category_with_color, it is up to you
1733 * links in the trace file are again identified by their names
1734 * trace contains the full platform hierarchy exactly as declared using the ASes
1735 * Options tracing/msg/[task|process]:1 groups the process by hosts
1736 for both cases, tasks and processes must have names that are unique during the simulation
1737 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1738 * The experimental option tracing/msg/volume is deprecated
1739 its functionality may be reincorporated if needed
1741 The tracing generates a trace file with unordered timestamped events,
1742 because of the way the core simulator (surf) works. A script available
1743 at the tools directory (fix-paje-trace.sh) can be used to put the events
1744 in order. We have changed the tracing so it can generate ordered timestamped
1745 events in the final trace, but depending on the simulator (and how much time
1746 is simulated) that can lead to a huge memory utilization. It is deactivated
1747 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1749 Build Infrastructure
1750 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1751 - We are trying hard to keep the API stable, but it may happen that
1752 some things change (we're a research project after all, not a
1753 nuclear plant operating system). If such things should happen, you
1754 could rely on that macro to adapt.
1755 - current value: 30600 for 3.06.00, aka 3.6
1756 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1757 a number that can be compared with SIMGRID_VERSION.
1758 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1759 assertions and verbose/debug logging events are disabled at compile time.
1761 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1763 SimGrid (3.5) stable; urgency=medium
1766 * New feature to any SimGrid-based simulator: Model-Checking
1767 Check SIN#1 for more details.
1770 * New Model SMPI (three-interval linear regression for correction factors)
1771 See RR-7426, available at http://hal.inria.fr/inria-00527150
1772 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1773 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1774 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1775 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1776 * See also src/smpi/README
1781 - Tracing API changes: TRACE_start and TRACE_end should not be called
1782 by user-code. They are automatically called by simulators created
1783 with SimDAG, MSG and SMPI if the toolkit is compiled with
1784 tracing_enabled=ON. Categories declaration and utilization remain the
1785 same for MSG and SimDag.
1786 - A function was added to the tracing API to declare categories with
1788 - TRACE_category_with_color (char *category, char *color)
1789 where color must be in the following format
1790 "%f %f %f", red, green, blue
1791 and red, green, blue are float values in the interval [0, 1]
1792 - User can specify NULL as color parameter, or continue calling
1793 TRACE_category (cat)
1794 On that case, the tracing system will define random colors
1795 - The following command-line options are supported:
1796 --cfg=tracing/filename:msg.trace
1797 --cfg=tracing:1 (activate tracing, needed to use others)
1798 --cfg=tracing/platform:1 (categorized resource use)
1799 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1800 --cfg=tracing/msg/task:1 (task creation)
1801 --cfg=tracing/msg/process:1 (process creation, migration)
1802 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1803 --cfg=tracing/smpi:1 (SMPI interface tracing)
1804 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1805 - examples of examples/msg/tracing updated
1807 - DAXLoader and DOTLoader functions can generate tasks with categories
1808 - A new function to attribute a category to SD tasks:
1809 TRACE_sd_set_task_category (SD_task_t task, char *category)
1810 * Tracing the MPI interface implemented by SMPI
1811 - Collective operations are traced with states
1812 - Point-to-Point operations are traced with states/links
1813 - Tracing activated by a parameter "-trace filename" passed
1814 to smpirun during execution (considering that simgrid
1815 is compiled with tracing enabled)
1816 - To run the simulation with gdb, the simulator
1817 accepts --cfg=tracing/smpi:1 to trace SMPI
1818 - tesh files to check if smpi tracing is ok
1819 - See examples/smpi/NAS/DT-trace
1820 * GTNetS tracing re-worked
1821 - adaptation to the tracing system of GTNets to cope
1822 with modifications regarding the fullduplex mode
1823 - new tesh files to check if gtnets tracing is ok
1826 * Asynchronous communications through the functions:
1827 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1828 * New function: MSG_load_platform_script()
1829 to make possible using a lua script instead of XML files to set up platforms
1830 * New function: MSG_set_function
1831 to associate functions to processes, used when bypassing the parser
1832 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1834 Platforms: Add some more examples in examples/platforms
1835 * Grid'5000: see www.grid5000.fr
1836 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1839 * Change the XML format. This is a very important modification. SimGrid
1840 3.5 introduces a new hierarchical format based on the notion of
1841 Autonomous Systems. Compatibility with old format is ensured
1842 through the perl script provided in the install bin directory
1843 bin/simgrid_update_xml.
1844 It is now possible to build platforms with specific routing
1845 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1846 connect several platforms together. We will try to provide soon
1847 set of realistic platforms exploiting these properties (have a
1848 look at examples/platforms/ for the moment).
1849 * Take the opportunity of the XML format change to be a good XML citizen:
1850 rename link:ctn to link_ctn and similar changes (also dealt with by
1852 * Add a new routing scheme (rule-based) using regular expressions. It
1853 enables to have an extremely low memory footprint when the
1854 underlying routing is simple and can be compactly described. You
1855 need to have libpcre4-dev (perl regular expressions) installed if
1856 you want to use this routing scheme.
1857 * Revive the cluster TAG and allow to easily and efficiently (both in
1858 term of memory and speed) connect clusters together. Have a look
1859 at teshsuite/simdag/platforms/ to see how this can be done. With
1860 this tag, you can create clusters with thousands of tasks at no
1861 cost (have a look at examples/platforms/).
1862 Note: clusters are implemented as ASes, so there is no need for an
1863 enclosing AS tag if you have only one cluster in your platform.
1864 * Add new generic functions in the public interface that allows the user
1865 to call SURF 'create_resource' methods from your code (same
1866 functionality as the XML bypass mechanism but with a much lighter
1868 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1869 that uses a piecewise linear approximation to produce better
1870 results when exchanging small messages.
1871 * Add a new parameter to handle correctly full duplex link and account
1872 for interferences between uplink and downlink communications
1873 (activate with --cfg=fullduplex:1).
1876 * Rename the SD_READY (all dependencies are satisfied and task is
1877 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1878 dependencies are satisfied) state.
1879 This prevents a confusion between the notion of "ready to schedule"
1880 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1881 simulated" (SD_RUNNABLE) used by the simulation kernel.
1882 * Change the way a task is considered as ready. Instead of removing
1883 dependencies when a task is done, a counter is decreased. This way,
1884 it is always possible to reach ancestors thanks to the
1885 SD_taks_get_parents function (even after the end of the simulation.)
1886 * Change the return type of SD_Simulate from (SD_task_t*) into
1887 xbt_dynar_t. This function was in handling a dynar internally and
1888 converted it into a NULL terminated array for historical reasons.
1889 * New function SD_dotload(char*) to load a DAG described in dot
1890 format. This loader and the corresponding examples require the
1891 installation of the graphviz library.
1892 * Fix a bug in the management of tasks of size 0 in the surf network
1893 models. This problem was only visible with SIMDAG and you should
1894 thus disregard results produced with earlier versions if you
1895 relied on this feature (some tasks were blocked because of this).
1896 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1897 even though your applications were doing only point-to-point
1898 communications and sequential computations. Now you can really use any
1899 model you want (of course, if you create real parallel tasks, which are
1900 not implemented in most models beside ptaskL07, this will abort).
1901 * Add an example that schedules a DAX on an heterogeneous platform
1902 using a Min-Min strategy.
1903 * New function SD_workstation_get_current_task() that returns the kind
1904 of task currently running on a workstation in the sequential access
1906 * Raise some warnings when unexecuted tasks remains at the end of the
1907 simulation. This is usually caused by cycles in the DAG.
1910 * New function: SIMIX_process_set_function() called by MSG_set_function
1911 * Change the underlying waiting queue in semaphores so that a process
1912 can wait on several of them simultaneously (as in waitany).
1913 * Fix the way to handle tokens in semaphores so that all access patterns
1914 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1915 * kill the dirty pimple SIMIX_message_sizes_output()
1916 Please use (proper) visualization instead
1919 * New data container: setset (set of sets of elements)
1920 * New module: mmalloc (mapped malloc, allowing to have several
1921 independent segments of malloc)
1922 * New function: xbt_dict_cursor_set_data()
1923 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1924 * New function: xbt_dynar_is_empty()
1925 * New function: xbt_fifo_get_last_item()
1926 * Fix xbt_dynar_shrink(): use the right element size.
1927 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1928 uninitialized areas during expand.
1929 * Fix semaphores: previous implementation was severely broken.
1930 * Use library init/fini functions for our initialization.
1931 - you can use logs and other feature as soon as you want in your
1932 code (even before the xbt_init / MSG_init)
1933 - xbt_exit is now a no-op and produce a warning when used.
1936 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1937 benefit from the latest improvement to the simulation kernel.
1938 * Kill measurement sockets for now. If you rely on them, sorry. This
1939 release is not for you. This feature will be reintroduced in the
1940 future, but we cannot delay the release any further.
1941 * New function: gras_msgtype_get_name().
1942 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1943 * Fix (at last) the pmm example: it should not randomly fail anymore.
1945 Build chain: bug fixes and overall polishing
1946 * Cmake is now stable enough. Hence, we killed the autotools.
1947 * Port to windows ( TM :)
1948 * Fix the 'make install' target.
1949 No need to use 'make install-simgrid' anymore
1950 * Introduce a 'make dist' target compiling a *source* archive
1951 'make package' compiles a binary archive
1952 * Compile java files only on need
1953 * Add --cd and --setenv command line options to tesh
1954 * Out of source builds are not fully supported yet, but we are close
1955 * Enable supernovae and optimization flags by default for our users
1958 * Add layer to set up environment directly from lua, without XML.
1959 * The effect of gras_stub_generator can be achieved through
1960 lua too (check examples/gras/console/ping_generator.lua)
1962 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1964 SimGrid (3.4.1) stable; urgency=low
1966 The "Polishing easter eggs is probably a good idea" release.
1967 This is a bug fixes release only.
1971 * Fix a bug preventing the tasks from begin garbage collected.
1974 * Fix a bug occuring when a host involved in a communication fails.
1975 This was not detected properly by the other peer involved in the
1976 communication. Now, it's reported as a network error.
1979 * Warn the user about loop dependencies in data flow of DAX files
1980 * Obey the control-flow dependencies of DAX files
1983 * Add option "enable_smpi" allowing to not compile SMPI.
1984 Probably useful for the (Mac) users experiencing a build error here
1985 * Improve the detection of lua5.1 and ruby1.8
1987 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1989 SimGrid (3.4) stable; urgency=low
1991 The "Easter in Cargese" release. Also known as (major changes):
1993 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1994 ~> bindings were greatly improved
1995 ~> new tracing infrastructure for better visualization introduced
1997 * the "Welcome to configury modernity" release.
1998 ~> we switched from autotools to cmake, and improved our cdash
2001 A more detailled list of changes follow (full detail in svn log).
2003 Java Bindings: Various Cleanups
2004 * (install java-gcj-compat-dev on debian-like to use them)
2005 * Remove put/get: no need to export deprecated interface in Java
2006 Use send/receive instead.
2007 * Cleanup the examples and add a README per directory
2008 * Remove example autoDestination (that's the only way to go now)
2009 * Remove example explicitDestination (was a plain copy of basic)
2010 * Make JniException a runtime exception, so that there is no need to
2011 declare the fact that you may encounter such a beast. I guess that
2012 nobody will ever want to survive such error.
2013 * Create specific errors for each MSG case of failure:
2014 host failure, transfer failure, timeout, task cancelled
2015 * Cleanup the exceptions that may get thrown by each function
2016 * Other internal cleanups in Java bindings. Performance still bad :/
2017 Ruby and Lua Bindings: create them
2018 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2019 * That's new and great, you should try them out.
2020 Same functionalities than Java bindings, only even less polished
2022 * Kill the useless "rate" argument of SD_task_get_execution_time()
2023 Everyone used to provide -1 as a value, it was not used, and the
2024 semantic of a possible use wasn't even clear.
2025 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2026 either as comm costs or compute costs to mean that there is no
2027 such thing for that specific task.
2028 * Add a SD_task_set_name() function
2029 * Fix SD_task_unschedule() on typed tasks
2030 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2031 * In DAX loader, accept useless 'level' attributes to <job> since
2032 LIGO DAGs have them (seem to be to ease graphical representation).
2034 * Add an example masterslave_mailbox.c using send/receive and not
2035 the deprecated put/get interface.
2036 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2037 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2039 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2040 (declare MSG_USE_DEPRECATED to still have the old name)
2041 * Add a MSG_task_set_data() function
2042 * About trace replay (see examples/msg/actions):
2044 - Allow to work with splitted trace files for each process
2045 Give the specific trace file as argument of each process,
2046 and call MSG_action_trace_run(NULL)
2047 You can still have one merged file for all processes.
2048 - Fix implementation of collective operations
2049 * Allow task_execute() on 0-sized tasks (closes #10063)
2051 * This is the first release of SimGrid where SMPI is not considered
2052 beta anymore (even if some corners should still be improved)
2053 * Port over the new SIMIX_network submodule (internal refactoring)
2054 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2055 * Implement more missing elements of the standard:
2057 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2059 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2060 MPI_Scan() MPI_Get_processor_name()
2061 - Added implementation of missing case for Alltoall (warning: it's
2062 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2063 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2064 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2065 SMPI_MPI_Allgatherv()
2066 * Bug fixes include:
2067 - MPI_Waitsome() was broken
2068 - Allow relative includes in smpicc
2069 - Command line cfg argument 'reference_speed' was ignored...
2070 - Some functions did not properly lead to auto-benching of user code
2071 - smpicc passes -O2 by default (just like openmpi one)
2073 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2074 * Bug fixes about timeouts during communications
2075 * add SIMIX_message_sizes_output() as a pimple to write to file the
2076 amount of messages per size. Use gnuplot to get histogram.
2077 Pimple because that's the only user-visible function of simix,
2078 defined directly in xbt.h (irk, sorry)
2080 - Add a SIMIX_sem_get_capacity() function
2081 - Fix interactions with processe resume/suspende
2082 - release_forever() was stupidly broken
2083 - Fix SIMIX_display_process_status() for processes in a semaphore
2084 - Make SIMIX_sem_block_onto() user-visible
2085 * Refactoring context stuff:
2086 - Use pseudo-OOP for better modularity
2087 - reimplement SIMIX_process_kill() without process_schedule() so
2088 that the latter can take as invariant that it is called from
2090 - Merge context_start into context_new for sake of simplicity
2092 * Add a Vivaldi network model, coded live during SUD'10 ;)
2093 * Rename configuration variables to start a hierarchy:
2094 o cpu_model -> cpu/model
2095 o network_model -> network/model
2096 o workstation_model -> workstation/model
2097 * New configuration variables:
2098 o network/bandwidth_factor: correction to bandwith
2099 o network/latency_factor: correction to latency
2100 o netwotk/weight_S: correction to the weight of competing streams
2101 * Add a long description to the models, that users can see with such
2102 argument on the command line: --cfg=cpu/model:help
2103 * --help-models display the long description of all known models
2105 * config: add the ability to set a default value after registration
2106 Does not override any previously set value (e.g. from cmd line)
2107 * dict: allow to have integer key and data.
2108 When so, you need to use the following functions
2109 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2110 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2111 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2112 In contrary to regular dicts, the key is not malloced before copy.
2113 Mixing scalar and regular elements in the same dict is not tested
2115 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2116 Tracing for Visualization:
2117 * SimGrid is now instrumented in order to generate a trace file for
2118 visualization analysis: to use it, need to compile SimGrid with the
2119 "tracing" option enabled, and instrument the program using SimGrid with
2120 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2121 (among other functions).
2122 * The instrumentation only traces the platform utilization for now
2123 * Documentation to use the tracing functions and how to analyze the
2124 traces with the Triva tool is written.
2125 * More information about: SimGrid FAQ (in the section Tracing Simulations
2128 * We moved to cmake as default build system. Autotools support will
2129 be dropped soon. Check the FAQ for more info about how to use it.
2130 * Greatly improved our cdash/ctest interactions
2131 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2132 * Added memory checking tests with valgrind; lot of memleak fixing.
2133 This may be the first release of simgrid with so few memory issues
2134 * Added code coverage tests.
2135 Our coverage is still improvable, but at least we see it on cdash.
2137 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2139 SimGrid (3.3.4) stable; urgency=low
2141 The "Desktop Grid needs love too" release (also called Xmas release).
2143 Models improvements:
2144 * Major speedup in the maxmin system solving by using lazy evaluation
2145 Instead of solving completely the maxmin system at each iteration,
2146 only invalidate (and recompute) the modified parts.
2147 This new feature is enabled in default models but you can try to
2148 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2149 * Cas01 IMproved as default CPU model
2150 This CPU model is the same Cas01 model, but it uses the
2151 maxmin-selective-update flag and a heap structure to manage
2152 actions on SURF kernel.
2153 It reduces the complexity to find the next action to finish and,
2154 consequently, it's faster than the old Cas01.
2155 This is the new default CPU model (Cas01).
2156 * Rename the old Cas01 model to Cas01_fullupdate
2157 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2158 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2159 * CpuTI (CPU Trace Integration)
2160 A new CPU model whose objective is simulate faster when using
2161 availability trace files.
2162 Instead of using a full featured, over engineered maxmin system for
2163 CPU modeling, this model does the pre-integration of traces files
2164 to calculate the amount of CPU power available, and so, executes
2165 faster than the old CPU models.
2166 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2167 * Use LV08 as default network model since it gives better accuracy
2168 for small messages and shouldn't change things for big ones.
2169 Use --cfg=network_model:CM02 to get the previous behavior.
2172 ******************************************
2173 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2174 ******************************************
2175 * The new CPU model may changes simulations!
2176 The point is that events occurring at the exact same timestamp
2177 are not scheduled in the same order with the old and new
2178 version. This may be enough to completely change the execution
2179 of simulations in some cases.
2180 * The new network model will change simulations!
2181 This new model is more realistic than the previous one, so you
2182 should consider redoing your old experiments with this model.
2183 Sorry for the inconvenience.
2186 * Introduce the supernovae compilation mode
2187 When compiled that way, the whole SimGrid (or almost) is put in a
2188 single compilation unit and compiled in one shoot.
2189 This is to help gcc which has difficulties to inline stuff from one
2191 The speedup seem to be above 15%, althrough more tests are needed on
2192 amd64 to confirm that gain.
2195 * Port of MSG's mailbox on top of SIMIX network
2196 The put/get mechanism was greatly simplified on the way.
2199 * New SIMIX network module. Provides:
2200 - Mailbox: rendez-vous mechanism to find with who you want to speak
2201 - Synchronous send/recv: easier and hopefully faster since the
2202 logic is handled in the maestro process directly now
2203 - Asynchronous send/recv: you dreamt of it? It's here now
2204 Too bad that nobody cared enough to propagate the change to MSG.
2205 * Add semaphores as SIMIX synchronization mechanism.
2208 * new function SD_daxload(char*) to load a DAX file
2209 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2210 * Introduce typed tasks. Specify its kind and cost at creation.
2211 At scheduling, just give where it should be placed, and the cost
2212 for each involved resource is automatically computed.
2213 Existing constructors so far (more to come of course):
2214 - SD_task_create_comm_e2e() for end-to-end communication
2215 - SD_task_create_comp_seq() for sequential computation
2216 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2217 * new function SD_task_dump() for debuging display
2218 * new function SD_task_dotty(task,FILE*) writing to file the info
2219 about the task in dotty format
2220 * SD_task_dependency_exists() can now cope with having one of its
2221 arguments NULL. If so, it tests whether the other argument has any
2223 * Add getters on list of preceding/following tasks:
2224 SD_task_get_parents(task) and SD_task_get_children(task)
2225 * Add getters on amount of workstations and list:
2226 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2227 * Add getter on task kind: SD_task_get_kind(task)
2228 * Update the start_time and finish_time of tasks on completion/failure
2229 * Bugfix: Remove task from state swags when destroyed
2232 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2235 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2236 dynar with one separate thread per value of the dynar.
2237 * Change the prototype of xbt_thread_create(), sorry.
2238 Added a boolean parameter indicating whether we want to join this
2239 thread (used in SG only for now)
2240 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2243 * GTNetS wrappers should now be usable again (and betterly tested too)
2244 * Fix a major regression from 3.2 where the timeout provided to
2245 MSG_task_put_with_timeout() was used as absolute time before which
2246 the comm should be done.
2247 * Start to fix the <cluster> tag.
2248 - Internal links should be good now (beside of the loopback, which
2249 use the private link instead)
2250 - paths to the external world is still rather broken
2251 - the <route:multi> tag is just broken. Actually that's brain-dead.
2252 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2253 to make it less stupid
2254 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2255 * Fix a source-level compatibility glitch from 3.2: after defining
2256 MSG_USE_DEPRECATED, you can use the old name
2257 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2258 * Allow to compile from the SVN with automake 1.11
2259 * Fix some problems when using the "start_time" tag in deployment XMLs.
2260 * Fix #8569: XBT/synchro.h has redundant declarations
2261 * Fix #8563: MSG return values and exceptions
2262 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2263 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2264 Thanks to Robson Peixoto.
2265 * Fix a vicious bug in dictionaries inducing that some elements were
2266 not freed on xbt_dict_free()
2268 Portability report of this version:
2269 * Main portability targets:
2270 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2271 on (amd64/i386/ia64)
2272 - mac leopard on i386
2273 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2274 but nothing critical.
2275 * Other platforms: windows, AIX and others were not tested for this release
2277 Timing report of this version:
2278 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2279 scenario parameters). From 8h to a few seconds in desktop grid settings.
2280 * Supernovae brings about 25% speedup on i386.
2282 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2284 SimGrid (3.3.3) stable; urgency=low
2286 The "Need for Speed" release.
2288 The timings done to validate the 3.3.2 were faulty.
2289 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2291 The problem was a conversion from a manually handled vector to
2292 xbt_dynar_t on the critical path.
2293 xbt_dynar_foreach calls functions, inducing stack management crap.
2295 We inlined these functions and xbt_dynar_foreach is now breath taking.
2296 We also inlined xbt_swag_belong on the way.
2298 Here are some approximate speedup measurements (on master/slaves
2299 simulations lasting between 10s and 20s each):
2300 3.3.1 -> 3.3.2: about same performance
2301 3.3.2 -> 3.3.3: 40% speedup
2302 3.3.1 -> 3.3.3: 40% speedup
2303 3.3.1 with inline patch -> 3.3.3: 30% speedup
2305 Our reading is that the refactoring which occurred in 3.3.2 made us
2306 suffer much more from the xbt_dynar_foreach low performance, but
2307 once we solved this, this refactoring proved to be very performance
2308 effective. From the 40% speedup, somehow, 10% are due to the
2309 inlining and 30% to the refactoring.
2311 That's a pitty that gcc cannot inline functions placed in other files
2312 alone. We have to choose between:
2313 - break the encapsulation (by putting private data structures and
2314 accessors in headers files to help gcc)
2315 - live with low performance
2316 - switch to a decent compiler such as icc (not quite possible).
2318 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2320 SimGrid (3.3.2) stable; urgency=low
2322 The "Simplicity does not preceed complexity, but follows it" release.
2324 The main contributors of this release were (lexical order):
2325 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2328 * Extract the routing logic into its own object.
2329 (was dupplicated in network.c and workstation_LV07.c;
2330 Allows to implement other ways of storing that info)
2331 => kill now useless network_card concept
2332 - Use dynar to represent routes (instead of void** + int*)
2333 - kill link_set (use surf_network_model->resource_set instead)
2334 - Add a command-line option to choose the routing schema to use
2335 - Add three new models:
2336 * Floyd (shortest path computed at initialization)
2337 * Dijikstra (shortest path recomputed all the time)
2338 * Cached Dijikstra (shortest path computed on need)
2339 All these models where contributed by Silas De Munck, and are
2340 described in his ICCS09 paper.
2342 * Simplify model declaration
2343 (less redirections, less function to write when defining a model)
2344 - Factorize stuff between models:
2347 surf_model_resource_set(model)
2348 surf_model_resource_by_name(model, name)
2349 - Unify the types of models in s_surf_model_t (using an union)
2350 - Embeed fields of common_public directly into s_surf_model_t
2351 - Rename model methods:
2352 action_free ~> action_unref
2353 action_change_state ~> action_state_set
2354 action_get_state ~> action_state_get
2355 - Change model methods into functions:
2356 (model)->common_public->action_use ~> surf_action_ref
2358 * Implement a generic resource; use it as ancestor to specific ones
2359 (allows to kill duplicated code in models)
2360 Drawback: timer command don't need no name nor properties;
2361 workstation_CLM03 don't need no properties
2362 (but I guess we can live with those few bytes wasted)
2364 * Improve the action object model
2365 - implement a constructor avoiding dupplicated code about field
2366 initialization in generic_action part.
2368 * Kill the SDP model: it has an external dependency, is deprecated
2369 in flavor of modern lmm models, and didn't compile since a while
2372 * Relocation of the context module from XBT to SIMIX.
2373 (the context were decoupled from the simix processes, duplicating a lot of code)
2374 => a lot of code was factorized
2375 - less overhead is introduced during scheduling
2376 - simpler API for the context factory
2377 - the logic for process creation,destruction and manipulation was simplified
2378 * Simplification of the s_smx_process_t data structure.
2379 => accesing the simix level data associated to a process is faster now,
2380 and the code is a lot more readable.
2383 * Implement some more MPI primitives:
2384 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2385 -implementation: Bcast: flat or 2-ary tree (default),
2386 Barrier: 4-ary tree,
2388 Allreduce: Reduce then Bcast
2389 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2390 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2391 Alltoallv: flat tree, like ompi
2393 * Add support for optimized collectives (Bcast is now binomial by default)
2394 * Port smpirun and smpicc to OS X
2397 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2398 Shout out if you used it.
2401 * Display the list of still queued messages in SG mode when existing
2405 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2406 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2407 * Add xbt_str_from_file(FILE*)
2408 * Add xbt_dict_get_key achieving a linear reverse search
2409 * Remove the context module
2411 Portability report of this version:
2412 * Main portability targets:
2413 - Linux(debian)/x86/context
2414 - Linux(debian)/x86/pthreads
2415 - Linux(debian)/amd64/context
2416 - Linux(debian)/amd64/pthreads
2417 On these, we still have the eratic breakages of gras/pmm and
2418 amok/saturate_sg reported in previous version. We still think
2419 that the tests are the cause of the fault, not the tested code.
2421 - Mac OSX Leopard/x86/context
2422 Still false negative in tesh autotesting.
2423 Smpi still fails, but this time because readlink does not accept -f
2424 Everything seems to work properly beside of that.
2427 - AIX version 5.3 (only tested contexts this time)
2428 Smpi still fails there because mktemp is not installed.
2429 Everything seems to work properly beside of that.
2431 I managed to compile it for the first time, but several breakages.
2432 Won't delay the release for this exotic platform.
2434 * Windows: it's still lagging behind. If you want to help, please
2437 Timing report of this version:
2438 This version seem to be more than 5% faster than 3.3.1 (on linux
2439 64bits with contextes). The gain is less than expected, we are
2440 investigating this for next release.
2442 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2444 SimGrid (3.3.1) stable; urgency=low
2447 * Implement a --cfg-help to show existing configuration variables
2448 * Build chain do not require doxygen in maintainer mode
2451 * fix a bug on struct sizeof computation, which prevented the
2452 exchange of arrays of structs in some conditions
2453 - added a regression test about this in datadesc_usage
2454 * Allow the exchange of 0-long dynamic vectors.
2455 - for that, use -1 as indicator of dynamic size instead of 0
2456 - This implied to change any size from unsigned long to long,
2457 reducing a bit communication abilities, but I guess that with
2458 64bits being quite common, this is more than enough.
2459 - This also induce a protocol change, thus bumping network protocol
2460 version from 0 to 1 (if we have external users, we have to get
2461 clean on that point too ;)
2462 - added two regression tests about this in datadesc_usage
2463 * Be more verbose when propagating local exceptions
2464 This helps debugging.
2465 * Display the status of simulated processes when receiving SIGINT in
2469 * Allow to control the simulation from a trace file.
2470 New functions MSG_action_register() and MSG_action_trace_run()
2471 The first one allows to associate a function execution to each
2472 kind of action while the second one parses a trace file and
2473 triggers the corresponding actions within the system.
2474 For now, only a toy example is provided in examples/msg/actions
2475 * Add an exemple of process migration in examples/msg/migration
2476 * Fix a bug in task exchange which broke MSG_task_get_sender()
2477 Add a teshsuite regression test for that.
2478 [Bug: if MSG_task_get_sender() is called after sender exit,
2480 * Fix a bug which prevented suspend/resume to work properly
2481 * Display the status of simulated processes when receiving SIGINT
2482 This fixes a regression of v3.3. due to the introduction of SIMIX
2483 * Bug fixing in failure management:
2484 - trace could not start by a failure at time 0
2485 - failure during communications were not working
2488 * Add SIMIX_process_set_name() to change the name of the current
2489 process in the log messages.
2490 * Store smx_hosts in a dict since we only retrieve them by name
2491 * Move the configuration infrastructure to surf
2494 * Move the configuration infrastructure to surf
2497 * Massive internal cleanups:
2498 - Store internal structures on processes instead of hosts (allows
2499 to have more than one process per host, in addition of being more
2501 - Cleanup the initialization/finalization process
2502 - Kill a whole bunch of unneeded synchronization:
2503 processes run in exclusive manner within the simulator
2504 - Move queues from global tables to process data fields
2506 - now accept -platform and -hostfile arguments
2507 - Pass the right rank value to processes according to the hostfile
2508 * Compile the examples by default, and use them as regression tests
2509 * Implement MPI_Wtime()
2510 * Change the reference speed to a command line option
2513 * TCP_gamma can now be specified as command line option using
2514 --cfg=TCP_gamma:10000000.0
2515 * Change the --surf-path cmd line option into --cfg=path:
2518 * Also include strbuff from xbt.h public header
2519 * xbt_ex_display(): do not free the exception after displaying
2520 This allows to do more with the given exception afterward.
2521 Users should call xbt_ex_free() themselves.
2525 Portability report of this version:
2526 * Main portability targets:
2527 - Linux(debian)/x86/context
2528 - Linux(debian)/x86/pthreads
2529 - Linux(debian)/amd64/context
2530 - Linux(debian)/amd64/pthreads
2531 These targets fail about 1/10 of times on gras/pmm, but we believe
2532 that this is because of the test, not because of simgrid.
2533 amok/saturate_sg fails even more rarely, and the test may not be
2536 - Mac OSX Leopard/x86/context
2537 The test suite still spits tons of errors because some obscure
2538 force prevents us from removing the temporary directories
2539 arguing that they still contain some metadata I've never heard of.
2540 Smpi fails because seq is not installed.
2541 Everything seems to work properly beside of that.
2544 - AIX version 5.3 (both contexts and pthread)
2545 Smpi still fails there because mktemp is not installed.
2546 XML inclusions seems rosty on AIX.
2548 * Windows: it's still lagging behind. If you want to help, please
2551 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2553 SimGrid (3.3) stable; urgency=high
2557 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2558 [Malek Cherier & Mt]
2560 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2563 Until now, GRAS were using MSG as an interface to SURF. It was
2564 quite difficult because both interface have several differences
2565 (MSG channels vs GRAS sockets were the most notable point).
2567 This also opens the gate to SMPI (which should occur soon) and speed
2568 up simulations by to 40% (even if it were not the main goal).
2570 **************************************
2571 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2572 **************************************
2573 The point is that events occuring at the exact same timestamp are
2574 not scheduled in the same order with the old and new version. This
2575 may be enough to completely change the execution of simulations in
2576 some cases. Sorry for the inconvenience.
2578 * Cleanup and upgrade the XML format to push further scalability
2579 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2581 * Improve the testing infrastructure with tesh. Now a very large part of
2582 the code is tested not only by being run but also by checking that the
2583 output match an expected output [Mt].
2585 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2586 is really less memory-demanding, which should allow you to use
2587 larger files in SimGrid [AL].
2589 * Inform valgrind about our contextes, so that it becomes usable
2590 with the default (and more effecient) version of SimGrid
2591 [contributed by Sékou Diakite, many thanks]
2594 * Introduce a listener thread in charge of receiving incoming
2595 messages from the network. It allows to overlap communication and
2596 computation but most notably, it removes some stupid deadlocks due
2597 to the fact that so far, a process could not send and receive at
2598 the same time. This made most non trivial communication schema
2600 * Convert the PIDs from long int to int to match the MSG ones (and
2601 linux ones too) [Mt]
2602 * New function: gras_agent_spawn() to launch a new process on
2603 current host. Only working in simulation for now. [Mt]
2604 * New function: gras_os_hostport() returning a constant form (ie,
2605 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2608 * Make the backtrace of exceptions more human readable [Mt]
2609 * New module: xbt/str [Mt]
2610 a ton of string utility functions (split, join, printf to a newly
2611 allocated buffer, trim, etc)
2612 * New module: xbt/hash [Mt]
2613 SHA1 hashing algorithm (more to come if needed)
2614 * New module: xbt/synchro [Mt]
2615 synchronization tools (mutex and conditions) working the same way
2616 in simulation and in real life (mainly useful for GRAS, but not
2618 * New module: xbt/queue [Mt]
2619 classical producer/consumer synchronization scheme
2620 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2621 (using the classical functions will get serialized) [Mt]
2622 * Make dictionary internal table dynamic. No need to specify its size
2623 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2625 * Make sure the log channels are organized as a tree under windows
2626 (because of ANSI C compatibility issue, any channel were child of
2627 root directly) [Mt].
2630 * Cleaned many thing in surf and fixed a few bugs [AL].
2631 * Add a nice command line configuration mechanism to compose models [AL].
2632 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2633 the previous one (KCCFLN05). It relies on something that looks like
2634 a max-min sharing mechanism but cannot be written as such. A new solver
2635 was thus designed [AL].
2636 * Add a new solver to lmm. Based on Lagrange optimization and
2637 gradient-based descent, it enables to efficiently maximise systems s.a
2639 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2641 This solver enables to propose two new network models for TCP Reno and
2642 TCP Vegas based on Low's work. These models still need to be fully
2643 tested though [Pedro Velho].
2646 * Bug fix in SD_simulate. Now the time bound given as argument is
2648 * Use the new parallel task model (ptask_L07) as default.
2649 * Use the SURF command line configuration mechanism.
2650 * 0-size tasks (for synchronization) should now work.
2652 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2654 SimGrid (3.2) stable; urgency=high
2658 We still experience issues on this platform, but we believe that at
2659 least MSG is usable.
2661 GRAS API BREAKAGE (for simplification purpose, sorry):
2662 * the gras_msgtype_by_name is not used anymore. Instead of
2663 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2664 you can write (and must)
2665 gras_msg_send(toserver, "request", &request);
2666 - If you still want to pass a gras_msgtype_t to the function (to cache
2667 the type and avoid the lookup time), use the gras_msg_send_() variant.
2668 - Impacted functions:
2669 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2670 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2671 * The callbacks are now expected to return 0 when everything went well
2672 (just like the main() function)
2674 GRAS new features and improvements:
2675 * New module mechanism where user code can use per process globals [Mt]
2676 This is similar to gras_userdata_*() functions, but for libraries. It
2677 factorize some code developped over and over in the examples and AMOK.
2678 It has still to be documented and used (only amok/peermanagement is
2680 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2681 loss when they were small enough to fit into the buffer and sent quickly
2682 enough so that they can all get received in one shoot.
2683 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2684 if not found. Use the *_or_null() variant for the old semantic.
2685 * In gras_msg_handle, do not discard messages without callback.
2686 They are probably messages to be explicitly awaited later (ie, proofs of
2687 mis-synchronization in userland since they are sent before being awaited)
2688 No big deal usually.
2689 * gras_socket_meas_send/recv: semantic changed!
2690 The numerical arguments used to be (1) the total amount of data to send
2691 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2692 messages. This was need for the fool willing to send more than MAXINT
2693 bytes on quite fat pipes.
2696 * Do really rename the hostmanagement module to peermanagement. [Mt]
2697 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2698 but this is rather new and this was documented in the module
2699 documentation (poor excuses, I admit)
2700 * Bandwidth measurement semantic changed! This follows the changes to
2701 gras_socket_meas_send/recv explained above.
2704 * A sequential mode has been added to the workstations. When a workstation
2705 is in sequential mode, it can execute only one task, and the other tasks
2706 are waiting in a FIFO. [Christophe Thiery]
2709 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2710 model for SIMDAG. [Christophe Thiery]
2711 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2712 equal to 0, which caused some bad behaviors in
2713 saturated_constraint_set_update. I now use a threshold mechanism like in
2717 * When running manually src/testall, you select specific units [Mt]
2718 testall is the result of our cunit mechanism, and should replace all
2719 the scripty thingy around since bash don't run easily on billware.
2721 * A mallocator system has been added. [Christophe Thiery]
2722 Mallocators allow you to recycle your unused objects instead of freeing them
2723 and allocating new ones.
2725 Documentation update:
2726 * FAQ reworking + New FAQs:
2727 - "Valgrind spits tons of errors!" [Mt]
2728 - "How to repport bugs" [Mt]
2729 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2730 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2731 - Communication time measurement within MSG [AL]
2732 - I experience weird communication times when I change the latency [AL]
2733 * GRAS tutorial [Mt]
2735 - an introduction to the framework and to the used communication model
2736 - an initiatic tour introducing the most proheminent features:
2738 . Lesson 0: Installing GRAS
2739 . Lesson 1: Setting up your own project
2740 o Part 2: Message passing
2741 . Lesson 2: Exchanging simple messages
2742 . Lesson 3: Passing arguments to the processes (in SG)
2743 . Lesson 4: Attaching callbacks to messages
2744 . Lesson 5: Using globals in processes
2745 . Lesson 6: Logging information properly
2746 . Lesson 7: Using internal timers
2747 . Lesson 8: Handling errors through exceptions
2748 . Lesson 9: Exchanging simple data
2749 . Lesson 10: Remote Procedure Calling (RPC)
2750 . Lesson 11: Explicitely waiting for messages
2751 . Recapping of message passing features in GRAS
2752 - A HOWTO section containing:
2753 o HOWTO design a GRAS application
2754 More are due, of course. They will come latter. In the meanwhile, you can
2755 check the examples which are still here.
2757 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2759 SimGrid (3.1) stable; urgency=high
2763 There was a stack corruption somewhere, visible only when optimizing
2764 with these versions. [Vince]
2767 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2768 module that enabled to play with Directed Acyclic Graphs. It is built
2769 directly on top of SURF and provides an API rather close to the old
2770 SG. Some old codes using SG are currently under rewrite to check that
2771 all needful functions are provided. [Christophe Thiery]
2774 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2775 extension of the classical CLM03 model that gracefully handles
2776 failures. This is now the default model for MSG and GRAS. It doesn't
2777 handle parallel tasks yet though. [AL]
2778 * Bug fix: Weights were not correctly set in the network part.
2779 WARNING: This may have resulted in incorrect results with simulations
2780 where there are more than one flow on a given link. [AL]
2783 * After a (long ?) discussion on simgrid-devel, we have decided that the
2784 convention we had on units was stupid. That is why it has been decided
2785 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2786 WARNING: This means that all previous platform files will not work as
2787 such with this version! A warning is issued to ask users to update
2789 A conversion script can be found in the contrib module of the CVS, under
2790 the name contrib/platform_generation/surfxml_update.pl [MQ]
2793 * Bug fix: Processes were started in reverse order, wrt deployment file.
2794 WARNING: if your code relies on this bug, please fix it. [AL]
2795 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2796 being executed on two different locations. [AL]
2797 * Bug fix: Failures are now better supported thanks to Derrick's tests
2798 (there was many failure situations I hadn't thought of and that weren't
2799 correctly handled). [AL]
2800 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2804 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2805 Exception occurring on server-side are propagated back to client (!).
2807 API CHANGE: the callback changed their prototype. Change:
2808 int my_handler(gras_socket_t expeditor, void *payload_data) {
2810 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2811 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2813 * New! function: gras_msg_handleall to deal with all messages arriving
2814 within a given period.
2815 * New! function: gras_socket_server_range to get a server socket in a
2816 range of port numbers (ease to avoid port number conflicts) [MQ]
2817 * New! gras processes display their backtrace when they get a SIGUSR1
2818 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2819 Sweet to debug RL processes [MQ]
2823 - Do not force experiment sizes to be expressed in kb, or it becomes
2824 impossible to measure the latency this way (needs one byte-long tests)
2825 WARNING: this changes the amok_bw_* function semantic. [MQ]
2826 - Implements the link saturation stuff. [MQ]
2827 * Peer management module:
2828 New! module factorizing code that we wrote over and over [MQ].
2831 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2832 - Test units are placed directly into the library code, they get extracted
2833 automatically and placed into the src/testall binary.
2834 - Convert most of the XBT tests to this system.
2835 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2836 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2837 live in the same C code. This required to change a little bit the API
2838 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2839 * New module: sparse graph structure with basic algorithms (this is work
2840 in progress and the API is not considered to be frozen yet). [AL]
2841 * Display more information on backtraces: source line & function names are
2842 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2843 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2844 * Reworked a little bit some #include statements to load only required
2845 headers. Some user code that relied on SimGrid to include stdlib or
2846 stdio may need to include it by themselves. [AL]
2847 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2849 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2850 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2851 but I guess that almost nobody used those parts. [MQ]
2853 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2855 SimGrid (3.0.1) stable; urgency=low
2858 * Unfortunately, I had missed 5 misnamed functions:
2859 xbt_fifo_item_t xbt_fifo_newitem(void);
2860 void xbt_fifo_freeitem(xbt_fifo_item_t);
2861 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2862 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2863 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2864 They're now deprecated. Please use their new versions:
2865 xbt_fifo_item_t xbt_fifo_new_item(void);
2866 void xbt_fifo_free_item(xbt_fifo_item_t);
2867 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2868 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2869 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2871 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2872 * Documentation: xbt_log module unmercifully reworked [MQ]
2873 * Bugfix: there was a problem with the ending of contexts with
2874 the pthread backend. It caused some weird deadlock or behavior
2875 depending on the pthread implementation. [AL]
2876 * Bugfix: get the exceptions raised in the simulator repport where
2877 and why they come from when they are not catched in time [AL, MQ]
2880 * Bugfix: Do repport the error when two non-connected hosts try to
2881 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2884 * Add additionnal checkings on communications. Assert that two
2885 communicating hosts are connected by a set of links... [AL]
2888 * Add additionnal checkings on channel values in communication [AL]
2889 * New: MSG_task_get_source to see on which host a task was generated [HC]
2890 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2891 the number of tasks waiting to be received on channel and sent
2893 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2894 waits for the first task coming from a given host.. [AL]
2896 GRAS new functionnalities: [MQ]
2897 * Enhance the parsing macro to allow the size of multidimentional objects
2898 to be given thru annotations.
2899 * New example (and documentation): Matrix Multiplication a la RPC
2900 (as when I was young!) and fix a bunch of bugs found on the way.
2902 GRAS performance improvements: [MQ]
2904 * Reduce the amount of cbps creation/destruction by making it static to
2905 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2908 * Change libdata to a set so that we can search for stuff by ID (and thus
2909 reduce the insane amount of dict lookups)
2912 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2913 RL communication on top of SG since it's so uneffective.
2914 It may also be used for inter-thread communication in RL, one day.
2915 * Use gras_datadesc_copy() to exchange messages on top of SG
2917 - improve message exchange performance on top of SG
2918 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2919 * Don't exchange on the network the size of the used part of buffer,
2920 instead, specify the possible buffer size to read().
2922 - reduces the amount of read/write calls (one pair per exchange)
2923 - reduces the amount of exchanged data (the size)
2924 - allows to retrieve all arrived data on receiver side, if we don't need
2925 it right now (subsequent read will peek the buffer)
2926 - allows the receiver to proceed with the begining of the stream before
2927 everything is arrived
2928 - make it possible to build an iov transport (using readv/writev)
2930 - take care of the data with non-stable storage (like stacked data),
2932 * If possible, TCP send uses vector I/O (when writev() is here)
2933 - Don't use it for receive since we send data sizes and data on the
2934 same stream, so we wouldn't be able to chain large amount of chunks
2935 before having to flush the stuff to read the size.
2936 * Rework the transport plugin mechanism to simplify it and reduce the
2937 amount of pointer dereferencement when searching for the right function
2940 * I guess that now, we do almost as few system calls as possible while
2941 doing as few data copy as possible.
2943 To improve it further, we could try to send all the sizes first and then
2944 all the data (to use iov on receiving size), but it's only a partial
2945 solution: when you have 2 dimensional data, the sizes of the second
2946 dimension is data of the first dimension, so you need 3 streams.
2948 I'm not sure the potential performance gains justify the coding burden.
2950 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2952 SimGrid (3.00) stable; urgency=high
2955 * New! Give the possibility to hijack the surf parser and thus bypass
2956 MSG_create_environment and MSG_launch_application. Have a look at
2957 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2959 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2961 SimGrid (2.96) unstable; urgency=low
2966 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2967 This deprecates the xbt_error_t mechanisms.
2968 It modifies (simplifies) all XBT and GRAS API.
2969 MSG API keeps unchanged (exceptions raised by XBT are catched from
2970 within MSG and masked with existing error handling facilities)
2973 * New! Add a FATPIPE model. [AL]
2974 * New! Add a parallel task model. [AL]
2975 * New! Add automatically a loopback interface (in the default
2976 network model) if none was precised.
2979 * Bugfix: MSG_process_resume now works with the current running process.
2981 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2982 * Modification of MSG_task_get_compute_duration. Once a task has been
2983 processed, the value returned by this function is now equal to 0. [AL]
2984 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2985 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2986 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2987 of a m_task. For now, MSG_task_cancel only works with computation
2989 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2990 of the processor (in Mflop/s) regardless of the current load on the
2992 * API Change: use proper naming convention for MSG_getClock and
2993 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2995 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2996 This function changes the priority of a computation task. This priority
2997 doesn't affect the transfer rate. A priority of 2 will make a task
2998 receive two times more cpu power than the other ones. This function
2999 has been added to suit the needs of Nguyen The Loc and hasn't been that
3000 much tested yet. So if it fails, please report it and send me your code.
3002 * API Change: removed all functions and types that were marked "deprecated"
3003 since many months. Renamed MSG_global_init_args to MSG_global_init.
3005 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3007 SimGrid (2.95) unstable; urgency=low
3010 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3011 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3012 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3015 * Bugfix: Allow absolute paths for platform description files [MQ]
3016 * Bugfix: do free the variables after use. Leads to drastic performance
3018 * Implement max_duration (ie, timeouts) on resources [AL]
3021 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3023 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3024 the needed support (timeouts) [AL]
3027 * Implement measurement sockets. You can now get the bandwidth between two
3028 hosts thanks to AMOK (see below). [MQ]
3029 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3030 dynar over the network (yeah) [MQ]
3031 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3032 * simple/static token-ring example. [Alexandre Colucci and MQ]
3033 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3034 in place before [MQ]
3037 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3038 type parsing. "short" and "long" modifiers were also ignored; other
3039 modifier (such as reference level) are still ignored. [MQ]
3040 * Embeed the buffer size within the buffer itself on SG. [MQ]
3041 That way, send() are atomic and cannot get intermixed anymore (at least
3042 the ones which are less than 100k; bigger messages still have the issue)
3043 * Array size pushed by the field, not by the field type (or each
3044 and every long int will push stuff to the cbps) [MQ]
3045 * use select() to sleep since it allows to portably sleep less than one
3048 GRAS (minor cleanups)
3049 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3052 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3053 * type category 'ignored' killed as it was never used and were difficult
3055 * whether a type can cycle or not is now a flag, leaving room for more
3056 flags (as "ignored", if we feel the need one day ;)
3057 * Rename raw sockets to measurement sockets since "raw" has another
3058 meaning in networking community.
3061 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3062 and offers features not belonging to GRAS but that most applications
3063 need. One day, it may be a set of plugins loadable at runtime.
3064 * New module: bandwidth
3065 bandwidth measurement between arbitrary nodes running this module. [MQ]
3067 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3069 SimGrid (2.94) unstable; urgency=low
3071 The first beta release of SimGrid 3 !
3075 * Update the main page and the FAQ. Adding references to gforge.
3078 * Add a gras_os_getpid function.
3081 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3082 * Extend the logs so that they also print PID, hostname, date, ... if
3084 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3085 and kill the old version which were in testsuite/
3086 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3090 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3091 for users. Instead, each programming environment (GRAS, MSG,...) use it
3092 when needed to provide such a feature to users.
3093 Don't shortcut the mechanism or you will also shortcut the virtualization
3094 you need on the simulator.
3098 * Cleanups in configury with regard to compile optimization/warning flags.
3099 Also add -fno-loop-optimize to any powerpc since it's the optimization
3100 killing gcc (< 3.4.0).
3101 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3102 needed by MSG examples complications
3103 * Borrow configury beautifications from PHP
3106 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3107 warning (thanks loris for stumbling into this one).
3108 * Bugfix: stop loading private headers (gras_config.h) from the public
3112 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3113 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3114 RL would imply a DNS resolver, which is *hard* to do in a portable way
3115 (and therefore delayed).
3116 * Implement a real timer mechanism and use it in timing macros. This allows
3117 to avoid rounding errors and get a 0.000005 sec precision in timing
3118 macros. While I was at it, various cleanups:
3119 - allow to declare more than one timed section per file (fix a stupid bug)
3120 - move some private declaration to the right place
3121 - merge conditional execution and timing macros into emulation module
3122 - document the module
3123 - make sure the module cleanups its mess on gras_exit
3124 * Documentation improvements:
3125 - (new) how to compile applications using GRAS
3126 - (new) emulation support (timing macros)
3128 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3130 SimGrid (2.93) unstable; urgency=low
3132 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3135 - Use Paje properly where used. Still to be sanitized properly.
3136 - Portability fix: Add an implementation of the contexts using pthread
3140 - Add xbt_procname(): returns the name of the current process.
3141 Use it to show the current process's name in all logging.
3143 - fix detection of older flex version and the reaction, since we do
3144 depend on modern ones (we use lex_destroy)
3145 - Better separation of SG and RL in the libs: remove all simulation code
3146 from libgras. As a result, this lib is now only 200k when stripped.
3147 Some of the xbt modules may also be duplicated (two sets and such) and
3148 should be cleaned/killed before SG3.
3149 - Insist on using xlC on AIX because of weird problems involving gcc there.
3150 - Cleanup the make remote stuff. This is now done by scripts
3151 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3152 mainly for our private use, but we're working on changing them to user
3155 - Bugfix: flush the socket on close only if there is some *output*.
3156 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3158 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3161 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3163 SimGrid (2.92) unstable; urgency=low
3165 Alpha 3 on the path to SimGrid 3
3169 - New! First try of benchmarking macros.
3170 - New! First try so that gras_stub_generator generate deployment and
3171 remote compilation helpers.
3173 - Bugfix: Initialization fix in msg_test.
3177 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3179 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3181 - API Change: gras_os_sleep to take the amount of seconds as a double.
3182 Accepting an int was error prone since it was the only location where
3183 seconds were coded as such. It leaded to damn rounding errors.
3184 - Bugfix: Hard to belive that timers ever worked before this.
3186 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3188 SimGrid (2.91) unstable; urgency=low
3190 Alpha 2 on the path to SimGrid 3
3194 - Bug fix in the lmm_solver.
3196 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3197 through the function MSG_paje_output.
3198 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3199 - It is possible to bound the rate of a communication in MSG with
3200 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3201 to write it in the changelog).
3202 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3207 - Various cleanups to the autotools stuff
3208 - Begin to move Gras examples to examples/gras/
3209 - Let make distcheck work again (yeah!)
3211 - documentation overhauled using doxygen.
3212 gtk-doc-tools is dead in SimGrid now.
3213 - Automatically extract all existing logging categories, and add the list
3214 to the documentation (long standing one, to say the less)
3216 - Cleanup the known architecture table. Reorder the entries to group what
3217 should be, and use a more consistent naming scheme.
3218 (some of the test dataset are still to be regenerated)
3219 - New! Allow library to register globals on each process just as userdata
3221 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3222 lookup time (for now).
3223 Use it in msg and trp.
3224 This cleans a lot the internals and helps enforcing privacy of the
3225 headers between the gras components.
3226 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3227 - Bugfix: gras_os_time was delirious in RL.
3228 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3230 - Reenable GRAS now that it works.
3232 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3234 SimGrid (2.90) unstable; urgency=low
3236 Alpha 1 on the path to SimGrid 3
3238 * It is a long time since the last release of SimGrid. I'm sorry about
3239 that but as I had told you, I was rewriting a lot of things. I apologize
3240 to those who had been reporting bugs to me and that I had not answered.
3241 If your bug is still in the new version, please tell me. Here is a
3242 summary of the main changes.
3244 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3245 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3246 portability, speed, and a lot more but you'll figure it out later.
3247 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3248 I wanted to release the new MSG as soon as possible and I have
3249 broken GRAS, which is the reason why, for now, only MSG is fully
3250 functional. A laconic description of these projects is available
3251 in the documentation.
3253 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3254 kernel optimized for our needs (called SURF but only the developers
3255 should use it). Hence, MSG is now roughly 30 times faster and I think
3256 that by rewriting a little bit MSG, I could event speed it up a little
3257 bit more. Beside the gain in speed, it is also much easier to encode a
3258 new platform model with SURF than it was with SG. More to come...
3260 * REVOLUTION 3: I have tried to change a little as possible the API of
3261 MSG but a few things really had to disappear. The main differences
3262 with the previous version are:
3263 1) no more m_links_t and the corresponding functions. Platforms are
3264 directly read from a XML description and cannot be hard-coded
3265 anymore. The same format is used for application deployment
3266 description. The new format is described in the documentation.
3267 Have a look in tools/platform_generation. There is a tiny script
3268 that converts from the old platform format to the new one. Concerning
3269 the application deployment format, parsing the old one is tricky.
3270 I think most of you should however be able to convert your files. If
3271 it is really an issue, I can write a C code that does the conversion.
3273 2) the toolbox tbx does not exist anymore. We now have a library
3274 with much more data-structures but without the hash-tables (we have
3275 dictionaries that are much faster).
3277 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3279 *****************************************************************************
3280 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3281 * don't feel like converting the oldies. (Mt) *
3282 *****************************************************************************
3285 Version 2.90: "the long awaited one"
3286 - Finished rewriting and debugging MSG. Rewrote the documentation.
3287 - disable GRAS for now since it needs to be ported to the newest SG
3290 - Finish the port to windows (using mingw32 for cross-compile)
3293 - Main loop and datastructures of SURF. A cpu resource object is
3294 functional. Surf can thus be used to create cpu's with variable
3295 performance on which you can execute some actions.
3297 2004-11-15 Martin Quinson
3298 - Port to ARM. Simply added the alignment and size descriptions. Should
3299 work, but the ARM machines are so slow that I didn't had the opportunity
3300 to 'make check' over there yet.
3302 2004-11-15 Arnaud Legrand
3303 - Trace manager now written. It uses a heap structure and is therefore
3304 expected to be efficient. It may however be speeded up (particularly
3305 when many events occur at the same date) by using red and black
3306 trees. One day maybe...
3307 - Max-min linear system solver written. It uses a sparse matrix
3308 structure taking advantage of its expected use. Most operations are
3309 O(1) and free/calloc are called as few as possible. The computation of
3310 the minimum could however be improved by using a red and black tree
3313 2004-11-03 Arnaud Legrand
3314 - Rename every gras_* function that was in xbt/ to its xbt_
3316 - Add a heap and a doubly-linked list to xbt
3317 - Added a dichotomy to the dictionaries. make check works as well before
3318 so I assume that the patch is correct. I do not know however if things
3319 run effectively faster than before now. :)
3321 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3322 SimGrid, and the version number is bumped to 2.x
3324 2004-10-29 Martin Quinson
3325 - Introduction of the remote errors.
3326 They are the result of a RMI/RPC on the remote machine.
3327 ErrCodes being scalar values, you can't get the host on which those
3328 errors did happen. Extending the error mechanism as in Gnome is possible.
3329 No idea yet whether it is a good idea.
3331 2004-10-28 Martin Quinson
3332 - Interface revolution: the Starred Structure Eradication.
3333 I used to do typedef struct {} toto_t; and then handle *toto_t.
3334 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3335 - ???_t is a valid type (builded with typedef)
3336 - s_toto_t is a structure (access to fields with .)
3337 - s_toto is a structure needing 'struct' keyword to be used
3338 - e_toto_t is an enum
3339 - toto_t is an 'object' (struct*)
3341 typedef struct s_toto {} s_toto_t, *toto_t;
3342 typedef enum {} e_toto_t;
3343 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3344 s_toto_t) is private.
3346 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3347 so that it can set the variable to NULL. It was so for dicts and sets,
3348 it changed for dynars.
3350 - Fix a bunch of memleaks in dict_remove
3351 - Fix a bug in sg/server_socket opening: it failed all the time.
3353 2004-10-07 Martin Quinson
3354 - Speed up dynar lookup operation a bit.
3356 gras_dynar_get is dead.
3358 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3359 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3360 (which gives you the address of the stored data).
3362 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3363 copy of the data using an affectation to do the job and not a memcpy.
3365 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3366 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3368 It does not really speedup the dynar test because they are
3369 setting elements all the time (and look them seldom). But the dict does
3370 far more lookup than setting.
3372 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3374 2004-10-05 Martin Quinson
3375 - Allow to (en/dis)able the cycle detection at run time.
3377 Whether we should check for cycle or not is now a property of each
3378 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3379 datadesc_cycle_unset allow to remove this property when previously set.
3381 Note that the cycle detection is off by default since it impacts the
3382 performance. Watch the data you feed GRAS with ;)
3384 This property is hereditary. Any element embedded in a structure having it
3385 set have it set for the time of this data exchange.
3387 You should set it both on sender and receiver side. If you don't set it on
3388 sender side, it will enter an endless loop. If you forget on receiver
3389 side, the cycles won't be recreated after communication.
3391 - Header reorganization.
3392 Kill gras_private.h, each submodule must load the headers it needs.
3394 2004-10-04 Martin Quinson
3395 - Interface revolution: do not try to survive to malloc failure.
3397 Now, gras_malloc and friends call gras_abort() on failure.
3398 As a conclusion, malloc_error is not a valid error anymore, and all
3399 functions for which it was the only gras_error_t return value are
3400 changed. They now return void, or there result directly.
3401 This simplify the API a lot.
3403 2004-09-29 Martin Quinson
3404 - Re-enable raw sockets.
3405 Created by gras_socket_{client,server}_ext;
3406 Used with gras_raw_{send,recv}
3409 It should allow to kill the last bits of gras first version soon.
3411 This is not completely satisfactory yet (duplicate code with
3412 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3415 - Simplify transport plugin (internal) interface by not passing any
3416 argument to _server and _client, but embedding them in the socket
3419 2004-09-28 Martin Quinson
3420 - Finish the port to AIX.
3421 autoconf was my problem (segfault within the malloc replacement
3422 function. No idea why)
3424 2004-09-16 Martin Quinson
3425 - Fix some size_t madness on 64bit architectures.
3427 2004-09-08 Martin Quinson
3428 - Reduce the number of system headers loaded, overload some more system
3429 calls (such as malloc to cast the result of the system one, and work
3431 - Fix and reintroduce the config support
3433 2004-09-07 Martin Quinson
3434 - Source code reorganization to allow Arnaud to surf all over there.
3435 - Allow to document the logging categories.
3436 - Remove all uppercase from logging categories and useless cleanup in names.
3438 2004-08-18 Martin Quinson
3439 Version 0.6.2 (protocol not changed; API changed)
3440 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3441 gras_error_t), and NULL when not found. Functions expecting a msgtype
3442 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3443 hopefully usefull message.
3444 - Portability to prehistoric sparcs again
3446 2004-08-17 Martin Quinson
3447 Version 0.6.1 (protocol not changed; ABI not changed)
3448 - prealloc some buffers to speed things up
3450 2004-08-11 Martin Quinson
3451 Version 0.6 (protocol not changed; ABI expended)
3452 - The parsing macro can deal with the references, provided that you add
3453 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3455 2004-08-09 Martin Quinson
3456 Version 0.5 (protocol not changed; ABI changed)
3457 - Allow to off turn the cycle detection code in data exchange at
3458 compilation time. It should be at run time, but I'm short of time (and
3459 the config stuff is still broken). That way, we keep dict out of the
3460 critical path, which is good because the performance is poor:
3461 - search not dichotomial yet
3462 - dynar give no way to access their content and memcpy everytime
3463 - In composed data description (struct, ref and so on), stop foolness of
3464 keeping the subtype's ID, but store the type itself. This keeps sets out
3465 of the critical path, which is good since they rely on dynar and
3466 dictionnaries. The only loose of that is that we cannot detect the
3467 redeclaration of a structure/union with another content (but I'm not sure
3468 the code detected well this error before anyway). We still can detect
3469 the redefinition discrepancy for the other types.
3470 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3471 it breaks the code because of type-punning used all over the place).
3472 This breaks on all non-gcc architectures (for now).
3474 All those changes (plus the buffer of last time) allow me to gain 2 order
3475 of magnitude on cruel tests consisting of 800000 array of integers on two
3476 level of a hierarchical structure (200 secondes -> 4 secondes)
3479 - the selector of reference must now return the type it points to, not
3480 the ID of this type.
3482 2004-08-06 Martin Quinson
3483 Version 0.4 (protocol changed; ABI not changed)
3484 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3485 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3486 - Implement a buffer transport, and use it by default (it relies on tcp in
3487 real life and on sg in simulation).
3488 That's a bit hackish since I had a new field to the structure to store
3489 its data without interfering with the subtype ones. Inheritance
3490 is tricky in C. And that's a kind of reverse inheritance with one class
3491 derivating two classes. Or maybe a game with java interfaces. Anyway,
3492 that's damn hard in C (at least).
3493 Moreover, I got tired while trying to ensure plugin separation and
3494 genericity in SG mode. MSG wants me to do weird things, so let's go for
3495 cruel hacks (temporarily of course ;).
3496 See comment in transport_private.h:71
3497 - do not include all the _interface headers in private but in the files
3498 which really need them (to cut the compilation time when they are
3501 2004-07-26 Martin Quinson
3502 Version 0.3 (protocol not changed; ABI changed)
3503 - Major overhault of the datadesc interface to simplify it:
3504 - shorted the function names:
3505 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3506 - add a trivial way to push/pop integers into the cbps without malloc.
3507 This allows to make really generic sub_type description, which simply
3508 pop their size of the stack.
3509 - add a function gras_datadesc_ref_pop_arr() which does what users want
3510 most of the time: Declare a dynamic array (which pops its size of the
3511 stack) and declare a reference to it. Poor name, but anyway.
3512 - kill the post-send callback, add a post-receive one
3514 2004-07-23 Martin Quinson
3515 Version 0.2 (protocol changed; ABI changed)
3516 - add some testing for cpbs in the test cases, and fix some more bugs.
3517 This invalidate again the little64 data file, since I cannot regenerate
3519 - remove an awfull optimization in the logging stuff, allowing me to:
3520 - understand it again
3521 - learn gcc how to check that the argument match the provided format
3522 - fix all errors revealed by gcc after that
3523 - internal keys of dict are not \0 terminated. Deal with it properly in
3524 loggings instead of segfaulting when the user want to see the logs :-/
3526 2004-07-22 Martin Quinson
3527 - Fix some stupid bug preventing cbps (callback postit) from working
3529 2004-07-21 Martin Quinson
3530 - Some documentation cleanups
3531 - remove the useless last argument of msgtype_declare
3532 - rename the Virtu functions to fit into the 'os' namespace
3533 - move headers src/include -> src/include/gras/ and stop fooling with
3534 gras -> . symbolic link
3535 - make distcheck is now successful
3537 2004-07-19 Martin Quinson
3539 - Build shared library also
3540 - Install html doc to the right location
3541 - stop removing maintainer files in make clean
3542 - build tests only on make check
3544 2004-07-13 Martin Quinson
3546 - No major issue in previous version => change versioning schema
3547 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3548 corresponding dataset.
3550 2004-07-11 Martin Quinson
3552 - Get it working with any kind of structure (we can compute the padding
3553 bytes remotely for all the architectures I have access to)
3554 - Implement the structure parsing macro (still not quite robust/complete)
3555 - Improvement to the remote testing toysuite
3557 2004-07-10 Martin Quinson
3558 [autoconf mechanism]
3559 - get ride of a bunch of deprecated macros
3560 - actually run the test for two-compliment, not only compile it :-/
3561 - test whether the structures get packed (and bail out if yes. Damn.
3562 Alignment is a serious matter)
3563 - test whether the structures get compacted (but respecting the alignment
3564 constraints of each types)
3565 - test whether the array fields of structures can straddle alignment boundaries
3567 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3568 (took me 2 hours to find that bug, looking at the wrong place)
3569 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3570 before sending/receiving and are used to compute the offsets of fields
3571 - Given that padding size depend even on compiler options, keep track of
3572 alignment and aligned_size only for the current architecture. Not a big
3573 deal since we send structure fields one after the other (seems
3575 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3576 program, let it work on linux/gcc/little32. portability todo.
3578 2004-07-08 Martin Quinson
3579 - import and improve remote compilation support from FAST
3580 - make sure make check works on half a dozen of machines out there
3582 2004-07-07 Martin Quinson
3583 Let's say it's version 0.0.3 ;)
3584 - Implement conversions (yuhu!)
3585 - Let it work on solaris (beside conversion, of course)
3586 - Stupid me, using rand() to generate the conversion datatests in not wise.
3588 2004-07-06 Martin Quinson
3589 - Let make dist work, since I'm gonna need it to compile on remote hosts
3590 - Let Tests/datadesc_usage write the architecture on which the file was
3591 generated as first byte.
3592 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3593 architecture descriptions.
3594 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3595 on those architectures.
3596 - Optimization: send/recv array of scalar in one shoot
3598 2004-07-05 Martin Quinson
3599 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3601 - Plug a whole bunch of memleaks
3602 - each process now have to call gras_{init,exit}. One day, their log
3603 settings will be separated
3604 - Continue the code factorisation between SG, RL and common in Transport.
3606 2004-07-04 Martin Quinson
3608 - Redistribution between SG and RL.
3609 We wanna have to accept in SG, so move accepted related parts of RL in
3610 the common part. (more precisely, the dynar of all known sockets is no
3611 more a static in transport.c, but part of the process_data)
3613 [gras_stub_generator]
3614 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3616 2004-07-03 Martin Quinson
3617 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3618 them all up in one shot)
3620 - Fix the ugly case of reference to dynamic array.
3621 - New (semi-public) function gras_datadesc_size to allow the messaging
3622 layer to malloc the needed space for the buffer.
3624 - gras_socket_close now expect the socket to close (and not its address to
3625 put NULL in it after it). This is because the socket passed to handlers
3626 is one of their argument (=> not writable).
3628 - propagate the interface cleanup from last week in datadesc, ie remove a
3629 superfluous level of indirection. User pass adress of variable
3630 containing data (both when sending and receiving), and not of a variable
3631 being a pointer to the data. Let's say that I like it better ;)
3632 The price for that is constructs like "int msg=*(int*)payload" in
3633 handlers, but it's a fine price, IMHO.
3635 - Let it work in RL (yuhu)
3637 2004-06-21 Martin Quinson
3639 - porting SG plugin and SG select to new standards (works almost).
3640 - plug memleaks and fix bugs around.
3643 - cleanup the prototype of data recv and force users to specify when they
3644 want to handle references to objects. Test case working even for cycles.
3645 - plug memleaks. Valgrind is perfectly ok with this.
3647 2004-06-12 Martin Quinson
3649 - cleanup the separation between plugin and main code in plugin creation
3651 2004-06-11 Martin Quinson
3653 - Reput hook for raw sockets, needed for BW experiments
3654 - kill a few lines of dead code
3655 [Data description] Interface cleanup
3656 - gras_datadesc_by_name returns the searched type or NULL.
3657 That way, no variable is needed to use a type desc once, which makes
3659 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3660 two parameters were strings (field name, type name), leading to
3662 [Dicos] Interface cleanup
3663 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3664 This is consistant with the dynar API.
3666 2004-04-21 Martin Quinson
3668 - Porting to new standards.
3670 - interface cleanup.
3671 There is no bag anymore, no need to take extra provision to mask the
3672 pointers behind "ID".
3673 Better splitup of functions between files create/exchange/convert.
3674 This is still a bit artificial since convert and receive are so
3675 interleaved, but anyway.
3677 - add a queued message list to procdata (the ones not matching criteria
3679 - factorize some more code between SG and RL wrt procdata
3681 - use gras_exit in example to track memleaks
3682 - get rid of gs_example now that GS is properly integrated into gras
3683 - update run_test to integrate the latest tests (datadesc)
3685 - rename WARNINGn macros to WARNn since it prooved error-prone
3687 2004-04-19 Martin Quinson
3689 - register init/exit functions within gras module mechanism
3690 - send/receive function.
3691 Convertion is not implemented, but short-cutted if not needed.
3692 struct/array elements are sent one by one (instead of block-wise), but
3693 nobody really cares (yet). Get a prototype before optimizing.
3694 - tests (using a file socket) for DD send/receive on:
3695 - base types: int, float
3696 - array: fixed size, string (ie ref to dynamic string)
3697 - structure: homogeneous, heterogeneous
3698 - chained list, graph with cycle
3699 Believe it or not, valgrind is not too unhappy with the results. The
3700 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3704 - Bugfix when using a filename explicitely (instead of '-')
3706 2004-04-09 Martin Quinson
3708 - factorize more code between RL and SG in socket creation
3709 - Complete the implementation and tests of:
3711 o file (only in RL, and mainly for debugging)
3713 I lost 3 days to design a portable address resolver, and then decided
3714 that the prototype mainly have to run on my box.
3715 Addressing portability too early may be like optimizing too early :-/
3717 - use gras_init in the Tests instead of the crappy parse_log_opt
3718 (the latter function is removed)
3719 [Conditional execution]
3720 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3721 [Code reorganisation]
3722 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3723 Build examples against the RL library, since there is no way to disable
3724 its creation for now.
3726 For information, the beginning of coding on GRAS was back in june
3727 2003. I guess that every line has been rewritten at least twice since