1 SimGrid (3.21) NOT RELEASED (Release Target: September 23. 2018, 1:54 UTC)
4 - s4u::Io: IOs go asynchronous as activities. This comes with new methods in the
6 - io_init(sg_size_t, s4u::Io::OpType) to create a READ or WRITE asynchronous
7 IO operations that can be started, waited for, or canceled as a regular
9 - read_async(sg_size_t) and write_async(sg_size_t) which are wrappers on
11 - Fully reimplement auto_restart mechanism that was utterly broken
12 (fix #22, #131 and #281 that were all very related)
13 - When creating an actor from a function and its parameters,
14 move-only parameters are not allowed anymore, as it would prevent
15 the actor to be restartable if its parameters are consumed on run.
18 - Rename 'power' and 'power_used' variables into 'speed' and 'speed_used'
19 - New host variable: 'core_count'
22 - Remove xbt_os_thread_specific features
23 - Remove portability wrapper to condition variables
24 - Remove xbt_os_thread_yield()
27 - Due to an internal bug, Msg.run() must now be your last line.
28 We hope to fix it in a future release, and we are sorry for the inconvenience.
31 - #22: Process autorestart seem to only work with CAS01 cpus
32 - #93: simgrid should not eat --help
33 - #131: Java : autostart does not work
34 - #264: Add ptask L07 resource tracing
35 - #271: Dynamic and manual replay of SMPI traces
36 - #279: Breakpoints option uses time deltas instead of absolute time
37 - #281: Daemonized actor no longer daemonized after an auto-restart
38 - #282: TIT convertor erase traces when using absolute path in the trace list file
39 - #285: segfault when a s4u actor kills itself with msg process tracing activated
40 - #286: Pajé traces are not exposing the number of cores
41 - #287: Command-line parsing should stop when encountering '--'
42 - #288: MPI_Init(NULL, NULL)
43 - #296: DTD too permissive
45 ----------------------------------------------------------------------------
47 SimGrid (3.20) Released June 24. 2018
49 The proxy snake_case() release.
52 - The whole API is now using snake_case() instead of mixing it with camelCase.
53 Compatibility wrappers in place for one year.
54 - We took the opportunity to increase the consistency of the API.
55 Please report any remaining glitches.
56 - New function simgrid::s4u::Host::get_actor_count:
57 Returns the number of actors running on a specific host.
60 - SMPI is now tested with ~45 proxy apps from various sources, with none or
61 only minor patching needed: check https://github.com/simgrid/SMPI-proxy-apps
62 - Replay: The replay file has been re-written in C++.
63 - Replay: Tags used for messages sent via MPI_Send / MPI_Recv are now
64 supported. They are stored in the trace and used when replayed.
65 - Basic support of MPI_Cancel. Robustness not guaranteed.
66 - Support of MPI_Win_allocate_shared, MPI_Win_shared_query, MPI_Comm_split_type
67 (only for MPI_COMM_TYPE_SHARED).
68 - New option: smpi/privatize-libs, to add external shared libs to be privatized
69 by SMPI. They will be copied locally and loaded separately by each process.
70 Example --cfg=smpi/privatize-libs:"libgfortran.so.3;libscalapack.so".
71 - Tracing: add tracing for MPI_Start, Startall, Testall, Testany
72 - Interception of getopt, getopt_long and getopt_long_only calls to avoid issues
73 with internal index optind with multiple processes. Only works if MPI_Init has
75 - Fortran: SMPI builds a mpi.mod file which should allow use of "use mpi"
76 syntax without preprocessing tricks.
79 - Change --cfg=tracing/msg/vm to --cfg=tracing/vm as virtual machine
80 behavior tracing is no longer limited to MSG
81 - TIT (Time Independent Traces): We finally support tags. Unfortunately,
82 this means that traces now need to be updated or re-obtained; both Irecv
83 and Isend lines in your traces have now in total 5 mandatory fields:
84 <rankid> <command> <to/from rankid> <tag> <size>
85 To update your traces, it suffices to add a 0 for the tag here.
86 - TIT now also supports waiting for a distinct request via MPI_Wait.
87 Wait/Test now wait for a specific request, not just the last one that was
88 issued. This unfortunately means another update, because we need to
89 identify which request you want to wait for. We do this via the
90 triplet (sender, receiver, tag), which needs to be added:
91 <rankid> <command> <sender> <receiver> <tag>
92 - Actions are now in lower case (e.g., "allReduce" becomes "allreduce").
93 New script simgrid_convert_TI_traces.py: upgrade your TI traces.
96 - The deprecation of MSG is ongoing (but this should not impact you).
97 Many MSG functions are now simple wrappers on the C API of S4U. If
98 you wish to convert your code to S4U, find the S4U counterparts of
99 your MSG calls in src/msg/msg_legacy.cpp.
100 - MSG can still be used, but won't evolve anymore.
103 - Allow to run the Link energy plugin from the command line with
104 --cfg=plugin:link_energy
105 - Rename Energy plugin into host_energy
106 - Rename Load plugin into host_load
109 - Add parameter --cfg=simix/breakpoint to raise a SIGTRAP at given time.
110 - kill simix::onDeadlock() that was somewhat dupplicating s4u::on_deadlock()
111 - Improve performance when handling timeouts of simix synchros.
114 - Config: the C API is now deprecated (will be removed in 3.23), and
115 the C++ API has been slightly improved.
118 - Fix several build issues on OSX.
119 - Move simgrid_config.h to simgrid/config.h (old header still working)
122 - #143: Setting a breakpoint at a given time
123 - #150: Inconsistent event names in SMPI replay
124 - #258: daemonized actors hang after all non-daemonized actors have completed
125 - #267: Linker error on unit_tmgr
126 - #269: SMPI: tracing of MPI_Wait/all/any broken
127 - SMPI: Fix various crashes with combined use of MPI_PROC_NULL and MPI_IGNORE_STATUS
129 ----------------------------------------------------------------------------
131 SimGrid (3.19.1) Released March 22 2018
133 The Fixed ABI Release.
135 As you may know, we are currently refactoring SimGrid in deep.
136 Upcoming SimGrid4 will be really different from SimGrid3: modular,
137 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
139 Our goal is to smooth this transition, with backward compatibility
140 and automatic update paths, while still progressing toward SimGrid4.
142 SimGrid remains open during works: The last pure SimGrid3 release was
143 v3.12 while all subsequent versions are usable alpha versions of
144 SimGrid4: Existing interfaces remain unchanged, but the new S4U
145 interface is budding and the internals are deeply reorganized.
147 Since 2015, we work hard to reduce the changes to public APIs. When
148 we need to rename a public library symbol in S4U, we let your
149 compiler issue an explicative warning when you use the deprecated
150 function. These messages remain for four releases, i.e. for one full
151 year, before turning into an error. Starting with v3.15, your can
152 also adapt to API changes with the SIMGRID_VERSION macro, that is
153 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
155 Starting with this v3.19.1, our commitment to reduce the changes to
156 the public interfaces is extended from the API to the ABI: a program
157 using only MSG or SimDag and compiled against a given version of
158 simgrid can probably be used with a later version of SimGrid without
159 recompilation. We will do our best... but don't expect too much of
160 it, that's a really difficult goal during such profund refactoring.
162 The difference between v3.19 and v3.19.1 is that the former was
163 accidentally breaking the ABI of MSG, while the later is restoring
166 S4U and kernel APIs will still evolve until SimGrid4, with one-year
167 deprecation warnings as currently. In fact, cleaning up these
168 interfaces and converting them to snake_case() is one release goal of
169 v3.20. But don't worry, we are working to smooth this upgrade path.
171 In summary, new projects should start with S4U to benefit of the
172 future, but old MSG projects should still be usable with no change.
174 ----------------------------------------------------------------------------
176 SimGrid (3.19) Released March 20 2018
178 The Moscovitly-cold Spring release.
180 SMPI user-visible improvements:
181 - Support of MPI_Type_create_subarray
182 - Unimplemented MPI calls now abort (instead of issuing a warning)
183 Some harmless calls around errhandlers still only issue a warning.
184 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
185 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
186 is now possible (mmap privatization was refactored to this end).
187 Adding new SMPI instances should also be technically doable now.
189 SMPI internal cleanups:
190 - Removed index notion from SMPI -- no more "getPid() - 1"!
191 In tracing files, actors are called rank-<PROCESS-ID> but these are
192 NOT the real ranks (tracing several communicators is not supported yet).
193 - Deprecate smpi_process_index() (will be removed in 3.22)
194 - Better testing on classical proxy apps, through an external project:
195 https://github.com/simgrid/SMPI-proxy-apps/
198 - Use a graphical TOC to make it easier to find the documentation you need
199 - Revamp the MSG tutorial
202 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
203 - Remove parameter of MSG_process_killall().
204 Resetting the PID was bogus anyway (several actors could have the same PID).
207 - Execution->setHost() can be called after start(), to migrate it.
208 - Comm::test_any() is now implemented.
209 - s4u::Actor now has onCreation() and onDestruction() signals.
210 - Install some kernel header files for the users' plugins and more.
211 Warning, their API is really not stable yet.
212 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
213 include/simgrid/kernel/routing/*Zone.hpp
216 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
218 Build System and other cleanups:
219 - Remove unused run-time parameter "tracing/onelink-only".
220 - SimGrid now works with intel compilers (MC still buggy)
221 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
222 - tesh kills subprocesses on timeout (but not on windows)
225 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
228 - #194: Feature request: simgrid::s4u::Comm::test_any()
229 - #245: migrating an actor does not migrate its execution
230 - #253: Feature Request: expose clusters as objects
231 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
232 - #255: Tesh broken on Windows
233 - #256: Modernize FindSimGrid.cmake
234 - #257: Fix (ab)use of CMake install
236 ----------------------------------------------------------------------------
238 SimGrid (3.18) Released December 24 2017
240 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
242 MSG IS NOW DEPRECATED. Please use S4U for new projects.
243 - Support for MSG should not be removed from SimGrid before 2020,
244 but future evolutions will be limited to the integration of
245 user-provided patches.
248 - s4u::onDeadlock() signal, run before stopping the simulation.
249 - s4u::Exec: asynchronous executions (abstraction of background
251 - s4u::Host->getLoad() returns the achieved speed in flops/s
254 - Link::name() is deprecated and will be removed in v3.21, use
255 Link::getCname() instead.
256 - Mailbox::getName() changed to return a std::string, use
257 Mailbox::getCname() instead to get a char*.
258 - Storage::getName() changed to return a std::string, use
259 Storage::getCname() instead to get a char*.
260 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
261 to have both allocation and deallocation of the map in user space.
264 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
265 The amount of flops remaining to do is a vector, not a scalar.
266 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
267 name implies on both sequential and parallel tasks.
268 - Both changes fix GitHub's #223 using PR #237 as a basis.
269 Thanks Michael Mercier.
270 - Most examples were converted to S4U and hidden elsewhere at tests.
271 Rationale: we still want MSG to work; we want newcomers to use S4U.
274 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
275 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
278 - Switch to the faster dlopen privatization mechanism by default
279 - Documentation improvments
282 - Live migration is getting moved to a plugin. Dirty page tracking is
283 the first part of this plugin. This imply that VM migration is now
284 only possible if one this function is called:
285 - C/MSG: MSG_vm_live_migration_plugin_init()
286 - C/C++: sg_vm_live_migration_plugin_init()
287 - Java: Msg.liveMigrationInit()
288 For C and C++, "simgrid/plugins/live_migration.h" also has to be
292 - Define class simgrid::xbt::Path to manage file names.
293 - Removed unused functions:
294 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
295 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
296 xbt_graph_export_graphviz()
297 - xbt/str.h: xbt_str_join()
298 - Remove unused datatypes:
299 - xbt/heap.h: use std::priority_queue or boost::heap instead
300 - xbt/swag.h: use boost::intrusive::list instead
303 - New link_energy plugin for the consumption of the links.
304 - All of the operations on files and storage contents have been
305 packaged into a plugin (src/plugins/file_system). The current
306 public interface can be found in
307 include/simgrid/plugins/file_system.h
308 To use these functions you now have to initialize the plugin by
309 calling MSG_storage_file_system_init() just after calling
310 MSG_init() or sg_storage_file_system_init() just after creating
314 - Remove the undocumented/untested tag <include>
317 - Remove viva specific tracing as the tool is no longer maintained
320 - #248: Kill of finished processes leads to segfault
321 - #240: xbt_cond_wait_timeout should gracefully return for C apps
322 - #239: please implement signal s4u::onDeadlock()
323 - #230: segfaults when exit() without run()
324 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
325 - #223: MSG_task_get_flops_amount() not working with parallel tasks
326 - #222: Actor::kill() doesn't really kill and segfaults
327 - #221: odd LMM warning when killing an actor
328 - #120: Memory leak when the processes are forcefully killed
330 ----------------------------------------------------------------------------
332 SimGrid (3.17) Released October 8 2017
334 The Drained Leaks release: (almost) no known leaks despite the tests.
336 Even very long simulations will preserve your memory: our testsuite
337 stresses SimGrid for over 45mn on fast machines for a coverage of
338 over 80%, and there is only one single known leak, of about 4kb.
342 - Comm.detach(): start and forget about asynchronous emission. A cleanup
343 handler may be given to free resources if the comm cannot be completed.
344 - this_actor::send(mailbox) is now mailbox->put()
345 - New: simgrid::s4u::Comm::wait_all()
346 - New: Host.execute() for remote executions.
349 - Unused option network/sender-gap is removed.
352 - Deprecate MSG_task_isend_with_matching(): This unused feature
353 really complicates our internals. Will be removed in v3.20.
356 - Improved context termination. It is now possible to free resources, even
357 when a process is forcibly killed.
358 - ContextBoost: add support for Boost versions above 1.61.
361 - Bring back run-time option --cfg=exception/cutpath to remove exception
363 - Removed unused functions:
364 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
365 xbt_str_rtrim(), xbt_str_trim().
366 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
369 - Removed header files obsolete since SimGrid 3.12:
370 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
372 - #3: SD_exit should be made optional
373 - #120: Memory leak when the processes are forcefully killed
374 - #159: Threading test regression in Actor refcounting
375 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
376 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
377 - #186: Actor::killAll() segfaults if some process is blocked on wait()
378 - #191: VM migration and pstate
379 - #192: Updating the energy consumptions of all hosts crashes with VMs
380 - #195: All actors have PID=0 in the logs
381 - #204: Sometimes segfault with thread contexts and mmap privatization
382 - #222: Actor::kill() doesn't really kill and segfaults
383 - #225: Actor::kill() doesn't really kill when victims are doing a join()
385 ----------------------------------------------------------------------------
387 SimGrid (3.16) Released June 22. 2017.
389 The Blooming Spring Release: developments are budding.
392 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
393 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
394 - Java JAR file should be finally fully working on Mac OSX too.
396 XML platforms: Switch to platform v4.1 format.
397 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
398 - <zone> can be used as a synonym for the now deprecated <as>
399 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
400 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
401 - <actor> can be used as a synonym for the now deprecated <process>
402 - state_file and avail_file periodicity is now easier to express
403 (check the documentation) the old behavior should still work.
404 * Storage: not backward compatible, but it seems that nobody ever used it
405 - Remove attribute 'content_type' of <storage_type>: was never used
406 - Make attribute 'model' of <storage_type> optional: for future usage
407 - Remove Bconnection model property: was never (in)validated. Replaced by
408 the maximum of the read and write bandwidth as a resource constraint.
411 - New and Backwards Compatibility break:
412 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
413 When one wants to get the list of tasks whose states have changed during a
414 simulation round, s/he has to allocate and free a dynar and use it as
415 argument to this function. The former SD_simulate (double how_long)
419 - Allow multicore VMs, with the correct sharing (unless you overcommit)
420 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
423 - New function to update the consumption of all hosts at once.
424 - Fix the model for multi-core hosts, linear in the amount of busy
425 cores with an abnormality for pIdle (see paper).
426 BUG: Still not satisfactory for ptask on mono-cores.
429 - The netzone are now available from the MSG API.
430 The old names still work, but are now deprecated.
433 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
434 - New option: smpi/keep-temps to not cleanup temp files
435 - New option: smpi/shared-malloc-blocksize . Relevant only when global shared
436 mallocs mode is used, allows to change the size of the fake file used
437 (default 1MB), to potentially limit the number of mappings for large runs.
438 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
439 - Fortran ifort and flang compilers support
440 - New RMA calls supported (experimental):
441 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
442 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
443 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
444 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
445 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
446 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
449 - New: this_actor::isMaestro() returns whether we are in kernel mode.
450 - Behind the scene, ActivityImpl are now automatically refcounted.
451 Removing this nasty bottleneck will greatly simplify our internals.
454 - Replay: New function xbt_replay_action_get():
455 Retrieve the function previously associated to an event type.
456 - DROPPED FUNCTION: xbt_str_varsubst()
457 - DROPPED MODULE: strbuff. We don't need it anymore.
458 - DROPPED MODULE: matrix. We don't need it anymore.
459 - DROPPED MODULE: lib. We don't need it anymore.
461 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
463 SimGrid (3.15) Released March 22. 2017
465 The Spring Release: continuous integration servers become green
466 We fixed even the transient bugs on all target architectures:
467 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
468 Windows, FreeBSD, NetBSD.
470 - Rename NetCards to NetPoints.
471 This was intended to help NS3 users, but that's not a netcard.
472 That's a point in the routing algorithm, let's avoid wrong simplifications.
475 - New: MSG_process_yield(). Stop and yield to other processes.
476 - New: MSG_process_daemon(). Daemon processes are automatically killed
477 when the last non-daemon process terminates
478 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
479 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
480 to make room for the upcoming network energy plugin.
481 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
484 - Ensure that an actor can kill itself with Process::exit()
485 - Kill the obscure NativeException. Nobody want to survive the issues
486 it denotes, so use JniException that is a RuntimeException (not to
487 be caught explicitly).
488 - Partial bug fix in initialization. SimGrid flags on command line were
489 consumed at C level but stayed in the original Java String[] args.
490 This could mess users' args[i] if SG flags were not put at the end of
492 The SimGrid flags are now removed from the Java arguments. However,
493 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
494 if args.length is greater than the number of YOUR OWN ARGUMENTS.
495 It might be if you have --log or --cfg flags in the command line.
496 - Fix numerous memleaks all around the place. In particular, around VMs.
500 - s4u::onPlatformCreated: right before the simulation starts
501 - s4u::onSimulationEnd: right after the main simulation loop
502 - s4u::onTimeAdvance: right after a clock change (time is discrete)
503 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
504 event from the availability_file changes the avail speed.
505 - Links are now usable from s4u
506 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
507 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
508 - New examples: The conversion of MSG examples to S4U has begun
509 - Actors: create, kill, migration, and suspend
510 - Applications: master-worker and token-ring
511 - Action replay: communications and storage
512 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
515 - Backwards Compatibility breaks
516 - The SD_route_get_list and SD_route_get_size functions have been removed.
517 They are replaced by the unique
518 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
519 The route from src to dst is built in the links dynar whose size can
520 be obtained with xbt_dynar_length.
521 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
522 sg_host_route_bandwidth and sg_host_route_latency.
523 Macros ensure the backwards compatibility, but you should fix your code
526 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
527 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
528 - Bump our claim of support from MPI 1.1 to MPI 2.2.
529 We don't support 100% of it, but it should be enough. Contact us if not.
530 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
531 naive and probably inaccurate)
532 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
533 mapped onto a unique small file using some system magic.
534 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
535 - Fortran: cleanups, fixes, support of user-added operations
536 - RMA: MPI_Accumulate are applied in correct order.
537 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
538 - Better support for MPI_IN_PLACE
539 - Support for MPI_Win attrs and keyvals.
540 - Support MPI_Comm_group_create, MPI_Type_size_x
541 - MPI ops should be properly applied to all allowed types
542 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
543 algorithm by a non failing one from mpich.
546 - Kill the fifo data container: we don't use it anymore.
549 - A <cluster> can now be created with different speed values to represent pstates.
550 see examples/platforms/energy_cluster.xml for an example.
552 SimGrid (3.14.159) Released December 28. 2016
554 The previous release was only a crude approximation of a Pi release;
555 this one should be a bit better.
557 - Revert a last minute change that broke on Mac OSX.
558 - Fix the changelog and NEWS files.
559 - Improve the documentation, in particular of the routing module.
560 - Integrate some patches flying around in Debian.
562 SimGrid (3.14) Released December 24. 2016
565 * The whole documentation was reworked and reorganized. There is
566 still much room for improvement here, but we're on it.
569 * We now need python3 for our testing infrastructure.
570 * Model checking needs libevent
573 * Now works on FreeBSD
576 * Privatization now works on FreeBSD
578 * Privatization is now activated by default in smpirun
579 You can turn it off with -no-privatize if something goes wrong.
581 * Call-location tracing for SMPI.
582 You can add the exact location (filename / linenumber) of an MPI call to
583 your trace files and slow down or speed up the simulation between two
584 consecutive calls by using an adjustment file (see the documentation).
586 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
587 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
588 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
589 values were ignored in some cases. The timings of these functions can now
590 be significantly different.
592 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
593 smpi/running-power is renamed to smpi/host-speed
595 * smpi/grow-injected-times option to enable or disable multiplication of the
596 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
597 make simulation less precise (but also much faster).
599 * smpirun script should be (much) faster for large deployments.
601 * SMPI tracing: fixed issue with poor matching of send/receives.
603 * Replay: Fix broken waitall
605 New functions and features
606 * MSG_parallel_task_execute_with_timeout, to timeout computations.
608 Dropped / renamed functions and features
609 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
610 - MSG_mailbox_is_empty() -> Mailbox::empty()
611 - MSG_mailbox_front() -> Mailbox::front()
612 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
613 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
614 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
615 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
616 * Don't pass the free_f parameter to property related functions:
617 - MSG_host_set_property_value()
618 - MSG_as_router_set_property_value()
619 - MSG_storage_set_property_value()
620 * VM properties. Since msg_vm_t are msg_host_t, just use
621 MSG_host_get_property_value() and friends
622 * VM I/O related things:
623 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
624 - Unimplemented save/restore methods
625 * MSG_as_router_get_property_value() was redundent with
626 MSG_environment_as_get_property_value().
627 - Removed MSG_as_router_*propert*() functions
628 - Added MSG_environment_as_set_property_value() for consistency
629 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
630 Well, they are still there for now, but deprecated with a warning.
631 Please switch to xbt_dict_new_homogeneous() before this is removed
633 * Task affinity. Its intended behavior (that was very badly tested
634 and probably not really working) was deceiving what most users
635 would have hoped here.
636 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
637 * Option network/coordinates is now useless and should be dropped.
640 * Added option storage/max_file_descriptors to allow more than 1024 files opened
643 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
644 * Added Dragonfly topology support
647 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
648 EiBps, EBps, Eibps, Ebps for bandwidth.
649 They may become useful to some lucky ones.
652 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
653 * ex/app_tokenring: new example, very similar to the MSG Token Ring
654 * ex/async_waitAll: new example, on asynchronous communications
657 * Memory usage should be decreased for simulations with a large number
658 of processes. This also helps for SMPI.
660 SimGrid (3.13) stable; urgency=low
662 The Half Release, a.k.a. the Zealous Easter Trim.
664 - We removed half of the lines, that were mostly experimental cruft.
665 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
666 (not counting blanks and comments -- according to openhub.net)
667 - The internals are now compiled in C++ (and will soon be clean C++)
668 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
669 59 klines of C, etc. We added only 29 klines of C++ in replacement.
671 * Backwards Compatibility breaks
672 - Removed Lua simulation bindings (switch to C or Java for that).
673 Lua can still be used to describe platforms
674 - Removed Java kernel plug-ins.
675 Will be reintroduced after the ongoing major internals reorg.
677 - the following functions were removed.
678 They were too specific and should be reimplemented in a generic
679 way, with filter function.
680 - MSG_task_listen_from_host
681 - MSG_mailbox_get_count_host_waiting_tasks
682 - MSG_mailbox_put_with_timeout was removed.
683 Please use MSG_task_send_with_timeout instead.
685 - the SD_application_reinit function was removed. It has been a noop for a while.
686 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
687 replaced by a more flexible mechanism.
688 - The following functions thus do not exist anymore
689 - SD_workstation_get_access_mode
690 - SD_workstation_set_access_mode
691 - SD_workstation_get_current_task
692 - Basic estimation functions have been removed but can easily be replaced
693 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
694 - SD_workstation_get_computation_time => amount / sg_host_speed()
696 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
697 Use VM.getSpeed()*load/100 for the legacy behavior.
699 - option enable_tracing was removed. It was not doing anything for a while.
700 - In the ModelChecker:
701 - the model-checker now ptraces the model-checked process which means
702 you cannot use a debugger on the latter anymore (we might make this
703 optional in the feature);
704 - removed soft-dirty page tracking;
705 - remove model-checked side snapshot management,
706 MC_snapshot() and MC_compare_snapshot();
707 - keep the MC_cut() function as a stub (it was not really working
708 in the previous release).
711 * All options are consistently in kebab-case. Old names are kept as alias.
714 * Switch to platform v4 format.
715 - Rename from 'power' to 'speed' the attributes describing the amount of
716 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
717 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
718 - In <host> and <link>, attributes availability and state are gone.
719 It was redundent with state and availability traces, and with peak values.
720 - In <cluster>, attributes availability_file and state_file are gone.
721 It was too complex and unused.
722 - Kill <gpu>. Was not doing anything.
723 - The DOCTYPE points to the right URL:
724 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
725 (the file at this address now documents the changelog since its v1)
726 - A warning is emitted for unit-less values (they are still accepted).
727 - speed. Default: 'f' or 'flops'. Also defined:
728 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
729 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
730 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
731 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
732 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
733 - latency. Default: 's' second. Also defined:
734 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
736 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
737 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
740 * s4u::Host is now the preferred public interface to the Host features.
741 sg_host_* functions are C bindings to the exact same behavior
742 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
745 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
746 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
747 They were deprecated since a few years, and probably did not even compile anymore.
750 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
751 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
752 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
753 encouraged to update their codes. A list of the modified functions can be found at the end of
754 include/simgrid/simdag.h
757 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
758 the lambda or closure passed as a parameter will run in kernel mode.
759 All the callback functions should be rewritten to that interface at some point.
762 * Reorganizing and cleaning the internals all around the place.
765 * Remove old default barrier/bcast buggy algorithms (see #18407)
766 * Various bug fixes to handle more codes
767 * Remove the need for the --foreground option of smpirun (it is still
768 accepted for backward compatibility).
771 * Kill the setset data container: MC don't use it anymore.
772 * Kill the queue data container: it made more sense with GRAS.
773 * Kill the xbt_peer_t data type: it's useless without GRAS.
774 * Kill rm_cb feature of config sets: it was never useful.
775 * Kill graphxml parsing feature. It was not used.
776 * Kill the deprecated code protected by XBT_USE_DEPRECATED
778 - xbt_dynar_sort_strings(), when the content is char*
779 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
780 They throw exceptions on invalid input;
781 * C++ support for declaring CLI flags (simgrid::config::Flag);
782 * class for abstracting different signal backends (simgrid::xbt::signal).
783 with no external dependencies (we need very simple signals).
786 * refactoring and cleanup of the code;
787 * ongoing process to cleanly separate the model-checking algorithms
788 from the code model-checking support.
790 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
792 SimGrid (3.12) stable; urgency=low
794 The Facelift Release.
797 * Require g++ v4.7 at least to not speak prehistorical C++.
798 * Require Boost 1.48 (for signal2 component).
799 * Java must be version 7 at least when activated.
800 * Builds on Windows again (including Java bindings).
801 * Tracing is now always enabled (no way to turn it out)
802 * Remove GTNetS. It was not working anyway.
803 * Various cleanups in the cmake scripts.
804 * Move headers around to sort them out on installed systems:
805 - instr/instr.h -> simgrid/instr.h
806 - instr/jedule/* -> simgrid/jedule
807 - simdag/datatypes.h was removed
808 - simdag/simdag.h -> simgrid/simdag.h
809 - msg/datatypes.h was removed
810 - msg/msg.h -> simgrid/msg.h
813 * Interface improvement:
814 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
815 - Sanitize the interface in MSG_task_ module:
816 - Merge two functions that were close enough but misleading:
817 set_compute_duration(t) -> set_flops_amount(t)
818 get_remaining_computation(t) -> get_flops_amount(t)
819 - set_data_size(t) -> set_bytes_amount(t)
820 get_data_size(t) -> get_bytes_amount(t)
821 - Massive cleanups in the functions related to the energy
822 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
823 - New: MSG_host_get_pstate()
825 - msg/energy/onoff: switching hosts on and off
827 * Interface improvement:
828 - Rename Host.isAvail() to Host.isOn()
829 - Rename Process.currentProcess() to Process.getCurrentProcess()
830 - Rename Task.setDataSize() to Task.setBytesAmount()
831 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
833 - #18874: Actually allows the GC to reclaim tasks
837 - SIMIX_process_throw: raises an exception in a remote process
838 * Refactoring: Separate sub-modules
839 - libsmx: the public interface, as libc in a real system
840 - popping: the strange dance that converts a user request into a kernel handling
841 - smx_context_*: the virtualization mechanisms that embed the user code
842 - smx_*: the handling of each simcalls
844 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
845 * Rename smx_action_t into smx_synchro_t, making explicit that these
846 things are used to synchronize processes with their environment.
847 For example, a communication is a sort of synchronization involving
848 the communicating processes (that may block until the exchange) and
849 the platform. The same can be said from computations, etc.
851 - #18888: segfault when a process ends before its kill_time
855 - 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.
856 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
857 - MPI_Type_set_name, MPI_Type_get_name
858 - MPI_*_c2f and MPI_*_f2c functions
859 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
860 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
861 - Activate a lot of new tests from the mpich 3 testsuite
863 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
864 - InfiniBand network model added: Based on the works of Jerome Vienne
865 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
866 - When smpi/display_timing is set, also display global simulation time and application times
867 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
868 * Collective communications
869 - 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
870 - 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/).
871 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
873 - Replay now uses algorithms from wanted collective selector
874 - Replay can be used with SMP-aware algorithms
875 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
876 - 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.
878 - [#17799]: have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
879 - Correctly use loopback on fat-tree clusters
880 - Asynchronous small messages shouldn't trigger deadlocks anymore
881 * Energy/DVFS cleanup and improvement
882 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
883 - new: smpi_get_host_pstate()
886 - "Full" network optimization flag was broken since Surf++
887 - Better handling of precision flags in maxmin
888 - Fix bug causing sometimes "Impossible" errors
889 - Properly pass cluster properties to included hosts
890 * Improvement of the Energy plugin.
891 - Always update the consumption before returning that value
892 - New property: watt_off to denote the dissipation when the host is off
893 - New functions getWattMinAt and getWattMaxAt to retrieve the
894 dissipation of pstates that we are not currently at.
895 * Java: class NetworkLink renamed to Link
896 * New function: simcall_process_get_kill_time()
897 * Massive rename s/workstation/host/
898 - That's intrusive, but that's good for the project consistency. Sorry.
899 - Change config option "workstation/model" into "host/model"
902 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
903 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
904 - Add two new log appenders: rollfile and splitfile. Patch by Fabien Chaix.
905 - xbt_dirname and xbt_basename for non-POSIX systems
907 * The model checker now runs as a separate process.
908 * The model checker runs must now be launched with the new simgrid-mc program.
909 * Record/Replay: the MC can display a textual representation of a path in the
910 execution graph. It can then be replayed outside of the model checker.
912 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
914 SimGrid (3.11) stable; urgency=low
919 * Normalizing pointers addresses tool for better diff between logs
922 * Add cloud examples using new VMs
923 - examples/msg/cloud/two_tasks_vm.tesh
924 - examples/msg/cloud/simple_vm.tesh
925 - examples/java/cloud/cloud.tesh
926 - examples/java/cloud/migration/migration.tesh
927 * Add java surf examples:
928 - examples/java/surfPlugin/surf_plugin.tesh
929 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
930 - examples/java/surfCpuModel/surf_cpu_model.tesh
931 * Add SMPI+MSG example:
932 - examples/smpi/smpi_msg_masterslave/
937 - msg task destroy cancel test
938 - msg_host on/off test
939 * Move all tests in testsuite to teshsuite (adding tesh files)
940 * Restructure teshsuites
941 - one folder for each kind of test
942 * Restructure AddTests.cmake
944 - structure the order of tests (with sections)
947 * Add virtual machine
948 - creation of a VM on a PM
949 - migration of a VM from a PM to another PM
951 - MSG_process_join(msg_process_t process, double timeout)
952 - msg_bar_t MSG_barrier_init(unsigned int count)
953 - int MSG_barrier_wait(msg_bar_t barrier)
954 - void MSG_barrier_destroy(msg_bar_t barrier)
955 - msg_as_t MSG_environment_as_get_by_name(const char * name)
956 * New option "msg/debug_multiple_use" to help debugging when a task is used
959 - Improvements and finalization of MSG_storage, MSG_file APIs and their
961 - Increase code coverage in test suites
965 * Protect context stacks against stack overflow. The number of protected memory
966 pages allocated on the top of each stack (1 by default) can be configured
967 with the new command line option --cfg=contexts/guard_size:<value>.
968 * Simcalls are now generated by a python script that
969 - generates files included by SimGrid
970 - checks that all the functions exist, and proposes prototypes
972 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
973 host_set_data, host_get_data
975 - simcall_process_join(smx_process_t process, double timeout)
976 * Fix bug where sleeping processing could not be suspended.
979 * Translate surf models from C to C++
980 - Generic classes for all models: Model, Resource, Action
981 - A generic interface for each kind of model (CPU, Network, Storage
982 Workstation, WorkstationVM)
984 * Translate surf routings from C to C++
985 * Add callbacks using sigc++ or boost::signals2
986 - Add callback functions for resource creation/destruction
987 - Add callback functions for action state change
988 - Handle Energy as a plugin
989 * Replace swag by boost::intrusive
990 * Add new routing models for clusters. For documentation, see
991 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
992 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
993 parameters for cluster tag
994 - Fat trees, with topology="FAT_TREE" and
995 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
996 - see examples/platforms/torus_cluster.xml and
997 examples/platforms/fat_tree_cluster.xml
1001 * Hostfiles support host:nb_processes construct to deploy several processes on
1003 * Collective communication algorithms should not crash if used with
1004 improper number of nodes and report the error.
1005 * SMPI now partially supports MPI_Topologies: MPI_Cart_create, MPI_Cart_shift,
1006 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
1007 MPI_Dims_create, MPI_Cart_sub are supported.
1008 * New interface to use SMPI programmatically (still depends on MSG for
1009 some parts, see examples/smpi/smpi_msg_masterslave):
1010 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
1014 * Global variables privatization in MPI executables is now performed at runtime
1015 with the option smpi/privatize_global_variables (default:no).
1016 Limitations: Linux/BSD only, with mmap enabled. Global variables inside
1017 dynamic libraries loaded by the application are not privatized (static
1018 linking with these libraries is advised in this case)
1021 * Options defined in XML work correctly now.
1024 * New cmake option, enable_lib_in_jar, to control whether native libraries are
1025 copied into simgrid.jar or not (ON by default). Use this option if you want
1026 to reduce the size of the installed simgrid.jar, *and* the native libraries
1027 are kept installed elsewhere.
1028 * Surf binding with SWIG (code generated in maintainer mode only):
1029 - plugin to handle callbacks
1030 - CPU model only for the moment
1033 * Supernovae build mode is definitively removed. It was used to improve
1034 inlining and inter-module optimizations. It is nowadays superseded by
1035 link-time optimizations commonly available in compilers.
1036 * Update ns-3 find lib. Bindings for ns-3 should work again now.
1037 * Add boost dependency for surf++
1038 * Add new macro for tests
1039 - ADD_TESH(name <tesh_args>)
1040 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
1044 - xbt_bar_t XBT_barrier_init(unsigned int count)
1045 - int XBT_barrier_wait(xbt_bar_t barrier)
1046 - void XBT_barrier_destroy(xbt_bar_t barrier)
1047 * Make the xbt_os_time module public
1049 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1051 SimGrid (3.10) stable; urgency=low
1053 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
1056 * Reintegrate Java to the main archive as desynchronizing these
1057 package is not acceptable anymore (Java is now considered stable)
1058 * Add explicit synchronization facilities through semaphores
1059 * Bug fix: Task.setDataSize() only changed the C world, not the value
1060 cached in the Java world
1063 * Dramatically change the way files are handled. API and internals changed, but
1064 this part of MSG was not considered as production grade either.
1065 * Add explicit synchronization facilities through semaphores
1066 * Add a new function MSG_host_get_process_list()
1067 * Preliminary DVFS support (see examples/msg/energy/ for details)
1070 * SMPI is now included directly in the libsimgrid as the windows
1071 linker doesn't force us on splitting it anymore.
1072 * Improvements of the SMPI replay tool:
1073 - Most of the collective communications are now rooted in the same process as
1074 in the original application.
1075 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
1076 used until now). Multiple data types can now be used in a trace.
1077 - The replay tool now supports traces produce either by TAU or a modified
1079 - Bug Fix: the compute part of the reduce action is now taken into account.
1080 - Gatherv collective is now supported
1081 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
1082 of smpirun outputs time independent traces for the current run. One file
1083 is created per process. If too many processes are simulated, this behavior
1084 can be changed to one file for all processes by using the
1085 tracing/smpi/format/ti_one_file flag
1086 * smpirun generates the host file if needed (with given host count and platform)
1087 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
1088 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
1089 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
1090 --cfg=smpi/coll_selector:(mpich/ompi)
1091 * Support for new functions: MPI_Issend, MPI_Ssend, Commutative operations in
1093 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
1094 happening inside a collective SMPI call.
1095 * Fix the behavior of complex data types handling.
1096 * Make MPI_Wtime another synchronization point to take computations into
1098 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
1099 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
1100 * Add all missing Fortran bindings, SMPI should work with Fortran 90
1101 (no privatization of global variables yet)
1102 * Preliminary DVFS support (see examples/smpi/energy/ for details)
1105 * Verification of liveness properties is now available for SMPI applications
1106 (in addition to MSG applications)
1107 * Bugged examples using SMPI in examples/smpi/mc/
1108 * Add --cfg=model-check/visited option. Allows the verification of infinite
1109 programs. Detection of loops in the execution thanks to the system state
1110 comparison and reduction of the state space to explore. Can be combined with
1111 DPOR for safety properties.
1114 * Allow to change SimGrid configuration (see --help) within the code
1115 thanks to SD_config() as it can be done in MSG.
1116 * Add a new function SD_task_set_amount() upon user request.
1119 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
1120 * Remove rule based routing (no more PCRE dependency)
1121 * Add a limiter_link option to cluster tag, to specify a maximum reachable
1122 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
1123 * Add a loopback_bw and loopback_lat options to cluster tag.
1124 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
1125 routing. Now peers are encapsulated in an AS and have their own private
1126 router but this is transparent.
1129 * Our own implementation of getline is renamed xbt_getline, and gets
1130 used even if the OS provide a getline(). This should reduce the
1131 configuration complexity by using the same code on all platforms.
1132 * New type: xbt_cfg_elm_boolean.
1133 * Allow to use yes/no for boolean configuration options in the command line.
1134 * Allow to disable SimGrid cleanups at exit from command line option.
1135 There are situations where one may want a simulation to end with an exit.
1136 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1137 annoying when scripting around the simulator. Adding a
1138 --cfg=clean_atexit:no allows to circumvent this issue.
1141 * Lots of memory leaks were corrected in this release.
1142 There is no known memory leaks anymore, in all of our 600+ tests.
1143 * New command line option --version, to get SimGrid version information.
1144 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1146 * Supernovae builds are deprecated, and expected to be removed in the next
1149 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1151 SimGrid (3.9) stable; urgency=low
1153 The Grasgory release: GRAS is really dead now.
1155 * Complete overhaul of the internal host structures scheme.
1158 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1159 was considered as experimental and badly maintained since then.
1160 * Keeping it was thus a trap to our potential users, that could take
1161 it instead of MSG or SMPI by mistake despite is pity state.
1162 * GRAS seems to have very few users (if any), and no one volunteered
1163 to maintain it further. It also induces a lot of XBT code (for
1164 portability sake), that must be maintained too.
1165 * For all these reasons, we killed GRAS. If someone wants to revive it
1166 in the future, don't cry, our git history still remembers of GRAS.
1169 * Major overhaul. Merge our documentation again as time proved that
1170 splitting it was really not helping our users.
1171 * Further improve the developer documentation to help newcomers
1172 hacking on SimGrid itself. The user documentation (and in
1173 particular, the beginner documentation) is still in a sorry state.
1176 * Now works on Windows too!
1177 * Much more extensive test suite, from MPICH
1180 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1181 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1182 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1183 * Bug fix: let task be scheduled when the last dependency to be solved is
1184 a control dependency.
1185 * Remove SD_load_environment_script function.
1186 Use the C sg_platf function if you want to declare a platform
1190 * New function: MSG_process_get_number()
1191 * Old function documented: MSG_config()
1192 * Remove MSG_load_platform_script function
1193 Use the C sg_platf function if you want to declare a platform
1197 * Change the default value of the TCP_gamma constant (maximal size of TCP
1198 congestion window) to a more realistic 4MiB value. If you notice changes in
1199 your simulation results, you can fall back to the previous 20k tiny window
1200 by adding --cfg=network/TCP_gamma:20000 on command line.
1201 * (Hopefully) fix a bug wrt periodic availability and state traces
1202 * Bug fix: use default values at start when first event in availability/state
1203 trace is not at time 0.
1206 * remove the "new_" part of function name sg_platf_new_trace_connect
1207 (resulting in sg_platf_trace_connect), since it does not create
1211 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1212 use case where it's really mandatory, and maintaining it was a pain in
1214 * New: xbt_fifo_search(), search an item with a user-provided
1215 comparison function instead of dumb pointer comparison.
1218 * Fix the lua deployment:
1219 Use `simgrid.init_application()` before deployment instead of
1220 `simgrid.msg_register_application()` after.
1223 * Transfer the tracing files into the corresponding modules.
1225 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1227 SimGrid (3.8.1) stable; urgency=low
1229 The "we are told that some people want to also *install* the simgrid
1232 * Add missing file "tesh.1" to the archive.
1234 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1236 SimGrid (3.8) stable; urgency=low
1238 The Psssshiiiit release: SimGrid jumps into the Cloud.
1241 * Add an experimental interface to manipulate VMs. They are mainly
1242 process groups with very few intrinsic semantic, but they should
1243 allow you to build the semantic you want easily.
1244 * New function: MSG_host_set_property_value()
1245 * New function: MSG_process_on_exit(). To clean memory in all cases.
1246 * Bug fixes that made the host (and link) failures unusable.
1247 * Add a way to auto-restart process when the host in which they are
1248 executing comes back (ON_FAILURE="RESTART" on deployment file,
1249 MSG_process_auto_restart_set).
1250 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1251 please stop using the old ones, they are DEPRECATED.
1253 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1254 Please use MSG_init() instead. (reducing the amount of entry
1255 points in the library helps us).
1256 * Make it impossible to link against the wrong version of the lib
1257 * Deprecate MSG_clean(). No need to call it anymore.
1258 * Function MSG_get_host_number() is not deprecated anymore.
1261 * Split the doc into a user guide and a reference guide.
1262 * Start a developper guide to help people hacking on SimGrid.
1265 * Enable tracing by default. This modules rocks you should use it.
1266 * Remove option custom_flags. Now use environment variables CFLAGS
1268 * Use default cmake things to detect lua instead of home grown ones.
1269 * New option "enable_mallocators" to disable mallocators, for debugging
1270 purpose ("on" by default).
1273 * Bug fixes around the resource failures: don't let the processes
1274 survive the host they are running onto.
1275 * Add an interface to auto-restart processes when the host in which they are
1276 executing comes back.
1277 * Ensures that SIMIX_clean is called automatically. It's not part of
1278 the public interface anymore (bindings should be updated).
1281 * Bug fix for when SD_Simulate is called with a positive value: be careful
1282 when comparing doubles. Sometimes they are different for non significant
1284 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1285 parallel task whose initial work is distributed among host according
1286 to the Amdahl's law. Such tasks are created with a parameter alpha
1287 that corresponds to the non-parallelizable part of the computation.
1288 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1289 between two sets of workstations assuming a 1D block distribution (each
1290 workstation owns a similar share of data) on both sides.
1292 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1293 Data redistribution will be automatically scheduled once parent and child
1294 are both scheduled. The filling of computation_amount and
1295 communication_amount structures is now done seamlessly thanks to the chosen
1297 * New function SD_workstation_dump to display various information
1298 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1299 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1300 bandwidth on the route onto which the task is scheduled and the amount of
1302 To divide the nominal bandwidth by 2, the rate then has to be:
1303 rate = bandwidth/(2*amount)
1304 * Compute tasks that have failed can now be rescheduled and executed again
1305 (from their beginning)
1306 * Increasing source code coverage (src/simdag is now covered at 95.8%
1310 * Re-implement time-independent trace replay using SMPI (at the
1311 smpi_smp_* level) instead of MSG. This should replace
1312 examples/msg/actions/actions.c
1313 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1314 hindexed and structs)
1315 * Implement the exchange of non-contiguous data.
1316 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1317 * Correct behavior of smpi/sender_gap and set its default value to 0
1318 * Add option to asynchronously send small messages to allow better
1319 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1320 specifies the size in bytes under which messages will be asynchronously sent.
1321 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1322 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1325 * New C interface to define a platform: XML is now optional.
1326 For more info, please check include/simgrid/platf.h.
1327 * New interface to define random platforms from the C:
1328 For more info, please check include/simgrid/platf_generator.h and
1329 examples/msg/masterslave/masterslave_platfgen.c
1330 * Export a sg_cmdline dynar containing all the arguments we got from
1334 * Two new tracing options for adding comments to trace file so you
1335 can track your experiments (see --help-tracing for details).
1336 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1337 * Adding the SimGrid version that generated the trace file as a comment.
1338 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1339 * Fix to avoid key clashes on Paje links
1340 * Other minor fixes related to the Paje specification
1343 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1344 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1345 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1346 * Greatly improve the robustness of mmalloc to user errors (such as
1347 using an area after freeing it, or freeing it twice)
1349 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1351 SimGrid-java (3.8.1) stable; urgency=low
1353 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1354 Warning: all methods are visible, but only some of them are
1355 implemented so far. Check the source (src/jtrace.c)
1356 for further information.
1357 * New module: org.simgrid.msg.File (SimGrid File management functions)
1358 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1359 * Change the meaning of Process.restart: now restart the process from
1360 the begining, like MSG_process_restart in C.
1361 * Add Process.setAutoRestart: handling of process restart when failed
1363 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1364 you to retrieve the properties of the processes/hosts
1365 * Deprecate Msg.clean(): you can just forget about it now.
1366 * New function Process.getCount(), that only works when compiling
1367 with the not yet released version 3.9 of the C library.
1369 * New context factory based on Coroutines. It mandates a modified JVM
1370 but then, the simulations run about five times faster, and there is
1371 no limit to the amount of processes (beside of the available memory).
1373 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1375 SimGrid (3.7.1) stable; urgency=low
1378 * Restore the prototype of MSG_process_create_with_environment() to
1379 the pre-3.7 situation by removing the kill_time argument.
1380 * Add a MSG_process_set_kill_time() function instead.
1383 * Fix weird behaviors when dealing with parallel tasks.
1386 * Simgrid is now built as a dll.
1387 * Simgrid-java now works on Windows.
1388 * Simgrid-Java is now included into Windows package.
1391 * First pre-build package for MacOSX.
1394 * Fix compilation when using MSG_USE_DEPRECATED.
1395 * Fix some compilation issues on Macs and Windows.
1396 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1397 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1399 We discovered that the Lua console is broken, but we are missing the
1400 manpower to fix it right now. The problem existed in 3.7 too, so we
1401 are not blocking the release for that. Sorry if you depended on this
1402 feature, any help would be really welcome.
1404 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1407 SimGrid-java (3.7.1) stable; urgency=low
1409 The "Java aint got to be bloated and slow" release
1412 * Various internal cleanups and performance improvement
1413 Simulations are expected to run up to twice faster or so
1414 * Make Process.kill(process) an instance method, not a static one
1415 * User processes are not java.lang.Thread subclasses.
1416 This breaks the compatibility (sorry), but previous API was
1417 brain-dead, making it impossible to have non-trivial
1418 initializations in the process constructor.
1419 * Require a full constructor per Process sub-class.
1420 Kinda breaks the compatibility (sorry), but this allows a much more
1421 efficient way to launch the processes at simulation startup.
1422 * Do not embeed our version of semaphores, java 1.5 can be considered
1423 as sufficiently prevalent for us to not dupplicate its features.
1427 * Add examples for almost every part of the API
1428 We spotted and fixed a lot of bugs in the process
1430 * New module: asynchronous communication API
1431 * New function: Process.sleep()
1432 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1433 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1434 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1436 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1438 SimGrid (3.7) stable; urgency=low
1440 The "spring cleanups (before the next Big Project kicks in)" release.
1443 * We can specify the SMPI latency/bandwidth factor with command line
1444 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1445 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1446 You can also use the "config tag" from platform file by adding this line
1447 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1448 (see "example/platforms/tag_config.xml" to use "config tag").
1449 Note that the command line supersedes the platform file configuration.
1450 * Change the correction factors used in LMM model, according to
1451 the latest experiments described in INRIA RR-7821.
1452 Accuracy should be improved this way.
1453 * Use the partial invalidation optimization by default for the
1454 network too. Should produce the exact same results, only faster.
1455 * Major cleanup in surf to merge models and split some optimization
1456 mechanisms from the core of the models. As a result you can now
1457 specify which model to use (e.g., --cfg=network/model:LV08
1458 --cfg=cpu/model:Cas01) and which optimization mode to use
1459 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1460 Incompatible combinations should err at initialization. See
1461 --help-models for the list of all models and optimization modes.
1462 * The CLM03 workstation model was dropped for simplicity because it
1463 used the deprecated CM02 network model. Use default instead.
1464 * Rename the TCP_gamma configuration option to network/TCP_gamma
1465 * Rename the coordinates configuration option to
1466 network/coordinates, and document it
1467 * Use now crosstraffic keyword instead of the terribly misleading
1468 fullduplex keyword. It is activated by default now in the current
1469 default model, use --cfg=network/crosstraffic:0 to turn it off.
1470 * Ongoing refactoring the model parsing to make XML files optional
1471 See include/simgrid/platf.h for details (still to be completed)
1474 * Major overhaul of the documentation. Almost instructive now :/
1475 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1476 functions and friends. This interface was considered as
1477 deprecated since over 2 years, it's time to inform our users that it is.
1478 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1479 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1480 use these (crappy) functions in your code.
1481 These functions will be removed soon. Stop using them now.
1482 * Deprecate MSG_get_host_{table,number}
1483 Implement MSG_hosts_as_dynar() instead.
1484 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1485 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1488 * Stabilize the parallel execution mode of user contexts
1489 * Introduce configuration variables to control parallel execution:
1490 - contexts/synchro: Synchronization mode to use when running
1491 contexts in parallel (either futex, posix or busy_wait)
1492 - contexts/parallel_threshold: Minimal number of user contexts
1493 that must be part of a scheduling round to switch to parallel
1494 execution mode (raw contexts only)
1495 * Fix bugs that prevented to use suspend/resume along with
1496 synchronization structures.
1497 * Fix bugs in process termination that lead to invalid memory access
1498 in very specific conditions.
1501 * Introduce a parallel mode for the models (controlled by surf/nthreads
1502 configuration item). In our tests, running the models in parallel
1503 never lead to any speedups because they are so fast that the gain
1504 of computing each model in parallel does not amortizes the
1505 synchronization costs, even when ultra fast futexes are used.
1506 This is released anyway because YMMV.
1509 * Performance boost by using a swag internally to compute the set of
1510 tasks that are finished and should constitute the return value of
1514 * Enable it by default now that it is considered rather stable.
1517 * Documentation of the tracing functions.
1518 * Performance gains when tracing categorized/uncategorized resource
1519 utilization by avoiding calls to get route when updating resource
1520 variables. LMM constraints are being used instead.
1521 * API changed to set task categories. Use MSG_task_set_category instead
1522 of TRACE_msg_set_task_category, and SD_task_set_category instead
1523 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1524 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1525 * Examples for link user variables added (see at examples/msg/tracing/)
1526 * Deprecated function TRACE_msg_set_process_category completely removed
1527 * Trace header updated according to the latest Paje file format
1528 * Tracing network lazy updates, no longer obligate users to use full updates
1529 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1530 * Experimental: let user code declare/set/push/pop application states for hosts
1531 * API changed to allow the manual creation of graph configuration files
1532 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1535 * Improve the API of Lua MSG bindings, using the Lua spirit.
1536 * Each simulated process now lives in its own Lua world (globals are
1537 automatically duplicated). It helps writing simulators. Will allow
1538 to run Splay programs within SimGrid in the future.
1539 * Add a Chord example in Lua, equivalent to the MSG one.
1542 * Start the implementation of a solution to express temporal
1543 properties, not only local assertions. This is still an
1544 experimental work in progress, stay clear from it to be safe.
1548 - Add new runtime parameters --help-logs and --help-log-categories
1549 to display information about supported logging parameters and
1551 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1552 don't exists anymore.
1553 * Mallocators: allow value NULL for the reset function.
1555 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1556 create homogeneous dictionaries, where all the elements share the
1557 same free function. Non homogeneous dictionaries will be
1558 deprecated in the next release.
1559 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1560 - Multi-level dictionaries are deprecated.
1562 - new function xbt_dynar_search_or_negative() that is useful when
1563 you have less than 2 million elements in your dynar and don't
1564 want of the extra complexity of catching exceptions when the
1565 element is not found.
1567 - Make xbt_os_thread module (for thread portability) public.
1568 Documentation is still to come, sorry.
1570 - Cleanups and simplifications to make it maintainable again.
1571 - Exotic features (such as memalign and valloc) were removed.
1572 - The metadata were extended and improved so that the
1573 model-checker becomes able to explore and inspect the heaps.
1574 - This may induce a performance drop when enable_model-checking is
1575 ON in cmake (even if it's not used in the simulation), but it is
1576 necessary at this point to get MC working.
1578 Turn model-checking OFF if simulation performance matters to you.
1579 Not enabling it at runtime is not enough, disable it in cmake.
1581 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1584 SimGrid (3.6.2) stable; urgency=low
1586 The "Not coding new stuff allows to polish old things" release.
1589 * New bindings to the NS3 packet level simulator (experimental)
1590 * Use the raw (efficient) execution contexts instead of the sysv
1591 (portable) ones when possible.
1592 * libpcre is now mandatory in any cases since not using it led to
1593 severe performance loss and possibly other issues
1594 * Update the XML platforms:
1595 - G5K: include the latest machine in Nancy
1596 - GridPP and LCG: new platforms
1597 * Documentation was partially updated, at least (more to come)
1599 Bug fixes, cosmetics and small improvements
1600 * Free terminated processes before the end of the simulation to avoid
1601 exhausting the memory of users having very dynamic amount of
1603 * Bug fix and cosmetics about canceling non-running tasks
1604 * Bug fix about the dot loader's issues when using libcgraph
1607 * Create an installer for windows with nsis (amd64 and win32)
1608 - Add an hello world project to illustrate simgrid project creation.
1609 - Embed libpcre into the Simgrid installer to avoid
1610 its compilation burden
1611 * The raw execution contexts should work on Apple now
1612 * Port to Windows 64 bits
1613 - Sysv contexts now have an implementation for this arch
1614 - GRAS communication features now support this arch
1615 * Drop support for borland compiler on windows
1616 - this code was not maintained, and we kinda depend on gcc nowadays
1617 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1618 * Fix portability issue on unstable ubuntu: linker became picky on
1621 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1623 SimGrid (3.6.1) stable; urgency=low
1625 The "Oops, we broke Macs too" release
1628 * Fixed contexts detection so that raw ones are used when possible
1629 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1630 leads to a strange error, with user code segfaulting sometimes when
1631 the generated code is not perfectly aligned (which is not
1632 controllable from the user side, depends on the amount of code)
1635 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1637 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1639 SimGrid-java (3.6) unstable; urgency=low
1642 * Split of every thing from simgrid v3.5 into a separate package.
1644 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1646 SimGrid (3.6) stable; urgency=medium
1648 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1651 * Bindings now constitute their own package, separated from the main one.
1652 Rationale: reduce our maintenance nightmare by reducing the module coupling
1653 They will soon be released on their own on gforge.
1655 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1656 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1658 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1659 * It's not quite deprecated for now because we have no replacement,
1660 but it may soon become the case.
1663 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1664 * Fortran: New user-level cache variable to store the rank of the running
1665 process. This improves performance by an order of magnitude.
1666 * C: New coccinelle script to automatically locate and modify global and
1667 local static variables.
1668 * Improved SMPI network model with a sender-side gap to account for multiple
1672 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1673 only say if a communication is finished, no matter whether it succeeded or
1674 failed. You can call MSG_comm_get_status() to know the status of a finished
1676 * New function MSG_task_dsend() to send a task and detach it. When a
1677 communication is detached, you are never notified of its success or failure
1678 and the memory is released automatically once it is finished. This function
1679 is useful when you don't care about the end nor the success of a
1681 * Change the prototypes of action replay. Sorry for inconvenience,
1682 but this is really more efficient this way (and to adapt your code,
1683 you just have to fix the initialization, that shouldn't be too long)
1684 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1685 ever managed to do anything useful with it.
1686 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1687 takes a dynar of communications. It returns immediately and gives the
1688 index of a finished communication (if any).
1689 * New example: a basic implementation of the Chord P2P algorithm.
1692 * New model for multi-core CPUs. You can now use the core attribute to
1693 precise the number of cores of a host. This is a basic model. Every
1694 process running on the host receives at most the power provided in
1695 the DTD (throughput<=power). Total throughput of process cannot exceed
1697 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1698 router linked by an up-link and a down-link (possibly asymmetrical).
1699 This kind of pattern allows to easily build last-mile model style platforms.
1700 Aggregating such patterns in a rule-based AS is thus the technique of
1701 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1702 * New model for Vivaldi routing. We transformed the Vivaldi network model
1703 into a Vivaldi routing model (based on the rule-based model). This allows to
1704 combine Vivaldi based latencies with last-mile platforms.
1707 * Added a check for NaN of IEEE754 infinite in the double entries of
1709 * Introduce a new context factory "raw", highly inspirated from the
1710 ucontext factory, but using manually crafted functions in assembly to
1711 do the work in an efficient manner.
1712 * Allow to change the used context factory at run time, not only at
1713 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1714 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1715 contexts at runtime (only with raw contexts or ucontexts).
1716 * Completely rewrote this module to allow parallel execution of user
1717 processes. Use --cfg=contexts/nthreads:N to execute user processes
1718 with N parallel threads (the default is 1, meaning no parallelism).
1719 * Allow to decide dynamically between sequential and parallel modes.
1720 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1721 processes in parallel only when their number is greater than or equal to P
1723 * Added a check for NaN of IEEE754 infinite in the double entries of
1727 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1728 won't output the state of processes when interrupted with Ctrl-C
1729 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1730 NULL-terminated array. This may solve backward compatibility issues
1731 due to the change to return type of SD_simulate. See also:
1732 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1733 * Add new macros with variable number of arguments.
1734 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1735 - in xbt/asserts.h: xbt_assert
1736 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1737 - in xbt/ex.h: THROWF and RETHROWF.
1738 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1740 * Change xbt_die() to accept a format string with arguments, just like printf.
1741 * New data structure: xbt_lib_t, like a dict but more general and with better
1745 * New configuration options
1746 Options triva/categorized and triva/uncategorized can be used to generate
1747 graph configuration files for Triva visualization tool.
1748 * Configuration option tracing/platform is renamed to tracing/categorized
1749 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1750 * New TRACE_user_link_variable interface:
1751 User provides the name of the link and the tracing variable to attach to it
1752 * the declaration of tracing categories must be done after the environment
1754 * simpler tracing interface, just one way to declare categories
1755 TRACE_category or TRACE_category_with_color, it is up to you
1756 * links in the trace file are again identified by their names
1757 * trace contains the full platform hierarchy exactly as declared using the ASes
1758 * Options tracing/msg/[task|process]:1 groups the process by hosts
1759 for both cases, tasks and processes must have names that are unique during the simulation
1760 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1761 * The experimental option tracing/msg/volume is deprecated
1762 its functionality may be reincorporated if needed
1764 The tracing generates a trace file with unordered timestamped events,
1765 because of the way the core simulator (surf) works. A script available
1766 at the tools directory (fix-paje-trace.sh) can be used to put the events
1767 in order. We have changed the tracing so it can generate ordered timestamped
1768 events in the final trace, but depending on the simulator (and how much time
1769 is simulated) that can lead to a huge memory utilization. It is deactivated
1770 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1772 Build Infrastructure
1773 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1774 - We are trying hard to keep the API stable, but it may happen that
1775 some things change (we're a research project after all, not a
1776 nuclear plant operating system). If such things should happen, you
1777 could rely on that macro to adapt.
1778 - current value: 30600 for 3.06.00, aka 3.6
1779 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1780 a number that can be compared with SIMGRID_VERSION.
1781 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1782 assertions and verbose/debug logging events are disabled at compile time.
1784 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1786 SimGrid (3.5) stable; urgency=medium
1789 * New feature to any SimGrid-based simulator: Model-Checking
1790 Check SIN#1 for more details.
1793 * New Model SMPI (three-interval linear regression for correction factors)
1794 See RR-7426, available at http://hal.inria.fr/inria-00527150
1795 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1796 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1797 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1798 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1799 * See also src/smpi/README
1804 - Tracing API changes: TRACE_start and TRACE_end should not be called
1805 by user-code. They are automatically called by simulators created
1806 with SimDAG, MSG and SMPI if the toolkit is compiled with
1807 tracing_enabled=ON. Categories declaration and utilization remain the
1808 same for MSG and SimDag.
1809 - A function was added to the tracing API to declare categories with
1811 - TRACE_category_with_color (char *category, char *color)
1812 where color must be in the following format
1813 "%f %f %f", red, green, blue
1814 and red, green, blue are float values in the interval [0, 1]
1815 - User can specify NULL as color parameter, or continue calling
1816 TRACE_category (cat)
1817 On that case, the tracing system will define random colors
1818 - The following command-line options are supported:
1819 --cfg=tracing/filename:msg.trace
1820 --cfg=tracing:1 (activate tracing, needed to use others)
1821 --cfg=tracing/platform:1 (categorized resource use)
1822 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1823 --cfg=tracing/msg/task:1 (task creation)
1824 --cfg=tracing/msg/process:1 (process creation, migration)
1825 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1826 --cfg=tracing/smpi:1 (SMPI interface tracing)
1827 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1828 - examples of examples/msg/tracing updated
1830 - DAXLoader and DOTLoader functions can generate tasks with categories
1831 - A new function to attribute a category to SD tasks:
1832 TRACE_sd_set_task_category (SD_task_t task, char *category)
1833 * Tracing the MPI interface implemented by SMPI
1834 - Collective operations are traced with states
1835 - Point-to-Point operations are traced with states/links
1836 - Tracing activated by a parameter "-trace filename" passed
1837 to smpirun during execution (considering that simgrid
1838 is compiled with tracing enabled)
1839 - To run the simulation with gdb, the simulator
1840 accepts --cfg=tracing/smpi:1 to trace SMPI
1841 - tesh files to check if smpi tracing is ok
1842 - See examples/smpi/NAS/DT-trace
1843 * GTNetS tracing re-worked
1844 - adaptation to the tracing system of GTNets to cope
1845 with modifications regarding the fullduplex mode
1846 - new tesh files to check if gtnets tracing is ok
1849 * Asynchronous communications through the functions:
1850 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1851 * New function: MSG_load_platform_script()
1852 to make possible using a lua script instead of XML files to set up platforms
1853 * New function: MSG_set_function
1854 to associate functions to processes, used when bypassing the parser
1855 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1857 Platforms: Add some more examples in examples/platforms
1858 * Grid'5000: see www.grid5000.fr
1859 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1862 * Change the XML format. This is a very important modification. SimGrid
1863 3.5 introduces a new hierarchical format based on the notion of
1864 Autonomous Systems. Compatibility with old format is ensured
1865 through the perl script provided in the install bin directory
1866 bin/simgrid_update_xml.
1867 It is now possible to build platforms with specific routing
1868 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1869 connect several platforms together. We will try to provide soon
1870 set of realistic platforms exploiting these properties (have a
1871 look at examples/platforms/ for the moment).
1872 * Take the opportunity of the XML format change to be a good XML citizen:
1873 rename link:ctn to link_ctn and similar changes (also dealt with by
1875 * Add a new routing scheme (rule-based) using regular expressions. It
1876 enables to have an extremely low memory footprint when the
1877 underlying routing is simple and can be compactly described. You
1878 need to have libpcre4-dev (perl regular expressions) installed if
1879 you want to use this routing scheme.
1880 * Revive the cluster TAG and allow to easily and efficiently (both in
1881 term of memory and speed) connect clusters together. Have a look
1882 at teshsuite/simdag/platforms/ to see how this can be done. With
1883 this tag, you can create clusters with thousands of tasks at no
1884 cost (have a look at examples/platforms/).
1885 Note: clusters are implemented as ASes, so there is no need for an
1886 enclosing AS tag if you have only one cluster in your platform.
1887 * Add new generic functions in the public interface that allows the user
1888 to call SURF 'create_resource' methods from your code (same
1889 functionality as the XML bypass mechanism but with a much lighter
1891 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1892 that uses a piecewise linear approximation to produce better
1893 results when exchanging small messages.
1894 * Add a new parameter to handle correctly full duplex link and account
1895 for interferences between uplink and downlink communications
1896 (activate with --cfg=fullduplex:1).
1899 * Rename the SD_READY (all dependencies are satisfied and task is
1900 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1901 dependencies are satisfied) state.
1902 This prevents a confusion between the notion of "ready to schedule"
1903 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1904 simulated" (SD_RUNNABLE) used by the simulation kernel.
1905 * Change the way a task is considered as ready. Instead of removing
1906 dependencies when a task is done, a counter is decreased. This way,
1907 it is always possible to reach ancestors thanks to the
1908 SD_taks_get_parents function (even after the end of the simulation.)
1909 * Change the return type of SD_Simulate from (SD_task_t*) into
1910 xbt_dynar_t. This function was in handling a dynar internally and
1911 converted it into a NULL terminated array for historical reasons.
1912 * New function SD_dotload(char*) to load a DAG described in dot
1913 format. This loader and the corresponding examples require the
1914 installation of the graphviz library.
1915 * Fix a bug in the management of tasks of size 0 in the surf network
1916 models. This problem was only visible with SIMDAG and you should
1917 thus disregard results produced with earlier versions if you
1918 relied on this feature (some tasks were blocked because of this).
1919 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1920 even though your applications were doing only point-to-point
1921 communications and sequential computations. Now you can really use any
1922 model you want (of course, if you create real parallel tasks, which are
1923 not implemented in most models beside ptaskL07, this will abort).
1924 * Add an example that schedules a DAX on an heterogeneous platform
1925 using a Min-Min strategy.
1926 * New function SD_workstation_get_current_task() that returns the kind
1927 of task currently running on a workstation in the sequential access
1929 * Raise some warnings when unexecuted tasks remains at the end of the
1930 simulation. This is usually caused by cycles in the DAG.
1933 * New function: SIMIX_process_set_function() called by MSG_set_function
1934 * Change the underlying waiting queue in semaphores so that a process
1935 can wait on several of them simultaneously (as in waitany).
1936 * Fix the way to handle tokens in semaphores so that all access patterns
1937 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1938 * kill the dirty pimple SIMIX_message_sizes_output()
1939 Please use (proper) visualization instead
1942 * New data container: setset (set of sets of elements)
1943 * New module: mmalloc (mapped malloc, allowing to have several
1944 independent segments of malloc)
1945 * New function: xbt_dict_cursor_set_data()
1946 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1947 * New function: xbt_dynar_is_empty()
1948 * New function: xbt_fifo_get_last_item()
1949 * Fix xbt_dynar_shrink(): use the right element size.
1950 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1951 uninitialized areas during expand.
1952 * Fix semaphores: previous implementation was severely broken.
1953 * Use library init/fini functions for our initialization.
1954 - you can use logs and other feature as soon as you want in your
1955 code (even before the xbt_init / MSG_init)
1956 - xbt_exit is now a no-op and produce a warning when used.
1959 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1960 benefit from the latest improvement to the simulation kernel.
1961 * Kill measurement sockets for now. If you rely on them, sorry. This
1962 release is not for you. This feature will be reintroduced in the
1963 future, but we cannot delay the release any further.
1964 * New function: gras_msgtype_get_name().
1965 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1966 * Fix (at last) the pmm example: it should not randomly fail anymore.
1968 Build chain: bug fixes and overall polishing
1969 * Cmake is now stable enough. Hence, we killed the autotools.
1970 * Port to windows ( TM :)
1971 * Fix the 'make install' target.
1972 No need to use 'make install-simgrid' anymore
1973 * Introduce a 'make dist' target compiling a *source* archive
1974 'make package' compiles a binary archive
1975 * Compile java files only on need
1976 * Add --cd and --setenv command line options to tesh
1977 * Out of source builds are not fully supported yet, but we are close
1978 * Enable supernovae and optimization flags by default for our users
1981 * Add layer to set up environment directly from lua, without XML.
1982 * The effect of gras_stub_generator can be achieved through
1983 lua too (check examples/gras/console/ping_generator.lua)
1985 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1987 SimGrid (3.4.1) stable; urgency=low
1989 The "Polishing easter eggs is probably a good idea" release.
1990 This is a bug fixes release only.
1994 * Fix a bug preventing the tasks from begin garbage collected.
1997 * Fix a bug occuring when a host involved in a communication fails.
1998 This was not detected properly by the other peer involved in the
1999 communication. Now, it's reported as a network error.
2002 * Warn the user about loop dependencies in data flow of DAX files
2003 * Obey the control-flow dependencies of DAX files
2006 * Add option "enable_smpi" allowing to not compile SMPI.
2007 Probably useful for the (Mac) users experiencing a build error here
2008 * Improve the detection of lua5.1 and ruby1.8
2010 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
2012 SimGrid (3.4) stable; urgency=low
2014 The "Easter in Cargese" release. Also known as (major changes):
2016 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
2017 ~> bindings were greatly improved
2018 ~> new tracing infrastructure for better visualization introduced
2020 * the "Welcome to configury modernity" release.
2021 ~> we switched from autotools to cmake, and improved our cdash
2024 A more detailled list of changes follow (full detail in svn log).
2026 Java Bindings: Various Cleanups
2027 * (install java-gcj-compat-dev on debian-like to use them)
2028 * Remove put/get: no need to export deprecated interface in Java
2029 Use send/receive instead.
2030 * Cleanup the examples and add a README per directory
2031 * Remove example autoDestination (that's the only way to go now)
2032 * Remove example explicitDestination (was a plain copy of basic)
2033 * Make JniException a runtime exception, so that there is no need to
2034 declare the fact that you may encounter such a beast. I guess that
2035 nobody will ever want to survive such error.
2036 * Create specific errors for each MSG case of failure:
2037 host failure, transfer failure, timeout, task cancelled
2038 * Cleanup the exceptions that may get thrown by each function
2039 * Other internal cleanups in Java bindings. Performance still bad :/
2040 Ruby and Lua Bindings: create them
2041 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
2042 * That's new and great, you should try them out.
2043 Same functionalities than Java bindings, only even less polished
2045 * Kill the useless "rate" argument of SD_task_get_execution_time()
2046 Everyone used to provide -1 as a value, it was not used, and the
2047 semantic of a possible use wasn't even clear.
2048 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
2049 either as comm costs or compute costs to mean that there is no
2050 such thing for that specific task.
2051 * Add a SD_task_set_name() function
2052 * Fix SD_task_unschedule() on typed tasks
2053 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
2054 * In DAX loader, accept useless 'level' attributes to <job> since
2055 LIGO DAGs have them (seem to be to ease graphical representation).
2057 * Add an example masterslave_mailbox.c using send/receive and not
2058 the deprecated put/get interface.
2059 * Kill the MSG_paje_output() function. It's a noop since 2 years.
2060 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
2062 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
2063 (declare MSG_USE_DEPRECATED to still have the old name)
2064 * Add a MSG_task_set_data() function
2065 * About trace replay (see examples/msg/actions):
2067 - Allow to work with splitted trace files for each process
2068 Give the specific trace file as argument of each process,
2069 and call MSG_action_trace_run(NULL)
2070 You can still have one merged file for all processes.
2071 - Fix implementation of collective operations
2072 * Allow task_execute() on 0-sized tasks (closes #10063)
2074 * This is the first release of SimGrid where SMPI is not considered
2075 beta anymore (even if some corners should still be improved)
2076 * Port over the new SIMIX_network submodule (internal refactoring)
2077 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
2078 * Implement more missing elements of the standard:
2080 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
2082 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
2083 MPI_Scan() MPI_Get_processor_name()
2084 - Added implementation of missing case for Alltoall (warning: it's
2085 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
2086 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
2087 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
2088 SMPI_MPI_Allgatherv()
2089 * Bug fixes include:
2090 - MPI_Waitsome() was broken
2091 - Allow relative includes in smpicc
2092 - Command line cfg argument 'reference_speed' was ignored...
2093 - Some functions did not properly lead to auto-benching of user code
2094 - smpicc passes -O2 by default (just like openmpi one)
2096 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
2097 * Bug fixes about timeouts during communications
2098 * add SIMIX_message_sizes_output() as a pimple to write to file the
2099 amount of messages per size. Use gnuplot to get histogram.
2100 Pimple because that's the only user-visible function of simix,
2101 defined directly in xbt.h (irk, sorry)
2103 - Add a SIMIX_sem_get_capacity() function
2104 - Fix interactions with processe resume/suspende
2105 - release_forever() was stupidly broken
2106 - Fix SIMIX_display_process_status() for processes in a semaphore
2107 - Make SIMIX_sem_block_onto() user-visible
2108 * Refactoring context stuff:
2109 - Use pseudo-OOP for better modularity
2110 - reimplement SIMIX_process_kill() without process_schedule() so
2111 that the latter can take as invariant that it is called from
2113 - Merge context_start into context_new for sake of simplicity
2115 * Add a Vivaldi network model, coded live during SUD'10 ;)
2116 * Rename configuration variables to start a hierarchy:
2117 o cpu_model -> cpu/model
2118 o network_model -> network/model
2119 o workstation_model -> workstation/model
2120 * New configuration variables:
2121 o network/bandwidth_factor: correction to bandwith
2122 o network/latency_factor: correction to latency
2123 o netwotk/weight_S: correction to the weight of competing streams
2124 * Add a long description to the models, that users can see with such
2125 argument on the command line: --cfg=cpu/model:help
2126 * --help-models display the long description of all known models
2128 * config: add the ability to set a default value after registration
2129 Does not override any previously set value (e.g. from cmd line)
2130 * dict: allow to have integer key and data.
2131 When so, you need to use the following functions
2132 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
2133 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
2134 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2135 In contrary to regular dicts, the key is not malloced before copy.
2136 Mixing scalar and regular elements in the same dict is not tested
2138 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2139 Tracing for Visualization:
2140 * SimGrid is now instrumented in order to generate a trace file for
2141 visualization analysis: to use it, need to compile SimGrid with the
2142 "tracing" option enabled, and instrument the program using SimGrid with
2143 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2144 (among other functions).
2145 * The instrumentation only traces the platform utilization for now
2146 * Documentation to use the tracing functions and how to analyze the
2147 traces with the Triva tool is written.
2148 * More information about: SimGrid FAQ (in the section Tracing Simulations
2151 * We moved to cmake as default build system. Autotools support will
2152 be dropped soon. Check the FAQ for more info about how to use it.
2153 * Greatly improved our cdash/ctest interactions
2154 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2155 * Added memory checking tests with valgrind; lot of memleak fixing.
2156 This may be the first release of simgrid with so few memory issues
2157 * Added code coverage tests.
2158 Our coverage is still improvable, but at least we see it on cdash.
2160 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2162 SimGrid (3.3.4) stable; urgency=low
2164 The "Desktop Grid needs love too" release (also called Xmas release).
2166 Models improvements:
2167 * Major speedup in the maxmin system solving by using lazy evaluation
2168 Instead of solving completely the maxmin system at each iteration,
2169 only invalidate (and recompute) the modified parts.
2170 This new feature is enabled in default models but you can try to
2171 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2172 * Cas01 IMproved as default CPU model
2173 This CPU model is the same Cas01 model, but it uses the
2174 maxmin-selective-update flag and a heap structure to manage
2175 actions on SURF kernel.
2176 It reduces the complexity to find the next action to finish and,
2177 consequently, it's faster than the old Cas01.
2178 This is the new default CPU model (Cas01).
2179 * Rename the old Cas01 model to Cas01_fullupdate
2180 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2181 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2182 * CpuTI (CPU Trace Integration)
2183 A new CPU model whose objective is simulate faster when using
2184 availability trace files.
2185 Instead of using a full featured, over engineered maxmin system for
2186 CPU modeling, this model does the pre-integration of traces files
2187 to calculate the amount of CPU power available, and so, executes
2188 faster than the old CPU models.
2189 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2190 * Use LV08 as default network model since it gives better accuracy
2191 for small messages and shouldn't change things for big ones.
2192 Use --cfg=network_model:CM02 to get the previous behavior.
2195 ******************************************
2196 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2197 ******************************************
2198 * The new CPU model may changes simulations!
2199 The point is that events occurring at the exact same timestamp
2200 are not scheduled in the same order with the old and new
2201 version. This may be enough to completely change the execution
2202 of simulations in some cases.
2203 * The new network model will change simulations!
2204 This new model is more realistic than the previous one, so you
2205 should consider redoing your old experiments with this model.
2206 Sorry for the inconvenience.
2209 * Introduce the supernovae compilation mode
2210 When compiled that way, the whole SimGrid (or almost) is put in a
2211 single compilation unit and compiled in one shoot.
2212 This is to help gcc which has difficulties to inline stuff from one
2214 The speedup seem to be above 15%, althrough more tests are needed on
2215 amd64 to confirm that gain.
2218 * Port of MSG's mailbox on top of SIMIX network
2219 The put/get mechanism was greatly simplified on the way.
2222 * New SIMIX network module. Provides:
2223 - Mailbox: rendez-vous mechanism to find with who you want to speak
2224 - Synchronous send/recv: easier and hopefully faster since the
2225 logic is handled in the maestro process directly now
2226 - Asynchronous send/recv: you dreamt of it? It's here now
2227 Too bad that nobody cared enough to propagate the change to MSG.
2228 * Add semaphores as SIMIX synchronization mechanism.
2231 * new function SD_daxload(char*) to load a DAX file
2232 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2233 * Introduce typed tasks. Specify its kind and cost at creation.
2234 At scheduling, just give where it should be placed, and the cost
2235 for each involved resource is automatically computed.
2236 Existing constructors so far (more to come of course):
2237 - SD_task_create_comm_e2e() for end-to-end communication
2238 - SD_task_create_comp_seq() for sequential computation
2239 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2240 * new function SD_task_dump() for debuging display
2241 * new function SD_task_dotty(task,FILE*) writing to file the info
2242 about the task in dotty format
2243 * SD_task_dependency_exists() can now cope with having one of its
2244 arguments NULL. If so, it tests whether the other argument has any
2246 * Add getters on list of preceding/following tasks:
2247 SD_task_get_parents(task) and SD_task_get_children(task)
2248 * Add getters on amount of workstations and list:
2249 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2250 * Add getter on task kind: SD_task_get_kind(task)
2251 * Update the start_time and finish_time of tasks on completion/failure
2252 * Bugfix: Remove task from state swags when destroyed
2255 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2258 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2259 dynar with one separate thread per value of the dynar.
2260 * Change the prototype of xbt_thread_create(), sorry.
2261 Added a boolean parameter indicating whether we want to join this
2262 thread (used in SG only for now)
2263 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2266 * GTNetS wrappers should now be usable again (and betterly tested too)
2267 * Fix a major regression from 3.2 where the timeout provided to
2268 MSG_task_put_with_timeout() was used as absolute time before which
2269 the comm should be done.
2270 * Start to fix the <cluster> tag.
2271 - Internal links should be good now (beside of the loopback, which
2272 use the private link instead)
2273 - paths to the external world is still rather broken
2274 - the <route:multi> tag is just broken. Actually that's brain-dead.
2275 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2276 to make it less stupid
2277 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2278 * Fix a source-level compatibility glitch from 3.2: after defining
2279 MSG_USE_DEPRECATED, you can use the old name
2280 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2281 * Allow to compile from the SVN with automake 1.11
2282 * Fix some problems when using the "start_time" tag in deployment XMLs.
2283 * Fix #8569: XBT/synchro.h has redundant declarations
2284 * Fix #8563: MSG return values and exceptions
2285 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2286 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2287 Thanks to Robson Peixoto.
2288 * Fix a vicious bug in dictionaries inducing that some elements were
2289 not freed on xbt_dict_free()
2291 Portability report of this version:
2292 * Main portability targets:
2293 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2294 on (amd64/i386/ia64)
2295 - mac leopard on i386
2296 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2297 but nothing critical.
2298 * Other platforms: windows, AIX and others were not tested for this release
2300 Timing report of this version:
2301 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2302 scenario parameters). From 8h to a few seconds in desktop grid settings.
2303 * Supernovae brings about 25% speedup on i386.
2305 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2307 SimGrid (3.3.3) stable; urgency=low
2309 The "Need for Speed" release.
2311 The timings done to validate the 3.3.2 were faulty.
2312 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2314 The problem was a conversion from a manually handled vector to
2315 xbt_dynar_t on the critical path.
2316 xbt_dynar_foreach calls functions, inducing stack management crap.
2318 We inlined these functions and xbt_dynar_foreach is now breath taking.
2319 We also inlined xbt_swag_belong on the way.
2321 Here are some approximate speedup measurements (on master/slaves
2322 simulations lasting between 10s and 20s each):
2323 3.3.1 -> 3.3.2: about same performance
2324 3.3.2 -> 3.3.3: 40% speedup
2325 3.3.1 -> 3.3.3: 40% speedup
2326 3.3.1 with inline patch -> 3.3.3: 30% speedup
2328 Our reading is that the refactoring which occurred in 3.3.2 made us
2329 suffer much more from the xbt_dynar_foreach low performance, but
2330 once we solved this, this refactoring proved to be very performance
2331 effective. From the 40% speedup, somehow, 10% are due to the
2332 inlining and 30% to the refactoring.
2334 That's a pitty that gcc cannot inline functions placed in other files
2335 alone. We have to choose between:
2336 - break the encapsulation (by putting private data structures and
2337 accessors in headers files to help gcc)
2338 - live with low performance
2339 - switch to a decent compiler such as icc (not quite possible).
2341 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2343 SimGrid (3.3.2) stable; urgency=low
2345 The "Simplicity does not preceed complexity, but follows it" release.
2347 The main contributors of this release were (lexical order):
2348 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2351 * Extract the routing logic into its own object.
2352 (was dupplicated in network.c and workstation_LV07.c;
2353 Allows to implement other ways of storing that info)
2354 => kill now useless network_card concept
2355 - Use dynar to represent routes (instead of void** + int*)
2356 - kill link_set (use surf_network_model->resource_set instead)
2357 - Add a command-line option to choose the routing schema to use
2358 - Add three new models:
2359 * Floyd (shortest path computed at initialization)
2360 * Dijikstra (shortest path recomputed all the time)
2361 * Cached Dijikstra (shortest path computed on need)
2362 All these models where contributed by Silas De Munck, and are
2363 described in his ICCS09 paper.
2365 * Simplify model declaration
2366 (less redirections, less function to write when defining a model)
2367 - Factorize stuff between models:
2370 surf_model_resource_set(model)
2371 surf_model_resource_by_name(model, name)
2372 - Unify the types of models in s_surf_model_t (using an union)
2373 - Embeed fields of common_public directly into s_surf_model_t
2374 - Rename model methods:
2375 action_free ~> action_unref
2376 action_change_state ~> action_state_set
2377 action_get_state ~> action_state_get
2378 - Change model methods into functions:
2379 (model)->common_public->action_use ~> surf_action_ref
2381 * Implement a generic resource; use it as ancestor to specific ones
2382 (allows to kill duplicated code in models)
2383 Drawback: timer command don't need no name nor properties;
2384 workstation_CLM03 don't need no properties
2385 (but I guess we can live with those few bytes wasted)
2387 * Improve the action object model
2388 - implement a constructor avoiding dupplicated code about field
2389 initialization in generic_action part.
2391 * Kill the SDP model: it has an external dependency, is deprecated
2392 in flavor of modern lmm models, and didn't compile since a while
2395 * Relocation of the context module from XBT to SIMIX.
2396 (the context were decoupled from the simix processes, duplicating a lot of code)
2397 => a lot of code was factorized
2398 - less overhead is introduced during scheduling
2399 - simpler API for the context factory
2400 - the logic for process creation,destruction and manipulation was simplified
2401 * Simplification of the s_smx_process_t data structure.
2402 => accesing the simix level data associated to a process is faster now,
2403 and the code is a lot more readable.
2406 * Implement some more MPI primitives:
2407 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2408 -implementation: Bcast: flat or 2-ary tree (default),
2409 Barrier: 4-ary tree,
2411 Allreduce: Reduce then Bcast
2412 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2413 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2414 Alltoallv: flat tree, like ompi
2416 * Add support for optimized collectives (Bcast is now binomial by default)
2417 * Port smpirun and smpicc to OS X
2420 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2421 Shout out if you used it.
2424 * Display the list of still queued messages in SG mode when existing
2428 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2429 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2430 * Add xbt_str_from_file(FILE*)
2431 * Add xbt_dict_get_key achieving a linear reverse search
2432 * Remove the context module
2434 Portability report of this version:
2435 * Main portability targets:
2436 - Linux(debian)/x86/context
2437 - Linux(debian)/x86/pthreads
2438 - Linux(debian)/amd64/context
2439 - Linux(debian)/amd64/pthreads
2440 On these, we still have the eratic breakages of gras/pmm and
2441 amok/saturate_sg reported in previous version. We still think
2442 that the tests are the cause of the fault, not the tested code.
2444 - Mac OSX Leopard/x86/context
2445 Still false negative in tesh autotesting.
2446 Smpi still fails, but this time because readlink does not accept -f
2447 Everything seems to work properly beside of that.
2450 - AIX version 5.3 (only tested contexts this time)
2451 Smpi still fails there because mktemp is not installed.
2452 Everything seems to work properly beside of that.
2454 I managed to compile it for the first time, but several breakages.
2455 Won't delay the release for this exotic platform.
2457 * Windows: it's still lagging behind. If you want to help, please
2460 Timing report of this version:
2461 This version seem to be more than 5% faster than 3.3.1 (on linux
2462 64bits with contextes). The gain is less than expected, we are
2463 investigating this for next release.
2465 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2467 SimGrid (3.3.1) stable; urgency=low
2470 * Implement a --cfg-help to show existing configuration variables
2471 * Build chain do not require doxygen in maintainer mode
2474 * fix a bug on struct sizeof computation, which prevented the
2475 exchange of arrays of structs in some conditions
2476 - added a regression test about this in datadesc_usage
2477 * Allow the exchange of 0-long dynamic vectors.
2478 - for that, use -1 as indicator of dynamic size instead of 0
2479 - This implied to change any size from unsigned long to long,
2480 reducing a bit communication abilities, but I guess that with
2481 64bits being quite common, this is more than enough.
2482 - This also induce a protocol change, thus bumping network protocol
2483 version from 0 to 1 (if we have external users, we have to get
2484 clean on that point too ;)
2485 - added two regression tests about this in datadesc_usage
2486 * Be more verbose when propagating local exceptions
2487 This helps debugging.
2488 * Display the status of simulated processes when receiving SIGINT in
2492 * Allow to control the simulation from a trace file.
2493 New functions MSG_action_register() and MSG_action_trace_run()
2494 The first one allows to associate a function execution to each
2495 kind of action while the second one parses a trace file and
2496 triggers the corresponding actions within the system.
2497 For now, only a toy example is provided in examples/msg/actions
2498 * Add an exemple of process migration in examples/msg/migration
2499 * Fix a bug in task exchange which broke MSG_task_get_sender()
2500 Add a teshsuite regression test for that.
2501 [Bug: if MSG_task_get_sender() is called after sender exit,
2503 * Fix a bug which prevented suspend/resume to work properly
2504 * Display the status of simulated processes when receiving SIGINT
2505 This fixes a regression of v3.3. due to the introduction of SIMIX
2506 * Bug fixing in failure management:
2507 - trace could not start by a failure at time 0
2508 - failure during communications were not working
2511 * Add SIMIX_process_set_name() to change the name of the current
2512 process in the log messages.
2513 * Store smx_hosts in a dict since we only retrieve them by name
2514 * Move the configuration infrastructure to surf
2517 * Move the configuration infrastructure to surf
2520 * Massive internal cleanups:
2521 - Store internal structures on processes instead of hosts (allows
2522 to have more than one process per host, in addition of being more
2524 - Cleanup the initialization/finalization process
2525 - Kill a whole bunch of unneeded synchronization:
2526 processes run in exclusive manner within the simulator
2527 - Move queues from global tables to process data fields
2529 - now accept -platform and -hostfile arguments
2530 - Pass the right rank value to processes according to the hostfile
2531 * Compile the examples by default, and use them as regression tests
2532 * Implement MPI_Wtime()
2533 * Change the reference speed to a command line option
2536 * TCP_gamma can now be specified as command line option using
2537 --cfg=TCP_gamma:10000000.0
2538 * Change the --surf-path cmd line option into --cfg=path:
2541 * Also include strbuff from xbt.h public header
2542 * xbt_ex_display(): do not free the exception after displaying
2543 This allows to do more with the given exception afterward.
2544 Users should call xbt_ex_free() themselves.
2548 Portability report of this version:
2549 * Main portability targets:
2550 - Linux(debian)/x86/context
2551 - Linux(debian)/x86/pthreads
2552 - Linux(debian)/amd64/context
2553 - Linux(debian)/amd64/pthreads
2554 These targets fail about 1/10 of times on gras/pmm, but we believe
2555 that this is because of the test, not because of simgrid.
2556 amok/saturate_sg fails even more rarely, and the test may not be
2559 - Mac OSX Leopard/x86/context
2560 The test suite still spits tons of errors because some obscure
2561 force prevents us from removing the temporary directories
2562 arguing that they still contain some metadata I've never heard of.
2563 Smpi fails because seq is not installed.
2564 Everything seems to work properly beside of that.
2567 - AIX version 5.3 (both contexts and pthread)
2568 Smpi still fails there because mktemp is not installed.
2569 XML inclusions seems rosty on AIX.
2571 * Windows: it's still lagging behind. If you want to help, please
2574 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2576 SimGrid (3.3) stable; urgency=high
2580 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2581 [Malek Cherier & Mt]
2583 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2586 Until now, GRAS were using MSG as an interface to SURF. It was
2587 quite difficult because both interface have several differences
2588 (MSG channels vs GRAS sockets were the most notable point).
2590 This also opens the gate to SMPI (which should occur soon) and speed
2591 up simulations by to 40% (even if it were not the main goal).
2593 **************************************
2594 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2595 **************************************
2596 The point is that events occuring at the exact same timestamp are
2597 not scheduled in the same order with the old and new version. This
2598 may be enough to completely change the execution of simulations in
2599 some cases. Sorry for the inconvenience.
2601 * Cleanup and upgrade the XML format to push further scalability
2602 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2604 * Improve the testing infrastructure with tesh. Now a very large part of
2605 the code is tested not only by being run but also by checking that the
2606 output match an expected output [Mt].
2608 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2609 is really less memory-demanding, which should allow you to use
2610 larger files in SimGrid [AL].
2612 * Inform valgrind about our contextes, so that it becomes usable
2613 with the default (and more effecient) version of SimGrid
2614 [contributed by Sékou Diakite, many thanks]
2617 * Introduce a listener thread in charge of receiving incoming
2618 messages from the network. It allows to overlap communication and
2619 computation but most notably, it removes some stupid deadlocks due
2620 to the fact that so far, a process could not send and receive at
2621 the same time. This made most non trivial communication schema
2623 * Convert the PIDs from long int to int to match the MSG ones (and
2624 linux ones too) [Mt]
2625 * New function: gras_agent_spawn() to launch a new process on
2626 current host. Only working in simulation for now. [Mt]
2627 * New function: gras_os_hostport() returning a constant form (ie,
2628 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2631 * Make the backtrace of exceptions more human readable [Mt]
2632 * New module: xbt/str [Mt]
2633 a ton of string utility functions (split, join, printf to a newly
2634 allocated buffer, trim, etc)
2635 * New module: xbt/hash [Mt]
2636 SHA1 hashing algorithm (more to come if needed)
2637 * New module: xbt/synchro [Mt]
2638 synchronization tools (mutex and conditions) working the same way
2639 in simulation and in real life (mainly useful for GRAS, but not
2641 * New module: xbt/queue [Mt]
2642 classical producer/consumer synchronization scheme
2643 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2644 (using the classical functions will get serialized) [Mt]
2645 * Make dictionary internal table dynamic. No need to specify its size
2646 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2648 * Make sure the log channels are organized as a tree under windows
2649 (because of ANSI C compatibility issue, any channel were child of
2650 root directly) [Mt].
2653 * Cleaned many thing in surf and fixed a few bugs [AL].
2654 * Add a nice command line configuration mechanism to compose models [AL].
2655 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2656 the previous one (KCCFLN05). It relies on something that looks like
2657 a max-min sharing mechanism but cannot be written as such. A new solver
2658 was thus designed [AL].
2659 * Add a new solver to lmm. Based on Lagrange optimization and
2660 gradient-based descent, it enables to efficiently maximise systems s.a
2662 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2664 This solver enables to propose two new network models for TCP Reno and
2665 TCP Vegas based on Low's work. These models still need to be fully
2666 tested though [Pedro Velho].
2669 * Bug fix in SD_simulate. Now the time bound given as argument is
2671 * Use the new parallel task model (ptask_L07) as default.
2672 * Use the SURF command line configuration mechanism.
2673 * 0-size tasks (for synchronization) should now work.
2675 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2677 SimGrid (3.2) stable; urgency=high
2681 We still experience issues on this platform, but we believe that at
2682 least MSG is usable.
2684 GRAS API BREAKAGE (for simplification purpose, sorry):
2685 * the gras_msgtype_by_name is not used anymore. Instead of
2686 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2687 you can write (and must)
2688 gras_msg_send(toserver, "request", &request);
2689 - If you still want to pass a gras_msgtype_t to the function (to cache
2690 the type and avoid the lookup time), use the gras_msg_send_() variant.
2691 - Impacted functions:
2692 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2693 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2694 * The callbacks are now expected to return 0 when everything went well
2695 (just like the main() function)
2697 GRAS new features and improvements:
2698 * New module mechanism where user code can use per process globals [Mt]
2699 This is similar to gras_userdata_*() functions, but for libraries. It
2700 factorize some code developped over and over in the examples and AMOK.
2701 It has still to be documented and used (only amok/peermanagement is
2703 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2704 loss when they were small enough to fit into the buffer and sent quickly
2705 enough so that they can all get received in one shoot.
2706 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2707 if not found. Use the *_or_null() variant for the old semantic.
2708 * In gras_msg_handle, do not discard messages without callback.
2709 They are probably messages to be explicitly awaited later (ie, proofs of
2710 mis-synchronization in userland since they are sent before being awaited)
2711 No big deal usually.
2712 * gras_socket_meas_send/recv: semantic changed!
2713 The numerical arguments used to be (1) the total amount of data to send
2714 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2715 messages. This was need for the fool willing to send more than MAXINT
2716 bytes on quite fat pipes.
2719 * Do really rename the hostmanagement module to peermanagement. [Mt]
2720 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2721 but this is rather new and this was documented in the module
2722 documentation (poor excuses, I admit)
2723 * Bandwidth measurement semantic changed! This follows the changes to
2724 gras_socket_meas_send/recv explained above.
2727 * A sequential mode has been added to the workstations. When a workstation
2728 is in sequential mode, it can execute only one task, and the other tasks
2729 are waiting in a FIFO. [Christophe Thiery]
2732 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2733 model for SIMDAG. [Christophe Thiery]
2734 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2735 equal to 0, which caused some bad behaviors in
2736 saturated_constraint_set_update. I now use a threshold mechanism like in
2740 * When running manually src/testall, you select specific units [Mt]
2741 testall is the result of our cunit mechanism, and should replace all
2742 the scripty thingy around since bash don't run easily on billware.
2744 * A mallocator system has been added. [Christophe Thiery]
2745 Mallocators allow you to recycle your unused objects instead of freeing them
2746 and allocating new ones.
2748 Documentation update:
2749 * FAQ reworking + New FAQs:
2750 - "Valgrind spits tons of errors!" [Mt]
2751 - "How to repport bugs" [Mt]
2752 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2753 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2754 - Communication time measurement within MSG [AL]
2755 - I experience weird communication times when I change the latency [AL]
2756 * GRAS tutorial [Mt]
2758 - an introduction to the framework and to the used communication model
2759 - an initiatic tour introducing the most proheminent features:
2761 . Lesson 0: Installing GRAS
2762 . Lesson 1: Setting up your own project
2763 o Part 2: Message passing
2764 . Lesson 2: Exchanging simple messages
2765 . Lesson 3: Passing arguments to the processes (in SG)
2766 . Lesson 4: Attaching callbacks to messages
2767 . Lesson 5: Using globals in processes
2768 . Lesson 6: Logging information properly
2769 . Lesson 7: Using internal timers
2770 . Lesson 8: Handling errors through exceptions
2771 . Lesson 9: Exchanging simple data
2772 . Lesson 10: Remote Procedure Calling (RPC)
2773 . Lesson 11: Explicitely waiting for messages
2774 . Recapping of message passing features in GRAS
2775 - A HOWTO section containing:
2776 o HOWTO design a GRAS application
2777 More are due, of course. They will come latter. In the meanwhile, you can
2778 check the examples which are still here.
2780 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2782 SimGrid (3.1) stable; urgency=high
2786 There was a stack corruption somewhere, visible only when optimizing
2787 with these versions. [Vince]
2790 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2791 module that enabled to play with Directed Acyclic Graphs. It is built
2792 directly on top of SURF and provides an API rather close to the old
2793 SG. Some old codes using SG are currently under rewrite to check that
2794 all needful functions are provided. [Christophe Thiery]
2797 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2798 extension of the classical CLM03 model that gracefully handles
2799 failures. This is now the default model for MSG and GRAS. It doesn't
2800 handle parallel tasks yet though. [AL]
2801 * Bug fix: Weights were not correctly set in the network part.
2802 WARNING: This may have resulted in incorrect results with simulations
2803 where there are more than one flow on a given link. [AL]
2806 * After a (long ?) discussion on simgrid-devel, we have decided that the
2807 convention we had on units was stupid. That is why it has been decided
2808 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2809 WARNING: This means that all previous platform files will not work as
2810 such with this version! A warning is issued to ask users to update
2812 A conversion script can be found in the contrib module of the CVS, under
2813 the name contrib/platform_generation/surfxml_update.pl [MQ]
2816 * Bug fix: Processes were started in reverse order, wrt deployment file.
2817 WARNING: if your code relies on this bug, please fix it. [AL]
2818 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2819 being executed on two different locations. [AL]
2820 * Bug fix: Failures are now better supported thanks to Derrick's tests
2821 (there was many failure situations I hadn't thought of and that weren't
2822 correctly handled). [AL]
2823 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2827 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2828 Exception occurring on server-side are propagated back to client (!).
2830 API CHANGE: the callback changed their prototype. Change:
2831 int my_handler(gras_socket_t expeditor, void *payload_data) {
2833 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2834 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2836 * New! function: gras_msg_handleall to deal with all messages arriving
2837 within a given period.
2838 * New! function: gras_socket_server_range to get a server socket in a
2839 range of port numbers (ease to avoid port number conflicts) [MQ]
2840 * New! gras processes display their backtrace when they get a SIGUSR1
2841 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2842 Sweet to debug RL processes [MQ]
2846 - Do not force experiment sizes to be expressed in kb, or it becomes
2847 impossible to measure the latency this way (needs one byte-long tests)
2848 WARNING: this changes the amok_bw_* function semantic. [MQ]
2849 - Implements the link saturation stuff. [MQ]
2850 * Peer management module:
2851 New! module factorizing code that we wrote over and over [MQ].
2854 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2855 - Test units are placed directly into the library code, they get extracted
2856 automatically and placed into the src/testall binary.
2857 - Convert most of the XBT tests to this system.
2858 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2859 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2860 live in the same C code. This required to change a little bit the API
2861 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2862 * New module: sparse graph structure with basic algorithms (this is work
2863 in progress and the API is not considered to be frozen yet). [AL]
2864 * Display more information on backtraces: source line & function names are
2865 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2866 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2867 * Reworked a little bit some #include statements to load only required
2868 headers. Some user code that relied on SimGrid to include stdlib or
2869 stdio may need to include it by themselves. [AL]
2870 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2872 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2873 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2874 but I guess that almost nobody used those parts. [MQ]
2876 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2878 SimGrid (3.0.1) stable; urgency=low
2881 * Unfortunately, I had missed 5 misnamed functions:
2882 xbt_fifo_item_t xbt_fifo_newitem(void);
2883 void xbt_fifo_freeitem(xbt_fifo_item_t);
2884 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2885 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2886 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2887 They're now deprecated. Please use their new versions:
2888 xbt_fifo_item_t xbt_fifo_new_item(void);
2889 void xbt_fifo_free_item(xbt_fifo_item_t);
2890 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2891 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2892 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2894 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2895 * Documentation: xbt_log module unmercifully reworked [MQ]
2896 * Bugfix: there was a problem with the ending of contexts with
2897 the pthread backend. It caused some weird deadlock or behavior
2898 depending on the pthread implementation. [AL]
2899 * Bugfix: get the exceptions raised in the simulator repport where
2900 and why they come from when they are not catched in time [AL, MQ]
2903 * Bugfix: Do repport the error when two non-connected hosts try to
2904 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2907 * Add additionnal checkings on communications. Assert that two
2908 communicating hosts are connected by a set of links... [AL]
2911 * Add additionnal checkings on channel values in communication [AL]
2912 * New: MSG_task_get_source to see on which host a task was generated [HC]
2913 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2914 the number of tasks waiting to be received on channel and sent
2916 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2917 waits for the first task coming from a given host.. [AL]
2919 GRAS new functionnalities: [MQ]
2920 * Enhance the parsing macro to allow the size of multidimentional objects
2921 to be given thru annotations.
2922 * New example (and documentation): Matrix Multiplication a la RPC
2923 (as when I was young!) and fix a bunch of bugs found on the way.
2925 GRAS performance improvements: [MQ]
2927 * Reduce the amount of cbps creation/destruction by making it static to
2928 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2931 * Change libdata to a set so that we can search for stuff by ID (and thus
2932 reduce the insane amount of dict lookups)
2935 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2936 RL communication on top of SG since it's so uneffective.
2937 It may also be used for inter-thread communication in RL, one day.
2938 * Use gras_datadesc_copy() to exchange messages on top of SG
2940 - improve message exchange performance on top of SG
2941 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2942 * Don't exchange on the network the size of the used part of buffer,
2943 instead, specify the possible buffer size to read().
2945 - reduces the amount of read/write calls (one pair per exchange)
2946 - reduces the amount of exchanged data (the size)
2947 - allows to retrieve all arrived data on receiver side, if we don't need
2948 it right now (subsequent read will peek the buffer)
2949 - allows the receiver to proceed with the begining of the stream before
2950 everything is arrived
2951 - make it possible to build an iov transport (using readv/writev)
2953 - take care of the data with non-stable storage (like stacked data),
2955 * If possible, TCP send uses vector I/O (when writev() is here)
2956 - Don't use it for receive since we send data sizes and data on the
2957 same stream, so we wouldn't be able to chain large amount of chunks
2958 before having to flush the stuff to read the size.
2959 * Rework the transport plugin mechanism to simplify it and reduce the
2960 amount of pointer dereferencement when searching for the right function
2963 * I guess that now, we do almost as few system calls as possible while
2964 doing as few data copy as possible.
2966 To improve it further, we could try to send all the sizes first and then
2967 all the data (to use iov on receiving size), but it's only a partial
2968 solution: when you have 2 dimensional data, the sizes of the second
2969 dimension is data of the first dimension, so you need 3 streams.
2971 I'm not sure the potential performance gains justify the coding burden.
2973 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2975 SimGrid (3.00) stable; urgency=high
2978 * New! Give the possibility to hijack the surf parser and thus bypass
2979 MSG_create_environment and MSG_launch_application. Have a look at
2980 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2982 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2984 SimGrid (2.96) unstable; urgency=low
2989 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2990 This deprecates the xbt_error_t mechanisms.
2991 It modifies (simplifies) all XBT and GRAS API.
2992 MSG API keeps unchanged (exceptions raised by XBT are catched from
2993 within MSG and masked with existing error handling facilities)
2996 * New! Add a FATPIPE model. [AL]
2997 * New! Add a parallel task model. [AL]
2998 * New! Add automatically a loopback interface (in the default
2999 network model) if none was precised.
3002 * Bugfix: MSG_process_resume now works with the current running process.
3004 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
3005 * Modification of MSG_task_get_compute_duration. Once a task has been
3006 processed, the value returned by this function is now equal to 0. [AL]
3007 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
3008 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
3009 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
3010 of a m_task. For now, MSG_task_cancel only works with computation
3012 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
3013 of the processor (in Mflop/s) regardless of the current load on the
3015 * API Change: use proper naming convention for MSG_getClock and
3016 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
3018 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
3019 This function changes the priority of a computation task. This priority
3020 doesn't affect the transfer rate. A priority of 2 will make a task
3021 receive two times more cpu power than the other ones. This function
3022 has been added to suit the needs of Nguyen The Loc and hasn't been that
3023 much tested yet. So if it fails, please report it and send me your code.
3025 * API Change: removed all functions and types that were marked "deprecated"
3026 since many months. Renamed MSG_global_init_args to MSG_global_init.
3028 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
3030 SimGrid (2.95) unstable; urgency=low
3033 * Steal some nice code to GNU pth to fix context detection and usage [AL]
3034 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
3035 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
3038 * Bugfix: Allow absolute paths for platform description files [MQ]
3039 * Bugfix: do free the variables after use. Leads to drastic performance
3041 * Implement max_duration (ie, timeouts) on resources [AL]
3044 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
3046 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
3047 the needed support (timeouts) [AL]
3050 * Implement measurement sockets. You can now get the bandwidth between two
3051 hosts thanks to AMOK (see below). [MQ]
3052 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
3053 dynar over the network (yeah) [MQ]
3054 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
3055 * simple/static token-ring example. [Alexandre Colucci and MQ]
3056 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
3057 in place before [MQ]
3060 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
3061 type parsing. "short" and "long" modifiers were also ignored; other
3062 modifier (such as reference level) are still ignored. [MQ]
3063 * Embeed the buffer size within the buffer itself on SG. [MQ]
3064 That way, send() are atomic and cannot get intermixed anymore (at least
3065 the ones which are less than 100k; bigger messages still have the issue)
3066 * Array size pushed by the field, not by the field type (or each
3067 and every long int will push stuff to the cbps) [MQ]
3068 * use select() to sleep since it allows to portably sleep less than one
3071 GRAS (minor cleanups)
3072 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
3075 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
3076 * type category 'ignored' killed as it was never used and were difficult
3078 * whether a type can cycle or not is now a flag, leaving room for more
3079 flags (as "ignored", if we feel the need one day ;)
3080 * Rename raw sockets to measurement sockets since "raw" has another
3081 meaning in networking community.
3084 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
3085 and offers features not belonging to GRAS but that most applications
3086 need. One day, it may be a set of plugins loadable at runtime.
3087 * New module: bandwidth
3088 bandwidth measurement between arbitrary nodes running this module. [MQ]
3090 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
3092 SimGrid (2.94) unstable; urgency=low
3094 The first beta release of SimGrid 3 !
3098 * Update the main page and the FAQ. Adding references to gforge.
3101 * Add a gras_os_getpid function.
3104 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
3105 * Extend the logs so that they also print PID, hostname, date, ... if
3107 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
3108 and kill the old version which were in testsuite/
3109 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
3113 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
3114 for users. Instead, each programming environment (GRAS, MSG,...) use it
3115 when needed to provide such a feature to users.
3116 Don't shortcut the mechanism or you will also shortcut the virtualization
3117 you need on the simulator.
3121 * Cleanups in configury with regard to compile optimization/warning flags.
3122 Also add -fno-loop-optimize to any powerpc since it's the optimization
3123 killing gcc (< 3.4.0).
3124 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
3125 needed by MSG examples complications
3126 * Borrow configury beautifications from PHP
3129 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
3130 warning (thanks loris for stumbling into this one).
3131 * Bugfix: stop loading private headers (gras_config.h) from the public
3135 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3136 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3137 RL would imply a DNS resolver, which is *hard* to do in a portable way
3138 (and therefore delayed).
3139 * Implement a real timer mechanism and use it in timing macros. This allows
3140 to avoid rounding errors and get a 0.000005 sec precision in timing
3141 macros. While I was at it, various cleanups:
3142 - allow to declare more than one timed section per file (fix a stupid bug)
3143 - move some private declaration to the right place
3144 - merge conditional execution and timing macros into emulation module
3145 - document the module
3146 - make sure the module cleanups its mess on gras_exit
3147 * Documentation improvements:
3148 - (new) how to compile applications using GRAS
3149 - (new) emulation support (timing macros)
3151 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3153 SimGrid (2.93) unstable; urgency=low
3155 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3158 - Use Paje properly where used. Still to be sanitized properly.
3159 - Portability fix: Add an implementation of the contexts using pthread
3163 - Add xbt_procname(): returns the name of the current process.
3164 Use it to show the current process's name in all logging.
3166 - fix detection of older flex version and the reaction, since we do
3167 depend on modern ones (we use lex_destroy)
3168 - Better separation of SG and RL in the libs: remove all simulation code
3169 from libgras. As a result, this lib is now only 200k when stripped.
3170 Some of the xbt modules may also be duplicated (two sets and such) and
3171 should be cleaned/killed before SG3.
3172 - Insist on using xlC on AIX because of weird problems involving gcc there.
3173 - Cleanup the make remote stuff. This is now done by scripts
3174 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3175 mainly for our private use, but we're working on changing them to user
3178 - Bugfix: flush the socket on close only if there is some *output*.
3179 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3181 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3184 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3186 SimGrid (2.92) unstable; urgency=low
3188 Alpha 3 on the path to SimGrid 3
3192 - New! First try of benchmarking macros.
3193 - New! First try so that gras_stub_generator generate deployment and
3194 remote compilation helpers.
3196 - Bugfix: Initialization fix in msg_test.
3200 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3202 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3204 - API Change: gras_os_sleep to take the amount of seconds as a double.
3205 Accepting an int was error prone since it was the only location where
3206 seconds were coded as such. It leaded to damn rounding errors.
3207 - Bugfix: Hard to belive that timers ever worked before this.
3209 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3211 SimGrid (2.91) unstable; urgency=low
3213 Alpha 2 on the path to SimGrid 3
3217 - Bug fix in the lmm_solver.
3219 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3220 through the function MSG_paje_output.
3221 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3222 - It is possible to bound the rate of a communication in MSG with
3223 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3224 to write it in the changelog).
3225 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3230 - Various cleanups to the autotools stuff
3231 - Begin to move Gras examples to examples/gras/
3232 - Let make distcheck work again (yeah!)
3234 - documentation overhauled using doxygen.
3235 gtk-doc-tools is dead in SimGrid now.
3236 - Automatically extract all existing logging categories, and add the list
3237 to the documentation (long standing one, to say the less)
3239 - Cleanup the known architecture table. Reorder the entries to group what
3240 should be, and use a more consistent naming scheme.
3241 (some of the test dataset are still to be regenerated)
3242 - New! Allow library to register globals on each process just as userdata
3244 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3245 lookup time (for now).
3246 Use it in msg and trp.
3247 This cleans a lot the internals and helps enforcing privacy of the
3248 headers between the gras components.
3249 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3250 - Bugfix: gras_os_time was delirious in RL.
3251 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3253 - Reenable GRAS now that it works.
3255 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3257 SimGrid (2.90) unstable; urgency=low
3259 Alpha 1 on the path to SimGrid 3
3261 * It is a long time since the last release of SimGrid. I'm sorry about
3262 that but as I had told you, I was rewriting a lot of things. I apologize
3263 to those who had been reporting bugs to me and that I had not answered.
3264 If your bug is still in the new version, please tell me. Here is a
3265 summary of the main changes.
3267 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3268 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3269 portability, speed, and a lot more but you'll figure it out later.
3270 SimGrid now comprises 3 different projects: MSG, GRAS and SMPI.
3271 I wanted to release the new MSG as soon as possible and I have
3272 broken GRAS, which is the reason why, for now, only MSG is fully
3273 functional. A laconic description of these projects is available
3274 in the documentation.
3276 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3277 kernel optimized for our needs (called SURF but only the developers
3278 should use it). Hence, MSG is now roughly 30 times faster and I think
3279 that by rewriting a little bit MSG, I could event speed it up a little
3280 bit more. Beside the gain in speed, it is also much easier to encode a
3281 new platform model with SURF than it was with SG. More to come...
3283 * REVOLUTION 3: I have tried to change a little as possible the API of
3284 MSG but a few things really had to disappear. The main differences
3285 with the previous version are:
3286 1) no more m_links_t and the corresponding functions. Platforms are
3287 directly read from a XML description and cannot be hard-coded
3288 anymore. The same format is used for application deployment
3289 description. The new format is described in the documentation.
3290 Have a look in tools/platform_generation. There is a tiny script
3291 that converts from the old platform format to the new one. Concerning
3292 the application deployment format, parsing the old one is tricky.
3293 I think most of you should however be able to convert your files. If
3294 it is really an issue, I can write a C code that does the conversion.
3296 2) the toolbox tbx does not exist anymore. We now have a library
3297 with much more data-structures but without the hash-tables (we have
3298 dictionaries that are much faster).
3300 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3302 *****************************************************************************
3303 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3304 * don't feel like converting the oldies. (Mt) *
3305 *****************************************************************************
3308 Version 2.90: "the long awaited one"
3309 - Finished rewriting and debugging MSG. Rewrote the documentation.
3310 - disable GRAS for now since it needs to be ported to the newest SG
3313 - Finish the port to windows (using mingw32 for cross-compile)
3316 - Main loop and datastructures of SURF. A cpu resource object is
3317 functional. Surf can thus be used to create cpu's with variable
3318 performance on which you can execute some actions.
3320 2004-11-15 Martin Quinson
3321 - Port to ARM. Simply added the alignment and size descriptions. Should
3322 work, but the ARM machines are so slow that I didn't had the opportunity
3323 to 'make check' over there yet.
3325 2004-11-15 Arnaud Legrand
3326 - Trace manager now written. It uses a heap structure and is therefore
3327 expected to be efficient. It may however be speeded up (particularly
3328 when many events occur at the same date) by using red and black
3329 trees. One day maybe...
3330 - Max-min linear system solver written. It uses a sparse matrix
3331 structure taking advantage of its expected use. Most operations are
3332 O(1) and free/calloc are called as few as possible. The computation of
3333 the minimum could however be improved by using a red and black tree
3336 2004-11-03 Arnaud Legrand
3337 - Rename every gras_* function that was in xbt/ to its xbt_
3339 - Add a heap and a doubly-linked list to xbt
3340 - Added a dichotomy to the dictionaries. make check works as well before
3341 so I assume that the patch is correct. I do not know however if things
3342 run effectively faster than before now. :)
3344 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3345 SimGrid, and the version number is bumped to 2.x
3347 2004-10-29 Martin Quinson
3348 - Introduction of the remote errors.
3349 They are the result of a RMI/RPC on the remote machine.
3350 ErrCodes being scalar values, you can't get the host on which those
3351 errors did happen. Extending the error mechanism as in Gnome is possible.
3352 No idea yet whether it is a good idea.
3354 2004-10-28 Martin Quinson
3355 - Interface revolution: the Starred Structure Eradication.
3356 I used to do typedef struct {} toto_t; and then handle *toto_t.
3357 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3358 - ???_t is a valid type (builded with typedef)
3359 - s_toto_t is a structure (access to fields with .)
3360 - s_toto is a structure needing 'struct' keyword to be used
3361 - e_toto_t is an enum
3362 - toto_t is an 'object' (struct*)
3364 typedef struct s_toto {} s_toto_t, *toto_t;
3365 typedef enum {} e_toto_t;
3366 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3367 s_toto_t) is private.
3369 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3370 so that it can set the variable to NULL. It was so for dicts and sets,
3371 it changed for dynars.
3373 - Fix a bunch of memleaks in dict_remove
3374 - Fix a bug in sg/server_socket opening: it failed all the time.
3376 2004-10-07 Martin Quinson
3377 - Speed up dynar lookup operation a bit.
3379 gras_dynar_get is dead.
3381 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3382 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3383 (which gives you the address of the stored data).
3385 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3386 copy of the data using an affectation to do the job and not a memcpy.
3388 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3389 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3391 It does not really speedup the dynar test because they are
3392 setting elements all the time (and look them seldom). But the dict does
3393 far more lookup than setting.
3395 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3397 2004-10-05 Martin Quinson
3398 - Allow to (en/dis)able the cycle detection at run time.
3400 Whether we should check for cycle or not is now a property of each
3401 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3402 datadesc_cycle_unset allow to remove this property when previously set.
3404 Note that the cycle detection is off by default since it impacts the
3405 performance. Watch the data you feed GRAS with ;)
3407 This property is hereditary. Any element embedded in a structure having it
3408 set have it set for the time of this data exchange.
3410 You should set it both on sender and receiver side. If you don't set it on
3411 sender side, it will enter an endless loop. If you forget on receiver
3412 side, the cycles won't be recreated after communication.
3414 - Header reorganization.
3415 Kill gras_private.h, each submodule must load the headers it needs.
3417 2004-10-04 Martin Quinson
3418 - Interface revolution: do not try to survive to malloc failure.
3420 Now, gras_malloc and friends call gras_abort() on failure.
3421 As a conclusion, malloc_error is not a valid error anymore, and all
3422 functions for which it was the only gras_error_t return value are
3423 changed. They now return void, or there result directly.
3424 This simplify the API a lot.
3426 2004-09-29 Martin Quinson
3427 - Re-enable raw sockets.
3428 Created by gras_socket_{client,server}_ext;
3429 Used with gras_raw_{send,recv}
3432 It should allow to kill the last bits of gras first version soon.
3434 This is not completely satisfactory yet (duplicate code with
3435 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3438 - Simplify transport plugin (internal) interface by not passing any
3439 argument to _server and _client, but embedding them in the socket
3442 2004-09-28 Martin Quinson
3443 - Finish the port to AIX.
3444 autoconf was my problem (segfault within the malloc replacement
3445 function. No idea why)
3447 2004-09-16 Martin Quinson
3448 - Fix some size_t madness on 64bit architectures.
3450 2004-09-08 Martin Quinson
3451 - Reduce the number of system headers loaded, overload some more system
3452 calls (such as malloc to cast the result of the system one, and work
3454 - Fix and reintroduce the config support
3456 2004-09-07 Martin Quinson
3457 - Source code reorganization to allow Arnaud to surf all over there.
3458 - Allow to document the logging categories.
3459 - Remove all uppercase from logging categories and useless cleanup in names.
3461 2004-08-18 Martin Quinson
3462 Version 0.6.2 (protocol not changed; API changed)
3463 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3464 gras_error_t), and NULL when not found. Functions expecting a msgtype
3465 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3466 hopefully usefull message.
3467 - Portability to prehistoric sparcs again
3469 2004-08-17 Martin Quinson
3470 Version 0.6.1 (protocol not changed; ABI not changed)
3471 - prealloc some buffers to speed things up
3473 2004-08-11 Martin Quinson
3474 Version 0.6 (protocol not changed; ABI expended)
3475 - The parsing macro can deal with the references, provided that you add
3476 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3478 2004-08-09 Martin Quinson
3479 Version 0.5 (protocol not changed; ABI changed)
3480 - Allow to off turn the cycle detection code in data exchange at
3481 compilation time. It should be at run time, but I'm short of time (and
3482 the config stuff is still broken). That way, we keep dict out of the
3483 critical path, which is good because the performance is poor:
3484 - search not dichotomial yet
3485 - dynar give no way to access their content and memcpy everytime
3486 - In composed data description (struct, ref and so on), stop foolness of
3487 keeping the subtype's ID, but store the type itself. This keeps sets out
3488 of the critical path, which is good since they rely on dynar and
3489 dictionnaries. The only loose of that is that we cannot detect the
3490 redeclaration of a structure/union with another content (but I'm not sure
3491 the code detected well this error before anyway). We still can detect
3492 the redefinition discrepancy for the other types.
3493 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3494 it breaks the code because of type-punning used all over the place).
3495 This breaks on all non-gcc architectures (for now).
3497 All those changes (plus the buffer of last time) allow me to gain 2 order
3498 of magnitude on cruel tests consisting of 800000 array of integers on two
3499 level of a hierarchical structure (200 secondes -> 4 secondes)
3502 - the selector of reference must now return the type it points to, not
3503 the ID of this type.
3505 2004-08-06 Martin Quinson
3506 Version 0.4 (protocol changed; ABI not changed)
3507 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3508 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3509 - Implement a buffer transport, and use it by default (it relies on tcp in
3510 real life and on sg in simulation).
3511 That's a bit hackish since I had a new field to the structure to store
3512 its data without interfering with the subtype ones. Inheritance
3513 is tricky in C. And that's a kind of reverse inheritance with one class
3514 derivating two classes. Or maybe a game with java interfaces. Anyway,
3515 that's damn hard in C (at least).
3516 Moreover, I got tired while trying to ensure plugin separation and
3517 genericity in SG mode. MSG wants me to do weird things, so let's go for
3518 cruel hacks (temporarily of course ;).
3519 See comment in transport_private.h:71
3520 - do not include all the _interface headers in private but in the files
3521 which really need them (to cut the compilation time when they are
3524 2004-07-26 Martin Quinson
3525 Version 0.3 (protocol not changed; ABI changed)
3526 - Major overhault of the datadesc interface to simplify it:
3527 - shorted the function names:
3528 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3529 - add a trivial way to push/pop integers into the cbps without malloc.
3530 This allows to make really generic sub_type description, which simply
3531 pop their size of the stack.
3532 - add a function gras_datadesc_ref_pop_arr() which does what users want
3533 most of the time: Declare a dynamic array (which pops its size of the
3534 stack) and declare a reference to it. Poor name, but anyway.
3535 - kill the post-send callback, add a post-receive one
3537 2004-07-23 Martin Quinson
3538 Version 0.2 (protocol changed; ABI changed)
3539 - add some testing for cpbs in the test cases, and fix some more bugs.
3540 This invalidate again the little64 data file, since I cannot regenerate
3542 - remove an awfull optimization in the logging stuff, allowing me to:
3543 - understand it again
3544 - learn gcc how to check that the argument match the provided format
3545 - fix all errors revealed by gcc after that
3546 - internal keys of dict are not \0 terminated. Deal with it properly in
3547 loggings instead of segfaulting when the user want to see the logs :-/
3549 2004-07-22 Martin Quinson
3550 - Fix some stupid bug preventing cbps (callback postit) from working
3552 2004-07-21 Martin Quinson
3553 - Some documentation cleanups
3554 - remove the useless last argument of msgtype_declare
3555 - rename the Virtu functions to fit into the 'os' namespace
3556 - move headers src/include -> src/include/gras/ and stop fooling with
3557 gras -> . symbolic link
3558 - make distcheck is now successful
3560 2004-07-19 Martin Quinson
3562 - Build shared library also
3563 - Install html doc to the right location
3564 - stop removing maintainer files in make clean
3565 - build tests only on make check
3567 2004-07-13 Martin Quinson
3569 - No major issue in previous version => change versioning schema
3570 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3571 corresponding dataset.
3573 2004-07-11 Martin Quinson
3575 - Get it working with any kind of structure (we can compute the padding
3576 bytes remotely for all the architectures I have access to)
3577 - Implement the structure parsing macro (still not quite robust/complete)
3578 - Improvement to the remote testing toysuite
3580 2004-07-10 Martin Quinson
3581 [autoconf mechanism]
3582 - get ride of a bunch of deprecated macros
3583 - actually run the test for two-compliment, not only compile it :-/
3584 - test whether the structures get packed (and bail out if yes. Damn.
3585 Alignment is a serious matter)
3586 - test whether the structures get compacted (but respecting the alignment
3587 constraints of each types)
3588 - test whether the array fields of structures can straddle alignment boundaries
3590 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3591 (took me 2 hours to find that bug, looking at the wrong place)
3592 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3593 before sending/receiving and are used to compute the offsets of fields
3594 - Given that padding size depend even on compiler options, keep track of
3595 alignment and aligned_size only for the current architecture. Not a big
3596 deal since we send structure fields one after the other (seems
3598 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3599 program, let it work on linux/gcc/little32. portability todo.
3601 2004-07-08 Martin Quinson
3602 - import and improve remote compilation support from FAST
3603 - make sure make check works on half a dozen of machines out there
3605 2004-07-07 Martin Quinson
3606 Let's say it's version 0.0.3 ;)
3607 - Implement conversions (yuhu!)
3608 - Let it work on solaris (beside conversion, of course)
3609 - Stupid me, using rand() to generate the conversion datatests in not wise.
3611 2004-07-06 Martin Quinson
3612 - Let make dist work, since I'm gonna need it to compile on remote hosts
3613 - Let Tests/datadesc_usage write the architecture on which the file was
3614 generated as first byte.
3615 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3616 architecture descriptions.
3617 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3618 on those architectures.
3619 - Optimization: send/recv array of scalar in one shoot
3621 2004-07-05 Martin Quinson
3622 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3624 - Plug a whole bunch of memleaks
3625 - each process now have to call gras_{init,exit}. One day, their log
3626 settings will be separated
3627 - Continue the code factorisation between SG, RL and common in Transport.
3629 2004-07-04 Martin Quinson
3631 - Redistribution between SG and RL.
3632 We wanna have to accept in SG, so move accepted related parts of RL in
3633 the common part. (more precisely, the dynar of all known sockets is no
3634 more a static in transport.c, but part of the process_data)
3636 [gras_stub_generator]
3637 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3639 2004-07-03 Martin Quinson
3640 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3641 them all up in one shot)
3643 - Fix the ugly case of reference to dynamic array.
3644 - New (semi-public) function gras_datadesc_size to allow the messaging
3645 layer to malloc the needed space for the buffer.
3647 - gras_socket_close now expect the socket to close (and not its address to
3648 put NULL in it after it). This is because the socket passed to handlers
3649 is one of their argument (=> not writable).
3651 - propagate the interface cleanup from last week in datadesc, ie remove a
3652 superfluous level of indirection. User pass adress of variable
3653 containing data (both when sending and receiving), and not of a variable
3654 being a pointer to the data. Let's say that I like it better ;)
3655 The price for that is constructs like "int msg=*(int*)payload" in
3656 handlers, but it's a fine price, IMHO.
3658 - Let it work in RL (yuhu)
3660 2004-06-21 Martin Quinson
3662 - porting SG plugin and SG select to new standards (works almost).
3663 - plug memleaks and fix bugs around.
3666 - cleanup the prototype of data recv and force users to specify when they
3667 want to handle references to objects. Test case working even for cycles.
3668 - plug memleaks. Valgrind is perfectly ok with this.
3670 2004-06-12 Martin Quinson
3672 - cleanup the separation between plugin and main code in plugin creation
3674 2004-06-11 Martin Quinson
3676 - Reput hook for raw sockets, needed for BW experiments
3677 - kill a few lines of dead code
3678 [Data description] Interface cleanup
3679 - gras_datadesc_by_name returns the searched type or NULL.
3680 That way, no variable is needed to use a type desc once, which makes
3682 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3683 two parameters were strings (field name, type name), leading to
3685 [Dicos] Interface cleanup
3686 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3687 This is consistant with the dynar API.
3689 2004-04-21 Martin Quinson
3691 - Porting to new standards.
3693 - interface cleanup.
3694 There is no bag anymore, no need to take extra provision to mask the
3695 pointers behind "ID".
3696 Better splitup of functions between files create/exchange/convert.
3697 This is still a bit artificial since convert and receive are so
3698 interleaved, but anyway.
3700 - add a queued message list to procdata (the ones not matching criteria
3702 - factorize some more code between SG and RL wrt procdata
3704 - use gras_exit in example to track memleaks
3705 - get rid of gs_example now that GS is properly integrated into gras
3706 - update run_test to integrate the latest tests (datadesc)
3708 - rename WARNINGn macros to WARNn since it prooved error-prone
3710 2004-04-19 Martin Quinson
3712 - register init/exit functions within gras module mechanism
3713 - send/receive function.
3714 Convertion is not implemented, but short-cutted if not needed.
3715 struct/array elements are sent one by one (instead of block-wise), but
3716 nobody really cares (yet). Get a prototype before optimizing.
3717 - tests (using a file socket) for DD send/receive on:
3718 - base types: int, float
3719 - array: fixed size, string (ie ref to dynamic string)
3720 - structure: homogeneous, heterogeneous
3721 - chained list, graph with cycle
3722 Believe it or not, valgrind is not too unhappy with the results. The
3723 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3727 - Bugfix when using a filename explicitely (instead of '-')
3729 2004-04-09 Martin Quinson
3731 - factorize more code between RL and SG in socket creation
3732 - Complete the implementation and tests of:
3734 o file (only in RL, and mainly for debugging)
3736 I lost 3 days to design a portable address resolver, and then decided
3737 that the prototype mainly have to run on my box.
3738 Addressing portability too early may be like optimizing too early :-/
3740 - use gras_init in the Tests instead of the crappy parse_log_opt
3741 (the latter function is removed)
3742 [Conditional execution]
3743 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3744 [Code reorganisation]
3745 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3746 Build examples against the RL library, since there is no way to disable
3747 its creation for now.
3749 For information, the beginning of coding on GRAS was back in june
3750 2003. I guess that every line has been rewritten at least twice since