1 SimGrid (3.19.1) NOT RELEASED YET
5 As you may know, we are currently refactoring SimGrid in deep.
6 Upcoming SimGrid4 will be really different from SimGrid3: modular,
7 standard and extensible vs. layered, homegrown and rigid. C++ vs. C.
9 Our goal is to smooth this transition, with backward compatibility
10 and automatic update paths, while still progressing toward SimGrid4.
12 SimGrid remains open during works: The last pure SimGrid3 release was
13 v3.12 while all subsequent versions are usable alpha versions of
14 SimGrid4: Existing interfaces remain unchanged, but the new S4U
15 interface is budding and the internals are deeply reorganized.
17 Since 2015, we work hard to reduce the changes to public APIs. When
18 we need to rename a public library symbol in S4U, we let your
19 compiler issue an explicative warning when you use the deprecated
20 function. These messages remain for four releases, i.e. for one full
21 year, before turning into an error. Starting with v3.15, your can
22 also adapt to API changes with the SIMGRID_VERSION macro, that is
23 defined to 31500 for v3.15, to 31901 for v3.19.1 and so on.
25 Starting with this v3.19.1, our commitment to reduce the changes to
26 the public interfaces is extended from the API to the ABI: a program
27 using only MSG or SimDag and compiled against a given version of
28 simgrid can probably be used with a later version of SimGrid without
29 recompilation. We will do our best... but don't expect too much of
30 it, that's a really difficult goal during such profund refactoring.
32 The difference between v3.19 and v3.19.1 is that the former was
33 accidentally breaking the ABI of MSG, while the later is restoring
36 S4U and kernel APIs will still evolve until SimGrid4, with one-year
37 deprecation warnings as currently. In fact, cleaning up these
38 interfaces and converting them to snake_case() is one release goal of
39 v3.20. But don't worry, we are working to smooth this upgrade path.
41 ----------------------------------------------------------------------------
43 SimGrid (3.19) Released March 20 2018
45 The Moscovitly-cold Spring release.
47 SMPI user-visible improvements:
48 - Support of MPI_Type_create_subarray
49 - Unimplemented MPI calls now abort (instead of issuing a warning)
50 Some harmless calls around errhandlers still only issue a warning.
51 - MPI_Sendrecv was behaving badly when MPI_PROC_NULL was sender or receiver.
52 - Starting MPI ranks (or even MSG/S4U actors) once the simulation started
53 is now possible (mmap privatization was refactored to this end).
54 Adding new SMPI instances should also be technically doable now.
56 SMPI internal cleanups:
57 - Removed index notion from SMPI -- no more "getPid() - 1"!
58 In tracing files, actors are called rank-<PROCESS-ID> but these are
59 NOT the real ranks (tracing several communicators is not supported yet).
60 - Deprecate smpi_process_index() (will be removed in 3.22)
61 - Better testing on classical proxy apps, through an external project:
62 https://github.com/simgrid/SMPI-proxy-apps/
65 - Use a graphical TOC to make it easier to find the documentation you need
66 - Revamp the MSG tutorial
69 - Fix MSG_task_get_remaining_work_ratio(): return 1.0 for unstarted tasks.
70 - Remove parameter of MSG_process_killall().
71 Resetting the PID was bogus anyway (several actors could have the same PID).
74 - Execution->setHost() can be called after start(), to migrate it.
75 - Comm::test_any() is now implemented.
76 - s4u::Actor now has onCreation() and onDestruction() signals.
77 - Install some kernel header files for the users' plugins and more.
78 Warning, their API is really not stable yet.
79 include/simgrid/kernel/resource/{Action,Resource,Model}.hpp
80 include/simgrid/kernel/routing/*Zone.hpp
83 - SD_task_dependency_add(): remove unused parameters 'name' and 'data'.
85 Build System and other cleanups:
86 - Remove unused run-time parameter "tracing/onelink-only".
87 - SimGrid now works with intel compilers (MC still buggy)
88 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
89 - tesh kills subprocesses on timeout (but not on windows)
92 - Rename FULLDUPLEX into SPLITDUPLEX (old name still accepted)
95 - #194: Feature request: simgrid::s4u::Comm::test_any()
96 - #245: migrating an actor does not migrate its execution
97 - #254: Something seems wrong with s4u::Actor::kill(aid_t)
98 - #256: Modernize FindSimGrid.cmake
99 - #257: Fix (ab)use of CMake install
101 ----------------------------------------------------------------------------
103 SimGrid (3.18) Released December 24 2017
105 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
107 MSG IS NOW DEPRECATED. Please use S4U for new projects.
108 - Support for MSG should not be removed from SimGrid before 2020,
109 but future evolutions will be limited to the integration of
110 user-provided patches.
113 - s4u::onDeadlock() signal, run before stopping the simulation.
114 - s4u::Exec: asynchronous executions (abstraction of background
116 - s4u::Host->getLoad() returns the achieved speed in flops/s
119 - Link::name() is deprecated and will be removed in v3.21, use
120 Link::getCname() instead.
121 - Mailbox::getName() changed to return a std::string, use
122 Mailbox::getCname() instead to get a char*.
123 - Storage::getName() changed to return a std::string, use
124 Storage::getCname() instead to get a char*.
125 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
126 to have both allocation and deallocation of the map in user space.
129 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
130 The amount of flops remaining to do is a vector, not a scalar.
131 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
132 name implies on both sequential and parallel tasks.
133 - Both changes fix GitHub's #223 using PR #237 as a basis.
134 Thanks Michael Mercier.
135 - Most examples were converted to S4U and hidden elsewhere at tests.
136 Rationale: we still want MSG to work; we want newcomers to use S4U.
139 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
140 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
143 - Switch to the faster dlopen privatization mechanism by default
144 - Documentation improvments
147 - Live migration is getting moved to a plugin. Dirty page tracking is
148 the first part of this plugin. This imply that VM migration is now
149 only possible if one this function is called :
150 - C/MSG: MSG_vm_live_migration_plugin_init()
151 - C/C++: sg_vm_live_migration_plugin_init()
152 - Java: Msg.liveMigrationInit()
153 For C and C++, "simgrid/plugins/live_migration.h" also has to be
157 - Define class simgrid::xbt::Path to manage file names.
158 - Removed unused functions:
159 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
160 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
161 xbt_graph_export_graphviz()
162 - xbt/str.h: xbt_str_join()
163 - Remove unused datatypes:
164 - xbt/heap.h: use std::priority_queue or boost::heap instead
165 - xbt/swag.h: use boost::intrusive::list instead
168 - New link_energy plugin for the consumption of the links.
169 - All of the operations on files and storage contents have been
170 packaged into a plugin (src/plugins/file_system). The current
171 public interface can be found in
172 include/simgrid/plugins/file_system.h
173 To use these functions you now have to initialize the plugin by
174 calling MSG_storage_file_system_init() just after calling
175 MSG_init() or sg_storage_file_system_init() just after creating
179 - Remove the undocumented/untested tag <include>
182 - Remove viva specific tracing as the tool is no longer maintained
185 - #248: Kill of finished processes leads to segfault
186 - #240: xbt_cond_wait_timeout should gracefully return for C apps
187 - #239: please implement signal s4u::onDeadlock()
188 - #230: segfaults when exit() without run()
189 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
190 - #223: MSG_task_get_flops_amount() not working with parallel tasks
191 - #222: Actor::kill() doesn't really kill and segfaults
192 - #221: odd LMM warning when killing an actor
193 - #120: Memory leak when the processes are forcefully killed
195 ----------------------------------------------------------------------------
197 SimGrid (3.17) Released October 8 2017
199 The Drained Leaks release: (almost) no known leaks despite the tests.
201 Even very long simulations will preserve your memory: our testsuite
202 stresses SimGrid for over 45mn on fast machines for a coverage of
203 over 80%, and there is only one single known leak, of about 4kb.
207 - Comm.detach(): start and forget about asynchronous emission. A cleanup
208 handler may be given to free resources if the comm cannot be completed.
209 - this_actor::send(mailbox) is now mailbox->put()
210 - New: simgrid::s4u::Comm::wait_all()
211 - New: Host.execute() for remote executions.
214 - Unused option network/sender-gap is removed.
217 - Deprecate MSG_task_isend_with_matching(): This unused feature
218 really complicates our internals. Will be removed in v3.20.
221 - Improved context termination. It is now possible to free resources, even
222 when a process is forcibly killed.
223 - ContextBoost: add support for Boost versions above 1.61.
226 - Bring back run-time option --cfg=exception/cutpath to remove exception
228 - Removed unused functions:
229 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
230 xbt_str_rtrim(), xbt_str_trim().
231 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
234 - Removed header files obsolete since SimGrid 3.12:
235 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
237 - #3: SD_exit should be made optional
238 - #120: Memory leak when the processes are forcefully killed
239 - #159: Threading test regression in Actor refcounting
240 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
241 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
242 - #186: Actor::killAll() segfaults if some process is blocked on wait()
243 - #191: VM migration and pstate
244 - #192: Updating the energy consumptions of all hosts crashes with VMs
245 - #195: All actors have PID=0 in the logs
246 - #204: Sometimes segfault with thread contexts and mmap privatization
247 - #222: Actor::kill() doesn't really kill and segfaults
248 - #225: Actor::kill() doesn't really kill when victims are doing a join()
250 ----------------------------------------------------------------------------
252 SimGrid (3.16) Released June 22. 2017.
254 The Blooming Spring Release: developments are budding.
257 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
258 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
259 - Java JAR file should be finally fully working on Mac OSX too.
261 XML platforms: Switch to platform v4.1 format.
262 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
263 - <zone> can be used as a synonym for the now deprecated <as>
264 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
265 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
266 - <actor> can be used as a synonym for the now deprecated <process>
267 - state_file and avail_file periodicity is now easier to express
268 (check the documentation) the old behavior should still work.
269 * Storage: not backward compatible, but it seems that nobody ever used it
270 - Remove attribute 'content_type' of <storage_type>: was never used
271 - Make attribute 'model' of <storage_type> optional: for future usage
272 - Remove Bconnection model property: was never (in)validated. Replaced by
273 the maximum of the read and write bandwidth as a resource constraint.
276 - New and Backwards Compatibility break:
277 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
278 When one wants to get the list of tasks whose states have changed during a
279 simulation round, s/he has to allocate and free a dynar and use it as
280 argument to this function. The former SD_simulate (double how_long)
284 - Allow multicore VMs, with the correct sharing (unless you overcommit)
285 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
288 - New function to update the consumption of all hosts at once.
289 - Fix the model for multi-core hosts, linear in the amount of busy
290 cores with an abnormality for pIdle (see paper).
291 BUG: Still not satisfactory for ptask on mono-cores.
294 - The netzone are now available from the MSG API.
295 The old names still work, but are now deprecated.
298 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
299 - New option: smpi/keep-temps to not cleanup temp files
300 - New option : smpi/shared-malloc-blocksize . Relevant only when global shared
301 mallocs mode is used, allows to change the size of the fake file used
302 (default 1MB), to potentially limit the number of mappings for large runs.
303 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
304 - Fortran ifort and flang compilers support
305 - New RMA calls supported (experimental) :
306 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
307 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
308 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
309 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
310 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
311 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
314 - New: this_actor::isMaestro() returns whether we are in kernel mode.
315 - Behind the scene, ActivityImpl are now automatically refcounted.
316 Removing this nasty bottleneck will greatly simplify our internals.
319 - Replay: New function xbt_replay_action_get():
320 Retrieve the function previously associated to an event type.
321 - DROPPED FUNCTION: xbt_str_varsubst()
322 - DROPPED MODULE: strbuff. We don't need it anymore.
323 - DROPPED MODULE: matrix. We don't need it anymore.
324 - DROPPED MODULE: lib. We don't need it anymore.
326 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
328 SimGrid (3.15) Released March 22. 2017
330 The Spring Release: continuous integration servers become green
331 We fixed even the transient bugs on all target architectures:
332 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
333 Windows, FreeBSD, NetBSD.
335 - Rename NetCards to NetPoints.
336 This was intended to help NS3 users, but that's not a netcard.
337 That's a point in the routing algorithm, let's avoid wrong simplifications.
340 - New: MSG_process_yield(). Stop and yield to other processes.
341 - New: MSG_process_daemon(). Daemon processes are automatically killed
342 when the last non-daemon process terminates
343 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
344 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
345 to make room for the upcoming network energy plugin.
346 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
349 - Ensure that an actor can kill itself with Process::exit()
350 - Kill the obscure NativeException. Nobody want to survive the issues
351 it denotes, so use JniException that is a RuntimeException (not to
352 be caught explicitly).
353 - Partial bug fix in initialization. SimGrid flags on command line were
354 consumed at C level but stayed in the original Java String[] args.
355 This could mess users' args[i] if SG flags were not put at the end of
357 The SimGrid flags are now removed from the Java arguments. However,
358 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
359 if args.length is greater than the number of YOUR OWN ARGUMENTS.
360 It might be if you have --log or --cfg flags in the command line.
361 - Fix numerous memleaks all around the place. In particular, around VMs.
365 - s4u::onPlatformCreated: right before the simulation starts
366 - s4u::onSimulationEnd: right after the main simulation loop
367 - s4u::onTimeAdvance: right after a clock change (time is discrete)
368 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
369 event from the availability_file changes the avail speed.
370 - Links are now usable from s4u
371 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
372 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
373 - New examples: The conversion of MSG examples to S4U has begun
374 - Actors: create, kill, migration, and suspend
375 - Applications: master-worker and token-ring
376 - Action replay: communications and storage
377 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
380 - Backwards Compatibility breaks
381 - The SD_route_get_list and SD_route_get_size functions have been removed.
382 They are replaced by the unique
383 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
384 The route from src to dst is built in the links dynar whose size can
385 be obtained with xbt_dynar_length.
386 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
387 sg_host_route_bandwidth and sg_host_route_latency.
388 Macros ensure the backwards compatibility, but you should fix your code
391 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
392 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
393 - Bump our claim of support from MPI 1.1 to MPI 2.2.
394 We don't support 100% of it, but it should be enough. Contact us if not.
395 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
396 naive and probably inaccurate)
397 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
398 mapped onto a unique small file using some system magic.
399 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
400 - Fortran: cleanups, fixes, support of user-added operations
401 - RMA: MPI_Accumulate are applied in correct order.
402 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
403 - Better support for MPI_IN_PLACE
404 - Support for MPI_Win attrs and keyvals.
405 - Support MPI_Comm_group_create, MPI_Type_size_x
406 - MPI ops should be properly applied to all allowed types
407 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
408 algorithm by a non failing one from mpich.
411 - Kill the fifo data container: we don't use it anymore.
414 - A <cluster> can now be created with different speed values to represent pstates.
415 see examples/platforms/energy_cluster.xml for an example.
417 SimGrid (3.14.159) Released December 28. 2016
419 The previous release was only a crude approximation of a Pi release;
420 this one should be a bit better.
422 - Revert a last minute change that broke on Mac OSX.
423 - Fix the changelog and NEWS files.
424 - Improve the documentation, in particular of the routing module.
425 - Integrate some patches flying around in Debian.
427 SimGrid (3.14) Released December 24. 2016
430 * The whole documentation was reworked and reorganized. There is
431 still much room for improvement here, but we're on it.
434 * We now need python3 for our testing infrastructure.
435 * Model checking needs libevent
438 * Now works on FreeBSD
441 * Privatization now works on FreeBSD
443 * Privatization is now activated by default in smpirun
444 You can turn it off with -no-privatize if something goes wrong.
446 * Call-location tracing for SMPI.
447 You can add the exact location (filename / linenumber) of an MPI call to
448 your trace files and slow down or speed up the simulation between two
449 consecutive calls by using an adjustment file (see the documentation).
451 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
452 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
453 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
454 values were ignored in some cases. The timings of these functions can now
455 be significantly different.
457 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
458 smpi/running-power is renamed to smpi/host-speed
460 * smpi/grow-injected-times option to enable or disable multiplication of the
461 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
462 make simulation less precise (but also much faster).
464 * smpirun script should be (much) faster for large deployments.
466 * SMPI tracing : fixed issue with poor matching of send/receives.
468 * Replay : Fix broken waitall
470 New functions and features
471 * MSG_parallel_task_execute_with_timeout, to timeout computations.
473 Dropped / renamed functions and features
474 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
475 - MSG_mailbox_is_empty() -> Mailbox::empty()
476 - MSG_mailbox_front() -> Mailbox::front()
477 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
478 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
479 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
480 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
481 * Don't pass the free_f parameter to property related functions:
482 - MSG_host_set_property_value()
483 - MSG_as_router_set_property_value()
484 - MSG_storage_set_property_value()
485 * VM properties. Since msg_vm_t are msg_host_t, just use
486 MSG_host_get_property_value() and friends
487 * VM I/O related things:
488 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
489 - Unimplemented save/restore methods
490 * MSG_as_router_get_property_value() was redundent with
491 MSG_environment_as_get_property_value().
492 - Removed MSG_as_router_*propert*() functions
493 - Added MSG_environment_as_set_property_value() for consistency
494 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
495 Well, they are still there for now, but deprecated with a warning.
496 Please switch to xbt_dict_new_homogeneous() before this is removed
498 * Task affinity. Its intended behavior (that was very badly tested
499 and probably not really working) was deceiving what most users
500 would have hoped here.
501 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
502 * Option network/coordinates is now useless and should be dropped.
505 * Added option storage/max_file_descriptors to allow more than 1024 files opened
508 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
509 * Added Dragonfly topology support
512 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
513 EiBps, EBps, Eibps, Ebps for bandwidth.
514 They may become useful to some lucky ones.
517 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
518 * ex/app_tokenring: new example, very similar to the MSG Token Ring
519 * ex/async_waitAll: new example, on asynchronous communications
522 * Memory usage should be decreased for simulations with a large number
523 of processes. This also helps for SMPI.
525 SimGrid (3.13) stable; urgency=low
527 The Half Release, a.k.a. the Zealous Easter Trim.
529 - We removed half of the lines, that were mostly experimental cruft.
530 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
531 (not counting blanks and comments -- according to openhub.net)
532 - The internals are now compiled in C++ (and will soon be clean C++)
533 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
534 59 klines of C, etc. We added only 29 klines of C++ in replacement.
536 * Backwards Compatibility breaks
537 - Removed Lua simulation bindings (switch to C or Java for that).
538 Lua can still be used to describe platforms
539 - Removed Java kernel plug-ins.
540 Will be reintroduced after the ongoing major internals reorg.
542 - the following functions were removed.
543 They were too specific and should be reimplemented in a generic
544 way, with filter function.
545 - MSG_task_listen_from_host
546 - MSG_mailbox_get_count_host_waiting_tasks
547 - MSG_mailbox_put_with_timeout was removed.
548 Please use MSG_task_send_with_timeout instead.
550 - the SD_application_reinit function was removed. It has been a noop for a while.
551 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
552 replaced by a more flexible mechanism.
553 - The following functions thus do not exist anymore
554 - SD_workstation_get_access_mode
555 - SD_workstation_set_access_mode
556 - SD_workstation_get_current_task
557 - Basic estimation functions have been removed but can easily be replaced
558 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
559 - SD_workstation_get_computation_time => amount / sg_host_speed()
561 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
562 Use VM.getSpeed()*load/100 for the legacy behavior.
564 - option enable_tracing was removed. It was not doing anything for a while.
565 - In the ModelChecker:
566 - the model-checker now ptraces the model-checked process which means
567 you cannot use a debugger on the latter anymore (we might make this
568 optional in the feature);
569 - removed soft-dirty page tracking;
570 - remove model-checked side snapshot management,
571 MC_snapshot() and MC_compare_snapshot();
572 - keep the MC_cut() function as a stub (it was not really working
573 in the previous release).
576 * All options are consistently in kebab-case. Old names are kept as alias.
579 * Switch to platform v4 format.
580 - Rename from 'power' to 'speed' the attributes describing the amount of
581 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
582 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
583 - In <host> and <link>, attributes availability and state are gone.
584 It was redundent with state and availability traces, and with peak values.
585 - In <cluster>, attributes availability_file and state_file are gone.
586 It was too complex and unused.
587 - Kill <gpu>. Was not doing anything.
588 - The DOCTYPE points to the right URL:
589 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
590 (the file at this address now documents the changelog since its v1)
591 - A warning is emitted for unit-less values (they are still accepted).
592 - speed. Default: 'f' or 'flops'. Also defined:
593 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
594 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
595 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
596 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
597 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
598 - latency. Default: 's' second. Also defined:
599 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
601 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
602 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
605 * s4u::Host is now the preferred public interface to the Host features.
606 sg_host_* functions are C bindings to the exact same behavior
607 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
610 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
611 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
612 They were deprecated since a few years, and probably did not even compile anymore.
615 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
616 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
617 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
618 encouraged to update their codes. A list of the modified functions can be found at the end of
619 include/simgrid/simdag.h
622 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
623 the lambda or closure passed as a parameter will run in kernel mode.
624 All the callback functions should be rewritten to that interface at some point.
627 * Reorganizing and cleaning the internals all around the place.
630 * Remove old default barrier/bcast buggy algorithms (see #18407)
631 * Various bug fixes to handle more codes
632 * Remove the need for the --foreground option of smpirun (it is still
633 accepted for backward compatibility).
636 * Kill the setset data container: MC don't use it anymore.
637 * Kill the queue data container: it made more sense with GRAS.
638 * Kill the xbt_peer_t data type: it's useless without GRAS.
639 * Kill rm_cb feature of config sets: it was never useful.
640 * Kill graphxml parsing feature. It was not used.
641 * Kill the deprecated code protected by XBT_USE_DEPRECATED
643 - xbt_dynar_sort_strings(), when the content is char*
644 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
645 They throw exceptions on invalid input;
646 * C++ support for declaring CLI flags (simgrid::config::Flag);
647 * class for abstracting different signal backends (simgrid::xbt::signal).
648 with no external dependencies (we need very simple signals).
651 * refactoring and cleanup of the code;
652 * ongoing process to cleanly separate the model-checking algorithms
653 from the code model-checking support.
655 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
657 SimGrid (3.12) stable; urgency=low
659 The Facelift Release.
662 * Require g++ v4.7 at least to not speak prehistorical C++.
663 * Require Boost 1.48 (for signal2 component).
664 * Java must be version 7 at least when activated.
665 * Builds on Windows again (including Java bindings).
666 * Tracing is now always enabled (no way to turn it out)
667 * Remove GTNetS. It was not working anyway.
668 * Various cleanups in the cmake scripts.
669 * Move headers around to sort them out on installed systems:
670 - instr/instr.h -> simgrid/instr.h
671 - instr/jedule/* -> simgrid/jedule
672 - simdag/datatypes.h was removed
673 - simdag/simdag.h -> simgrid/simdag.h
674 - msg/datatypes.h was removed
675 - msg/msg.h -> simgrid/msg.h
678 * Interface improvement:
679 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
680 - Sanitize the interface in MSG_task_ module:
681 - Merge two functions that were close enough but misleading:
682 set_compute_duration(t) -> set_flops_amount(t)
683 get_remaining_computation(t) -> get_flops_amount(t)
684 - set_data_size(t) -> set_bytes_amount(t)
685 get_data_size(t) -> get_bytes_amount(t)
686 - Massive cleanups in the functions related to the energy
687 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
688 - New: MSG_host_get_pstate()
690 - msg/energy/onoff: switching hosts on and off
692 * Interface improvement:
693 - Rename Host.isAvail() to Host.isOn()
694 - Rename Process.currentProcess() to Process.getCurrentProcess()
695 - Rename Task.setDataSize() to Task.setBytesAmount()
696 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
698 - #18874: Actually allows the GC to reclaim tasks
702 - SIMIX_process_throw: raises an exception in a remote process
703 * Refactoring: Separate sub-modules
704 - libsmx: the public interface, as libc in a real system
705 - popping: the strange dance that converts a user request into a kernel handling
706 - smx_context_*: the virtualization mechanisms that embed the user code
707 - smx_*: the handling of each simcalls
709 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
710 * Rename smx_action_t into smx_synchro_t, making explicit that these
711 things are used to synchronize processes with their environment.
712 For example, a communication is a sort of synchronization involving
713 the communicating processes (that may block until the exchange) and
714 the platform. The same can be said from computations, etc.
716 - #18888: segfault when a process ends before its kill_time
720 - 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.
721 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
722 - MPI_Type_set_name, MPI_Type_get_name
723 - MPI_*_c2f and MPI_*_f2c functions
724 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
725 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
726 - Activate a lot of new tests from the mpich 3 testsuite
728 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
729 - InfiniBand network model added : Based on the works of Jerome Vienne
730 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
731 - When smpi/display_timing is set, also display global simulation time and application times
732 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
733 * Collective communications
734 - 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
735 - 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/).
736 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
738 - Replay now uses algorithms from wanted collective selector
739 - Replay can be used with SMP-aware algorithms
740 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
741 - 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.
743 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
744 - Correctly use loopback on fat-tree clusters
745 - Asynchronous small messages shouldn't trigger deadlocks anymore
746 * Energy/DVFS cleanup and improvement
747 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
748 - new: smpi_get_host_pstate()
751 - "Full" network optimization flag was broken since Surf++
752 - Better handling of precision flags in maxmin
753 - Fix bug causing sometimes "Impossible" errors
754 - Properly pass cluster properties to included hosts
755 * Improvement of the Energy plugin.
756 - Always update the consumption before returning that value
757 - New property: watt_off to denote the dissipation when the host is off
758 - New functions getWattMinAt and getWattMaxAt to retrieve the
759 dissipation of pstates that we are not currently at.
760 * Java: class NetworkLink renamed to Link
761 * New function: simcall_process_get_kill_time()
762 * Massive rename s/workstation/host/
763 - That's intrusive, but that's good for the project consistency. Sorry.
764 - Change config option "workstation/model" into "host/model"
767 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
768 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
769 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
770 - xbt_dirname and xbt_basename for non-POSIX systems
772 * The model checker now runs as a separate process.
773 * The model checker runs must now be launched with the new simgrid-mc program.
774 * Record/Replay: the MC can display a textual representation of a path in the
775 execution graph. It can then be replayed outside of the model checker.
777 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
779 SimGrid (3.11) stable; urgency=low
784 * Normalizing pointers addresses tool for better diff between logs
787 * Add cloud examples using new VMs
788 - examples/msg/cloud/two_tasks_vm.tesh
789 - examples/msg/cloud/simple_vm.tesh
790 - examples/java/cloud/cloud.tesh
791 - examples/java/cloud/migration/migration.tesh
792 * Add java surf examples:
793 - examples/java/surfPlugin/surf_plugin.tesh
794 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
795 - examples/java/surfCpuModel/surf_cpu_model.tesh
796 * Add SMPI+MSG example:
797 - examples/smpi/smpi_msg_masterslave/
802 - msg task destroy cancel test
803 - msg_host on/off test
804 * Move all tests in testsuite to teshsuite (adding tesh files)
805 * Restructure teshsuites
806 - one folder for each kind of test
807 * Restructure AddTests.cmake
809 - structure the order of tests (with sections)
812 * Add virtual machine
813 - creation of a VM on a PM
814 - migration of a VM from a PM to another PM
816 - MSG_process_join(msg_process_t process, double timeout)
817 - msg_bar_t MSG_barrier_init(unsigned int count)
818 - int MSG_barrier_wait(msg_bar_t barrier)
819 - void MSG_barrier_destroy(msg_bar_t barrier)
820 - msg_as_t MSG_environment_as_get_by_name(const char * name)
821 * New option "msg/debug_multiple_use" to help debugging when a task is used
824 - Improvements and finalization of MSG_storage, MSG_file APIs and their
826 - Increase code coverage in test suites
830 * Protect context stacks against stack overflow. The number of protected memory
831 pages allocated on the top of each stack (1 by default) can be configured
832 with the new command line option --cfg=contexts/guard_size:<value>.
833 * Simcalls are now generated by a python script that
834 - generates files included by SimGrid
835 - checks that all the functions exist, and proposes prototypes
837 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
838 host_set_data, host_get_data
840 - simcall_process_join(smx_process_t process, double timeout)
841 * Fix bug where sleeping processing could not be suspended.
844 * Translate surf models from C to C++
845 - Generic classes for all models: Model, Resource, Action
846 - A generic interface for each kind of model (CPU, Network, Storage
847 Workstation, WorkstationVM)
849 * Translate surf routings from C to C++
850 * Add callbacks using sigc++ or boost::signals2
851 - Add callback functions for resource creation/destruction
852 - Add callback functions for action state change
853 - Handle Energy as a plugin
854 * Replace swag by boost::intrusive
855 * Add new routing models for clusters. For documentation, see
856 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
857 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
858 parameters for cluster tag
859 - Fat trees, with topology="FAT_TREE" and
860 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
861 - see examples/platforms/torus_cluster.xml and
862 examples/platforms/fat_tree_cluster.xml
866 * Hostfiles support host:nb_processes construct to deploy several processes on
868 * Collective communication algorithms should not crash if used with
869 improper number of nodes and report the error.
870 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
871 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
872 MPI_Dims_create, MPI_Cart_sub are supported.
873 * New interface to use SMPI programmatically (still depends on MSG for
874 some parts, see examples/smpi/smpi_msg_masterslave) :
875 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
879 * Global variables privatization in MPI executables is now performed at runtime
880 with the option smpi/privatize_global_variables (default:no).
881 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
882 dynamic libraries loaded by the application are not privatized (static
883 linking with these libraries is advised in this case)
886 * Options defined in XML work correctly now.
889 * New cmake option, enable_lib_in_jar, to control whether native libraries are
890 copied into simgrid.jar or not (ON by default). Use this option if you want
891 to reduce the size of the installed simgrid.jar, *and* the native libraries
892 are kept installed elsewhere.
893 * Surf binding with SWIG (code generated in maintainer mode only):
894 - plugin to handle callbacks
895 - CPU model only for the moment
898 * Supernovae build mode is definitively removed. It was used to improve
899 inlining and inter-module optimizations. It is nowadays superseded by
900 link-time optimizations commonly available in compilers.
901 * Update ns-3 find lib. Bindings for ns-3 should work again now.
902 * Add boost dependency for surf++
903 * Add new macro for tests
904 - ADD_TESH(name <tesh_args>)
905 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
909 - xbt_bar_t XBT_barrier_init(unsigned int count)
910 - int XBT_barrier_wait(xbt_bar_t barrier)
911 - void XBT_barrier_destroy(xbt_bar_t barrier)
912 * Make the xbt_os_time module public
914 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
916 SimGrid (3.10) stable; urgency=low
918 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
921 * Reintegrate Java to the main archive as desynchronizing these
922 package is not acceptable anymore (Java is now considered stable)
923 * Add explicit synchronization facilities through semaphores
924 * Bug fix: Task.setDataSize() only changed the C world, not the value
925 cached in the Java world
928 * Dramatically change the way files are handled. API and internals changed, but
929 this part of MSG was not considered as production grade either.
930 * Add explicit synchronization facilities through semaphores
931 * Add a new function MSG_host_get_process_list()
932 * Preliminary DVFS support (see examples/msg/energy/ for details)
935 * SMPI is now included directly in the libsimgrid as the windows
936 linker doesn't force us on splitting it anymore.
937 * Improvements of the SMPI replay tool:
938 - Most of the collective communications are now rooted in the same process as
939 in the original application.
940 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
941 used until now). Multiple data types can now be used in a trace.
942 - The replay tool now supports traces produce either by TAU or a modified
944 - Bug Fix: the compute part of the reduce action is now taken into account.
945 - Gatherv collective is now supported
946 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
947 of smpirun outputs time independent traces for the current run. One file
948 is created per process. If too many processes are simulated, this behavior
949 can be changed to one file for all processes by using the
950 tracing/smpi/format/ti_one_file flag
951 * smpirun generates the host file if needed (with given host count and platform)
952 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
953 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
954 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
955 --cfg=smpi/coll_selector:(mpich/ompi)
956 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
958 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
959 happening inside a collective SMPI call.
960 * Fix the behavior of complex data types handling.
961 * Make MPI_Wtime another synchronization point to take computations into
963 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
964 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
965 * Add all missing Fortran bindings, SMPI should work with Fortran 90
966 (no privatization of global variables yet)
967 * Preliminary DVFS support (see examples/smpi/energy/ for details)
970 * Verification of liveness properties is now available for SMPI applications
971 (in addition to MSG applications)
972 * Bugged examples using SMPI in examples/smpi/mc/
973 * Add --cfg=model-check/visited option. Allows the verification of infinite
974 programs. Detection of loops in the execution thanks to the system state
975 comparison and reduction of the state space to explore. Can be combined with
976 DPOR for safety properties.
979 * Allow to change SimGrid configuration (see --help) within the code
980 thanks to SD_config() as it can be done in MSG.
981 * Add a new function SD_task_set_amount() upon user request.
984 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
985 * Remove rule based routing (no more PCRE dependency)
986 * Add a limiter_link option to cluster tag, to specify a maximum reachable
987 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
988 * Add a loopback_bw and loopback_lat options to cluster tag.
989 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
990 routing. Now peers are encapsulated in an AS and have their own private
991 router but this is transparent.
994 * Our own implementation of getline is renamed xbt_getline, and gets
995 used even if the OS provide a getline(). This should reduce the
996 configuration complexity by using the same code on all platforms.
997 * New type: xbt_cfg_elm_boolean.
998 * Allow to use yes/no for boolean configuration options in the command line.
999 * Allow to disable SimGrid cleanups at exit from command line option.
1000 There are situations where one may want a simulation to end with an exit.
1001 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
1002 annoying when scripting around the simulator. Adding a
1003 --cfg=clean_atexit:no allows to circumvent this issue.
1006 * Lots of memory leaks were corrected in this release.
1007 There is no known memory leaks anymore, in all of our 600+ tests.
1008 * New command line option --version, to get SimGrid version information.
1009 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
1011 * Supernovae builds are deprecated, and expected to be removed in the next
1014 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1016 SimGrid (3.9) stable; urgency=low
1018 The Grasgory release: GRAS is really dead now.
1020 * Complete overhaul of the internal host structures scheme.
1023 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
1024 was considered as experimental and badly maintained since then.
1025 * Keeping it was thus a trap to our potential users, that could take
1026 it instead of MSG or SMPI by mistake despite is pity state.
1027 * GRAS seems to have very few users (if any), and no one volunteered
1028 to maintain it further. It also induces a lot of XBT code (for
1029 portability sake), that must be maintained too.
1030 * For all these reasons, we killed GRAS. If someone wants to revive it
1031 in the future, don't cry, our git history still remembers of GRAS.
1034 * Major overhaul. Merge our documentation again as time proved that
1035 splitting it was really not helping our users.
1036 * Further improve the developer documentation to help newcomers
1037 hacking on SimGrid itself. The user documentation (and in
1038 particular, the beginner documentation) is still in a sorry state.
1041 * Now works on Windows too!
1042 * Much more extensive test suite, from MPICH
1045 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
1046 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
1047 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
1048 * Bug fix: let task be scheduled when the last dependency to be solved is
1049 a control dependency.
1050 * Remove SD_load_environment_script function.
1051 Use the C sg_platf function if you want to declare a platform
1055 * New function: MSG_process_get_number()
1056 * Old function documented: MSG_config()
1057 * Remove MSG_load_platform_script function
1058 Use the C sg_platf function if you want to declare a platform
1062 * Change the default value of the TCP_gamma constant (maximal size of TCP
1063 congestion window) to a more realistic 4MiB value. If you notice changes in
1064 your simulation results, you can fall back to the previous 20k tiny window
1065 by adding --cfg=network/TCP_gamma:20000 on command line.
1066 * (Hopefully) fix a bug wrt periodic availability and state traces
1067 * Bug fix: use default values at start when first event in availability/state
1068 trace is not at time 0.
1071 * remove the "new_" part of function name sg_platf_new_trace_connect
1072 (resulting in sg_platf_trace_connect), since it does not create
1076 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1077 use case where it's really mandatory, and maintaining it was a pain in
1079 * New: xbt_fifo_search(), search an item with a user-provided
1080 comparison function instead of dumb pointer comparison.
1083 * Fix the lua deployment:
1084 Use `simgrid.init_application()` before deployment instead of
1085 `simgrid.msg_register_application()` after.
1088 * Transfer the tracing files into the corresponding modules.
1090 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1092 SimGrid (3.8.1) stable; urgency=low
1094 The "we are told that some people want to also *install* the simgrid
1097 * Add missing file "tesh.1" to the archive.
1099 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1101 SimGrid (3.8) stable; urgency=low
1103 The Psssshiiiit release: SimGrid jumps into the Cloud.
1106 * Add an experimental interface to manipulate VMs. They are mainly
1107 process groups with very few intrinsic semantic, but they should
1108 allow you to build the semantic you want easily.
1109 * New function: MSG_host_set_property_value()
1110 * New function: MSG_process_on_exit(). To clean memory in all cases.
1111 * Bug fixes that made the host (and link) failures unusable.
1112 * Add a way to auto-restart process when the host in which they are
1113 executing comes back (ON_FAILURE="RESTART" on deployment file,
1114 MSG_process_auto_restart_set).
1115 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1116 please stop using the old ones, they are DEPRECATED.
1118 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1119 Please use MSG_init() instead. (reducing the amount of entry
1120 points in the library helps us).
1121 * Make it impossible to link against the wrong version of the lib
1122 * Deprecate MSG_clean(). No need to call it anymore.
1123 * Function MSG_get_host_number() is not deprecated anymore.
1126 * Split the doc into a user guide and a reference guide.
1127 * Start a developper guide to help people hacking on SimGrid.
1130 * Enable tracing by default. This modules rocks you should use it.
1131 * Remove option custom_flags. Now use environment variables CFLAGS
1133 * Use default cmake things to detect lua instead of home grown ones.
1134 * New option "enable_mallocators" to disable mallocators, for debugging
1135 purpose ("on" by default).
1138 * Bug fixes around the resource failures: don't let the processes
1139 survive the host they are running onto.
1140 * Add an interface to auto-restart processes when the host in which they are
1141 executing comes back.
1142 * Ensures that SIMIX_clean is called automatically. It's not part of
1143 the public interface anymore (bindings should be updated).
1146 * Bug fix for when SD_Simulate is called with a positive value: be careful
1147 when comparing doubles. Sometimes they are different for non significant
1149 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1150 parallel task whose initial work is distributed among host according
1151 to the Amdahl's law. Such tasks are created with a parameter alpha
1152 that corresponds to the non-parallelizable part of the computation.
1153 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1154 between two sets of workstations assuming a 1D block distribution (each
1155 workstation owns a similar share of data) on both sides.
1157 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1158 Data redistribution will be automatically scheduled once parent and child
1159 are both scheduled. The filling of computation_amount and
1160 communication_amount structures is now done seamlessly thanks to the chosen
1162 * New function SD_workstation_dump to display various information
1163 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1164 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1165 bandwidth on the route onto which the task is scheduled and the amount of
1167 To divide the nominal bandwidth by 2, the rate then has to be :
1168 rate = bandwidth/(2*amount)
1169 * Compute tasks that have failed can now be rescheduled and executed again
1170 (from their beginning)
1171 * Increasing source code coverage (src/simdag is now covered at 95.8%
1175 * Re-implement time-independent trace replay using SMPI (at the
1176 smpi_smp_* level) instead of MSG. This should replace
1177 examples/msg/actions/actions.c
1178 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1179 hindexed and structs)
1180 * Implement the exchange of non-contiguous data.
1181 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1182 * Correct behavior of smpi/sender_gap and set its default value to 0
1183 * Add option to asynchronously send small messages to allow better
1184 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1185 specifies the size in bytes under which messages will be asynchronously sent.
1186 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1187 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1190 * New C interface to define a platform: XML is now optional.
1191 For more info, please check include/simgrid/platf.h.
1192 * New interface to define random platforms from the C:
1193 For more info, please check include/simgrid/platf_generator.h and
1194 examples/msg/masterslave/masterslave_platfgen.c
1195 * Export a sg_cmdline dynar containing all the arguments we got from
1199 * Two new tracing options for adding comments to trace file so you
1200 can track your experiments (see --help-tracing for details).
1201 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1202 * Adding the SimGrid version that generated the trace file as a comment.
1203 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1204 * Fix to avoid key clashes on Paje links
1205 * Other minor fixes related to the Paje specification
1208 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1209 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1210 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1211 * Greatly improve the robustness of mmalloc to user errors (such as
1212 using an area after freeing it, or freeing it twice)
1214 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1216 SimGrid-java (3.8.1) stable; urgency=low
1218 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1219 Warning: all methods are visible, but only some of them are
1220 implemented so far. Check the source (src/jtrace.c)
1221 for further information.
1222 * New module: org.simgrid.msg.File (SimGrid File management functions)
1223 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1224 * Change the meaning of Process.restart: now restart the process from
1225 the begining, like MSG_process_restart in C.
1226 * Add Process.setAutoRestart: handling of process restart when failed
1228 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1229 you to retrieve the properties of the processes/hosts
1230 * Deprecate Msg.clean(): you can just forget about it now.
1231 * New function Process.getCount(), that only works when compiling
1232 with the not yet released version 3.9 of the C library.
1234 * New context factory based on Coroutines. It mandates a modified JVM
1235 but then, the simulations run about five times faster, and there is
1236 no limit to the amount of processes (beside of the available memory).
1238 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1240 SimGrid (3.7.1) stable; urgency=low
1243 * Restore the prototype of MSG_process_create_with_environment() to
1244 the pre-3.7 situation by removing the kill_time argument.
1245 * Add a MSG_process_set_kill_time() function instead.
1248 * Fix weird behaviors when dealing with parallel tasks.
1251 * Simgrid is now built as a dll.
1252 * Simgrid-java now works on Windows.
1253 * Simgrid-Java is now included into Windows package.
1256 * First pre-build package for MacOSX.
1259 * Fix compilation when using MSG_USE_DEPRECATED.
1260 * Fix some compilation issues on Macs and Windows.
1261 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1262 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1264 We discovered that the Lua console is broken, but we are missing the
1265 manpower to fix it right now. The problem existed in 3.7 too, so we
1266 are not blocking the release for that. Sorry if you depended on this
1267 feature, any help would be really welcome.
1269 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1272 SimGrid-java (3.7.1) stable; urgency=low
1274 The "Java aint got to be bloated and slow" release
1277 * Various internal cleanups and performance improvement
1278 Simulations are expected to run up to twice faster or so
1279 * Make Process.kill(process) an instance method, not a static one
1280 * User processes are not java.lang.Thread subclasses.
1281 This breaks the compatibility (sorry), but previous API was
1282 brain-dead, making it impossible to have non-trivial
1283 initializations in the process constructor.
1284 * Require a full constructor per Process sub-class.
1285 Kinda breaks the compatibility (sorry), but this allows a much more
1286 efficient way to launch the processes at simulation startup.
1287 * Do not embeed our version of semaphores, java 1.5 can be considered
1288 as sufficiently prevalent for us to not dupplicate its features.
1292 * Add examples for almost every part of the API
1293 We spotted and fixed a lot of bugs in the process
1295 * New module: asynchronous communication API
1296 * New function: Process.sleep()
1297 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1298 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1299 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1301 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1303 SimGrid (3.7) stable; urgency=low
1305 The "spring cleanups (before the next Big Project kicks in)" release.
1308 * We can specify the SMPI latency/bandwidth factor with command line
1309 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1310 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1311 You can also use the "config tag" from platform file by adding this line
1312 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1313 (see "example/platforms/tag_config.xml" to use "config tag").
1314 Note that the command line supersedes the platform file configuration.
1315 * Change the correction factors used in LMM model, according to
1316 the latest experiments described in INRIA RR-7821.
1317 Accuracy should be improved this way.
1318 * Use the partial invalidation optimization by default for the
1319 network too. Should produce the exact same results, only faster.
1320 * Major cleanup in surf to merge models and split some optimization
1321 mechanisms from the core of the models. As a result you can now
1322 specify which model to use (e.g., --cfg=network/model:LV08
1323 --cfg=cpu/model:Cas01) and which optimization mode to use
1324 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1325 Incompatible combinations should err at initialization. See
1326 --help-models for the list of all models and optimization modes.
1327 * The CLM03 workstation model was dropped for simplicity because it
1328 used the deprecated CM02 network model. Use default instead.
1329 * Rename the TCP_gamma configuration option to network/TCP_gamma
1330 * Rename the coordinates configuration option to
1331 network/coordinates, and document it
1332 * Use now crosstraffic keyword instead of the terribly misleading
1333 fullduplex keyword. It is activated by default now in the current
1334 default model, use --cfg=network/crosstraffic:0 to turn it off.
1335 * Ongoing refactoring the model parsing to make XML files optional
1336 See include/simgrid/platf.h for details (still to be completed)
1339 * Major overhaul of the documentation. Almost instructive now :/
1340 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1341 functions and friends. This interface was considered as
1342 deprecated since over 2 years, it's time to inform our users that it is.
1343 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1344 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1345 use these (crappy) functions in your code.
1346 These functions will be removed soon. Stop using them now.
1347 * Deprecate MSG_get_host_{table,number}
1348 Implement MSG_hosts_as_dynar() instead.
1349 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1350 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1353 * Stabilize the parallel execution mode of user contexts
1354 * Introduce configuration variables to control parallel execution:
1355 - contexts/synchro: Synchronization mode to use when running
1356 contexts in parallel (either futex, posix or busy_wait)
1357 - contexts/parallel_threshold: Minimal number of user contexts
1358 that must be part of a scheduling round to switch to parallel
1359 execution mode (raw contexts only)
1360 * Fix bugs that prevented to use suspend/resume along with
1361 synchronization structures.
1362 * Fix bugs in process termination that lead to invalid memory access
1363 in very specific conditions.
1366 * Introduce a parallel mode for the models (controlled by surf/nthreads
1367 configuration item). In our tests, running the models in parallel
1368 never lead to any speedups because they are so fast that the gain
1369 of computing each model in parallel does not amortizes the
1370 synchronization costs, even when ultra fast futexes are used.
1371 This is released anyway because YMMV.
1374 * Performance boost by using a swag internally to compute the set of
1375 tasks that are finished and should constitute the return value of
1379 * Enable it by default now that it is considered rather stable.
1382 * Documentation of the tracing functions.
1383 * Performance gains when tracing categorized/uncategorized resource
1384 utilization by avoiding calls to get route when updating resource
1385 variables. LMM constraints are being used instead.
1386 * API changed to set task categories. Use MSG_task_set_category instead
1387 of TRACE_msg_set_task_category, and SD_task_set_category instead
1388 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1389 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1390 * Examples for link user variables added (see at examples/msg/tracing/)
1391 * Deprecated function TRACE_msg_set_process_category completely removed
1392 * Trace header updated according to the latest Paje file format
1393 * Tracing network lazy updates, no longer obligate users to use full updates
1394 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1395 * Experimental: let user code declare/set/push/pop application states for hosts
1396 * API changed to allow the manual creation of graph configuration files
1397 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1400 * Improve the API of Lua MSG bindings, using the Lua spirit.
1401 * Each simulated process now lives in its own Lua world (globals are
1402 automatically duplicated). It helps writing simulators. Will allow
1403 to run Splay programs within SimGrid in the future.
1404 * Add a Chord example in Lua, equivalent to the MSG one.
1407 * Start the implementation of a solution to express temporal
1408 properties, not only local assertions. This is still an
1409 experimental work in progress, stay clear from it to be safe.
1413 - Add new runtime parameters --help-logs and --help-log-categories
1414 to display information about supported logging parameters and
1416 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1417 don't exists anymore.
1418 * Mallocators: allow value NULL for the reset function.
1420 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1421 create homogeneous dictionaries, where all the elements share the
1422 same free function. Non homogeneous dictionaries will be
1423 deprecated in the next release.
1424 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1425 - Multi-level dictionaries are deprecated.
1427 - new function xbt_dynar_search_or_negative() that is useful when
1428 you have less than 2 million elements in your dynar and don't
1429 want of the extra complexity of catching exceptions when the
1430 element is not found.
1432 - Make xbt_os_thread module (for thread portability) public.
1433 Documentation is still to come, sorry.
1435 - Cleanups and simplifications to make it maintainable again.
1436 - Exotic features (such as memalign and valloc) were removed.
1437 - The metadata were extended and improved so that the
1438 model-checker becomes able to explore and inspect the heaps.
1439 - This may induce a performance drop when enable_model-checking is
1440 ON in cmake (even if it's not used in the simulation), but it is
1441 necessary at this point to get MC working.
1443 Turn model-checking OFF if simulation performance matters to you.
1444 Not enabling it at runtime is not enough, disable it in cmake.
1446 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1449 SimGrid (3.6.2) stable; urgency=low
1451 The "Not coding new stuff allows to polish old things" release.
1454 * New bindings to the NS3 packet level simulator (experimental)
1455 * Use the raw (efficient) execution contexts instead of the sysv
1456 (portable) ones when possible.
1457 * libpcre is now mandatory in any cases since not using it led to
1458 severe performance loss and possibly other issues
1459 * Update the XML platforms:
1460 - G5K: include the latest machine in Nancy
1461 - GridPP and LCG: new platforms
1462 * Documentation was partially updated, at least (more to come)
1464 Bug fixes, cosmetics and small improvements
1465 * Free terminated processes before the end of the simulation to avoid
1466 exhausting the memory of users having very dynamic amount of
1468 * Bug fix and cosmetics about canceling non-running tasks
1469 * Bug fix about the dot loader's issues when using libcgraph
1472 * Create an installer for windows with nsis (amd64 and win32)
1473 - Add an hello world project to illustrate simgrid project creation.
1474 - Embed libpcre into the Simgrid installer to avoid
1475 its compilation burden
1476 * The raw execution contexts should work on Apple now
1477 * Port to Windows 64 bits
1478 - Sysv contexts now have an implementation for this arch
1479 - GRAS communication features now support this arch
1480 * Drop support for borland compiler on windows
1481 - this code was not maintained, and we kinda depend on gcc nowadays
1482 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1483 * Fix portability issue on unstable ubuntu: linker became picky on
1486 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1488 SimGrid (3.6.1) stable; urgency=low
1490 The "Oops, we broke Macs too" release
1493 * Fixed contexts detection so that raw ones are used when possible
1494 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1495 leads to a strange error, with user code segfaulting sometimes when
1496 the generated code is not perfectly aligned (which is not
1497 controllable from the user side, depends on the amount of code)
1500 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1502 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1504 SimGrid-java (3.6) unstable; urgency=low
1507 * Split of every thing from simgrid v3.5 into a separate package.
1509 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1511 SimGrid (3.6) stable; urgency=medium
1513 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1516 * Bindings now constitute their own package, separated from the main one.
1517 Rationale: reduce our maintenance nightmare by reducing the module coupling
1518 They will soon be released on their own on gforge.
1520 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1521 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1523 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1524 * It's not quite deprecated for now because we have no replacement,
1525 but it may soon become the case.
1528 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1529 * Fortran: New user-level cache variable to store the rank of the running
1530 process. This improves performance by an order of magnitude.
1531 * C: New coccinelle script to automatically locate and modify global and
1532 local static variables.
1533 * Improved SMPI network model with a sender-side gap to account for multiple
1537 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1538 only say if a communication is finished, no matter whether it succeeded or
1539 failed. You can call MSG_comm_get_status() to know the status of a finished
1541 * New function MSG_task_dsend() to send a task and detach it. When a
1542 communication is detached, you are never notified of its success or failure
1543 and the memory is released automatically once it is finished. This function
1544 is useful when you don't care about the end nor the success of a
1546 * Change the prototypes of action replay. Sorry for inconvenience,
1547 but this is really more efficient this way (and to adapt your code,
1548 you just have to fix the initialization, that shouldn't be too long)
1549 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1550 ever managed to do anything useful with it.
1551 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1552 takes a dynar of communications. It returns immediately and gives the
1553 index of a finished communication (if any).
1554 * New example: a basic implementation of the Chord P2P algorithm.
1557 * New model for multi-core CPUs. You can now use the core attribute to
1558 precise the number of cores of a host. This is a basic model. Every
1559 process running on the host receives at most the power provided in
1560 the DTD (throughput<=power). Total throughput of process cannot exceed
1562 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1563 router linked by an up-link and a down-link (possibly asymmetrical).
1564 This kind of pattern allows to easily build last-mile model style platforms.
1565 Aggregating such patterns in a rule-based AS is thus the technique of
1566 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1567 * New model for Vivaldi routing. We transformed the Vivaldi network model
1568 into a Vivaldi routing model (based on the rule-based model). This allows to
1569 combine Vivaldi based latencies with last-mile platforms.
1572 * Added a check for NaN of IEEE754 infinite in the double entries of
1574 * Introduce a new context factory "raw", highly inspirated from the
1575 ucontext factory, but using manually crafted functions in assembly to
1576 do the work in an efficient manner.
1577 * Allow to change the used context factory at run time, not only at
1578 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1579 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1580 contexts at runtime (only with raw contexts or ucontexts).
1581 * Completely rewrote this module to allow parallel execution of user
1582 processes. Use --cfg=contexts/nthreads:N to execute user processes
1583 with N parallel threads (the default is 1, meaning no parallelism).
1584 * Allow to decide dynamically between sequential and parallel modes.
1585 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1586 processes in parallel only when their number is greater than or equal to P
1588 * Added a check for NaN of IEEE754 infinite in the double entries of
1592 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1593 won't output the state of processes when interrupted with Ctrl-C
1594 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1595 NULL-terminated array. This may solve backward compatibility issues
1596 due to the change to return type of SD_simulate. See also:
1597 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1598 * Add new macros with variable number of arguments.
1599 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1600 - in xbt/asserts.h: xbt_assert
1601 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1602 - in xbt/ex.h: THROWF and RETHROWF.
1603 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1605 * Change xbt_die() to accept a format string with arguments, just like printf.
1606 * New data structure: xbt_lib_t, like a dict but more general and with better
1610 * New configuration options
1611 Options triva/categorized and triva/uncategorized can be used to generate
1612 graph configuration files for Triva visualization tool.
1613 * Configuration option tracing/platform is renamed to tracing/categorized
1614 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1615 * New TRACE_user_link_variable interface:
1616 User provides the name of the link and the tracing variable to attach to it
1617 * the declaration of tracing categories must be done after the environment
1619 * simpler tracing interface, just one way to declare categories
1620 TRACE_category or TRACE_category_with_color, it is up to you
1621 * links in the trace file are again identified by their names
1622 * trace contains the full platform hierarchy exactly as declared using the ASes
1623 * Options tracing/msg/[task|process]:1 groups the process by hosts
1624 for both cases, tasks and processes must have names that are unique during the simulation
1625 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1626 * The experimental option tracing/msg/volume is deprecated
1627 its functionality may be reincorporated if needed
1629 The tracing generates a trace file with unordered timestamped events,
1630 because of the way the core simulator (surf) works. A script available
1631 at the tools directory (fix-paje-trace.sh) can be used to put the events
1632 in order. We have changed the tracing so it can generate ordered timestamped
1633 events in the final trace, but depending on the simulator (and how much time
1634 is simulated) that can lead to a huge memory utilization. It is deactivated
1635 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1637 Build Infrastructure
1638 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1639 - We are trying hard to keep the API stable, but it may happen that
1640 some things change (we're a research project after all, not a
1641 nuclear plant operating system). If such things should happen, you
1642 could rely on that macro to adapt.
1643 - current value: 30600 for 3.06.00, aka 3.6
1644 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1645 a number that can be compared with SIMGRID_VERSION.
1646 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1647 assertions and verbose/debug logging events are disabled at compile time.
1649 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1651 SimGrid (3.5) stable; urgency=medium
1654 * New feature to any SimGrid-based simulator: Model-Checking
1655 Check SIN#1 for more details.
1658 * New Model SMPI (three-interval linear regression for correction factors)
1659 See RR-7426, available at http://hal.inria.fr/inria-00527150
1660 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1661 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1662 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1663 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1664 * See also src/smpi/README
1669 - Tracing API changes: TRACE_start and TRACE_end should not be called
1670 by user-code. They are automatically called by simulators created
1671 with SimDAG, MSG and SMPI if the toolkit is compiled with
1672 tracing_enabled=ON. Categories declaration and utilization remain the
1673 same for MSG and SimDag.
1674 - A function was added to the tracing API to declare categories with
1676 - TRACE_category_with_color (char *category, char *color)
1677 where color must be in the following format
1678 "%f %f %f", red, green, blue
1679 and red, green, blue are float values in the interval [0, 1]
1680 - User can specify NULL as color parameter, or continue calling
1681 TRACE_category (cat)
1682 On that case, the tracing system will define random colors
1683 - The following command-line options are supported:
1684 --cfg=tracing/filename:msg.trace
1685 --cfg=tracing:1 (activate tracing, needed to use others)
1686 --cfg=tracing/platform:1 (categorized resource use)
1687 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1688 --cfg=tracing/msg/task:1 (task creation)
1689 --cfg=tracing/msg/process:1 (process creation, migration)
1690 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1691 --cfg=tracing/smpi:1 (SMPI interface tracing)
1692 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1693 - examples of examples/msg/tracing updated
1695 - DAXLoader and DOTLoader functions can generate tasks with categories
1696 - A new function to attribute a category to SD tasks:
1697 TRACE_sd_set_task_category (SD_task_t task, char *category)
1698 * Tracing the MPI interface implemented by SMPI
1699 - Collective operations are traced with states
1700 - Point-to-Point operations are traced with states/links
1701 - Tracing activated by a parameter "-trace filename" passed
1702 to smpirun during execution (considering that simgrid
1703 is compiled with tracing enabled)
1704 - To run the simulation with gdb, the simulator
1705 accepts --cfg=tracing/smpi:1 to trace SMPI
1706 - tesh files to check if smpi tracing is ok
1707 - See examples/smpi/NAS/DT-trace
1708 * GTNetS tracing re-worked
1709 - adaptation to the tracing system of GTNets to cope
1710 with modifications regarding the fullduplex mode
1711 - new tesh files to check if gtnets tracing is ok
1714 * Asynchronous communications through the functions:
1715 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1716 * New function: MSG_load_platform_script()
1717 to make possible using a lua script instead of XML files to set up platforms
1718 * New function: MSG_set_function
1719 to associate functions to processes, used when bypassing the parser
1720 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1722 Platforms: Add some more examples in examples/platforms
1723 * Grid'5000: see www.grid5000.fr
1724 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1727 * Change the XML format. This is a very important modification. SimGrid
1728 3.5 introduces a new hierarchical format based on the notion of
1729 Autonomous Systems. Compatibility with old format is ensured
1730 through the perl script provided in the install bin directory
1731 bin/simgrid_update_xml.
1732 It is now possible to build platforms with specific routing
1733 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1734 connect several platforms together. We will try to provide soon
1735 set of realistic platforms exploiting these properties (have a
1736 look at examples/platforms/ for the moment).
1737 * Take the opportunity of the XML format change to be a good XML citizen:
1738 rename link:ctn to link_ctn and similar changes (also dealt with by
1740 * Add a new routing scheme (rule-based) using regular expressions. It
1741 enables to have an extremely low memory footprint when the
1742 underlying routing is simple and can be compactly described. You
1743 need to have libpcre4-dev (perl regular expressions) installed if
1744 you want to use this routing scheme.
1745 * Revive the cluster TAG and allow to easily and efficiently (both in
1746 term of memory and speed) connect clusters together. Have a look
1747 at teshsuite/simdag/platforms/ to see how this can be done. With
1748 this tag, you can create clusters with thousands of tasks at no
1749 cost (have a look at examples/platforms/).
1750 Note: clusters are implemented as ASes, so there is no need for an
1751 enclosing AS tag if you have only one cluster in your platform.
1752 * Add new generic functions in the public interface that allows the user
1753 to call SURF 'create_resource' methods from your code (same
1754 functionality as the XML bypass mechanism but with a much lighter
1756 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1757 that uses a piecewise linear approximation to produce better
1758 results when exchanging small messages.
1759 * Add a new parameter to handle correctly full duplex link and account
1760 for interferences between uplink and downlink communications
1761 (activate with --cfg=fullduplex:1).
1764 * Rename the SD_READY (all dependencies are satisfied and task is
1765 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1766 dependencies are satisfied) state.
1767 This prevents a confusion between the notion of "ready to schedule"
1768 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1769 simulated" (SD_RUNNABLE) used by the simulation kernel.
1770 * Change the way a task is considered as ready. Instead of removing
1771 dependencies when a task is done, a counter is decreased. This way,
1772 it is always possible to reach ancestors thanks to the
1773 SD_taks_get_parents function (even after the end of the simulation.)
1774 * Change the return type of SD_Simulate from (SD_task_t*) into
1775 xbt_dynar_t. This function was in handling a dynar internally and
1776 converted it into a NULL terminated array for historical reasons.
1777 * New function SD_dotload(char*) to load a DAG described in dot
1778 format. This loader and the corresponding examples require the
1779 installation of the graphviz library.
1780 * Fix a bug in the management of tasks of size 0 in the surf network
1781 models. This problem was only visible with SIMDAG and you should
1782 thus disregard results produced with earlier versions if you
1783 relied on this feature (some tasks were blocked because of this).
1784 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1785 even though your applications were doing only point-to-point
1786 communications and sequential computations. Now you can really use any
1787 model you want (of course, if you create real parallel tasks, which are
1788 not implemented in most models beside ptaskL07, this will abort).
1789 * Add an example that schedules a DAX on an heterogeneous platform
1790 using a Min-Min strategy.
1791 * New function SD_workstation_get_current_task() that returns the kind
1792 of task currently running on a workstation in the sequential access
1794 * Raise some warnings when unexecuted tasks remains at the end of the
1795 simulation. This is usually caused by cycles in the DAG.
1798 * New function: SIMIX_process_set_function() called by MSG_set_function
1799 * Change the underlying waiting queue in semaphores so that a process
1800 can wait on several of them simultaneously (as in waitany).
1801 * Fix the way to handle tokens in semaphores so that all access patterns
1802 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1803 * kill the dirty pimple SIMIX_message_sizes_output()
1804 Please use (proper) visualization instead
1807 * New data container: setset (set of sets of elements)
1808 * New module: mmalloc (mapped malloc, allowing to have several
1809 independent segments of malloc)
1810 * New function: xbt_dict_cursor_set_data()
1811 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1812 * New function: xbt_dynar_is_empty()
1813 * New function: xbt_fifo_get_last_item()
1814 * Fix xbt_dynar_shrink(): use the right element size.
1815 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1816 uninitialized areas during expand.
1817 * Fix semaphores: previous implementation was severely broken.
1818 * Use library init/fini functions for our initialization.
1819 - you can use logs and other feature as soon as you want in your
1820 code (even before the xbt_init / MSG_init)
1821 - xbt_exit is now a no-op and produce a warning when used.
1824 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1825 benefit from the latest improvement to the simulation kernel.
1826 * Kill measurement sockets for now. If you rely on them, sorry. This
1827 release is not for you. This feature will be reintroduced in the
1828 future, but we cannot delay the release any further.
1829 * New function: gras_msgtype_get_name().
1830 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1831 * Fix (at last) the pmm example: it should not randomly fail anymore.
1833 Build chain: bug fixes and overall polishing
1834 * Cmake is now stable enough. Hence, we killed the autotools.
1835 * Port to windows ( TM :)
1836 * Fix the 'make install' target.
1837 No need to use 'make install-simgrid' anymore
1838 * Introduce a 'make dist' target compiling a *source* archive
1839 'make package' compiles a binary archive
1840 * Compile java files only on need
1841 * Add --cd and --setenv command line options to tesh
1842 * Out of source builds are not fully supported yet, but we are close
1843 * Enable supernovae and optimization flags by default for our users
1846 * Add layer to set up environment directly from lua, without XML.
1847 * The effect of gras_stub_generator can be achieved through
1848 lua too (check examples/gras/console/ping_generator.lua)
1850 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1852 SimGrid (3.4.1) stable; urgency=low
1854 The "Polishing easter eggs is probably a good idea" release.
1855 This is a bug fixes release only.
1859 * Fix a bug preventing the tasks from begin garbage collected.
1862 * Fix a bug occuring when a host involved in a communication fails.
1863 This was not detected properly by the other peer involved in the
1864 communication. Now, it's reported as a network error.
1867 * Warn the user about loop dependencies in data flow of DAX files
1868 * Obey the control-flow dependencies of DAX files
1871 * Add option "enable_smpi" allowing to not compile SMPI.
1872 Probably useful for the (Mac) users experiencing a build error here
1873 * Improve the detection of lua5.1 and ruby1.8
1875 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1877 SimGrid (3.4) stable; urgency=low
1879 The "Easter in Cargese" release. Also known as (major changes):
1881 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1882 ~> bindings were greatly improved
1883 ~> new tracing infrastructure for better visualization introduced
1885 * the "Welcome to configury modernity" release.
1886 ~> we switched from autotools to cmake, and improved our cdash
1889 A more detailled list of changes follow (full detail in svn log).
1891 Java Bindings: Various Cleanups
1892 * (install java-gcj-compat-dev on debian-like to use them)
1893 * Remove put/get: no need to export deprecated interface in Java
1894 Use send/receive instead.
1895 * Cleanup the examples and add a README per directory
1896 * Remove example autoDestination (that's the only way to go now)
1897 * Remove example explicitDestination (was a plain copy of basic)
1898 * Make JniException a runtime exception, so that there is no need to
1899 declare the fact that you may encounter such a beast. I guess that
1900 nobody will ever want to survive such error.
1901 * Create specific errors for each MSG case of failure:
1902 host failure, transfer failure, timeout, task cancelled
1903 * Cleanup the exceptions that may get thrown by each function
1904 * Other internal cleanups in Java bindings. Performance still bad :/
1905 Ruby and Lua Bindings: create them
1906 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1907 * That's new and great, you should try them out.
1908 Same functionalities than Java bindings, only even less polished
1910 * Kill the useless "rate" argument of SD_task_get_execution_time()
1911 Everyone used to provide -1 as a value, it was not used, and the
1912 semantic of a possible use wasn't even clear.
1913 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1914 either as comm costs or compute costs to mean that there is no
1915 such thing for that specific task.
1916 * Add a SD_task_set_name() function
1917 * Fix SD_task_unschedule() on typed tasks
1918 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1919 * In DAX loader, accept useless 'level' attributes to <job> since
1920 LIGO DAGs have them (seem to be to ease graphical representation).
1922 * Add an example masterslave_mailbox.c using send/receive and not
1923 the deprecated put/get interface.
1924 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1925 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1927 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1928 (declare MSG_USE_DEPRECATED to still have the old name)
1929 * Add a MSG_task_set_data() function
1930 * About trace replay (see examples/msg/actions):
1932 - Allow to work with splitted trace files for each process
1933 Give the specific trace file as argument of each process,
1934 and call MSG_action_trace_run(NULL)
1935 You can still have one merged file for all processes.
1936 - Fix implementation of collective operations
1937 * Allow task_execute() on 0-sized tasks (closes #10063)
1939 * This is the first release of SimGrid where SMPI is not considered
1940 beta anymore (even if some corners should still be improved)
1941 * Port over the new SIMIX_network submodule (internal refactoring)
1942 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1943 * Implement more missing elements of the standard:
1945 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1947 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1948 MPI_Scan() MPI_Get_processor_name()
1949 - Added implementation of missing case for Alltoall (warning: it's
1950 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1951 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1952 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1953 SMPI_MPI_Allgatherv()
1954 * Bug fixes include:
1955 - MPI_Waitsome() was broken
1956 - Allow relative includes in smpicc
1957 - Command line cfg argument 'reference_speed' was ignored...
1958 - Some functions did not properly lead to auto-benching of user code
1959 - smpicc passes -O2 by default (just like openmpi one)
1961 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1962 * Bug fixes about timeouts during communications
1963 * add SIMIX_message_sizes_output() as a pimple to write to file the
1964 amount of messages per size. Use gnuplot to get histogram.
1965 Pimple because that's the only user-visible function of simix,
1966 defined directly in xbt.h (irk, sorry)
1968 - Add a SIMIX_sem_get_capacity() function
1969 - Fix interactions with processe resume/suspende
1970 - release_forever() was stupidly broken
1971 - Fix SIMIX_display_process_status() for processes in a semaphore
1972 - Make SIMIX_sem_block_onto() user-visible
1973 * Refactoring context stuff:
1974 - Use pseudo-OOP for better modularity
1975 - reimplement SIMIX_process_kill() without process_schedule() so
1976 that the latter can take as invariant that it is called from
1978 - Merge context_start into context_new for sake of simplicity
1980 * Add a Vivaldi network model, coded live during SUD'10 ;)
1981 * Rename configuration variables to start a hierarchy:
1982 o cpu_model -> cpu/model
1983 o network_model -> network/model
1984 o workstation_model -> workstation/model
1985 * New configuration variables:
1986 o network/bandwidth_factor: correction to bandwith
1987 o network/latency_factor: correction to latency
1988 o netwotk/weight_S: correction to the weight of competing streams
1989 * Add a long description to the models, that users can see with such
1990 argument on the command line: --cfg=cpu/model:help
1991 * --help-models display the long description of all known models
1993 * config: add the ability to set a default value after registration
1994 Does not override any previously set value (e.g. from cmd line)
1995 * dict: allow to have integer key and data.
1996 When so, you need to use the following functions
1997 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1998 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1999 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
2000 In contrary to regular dicts, the key is not malloced before copy.
2001 Mixing scalar and regular elements in the same dict is not tested
2003 * Allow to use xbt_dynar_shrink() to expend the dynar instead
2004 Tracing for Visualization:
2005 * SimGrid is now instrumented in order to generate a trace file for
2006 visualization analysis: to use it, need to compile SimGrid with the
2007 "tracing" option enabled, and instrument the program using SimGrid with
2008 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
2009 (among other functions).
2010 * The instrumentation only traces the platform utilization for now
2011 * Documentation to use the tracing functions and how to analyze the
2012 traces with the Triva tool is written.
2013 * More information about: SimGrid FAQ (in the section Tracing Simulations
2016 * We moved to cmake as default build system. Autotools support will
2017 be dropped soon. Check the FAQ for more info about how to use it.
2018 * Greatly improved our cdash/ctest interactions
2019 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
2020 * Added memory checking tests with valgrind; lot of memleak fixing.
2021 This may be the first release of simgrid with so few memory issues
2022 * Added code coverage tests.
2023 Our coverage is still improvable, but at least we see it on cdash.
2025 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
2027 SimGrid (3.3.4) stable; urgency=low
2029 The "Desktop Grid needs love too" release (also called Xmas release).
2031 Models improvements:
2032 * Major speedup in the maxmin system solving by using lazy evaluation
2033 Instead of solving completely the maxmin system at each iteration,
2034 only invalidate (and recompute) the modified parts.
2035 This new feature is enabled in default models but you can try to
2036 turn it on with "--cfg:maxmin-selective-update=1" for other models.
2037 * Cas01 IMproved as default CPU model
2038 This CPU model is the same Cas01 model, but it uses the
2039 maxmin-selective-update flag and a heap structure to manage
2040 actions on SURF kernel.
2041 It reduces the complexity to find the next action to finish and,
2042 consequently, it's faster than the old Cas01.
2043 This is the new default CPU model (Cas01).
2044 * Rename the old Cas01 model to Cas01_fullupdate
2045 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
2046 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
2047 * CpuTI (CPU Trace Integration)
2048 A new CPU model whose objective is simulate faster when using
2049 availability trace files.
2050 Instead of using a full featured, over engineered maxmin system for
2051 CPU modeling, this model does the pre-integration of traces files
2052 to calculate the amount of CPU power available, and so, executes
2053 faster than the old CPU models.
2054 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
2055 * Use LV08 as default network model since it gives better accuracy
2056 for small messages and shouldn't change things for big ones.
2057 Use --cfg=network_model:CM02 to get the previous behavior.
2060 ******************************************
2061 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
2062 ******************************************
2063 * The new CPU model may changes simulations!
2064 The point is that events occurring at the exact same timestamp
2065 are not scheduled in the same order with the old and new
2066 version. This may be enough to completely change the execution
2067 of simulations in some cases.
2068 * The new network model will change simulations!
2069 This new model is more realistic than the previous one, so you
2070 should consider redoing your old experiments with this model.
2071 Sorry for the inconvenience.
2074 * Introduce the supernovae compilation mode
2075 When compiled that way, the whole SimGrid (or almost) is put in a
2076 single compilation unit and compiled in one shoot.
2077 This is to help gcc which has difficulties to inline stuff from one
2079 The speedup seem to be above 15%, althrough more tests are needed on
2080 amd64 to confirm that gain.
2083 * Port of MSG's mailbox on top of SIMIX network
2084 The put/get mechanism was greatly simplified on the way.
2087 * New SIMIX network module. Provides:
2088 - Mailbox: rendez-vous mechanism to find with who you want to speak
2089 - Synchronous send/recv: easier and hopefully faster since the
2090 logic is handled in the maestro process directly now
2091 - Asynchronous send/recv: you dreamt of it? It's here now
2092 Too bad that nobody cared enough to propagate the change to MSG.
2093 * Add semaphores as SIMIX synchronization mechanism.
2096 * new function SD_daxload(char*) to load a DAX file
2097 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2098 * Introduce typed tasks. Specify its kind and cost at creation.
2099 At scheduling, just give where it should be placed, and the cost
2100 for each involved resource is automatically computed.
2101 Existing constructors so far (more to come of course):
2102 - SD_task_create_comm_e2e() for end-to-end communication
2103 - SD_task_create_comp_seq() for sequential computation
2104 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2105 * new function SD_task_dump() for debuging display
2106 * new function SD_task_dotty(task,FILE*) writing to file the info
2107 about the task in dotty format
2108 * SD_task_dependency_exists() can now cope with having one of its
2109 arguments NULL. If so, it tests whether the other argument has any
2111 * Add getters on list of preceding/following tasks:
2112 SD_task_get_parents(task) and SD_task_get_children(task)
2113 * Add getters on amount of workstations and list:
2114 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2115 * Add getter on task kind: SD_task_get_kind(task)
2116 * Update the start_time and finish_time of tasks on completion/failure
2117 * Bugfix: Remove task from state swags when destroyed
2120 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2123 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2124 dynar with one separate thread per value of the dynar.
2125 * Change the prototype of xbt_thread_create(), sorry.
2126 Added a boolean parameter indicating whether we want to join this
2127 thread (used in SG only for now)
2128 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2131 * GTNetS wrappers should now be usable again (and betterly tested too)
2132 * Fix a major regression from 3.2 where the timeout provided to
2133 MSG_task_put_with_timeout() was used as absolute time before which
2134 the comm should be done.
2135 * Start to fix the <cluster> tag.
2136 - Internal links should be good now (beside of the loopback, which
2137 use the private link instead)
2138 - paths to the external world is still rather broken
2139 - the <route:multi> tag is just broken. Actually that's brain-dead.
2140 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2141 to make it less stupid
2142 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2143 * Fix a source-level compatibility glitch from 3.2: after defining
2144 MSG_USE_DEPRECATED, you can use the old name
2145 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2146 * Allow to compile from the SVN with automake 1.11
2147 * Fix some problems when using the "start_time" tag in deployment XMLs.
2148 * Fix #8569: XBT/synchro.h has redundant declarations
2149 * Fix #8563: MSG return values and exceptions
2150 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2151 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2152 Thanks to Robson Peixoto.
2153 * Fix a vicious bug in dictionaries inducing that some elements were
2154 not freed on xbt_dict_free()
2156 Portability report of this version:
2157 * Main portability targets:
2158 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2159 on (amd64/i386/ia64)
2160 - mac leopard on i386
2161 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2162 but nothing critical.
2163 * Other platforms: windows, AIX and others were not tested for this release
2165 Timing report of this version:
2166 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2167 scenario parameters). From 8h to a few seconds in desktop grid settings.
2168 * Supernovae brings about 25% speedup on i386.
2170 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2172 SimGrid (3.3.3) stable; urgency=low
2174 The "Need for Speed" release.
2176 The timings done to validate the 3.3.2 were faulty.
2177 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2179 The problem was a conversion from a manually handled vector to
2180 xbt_dynar_t on the critical path.
2181 xbt_dynar_foreach calls functions, inducing stack management crap.
2183 We inlined these functions and xbt_dynar_foreach is now breath taking.
2184 We also inlined xbt_swag_belong on the way.
2186 Here are some approximate speedup measurements (on master/slaves
2187 simulations lasting between 10s and 20s each):
2188 3.3.1 -> 3.3.2: about same performance
2189 3.3.2 -> 3.3.3: 40% speedup
2190 3.3.1 -> 3.3.3: 40% speedup
2191 3.3.1 with inline patch -> 3.3.3: 30% speedup
2193 Our reading is that the refactoring which occurred in 3.3.2 made us
2194 suffer much more from the xbt_dynar_foreach low performance, but
2195 once we solved this, this refactoring proved to be very performance
2196 effective. From the 40% speedup, somehow, 10% are due to the
2197 inlining and 30% to the refactoring.
2199 That's a pitty that gcc cannot inline functions placed in other files
2200 alone. We have to choose between:
2201 - break the encapsulation (by putting private data structures and
2202 accessors in headers files to help gcc)
2203 - live with low performance
2204 - switch to a decent compiler such as icc (not quite possible).
2206 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2208 SimGrid (3.3.2) stable; urgency=low
2210 The "Simplicity does not preceed complexity, but follows it" release.
2212 The main contributors of this release were (lexical order):
2213 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2216 * Extract the routing logic into its own object.
2217 (was dupplicated in network.c and workstation_LV07.c;
2218 Allows to implement other ways of storing that info)
2219 => kill now useless network_card concept
2220 - Use dynar to represent routes (instead of void** + int*)
2221 - kill link_set (use surf_network_model->resource_set instead)
2222 - Add a command-line option to choose the routing schema to use
2223 - Add three new models:
2224 * Floyd (shortest path computed at initialization)
2225 * Dijikstra (shortest path recomputed all the time)
2226 * Cached Dijikstra (shortest path computed on need)
2227 All these models where contributed by Silas De Munck, and are
2228 described in his ICCS09 paper.
2230 * Simplify model declaration
2231 (less redirections, less function to write when defining a model)
2232 - Factorize stuff between models:
2235 surf_model_resource_set(model)
2236 surf_model_resource_by_name(model, name)
2237 - Unify the types of models in s_surf_model_t (using an union)
2238 - Embeed fields of common_public directly into s_surf_model_t
2239 - Rename model methods:
2240 action_free ~> action_unref
2241 action_change_state ~> action_state_set
2242 action_get_state ~> action_state_get
2243 - Change model methods into functions :
2244 (model)->common_public->action_use ~> surf_action_ref
2246 * Implement a generic resource; use it as ancestor to specific ones
2247 (allows to kill duplicated code in models)
2248 Drawback: timer command don't need no name nor properties;
2249 workstation_CLM03 don't need no properties
2250 (but I guess we can live with those few bytes wasted)
2252 * Improve the action object model
2253 - implement a constructor avoiding dupplicated code about field
2254 initialization in generic_action part.
2256 * Kill the SDP model: it has an external dependency, is deprecated
2257 in flavor of modern lmm models, and didn't compile since a while
2260 * Relocation of the context module from XBT to SIMIX.
2261 (the context were decoupled from the simix processes, duplicating a lot of code)
2262 => a lot of code was factorized
2263 - less overhead is introduced during scheduling
2264 - simpler API for the context factory
2265 - the logic for process creation,destruction and manipulation was simplified
2266 * Simplification of the s_smx_process_t data structure.
2267 => accesing the simix level data associated to a process is faster now,
2268 and the code is a lot more readable.
2271 * Implement some more MPI primitives:
2272 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2273 -implementation: Bcast: flat or 2-ary tree (default),
2274 Barrier: 4-ary tree,
2276 Allreduce: Reduce then Bcast
2277 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2278 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2279 Alltoallv: flat tree, like ompi
2281 * Add support for optimized collectives (Bcast is now binomial by default)
2282 * Port smpirun and smpicc to OS X
2285 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2286 Shout out if you used it.
2289 * Display the list of still queued messages in SG mode when existing
2293 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2294 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2295 * Add xbt_str_from_file(FILE*)
2296 * Add xbt_dict_get_key achieving a linear reverse search
2297 * Remove the context module
2299 Portability report of this version:
2300 * Main portability targets:
2301 - Linux(debian)/x86/context
2302 - Linux(debian)/x86/pthreads
2303 - Linux(debian)/amd64/context
2304 - Linux(debian)/amd64/pthreads
2305 On these, we still have the eratic breakages of gras/pmm and
2306 amok/saturate_sg reported in previous version. We still think
2307 that the tests are the cause of the fault, not the tested code.
2309 - Mac OSX Leopard/x86/context
2310 Still false negative in tesh autotesting.
2311 Smpi still fails, but this time because readlink does not accept -f
2312 Everything seems to work properly beside of that.
2315 - AIX version 5.3 (only tested contexts this time)
2316 Smpi still fails there because mktemp is not installed.
2317 Everything seems to work properly beside of that.
2319 I managed to compile it for the first time, but several breakages.
2320 Won't delay the release for this exotic platform.
2322 * Windows: it's still lagging behind. If you want to help, please
2325 Timing report of this version:
2326 This version seem to be more than 5% faster than 3.3.1 (on linux
2327 64bits with contextes). The gain is less than expected, we are
2328 investigating this for next release.
2330 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2332 SimGrid (3.3.1) stable; urgency=low
2335 * Implement a --cfg-help to show existing configuration variables
2336 * Build chain do not require doxygen in maintainer mode
2339 * fix a bug on struct sizeof computation, which prevented the
2340 exchange of arrays of structs in some conditions
2341 - added a regression test about this in datadesc_usage
2342 * Allow the exchange of 0-long dynamic vectors.
2343 - for that, use -1 as indicator of dynamic size instead of 0
2344 - This implied to change any size from unsigned long to long,
2345 reducing a bit communication abilities, but I guess that with
2346 64bits being quite common, this is more than enough.
2347 - This also induce a protocol change, thus bumping network protocol
2348 version from 0 to 1 (if we have external users, we have to get
2349 clean on that point too ;)
2350 - added two regression tests about this in datadesc_usage
2351 * Be more verbose when propagating local exceptions
2352 This helps debugging.
2353 * Display the status of simulated processes when receiving SIGINT in
2357 * Allow to control the simulation from a trace file.
2358 New functions MSG_action_register() and MSG_action_trace_run()
2359 The first one allows to associate a function execution to each
2360 kind of action while the second one parses a trace file and
2361 triggers the corresponding actions within the system.
2362 For now, only a toy example is provided in examples/msg/actions
2363 * Add an exemple of process migration in examples/msg/migration
2364 * Fix a bug in task exchange which broke MSG_task_get_sender()
2365 Add a teshsuite regression test for that.
2366 [Bug: if MSG_task_get_sender() is called after sender exit,
2368 * Fix a bug which prevented suspend/resume to work properly
2369 * Display the status of simulated processes when receiving SIGINT
2370 This fixes a regression of v3.3. due to the introduction of SIMIX
2371 * Bug fixing in failure management:
2372 - trace could not start by a failure at time 0
2373 - failure during communications were not working
2376 * Add SIMIX_process_set_name() to change the name of the current
2377 process in the log messages.
2378 * Store smx_hosts in a dict since we only retrieve them by name
2379 * Move the configuration infrastructure to surf
2382 * Move the configuration infrastructure to surf
2385 * Massive internal cleanups:
2386 - Store internal structures on processes instead of hosts (allows
2387 to have more than one process per host, in addition of being more
2389 - Cleanup the initialization/finalization process
2390 - Kill a whole bunch of unneeded synchronization:
2391 processes run in exclusive manner within the simulator
2392 - Move queues from global tables to process data fields
2394 - now accept -platform and -hostfile arguments
2395 - Pass the right rank value to processes according to the hostfile
2396 * Compile the examples by default, and use them as regression tests
2397 * Implement MPI_Wtime()
2398 * Change the reference speed to a command line option
2401 * TCP_gamma can now be specified as command line option using
2402 --cfg=TCP_gamma:10000000.0
2403 * Change the --surf-path cmd line option into --cfg=path:
2406 * Also include strbuff from xbt.h public header
2407 * xbt_ex_display(): do not free the exception after displaying
2408 This allows to do more with the given exception afterward.
2409 Users should call xbt_ex_free() themselves.
2413 Portability report of this version:
2414 * Main portability targets:
2415 - Linux(debian)/x86/context
2416 - Linux(debian)/x86/pthreads
2417 - Linux(debian)/amd64/context
2418 - Linux(debian)/amd64/pthreads
2419 These targets fail about 1/10 of times on gras/pmm, but we believe
2420 that this is because of the test, not because of simgrid.
2421 amok/saturate_sg fails even more rarely, and the test may not be
2424 - Mac OSX Leopard/x86/context
2425 The test suite still spits tons of errors because some obscure
2426 force prevents us from removing the temporary directories
2427 arguing that they still contain some metadata I've never heard of.
2428 Smpi fails because seq is not installed.
2429 Everything seems to work properly beside of that.
2432 - AIX version 5.3 (both contexts and pthread)
2433 Smpi still fails there because mktemp is not installed.
2434 XML inclusions seems rosty on AIX.
2436 * Windows: it's still lagging behind. If you want to help, please
2439 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2441 SimGrid (3.3) stable; urgency=high
2445 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2446 [Malek Cherier & Mt]
2448 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2451 Until now, GRAS were using MSG as an interface to SURF. It was
2452 quite difficult because both interface have several differences
2453 (MSG channels vs GRAS sockets were the most notable point).
2455 This also opens the gate to SMPI (which should occur soon) and speed
2456 up simulations by to 40% (even if it were not the main goal).
2458 **************************************
2459 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2460 **************************************
2461 The point is that events occuring at the exact same timestamp are
2462 not scheduled in the same order with the old and new version. This
2463 may be enough to completely change the execution of simulations in
2464 some cases. Sorry for the inconvenience.
2466 * Cleanup and upgrade the XML format to push further scalability
2467 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2469 * Improve the testing infrastructure with tesh. Now a very large part of
2470 the code is tested not only by being run but also by checking that the
2471 output match an expected output [Mt].
2473 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2474 is really less memory-demanding, which should allow you to use
2475 larger files in SimGrid [AL].
2477 * Inform valgrind about our contextes, so that it becomes usable
2478 with the default (and more effecient) version of SimGrid
2479 [contributed by Sékou Diakite, many thanks]
2482 * Introduce a listener thread in charge of receiving incoming
2483 messages from the network. It allows to overlap communication and
2484 computation but most notably, it removes some stupid deadlocks due
2485 to the fact that so far, a process could not send and receive at
2486 the same time. This made most non trivial communication schema
2488 * Convert the PIDs from long int to int to match the MSG ones (and
2489 linux ones too) [Mt]
2490 * New function: gras_agent_spawn() to launch a new process on
2491 current host. Only working in simulation for now. [Mt]
2492 * New function: gras_os_hostport() returning a constant form (ie,
2493 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2496 * Make the backtrace of exceptions more human readable [Mt]
2497 * New module: xbt/str [Mt]
2498 a ton of string utility functions (split, join, printf to a newly
2499 allocated buffer, trim, etc)
2500 * New module: xbt/hash [Mt]
2501 SHA1 hashing algorithm (more to come if needed)
2502 * New module: xbt/synchro [Mt]
2503 synchronization tools (mutex and conditions) working the same way
2504 in simulation and in real life (mainly useful for GRAS, but not
2506 * New module: xbt/queue [Mt]
2507 classical producer/consumer synchronization scheme
2508 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2509 (using the classical functions will get serialized) [Mt]
2510 * Make dictionary internal table dynamic. No need to specify its size
2511 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2513 * Make sure the log channels are organized as a tree under windows
2514 (because of ANSI C compatibility issue, any channel were child of
2515 root directly) [Mt].
2518 * Cleaned many thing in surf and fixed a few bugs [AL].
2519 * Add a nice command line configuration mechanism to compose models [AL].
2520 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2521 the previous one (KCCFLN05). It relies on something that looks like
2522 a max-min sharing mechanism but cannot be written as such. A new solver
2523 was thus designed [AL].
2524 * Add a new solver to lmm. Based on Lagrange optimization and
2525 gradient-based descent, it enables to efficiently maximise systems s.a
2527 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2529 This solver enables to propose two new network models for TCP Reno and
2530 TCP Vegas based on Low's work. These models still need to be fully
2531 tested though [Pedro Velho].
2534 * Bug fix in SD_simulate. Now the time bound given as argument is
2536 * Use the new parallel task model (ptask_L07) as default.
2537 * Use the SURF command line configuration mechanism.
2538 * 0-size tasks (for synchronization) should now work.
2540 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2542 SimGrid (3.2) stable; urgency=high
2546 We still experience issues on this platform, but we believe that at
2547 least MSG is usable.
2549 GRAS API BREAKAGE (for simplification purpose, sorry):
2550 * the gras_msgtype_by_name is not used anymore. Instead of
2551 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2552 you can write (and must)
2553 gras_msg_send(toserver, "request", &request);
2554 - If you still want to pass a gras_msgtype_t to the function (to cache
2555 the type and avoid the lookup time), use the gras_msg_send_() variant.
2556 - Impacted functions:
2557 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2558 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2559 * The callbacks are now expected to return 0 when everything went well
2560 (just like the main() function)
2562 GRAS new features and improvements:
2563 * New module mechanism where user code can use per process globals [Mt]
2564 This is similar to gras_userdata_*() functions, but for libraries. It
2565 factorize some code developped over and over in the examples and AMOK.
2566 It has still to be documented and used (only amok/peermanagement is
2568 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2569 loss when they were small enough to fit into the buffer and sent quickly
2570 enough so that they can all get received in one shoot.
2571 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2572 if not found. Use the *_or_null() variant for the old semantic.
2573 * In gras_msg_handle, do not discard messages without callback.
2574 They are probably messages to be explicitly awaited later (ie, proofs of
2575 mis-synchronization in userland since they are sent before being awaited)
2576 No big deal usually.
2577 * gras_socket_meas_send/recv: semantic changed!
2578 The numerical arguments used to be (1) the total amount of data to send
2579 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2580 messages. This was need for the fool willing to send more than MAXINT
2581 bytes on quite fat pipes.
2584 * Do really rename the hostmanagement module to peermanagement. [Mt]
2585 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2586 but this is rather new and this was documented in the module
2587 documentation (poor excuses, I admit)
2588 * Bandwidth measurement semantic changed! This follows the changes to
2589 gras_socket_meas_send/recv explained above.
2592 * A sequential mode has been added to the workstations. When a workstation
2593 is in sequential mode, it can execute only one task, and the other tasks
2594 are waiting in a FIFO. [Christophe Thiery]
2597 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2598 model for SIMDAG. [Christophe Thiery]
2599 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2600 equal to 0, which caused some bad behaviors in
2601 saturated_constraint_set_update. I now use a threshold mechanism like in
2605 * When running manually src/testall, you select specific units [Mt]
2606 testall is the result of our cunit mechanism, and should replace all
2607 the scripty thingy around since bash don't run easily on billware.
2609 * A mallocator system has been added. [Christophe Thiery]
2610 Mallocators allow you to recycle your unused objects instead of freeing them
2611 and allocating new ones.
2613 Documentation update:
2614 * FAQ reworking + New FAQs:
2615 - "Valgrind spits tons of errors!" [Mt]
2616 - "How to repport bugs" [Mt]
2617 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2618 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2619 - Communication time measurement within MSG [AL]
2620 - I experience weird communication times when I change the latency [AL]
2621 * GRAS tutorial [Mt]
2623 - an introduction to the framework and to the used communication model
2624 - an initiatic tour introducing the most proheminent features:
2626 . Lesson 0: Installing GRAS
2627 . Lesson 1: Setting up your own project
2628 o Part 2: Message passing
2629 . Lesson 2: Exchanging simple messages
2630 . Lesson 3: Passing arguments to the processes (in SG)
2631 . Lesson 4: Attaching callbacks to messages
2632 . Lesson 5: Using globals in processes
2633 . Lesson 6: Logging information properly
2634 . Lesson 7: Using internal timers
2635 . Lesson 8: Handling errors through exceptions
2636 . Lesson 9: Exchanging simple data
2637 . Lesson 10: Remote Procedure Calling (RPC)
2638 . Lesson 11: Explicitely waiting for messages
2639 . Recapping of message passing features in GRAS
2640 - A HOWTO section containing:
2641 o HOWTO design a GRAS application
2642 More are due, of course. They will come latter. In the meanwhile, you can
2643 check the examples which are still here.
2645 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2647 SimGrid (3.1) stable; urgency=high
2651 There was a stack corruption somewhere, visible only when optimizing
2652 with these versions. [Vince]
2655 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2656 module that enabled to play with Directed Acyclic Graphs. It is built
2657 directly on top of SURF and provides an API rather close to the old
2658 SG. Some old codes using SG are currently under rewrite to check that
2659 all needful functions are provided. [Christophe Thiery]
2662 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2663 extension of the classical CLM03 model that gracefully handles
2664 failures. This is now the default model for MSG and GRAS. It doesn't
2665 handle parallel tasks yet though. [AL]
2666 * Bug fix: Weights were not correctly set in the network part.
2667 WARNING: This may have resulted in incorrect results with simulations
2668 where there are more than one flow on a given link. [AL]
2671 * After a (long ?) discussion on simgrid-devel, we have decided that the
2672 convention we had on units was stupid. That is why it has been decided
2673 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2674 WARNING : This means that all previous platform files will not work as
2675 such with this version! A warning is issued to ask users to update
2677 A conversion script can be found in the contrib module of the CVS, under
2678 the name contrib/platform_generation/surfxml_update.pl [MQ]
2681 * Bug fix: Processes were started in reverse order, wrt deployment file.
2682 WARNING: if your code relies on this bug, please fix it. [AL]
2683 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2684 being executed on two different locations. [AL]
2685 * Bug fix: Failures are now better supported thanks to Derrick's tests
2686 (there was many failure situations I hadn't thought of and that weren't
2687 correctly handled). [AL]
2688 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2692 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2693 Exception occurring on server-side are propagated back to client (!).
2695 API CHANGE: the callback changed their prototype. Change:
2696 int my_handler(gras_socket_t expeditor, void *payload_data) {
2698 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2699 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2701 * New! function: gras_msg_handleall to deal with all messages arriving
2702 within a given period.
2703 * New! function: gras_socket_server_range to get a server socket in a
2704 range of port numbers (ease to avoid port number conflicts) [MQ]
2705 * New! gras processes display their backtrace when they get a SIGUSR1
2706 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2707 Sweet to debug RL processes [MQ]
2711 - Do not force experiment sizes to be expressed in kb, or it becomes
2712 impossible to measure the latency this way (needs one byte-long tests)
2713 WARNING: this changes the amok_bw_* function semantic. [MQ]
2714 - Implements the link saturation stuff. [MQ]
2715 * Peer management module:
2716 New! module factorizing code that we wrote over and over [MQ].
2719 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2720 - Test units are placed directly into the library code, they get extracted
2721 automatically and placed into the src/testall binary.
2722 - Convert most of the XBT tests to this system.
2723 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2724 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2725 live in the same C code. This required to change a little bit the API
2726 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2727 * New module: sparse graph structure with basic algorithms (this is work
2728 in progress and the API is not considered to be frozen yet). [AL]
2729 * Display more information on backtraces: source line & function names are
2730 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2731 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2732 * Reworked a little bit some #include statements to load only required
2733 headers. Some user code that relied on SimGrid to include stdlib or
2734 stdio may need to include it by themselves. [AL]
2735 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2737 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2738 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2739 but I guess that almost nobody used those parts. [MQ]
2741 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2743 SimGrid (3.0.1) stable; urgency=low
2746 * Unfortunately, I had missed 5 misnamed functions:
2747 xbt_fifo_item_t xbt_fifo_newitem(void);
2748 void xbt_fifo_freeitem(xbt_fifo_item_t);
2749 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2750 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2751 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2752 They're now deprecated. Please use their new versions:
2753 xbt_fifo_item_t xbt_fifo_new_item(void);
2754 void xbt_fifo_free_item(xbt_fifo_item_t);
2755 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2756 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2757 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2759 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2760 * Documentation: xbt_log module unmercifully reworked [MQ]
2761 * Bugfix: there was a problem with the ending of contexts with
2762 the pthread backend. It caused some weird deadlock or behavior
2763 depending on the pthread implementation. [AL]
2764 * Bugfix: get the exceptions raised in the simulator repport where
2765 and why they come from when they are not catched in time [AL, MQ]
2768 * Bugfix: Do repport the error when two non-connected hosts try to
2769 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2772 * Add additionnal checkings on communications. Assert that two
2773 communicating hosts are connected by a set of links... [AL]
2776 * Add additionnal checkings on channel values in communication [AL]
2777 * New: MSG_task_get_source to see on which host a task was generated [HC]
2778 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2779 the number of tasks waiting to be received on channel and sent
2781 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2782 waits for the first task coming from a given host.. [AL]
2784 GRAS new functionnalities: [MQ]
2785 * Enhance the parsing macro to allow the size of multidimentional objects
2786 to be given thru annotations.
2787 * New example (and documentation): Matrix Multiplication a la RPC
2788 (as when I was young!) and fix a bunch of bugs found on the way.
2790 GRAS performance improvements: [MQ]
2792 * Reduce the amount of cbps creation/destruction by making it static to
2793 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2796 * Change libdata to a set so that we can search for stuff by ID (and thus
2797 reduce the insane amount of dict lookups)
2800 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2801 RL communication on top of SG since it's so uneffective.
2802 It may also be used for inter-thread communication in RL, one day.
2803 * Use gras_datadesc_copy() to exchange messages on top of SG
2805 - improve message exchange performance on top of SG
2806 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2807 * Don't exchange on the network the size of the used part of buffer,
2808 instead, specify the possible buffer size to read().
2810 - reduces the amount of read/write calls (one pair per exchange)
2811 - reduces the amount of exchanged data (the size)
2812 - allows to retrieve all arrived data on receiver side, if we don't need
2813 it right now (subsequent read will peek the buffer)
2814 - allows the receiver to proceed with the begining of the stream before
2815 everything is arrived
2816 - make it possible to build an iov transport (using readv/writev)
2818 - take care of the data with non-stable storage (like stacked data),
2820 * If possible, TCP send uses vector I/O (when writev() is here)
2821 - Don't use it for receive since we send data sizes and data on the
2822 same stream, so we wouldn't be able to chain large amount of chunks
2823 before having to flush the stuff to read the size.
2824 * Rework the transport plugin mechanism to simplify it and reduce the
2825 amount of pointer dereferencement when searching for the right function
2828 * I guess that now, we do almost as few system calls as possible while
2829 doing as few data copy as possible.
2831 To improve it further, we could try to send all the sizes first and then
2832 all the data (to use iov on receiving size), but it's only a partial
2833 solution: when you have 2 dimensional data, the sizes of the second
2834 dimension is data of the first dimension, so you need 3 streams.
2836 I'm not sure the potential performance gains justify the coding burden.
2838 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2840 SimGrid (3.00) stable; urgency=high
2843 * New! Give the possibility to hijack the surf parser and thus bypass
2844 MSG_create_environment and MSG_launch_application. Have a look at
2845 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2847 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2849 SimGrid (2.96) unstable; urgency=low
2854 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2855 This deprecates the xbt_error_t mechanisms.
2856 It modifies (simplifies) all XBT and GRAS API.
2857 MSG API keeps unchanged (exceptions raised by XBT are catched from
2858 within MSG and masked with existing error handling facilities)
2861 * New! Add a FATPIPE model. [AL]
2862 * New! Add a parallel task model. [AL]
2863 * New! Add automatically a loopback interface (in the default
2864 network model) if none was precised.
2867 * Bugfix: MSG_process_resume now works with the current running process.
2869 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2870 * Modification of MSG_task_get_compute_duration. Once a task has been
2871 processed, the value returned by this function is now equal to 0. [AL]
2872 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2873 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2874 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2875 of a m_task. For now, MSG_task_cancel only works with computation
2877 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2878 of the processor (in Mflop/s) regardless of the current load on the
2880 * API Change: use proper naming convention for MSG_getClock and
2881 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2883 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2884 This function changes the priority of a computation task. This priority
2885 doesn't affect the transfer rate. A priority of 2 will make a task
2886 receive two times more cpu power than the other ones. This function
2887 has been added to suit the needs of Nguyen The Loc and hasn't been that
2888 much tested yet. So if it fails, please report it and send me your code.
2890 * API Change: removed all functions and types that were marked "deprecated"
2891 since many months. Renamed MSG_global_init_args to MSG_global_init.
2893 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2895 SimGrid (2.95) unstable; urgency=low
2898 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2899 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2900 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2903 * Bugfix: Allow absolute paths for platform description files [MQ]
2904 * Bugfix: do free the variables after use. Leads to drastic performance
2906 * Implement max_duration (ie, timeouts) on resources [AL]
2909 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2911 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2912 the needed support (timeouts) [AL]
2915 * Implement measurement sockets. You can now get the bandwidth between two
2916 hosts thanks to AMOK (see below). [MQ]
2917 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2918 dynar over the network (yeah) [MQ]
2919 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2920 * simple/static token-ring example. [Alexandre Colucci and MQ]
2921 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2922 in place before [MQ]
2925 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2926 type parsing. "short" and "long" modifiers were also ignored; other
2927 modifier (such as reference level) are still ignored. [MQ]
2928 * Embeed the buffer size within the buffer itself on SG. [MQ]
2929 That way, send() are atomic and cannot get intermixed anymore (at least
2930 the ones which are less than 100k; bigger messages still have the issue)
2931 * Array size pushed by the field, not by the field type (or each
2932 and every long int will push stuff to the cbps) [MQ]
2933 * use select() to sleep since it allows to portably sleep less than one
2936 GRAS (minor cleanups)
2937 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2940 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2941 * type category 'ignored' killed as it was never used and were difficult
2943 * whether a type can cycle or not is now a flag, leaving room for more
2944 flags (as "ignored", if we feel the need one day ;)
2945 * Rename raw sockets to measurement sockets since "raw" has another
2946 meaning in networking community.
2949 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2950 and offers features not belonging to GRAS but that most applications
2951 need. One day, it may be a set of plugins loadable at runtime.
2952 * New module: bandwidth
2953 bandwidth measurement between arbitrary nodes running this module. [MQ]
2955 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2957 SimGrid (2.94) unstable; urgency=low
2959 The first beta release of SimGrid 3 !
2963 * Update the main page and the FAQ. Adding references to gforge.
2966 * Add a gras_os_getpid function.
2969 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2970 * Extend the logs so that they also print PID, hostname, date, ... if
2972 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2973 and kill the old version which were in testsuite/
2974 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2978 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2979 for users. Instead, each programming environment (GRAS, MSG,...) use it
2980 when needed to provide such a feature to users.
2981 Don't shortcut the mechanism or you will also shortcut the virtualization
2982 you need on the simulator.
2986 * Cleanups in configury with regard to compile optimization/warning flags.
2987 Also add -fno-loop-optimize to any powerpc since it's the optimization
2988 killing gcc (< 3.4.0).
2989 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2990 needed by MSG examples complications
2991 * Borrow configury beautifications from PHP
2994 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2995 warning (thanks loris for stumbling into this one).
2996 * Bugfix: stop loading private headers (gras_config.h) from the public
3000 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
3001 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
3002 RL would imply a DNS resolver, which is *hard* to do in a portable way
3003 (and therefore delayed).
3004 * Implement a real timer mechanism and use it in timing macros. This allows
3005 to avoid rounding errors and get a 0.000005 sec precision in timing
3006 macros. While I was at it, various cleanups:
3007 - allow to declare more than one timed section per file (fix a stupid bug)
3008 - move some private declaration to the right place
3009 - merge conditional execution and timing macros into emulation module
3010 - document the module
3011 - make sure the module cleanups its mess on gras_exit
3012 * Documentation improvements:
3013 - (new) how to compile applications using GRAS
3014 - (new) emulation support (timing macros)
3016 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
3018 SimGrid (2.93) unstable; urgency=low
3020 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
3023 - Use Paje properly where used. Still to be sanitized properly.
3024 - Portability fix: Add an implementation of the contexts using pthread
3028 - Add xbt_procname(): returns the name of the current process.
3029 Use it to show the current process's name in all logging.
3031 - fix detection of older flex version and the reaction, since we do
3032 depend on modern ones (we use lex_destroy)
3033 - Better separation of SG and RL in the libs: remove all simulation code
3034 from libgras. As a result, this lib is now only 200k when stripped.
3035 Some of the xbt modules may also be duplicated (two sets and such) and
3036 should be cleaned/killed before SG3.
3037 - Insist on using xlC on AIX because of weird problems involving gcc there.
3038 - Cleanup the make remote stuff. This is now done by scripts
3039 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
3040 mainly for our private use, but we're working on changing them to user
3043 - Bugfix: flush the socket on close only if there is some *output*.
3044 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
3046 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
3049 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
3051 SimGrid (2.92) unstable; urgency=low
3053 Alpha 3 on the path to SimGrid 3
3057 - New! First try of benchmarking macros.
3058 - New! First try so that gras_stub_generator generate deployment and
3059 remote compilation helpers.
3061 - Bugfix: Initialization fix in msg_test.
3065 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
3067 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
3069 - API Change: gras_os_sleep to take the amount of seconds as a double.
3070 Accepting an int was error prone since it was the only location where
3071 seconds were coded as such. It leaded to damn rounding errors.
3072 - Bugfix: Hard to belive that timers ever worked before this.
3074 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3076 SimGrid (2.91) unstable; urgency=low
3078 Alpha 2 on the path to SimGrid 3
3082 - Bug fix in the lmm_solver.
3084 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3085 through the function MSG_paje_output.
3086 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3087 - It is possible to bound the rate of a communication in MSG with
3088 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3089 to write it in the changelog).
3090 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3095 - Various cleanups to the autotools stuff
3096 - Begin to move Gras examples to examples/gras/
3097 - Let make distcheck work again (yeah!)
3099 - documentation overhauled using doxygen.
3100 gtk-doc-tools is dead in SimGrid now.
3101 - Automatically extract all existing logging categories, and add the list
3102 to the documentation (long standing one, to say the less)
3104 - Cleanup the known architecture table. Reorder the entries to group what
3105 should be, and use a more consistent naming scheme.
3106 (some of the test dataset are still to be regenerated)
3107 - New! Allow library to register globals on each process just as userdata
3109 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3110 lookup time (for now).
3111 Use it in msg and trp.
3112 This cleans a lot the internals and helps enforcing privacy of the
3113 headers between the gras components.
3114 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3115 - Bugfix: gras_os_time was delirious in RL.
3116 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3118 - Reenable GRAS now that it works.
3120 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3122 SimGrid (2.90) unstable; urgency=low
3124 Alpha 1 on the path to SimGrid 3
3126 * It is a long time since the last release of SimGrid. I'm sorry about
3127 that but as I had told you, I was rewriting a lot of things. I apologize
3128 to those who had been reporting bugs to me and that I had not answered.
3129 If your bug is still in the new version, please tell me. Here is a
3130 summary of the main changes.
3132 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3133 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3134 portability, speed, and a lot more but you'll figure it out later.
3135 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
3136 I wanted to release the new MSG as soon as possible and I have
3137 broken GRAS, which is the reason why, for now, only MSG is fully
3138 functional. A laconic description of these projects is available
3139 in the documentation.
3141 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3142 kernel optimized for our needs (called SURF but only the developers
3143 should use it). Hence, MSG is now roughly 30 times faster and I think
3144 that by rewriting a little bit MSG, I could event speed it up a little
3145 bit more. Beside the gain in speed, it is also much easier to encode a
3146 new platform model with SURF than it was with SG. More to come...
3148 * REVOLUTION 3: I have tried to change a little as possible the API of
3149 MSG but a few things really had to disappear. The main differences
3150 with the previous version are :
3151 1) no more m_links_t and the corresponding functions. Platforms are
3152 directly read from a XML description and cannot be hard-coded
3153 anymore. The same format is used for application deployment
3154 description. The new format is described in the documentation.
3155 Have a look in tools/platform_generation. There is a tiny script
3156 that converts from the old platform format to the new one. Concerning
3157 the application deployment format, parsing the old one is tricky.
3158 I think most of you should however be able to convert your files. If
3159 it is really an issue, I can write a C code that does the conversion.
3161 2) the toolbox tbx does not exist anymore. We now have a library
3162 with much more data-structures but without the hash-tables (we have
3163 dictionaries that are much faster).
3165 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3167 *****************************************************************************
3168 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3169 * don't feel like converting the oldies. (Mt) *
3170 *****************************************************************************
3173 Version 2.90: "the long awaited one"
3174 - Finished rewriting and debugging MSG. Rewrote the documentation.
3175 - disable GRAS for now since it needs to be ported to the newest SG
3178 - Finish the port to windows (using mingw32 for cross-compile)
3181 - Main loop and datastructures of SURF. A cpu resource object is
3182 functional. Surf can thus be used to create cpu's with variable
3183 performance on which you can execute some actions.
3185 2004-11-15 Martin Quinson
3186 - Port to ARM. Simply added the alignment and size descriptions. Should
3187 work, but the ARM machines are so slow that I didn't had the opportunity
3188 to 'make check' over there yet.
3190 2004-11-15 Arnaud Legrand
3191 - Trace manager now written. It uses a heap structure and is therefore
3192 expected to be efficient. It may however be speeded up (particularly
3193 when many events occur at the same date) by using red and black
3194 trees. One day maybe...
3195 - Max-min linear system solver written. It uses a sparse matrix
3196 structure taking advantage of its expected use. Most operations are
3197 O(1) and free/calloc are called as few as possible. The computation of
3198 the minimum could however be improved by using a red and black tree
3201 2004-11-03 Arnaud Legrand
3202 - Rename every gras_* function that was in xbt/ to its xbt_
3204 - Add a heap and a doubly-linked list to xbt
3205 - Added a dichotomy to the dictionaries. make check works as well before
3206 so I assume that the patch is correct. I do not know however if things
3207 run effectively faster than before now. :)
3209 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3210 SimGrid, and the version number is bumped to 2.x
3212 2004-10-29 Martin Quinson
3213 - Introduction of the remote errors.
3214 They are the result of a RMI/RPC on the remote machine.
3215 ErrCodes being scalar values, you can't get the host on which those
3216 errors did happen. Extending the error mechanism as in Gnome is possible.
3217 No idea yet whether it is a good idea.
3219 2004-10-28 Martin Quinson
3220 - Interface revolution: the Starred Structure Eradication.
3221 I used to do typedef struct {} toto_t; and then handle *toto_t.
3222 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3223 - ???_t is a valid type (builded with typedef)
3224 - s_toto_t is a structure (access to fields with .)
3225 - s_toto is a structure needing 'struct' keyword to be used
3226 - e_toto_t is an enum
3227 - toto_t is an 'object' (struct*)
3229 typedef struct s_toto {} s_toto_t, *toto_t;
3230 typedef enum {} e_toto_t;
3231 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3232 s_toto_t) is private.
3234 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3235 so that it can set the variable to NULL. It was so for dicts and sets,
3236 it changed for dynars.
3238 - Fix a bunch of memleaks in dict_remove
3239 - Fix a bug in sg/server_socket opening: it failed all the time.
3241 2004-10-07 Martin Quinson
3242 - Speed up dynar lookup operation a bit.
3244 gras_dynar_get is dead.
3246 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3247 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3248 (which gives you the address of the stored data).
3250 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3251 copy of the data using an affectation to do the job and not a memcpy.
3253 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3254 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3256 It does not really speedup the dynar test because they are
3257 setting elements all the time (and look them seldom). But the dict does
3258 far more lookup than setting.
3260 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3262 2004-10-05 Martin Quinson
3263 - Allow to (en/dis)able the cycle detection at run time.
3265 Whether we should check for cycle or not is now a property of each
3266 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3267 datadesc_cycle_unset allow to remove this property when previously set.
3269 Note that the cycle detection is off by default since it impacts the
3270 performance. Watch the data you feed GRAS with ;)
3272 This property is hereditary. Any element embedded in a structure having it
3273 set have it set for the time of this data exchange.
3275 You should set it both on sender and receiver side. If you don't set it on
3276 sender side, it will enter an endless loop. If you forget on receiver
3277 side, the cycles won't be recreated after communication.
3279 - Header reorganization.
3280 Kill gras_private.h, each submodule must load the headers it needs.
3282 2004-10-04 Martin Quinson
3283 - Interface revolution: do not try to survive to malloc failure.
3285 Now, gras_malloc and friends call gras_abort() on failure.
3286 As a conclusion, malloc_error is not a valid error anymore, and all
3287 functions for which it was the only gras_error_t return value are
3288 changed. They now return void, or there result directly.
3289 This simplify the API a lot.
3291 2004-09-29 Martin Quinson
3292 - Re-enable raw sockets.
3293 Created by gras_socket_{client,server}_ext;
3294 Used with gras_raw_{send,recv}
3297 It should allow to kill the last bits of gras first version soon.
3299 This is not completely satisfactory yet (duplicate code with
3300 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3303 - Simplify transport plugin (internal) interface by not passing any
3304 argument to _server and _client, but embedding them in the socket
3307 2004-09-28 Martin Quinson
3308 - Finish the port to AIX.
3309 autoconf was my problem (segfault within the malloc replacement
3310 function. No idea why)
3312 2004-09-16 Martin Quinson
3313 - Fix some size_t madness on 64bit architectures.
3315 2004-09-08 Martin Quinson
3316 - Reduce the number of system headers loaded, overload some more system
3317 calls (such as malloc to cast the result of the system one, and work
3319 - Fix and reintroduce the config support
3321 2004-09-07 Martin Quinson
3322 - Source code reorganization to allow Arnaud to surf all over there.
3323 - Allow to document the logging categories.
3324 - Remove all uppercase from logging categories and useless cleanup in names.
3326 2004-08-18 Martin Quinson
3327 Version 0.6.2 (protocol not changed; API changed)
3328 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3329 gras_error_t), and NULL when not found. Functions expecting a msgtype
3330 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3331 hopefully usefull message.
3332 - Portability to prehistoric sparcs again
3334 2004-08-17 Martin Quinson
3335 Version 0.6.1 (protocol not changed; ABI not changed)
3336 - prealloc some buffers to speed things up
3338 2004-08-11 Martin Quinson
3339 Version 0.6 (protocol not changed; ABI expended)
3340 - The parsing macro can deal with the references, provided that you add
3341 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3343 2004-08-09 Martin Quinson
3344 Version 0.5 (protocol not changed; ABI changed)
3345 - Allow to off turn the cycle detection code in data exchange at
3346 compilation time. It should be at run time, but I'm short of time (and
3347 the config stuff is still broken). That way, we keep dict out of the
3348 critical path, which is good because the performance is poor:
3349 - search not dichotomial yet
3350 - dynar give no way to access their content and memcpy everytime
3351 - In composed data description (struct, ref and so on), stop foolness of
3352 keeping the subtype's ID, but store the type itself. This keeps sets out
3353 of the critical path, which is good since they rely on dynar and
3354 dictionnaries. The only loose of that is that we cannot detect the
3355 redeclaration of a structure/union with another content (but I'm not sure
3356 the code detected well this error before anyway). We still can detect
3357 the redefinition discrepancy for the other types.
3358 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3359 it breaks the code because of type-punning used all over the place).
3360 This breaks on all non-gcc architectures (for now).
3362 All those changes (plus the buffer of last time) allow me to gain 2 order
3363 of magnitude on cruel tests consisting of 800000 array of integers on two
3364 level of a hierarchical structure (200 secondes -> 4 secondes)
3367 - the selector of reference must now return the type it points to, not
3368 the ID of this type.
3370 2004-08-06 Martin Quinson
3371 Version 0.4 (protocol changed; ABI not changed)
3372 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3373 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3374 - Implement a buffer transport, and use it by default (it relies on tcp in
3375 real life and on sg in simulation).
3376 That's a bit hackish since I had a new field to the structure to store
3377 its data without interfering with the subtype ones. Inheritance
3378 is tricky in C. And that's a kind of reverse inheritance with one class
3379 derivating two classes. Or maybe a game with java interfaces. Anyway,
3380 that's damn hard in C (at least).
3381 Moreover, I got tired while trying to ensure plugin separation and
3382 genericity in SG mode. MSG wants me to do weird things, so let's go for
3383 cruel hacks (temporarily of course ;).
3384 See comment in transport_private.h:71
3385 - do not include all the _interface headers in private but in the files
3386 which really need them (to cut the compilation time when they are
3389 2004-07-26 Martin Quinson
3390 Version 0.3 (protocol not changed; ABI changed)
3391 - Major overhault of the datadesc interface to simplify it:
3392 - shorted the function names:
3393 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3394 - add a trivial way to push/pop integers into the cbps without malloc.
3395 This allows to make really generic sub_type description, which simply
3396 pop their size of the stack.
3397 - add a function gras_datadesc_ref_pop_arr() which does what users want
3398 most of the time: Declare a dynamic array (which pops its size of the
3399 stack) and declare a reference to it. Poor name, but anyway.
3400 - kill the post-send callback, add a post-receive one
3402 2004-07-23 Martin Quinson
3403 Version 0.2 (protocol changed; ABI changed)
3404 - add some testing for cpbs in the test cases, and fix some more bugs.
3405 This invalidate again the little64 data file, since I cannot regenerate
3407 - remove an awfull optimization in the logging stuff, allowing me to:
3408 - understand it again
3409 - learn gcc how to check that the argument match the provided format
3410 - fix all errors revealed by gcc after that
3411 - internal keys of dict are not \0 terminated. Deal with it properly in
3412 loggings instead of segfaulting when the user want to see the logs :-/
3414 2004-07-22 Martin Quinson
3415 - Fix some stupid bug preventing cbps (callback postit) from working
3417 2004-07-21 Martin Quinson
3418 - Some documentation cleanups
3419 - remove the useless last argument of msgtype_declare
3420 - rename the Virtu functions to fit into the 'os' namespace
3421 - move headers src/include -> src/include/gras/ and stop fooling with
3422 gras -> . symbolic link
3423 - make distcheck is now successful
3425 2004-07-19 Martin Quinson
3427 - Build shared library also
3428 - Install html doc to the right location
3429 - stop removing maintainer files in make clean
3430 - build tests only on make check
3432 2004-07-13 Martin Quinson
3434 - No major issue in previous version => change versioning schema
3435 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3436 corresponding dataset.
3438 2004-07-11 Martin Quinson
3440 - Get it working with any kind of structure (we can compute the padding
3441 bytes remotely for all the architectures I have access to)
3442 - Implement the structure parsing macro (still not quite robust/complete)
3443 - Improvement to the remote testing toysuite
3445 2004-07-10 Martin Quinson
3446 [autoconf mechanism]
3447 - get ride of a bunch of deprecated macros
3448 - actually run the test for two-compliment, not only compile it :-/
3449 - test whether the structures get packed (and bail out if yes. Damn.
3450 Alignment is a serious matter)
3451 - test whether the structures get compacted (but respecting the alignment
3452 constraints of each types)
3453 - test whether the array fields of structures can straddle alignment boundaries
3455 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3456 (took me 2 hours to find that bug, looking at the wrong place)
3457 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3458 before sending/receiving and are used to compute the offsets of fields
3459 - Given that padding size depend even on compiler options, keep track of
3460 alignment and aligned_size only for the current architecture. Not a big
3461 deal since we send structure fields one after the other (seems
3463 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3464 program, let it work on linux/gcc/little32. portability todo.
3466 2004-07-08 Martin Quinson
3467 - import and improve remote compilation support from FAST
3468 - make sure make check works on half a dozen of machines out there
3470 2004-07-07 Martin Quinson
3471 Let's say it's version 0.0.3 ;)
3472 - Implement conversions (yuhu!)
3473 - Let it work on solaris (beside conversion, of course)
3474 - Stupid me, using rand() to generate the conversion datatests in not wise.
3476 2004-07-06 Martin Quinson
3477 - Let make dist work, since I'm gonna need it to compile on remote hosts
3478 - Let Tests/datadesc_usage write the architecture on which the file was
3479 generated as first byte.
3480 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3481 architecture descriptions.
3482 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3483 on those architectures.
3484 - Optimization: send/recv array of scalar in one shoot
3486 2004-07-05 Martin Quinson
3487 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3489 - Plug a whole bunch of memleaks
3490 - each process now have to call gras_{init,exit}. One day, their log
3491 settings will be separated
3492 - Continue the code factorisation between SG, RL and common in Transport.
3494 2004-07-04 Martin Quinson
3496 - Redistribution between SG and RL.
3497 We wanna have to accept in SG, so move accepted related parts of RL in
3498 the common part. (more precisely, the dynar of all known sockets is no
3499 more a static in transport.c, but part of the process_data)
3501 [gras_stub_generator]
3502 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3504 2004-07-03 Martin Quinson
3505 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3506 them all up in one shot)
3508 - Fix the ugly case of reference to dynamic array.
3509 - New (semi-public) function gras_datadesc_size to allow the messaging
3510 layer to malloc the needed space for the buffer.
3512 - gras_socket_close now expect the socket to close (and not its address to
3513 put NULL in it after it). This is because the socket passed to handlers
3514 is one of their argument (=> not writable).
3516 - propagate the interface cleanup from last week in datadesc, ie remove a
3517 superfluous level of indirection. User pass adress of variable
3518 containing data (both when sending and receiving), and not of a variable
3519 being a pointer to the data. Let's say that I like it better ;)
3520 The price for that is constructs like "int msg=*(int*)payload" in
3521 handlers, but it's a fine price, IMHO.
3523 - Let it work in RL (yuhu)
3525 2004-06-21 Martin Quinson
3527 - porting SG plugin and SG select to new standards (works almost).
3528 - plug memleaks and fix bugs around.
3531 - cleanup the prototype of data recv and force users to specify when they
3532 want to handle references to objects. Test case working even for cycles.
3533 - plug memleaks. Valgrind is perfectly ok with this.
3535 2004-06-12 Martin Quinson
3537 - cleanup the separation between plugin and main code in plugin creation
3539 2004-06-11 Martin Quinson
3541 - Reput hook for raw sockets, needed for BW experiments
3542 - kill a few lines of dead code
3543 [Data description] Interface cleanup
3544 - gras_datadesc_by_name returns the searched type or NULL.
3545 That way, no variable is needed to use a type desc once, which makes
3547 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3548 two parameters were strings (field name, type name), leading to
3550 [Dicos] Interface cleanup
3551 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3552 This is consistant with the dynar API.
3554 2004-04-21 Martin Quinson
3556 - Porting to new standards.
3558 - interface cleanup.
3559 There is no bag anymore, no need to take extra provision to mask the
3560 pointers behind "ID".
3561 Better splitup of functions between files create/exchange/convert.
3562 This is still a bit artificial since convert and receive are so
3563 interleaved, but anyway.
3565 - add a queued message list to procdata (the ones not matching criteria
3567 - factorize some more code between SG and RL wrt procdata
3569 - use gras_exit in example to track memleaks
3570 - get rid of gs_example now that GS is properly integrated into gras
3571 - update run_test to integrate the latest tests (datadesc)
3573 - rename WARNINGn macros to WARNn since it prooved error-prone
3575 2004-04-19 Martin Quinson
3577 - register init/exit functions within gras module mechanism
3578 - send/receive function.
3579 Convertion is not implemented, but short-cutted if not needed.
3580 struct/array elements are sent one by one (instead of block-wise), but
3581 nobody really cares (yet). Get a prototype before optimizing.
3582 - tests (using a file socket) for DD send/receive on:
3583 - base types: int, float
3584 - array: fixed size, string (ie ref to dynamic string)
3585 - structure: homogeneous, heterogeneous
3586 - chained list, graph with cycle
3587 Believe it or not, valgrind is not too unhappy with the results. The
3588 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3592 - Bugfix when using a filename explicitely (instead of '-')
3594 2004-04-09 Martin Quinson
3596 - factorize more code between RL and SG in socket creation
3597 - Complete the implementation and tests of:
3599 o file (only in RL, and mainly for debugging)
3601 I lost 3 days to design a portable address resolver, and then decided
3602 that the prototype mainly have to run on my box.
3603 Addressing portability too early may be like optimizing too early :-/
3605 - use gras_init in the Tests instead of the crappy parse_log_opt
3606 (the latter function is removed)
3607 [Conditional execution]
3608 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3609 [Code reorganisation]
3610 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3611 Build examples against the RL library, since there is no way to disable
3612 its creation for now.
3614 For information, the beginning of coding on GRAS was back in june
3615 2003. I guess that every line has been rewritten at least twice since