1 SimGrid (3.19) NOT RELEASED YET (target: March 20 2018, 16:15:27 UTC)
4 - Fix MSG_task_get_remaining_work_ratio() to return 1.0 for tasks that have
8 - Execution->setHost() can be called after start() to migrate it.
9 - Comm::test_any() is now implemented.
10 - s4u::Actor now has onCreation() and onDestruction() signals.
13 - Remove XBT_LOG_CONNECT, it should be useless nowadays.
16 - The privatization of global variables via mmap() is now significantly more
17 flexible and allows the use of daemons with SMPI and to add new processes
18 while the simulation is running. (Adding new SMPI instances after the
19 simulation has already started should also be technically doable now.)
20 - Removed index notion from SMPI -- no more "getPid() - 1"!
21 This also means that the trace files
22 now contain actors called rank-<PROCESS-ID>. These are NOT the real
23 ranks, because tracing several communicators right now is not supported.
24 - Because of this, smpi_process_index() has now been deprecated and will
27 - #245: migrating an actor does not migrate its execution
30 - Remove unused run-time parameter "tracing/onelink-only".
32 SimGrid (3.18) Released December 24 2017
34 The "Ho Ho Ho! SimGrid 4 beta is coming to town" release.
36 MSG IS NOW DEPRECATED. Please use S4U for new projects.
37 - Support for MSG should not be removed from SimGrid before 2020,
38 but future evolutions will be limited to the integration of
39 user-provided patches.
42 - s4u::onDeadlock() signal, run before stopping the simulation.
43 - s4u::Exec: asynchronous executions (abstraction of background
45 - s4u::Host->getLoad() returns the achieved speed in flops/s
48 - Link::name() is deprecated and will be removed in v3.21, use
49 Link::getCname() instead.
50 - Mailbox::getName() changed to return a std::string, use
51 Mailbox::getCname() instead to get a char*.
52 - Storage::getName() changed to return a std::string, use
53 Storage::getCname() instead to get a char*.
54 - s4u::allStorages() becomes s4u::getStorageList(whereTo)
55 to have both allocation and deallocation of the map in user space.
58 - Forbid the use of MSG_task_get_flops_amount() on parallel tasks:
59 The amount of flops remaining to do is a vector, not a scalar.
60 - Introduce MSG_task_get_remaining_work_ratio(), which does what its
61 name implies on both sequential and parallel tasks.
62 - Both changes fix GitHub's #223 using PR #237 as a basis.
63 Thanks Michael Mercier.
64 - Most examples were converted to S4U and hidden elsewhere at tests.
65 Rationale: we still want MSG to work; we want newcomers to use S4U.
68 - LMM stuff moved to its own namespace: simgrid::kernel::lmm.
69 - Renamed LMM classes (e.g. s_lmm_system_t -> System).
72 - Switch to the faster dlopen privatization mechanism by default
73 - Documentation improvments
76 - Live migration is getting moved to a plugin. Dirty page tracking is
77 the first part of this plugin. This imply that VM migration is now
78 only possible if one this function is called :
79 - C/MSG: MSG_vm_live_migration_plugin_init()
80 - C/C++: sg_vm_live_migration_plugin_init()
81 - Java: Msg.liveMigrationInit()
82 For C and C++, "simgrid/plugins/live_migration.h" also has to be
86 - Define class simgrid::xbt::Path to manage file names.
87 - Removed unused functions:
88 - xbt/file.h: xbt_basename(), xbt_dirname(), xbt_getline()
89 - xbt/graph.h: xbt_graph_edge_get_length(), xbt_graph_edge_set_length,
90 xbt_graph_export_graphviz()
91 - xbt/str.h: xbt_str_join()
92 - Remove unused datatypes:
93 - xbt/heap.h: use std::priority_queue or boost::heap instead
94 - xbt/swag.h: use boost::intrusive::list instead
97 - New link_energy plugin for the consumption of the links.
98 - All of the operations on files and storage contents have been
99 packaged into a plugin (src/plugins/file_system). The current
100 public interface can be found in
101 include/simgrid/plugins/file_system.h
102 To use these functions you now have to initialize the plugin by
103 calling MSG_storage_file_system_init() just after calling
104 MSG_init() or sg_storage_file_system_init() just after creating
108 - Remove the undocumented/untested tag <include>
111 - Remove viva specific tracing as the tool is no longer maintained
114 - #248: Kill of finished processes leads to segfault
115 - #240: xbt_cond_wait_timeout should gracefully return for C apps
116 - #239: please implement signal s4u::onDeadlock()
117 - #230: segfaults when exit() without run()
118 - #225: s4u::Actor::kill() doesn not really kill victims in a join()
119 - #223: MSG_task_get_flops_amount() not working with parallel tasks
120 - #222: Actor::kill() doesn't really kill and segfaults
121 - #221: odd LMM warning when killing an actor
122 - #120: Memory leak when the processes are forcefully killed
124 SimGrid (3.17) Released October 8 2017
126 The Drained Leaks release: (almost) no known leaks despite the tests.
128 Even very long simulations will preserve your memory: our testsuite
129 stresses SimGrid for over 45mn on fast machines for a coverage of
130 over 80%, and there is only one single known leak, of about 4kb.
134 - Comm.detach(): start and forget about asynchronous emission. A cleanup
135 handler may be given to free resources if the comm cannot be completed.
136 - this_actor::send(mailbox) is now mailbox->put()
137 - New: simgrid::s4u::Comm::wait_all()
138 - New: Host.execute() for remote executions.
141 - Unused option network/sender-gap is removed.
144 - Deprecate MSG_task_isend_with_matching(): This unused feature
145 really complicates our internals. Will be removed in v3.20.
148 - Improved context termination. It is now possible to free resources, even
149 when a process is forcibly killed.
150 - ContextBoost: add support for Boost versions above 1.61.
153 - Bring back run-time option --cfg=exception/cutpath to remove exception
155 - Removed unused functions:
156 - xbt/str.h: xbt_str_split_str(), xbt_str_subst(), xbt_str_ltrim(),
157 xbt_str_rtrim(), xbt_str_trim().
158 - xbt/xbt_os_thread.h: xbt_os_thread_cancel(), xbt_os_thread_detach().
161 - Removed header files obsolete since SimGrid 3.12:
162 msg/datatypes.h, msg/msg.h, simdag/datatypes.h, simdag/simdag.h.
164 - #3: SD_exit should be made optional
165 - #120: Memory leak when the processes are forcefully killed
166 - #159: Threading test regression in Actor refcounting
167 - #170: simgrid::s4u::Comm::wait_any() returns too many comms
168 - #185: simgrid::s4u::Engine::instance()->shutdown() segfaults
169 - #186: Actor::killAll() segfaults if some process is blocked on wait()
170 - #191: VM migration and pstate
171 - #192: Updating the energy consumptions of all hosts crashes with VMs
172 - #195: All actors have PID=0 in the logs
173 - #204: Sometimes segfault with thread contexts and mmap privatization
174 - #222: Actor::kill() doesn't really kill and segfaults
175 - #225: Actor::kill() doesn't really kill when victims are doing a join()
177 ----------------------------------------------------------------------------
179 SimGrid (3.16) Released June 22. 2017.
181 The Blooming Spring Release: developments are budding.
184 - FreeBSD: Disable SMPI mmap privatization, switch automatically to dlopen.
185 - Mac, BSD: dlopen+thread broken, switch automatically to raw contexts.
186 - Java JAR file should be finally fully working on Mac OSX too.
188 XML platforms: Switch to platform v4.1 format.
189 * This is (mainly) a backward compatible change: v4 are valid v4.1 files
190 - <zone> can be used as a synonym for the now deprecated <as>
191 - <zoneRoute> can be used as a synonym for the now deprecated <asroute>
192 - <bypassZoneRoute> an be used as a synonym for the now deprecated <bypassAsRoute>
193 - <actor> can be used as a synonym for the now deprecated <process>
194 - state_file and avail_file periodicity is now easier to express
195 (check the documentation) the old behavior should still work.
196 * Storage: not backward compatible, but it seems that nobody ever used it
197 - Remove attribute 'content_type' of <storage_type>: was never used
198 - Make attribute 'model' of <storage_type> optional: for future usage
199 - Remove Bconnection model property: was never (in)validated. Replaced by
200 the maximum of the read and write bandwidth as a resource constraint.
203 - New and Backwards Compatibility break:
204 SD_simulate_with_update (double how_long, xbt_dynar_t changed_tasks_dynar)
205 When one wants to get the list of tasks whose states have changed during a
206 simulation round, s/he has to allocate and free a dynar and use it as
207 argument to this function. The former SD_simulate (double how_long)
211 - Allow multicore VMs, with the correct sharing (unless you overcommit)
212 BUG: vCPU overcommitting still leads to buggy sharing, though. WIP.
215 - New function to update the consumption of all hosts at once.
216 - Fix the model for multi-core hosts, linear in the amount of busy
217 cores with an abnormality for pIdle (see paper).
218 BUG: Still not satisfactory for ptask on mono-cores.
221 - The netzone are now available from the MSG API.
222 The old names still work, but are now deprecated.
225 - New algorithm to privatize globals: dlopen, with dynamic loading tricks
226 - New option: smpi/keep-temps to not cleanup temp files
227 - New option : smpi/shared-malloc-blocksize . Relevant only when global shared
228 mallocs mode is used, allows to change the size of the fake file used
229 (default 1MB), to potentially limit the number of mappings for large runs.
230 - Support for sparse privatized malloc with SMPI_PARTIAL_SHARED_MALLOC()
231 - Fortran ifort and flang compilers support
232 - New RMA calls supported (experimental) :
233 - MPI_Win_allocate, MPI_Win_create_dynamic, MPI_Win_attach
234 - MPI_Win_detach, MPI_Win_set_info, MPI_Win_get_info
235 - MPI_Win_lock_all, MPI_Win_unlock_all, MPI_Win_flush
236 - MPI_Win_flush_local, MPI_Win_flush_all, MPI_Win_flush_local_all
237 - MPI_Op_commutative, MPI_Fetch_and_Op, MPI_Compare_and_swap
238 - MPI_Rput, MPI_Rget, MPI_Raccumulate, MPI_Rget_accumulate
241 - New: this_actor::isMaestro() returns whether we are in kernel mode.
242 - Behind the scene, ActivityImpl are now automatically refcounted.
243 Removing this nasty bottleneck will greatly simplify our internals.
246 - Replay: New function xbt_replay_action_get():
247 Retrieve the function previously associated to an event type.
248 - DROPPED FUNCTION: xbt_str_varsubst()
249 - DROPPED MODULE: strbuff. We don't need it anymore.
250 - DROPPED MODULE: matrix. We don't need it anymore.
251 - DROPPED MODULE: lib. We don't need it anymore.
253 -- Thu Jun 22 22:48:12 CEST 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
255 SimGrid (3.15) Released March 22. 2017
257 The Spring Release: continuous integration servers become green
258 We fixed even the transient bugs on all target architectures:
259 Linux (CentOS, Debian, Ubuntu, Fedora), Mac OSX (Mavericks, El Capitan)
260 Windows, FreeBSD, NetBSD.
262 - Rename NetCards to NetPoints.
263 This was intended to help NS3 users, but that's not a netcard.
264 That's a point in the routing algorithm, let's avoid wrong simplifications.
267 - New: MSG_process_yield(). Stop and yield to other processes.
268 - New: MSG_process_daemon(). Daemon processes are automatically killed
269 when the last non-daemon process terminates
270 - New: MSG_process_ref/unref(). Fiddle with the process refcounting.
271 - Renamed MSG_energy_plugin_init() -> MSG_host_energy_plugin_init()
272 to make room for the upcoming network energy plugin.
273 - Drop MSG_host_get_current_power_peak: duplicates MSG_host_get_speed
276 - Ensure that an actor can kill itself with Process::exit()
277 - Kill the obscure NativeException. Nobody want to survive the issues
278 it denotes, so use JniException that is a RuntimeException (not to
279 be caught explicitly).
280 - Partial bug fix in initialization. SimGrid flags on command line were
281 consumed at C level but stayed in the original Java String[] args.
282 This could mess users' args[i] if SG flags were not put at the end of
284 The SimGrid flags are now removed from the Java arguments. However,
285 the number of arguments REMAINS UNCHANGED. It is then UNSAFE to test
286 if args.length is greater than the number of YOUR OWN ARGUMENTS.
287 It might be if you have --log or --cfg flags in the command line.
288 - Fix numerous memleaks all around the place. In particular, around VMs.
292 - s4u::onPlatformCreated: right before the simulation starts
293 - s4u::onSimulationEnd: right after the main simulation loop
294 - s4u::onTimeAdvance: right after a clock change (time is discrete)
295 - s4u::Host::onSpeedChange: when the pstate is changed, or when an
296 event from the availability_file changes the avail speed.
297 - Links are now usable from s4u
298 - New: Engine::hostList() and Engine::hostCount(). Still clumsy.
299 - New: Actor::suspend(), Actor::resume(), and Actor::migrate(new_host)
300 - New examples: The conversion of MSG examples to S4U has begun
301 - Actors: create, kill, migration, and suspend
302 - Applications: master-worker and token-ring
303 - Action replay: communications and storage
304 - Drop Host::getPstateSpeedCurrent() which duplicates Host::speed()
307 - Backwards Compatibility breaks
308 - The SD_route_get_list and SD_route_get_size functions have been removed.
309 They are replaced by the unique
310 void sg_host_route(sg_host_t src, sg_host_t dst, xbt_dynar_t links)
311 The route from src to dst is built in the links dynar whose size can
312 be obtained with xbt_dynar_length.
313 - The SD_route_bandwidth and SD_route_latency functions have been replaced by
314 sg_host_route_bandwidth and sg_host_route_latency.
315 Macros ensure the backwards compatibility, but you should fix your code
318 - Major C++ rewrite ongoing (SMPI used to be C compiled in C++).
319 This can break codes using SMPI internals (from private.h instead of the public smpi.h).
320 - Bump our claim of support from MPI 1.1 to MPI 2.2.
321 We don't support 100% of it, but it should be enough. Contact us if not.
322 - MPI_Win_lock/unlock, MPI_Get_accumulate support added (as for all RMA, implementation is
323 naive and probably inaccurate)
324 - New algorithm for SMPI_SHARED_MALLOC: global, where all blocks are
325 mapped onto a unique small file using some system magic.
326 - Bugfix: smpirun was sometimes misusing hostfile when wrongly terminated
327 - Fortran: cleanups, fixes, support of user-added operations
328 - RMA: MPI_Accumulate are applied in correct order.
329 - RMA: MPI_Win_{complete/post} shouldn't miss messages anymore.
330 - Better support for MPI_IN_PLACE
331 - Support for MPI_Win attrs and keyvals.
332 - Support MPI_Comm_group_create, MPI_Type_size_x
333 - MPI ops should be properly applied to all allowed types
334 - Replace implementation of MPI_Bcast "scatter followed by rdb allgather"
335 algorithm by a non failing one from mpich.
338 - Kill the fifo data container: we don't use it anymore.
341 - A <cluster> can now be created with different speed values to represent pstates.
342 see examples/platforms/energy_cluster.xml for an example.
344 SimGrid (3.14.159) Released December 28. 2016
346 The previous release was only a crude approximation of a Pi release;
347 this one should be a bit better.
349 - Revert a last minute change that broke on Mac OSX.
350 - Fix the changelog and NEWS files.
351 - Improve the documentation, in particular of the routing module.
352 - Integrate some patches flying around in Debian.
354 SimGrid (3.14) Released December 24. 2016
357 * The whole documentation was reworked and reorganized. There is
358 still much room for improvement here, but we're on it.
361 * We now need python3 for our testing infrastructure.
362 * Model checking needs libevent
365 * Now works on FreeBSD
368 * Privatization now works on FreeBSD
370 * Privatization is now activated by default in smpirun
371 You can turn it off with -no-privatize if something goes wrong.
373 * Call-location tracing for SMPI.
374 You can add the exact location (filename / linenumber) of an MPI call to
375 your trace files and slow down or speed up the simulation between two
376 consecutive calls by using an adjustment file (see the documentation).
378 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
379 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
380 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
381 values were ignored in some cases. The timings of these functions can now
382 be significantly different.
384 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
385 smpi/running-power is renamed to smpi/host-speed
387 * smpi/grow-injected-times option to enable or disable multiplication of the
388 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
389 make simulation less precise (but also much faster).
391 * smpirun script should be (much) faster for large deployments.
393 * SMPI tracing : fixed issue with poor matching of send/receives.
395 * Replay : Fix broken waitall
397 New functions and features
398 * MSG_parallel_task_execute_with_timeout, to timeout computations.
400 Dropped / renamed functions and features
401 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
402 - MSG_mailbox_is_empty() -> Mailbox::empty()
403 - MSG_mailbox_front() -> Mailbox::front()
404 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
405 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
406 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
407 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
408 * Don't pass the free_f parameter to property related functions:
409 - MSG_host_set_property_value()
410 - MSG_as_router_set_property_value()
411 - MSG_storage_set_property_value()
412 * VM properties. Since msg_vm_t are msg_host_t, just use
413 MSG_host_get_property_value() and friends
414 * VM I/O related things:
415 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
416 - Unimplemented save/restore methods
417 * MSG_as_router_get_property_value() was redundent with
418 MSG_environment_as_get_property_value().
419 - Removed MSG_as_router_*propert*() functions
420 - Added MSG_environment_as_set_property_value() for consistency
421 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
422 Well, they are still there for now, but deprecated with a warning.
423 Please switch to xbt_dict_new_homogeneous() before this is removed
425 * Task affinity. Its intended behavior (that was very badly tested
426 and probably not really working) was deceiving what most users
427 would have hoped here.
428 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
429 * Option network/coordinates is now useless and should be dropped.
432 * Added option storage/max_file_descriptors to allow more than 1024 files opened
435 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
436 * Added Dragonfly topology support
439 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
440 EiBps, EBps, Eibps, Ebps for bandwidth.
441 They may become useful to some lucky ones.
444 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
445 * ex/app_tokenring: new example, very similar to the MSG Token Ring
446 * ex/async_waitAll: new example, on asynchronous communications
449 * Memory usage should be decreased for simulations with a large number
450 of processes. This also helps for SMPI.
452 SimGrid (3.13) stable; urgency=low
454 The Half Release, a.k.a. the Zealous Easter Trim.
456 - We removed half of the lines, that were mostly experimental cruft.
457 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
458 (not counting blanks and comments -- according to openhub.net)
459 - The internals are now compiled in C++ (and will soon be clean C++)
460 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
461 59 klines of C, etc. We added only 29 klines of C++ in replacement.
463 * Backwards Compatibility breaks
464 - Removed Lua simulation bindings (switch to C or Java for that).
465 Lua can still be used to describe platforms
466 - Removed Java kernel plug-ins.
467 Will be reintroduced after the ongoing major internals reorg.
469 - the following functions were removed.
470 They were too specific and should be reimplemented in a generic
471 way, with filter function.
472 - MSG_task_listen_from_host
473 - MSG_mailbox_get_count_host_waiting_tasks
474 - MSG_mailbox_put_with_timeout was removed.
475 Please use MSG_task_send_with_timeout instead.
477 - the SD_application_reinit function was removed. It has been a noop for a while.
478 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
479 replaced by a more flexible mechanism.
480 - The following functions thus do not exist anymore
481 - SD_workstation_get_access_mode
482 - SD_workstation_set_access_mode
483 - SD_workstation_get_current_task
484 - Basic estimation functions have been removed but can easily be replaced
485 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
486 - SD_workstation_get_computation_time => amount / sg_host_speed()
488 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
489 Use VM.getSpeed()*load/100 for the legacy behavior.
491 - option enable_tracing was removed. It was not doing anything for a while.
492 - In the ModelChecker:
493 - the model-checker now ptraces the model-checked process which means
494 you cannot use a debugger on the latter anymore (we might make this
495 optional in the feature);
496 - removed soft-dirty page tracking;
497 - remove model-checked side snapshot management,
498 MC_snapshot() and MC_compare_snapshot();
499 - keep the MC_cut() function as a stub (it was not really working
500 in the previous release).
503 * All options are consistently in kebab-case. Old names are kept as alias.
506 * Switch to platform v4 format.
507 - Rename from 'power' to 'speed' the attributes describing the amount of
508 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
509 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
510 - In <host> and <link>, attributes availability and state are gone.
511 It was redundent with state and availability traces, and with peak values.
512 - In <cluster>, attributes availability_file and state_file are gone.
513 It was too complex and unused.
514 - Kill <gpu>. Was not doing anything.
515 - The DOCTYPE points to the right URL:
516 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
517 (the file at this address now documents the changelog since its v1)
518 - A warning is emitted for unit-less values (they are still accepted).
519 - speed. Default: 'f' or 'flops'. Also defined:
520 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
521 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
522 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
523 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
524 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
525 - latency. Default: 's' second. Also defined:
526 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
528 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
529 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
532 * s4u::Host is now the preferred public interface to the Host features.
533 sg_host_* functions are C bindings to the exact same behavior
534 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
537 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
538 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
539 They were deprecated since a few years, and probably did not even compile anymore.
542 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
543 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
544 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
545 encouraged to update their codes. A list of the modified functions can be found at the end of
546 include/simgrid/simdag.h
549 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
550 the lambda or closure passed as a parameter will run in kernel mode.
551 All the callback functions should be rewritten to that interface at some point.
554 * Reorganizing and cleaning the internals all around the place.
557 * Remove old default barrier/bcast buggy algorithms (see #18407)
558 * Various bug fixes to handle more codes
559 * Remove the need for the --foreground option of smpirun (it is still
560 accepted for backward compatibility).
563 * Kill the setset data container: MC don't use it anymore.
564 * Kill the queue data container: it made more sense with GRAS.
565 * Kill the xbt_peer_t data type: it's useless without GRAS.
566 * Kill rm_cb feature of config sets: it was never useful.
567 * Kill graphxml parsing feature. It was not used.
568 * Kill the deprecated code protected by XBT_USE_DEPRECATED
570 - xbt_dynar_sort_strings(), when the content is char*
571 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
572 They throw exceptions on invalid input;
573 * C++ support for declaring CLI flags (simgrid::config::Flag);
574 * class for abstracting different signal backends (simgrid::xbt::signal).
575 with no external dependencies (we need very simple signals).
578 * refactoring and cleanup of the code;
579 * ongoing process to cleanly separate the model-checking algorithms
580 from the code model-checking support.
582 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
584 SimGrid (3.12) stable; urgency=low
586 The Facelift Release.
589 * Require g++ v4.7 at least to not speak prehistorical C++.
590 * Require Boost 1.48 (for signal2 component).
591 * Java must be version 7 at least when activated.
592 * Builds on Windows again (including Java bindings).
593 * Tracing is now always enabled (no way to turn it out)
594 * Remove GTNetS. It was not working anyway.
595 * Various cleanups in the cmake scripts.
596 * Move headers around to sort them out on installed systems:
597 - instr/instr.h -> simgrid/instr.h
598 - instr/jedule/* -> simgrid/jedule
599 - simdag/datatypes.h was removed
600 - simdag/simdag.h -> simgrid/simdag.h
601 - msg/datatypes.h was removed
602 - msg/msg.h -> simgrid/msg.h
605 * Interface improvement:
606 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
607 - Sanitize the interface in MSG_task_ module:
608 - Merge two functions that were close enough but misleading:
609 set_compute_duration(t) -> set_flops_amount(t)
610 get_remaining_computation(t) -> get_flops_amount(t)
611 - set_data_size(t) -> set_bytes_amount(t)
612 get_data_size(t) -> get_bytes_amount(t)
613 - Massive cleanups in the functions related to the energy
614 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
615 - New: MSG_host_get_pstate()
617 - msg/energy/onoff: switching hosts on and off
619 * Interface improvement:
620 - Rename Host.isAvail() to Host.isOn()
621 - Rename Process.currentProcess() to Process.getCurrentProcess()
622 - Rename Task.setDataSize() to Task.setBytesAmount()
623 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
625 - #18874: Actually allows the GC to reclaim tasks
629 - SIMIX_process_throw: raises an exception in a remote process
630 * Refactoring: Separate sub-modules
631 - libsmx: the public interface, as libc in a real system
632 - popping: the strange dance that converts a user request into a kernel handling
633 - smx_context_*: the virtualization mechanisms that embed the user code
634 - smx_*: the handling of each simcalls
636 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
637 * Rename smx_action_t into smx_synchro_t, making explicit that these
638 things are used to synchronize processes with their environment.
639 For example, a communication is a sort of synchronization involving
640 the communicating processes (that may block until the exchange) and
641 the platform. The same can be said from computations, etc.
643 - #18888: segfault when a process ends before its kill_time
647 - 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.
648 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
649 - MPI_Type_set_name, MPI_Type_get_name
650 - MPI_*_c2f and MPI_*_f2c functions
651 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
652 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
653 - Activate a lot of new tests from the mpich 3 testsuite
655 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
656 - InfiniBand network model added : Based on the works of Jerome Vienne
657 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
658 - When smpi/display_timing is set, also display global simulation time and application times
659 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
660 * Collective communications
661 - 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
662 - 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/).
663 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
665 - Replay now uses algorithms from wanted collective selector
666 - Replay can be used with SMP-aware algorithms
667 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
668 - 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.
670 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
671 - Correctly use loopback on fat-tree clusters
672 - Asynchronous small messages shouldn't trigger deadlocks anymore
673 * Energy/DVFS cleanup and improvement
674 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
675 - new: smpi_get_host_pstate()
678 - "Full" network optimization flag was broken since Surf++
679 - Better handling of precision flags in maxmin
680 - Fix bug causing sometimes "Impossible" errors
681 - Properly pass cluster properties to included hosts
682 * Improvement of the Energy plugin.
683 - Always update the consumption before returning that value
684 - New property: watt_off to denote the dissipation when the host is off
685 - New functions getWattMinAt and getWattMaxAt to retrieve the
686 dissipation of pstates that we are not currently at.
687 * Java: class NetworkLink renamed to Link
688 * New function: simcall_process_get_kill_time()
689 * Massive rename s/workstation/host/
690 - That's intrusive, but that's good for the project consistency. Sorry.
691 - Change config option "workstation/model" into "host/model"
694 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
695 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
696 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
697 - xbt_dirname and xbt_basename for non-POSIX systems
699 * The model checker now runs as a separate process.
700 * The model checker runs must now be launched with the new simgrid-mc program.
701 * Record/Replay: the MC can display a textual representation of a path in the
702 execution graph. It can then be replayed outside of the model checker.
704 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
706 SimGrid (3.11) stable; urgency=low
711 * Normalizing pointers addresses tool for better diff between logs
714 * Add cloud examples using new VMs
715 - examples/msg/cloud/two_tasks_vm.tesh
716 - examples/msg/cloud/simple_vm.tesh
717 - examples/java/cloud/cloud.tesh
718 - examples/java/cloud/migration/migration.tesh
719 * Add java surf examples:
720 - examples/java/surfPlugin/surf_plugin.tesh
721 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
722 - examples/java/surfCpuModel/surf_cpu_model.tesh
723 * Add SMPI+MSG example:
724 - examples/smpi/smpi_msg_masterslave/
729 - msg task destroy cancel test
730 - msg_host on/off test
731 * Move all tests in testsuite to teshsuite (adding tesh files)
732 * Restructure teshsuites
733 - one folder for each kind of test
734 * Restructure AddTests.cmake
736 - structure the order of tests (with sections)
739 * Add virtual machine
740 - creation of a VM on a PM
741 - migration of a VM from a PM to another PM
743 - MSG_process_join(msg_process_t process, double timeout)
744 - msg_bar_t MSG_barrier_init(unsigned int count)
745 - int MSG_barrier_wait(msg_bar_t barrier)
746 - void MSG_barrier_destroy(msg_bar_t barrier)
747 - msg_as_t MSG_environment_as_get_by_name(const char * name)
748 * New option "msg/debug_multiple_use" to help debugging when a task is used
751 - Improvements and finalization of MSG_storage, MSG_file APIs and their
753 - Increase code coverage in test suites
757 * Protect context stacks against stack overflow. The number of protected memory
758 pages allocated on the top of each stack (1 by default) can be configured
759 with the new command line option --cfg=contexts/guard_size:<value>.
760 * Simcalls are now generated by a python script that
761 - generates files included by SimGrid
762 - checks that all the functions exist, and proposes prototypes
764 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
765 host_set_data, host_get_data
767 - simcall_process_join(smx_process_t process, double timeout)
768 * Fix bug where sleeping processing could not be suspended.
771 * Translate surf models from C to C++
772 - Generic classes for all models: Model, Resource, Action
773 - A generic interface for each kind of model (CPU, Network, Storage
774 Workstation, WorkstationVM)
776 * Translate surf routings from C to C++
777 * Add callbacks using sigc++ or boost::signals2
778 - Add callback functions for resource creation/destruction
779 - Add callback functions for action state change
780 - Handle Energy as a plugin
781 * Replace swag by boost::intrusive
782 * Add new routing models for clusters. For documentation, see
783 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
784 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
785 parameters for cluster tag
786 - Fat trees, with topology="FAT_TREE" and
787 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
788 - see examples/platforms/torus_cluster.xml and
789 examples/platforms/fat_tree_cluster.xml
793 * Hostfiles support host:nb_processes construct to deploy several processes on
795 * Collective communication algorithms should not crash if used with
796 improper number of nodes and report the error.
797 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
798 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
799 MPI_Dims_create, MPI_Cart_sub are supported.
800 * New interface to use SMPI programmatically (still depends on MSG for
801 some parts, see examples/smpi/smpi_msg_masterslave) :
802 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
806 * Global variables privatization in MPI executables is now performed at runtime
807 with the option smpi/privatize_global_variables (default:no).
808 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
809 dynamic libraries loaded by the application are not privatized (static
810 linking with these libraries is advised in this case)
813 * Options defined in XML work correctly now.
816 * New cmake option, enable_lib_in_jar, to control whether native libraries are
817 copied into simgrid.jar or not (ON by default). Use this option if you want
818 to reduce the size of the installed simgrid.jar, *and* the native libraries
819 are kept installed elsewhere.
820 * Surf binding with SWIG (code generated in maintainer mode only):
821 - plugin to handle callbacks
822 - CPU model only for the moment
825 * Supernovae build mode is definitively removed. It was used to improve
826 inlining and inter-module optimizations. It is nowadays superseded by
827 link-time optimizations commonly available in compilers.
828 * Update ns-3 find lib. Bindings for ns-3 should work again now.
829 * Add boost dependency for surf++
830 * Add new macro for tests
831 - ADD_TESH(name <tesh_args>)
832 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
836 - xbt_bar_t XBT_barrier_init(unsigned int count)
837 - int XBT_barrier_wait(xbt_bar_t barrier)
838 - void XBT_barrier_destroy(xbt_bar_t barrier)
839 * Make the xbt_os_time module public
841 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
843 SimGrid (3.10) stable; urgency=low
845 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
848 * Reintegrate Java to the main archive as desynchronizing these
849 package is not acceptable anymore (Java is now considered stable)
850 * Add explicit synchronization facilities through semaphores
851 * Bug fix: Task.setDataSize() only changed the C world, not the value
852 cached in the Java world
855 * Dramatically change the way files are handled. API and internals changed, but
856 this part of MSG was not considered as production grade either.
857 * Add explicit synchronization facilities through semaphores
858 * Add a new function MSG_host_get_process_list()
859 * Preliminary DVFS support (see examples/msg/energy/ for details)
862 * SMPI is now included directly in the libsimgrid as the windows
863 linker doesn't force us on splitting it anymore.
864 * Improvements of the SMPI replay tool:
865 - Most of the collective communications are now rooted in the same process as
866 in the original application.
867 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
868 used until now). Multiple data types can now be used in a trace.
869 - The replay tool now supports traces produce either by TAU or a modified
871 - Bug Fix: the compute part of the reduce action is now taken into account.
872 - Gatherv collective is now supported
873 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
874 of smpirun outputs time independent traces for the current run. One file
875 is created per process. If too many processes are simulated, this behavior
876 can be changed to one file for all processes by using the
877 tracing/smpi/format/ti_one_file flag
878 * smpirun generates the host file if needed (with given host count and platform)
879 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
880 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
881 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
882 --cfg=smpi/coll_selector:(mpich/ompi)
883 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
885 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
886 happening inside a collective SMPI call.
887 * Fix the behavior of complex data types handling.
888 * Make MPI_Wtime another synchronization point to take computations into
890 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
891 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
892 * Add all missing Fortran bindings, SMPI should work with Fortran 90
893 (no privatization of global variables yet)
894 * Preliminary DVFS support (see examples/smpi/energy/ for details)
897 * Verification of liveness properties is now available for SMPI applications
898 (in addition to MSG applications)
899 * Bugged examples using SMPI in examples/smpi/mc/
900 * Add --cfg=model-check/visited option. Allows the verification of infinite
901 programs. Detection of loops in the execution thanks to the system state
902 comparison and reduction of the state space to explore. Can be combined with
903 DPOR for safety properties.
906 * Allow to change SimGrid configuration (see --help) within the code
907 thanks to SD_config() as it can be done in MSG.
908 * Add a new function SD_task_set_amount() upon user request.
911 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
912 * Remove rule based routing (no more PCRE dependency)
913 * Add a limiter_link option to cluster tag, to specify a maximum reachable
914 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
915 * Add a loopback_bw and loopback_lat options to cluster tag.
916 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
917 routing. Now peers are encapsulated in an AS and have their own private
918 router but this is transparent.
921 * Our own implementation of getline is renamed xbt_getline, and gets
922 used even if the OS provide a getline(). This should reduce the
923 configuration complexity by using the same code on all platforms.
924 * New type: xbt_cfg_elm_boolean.
925 * Allow to use yes/no for boolean configuration options in the command line.
926 * Allow to disable SimGrid cleanups at exit from command line option.
927 There are situations where one may want a simulation to end with an exit.
928 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
929 annoying when scripting around the simulator. Adding a
930 --cfg=clean_atexit:no allows to circumvent this issue.
933 * Lots of memory leaks were corrected in this release.
934 There is no known memory leaks anymore, in all of our 600+ tests.
935 * New command line option --version, to get SimGrid version information.
936 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
938 * Supernovae builds are deprecated, and expected to be removed in the next
941 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
943 SimGrid (3.9) stable; urgency=low
945 The Grasgory release: GRAS is really dead now.
947 * Complete overhaul of the internal host structures scheme.
950 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
951 was considered as experimental and badly maintained since then.
952 * Keeping it was thus a trap to our potential users, that could take
953 it instead of MSG or SMPI by mistake despite is pity state.
954 * GRAS seems to have very few users (if any), and no one volunteered
955 to maintain it further. It also induces a lot of XBT code (for
956 portability sake), that must be maintained too.
957 * For all these reasons, we killed GRAS. If someone wants to revive it
958 in the future, don't cry, our git history still remembers of GRAS.
961 * Major overhaul. Merge our documentation again as time proved that
962 splitting it was really not helping our users.
963 * Further improve the developer documentation to help newcomers
964 hacking on SimGrid itself. The user documentation (and in
965 particular, the beginner documentation) is still in a sorry state.
968 * Now works on Windows too!
969 * Much more extensive test suite, from MPICH
972 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
973 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
974 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
975 * Bug fix: let task be scheduled when the last dependency to be solved is
976 a control dependency.
977 * Remove SD_load_environment_script function.
978 Use the C sg_platf function if you want to declare a platform
982 * New function: MSG_process_get_number()
983 * Old function documented: MSG_config()
984 * Remove MSG_load_platform_script function
985 Use the C sg_platf function if you want to declare a platform
989 * Change the default value of the TCP_gamma constant (maximal size of TCP
990 congestion window) to a more realistic 4MiB value. If you notice changes in
991 your simulation results, you can fall back to the previous 20k tiny window
992 by adding --cfg=network/TCP_gamma:20000 on command line.
993 * (Hopefully) fix a bug wrt periodic availability and state traces
994 * Bug fix: use default values at start when first event in availability/state
995 trace is not at time 0.
998 * remove the "new_" part of function name sg_platf_new_trace_connect
999 (resulting in sg_platf_trace_connect), since it does not create
1003 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
1004 use case where it's really mandatory, and maintaining it was a pain in
1006 * New: xbt_fifo_search(), search an item with a user-provided
1007 comparison function instead of dumb pointer comparison.
1010 * Fix the lua deployment:
1011 Use `simgrid.init_application()` before deployment instead of
1012 `simgrid.msg_register_application()` after.
1015 * Transfer the tracing files into the corresponding modules.
1017 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1019 SimGrid (3.8.1) stable; urgency=low
1021 The "we are told that some people want to also *install* the simgrid
1024 * Add missing file "tesh.1" to the archive.
1026 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1028 SimGrid (3.8) stable; urgency=low
1030 The Psssshiiiit release: SimGrid jumps into the Cloud.
1033 * Add an experimental interface to manipulate VMs. They are mainly
1034 process groups with very few intrinsic semantic, but they should
1035 allow you to build the semantic you want easily.
1036 * New function: MSG_host_set_property_value()
1037 * New function: MSG_process_on_exit(). To clean memory in all cases.
1038 * Bug fixes that made the host (and link) failures unusable.
1039 * Add a way to auto-restart process when the host in which they are
1040 executing comes back (ON_FAILURE="RESTART" on deployment file,
1041 MSG_process_auto_restart_set).
1042 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
1043 please stop using the old ones, they are DEPRECATED.
1045 * Deprecate functions MSG_global_init() / MSG_global_init_args()
1046 Please use MSG_init() instead. (reducing the amount of entry
1047 points in the library helps us).
1048 * Make it impossible to link against the wrong version of the lib
1049 * Deprecate MSG_clean(). No need to call it anymore.
1050 * Function MSG_get_host_number() is not deprecated anymore.
1053 * Split the doc into a user guide and a reference guide.
1054 * Start a developper guide to help people hacking on SimGrid.
1057 * Enable tracing by default. This modules rocks you should use it.
1058 * Remove option custom_flags. Now use environment variables CFLAGS
1060 * Use default cmake things to detect lua instead of home grown ones.
1061 * New option "enable_mallocators" to disable mallocators, for debugging
1062 purpose ("on" by default).
1065 * Bug fixes around the resource failures: don't let the processes
1066 survive the host they are running onto.
1067 * Add an interface to auto-restart processes when the host in which they are
1068 executing comes back.
1069 * Ensures that SIMIX_clean is called automatically. It's not part of
1070 the public interface anymore (bindings should be updated).
1073 * Bug fix for when SD_Simulate is called with a positive value: be careful
1074 when comparing doubles. Sometimes they are different for non significant
1076 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
1077 parallel task whose initial work is distributed among host according
1078 to the Amdahl's law. Such tasks are created with a parameter alpha
1079 that corresponds to the non-parallelizable part of the computation.
1080 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
1081 between two sets of workstations assuming a 1D block distribution (each
1082 workstation owns a similar share of data) on both sides.
1084 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
1085 Data redistribution will be automatically scheduled once parent and child
1086 are both scheduled. The filling of computation_amount and
1087 communication_amount structures is now done seamlessly thanks to the chosen
1089 * New function SD_workstation_dump to display various information
1090 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
1091 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
1092 bandwidth on the route onto which the task is scheduled and the amount of
1094 To divide the nominal bandwidth by 2, the rate then has to be :
1095 rate = bandwidth/(2*amount)
1096 * Compute tasks that have failed can now be rescheduled and executed again
1097 (from their beginning)
1098 * Increasing source code coverage (src/simdag is now covered at 95.8%
1102 * Re-implement time-independent trace replay using SMPI (at the
1103 smpi_smp_* level) instead of MSG. This should replace
1104 examples/msg/actions/actions.c
1105 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
1106 hindexed and structs)
1107 * Implement the exchange of non-contiguous data.
1108 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
1109 * Correct behavior of smpi/sender_gap and set its default value to 0
1110 * Add option to asynchronously send small messages to allow better
1111 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
1112 specifies the size in bytes under which messages will be asynchronously sent.
1113 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
1114 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
1117 * New C interface to define a platform: XML is now optional.
1118 For more info, please check include/simgrid/platf.h.
1119 * New interface to define random platforms from the C:
1120 For more info, please check include/simgrid/platf_generator.h and
1121 examples/msg/masterslave/masterslave_platfgen.c
1122 * Export a sg_cmdline dynar containing all the arguments we got from
1126 * Two new tracing options for adding comments to trace file so you
1127 can track your experiments (see --help-tracing for details).
1128 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
1129 * Adding the SimGrid version that generated the trace file as a comment.
1130 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
1131 * Fix to avoid key clashes on Paje links
1132 * Other minor fixes related to the Paje specification
1135 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
1136 and renamed to xbt_str_hash() and xbt_str_hash_ext().
1137 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
1138 * Greatly improve the robustness of mmalloc to user errors (such as
1139 using an area after freeing it, or freeing it twice)
1141 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1143 SimGrid-java (3.8.1) stable; urgency=low
1145 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
1146 Warning: all methods are visible, but only some of them are
1147 implemented so far. Check the source (src/jtrace.c)
1148 for further information.
1149 * New module: org.simgrid.msg.File (SimGrid File management functions)
1150 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
1151 * Change the meaning of Process.restart: now restart the process from
1152 the begining, like MSG_process_restart in C.
1153 * Add Process.setAutoRestart: handling of process restart when failed
1155 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
1156 you to retrieve the properties of the processes/hosts
1157 * Deprecate Msg.clean(): you can just forget about it now.
1158 * New function Process.getCount(), that only works when compiling
1159 with the not yet released version 3.9 of the C library.
1161 * New context factory based on Coroutines. It mandates a modified JVM
1162 but then, the simulations run about five times faster, and there is
1163 no limit to the amount of processes (beside of the available memory).
1165 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1167 SimGrid (3.7.1) stable; urgency=low
1170 * Restore the prototype of MSG_process_create_with_environment() to
1171 the pre-3.7 situation by removing the kill_time argument.
1172 * Add a MSG_process_set_kill_time() function instead.
1175 * Fix weird behaviors when dealing with parallel tasks.
1178 * Simgrid is now built as a dll.
1179 * Simgrid-java now works on Windows.
1180 * Simgrid-Java is now included into Windows package.
1183 * First pre-build package for MacOSX.
1186 * Fix compilation when using MSG_USE_DEPRECATED.
1187 * Fix some compilation issues on Macs and Windows.
1188 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
1189 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
1191 We discovered that the Lua console is broken, but we are missing the
1192 manpower to fix it right now. The problem existed in 3.7 too, so we
1193 are not blocking the release for that. Sorry if you depended on this
1194 feature, any help would be really welcome.
1196 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1199 SimGrid-java (3.7.1) stable; urgency=low
1201 The "Java aint got to be bloated and slow" release
1204 * Various internal cleanups and performance improvement
1205 Simulations are expected to run up to twice faster or so
1206 * Make Process.kill(process) an instance method, not a static one
1207 * User processes are not java.lang.Thread subclasses.
1208 This breaks the compatibility (sorry), but previous API was
1209 brain-dead, making it impossible to have non-trivial
1210 initializations in the process constructor.
1211 * Require a full constructor per Process sub-class.
1212 Kinda breaks the compatibility (sorry), but this allows a much more
1213 efficient way to launch the processes at simulation startup.
1214 * Do not embeed our version of semaphores, java 1.5 can be considered
1215 as sufficiently prevalent for us to not dupplicate its features.
1219 * Add examples for almost every part of the API
1220 We spotted and fixed a lot of bugs in the process
1222 * New module: asynchronous communication API
1223 * New function: Process.sleep()
1224 It takes milliseconds as argument, just as java.lang.Thread.sleep()
1225 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
1226 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
1228 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1230 SimGrid (3.7) stable; urgency=low
1232 The "spring cleanups (before the next Big Project kicks in)" release.
1235 * We can specify the SMPI latency/bandwidth factor with command line
1236 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
1237 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
1238 You can also use the "config tag" from platform file by adding this line
1239 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
1240 (see "example/platforms/tag_config.xml" to use "config tag").
1241 Note that the command line supersedes the platform file configuration.
1242 * Change the correction factors used in LMM model, according to
1243 the latest experiments described in INRIA RR-7821.
1244 Accuracy should be improved this way.
1245 * Use the partial invalidation optimization by default for the
1246 network too. Should produce the exact same results, only faster.
1247 * Major cleanup in surf to merge models and split some optimization
1248 mechanisms from the core of the models. As a result you can now
1249 specify which model to use (e.g., --cfg=network/model:LV08
1250 --cfg=cpu/model:Cas01) and which optimization mode to use
1251 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
1252 Incompatible combinations should err at initialization. See
1253 --help-models for the list of all models and optimization modes.
1254 * The CLM03 workstation model was dropped for simplicity because it
1255 used the deprecated CM02 network model. Use default instead.
1256 * Rename the TCP_gamma configuration option to network/TCP_gamma
1257 * Rename the coordinates configuration option to
1258 network/coordinates, and document it
1259 * Use now crosstraffic keyword instead of the terribly misleading
1260 fullduplex keyword. It is activated by default now in the current
1261 default model, use --cfg=network/crosstraffic:0 to turn it off.
1262 * Ongoing refactoring the model parsing to make XML files optional
1263 See include/simgrid/platf.h for details (still to be completed)
1266 * Major overhaul of the documentation. Almost instructive now :/
1267 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
1268 functions and friends. This interface was considered as
1269 deprecated since over 2 years, it's time to inform our users that it is.
1270 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
1271 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
1272 use these (crappy) functions in your code.
1273 These functions will be removed soon. Stop using them now.
1274 * Deprecate MSG_get_host_{table,number}
1275 Implement MSG_hosts_as_dynar() instead.
1276 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
1277 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
1280 * Stabilize the parallel execution mode of user contexts
1281 * Introduce configuration variables to control parallel execution:
1282 - contexts/synchro: Synchronization mode to use when running
1283 contexts in parallel (either futex, posix or busy_wait)
1284 - contexts/parallel_threshold: Minimal number of user contexts
1285 that must be part of a scheduling round to switch to parallel
1286 execution mode (raw contexts only)
1287 * Fix bugs that prevented to use suspend/resume along with
1288 synchronization structures.
1289 * Fix bugs in process termination that lead to invalid memory access
1290 in very specific conditions.
1293 * Introduce a parallel mode for the models (controlled by surf/nthreads
1294 configuration item). In our tests, running the models in parallel
1295 never lead to any speedups because they are so fast that the gain
1296 of computing each model in parallel does not amortizes the
1297 synchronization costs, even when ultra fast futexes are used.
1298 This is released anyway because YMMV.
1301 * Performance boost by using a swag internally to compute the set of
1302 tasks that are finished and should constitute the return value of
1306 * Enable it by default now that it is considered rather stable.
1309 * Documentation of the tracing functions.
1310 * Performance gains when tracing categorized/uncategorized resource
1311 utilization by avoiding calls to get route when updating resource
1312 variables. LMM constraints are being used instead.
1313 * API changed to set task categories. Use MSG_task_set_category instead
1314 of TRACE_msg_set_task_category, and SD_task_set_category instead
1315 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
1316 * Bugfix for graphicator, routes not correctly obtained, memory leaks
1317 * Examples for link user variables added (see at examples/msg/tracing/)
1318 * Deprecated function TRACE_msg_set_process_category completely removed
1319 * Trace header updated according to the latest Paje file format
1320 * Tracing network lazy updates, no longer obligate users to use full updates
1321 * --cfg=tracing/platform:1 also registers power/bandwidth variables
1322 * Experimental: let user code declare/set/push/pop application states for hosts
1323 * API changed to allow the manual creation of graph configuration files
1324 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
1327 * Improve the API of Lua MSG bindings, using the Lua spirit.
1328 * Each simulated process now lives in its own Lua world (globals are
1329 automatically duplicated). It helps writing simulators. Will allow
1330 to run Splay programs within SimGrid in the future.
1331 * Add a Chord example in Lua, equivalent to the MSG one.
1334 * Start the implementation of a solution to express temporal
1335 properties, not only local assertions. This is still an
1336 experimental work in progress, stay clear from it to be safe.
1340 - Add new runtime parameters --help-logs and --help-log-categories
1341 to display information about supported logging parameters and
1343 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1344 don't exists anymore.
1345 * Mallocators: allow value NULL for the reset function.
1347 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1348 create homogeneous dictionaries, where all the elements share the
1349 same free function. Non homogeneous dictionaries will be
1350 deprecated in the next release.
1351 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1352 - Multi-level dictionaries are deprecated.
1354 - new function xbt_dynar_search_or_negative() that is useful when
1355 you have less than 2 million elements in your dynar and don't
1356 want of the extra complexity of catching exceptions when the
1357 element is not found.
1359 - Make xbt_os_thread module (for thread portability) public.
1360 Documentation is still to come, sorry.
1362 - Cleanups and simplifications to make it maintainable again.
1363 - Exotic features (such as memalign and valloc) were removed.
1364 - The metadata were extended and improved so that the
1365 model-checker becomes able to explore and inspect the heaps.
1366 - This may induce a performance drop when enable_model-checking is
1367 ON in cmake (even if it's not used in the simulation), but it is
1368 necessary at this point to get MC working.
1370 Turn model-checking OFF if simulation performance matters to you.
1371 Not enabling it at runtime is not enough, disable it in cmake.
1373 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1376 SimGrid (3.6.2) stable; urgency=low
1378 The "Not coding new stuff allows to polish old things" release.
1381 * New bindings to the NS3 packet level simulator (experimental)
1382 * Use the raw (efficient) execution contexts instead of the sysv
1383 (portable) ones when possible.
1384 * libpcre is now mandatory in any cases since not using it led to
1385 severe performance loss and possibly other issues
1386 * Update the XML platforms:
1387 - G5K: include the latest machine in Nancy
1388 - GridPP and LCG: new platforms
1389 * Documentation was partially updated, at least (more to come)
1391 Bug fixes, cosmetics and small improvements
1392 * Free terminated processes before the end of the simulation to avoid
1393 exhausting the memory of users having very dynamic amount of
1395 * Bug fix and cosmetics about canceling non-running tasks
1396 * Bug fix about the dot loader's issues when using libcgraph
1399 * Create an installer for windows with nsis (amd64 and win32)
1400 - Add an hello world project to illustrate simgrid project creation.
1401 - Embed libpcre into the Simgrid installer to avoid
1402 its compilation burden
1403 * The raw execution contexts should work on Apple now
1404 * Port to Windows 64 bits
1405 - Sysv contexts now have an implementation for this arch
1406 - GRAS communication features now support this arch
1407 * Drop support for borland compiler on windows
1408 - this code was not maintained, and we kinda depend on gcc nowadays
1409 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1410 * Fix portability issue on unstable ubuntu: linker became picky on
1413 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1415 SimGrid (3.6.1) stable; urgency=low
1417 The "Oops, we broke Macs too" release
1420 * Fixed contexts detection so that raw ones are used when possible
1421 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1422 leads to a strange error, with user code segfaulting sometimes when
1423 the generated code is not perfectly aligned (which is not
1424 controllable from the user side, depends on the amount of code)
1427 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1429 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1431 SimGrid-java (3.6) unstable; urgency=low
1434 * Split of every thing from simgrid v3.5 into a separate package.
1436 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1438 SimGrid (3.6) stable; urgency=medium
1440 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1443 * Bindings now constitute their own package, separated from the main one.
1444 Rationale: reduce our maintenance nightmare by reducing the module coupling
1445 They will soon be released on their own on gforge.
1447 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1448 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1450 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1451 * It's not quite deprecated for now because we have no replacement,
1452 but it may soon become the case.
1455 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1456 * Fortran: New user-level cache variable to store the rank of the running
1457 process. This improves performance by an order of magnitude.
1458 * C: New coccinelle script to automatically locate and modify global and
1459 local static variables.
1460 * Improved SMPI network model with a sender-side gap to account for multiple
1464 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1465 only say if a communication is finished, no matter whether it succeeded or
1466 failed. You can call MSG_comm_get_status() to know the status of a finished
1468 * New function MSG_task_dsend() to send a task and detach it. When a
1469 communication is detached, you are never notified of its success or failure
1470 and the memory is released automatically once it is finished. This function
1471 is useful when you don't care about the end nor the success of a
1473 * Change the prototypes of action replay. Sorry for inconvenience,
1474 but this is really more efficient this way (and to adapt your code,
1475 you just have to fix the initialization, that shouldn't be too long)
1476 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1477 ever managed to do anything useful with it.
1478 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1479 takes a dynar of communications. It returns immediately and gives the
1480 index of a finished communication (if any).
1481 * New example: a basic implementation of the Chord P2P algorithm.
1484 * New model for multi-core CPUs. You can now use the core attribute to
1485 precise the number of cores of a host. This is a basic model. Every
1486 process running on the host receives at most the power provided in
1487 the DTD (throughput<=power). Total throughput of process cannot exceed
1489 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1490 router linked by an up-link and a down-link (possibly asymmetrical).
1491 This kind of pattern allows to easily build last-mile model style platforms.
1492 Aggregating such patterns in a rule-based AS is thus the technique of
1493 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1494 * New model for Vivaldi routing. We transformed the Vivaldi network model
1495 into a Vivaldi routing model (based on the rule-based model). This allows to
1496 combine Vivaldi based latencies with last-mile platforms.
1499 * Added a check for NaN of IEEE754 infinite in the double entries of
1501 * Introduce a new context factory "raw", highly inspirated from the
1502 ucontext factory, but using manually crafted functions in assembly to
1503 do the work in an efficient manner.
1504 * Allow to change the used context factory at run time, not only at
1505 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1506 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1507 contexts at runtime (only with raw contexts or ucontexts).
1508 * Completely rewrote this module to allow parallel execution of user
1509 processes. Use --cfg=contexts/nthreads:N to execute user processes
1510 with N parallel threads (the default is 1, meaning no parallelism).
1511 * Allow to decide dynamically between sequential and parallel modes.
1512 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1513 processes in parallel only when their number is greater than or equal to P
1515 * Added a check for NaN of IEEE754 infinite in the double entries of
1519 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1520 won't output the state of processes when interrupted with Ctrl-C
1521 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1522 NULL-terminated array. This may solve backward compatibility issues
1523 due to the change to return type of SD_simulate. See also:
1524 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1525 * Add new macros with variable number of arguments.
1526 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1527 - in xbt/asserts.h: xbt_assert
1528 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1529 - in xbt/ex.h: THROWF and RETHROWF.
1530 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1532 * Change xbt_die() to accept a format string with arguments, just like printf.
1533 * New data structure: xbt_lib_t, like a dict but more general and with better
1537 * New configuration options
1538 Options triva/categorized and triva/uncategorized can be used to generate
1539 graph configuration files for Triva visualization tool.
1540 * Configuration option tracing/platform is renamed to tracing/categorized
1541 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1542 * New TRACE_user_link_variable interface:
1543 User provides the name of the link and the tracing variable to attach to it
1544 * the declaration of tracing categories must be done after the environment
1546 * simpler tracing interface, just one way to declare categories
1547 TRACE_category or TRACE_category_with_color, it is up to you
1548 * links in the trace file are again identified by their names
1549 * trace contains the full platform hierarchy exactly as declared using the ASes
1550 * Options tracing/msg/[task|process]:1 groups the process by hosts
1551 for both cases, tasks and processes must have names that are unique during the simulation
1552 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1553 * The experimental option tracing/msg/volume is deprecated
1554 its functionality may be reincorporated if needed
1556 The tracing generates a trace file with unordered timestamped events,
1557 because of the way the core simulator (surf) works. A script available
1558 at the tools directory (fix-paje-trace.sh) can be used to put the events
1559 in order. We have changed the tracing so it can generate ordered timestamped
1560 events in the final trace, but depending on the simulator (and how much time
1561 is simulated) that can lead to a huge memory utilization. It is deactivated
1562 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1564 Build Infrastructure
1565 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1566 - We are trying hard to keep the API stable, but it may happen that
1567 some things change (we're a research project after all, not a
1568 nuclear plant operating system). If such things should happen, you
1569 could rely on that macro to adapt.
1570 - current value: 30600 for 3.06.00, aka 3.6
1571 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1572 a number that can be compared with SIMGRID_VERSION.
1573 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1574 assertions and verbose/debug logging events are disabled at compile time.
1576 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1578 SimGrid (3.5) stable; urgency=medium
1581 * New feature to any SimGrid-based simulator: Model-Checking
1582 Check SIN#1 for more details.
1585 * New Model SMPI (three-interval linear regression for correction factors)
1586 See RR-7426, available at http://hal.inria.fr/inria-00527150
1587 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1588 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1589 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1590 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1591 * See also src/smpi/README
1596 - Tracing API changes: TRACE_start and TRACE_end should not be called
1597 by user-code. They are automatically called by simulators created
1598 with SimDAG, MSG and SMPI if the toolkit is compiled with
1599 tracing_enabled=ON. Categories declaration and utilization remain the
1600 same for MSG and SimDag.
1601 - A function was added to the tracing API to declare categories with
1603 - TRACE_category_with_color (char *category, char *color)
1604 where color must be in the following format
1605 "%f %f %f", red, green, blue
1606 and red, green, blue are float values in the interval [0, 1]
1607 - User can specify NULL as color parameter, or continue calling
1608 TRACE_category (cat)
1609 On that case, the tracing system will define random colors
1610 - The following command-line options are supported:
1611 --cfg=tracing/filename:msg.trace
1612 --cfg=tracing:1 (activate tracing, needed to use others)
1613 --cfg=tracing/platform:1 (categorized resource use)
1614 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1615 --cfg=tracing/msg/task:1 (task creation)
1616 --cfg=tracing/msg/process:1 (process creation, migration)
1617 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1618 --cfg=tracing/smpi:1 (SMPI interface tracing)
1619 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1620 - examples of examples/msg/tracing updated
1622 - DAXLoader and DOTLoader functions can generate tasks with categories
1623 - A new function to attribute a category to SD tasks:
1624 TRACE_sd_set_task_category (SD_task_t task, char *category)
1625 * Tracing the MPI interface implemented by SMPI
1626 - Collective operations are traced with states
1627 - Point-to-Point operations are traced with states/links
1628 - Tracing activated by a parameter "-trace filename" passed
1629 to smpirun during execution (considering that simgrid
1630 is compiled with tracing enabled)
1631 - To run the simulation with gdb, the simulator
1632 accepts --cfg=tracing/smpi:1 to trace SMPI
1633 - tesh files to check if smpi tracing is ok
1634 - See examples/smpi/NAS/DT-trace
1635 * GTNetS tracing re-worked
1636 - adaptation to the tracing system of GTNets to cope
1637 with modifications regarding the fullduplex mode
1638 - new tesh files to check if gtnets tracing is ok
1641 * Asynchronous communications through the functions:
1642 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1643 * New function: MSG_load_platform_script()
1644 to make possible using a lua script instead of XML files to set up platforms
1645 * New function: MSG_set_function
1646 to associate functions to processes, used when bypassing the parser
1647 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1649 Platforms: Add some more examples in examples/platforms
1650 * Grid'5000: see www.grid5000.fr
1651 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1654 * Change the XML format. This is a very important modification. SimGrid
1655 3.5 introduces a new hierarchical format based on the notion of
1656 Autonomous Systems. Compatibility with old format is ensured
1657 through the perl script provided in the install bin directory
1658 bin/simgrid_update_xml.
1659 It is now possible to build platforms with specific routing
1660 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1661 connect several platforms together. We will try to provide soon
1662 set of realistic platforms exploiting these properties (have a
1663 look at examples/platforms/ for the moment).
1664 * Take the opportunity of the XML format change to be a good XML citizen:
1665 rename link:ctn to link_ctn and similar changes (also dealt with by
1667 * Add a new routing scheme (rule-based) using regular expressions. It
1668 enables to have an extremely low memory footprint when the
1669 underlying routing is simple and can be compactly described. You
1670 need to have libpcre4-dev (perl regular expressions) installed if
1671 you want to use this routing scheme.
1672 * Revive the cluster TAG and allow to easily and efficiently (both in
1673 term of memory and speed) connect clusters together. Have a look
1674 at teshsuite/simdag/platforms/ to see how this can be done. With
1675 this tag, you can create clusters with thousands of tasks at no
1676 cost (have a look at examples/platforms/).
1677 Note: clusters are implemented as ASes, so there is no need for an
1678 enclosing AS tag if you have only one cluster in your platform.
1679 * Add new generic functions in the public interface that allows the user
1680 to call SURF 'create_resource' methods from your code (same
1681 functionality as the XML bypass mechanism but with a much lighter
1683 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1684 that uses a piecewise linear approximation to produce better
1685 results when exchanging small messages.
1686 * Add a new parameter to handle correctly full duplex link and account
1687 for interferences between uplink and downlink communications
1688 (activate with --cfg=fullduplex:1).
1691 * Rename the SD_READY (all dependencies are satisfied and task is
1692 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1693 dependencies are satisfied) state.
1694 This prevents a confusion between the notion of "ready to schedule"
1695 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1696 simulated" (SD_RUNNABLE) used by the simulation kernel.
1697 * Change the way a task is considered as ready. Instead of removing
1698 dependencies when a task is done, a counter is decreased. This way,
1699 it is always possible to reach ancestors thanks to the
1700 SD_taks_get_parents function (even after the end of the simulation.)
1701 * Change the return type of SD_Simulate from (SD_task_t*) into
1702 xbt_dynar_t. This function was in handling a dynar internally and
1703 converted it into a NULL terminated array for historical reasons.
1704 * New function SD_dotload(char*) to load a DAG described in dot
1705 format. This loader and the corresponding examples require the
1706 installation of the graphviz library.
1707 * Fix a bug in the management of tasks of size 0 in the surf network
1708 models. This problem was only visible with SIMDAG and you should
1709 thus disregard results produced with earlier versions if you
1710 relied on this feature (some tasks were blocked because of this).
1711 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1712 even though your applications were doing only point-to-point
1713 communications and sequential computations. Now you can really use any
1714 model you want (of course, if you create real parallel tasks, which are
1715 not implemented in most models beside ptaskL07, this will abort).
1716 * Add an example that schedules a DAX on an heterogeneous platform
1717 using a Min-Min strategy.
1718 * New function SD_workstation_get_current_task() that returns the kind
1719 of task currently running on a workstation in the sequential access
1721 * Raise some warnings when unexecuted tasks remains at the end of the
1722 simulation. This is usually caused by cycles in the DAG.
1725 * New function: SIMIX_process_set_function() called by MSG_set_function
1726 * Change the underlying waiting queue in semaphores so that a process
1727 can wait on several of them simultaneously (as in waitany).
1728 * Fix the way to handle tokens in semaphores so that all access patterns
1729 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1730 * kill the dirty pimple SIMIX_message_sizes_output()
1731 Please use (proper) visualization instead
1734 * New data container: setset (set of sets of elements)
1735 * New module: mmalloc (mapped malloc, allowing to have several
1736 independent segments of malloc)
1737 * New function: xbt_dict_cursor_set_data()
1738 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1739 * New function: xbt_dynar_is_empty()
1740 * New function: xbt_fifo_get_last_item()
1741 * Fix xbt_dynar_shrink(): use the right element size.
1742 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1743 uninitialized areas during expand.
1744 * Fix semaphores: previous implementation was severely broken.
1745 * Use library init/fini functions for our initialization.
1746 - you can use logs and other feature as soon as you want in your
1747 code (even before the xbt_init / MSG_init)
1748 - xbt_exit is now a no-op and produce a warning when used.
1751 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1752 benefit from the latest improvement to the simulation kernel.
1753 * Kill measurement sockets for now. If you rely on them, sorry. This
1754 release is not for you. This feature will be reintroduced in the
1755 future, but we cannot delay the release any further.
1756 * New function: gras_msgtype_get_name().
1757 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1758 * Fix (at last) the pmm example: it should not randomly fail anymore.
1760 Build chain: bug fixes and overall polishing
1761 * Cmake is now stable enough. Hence, we killed the autotools.
1762 * Port to windows ( TM :)
1763 * Fix the 'make install' target.
1764 No need to use 'make install-simgrid' anymore
1765 * Introduce a 'make dist' target compiling a *source* archive
1766 'make package' compiles a binary archive
1767 * Compile java files only on need
1768 * Add --cd and --setenv command line options to tesh
1769 * Out of source builds are not fully supported yet, but we are close
1770 * Enable supernovae and optimization flags by default for our users
1773 * Add layer to set up environment directly from lua, without XML.
1774 * The effect of gras_stub_generator can be achieved through
1775 lua too (check examples/gras/console/ping_generator.lua)
1777 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1779 SimGrid (3.4.1) stable; urgency=low
1781 The "Polishing easter eggs is probably a good idea" release.
1782 This is a bug fixes release only.
1786 * Fix a bug preventing the tasks from begin garbage collected.
1789 * Fix a bug occuring when a host involved in a communication fails.
1790 This was not detected properly by the other peer involved in the
1791 communication. Now, it's reported as a network error.
1794 * Warn the user about loop dependencies in data flow of DAX files
1795 * Obey the control-flow dependencies of DAX files
1798 * Add option "enable_smpi" allowing to not compile SMPI.
1799 Probably useful for the (Mac) users experiencing a build error here
1800 * Improve the detection of lua5.1 and ruby1.8
1802 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1804 SimGrid (3.4) stable; urgency=low
1806 The "Easter in Cargese" release. Also known as (major changes):
1808 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1809 ~> bindings were greatly improved
1810 ~> new tracing infrastructure for better visualization introduced
1812 * the "Welcome to configury modernity" release.
1813 ~> we switched from autotools to cmake, and improved our cdash
1816 A more detailled list of changes follow (full detail in svn log).
1818 Java Bindings: Various Cleanups
1819 * (install java-gcj-compat-dev on debian-like to use them)
1820 * Remove put/get: no need to export deprecated interface in Java
1821 Use send/receive instead.
1822 * Cleanup the examples and add a README per directory
1823 * Remove example autoDestination (that's the only way to go now)
1824 * Remove example explicitDestination (was a plain copy of basic)
1825 * Make JniException a runtime exception, so that there is no need to
1826 declare the fact that you may encounter such a beast. I guess that
1827 nobody will ever want to survive such error.
1828 * Create specific errors for each MSG case of failure:
1829 host failure, transfer failure, timeout, task cancelled
1830 * Cleanup the exceptions that may get thrown by each function
1831 * Other internal cleanups in Java bindings. Performance still bad :/
1832 Ruby and Lua Bindings: create them
1833 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1834 * That's new and great, you should try them out.
1835 Same functionalities than Java bindings, only even less polished
1837 * Kill the useless "rate" argument of SD_task_get_execution_time()
1838 Everyone used to provide -1 as a value, it was not used, and the
1839 semantic of a possible use wasn't even clear.
1840 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1841 either as comm costs or compute costs to mean that there is no
1842 such thing for that specific task.
1843 * Add a SD_task_set_name() function
1844 * Fix SD_task_unschedule() on typed tasks
1845 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1846 * In DAX loader, accept useless 'level' attributes to <job> since
1847 LIGO DAGs have them (seem to be to ease graphical representation).
1849 * Add an example masterslave_mailbox.c using send/receive and not
1850 the deprecated put/get interface.
1851 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1852 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1854 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1855 (declare MSG_USE_DEPRECATED to still have the old name)
1856 * Add a MSG_task_set_data() function
1857 * About trace replay (see examples/msg/actions):
1859 - Allow to work with splitted trace files for each process
1860 Give the specific trace file as argument of each process,
1861 and call MSG_action_trace_run(NULL)
1862 You can still have one merged file for all processes.
1863 - Fix implementation of collective operations
1864 * Allow task_execute() on 0-sized tasks (closes #10063)
1866 * This is the first release of SimGrid where SMPI is not considered
1867 beta anymore (even if some corners should still be improved)
1868 * Port over the new SIMIX_network submodule (internal refactoring)
1869 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1870 * Implement more missing elements of the standard:
1872 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1874 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1875 MPI_Scan() MPI_Get_processor_name()
1876 - Added implementation of missing case for Alltoall (warning: it's
1877 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1878 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1879 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1880 SMPI_MPI_Allgatherv()
1881 * Bug fixes include:
1882 - MPI_Waitsome() was broken
1883 - Allow relative includes in smpicc
1884 - Command line cfg argument 'reference_speed' was ignored...
1885 - Some functions did not properly lead to auto-benching of user code
1886 - smpicc passes -O2 by default (just like openmpi one)
1888 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1889 * Bug fixes about timeouts during communications
1890 * add SIMIX_message_sizes_output() as a pimple to write to file the
1891 amount of messages per size. Use gnuplot to get histogram.
1892 Pimple because that's the only user-visible function of simix,
1893 defined directly in xbt.h (irk, sorry)
1895 - Add a SIMIX_sem_get_capacity() function
1896 - Fix interactions with processe resume/suspende
1897 - release_forever() was stupidly broken
1898 - Fix SIMIX_display_process_status() for processes in a semaphore
1899 - Make SIMIX_sem_block_onto() user-visible
1900 * Refactoring context stuff:
1901 - Use pseudo-OOP for better modularity
1902 - reimplement SIMIX_process_kill() without process_schedule() so
1903 that the latter can take as invariant that it is called from
1905 - Merge context_start into context_new for sake of simplicity
1907 * Add a Vivaldi network model, coded live during SUD'10 ;)
1908 * Rename configuration variables to start a hierarchy:
1909 o cpu_model -> cpu/model
1910 o network_model -> network/model
1911 o workstation_model -> workstation/model
1912 * New configuration variables:
1913 o network/bandwidth_factor: correction to bandwith
1914 o network/latency_factor: correction to latency
1915 o netwotk/weight_S: correction to the weight of competing streams
1916 * Add a long description to the models, that users can see with such
1917 argument on the command line: --cfg=cpu/model:help
1918 * --help-models display the long description of all known models
1920 * config: add the ability to set a default value after registration
1921 Does not override any previously set value (e.g. from cmd line)
1922 * dict: allow to have integer key and data.
1923 When so, you need to use the following functions
1924 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1925 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1926 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1927 In contrary to regular dicts, the key is not malloced before copy.
1928 Mixing scalar and regular elements in the same dict is not tested
1930 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1931 Tracing for Visualization:
1932 * SimGrid is now instrumented in order to generate a trace file for
1933 visualization analysis: to use it, need to compile SimGrid with the
1934 "tracing" option enabled, and instrument the program using SimGrid with
1935 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1936 (among other functions).
1937 * The instrumentation only traces the platform utilization for now
1938 * Documentation to use the tracing functions and how to analyze the
1939 traces with the Triva tool is written.
1940 * More information about: SimGrid FAQ (in the section Tracing Simulations
1943 * We moved to cmake as default build system. Autotools support will
1944 be dropped soon. Check the FAQ for more info about how to use it.
1945 * Greatly improved our cdash/ctest interactions
1946 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1947 * Added memory checking tests with valgrind; lot of memleak fixing.
1948 This may be the first release of simgrid with so few memory issues
1949 * Added code coverage tests.
1950 Our coverage is still improvable, but at least we see it on cdash.
1952 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1954 SimGrid (3.3.4) stable; urgency=low
1956 The "Desktop Grid needs love too" release (also called Xmas release).
1958 Models improvements:
1959 * Major speedup in the maxmin system solving by using lazy evaluation
1960 Instead of solving completely the maxmin system at each iteration,
1961 only invalidate (and recompute) the modified parts.
1962 This new feature is enabled in default models but you can try to
1963 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1964 * Cas01 IMproved as default CPU model
1965 This CPU model is the same Cas01 model, but it uses the
1966 maxmin-selective-update flag and a heap structure to manage
1967 actions on SURF kernel.
1968 It reduces the complexity to find the next action to finish and,
1969 consequently, it's faster than the old Cas01.
1970 This is the new default CPU model (Cas01).
1971 * Rename the old Cas01 model to Cas01_fullupdate
1972 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1973 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1974 * CpuTI (CPU Trace Integration)
1975 A new CPU model whose objective is simulate faster when using
1976 availability trace files.
1977 Instead of using a full featured, over engineered maxmin system for
1978 CPU modeling, this model does the pre-integration of traces files
1979 to calculate the amount of CPU power available, and so, executes
1980 faster than the old CPU models.
1981 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1982 * Use LV08 as default network model since it gives better accuracy
1983 for small messages and shouldn't change things for big ones.
1984 Use --cfg=network_model:CM02 to get the previous behavior.
1987 ******************************************
1988 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1989 ******************************************
1990 * The new CPU model may changes simulations!
1991 The point is that events occurring at the exact same timestamp
1992 are not scheduled in the same order with the old and new
1993 version. This may be enough to completely change the execution
1994 of simulations in some cases.
1995 * The new network model will change simulations!
1996 This new model is more realistic than the previous one, so you
1997 should consider redoing your old experiments with this model.
1998 Sorry for the inconvenience.
2001 * Introduce the supernovae compilation mode
2002 When compiled that way, the whole SimGrid (or almost) is put in a
2003 single compilation unit and compiled in one shoot.
2004 This is to help gcc which has difficulties to inline stuff from one
2006 The speedup seem to be above 15%, althrough more tests are needed on
2007 amd64 to confirm that gain.
2010 * Port of MSG's mailbox on top of SIMIX network
2011 The put/get mechanism was greatly simplified on the way.
2014 * New SIMIX network module. Provides:
2015 - Mailbox: rendez-vous mechanism to find with who you want to speak
2016 - Synchronous send/recv: easier and hopefully faster since the
2017 logic is handled in the maestro process directly now
2018 - Asynchronous send/recv: you dreamt of it? It's here now
2019 Too bad that nobody cared enough to propagate the change to MSG.
2020 * Add semaphores as SIMIX synchronization mechanism.
2023 * new function SD_daxload(char*) to load a DAX file
2024 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
2025 * Introduce typed tasks. Specify its kind and cost at creation.
2026 At scheduling, just give where it should be placed, and the cost
2027 for each involved resource is automatically computed.
2028 Existing constructors so far (more to come of course):
2029 - SD_task_create_comm_e2e() for end-to-end communication
2030 - SD_task_create_comp_seq() for sequential computation
2031 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
2032 * new function SD_task_dump() for debuging display
2033 * new function SD_task_dotty(task,FILE*) writing to file the info
2034 about the task in dotty format
2035 * SD_task_dependency_exists() can now cope with having one of its
2036 arguments NULL. If so, it tests whether the other argument has any
2038 * Add getters on list of preceding/following tasks:
2039 SD_task_get_parents(task) and SD_task_get_children(task)
2040 * Add getters on amount of workstations and list:
2041 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
2042 * Add getter on task kind: SD_task_get_kind(task)
2043 * Update the start_time and finish_time of tasks on completion/failure
2044 * Bugfix: Remove task from state swags when destroyed
2047 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
2050 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
2051 dynar with one separate thread per value of the dynar.
2052 * Change the prototype of xbt_thread_create(), sorry.
2053 Added a boolean parameter indicating whether we want to join this
2054 thread (used in SG only for now)
2055 * Implement xbt_thread_join and xbt_thread_yield in SG also.
2058 * GTNetS wrappers should now be usable again (and betterly tested too)
2059 * Fix a major regression from 3.2 where the timeout provided to
2060 MSG_task_put_with_timeout() was used as absolute time before which
2061 the comm should be done.
2062 * Start to fix the <cluster> tag.
2063 - Internal links should be good now (beside of the loopback, which
2064 use the private link instead)
2065 - paths to the external world is still rather broken
2066 - the <route:multi> tag is just broken. Actually that's brain-dead.
2067 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
2068 to make it less stupid
2069 ** Check your platform with teshsuite/simdag/platforms/flatifier **
2070 * Fix a source-level compatibility glitch from 3.2: after defining
2071 MSG_USE_DEPRECATED, you can use the old name
2072 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
2073 * Allow to compile from the SVN with automake 1.11
2074 * Fix some problems when using the "start_time" tag in deployment XMLs.
2075 * Fix #8569: XBT/synchro.h has redundant declarations
2076 * Fix #8563: MSG return values and exceptions
2077 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
2078 * Integrate patch #8636: Obey DESTDIR when installing documentation.
2079 Thanks to Robson Peixoto.
2080 * Fix a vicious bug in dictionaries inducing that some elements were
2081 not freed on xbt_dict_free()
2083 Portability report of this version:
2084 * Main portability targets:
2085 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
2086 on (amd64/i386/ia64)
2087 - mac leopard on i386
2088 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
2089 but nothing critical.
2090 * Other platforms: windows, AIX and others were not tested for this release
2092 Timing report of this version:
2093 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
2094 scenario parameters). From 8h to a few seconds in desktop grid settings.
2095 * Supernovae brings about 25% speedup on i386.
2097 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
2099 SimGrid (3.3.3) stable; urgency=low
2101 The "Need for Speed" release.
2103 The timings done to validate the 3.3.2 were faulty.
2104 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
2106 The problem was a conversion from a manually handled vector to
2107 xbt_dynar_t on the critical path.
2108 xbt_dynar_foreach calls functions, inducing stack management crap.
2110 We inlined these functions and xbt_dynar_foreach is now breath taking.
2111 We also inlined xbt_swag_belong on the way.
2113 Here are some approximate speedup measurements (on master/slaves
2114 simulations lasting between 10s and 20s each):
2115 3.3.1 -> 3.3.2: about same performance
2116 3.3.2 -> 3.3.3: 40% speedup
2117 3.3.1 -> 3.3.3: 40% speedup
2118 3.3.1 with inline patch -> 3.3.3: 30% speedup
2120 Our reading is that the refactoring which occurred in 3.3.2 made us
2121 suffer much more from the xbt_dynar_foreach low performance, but
2122 once we solved this, this refactoring proved to be very performance
2123 effective. From the 40% speedup, somehow, 10% are due to the
2124 inlining and 30% to the refactoring.
2126 That's a pitty that gcc cannot inline functions placed in other files
2127 alone. We have to choose between:
2128 - break the encapsulation (by putting private data structures and
2129 accessors in headers files to help gcc)
2130 - live with low performance
2131 - switch to a decent compiler such as icc (not quite possible).
2133 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
2135 SimGrid (3.3.2) stable; urgency=low
2137 The "Simplicity does not preceed complexity, but follows it" release.
2139 The main contributors of this release were (lexical order):
2140 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
2143 * Extract the routing logic into its own object.
2144 (was dupplicated in network.c and workstation_LV07.c;
2145 Allows to implement other ways of storing that info)
2146 => kill now useless network_card concept
2147 - Use dynar to represent routes (instead of void** + int*)
2148 - kill link_set (use surf_network_model->resource_set instead)
2149 - Add a command-line option to choose the routing schema to use
2150 - Add three new models:
2151 * Floyd (shortest path computed at initialization)
2152 * Dijikstra (shortest path recomputed all the time)
2153 * Cached Dijikstra (shortest path computed on need)
2154 All these models where contributed by Silas De Munck, and are
2155 described in his ICCS09 paper.
2157 * Simplify model declaration
2158 (less redirections, less function to write when defining a model)
2159 - Factorize stuff between models:
2162 surf_model_resource_set(model)
2163 surf_model_resource_by_name(model, name)
2164 - Unify the types of models in s_surf_model_t (using an union)
2165 - Embeed fields of common_public directly into s_surf_model_t
2166 - Rename model methods:
2167 action_free ~> action_unref
2168 action_change_state ~> action_state_set
2169 action_get_state ~> action_state_get
2170 - Change model methods into functions :
2171 (model)->common_public->action_use ~> surf_action_ref
2173 * Implement a generic resource; use it as ancestor to specific ones
2174 (allows to kill duplicated code in models)
2175 Drawback: timer command don't need no name nor properties;
2176 workstation_CLM03 don't need no properties
2177 (but I guess we can live with those few bytes wasted)
2179 * Improve the action object model
2180 - implement a constructor avoiding dupplicated code about field
2181 initialization in generic_action part.
2183 * Kill the SDP model: it has an external dependency, is deprecated
2184 in flavor of modern lmm models, and didn't compile since a while
2187 * Relocation of the context module from XBT to SIMIX.
2188 (the context were decoupled from the simix processes, duplicating a lot of code)
2189 => a lot of code was factorized
2190 - less overhead is introduced during scheduling
2191 - simpler API for the context factory
2192 - the logic for process creation,destruction and manipulation was simplified
2193 * Simplification of the s_smx_process_t data structure.
2194 => accesing the simix level data associated to a process is faster now,
2195 and the code is a lot more readable.
2198 * Implement some more MPI primitives:
2199 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
2200 -implementation: Bcast: flat or 2-ary tree (default),
2201 Barrier: 4-ary tree,
2203 Allreduce: Reduce then Bcast
2204 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
2205 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
2206 Alltoallv: flat tree, like ompi
2208 * Add support for optimized collectives (Bcast is now binomial by default)
2209 * Port smpirun and smpicc to OS X
2212 * Kill SD_link_get_properties: hard to maintain and makes very little sense
2213 Shout out if you used it.
2216 * Display the list of still queued messages in SG mode when existing
2220 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
2221 * Add xbt_graph_node_get_outedges() [Silas De Munck]
2222 * Add xbt_str_from_file(FILE*)
2223 * Add xbt_dict_get_key achieving a linear reverse search
2224 * Remove the context module
2226 Portability report of this version:
2227 * Main portability targets:
2228 - Linux(debian)/x86/context
2229 - Linux(debian)/x86/pthreads
2230 - Linux(debian)/amd64/context
2231 - Linux(debian)/amd64/pthreads
2232 On these, we still have the eratic breakages of gras/pmm and
2233 amok/saturate_sg reported in previous version. We still think
2234 that the tests are the cause of the fault, not the tested code.
2236 - Mac OSX Leopard/x86/context
2237 Still false negative in tesh autotesting.
2238 Smpi still fails, but this time because readlink does not accept -f
2239 Everything seems to work properly beside of that.
2242 - AIX version 5.3 (only tested contexts this time)
2243 Smpi still fails there because mktemp is not installed.
2244 Everything seems to work properly beside of that.
2246 I managed to compile it for the first time, but several breakages.
2247 Won't delay the release for this exotic platform.
2249 * Windows: it's still lagging behind. If you want to help, please
2252 Timing report of this version:
2253 This version seem to be more than 5% faster than 3.3.1 (on linux
2254 64bits with contextes). The gain is less than expected, we are
2255 investigating this for next release.
2257 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
2259 SimGrid (3.3.1) stable; urgency=low
2262 * Implement a --cfg-help to show existing configuration variables
2263 * Build chain do not require doxygen in maintainer mode
2266 * fix a bug on struct sizeof computation, which prevented the
2267 exchange of arrays of structs in some conditions
2268 - added a regression test about this in datadesc_usage
2269 * Allow the exchange of 0-long dynamic vectors.
2270 - for that, use -1 as indicator of dynamic size instead of 0
2271 - This implied to change any size from unsigned long to long,
2272 reducing a bit communication abilities, but I guess that with
2273 64bits being quite common, this is more than enough.
2274 - This also induce a protocol change, thus bumping network protocol
2275 version from 0 to 1 (if we have external users, we have to get
2276 clean on that point too ;)
2277 - added two regression tests about this in datadesc_usage
2278 * Be more verbose when propagating local exceptions
2279 This helps debugging.
2280 * Display the status of simulated processes when receiving SIGINT in
2284 * Allow to control the simulation from a trace file.
2285 New functions MSG_action_register() and MSG_action_trace_run()
2286 The first one allows to associate a function execution to each
2287 kind of action while the second one parses a trace file and
2288 triggers the corresponding actions within the system.
2289 For now, only a toy example is provided in examples/msg/actions
2290 * Add an exemple of process migration in examples/msg/migration
2291 * Fix a bug in task exchange which broke MSG_task_get_sender()
2292 Add a teshsuite regression test for that.
2293 [Bug: if MSG_task_get_sender() is called after sender exit,
2295 * Fix a bug which prevented suspend/resume to work properly
2296 * Display the status of simulated processes when receiving SIGINT
2297 This fixes a regression of v3.3. due to the introduction of SIMIX
2298 * Bug fixing in failure management:
2299 - trace could not start by a failure at time 0
2300 - failure during communications were not working
2303 * Add SIMIX_process_set_name() to change the name of the current
2304 process in the log messages.
2305 * Store smx_hosts in a dict since we only retrieve them by name
2306 * Move the configuration infrastructure to surf
2309 * Move the configuration infrastructure to surf
2312 * Massive internal cleanups:
2313 - Store internal structures on processes instead of hosts (allows
2314 to have more than one process per host, in addition of being more
2316 - Cleanup the initialization/finalization process
2317 - Kill a whole bunch of unneeded synchronization:
2318 processes run in exclusive manner within the simulator
2319 - Move queues from global tables to process data fields
2321 - now accept -platform and -hostfile arguments
2322 - Pass the right rank value to processes according to the hostfile
2323 * Compile the examples by default, and use them as regression tests
2324 * Implement MPI_Wtime()
2325 * Change the reference speed to a command line option
2328 * TCP_gamma can now be specified as command line option using
2329 --cfg=TCP_gamma:10000000.0
2330 * Change the --surf-path cmd line option into --cfg=path:
2333 * Also include strbuff from xbt.h public header
2334 * xbt_ex_display(): do not free the exception after displaying
2335 This allows to do more with the given exception afterward.
2336 Users should call xbt_ex_free() themselves.
2340 Portability report of this version:
2341 * Main portability targets:
2342 - Linux(debian)/x86/context
2343 - Linux(debian)/x86/pthreads
2344 - Linux(debian)/amd64/context
2345 - Linux(debian)/amd64/pthreads
2346 These targets fail about 1/10 of times on gras/pmm, but we believe
2347 that this is because of the test, not because of simgrid.
2348 amok/saturate_sg fails even more rarely, and the test may not be
2351 - Mac OSX Leopard/x86/context
2352 The test suite still spits tons of errors because some obscure
2353 force prevents us from removing the temporary directories
2354 arguing that they still contain some metadata I've never heard of.
2355 Smpi fails because seq is not installed.
2356 Everything seems to work properly beside of that.
2359 - AIX version 5.3 (both contexts and pthread)
2360 Smpi still fails there because mktemp is not installed.
2361 XML inclusions seems rosty on AIX.
2363 * Windows: it's still lagging behind. If you want to help, please
2366 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2368 SimGrid (3.3) stable; urgency=high
2372 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2373 [Malek Cherier & Mt]
2375 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2378 Until now, GRAS were using MSG as an interface to SURF. It was
2379 quite difficult because both interface have several differences
2380 (MSG channels vs GRAS sockets were the most notable point).
2382 This also opens the gate to SMPI (which should occur soon) and speed
2383 up simulations by to 40% (even if it were not the main goal).
2385 **************************************
2386 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2387 **************************************
2388 The point is that events occuring at the exact same timestamp are
2389 not scheduled in the same order with the old and new version. This
2390 may be enough to completely change the execution of simulations in
2391 some cases. Sorry for the inconvenience.
2393 * Cleanup and upgrade the XML format to push further scalability
2394 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2396 * Improve the testing infrastructure with tesh. Now a very large part of
2397 the code is tested not only by being run but also by checking that the
2398 output match an expected output [Mt].
2400 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2401 is really less memory-demanding, which should allow you to use
2402 larger files in SimGrid [AL].
2404 * Inform valgrind about our contextes, so that it becomes usable
2405 with the default (and more effecient) version of SimGrid
2406 [contributed by Sékou Diakite, many thanks]
2409 * Introduce a listener thread in charge of receiving incoming
2410 messages from the network. It allows to overlap communication and
2411 computation but most notably, it removes some stupid deadlocks due
2412 to the fact that so far, a process could not send and receive at
2413 the same time. This made most non trivial communication schema
2415 * Convert the PIDs from long int to int to match the MSG ones (and
2416 linux ones too) [Mt]
2417 * New function: gras_agent_spawn() to launch a new process on
2418 current host. Only working in simulation for now. [Mt]
2419 * New function: gras_os_hostport() returning a constant form (ie,
2420 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2423 * Make the backtrace of exceptions more human readable [Mt]
2424 * New module: xbt/str [Mt]
2425 a ton of string utility functions (split, join, printf to a newly
2426 allocated buffer, trim, etc)
2427 * New module: xbt/hash [Mt]
2428 SHA1 hashing algorithm (more to come if needed)
2429 * New module: xbt/synchro [Mt]
2430 synchronization tools (mutex and conditions) working the same way
2431 in simulation and in real life (mainly useful for GRAS, but not
2433 * New module: xbt/queue [Mt]
2434 classical producer/consumer synchronization scheme
2435 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2436 (using the classical functions will get serialized) [Mt]
2437 * Make dictionary internal table dynamic. No need to specify its size
2438 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2440 * Make sure the log channels are organized as a tree under windows
2441 (because of ANSI C compatibility issue, any channel were child of
2442 root directly) [Mt].
2445 * Cleaned many thing in surf and fixed a few bugs [AL].
2446 * Add a nice command line configuration mechanism to compose models [AL].
2447 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2448 the previous one (KCCFLN05). It relies on something that looks like
2449 a max-min sharing mechanism but cannot be written as such. A new solver
2450 was thus designed [AL].
2451 * Add a new solver to lmm. Based on Lagrange optimization and
2452 gradient-based descent, it enables to efficiently maximise systems s.a
2454 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2456 This solver enables to propose two new network models for TCP Reno and
2457 TCP Vegas based on Low's work. These models still need to be fully
2458 tested though [Pedro Velho].
2461 * Bug fix in SD_simulate. Now the time bound given as argument is
2463 * Use the new parallel task model (ptask_L07) as default.
2464 * Use the SURF command line configuration mechanism.
2465 * 0-size tasks (for synchronization) should now work.
2467 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2469 SimGrid (3.2) stable; urgency=high
2473 We still experience issues on this platform, but we believe that at
2474 least MSG is usable.
2476 GRAS API BREAKAGE (for simplification purpose, sorry):
2477 * the gras_msgtype_by_name is not used anymore. Instead of
2478 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2479 you can write (and must)
2480 gras_msg_send(toserver, "request", &request);
2481 - If you still want to pass a gras_msgtype_t to the function (to cache
2482 the type and avoid the lookup time), use the gras_msg_send_() variant.
2483 - Impacted functions:
2484 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2485 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2486 * The callbacks are now expected to return 0 when everything went well
2487 (just like the main() function)
2489 GRAS new features and improvements:
2490 * New module mechanism where user code can use per process globals [Mt]
2491 This is similar to gras_userdata_*() functions, but for libraries. It
2492 factorize some code developped over and over in the examples and AMOK.
2493 It has still to be documented and used (only amok/peermanagement is
2495 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2496 loss when they were small enough to fit into the buffer and sent quickly
2497 enough so that they can all get received in one shoot.
2498 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2499 if not found. Use the *_or_null() variant for the old semantic.
2500 * In gras_msg_handle, do not discard messages without callback.
2501 They are probably messages to be explicitly awaited later (ie, proofs of
2502 mis-synchronization in userland since they are sent before being awaited)
2503 No big deal usually.
2504 * gras_socket_meas_send/recv: semantic changed!
2505 The numerical arguments used to be (1) the total amount of data to send
2506 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2507 messages. This was need for the fool willing to send more than MAXINT
2508 bytes on quite fat pipes.
2511 * Do really rename the hostmanagement module to peermanagement. [Mt]
2512 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2513 but this is rather new and this was documented in the module
2514 documentation (poor excuses, I admit)
2515 * Bandwidth measurement semantic changed! This follows the changes to
2516 gras_socket_meas_send/recv explained above.
2519 * A sequential mode has been added to the workstations. When a workstation
2520 is in sequential mode, it can execute only one task, and the other tasks
2521 are waiting in a FIFO. [Christophe Thiery]
2524 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2525 model for SIMDAG. [Christophe Thiery]
2526 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2527 equal to 0, which caused some bad behaviors in
2528 saturated_constraint_set_update. I now use a threshold mechanism like in
2532 * When running manually src/testall, you select specific units [Mt]
2533 testall is the result of our cunit mechanism, and should replace all
2534 the scripty thingy around since bash don't run easily on billware.
2536 * A mallocator system has been added. [Christophe Thiery]
2537 Mallocators allow you to recycle your unused objects instead of freeing them
2538 and allocating new ones.
2540 Documentation update:
2541 * FAQ reworking + New FAQs:
2542 - "Valgrind spits tons of errors!" [Mt]
2543 - "How to repport bugs" [Mt]
2544 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2545 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2546 - Communication time measurement within MSG [AL]
2547 - I experience weird communication times when I change the latency [AL]
2548 * GRAS tutorial [Mt]
2550 - an introduction to the framework and to the used communication model
2551 - an initiatic tour introducing the most proheminent features:
2553 . Lesson 0: Installing GRAS
2554 . Lesson 1: Setting up your own project
2555 o Part 2: Message passing
2556 . Lesson 2: Exchanging simple messages
2557 . Lesson 3: Passing arguments to the processes (in SG)
2558 . Lesson 4: Attaching callbacks to messages
2559 . Lesson 5: Using globals in processes
2560 . Lesson 6: Logging information properly
2561 . Lesson 7: Using internal timers
2562 . Lesson 8: Handling errors through exceptions
2563 . Lesson 9: Exchanging simple data
2564 . Lesson 10: Remote Procedure Calling (RPC)
2565 . Lesson 11: Explicitely waiting for messages
2566 . Recapping of message passing features in GRAS
2567 - A HOWTO section containing:
2568 o HOWTO design a GRAS application
2569 More are due, of course. They will come latter. In the meanwhile, you can
2570 check the examples which are still here.
2572 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2574 SimGrid (3.1) stable; urgency=high
2578 There was a stack corruption somewhere, visible only when optimizing
2579 with these versions. [Vince]
2582 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2583 module that enabled to play with Directed Acyclic Graphs. It is built
2584 directly on top of SURF and provides an API rather close to the old
2585 SG. Some old codes using SG are currently under rewrite to check that
2586 all needful functions are provided. [Christophe Thiery]
2589 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2590 extension of the classical CLM03 model that gracefully handles
2591 failures. This is now the default model for MSG and GRAS. It doesn't
2592 handle parallel tasks yet though. [AL]
2593 * Bug fix: Weights were not correctly set in the network part.
2594 WARNING: This may have resulted in incorrect results with simulations
2595 where there are more than one flow on a given link. [AL]
2598 * After a (long ?) discussion on simgrid-devel, we have decided that the
2599 convention we had on units was stupid. That is why it has been decided
2600 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2601 WARNING : This means that all previous platform files will not work as
2602 such with this version! A warning is issued to ask users to update
2604 A conversion script can be found in the contrib module of the CVS, under
2605 the name contrib/platform_generation/surfxml_update.pl [MQ]
2608 * Bug fix: Processes were started in reverse order, wrt deployment file.
2609 WARNING: if your code relies on this bug, please fix it. [AL]
2610 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2611 being executed on two different locations. [AL]
2612 * Bug fix: Failures are now better supported thanks to Derrick's tests
2613 (there was many failure situations I hadn't thought of and that weren't
2614 correctly handled). [AL]
2615 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2619 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2620 Exception occurring on server-side are propagated back to client (!).
2622 API CHANGE: the callback changed their prototype. Change:
2623 int my_handler(gras_socket_t expeditor, void *payload_data) {
2625 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2626 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2628 * New! function: gras_msg_handleall to deal with all messages arriving
2629 within a given period.
2630 * New! function: gras_socket_server_range to get a server socket in a
2631 range of port numbers (ease to avoid port number conflicts) [MQ]
2632 * New! gras processes display their backtrace when they get a SIGUSR1
2633 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2634 Sweet to debug RL processes [MQ]
2638 - Do not force experiment sizes to be expressed in kb, or it becomes
2639 impossible to measure the latency this way (needs one byte-long tests)
2640 WARNING: this changes the amok_bw_* function semantic. [MQ]
2641 - Implements the link saturation stuff. [MQ]
2642 * Peer management module:
2643 New! module factorizing code that we wrote over and over [MQ].
2646 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2647 - Test units are placed directly into the library code, they get extracted
2648 automatically and placed into the src/testall binary.
2649 - Convert most of the XBT tests to this system.
2650 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2651 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2652 live in the same C code. This required to change a little bit the API
2653 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2654 * New module: sparse graph structure with basic algorithms (this is work
2655 in progress and the API is not considered to be frozen yet). [AL]
2656 * Display more information on backtraces: source line & function names are
2657 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2658 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2659 * Reworked a little bit some #include statements to load only required
2660 headers. Some user code that relied on SimGrid to include stdlib or
2661 stdio may need to include it by themselves. [AL]
2662 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2664 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2665 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2666 but I guess that almost nobody used those parts. [MQ]
2668 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2670 SimGrid (3.0.1) stable; urgency=low
2673 * Unfortunately, I had missed 5 misnamed functions:
2674 xbt_fifo_item_t xbt_fifo_newitem(void);
2675 void xbt_fifo_freeitem(xbt_fifo_item_t);
2676 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2677 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2678 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2679 They're now deprecated. Please use their new versions:
2680 xbt_fifo_item_t xbt_fifo_new_item(void);
2681 void xbt_fifo_free_item(xbt_fifo_item_t);
2682 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2683 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2684 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2686 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2687 * Documentation: xbt_log module unmercifully reworked [MQ]
2688 * Bugfix: there was a problem with the ending of contexts with
2689 the pthread backend. It caused some weird deadlock or behavior
2690 depending on the pthread implementation. [AL]
2691 * Bugfix: get the exceptions raised in the simulator repport where
2692 and why they come from when they are not catched in time [AL, MQ]
2695 * Bugfix: Do repport the error when two non-connected hosts try to
2696 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2699 * Add additionnal checkings on communications. Assert that two
2700 communicating hosts are connected by a set of links... [AL]
2703 * Add additionnal checkings on channel values in communication [AL]
2704 * New: MSG_task_get_source to see on which host a task was generated [HC]
2705 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2706 the number of tasks waiting to be received on channel and sent
2708 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2709 waits for the first task coming from a given host.. [AL]
2711 GRAS new functionnalities: [MQ]
2712 * Enhance the parsing macro to allow the size of multidimentional objects
2713 to be given thru annotations.
2714 * New example (and documentation): Matrix Multiplication a la RPC
2715 (as when I was young!) and fix a bunch of bugs found on the way.
2717 GRAS performance improvements: [MQ]
2719 * Reduce the amount of cbps creation/destruction by making it static to
2720 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2723 * Change libdata to a set so that we can search for stuff by ID (and thus
2724 reduce the insane amount of dict lookups)
2727 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2728 RL communication on top of SG since it's so uneffective.
2729 It may also be used for inter-thread communication in RL, one day.
2730 * Use gras_datadesc_copy() to exchange messages on top of SG
2732 - improve message exchange performance on top of SG
2733 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2734 * Don't exchange on the network the size of the used part of buffer,
2735 instead, specify the possible buffer size to read().
2737 - reduces the amount of read/write calls (one pair per exchange)
2738 - reduces the amount of exchanged data (the size)
2739 - allows to retrieve all arrived data on receiver side, if we don't need
2740 it right now (subsequent read will peek the buffer)
2741 - allows the receiver to proceed with the begining of the stream before
2742 everything is arrived
2743 - make it possible to build an iov transport (using readv/writev)
2745 - take care of the data with non-stable storage (like stacked data),
2747 * If possible, TCP send uses vector I/O (when writev() is here)
2748 - Don't use it for receive since we send data sizes and data on the
2749 same stream, so we wouldn't be able to chain large amount of chunks
2750 before having to flush the stuff to read the size.
2751 * Rework the transport plugin mechanism to simplify it and reduce the
2752 amount of pointer dereferencement when searching for the right function
2755 * I guess that now, we do almost as few system calls as possible while
2756 doing as few data copy as possible.
2758 To improve it further, we could try to send all the sizes first and then
2759 all the data (to use iov on receiving size), but it's only a partial
2760 solution: when you have 2 dimensional data, the sizes of the second
2761 dimension is data of the first dimension, so you need 3 streams.
2763 I'm not sure the potential performance gains justify the coding burden.
2765 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2767 SimGrid (3.00) stable; urgency=high
2770 * New! Give the possibility to hijack the surf parser and thus bypass
2771 MSG_create_environment and MSG_launch_application. Have a look at
2772 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2774 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2776 SimGrid (2.96) unstable; urgency=low
2781 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2782 This deprecates the xbt_error_t mechanisms.
2783 It modifies (simplifies) all XBT and GRAS API.
2784 MSG API keeps unchanged (exceptions raised by XBT are catched from
2785 within MSG and masked with existing error handling facilities)
2788 * New! Add a FATPIPE model. [AL]
2789 * New! Add a parallel task model. [AL]
2790 * New! Add automatically a loopback interface (in the default
2791 network model) if none was precised.
2794 * Bugfix: MSG_process_resume now works with the current running process.
2796 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2797 * Modification of MSG_task_get_compute_duration. Once a task has been
2798 processed, the value returned by this function is now equal to 0. [AL]
2799 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2800 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2801 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2802 of a m_task. For now, MSG_task_cancel only works with computation
2804 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2805 of the processor (in Mflop/s) regardless of the current load on the
2807 * API Change: use proper naming convention for MSG_getClock and
2808 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2810 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2811 This function changes the priority of a computation task. This priority
2812 doesn't affect the transfer rate. A priority of 2 will make a task
2813 receive two times more cpu power than the other ones. This function
2814 has been added to suit the needs of Nguyen The Loc and hasn't been that
2815 much tested yet. So if it fails, please report it and send me your code.
2817 * API Change: removed all functions and types that were marked "deprecated"
2818 since many months. Renamed MSG_global_init_args to MSG_global_init.
2820 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2822 SimGrid (2.95) unstable; urgency=low
2825 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2826 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2827 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2830 * Bugfix: Allow absolute paths for platform description files [MQ]
2831 * Bugfix: do free the variables after use. Leads to drastic performance
2833 * Implement max_duration (ie, timeouts) on resources [AL]
2836 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2838 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2839 the needed support (timeouts) [AL]
2842 * Implement measurement sockets. You can now get the bandwidth between two
2843 hosts thanks to AMOK (see below). [MQ]
2844 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2845 dynar over the network (yeah) [MQ]
2846 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2847 * simple/static token-ring example. [Alexandre Colucci and MQ]
2848 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2849 in place before [MQ]
2852 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2853 type parsing. "short" and "long" modifiers were also ignored; other
2854 modifier (such as reference level) are still ignored. [MQ]
2855 * Embeed the buffer size within the buffer itself on SG. [MQ]
2856 That way, send() are atomic and cannot get intermixed anymore (at least
2857 the ones which are less than 100k; bigger messages still have the issue)
2858 * Array size pushed by the field, not by the field type (or each
2859 and every long int will push stuff to the cbps) [MQ]
2860 * use select() to sleep since it allows to portably sleep less than one
2863 GRAS (minor cleanups)
2864 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2867 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2868 * type category 'ignored' killed as it was never used and were difficult
2870 * whether a type can cycle or not is now a flag, leaving room for more
2871 flags (as "ignored", if we feel the need one day ;)
2872 * Rename raw sockets to measurement sockets since "raw" has another
2873 meaning in networking community.
2876 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2877 and offers features not belonging to GRAS but that most applications
2878 need. One day, it may be a set of plugins loadable at runtime.
2879 * New module: bandwidth
2880 bandwidth measurement between arbitrary nodes running this module. [MQ]
2882 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2884 SimGrid (2.94) unstable; urgency=low
2886 The first beta release of SimGrid 3 !
2890 * Update the main page and the FAQ. Adding references to gforge.
2893 * Add a gras_os_getpid function.
2896 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2897 * Extend the logs so that they also print PID, hostname, date, ... if
2899 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2900 and kill the old version which were in testsuite/
2901 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2905 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2906 for users. Instead, each programming environment (GRAS, MSG,...) use it
2907 when needed to provide such a feature to users.
2908 Don't shortcut the mechanism or you will also shortcut the virtualization
2909 you need on the simulator.
2913 * Cleanups in configury with regard to compile optimization/warning flags.
2914 Also add -fno-loop-optimize to any powerpc since it's the optimization
2915 killing gcc (< 3.4.0).
2916 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2917 needed by MSG examples complications
2918 * Borrow configury beautifications from PHP
2921 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2922 warning (thanks loris for stumbling into this one).
2923 * Bugfix: stop loading private headers (gras_config.h) from the public
2927 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2928 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2929 RL would imply a DNS resolver, which is *hard* to do in a portable way
2930 (and therefore delayed).
2931 * Implement a real timer mechanism and use it in timing macros. This allows
2932 to avoid rounding errors and get a 0.000005 sec precision in timing
2933 macros. While I was at it, various cleanups:
2934 - allow to declare more than one timed section per file (fix a stupid bug)
2935 - move some private declaration to the right place
2936 - merge conditional execution and timing macros into emulation module
2937 - document the module
2938 - make sure the module cleanups its mess on gras_exit
2939 * Documentation improvements:
2940 - (new) how to compile applications using GRAS
2941 - (new) emulation support (timing macros)
2943 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2945 SimGrid (2.93) unstable; urgency=low
2947 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2950 - Use Paje properly where used. Still to be sanitized properly.
2951 - Portability fix: Add an implementation of the contexts using pthread
2955 - Add xbt_procname(): returns the name of the current process.
2956 Use it to show the current process's name in all logging.
2958 - fix detection of older flex version and the reaction, since we do
2959 depend on modern ones (we use lex_destroy)
2960 - Better separation of SG and RL in the libs: remove all simulation code
2961 from libgras. As a result, this lib is now only 200k when stripped.
2962 Some of the xbt modules may also be duplicated (two sets and such) and
2963 should be cleaned/killed before SG3.
2964 - Insist on using xlC on AIX because of weird problems involving gcc there.
2965 - Cleanup the make remote stuff. This is now done by scripts
2966 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2967 mainly for our private use, but we're working on changing them to user
2970 - Bugfix: flush the socket on close only if there is some *output*.
2971 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2973 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2976 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2978 SimGrid (2.92) unstable; urgency=low
2980 Alpha 3 on the path to SimGrid 3
2984 - New! First try of benchmarking macros.
2985 - New! First try so that gras_stub_generator generate deployment and
2986 remote compilation helpers.
2988 - Bugfix: Initialization fix in msg_test.
2992 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2994 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2996 - API Change: gras_os_sleep to take the amount of seconds as a double.
2997 Accepting an int was error prone since it was the only location where
2998 seconds were coded as such. It leaded to damn rounding errors.
2999 - Bugfix: Hard to belive that timers ever worked before this.
3001 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
3003 SimGrid (2.91) unstable; urgency=low
3005 Alpha 2 on the path to SimGrid 3
3009 - Bug fix in the lmm_solver.
3011 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
3012 through the function MSG_paje_output.
3013 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
3014 - It is possible to bound the rate of a communication in MSG with
3015 MSG_task_put_bounded() (was already in the previous version but I had forgotten
3016 to write it in the changelog).
3017 - Bug fix to let GRAS run on top of MSG until we move it directly on top
3022 - Various cleanups to the autotools stuff
3023 - Begin to move Gras examples to examples/gras/
3024 - Let make distcheck work again (yeah!)
3026 - documentation overhauled using doxygen.
3027 gtk-doc-tools is dead in SimGrid now.
3028 - Automatically extract all existing logging categories, and add the list
3029 to the documentation (long standing one, to say the less)
3031 - Cleanup the known architecture table. Reorder the entries to group what
3032 should be, and use a more consistent naming scheme.
3033 (some of the test dataset are still to be regenerated)
3034 - New! Allow library to register globals on each process just as userdata
3036 This is implemented using a xbt_dict and not a xbt_set, so we loose the
3037 lookup time (for now).
3038 Use it in msg and trp.
3039 This cleans a lot the internals and helps enforcing privacy of the
3040 headers between the gras components.
3041 - New! Add a timer mechanism, not unlike cron(8) and at(1).
3042 - Bugfix: gras_os_time was delirious in RL.
3043 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
3045 - Reenable GRAS now that it works.
3047 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
3049 SimGrid (2.90) unstable; urgency=low
3051 Alpha 1 on the path to SimGrid 3
3053 * It is a long time since the last release of SimGrid. I'm sorry about
3054 that but as I had told you, I was rewriting a lot of things. I apologize
3055 to those who had been reporting bugs to me and that I had not answered.
3056 If your bug is still in the new version, please tell me. Here is a
3057 summary of the main changes.
3059 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
3060 lead by Martin Quinson. As a consequence SimGrid gains a lot in
3061 portability, speed, and a lot more but you'll figure it out later.
3062 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
3063 I wanted to release the new MSG as soon as possible and I have
3064 broken GRAS, which is the reason why, for now, only MSG is fully
3065 functional. A laconic description of these projects is available
3066 in the documentation.
3068 * REVOLUTION 2: I have removed SG and I am now using a new simulation
3069 kernel optimized for our needs (called SURF but only the developers
3070 should use it). Hence, MSG is now roughly 30 times faster and I think
3071 that by rewriting a little bit MSG, I could event speed it up a little
3072 bit more. Beside the gain in speed, it is also much easier to encode a
3073 new platform model with SURF than it was with SG. More to come...
3075 * REVOLUTION 3: I have tried to change a little as possible the API of
3076 MSG but a few things really had to disappear. The main differences
3077 with the previous version are :
3078 1) no more m_links_t and the corresponding functions. Platforms are
3079 directly read from a XML description and cannot be hard-coded
3080 anymore. The same format is used for application deployment
3081 description. The new format is described in the documentation.
3082 Have a look in tools/platform_generation. There is a tiny script
3083 that converts from the old platform format to the new one. Concerning
3084 the application deployment format, parsing the old one is tricky.
3085 I think most of you should however be able to convert your files. If
3086 it is really an issue, I can write a C code that does the conversion.
3088 2) the toolbox tbx does not exist anymore. We now have a library
3089 with much more data-structures but without the hash-tables (we have
3090 dictionaries that are much faster).
3092 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
3094 *****************************************************************************
3095 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
3096 * don't feel like converting the oldies. (Mt) *
3097 *****************************************************************************
3100 Version 2.90: "the long awaited one"
3101 - Finished rewriting and debugging MSG. Rewrote the documentation.
3102 - disable GRAS for now since it needs to be ported to the newest SG
3105 - Finish the port to windows (using mingw32 for cross-compile)
3108 - Main loop and datastructures of SURF. A cpu resource object is
3109 functional. Surf can thus be used to create cpu's with variable
3110 performance on which you can execute some actions.
3112 2004-11-15 Martin Quinson
3113 - Port to ARM. Simply added the alignment and size descriptions. Should
3114 work, but the ARM machines are so slow that I didn't had the opportunity
3115 to 'make check' over there yet.
3117 2004-11-15 Arnaud Legrand
3118 - Trace manager now written. It uses a heap structure and is therefore
3119 expected to be efficient. It may however be speeded up (particularly
3120 when many events occur at the same date) by using red and black
3121 trees. One day maybe...
3122 - Max-min linear system solver written. It uses a sparse matrix
3123 structure taking advantage of its expected use. Most operations are
3124 O(1) and free/calloc are called as few as possible. The computation of
3125 the minimum could however be improved by using a red and black tree
3128 2004-11-03 Arnaud Legrand
3129 - Rename every gras_* function that was in xbt/ to its xbt_
3131 - Add a heap and a doubly-linked list to xbt
3132 - Added a dichotomy to the dictionaries. make check works as well before
3133 so I assume that the patch is correct. I do not know however if things
3134 run effectively faster than before now. :)
3136 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
3137 SimGrid, and the version number is bumped to 2.x
3139 2004-10-29 Martin Quinson
3140 - Introduction of the remote errors.
3141 They are the result of a RMI/RPC on the remote machine.
3142 ErrCodes being scalar values, you can't get the host on which those
3143 errors did happen. Extending the error mechanism as in Gnome is possible.
3144 No idea yet whether it is a good idea.
3146 2004-10-28 Martin Quinson
3147 - Interface revolution: the Starred Structure Eradication.
3148 I used to do typedef struct {} toto_t; and then handle *toto_t.
3149 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
3150 - ???_t is a valid type (builded with typedef)
3151 - s_toto_t is a structure (access to fields with .)
3152 - s_toto is a structure needing 'struct' keyword to be used
3153 - e_toto_t is an enum
3154 - toto_t is an 'object' (struct*)
3156 typedef struct s_toto {} s_toto_t, *toto_t;
3157 typedef enum {} e_toto_t;
3158 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
3159 s_toto_t) is private.
3161 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
3162 so that it can set the variable to NULL. It was so for dicts and sets,
3163 it changed for dynars.
3165 - Fix a bunch of memleaks in dict_remove
3166 - Fix a bug in sg/server_socket opening: it failed all the time.
3168 2004-10-07 Martin Quinson
3169 - Speed up dynar lookup operation a bit.
3171 gras_dynar_get is dead.
3173 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
3174 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
3175 (which gives you the address of the stored data).
3177 gras_dynar_get_as is an helpful macro which allows you to retrieve a
3178 copy of the data using an affectation to do the job and not a memcpy.
3180 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
3181 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
3183 It does not really speedup the dynar test because they are
3184 setting elements all the time (and look them seldom). But the dict does
3185 far more lookup than setting.
3187 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
3189 2004-10-05 Martin Quinson
3190 - Allow to (en/dis)able the cycle detection at run time.
3192 Whether we should check for cycle or not is now a property of each
3193 datatype. When you think there may be some cycle, use datadesc_cycle_set.
3194 datadesc_cycle_unset allow to remove this property when previously set.
3196 Note that the cycle detection is off by default since it impacts the
3197 performance. Watch the data you feed GRAS with ;)
3199 This property is hereditary. Any element embedded in a structure having it
3200 set have it set for the time of this data exchange.
3202 You should set it both on sender and receiver side. If you don't set it on
3203 sender side, it will enter an endless loop. If you forget on receiver
3204 side, the cycles won't be recreated after communication.
3206 - Header reorganization.
3207 Kill gras_private.h, each submodule must load the headers it needs.
3209 2004-10-04 Martin Quinson
3210 - Interface revolution: do not try to survive to malloc failure.
3212 Now, gras_malloc and friends call gras_abort() on failure.
3213 As a conclusion, malloc_error is not a valid error anymore, and all
3214 functions for which it was the only gras_error_t return value are
3215 changed. They now return void, or there result directly.
3216 This simplify the API a lot.
3218 2004-09-29 Martin Quinson
3219 - Re-enable raw sockets.
3220 Created by gras_socket_{client,server}_ext;
3221 Used with gras_raw_{send,recv}
3224 It should allow to kill the last bits of gras first version soon.
3226 This is not completely satisfactory yet (duplicate code with
3227 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
3230 - Simplify transport plugin (internal) interface by not passing any
3231 argument to _server and _client, but embedding them in the socket
3234 2004-09-28 Martin Quinson
3235 - Finish the port to AIX.
3236 autoconf was my problem (segfault within the malloc replacement
3237 function. No idea why)
3239 2004-09-16 Martin Quinson
3240 - Fix some size_t madness on 64bit architectures.
3242 2004-09-08 Martin Quinson
3243 - Reduce the number of system headers loaded, overload some more system
3244 calls (such as malloc to cast the result of the system one, and work
3246 - Fix and reintroduce the config support
3248 2004-09-07 Martin Quinson
3249 - Source code reorganization to allow Arnaud to surf all over there.
3250 - Allow to document the logging categories.
3251 - Remove all uppercase from logging categories and useless cleanup in names.
3253 2004-08-18 Martin Quinson
3254 Version 0.6.2 (protocol not changed; API changed)
3255 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
3256 gras_error_t), and NULL when not found. Functions expecting a msgtype
3257 as argument (msg_wait; msg_send) deal with NULL argument by providing a
3258 hopefully usefull message.
3259 - Portability to prehistoric sparcs again
3261 2004-08-17 Martin Quinson
3262 Version 0.6.1 (protocol not changed; ABI not changed)
3263 - prealloc some buffers to speed things up
3265 2004-08-11 Martin Quinson
3266 Version 0.6 (protocol not changed; ABI expended)
3267 - The parsing macro can deal with the references, provided that you add
3268 the relevant annotations (using GRAS_ANNOTE(size,field_name))
3270 2004-08-09 Martin Quinson
3271 Version 0.5 (protocol not changed; ABI changed)
3272 - Allow to off turn the cycle detection code in data exchange at
3273 compilation time. It should be at run time, but I'm short of time (and
3274 the config stuff is still broken). That way, we keep dict out of the
3275 critical path, which is good because the performance is poor:
3276 - search not dichotomial yet
3277 - dynar give no way to access their content and memcpy everytime
3278 - In composed data description (struct, ref and so on), stop foolness of
3279 keeping the subtype's ID, but store the type itself. This keeps sets out
3280 of the critical path, which is good since they rely on dynar and
3281 dictionnaries. The only loose of that is that we cannot detect the
3282 redeclaration of a structure/union with another content (but I'm not sure
3283 the code detected well this error before anyway). We still can detect
3284 the redefinition discrepancy for the other types.
3285 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
3286 it breaks the code because of type-punning used all over the place).
3287 This breaks on all non-gcc architectures (for now).
3289 All those changes (plus the buffer of last time) allow me to gain 2 order
3290 of magnitude on cruel tests consisting of 800000 array of integers on two
3291 level of a hierarchical structure (200 secondes -> 4 secondes)
3294 - the selector of reference must now return the type it points to, not
3295 the ID of this type.
3297 2004-08-06 Martin Quinson
3298 Version 0.4 (protocol changed; ABI not changed)
3299 - Allow to pass --gras-log argument to processes in simulation mode. Really.
3300 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
3301 - Implement a buffer transport, and use it by default (it relies on tcp in
3302 real life and on sg in simulation).
3303 That's a bit hackish since I had a new field to the structure to store
3304 its data without interfering with the subtype ones. Inheritance
3305 is tricky in C. And that's a kind of reverse inheritance with one class
3306 derivating two classes. Or maybe a game with java interfaces. Anyway,
3307 that's damn hard in C (at least).
3308 Moreover, I got tired while trying to ensure plugin separation and
3309 genericity in SG mode. MSG wants me to do weird things, so let's go for
3310 cruel hacks (temporarily of course ;).
3311 See comment in transport_private.h:71
3312 - do not include all the _interface headers in private but in the files
3313 which really need them (to cut the compilation time when they are
3316 2004-07-26 Martin Quinson
3317 Version 0.3 (protocol not changed; ABI changed)
3318 - Major overhault of the datadesc interface to simplify it:
3319 - shorted the function names:
3320 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
3321 - add a trivial way to push/pop integers into the cbps without malloc.
3322 This allows to make really generic sub_type description, which simply
3323 pop their size of the stack.
3324 - add a function gras_datadesc_ref_pop_arr() which does what users want
3325 most of the time: Declare a dynamic array (which pops its size of the
3326 stack) and declare a reference to it. Poor name, but anyway.
3327 - kill the post-send callback, add a post-receive one
3329 2004-07-23 Martin Quinson
3330 Version 0.2 (protocol changed; ABI changed)
3331 - add some testing for cpbs in the test cases, and fix some more bugs.
3332 This invalidate again the little64 data file, since I cannot regenerate
3334 - remove an awfull optimization in the logging stuff, allowing me to:
3335 - understand it again
3336 - learn gcc how to check that the argument match the provided format
3337 - fix all errors revealed by gcc after that
3338 - internal keys of dict are not \0 terminated. Deal with it properly in
3339 loggings instead of segfaulting when the user want to see the logs :-/
3341 2004-07-22 Martin Quinson
3342 - Fix some stupid bug preventing cbps (callback postit) from working
3344 2004-07-21 Martin Quinson
3345 - Some documentation cleanups
3346 - remove the useless last argument of msgtype_declare
3347 - rename the Virtu functions to fit into the 'os' namespace
3348 - move headers src/include -> src/include/gras/ and stop fooling with
3349 gras -> . symbolic link
3350 - make distcheck is now successful
3352 2004-07-19 Martin Quinson
3354 - Build shared library also
3355 - Install html doc to the right location
3356 - stop removing maintainer files in make clean
3357 - build tests only on make check
3359 2004-07-13 Martin Quinson
3361 - No major issue in previous version => change versioning schema
3362 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3363 corresponding dataset.
3365 2004-07-11 Martin Quinson
3367 - Get it working with any kind of structure (we can compute the padding
3368 bytes remotely for all the architectures I have access to)
3369 - Implement the structure parsing macro (still not quite robust/complete)
3370 - Improvement to the remote testing toysuite
3372 2004-07-10 Martin Quinson
3373 [autoconf mechanism]
3374 - get ride of a bunch of deprecated macros
3375 - actually run the test for two-compliment, not only compile it :-/
3376 - test whether the structures get packed (and bail out if yes. Damn.
3377 Alignment is a serious matter)
3378 - test whether the structures get compacted (but respecting the alignment
3379 constraints of each types)
3380 - test whether the array fields of structures can straddle alignment boundaries
3382 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3383 (took me 2 hours to find that bug, looking at the wrong place)
3384 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3385 before sending/receiving and are used to compute the offsets of fields
3386 - Given that padding size depend even on compiler options, keep track of
3387 alignment and aligned_size only for the current architecture. Not a big
3388 deal since we send structure fields one after the other (seems
3390 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3391 program, let it work on linux/gcc/little32. portability todo.
3393 2004-07-08 Martin Quinson
3394 - import and improve remote compilation support from FAST
3395 - make sure make check works on half a dozen of machines out there
3397 2004-07-07 Martin Quinson
3398 Let's say it's version 0.0.3 ;)
3399 - Implement conversions (yuhu!)
3400 - Let it work on solaris (beside conversion, of course)
3401 - Stupid me, using rand() to generate the conversion datatests in not wise.
3403 2004-07-06 Martin Quinson
3404 - Let make dist work, since I'm gonna need it to compile on remote hosts
3405 - Let Tests/datadesc_usage write the architecture on which the file was
3406 generated as first byte.
3407 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3408 architecture descriptions.
3409 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3410 on those architectures.
3411 - Optimization: send/recv array of scalar in one shoot
3413 2004-07-05 Martin Quinson
3414 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3416 - Plug a whole bunch of memleaks
3417 - each process now have to call gras_{init,exit}. One day, their log
3418 settings will be separated
3419 - Continue the code factorisation between SG, RL and common in Transport.
3421 2004-07-04 Martin Quinson
3423 - Redistribution between SG and RL.
3424 We wanna have to accept in SG, so move accepted related parts of RL in
3425 the common part. (more precisely, the dynar of all known sockets is no
3426 more a static in transport.c, but part of the process_data)
3428 [gras_stub_generator]
3429 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3431 2004-07-03 Martin Quinson
3432 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3433 them all up in one shot)
3435 - Fix the ugly case of reference to dynamic array.
3436 - New (semi-public) function gras_datadesc_size to allow the messaging
3437 layer to malloc the needed space for the buffer.
3439 - gras_socket_close now expect the socket to close (and not its address to
3440 put NULL in it after it). This is because the socket passed to handlers
3441 is one of their argument (=> not writable).
3443 - propagate the interface cleanup from last week in datadesc, ie remove a
3444 superfluous level of indirection. User pass adress of variable
3445 containing data (both when sending and receiving), and not of a variable
3446 being a pointer to the data. Let's say that I like it better ;)
3447 The price for that is constructs like "int msg=*(int*)payload" in
3448 handlers, but it's a fine price, IMHO.
3450 - Let it work in RL (yuhu)
3452 2004-06-21 Martin Quinson
3454 - porting SG plugin and SG select to new standards (works almost).
3455 - plug memleaks and fix bugs around.
3458 - cleanup the prototype of data recv and force users to specify when they
3459 want to handle references to objects. Test case working even for cycles.
3460 - plug memleaks. Valgrind is perfectly ok with this.
3462 2004-06-12 Martin Quinson
3464 - cleanup the separation between plugin and main code in plugin creation
3466 2004-06-11 Martin Quinson
3468 - Reput hook for raw sockets, needed for BW experiments
3469 - kill a few lines of dead code
3470 [Data description] Interface cleanup
3471 - gras_datadesc_by_name returns the searched type or NULL.
3472 That way, no variable is needed to use a type desc once, which makes
3474 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3475 two parameters were strings (field name, type name), leading to
3477 [Dicos] Interface cleanup
3478 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3479 This is consistant with the dynar API.
3481 2004-04-21 Martin Quinson
3483 - Porting to new standards.
3485 - interface cleanup.
3486 There is no bag anymore, no need to take extra provision to mask the
3487 pointers behind "ID".
3488 Better splitup of functions between files create/exchange/convert.
3489 This is still a bit artificial since convert and receive are so
3490 interleaved, but anyway.
3492 - add a queued message list to procdata (the ones not matching criteria
3494 - factorize some more code between SG and RL wrt procdata
3496 - use gras_exit in example to track memleaks
3497 - get rid of gs_example now that GS is properly integrated into gras
3498 - update run_test to integrate the latest tests (datadesc)
3500 - rename WARNINGn macros to WARNn since it prooved error-prone
3502 2004-04-19 Martin Quinson
3504 - register init/exit functions within gras module mechanism
3505 - send/receive function.
3506 Convertion is not implemented, but short-cutted if not needed.
3507 struct/array elements are sent one by one (instead of block-wise), but
3508 nobody really cares (yet). Get a prototype before optimizing.
3509 - tests (using a file socket) for DD send/receive on:
3510 - base types: int, float
3511 - array: fixed size, string (ie ref to dynamic string)
3512 - structure: homogeneous, heterogeneous
3513 - chained list, graph with cycle
3514 Believe it or not, valgrind is not too unhappy with the results. The
3515 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3519 - Bugfix when using a filename explicitely (instead of '-')
3521 2004-04-09 Martin Quinson
3523 - factorize more code between RL and SG in socket creation
3524 - Complete the implementation and tests of:
3526 o file (only in RL, and mainly for debugging)
3528 I lost 3 days to design a portable address resolver, and then decided
3529 that the prototype mainly have to run on my box.
3530 Addressing portability too early may be like optimizing too early :-/
3532 - use gras_init in the Tests instead of the crappy parse_log_opt
3533 (the latter function is removed)
3534 [Conditional execution]
3535 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3536 [Code reorganisation]
3537 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3538 Build examples against the RL library, since there is no way to disable
3539 its creation for now.
3541 For information, the beginning of coding on GRAS was back in june
3542 2003. I guess that every line has been rewritten at least twice since